Original Research

According to the Centers for Disease Control and Prevention National Center for Health Statistics (NCHS), the use of prescription drugs has increased in the past half century. Although prescription drugs have played an important role in preventing, controlling, and delaying onset or progression of disease, their growth in use also has posed many risks.¹ One ramification of this growth is the occurrence of polypharmacy, which does not have a universal, clear definition. In general, it can be described as the concurrent use of multiple medications by a single patient to treat one or more medical ailments. Five or more medications taken simultaneously is the most common definition to date, but this is just one of many acceptable definitions and that varies from one health care facility to another.^1,2

Regardless of the cutoffs established to indicate polypharmacy, its incidence can result in poor and potentially harmful health outcomes. Polypharmacy increases the risk of experiencing adverse drug events (ADEs), drug-drug interactions (DDIs), geriatric-related syndromes, falls, hospitalization, and mortality. Issues with adherence may begin to unfold secondary to increased pill burden. Both the patient and the health care system may encounter financial strain, as polypharmacy can lead to unnecessary and essentially preventable costs of care. When evaluating the likelihood of polypharmacy based on age group, NCHS found that 47.5% of patients taking ≥ 5 medications were aged ≥ 65 years.^1-5 This indicates that polypharmacy is of great concern in the geriatric population, which also represents a large proportion of individuals accessing Veterans Health Administration (VHA) care.

Deprescibing

Deprescribing is the act of withdrawing or discontinuing potentially inappropriate medications (PIM), or medications used by older patients harboring ADEs that generally outweigh the clinical benefits of the drug. Deprescribing is an effective tool for managing or reducing polypharmacy. A variety of tools have been created whose sole purpose is to simplify deprescribing. Some tools explicitly identify PIM and are widely familiar in medical practice. Examples are the Beers Criteria developed in 1991 or Screening Tool to Alert Right Treatment/Screening Tool of Older Persons Prescriptions (START/STOPP) criteria created in 2003. Other tools that are less commonplace but equally as resourceful are MedStopper and Deprescribing.org. The former was launched in 2015 and is a Canadian online system that provides risk assessments for medications with guidance for tapering or stopping medications if continuation of the drug presents higher risk than benefit.^5-7 The latter is a full-fledged website developed by a physician, a pharmacist, and their research teams that serves as an exchange hub for deprescribing information.

In 2016, the VIONE (Vital, Important, Optional, Not indicated/treatment complete, and Every medication has an indication) deprescribing tool was developed by Saraswathy Battar, MD, at Central Arkansas Veterans Healthcare System (CAVHS) in Little Rock, as a system that could go beyond medication reconciliation (Table 1). Health care providers (HCPs) and pharmacists evaluate each medication that a patient has been prescribed and places each medication in a VIONE category. Prescribers may then take the opportunity to deprescribe or discontinue medications if deemed appropriate based on their clinical assessments and shared decision making.⁸ Traditionally, medication reconciliation involves the process of obtaining a complete and accurate list of medications as reported by a patient or caregiver to a HCP. VIONE encourages HCPs and pharmacists not only to ensure medication lists are accurate, but also that each medication reported is appropriate for continued use. In other words, VIONE is meant to help implement deprescribing at opportune times. More than 14,000 medications have been deprescribed using the VIONE method, resulting in more than $2,000,000 of annualized cost avoidance after just 1 year of implementation at CAVHS.⁹

VIONE consists of 2 major components in the Computerized Patient Record System (CPRS): a template and a dropdown discontinuation menu. The template captured patient allergies, pertinent laboratory data, the patient’s active problem list and applicable diagnoses, and active medication list. Patient aligned care team (PACT) pharmacists used the information captured in the template to conduct medication reconciliations and polypharmacy reviews. Each medication is categorized in VIONE using data collected during reviews. A menu delineates reasons for discontinuation: optional, dose decrease, no diagnosis, not indicated/treatment complete, discontinue alternate medication prescribed, and patient reported no longer taking. The discontinuation menu allowed PACT pharmacists and physicians to choose 1 VIONE option per medication to clarify the reason for discontinuation. VIONE-based discontinuations are recorded in CPRS and identified as deprescribed.

At the time of this project, > 30 US Department of Veterans Affairs (VA) facilities had adopted VIONE. Use of VIONE at VA Southern Nevada Healthcare System (VASNHS) in North Las Vegas has been incorporated in the everyday practices of home-based primary care pharmacists and physicians but has yet to be implemented in other areas of the facility. The purpose of this project was to determine the impact of the VIONE tool on polypharmacy and cost avoidance at VASNHS when used by primary care physicians (PCPs) and PACT primary care clinics.

 

Methods

Veterans receiving care at VASNHS aged ≥ 65 years with ≥ 10 active medications noted in CPRS were included in this project. PACT pharmacists and physicians were educated on the proper use of the VIONE tool prior to its implementation. Education included a 15-minute slide presentation followed by dissemination of a 1-page VIONE tool handout during a PACT all-staff clinic meeting.

Data were collected for 3 months before and after the intervention. Data were made available for assessment by the Automated Data Processing Application Coordinator (ADPAC) at VASNHS. The ADPAC created and generated an Excel spreadsheet report, which listed all medications deprescribed using the VIONE method. The primary endpoint was the total number of medications discontinued using the VIONE template and/or discontinuation menu. For the purpose of this project, appropriate discontinuation was considered any prescription deprescribed, excluding medical supplies, by pharmacists and PCPs who received VIONE education.

The secondary endpoint was the estimated annualized cost avoidance for the facility (Figure). The calculation does not include medications discontinued due to the prescription of an alternative medication or dose decreases since these VIONE selections imply that a new prescription or order was placed and the original prescription was not deprescribed. Annualized cost avoidance was determined with use of the VIONE dashboard, a database that retrospectively gathers information regarding patients at risk of polypharmacy, polypharmacy-related ADEs, and cost. Manual adjustments were made to various parameters on the Veterans Integrated Service Network 15 VIONE dashboard by the author in order to obtain data specific to this project. These parameters allowed selection of service sections, specific staff members or the option to include or exclude chronic or nonchronic medications. The annualized cost avoidance figure was then compared to raw data pulled by a VIONE dashboard correspondent to ensure the manual calculation was accurate. Finally, the 5 most common classes of medications deprescribed were identified for information purposes and to provide a better postulation on the types of medications being discontinued using the VIONE method.

Results

A total of 2,442 veterans met inclusion criteria, and the VIONE method was applied to 598 between late October 2018 and January 2019. The 13 PACT pharmacists contacted at least 10 veterans each, thus at least 130 were randomly selected for telephone calls to perform polypharmacy reviews using the VIONE note template. The discontinuation menu was used if a medication qualified to be deprescribed. After 3 months, 1986 prescriptions were deprescribed using VIONE; however, 1060 prescriptions were considered appropriately deprescribed (Table 2). The 13 PACT pharmacists deprescribed 361 medications, and the 29 PACT physicians deprescribed 699 medications. These prescriptions were then separated into medication categories to determine the most common discontinued classes. Vitamins and supplements were the medication class most frequently deprescribed (19.4%), followed by pain medications (15.5%), antimicrobial agents (9.6%), antihypertensive medications (9.2%), and diabetes medications (6.4%) (Table 3). The top 5 medication categories accounted for 60% of all medications appropriately deprescribed.

The estimated annualized cost avoidance for all medications deprescribed in the 3-month project period was $84,030.46. To provide the most appropriate and accurate calculation, medication classes excluded from this figure were acute or short-term prescriptions and antimicrobial agents. Medications prescribed short-term typically are not suitable to continue for an extended period, and antimicrobial agents were excluded since they are normally associated with higher costs, and may overestimate the cost avoidance calculation for the facility.

 

Discussion

The outcomes for the primary and secondary endpoints of this project illustrate that using VIONE in PACT primary care clinics had a notable impact on polypharmacy and cost avoidance over a short period. This outcome can be attributed to 2 significant effects of using the deprescribing tool. VIONE’s simplicity in application allowed clinicians to incorporate daily use of the tool with minimal effort. Education was all that was required to fully enable clinicians to work together successfully and exercise collaborative practice to promote deprescribing. VIONE also elicited a cascade of favorable effects that improve patient safety and health outcomes. The tool aided in identification of PIM, which helped reduce polypharmacy and medication burden. The risk for DDIs and ADEs may decrease; therefore, the incidence of falls, need for emergency department visits or inpatient care related to polypharmacy may decline. Less complex medication regimens may alleviate issues with adherence and avoid the various consequences of polypharmacy in theory. Simplified regimens can potentially improve disease management and quality of life for patients. Further studies are needed to substantiate deprescribing and its true effect on patient adherence and better health outcomes at this time.¹⁰

Reducing polypharmacy can lead to cost savings. Based on the results of this 3-month study, we expect that VASNHS would save more than $84,000 by reducing polypharmacy among its patients. Those savings can be funneled back into the health care system, and allotted to necessary patient care, prescriptions, and health care facility needs.

Limitations

There are some important limitations to this study. Definitions of polypharmacy may vary from one health care facility to another. The cutoffs for polypharmacy may differ, causing the prevalence of polypharmacy and potential costs savings to vary. Use of VIONE may be inconsistent among users if not previously educated or properly trained. For instance, VIONE selections are listed in the same menu as the standard CPRS discontinuation options, which may lead to discontinuation of medical supplies or laboratory orders instead of prescriptions.

The method of data analysis and project design used in this study may have been subject to error. For example, the list of PCPs may have been inaccurate or outdated, which would result in an over- or underrepresentation of those who contributed to data collection. Furthermore, there is some volatility in calculating the total cost avoidance. For example, medications for chronic conditions that were only taken on an as needed basis may have overestimated savings. Either under- or overestimations could occur when parameters are adjusted on the VIONE discontinuation dashboard without appropriate guidance. With the ability to manually adjust the dashboard parameters, dissimilarities in calculations may follow.

Conclusions

The VIONE tool may be useful in improving patient safety through deprescribing and discontinuing PIM. Decreasing the number of medications being taken concomitantly by a patient and continuing only those that are imperative in their medical treatment is the first step to reducing the incidence of polypharmacy. Consequently, chances of ADEs or DDIs are lessened, especially among older individuals who are considered high risk for experiencing the detrimental effects that may ensue. These effects include geriatric-related syndromes, increased risk of fall, hospital visits or admissions, or death. Use of VIONE easily promotes collaboration among clinicians to evaluate medications eligible for discontinuation more regularly. If this deprescribing tool is continuously used, costs avoided can likely be maximized within VA health care systems.

The results of this project should serve as an incentive to push for better prescribing practices and increase deprescribing efforts. It should provoke the need for change in regimens and the subsequent discontinuation of prescriptions that are not considered vital to continue. Finally, the result of this project should substantiate the positive impact a deprescribing tool can possess to avert the issues commonly associated with polypharmacy.

According to the Centers for Disease Control and Prevention National Center for Health Statistics (NCHS), the use of prescription drugs has increased in the past half century. Although prescription drugs have played an important role in preventing, controlling, and delaying onset or progression of disease, their growth in use also has posed many risks.¹ One ramification of this growth is the occurrence of polypharmacy, which does not have a universal, clear definition. In general, it can be described as the concurrent use of multiple medications by a single patient to treat one or more medical ailments. Five or more medications taken simultaneously is the most common definition to date, but this is just one of many acceptable definitions and that varies from one health care facility to another.^1,2

Regardless of the cutoffs established to indicate polypharmacy, its incidence can result in poor and potentially harmful health outcomes. Polypharmacy increases the risk of experiencing adverse drug events (ADEs), drug-drug interactions (DDIs), geriatric-related syndromes, falls, hospitalization, and mortality. Issues with adherence may begin to unfold secondary to increased pill burden. Both the patient and the health care system may encounter financial strain, as polypharmacy can lead to unnecessary and essentially preventable costs of care. When evaluating the likelihood of polypharmacy based on age group, NCHS found that 47.5% of patients taking ≥ 5 medications were aged ≥ 65 years.^1-5 This indicates that polypharmacy is of great concern in the geriatric population, which also represents a large proportion of individuals accessing Veterans Health Administration (VHA) care.

Deprescibing

Deprescribing is the act of withdrawing or discontinuing potentially inappropriate medications (PIM), or medications used by older patients harboring ADEs that generally outweigh the clinical benefits of the drug. Deprescribing is an effective tool for managing or reducing polypharmacy. A variety of tools have been created whose sole purpose is to simplify deprescribing. Some tools explicitly identify PIM and are widely familiar in medical practice. Examples are the Beers Criteria developed in 1991 or Screening Tool to Alert Right Treatment/Screening Tool of Older Persons Prescriptions (START/STOPP) criteria created in 2003. Other tools that are less commonplace but equally as resourceful are MedStopper and Deprescribing.org. The former was launched in 2015 and is a Canadian online system that provides risk assessments for medications with guidance for tapering or stopping medications if continuation of the drug presents higher risk than benefit.^5-7 The latter is a full-fledged website developed by a physician, a pharmacist, and their research teams that serves as an exchange hub for deprescribing information.

In 2016, the VIONE (Vital, Important, Optional, Not indicated/treatment complete, and Every medication has an indication) deprescribing tool was developed by Saraswathy Battar, MD, at Central Arkansas Veterans Healthcare System (CAVHS) in Little Rock, as a system that could go beyond medication reconciliation (Table 1). Health care providers (HCPs) and pharmacists evaluate each medication that a patient has been prescribed and places each medication in a VIONE category. Prescribers may then take the opportunity to deprescribe or discontinue medications if deemed appropriate based on their clinical assessments and shared decision making.⁸ Traditionally, medication reconciliation involves the process of obtaining a complete and accurate list of medications as reported by a patient or caregiver to a HCP. VIONE encourages HCPs and pharmacists not only to ensure medication lists are accurate, but also that each medication reported is appropriate for continued use. In other words, VIONE is meant to help implement deprescribing at opportune times. More than 14,000 medications have been deprescribed using the VIONE method, resulting in more than $2,000,000 of annualized cost avoidance after just 1 year of implementation at CAVHS.⁹

VIONE consists of 2 major components in the Computerized Patient Record System (CPRS): a template and a dropdown discontinuation menu. The template captured patient allergies, pertinent laboratory data, the patient’s active problem list and applicable diagnoses, and active medication list. Patient aligned care team (PACT) pharmacists used the information captured in the template to conduct medication reconciliations and polypharmacy reviews. Each medication is categorized in VIONE using data collected during reviews. A menu delineates reasons for discontinuation: optional, dose decrease, no diagnosis, not indicated/treatment complete, discontinue alternate medication prescribed, and patient reported no longer taking. The discontinuation menu allowed PACT pharmacists and physicians to choose 1 VIONE option per medication to clarify the reason for discontinuation. VIONE-based discontinuations are recorded in CPRS and identified as deprescribed.

At the time of this project, > 30 US Department of Veterans Affairs (VA) facilities had adopted VIONE. Use of VIONE at VA Southern Nevada Healthcare System (VASNHS) in North Las Vegas has been incorporated in the everyday practices of home-based primary care pharmacists and physicians but has yet to be implemented in other areas of the facility. The purpose of this project was to determine the impact of the VIONE tool on polypharmacy and cost avoidance at VASNHS when used by primary care physicians (PCPs) and PACT primary care clinics.

 

Methods

Veterans receiving care at VASNHS aged ≥ 65 years with ≥ 10 active medications noted in CPRS were included in this project. PACT pharmacists and physicians were educated on the proper use of the VIONE tool prior to its implementation. Education included a 15-minute slide presentation followed by dissemination of a 1-page VIONE tool handout during a PACT all-staff clinic meeting.

Data were collected for 3 months before and after the intervention. Data were made available for assessment by the Automated Data Processing Application Coordinator (ADPAC) at VASNHS. The ADPAC created and generated an Excel spreadsheet report, which listed all medications deprescribed using the VIONE method. The primary endpoint was the total number of medications discontinued using the VIONE template and/or discontinuation menu. For the purpose of this project, appropriate discontinuation was considered any prescription deprescribed, excluding medical supplies, by pharmacists and PCPs who received VIONE education.

The secondary endpoint was the estimated annualized cost avoidance for the facility (Figure). The calculation does not include medications discontinued due to the prescription of an alternative medication or dose decreases since these VIONE selections imply that a new prescription or order was placed and the original prescription was not deprescribed. Annualized cost avoidance was determined with use of the VIONE dashboard, a database that retrospectively gathers information regarding patients at risk of polypharmacy, polypharmacy-related ADEs, and cost. Manual adjustments were made to various parameters on the Veterans Integrated Service Network 15 VIONE dashboard by the author in order to obtain data specific to this project. These parameters allowed selection of service sections, specific staff members or the option to include or exclude chronic or nonchronic medications. The annualized cost avoidance figure was then compared to raw data pulled by a VIONE dashboard correspondent to ensure the manual calculation was accurate. Finally, the 5 most common classes of medications deprescribed were identified for information purposes and to provide a better postulation on the types of medications being discontinued using the VIONE method.

Results

A total of 2,442 veterans met inclusion criteria, and the VIONE method was applied to 598 between late October 2018 and January 2019. The 13 PACT pharmacists contacted at least 10 veterans each, thus at least 130 were randomly selected for telephone calls to perform polypharmacy reviews using the VIONE note template. The discontinuation menu was used if a medication qualified to be deprescribed. After 3 months, 1986 prescriptions were deprescribed using VIONE; however, 1060 prescriptions were considered appropriately deprescribed (Table 2). The 13 PACT pharmacists deprescribed 361 medications, and the 29 PACT physicians deprescribed 699 medications. These prescriptions were then separated into medication categories to determine the most common discontinued classes. Vitamins and supplements were the medication class most frequently deprescribed (19.4%), followed by pain medications (15.5%), antimicrobial agents (9.6%), antihypertensive medications (9.2%), and diabetes medications (6.4%) (Table 3). The top 5 medication categories accounted for 60% of all medications appropriately deprescribed.

The estimated annualized cost avoidance for all medications deprescribed in the 3-month project period was $84,030.46. To provide the most appropriate and accurate calculation, medication classes excluded from this figure were acute or short-term prescriptions and antimicrobial agents. Medications prescribed short-term typically are not suitable to continue for an extended period, and antimicrobial agents were excluded since they are normally associated with higher costs, and may overestimate the cost avoidance calculation for the facility.

 

Discussion

The outcomes for the primary and secondary endpoints of this project illustrate that using VIONE in PACT primary care clinics had a notable impact on polypharmacy and cost avoidance over a short period. This outcome can be attributed to 2 significant effects of using the deprescribing tool. VIONE’s simplicity in application allowed clinicians to incorporate daily use of the tool with minimal effort. Education was all that was required to fully enable clinicians to work together successfully and exercise collaborative practice to promote deprescribing. VIONE also elicited a cascade of favorable effects that improve patient safety and health outcomes. The tool aided in identification of PIM, which helped reduce polypharmacy and medication burden. The risk for DDIs and ADEs may decrease; therefore, the incidence of falls, need for emergency department visits or inpatient care related to polypharmacy may decline. Less complex medication regimens may alleviate issues with adherence and avoid the various consequences of polypharmacy in theory. Simplified regimens can potentially improve disease management and quality of life for patients. Further studies are needed to substantiate deprescribing and its true effect on patient adherence and better health outcomes at this time.¹⁰

Reducing polypharmacy can lead to cost savings. Based on the results of this 3-month study, we expect that VASNHS would save more than $84,000 by reducing polypharmacy among its patients. Those savings can be funneled back into the health care system, and allotted to necessary patient care, prescriptions, and health care facility needs.

Limitations

There are some important limitations to this study. Definitions of polypharmacy may vary from one health care facility to another. The cutoffs for polypharmacy may differ, causing the prevalence of polypharmacy and potential costs savings to vary. Use of VIONE may be inconsistent among users if not previously educated or properly trained. For instance, VIONE selections are listed in the same menu as the standard CPRS discontinuation options, which may lead to discontinuation of medical supplies or laboratory orders instead of prescriptions.

The method of data analysis and project design used in this study may have been subject to error. For example, the list of PCPs may have been inaccurate or outdated, which would result in an over- or underrepresentation of those who contributed to data collection. Furthermore, there is some volatility in calculating the total cost avoidance. For example, medications for chronic conditions that were only taken on an as needed basis may have overestimated savings. Either under- or overestimations could occur when parameters are adjusted on the VIONE discontinuation dashboard without appropriate guidance. With the ability to manually adjust the dashboard parameters, dissimilarities in calculations may follow.

Conclusions

The VIONE tool may be useful in improving patient safety through deprescribing and discontinuing PIM. Decreasing the number of medications being taken concomitantly by a patient and continuing only those that are imperative in their medical treatment is the first step to reducing the incidence of polypharmacy. Consequently, chances of ADEs or DDIs are lessened, especially among older individuals who are considered high risk for experiencing the detrimental effects that may ensue. These effects include geriatric-related syndromes, increased risk of fall, hospital visits or admissions, or death. Use of VIONE easily promotes collaboration among clinicians to evaluate medications eligible for discontinuation more regularly. If this deprescribing tool is continuously used, costs avoided can likely be maximized within VA health care systems.

The results of this project should serve as an incentive to push for better prescribing practices and increase deprescribing efforts. It should provoke the need for change in regimens and the subsequent discontinuation of prescriptions that are not considered vital to continue. Finally, the result of this project should substantiate the positive impact a deprescribing tool can possess to avert the issues commonly associated with polypharmacy.

According to the Centers for Disease Control and Prevention National Center for Health Statistics (NCHS), the use of prescription drugs has increased in the past half century. Although prescription drugs have played an important role in preventing, controlling, and delaying onset or progression of disease, their growth in use also has posed many risks.¹ One ramification of this growth is the occurrence of polypharmacy, which does not have a universal, clear definition. In general, it can be described as the concurrent use of multiple medications by a single patient to treat one or more medical ailments. Five or more medications taken simultaneously is the most common definition to date, but this is just one of many acceptable definitions and that varies from one health care facility to another.^1,2

Regardless of the cutoffs established to indicate polypharmacy, its incidence can result in poor and potentially harmful health outcomes. Polypharmacy increases the risk of experiencing adverse drug events (ADEs), drug-drug interactions (DDIs), geriatric-related syndromes, falls, hospitalization, and mortality. Issues with adherence may begin to unfold secondary to increased pill burden. Both the patient and the health care system may encounter financial strain, as polypharmacy can lead to unnecessary and essentially preventable costs of care. When evaluating the likelihood of polypharmacy based on age group, NCHS found that 47.5% of patients taking ≥ 5 medications were aged ≥ 65 years.^1-5 This indicates that polypharmacy is of great concern in the geriatric population, which also represents a large proportion of individuals accessing Veterans Health Administration (VHA) care.

Deprescibing

Deprescribing is the act of withdrawing or discontinuing potentially inappropriate medications (PIM), or medications used by older patients harboring ADEs that generally outweigh the clinical benefits of the drug. Deprescribing is an effective tool for managing or reducing polypharmacy. A variety of tools have been created whose sole purpose is to simplify deprescribing. Some tools explicitly identify PIM and are widely familiar in medical practice. Examples are the Beers Criteria developed in 1991 or Screening Tool to Alert Right Treatment/Screening Tool of Older Persons Prescriptions (START/STOPP) criteria created in 2003. Other tools that are less commonplace but equally as resourceful are MedStopper and Deprescribing.org. The former was launched in 2015 and is a Canadian online system that provides risk assessments for medications with guidance for tapering or stopping medications if continuation of the drug presents higher risk than benefit.^5-7 The latter is a full-fledged website developed by a physician, a pharmacist, and their research teams that serves as an exchange hub for deprescribing information.

In 2016, the VIONE (Vital, Important, Optional, Not indicated/treatment complete, and Every medication has an indication) deprescribing tool was developed by Saraswathy Battar, MD, at Central Arkansas Veterans Healthcare System (CAVHS) in Little Rock, as a system that could go beyond medication reconciliation (Table 1). Health care providers (HCPs) and pharmacists evaluate each medication that a patient has been prescribed and places each medication in a VIONE category. Prescribers may then take the opportunity to deprescribe or discontinue medications if deemed appropriate based on their clinical assessments and shared decision making.⁸ Traditionally, medication reconciliation involves the process of obtaining a complete and accurate list of medications as reported by a patient or caregiver to a HCP. VIONE encourages HCPs and pharmacists not only to ensure medication lists are accurate, but also that each medication reported is appropriate for continued use. In other words, VIONE is meant to help implement deprescribing at opportune times. More than 14,000 medications have been deprescribed using the VIONE method, resulting in more than $2,000,000 of annualized cost avoidance after just 1 year of implementation at CAVHS.⁹

VIONE consists of 2 major components in the Computerized Patient Record System (CPRS): a template and a dropdown discontinuation menu. The template captured patient allergies, pertinent laboratory data, the patient’s active problem list and applicable diagnoses, and active medication list. Patient aligned care team (PACT) pharmacists used the information captured in the template to conduct medication reconciliations and polypharmacy reviews. Each medication is categorized in VIONE using data collected during reviews. A menu delineates reasons for discontinuation: optional, dose decrease, no diagnosis, not indicated/treatment complete, discontinue alternate medication prescribed, and patient reported no longer taking. The discontinuation menu allowed PACT pharmacists and physicians to choose 1 VIONE option per medication to clarify the reason for discontinuation. VIONE-based discontinuations are recorded in CPRS and identified as deprescribed.

At the time of this project, > 30 US Department of Veterans Affairs (VA) facilities had adopted VIONE. Use of VIONE at VA Southern Nevada Healthcare System (VASNHS) in North Las Vegas has been incorporated in the everyday practices of home-based primary care pharmacists and physicians but has yet to be implemented in other areas of the facility. The purpose of this project was to determine the impact of the VIONE tool on polypharmacy and cost avoidance at VASNHS when used by primary care physicians (PCPs) and PACT primary care clinics.

 

Methods

Veterans receiving care at VASNHS aged ≥ 65 years with ≥ 10 active medications noted in CPRS were included in this project. PACT pharmacists and physicians were educated on the proper use of the VIONE tool prior to its implementation. Education included a 15-minute slide presentation followed by dissemination of a 1-page VIONE tool handout during a PACT all-staff clinic meeting.

Data were collected for 3 months before and after the intervention. Data were made available for assessment by the Automated Data Processing Application Coordinator (ADPAC) at VASNHS. The ADPAC created and generated an Excel spreadsheet report, which listed all medications deprescribed using the VIONE method. The primary endpoint was the total number of medications discontinued using the VIONE template and/or discontinuation menu. For the purpose of this project, appropriate discontinuation was considered any prescription deprescribed, excluding medical supplies, by pharmacists and PCPs who received VIONE education.

The secondary endpoint was the estimated annualized cost avoidance for the facility (Figure). The calculation does not include medications discontinued due to the prescription of an alternative medication or dose decreases since these VIONE selections imply that a new prescription or order was placed and the original prescription was not deprescribed. Annualized cost avoidance was determined with use of the VIONE dashboard, a database that retrospectively gathers information regarding patients at risk of polypharmacy, polypharmacy-related ADEs, and cost. Manual adjustments were made to various parameters on the Veterans Integrated Service Network 15 VIONE dashboard by the author in order to obtain data specific to this project. These parameters allowed selection of service sections, specific staff members or the option to include or exclude chronic or nonchronic medications. The annualized cost avoidance figure was then compared to raw data pulled by a VIONE dashboard correspondent to ensure the manual calculation was accurate. Finally, the 5 most common classes of medications deprescribed were identified for information purposes and to provide a better postulation on the types of medications being discontinued using the VIONE method.

Results

A total of 2,442 veterans met inclusion criteria, and the VIONE method was applied to 598 between late October 2018 and January 2019. The 13 PACT pharmacists contacted at least 10 veterans each, thus at least 130 were randomly selected for telephone calls to perform polypharmacy reviews using the VIONE note template. The discontinuation menu was used if a medication qualified to be deprescribed. After 3 months, 1986 prescriptions were deprescribed using VIONE; however, 1060 prescriptions were considered appropriately deprescribed (Table 2). The 13 PACT pharmacists deprescribed 361 medications, and the 29 PACT physicians deprescribed 699 medications. These prescriptions were then separated into medication categories to determine the most common discontinued classes. Vitamins and supplements were the medication class most frequently deprescribed (19.4%), followed by pain medications (15.5%), antimicrobial agents (9.6%), antihypertensive medications (9.2%), and diabetes medications (6.4%) (Table 3). The top 5 medication categories accounted for 60% of all medications appropriately deprescribed.

The estimated annualized cost avoidance for all medications deprescribed in the 3-month project period was $84,030.46. To provide the most appropriate and accurate calculation, medication classes excluded from this figure were acute or short-term prescriptions and antimicrobial agents. Medications prescribed short-term typically are not suitable to continue for an extended period, and antimicrobial agents were excluded since they are normally associated with higher costs, and may overestimate the cost avoidance calculation for the facility.

 

Discussion

The outcomes for the primary and secondary endpoints of this project illustrate that using VIONE in PACT primary care clinics had a notable impact on polypharmacy and cost avoidance over a short period. This outcome can be attributed to 2 significant effects of using the deprescribing tool. VIONE’s simplicity in application allowed clinicians to incorporate daily use of the tool with minimal effort. Education was all that was required to fully enable clinicians to work together successfully and exercise collaborative practice to promote deprescribing. VIONE also elicited a cascade of favorable effects that improve patient safety and health outcomes. The tool aided in identification of PIM, which helped reduce polypharmacy and medication burden. The risk for DDIs and ADEs may decrease; therefore, the incidence of falls, need for emergency department visits or inpatient care related to polypharmacy may decline. Less complex medication regimens may alleviate issues with adherence and avoid the various consequences of polypharmacy in theory. Simplified regimens can potentially improve disease management and quality of life for patients. Further studies are needed to substantiate deprescribing and its true effect on patient adherence and better health outcomes at this time.¹⁰

Reducing polypharmacy can lead to cost savings. Based on the results of this 3-month study, we expect that VASNHS would save more than $84,000 by reducing polypharmacy among its patients. Those savings can be funneled back into the health care system, and allotted to necessary patient care, prescriptions, and health care facility needs.

Limitations

There are some important limitations to this study. Definitions of polypharmacy may vary from one health care facility to another. The cutoffs for polypharmacy may differ, causing the prevalence of polypharmacy and potential costs savings to vary. Use of VIONE may be inconsistent among users if not previously educated or properly trained. For instance, VIONE selections are listed in the same menu as the standard CPRS discontinuation options, which may lead to discontinuation of medical supplies or laboratory orders instead of prescriptions.

The method of data analysis and project design used in this study may have been subject to error. For example, the list of PCPs may have been inaccurate or outdated, which would result in an over- or underrepresentation of those who contributed to data collection. Furthermore, there is some volatility in calculating the total cost avoidance. For example, medications for chronic conditions that were only taken on an as needed basis may have overestimated savings. Either under- or overestimations could occur when parameters are adjusted on the VIONE discontinuation dashboard without appropriate guidance. With the ability to manually adjust the dashboard parameters, dissimilarities in calculations may follow.

Conclusions

The VIONE tool may be useful in improving patient safety through deprescribing and discontinuing PIM. Decreasing the number of medications being taken concomitantly by a patient and continuing only those that are imperative in their medical treatment is the first step to reducing the incidence of polypharmacy. Consequently, chances of ADEs or DDIs are lessened, especially among older individuals who are considered high risk for experiencing the detrimental effects that may ensue. These effects include geriatric-related syndromes, increased risk of fall, hospital visits or admissions, or death. Use of VIONE easily promotes collaboration among clinicians to evaluate medications eligible for discontinuation more regularly. If this deprescribing tool is continuously used, costs avoided can likely be maximized within VA health care systems.

The results of this project should serve as an incentive to push for better prescribing practices and increase deprescribing efforts. It should provoke the need for change in regimens and the subsequent discontinuation of prescriptions that are not considered vital to continue. Finally, the result of this project should substantiate the positive impact a deprescribing tool can possess to avert the issues commonly associated with polypharmacy.

Acne rosacea is a chronic inflammatory disease that may affect the facial skin, eyes, and eyelids.¹ It is characterized by transient or persistent flushing, facial erythema, and telangiectases, generally located on the central portion of the face, and may progress to papules and pustules.^2,3 At the late stage of the disease, dermal edema or fibroplasia and sebaceous gland hypertrophy may cause phymatous alterations in the skin. In 2004, the National Rosacea Society Expert Committee developed a classification system for rosacea to standardize subtypes and variants that has since been widely accepted and continues to aid in research and epidemiologic studies.⁴ The committee defined 4 subtypes based on clinical characteristics: erythematotelangiectatic (ETR), papulopustular (PPR), phymatous, and ocular rosacea.^2,3

Ocular rosacea may accompany mild, moderate, and severe dermatologic disease or may occur in the absence of diagnostic skin disease.⁵ Ocular signs include eyelid margin telangiectasia, spade-shaped infiltrates in the cornea, scleritis, and sclerokeratitis. Common symptoms include burning, stinging, light sensitivity, and foreign-body sensation. Ocular signs commonly seen in rosacea are meibomian gland dysfunction characterized by inspissation and inflammation of the meibomian glands (chalazia), conjunctivitis, honey crust and cylindrical collarette accumulation at the base of the eyelashes, irregularity of the eyelid margin architecture, and evaporative tear dysfunction.^5,6

The physiopathology of rosacea is still unknown. Potential factors include genetic predisposition, abnormal inflammation, vascular dysfunction, and involvement of several microbial agents, such as commensal Demodex mites. The number of Demodex mites on normal skin flora is less than 5/cm²; however, the increased vascular dilation and capillary permeability associated with rosacea that result from sunlight and heat exposure increase the density of Demodex folliculorum.⁷ Elevated Demodex mite density has been observed in the lumens of the sebaceous follicles in patients with rosacea. However, because the severity of the clinical manifestations of the disease is not directly associated with the density of D folliculorum, it generally is accepted that D folliculorum is not a pathogenetic but rather an exacerbating factor.⁸ It has been reported that this species of mite is mostly found on the face and around the eyelashes and scalp of patients and that it can cause ocular surface inflammation.⁸

Most studies have researched ocular manifestations of rosacea but not ocular involvement in rosacea patients with and without Demodex mite infestation. In our study, we sought to compare the ocular surface, meibomian gland characteristics, and tear film abnormalities among patients with cutaneous rosacea with and without Demodex infestation.

Materials and Methods

We conducted a retrospective study of 60 patients with cutaneous rosacea. This study was approved by the ethics committee of the local hospital (2018/002-003), and all patients provided verbal and written informed consent before participating in the study. The study was carried out according to the guidelines of the Declaration of Helsinki.

Patient Selection and Evaluation
Patients diagnosed with rosacea by a dermatologist within 6 months were included in the study. Diagnosis of the disease was made after a detailed anamnesis and dermatologic examination. Rosacea was diagnosed if patients had an itching sensation, erythema and/or erythema attacks, and papules and pustules, and fulfilled the diagnostic criteria according to the National Rosacea Society. The skin disease was classified according to the subtypes as ETR, PPR, phymatous rosacea, or ocular rosacea.

The standard skin surface biopsy method was used in 60 patients for detecting Demodex density. When more than 5 mites were detected per square centimeter, the result was recorded as positive. Thirty consecutive, newly diagnosed patients with cutaneous acne rosacea with Demodex infestation and 30 consecutive, newly diagnosed sex- and age-matched patients with acne rosacea without Demodex infestation admitted to the dermatology outpatient clinic were included to this study. The patients who did not have any known dermatologic, systemic, or ocular diseases were included in the study. Patients who met any of the following criteria were excluded from the study: prior anti-inflammatory topical and/or systemic treatment for rosacea during the last 3 months, contact lens wear, eyelid surgery, or autoimmune disease requiring treatment.

 

Microscopic Demodex Examination
Demodex count was determined using a standardized skin surface biopsy, which is a noninvasive method. Every patient gave samples from the cheeks. This biopsy was repeated from the same site. A drop of cyanoacrylate was placed on a clean slide, pressed against a skin lesion, held in place for 1 minute, and removed. The obtained samples were evaluated under a light microscope (Nikon E200) with oil immersion. When more than 5 mites were detected per square centimeter, the result was recorded as positive.

Ophthalmologic Examination
A complete ophthalmologic examination including visual acuity assessment, standardized slit lamp examination, and fundus examination was done for all patients. Ocular rosacea was diagnosed on detection of 1 or more of the following: watery or bloodshot appearance, foreign-body sensation, burning or stinging, dryness, itching, light sensitivity, blurred vision, telangiectases of the conjunctiva and eyelid margin, eyelid lid and periocular erythema, anterior blepharitis, meibomian gland dysfunction, or irregularity of eyelid margins. All patients were screened for the signs and symptoms of ocular rosacea and underwent other ophthalmologic examinations, including tear function tests. Tear functions were evaluated with Schirmer tests without anesthesia and fluorescein tear breakup time (TBUT). Tear film breakup time was assessed after instillation of 2% fluorescein staining under a cobalt blue filter. The time interval between the last complete blink and the appearance of the first dry spot was recorded. The mean of 3 consecutive measurements was obtained. The Schirmer test was performed without topical anesthesia using a standardized filter strip (Bio-Tech Vision Care). The amount of wetting was measured after 5 minutes. Meibomian gland expressibility was assessed by applying digital pressure to the eyelid margin.

Statistical Analysis
Statistical analysis of the study was performed with SPSS Statistics Version 22.0 (SPSS Inc). Continuous variables were reported as mean (SD), and categorical variables were reported as percentages and counts. Descriptive statistics for numerical variables were created. An independent sample t test was used for normally distributed continuous variables. The Kolmogorov-Smirnov test was used to determine normality. The Schirmer test without anesthesia and TBUT values among groups were compared using one-way analysis of variance. The differences were calculated using the multiple comparison Tukey test. P<.05 was considered statistically significant.

Results

Demographic Characteristics of Rosacea Patients
Sixty eyes of 30 newly diagnosed patients with acne rosacea with Demodex infestation and 60 eyes of 30 newly diagnosed patients with acne rosacea without Demodex infestation were enrolled in this study. The mean age (SD) of the 60 patients was 37.63 (10.01) years. The mean TBUT (SD) of the 120 eyes was 6.65 (3.44) seconds, and the mean Schirmer score (SD) was 12.59 (6.71) mm (Table 1).

Meibomian Gland Dysfunction vs Subgroup of Rosacea Patients
Thirty-four (57%) patients had blepharitis, and 18 (30%) patients had meibomitis. Thirty-five (58.3%) patients had ETR, 5 (8.3%) patients had phymatous rosacea, and 20 (33.4%) patients had PPR (Table 2). Of the Demodex-negative patients, 73.3% (22/30) had ETR, 20% (6/30) had PPR, and 6.7% (2/30) had phymatous rosacea. Of the Demodex-positive patients, 43.3% (13/30) had ETR, 46.7% (14/30) had PPR, and 10% (3/30) had phymatous rosacea (Table 3). Papulopustular rosacea was found to be significantly associated with Demodex positivity (P=.003); neither ETR nor phymatous rosacea was found to be significantly associated with Demodex infestation (P=.66 and P=.13, respectively)(Table 3).

There was no statistically significant difference between the Demodex-negative and Demodex-positive groups for mean age (SD)(37.4 [11.54] years vs 37.87 [8.41] years; P=.85), mean TBUT (SD)(6.73 [3.62] seconds vs 6.57 [3.33] seconds; P=.85), and mean Schirmer score (SD)(13.68 [7.23] mm vs 11.5 [6.08] mm; P=.21)(Table 4).

Fifteen (50%) patients (30 eyes) in the Demodex-negative group and 19 (63.3%) patients (38 eyes) in the Demodex-positive group had blepharitis, with no statistically significant difference between the groups (P=.43). Seven (23.3%) patients (14 eyes) in the Demodex-negative group and 11 (36.7%) patients (22 eyes) in the Demodex-positive group had meibomitis, with no statistically significant difference between the groups (P=.39)(Table 3).

 

Sixteen (53.3%) patients (32 eyes) in the Demodex-negative group and 21 (70%) patients (42 eyes) in the Demodex-positive group had TBUT values less than 10 seconds. Eighteen (60%) patients (36 eyes) in the Demodex-negative group and 25 (83.3%) patients (50 eyes) in the Demodex-positive group had Schirmer scores less than 10 mm (Table 3). The 2 groups were not significantly different in dry eye findings (P=.25 and P=.29, respectively).

Comment

Inflammation in Rosacea
It is known that the density of nonfloral bacteria as well as D folliculorum and Demodex brevis increases in skin affected by rosacea compared to normal skin. Vascular dilation associated with rosacea that results from sunlight and heat causes increased capillary permeability and creates the ideal environment for the proliferation of D folliculorum. Demodex is thought to act as a vector for the activity of certain other microorganisms, particularly Bacillus oleronius, and thus initiates the inflammatory response associated with rosacea.⁹

One study reported that the inflammation associated with rosacea that was caused by Demodex and other environmental stimuli occurred through toll-like receptor 2 and various cytokines.¹⁰ It has been reported that the abnormal function of toll-like receptor 2 in the epidermis leads to the increased production of cathelicidin. Cathelicidin is an antimicrobial peptide with both vasoactive and proinflammatory activity and has been used as a basis to explain the pathogenesis of facial erythema, flushing, and telangiectasia in the context of rosacea.^11,12 In addition, it has been reported that the increased secretion of proinflammatory cytokines such as IL-1 and gelatinase B in ocular rosacea leads to tearing film abnormalities that result from increased bacterial flora in the eyelids, which subsequently leads to decreased tear drainage and dry eyes.¹³ In addition, B oleronius isolated from a D folliculorum mite from patients with PPR produced proteins that induced an inflammatory immune response in 73% (16/22) of patients with rosacea.¹⁴

Ocular Findings in Rosacea Patients
In our study, PPR was found to be significantly associated with Demodex positivity compared to ETR and phymatous rosacea (P=.003). However, ocular inflammation findings such as blepharitis and meibomitis were not significantly different between Demodex-positive and Demodex-negative patients. Although the mean Schirmer score of Demodex-positive patients was lower than Demodex-negative patients, this difference was not statistically significant. We evaluated a TBUT of less than 10 seconds and a Schirmer score less than 10 mm as dry eye. Accordingly, the number of patients with dry eye was higher in the Demodex-positive group, but this difference was not statistically significant.

Chronic blepharitis, conjunctival inflammation, and meibomian gland dysfunction are among the most common findings of ocular rosacea.^15,16 Patients with ocular rosacea commonly have dry eye and abnormal TBUT and Schirmer scores.¹⁷ In our study, we found that the fluorescein TBUT and Schirmer scores were more likely to be abnormal in the Demodex-positive group, but the difference between the 2 groups was not statistically significant.

It has been reported that proinflammatory cytokines due to a weakened immune system in rosacea patients were increased. The weakened immune system was further supported by the increased concentrations of proinflammatory cytokines such as IL-1 and matrix metalloproteinase 9 in these patients’ tears and the improvement of symptoms after the inhibition of these cytokines.¹¹ Luo et al¹⁸ reported that Demodex inflammation causes dry eye, particularly with D brevis. Ayyildiz and Sezgin¹⁹ reported that Schirmer scores were significantly lower and that the Ocular Surface Disease Index had significantly increased in the Demodex-positive group compared to the Demodex-negative group (P=.001 for both). A Korean study reported that Demodex density was correlated with age, sex, and TBUT results, but there was no significant relationship between Demodex density and Schirmer scores.¹⁶

Sobolewska et al²⁰ administered ivermectin cream 1% to 10 patients with cutaneous and ocular rosacea, but only to the forehead, chin, nose, cheeks, and regions close to the eyelids, and observed a significant improvement in blepharitis (P=.004). They stated that ivermectin, as applied only to the face, suppressed the proinflammatory cytokines associated with rosacea and showed anti-inflammatory effects by reducing Demodex mites.²⁰Li et al²¹ demonstrated a strong correlation between ocular Demodex inflammation and serum reactivity to these bacterial proteins in patients with ocular rosacea, and they found that eyelid margin inflammation and facial rosacea correlated with reactivity to these proteins. These studies suggest a possible role for Demodex infestation and bacterial proteins in the etiology of rosacea.

Gonzalez-Hinojosa et al²² demonstrated that even though eyelash blepharitis was more common in PPR than ETR, there was no statistically significant association between rosacea and Demodex blepharitis. In our study, we found a significant correlation between PPR and Demodex positivity. Also, meibomian gland dysfunction was more common in the Demodex-positive group; however, this result was not statistically significant. One study compared patients with primary demodicosis and patients with rosacea with Demodex-induced blepharitis to healthy controls and found that patients with primary demodicosis and patients with rosacea did not have significantly different ocular findings.²³ In contrast, Forton and De Maertelaer²⁴ reported that patients with PPR had significantly more severe ocular manifestations compared with patients with demodicosis (P=.004).

Mizuno et al²⁵ compared the normal (nonrosacea) population with and without Demodex-infested eyelashes and found that the 2 groups were not significantly different for meibomian gland dysfunction, fluorescein TBUT, or ocular surface discomfort.

Varying results have been reported regarding the association between Demodex and blepharitis or ocular surface discomfort with or without rosacea. In our study, we found that Demodex did not affect tear function tests or meibomian gland function in patients with rosacea. We believe this study is important because it demonstrates the effects of Demodex on ocular findings in patients with cutaneous rosacea.

Limitations
Our study has some limitations. The number of patients was relatively small, resulting in few significant differences between the comparison groups. A larger prospective research study is required to assess the prevalence of Demodex mites in the ocular rosacea population along with associated symptoms and findings.

Conclusion

Rosacea is a chronic disease associated with skin and ocular manifestations that range from mild to severe, that progresses in the form of attacks, and that requires long-term follow-up and treatment. Rosacea most often presents as a disease that causes ocular surface inflammation of varying degrees. Demodex infestation may increase cutaneous or ocular inflammation in rosacea. Therefore, every patient diagnosed with rosacea should be given a dermatologic examination to determine Demodex positivity and an ophthalmologic examination to determine ocular manifestations.

Acne rosacea is a chronic inflammatory disease that may affect the facial skin, eyes, and eyelids.¹ It is characterized by transient or persistent flushing, facial erythema, and telangiectases, generally located on the central portion of the face, and may progress to papules and pustules.^2,3 At the late stage of the disease, dermal edema or fibroplasia and sebaceous gland hypertrophy may cause phymatous alterations in the skin. In 2004, the National Rosacea Society Expert Committee developed a classification system for rosacea to standardize subtypes and variants that has since been widely accepted and continues to aid in research and epidemiologic studies.⁴ The committee defined 4 subtypes based on clinical characteristics: erythematotelangiectatic (ETR), papulopustular (PPR), phymatous, and ocular rosacea.^2,3

Ocular rosacea may accompany mild, moderate, and severe dermatologic disease or may occur in the absence of diagnostic skin disease.⁵ Ocular signs include eyelid margin telangiectasia, spade-shaped infiltrates in the cornea, scleritis, and sclerokeratitis. Common symptoms include burning, stinging, light sensitivity, and foreign-body sensation. Ocular signs commonly seen in rosacea are meibomian gland dysfunction characterized by inspissation and inflammation of the meibomian glands (chalazia), conjunctivitis, honey crust and cylindrical collarette accumulation at the base of the eyelashes, irregularity of the eyelid margin architecture, and evaporative tear dysfunction.^5,6

The physiopathology of rosacea is still unknown. Potential factors include genetic predisposition, abnormal inflammation, vascular dysfunction, and involvement of several microbial agents, such as commensal Demodex mites. The number of Demodex mites on normal skin flora is less than 5/cm²; however, the increased vascular dilation and capillary permeability associated with rosacea that result from sunlight and heat exposure increase the density of Demodex folliculorum.⁷ Elevated Demodex mite density has been observed in the lumens of the sebaceous follicles in patients with rosacea. However, because the severity of the clinical manifestations of the disease is not directly associated with the density of D folliculorum, it generally is accepted that D folliculorum is not a pathogenetic but rather an exacerbating factor.⁸ It has been reported that this species of mite is mostly found on the face and around the eyelashes and scalp of patients and that it can cause ocular surface inflammation.⁸

Most studies have researched ocular manifestations of rosacea but not ocular involvement in rosacea patients with and without Demodex mite infestation. In our study, we sought to compare the ocular surface, meibomian gland characteristics, and tear film abnormalities among patients with cutaneous rosacea with and without Demodex infestation.

Materials and Methods

We conducted a retrospective study of 60 patients with cutaneous rosacea. This study was approved by the ethics committee of the local hospital (2018/002-003), and all patients provided verbal and written informed consent before participating in the study. The study was carried out according to the guidelines of the Declaration of Helsinki.

Patient Selection and Evaluation
Patients diagnosed with rosacea by a dermatologist within 6 months were included in the study. Diagnosis of the disease was made after a detailed anamnesis and dermatologic examination. Rosacea was diagnosed if patients had an itching sensation, erythema and/or erythema attacks, and papules and pustules, and fulfilled the diagnostic criteria according to the National Rosacea Society. The skin disease was classified according to the subtypes as ETR, PPR, phymatous rosacea, or ocular rosacea.

The standard skin surface biopsy method was used in 60 patients for detecting Demodex density. When more than 5 mites were detected per square centimeter, the result was recorded as positive. Thirty consecutive, newly diagnosed patients with cutaneous acne rosacea with Demodex infestation and 30 consecutive, newly diagnosed sex- and age-matched patients with acne rosacea without Demodex infestation admitted to the dermatology outpatient clinic were included to this study. The patients who did not have any known dermatologic, systemic, or ocular diseases were included in the study. Patients who met any of the following criteria were excluded from the study: prior anti-inflammatory topical and/or systemic treatment for rosacea during the last 3 months, contact lens wear, eyelid surgery, or autoimmune disease requiring treatment.

 

Microscopic Demodex Examination
Demodex count was determined using a standardized skin surface biopsy, which is a noninvasive method. Every patient gave samples from the cheeks. This biopsy was repeated from the same site. A drop of cyanoacrylate was placed on a clean slide, pressed against a skin lesion, held in place for 1 minute, and removed. The obtained samples were evaluated under a light microscope (Nikon E200) with oil immersion. When more than 5 mites were detected per square centimeter, the result was recorded as positive.

Ophthalmologic Examination
A complete ophthalmologic examination including visual acuity assessment, standardized slit lamp examination, and fundus examination was done for all patients. Ocular rosacea was diagnosed on detection of 1 or more of the following: watery or bloodshot appearance, foreign-body sensation, burning or stinging, dryness, itching, light sensitivity, blurred vision, telangiectases of the conjunctiva and eyelid margin, eyelid lid and periocular erythema, anterior blepharitis, meibomian gland dysfunction, or irregularity of eyelid margins. All patients were screened for the signs and symptoms of ocular rosacea and underwent other ophthalmologic examinations, including tear function tests. Tear functions were evaluated with Schirmer tests without anesthesia and fluorescein tear breakup time (TBUT). Tear film breakup time was assessed after instillation of 2% fluorescein staining under a cobalt blue filter. The time interval between the last complete blink and the appearance of the first dry spot was recorded. The mean of 3 consecutive measurements was obtained. The Schirmer test was performed without topical anesthesia using a standardized filter strip (Bio-Tech Vision Care). The amount of wetting was measured after 5 minutes. Meibomian gland expressibility was assessed by applying digital pressure to the eyelid margin.

Statistical Analysis
Statistical analysis of the study was performed with SPSS Statistics Version 22.0 (SPSS Inc). Continuous variables were reported as mean (SD), and categorical variables were reported as percentages and counts. Descriptive statistics for numerical variables were created. An independent sample t test was used for normally distributed continuous variables. The Kolmogorov-Smirnov test was used to determine normality. The Schirmer test without anesthesia and TBUT values among groups were compared using one-way analysis of variance. The differences were calculated using the multiple comparison Tukey test. P<.05 was considered statistically significant.

Results

Demographic Characteristics of Rosacea Patients
Sixty eyes of 30 newly diagnosed patients with acne rosacea with Demodex infestation and 60 eyes of 30 newly diagnosed patients with acne rosacea without Demodex infestation were enrolled in this study. The mean age (SD) of the 60 patients was 37.63 (10.01) years. The mean TBUT (SD) of the 120 eyes was 6.65 (3.44) seconds, and the mean Schirmer score (SD) was 12.59 (6.71) mm (Table 1).

Meibomian Gland Dysfunction vs Subgroup of Rosacea Patients
Thirty-four (57%) patients had blepharitis, and 18 (30%) patients had meibomitis. Thirty-five (58.3%) patients had ETR, 5 (8.3%) patients had phymatous rosacea, and 20 (33.4%) patients had PPR (Table 2). Of the Demodex-negative patients, 73.3% (22/30) had ETR, 20% (6/30) had PPR, and 6.7% (2/30) had phymatous rosacea. Of the Demodex-positive patients, 43.3% (13/30) had ETR, 46.7% (14/30) had PPR, and 10% (3/30) had phymatous rosacea (Table 3). Papulopustular rosacea was found to be significantly associated with Demodex positivity (P=.003); neither ETR nor phymatous rosacea was found to be significantly associated with Demodex infestation (P=.66 and P=.13, respectively)(Table 3).

There was no statistically significant difference between the Demodex-negative and Demodex-positive groups for mean age (SD)(37.4 [11.54] years vs 37.87 [8.41] years; P=.85), mean TBUT (SD)(6.73 [3.62] seconds vs 6.57 [3.33] seconds; P=.85), and mean Schirmer score (SD)(13.68 [7.23] mm vs 11.5 [6.08] mm; P=.21)(Table 4).

Fifteen (50%) patients (30 eyes) in the Demodex-negative group and 19 (63.3%) patients (38 eyes) in the Demodex-positive group had blepharitis, with no statistically significant difference between the groups (P=.43). Seven (23.3%) patients (14 eyes) in the Demodex-negative group and 11 (36.7%) patients (22 eyes) in the Demodex-positive group had meibomitis, with no statistically significant difference between the groups (P=.39)(Table 3).

 

Sixteen (53.3%) patients (32 eyes) in the Demodex-negative group and 21 (70%) patients (42 eyes) in the Demodex-positive group had TBUT values less than 10 seconds. Eighteen (60%) patients (36 eyes) in the Demodex-negative group and 25 (83.3%) patients (50 eyes) in the Demodex-positive group had Schirmer scores less than 10 mm (Table 3). The 2 groups were not significantly different in dry eye findings (P=.25 and P=.29, respectively).

Comment

Inflammation in Rosacea
It is known that the density of nonfloral bacteria as well as D folliculorum and Demodex brevis increases in skin affected by rosacea compared to normal skin. Vascular dilation associated with rosacea that results from sunlight and heat causes increased capillary permeability and creates the ideal environment for the proliferation of D folliculorum. Demodex is thought to act as a vector for the activity of certain other microorganisms, particularly Bacillus oleronius, and thus initiates the inflammatory response associated with rosacea.⁹

One study reported that the inflammation associated with rosacea that was caused by Demodex and other environmental stimuli occurred through toll-like receptor 2 and various cytokines.¹⁰ It has been reported that the abnormal function of toll-like receptor 2 in the epidermis leads to the increased production of cathelicidin. Cathelicidin is an antimicrobial peptide with both vasoactive and proinflammatory activity and has been used as a basis to explain the pathogenesis of facial erythema, flushing, and telangiectasia in the context of rosacea.^11,12 In addition, it has been reported that the increased secretion of proinflammatory cytokines such as IL-1 and gelatinase B in ocular rosacea leads to tearing film abnormalities that result from increased bacterial flora in the eyelids, which subsequently leads to decreased tear drainage and dry eyes.¹³ In addition, B oleronius isolated from a D folliculorum mite from patients with PPR produced proteins that induced an inflammatory immune response in 73% (16/22) of patients with rosacea.¹⁴

Ocular Findings in Rosacea Patients
In our study, PPR was found to be significantly associated with Demodex positivity compared to ETR and phymatous rosacea (P=.003). However, ocular inflammation findings such as blepharitis and meibomitis were not significantly different between Demodex-positive and Demodex-negative patients. Although the mean Schirmer score of Demodex-positive patients was lower than Demodex-negative patients, this difference was not statistically significant. We evaluated a TBUT of less than 10 seconds and a Schirmer score less than 10 mm as dry eye. Accordingly, the number of patients with dry eye was higher in the Demodex-positive group, but this difference was not statistically significant.

Chronic blepharitis, conjunctival inflammation, and meibomian gland dysfunction are among the most common findings of ocular rosacea.^15,16 Patients with ocular rosacea commonly have dry eye and abnormal TBUT and Schirmer scores.¹⁷ In our study, we found that the fluorescein TBUT and Schirmer scores were more likely to be abnormal in the Demodex-positive group, but the difference between the 2 groups was not statistically significant.

It has been reported that proinflammatory cytokines due to a weakened immune system in rosacea patients were increased. The weakened immune system was further supported by the increased concentrations of proinflammatory cytokines such as IL-1 and matrix metalloproteinase 9 in these patients’ tears and the improvement of symptoms after the inhibition of these cytokines.¹¹ Luo et al¹⁸ reported that Demodex inflammation causes dry eye, particularly with D brevis. Ayyildiz and Sezgin¹⁹ reported that Schirmer scores were significantly lower and that the Ocular Surface Disease Index had significantly increased in the Demodex-positive group compared to the Demodex-negative group (P=.001 for both). A Korean study reported that Demodex density was correlated with age, sex, and TBUT results, but there was no significant relationship between Demodex density and Schirmer scores.¹⁶

Sobolewska et al²⁰ administered ivermectin cream 1% to 10 patients with cutaneous and ocular rosacea, but only to the forehead, chin, nose, cheeks, and regions close to the eyelids, and observed a significant improvement in blepharitis (P=.004). They stated that ivermectin, as applied only to the face, suppressed the proinflammatory cytokines associated with rosacea and showed anti-inflammatory effects by reducing Demodex mites.²⁰Li et al²¹ demonstrated a strong correlation between ocular Demodex inflammation and serum reactivity to these bacterial proteins in patients with ocular rosacea, and they found that eyelid margin inflammation and facial rosacea correlated with reactivity to these proteins. These studies suggest a possible role for Demodex infestation and bacterial proteins in the etiology of rosacea.

Gonzalez-Hinojosa et al²² demonstrated that even though eyelash blepharitis was more common in PPR than ETR, there was no statistically significant association between rosacea and Demodex blepharitis. In our study, we found a significant correlation between PPR and Demodex positivity. Also, meibomian gland dysfunction was more common in the Demodex-positive group; however, this result was not statistically significant. One study compared patients with primary demodicosis and patients with rosacea with Demodex-induced blepharitis to healthy controls and found that patients with primary demodicosis and patients with rosacea did not have significantly different ocular findings.²³ In contrast, Forton and De Maertelaer²⁴ reported that patients with PPR had significantly more severe ocular manifestations compared with patients with demodicosis (P=.004).

Mizuno et al²⁵ compared the normal (nonrosacea) population with and without Demodex-infested eyelashes and found that the 2 groups were not significantly different for meibomian gland dysfunction, fluorescein TBUT, or ocular surface discomfort.

Varying results have been reported regarding the association between Demodex and blepharitis or ocular surface discomfort with or without rosacea. In our study, we found that Demodex did not affect tear function tests or meibomian gland function in patients with rosacea. We believe this study is important because it demonstrates the effects of Demodex on ocular findings in patients with cutaneous rosacea.

Limitations
Our study has some limitations. The number of patients was relatively small, resulting in few significant differences between the comparison groups. A larger prospective research study is required to assess the prevalence of Demodex mites in the ocular rosacea population along with associated symptoms and findings.

Conclusion

Rosacea is a chronic disease associated with skin and ocular manifestations that range from mild to severe, that progresses in the form of attacks, and that requires long-term follow-up and treatment. Rosacea most often presents as a disease that causes ocular surface inflammation of varying degrees. Demodex infestation may increase cutaneous or ocular inflammation in rosacea. Therefore, every patient diagnosed with rosacea should be given a dermatologic examination to determine Demodex positivity and an ophthalmologic examination to determine ocular manifestations.

Acne rosacea is a chronic inflammatory disease that may affect the facial skin, eyes, and eyelids.¹ It is characterized by transient or persistent flushing, facial erythema, and telangiectases, generally located on the central portion of the face, and may progress to papules and pustules.^2,3 At the late stage of the disease, dermal edema or fibroplasia and sebaceous gland hypertrophy may cause phymatous alterations in the skin. In 2004, the National Rosacea Society Expert Committee developed a classification system for rosacea to standardize subtypes and variants that has since been widely accepted and continues to aid in research and epidemiologic studies.⁴ The committee defined 4 subtypes based on clinical characteristics: erythematotelangiectatic (ETR), papulopustular (PPR), phymatous, and ocular rosacea.^2,3

Ocular rosacea may accompany mild, moderate, and severe dermatologic disease or may occur in the absence of diagnostic skin disease.⁵ Ocular signs include eyelid margin telangiectasia, spade-shaped infiltrates in the cornea, scleritis, and sclerokeratitis. Common symptoms include burning, stinging, light sensitivity, and foreign-body sensation. Ocular signs commonly seen in rosacea are meibomian gland dysfunction characterized by inspissation and inflammation of the meibomian glands (chalazia), conjunctivitis, honey crust and cylindrical collarette accumulation at the base of the eyelashes, irregularity of the eyelid margin architecture, and evaporative tear dysfunction.^5,6

The physiopathology of rosacea is still unknown. Potential factors include genetic predisposition, abnormal inflammation, vascular dysfunction, and involvement of several microbial agents, such as commensal Demodex mites. The number of Demodex mites on normal skin flora is less than 5/cm²; however, the increased vascular dilation and capillary permeability associated with rosacea that result from sunlight and heat exposure increase the density of Demodex folliculorum.⁷ Elevated Demodex mite density has been observed in the lumens of the sebaceous follicles in patients with rosacea. However, because the severity of the clinical manifestations of the disease is not directly associated with the density of D folliculorum, it generally is accepted that D folliculorum is not a pathogenetic but rather an exacerbating factor.⁸ It has been reported that this species of mite is mostly found on the face and around the eyelashes and scalp of patients and that it can cause ocular surface inflammation.⁸

Most studies have researched ocular manifestations of rosacea but not ocular involvement in rosacea patients with and without Demodex mite infestation. In our study, we sought to compare the ocular surface, meibomian gland characteristics, and tear film abnormalities among patients with cutaneous rosacea with and without Demodex infestation.

Materials and Methods

We conducted a retrospective study of 60 patients with cutaneous rosacea. This study was approved by the ethics committee of the local hospital (2018/002-003), and all patients provided verbal and written informed consent before participating in the study. The study was carried out according to the guidelines of the Declaration of Helsinki.

Patient Selection and Evaluation
Patients diagnosed with rosacea by a dermatologist within 6 months were included in the study. Diagnosis of the disease was made after a detailed anamnesis and dermatologic examination. Rosacea was diagnosed if patients had an itching sensation, erythema and/or erythema attacks, and papules and pustules, and fulfilled the diagnostic criteria according to the National Rosacea Society. The skin disease was classified according to the subtypes as ETR, PPR, phymatous rosacea, or ocular rosacea.

The standard skin surface biopsy method was used in 60 patients for detecting Demodex density. When more than 5 mites were detected per square centimeter, the result was recorded as positive. Thirty consecutive, newly diagnosed patients with cutaneous acne rosacea with Demodex infestation and 30 consecutive, newly diagnosed sex- and age-matched patients with acne rosacea without Demodex infestation admitted to the dermatology outpatient clinic were included to this study. The patients who did not have any known dermatologic, systemic, or ocular diseases were included in the study. Patients who met any of the following criteria were excluded from the study: prior anti-inflammatory topical and/or systemic treatment for rosacea during the last 3 months, contact lens wear, eyelid surgery, or autoimmune disease requiring treatment.

 

Microscopic Demodex Examination
Demodex count was determined using a standardized skin surface biopsy, which is a noninvasive method. Every patient gave samples from the cheeks. This biopsy was repeated from the same site. A drop of cyanoacrylate was placed on a clean slide, pressed against a skin lesion, held in place for 1 minute, and removed. The obtained samples were evaluated under a light microscope (Nikon E200) with oil immersion. When more than 5 mites were detected per square centimeter, the result was recorded as positive.

Ophthalmologic Examination
A complete ophthalmologic examination including visual acuity assessment, standardized slit lamp examination, and fundus examination was done for all patients. Ocular rosacea was diagnosed on detection of 1 or more of the following: watery or bloodshot appearance, foreign-body sensation, burning or stinging, dryness, itching, light sensitivity, blurred vision, telangiectases of the conjunctiva and eyelid margin, eyelid lid and periocular erythema, anterior blepharitis, meibomian gland dysfunction, or irregularity of eyelid margins. All patients were screened for the signs and symptoms of ocular rosacea and underwent other ophthalmologic examinations, including tear function tests. Tear functions were evaluated with Schirmer tests without anesthesia and fluorescein tear breakup time (TBUT). Tear film breakup time was assessed after instillation of 2% fluorescein staining under a cobalt blue filter. The time interval between the last complete blink and the appearance of the first dry spot was recorded. The mean of 3 consecutive measurements was obtained. The Schirmer test was performed without topical anesthesia using a standardized filter strip (Bio-Tech Vision Care). The amount of wetting was measured after 5 minutes. Meibomian gland expressibility was assessed by applying digital pressure to the eyelid margin.

Statistical Analysis
Statistical analysis of the study was performed with SPSS Statistics Version 22.0 (SPSS Inc). Continuous variables were reported as mean (SD), and categorical variables were reported as percentages and counts. Descriptive statistics for numerical variables were created. An independent sample t test was used for normally distributed continuous variables. The Kolmogorov-Smirnov test was used to determine normality. The Schirmer test without anesthesia and TBUT values among groups were compared using one-way analysis of variance. The differences were calculated using the multiple comparison Tukey test. P<.05 was considered statistically significant.

Results

Demographic Characteristics of Rosacea Patients
Sixty eyes of 30 newly diagnosed patients with acne rosacea with Demodex infestation and 60 eyes of 30 newly diagnosed patients with acne rosacea without Demodex infestation were enrolled in this study. The mean age (SD) of the 60 patients was 37.63 (10.01) years. The mean TBUT (SD) of the 120 eyes was 6.65 (3.44) seconds, and the mean Schirmer score (SD) was 12.59 (6.71) mm (Table 1).

Meibomian Gland Dysfunction vs Subgroup of Rosacea Patients
Thirty-four (57%) patients had blepharitis, and 18 (30%) patients had meibomitis. Thirty-five (58.3%) patients had ETR, 5 (8.3%) patients had phymatous rosacea, and 20 (33.4%) patients had PPR (Table 2). Of the Demodex-negative patients, 73.3% (22/30) had ETR, 20% (6/30) had PPR, and 6.7% (2/30) had phymatous rosacea. Of the Demodex-positive patients, 43.3% (13/30) had ETR, 46.7% (14/30) had PPR, and 10% (3/30) had phymatous rosacea (Table 3). Papulopustular rosacea was found to be significantly associated with Demodex positivity (P=.003); neither ETR nor phymatous rosacea was found to be significantly associated with Demodex infestation (P=.66 and P=.13, respectively)(Table 3).

There was no statistically significant difference between the Demodex-negative and Demodex-positive groups for mean age (SD)(37.4 [11.54] years vs 37.87 [8.41] years; P=.85), mean TBUT (SD)(6.73 [3.62] seconds vs 6.57 [3.33] seconds; P=.85), and mean Schirmer score (SD)(13.68 [7.23] mm vs 11.5 [6.08] mm; P=.21)(Table 4).

Fifteen (50%) patients (30 eyes) in the Demodex-negative group and 19 (63.3%) patients (38 eyes) in the Demodex-positive group had blepharitis, with no statistically significant difference between the groups (P=.43). Seven (23.3%) patients (14 eyes) in the Demodex-negative group and 11 (36.7%) patients (22 eyes) in the Demodex-positive group had meibomitis, with no statistically significant difference between the groups (P=.39)(Table 3).

 

Sixteen (53.3%) patients (32 eyes) in the Demodex-negative group and 21 (70%) patients (42 eyes) in the Demodex-positive group had TBUT values less than 10 seconds. Eighteen (60%) patients (36 eyes) in the Demodex-negative group and 25 (83.3%) patients (50 eyes) in the Demodex-positive group had Schirmer scores less than 10 mm (Table 3). The 2 groups were not significantly different in dry eye findings (P=.25 and P=.29, respectively).

Comment

Inflammation in Rosacea
It is known that the density of nonfloral bacteria as well as D folliculorum and Demodex brevis increases in skin affected by rosacea compared to normal skin. Vascular dilation associated with rosacea that results from sunlight and heat causes increased capillary permeability and creates the ideal environment for the proliferation of D folliculorum. Demodex is thought to act as a vector for the activity of certain other microorganisms, particularly Bacillus oleronius, and thus initiates the inflammatory response associated with rosacea.⁹

One study reported that the inflammation associated with rosacea that was caused by Demodex and other environmental stimuli occurred through toll-like receptor 2 and various cytokines.¹⁰ It has been reported that the abnormal function of toll-like receptor 2 in the epidermis leads to the increased production of cathelicidin. Cathelicidin is an antimicrobial peptide with both vasoactive and proinflammatory activity and has been used as a basis to explain the pathogenesis of facial erythema, flushing, and telangiectasia in the context of rosacea.^11,12 In addition, it has been reported that the increased secretion of proinflammatory cytokines such as IL-1 and gelatinase B in ocular rosacea leads to tearing film abnormalities that result from increased bacterial flora in the eyelids, which subsequently leads to decreased tear drainage and dry eyes.¹³ In addition, B oleronius isolated from a D folliculorum mite from patients with PPR produced proteins that induced an inflammatory immune response in 73% (16/22) of patients with rosacea.¹⁴

Ocular Findings in Rosacea Patients
In our study, PPR was found to be significantly associated with Demodex positivity compared to ETR and phymatous rosacea (P=.003). However, ocular inflammation findings such as blepharitis and meibomitis were not significantly different between Demodex-positive and Demodex-negative patients. Although the mean Schirmer score of Demodex-positive patients was lower than Demodex-negative patients, this difference was not statistically significant. We evaluated a TBUT of less than 10 seconds and a Schirmer score less than 10 mm as dry eye. Accordingly, the number of patients with dry eye was higher in the Demodex-positive group, but this difference was not statistically significant.

Chronic blepharitis, conjunctival inflammation, and meibomian gland dysfunction are among the most common findings of ocular rosacea.^15,16 Patients with ocular rosacea commonly have dry eye and abnormal TBUT and Schirmer scores.¹⁷ In our study, we found that the fluorescein TBUT and Schirmer scores were more likely to be abnormal in the Demodex-positive group, but the difference between the 2 groups was not statistically significant.

It has been reported that proinflammatory cytokines due to a weakened immune system in rosacea patients were increased. The weakened immune system was further supported by the increased concentrations of proinflammatory cytokines such as IL-1 and matrix metalloproteinase 9 in these patients’ tears and the improvement of symptoms after the inhibition of these cytokines.¹¹ Luo et al¹⁸ reported that Demodex inflammation causes dry eye, particularly with D brevis. Ayyildiz and Sezgin¹⁹ reported that Schirmer scores were significantly lower and that the Ocular Surface Disease Index had significantly increased in the Demodex-positive group compared to the Demodex-negative group (P=.001 for both). A Korean study reported that Demodex density was correlated with age, sex, and TBUT results, but there was no significant relationship between Demodex density and Schirmer scores.¹⁶

Sobolewska et al²⁰ administered ivermectin cream 1% to 10 patients with cutaneous and ocular rosacea, but only to the forehead, chin, nose, cheeks, and regions close to the eyelids, and observed a significant improvement in blepharitis (P=.004). They stated that ivermectin, as applied only to the face, suppressed the proinflammatory cytokines associated with rosacea and showed anti-inflammatory effects by reducing Demodex mites.²⁰Li et al²¹ demonstrated a strong correlation between ocular Demodex inflammation and serum reactivity to these bacterial proteins in patients with ocular rosacea, and they found that eyelid margin inflammation and facial rosacea correlated with reactivity to these proteins. These studies suggest a possible role for Demodex infestation and bacterial proteins in the etiology of rosacea.

Gonzalez-Hinojosa et al²² demonstrated that even though eyelash blepharitis was more common in PPR than ETR, there was no statistically significant association between rosacea and Demodex blepharitis. In our study, we found a significant correlation between PPR and Demodex positivity. Also, meibomian gland dysfunction was more common in the Demodex-positive group; however, this result was not statistically significant. One study compared patients with primary demodicosis and patients with rosacea with Demodex-induced blepharitis to healthy controls and found that patients with primary demodicosis and patients with rosacea did not have significantly different ocular findings.²³ In contrast, Forton and De Maertelaer²⁴ reported that patients with PPR had significantly more severe ocular manifestations compared with patients with demodicosis (P=.004).

Mizuno et al²⁵ compared the normal (nonrosacea) population with and without Demodex-infested eyelashes and found that the 2 groups were not significantly different for meibomian gland dysfunction, fluorescein TBUT, or ocular surface discomfort.

Varying results have been reported regarding the association between Demodex and blepharitis or ocular surface discomfort with or without rosacea. In our study, we found that Demodex did not affect tear function tests or meibomian gland function in patients with rosacea. We believe this study is important because it demonstrates the effects of Demodex on ocular findings in patients with cutaneous rosacea.

Limitations
Our study has some limitations. The number of patients was relatively small, resulting in few significant differences between the comparison groups. A larger prospective research study is required to assess the prevalence of Demodex mites in the ocular rosacea population along with associated symptoms and findings.

Conclusion

Rosacea is a chronic disease associated with skin and ocular manifestations that range from mild to severe, that progresses in the form of attacks, and that requires long-term follow-up and treatment. Rosacea most often presents as a disease that causes ocular surface inflammation of varying degrees. Demodex infestation may increase cutaneous or ocular inflammation in rosacea. Therefore, every patient diagnosed with rosacea should be given a dermatologic examination to determine Demodex positivity and an ophthalmologic examination to determine ocular manifestations.

Intraosseous (IO) access can afford a lifesaving means of vascular access in emergency settings, as it allows for the administration of large volumes of fluids, blood products, and medications at high flow rates directly into the highly vascularized osseous medullary cavity.¹ Fortunately, the complication rate with this resuscitative effort is low, with many reports demonstrating complication rates of less than 1%.² The most commonly reported complications include fluid extravasation, osteomyelitis, traumatic bone fracture, and epiphyseal plate damage.^1-3 Although compartment syndrome and skin necrosis have been reported,^4,5 there is no comprehensive list of sequelae resulting from fluid extravasation in the literature, and there are no known studies examining the incidence and types of cutaneous complications. In this study, we sought to evaluate the dermatologic impacts of this procedure.

Methods

We performed a retrospective chart review approved by the institutional review board at a large metropolitan level I trauma center in the Midwestern United States spanning 18 consecutive months to identify all patients who underwent IO line placement, either en route to or upon arrival at the trauma center. The electronic medical records of 113 patients (age range, 10 days–94 years) were identified using either an automated natural language look-up program with keywords including intraosseous access and IO or a Current Procedural Terminology code 36680. Data including patient age, reason for IO insertion, anatomic location of the IO, and complications secondary to IO line placement were recorded.

Results

We identified an overall complication rate of 2.7% (3/113), with only 1 patient showing isolated cutaneous complications from IO line placement. The complications in the first 2 patients included compartment syndrome following IO line placement in the right tibia and needle breakage during IO line placement. The third patient, a 30-year-old heart transplant recipient, developed tense bullae on the left leg 5 days after a resuscitative effort required IO access through the bilateral tibiae. The patient had received vasopressors as well as 750 mL of normal saline through these access points. Two days after resuscitation, she developed an enlarging ecchymosis around the left IO access point. On day 5, cutaneous findings included 2 large firm bullae on a purpuric base overlying the left proximal tibia and patella (Figure 1). After an ultrasound revealed no connection to the underlying joint space, the bullae were incised and drained for patient comfort, as she reported pain and undue pressure at the knee joint on initial dermatologic consultation. These symptoms abated after the procedure. Cultures of bullous fluid were negative for infection. Histopathologic examination revealed a subepidermal split with underlying minimal mixed inflammation, favoring the diagnosis of traumatic bullae.

At a scheduled 7-month dermatology follow-up, the wound bed appeared to be healing well with surrounding scarring with no residual bleeding or drainage (Figure 2) despite the patient reporting a protracted course of wound healing requiring debridement due to eschar formation and multiple follow-up appointments with the wound care service.

Comment

The most commonly reported complications with IO line placement result from fluid infiltration of the subcutaneous tissue secondary to catheter misplacement.^1,3 Extravasated fluid may lead to tissue damage, compartment syndrome, and even tissue necrosis in some cases.^1,4,5 Localized cellulitis and the formation of subcutaneous abscesses also have been reported, albeit rarely.^3,5

In our retrospective cohort review, we identified an additional potential complication of IO line placement that has not been widely reported—development of large traumatic bullae. It is most likely that this patient’s IO catheter became dislodged, resulting in extravasation of fluids into the dermal and subcutaneous tissues.

Our findings support the previously noted complication rate of less than 1% following IO line placement, with an overall complication rate of 2.7% that included only 1 patient with a cutaneous complication.² Given this low incidence, providers may not be used to recognizing such complications, leading to delayed or incorrect diagnosis of these entities. While there are certain conditions in which IO insertion is contraindicated, including severe bone diseases (eg, osteogenesis imperfecta, osteomyelitis), overlying cellulitis, and bone fracture, these conditions are rare and can be avoided in most cases by use of an alternative site for needle insertion.² Due to the widespread utility of this tool and its few contraindications, its use in hospitalized patients is rapidly increasing, necessitating a need for quick recognition of potential complications.

From previous data on the incidence of traumatic blisters with underlying bone fractures, there are several identifiable risk factors that could be extended to patients at high risk for developing cutaneous IO complications secondary to the trauma associated with needle insertion,⁶ including wound-healing impairments in patients with fragile lymphatics, peripheral vascular disease, diabetes, or collagen vascular diseases (eg, lupus, rheumatoid arthritis, Sjögren syndrome). Patients with these conditions should be closely monitored for the development of bullae.⁶ While the patient we highlighted in our study did not have a history of such conditions, her history of cardiac disease, recent resuscitation attempts, and immunosuppression certainly could have contributed to suboptimal tissue agility and repair after IO line placement.

Conclusion

Intraosseous access is a safe, effective, and reliable option for vascular access in both pediatric and adult populations that is widely used in both prehospital (ie, paramedic administered) and hospital settings, including intensive care units, emergency departments, and any acute situation where rapid vascular access is necessary. This retrospective chart review examining the incidence and types of cutaneous complications associated with IO line placement at a level I trauma center revealed a total complication rate similar to those reported in previous studies and also highlighted a unique postprocedural cutaneous finding of traumatic bullae. Although no unified management recommendations currently exist, providers should consider this complication in the differential for hospitalized patients with large, atypical, asymmetric bullae in the absence of an alternative explanation for such skin findings.

Intraosseous (IO) access can afford a lifesaving means of vascular access in emergency settings, as it allows for the administration of large volumes of fluids, blood products, and medications at high flow rates directly into the highly vascularized osseous medullary cavity.¹ Fortunately, the complication rate with this resuscitative effort is low, with many reports demonstrating complication rates of less than 1%.² The most commonly reported complications include fluid extravasation, osteomyelitis, traumatic bone fracture, and epiphyseal plate damage.^1-3 Although compartment syndrome and skin necrosis have been reported,^4,5 there is no comprehensive list of sequelae resulting from fluid extravasation in the literature, and there are no known studies examining the incidence and types of cutaneous complications. In this study, we sought to evaluate the dermatologic impacts of this procedure.

Methods

We performed a retrospective chart review approved by the institutional review board at a large metropolitan level I trauma center in the Midwestern United States spanning 18 consecutive months to identify all patients who underwent IO line placement, either en route to or upon arrival at the trauma center. The electronic medical records of 113 patients (age range, 10 days–94 years) were identified using either an automated natural language look-up program with keywords including intraosseous access and IO or a Current Procedural Terminology code 36680. Data including patient age, reason for IO insertion, anatomic location of the IO, and complications secondary to IO line placement were recorded.

Results

We identified an overall complication rate of 2.7% (3/113), with only 1 patient showing isolated cutaneous complications from IO line placement. The complications in the first 2 patients included compartment syndrome following IO line placement in the right tibia and needle breakage during IO line placement. The third patient, a 30-year-old heart transplant recipient, developed tense bullae on the left leg 5 days after a resuscitative effort required IO access through the bilateral tibiae. The patient had received vasopressors as well as 750 mL of normal saline through these access points. Two days after resuscitation, she developed an enlarging ecchymosis around the left IO access point. On day 5, cutaneous findings included 2 large firm bullae on a purpuric base overlying the left proximal tibia and patella (Figure 1). After an ultrasound revealed no connection to the underlying joint space, the bullae were incised and drained for patient comfort, as she reported pain and undue pressure at the knee joint on initial dermatologic consultation. These symptoms abated after the procedure. Cultures of bullous fluid were negative for infection. Histopathologic examination revealed a subepidermal split with underlying minimal mixed inflammation, favoring the diagnosis of traumatic bullae.

At a scheduled 7-month dermatology follow-up, the wound bed appeared to be healing well with surrounding scarring with no residual bleeding or drainage (Figure 2) despite the patient reporting a protracted course of wound healing requiring debridement due to eschar formation and multiple follow-up appointments with the wound care service.

Comment

The most commonly reported complications with IO line placement result from fluid infiltration of the subcutaneous tissue secondary to catheter misplacement.^1,3 Extravasated fluid may lead to tissue damage, compartment syndrome, and even tissue necrosis in some cases.^1,4,5 Localized cellulitis and the formation of subcutaneous abscesses also have been reported, albeit rarely.^3,5

In our retrospective cohort review, we identified an additional potential complication of IO line placement that has not been widely reported—development of large traumatic bullae. It is most likely that this patient’s IO catheter became dislodged, resulting in extravasation of fluids into the dermal and subcutaneous tissues.

Our findings support the previously noted complication rate of less than 1% following IO line placement, with an overall complication rate of 2.7% that included only 1 patient with a cutaneous complication.² Given this low incidence, providers may not be used to recognizing such complications, leading to delayed or incorrect diagnosis of these entities. While there are certain conditions in which IO insertion is contraindicated, including severe bone diseases (eg, osteogenesis imperfecta, osteomyelitis), overlying cellulitis, and bone fracture, these conditions are rare and can be avoided in most cases by use of an alternative site for needle insertion.² Due to the widespread utility of this tool and its few contraindications, its use in hospitalized patients is rapidly increasing, necessitating a need for quick recognition of potential complications.

From previous data on the incidence of traumatic blisters with underlying bone fractures, there are several identifiable risk factors that could be extended to patients at high risk for developing cutaneous IO complications secondary to the trauma associated with needle insertion,⁶ including wound-healing impairments in patients with fragile lymphatics, peripheral vascular disease, diabetes, or collagen vascular diseases (eg, lupus, rheumatoid arthritis, Sjögren syndrome). Patients with these conditions should be closely monitored for the development of bullae.⁶ While the patient we highlighted in our study did not have a history of such conditions, her history of cardiac disease, recent resuscitation attempts, and immunosuppression certainly could have contributed to suboptimal tissue agility and repair after IO line placement.

Conclusion

Intraosseous access is a safe, effective, and reliable option for vascular access in both pediatric and adult populations that is widely used in both prehospital (ie, paramedic administered) and hospital settings, including intensive care units, emergency departments, and any acute situation where rapid vascular access is necessary. This retrospective chart review examining the incidence and types of cutaneous complications associated with IO line placement at a level I trauma center revealed a total complication rate similar to those reported in previous studies and also highlighted a unique postprocedural cutaneous finding of traumatic bullae. Although no unified management recommendations currently exist, providers should consider this complication in the differential for hospitalized patients with large, atypical, asymmetric bullae in the absence of an alternative explanation for such skin findings.

Intraosseous (IO) access can afford a lifesaving means of vascular access in emergency settings, as it allows for the administration of large volumes of fluids, blood products, and medications at high flow rates directly into the highly vascularized osseous medullary cavity.¹ Fortunately, the complication rate with this resuscitative effort is low, with many reports demonstrating complication rates of less than 1%.² The most commonly reported complications include fluid extravasation, osteomyelitis, traumatic bone fracture, and epiphyseal plate damage.^1-3 Although compartment syndrome and skin necrosis have been reported,^4,5 there is no comprehensive list of sequelae resulting from fluid extravasation in the literature, and there are no known studies examining the incidence and types of cutaneous complications. In this study, we sought to evaluate the dermatologic impacts of this procedure.

Methods

We performed a retrospective chart review approved by the institutional review board at a large metropolitan level I trauma center in the Midwestern United States spanning 18 consecutive months to identify all patients who underwent IO line placement, either en route to or upon arrival at the trauma center. The electronic medical records of 113 patients (age range, 10 days–94 years) were identified using either an automated natural language look-up program with keywords including intraosseous access and IO or a Current Procedural Terminology code 36680. Data including patient age, reason for IO insertion, anatomic location of the IO, and complications secondary to IO line placement were recorded.

Results

We identified an overall complication rate of 2.7% (3/113), with only 1 patient showing isolated cutaneous complications from IO line placement. The complications in the first 2 patients included compartment syndrome following IO line placement in the right tibia and needle breakage during IO line placement. The third patient, a 30-year-old heart transplant recipient, developed tense bullae on the left leg 5 days after a resuscitative effort required IO access through the bilateral tibiae. The patient had received vasopressors as well as 750 mL of normal saline through these access points. Two days after resuscitation, she developed an enlarging ecchymosis around the left IO access point. On day 5, cutaneous findings included 2 large firm bullae on a purpuric base overlying the left proximal tibia and patella (Figure 1). After an ultrasound revealed no connection to the underlying joint space, the bullae were incised and drained for patient comfort, as she reported pain and undue pressure at the knee joint on initial dermatologic consultation. These symptoms abated after the procedure. Cultures of bullous fluid were negative for infection. Histopathologic examination revealed a subepidermal split with underlying minimal mixed inflammation, favoring the diagnosis of traumatic bullae.

At a scheduled 7-month dermatology follow-up, the wound bed appeared to be healing well with surrounding scarring with no residual bleeding or drainage (Figure 2) despite the patient reporting a protracted course of wound healing requiring debridement due to eschar formation and multiple follow-up appointments with the wound care service.

Comment

The most commonly reported complications with IO line placement result from fluid infiltration of the subcutaneous tissue secondary to catheter misplacement.^1,3 Extravasated fluid may lead to tissue damage, compartment syndrome, and even tissue necrosis in some cases.^1,4,5 Localized cellulitis and the formation of subcutaneous abscesses also have been reported, albeit rarely.^3,5

In our retrospective cohort review, we identified an additional potential complication of IO line placement that has not been widely reported—development of large traumatic bullae. It is most likely that this patient’s IO catheter became dislodged, resulting in extravasation of fluids into the dermal and subcutaneous tissues.

Our findings support the previously noted complication rate of less than 1% following IO line placement, with an overall complication rate of 2.7% that included only 1 patient with a cutaneous complication.² Given this low incidence, providers may not be used to recognizing such complications, leading to delayed or incorrect diagnosis of these entities. While there are certain conditions in which IO insertion is contraindicated, including severe bone diseases (eg, osteogenesis imperfecta, osteomyelitis), overlying cellulitis, and bone fracture, these conditions are rare and can be avoided in most cases by use of an alternative site for needle insertion.² Due to the widespread utility of this tool and its few contraindications, its use in hospitalized patients is rapidly increasing, necessitating a need for quick recognition of potential complications.

From previous data on the incidence of traumatic blisters with underlying bone fractures, there are several identifiable risk factors that could be extended to patients at high risk for developing cutaneous IO complications secondary to the trauma associated with needle insertion,⁶ including wound-healing impairments in patients with fragile lymphatics, peripheral vascular disease, diabetes, or collagen vascular diseases (eg, lupus, rheumatoid arthritis, Sjögren syndrome). Patients with these conditions should be closely monitored for the development of bullae.⁶ While the patient we highlighted in our study did not have a history of such conditions, her history of cardiac disease, recent resuscitation attempts, and immunosuppression certainly could have contributed to suboptimal tissue agility and repair after IO line placement.

Conclusion

Intraosseous access is a safe, effective, and reliable option for vascular access in both pediatric and adult populations that is widely used in both prehospital (ie, paramedic administered) and hospital settings, including intensive care units, emergency departments, and any acute situation where rapid vascular access is necessary. This retrospective chart review examining the incidence and types of cutaneous complications associated with IO line placement at a level I trauma center revealed a total complication rate similar to those reported in previous studies and also highlighted a unique postprocedural cutaneous finding of traumatic bullae. Although no unified management recommendations currently exist, providers should consider this complication in the differential for hospitalized patients with large, atypical, asymmetric bullae in the absence of an alternative explanation for such skin findings.

Regarded as dermatologic emergencies, Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) represent a spectrum of blistering skin diseases that have a high mortality rate. Because of a misguided immune response to medications or infections, CD8⁺ T lymphocytes release proinflammatory cytokines, giving rise to the extensive epidermal destruction seen in SJS and TEN. The exact pathogenesis of SJS and TEN is still poorly defined, but studies have proposed that T cells mediate keratinocyte (KC) apoptosis through perforin and granzyme release and activation of the Fas/Fas ligand (FasL). Functioning as a transmembrane death receptor in the tumor necrosis factor (TNF) superfamily, Fas (CD95) activates Fas-associated death domain protein, caspases, and nucleases, resulting in organized cell destruction. Likewise, perforin and granzymes also have been shown to play a similar role in apoptosis via activation of caspases.¹

Evidence for the role of TNF-α in SJS and TEN has been supported by findings of elevated levels of TNF-α within the blister fluid, serum, and KC cell surface. Additionally, TNF-α has been shown to upregulate inducible nitric oxide synthase in KCs, causing an accumulation of nitric oxide and subsequent FasL-mediated cell death.^1-3 Notably, studies have demonstrated a relative lack of lymphocytes in the tissue of TEN patients despite the extensive destruction that is observed, thus emphasizing the importance of amplification and cell signaling via inflammatory mediators such as TNF-α.¹ In this proposed model, T cells release IFN-γ, causing KCs to release TNF-α that subsequently promotes the upregulation of the aforementioned FasL.¹ Tumor necrosis factor α also may promote increased MHC class I complex deposition on KC surfaces that may play a role in perforin and granzyme-mediated apoptosis of KCs.¹

There is still debate on the standard of care for the treatment of SJS and TEN, attributed to the absence of randomized controlled trials and the rarity of the disease as well as the numerous conflicting studies evaluating potential treatments.^1,4 Despite conflicting data to support their use, supportive care and intravenous immunoglobulin (IVIG) continue to be common treatments for SJS and TEN in hospitals worldwide. Elucidation of the role of TNF-α has prompted the use of infliximab and etanercept. In a case series of Italian patients with TEN (average SCORTEN, 3.6) treated with the TNF-α antagonist etanercept, no mortality was observed, which was well below the calculated expected mortality of 46.9%.² Our retrospective study compared the use of a TNF antagonist to other therapies in the treatment of SJS/TEN. Our data suggest that etanercept is a lifesaving and disease-modifying therapy.

Methods

Twenty-two patients with SJS/TEN were included in this analysis. This included all patients who carried a clinical diagnosis of SJS/TEN with a confirmatory biopsy at our 2 university centers—University of California, Los Angeles, and Keck-LA County-Norris Hospital at the University of Southern California, Los Angeles—from 2013 to 2016. The diagnosis was rendered when a clinical diagnosis of SJS/TEN was given by a dermatologist and a confirmatory biopsy was performed. Every patient given a diagnosis of SJS/TEN at either university system from 2015 onward received an injection of etanercept given the positive results reported by Paradisi et al.²

The 9 patients who presented from 2013 to 2014 to our 2 hospital systems and were given a diagnosis of SJS/TEN received either IVIG or supportive care alone and had an average body surface area (BSA) affected of 23%. The 13 patients who presented from 2015 to 2016 were treated with etanercept in the form of a 50-mg subcutaneous injection given once to the right upper arm. Of this group, 4 patients received dual therapy with both IVIG and etanercept. In the etanercept-treated group (etanercept alone and etanercept plus IVIG), the average BSA affected was 30%. At the time of preliminary diagnosis, all patient medications were evaluated for a possible temporal relationship to the onset of rash and were discontinued if felt to be causative. The causative agent and treatment course for each patient is summarized in Table 1.

Patients were monitored daily in the hospital for improvement, and time to re-epithelialization was measured. Re-epithelialization was defined as progressive healing with residual lesions (erosions, ulcers, or bullae) covering no more than 5% BSA and was contingent on the patient having no new lesions within 24 hours.⁵ SCORe of Toxic Epidermal Necrosis (SCORTEN), a validated severity-of-illness score,⁶ was calculated by giving 1 point for each of the following criteria at the time of diagnosis: age ≥40 years, concurrent malignancy, heart rate ≥120 beats/min, serum blood urea nitrogen >27 mg/dL, serum bicarbonate <20 mEq/L, serum glucose >250 mg/dL, and detached or compromised BSA >10%. The total SCORTEN was correlated with the following risk of mortality as supported by prior validation studies: SCORTEN of 0 to 1, 3.2%; SCORTEN of 2, 12.1%; SCORTEN of 3, 35.3%; SCORTEN of 4, 58.3%; SCORTEN of ≥5, >90%.

 

Results

A total of 13 patients received etanercept. The mean SCORTEN was 2.2. The observed mortality was 0%, which was markedly lower than the predicted mortality of 24.3% (as determined by linear interpolation). Of this cohort, 9 patients received etanercept alone (mean SCORTEN of 2.1, predicted mortality of 22.9%), whereas 4 patients received a combination of etanercept and IVIG (mean SCORTEN of 2.3, predicted mortality of 27.2%).

The 4 patients who received both etanercept and IVIG received dual therapy for varying reasons. In patient 2 (Table 1), the perceived severity of this case ultimately led to the decision to start IVIG in addition to etanercept, resulting in rapid recovery and discharge after only 1 week of hospitalization. Intravenous immunoglobulin also was given in patient 3 (SCORTEN of 4) and patient 6 (SCORTEN of 2) for progression of disease despite administration of etanercept, with subsequent cessation of progression after the addition of the second agent (IVIG). Patient 12 might have done well on etanercept monotherapy but was administered IVIG as a precautionary measure because of hospital treatment algorithms.

Nine patients did not receive etanercept. Of this group, 5 received IVIG and 4 were managed with supportive care alone. The average SCORTEN for this group was 2.4, only slightly higher than the group that received etanercept (Table 2). The mortality rate in this group was 33%, which was higher than the predicted mortality of 28.1%.

Re-epithelialization data were available for 8 patients who received etanercept. The average time to re-epithelialization for these patients was 8.9 days and ranged from 3 to 19 days. Of these patients, 2 received both IVIG and etanercept, with an average time to re-epithelialization of 13 days. For the 6 patients who received etanercept alone, the average time to re-epithelialization was 7.5 days. Re-epithelialization data were not available for any of the patients who received only IVIG or supportive care but to our recollection ranged from 14 to 21 days.

The clinical course of the 13 patients after the administration of a single dose of etanercept was remarkable, as there was complete absence of mortality and an increase in speed of recovery in most patients receiving this intervention (time to re-epithelialization, 3–19 days). We also observed another interesting trend from our patients treated with etanercept, which was the suggestion that treatment with etanercept may be less effective if IVIG and/or steroids are given prior to etanercept; likewise, treatment is more effective when etanercept is given quickly. For patients 1, 4, 5, 7, 9, and 11 (as shown in Table 1), no prior IVIG therapy or other immunosuppressive therapy had been given before etanercept was administered. In these 6 patients, the average time to re-epithelialization after etanercept administration was 7.5 days; average time to re-epithelialization, unfortunately, is not available for the patients who were not treated with etanercept. In addition, as shown in the Figure, it was noted in some patients that the depth of denudation was markedly more superficial than what would typically be clinically observed with TEN after administration of other immunomodulatory therapies such as IVIG or prednisone or with supportive care alone. In these 2 patients with superficial desquamation—patients 7 and 9—etanercept notably was given within 6 hours of onset of skin pain.

 

Comment

There is no definitive gold standard treatment of SJS, SJS/TEN overlap, or TEN. However, generally agreed upon management includes immediate discontinuation of the offending medication and supportive therapy with aggressive electrolyte replacement and wound care. Management in a burn unit or intensive care unit is recommended in severe cases. Contention over the efficacy of various medications in the treatment of SJS and TEN continues and largely is due to the rarity of SJS and TEN; studies are small and almost all lack randomization. Therapies that have been used include high-dose steroids, IVIG, plasmapheresis, cyclophosphamide, cyclosporine A, and TNF inhibitors (eg, etanercept, infliximab).¹

Evidence for the use of anti–TNF-α antibodies has been limited thus far, with most of the literature focusing on infliximab and etanercept. Adalimumab, a fully humanized clonal antibody, has no reported cases in the dermatologic literature for use in patients with SJS/TEN. Two case reports of adalimumab paradoxically causing SJS have been documented. In both cases, adalimumab was stopped and patients responded to intravenous corticosteroids and infliximab.^7,8 Similarly, thalidomide has not proven to be a promising anti–TNF-α agent for the treatment of SJS/TEN. In the only attempted randomized controlled trial for SJS and TEN, thalidomide appeared to increase mortality, eventuating in this trial being terminated prior to the planned end date.⁹Infliximab and etanercept have several case reports and a few case series highlighting potentially efficacious application of TNF-α inhibitors for the treatment of SJS/TEN.^10-13 In 2002, Fischer et al¹⁰ reported the first case of TEN treated successfully with a single dose of infliximab 5 mg/kg. Kreft et al¹⁴ reported on etoricoxib-induced TEN that was treated with infliximab 5 mg/kg, which led to re-epithelialization within 5 weeks (notably a 5-week re-epithelialization time is not necessarily an improvement).

In 2005, Hunger et al³ demonstrated TNF-α’s release by KCs in the epidermis and by inflammatory cells in the dermis of a TEN patient. Twenty-four hours after the administration of infliximab 5 mg/kg in these patients, TNF-α was found to be below normal and epidermal detachment ceased.³ Wojtkietwicz et al¹³ demonstrated benefit following an infusion of infliximab 5 mg/kg in a patient whose disease continued to progress despite treatment with dexamethasone and 1.8 g/kg of IVIG.

Then 2 subsequent case series added further support for the efficacy of infliximab in the treatment of TEN. Patmanidis et al¹⁵ and Gaitanis et al¹⁶ reported similar results in 4 patients, each treated with infliximab 5 mg/kg immediately followed by initiation of high-dose IVIG (2 g/kg over 5 days). Zárate-Correa et al¹⁷ reported a 0% mortality rate and near-complete re-epithelialization after 5 to 14 days in 4 patients treated with a single 300-mg dose of infliximab.

However, the success of infliximab in the treatment of TEN has been countered by the pilot study by Paquet et al,¹⁸ which compared the efficacy of 150 mg/kg of N-acetylcysteine alone vs adding infliximab 5 mg/kg to treat 10 TEN patients. The study demonstrated no benefit at 48 hours in the group given infliximab, the time frame in which prior case reports touting infliximab’s benefit claimed the benefit was observed. Similarly, there was no effect on mortality for either treatment modality as assessed by illness auxiliary score.¹⁸

Evidence in support of the use of etanercept in the treatment of SJS/TEN is mounting, and some centers have begun to use it as the first-choice therapy for SJS/TEN. The first case was reported by Famularo et al,¹⁹ in which a patient with TEN was given 2 doses of etanercept 25 mg after failure to improve with prednisolone 1 mg/kg. The patient showed near-complete and rapid re-epithelization in 6 days before death due to disseminated intravascular coagulation 10 days after admission.¹⁹ Gubinelli et al²⁰ and Sadighha²¹ independently reported cases of TEN and TEN/acute generalized exanthematous pustulosis overlap treated with a total of 50 mg of etanercept, demonstrating rapid cessation of lesion progression. Didona et al²² found similar benefit using etanercept 50 mg to treat TEN secondary to rituximab after failure to improve with prednisone and cyclophosphamide. Treatment of TEN with etanercept in an HIV-positive patient also has been reported. Lee et al²³ described a patient who was administered 50-mg and 25-mg injections on days 3 and 5 of hospitalization, respectively, with re-epithelialization occurring by day 8. Finally, Owczarczyk-Saczonek et al²⁴ reported a case of SJS in a patient with a 4-year history of etanercept and sulfasalazine treatment of rheumatoid arthritis; sulfasalazine was stopped, but this patient was continued on etanercept until resolution of skin and mucosal symptoms. However, it is important to consider the possibility of publication bias among these cases selected for their positive outcomes.

Perhaps the most compelling literature regarding the use of etanercept for TEN was described in a case series by Paradisi et al.² This study included 10 patients with TEN, all of whom demonstrated complete re-epithelialization shortly after receiving etanercept 50 mg. Average SCORTEN was 3.6 with a range of 2 to 6. Eight patients in this study had severe comorbidities and all 10 patients survived, with a time to re-epithelialization ranging from 7 to 20 days.² Additionally, a randomized controlled trial showed that 38 etanercept-treated patients had improved mortality (P=.266) and re-epithelialization time (P=.01) compared to patients treated with intravenous methylprednisolone.²⁵Limitations to our study are similar to other reports of SJS/TEN and included the small number of cases and lack of randomization. Additionally, we do not have data available for all patients for time between onset of disease and treatment initiation. Because of these challenges, data presented in this case series is observational only. Additionally, the patients treated with etanercept alone had a slightly lower SCORTEN compared to the group that received IVIG or supportive care alone (2.1 and 2.4 respectively). However, the etanercept-only group actually had higher involvement of epidermal detachment (33%) compared to the non-etanercept group (23%).

Conclusion

Although treatment with etanercept lacks the support of a randomized controlled trial, similar to all other treatments currently used for SJS and TEN, preliminary reports highlight a benefit in disease progression and improvement in time to re-epithelialization. In particular, if etanercept 50 mg subcutaneously is given as monotherapy or is given early in the disease course (prior to other therapies being attempted and ideally within 6 hours of presentation), our data suggest an even greater trend toward improved mortality and decreased time to re-epithelialization. Additionally, our findings may suggest that in some patients, etanercept monotherapy is not an adequate intervention but the addition of IVIG may be helpful; however, the senior author (S.W.) notes anecdotally that in his experience with the patients treated at the University of California Los Angeles, the order of administration of combination therapies—etanercept followed by IVIG—was important in addition to the choice of therapy. These findings are promising enough to warrant a multicenter randomized controlled trial comparing the efficacy of etanercept to other more commonly used treatments for this spectrum of disease, including IVIG and/or cyclosporine. Based on the data presented in this case series, including the 13 patients who received etanercept and had a 0% mortality rate, etanercept may be viewed as a targeted therapeutic intervention for patients with SJS and TEN.

Regarded as dermatologic emergencies, Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) represent a spectrum of blistering skin diseases that have a high mortality rate. Because of a misguided immune response to medications or infections, CD8⁺ T lymphocytes release proinflammatory cytokines, giving rise to the extensive epidermal destruction seen in SJS and TEN. The exact pathogenesis of SJS and TEN is still poorly defined, but studies have proposed that T cells mediate keratinocyte (KC) apoptosis through perforin and granzyme release and activation of the Fas/Fas ligand (FasL). Functioning as a transmembrane death receptor in the tumor necrosis factor (TNF) superfamily, Fas (CD95) activates Fas-associated death domain protein, caspases, and nucleases, resulting in organized cell destruction. Likewise, perforin and granzymes also have been shown to play a similar role in apoptosis via activation of caspases.¹

Evidence for the role of TNF-α in SJS and TEN has been supported by findings of elevated levels of TNF-α within the blister fluid, serum, and KC cell surface. Additionally, TNF-α has been shown to upregulate inducible nitric oxide synthase in KCs, causing an accumulation of nitric oxide and subsequent FasL-mediated cell death.^1-3 Notably, studies have demonstrated a relative lack of lymphocytes in the tissue of TEN patients despite the extensive destruction that is observed, thus emphasizing the importance of amplification and cell signaling via inflammatory mediators such as TNF-α.¹ In this proposed model, T cells release IFN-γ, causing KCs to release TNF-α that subsequently promotes the upregulation of the aforementioned FasL.¹ Tumor necrosis factor α also may promote increased MHC class I complex deposition on KC surfaces that may play a role in perforin and granzyme-mediated apoptosis of KCs.¹

There is still debate on the standard of care for the treatment of SJS and TEN, attributed to the absence of randomized controlled trials and the rarity of the disease as well as the numerous conflicting studies evaluating potential treatments.^1,4 Despite conflicting data to support their use, supportive care and intravenous immunoglobulin (IVIG) continue to be common treatments for SJS and TEN in hospitals worldwide. Elucidation of the role of TNF-α has prompted the use of infliximab and etanercept. In a case series of Italian patients with TEN (average SCORTEN, 3.6) treated with the TNF-α antagonist etanercept, no mortality was observed, which was well below the calculated expected mortality of 46.9%.² Our retrospective study compared the use of a TNF antagonist to other therapies in the treatment of SJS/TEN. Our data suggest that etanercept is a lifesaving and disease-modifying therapy.

Methods

Twenty-two patients with SJS/TEN were included in this analysis. This included all patients who carried a clinical diagnosis of SJS/TEN with a confirmatory biopsy at our 2 university centers—University of California, Los Angeles, and Keck-LA County-Norris Hospital at the University of Southern California, Los Angeles—from 2013 to 2016. The diagnosis was rendered when a clinical diagnosis of SJS/TEN was given by a dermatologist and a confirmatory biopsy was performed. Every patient given a diagnosis of SJS/TEN at either university system from 2015 onward received an injection of etanercept given the positive results reported by Paradisi et al.²

The 9 patients who presented from 2013 to 2014 to our 2 hospital systems and were given a diagnosis of SJS/TEN received either IVIG or supportive care alone and had an average body surface area (BSA) affected of 23%. The 13 patients who presented from 2015 to 2016 were treated with etanercept in the form of a 50-mg subcutaneous injection given once to the right upper arm. Of this group, 4 patients received dual therapy with both IVIG and etanercept. In the etanercept-treated group (etanercept alone and etanercept plus IVIG), the average BSA affected was 30%. At the time of preliminary diagnosis, all patient medications were evaluated for a possible temporal relationship to the onset of rash and were discontinued if felt to be causative. The causative agent and treatment course for each patient is summarized in Table 1.

Patients were monitored daily in the hospital for improvement, and time to re-epithelialization was measured. Re-epithelialization was defined as progressive healing with residual lesions (erosions, ulcers, or bullae) covering no more than 5% BSA and was contingent on the patient having no new lesions within 24 hours.⁵ SCORe of Toxic Epidermal Necrosis (SCORTEN), a validated severity-of-illness score,⁶ was calculated by giving 1 point for each of the following criteria at the time of diagnosis: age ≥40 years, concurrent malignancy, heart rate ≥120 beats/min, serum blood urea nitrogen >27 mg/dL, serum bicarbonate <20 mEq/L, serum glucose >250 mg/dL, and detached or compromised BSA >10%. The total SCORTEN was correlated with the following risk of mortality as supported by prior validation studies: SCORTEN of 0 to 1, 3.2%; SCORTEN of 2, 12.1%; SCORTEN of 3, 35.3%; SCORTEN of 4, 58.3%; SCORTEN of ≥5, >90%.

 

Results

A total of 13 patients received etanercept. The mean SCORTEN was 2.2. The observed mortality was 0%, which was markedly lower than the predicted mortality of 24.3% (as determined by linear interpolation). Of this cohort, 9 patients received etanercept alone (mean SCORTEN of 2.1, predicted mortality of 22.9%), whereas 4 patients received a combination of etanercept and IVIG (mean SCORTEN of 2.3, predicted mortality of 27.2%).

The 4 patients who received both etanercept and IVIG received dual therapy for varying reasons. In patient 2 (Table 1), the perceived severity of this case ultimately led to the decision to start IVIG in addition to etanercept, resulting in rapid recovery and discharge after only 1 week of hospitalization. Intravenous immunoglobulin also was given in patient 3 (SCORTEN of 4) and patient 6 (SCORTEN of 2) for progression of disease despite administration of etanercept, with subsequent cessation of progression after the addition of the second agent (IVIG). Patient 12 might have done well on etanercept monotherapy but was administered IVIG as a precautionary measure because of hospital treatment algorithms.

Nine patients did not receive etanercept. Of this group, 5 received IVIG and 4 were managed with supportive care alone. The average SCORTEN for this group was 2.4, only slightly higher than the group that received etanercept (Table 2). The mortality rate in this group was 33%, which was higher than the predicted mortality of 28.1%.

Re-epithelialization data were available for 8 patients who received etanercept. The average time to re-epithelialization for these patients was 8.9 days and ranged from 3 to 19 days. Of these patients, 2 received both IVIG and etanercept, with an average time to re-epithelialization of 13 days. For the 6 patients who received etanercept alone, the average time to re-epithelialization was 7.5 days. Re-epithelialization data were not available for any of the patients who received only IVIG or supportive care but to our recollection ranged from 14 to 21 days.

The clinical course of the 13 patients after the administration of a single dose of etanercept was remarkable, as there was complete absence of mortality and an increase in speed of recovery in most patients receiving this intervention (time to re-epithelialization, 3–19 days). We also observed another interesting trend from our patients treated with etanercept, which was the suggestion that treatment with etanercept may be less effective if IVIG and/or steroids are given prior to etanercept; likewise, treatment is more effective when etanercept is given quickly. For patients 1, 4, 5, 7, 9, and 11 (as shown in Table 1), no prior IVIG therapy or other immunosuppressive therapy had been given before etanercept was administered. In these 6 patients, the average time to re-epithelialization after etanercept administration was 7.5 days; average time to re-epithelialization, unfortunately, is not available for the patients who were not treated with etanercept. In addition, as shown in the Figure, it was noted in some patients that the depth of denudation was markedly more superficial than what would typically be clinically observed with TEN after administration of other immunomodulatory therapies such as IVIG or prednisone or with supportive care alone. In these 2 patients with superficial desquamation—patients 7 and 9—etanercept notably was given within 6 hours of onset of skin pain.

 

Comment

There is no definitive gold standard treatment of SJS, SJS/TEN overlap, or TEN. However, generally agreed upon management includes immediate discontinuation of the offending medication and supportive therapy with aggressive electrolyte replacement and wound care. Management in a burn unit or intensive care unit is recommended in severe cases. Contention over the efficacy of various medications in the treatment of SJS and TEN continues and largely is due to the rarity of SJS and TEN; studies are small and almost all lack randomization. Therapies that have been used include high-dose steroids, IVIG, plasmapheresis, cyclophosphamide, cyclosporine A, and TNF inhibitors (eg, etanercept, infliximab).¹

Evidence for the use of anti–TNF-α antibodies has been limited thus far, with most of the literature focusing on infliximab and etanercept. Adalimumab, a fully humanized clonal antibody, has no reported cases in the dermatologic literature for use in patients with SJS/TEN. Two case reports of adalimumab paradoxically causing SJS have been documented. In both cases, adalimumab was stopped and patients responded to intravenous corticosteroids and infliximab.^7,8 Similarly, thalidomide has not proven to be a promising anti–TNF-α agent for the treatment of SJS/TEN. In the only attempted randomized controlled trial for SJS and TEN, thalidomide appeared to increase mortality, eventuating in this trial being terminated prior to the planned end date.⁹Infliximab and etanercept have several case reports and a few case series highlighting potentially efficacious application of TNF-α inhibitors for the treatment of SJS/TEN.^10-13 In 2002, Fischer et al¹⁰ reported the first case of TEN treated successfully with a single dose of infliximab 5 mg/kg. Kreft et al¹⁴ reported on etoricoxib-induced TEN that was treated with infliximab 5 mg/kg, which led to re-epithelialization within 5 weeks (notably a 5-week re-epithelialization time is not necessarily an improvement).

In 2005, Hunger et al³ demonstrated TNF-α’s release by KCs in the epidermis and by inflammatory cells in the dermis of a TEN patient. Twenty-four hours after the administration of infliximab 5 mg/kg in these patients, TNF-α was found to be below normal and epidermal detachment ceased.³ Wojtkietwicz et al¹³ demonstrated benefit following an infusion of infliximab 5 mg/kg in a patient whose disease continued to progress despite treatment with dexamethasone and 1.8 g/kg of IVIG.

Then 2 subsequent case series added further support for the efficacy of infliximab in the treatment of TEN. Patmanidis et al¹⁵ and Gaitanis et al¹⁶ reported similar results in 4 patients, each treated with infliximab 5 mg/kg immediately followed by initiation of high-dose IVIG (2 g/kg over 5 days). Zárate-Correa et al¹⁷ reported a 0% mortality rate and near-complete re-epithelialization after 5 to 14 days in 4 patients treated with a single 300-mg dose of infliximab.

However, the success of infliximab in the treatment of TEN has been countered by the pilot study by Paquet et al,¹⁸ which compared the efficacy of 150 mg/kg of N-acetylcysteine alone vs adding infliximab 5 mg/kg to treat 10 TEN patients. The study demonstrated no benefit at 48 hours in the group given infliximab, the time frame in which prior case reports touting infliximab’s benefit claimed the benefit was observed. Similarly, there was no effect on mortality for either treatment modality as assessed by illness auxiliary score.¹⁸

Evidence in support of the use of etanercept in the treatment of SJS/TEN is mounting, and some centers have begun to use it as the first-choice therapy for SJS/TEN. The first case was reported by Famularo et al,¹⁹ in which a patient with TEN was given 2 doses of etanercept 25 mg after failure to improve with prednisolone 1 mg/kg. The patient showed near-complete and rapid re-epithelization in 6 days before death due to disseminated intravascular coagulation 10 days after admission.¹⁹ Gubinelli et al²⁰ and Sadighha²¹ independently reported cases of TEN and TEN/acute generalized exanthematous pustulosis overlap treated with a total of 50 mg of etanercept, demonstrating rapid cessation of lesion progression. Didona et al²² found similar benefit using etanercept 50 mg to treat TEN secondary to rituximab after failure to improve with prednisone and cyclophosphamide. Treatment of TEN with etanercept in an HIV-positive patient also has been reported. Lee et al²³ described a patient who was administered 50-mg and 25-mg injections on days 3 and 5 of hospitalization, respectively, with re-epithelialization occurring by day 8. Finally, Owczarczyk-Saczonek et al²⁴ reported a case of SJS in a patient with a 4-year history of etanercept and sulfasalazine treatment of rheumatoid arthritis; sulfasalazine was stopped, but this patient was continued on etanercept until resolution of skin and mucosal symptoms. However, it is important to consider the possibility of publication bias among these cases selected for their positive outcomes.

Perhaps the most compelling literature regarding the use of etanercept for TEN was described in a case series by Paradisi et al.² This study included 10 patients with TEN, all of whom demonstrated complete re-epithelialization shortly after receiving etanercept 50 mg. Average SCORTEN was 3.6 with a range of 2 to 6. Eight patients in this study had severe comorbidities and all 10 patients survived, with a time to re-epithelialization ranging from 7 to 20 days.² Additionally, a randomized controlled trial showed that 38 etanercept-treated patients had improved mortality (P=.266) and re-epithelialization time (P=.01) compared to patients treated with intravenous methylprednisolone.²⁵Limitations to our study are similar to other reports of SJS/TEN and included the small number of cases and lack of randomization. Additionally, we do not have data available for all patients for time between onset of disease and treatment initiation. Because of these challenges, data presented in this case series is observational only. Additionally, the patients treated with etanercept alone had a slightly lower SCORTEN compared to the group that received IVIG or supportive care alone (2.1 and 2.4 respectively). However, the etanercept-only group actually had higher involvement of epidermal detachment (33%) compared to the non-etanercept group (23%).

Conclusion

Although treatment with etanercept lacks the support of a randomized controlled trial, similar to all other treatments currently used for SJS and TEN, preliminary reports highlight a benefit in disease progression and improvement in time to re-epithelialization. In particular, if etanercept 50 mg subcutaneously is given as monotherapy or is given early in the disease course (prior to other therapies being attempted and ideally within 6 hours of presentation), our data suggest an even greater trend toward improved mortality and decreased time to re-epithelialization. Additionally, our findings may suggest that in some patients, etanercept monotherapy is not an adequate intervention but the addition of IVIG may be helpful; however, the senior author (S.W.) notes anecdotally that in his experience with the patients treated at the University of California Los Angeles, the order of administration of combination therapies—etanercept followed by IVIG—was important in addition to the choice of therapy. These findings are promising enough to warrant a multicenter randomized controlled trial comparing the efficacy of etanercept to other more commonly used treatments for this spectrum of disease, including IVIG and/or cyclosporine. Based on the data presented in this case series, including the 13 patients who received etanercept and had a 0% mortality rate, etanercept may be viewed as a targeted therapeutic intervention for patients with SJS and TEN.

Regarded as dermatologic emergencies, Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) represent a spectrum of blistering skin diseases that have a high mortality rate. Because of a misguided immune response to medications or infections, CD8⁺ T lymphocytes release proinflammatory cytokines, giving rise to the extensive epidermal destruction seen in SJS and TEN. The exact pathogenesis of SJS and TEN is still poorly defined, but studies have proposed that T cells mediate keratinocyte (KC) apoptosis through perforin and granzyme release and activation of the Fas/Fas ligand (FasL). Functioning as a transmembrane death receptor in the tumor necrosis factor (TNF) superfamily, Fas (CD95) activates Fas-associated death domain protein, caspases, and nucleases, resulting in organized cell destruction. Likewise, perforin and granzymes also have been shown to play a similar role in apoptosis via activation of caspases.¹

Evidence for the role of TNF-α in SJS and TEN has been supported by findings of elevated levels of TNF-α within the blister fluid, serum, and KC cell surface. Additionally, TNF-α has been shown to upregulate inducible nitric oxide synthase in KCs, causing an accumulation of nitric oxide and subsequent FasL-mediated cell death.^1-3 Notably, studies have demonstrated a relative lack of lymphocytes in the tissue of TEN patients despite the extensive destruction that is observed, thus emphasizing the importance of amplification and cell signaling via inflammatory mediators such as TNF-α.¹ In this proposed model, T cells release IFN-γ, causing KCs to release TNF-α that subsequently promotes the upregulation of the aforementioned FasL.¹ Tumor necrosis factor α also may promote increased MHC class I complex deposition on KC surfaces that may play a role in perforin and granzyme-mediated apoptosis of KCs.¹

There is still debate on the standard of care for the treatment of SJS and TEN, attributed to the absence of randomized controlled trials and the rarity of the disease as well as the numerous conflicting studies evaluating potential treatments.^1,4 Despite conflicting data to support their use, supportive care and intravenous immunoglobulin (IVIG) continue to be common treatments for SJS and TEN in hospitals worldwide. Elucidation of the role of TNF-α has prompted the use of infliximab and etanercept. In a case series of Italian patients with TEN (average SCORTEN, 3.6) treated with the TNF-α antagonist etanercept, no mortality was observed, which was well below the calculated expected mortality of 46.9%.² Our retrospective study compared the use of a TNF antagonist to other therapies in the treatment of SJS/TEN. Our data suggest that etanercept is a lifesaving and disease-modifying therapy.

Methods

Twenty-two patients with SJS/TEN were included in this analysis. This included all patients who carried a clinical diagnosis of SJS/TEN with a confirmatory biopsy at our 2 university centers—University of California, Los Angeles, and Keck-LA County-Norris Hospital at the University of Southern California, Los Angeles—from 2013 to 2016. The diagnosis was rendered when a clinical diagnosis of SJS/TEN was given by a dermatologist and a confirmatory biopsy was performed. Every patient given a diagnosis of SJS/TEN at either university system from 2015 onward received an injection of etanercept given the positive results reported by Paradisi et al.²

The 9 patients who presented from 2013 to 2014 to our 2 hospital systems and were given a diagnosis of SJS/TEN received either IVIG or supportive care alone and had an average body surface area (BSA) affected of 23%. The 13 patients who presented from 2015 to 2016 were treated with etanercept in the form of a 50-mg subcutaneous injection given once to the right upper arm. Of this group, 4 patients received dual therapy with both IVIG and etanercept. In the etanercept-treated group (etanercept alone and etanercept plus IVIG), the average BSA affected was 30%. At the time of preliminary diagnosis, all patient medications were evaluated for a possible temporal relationship to the onset of rash and were discontinued if felt to be causative. The causative agent and treatment course for each patient is summarized in Table 1.

Patients were monitored daily in the hospital for improvement, and time to re-epithelialization was measured. Re-epithelialization was defined as progressive healing with residual lesions (erosions, ulcers, or bullae) covering no more than 5% BSA and was contingent on the patient having no new lesions within 24 hours.⁵ SCORe of Toxic Epidermal Necrosis (SCORTEN), a validated severity-of-illness score,⁶ was calculated by giving 1 point for each of the following criteria at the time of diagnosis: age ≥40 years, concurrent malignancy, heart rate ≥120 beats/min, serum blood urea nitrogen >27 mg/dL, serum bicarbonate <20 mEq/L, serum glucose >250 mg/dL, and detached or compromised BSA >10%. The total SCORTEN was correlated with the following risk of mortality as supported by prior validation studies: SCORTEN of 0 to 1, 3.2%; SCORTEN of 2, 12.1%; SCORTEN of 3, 35.3%; SCORTEN of 4, 58.3%; SCORTEN of ≥5, >90%.

 

Results

A total of 13 patients received etanercept. The mean SCORTEN was 2.2. The observed mortality was 0%, which was markedly lower than the predicted mortality of 24.3% (as determined by linear interpolation). Of this cohort, 9 patients received etanercept alone (mean SCORTEN of 2.1, predicted mortality of 22.9%), whereas 4 patients received a combination of etanercept and IVIG (mean SCORTEN of 2.3, predicted mortality of 27.2%).

The 4 patients who received both etanercept and IVIG received dual therapy for varying reasons. In patient 2 (Table 1), the perceived severity of this case ultimately led to the decision to start IVIG in addition to etanercept, resulting in rapid recovery and discharge after only 1 week of hospitalization. Intravenous immunoglobulin also was given in patient 3 (SCORTEN of 4) and patient 6 (SCORTEN of 2) for progression of disease despite administration of etanercept, with subsequent cessation of progression after the addition of the second agent (IVIG). Patient 12 might have done well on etanercept monotherapy but was administered IVIG as a precautionary measure because of hospital treatment algorithms.

Nine patients did not receive etanercept. Of this group, 5 received IVIG and 4 were managed with supportive care alone. The average SCORTEN for this group was 2.4, only slightly higher than the group that received etanercept (Table 2). The mortality rate in this group was 33%, which was higher than the predicted mortality of 28.1%.

Re-epithelialization data were available for 8 patients who received etanercept. The average time to re-epithelialization for these patients was 8.9 days and ranged from 3 to 19 days. Of these patients, 2 received both IVIG and etanercept, with an average time to re-epithelialization of 13 days. For the 6 patients who received etanercept alone, the average time to re-epithelialization was 7.5 days. Re-epithelialization data were not available for any of the patients who received only IVIG or supportive care but to our recollection ranged from 14 to 21 days.

The clinical course of the 13 patients after the administration of a single dose of etanercept was remarkable, as there was complete absence of mortality and an increase in speed of recovery in most patients receiving this intervention (time to re-epithelialization, 3–19 days). We also observed another interesting trend from our patients treated with etanercept, which was the suggestion that treatment with etanercept may be less effective if IVIG and/or steroids are given prior to etanercept; likewise, treatment is more effective when etanercept is given quickly. For patients 1, 4, 5, 7, 9, and 11 (as shown in Table 1), no prior IVIG therapy or other immunosuppressive therapy had been given before etanercept was administered. In these 6 patients, the average time to re-epithelialization after etanercept administration was 7.5 days; average time to re-epithelialization, unfortunately, is not available for the patients who were not treated with etanercept. In addition, as shown in the Figure, it was noted in some patients that the depth of denudation was markedly more superficial than what would typically be clinically observed with TEN after administration of other immunomodulatory therapies such as IVIG or prednisone or with supportive care alone. In these 2 patients with superficial desquamation—patients 7 and 9—etanercept notably was given within 6 hours of onset of skin pain.

 

Comment

There is no definitive gold standard treatment of SJS, SJS/TEN overlap, or TEN. However, generally agreed upon management includes immediate discontinuation of the offending medication and supportive therapy with aggressive electrolyte replacement and wound care. Management in a burn unit or intensive care unit is recommended in severe cases. Contention over the efficacy of various medications in the treatment of SJS and TEN continues and largely is due to the rarity of SJS and TEN; studies are small and almost all lack randomization. Therapies that have been used include high-dose steroids, IVIG, plasmapheresis, cyclophosphamide, cyclosporine A, and TNF inhibitors (eg, etanercept, infliximab).¹

Evidence for the use of anti–TNF-α antibodies has been limited thus far, with most of the literature focusing on infliximab and etanercept. Adalimumab, a fully humanized clonal antibody, has no reported cases in the dermatologic literature for use in patients with SJS/TEN. Two case reports of adalimumab paradoxically causing SJS have been documented. In both cases, adalimumab was stopped and patients responded to intravenous corticosteroids and infliximab.^7,8 Similarly, thalidomide has not proven to be a promising anti–TNF-α agent for the treatment of SJS/TEN. In the only attempted randomized controlled trial for SJS and TEN, thalidomide appeared to increase mortality, eventuating in this trial being terminated prior to the planned end date.⁹Infliximab and etanercept have several case reports and a few case series highlighting potentially efficacious application of TNF-α inhibitors for the treatment of SJS/TEN.^10-13 In 2002, Fischer et al¹⁰ reported the first case of TEN treated successfully with a single dose of infliximab 5 mg/kg. Kreft et al¹⁴ reported on etoricoxib-induced TEN that was treated with infliximab 5 mg/kg, which led to re-epithelialization within 5 weeks (notably a 5-week re-epithelialization time is not necessarily an improvement).

In 2005, Hunger et al³ demonstrated TNF-α’s release by KCs in the epidermis and by inflammatory cells in the dermis of a TEN patient. Twenty-four hours after the administration of infliximab 5 mg/kg in these patients, TNF-α was found to be below normal and epidermal detachment ceased.³ Wojtkietwicz et al¹³ demonstrated benefit following an infusion of infliximab 5 mg/kg in a patient whose disease continued to progress despite treatment with dexamethasone and 1.8 g/kg of IVIG.

Then 2 subsequent case series added further support for the efficacy of infliximab in the treatment of TEN. Patmanidis et al¹⁵ and Gaitanis et al¹⁶ reported similar results in 4 patients, each treated with infliximab 5 mg/kg immediately followed by initiation of high-dose IVIG (2 g/kg over 5 days). Zárate-Correa et al¹⁷ reported a 0% mortality rate and near-complete re-epithelialization after 5 to 14 days in 4 patients treated with a single 300-mg dose of infliximab.

However, the success of infliximab in the treatment of TEN has been countered by the pilot study by Paquet et al,¹⁸ which compared the efficacy of 150 mg/kg of N-acetylcysteine alone vs adding infliximab 5 mg/kg to treat 10 TEN patients. The study demonstrated no benefit at 48 hours in the group given infliximab, the time frame in which prior case reports touting infliximab’s benefit claimed the benefit was observed. Similarly, there was no effect on mortality for either treatment modality as assessed by illness auxiliary score.¹⁸

Evidence in support of the use of etanercept in the treatment of SJS/TEN is mounting, and some centers have begun to use it as the first-choice therapy for SJS/TEN. The first case was reported by Famularo et al,¹⁹ in which a patient with TEN was given 2 doses of etanercept 25 mg after failure to improve with prednisolone 1 mg/kg. The patient showed near-complete and rapid re-epithelization in 6 days before death due to disseminated intravascular coagulation 10 days after admission.¹⁹ Gubinelli et al²⁰ and Sadighha²¹ independently reported cases of TEN and TEN/acute generalized exanthematous pustulosis overlap treated with a total of 50 mg of etanercept, demonstrating rapid cessation of lesion progression. Didona et al²² found similar benefit using etanercept 50 mg to treat TEN secondary to rituximab after failure to improve with prednisone and cyclophosphamide. Treatment of TEN with etanercept in an HIV-positive patient also has been reported. Lee et al²³ described a patient who was administered 50-mg and 25-mg injections on days 3 and 5 of hospitalization, respectively, with re-epithelialization occurring by day 8. Finally, Owczarczyk-Saczonek et al²⁴ reported a case of SJS in a patient with a 4-year history of etanercept and sulfasalazine treatment of rheumatoid arthritis; sulfasalazine was stopped, but this patient was continued on etanercept until resolution of skin and mucosal symptoms. However, it is important to consider the possibility of publication bias among these cases selected for their positive outcomes.

Perhaps the most compelling literature regarding the use of etanercept for TEN was described in a case series by Paradisi et al.² This study included 10 patients with TEN, all of whom demonstrated complete re-epithelialization shortly after receiving etanercept 50 mg. Average SCORTEN was 3.6 with a range of 2 to 6. Eight patients in this study had severe comorbidities and all 10 patients survived, with a time to re-epithelialization ranging from 7 to 20 days.² Additionally, a randomized controlled trial showed that 38 etanercept-treated patients had improved mortality (P=.266) and re-epithelialization time (P=.01) compared to patients treated with intravenous methylprednisolone.²⁵Limitations to our study are similar to other reports of SJS/TEN and included the small number of cases and lack of randomization. Additionally, we do not have data available for all patients for time between onset of disease and treatment initiation. Because of these challenges, data presented in this case series is observational only. Additionally, the patients treated with etanercept alone had a slightly lower SCORTEN compared to the group that received IVIG or supportive care alone (2.1 and 2.4 respectively). However, the etanercept-only group actually had higher involvement of epidermal detachment (33%) compared to the non-etanercept group (23%).

Conclusion

Although treatment with etanercept lacks the support of a randomized controlled trial, similar to all other treatments currently used for SJS and TEN, preliminary reports highlight a benefit in disease progression and improvement in time to re-epithelialization. In particular, if etanercept 50 mg subcutaneously is given as monotherapy or is given early in the disease course (prior to other therapies being attempted and ideally within 6 hours of presentation), our data suggest an even greater trend toward improved mortality and decreased time to re-epithelialization. Additionally, our findings may suggest that in some patients, etanercept monotherapy is not an adequate intervention but the addition of IVIG may be helpful; however, the senior author (S.W.) notes anecdotally that in his experience with the patients treated at the University of California Los Angeles, the order of administration of combination therapies—etanercept followed by IVIG—was important in addition to the choice of therapy. These findings are promising enough to warrant a multicenter randomized controlled trial comparing the efficacy of etanercept to other more commonly used treatments for this spectrum of disease, including IVIG and/or cyclosporine. Based on the data presented in this case series, including the 13 patients who received etanercept and had a 0% mortality rate, etanercept may be viewed as a targeted therapeutic intervention for patients with SJS and TEN.

The prospect of facing a medical malpractice claim is a source of apprehension for physicians that affects physician behavior, including leading to defensive medicine.^1-3 Overall, annual defensive medicine costs have been estimated at $45.6 billion,⁴ and surveys of hospitalists indicate that 13.0% to 37.5% of hospitalist healthcare costs involve defensive medicine.^5,6 Despite the impact of malpractice concerns on hospitalist practice and the unprecedented growth of the field of hospital medicine, relatively few studies have examined the liability environment surrounding hospitalist practice.^7,8 The specific issue of malpractice claims rates faced by hospitalists has received even less attention in the medical literature.⁸

A better understanding of the contributing factors and other attributes of malpractice claims can help guide patient safety initiatives and inform hospitalists’ level of concern regarding liability. Although most medical errors do not result in a malpractice claim,^9,10 the majority of malpractice claims in which there is an indemnity payment involve medical injury due to clinician error.¹¹ Even malpractice claims that do not result in an indemnity payment represent opportunities to identify patient safety and risk management vulnerabilities.¹²

We used a national malpractice claims database to analyze the characteristics of claims made against hospitalists, including claims rates. In addition to claims rates, we also analyzed the other types of providers named in hospitalist claims given the importance of interdisciplinary collaboration to hospital medicine.^13,14 To provide context for understanding hospitalist liability data, we present data on other specialties. We also describe a model to predict whether hospitalist malpractice claims will close with an indemnity payment.

METHODS

Data Sources and Elements

This analysis used a repository of malpractice claims maintained by CRICO, the captive malpractice insurer of the Harvard-affiliated medical institutions. This database, the Comparative Benchmarking System, aggregates malpractice claims from multiple other malpractice carriers (both commercial and captive). This database comprises approximately 31% of all malpractice claims in the United States (both paid and unpaid) according to a comparison with claims in the National Practitioner Data Bank (NPDB), which is a government-run database to which all paid malpractice claims against physicians must be reported.¹⁵ To compile this national database, trained nurse coders, who have access to the medical records and legal documents from each case, abstract information into coded fields, such as the type of allegation, contributing factors, and injury severity. The information in these coded fields is based on a taxonomy that was developed specifically for use in malpractice cases. The nurse coders participate in regular meetings in which cases are discussed to ensure that they are applying the coding and taxonomy uniformly.

Injury severity was based on a widely used scale developed for malpractice claims by the National Association of Insurance Commissioners.¹⁶ Low injury severity included emotional injury and temporary insignificant injury. Medium injury severity included temporary minor, temporary major, and permanent minor injury. High injury severity included permanent significant injury through death. Because this study used a database assembled for operational and patient safety purposes and was not human subjects research, institutional review board approval was not needed.

Study Cohort

Malpractice claims included formal lawsuits or written requests for compensation for negligent medical care. Hospitalist malpractice claims were defined as those in which the nurse coders determined that the internal medicine hospitalist service was the clinical service that was primarily responsible for the care of the patient at the time of the clinical event giving rise to the claim. Malpractice claims closed during the period 2009-2018 were included, regardless of whether an indemnity payment was made. Claims against select other medical specialties (nonhospitalist general internal medicine, internal medicine subspecialists, emergency medicine, neurosurgery, and psychiatry), defined by the specialty of the physician named in the claim, were included for comparison with the characteristics of the claims against hospitalists. Physician specialty was available for all physicians in the study cohort; however, given that malpractice cases were aggregated from multiple different insurers, full demographic information (eg, age and sex) was not available for all physicians.

Statistical Analysis

Malpractice claims rates were treated as Poisson rates and compared using a Z-test. Malpractice claims rates are expressed as claims per 100 physician-years. Each physician-year represents 1 year of coverage of one physician by the medical malpractice carrier whose data were used. Physician-years represent the duration of time physicians practiced during which they were insured by the malpractice carrier and, as such, could have been subject to a malpractice claim that would have been included in our data. Claims rates are based on the subset of the malpractice claims in the study for which the number of physician-years of coverage is available, representing 8.2% of hospitalist claims and 11.6% of all claims.

Comparisons of the percentages of cases closing with an indemnity payment, as well as the percentages of cases in different allegation type and clinical severity categories, were made using the Fisher exact test. Indemnity payment amounts were inflation-adjusted to 2018 dollars using the Consumer Price Index. Comparisons of indemnity payment amounts between physician specialties were carried out using the Wilcoxon rank sum test given that the distribution of the payment amounts appeared nonnormal; this was confirmed with the Shapiro-Wilk test. A multivariable logistic regression model was developed to predict the binary outcome of whether a hospitalist case would close with an indemnity payment (compared with no payment), based on the 1,216 hospitalist claims. The predictors used in this regression model were chosen a priori based on hypotheses about what factors drive the likelihood that a case closes with payment. Both the unadjusted and adjusted odds ratios for the predictors are presented. The adjusted model is adjusted for all the other predictors contained in the model. All reported P values are two-sided. The statistical analysis was carried out using JMP Pro version 15 (SAS Institute Inc) and Minitab version 19 (Minitab LLC).

RESULTS

We identified 1,216 hospital medicine malpractice claims from our database. Claims rates were calculated from the subset of our data for which physician-years were available—including 5,140 physician-years encompassing 100 claims, representing 8.2% of all hospitalist claims studied. An additional 18,644 malpractice claims from five other specialties—nonhospitalist general internal medicine, internal medicine subspecialists, emergency medicine, neurosurgery, and psychiatry—were analyzed to provide context for the hospitalist claims.

The malpractice claims rate for hospitalists was significantly higher than the rate for internal medicine subspecialists (1.95 vs 1.30 claims per 100 physician-years; P < .001), though they were not significantly different from the rate for nonhospitalist general internal medicine physicians (1.95 vs 1.92 claims per 100 physician-years; P = .93) (Table 1). Compared with emergency medicine physicians, with whom hospitalists are sometimes compared due to both specialties being defined by their site of practice and the absence of longitudinal patient relationships, hospitalists had a significantly lower claims rate (1.95 vs 4.07 claims per 100 physician-years; P < .001).

An assessment of the temporal trends in the claims rates, based on a comparison between the two halves of the study period (2014-2018 vs 2009-2013), showed that the claims rate for hospitalists was increasing, but at a rate that did not reach statistical significance (Table 1). In contrast, the claims rates for the five other specialties assessed decreased over time, and the decreases were significant for four of these five other specialties (internal medicine subspecialties, emergency medicine, neurosurgery, and psychiatry).

Multiple claims against a single physician were uncommon in our hospitalist malpractice claims data. Among the 100 claims that were used to calculate the claims rates, one physician was named in 2 claims, and all the other physicians were named in only a single claim. Among all of the 1,216 hospitalist malpractice claims we analyzed, there were eight physicians who were named in more than 1 claim, seven of whom were named in 2 claims, and one of whom was named in 3 claims.

The median indemnity payment for hospitalist claims was $231,454 (interquartile range [IQR], $100,000-$503,015), similar to the median indemnity payment for neurosurgery ($233,723; IQR, $85,292-$697,872), though significantly greater than the median indemnity payment for the other four specialties studied (Table 2). Among the hospitalist claims, 29.9% resulted in an indemnity payment, not significantly different from the rate for nonhospitalist general internal medicine, internal medicine subspecialties, or neurosurgery, but significantly lower than the rate for emergency medicine (33.8%; P = .011). No significant temporal trend in the percentage of hospitalist claims that resulted in an indemnity payment was observed.

We performed a multivariable logistic regression analysis to assess the effect of different factors on the likelihood of a hospitalist case closing with an indemnity payment, compared with no payment (Table 3). In the multivariable model, the presence of an error in clinical judgment had an adjusted odds ratio (AOR) of 5.01 (95% CI, 3.37-7.45; P < .001) for a claim closing with payment, the largest effect found. The presence of problems with communication (AOR, 1.89; 95% CI, 1.42-2.51; P < .001), the clinical environment (eg, weekend/holiday or clinical busyness; AOR, 1.70; 95% CI, 1.20-2.40; P = .0026), and documentation (AOR, 1.65; 95% CI, 1.18-2.31; P = .0038) were also positive predictors of claims closing with payment. Greater patient age (per decade) was a negative predictor of the likelihood of a claim closing with payment (AOR, 0.92; 95% CI, 0.86-0.998), though it was of borderline statistical significance (P = .044).

We also assessed multiple clinical attributes of hospitalist malpractice claims, including the major allegation type and injury severity (Appendix Table). Among the 1,216 hospitalist malpractice claims studied, the most common allegation types were for errors related to medical treatment (n = 482; 39.6%), diagnosis (n = 446; 36.7%), and medications (n = 157; 12.9%). Among the hospitalist claims, 888 (73.0%) involved high-severity injury, and 674 (55.4%) involved the death of the patient. The percentages of cases involving high-severity injury and death were significantly greater for hospitalists, compared with that of the other specialties studied (P < .001 for all pairwise comparisons). Of the six specialties studied, hospital medicine was the only one in which the percentage of cases involving death exceeded 50%.

Hospital medicine is typically team-based, and we evaluated which other services were named in claims with hospital medicine as the primary responsible service. The clinician groups most commonly named in hospitalist claims were nursing (n = 269; 22.1%), followed by emergency medicine (n = 91; 7.5%), general surgery (n = 51; 4.2%), cardiology (n = 49; 4.0%), and orthopedic surgery (n = 46; 3.8%) (Appendix Figure). During the first 2 years of the study period, no physician assistants (PAs) or nurse practitioners (NPs) were named in hospitalist claims. Over the study period, the proportion of hospitalist cases also naming PAs and NPs increased steadily, reaching 6.9% and 6.2% of claims, respectively, in 2018 (Figure) (P < .001 for NPs and P = .037 for PAs based on a comparison between the two halves of the study period).

DISCUSSION

We found that the average annual claims rate for hospitalists was similar to that for nonhospitalist general internists (1.95 vs 1.92 claims per 100 physician-years) but significantly greater than that for internal medicine subspecialists (1.95 vs 1.30 claims per 100 physician-years). Hospitalist claims rates showed a notable temporal trend—a nonsignificant increase—over the study period (2009-2018). This contrasts with the five other specialties studied, all of which had decreasing claims rates, four of which were significant. An analysis of a different national malpractice claims database, the NPDB, found that the rate of paid malpractice claims overall decreased 55.7% during the period 1992-2014, again contrasting with the trend we found for hospitalist claims rates.¹⁷

We posit several explanations for why the malpractice claims rate trend for hospitalists has diverged from that of other specialties. There has been a large expansion in the number of hospitalists in the United States.¹⁸ With this increasing demand, many young physicians have entered the hospital medicine field. In a survey of general internal medicine physicians conducted by the Society of General Internal Medicine, 73% of hospitalists were aged 25 to 44 years, significantly greater than the 45% in this age range among nonhospitalist general internal medicine physicians.¹⁹ Hospitalists in their first year of practice have higher mortality rates than more experienced hospitalists.²⁰ Therefore, the relative inexperience of hospitalists, driven by this high demand, could be putting them at increased risk of medical errors and resulting malpractice claims. The higher mortality rate among hospitalists in their first year of practice could be due to a lack of familiarity with the systems of care, such as managing test results and obtaining appropriate consults.²⁰ This possibility suggests that enhanced training and mentorship could be valuable as a strategy to both improve the quality of care and reduce medicolegal risk. The increasing demand for hospitalists could also be affecting the qualification level of physicians entering the field.

Our analysis also showed that the severity of injury in hospitalist claims was greater than that for the other specialties studied. In addition, the percentage of claims involving death was greater for hospitalists than that for the other specialties. The increased acuity of inpatients, compared with that of outpatients—and the trend, at least for some conditions, of increased inpatient acuity over time^21,22— could account for the high injury severity seen among hospitalist claims. Given the positive correlation between injury severity and the size of indemnity payments made on malpractice claims,¹² the high injury severity seen in hospitalist claims was very likely a driver of the high indemnity payments observed among the hospitalist claims.

The relationship between injury severity and financial outcomes is supported by the results of our multivariable regression model (Table 3). Compared with medium-severity injury claims, both death and high-severity injury cases were significantly more likely to close with an indemnity payment (compared with no payment), with AORs of 1.79 (95% CI, 1.21-2.65) and 2.44 (95% CI, 1.54-3.87), respectively.

The most striking finding in our regression model was the magnitude of the effect of an error in clinical judgement. Cases coded with this contributing factor had five times the AOR of closing with payment (compared with no payment) (AOR, 5.01; 95% CI, 3.37-7.45). A clinical judgment call may be difficult to defend when it is ultimately associated with a bad patient outcome. The importance of clinical judgment in our analysis suggests a risk management strategy: clearly and contemporaneously documenting the rationale behind one’s clinical decision-making. This may help make a claim more defensible in the event of an adverse outcome by demonstrating that the clinician was acting reasonably based on the information available at the time. The importance of specifying a rationale for a clinical decision may be especially important in the era of electronic health records (EHRs). EHRs are not structured as chronologically linear charts, which can make it challenging during a trial to retrospectively show what information was available to the physician at the time the clinical decision was made. The importance of clinical judgment also affirms the importance of effective clinical decision support as a patient safety tool.²³

More broadly, it is notable that several contributing factors, including errors in clinical judgment (as discussed previously), problems with communication, and issues with the clinical environment, were significantly associated with malpractice cases closing with payment. This demonstrates that systematically examining malpractice claims to determine the underlying contributing factors can generate predictive analytics, as well as suggest risk management and patient safety strategies.

Interdisciplinary collaboration, as a component of systems-based practice, is a core principle of hospital medicine,¹³ and so we analyzed the involvement of other clinicians in hospitalist claims. Of the five specialties most frequently named in claims with hospitalists, two were surgical services: general surgery (n = 51; 4.2%) and orthopedic surgery (n = 46; 3.8%). With hospitalists being asked to play an increasing role in the care of surgical patients, they may be providing care to patient populations with whom they have less experience, which could put them at risk of adverse outcomes, leading to malpractice claims.^24,25 Hospitalists need to be attuned to the liability risks related to the care of patients requiring surgical management and ensure areas of responsibility are clearly delineated between the hospital medicine and surgical services.²⁶ We also found that hospitalist claims increasingly involve PAs and NPs, likely reflecting their increasing role in providing care on hospitalist services.^27,28The naming of these other services and provider types does not necessarily mean a breakdown in interdisciplinary collaboration occurred. Rather, these findings highlight the importance of including these surgical services, PAs, and NPs as integral participants in the patient safety and risk management efforts undertaken by hospitalists.

A prior analysis of claims rates for hospitalists that covered injury dates from 1997 to 2011 found that hospitalists had a relatively low claims rate, significantly lower than that for other internal medicine physicians.⁸ In addition to covering an earlier time period, that analysis based its claims rates on data from academic medical centers covered by a single insurer, and physicians at academic medical centers generally have lower claims rates, likely due, at least in part, to their spending a smaller proportion of their time on patient care, compared with nonacademic physicians.²⁹ Another analysis of hospitalist closed claims, which shared some cases with the cohort we analyzed, was performed by The Doctors Company, a commercial liability insurer.⁷ That analysis astutely emphasized the importance of breakdowns in diagnostic processes as a factor underlying hospitalist claims.

Our study has several limitations. First, although our database of malpractice claims includes approximately 31% of all the claims in the country and includes claims from every state, it may not be nationally representative. Another limitation relates to calculating the claims rates for physicians. Detailed information on the number of years of clinical activity, which is necessary to calculate claims rates, was available for only a subset of our data (8.2% of the hospitalist cases and 11.6% of all cases), so claims rates are based on this subset of our data (among which academic centers are overrepresented). Therefore, the claims rates should be interpreted with caution, especially regarding their application to the community hospital setting. The institutions included in the subset of our data used for determining claims rates were stable over time, so the use of a subset of our data for calculating claims rates reduces the generalizability of our claims rates but should not be a source of bias.

Potentially offsetting strengths of our claims database and study include the availability of unpaid claims (which outnumber paid claims roughly 2:1)^11,12; the presence of information on contributing factors and other case characteristics obtained through structured manual review of the cases; and the availability of the specialties of the clinicians involved. These features distinguish the database we used from the NPDB, another national database of malpractice claims, which does not include unpaid claims and which does not include information on contributing factors or physician specialty.

CONCLUSION

First described in 1996, the hospitalist field is the fastest growing specialty in modern medical history.^18,30 Therefore, an understanding of the malpractice risk of hospitalists is important and can shed light on the patient safety environment in hospitals. Our analysis showed that hospitalist malpractice claims rates remain roughly stable, in contrast to most other specialties, which have seen a fall in malpractice claims rates.¹⁷ In addition, unlike a previous analysis,⁸ we found that claims rates for hospitalists were essentially equal to those of other general internal medicine physicians (not lower, as had been previously reported), and higher than those of the internal medicine subspecialties. Hospitalist claims also have relatively high severity of injury. Potential factors driving these trends include the increasing demand for hospitalists, which results in a higher proportion of less-experienced physicians entering the field, and the expanding clinical scope of hospitalists, which may lead to their managing patients with conditions with which they may be less comfortable. Overall, our analysis suggests that the malpractice environment for hospitalists is becoming less favorable, and therefore, hospitalists should explore opportunities for mitigating liability risk and enhancing patient safety.

The prospect of facing a medical malpractice claim is a source of apprehension for physicians that affects physician behavior, including leading to defensive medicine.^1-3 Overall, annual defensive medicine costs have been estimated at $45.6 billion,⁴ and surveys of hospitalists indicate that 13.0% to 37.5% of hospitalist healthcare costs involve defensive medicine.^5,6 Despite the impact of malpractice concerns on hospitalist practice and the unprecedented growth of the field of hospital medicine, relatively few studies have examined the liability environment surrounding hospitalist practice.^7,8 The specific issue of malpractice claims rates faced by hospitalists has received even less attention in the medical literature.⁸

A better understanding of the contributing factors and other attributes of malpractice claims can help guide patient safety initiatives and inform hospitalists’ level of concern regarding liability. Although most medical errors do not result in a malpractice claim,^9,10 the majority of malpractice claims in which there is an indemnity payment involve medical injury due to clinician error.¹¹ Even malpractice claims that do not result in an indemnity payment represent opportunities to identify patient safety and risk management vulnerabilities.¹²

We used a national malpractice claims database to analyze the characteristics of claims made against hospitalists, including claims rates. In addition to claims rates, we also analyzed the other types of providers named in hospitalist claims given the importance of interdisciplinary collaboration to hospital medicine.^13,14 To provide context for understanding hospitalist liability data, we present data on other specialties. We also describe a model to predict whether hospitalist malpractice claims will close with an indemnity payment.

METHODS

Data Sources and Elements

This analysis used a repository of malpractice claims maintained by CRICO, the captive malpractice insurer of the Harvard-affiliated medical institutions. This database, the Comparative Benchmarking System, aggregates malpractice claims from multiple other malpractice carriers (both commercial and captive). This database comprises approximately 31% of all malpractice claims in the United States (both paid and unpaid) according to a comparison with claims in the National Practitioner Data Bank (NPDB), which is a government-run database to which all paid malpractice claims against physicians must be reported.¹⁵ To compile this national database, trained nurse coders, who have access to the medical records and legal documents from each case, abstract information into coded fields, such as the type of allegation, contributing factors, and injury severity. The information in these coded fields is based on a taxonomy that was developed specifically for use in malpractice cases. The nurse coders participate in regular meetings in which cases are discussed to ensure that they are applying the coding and taxonomy uniformly.

Injury severity was based on a widely used scale developed for malpractice claims by the National Association of Insurance Commissioners.¹⁶ Low injury severity included emotional injury and temporary insignificant injury. Medium injury severity included temporary minor, temporary major, and permanent minor injury. High injury severity included permanent significant injury through death. Because this study used a database assembled for operational and patient safety purposes and was not human subjects research, institutional review board approval was not needed.

Study Cohort

Malpractice claims included formal lawsuits or written requests for compensation for negligent medical care. Hospitalist malpractice claims were defined as those in which the nurse coders determined that the internal medicine hospitalist service was the clinical service that was primarily responsible for the care of the patient at the time of the clinical event giving rise to the claim. Malpractice claims closed during the period 2009-2018 were included, regardless of whether an indemnity payment was made. Claims against select other medical specialties (nonhospitalist general internal medicine, internal medicine subspecialists, emergency medicine, neurosurgery, and psychiatry), defined by the specialty of the physician named in the claim, were included for comparison with the characteristics of the claims against hospitalists. Physician specialty was available for all physicians in the study cohort; however, given that malpractice cases were aggregated from multiple different insurers, full demographic information (eg, age and sex) was not available for all physicians.

Statistical Analysis

Malpractice claims rates were treated as Poisson rates and compared using a Z-test. Malpractice claims rates are expressed as claims per 100 physician-years. Each physician-year represents 1 year of coverage of one physician by the medical malpractice carrier whose data were used. Physician-years represent the duration of time physicians practiced during which they were insured by the malpractice carrier and, as such, could have been subject to a malpractice claim that would have been included in our data. Claims rates are based on the subset of the malpractice claims in the study for which the number of physician-years of coverage is available, representing 8.2% of hospitalist claims and 11.6% of all claims.

Comparisons of the percentages of cases closing with an indemnity payment, as well as the percentages of cases in different allegation type and clinical severity categories, were made using the Fisher exact test. Indemnity payment amounts were inflation-adjusted to 2018 dollars using the Consumer Price Index. Comparisons of indemnity payment amounts between physician specialties were carried out using the Wilcoxon rank sum test given that the distribution of the payment amounts appeared nonnormal; this was confirmed with the Shapiro-Wilk test. A multivariable logistic regression model was developed to predict the binary outcome of whether a hospitalist case would close with an indemnity payment (compared with no payment), based on the 1,216 hospitalist claims. The predictors used in this regression model were chosen a priori based on hypotheses about what factors drive the likelihood that a case closes with payment. Both the unadjusted and adjusted odds ratios for the predictors are presented. The adjusted model is adjusted for all the other predictors contained in the model. All reported P values are two-sided. The statistical analysis was carried out using JMP Pro version 15 (SAS Institute Inc) and Minitab version 19 (Minitab LLC).

RESULTS

We identified 1,216 hospital medicine malpractice claims from our database. Claims rates were calculated from the subset of our data for which physician-years were available—including 5,140 physician-years encompassing 100 claims, representing 8.2% of all hospitalist claims studied. An additional 18,644 malpractice claims from five other specialties—nonhospitalist general internal medicine, internal medicine subspecialists, emergency medicine, neurosurgery, and psychiatry—were analyzed to provide context for the hospitalist claims.

The malpractice claims rate for hospitalists was significantly higher than the rate for internal medicine subspecialists (1.95 vs 1.30 claims per 100 physician-years; P < .001), though they were not significantly different from the rate for nonhospitalist general internal medicine physicians (1.95 vs 1.92 claims per 100 physician-years; P = .93) (Table 1). Compared with emergency medicine physicians, with whom hospitalists are sometimes compared due to both specialties being defined by their site of practice and the absence of longitudinal patient relationships, hospitalists had a significantly lower claims rate (1.95 vs 4.07 claims per 100 physician-years; P < .001).

An assessment of the temporal trends in the claims rates, based on a comparison between the two halves of the study period (2014-2018 vs 2009-2013), showed that the claims rate for hospitalists was increasing, but at a rate that did not reach statistical significance (Table 1). In contrast, the claims rates for the five other specialties assessed decreased over time, and the decreases were significant for four of these five other specialties (internal medicine subspecialties, emergency medicine, neurosurgery, and psychiatry).

Multiple claims against a single physician were uncommon in our hospitalist malpractice claims data. Among the 100 claims that were used to calculate the claims rates, one physician was named in 2 claims, and all the other physicians were named in only a single claim. Among all of the 1,216 hospitalist malpractice claims we analyzed, there were eight physicians who were named in more than 1 claim, seven of whom were named in 2 claims, and one of whom was named in 3 claims.

The median indemnity payment for hospitalist claims was $231,454 (interquartile range [IQR], $100,000-$503,015), similar to the median indemnity payment for neurosurgery ($233,723; IQR, $85,292-$697,872), though significantly greater than the median indemnity payment for the other four specialties studied (Table 2). Among the hospitalist claims, 29.9% resulted in an indemnity payment, not significantly different from the rate for nonhospitalist general internal medicine, internal medicine subspecialties, or neurosurgery, but significantly lower than the rate for emergency medicine (33.8%; P = .011). No significant temporal trend in the percentage of hospitalist claims that resulted in an indemnity payment was observed.

We performed a multivariable logistic regression analysis to assess the effect of different factors on the likelihood of a hospitalist case closing with an indemnity payment, compared with no payment (Table 3). In the multivariable model, the presence of an error in clinical judgment had an adjusted odds ratio (AOR) of 5.01 (95% CI, 3.37-7.45; P < .001) for a claim closing with payment, the largest effect found. The presence of problems with communication (AOR, 1.89; 95% CI, 1.42-2.51; P < .001), the clinical environment (eg, weekend/holiday or clinical busyness; AOR, 1.70; 95% CI, 1.20-2.40; P = .0026), and documentation (AOR, 1.65; 95% CI, 1.18-2.31; P = .0038) were also positive predictors of claims closing with payment. Greater patient age (per decade) was a negative predictor of the likelihood of a claim closing with payment (AOR, 0.92; 95% CI, 0.86-0.998), though it was of borderline statistical significance (P = .044).

We also assessed multiple clinical attributes of hospitalist malpractice claims, including the major allegation type and injury severity (Appendix Table). Among the 1,216 hospitalist malpractice claims studied, the most common allegation types were for errors related to medical treatment (n = 482; 39.6%), diagnosis (n = 446; 36.7%), and medications (n = 157; 12.9%). Among the hospitalist claims, 888 (73.0%) involved high-severity injury, and 674 (55.4%) involved the death of the patient. The percentages of cases involving high-severity injury and death were significantly greater for hospitalists, compared with that of the other specialties studied (P < .001 for all pairwise comparisons). Of the six specialties studied, hospital medicine was the only one in which the percentage of cases involving death exceeded 50%.

Hospital medicine is typically team-based, and we evaluated which other services were named in claims with hospital medicine as the primary responsible service. The clinician groups most commonly named in hospitalist claims were nursing (n = 269; 22.1%), followed by emergency medicine (n = 91; 7.5%), general surgery (n = 51; 4.2%), cardiology (n = 49; 4.0%), and orthopedic surgery (n = 46; 3.8%) (Appendix Figure). During the first 2 years of the study period, no physician assistants (PAs) or nurse practitioners (NPs) were named in hospitalist claims. Over the study period, the proportion of hospitalist cases also naming PAs and NPs increased steadily, reaching 6.9% and 6.2% of claims, respectively, in 2018 (Figure) (P < .001 for NPs and P = .037 for PAs based on a comparison between the two halves of the study period).

DISCUSSION

We found that the average annual claims rate for hospitalists was similar to that for nonhospitalist general internists (1.95 vs 1.92 claims per 100 physician-years) but significantly greater than that for internal medicine subspecialists (1.95 vs 1.30 claims per 100 physician-years). Hospitalist claims rates showed a notable temporal trend—a nonsignificant increase—over the study period (2009-2018). This contrasts with the five other specialties studied, all of which had decreasing claims rates, four of which were significant. An analysis of a different national malpractice claims database, the NPDB, found that the rate of paid malpractice claims overall decreased 55.7% during the period 1992-2014, again contrasting with the trend we found for hospitalist claims rates.¹⁷

We posit several explanations for why the malpractice claims rate trend for hospitalists has diverged from that of other specialties. There has been a large expansion in the number of hospitalists in the United States.¹⁸ With this increasing demand, many young physicians have entered the hospital medicine field. In a survey of general internal medicine physicians conducted by the Society of General Internal Medicine, 73% of hospitalists were aged 25 to 44 years, significantly greater than the 45% in this age range among nonhospitalist general internal medicine physicians.¹⁹ Hospitalists in their first year of practice have higher mortality rates than more experienced hospitalists.²⁰ Therefore, the relative inexperience of hospitalists, driven by this high demand, could be putting them at increased risk of medical errors and resulting malpractice claims. The higher mortality rate among hospitalists in their first year of practice could be due to a lack of familiarity with the systems of care, such as managing test results and obtaining appropriate consults.²⁰ This possibility suggests that enhanced training and mentorship could be valuable as a strategy to both improve the quality of care and reduce medicolegal risk. The increasing demand for hospitalists could also be affecting the qualification level of physicians entering the field.

Our analysis also showed that the severity of injury in hospitalist claims was greater than that for the other specialties studied. In addition, the percentage of claims involving death was greater for hospitalists than that for the other specialties. The increased acuity of inpatients, compared with that of outpatients—and the trend, at least for some conditions, of increased inpatient acuity over time^21,22— could account for the high injury severity seen among hospitalist claims. Given the positive correlation between injury severity and the size of indemnity payments made on malpractice claims,¹² the high injury severity seen in hospitalist claims was very likely a driver of the high indemnity payments observed among the hospitalist claims.

The relationship between injury severity and financial outcomes is supported by the results of our multivariable regression model (Table 3). Compared with medium-severity injury claims, both death and high-severity injury cases were significantly more likely to close with an indemnity payment (compared with no payment), with AORs of 1.79 (95% CI, 1.21-2.65) and 2.44 (95% CI, 1.54-3.87), respectively.

The most striking finding in our regression model was the magnitude of the effect of an error in clinical judgement. Cases coded with this contributing factor had five times the AOR of closing with payment (compared with no payment) (AOR, 5.01; 95% CI, 3.37-7.45). A clinical judgment call may be difficult to defend when it is ultimately associated with a bad patient outcome. The importance of clinical judgment in our analysis suggests a risk management strategy: clearly and contemporaneously documenting the rationale behind one’s clinical decision-making. This may help make a claim more defensible in the event of an adverse outcome by demonstrating that the clinician was acting reasonably based on the information available at the time. The importance of specifying a rationale for a clinical decision may be especially important in the era of electronic health records (EHRs). EHRs are not structured as chronologically linear charts, which can make it challenging during a trial to retrospectively show what information was available to the physician at the time the clinical decision was made. The importance of clinical judgment also affirms the importance of effective clinical decision support as a patient safety tool.²³

More broadly, it is notable that several contributing factors, including errors in clinical judgment (as discussed previously), problems with communication, and issues with the clinical environment, were significantly associated with malpractice cases closing with payment. This demonstrates that systematically examining malpractice claims to determine the underlying contributing factors can generate predictive analytics, as well as suggest risk management and patient safety strategies.

Interdisciplinary collaboration, as a component of systems-based practice, is a core principle of hospital medicine,¹³ and so we analyzed the involvement of other clinicians in hospitalist claims. Of the five specialties most frequently named in claims with hospitalists, two were surgical services: general surgery (n = 51; 4.2%) and orthopedic surgery (n = 46; 3.8%). With hospitalists being asked to play an increasing role in the care of surgical patients, they may be providing care to patient populations with whom they have less experience, which could put them at risk of adverse outcomes, leading to malpractice claims.^24,25 Hospitalists need to be attuned to the liability risks related to the care of patients requiring surgical management and ensure areas of responsibility are clearly delineated between the hospital medicine and surgical services.²⁶ We also found that hospitalist claims increasingly involve PAs and NPs, likely reflecting their increasing role in providing care on hospitalist services.^27,28The naming of these other services and provider types does not necessarily mean a breakdown in interdisciplinary collaboration occurred. Rather, these findings highlight the importance of including these surgical services, PAs, and NPs as integral participants in the patient safety and risk management efforts undertaken by hospitalists.

A prior analysis of claims rates for hospitalists that covered injury dates from 1997 to 2011 found that hospitalists had a relatively low claims rate, significantly lower than that for other internal medicine physicians.⁸ In addition to covering an earlier time period, that analysis based its claims rates on data from academic medical centers covered by a single insurer, and physicians at academic medical centers generally have lower claims rates, likely due, at least in part, to their spending a smaller proportion of their time on patient care, compared with nonacademic physicians.²⁹ Another analysis of hospitalist closed claims, which shared some cases with the cohort we analyzed, was performed by The Doctors Company, a commercial liability insurer.⁷ That analysis astutely emphasized the importance of breakdowns in diagnostic processes as a factor underlying hospitalist claims.

Our study has several limitations. First, although our database of malpractice claims includes approximately 31% of all the claims in the country and includes claims from every state, it may not be nationally representative. Another limitation relates to calculating the claims rates for physicians. Detailed information on the number of years of clinical activity, which is necessary to calculate claims rates, was available for only a subset of our data (8.2% of the hospitalist cases and 11.6% of all cases), so claims rates are based on this subset of our data (among which academic centers are overrepresented). Therefore, the claims rates should be interpreted with caution, especially regarding their application to the community hospital setting. The institutions included in the subset of our data used for determining claims rates were stable over time, so the use of a subset of our data for calculating claims rates reduces the generalizability of our claims rates but should not be a source of bias.

Potentially offsetting strengths of our claims database and study include the availability of unpaid claims (which outnumber paid claims roughly 2:1)^11,12; the presence of information on contributing factors and other case characteristics obtained through structured manual review of the cases; and the availability of the specialties of the clinicians involved. These features distinguish the database we used from the NPDB, another national database of malpractice claims, which does not include unpaid claims and which does not include information on contributing factors or physician specialty.

CONCLUSION

First described in 1996, the hospitalist field is the fastest growing specialty in modern medical history.^18,30 Therefore, an understanding of the malpractice risk of hospitalists is important and can shed light on the patient safety environment in hospitals. Our analysis showed that hospitalist malpractice claims rates remain roughly stable, in contrast to most other specialties, which have seen a fall in malpractice claims rates.¹⁷ In addition, unlike a previous analysis,⁸ we found that claims rates for hospitalists were essentially equal to those of other general internal medicine physicians (not lower, as had been previously reported), and higher than those of the internal medicine subspecialties. Hospitalist claims also have relatively high severity of injury. Potential factors driving these trends include the increasing demand for hospitalists, which results in a higher proportion of less-experienced physicians entering the field, and the expanding clinical scope of hospitalists, which may lead to their managing patients with conditions with which they may be less comfortable. Overall, our analysis suggests that the malpractice environment for hospitalists is becoming less favorable, and therefore, hospitalists should explore opportunities for mitigating liability risk and enhancing patient safety.

The prospect of facing a medical malpractice claim is a source of apprehension for physicians that affects physician behavior, including leading to defensive medicine.^1-3 Overall, annual defensive medicine costs have been estimated at $45.6 billion,⁴ and surveys of hospitalists indicate that 13.0% to 37.5% of hospitalist healthcare costs involve defensive medicine.^5,6 Despite the impact of malpractice concerns on hospitalist practice and the unprecedented growth of the field of hospital medicine, relatively few studies have examined the liability environment surrounding hospitalist practice.^7,8 The specific issue of malpractice claims rates faced by hospitalists has received even less attention in the medical literature.⁸

A better understanding of the contributing factors and other attributes of malpractice claims can help guide patient safety initiatives and inform hospitalists’ level of concern regarding liability. Although most medical errors do not result in a malpractice claim,^9,10 the majority of malpractice claims in which there is an indemnity payment involve medical injury due to clinician error.¹¹ Even malpractice claims that do not result in an indemnity payment represent opportunities to identify patient safety and risk management vulnerabilities.¹²

We used a national malpractice claims database to analyze the characteristics of claims made against hospitalists, including claims rates. In addition to claims rates, we also analyzed the other types of providers named in hospitalist claims given the importance of interdisciplinary collaboration to hospital medicine.^13,14 To provide context for understanding hospitalist liability data, we present data on other specialties. We also describe a model to predict whether hospitalist malpractice claims will close with an indemnity payment.

METHODS

Data Sources and Elements

This analysis used a repository of malpractice claims maintained by CRICO, the captive malpractice insurer of the Harvard-affiliated medical institutions. This database, the Comparative Benchmarking System, aggregates malpractice claims from multiple other malpractice carriers (both commercial and captive). This database comprises approximately 31% of all malpractice claims in the United States (both paid and unpaid) according to a comparison with claims in the National Practitioner Data Bank (NPDB), which is a government-run database to which all paid malpractice claims against physicians must be reported.¹⁵ To compile this national database, trained nurse coders, who have access to the medical records and legal documents from each case, abstract information into coded fields, such as the type of allegation, contributing factors, and injury severity. The information in these coded fields is based on a taxonomy that was developed specifically for use in malpractice cases. The nurse coders participate in regular meetings in which cases are discussed to ensure that they are applying the coding and taxonomy uniformly.

Injury severity was based on a widely used scale developed for malpractice claims by the National Association of Insurance Commissioners.¹⁶ Low injury severity included emotional injury and temporary insignificant injury. Medium injury severity included temporary minor, temporary major, and permanent minor injury. High injury severity included permanent significant injury through death. Because this study used a database assembled for operational and patient safety purposes and was not human subjects research, institutional review board approval was not needed.

Study Cohort

Malpractice claims included formal lawsuits or written requests for compensation for negligent medical care. Hospitalist malpractice claims were defined as those in which the nurse coders determined that the internal medicine hospitalist service was the clinical service that was primarily responsible for the care of the patient at the time of the clinical event giving rise to the claim. Malpractice claims closed during the period 2009-2018 were included, regardless of whether an indemnity payment was made. Claims against select other medical specialties (nonhospitalist general internal medicine, internal medicine subspecialists, emergency medicine, neurosurgery, and psychiatry), defined by the specialty of the physician named in the claim, were included for comparison with the characteristics of the claims against hospitalists. Physician specialty was available for all physicians in the study cohort; however, given that malpractice cases were aggregated from multiple different insurers, full demographic information (eg, age and sex) was not available for all physicians.

Statistical Analysis

Malpractice claims rates were treated as Poisson rates and compared using a Z-test. Malpractice claims rates are expressed as claims per 100 physician-years. Each physician-year represents 1 year of coverage of one physician by the medical malpractice carrier whose data were used. Physician-years represent the duration of time physicians practiced during which they were insured by the malpractice carrier and, as such, could have been subject to a malpractice claim that would have been included in our data. Claims rates are based on the subset of the malpractice claims in the study for which the number of physician-years of coverage is available, representing 8.2% of hospitalist claims and 11.6% of all claims.

Comparisons of the percentages of cases closing with an indemnity payment, as well as the percentages of cases in different allegation type and clinical severity categories, were made using the Fisher exact test. Indemnity payment amounts were inflation-adjusted to 2018 dollars using the Consumer Price Index. Comparisons of indemnity payment amounts between physician specialties were carried out using the Wilcoxon rank sum test given that the distribution of the payment amounts appeared nonnormal; this was confirmed with the Shapiro-Wilk test. A multivariable logistic regression model was developed to predict the binary outcome of whether a hospitalist case would close with an indemnity payment (compared with no payment), based on the 1,216 hospitalist claims. The predictors used in this regression model were chosen a priori based on hypotheses about what factors drive the likelihood that a case closes with payment. Both the unadjusted and adjusted odds ratios for the predictors are presented. The adjusted model is adjusted for all the other predictors contained in the model. All reported P values are two-sided. The statistical analysis was carried out using JMP Pro version 15 (SAS Institute Inc) and Minitab version 19 (Minitab LLC).

RESULTS

We identified 1,216 hospital medicine malpractice claims from our database. Claims rates were calculated from the subset of our data for which physician-years were available—including 5,140 physician-years encompassing 100 claims, representing 8.2% of all hospitalist claims studied. An additional 18,644 malpractice claims from five other specialties—nonhospitalist general internal medicine, internal medicine subspecialists, emergency medicine, neurosurgery, and psychiatry—were analyzed to provide context for the hospitalist claims.

The malpractice claims rate for hospitalists was significantly higher than the rate for internal medicine subspecialists (1.95 vs 1.30 claims per 100 physician-years; P < .001), though they were not significantly different from the rate for nonhospitalist general internal medicine physicians (1.95 vs 1.92 claims per 100 physician-years; P = .93) (Table 1). Compared with emergency medicine physicians, with whom hospitalists are sometimes compared due to both specialties being defined by their site of practice and the absence of longitudinal patient relationships, hospitalists had a significantly lower claims rate (1.95 vs 4.07 claims per 100 physician-years; P < .001).

An assessment of the temporal trends in the claims rates, based on a comparison between the two halves of the study period (2014-2018 vs 2009-2013), showed that the claims rate for hospitalists was increasing, but at a rate that did not reach statistical significance (Table 1). In contrast, the claims rates for the five other specialties assessed decreased over time, and the decreases were significant for four of these five other specialties (internal medicine subspecialties, emergency medicine, neurosurgery, and psychiatry).

Multiple claims against a single physician were uncommon in our hospitalist malpractice claims data. Among the 100 claims that were used to calculate the claims rates, one physician was named in 2 claims, and all the other physicians were named in only a single claim. Among all of the 1,216 hospitalist malpractice claims we analyzed, there were eight physicians who were named in more than 1 claim, seven of whom were named in 2 claims, and one of whom was named in 3 claims.

The median indemnity payment for hospitalist claims was $231,454 (interquartile range [IQR], $100,000-$503,015), similar to the median indemnity payment for neurosurgery ($233,723; IQR, $85,292-$697,872), though significantly greater than the median indemnity payment for the other four specialties studied (Table 2). Among the hospitalist claims, 29.9% resulted in an indemnity payment, not significantly different from the rate for nonhospitalist general internal medicine, internal medicine subspecialties, or neurosurgery, but significantly lower than the rate for emergency medicine (33.8%; P = .011). No significant temporal trend in the percentage of hospitalist claims that resulted in an indemnity payment was observed.

We performed a multivariable logistic regression analysis to assess the effect of different factors on the likelihood of a hospitalist case closing with an indemnity payment, compared with no payment (Table 3). In the multivariable model, the presence of an error in clinical judgment had an adjusted odds ratio (AOR) of 5.01 (95% CI, 3.37-7.45; P < .001) for a claim closing with payment, the largest effect found. The presence of problems with communication (AOR, 1.89; 95% CI, 1.42-2.51; P < .001), the clinical environment (eg, weekend/holiday or clinical busyness; AOR, 1.70; 95% CI, 1.20-2.40; P = .0026), and documentation (AOR, 1.65; 95% CI, 1.18-2.31; P = .0038) were also positive predictors of claims closing with payment. Greater patient age (per decade) was a negative predictor of the likelihood of a claim closing with payment (AOR, 0.92; 95% CI, 0.86-0.998), though it was of borderline statistical significance (P = .044).

We also assessed multiple clinical attributes of hospitalist malpractice claims, including the major allegation type and injury severity (Appendix Table). Among the 1,216 hospitalist malpractice claims studied, the most common allegation types were for errors related to medical treatment (n = 482; 39.6%), diagnosis (n = 446; 36.7%), and medications (n = 157; 12.9%). Among the hospitalist claims, 888 (73.0%) involved high-severity injury, and 674 (55.4%) involved the death of the patient. The percentages of cases involving high-severity injury and death were significantly greater for hospitalists, compared with that of the other specialties studied (P < .001 for all pairwise comparisons). Of the six specialties studied, hospital medicine was the only one in which the percentage of cases involving death exceeded 50%.

Hospital medicine is typically team-based, and we evaluated which other services were named in claims with hospital medicine as the primary responsible service. The clinician groups most commonly named in hospitalist claims were nursing (n = 269; 22.1%), followed by emergency medicine (n = 91; 7.5%), general surgery (n = 51; 4.2%), cardiology (n = 49; 4.0%), and orthopedic surgery (n = 46; 3.8%) (Appendix Figure). During the first 2 years of the study period, no physician assistants (PAs) or nurse practitioners (NPs) were named in hospitalist claims. Over the study period, the proportion of hospitalist cases also naming PAs and NPs increased steadily, reaching 6.9% and 6.2% of claims, respectively, in 2018 (Figure) (P < .001 for NPs and P = .037 for PAs based on a comparison between the two halves of the study period).

DISCUSSION

We found that the average annual claims rate for hospitalists was similar to that for nonhospitalist general internists (1.95 vs 1.92 claims per 100 physician-years) but significantly greater than that for internal medicine subspecialists (1.95 vs 1.30 claims per 100 physician-years). Hospitalist claims rates showed a notable temporal trend—a nonsignificant increase—over the study period (2009-2018). This contrasts with the five other specialties studied, all of which had decreasing claims rates, four of which were significant. An analysis of a different national malpractice claims database, the NPDB, found that the rate of paid malpractice claims overall decreased 55.7% during the period 1992-2014, again contrasting with the trend we found for hospitalist claims rates.¹⁷

We posit several explanations for why the malpractice claims rate trend for hospitalists has diverged from that of other specialties. There has been a large expansion in the number of hospitalists in the United States.¹⁸ With this increasing demand, many young physicians have entered the hospital medicine field. In a survey of general internal medicine physicians conducted by the Society of General Internal Medicine, 73% of hospitalists were aged 25 to 44 years, significantly greater than the 45% in this age range among nonhospitalist general internal medicine physicians.¹⁹ Hospitalists in their first year of practice have higher mortality rates than more experienced hospitalists.²⁰ Therefore, the relative inexperience of hospitalists, driven by this high demand, could be putting them at increased risk of medical errors and resulting malpractice claims. The higher mortality rate among hospitalists in their first year of practice could be due to a lack of familiarity with the systems of care, such as managing test results and obtaining appropriate consults.²⁰ This possibility suggests that enhanced training and mentorship could be valuable as a strategy to both improve the quality of care and reduce medicolegal risk. The increasing demand for hospitalists could also be affecting the qualification level of physicians entering the field.

Our analysis also showed that the severity of injury in hospitalist claims was greater than that for the other specialties studied. In addition, the percentage of claims involving death was greater for hospitalists than that for the other specialties. The increased acuity of inpatients, compared with that of outpatients—and the trend, at least for some conditions, of increased inpatient acuity over time^21,22— could account for the high injury severity seen among hospitalist claims. Given the positive correlation between injury severity and the size of indemnity payments made on malpractice claims,¹² the high injury severity seen in hospitalist claims was very likely a driver of the high indemnity payments observed among the hospitalist claims.

The relationship between injury severity and financial outcomes is supported by the results of our multivariable regression model (Table 3). Compared with medium-severity injury claims, both death and high-severity injury cases were significantly more likely to close with an indemnity payment (compared with no payment), with AORs of 1.79 (95% CI, 1.21-2.65) and 2.44 (95% CI, 1.54-3.87), respectively.

The most striking finding in our regression model was the magnitude of the effect of an error in clinical judgement. Cases coded with this contributing factor had five times the AOR of closing with payment (compared with no payment) (AOR, 5.01; 95% CI, 3.37-7.45). A clinical judgment call may be difficult to defend when it is ultimately associated with a bad patient outcome. The importance of clinical judgment in our analysis suggests a risk management strategy: clearly and contemporaneously documenting the rationale behind one’s clinical decision-making. This may help make a claim more defensible in the event of an adverse outcome by demonstrating that the clinician was acting reasonably based on the information available at the time. The importance of specifying a rationale for a clinical decision may be especially important in the era of electronic health records (EHRs). EHRs are not structured as chronologically linear charts, which can make it challenging during a trial to retrospectively show what information was available to the physician at the time the clinical decision was made. The importance of clinical judgment also affirms the importance of effective clinical decision support as a patient safety tool.²³

More broadly, it is notable that several contributing factors, including errors in clinical judgment (as discussed previously), problems with communication, and issues with the clinical environment, were significantly associated with malpractice cases closing with payment. This demonstrates that systematically examining malpractice claims to determine the underlying contributing factors can generate predictive analytics, as well as suggest risk management and patient safety strategies.

Interdisciplinary collaboration, as a component of systems-based practice, is a core principle of hospital medicine,¹³ and so we analyzed the involvement of other clinicians in hospitalist claims. Of the five specialties most frequently named in claims with hospitalists, two were surgical services: general surgery (n = 51; 4.2%) and orthopedic surgery (n = 46; 3.8%). With hospitalists being asked to play an increasing role in the care of surgical patients, they may be providing care to patient populations with whom they have less experience, which could put them at risk of adverse outcomes, leading to malpractice claims.^24,25 Hospitalists need to be attuned to the liability risks related to the care of patients requiring surgical management and ensure areas of responsibility are clearly delineated between the hospital medicine and surgical services.²⁶ We also found that hospitalist claims increasingly involve PAs and NPs, likely reflecting their increasing role in providing care on hospitalist services.^27,28The naming of these other services and provider types does not necessarily mean a breakdown in interdisciplinary collaboration occurred. Rather, these findings highlight the importance of including these surgical services, PAs, and NPs as integral participants in the patient safety and risk management efforts undertaken by hospitalists.

A prior analysis of claims rates for hospitalists that covered injury dates from 1997 to 2011 found that hospitalists had a relatively low claims rate, significantly lower than that for other internal medicine physicians.⁸ In addition to covering an earlier time period, that analysis based its claims rates on data from academic medical centers covered by a single insurer, and physicians at academic medical centers generally have lower claims rates, likely due, at least in part, to their spending a smaller proportion of their time on patient care, compared with nonacademic physicians.²⁹ Another analysis of hospitalist closed claims, which shared some cases with the cohort we analyzed, was performed by The Doctors Company, a commercial liability insurer.⁷ That analysis astutely emphasized the importance of breakdowns in diagnostic processes as a factor underlying hospitalist claims.

Our study has several limitations. First, although our database of malpractice claims includes approximately 31% of all the claims in the country and includes claims from every state, it may not be nationally representative. Another limitation relates to calculating the claims rates for physicians. Detailed information on the number of years of clinical activity, which is necessary to calculate claims rates, was available for only a subset of our data (8.2% of the hospitalist cases and 11.6% of all cases), so claims rates are based on this subset of our data (among which academic centers are overrepresented). Therefore, the claims rates should be interpreted with caution, especially regarding their application to the community hospital setting. The institutions included in the subset of our data used for determining claims rates were stable over time, so the use of a subset of our data for calculating claims rates reduces the generalizability of our claims rates but should not be a source of bias.

Potentially offsetting strengths of our claims database and study include the availability of unpaid claims (which outnumber paid claims roughly 2:1)^11,12; the presence of information on contributing factors and other case characteristics obtained through structured manual review of the cases; and the availability of the specialties of the clinicians involved. These features distinguish the database we used from the NPDB, another national database of malpractice claims, which does not include unpaid claims and which does not include information on contributing factors or physician specialty.

CONCLUSION

First described in 1996, the hospitalist field is the fastest growing specialty in modern medical history.^18,30 Therefore, an understanding of the malpractice risk of hospitalists is important and can shed light on the patient safety environment in hospitals. Our analysis showed that hospitalist malpractice claims rates remain roughly stable, in contrast to most other specialties, which have seen a fall in malpractice claims rates.¹⁷ In addition, unlike a previous analysis,⁸ we found that claims rates for hospitalists were essentially equal to those of other general internal medicine physicians (not lower, as had been previously reported), and higher than those of the internal medicine subspecialties. Hospitalist claims also have relatively high severity of injury. Potential factors driving these trends include the increasing demand for hospitalists, which results in a higher proportion of less-experienced physicians entering the field, and the expanding clinical scope of hospitalists, which may lead to their managing patients with conditions with which they may be less comfortable. Overall, our analysis suggests that the malpractice environment for hospitalists is becoming less favorable, and therefore, hospitalists should explore opportunities for mitigating liability risk and enhancing patient safety.

The Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines define acute exacerbations of chronic obstructive pulmonary disease (AECOPD) as a sudden worsening of respiratory symptoms that require additional interventions. Exacerbations are classified as mild (treated with short-acting bronchodilators only), moderate (treated with antibiotics and/or oral corticosteroids), or severe (treatment requiring hospitalization). Exacerbations must include increased dyspnea, and other symptoms may involve increased sputum volume and purulence, cough, and a change in sputum color. These symptoms can be due to viral, bacterial, or environmental causes, with viral respiratory infections being the most common cause.^1-4 However, determining the etiology of an exacerbation can be difficult based on symptoms alone and can lead to an excessive and unnecessary use of antibiotics. Only the change in sputum color is considered highly sensitive and specific for bacterial causes.¹ As a result, there has been an increased interest in the use of acute biomarkers to determine whether antibiotics are necessary.

Procalcitonin (PCT) is an acute phase reactant that increases in response to inflammation, especially inflammation caused by a bacterial infection. Recent studies have suggested that PCT may be used in patients experiencing an AECOPD to reduce antibiotic use without impacting rates of treatment failure.^5-9 A majority of these studies have been in the inpatient setting or a combination of inpatient and outpatient settings.

The purpose of this study was to create and to evaluate the efficacy and practicality of a PCT-based algorithm to aid emergency department (ED) clinicians in the evaluation of patients with AECOPD who do not require hospitalization. The primary outcome of this project was the rate of antibiotic prescriptions before and after the initiation of the algorithm.

Methods

This was an observational, retrospective, pre/post assessment at the Phoenix Veterans Affairs Health Care System (PVAHCS) in Arizona. Patients who were discharged from the ED with a diagnosis of an AECOPD were identified using International Classification of Diseases, Tenth Revision (ICD-10) codes. Patient charts were reviewed from November 2018 to March 2019 for the preimplementation group and from November 2019 for March 2020 in the postimplementation group. The periods were chosen to reflect similar seasons for both the pre- and postimplementation interventions. Patients were excluded from analysis if they required hospital admission, were immunocompromised, on chronic antimicrobial therapy, had no documented medical history of COPD, or if they were presenting primarily for medication refills. Information collected included the rate of antibiotic prescriptions, procalcitonin test orders, COPD GOLD classification, and 30-, 60-, and 90-day reexacerbation rates.

A PCT-based algorithm (Figure 1) was developed and approved by the PVAHCS Antimicrobial Stewardship Program, the Pharmacy and Therapeutics committee, and ED leadership. PCT threshold values were based on values approved by the US Food and Drug Administration and previous studies—antibiotics were discouraged for PCT levels ≤ 0.25 ng/mL but could be considered for PCT levels > 0.25 ng/mL.^5,8,9 Clinicians were not required to use the algorithm, and the use of clinical judgement was encouraged. The recommended antibiotic therapies were based on previously approved PVAHCS antimicrobial stewardship guidance. To promote utilization, a PCT quick order option was added to the ED laboratory order menu.

ED clinicians were individually educated by the antimicrobial stewardship and emergency medicine pharmacists, an infectious disease physician champion, and the pharmacy resident. Clinicians were educated about PCT and its use in the setting of AECOPD to aid in the determination of bacterial infections. Each clinician received an electronic copy the algorithm and summary of the study protocol before implementation and 3 months after implementation for follow-up education. In addition, a printed copy of the algorithm was posted in multiple clinician workstations within the ED. For the first month of implementation, the project lead was available full-time in the ED to encourage algorithm use and to field questions or concerns from clinicians.

Outcome Measures

The primary outcome was the rate of antibiotic prescriptions pre- and postintervention. The safety endpoints were 30-, 60-, and 90-day reexacerbation rates. Reexacerbation rates were defined by ICD-10 codes and documentation from a primary care visit or subsequent ED visit. The secondary outcomes were the rate of PCT tests ordered and used for treatment decisions. Other areas of interest were antibiotic prescribing trends, duration of therapy, and patient COPD GOLD classification.

Statistical analysis

It was estimated that a sample size of 146 patients (73 patients/group) would provide 80% power to detect a between-group difference of 10% in the percentage of patients who were prescribed antibiotics. Categorical variables were expressed using estimates of frequency and percentages. Percentages were compared using Fisher exact tests. For all tests, the significance level was set at 0.05.

 

Results

Seventy-three patients were included in the preintervention group and 77 in the postintervention group. The GOLD classification rates were similar between the groups (Table 1). In addition, > 90% of patients were White males and all patients were aged ≥ 50 years, which is characteristic of the US Department of Veterans Affairs (VA) population.

The percentage of antibiotic prescriptions decreased by 20% after implementation, falling from 83.6% before to 63.6% after the implementation (P =.01). The documented change in sputum color remained low compared with antibiotic prescriptions: 17.8% preimplementation and 16.9% postimplementation. The reduction in antibiotic prescriptions was associated with limited differences observed in 30-, 60-, and 90-day reexacerbation rates pre- and postintervention: 19.2% vs 23.4%, 12.3% vs 11.7%, and 4.1% vs 9.1%, respectively.

Prior to the education, introduction of the algorithm, and implementation of the PCT quick-order menu, PCT was ordered for 1.4% of AECOPD cases. Postintervention, PCT was ordered for 28.6% of mild-to-moderate AECOPD cases and used in clinical decision making per clinical documentation 81.8% of the time. PCT was used in 5 GOLD group B patients, 5 GOLD group C patients, and 7 GOLD group D patients. In all cases, PCT was < 0.25 ng/mL. In 4 cases PCT was ordered but not used: 1 GOLD group D patient refused traditional treatment with oral corticosteroids, which resulted in the clinician prescribing antibiotics, and the other 3 cases did not use PCT based on clinical decision making. The rate of PCT tests ordered for mild-to-moderate AECOPD over time is depicted in Figure 2.

The average duration of antibiotic therapy was about 6 days pre- and postintervention. This is longer than the PVAHCS recommended duration of 5 days but is consistent with the GOLD guidelines recommended duration of 5 to 7 days.¹ Azithromycin is recommended as a first-line treatment option at the PVAHCS based on the local antibiogram, and it remained the most commonly prescribed antibiotic pre- and postintervention. Outcomes of interest are detailed in Table 2.

Discussion

The implementation of PCT-guided antibiotic prescribing for patients with mild and moderate AECOPD who presented to the ED resulted in a 20% reduction in antibiotic prescriptions, falling from 83.6% before the intervention to 63.6% afterward (P = .01). The measured decrease in antibiotic prescriptions is consistent with other studies evaluating the use of acute phase reactants to guide antibiotic prescribing for AECOPD.^10,11 In addition, there was no observed difference in reexacerbation rates. This adds to the increasing body of evidence that antibiotics are overprescribed in mild and moderate AECOPD.¹² This is exemplified in our data by the low percentage of patients who had a documented change in sputum color; symptoms that are well known to be highly specific and sensitive for a bacterial infection in AECOPD.

Many health care providers (HCPs) in the ED were unfamiliar with PCT prior to implementation. A primary concern with this study was its impact on diagnostic stewardship. Preimplementation, ED clinicians ordered PCT 8 times for any cause. Postintervention, ED clinicians ordered PCT 180 times for any cause: 36% of these orders were for patients with AECOPD who were discharged from the ED or who required hospital admission. The other orders were for other respiratory conditions, including asthma exacerbations, pneumonia, bronchitis, sinusitis, pharyngitis, nonspecific respiratory infections, and respiratory failure.

The early phase of the COVID-19 pandemic coincided with the postintervention phase of this project. PVAHCS started preparing for the pandemic in March 2020, and the first confirmed diagnosis at the facility occurred mid-March. COVID-19 may have contributed to the sharp increase in PCT tests. There is currently no well-defined role for PCT in the diagnosis or management of COVID-19, but ED clinicians may have increased their use of PCT tests to help characterize the etiology of the large influx of patients presenting with respiratory symptoms.¹³

 

Strengths

Strengths of this project include its multimodal implementation and overall pragmatic design, which reflects real-world utilization of procalcitonin by ED HCPs. The HCPs were not mandated to follow the procalcitonin algorithm, and the use of clinical judgment was strongly encouraged. This project occurred concomitantly with the VA Infectious Disease Academic Detailing education program. The program focused on clinician education for the proper diagnosis and treatment of respiratory tract infections. In addition, viral illness packs were introduced as part of this initiative to reduce unnecessary antibiotic prescribing. The viral illness pack included standard items for symptom relief, such as saline nasal spray, cough drops, and hand sanitizer, as well as an explanation card of why the patient was not receiving antibiotics. Several studies have suggested that patients expect a prescription for an antibiotic when they present with respiratory tract symptoms, and HCPs often are compelled to maintain patient satisfaction, thus leading to unnecessary antibiotic prescriptions.¹⁴ The viral illness pack helped fulfill the patient’s expectation to receive treatment after seeking care. In addition, the project lead was available full time during the first month of PCT algorithm implementation to address questions and concerns, which may have improved HCPs overall confidence in using PCT.

Limitations

Limitations of this project include its population and its retrospective nature. The PVAHCS patient population is predominantly older, more White, and more male compared with the general civilian population, and results may not be generalizable to other populations. Data were limited to documentation in the electronic health record. The population was based on data extraction by the ICD-10 code, which may not be an accurate capture of the total population as HCPs may not select the most accurate ICD-10 code on documentation. Another potential limitation was the COVID-19 pandemic which may have resulted in HCPs ordering PCT more frequently as more patients presented to the ED with undifferentiated respiratory symptoms. Finally, there were minimal differences observed in reexacerbation rates; however, although the sample size was powered to detect a difference in antibiotic prescriptions, the sample size was not powered to detect a statistically significant difference in the primary safety outcome.

Conclusions

PCT-guided antibiotic prescribing significantly reduced the number of antibiotic prescriptions without an observable increase in reexacerbation rates for patients with mild and moderate AECOPD in the ED. This study provides a pragmatic evaluation of PCT-guided antibiotic prescribing for patients with AECOPD solely in the outpatient setting. Acute phase reactants like PCT can play a role in the management of AECOPD to reduce unnecessary antibiotic prescriptions.

The Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines define acute exacerbations of chronic obstructive pulmonary disease (AECOPD) as a sudden worsening of respiratory symptoms that require additional interventions. Exacerbations are classified as mild (treated with short-acting bronchodilators only), moderate (treated with antibiotics and/or oral corticosteroids), or severe (treatment requiring hospitalization). Exacerbations must include increased dyspnea, and other symptoms may involve increased sputum volume and purulence, cough, and a change in sputum color. These symptoms can be due to viral, bacterial, or environmental causes, with viral respiratory infections being the most common cause.^1-4 However, determining the etiology of an exacerbation can be difficult based on symptoms alone and can lead to an excessive and unnecessary use of antibiotics. Only the change in sputum color is considered highly sensitive and specific for bacterial causes.¹ As a result, there has been an increased interest in the use of acute biomarkers to determine whether antibiotics are necessary.

Procalcitonin (PCT) is an acute phase reactant that increases in response to inflammation, especially inflammation caused by a bacterial infection. Recent studies have suggested that PCT may be used in patients experiencing an AECOPD to reduce antibiotic use without impacting rates of treatment failure.^5-9 A majority of these studies have been in the inpatient setting or a combination of inpatient and outpatient settings.

The purpose of this study was to create and to evaluate the efficacy and practicality of a PCT-based algorithm to aid emergency department (ED) clinicians in the evaluation of patients with AECOPD who do not require hospitalization. The primary outcome of this project was the rate of antibiotic prescriptions before and after the initiation of the algorithm.

Methods

This was an observational, retrospective, pre/post assessment at the Phoenix Veterans Affairs Health Care System (PVAHCS) in Arizona. Patients who were discharged from the ED with a diagnosis of an AECOPD were identified using International Classification of Diseases, Tenth Revision (ICD-10) codes. Patient charts were reviewed from November 2018 to March 2019 for the preimplementation group and from November 2019 for March 2020 in the postimplementation group. The periods were chosen to reflect similar seasons for both the pre- and postimplementation interventions. Patients were excluded from analysis if they required hospital admission, were immunocompromised, on chronic antimicrobial therapy, had no documented medical history of COPD, or if they were presenting primarily for medication refills. Information collected included the rate of antibiotic prescriptions, procalcitonin test orders, COPD GOLD classification, and 30-, 60-, and 90-day reexacerbation rates.

A PCT-based algorithm (Figure 1) was developed and approved by the PVAHCS Antimicrobial Stewardship Program, the Pharmacy and Therapeutics committee, and ED leadership. PCT threshold values were based on values approved by the US Food and Drug Administration and previous studies—antibiotics were discouraged for PCT levels ≤ 0.25 ng/mL but could be considered for PCT levels > 0.25 ng/mL.^5,8,9 Clinicians were not required to use the algorithm, and the use of clinical judgement was encouraged. The recommended antibiotic therapies were based on previously approved PVAHCS antimicrobial stewardship guidance. To promote utilization, a PCT quick order option was added to the ED laboratory order menu.

ED clinicians were individually educated by the antimicrobial stewardship and emergency medicine pharmacists, an infectious disease physician champion, and the pharmacy resident. Clinicians were educated about PCT and its use in the setting of AECOPD to aid in the determination of bacterial infections. Each clinician received an electronic copy the algorithm and summary of the study protocol before implementation and 3 months after implementation for follow-up education. In addition, a printed copy of the algorithm was posted in multiple clinician workstations within the ED. For the first month of implementation, the project lead was available full-time in the ED to encourage algorithm use and to field questions or concerns from clinicians.

Outcome Measures

The primary outcome was the rate of antibiotic prescriptions pre- and postintervention. The safety endpoints were 30-, 60-, and 90-day reexacerbation rates. Reexacerbation rates were defined by ICD-10 codes and documentation from a primary care visit or subsequent ED visit. The secondary outcomes were the rate of PCT tests ordered and used for treatment decisions. Other areas of interest were antibiotic prescribing trends, duration of therapy, and patient COPD GOLD classification.

Statistical analysis

It was estimated that a sample size of 146 patients (73 patients/group) would provide 80% power to detect a between-group difference of 10% in the percentage of patients who were prescribed antibiotics. Categorical variables were expressed using estimates of frequency and percentages. Percentages were compared using Fisher exact tests. For all tests, the significance level was set at 0.05.

 

Results

Seventy-three patients were included in the preintervention group and 77 in the postintervention group. The GOLD classification rates were similar between the groups (Table 1). In addition, > 90% of patients were White males and all patients were aged ≥ 50 years, which is characteristic of the US Department of Veterans Affairs (VA) population.

The percentage of antibiotic prescriptions decreased by 20% after implementation, falling from 83.6% before to 63.6% after the implementation (P =.01). The documented change in sputum color remained low compared with antibiotic prescriptions: 17.8% preimplementation and 16.9% postimplementation. The reduction in antibiotic prescriptions was associated with limited differences observed in 30-, 60-, and 90-day reexacerbation rates pre- and postintervention: 19.2% vs 23.4%, 12.3% vs 11.7%, and 4.1% vs 9.1%, respectively.

Prior to the education, introduction of the algorithm, and implementation of the PCT quick-order menu, PCT was ordered for 1.4% of AECOPD cases. Postintervention, PCT was ordered for 28.6% of mild-to-moderate AECOPD cases and used in clinical decision making per clinical documentation 81.8% of the time. PCT was used in 5 GOLD group B patients, 5 GOLD group C patients, and 7 GOLD group D patients. In all cases, PCT was < 0.25 ng/mL. In 4 cases PCT was ordered but not used: 1 GOLD group D patient refused traditional treatment with oral corticosteroids, which resulted in the clinician prescribing antibiotics, and the other 3 cases did not use PCT based on clinical decision making. The rate of PCT tests ordered for mild-to-moderate AECOPD over time is depicted in Figure 2.

The average duration of antibiotic therapy was about 6 days pre- and postintervention. This is longer than the PVAHCS recommended duration of 5 days but is consistent with the GOLD guidelines recommended duration of 5 to 7 days.¹ Azithromycin is recommended as a first-line treatment option at the PVAHCS based on the local antibiogram, and it remained the most commonly prescribed antibiotic pre- and postintervention. Outcomes of interest are detailed in Table 2.

Discussion

The implementation of PCT-guided antibiotic prescribing for patients with mild and moderate AECOPD who presented to the ED resulted in a 20% reduction in antibiotic prescriptions, falling from 83.6% before the intervention to 63.6% afterward (P = .01). The measured decrease in antibiotic prescriptions is consistent with other studies evaluating the use of acute phase reactants to guide antibiotic prescribing for AECOPD.^10,11 In addition, there was no observed difference in reexacerbation rates. This adds to the increasing body of evidence that antibiotics are overprescribed in mild and moderate AECOPD.¹² This is exemplified in our data by the low percentage of patients who had a documented change in sputum color; symptoms that are well known to be highly specific and sensitive for a bacterial infection in AECOPD.

Many health care providers (HCPs) in the ED were unfamiliar with PCT prior to implementation. A primary concern with this study was its impact on diagnostic stewardship. Preimplementation, ED clinicians ordered PCT 8 times for any cause. Postintervention, ED clinicians ordered PCT 180 times for any cause: 36% of these orders were for patients with AECOPD who were discharged from the ED or who required hospital admission. The other orders were for other respiratory conditions, including asthma exacerbations, pneumonia, bronchitis, sinusitis, pharyngitis, nonspecific respiratory infections, and respiratory failure.

The early phase of the COVID-19 pandemic coincided with the postintervention phase of this project. PVAHCS started preparing for the pandemic in March 2020, and the first confirmed diagnosis at the facility occurred mid-March. COVID-19 may have contributed to the sharp increase in PCT tests. There is currently no well-defined role for PCT in the diagnosis or management of COVID-19, but ED clinicians may have increased their use of PCT tests to help characterize the etiology of the large influx of patients presenting with respiratory symptoms.¹³

 

Strengths

Strengths of this project include its multimodal implementation and overall pragmatic design, which reflects real-world utilization of procalcitonin by ED HCPs. The HCPs were not mandated to follow the procalcitonin algorithm, and the use of clinical judgment was strongly encouraged. This project occurred concomitantly with the VA Infectious Disease Academic Detailing education program. The program focused on clinician education for the proper diagnosis and treatment of respiratory tract infections. In addition, viral illness packs were introduced as part of this initiative to reduce unnecessary antibiotic prescribing. The viral illness pack included standard items for symptom relief, such as saline nasal spray, cough drops, and hand sanitizer, as well as an explanation card of why the patient was not receiving antibiotics. Several studies have suggested that patients expect a prescription for an antibiotic when they present with respiratory tract symptoms, and HCPs often are compelled to maintain patient satisfaction, thus leading to unnecessary antibiotic prescriptions.¹⁴ The viral illness pack helped fulfill the patient’s expectation to receive treatment after seeking care. In addition, the project lead was available full time during the first month of PCT algorithm implementation to address questions and concerns, which may have improved HCPs overall confidence in using PCT.

Limitations

Limitations of this project include its population and its retrospective nature. The PVAHCS patient population is predominantly older, more White, and more male compared with the general civilian population, and results may not be generalizable to other populations. Data were limited to documentation in the electronic health record. The population was based on data extraction by the ICD-10 code, which may not be an accurate capture of the total population as HCPs may not select the most accurate ICD-10 code on documentation. Another potential limitation was the COVID-19 pandemic which may have resulted in HCPs ordering PCT more frequently as more patients presented to the ED with undifferentiated respiratory symptoms. Finally, there were minimal differences observed in reexacerbation rates; however, although the sample size was powered to detect a difference in antibiotic prescriptions, the sample size was not powered to detect a statistically significant difference in the primary safety outcome.

Conclusions

PCT-guided antibiotic prescribing significantly reduced the number of antibiotic prescriptions without an observable increase in reexacerbation rates for patients with mild and moderate AECOPD in the ED. This study provides a pragmatic evaluation of PCT-guided antibiotic prescribing for patients with AECOPD solely in the outpatient setting. Acute phase reactants like PCT can play a role in the management of AECOPD to reduce unnecessary antibiotic prescriptions.

The Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines define acute exacerbations of chronic obstructive pulmonary disease (AECOPD) as a sudden worsening of respiratory symptoms that require additional interventions. Exacerbations are classified as mild (treated with short-acting bronchodilators only), moderate (treated with antibiotics and/or oral corticosteroids), or severe (treatment requiring hospitalization). Exacerbations must include increased dyspnea, and other symptoms may involve increased sputum volume and purulence, cough, and a change in sputum color. These symptoms can be due to viral, bacterial, or environmental causes, with viral respiratory infections being the most common cause.^1-4 However, determining the etiology of an exacerbation can be difficult based on symptoms alone and can lead to an excessive and unnecessary use of antibiotics. Only the change in sputum color is considered highly sensitive and specific for bacterial causes.¹ As a result, there has been an increased interest in the use of acute biomarkers to determine whether antibiotics are necessary.

Procalcitonin (PCT) is an acute phase reactant that increases in response to inflammation, especially inflammation caused by a bacterial infection. Recent studies have suggested that PCT may be used in patients experiencing an AECOPD to reduce antibiotic use without impacting rates of treatment failure.^5-9 A majority of these studies have been in the inpatient setting or a combination of inpatient and outpatient settings.

The purpose of this study was to create and to evaluate the efficacy and practicality of a PCT-based algorithm to aid emergency department (ED) clinicians in the evaluation of patients with AECOPD who do not require hospitalization. The primary outcome of this project was the rate of antibiotic prescriptions before and after the initiation of the algorithm.

Methods

This was an observational, retrospective, pre/post assessment at the Phoenix Veterans Affairs Health Care System (PVAHCS) in Arizona. Patients who were discharged from the ED with a diagnosis of an AECOPD were identified using International Classification of Diseases, Tenth Revision (ICD-10) codes. Patient charts were reviewed from November 2018 to March 2019 for the preimplementation group and from November 2019 for March 2020 in the postimplementation group. The periods were chosen to reflect similar seasons for both the pre- and postimplementation interventions. Patients were excluded from analysis if they required hospital admission, were immunocompromised, on chronic antimicrobial therapy, had no documented medical history of COPD, or if they were presenting primarily for medication refills. Information collected included the rate of antibiotic prescriptions, procalcitonin test orders, COPD GOLD classification, and 30-, 60-, and 90-day reexacerbation rates.

A PCT-based algorithm (Figure 1) was developed and approved by the PVAHCS Antimicrobial Stewardship Program, the Pharmacy and Therapeutics committee, and ED leadership. PCT threshold values were based on values approved by the US Food and Drug Administration and previous studies—antibiotics were discouraged for PCT levels ≤ 0.25 ng/mL but could be considered for PCT levels > 0.25 ng/mL.^5,8,9 Clinicians were not required to use the algorithm, and the use of clinical judgement was encouraged. The recommended antibiotic therapies were based on previously approved PVAHCS antimicrobial stewardship guidance. To promote utilization, a PCT quick order option was added to the ED laboratory order menu.

ED clinicians were individually educated by the antimicrobial stewardship and emergency medicine pharmacists, an infectious disease physician champion, and the pharmacy resident. Clinicians were educated about PCT and its use in the setting of AECOPD to aid in the determination of bacterial infections. Each clinician received an electronic copy the algorithm and summary of the study protocol before implementation and 3 months after implementation for follow-up education. In addition, a printed copy of the algorithm was posted in multiple clinician workstations within the ED. For the first month of implementation, the project lead was available full-time in the ED to encourage algorithm use and to field questions or concerns from clinicians.

Outcome Measures

The primary outcome was the rate of antibiotic prescriptions pre- and postintervention. The safety endpoints were 30-, 60-, and 90-day reexacerbation rates. Reexacerbation rates were defined by ICD-10 codes and documentation from a primary care visit or subsequent ED visit. The secondary outcomes were the rate of PCT tests ordered and used for treatment decisions. Other areas of interest were antibiotic prescribing trends, duration of therapy, and patient COPD GOLD classification.

Statistical analysis

It was estimated that a sample size of 146 patients (73 patients/group) would provide 80% power to detect a between-group difference of 10% in the percentage of patients who were prescribed antibiotics. Categorical variables were expressed using estimates of frequency and percentages. Percentages were compared using Fisher exact tests. For all tests, the significance level was set at 0.05.

 

Results

Seventy-three patients were included in the preintervention group and 77 in the postintervention group. The GOLD classification rates were similar between the groups (Table 1). In addition, > 90% of patients were White males and all patients were aged ≥ 50 years, which is characteristic of the US Department of Veterans Affairs (VA) population.

The percentage of antibiotic prescriptions decreased by 20% after implementation, falling from 83.6% before to 63.6% after the implementation (P =.01). The documented change in sputum color remained low compared with antibiotic prescriptions: 17.8% preimplementation and 16.9% postimplementation. The reduction in antibiotic prescriptions was associated with limited differences observed in 30-, 60-, and 90-day reexacerbation rates pre- and postintervention: 19.2% vs 23.4%, 12.3% vs 11.7%, and 4.1% vs 9.1%, respectively.

Prior to the education, introduction of the algorithm, and implementation of the PCT quick-order menu, PCT was ordered for 1.4% of AECOPD cases. Postintervention, PCT was ordered for 28.6% of mild-to-moderate AECOPD cases and used in clinical decision making per clinical documentation 81.8% of the time. PCT was used in 5 GOLD group B patients, 5 GOLD group C patients, and 7 GOLD group D patients. In all cases, PCT was < 0.25 ng/mL. In 4 cases PCT was ordered but not used: 1 GOLD group D patient refused traditional treatment with oral corticosteroids, which resulted in the clinician prescribing antibiotics, and the other 3 cases did not use PCT based on clinical decision making. The rate of PCT tests ordered for mild-to-moderate AECOPD over time is depicted in Figure 2.

The average duration of antibiotic therapy was about 6 days pre- and postintervention. This is longer than the PVAHCS recommended duration of 5 days but is consistent with the GOLD guidelines recommended duration of 5 to 7 days.¹ Azithromycin is recommended as a first-line treatment option at the PVAHCS based on the local antibiogram, and it remained the most commonly prescribed antibiotic pre- and postintervention. Outcomes of interest are detailed in Table 2.

Discussion

The implementation of PCT-guided antibiotic prescribing for patients with mild and moderate AECOPD who presented to the ED resulted in a 20% reduction in antibiotic prescriptions, falling from 83.6% before the intervention to 63.6% afterward (P = .01). The measured decrease in antibiotic prescriptions is consistent with other studies evaluating the use of acute phase reactants to guide antibiotic prescribing for AECOPD.^10,11 In addition, there was no observed difference in reexacerbation rates. This adds to the increasing body of evidence that antibiotics are overprescribed in mild and moderate AECOPD.¹² This is exemplified in our data by the low percentage of patients who had a documented change in sputum color; symptoms that are well known to be highly specific and sensitive for a bacterial infection in AECOPD.

Many health care providers (HCPs) in the ED were unfamiliar with PCT prior to implementation. A primary concern with this study was its impact on diagnostic stewardship. Preimplementation, ED clinicians ordered PCT 8 times for any cause. Postintervention, ED clinicians ordered PCT 180 times for any cause: 36% of these orders were for patients with AECOPD who were discharged from the ED or who required hospital admission. The other orders were for other respiratory conditions, including asthma exacerbations, pneumonia, bronchitis, sinusitis, pharyngitis, nonspecific respiratory infections, and respiratory failure.

The early phase of the COVID-19 pandemic coincided with the postintervention phase of this project. PVAHCS started preparing for the pandemic in March 2020, and the first confirmed diagnosis at the facility occurred mid-March. COVID-19 may have contributed to the sharp increase in PCT tests. There is currently no well-defined role for PCT in the diagnosis or management of COVID-19, but ED clinicians may have increased their use of PCT tests to help characterize the etiology of the large influx of patients presenting with respiratory symptoms.¹³

 

Strengths

Strengths of this project include its multimodal implementation and overall pragmatic design, which reflects real-world utilization of procalcitonin by ED HCPs. The HCPs were not mandated to follow the procalcitonin algorithm, and the use of clinical judgment was strongly encouraged. This project occurred concomitantly with the VA Infectious Disease Academic Detailing education program. The program focused on clinician education for the proper diagnosis and treatment of respiratory tract infections. In addition, viral illness packs were introduced as part of this initiative to reduce unnecessary antibiotic prescribing. The viral illness pack included standard items for symptom relief, such as saline nasal spray, cough drops, and hand sanitizer, as well as an explanation card of why the patient was not receiving antibiotics. Several studies have suggested that patients expect a prescription for an antibiotic when they present with respiratory tract symptoms, and HCPs often are compelled to maintain patient satisfaction, thus leading to unnecessary antibiotic prescriptions.¹⁴ The viral illness pack helped fulfill the patient’s expectation to receive treatment after seeking care. In addition, the project lead was available full time during the first month of PCT algorithm implementation to address questions and concerns, which may have improved HCPs overall confidence in using PCT.

Limitations

Limitations of this project include its population and its retrospective nature. The PVAHCS patient population is predominantly older, more White, and more male compared with the general civilian population, and results may not be generalizable to other populations. Data were limited to documentation in the electronic health record. The population was based on data extraction by the ICD-10 code, which may not be an accurate capture of the total population as HCPs may not select the most accurate ICD-10 code on documentation. Another potential limitation was the COVID-19 pandemic which may have resulted in HCPs ordering PCT more frequently as more patients presented to the ED with undifferentiated respiratory symptoms. Finally, there were minimal differences observed in reexacerbation rates; however, although the sample size was powered to detect a difference in antibiotic prescriptions, the sample size was not powered to detect a statistically significant difference in the primary safety outcome.

Conclusions

PCT-guided antibiotic prescribing significantly reduced the number of antibiotic prescriptions without an observable increase in reexacerbation rates for patients with mild and moderate AECOPD in the ED. This study provides a pragmatic evaluation of PCT-guided antibiotic prescribing for patients with AECOPD solely in the outpatient setting. Acute phase reactants like PCT can play a role in the management of AECOPD to reduce unnecessary antibiotic prescriptions.

Antibiotics are commonly overused for several viral respiratory conditions where antibiotic treatment is not clinically indicated. For example, a 2016 study by Fleming-Dutra and colleagues showed that at least 30% of all antibiotics prescribed in an outpatient setting were inappropriate and for acute bronchitis, antibiotic prescriptions were inappropriate in 50% of cases.¹ Acute bronchitis is predominantly a viral illness where antibiotics should be rarely used.^2-8 The Healthcare Effectiveness Data and Information Set has measured the avoidance of antibiotic treatment in adults with acute bronchitis since 2006. The National Committee for Quality Assurance reported in 2018 that about 75% of adults received antibiotics for acute bronchitis.⁹ Inappropriate antibiotic use contributes to antimicrobial resistance, resulting in the increase of morbidity and mortality of treatable infections.¹⁰ Reducing inappropriate antibiotic use in outpatient settings is a high-priority public health issue and is a Healthy People 2030 objective.¹¹

Antimicrobial Stewardship

Antimicrobial stewardship programs measure and track how antibiotics are prescribed by health care providers (HCPs) and used by patients. The Centers for Disease Control and Prevention (CDC) created a framework for outpatient antimicrobial stewardship programs by outlining 4 core elements: (1) commitment from every person involved in patient care to act as an antibiotic steward; (2) policies and interventions to promote appropriate antibiotic prescribing practices; (3) antibiotic prescription tracking and reporting; and (4) appropriate antibiotic use education.¹²

Audit and feedback (A&F) is a form of antibiotic prescription tracking and reporting that involves measuring and comparing a HCP’s performance (ie, antibiotic prescribing) with a standard, and the results of this audit are shared with the HCP. This strategy is based on the belief that a HCP is motivated to modify practice habits when given feedback showing that his or her performance is inconsistent with targeted expectations. A&F is most effective when feedback is provided by a supervisor or respected peer, presented more than once, individualized, delivered in both verbal and written formats, and includes explicit targets and an action plan.^13,14

This study focuses on an antimicrobial stewardship program implemented in an outpatient Indian Health Service ambulatory care clinic in the Pacific Northwest. The clinic was staffed by 9 HCPs serving about 12,000 American Indian and Alaskan Native patients. The clinic includes a full-service pharmacy where nearly all prescriptions issued by in-house HCPs are filled. The clinic’s antibiotic prescribing rate for adult patients with acute bronchitis was similar to the national mean in 2018 (75%).⁹ The study objective was to reduce the rate of potentially inappropriate (not guideline-concordant) antibiotic prescribing in patients with acute bronchitis without underlying chronic lung disease or evidence of bacterial infection through A&F.

Methods

The antimicrobial stewardship program was implemented by 3 pharmacists, including a pharmacy resident. HCPs received education by pharmacy staff on evidence-based prescribing for adult acute bronchitis and quarterly feedback on antibiotic prescribing rates. All prescribing and dispensing records necessary for the program were available in the clinic electronic health record. The rate of potentially inappropriate antibiotic prescribing was calculated as the proportion of eligible bronchitis cases who received antibiotics.

In October 2018, a 60-minute educational session was provided by 2 pharmacists to HCPs. The material covered an overview of acute bronchitis presentation, diagnosis, treatment (Table 1), and a comparison of national and local prescribing data (baseline audit).^2-4 The educational session concluded with prescription strategies to reduce inappropriate antibiotic prescribing, including but not limited to: delayed prescriptions, patient and caregiver education, use of nonantibiotic medications to control symptoms, and use of A&F reports.^5-8 At the conclusion of the session, HCPs committed to engage in the antimicrobial stewardship program.

Audit

To determine the total number of eligible bronchitis cases (denominator), a visit report was generated by a pharmacist for a primary diagnosis of acute bronchitis using International Statistical Classification of Diseases, Tenth Revision (ICD 10) codes (J20.3 - J20.9) for the review period. Only adults aged ≥ 18 years were included. Patients with a chronic lung disease (eg, chronic obstructive pulmonary disease, asthma) and those who had a concomitant bacterial infection (eg, urinary tract infection, cellulitis) were excluded. A visit for acute bronchitis that included additional ICD 10 codes indicating the patient had a chronic lung disease or concomitant bacterial infection were used to determine exclusion. The remaining patients who received a potentially inappropriate antibiotic prescription (numerator) were those who were prescribed or dispensed antibiotics on the date of service.

 

Feedback

Baseline data were presented to HCPs during the educational session in October 2018. Prospective audits were performed quarterly thereafter (January, April, and July) by the pharmacy resident using the criteria described above. Audit data were compiled into personalized reports and provided to HCPs by the pharmacy resident with written and verbal individual feedback. Written feedback was sent by email to each HCP containing the HCP’s rate, the clinic rate in aggregate, rates from the prior year and quarter(s) for comparison, and clinical pearls from the guidelines (Figure). Verbal feedback included a review of the written feedback and answering any questions concerning the report.

Implementation

Study periods were chosen to coincide with the pharmacy residency training year, which starts in July and ends in June. The start date of October 2018 differed from the start of the residency year (July 2018) owing to delays in obtaining permissions. A&F and analysis of prescribing rates continued through the end of the residency year, for total duration of 9 months (October 1, 2018 to June 30, 2019). For ease of reporting, quarterly reports followed the federal government’s fiscal year (FY) which runs from October 1 of the prior calendar year through September 30 of the year being described. HCPs received 4 feedback reports: baseline (October 1, 2018 - June 30, 2018) in October 2018, quarter 1 (October 1, 2018 - December 31, 2018) in January 2019, quarter 2 (January 1, 2019 - March 31, 2019) in April 2019, and quarter 3 (April 1, 2019 - June 30, 2019) in July 2019.

Statistical Analysis

Prescribing rates were compared between identical 9 -month periods. A 2-sample binomial test for proportions was used to derive an approximate CI of prescribing rates at the patient level. However, to account for clustering of patients within HCP panels and dependence of observations over study periods stemming from examining the same HCPs within each of the periods, the Wilcoxon signed rank test for paired data was used to evaluate prescribing rates at the HCP level. Statistical analysis was performed using R statistical software version 4.0.3. Differences were considered significant at P < .05 set a priori.

This study was approved by the Portland Area Indian Health Service Institutional Review Board (Study ID: 1316730).

Results

All 9 HCPs who see adult patients at the clinic agreed to participate and were all fully present in each study period. Among HCPs, there were 5 physicians and 4 physician assistants or nurse practitioners. There was a total of 213 visits that met study criteria during the baseline period (October 1, 2017 to June 30, 2018) and 177 visits in the posteducation period (October 1, 2018 to June 30, 2019). The total number of acute bronchitis encounters varied by HCP (Ranges, 5-63 [baseline] and 2-57 [posteducation]); however, the relative number of encounters each HCP contributed was similar in each study period (Table 2). The pharmacy resident spent about 2 hours each quarter to generate 9 feedback reports, 1 for each HCP.

Antibiotic Prescribing

Antibiotic prescribing rates decreased from 75% at baseline to 60% at posteducation month 9 (absolute difference, -15% [95% CI, 5 - 24%]; P ≤ .01) (Table 3). The clinic rate was lower for each quarter in FY 2019 (posteducation) compared with the same quarter of FY 2018 (baseline), with the lowest rate observed in the final quarter of the study. Comparing pre- and post- A&F, the rates for HCPs prescribing antibiotics were lower for 7 HCPs, unchanged for 1 HCP, and slightly increased for 1 HCP(P = .02).

 

Discussion

Acute bronchitis remains a common diagnosis where antibiotics are prescribed despite being a predominately viral illness. Guidelines and evidence-based practices advise against antibiotics for this diagnosis. According to the American Academy of Family Physicians, antibiotics are reserved for cases where chronic lung disease is present as these patients are at a high risk of developing pneumonia.³ The decision to prescribe antibiotics is complex and driven by several interdependent factors, such as patient expectations, health system limitations, clinician training, and specialty.¹⁵ HCPs may more aggressively treat acute bronchitis among American Indian/Alaskan Native (AI/AN) people due to a high risk of developing serious complications from respiratory illnesses.¹⁶ A clinician’s background, usual patient cohort (ie, mostly pediatric or geriatric), and time spent in urgent care or in activities outside of patient care (administration) may account for the difference in patient encounters by HCP for acute bronchitis.

Following the CDC framework, this antimicrobial stewardship program helped empower people involved in patient care (eg, pharmacists, HCPs), educate staff on proper use of antibiotics for acute bronchitis, and track and report antibiotic prescribing through the A&F process. Educational interventions coupled with ongoing A&F are reproducible by other health care facilities and are not usually time consuming. This study showcases a successful example of implementing A&F in an antimicrobial stewardship quality improvement project that could be translated toward other conditions (eg, sinusitis, urinary tract infection, community-acquired pneumonia).

In a similar study, Meeker and colleagues used a variation of an A&F intervention using a monthly email showing peer comparisons to notify clinicians who were prescribing too many unnecessary antibiotics for common respiratory illnesses that did not require antibiotics, such as the common cold.¹⁷ The peer comparison intervention arm emailed a rank order that listed prescribers by the number of prescriptions for common respiratory illnesses. This intervention demonstrated a reduction of 5.2% in inappropriate antibiotic prescribing.

Limitations

This quality improvement study had several limitations. The study did not account for the duration of symptoms as a factor to judge appropriateness. Although this was identified early in the study, it was unavoidable since there was no report that could extract the duration of symptoms in the electronic health record. Future studies should consider a manual review of each encounter to overcome this limitation. Another limitation was that only three-quarters of the year and not the entire year were reviewed. Future studies should include longer time frames to measure the durability of changes to antibiotic prescriptions. Lastly, the study did not assess diagnosis shifting (the practice of changing the proportion of antibiotic-appropriate acute respiratory tract infection diagnosis over time), effects of patient demographics (patient age and sex were not recorded), or any sustained effect on prescribing rates after the study ended.

Conclusions

Clinician education coupled with A&F are components of the CDC’s framework for an effective antimicrobial stewardship program. The intervention seem to be an effective means toward reducing inappropriate antibiotic prescribing for acute bronchitis and has the potential for application to other antimicrobial stewardship initiatives. The present study adds to the growing body of evidence on the importance and impact an antimicrobial stewardship program has on a clinic or health system.

Acknowledgment

The results of this study have been reported at the 2019 IHS Southwest Regional Pharmacy Continuing Education Seminar, April 12-14, 2019.

Antibiotics are commonly overused for several viral respiratory conditions where antibiotic treatment is not clinically indicated. For example, a 2016 study by Fleming-Dutra and colleagues showed that at least 30% of all antibiotics prescribed in an outpatient setting were inappropriate and for acute bronchitis, antibiotic prescriptions were inappropriate in 50% of cases.¹ Acute bronchitis is predominantly a viral illness where antibiotics should be rarely used.^2-8 The Healthcare Effectiveness Data and Information Set has measured the avoidance of antibiotic treatment in adults with acute bronchitis since 2006. The National Committee for Quality Assurance reported in 2018 that about 75% of adults received antibiotics for acute bronchitis.⁹ Inappropriate antibiotic use contributes to antimicrobial resistance, resulting in the increase of morbidity and mortality of treatable infections.¹⁰ Reducing inappropriate antibiotic use in outpatient settings is a high-priority public health issue and is a Healthy People 2030 objective.¹¹

Antimicrobial Stewardship

Antimicrobial stewardship programs measure and track how antibiotics are prescribed by health care providers (HCPs) and used by patients. The Centers for Disease Control and Prevention (CDC) created a framework for outpatient antimicrobial stewardship programs by outlining 4 core elements: (1) commitment from every person involved in patient care to act as an antibiotic steward; (2) policies and interventions to promote appropriate antibiotic prescribing practices; (3) antibiotic prescription tracking and reporting; and (4) appropriate antibiotic use education.¹²

Audit and feedback (A&F) is a form of antibiotic prescription tracking and reporting that involves measuring and comparing a HCP’s performance (ie, antibiotic prescribing) with a standard, and the results of this audit are shared with the HCP. This strategy is based on the belief that a HCP is motivated to modify practice habits when given feedback showing that his or her performance is inconsistent with targeted expectations. A&F is most effective when feedback is provided by a supervisor or respected peer, presented more than once, individualized, delivered in both verbal and written formats, and includes explicit targets and an action plan.^13,14

This study focuses on an antimicrobial stewardship program implemented in an outpatient Indian Health Service ambulatory care clinic in the Pacific Northwest. The clinic was staffed by 9 HCPs serving about 12,000 American Indian and Alaskan Native patients. The clinic includes a full-service pharmacy where nearly all prescriptions issued by in-house HCPs are filled. The clinic’s antibiotic prescribing rate for adult patients with acute bronchitis was similar to the national mean in 2018 (75%).⁹ The study objective was to reduce the rate of potentially inappropriate (not guideline-concordant) antibiotic prescribing in patients with acute bronchitis without underlying chronic lung disease or evidence of bacterial infection through A&F.

Methods

The antimicrobial stewardship program was implemented by 3 pharmacists, including a pharmacy resident. HCPs received education by pharmacy staff on evidence-based prescribing for adult acute bronchitis and quarterly feedback on antibiotic prescribing rates. All prescribing and dispensing records necessary for the program were available in the clinic electronic health record. The rate of potentially inappropriate antibiotic prescribing was calculated as the proportion of eligible bronchitis cases who received antibiotics.

In October 2018, a 60-minute educational session was provided by 2 pharmacists to HCPs. The material covered an overview of acute bronchitis presentation, diagnosis, treatment (Table 1), and a comparison of national and local prescribing data (baseline audit).^2-4 The educational session concluded with prescription strategies to reduce inappropriate antibiotic prescribing, including but not limited to: delayed prescriptions, patient and caregiver education, use of nonantibiotic medications to control symptoms, and use of A&F reports.^5-8 At the conclusion of the session, HCPs committed to engage in the antimicrobial stewardship program.

Audit

To determine the total number of eligible bronchitis cases (denominator), a visit report was generated by a pharmacist for a primary diagnosis of acute bronchitis using International Statistical Classification of Diseases, Tenth Revision (ICD 10) codes (J20.3 - J20.9) for the review period. Only adults aged ≥ 18 years were included. Patients with a chronic lung disease (eg, chronic obstructive pulmonary disease, asthma) and those who had a concomitant bacterial infection (eg, urinary tract infection, cellulitis) were excluded. A visit for acute bronchitis that included additional ICD 10 codes indicating the patient had a chronic lung disease or concomitant bacterial infection were used to determine exclusion. The remaining patients who received a potentially inappropriate antibiotic prescription (numerator) were those who were prescribed or dispensed antibiotics on the date of service.

 

Feedback

Baseline data were presented to HCPs during the educational session in October 2018. Prospective audits were performed quarterly thereafter (January, April, and July) by the pharmacy resident using the criteria described above. Audit data were compiled into personalized reports and provided to HCPs by the pharmacy resident with written and verbal individual feedback. Written feedback was sent by email to each HCP containing the HCP’s rate, the clinic rate in aggregate, rates from the prior year and quarter(s) for comparison, and clinical pearls from the guidelines (Figure). Verbal feedback included a review of the written feedback and answering any questions concerning the report.

Implementation

Study periods were chosen to coincide with the pharmacy residency training year, which starts in July and ends in June. The start date of October 2018 differed from the start of the residency year (July 2018) owing to delays in obtaining permissions. A&F and analysis of prescribing rates continued through the end of the residency year, for total duration of 9 months (October 1, 2018 to June 30, 2019). For ease of reporting, quarterly reports followed the federal government’s fiscal year (FY) which runs from October 1 of the prior calendar year through September 30 of the year being described. HCPs received 4 feedback reports: baseline (October 1, 2018 - June 30, 2018) in October 2018, quarter 1 (October 1, 2018 - December 31, 2018) in January 2019, quarter 2 (January 1, 2019 - March 31, 2019) in April 2019, and quarter 3 (April 1, 2019 - June 30, 2019) in July 2019.

Statistical Analysis

Prescribing rates were compared between identical 9 -month periods. A 2-sample binomial test for proportions was used to derive an approximate CI of prescribing rates at the patient level. However, to account for clustering of patients within HCP panels and dependence of observations over study periods stemming from examining the same HCPs within each of the periods, the Wilcoxon signed rank test for paired data was used to evaluate prescribing rates at the HCP level. Statistical analysis was performed using R statistical software version 4.0.3. Differences were considered significant at P < .05 set a priori.

This study was approved by the Portland Area Indian Health Service Institutional Review Board (Study ID: 1316730).

Results

All 9 HCPs who see adult patients at the clinic agreed to participate and were all fully present in each study period. Among HCPs, there were 5 physicians and 4 physician assistants or nurse practitioners. There was a total of 213 visits that met study criteria during the baseline period (October 1, 2017 to June 30, 2018) and 177 visits in the posteducation period (October 1, 2018 to June 30, 2019). The total number of acute bronchitis encounters varied by HCP (Ranges, 5-63 [baseline] and 2-57 [posteducation]); however, the relative number of encounters each HCP contributed was similar in each study period (Table 2). The pharmacy resident spent about 2 hours each quarter to generate 9 feedback reports, 1 for each HCP.

Antibiotic Prescribing

Antibiotic prescribing rates decreased from 75% at baseline to 60% at posteducation month 9 (absolute difference, -15% [95% CI, 5 - 24%]; P ≤ .01) (Table 3). The clinic rate was lower for each quarter in FY 2019 (posteducation) compared with the same quarter of FY 2018 (baseline), with the lowest rate observed in the final quarter of the study. Comparing pre- and post- A&F, the rates for HCPs prescribing antibiotics were lower for 7 HCPs, unchanged for 1 HCP, and slightly increased for 1 HCP(P = .02).

 

Discussion

Acute bronchitis remains a common diagnosis where antibiotics are prescribed despite being a predominately viral illness. Guidelines and evidence-based practices advise against antibiotics for this diagnosis. According to the American Academy of Family Physicians, antibiotics are reserved for cases where chronic lung disease is present as these patients are at a high risk of developing pneumonia.³ The decision to prescribe antibiotics is complex and driven by several interdependent factors, such as patient expectations, health system limitations, clinician training, and specialty.¹⁵ HCPs may more aggressively treat acute bronchitis among American Indian/Alaskan Native (AI/AN) people due to a high risk of developing serious complications from respiratory illnesses.¹⁶ A clinician’s background, usual patient cohort (ie, mostly pediatric or geriatric), and time spent in urgent care or in activities outside of patient care (administration) may account for the difference in patient encounters by HCP for acute bronchitis.

Following the CDC framework, this antimicrobial stewardship program helped empower people involved in patient care (eg, pharmacists, HCPs), educate staff on proper use of antibiotics for acute bronchitis, and track and report antibiotic prescribing through the A&F process. Educational interventions coupled with ongoing A&F are reproducible by other health care facilities and are not usually time consuming. This study showcases a successful example of implementing A&F in an antimicrobial stewardship quality improvement project that could be translated toward other conditions (eg, sinusitis, urinary tract infection, community-acquired pneumonia).

In a similar study, Meeker and colleagues used a variation of an A&F intervention using a monthly email showing peer comparisons to notify clinicians who were prescribing too many unnecessary antibiotics for common respiratory illnesses that did not require antibiotics, such as the common cold.¹⁷ The peer comparison intervention arm emailed a rank order that listed prescribers by the number of prescriptions for common respiratory illnesses. This intervention demonstrated a reduction of 5.2% in inappropriate antibiotic prescribing.

Limitations

This quality improvement study had several limitations. The study did not account for the duration of symptoms as a factor to judge appropriateness. Although this was identified early in the study, it was unavoidable since there was no report that could extract the duration of symptoms in the electronic health record. Future studies should consider a manual review of each encounter to overcome this limitation. Another limitation was that only three-quarters of the year and not the entire year were reviewed. Future studies should include longer time frames to measure the durability of changes to antibiotic prescriptions. Lastly, the study did not assess diagnosis shifting (the practice of changing the proportion of antibiotic-appropriate acute respiratory tract infection diagnosis over time), effects of patient demographics (patient age and sex were not recorded), or any sustained effect on prescribing rates after the study ended.

Conclusions

Clinician education coupled with A&F are components of the CDC’s framework for an effective antimicrobial stewardship program. The intervention seem to be an effective means toward reducing inappropriate antibiotic prescribing for acute bronchitis and has the potential for application to other antimicrobial stewardship initiatives. The present study adds to the growing body of evidence on the importance and impact an antimicrobial stewardship program has on a clinic or health system.

Acknowledgment

The results of this study have been reported at the 2019 IHS Southwest Regional Pharmacy Continuing Education Seminar, April 12-14, 2019.

Antibiotics are commonly overused for several viral respiratory conditions where antibiotic treatment is not clinically indicated. For example, a 2016 study by Fleming-Dutra and colleagues showed that at least 30% of all antibiotics prescribed in an outpatient setting were inappropriate and for acute bronchitis, antibiotic prescriptions were inappropriate in 50% of cases.¹ Acute bronchitis is predominantly a viral illness where antibiotics should be rarely used.^2-8 The Healthcare Effectiveness Data and Information Set has measured the avoidance of antibiotic treatment in adults with acute bronchitis since 2006. The National Committee for Quality Assurance reported in 2018 that about 75% of adults received antibiotics for acute bronchitis.⁹ Inappropriate antibiotic use contributes to antimicrobial resistance, resulting in the increase of morbidity and mortality of treatable infections.¹⁰ Reducing inappropriate antibiotic use in outpatient settings is a high-priority public health issue and is a Healthy People 2030 objective.¹¹

Antimicrobial Stewardship

Antimicrobial stewardship programs measure and track how antibiotics are prescribed by health care providers (HCPs) and used by patients. The Centers for Disease Control and Prevention (CDC) created a framework for outpatient antimicrobial stewardship programs by outlining 4 core elements: (1) commitment from every person involved in patient care to act as an antibiotic steward; (2) policies and interventions to promote appropriate antibiotic prescribing practices; (3) antibiotic prescription tracking and reporting; and (4) appropriate antibiotic use education.¹²

Audit and feedback (A&F) is a form of antibiotic prescription tracking and reporting that involves measuring and comparing a HCP’s performance (ie, antibiotic prescribing) with a standard, and the results of this audit are shared with the HCP. This strategy is based on the belief that a HCP is motivated to modify practice habits when given feedback showing that his or her performance is inconsistent with targeted expectations. A&F is most effective when feedback is provided by a supervisor or respected peer, presented more than once, individualized, delivered in both verbal and written formats, and includes explicit targets and an action plan.^13,14

This study focuses on an antimicrobial stewardship program implemented in an outpatient Indian Health Service ambulatory care clinic in the Pacific Northwest. The clinic was staffed by 9 HCPs serving about 12,000 American Indian and Alaskan Native patients. The clinic includes a full-service pharmacy where nearly all prescriptions issued by in-house HCPs are filled. The clinic’s antibiotic prescribing rate for adult patients with acute bronchitis was similar to the national mean in 2018 (75%).⁹ The study objective was to reduce the rate of potentially inappropriate (not guideline-concordant) antibiotic prescribing in patients with acute bronchitis without underlying chronic lung disease or evidence of bacterial infection through A&F.

Methods

The antimicrobial stewardship program was implemented by 3 pharmacists, including a pharmacy resident. HCPs received education by pharmacy staff on evidence-based prescribing for adult acute bronchitis and quarterly feedback on antibiotic prescribing rates. All prescribing and dispensing records necessary for the program were available in the clinic electronic health record. The rate of potentially inappropriate antibiotic prescribing was calculated as the proportion of eligible bronchitis cases who received antibiotics.

In October 2018, a 60-minute educational session was provided by 2 pharmacists to HCPs. The material covered an overview of acute bronchitis presentation, diagnosis, treatment (Table 1), and a comparison of national and local prescribing data (baseline audit).^2-4 The educational session concluded with prescription strategies to reduce inappropriate antibiotic prescribing, including but not limited to: delayed prescriptions, patient and caregiver education, use of nonantibiotic medications to control symptoms, and use of A&F reports.^5-8 At the conclusion of the session, HCPs committed to engage in the antimicrobial stewardship program.

Audit

To determine the total number of eligible bronchitis cases (denominator), a visit report was generated by a pharmacist for a primary diagnosis of acute bronchitis using International Statistical Classification of Diseases, Tenth Revision (ICD 10) codes (J20.3 - J20.9) for the review period. Only adults aged ≥ 18 years were included. Patients with a chronic lung disease (eg, chronic obstructive pulmonary disease, asthma) and those who had a concomitant bacterial infection (eg, urinary tract infection, cellulitis) were excluded. A visit for acute bronchitis that included additional ICD 10 codes indicating the patient had a chronic lung disease or concomitant bacterial infection were used to determine exclusion. The remaining patients who received a potentially inappropriate antibiotic prescription (numerator) were those who were prescribed or dispensed antibiotics on the date of service.

 

Feedback

Baseline data were presented to HCPs during the educational session in October 2018. Prospective audits were performed quarterly thereafter (January, April, and July) by the pharmacy resident using the criteria described above. Audit data were compiled into personalized reports and provided to HCPs by the pharmacy resident with written and verbal individual feedback. Written feedback was sent by email to each HCP containing the HCP’s rate, the clinic rate in aggregate, rates from the prior year and quarter(s) for comparison, and clinical pearls from the guidelines (Figure). Verbal feedback included a review of the written feedback and answering any questions concerning the report.

Implementation

Study periods were chosen to coincide with the pharmacy residency training year, which starts in July and ends in June. The start date of October 2018 differed from the start of the residency year (July 2018) owing to delays in obtaining permissions. A&F and analysis of prescribing rates continued through the end of the residency year, for total duration of 9 months (October 1, 2018 to June 30, 2019). For ease of reporting, quarterly reports followed the federal government’s fiscal year (FY) which runs from October 1 of the prior calendar year through September 30 of the year being described. HCPs received 4 feedback reports: baseline (October 1, 2018 - June 30, 2018) in October 2018, quarter 1 (October 1, 2018 - December 31, 2018) in January 2019, quarter 2 (January 1, 2019 - March 31, 2019) in April 2019, and quarter 3 (April 1, 2019 - June 30, 2019) in July 2019.

Statistical Analysis

Prescribing rates were compared between identical 9 -month periods. A 2-sample binomial test for proportions was used to derive an approximate CI of prescribing rates at the patient level. However, to account for clustering of patients within HCP panels and dependence of observations over study periods stemming from examining the same HCPs within each of the periods, the Wilcoxon signed rank test for paired data was used to evaluate prescribing rates at the HCP level. Statistical analysis was performed using R statistical software version 4.0.3. Differences were considered significant at P < .05 set a priori.

This study was approved by the Portland Area Indian Health Service Institutional Review Board (Study ID: 1316730).

Results

All 9 HCPs who see adult patients at the clinic agreed to participate and were all fully present in each study period. Among HCPs, there were 5 physicians and 4 physician assistants or nurse practitioners. There was a total of 213 visits that met study criteria during the baseline period (October 1, 2017 to June 30, 2018) and 177 visits in the posteducation period (October 1, 2018 to June 30, 2019). The total number of acute bronchitis encounters varied by HCP (Ranges, 5-63 [baseline] and 2-57 [posteducation]); however, the relative number of encounters each HCP contributed was similar in each study period (Table 2). The pharmacy resident spent about 2 hours each quarter to generate 9 feedback reports, 1 for each HCP.

Antibiotic Prescribing

Antibiotic prescribing rates decreased from 75% at baseline to 60% at posteducation month 9 (absolute difference, -15% [95% CI, 5 - 24%]; P ≤ .01) (Table 3). The clinic rate was lower for each quarter in FY 2019 (posteducation) compared with the same quarter of FY 2018 (baseline), with the lowest rate observed in the final quarter of the study. Comparing pre- and post- A&F, the rates for HCPs prescribing antibiotics were lower for 7 HCPs, unchanged for 1 HCP, and slightly increased for 1 HCP(P = .02).

 

Discussion

Acute bronchitis remains a common diagnosis where antibiotics are prescribed despite being a predominately viral illness. Guidelines and evidence-based practices advise against antibiotics for this diagnosis. According to the American Academy of Family Physicians, antibiotics are reserved for cases where chronic lung disease is present as these patients are at a high risk of developing pneumonia.³ The decision to prescribe antibiotics is complex and driven by several interdependent factors, such as patient expectations, health system limitations, clinician training, and specialty.¹⁵ HCPs may more aggressively treat acute bronchitis among American Indian/Alaskan Native (AI/AN) people due to a high risk of developing serious complications from respiratory illnesses.¹⁶ A clinician’s background, usual patient cohort (ie, mostly pediatric or geriatric), and time spent in urgent care or in activities outside of patient care (administration) may account for the difference in patient encounters by HCP for acute bronchitis.

Following the CDC framework, this antimicrobial stewardship program helped empower people involved in patient care (eg, pharmacists, HCPs), educate staff on proper use of antibiotics for acute bronchitis, and track and report antibiotic prescribing through the A&F process. Educational interventions coupled with ongoing A&F are reproducible by other health care facilities and are not usually time consuming. This study showcases a successful example of implementing A&F in an antimicrobial stewardship quality improvement project that could be translated toward other conditions (eg, sinusitis, urinary tract infection, community-acquired pneumonia).

In a similar study, Meeker and colleagues used a variation of an A&F intervention using a monthly email showing peer comparisons to notify clinicians who were prescribing too many unnecessary antibiotics for common respiratory illnesses that did not require antibiotics, such as the common cold.¹⁷ The peer comparison intervention arm emailed a rank order that listed prescribers by the number of prescriptions for common respiratory illnesses. This intervention demonstrated a reduction of 5.2% in inappropriate antibiotic prescribing.

Limitations

This quality improvement study had several limitations. The study did not account for the duration of symptoms as a factor to judge appropriateness. Although this was identified early in the study, it was unavoidable since there was no report that could extract the duration of symptoms in the electronic health record. Future studies should consider a manual review of each encounter to overcome this limitation. Another limitation was that only three-quarters of the year and not the entire year were reviewed. Future studies should include longer time frames to measure the durability of changes to antibiotic prescriptions. Lastly, the study did not assess diagnosis shifting (the practice of changing the proportion of antibiotic-appropriate acute respiratory tract infection diagnosis over time), effects of patient demographics (patient age and sex were not recorded), or any sustained effect on prescribing rates after the study ended.

Conclusions

Clinician education coupled with A&F are components of the CDC’s framework for an effective antimicrobial stewardship program. The intervention seem to be an effective means toward reducing inappropriate antibiotic prescribing for acute bronchitis and has the potential for application to other antimicrobial stewardship initiatives. The present study adds to the growing body of evidence on the importance and impact an antimicrobial stewardship program has on a clinic or health system.

Acknowledgment

The results of this study have been reported at the 2019 IHS Southwest Regional Pharmacy Continuing Education Seminar, April 12-14, 2019.

Colonoscopies and annual fecal immunochemical tests (FITs), are 2 of the preferred modalities for colorectal cancer (CRC) screening endorsed by the US Preventive Services Task Forces as well as the US Multi-Society Task Force of Colorectal Cancer, which represents the American Gastroenterological Association, American College of Gastroenterology, and the American Society of Gastrointestinal Endoscopy.^1,2 The recommendations include proper patient selection (patients aged 50 - 75 years with a life expectancy of at least 10 years), and a discussion with the patient regarding both options.

Background

It is known that patients with a positive FIT are at an increased risk for CRC. Lee and colleagues found that patients who do not undergo subsequent colonoscopy after a positive FIT have a 1.64 relative risk of death from colon cancer compared with those who undergo follow-up colonoscopy.³ Studies also have shown that longer wait times (10 months vs 1 month) between a positive FIT and colonoscopy also are associated with a higher risk of CRC.⁴ FIT utilize antibodies specific for the globin moiety of human hemoglobin and measure the development of antibody-globin complexes using immunoassay techniques. FIT has largely replaced the fecal occult blood test (FOBT), which depends on the detection of heme in feces through oxidation.

A US Department of Veterans Affairs (VA) study found that a longer time to colonoscopy was associated with a higher risk of neoplasia in veterans with a positive FOBT (odds ratio [OR], 1.10).⁵ It is thus crucial that a positive FOBT or FIT be investigated with follow-up colonoscopy. However, a retrospective study at a single safety-net hospital in San Francisco found that only 55.6% of patients with a positive FIT completed colonoscopy within 1 year.⁶ Importantly, almost half the patients examined in this study lacked documentation of the result of the FIT or counseling regarding the significance of the positive FIT by the patient’s primary care provider who ordered the test. A VA study looked at veterans aged > 70 years at 4 VA medical centers who did not receive a follow-up colonoscopy within 1 year and reported that 26% of patients studied had a documented refusal to undergo colonoscopy.⁷

It also is clear that FOBT is used inappropriately for colon cancer screening in some patients. A 2005 single-center VA study looked at inappropriate fecal occult blood tests and found that 18% of veterans for whom FOBTs were ordered had a severe comorbid illness, 13% had signs or symptoms of gastrointestinal (GI) blood loss, and 7% had a history of colorectal neoplasia or inflammatory bowel disease.⁸ An additional national VA study looked at all veterans aged ≥ 50 years who underwent FOBT or screening colonoscopy between 2009 and 2011 and found 26% to be inappropriate (13.9% of veterans not due for screening, 7.8% with limited life expectancy, and 11% receiving a FOBT when colonoscopy was indicated).⁹

An often-misunderstood additional requirement in utilizing FIT for CRC screening is that negative tests should be repeated annually.² A study from Kaiser Permanente in California found that 75.3 to 86.1% of eligible patients underwent yearly FIT.¹⁰ In this study, programmatic FIT detected 80.4% of all patients with CRC detected within 1 year of testing.

Since most of the VA-specific studies are based on inappropriate or inadequate use of FOBT, we feel it is essential that further data be gained on appropriate and inappropriate testing. The aim of this study is to determine the frequency at which improper FIT occurs because of failure to obtain serial FIT over time with a negative result, failure to follow-up a positive FIT result with a diagnostic colonoscopy, or performance of FIT in veterans undergoing a recent colonoscopy with adequate bowel preparation. This quality assurance study received an institutional review board exemption from the VA Pittsburgh Healthcare System (VAPHS) in Pennsylvania.

Methods

VAPHS has a data repository of all veterans served within the health care system, which was queried for all veterans who underwent a FIT in the system from January 1, 2015 through December 31, 2017 as well as the number and results of FITs during the interval. In addition, the data repository was also queried specifically for veterans who had at least 1 colonoscopy as well as FIT between 2015 and 2017. The ordering location for each FIT also was queried.

 

We made 3 calculations for this study. First, we measured the rate of a negative initial FIT in 2015 and/or 2016 followed by a second FIT in 2016 and/or 2017 in a random selection of veterans (3% SE, 95% CI). Demographics were compared in an equal random number of veterans who did and did not have a follow up FIT (5% SE, 95% CI of all negative FIT). Second, we measured the rate of completing colonoscopy following a positive FIT in a random selection of veterans (3% SE, 95% SI). Finally, we calculated FITs following a colonoscopy for all veterans.

Using a power analysis with a 3% SE and 95% CI for sample size calculation and accounting for the approximate 50% exclusion rate from the final eligible population of veterans with at least 1 negative FIT, a random sample of 1,742 patient charts with a negative FIT in the interval were then reviewed to determine the frequency with which they underwent multiple FITs in the interval as well as for the presence of exclusionary factors. Because of the large number of veterans involved in this category, a more detailed demographics review was performed of a subset of these patients using a 95% CI and 5% SE. Using a 95% CI and 3% SE, 445 veterans with a positive FIT in the interval were reviewed to determine the frequency at which they underwent a follow-up diagnostic colonoscopy.

Because of a relatively small sample size, all 108 veterans who underwent a colonoscopy followed by a FIT were reviewed to determine the reason for follow-up FIT. In addition, in veterans who then went on to have a subsequent repeat colonoscopy, the examination findings were recorded.

Results

From January 1, 2015 to December 31, 2017, 6,766 FIT, were ordered at VAPHS. Of these, 4,391 unique veterans had at least 1 negative FIT during the period and 709 unique veterans had a positive FIT. There were 832 veterans who had both a FIT and colonoscopy during the study period. Of these, 108 had a colonoscopy with a subsequent FIT (Figure).

Of 1,742 randomly selected veterans with at least 1 negative FIT in the study interval, 870 were eligible for multiple FITs during this period as they were in the appropriate screening age (50-75 years or 85 years based on an assessment of life expectancy by the ordering health care provider [HCP]), did not have exclusionary comorbidities to multiple FIT, were not lost to follow-up, and had at least 1 negative FIT collected from 2015 to 2016 (veterans who only had a FIT in 2017 were excluded from this aim to avoid confounding). Of these 870 veterans, 543 (62.4%) underwent at least 2 FITs during the study period. In a demographic comparison of 110 veterans with 1 FIT and 110 veterans with > 1 FIT, there were no statistically significant differences in demographics (Table 1).

In a random chart review of 410 veterans with a positive FIT, 113 (27.5%) veterans did not undergo a subsequent colonoscopy within 1 year due to patient refusal, failure to schedule, or failure to keep colonoscopy appointment. There were no differences in demographics between those that underwent a diagnostic colonoscopy and those that did not (Table 2).

Of the 108 patients with a FIT following colonoscopy in the study interval, 97 FITs were negative. Ninety-five of the 108 FITs (88%) were judged to be inappropriate, having been performed for indications, including 38 for colon cancer screening, 23 for anemia, 32 for GI symptoms (eg, diarrhea, rectal bleeding, possible GI bleeding), and 2 for unclear indications. Thirteen FITs were deemed appropriate, as they were performed on veterans who refused to have a repeat colonoscopy following an examination with inadequate bowel preparation (Table 3). There was no difference in age or race between these 2 groups, although there was a statistically significant difference in gender (Table 4).

There were 19 patients who had a colonoscopy following a prior colonoscopy and subsequent positive FIT in the interval. Eight patients had no significant findings, 10 had nonadvanced adenomas, and 1 had an advanced adenoma (this patient had inadequate preparation with recommendation to repeat colonoscopy in 1 year).

 

While not a specific aim of the study we were able to identify certain HCPs by clinic location who systematically performed inappropriate or appropriate FIT. There were 47 separate ordering locations for the 95 inappropriate FIT following recent colonoscopy. Of these, 1 location was responsible for ordering 20 (21%) inappropriate FIT. Eight locations accounted for 51% of all the inappropriately ordered FIT. Two clinics seemed to be high performers in regard to overall appropriate vs inappropriate FIT use. The appropriate FIT rate for these locations was 30 of 33 (90.9%) and 26 of 28 (92.8%), respectively.

Discussion

In this retrospective study, we found that a large percentage of veterans eligible for colon cancer screening utilizing FIT did not undergo appropriate screening. Almost 40% of veterans in a 3-year interval received only 1 FIT. This seemed to occur due to a combination of patient refusal and inadequate education by HCPs regarding how to screen appropriately for CRC using FIT. This occurred despite a reminder in the VA Computerized Patient Record System regarding CRC screening.

There did not seem to be significant differences in demographics between those who were screened appropriately vs inappropriately. While there was a statistically significant difference in gender between those who had an appropriate FIT following recent colonoscopy (2 of 13 were female) and those who had an inappropriate FIT after recent colonoscopy (1 of 95 was a female), we are uncertain of the significance of this finding given the small number of female veterans in the analysis.

We do believe that the ratio of veterans in our study with a single FIT likely underestimates the true prevalence. To avoid confounding from factors such as inadequate prior follow-up in the study interval, we excluded veterans who underwent FIT only in 2017 for this analysis. As such, a significant percentage of these veterans were actually eligible to be screened throughout the study interval.

In spite of recommendations regarding the need for diagnostic colonoscopy following a positive FIT, we found that more than one-quarter of patients did not undergo colonoscopy. Although this number is an improvement over previously published literature that found almost half of patients at a safety-net hospital did not undergo diagnostic colonoscopy following a positive FIT, this is still clearly suboptimal.⁶

VAPHS has a mandate that all patients with a positive FIT be scheduled for colonoscopy within 30 days, either at VAPHS or in the community. An alert is sent to both ordering HCP regarding the positive FIT as well as to the GI department. In addition to contact from the ordering HCP, all veterans also are contacted by either a physician or nurse practitioner GI provider to provide test results and an explanation of its clinical significance and to facilitate colonoscopy scheduling. If a patient cannot be reached by telephone, the patient is sent a certified letter from the GI department regarding the significance of a positive FIT and instructions for scheduling a colonoscopy.

 

Despite this outreach, 27.5% of veterans did not have a diagnostic colonoscopy following a positive FIT. This suggests that there may be inadequate education and counseling of veterans at the time of the FIT order about the subsequent series of events and need for diagnostic colonoscopy following a positive FIT. If a patient refuses to undergo a colonoscopy under any circumstances (including after a positive FIT), the utility of placing a FIT order is questionable.

There is also a need for more education of ordering HCPs on appropriate indications for FITs. We found that 35% of FIT ordered after a recent colonoscopy were done for the purpose of CRC screening, despite clear guidelines recommending against this. In addition, another 50% of FIT ordered after recent colonoscopy was done either for evaluation of GI symptoms like diarrhea and rectal bleeding or in the evaluation of anemia, both of which are inappropriate uses for FIT. Since FIT is an antibody test against globin, the protein component of hemoglobin that degrades during passage through the small bowel, it is not a useful test for the evaluation of upper GI or small bowel bleeding. A relatively recent database study in the Netherlands looking at the diagnosis of upper GI malignancies within 3 years of a positive FIT found a < 1% rate.¹¹

In our study, albeit limited by the small number of veterans undergoing a repeat colonoscopy following a prior colonoscopy and subsequent positive FIT, there were few significant findings. Only 1 veteran had an advanced adenoma detected, and this veteran had already been recommended a repeat colonoscopy in 1 year due to an inadequate bowel preparation on the last examination.

Lastly, we found that certain HCPs (based on ordering clinic location) systematically performed improper FIT compared with other HCPs. This presumably is due to a lack of education on appropriate FIT usage and suggests opportunity for educational and/or systems interventions.

Limitations

While our study strengths include a relatively large number of veterans and detailed review of individual patient data, it has multiple limitations. As a retrospective chart review-based study, incomplete or inaccurate data are a possibility. It is possible that patients underwent repeat FIT or underwent colonoscopy outside of the VA system and never recorded into the VA records. In addition, there is likely a sampling bias in this study as only veterans who underwent at least 1 FIT in the interval were included. These patients may be different from those who choose to undergo colonoscopy for CRC screening or from those who do not undergo screening at all.

Conclusions

A large percentage of patients underwent improper FIT at a tertiary referral academic VA medical center. Additional education and systems interventions are necessary to improve both provider and patient adherence to appropriate CRC screening. For example, one measure may include providing HCPs with a list of their patients not up-to-date with CRC screening that was shown to increase patient participation in FIT screening compared with patients who received usual care in a 2017 study.¹² In addition, a 2018 study showed that a digital health intervention that allows patients to self-order tests (eg, on an iPad) can increase CRC screening rates.¹³

Author Contributions

Adam Gluskin: Study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript. Jeffrey Dueker: Study concept and design; analysis and interpretation of data; statistical analysis; critical revision of the manuscript for important intellectual content. Asif Khalid: Study concept and design; analysis and interpretation of data; drafting of the manuscripts; critical revision of the manuscript for important intellectual content; study supervision.

Colonoscopies and annual fecal immunochemical tests (FITs), are 2 of the preferred modalities for colorectal cancer (CRC) screening endorsed by the US Preventive Services Task Forces as well as the US Multi-Society Task Force of Colorectal Cancer, which represents the American Gastroenterological Association, American College of Gastroenterology, and the American Society of Gastrointestinal Endoscopy.^1,2 The recommendations include proper patient selection (patients aged 50 - 75 years with a life expectancy of at least 10 years), and a discussion with the patient regarding both options.

Background

It is known that patients with a positive FIT are at an increased risk for CRC. Lee and colleagues found that patients who do not undergo subsequent colonoscopy after a positive FIT have a 1.64 relative risk of death from colon cancer compared with those who undergo follow-up colonoscopy.³ Studies also have shown that longer wait times (10 months vs 1 month) between a positive FIT and colonoscopy also are associated with a higher risk of CRC.⁴ FIT utilize antibodies specific for the globin moiety of human hemoglobin and measure the development of antibody-globin complexes using immunoassay techniques. FIT has largely replaced the fecal occult blood test (FOBT), which depends on the detection of heme in feces through oxidation.

A US Department of Veterans Affairs (VA) study found that a longer time to colonoscopy was associated with a higher risk of neoplasia in veterans with a positive FOBT (odds ratio [OR], 1.10).⁵ It is thus crucial that a positive FOBT or FIT be investigated with follow-up colonoscopy. However, a retrospective study at a single safety-net hospital in San Francisco found that only 55.6% of patients with a positive FIT completed colonoscopy within 1 year.⁶ Importantly, almost half the patients examined in this study lacked documentation of the result of the FIT or counseling regarding the significance of the positive FIT by the patient’s primary care provider who ordered the test. A VA study looked at veterans aged > 70 years at 4 VA medical centers who did not receive a follow-up colonoscopy within 1 year and reported that 26% of patients studied had a documented refusal to undergo colonoscopy.⁷

It also is clear that FOBT is used inappropriately for colon cancer screening in some patients. A 2005 single-center VA study looked at inappropriate fecal occult blood tests and found that 18% of veterans for whom FOBTs were ordered had a severe comorbid illness, 13% had signs or symptoms of gastrointestinal (GI) blood loss, and 7% had a history of colorectal neoplasia or inflammatory bowel disease.⁸ An additional national VA study looked at all veterans aged ≥ 50 years who underwent FOBT or screening colonoscopy between 2009 and 2011 and found 26% to be inappropriate (13.9% of veterans not due for screening, 7.8% with limited life expectancy, and 11% receiving a FOBT when colonoscopy was indicated).⁹

An often-misunderstood additional requirement in utilizing FIT for CRC screening is that negative tests should be repeated annually.² A study from Kaiser Permanente in California found that 75.3 to 86.1% of eligible patients underwent yearly FIT.¹⁰ In this study, programmatic FIT detected 80.4% of all patients with CRC detected within 1 year of testing.

Since most of the VA-specific studies are based on inappropriate or inadequate use of FOBT, we feel it is essential that further data be gained on appropriate and inappropriate testing. The aim of this study is to determine the frequency at which improper FIT occurs because of failure to obtain serial FIT over time with a negative result, failure to follow-up a positive FIT result with a diagnostic colonoscopy, or performance of FIT in veterans undergoing a recent colonoscopy with adequate bowel preparation. This quality assurance study received an institutional review board exemption from the VA Pittsburgh Healthcare System (VAPHS) in Pennsylvania.

Methods

VAPHS has a data repository of all veterans served within the health care system, which was queried for all veterans who underwent a FIT in the system from January 1, 2015 through December 31, 2017 as well as the number and results of FITs during the interval. In addition, the data repository was also queried specifically for veterans who had at least 1 colonoscopy as well as FIT between 2015 and 2017. The ordering location for each FIT also was queried.

 

We made 3 calculations for this study. First, we measured the rate of a negative initial FIT in 2015 and/or 2016 followed by a second FIT in 2016 and/or 2017 in a random selection of veterans (3% SE, 95% CI). Demographics were compared in an equal random number of veterans who did and did not have a follow up FIT (5% SE, 95% CI of all negative FIT). Second, we measured the rate of completing colonoscopy following a positive FIT in a random selection of veterans (3% SE, 95% SI). Finally, we calculated FITs following a colonoscopy for all veterans.

Using a power analysis with a 3% SE and 95% CI for sample size calculation and accounting for the approximate 50% exclusion rate from the final eligible population of veterans with at least 1 negative FIT, a random sample of 1,742 patient charts with a negative FIT in the interval were then reviewed to determine the frequency with which they underwent multiple FITs in the interval as well as for the presence of exclusionary factors. Because of the large number of veterans involved in this category, a more detailed demographics review was performed of a subset of these patients using a 95% CI and 5% SE. Using a 95% CI and 3% SE, 445 veterans with a positive FIT in the interval were reviewed to determine the frequency at which they underwent a follow-up diagnostic colonoscopy.

Because of a relatively small sample size, all 108 veterans who underwent a colonoscopy followed by a FIT were reviewed to determine the reason for follow-up FIT. In addition, in veterans who then went on to have a subsequent repeat colonoscopy, the examination findings were recorded.

Results

From January 1, 2015 to December 31, 2017, 6,766 FIT, were ordered at VAPHS. Of these, 4,391 unique veterans had at least 1 negative FIT during the period and 709 unique veterans had a positive FIT. There were 832 veterans who had both a FIT and colonoscopy during the study period. Of these, 108 had a colonoscopy with a subsequent FIT (Figure).

Of 1,742 randomly selected veterans with at least 1 negative FIT in the study interval, 870 were eligible for multiple FITs during this period as they were in the appropriate screening age (50-75 years or 85 years based on an assessment of life expectancy by the ordering health care provider [HCP]), did not have exclusionary comorbidities to multiple FIT, were not lost to follow-up, and had at least 1 negative FIT collected from 2015 to 2016 (veterans who only had a FIT in 2017 were excluded from this aim to avoid confounding). Of these 870 veterans, 543 (62.4%) underwent at least 2 FITs during the study period. In a demographic comparison of 110 veterans with 1 FIT and 110 veterans with > 1 FIT, there were no statistically significant differences in demographics (Table 1).

In a random chart review of 410 veterans with a positive FIT, 113 (27.5%) veterans did not undergo a subsequent colonoscopy within 1 year due to patient refusal, failure to schedule, or failure to keep colonoscopy appointment. There were no differences in demographics between those that underwent a diagnostic colonoscopy and those that did not (Table 2).

Of the 108 patients with a FIT following colonoscopy in the study interval, 97 FITs were negative. Ninety-five of the 108 FITs (88%) were judged to be inappropriate, having been performed for indications, including 38 for colon cancer screening, 23 for anemia, 32 for GI symptoms (eg, diarrhea, rectal bleeding, possible GI bleeding), and 2 for unclear indications. Thirteen FITs were deemed appropriate, as they were performed on veterans who refused to have a repeat colonoscopy following an examination with inadequate bowel preparation (Table 3). There was no difference in age or race between these 2 groups, although there was a statistically significant difference in gender (Table 4).

There were 19 patients who had a colonoscopy following a prior colonoscopy and subsequent positive FIT in the interval. Eight patients had no significant findings, 10 had nonadvanced adenomas, and 1 had an advanced adenoma (this patient had inadequate preparation with recommendation to repeat colonoscopy in 1 year).

 

While not a specific aim of the study we were able to identify certain HCPs by clinic location who systematically performed inappropriate or appropriate FIT. There were 47 separate ordering locations for the 95 inappropriate FIT following recent colonoscopy. Of these, 1 location was responsible for ordering 20 (21%) inappropriate FIT. Eight locations accounted for 51% of all the inappropriately ordered FIT. Two clinics seemed to be high performers in regard to overall appropriate vs inappropriate FIT use. The appropriate FIT rate for these locations was 30 of 33 (90.9%) and 26 of 28 (92.8%), respectively.

Discussion

In this retrospective study, we found that a large percentage of veterans eligible for colon cancer screening utilizing FIT did not undergo appropriate screening. Almost 40% of veterans in a 3-year interval received only 1 FIT. This seemed to occur due to a combination of patient refusal and inadequate education by HCPs regarding how to screen appropriately for CRC using FIT. This occurred despite a reminder in the VA Computerized Patient Record System regarding CRC screening.

There did not seem to be significant differences in demographics between those who were screened appropriately vs inappropriately. While there was a statistically significant difference in gender between those who had an appropriate FIT following recent colonoscopy (2 of 13 were female) and those who had an inappropriate FIT after recent colonoscopy (1 of 95 was a female), we are uncertain of the significance of this finding given the small number of female veterans in the analysis.

We do believe that the ratio of veterans in our study with a single FIT likely underestimates the true prevalence. To avoid confounding from factors such as inadequate prior follow-up in the study interval, we excluded veterans who underwent FIT only in 2017 for this analysis. As such, a significant percentage of these veterans were actually eligible to be screened throughout the study interval.

In spite of recommendations regarding the need for diagnostic colonoscopy following a positive FIT, we found that more than one-quarter of patients did not undergo colonoscopy. Although this number is an improvement over previously published literature that found almost half of patients at a safety-net hospital did not undergo diagnostic colonoscopy following a positive FIT, this is still clearly suboptimal.⁶

VAPHS has a mandate that all patients with a positive FIT be scheduled for colonoscopy within 30 days, either at VAPHS or in the community. An alert is sent to both ordering HCP regarding the positive FIT as well as to the GI department. In addition to contact from the ordering HCP, all veterans also are contacted by either a physician or nurse practitioner GI provider to provide test results and an explanation of its clinical significance and to facilitate colonoscopy scheduling. If a patient cannot be reached by telephone, the patient is sent a certified letter from the GI department regarding the significance of a positive FIT and instructions for scheduling a colonoscopy.

 

Despite this outreach, 27.5% of veterans did not have a diagnostic colonoscopy following a positive FIT. This suggests that there may be inadequate education and counseling of veterans at the time of the FIT order about the subsequent series of events and need for diagnostic colonoscopy following a positive FIT. If a patient refuses to undergo a colonoscopy under any circumstances (including after a positive FIT), the utility of placing a FIT order is questionable.

There is also a need for more education of ordering HCPs on appropriate indications for FITs. We found that 35% of FIT ordered after a recent colonoscopy were done for the purpose of CRC screening, despite clear guidelines recommending against this. In addition, another 50% of FIT ordered after recent colonoscopy was done either for evaluation of GI symptoms like diarrhea and rectal bleeding or in the evaluation of anemia, both of which are inappropriate uses for FIT. Since FIT is an antibody test against globin, the protein component of hemoglobin that degrades during passage through the small bowel, it is not a useful test for the evaluation of upper GI or small bowel bleeding. A relatively recent database study in the Netherlands looking at the diagnosis of upper GI malignancies within 3 years of a positive FIT found a < 1% rate.¹¹

In our study, albeit limited by the small number of veterans undergoing a repeat colonoscopy following a prior colonoscopy and subsequent positive FIT, there were few significant findings. Only 1 veteran had an advanced adenoma detected, and this veteran had already been recommended a repeat colonoscopy in 1 year due to an inadequate bowel preparation on the last examination.

Lastly, we found that certain HCPs (based on ordering clinic location) systematically performed improper FIT compared with other HCPs. This presumably is due to a lack of education on appropriate FIT usage and suggests opportunity for educational and/or systems interventions.

Limitations

While our study strengths include a relatively large number of veterans and detailed review of individual patient data, it has multiple limitations. As a retrospective chart review-based study, incomplete or inaccurate data are a possibility. It is possible that patients underwent repeat FIT or underwent colonoscopy outside of the VA system and never recorded into the VA records. In addition, there is likely a sampling bias in this study as only veterans who underwent at least 1 FIT in the interval were included. These patients may be different from those who choose to undergo colonoscopy for CRC screening or from those who do not undergo screening at all.

Conclusions

A large percentage of patients underwent improper FIT at a tertiary referral academic VA medical center. Additional education and systems interventions are necessary to improve both provider and patient adherence to appropriate CRC screening. For example, one measure may include providing HCPs with a list of their patients not up-to-date with CRC screening that was shown to increase patient participation in FIT screening compared with patients who received usual care in a 2017 study.¹² In addition, a 2018 study showed that a digital health intervention that allows patients to self-order tests (eg, on an iPad) can increase CRC screening rates.¹³

Author Contributions

Adam Gluskin: Study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript. Jeffrey Dueker: Study concept and design; analysis and interpretation of data; statistical analysis; critical revision of the manuscript for important intellectual content. Asif Khalid: Study concept and design; analysis and interpretation of data; drafting of the manuscripts; critical revision of the manuscript for important intellectual content; study supervision.

Colonoscopies and annual fecal immunochemical tests (FITs), are 2 of the preferred modalities for colorectal cancer (CRC) screening endorsed by the US Preventive Services Task Forces as well as the US Multi-Society Task Force of Colorectal Cancer, which represents the American Gastroenterological Association, American College of Gastroenterology, and the American Society of Gastrointestinal Endoscopy.^1,2 The recommendations include proper patient selection (patients aged 50 - 75 years with a life expectancy of at least 10 years), and a discussion with the patient regarding both options.

Background

It is known that patients with a positive FIT are at an increased risk for CRC. Lee and colleagues found that patients who do not undergo subsequent colonoscopy after a positive FIT have a 1.64 relative risk of death from colon cancer compared with those who undergo follow-up colonoscopy.³ Studies also have shown that longer wait times (10 months vs 1 month) between a positive FIT and colonoscopy also are associated with a higher risk of CRC.⁴ FIT utilize antibodies specific for the globin moiety of human hemoglobin and measure the development of antibody-globin complexes using immunoassay techniques. FIT has largely replaced the fecal occult blood test (FOBT), which depends on the detection of heme in feces through oxidation.

A US Department of Veterans Affairs (VA) study found that a longer time to colonoscopy was associated with a higher risk of neoplasia in veterans with a positive FOBT (odds ratio [OR], 1.10).⁵ It is thus crucial that a positive FOBT or FIT be investigated with follow-up colonoscopy. However, a retrospective study at a single safety-net hospital in San Francisco found that only 55.6% of patients with a positive FIT completed colonoscopy within 1 year.⁶ Importantly, almost half the patients examined in this study lacked documentation of the result of the FIT or counseling regarding the significance of the positive FIT by the patient’s primary care provider who ordered the test. A VA study looked at veterans aged > 70 years at 4 VA medical centers who did not receive a follow-up colonoscopy within 1 year and reported that 26% of patients studied had a documented refusal to undergo colonoscopy.⁷

It also is clear that FOBT is used inappropriately for colon cancer screening in some patients. A 2005 single-center VA study looked at inappropriate fecal occult blood tests and found that 18% of veterans for whom FOBTs were ordered had a severe comorbid illness, 13% had signs or symptoms of gastrointestinal (GI) blood loss, and 7% had a history of colorectal neoplasia or inflammatory bowel disease.⁸ An additional national VA study looked at all veterans aged ≥ 50 years who underwent FOBT or screening colonoscopy between 2009 and 2011 and found 26% to be inappropriate (13.9% of veterans not due for screening, 7.8% with limited life expectancy, and 11% receiving a FOBT when colonoscopy was indicated).⁹

An often-misunderstood additional requirement in utilizing FIT for CRC screening is that negative tests should be repeated annually.² A study from Kaiser Permanente in California found that 75.3 to 86.1% of eligible patients underwent yearly FIT.¹⁰ In this study, programmatic FIT detected 80.4% of all patients with CRC detected within 1 year of testing.

Since most of the VA-specific studies are based on inappropriate or inadequate use of FOBT, we feel it is essential that further data be gained on appropriate and inappropriate testing. The aim of this study is to determine the frequency at which improper FIT occurs because of failure to obtain serial FIT over time with a negative result, failure to follow-up a positive FIT result with a diagnostic colonoscopy, or performance of FIT in veterans undergoing a recent colonoscopy with adequate bowel preparation. This quality assurance study received an institutional review board exemption from the VA Pittsburgh Healthcare System (VAPHS) in Pennsylvania.

Methods

VAPHS has a data repository of all veterans served within the health care system, which was queried for all veterans who underwent a FIT in the system from January 1, 2015 through December 31, 2017 as well as the number and results of FITs during the interval. In addition, the data repository was also queried specifically for veterans who had at least 1 colonoscopy as well as FIT between 2015 and 2017. The ordering location for each FIT also was queried.

 

We made 3 calculations for this study. First, we measured the rate of a negative initial FIT in 2015 and/or 2016 followed by a second FIT in 2016 and/or 2017 in a random selection of veterans (3% SE, 95% CI). Demographics were compared in an equal random number of veterans who did and did not have a follow up FIT (5% SE, 95% CI of all negative FIT). Second, we measured the rate of completing colonoscopy following a positive FIT in a random selection of veterans (3% SE, 95% SI). Finally, we calculated FITs following a colonoscopy for all veterans.

Using a power analysis with a 3% SE and 95% CI for sample size calculation and accounting for the approximate 50% exclusion rate from the final eligible population of veterans with at least 1 negative FIT, a random sample of 1,742 patient charts with a negative FIT in the interval were then reviewed to determine the frequency with which they underwent multiple FITs in the interval as well as for the presence of exclusionary factors. Because of the large number of veterans involved in this category, a more detailed demographics review was performed of a subset of these patients using a 95% CI and 5% SE. Using a 95% CI and 3% SE, 445 veterans with a positive FIT in the interval were reviewed to determine the frequency at which they underwent a follow-up diagnostic colonoscopy.

Because of a relatively small sample size, all 108 veterans who underwent a colonoscopy followed by a FIT were reviewed to determine the reason for follow-up FIT. In addition, in veterans who then went on to have a subsequent repeat colonoscopy, the examination findings were recorded.

Results

From January 1, 2015 to December 31, 2017, 6,766 FIT, were ordered at VAPHS. Of these, 4,391 unique veterans had at least 1 negative FIT during the period and 709 unique veterans had a positive FIT. There were 832 veterans who had both a FIT and colonoscopy during the study period. Of these, 108 had a colonoscopy with a subsequent FIT (Figure).

Of 1,742 randomly selected veterans with at least 1 negative FIT in the study interval, 870 were eligible for multiple FITs during this period as they were in the appropriate screening age (50-75 years or 85 years based on an assessment of life expectancy by the ordering health care provider [HCP]), did not have exclusionary comorbidities to multiple FIT, were not lost to follow-up, and had at least 1 negative FIT collected from 2015 to 2016 (veterans who only had a FIT in 2017 were excluded from this aim to avoid confounding). Of these 870 veterans, 543 (62.4%) underwent at least 2 FITs during the study period. In a demographic comparison of 110 veterans with 1 FIT and 110 veterans with > 1 FIT, there were no statistically significant differences in demographics (Table 1).

In a random chart review of 410 veterans with a positive FIT, 113 (27.5%) veterans did not undergo a subsequent colonoscopy within 1 year due to patient refusal, failure to schedule, or failure to keep colonoscopy appointment. There were no differences in demographics between those that underwent a diagnostic colonoscopy and those that did not (Table 2).

Of the 108 patients with a FIT following colonoscopy in the study interval, 97 FITs were negative. Ninety-five of the 108 FITs (88%) were judged to be inappropriate, having been performed for indications, including 38 for colon cancer screening, 23 for anemia, 32 for GI symptoms (eg, diarrhea, rectal bleeding, possible GI bleeding), and 2 for unclear indications. Thirteen FITs were deemed appropriate, as they were performed on veterans who refused to have a repeat colonoscopy following an examination with inadequate bowel preparation (Table 3). There was no difference in age or race between these 2 groups, although there was a statistically significant difference in gender (Table 4).

There were 19 patients who had a colonoscopy following a prior colonoscopy and subsequent positive FIT in the interval. Eight patients had no significant findings, 10 had nonadvanced adenomas, and 1 had an advanced adenoma (this patient had inadequate preparation with recommendation to repeat colonoscopy in 1 year).

 

While not a specific aim of the study we were able to identify certain HCPs by clinic location who systematically performed inappropriate or appropriate FIT. There were 47 separate ordering locations for the 95 inappropriate FIT following recent colonoscopy. Of these, 1 location was responsible for ordering 20 (21%) inappropriate FIT. Eight locations accounted for 51% of all the inappropriately ordered FIT. Two clinics seemed to be high performers in regard to overall appropriate vs inappropriate FIT use. The appropriate FIT rate for these locations was 30 of 33 (90.9%) and 26 of 28 (92.8%), respectively.

Discussion

In this retrospective study, we found that a large percentage of veterans eligible for colon cancer screening utilizing FIT did not undergo appropriate screening. Almost 40% of veterans in a 3-year interval received only 1 FIT. This seemed to occur due to a combination of patient refusal and inadequate education by HCPs regarding how to screen appropriately for CRC using FIT. This occurred despite a reminder in the VA Computerized Patient Record System regarding CRC screening.

There did not seem to be significant differences in demographics between those who were screened appropriately vs inappropriately. While there was a statistically significant difference in gender between those who had an appropriate FIT following recent colonoscopy (2 of 13 were female) and those who had an inappropriate FIT after recent colonoscopy (1 of 95 was a female), we are uncertain of the significance of this finding given the small number of female veterans in the analysis.

We do believe that the ratio of veterans in our study with a single FIT likely underestimates the true prevalence. To avoid confounding from factors such as inadequate prior follow-up in the study interval, we excluded veterans who underwent FIT only in 2017 for this analysis. As such, a significant percentage of these veterans were actually eligible to be screened throughout the study interval.

In spite of recommendations regarding the need for diagnostic colonoscopy following a positive FIT, we found that more than one-quarter of patients did not undergo colonoscopy. Although this number is an improvement over previously published literature that found almost half of patients at a safety-net hospital did not undergo diagnostic colonoscopy following a positive FIT, this is still clearly suboptimal.⁶

VAPHS has a mandate that all patients with a positive FIT be scheduled for colonoscopy within 30 days, either at VAPHS or in the community. An alert is sent to both ordering HCP regarding the positive FIT as well as to the GI department. In addition to contact from the ordering HCP, all veterans also are contacted by either a physician or nurse practitioner GI provider to provide test results and an explanation of its clinical significance and to facilitate colonoscopy scheduling. If a patient cannot be reached by telephone, the patient is sent a certified letter from the GI department regarding the significance of a positive FIT and instructions for scheduling a colonoscopy.

 

Despite this outreach, 27.5% of veterans did not have a diagnostic colonoscopy following a positive FIT. This suggests that there may be inadequate education and counseling of veterans at the time of the FIT order about the subsequent series of events and need for diagnostic colonoscopy following a positive FIT. If a patient refuses to undergo a colonoscopy under any circumstances (including after a positive FIT), the utility of placing a FIT order is questionable.

There is also a need for more education of ordering HCPs on appropriate indications for FITs. We found that 35% of FIT ordered after a recent colonoscopy were done for the purpose of CRC screening, despite clear guidelines recommending against this. In addition, another 50% of FIT ordered after recent colonoscopy was done either for evaluation of GI symptoms like diarrhea and rectal bleeding or in the evaluation of anemia, both of which are inappropriate uses for FIT. Since FIT is an antibody test against globin, the protein component of hemoglobin that degrades during passage through the small bowel, it is not a useful test for the evaluation of upper GI or small bowel bleeding. A relatively recent database study in the Netherlands looking at the diagnosis of upper GI malignancies within 3 years of a positive FIT found a < 1% rate.¹¹

In our study, albeit limited by the small number of veterans undergoing a repeat colonoscopy following a prior colonoscopy and subsequent positive FIT, there were few significant findings. Only 1 veteran had an advanced adenoma detected, and this veteran had already been recommended a repeat colonoscopy in 1 year due to an inadequate bowel preparation on the last examination.

Lastly, we found that certain HCPs (based on ordering clinic location) systematically performed improper FIT compared with other HCPs. This presumably is due to a lack of education on appropriate FIT usage and suggests opportunity for educational and/or systems interventions.

Limitations

While our study strengths include a relatively large number of veterans and detailed review of individual patient data, it has multiple limitations. As a retrospective chart review-based study, incomplete or inaccurate data are a possibility. It is possible that patients underwent repeat FIT or underwent colonoscopy outside of the VA system and never recorded into the VA records. In addition, there is likely a sampling bias in this study as only veterans who underwent at least 1 FIT in the interval were included. These patients may be different from those who choose to undergo colonoscopy for CRC screening or from those who do not undergo screening at all.

Conclusions

A large percentage of patients underwent improper FIT at a tertiary referral academic VA medical center. Additional education and systems interventions are necessary to improve both provider and patient adherence to appropriate CRC screening. For example, one measure may include providing HCPs with a list of their patients not up-to-date with CRC screening that was shown to increase patient participation in FIT screening compared with patients who received usual care in a 2017 study.¹² In addition, a 2018 study showed that a digital health intervention that allows patients to self-order tests (eg, on an iPad) can increase CRC screening rates.¹³

Author Contributions

Adam Gluskin: Study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript. Jeffrey Dueker: Study concept and design; analysis and interpretation of data; statistical analysis; critical revision of the manuscript for important intellectual content. Asif Khalid: Study concept and design; analysis and interpretation of data; drafting of the manuscripts; critical revision of the manuscript for important intellectual content; study supervision.

Delirium is an acute disturbance in mental status characterized by fluctuations in cognition and attention that affects more than 2.6 million hospitalized older adults in the United States annually, a rate that is expected to increase as the population ages.^1-4 Hospital-acquired delirium is associated with poor outcomes, including prolonged hospital length of stay (LOS), loss of independence, cognitive impairment, and even death.^5-10 Individuals who develop delirium do poorly after hospital discharge and are more likely to be readmitted within 30 days.¹¹ Approximately 30% to 40% of hospital-acquired delirium cases are preventable.^10,12 However, programs designed to prevent delirium and associated complications, such as increased LOS, have demonstrated variable success.^12-14 Many studies are limited by small sample sizes, lack of generalizability to different hospitalized patient populations, poor adherence, or reliance on outside funding.^12,13,15-18

Delirium prevention programs face several challenges because delirium could be caused by a variety of risk factors and precipitants.^19,20 Some risk factors that occur frequently among hospitalized patients can be mitigated, such as sensory impairment, immobility from physical restraints or urinary catheters, and polypharmacy.^20,21 Effective delirium care pathways targeting these risk factors must be multifaceted, interdisciplinary, and interprofessional. Accurate risk assessment is critical to allocate resources to high-risk patients. Delirium affects patients in all medical and surgical disciplines, and often is underdiagnosed.^19,22 Comprehensive screening is necessary to identify cases early and track outcomes, and educational efforts must reach all providers in the hospital. These challenges require a systematic, pragmatic approach to change.

The purpose of this study was to evaluate the association between a delirium care pathway and clinical outcomes for hospitalized patients. We hypothesized that this program would be associated with reduced hospital LOS, with secondary benefits to hospitalization costs, odds of 30-day readmission, and delirium rates.

METHODS

Study Design

In this retrospective cohort study, we compared clinical outcomes the year before and after implementation of a delirium care pathway across seven hospital units. The study period spanned October 1, 2015, through February 28, 2019. The study was approved by the University of California, San Francisco Institutional Review Board (#13-12500).

Multicomponent Delirium Care Pathway

The delirium care pathway was developed collaboratively among geriatrics, hospital medicine, neurology, anesthesiology, surgery, and psychiatry services, with an interprofessional team of physicians, nurses, pharmacists, and physical and occupational therapists. This pathway was implemented in units consecutively, approximately every 4 months in the following order: neurosciences, medicine, cardiology, general surgery, specialty surgery, hematology-oncology, and transplant. The same implementation education protocols were performed in each unit. The pathway consisted of several components targeting delirium prevention and management (Appendix Figure 1 and Appendix Figure 2). Systematic screening for delirium was introduced as part of the multicomponent intervention. Nursing staff assessed each patient’s risk of developing delirium at admission using the AWOL score, a validated delirium prediction tool.²³ AWOL consists of: patient Age, spelling “World” backwards correctly, Orientation, and assessment of iLlness severity by the nurse. For patients who spoke a language other than English, spelling of “world” backwards was translated to his or her primary language, or if this was not possible, the task was modified to serial 7s (subtracting 7 from 100 in a serial fashion). This modification has been validated for use in other languages.²⁴ Patients at high risk for delirium based on an AWOL score ≥2 received a multidisciplinary intervention with four components: (1) notifying the primary team by pager and electronic medical record (EMR), (2) a nurse-led, evidence-based, nonpharmacologic multicomponent intervention,²⁵ (3) placement of a delirium order set by the physician, and (4) review of medications by the unit pharmacist who adjusted administration timing to occur during waking hours and placed a note in the EMR notifying the primary team of potentially deliriogenic medications. The delirium order set reinforced the nonpharmacologic multicomponent intervention through a nursing order, placed an automatic consult to occupational therapy, and included options to order physical therapy, order speech/language therapy, obtain vital signs three times daily with minimal night interruptions, remove an indwelling bladder catheter, and prescribe melatonin as a sleep aid.

The bedside nurse screened all patients for active delirium every 12-hour shift using the Nursing Delirium Screening Scale (NuDESC) and entered the results into the EMR.^23,26 Capturing NuDESC results in the EMR allowed communication across medical providers as well as monitoring of screening adherence. Each nurse received two in-person trainings in staff meetings and one-to-one instruction during the first week of implementation. All nurses were required to complete a 15-minute training module and had the option of completing an additional 1-hour continuing medical education module. If a patient was transferred to the intensive care unit (ICU), delirium was identified through use of the ICU-specific Confusion Assessment Method (CAM-ICU) assessments, which the bedside nurse performed each shift throughout the intervention period.²⁷ Nurses were instructed to call the primary team physician after every positive screen. Before each unit’s implementation start date, physicians with patients on that unit received education through a combination of grand rounds, resident lectures and seminars, and a pocket card on delirium evaluation and management.

Participants and Eligibility Criteria

We included all patients aged ≥50 years hospitalized for >1 day on each hospital unit (Figure). We included adults aged ≥50 years to maximize the number of participants for this study while also capturing a population at risk for delirium. Because the delirium care pathway was unit-based and the pathway was rolled out sequentially across units, only patients who were admitted to and discharged from the same unit were included to better isolate the effect of the pathway. Patients who were transferred to the ICU were only included if they were discharged from the original unit of admission. Only the first hospitalization was included for patients with multiple hospitalizations during the study period.

Patient Characteristics

Patient demographics and clinical data were collected after discharge through Clarity and Vizient electronic databases (Table 1 and Table 2). All Elixhauser comorbidities were included except for the following International Classification of Disease, Tenth Revision, Clinical Modification (ICD-10) codes that overlapped with a delirium diagnosis: G31.2, G93.89, G93.9, G94, R41.0, and R41.82 (Appendix Table 1). Severity of illness was obtained from Vizient, which calculates illness severity based on clinical and claims data (Appendix Table 1).

Delirium Metrics

Delirium screening was introduced as part of the multicomponent intervention, and therefore delirium rates before the intervention could not be determined. Trends in delirium prevalence and incidence after the intervention are reported. Prevalent delirium was defined as a single score of ≥2 on the nurse-administered NuDESC or a positive CAM-ICU at any point during the hospital stay. Incident delirium was identified if the first NuDESC score was negative and any subsequent NuDESC or CAM-ICU score was positive.

Outcomes

The primary study outcome was hospital LOS across all participants. Secondary outcomes included total direct cost and odds of 30-day hospital readmission. Readmissions tracked as part of hospital quality reporting were obtained from Vizient and were not captured if they occurred at another hospital. We also examined rates of safety attendant and restraint use during the study period, defined as the number of safety attendant days or restraint days per 1,000 patient days.

Because previous studies have demonstrated the effectiveness of multicomponent delirium interventions among elderly general medical patients,¹² we also investigated these same outcomes in the medicine unit alone.

Statistical Analysis

The date of intervention implementation was determined for each hospital unit, which was defined as time(0) [t(0)]. The 12-month postintervention period was divided into four 3-month epochs to assess for trends. Data were aggregated across the seven units using t(0) as the start date, agnostic to the calendar month. Demographic and clinical characteristics were collected for the 12-months before t(0) and the four 3-month epochs after t(0). Univariate analysis of outcome variables comparing trends across the same epochs were conducted in the same manner, except for the rate of delirium, which was measured after t(0) and therefore could not be compared with the preintervention period.

Multivariable models were adjusted for age, sex, race/ethnicity, admission category, Elixhauser comorbidities, severity of illness quartile, and number days spent in the ICU. Admission category referred to whether the admission was emergent, urgent, or elective/unknown. Because it took 3 months after t(0) for each unit to reach a delirium screening compliance rate of 90%, the intervention was only considered fully implemented after this period. A ramp-up variable was set to 0 for admissions occurring prior to the intervention to t(0), 1/3 for admissions occurring 1 month post intervention, 2/3 for 2 months post intervention, and 1 for admissions occurring 3 to 12 months post intervention. In this way, the coefficient for the ramp-up variable estimated the postintervention versus preintervention effect. Numerical outcomes (LOS, cost) were log transformed to reduce skewness and analyzed using linear models. Coefficients were back-transformed to provide interpretations as proportional change in the median outcomes.

For LOS and readmission, we assessed secular trends by including admission date and admission date squared, in case the trend was nonlinear, as possible predictors; admission date was the specific date—not time from t(0)—to account for secular trends and allow contemporaneous controls in the analysis. To be conservative, we retained secular terms (first considering the quadratic and then the linear) if P <.10. The categorical outcome (30-day readmission) was analyzed using a logistic model. Count variables (delirium, safety attendants, restraints) were analyzed using Poisson regression models with a log link, and coefficients were back-transformed to provide rate ratio interpretations. Because delirium was not measured before t(0), and because the intervention was considered to take 3 months to become fully effective, baseline delirium rates were defined as those in the first 3 months adjusted by the ramp-up variable. For each outcome we included hospital unit, a ramp-up variable (measuring the pre- vs postintervention effect), and their interaction. If there was no statistically significant interaction, we presented the outcome for all units combined. If the interaction was statistically significant, we looked for consistency across units and reported results for all units combined when consistent, along with site-specific results. If the results were not consistent across the units, we provided site-specific results only. All statistical analyses were performed using SAS software, version 9.4 (SAS Institute Inc).

RESULTS

Participant Demographics and Clinical Characteristics

A total of 22,708 individuals were included in this study, with 11,018 in the preintervention period (Table 1 and Table 2). Most patients were cared for on the general surgery unit (n = 5,899), followed by the medicine unit (n = 4,923). The smallest number of patients were cared for on the hematology-oncology unit (n = 1,709). Across the five epochs, patients were of similar age and sex, and spent a similar number of days in the ICU. The population was diverse with regard to race and ethnicity; there were minor differences in admission category. There were also minor differences in severity of illness and some comorbidities between timepoints (Appendix Table 1).

Delirium Metrics

Delirium prevalence was 13.0% during the first epoch post intervention, followed by 12.0%, 11.7%, and 13.0% in the subsequent epochs (P = .91). Incident delirium occurred in 6.1% of patients during the first epoch post intervention, followed by 5.3%, 5.3%, and 5.8% in the subsequent epochs (P = .63).

Primary Outcome

Epoch-level data for LOS before and after the intervention is shown in Appendix Table 2. The mean unadjusted LOS for all units combined did not decrease after the intervention, but in the adjusted model, the mean LOS decreased by 2% after the intervention (P = .0087; Table 3).

Secondary Outcomes

The odds of 30-day readmission decreased by 14% (P = .0002) in the adjusted models for all units combined (Table 3). There was no statistically significant reduction in adjusted total direct hospitalization cost or rate of restraint use. The safety attendant results showed strong effect modification across sites; the site-specific estimates are provided in Appendix Table 3. However, the estimated values all showed reductions, and a number were large and statistically significant.

Medicine Unit Outcomes

On the medicine unit alone, we observed a statistically significant reduction in LOS of 9% after implementation of the delirium care pathway (P = .028) in the adjusted model (Table 3). There was an associated 7% proportional decrease in total direct cost (P = .0002). Reductions in 30-day readmission and safety attendant use did not remain statistically significant in the adjusted models.

DISCUSSION

Implementation of a hospital-wide multicomponent delirium care pathway was associated with reduced hospital LOS and 30-day hospital readmission in a study of 22,708 hospitalized adults at a tertiary care, university hospital in Northern California, encompassing both medical and surgical acute care patients. When evaluating general medicine patients alone, pathway implementation was associated with reductions in LOS and total direct cost. The cost savings of 7% among medical patients translates to median savings of $1,237 per hospitalization. This study—one of the largest to date examining implementation of a hospital-wide delirium care pathway—supports use of a multicomponent delirium care pathway for older adults hospitalized for a range of conditions.

Multicomponent pathways for delirium prevention and management are increasingly being used in hospital settings. The United Kingdom National Institute for Health and Care Excellence guidelines recommend delirium assessment and intervention by a multidisciplinary team within 24 hours of hospital admission for those at risk.²⁵ These guidelines are based on evidence accumulated in clinical studies over the past 30 years suggesting that multicomponent interventions reduce incident delirium by 30% to 40% among medical and surgical patients.^12,13,25,28

Although multicomponent delirium care pathways are associated with improved patient outcomes, the specific clinical benefits might vary across patient populations. Here, we found larger reductions in LOS and total direct cost among medicine patients. Medical patients might respond more robustly to nonpharmacologic multicomponent delirium interventions because of differing delirium etiologies (eg, constipation and sleep deprivation in a medical patient vs seizures or encephalitis in a neurosciences patient). Another explanation for the difference observed in total direct cost might be the inclusion of surgical units in the total study population. For example, not all hospital days are equivalent in cost for patients on a surgical unit.²⁹ For patients requiring surgical care, most of the hospitalization cost might be incurred during the initial days of hospitalization, when there are perioperative costs; therefore, reduced LOS might have a lower economic impact.²⁹ Multicomponent, nonpharmacologic delirium interventions encourage discontinuing restraints. As a result, one might expect a need for more frequent safety attendant use and an associated cost increase. However, we found that the estimated unit-specific values for safety attendant use showed reductions, which were large and highly statistically significant. For all units combined and the medicine unit alone, we found that the rate of restraint use decreased, although the change was not statistically significant. It is possible that some of the interventions taught to nurses and physicians as part of care pathway implementation, such as the use of family support for at-risk and delirious patients, led to a reduction in both safety attendants and restraints.

Our study had several strengths. This is one of the largest hospital-based delirium interventions studied, both in terms of its scope across seven diverse medical and surgical hospital units and the number of hospitalized patients studied. This intervention did not require additional staff or creating a specialized ward. Adherence to the pathway, as measured by risk assessment and delirium screening, was high (>90%) 3 months after implementation. This allowed for robust outcome ascertainment. The patient population’s characteristics and rates of delirium were stable over time. Because different hospital units incorporated the multicomponent delirium care pathway at different times, limiting enrollment to patients admitted and discharged from the same unit isolated the analysis to patients exposed to the pathway on each unit. This design also limited potential influence of other hospital quality improvement projects that might have occurred at the same time.

The primary limitation of this study is that screening for delirium was introduced as part of the multicomponent intervention. This decision was made to maximize buy-in from bedside nurses performing delirium screening because this addition to their workflow was explicitly linked to delirium prevention and management measures. Delirium could not be ascertained preintervention from the EMR because it is a clinical diagnosis and is coded inadequately.³⁰ We could only measure the change in delirium metrics after implementation of the delirium care pathway. Because baseline delirium rates before the intervention were not measured systematically, conclusions about the intervention’s association with delirium metrics are limited. All other outcomes were measured before and after the intervention.

Although the comprehensive delirium screening program and high rate of adherence are a methodologic strength of this study, a second limitation is the use of the NuDESC. Our previous research demonstrated that the NuDESC has low sensitivity but high specificity and positive predictive value,²⁶ which might underestimate delirium rates in this study. However, any underestimation should be stable over time and temporal trends should remain meaningful. This could allow more widespread study of delirium among hospitalized individuals. Because this care pathway was hospital-wide, it was important to ensure both consistency of screening and longevity of the initiative, and it was necessary to select a delirium assessment tool that was efficient and validated for nursing implementation. For these reasons, the NuDESC was an appropriate choice.

It is possible that our results could be influenced by unmeasured confounders. For example, although we incorporated Elixhauser medical comorbidities and illness severity into our model, we were unable to adjust for baseline functional status or frailty. Baseline functional status and frailty were not reliably recorded in the EMR, although these are potential confounders when investigating clinical outcomes including hospital readmission.

CONCLUSION

Implementation of a systematic, hospital-wide multicomponent delirium care pathway is associated with reductions in hospital LOS and 30-day readmission. In general medicine units, the reduction in LOS and associated cost savings were robust. These results demonstrate the feasibility and effectiveness of implementing an interprofessional, multidisciplinary multicomponent delirium care pathway through medical center funding to benefit patients and the hospital system.

Acknowledgments

The authors thank the many hospital staff members, especially the nurses, pharmacists, therapists, and patient care assistants, who helped implement the multicomponent delirium care pathway. All persons who have contributed significantly to this work are listed as authors of this work.

Delirium is an acute disturbance in mental status characterized by fluctuations in cognition and attention that affects more than 2.6 million hospitalized older adults in the United States annually, a rate that is expected to increase as the population ages.^1-4 Hospital-acquired delirium is associated with poor outcomes, including prolonged hospital length of stay (LOS), loss of independence, cognitive impairment, and even death.^5-10 Individuals who develop delirium do poorly after hospital discharge and are more likely to be readmitted within 30 days.¹¹ Approximately 30% to 40% of hospital-acquired delirium cases are preventable.^10,12 However, programs designed to prevent delirium and associated complications, such as increased LOS, have demonstrated variable success.^12-14 Many studies are limited by small sample sizes, lack of generalizability to different hospitalized patient populations, poor adherence, or reliance on outside funding.^12,13,15-18

Delirium prevention programs face several challenges because delirium could be caused by a variety of risk factors and precipitants.^19,20 Some risk factors that occur frequently among hospitalized patients can be mitigated, such as sensory impairment, immobility from physical restraints or urinary catheters, and polypharmacy.^20,21 Effective delirium care pathways targeting these risk factors must be multifaceted, interdisciplinary, and interprofessional. Accurate risk assessment is critical to allocate resources to high-risk patients. Delirium affects patients in all medical and surgical disciplines, and often is underdiagnosed.^19,22 Comprehensive screening is necessary to identify cases early and track outcomes, and educational efforts must reach all providers in the hospital. These challenges require a systematic, pragmatic approach to change.

The purpose of this study was to evaluate the association between a delirium care pathway and clinical outcomes for hospitalized patients. We hypothesized that this program would be associated with reduced hospital LOS, with secondary benefits to hospitalization costs, odds of 30-day readmission, and delirium rates.

METHODS

Study Design

In this retrospective cohort study, we compared clinical outcomes the year before and after implementation of a delirium care pathway across seven hospital units. The study period spanned October 1, 2015, through February 28, 2019. The study was approved by the University of California, San Francisco Institutional Review Board (#13-12500).

Multicomponent Delirium Care Pathway

The delirium care pathway was developed collaboratively among geriatrics, hospital medicine, neurology, anesthesiology, surgery, and psychiatry services, with an interprofessional team of physicians, nurses, pharmacists, and physical and occupational therapists. This pathway was implemented in units consecutively, approximately every 4 months in the following order: neurosciences, medicine, cardiology, general surgery, specialty surgery, hematology-oncology, and transplant. The same implementation education protocols were performed in each unit. The pathway consisted of several components targeting delirium prevention and management (Appendix Figure 1 and Appendix Figure 2). Systematic screening for delirium was introduced as part of the multicomponent intervention. Nursing staff assessed each patient’s risk of developing delirium at admission using the AWOL score, a validated delirium prediction tool.²³ AWOL consists of: patient Age, spelling “World” backwards correctly, Orientation, and assessment of iLlness severity by the nurse. For patients who spoke a language other than English, spelling of “world” backwards was translated to his or her primary language, or if this was not possible, the task was modified to serial 7s (subtracting 7 from 100 in a serial fashion). This modification has been validated for use in other languages.²⁴ Patients at high risk for delirium based on an AWOL score ≥2 received a multidisciplinary intervention with four components: (1) notifying the primary team by pager and electronic medical record (EMR), (2) a nurse-led, evidence-based, nonpharmacologic multicomponent intervention,²⁵ (3) placement of a delirium order set by the physician, and (4) review of medications by the unit pharmacist who adjusted administration timing to occur during waking hours and placed a note in the EMR notifying the primary team of potentially deliriogenic medications. The delirium order set reinforced the nonpharmacologic multicomponent intervention through a nursing order, placed an automatic consult to occupational therapy, and included options to order physical therapy, order speech/language therapy, obtain vital signs three times daily with minimal night interruptions, remove an indwelling bladder catheter, and prescribe melatonin as a sleep aid.

The bedside nurse screened all patients for active delirium every 12-hour shift using the Nursing Delirium Screening Scale (NuDESC) and entered the results into the EMR.^23,26 Capturing NuDESC results in the EMR allowed communication across medical providers as well as monitoring of screening adherence. Each nurse received two in-person trainings in staff meetings and one-to-one instruction during the first week of implementation. All nurses were required to complete a 15-minute training module and had the option of completing an additional 1-hour continuing medical education module. If a patient was transferred to the intensive care unit (ICU), delirium was identified through use of the ICU-specific Confusion Assessment Method (CAM-ICU) assessments, which the bedside nurse performed each shift throughout the intervention period.²⁷ Nurses were instructed to call the primary team physician after every positive screen. Before each unit’s implementation start date, physicians with patients on that unit received education through a combination of grand rounds, resident lectures and seminars, and a pocket card on delirium evaluation and management.

Participants and Eligibility Criteria

We included all patients aged ≥50 years hospitalized for >1 day on each hospital unit (Figure). We included adults aged ≥50 years to maximize the number of participants for this study while also capturing a population at risk for delirium. Because the delirium care pathway was unit-based and the pathway was rolled out sequentially across units, only patients who were admitted to and discharged from the same unit were included to better isolate the effect of the pathway. Patients who were transferred to the ICU were only included if they were discharged from the original unit of admission. Only the first hospitalization was included for patients with multiple hospitalizations during the study period.

Patient Characteristics

Patient demographics and clinical data were collected after discharge through Clarity and Vizient electronic databases (Table 1 and Table 2). All Elixhauser comorbidities were included except for the following International Classification of Disease, Tenth Revision, Clinical Modification (ICD-10) codes that overlapped with a delirium diagnosis: G31.2, G93.89, G93.9, G94, R41.0, and R41.82 (Appendix Table 1). Severity of illness was obtained from Vizient, which calculates illness severity based on clinical and claims data (Appendix Table 1).

Delirium Metrics

Delirium screening was introduced as part of the multicomponent intervention, and therefore delirium rates before the intervention could not be determined. Trends in delirium prevalence and incidence after the intervention are reported. Prevalent delirium was defined as a single score of ≥2 on the nurse-administered NuDESC or a positive CAM-ICU at any point during the hospital stay. Incident delirium was identified if the first NuDESC score was negative and any subsequent NuDESC or CAM-ICU score was positive.

Outcomes

The primary study outcome was hospital LOS across all participants. Secondary outcomes included total direct cost and odds of 30-day hospital readmission. Readmissions tracked as part of hospital quality reporting were obtained from Vizient and were not captured if they occurred at another hospital. We also examined rates of safety attendant and restraint use during the study period, defined as the number of safety attendant days or restraint days per 1,000 patient days.

Because previous studies have demonstrated the effectiveness of multicomponent delirium interventions among elderly general medical patients,¹² we also investigated these same outcomes in the medicine unit alone.

Statistical Analysis

The date of intervention implementation was determined for each hospital unit, which was defined as time(0) [t(0)]. The 12-month postintervention period was divided into four 3-month epochs to assess for trends. Data were aggregated across the seven units using t(0) as the start date, agnostic to the calendar month. Demographic and clinical characteristics were collected for the 12-months before t(0) and the four 3-month epochs after t(0). Univariate analysis of outcome variables comparing trends across the same epochs were conducted in the same manner, except for the rate of delirium, which was measured after t(0) and therefore could not be compared with the preintervention period.

Multivariable models were adjusted for age, sex, race/ethnicity, admission category, Elixhauser comorbidities, severity of illness quartile, and number days spent in the ICU. Admission category referred to whether the admission was emergent, urgent, or elective/unknown. Because it took 3 months after t(0) for each unit to reach a delirium screening compliance rate of 90%, the intervention was only considered fully implemented after this period. A ramp-up variable was set to 0 for admissions occurring prior to the intervention to t(0), 1/3 for admissions occurring 1 month post intervention, 2/3 for 2 months post intervention, and 1 for admissions occurring 3 to 12 months post intervention. In this way, the coefficient for the ramp-up variable estimated the postintervention versus preintervention effect. Numerical outcomes (LOS, cost) were log transformed to reduce skewness and analyzed using linear models. Coefficients were back-transformed to provide interpretations as proportional change in the median outcomes.

For LOS and readmission, we assessed secular trends by including admission date and admission date squared, in case the trend was nonlinear, as possible predictors; admission date was the specific date—not time from t(0)—to account for secular trends and allow contemporaneous controls in the analysis. To be conservative, we retained secular terms (first considering the quadratic and then the linear) if P <.10. The categorical outcome (30-day readmission) was analyzed using a logistic model. Count variables (delirium, safety attendants, restraints) were analyzed using Poisson regression models with a log link, and coefficients were back-transformed to provide rate ratio interpretations. Because delirium was not measured before t(0), and because the intervention was considered to take 3 months to become fully effective, baseline delirium rates were defined as those in the first 3 months adjusted by the ramp-up variable. For each outcome we included hospital unit, a ramp-up variable (measuring the pre- vs postintervention effect), and their interaction. If there was no statistically significant interaction, we presented the outcome for all units combined. If the interaction was statistically significant, we looked for consistency across units and reported results for all units combined when consistent, along with site-specific results. If the results were not consistent across the units, we provided site-specific results only. All statistical analyses were performed using SAS software, version 9.4 (SAS Institute Inc).

RESULTS

Participant Demographics and Clinical Characteristics

A total of 22,708 individuals were included in this study, with 11,018 in the preintervention period (Table 1 and Table 2). Most patients were cared for on the general surgery unit (n = 5,899), followed by the medicine unit (n = 4,923). The smallest number of patients were cared for on the hematology-oncology unit (n = 1,709). Across the five epochs, patients were of similar age and sex, and spent a similar number of days in the ICU. The population was diverse with regard to race and ethnicity; there were minor differences in admission category. There were also minor differences in severity of illness and some comorbidities between timepoints (Appendix Table 1).

Delirium Metrics

Delirium prevalence was 13.0% during the first epoch post intervention, followed by 12.0%, 11.7%, and 13.0% in the subsequent epochs (P = .91). Incident delirium occurred in 6.1% of patients during the first epoch post intervention, followed by 5.3%, 5.3%, and 5.8% in the subsequent epochs (P = .63).

Primary Outcome

Epoch-level data for LOS before and after the intervention is shown in Appendix Table 2. The mean unadjusted LOS for all units combined did not decrease after the intervention, but in the adjusted model, the mean LOS decreased by 2% after the intervention (P = .0087; Table 3).

Secondary Outcomes

The odds of 30-day readmission decreased by 14% (P = .0002) in the adjusted models for all units combined (Table 3). There was no statistically significant reduction in adjusted total direct hospitalization cost or rate of restraint use. The safety attendant results showed strong effect modification across sites; the site-specific estimates are provided in Appendix Table 3. However, the estimated values all showed reductions, and a number were large and statistically significant.

Medicine Unit Outcomes

On the medicine unit alone, we observed a statistically significant reduction in LOS of 9% after implementation of the delirium care pathway (P = .028) in the adjusted model (Table 3). There was an associated 7% proportional decrease in total direct cost (P = .0002). Reductions in 30-day readmission and safety attendant use did not remain statistically significant in the adjusted models.

DISCUSSION

Implementation of a hospital-wide multicomponent delirium care pathway was associated with reduced hospital LOS and 30-day hospital readmission in a study of 22,708 hospitalized adults at a tertiary care, university hospital in Northern California, encompassing both medical and surgical acute care patients. When evaluating general medicine patients alone, pathway implementation was associated with reductions in LOS and total direct cost. The cost savings of 7% among medical patients translates to median savings of $1,237 per hospitalization. This study—one of the largest to date examining implementation of a hospital-wide delirium care pathway—supports use of a multicomponent delirium care pathway for older adults hospitalized for a range of conditions.

Multicomponent pathways for delirium prevention and management are increasingly being used in hospital settings. The United Kingdom National Institute for Health and Care Excellence guidelines recommend delirium assessment and intervention by a multidisciplinary team within 24 hours of hospital admission for those at risk.²⁵ These guidelines are based on evidence accumulated in clinical studies over the past 30 years suggesting that multicomponent interventions reduce incident delirium by 30% to 40% among medical and surgical patients.^12,13,25,28

Although multicomponent delirium care pathways are associated with improved patient outcomes, the specific clinical benefits might vary across patient populations. Here, we found larger reductions in LOS and total direct cost among medicine patients. Medical patients might respond more robustly to nonpharmacologic multicomponent delirium interventions because of differing delirium etiologies (eg, constipation and sleep deprivation in a medical patient vs seizures or encephalitis in a neurosciences patient). Another explanation for the difference observed in total direct cost might be the inclusion of surgical units in the total study population. For example, not all hospital days are equivalent in cost for patients on a surgical unit.²⁹ For patients requiring surgical care, most of the hospitalization cost might be incurred during the initial days of hospitalization, when there are perioperative costs; therefore, reduced LOS might have a lower economic impact.²⁹ Multicomponent, nonpharmacologic delirium interventions encourage discontinuing restraints. As a result, one might expect a need for more frequent safety attendant use and an associated cost increase. However, we found that the estimated unit-specific values for safety attendant use showed reductions, which were large and highly statistically significant. For all units combined and the medicine unit alone, we found that the rate of restraint use decreased, although the change was not statistically significant. It is possible that some of the interventions taught to nurses and physicians as part of care pathway implementation, such as the use of family support for at-risk and delirious patients, led to a reduction in both safety attendants and restraints.

Our study had several strengths. This is one of the largest hospital-based delirium interventions studied, both in terms of its scope across seven diverse medical and surgical hospital units and the number of hospitalized patients studied. This intervention did not require additional staff or creating a specialized ward. Adherence to the pathway, as measured by risk assessment and delirium screening, was high (>90%) 3 months after implementation. This allowed for robust outcome ascertainment. The patient population’s characteristics and rates of delirium were stable over time. Because different hospital units incorporated the multicomponent delirium care pathway at different times, limiting enrollment to patients admitted and discharged from the same unit isolated the analysis to patients exposed to the pathway on each unit. This design also limited potential influence of other hospital quality improvement projects that might have occurred at the same time.

The primary limitation of this study is that screening for delirium was introduced as part of the multicomponent intervention. This decision was made to maximize buy-in from bedside nurses performing delirium screening because this addition to their workflow was explicitly linked to delirium prevention and management measures. Delirium could not be ascertained preintervention from the EMR because it is a clinical diagnosis and is coded inadequately.³⁰ We could only measure the change in delirium metrics after implementation of the delirium care pathway. Because baseline delirium rates before the intervention were not measured systematically, conclusions about the intervention’s association with delirium metrics are limited. All other outcomes were measured before and after the intervention.

Although the comprehensive delirium screening program and high rate of adherence are a methodologic strength of this study, a second limitation is the use of the NuDESC. Our previous research demonstrated that the NuDESC has low sensitivity but high specificity and positive predictive value,²⁶ which might underestimate delirium rates in this study. However, any underestimation should be stable over time and temporal trends should remain meaningful. This could allow more widespread study of delirium among hospitalized individuals. Because this care pathway was hospital-wide, it was important to ensure both consistency of screening and longevity of the initiative, and it was necessary to select a delirium assessment tool that was efficient and validated for nursing implementation. For these reasons, the NuDESC was an appropriate choice.

It is possible that our results could be influenced by unmeasured confounders. For example, although we incorporated Elixhauser medical comorbidities and illness severity into our model, we were unable to adjust for baseline functional status or frailty. Baseline functional status and frailty were not reliably recorded in the EMR, although these are potential confounders when investigating clinical outcomes including hospital readmission.

CONCLUSION

Implementation of a systematic, hospital-wide multicomponent delirium care pathway is associated with reductions in hospital LOS and 30-day readmission. In general medicine units, the reduction in LOS and associated cost savings were robust. These results demonstrate the feasibility and effectiveness of implementing an interprofessional, multidisciplinary multicomponent delirium care pathway through medical center funding to benefit patients and the hospital system.

Acknowledgments

The authors thank the many hospital staff members, especially the nurses, pharmacists, therapists, and patient care assistants, who helped implement the multicomponent delirium care pathway. All persons who have contributed significantly to this work are listed as authors of this work.

Delirium is an acute disturbance in mental status characterized by fluctuations in cognition and attention that affects more than 2.6 million hospitalized older adults in the United States annually, a rate that is expected to increase as the population ages.^1-4 Hospital-acquired delirium is associated with poor outcomes, including prolonged hospital length of stay (LOS), loss of independence, cognitive impairment, and even death.^5-10 Individuals who develop delirium do poorly after hospital discharge and are more likely to be readmitted within 30 days.¹¹ Approximately 30% to 40% of hospital-acquired delirium cases are preventable.^10,12 However, programs designed to prevent delirium and associated complications, such as increased LOS, have demonstrated variable success.^12-14 Many studies are limited by small sample sizes, lack of generalizability to different hospitalized patient populations, poor adherence, or reliance on outside funding.^12,13,15-18

Delirium prevention programs face several challenges because delirium could be caused by a variety of risk factors and precipitants.^19,20 Some risk factors that occur frequently among hospitalized patients can be mitigated, such as sensory impairment, immobility from physical restraints or urinary catheters, and polypharmacy.^20,21 Effective delirium care pathways targeting these risk factors must be multifaceted, interdisciplinary, and interprofessional. Accurate risk assessment is critical to allocate resources to high-risk patients. Delirium affects patients in all medical and surgical disciplines, and often is underdiagnosed.^19,22 Comprehensive screening is necessary to identify cases early and track outcomes, and educational efforts must reach all providers in the hospital. These challenges require a systematic, pragmatic approach to change.

The purpose of this study was to evaluate the association between a delirium care pathway and clinical outcomes for hospitalized patients. We hypothesized that this program would be associated with reduced hospital LOS, with secondary benefits to hospitalization costs, odds of 30-day readmission, and delirium rates.

METHODS

Study Design

In this retrospective cohort study, we compared clinical outcomes the year before and after implementation of a delirium care pathway across seven hospital units. The study period spanned October 1, 2015, through February 28, 2019. The study was approved by the University of California, San Francisco Institutional Review Board (#13-12500).

Multicomponent Delirium Care Pathway

The delirium care pathway was developed collaboratively among geriatrics, hospital medicine, neurology, anesthesiology, surgery, and psychiatry services, with an interprofessional team of physicians, nurses, pharmacists, and physical and occupational therapists. This pathway was implemented in units consecutively, approximately every 4 months in the following order: neurosciences, medicine, cardiology, general surgery, specialty surgery, hematology-oncology, and transplant. The same implementation education protocols were performed in each unit. The pathway consisted of several components targeting delirium prevention and management (Appendix Figure 1 and Appendix Figure 2). Systematic screening for delirium was introduced as part of the multicomponent intervention. Nursing staff assessed each patient’s risk of developing delirium at admission using the AWOL score, a validated delirium prediction tool.²³ AWOL consists of: patient Age, spelling “World” backwards correctly, Orientation, and assessment of iLlness severity by the nurse. For patients who spoke a language other than English, spelling of “world” backwards was translated to his or her primary language, or if this was not possible, the task was modified to serial 7s (subtracting 7 from 100 in a serial fashion). This modification has been validated for use in other languages.²⁴ Patients at high risk for delirium based on an AWOL score ≥2 received a multidisciplinary intervention with four components: (1) notifying the primary team by pager and electronic medical record (EMR), (2) a nurse-led, evidence-based, nonpharmacologic multicomponent intervention,²⁵ (3) placement of a delirium order set by the physician, and (4) review of medications by the unit pharmacist who adjusted administration timing to occur during waking hours and placed a note in the EMR notifying the primary team of potentially deliriogenic medications. The delirium order set reinforced the nonpharmacologic multicomponent intervention through a nursing order, placed an automatic consult to occupational therapy, and included options to order physical therapy, order speech/language therapy, obtain vital signs three times daily with minimal night interruptions, remove an indwelling bladder catheter, and prescribe melatonin as a sleep aid.

The bedside nurse screened all patients for active delirium every 12-hour shift using the Nursing Delirium Screening Scale (NuDESC) and entered the results into the EMR.^23,26 Capturing NuDESC results in the EMR allowed communication across medical providers as well as monitoring of screening adherence. Each nurse received two in-person trainings in staff meetings and one-to-one instruction during the first week of implementation. All nurses were required to complete a 15-minute training module and had the option of completing an additional 1-hour continuing medical education module. If a patient was transferred to the intensive care unit (ICU), delirium was identified through use of the ICU-specific Confusion Assessment Method (CAM-ICU) assessments, which the bedside nurse performed each shift throughout the intervention period.²⁷ Nurses were instructed to call the primary team physician after every positive screen. Before each unit’s implementation start date, physicians with patients on that unit received education through a combination of grand rounds, resident lectures and seminars, and a pocket card on delirium evaluation and management.

Participants and Eligibility Criteria

We included all patients aged ≥50 years hospitalized for >1 day on each hospital unit (Figure). We included adults aged ≥50 years to maximize the number of participants for this study while also capturing a population at risk for delirium. Because the delirium care pathway was unit-based and the pathway was rolled out sequentially across units, only patients who were admitted to and discharged from the same unit were included to better isolate the effect of the pathway. Patients who were transferred to the ICU were only included if they were discharged from the original unit of admission. Only the first hospitalization was included for patients with multiple hospitalizations during the study period.

Patient Characteristics

Patient demographics and clinical data were collected after discharge through Clarity and Vizient electronic databases (Table 1 and Table 2). All Elixhauser comorbidities were included except for the following International Classification of Disease, Tenth Revision, Clinical Modification (ICD-10) codes that overlapped with a delirium diagnosis: G31.2, G93.89, G93.9, G94, R41.0, and R41.82 (Appendix Table 1). Severity of illness was obtained from Vizient, which calculates illness severity based on clinical and claims data (Appendix Table 1).

Delirium Metrics

Delirium screening was introduced as part of the multicomponent intervention, and therefore delirium rates before the intervention could not be determined. Trends in delirium prevalence and incidence after the intervention are reported. Prevalent delirium was defined as a single score of ≥2 on the nurse-administered NuDESC or a positive CAM-ICU at any point during the hospital stay. Incident delirium was identified if the first NuDESC score was negative and any subsequent NuDESC or CAM-ICU score was positive.

Outcomes

The primary study outcome was hospital LOS across all participants. Secondary outcomes included total direct cost and odds of 30-day hospital readmission. Readmissions tracked as part of hospital quality reporting were obtained from Vizient and were not captured if they occurred at another hospital. We also examined rates of safety attendant and restraint use during the study period, defined as the number of safety attendant days or restraint days per 1,000 patient days.

Because previous studies have demonstrated the effectiveness of multicomponent delirium interventions among elderly general medical patients,¹² we also investigated these same outcomes in the medicine unit alone.

Statistical Analysis

The date of intervention implementation was determined for each hospital unit, which was defined as time(0) [t(0)]. The 12-month postintervention period was divided into four 3-month epochs to assess for trends. Data were aggregated across the seven units using t(0) as the start date, agnostic to the calendar month. Demographic and clinical characteristics were collected for the 12-months before t(0) and the four 3-month epochs after t(0). Univariate analysis of outcome variables comparing trends across the same epochs were conducted in the same manner, except for the rate of delirium, which was measured after t(0) and therefore could not be compared with the preintervention period.

Multivariable models were adjusted for age, sex, race/ethnicity, admission category, Elixhauser comorbidities, severity of illness quartile, and number days spent in the ICU. Admission category referred to whether the admission was emergent, urgent, or elective/unknown. Because it took 3 months after t(0) for each unit to reach a delirium screening compliance rate of 90%, the intervention was only considered fully implemented after this period. A ramp-up variable was set to 0 for admissions occurring prior to the intervention to t(0), 1/3 for admissions occurring 1 month post intervention, 2/3 for 2 months post intervention, and 1 for admissions occurring 3 to 12 months post intervention. In this way, the coefficient for the ramp-up variable estimated the postintervention versus preintervention effect. Numerical outcomes (LOS, cost) were log transformed to reduce skewness and analyzed using linear models. Coefficients were back-transformed to provide interpretations as proportional change in the median outcomes.

For LOS and readmission, we assessed secular trends by including admission date and admission date squared, in case the trend was nonlinear, as possible predictors; admission date was the specific date—not time from t(0)—to account for secular trends and allow contemporaneous controls in the analysis. To be conservative, we retained secular terms (first considering the quadratic and then the linear) if P <.10. The categorical outcome (30-day readmission) was analyzed using a logistic model. Count variables (delirium, safety attendants, restraints) were analyzed using Poisson regression models with a log link, and coefficients were back-transformed to provide rate ratio interpretations. Because delirium was not measured before t(0), and because the intervention was considered to take 3 months to become fully effective, baseline delirium rates were defined as those in the first 3 months adjusted by the ramp-up variable. For each outcome we included hospital unit, a ramp-up variable (measuring the pre- vs postintervention effect), and their interaction. If there was no statistically significant interaction, we presented the outcome for all units combined. If the interaction was statistically significant, we looked for consistency across units and reported results for all units combined when consistent, along with site-specific results. If the results were not consistent across the units, we provided site-specific results only. All statistical analyses were performed using SAS software, version 9.4 (SAS Institute Inc).

RESULTS

Participant Demographics and Clinical Characteristics

A total of 22,708 individuals were included in this study, with 11,018 in the preintervention period (Table 1 and Table 2). Most patients were cared for on the general surgery unit (n = 5,899), followed by the medicine unit (n = 4,923). The smallest number of patients were cared for on the hematology-oncology unit (n = 1,709). Across the five epochs, patients were of similar age and sex, and spent a similar number of days in the ICU. The population was diverse with regard to race and ethnicity; there were minor differences in admission category. There were also minor differences in severity of illness and some comorbidities between timepoints (Appendix Table 1).

Delirium Metrics

Delirium prevalence was 13.0% during the first epoch post intervention, followed by 12.0%, 11.7%, and 13.0% in the subsequent epochs (P = .91). Incident delirium occurred in 6.1% of patients during the first epoch post intervention, followed by 5.3%, 5.3%, and 5.8% in the subsequent epochs (P = .63).

Primary Outcome

Epoch-level data for LOS before and after the intervention is shown in Appendix Table 2. The mean unadjusted LOS for all units combined did not decrease after the intervention, but in the adjusted model, the mean LOS decreased by 2% after the intervention (P = .0087; Table 3).

Secondary Outcomes

The odds of 30-day readmission decreased by 14% (P = .0002) in the adjusted models for all units combined (Table 3). There was no statistically significant reduction in adjusted total direct hospitalization cost or rate of restraint use. The safety attendant results showed strong effect modification across sites; the site-specific estimates are provided in Appendix Table 3. However, the estimated values all showed reductions, and a number were large and statistically significant.

Medicine Unit Outcomes

On the medicine unit alone, we observed a statistically significant reduction in LOS of 9% after implementation of the delirium care pathway (P = .028) in the adjusted model (Table 3). There was an associated 7% proportional decrease in total direct cost (P = .0002). Reductions in 30-day readmission and safety attendant use did not remain statistically significant in the adjusted models.

DISCUSSION

Implementation of a hospital-wide multicomponent delirium care pathway was associated with reduced hospital LOS and 30-day hospital readmission in a study of 22,708 hospitalized adults at a tertiary care, university hospital in Northern California, encompassing both medical and surgical acute care patients. When evaluating general medicine patients alone, pathway implementation was associated with reductions in LOS and total direct cost. The cost savings of 7% among medical patients translates to median savings of $1,237 per hospitalization. This study—one of the largest to date examining implementation of a hospital-wide delirium care pathway—supports use of a multicomponent delirium care pathway for older adults hospitalized for a range of conditions.

Multicomponent pathways for delirium prevention and management are increasingly being used in hospital settings. The United Kingdom National Institute for Health and Care Excellence guidelines recommend delirium assessment and intervention by a multidisciplinary team within 24 hours of hospital admission for those at risk.²⁵ These guidelines are based on evidence accumulated in clinical studies over the past 30 years suggesting that multicomponent interventions reduce incident delirium by 30% to 40% among medical and surgical patients.^12,13,25,28

Although multicomponent delirium care pathways are associated with improved patient outcomes, the specific clinical benefits might vary across patient populations. Here, we found larger reductions in LOS and total direct cost among medicine patients. Medical patients might respond more robustly to nonpharmacologic multicomponent delirium interventions because of differing delirium etiologies (eg, constipation and sleep deprivation in a medical patient vs seizures or encephalitis in a neurosciences patient). Another explanation for the difference observed in total direct cost might be the inclusion of surgical units in the total study population. For example, not all hospital days are equivalent in cost for patients on a surgical unit.²⁹ For patients requiring surgical care, most of the hospitalization cost might be incurred during the initial days of hospitalization, when there are perioperative costs; therefore, reduced LOS might have a lower economic impact.²⁹ Multicomponent, nonpharmacologic delirium interventions encourage discontinuing restraints. As a result, one might expect a need for more frequent safety attendant use and an associated cost increase. However, we found that the estimated unit-specific values for safety attendant use showed reductions, which were large and highly statistically significant. For all units combined and the medicine unit alone, we found that the rate of restraint use decreased, although the change was not statistically significant. It is possible that some of the interventions taught to nurses and physicians as part of care pathway implementation, such as the use of family support for at-risk and delirious patients, led to a reduction in both safety attendants and restraints.

Our study had several strengths. This is one of the largest hospital-based delirium interventions studied, both in terms of its scope across seven diverse medical and surgical hospital units and the number of hospitalized patients studied. This intervention did not require additional staff or creating a specialized ward. Adherence to the pathway, as measured by risk assessment and delirium screening, was high (>90%) 3 months after implementation. This allowed for robust outcome ascertainment. The patient population’s characteristics and rates of delirium were stable over time. Because different hospital units incorporated the multicomponent delirium care pathway at different times, limiting enrollment to patients admitted and discharged from the same unit isolated the analysis to patients exposed to the pathway on each unit. This design also limited potential influence of other hospital quality improvement projects that might have occurred at the same time.

The primary limitation of this study is that screening for delirium was introduced as part of the multicomponent intervention. This decision was made to maximize buy-in from bedside nurses performing delirium screening because this addition to their workflow was explicitly linked to delirium prevention and management measures. Delirium could not be ascertained preintervention from the EMR because it is a clinical diagnosis and is coded inadequately.³⁰ We could only measure the change in delirium metrics after implementation of the delirium care pathway. Because baseline delirium rates before the intervention were not measured systematically, conclusions about the intervention’s association with delirium metrics are limited. All other outcomes were measured before and after the intervention.

Although the comprehensive delirium screening program and high rate of adherence are a methodologic strength of this study, a second limitation is the use of the NuDESC. Our previous research demonstrated that the NuDESC has low sensitivity but high specificity and positive predictive value,²⁶ which might underestimate delirium rates in this study. However, any underestimation should be stable over time and temporal trends should remain meaningful. This could allow more widespread study of delirium among hospitalized individuals. Because this care pathway was hospital-wide, it was important to ensure both consistency of screening and longevity of the initiative, and it was necessary to select a delirium assessment tool that was efficient and validated for nursing implementation. For these reasons, the NuDESC was an appropriate choice.

It is possible that our results could be influenced by unmeasured confounders. For example, although we incorporated Elixhauser medical comorbidities and illness severity into our model, we were unable to adjust for baseline functional status or frailty. Baseline functional status and frailty were not reliably recorded in the EMR, although these are potential confounders when investigating clinical outcomes including hospital readmission.

CONCLUSION

Implementation of a systematic, hospital-wide multicomponent delirium care pathway is associated with reductions in hospital LOS and 30-day readmission. In general medicine units, the reduction in LOS and associated cost savings were robust. These results demonstrate the feasibility and effectiveness of implementing an interprofessional, multidisciplinary multicomponent delirium care pathway through medical center funding to benefit patients and the hospital system.

Acknowledgments

The authors thank the many hospital staff members, especially the nurses, pharmacists, therapists, and patient care assistants, who helped implement the multicomponent delirium care pathway. All persons who have contributed significantly to this work are listed as authors of this work.

From Titus Family Department of Clinical Pharmacy, School of Pharmacy, University of Southern California, Los Angeles, CA (Drs. Chu and Ma and Mimi Lou), and Department of Family Medicine, Keck Medicine, University of Southern California, Los Angeles, CA (Dr. Suh).

Objective: The objective of this study is to describe HbA_1c changes in patients who maintained continuous pharmacist care vs patients who had a gap in pharmacist care of 3 months or longer. 

Methods: This retrospective study was conducted from October 1, 2018, to September 30, 2019. Electronic health record data from an academic-affiliated, safety-net resident physician primary care clinic were collected to observe HbA_1c changes between patients with continuous pharmacist care and patients who had a gap of 3 months or longer in pharmacist care. A total of 189 patients met the inclusion criteria and were divided into 2 groups: those with continuous care and those with gaps in care. Data were analyzed using the Mann-Whitney test for continuous variables and the χ² (or Fisher exact) test for categorical variables. The differences-in-differences model was used to compare the changes in HbA_1c between the 2 groups.

Results: There was no significant difference in changes in HbA_1c between the continuous care group and the gaps in care group, although the mean magnitude of HbA_1c changes was numerically greater in the continuous care group (-1.48% vs -0.97%). Overall, both groups showed improvement in their HbA_1c levels and had similar numbers of primary care physician visits and acute care utilizations, while the gaps in care group had longer duration with pharmacists and between the adjacent pharmacist visits.

Conclusion: Maintaining continuous, regular visits with a pharmacist at a safety-net resident physician primary care clinic did not show a significant difference in HbA_1c changes compared to having gaps in pharmacist care. Future studies on socioeconomic and behavioral burden on HbA_1c improvement and on pharmacist visits in these populations should be explored.

Keywords: clinical pharmacist; diabetes management; continuous visit; primary care clinic.

Pharmacists have unique skills in identifying and resolving problems related to the safety and efficacy of drug therapy while addressing medication adherence and access for patients. Their expertise is especially important to meet the care needs of a growing population with chronic conditions amidst a primary care physician shortage.¹ As health care systems move toward value-based care, emphasis on improvement in quality and health measures have become central in care delivery. Pharmacists have been integrated into team-based care in primary care settings, but the value-based shift has opened more opportunities for pharmacists to address unmet quality standards.^2-5

 

Many studies have reported that the integration of pharmacists into team-based care improves health outcomes and reduces overall health care costs.^6-9 Specifically, when pharmacists were added to primary care teams to provide diabetes management, hemoglobin HbA_1c levels were reduced compared to teams without pharmacists.^10-13 Offering pharmacist visits as often as every 2 weeks to 3 months, with each patient having an average of 4.7 visits, resulted in improved therapeutic outcomes.^3,7 During visits, pharmacists address the need for additional drug therapy, deprescribe unnecessary therapy, correct insufficient doses or durations, and switch patients to more cost-efficient drug therapy.⁹ Likewise, patients who visit pharmacists in addition to seeing their primary care physician can have medication-related concerns resolved and improve their therapeutic outcomes.^10,11

Not much is known about the magnitude of HbA_1c change based on the regularity of pharmacist visits. Although pharmacists offer follow-up appointments in reasonable time intervals, patients do not keep every appointment for a variety of reasons, including forgetfulness, personal issues, and a lack of transportation.¹⁴ Such missed appointments can negatively impact health outcomes.^14-16 The purpose of this study is to describe HbA_1c changes in patients who maintained continuous, regular pharmacist visits without a 3-month gap and in patients who had history of inconsistent pharmacist visits with a gap of 3 months or longer. Furthermore, this study describes the frequency of health care utilization for these 2 groups.

Methods

Setting

The Internal Medicine resident physician primary care clinic is 1 of 2 adult primary care clinics at an academic, urban, public medical center. It is in the heart of East Los Angeles, where predominantly Spanish-speaking and minority populations reside. The clinic has approximately 19000 empaneled patients and is the largest resident primary care clinic in the public health system. The clinical pharmacy service addresses unmet quality standards, specifically HbA_1c. The clinical pharmacists are co-located and collaborate with resident physicians, attending physicians, care managers, nurses, social workers, and community health workers at the clinic. They operate under collaborative practice agreements with prescriptive authority, except for controlled substances, specialty drugs, and antipsychotic medications.

Pharmacist visit

Patients are primarily referred by resident physicians to clinical pharmacists when their HbA_1c level is above 8% for an extended period, when poor adherence and low health literacy are evident regardless of HbA_1c level, or when a complex medication regimen requires comprehensive medication review and reconciliation. The referral occurs through warm handoff by resident physicians as well as clinic nurses, and it is embedded in the clinic flow. Patients continue their visits with resident physicians for issues other than their referral to clinical pharmacists. The visits with pharmacists are appointment-based, occur independently from resident physician visits, and continue until the patient’s HbA_1c level or adherence is optimized. Clinical pharmacists continue to follow up with patients who may have reached their target HbA_1c level but still are deemed unstable due to inconsistency in their self-management and medication adherence.

After the desirable HbA_1c target is achieved along with full adherence to medications and self-management, clinical pharmacists will hand off patients back to resident physicians. At each visit, pharmacists perform a comprehensive medication assessment and reconciliation that includes adjusting medication therapy, placing orders for necessary laboratory tests and prescriptions, and assessing medication adherence. They also evaluate patients’ signs and symptoms for hyperglycemic complications, hypoglycemia, and other potential treatment-related adverse events. These are all within the pharmacist’s scope of practice in comprehensive medication management. Patient education is provided with the teach-back method and includes lifestyle modifications and medication counseling (Table 1). Pharmacists offer face-to-face visits as frequently as every 1 to 2 weeks to every 4 to 6 weeks, depending on the level of complexity and the severity of a patient’s conditions and medications. For patients whose HbA_1c has reached the target range but have not been deemed stable, pharmacists continue to check in with them every 2 months. Phone visits are also utilized as an additional care delivery method for patients having difficulty showing up for face-to-face visits or needing quick assessment of medication adherence and responses to changes in drug treatment in between the face-to-face visits. The maximal interval between pharmacist visits is offered no longer than every 8 weeks. Patients are contacted via phone or mail by the nursing staff to reschedule if they miss their appointments with pharmacists. Every pharmacy visit is documented in the patient’s electronic medical record.

 

Study design

This is a retrospective study describing the HbA_1c changes in a patient group that maintained pharmacist visits, with each interval less than 3 months, and in another group, who had a history of a 3-month or longer gap between pharmacist visits. The data were obtained from patients’ electronic medical records during the study period of October 1, 2018, and September 30, 2019, and collected using a HIPAA-compliant, electronic data storage website, REDCap. The institutional review board approval was obtained under HS-19-00929. Patients 18 years and older who were referred by primary care resident physicians for diabetes management, and had 2 or more visits with a pharmacist within the study period, were included. Patients were excluded if they had only 1 HbA_1c drawn during the study period, were referred to a pharmacist for reasons other than diabetes management, were concurrently managed by an endocrinologist, had only 1 visit with a pharmacist, or had no visits with their primary care resident physician for over a year. The patients were then divided into 2 groups: continuous care cohort (CCC) and gap in care cohort (GCC). Both face-to-face and phone visits were counted as pharmacist visits for each group.

Outcomes

The primary outcome was the change in HbA_1c from baseline between the 2 groups. Baseline HbA_1c was considered as the HbA_1c value obtained within 3 months prior to, or within 1 month, of the first visit with the pharmacist during the study period. The final HbA_1c was considered the value measured within 1 month of, or 3 months after, the patient’s last visit with the pharmacist during the study period.

Several subgroup analyses were conducted to examine the relationship between HbA_1c and each group. Among patients whose baseline HbA_1c was ≥ 8%, we looked at the percentage of patients reaching HbA_1c < 8%, the percentage of patients showing any level of improvement in HbA_1c, and the change in HbA_1c for each group. We also looked at the percentage of patients with baseline HbA_1c < 8% maintaining the level throughout the study period and the change in HbA_1c for each group. Additionally, we looked at health care utilization, which included pharmacist visits, primary care physician visits, emergency room and urgent care visits, and hospitalizations for each group. The latter 3 types of utilization were grouped as acute care utilization and further analyzed for visit reasons, which were subsequently categorized as diabetes related and non-diabetes related. The diabetes related reasons linking to acute care utilization were defined as any episodes related to hypoglycemia, diabetic ketoacidosis (DKA), hyperosmolar hyperglycemic state (HHS), foot ulcers, retinopathy, and osteomyelitis infection. All other reasons leading to acute care utilization were categorized as non-diabetes related.

Statistical analysis

Descriptive analyses were conducted using the Mann-Whitney test for continuous data and χ² (or Fisher exact) test for categorical data. A basic difference-in-differences (D-I-D) method was used to compare the changes of HbA_1c between the CCC and GCC over 2 time points: baseline and final measurements. The repeated measures ANOVA was used for analyzing D-I-D. P < .05 was considered significant. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC).

Results

Baseline data

A total of 1272 patients were identified within the study period, and 189 met the study inclusion criteria. The CCC included 132 patients, the GCC 57. The mean age of patients in both groups was similar at 57 years old (P = .39). Most patients had Medicaid as their primary insurance. About one-third of patients in each group experienced clinical atherosclerotic cardiovascular disease, and about 12% overall had chronic kidney disease stage 3 and higher. The average number of days that patients were under pharmacist care during the study period was longer in the GCC compared to the CCC, and it was statistically significant (P < .001) (Table 2). The mean ± SD baseline HbA_1c for the CCC and GCC was 10.0% ± 2.0% and 9.9% ± 1.7%, respectively, and the difference was not statistically significant (P = .93). About 86% of patients in the CCC and 90% in the GCC had a baseline HbA_1c of ≥ 8%.

 

 

HbA_1c

The mean change in HbA_1c between the 2 groups was not statistically significant (-1.5% ± 2.0% in the CCC vs -1.0% ± 2.1% in the GCC, P = .36) (Table 3). However, an absolute mean HbA_1c reduction of 1.3% was observed in both groups combined at the end of the study. Figure 1 shows a D-I-D model of the 2 groups. Based on the output, the P value of .11 on the interaction term (time*group) indicates that the D-I-D in HbA_1c change from baseline to final between the CCC and GCC is not statistically different. However, the magnitude of the difference calculated from the LSMEANS results showed a trend. The HbA_1c from baseline to final measurement of patients in the GCC declined by 0.97 percentage points (from 9.94% to 8.97%), while those in the CCC saw their HbA_1c decline by 1.48 percentage points (from 9.96% to 8.48%), for a D-I-D of 0.51. In other words, those in the GCC had an HbA_1c that decreased by 0.51% less than that of patients in the CCC, suggesting that the CCC shows a steeper line declining from baseline to final HbA_1c compared to the GCC, whose line declines less sharply.

In the subgroup analysis of patients whose baseline HbA_1c was ≥ 8%, about 42% in the CCC and 37% in the GCC achieved an HbA_1c < 8% (P = .56) (Table 4). Approximately 83% of patients in the CCC had some degree of HbA_1c improvement—the final HbA_1c was lower than their baseline HbA_1c—whereas this was observed in about 75% of patients in the GCC (P = .19). Of patients whose baseline HbA_1c was < 8%, there was no significant difference in proportion of patients maintaining an HbA_1c < 8% between the groups (P = .57), although some increases in HbA_1c and HbA_1c changes were observed in the GCC (Table 5).

Health care utilization

Patients in the CCC visited pharmacists 5 times on average over 12 months, whereas patients in the GCC had an average of 6 visits (5 ± 2.6 in the CCC vs 6 ± 2.6 in the GCC, P = .01) (Table 6). The mean length between any 2 adjacent visits was significantly different, averaging about 33 days in the CCC compared to 64 days in the GCC (33.2 ± 10 in the CCC vs 63.7 ± 39.4 in the GCC, P < .001). As shown in Figure 2, the GCC shows wider ranges between any adjacent pharmacy visits throughout until the 10th visit. Both groups had a similar number of visits with primary care physicians during the same time period (4.6 ± 1.86 in the CCC vs 4.3 ± 2.51 in the GCC, P = .44). About 30% of patients in the CCC and 47% in the GCC had at least 1 visit to the emergency room or urgent care or had at least 1 hospital admission, for a total of 124 acute care utilizations between the 2 groups combined. Only a small fraction of acute care visits with or without hospitalizations were related to diabetes and its complications (23.1% in the CCC vs 22.0% in the GCC).

Discussion

This is a real-world study that describes HbA_1c changes in patients who maintained pharmacy visits regularly and in those who had a history of a 3-month or longer gap in pharmacy visits. Although the study did not show statistically significant differences in HbA_1c reduction between the 2 groups, pharmacists’ care, overall, provided mean HbA_1c reductions of 1.3%. This result is consistent with those from multiple previous studies.^10-13 It is worth noting that the final HbA_1c was numerically lower in patients who followed up with pharmacists regularly than in patients with gaps in visits, with a difference of about 0.5 percentage points. This difference is considered clinically significant,¹⁷ and potentially could be even greater if the study duration was longer, as depicted by the slope of HbA_1c reductions in the D-I-D model (Figure 1).

Previous studies have shown that pharmacist visits are conducted in shorter intervals than primary care physician visits to provide closer follow-up and to resolve any medication-related problems that may hinder therapeutic outcome improvements.^3-4,7-9 Increasing access via pharmacists is particularly important in this clinic, where resident physician continuity and access is challenging. The pharmacist-driven program described in this study does not deviate from the norm, and this study confirms that pharmacist care, regardless of gaps in pharmacist visits, may still be beneficial.

 

Another notable finding from this study was that although the average number of pharmacist visits per patient was significantly different, this difference of 1 visit did not result in a statistically significant improvement in HbA_1c. In fact, the average number of pharmacist visits per patient seemed to be within the reported range by Choe et al in a similar setting.⁷ Conversely, patients with a history of a gap in pharmacist visits spent longer durations under pharmacist care compared to those who had continuous follow-up. This could mean that it may take longer times or 1 additional visit to achieve similar HbA_1c results with continuous pharmacist care. Higher number of visits with pharmacists in the group with the history of gaps between pharmacist visits could have been facilitated by resident physicians, as both groups had a similar number of visits with them. Although this is not conclusive, identifying the optimal number of visits with pharmacists in this underserved population could be beneficial in strategizing pharmacist visits. Acute care utilization was not different between the 2 groups, and most cases that led to acute care utilization were not directly related to diabetes or its complications.

The average HbA_1c at the end of the study did not measure < 8%, a target that was reached by less than half of patients from each group; however, this study is a snapshot of a series of ongoing clinical pharmacy services. About 25% of our patients started their first visit with a pharmacist less than 6 months from the study end date, and these patients may not have had enough time with pharmacists for their HbA_1c to reach below the target goal. In addition, most patients in this clinic were enrolled in public health plans and may carry a significant burden of social and behavioral factors that can affect diabetes management.^18,19 These patients may need longer care by pharmacists along with other integrated services, such as behavioral health and social work, to achieve optimal HbA_1c levels.²⁰

There are several limitations to this study, including the lack of a propensity matched control group of patients who only had resident physician visits; thus, it is hard to test the true impact of continuous or intermittent pharmacist visits on the therapeutic outcomes. The study also does not address potential social, economic, and physical environment factors that might have contributed to pharmacist visits and to overall diabetes care. These factors can negatively impact diabetes control and addressing them could help with an individualized diabetes management approach.^17,18 Additionally, by nature of being a descriptive study, the results may be subject to undetermined confounding factors.

Conclusion

Patients maintaining continuous pharmacist visits do not have statistically significant differences in change in HbA_1c compared to patients who had a history of 3-month or longer gaps in pharmacist visits at a resident physician primary care safety-net clinic. However, patients with diabetes will likely derive a benefit in HbA_1c reduction regardless of regularity of pharmacist care. This finding still holds true in collaboration with resident physicians who also regularly meet with patients.

The study highlights that it is important to integrate clinical pharmacists into primary care teams for improved therapeutic outcomes. It is our hope that regular visits to pharmacists can be a gateway for behavioral health and social work referrals, thereby addressing pharmacist-identified social barriers. Furthermore, exploration of socioeconomic and behavioral barriers to pharmacist visits is necessary to address and improve the patient experience, health care delivery, and health outcomes.

Acknowledgments: The authors thank Roxanna Perez, PharmD, Amy Li, and Julie Dopheide, PharmD, BCPP, FASHP for their contributions to this project.

Corresponding author: Michelle Koun Lee Chu, PharmD, BCACP, APh, Titus Family Department of Clinical Pharmacy, School of Pharmacy, University of Southern California, 1985 Zonal Ave, Los Angeles, CA 90089-9121; lee118@usc.edu.

Financial disclosures: None.

From Titus Family Department of Clinical Pharmacy, School of Pharmacy, University of Southern California, Los Angeles, CA (Drs. Chu and Ma and Mimi Lou), and Department of Family Medicine, Keck Medicine, University of Southern California, Los Angeles, CA (Dr. Suh).

Objective: The objective of this study is to describe HbA_1c changes in patients who maintained continuous pharmacist care vs patients who had a gap in pharmacist care of 3 months or longer. 

Methods: This retrospective study was conducted from October 1, 2018, to September 30, 2019. Electronic health record data from an academic-affiliated, safety-net resident physician primary care clinic were collected to observe HbA_1c changes between patients with continuous pharmacist care and patients who had a gap of 3 months or longer in pharmacist care. A total of 189 patients met the inclusion criteria and were divided into 2 groups: those with continuous care and those with gaps in care. Data were analyzed using the Mann-Whitney test for continuous variables and the χ² (or Fisher exact) test for categorical variables. The differences-in-differences model was used to compare the changes in HbA_1c between the 2 groups.

Results: There was no significant difference in changes in HbA_1c between the continuous care group and the gaps in care group, although the mean magnitude of HbA_1c changes was numerically greater in the continuous care group (-1.48% vs -0.97%). Overall, both groups showed improvement in their HbA_1c levels and had similar numbers of primary care physician visits and acute care utilizations, while the gaps in care group had longer duration with pharmacists and between the adjacent pharmacist visits.

Conclusion: Maintaining continuous, regular visits with a pharmacist at a safety-net resident physician primary care clinic did not show a significant difference in HbA_1c changes compared to having gaps in pharmacist care. Future studies on socioeconomic and behavioral burden on HbA_1c improvement and on pharmacist visits in these populations should be explored.

Keywords: clinical pharmacist; diabetes management; continuous visit; primary care clinic.

Pharmacists have unique skills in identifying and resolving problems related to the safety and efficacy of drug therapy while addressing medication adherence and access for patients. Their expertise is especially important to meet the care needs of a growing population with chronic conditions amidst a primary care physician shortage.¹ As health care systems move toward value-based care, emphasis on improvement in quality and health measures have become central in care delivery. Pharmacists have been integrated into team-based care in primary care settings, but the value-based shift has opened more opportunities for pharmacists to address unmet quality standards.^2-5

 

Many studies have reported that the integration of pharmacists into team-based care improves health outcomes and reduces overall health care costs.^6-9 Specifically, when pharmacists were added to primary care teams to provide diabetes management, hemoglobin HbA_1c levels were reduced compared to teams without pharmacists.^10-13 Offering pharmacist visits as often as every 2 weeks to 3 months, with each patient having an average of 4.7 visits, resulted in improved therapeutic outcomes.^3,7 During visits, pharmacists address the need for additional drug therapy, deprescribe unnecessary therapy, correct insufficient doses or durations, and switch patients to more cost-efficient drug therapy.⁹ Likewise, patients who visit pharmacists in addition to seeing their primary care physician can have medication-related concerns resolved and improve their therapeutic outcomes.^10,11

Not much is known about the magnitude of HbA_1c change based on the regularity of pharmacist visits. Although pharmacists offer follow-up appointments in reasonable time intervals, patients do not keep every appointment for a variety of reasons, including forgetfulness, personal issues, and a lack of transportation.¹⁴ Such missed appointments can negatively impact health outcomes.^14-16 The purpose of this study is to describe HbA_1c changes in patients who maintained continuous, regular pharmacist visits without a 3-month gap and in patients who had history of inconsistent pharmacist visits with a gap of 3 months or longer. Furthermore, this study describes the frequency of health care utilization for these 2 groups.

Methods

Setting

The Internal Medicine resident physician primary care clinic is 1 of 2 adult primary care clinics at an academic, urban, public medical center. It is in the heart of East Los Angeles, where predominantly Spanish-speaking and minority populations reside. The clinic has approximately 19000 empaneled patients and is the largest resident primary care clinic in the public health system. The clinical pharmacy service addresses unmet quality standards, specifically HbA_1c. The clinical pharmacists are co-located and collaborate with resident physicians, attending physicians, care managers, nurses, social workers, and community health workers at the clinic. They operate under collaborative practice agreements with prescriptive authority, except for controlled substances, specialty drugs, and antipsychotic medications.

Pharmacist visit

Patients are primarily referred by resident physicians to clinical pharmacists when their HbA_1c level is above 8% for an extended period, when poor adherence and low health literacy are evident regardless of HbA_1c level, or when a complex medication regimen requires comprehensive medication review and reconciliation. The referral occurs through warm handoff by resident physicians as well as clinic nurses, and it is embedded in the clinic flow. Patients continue their visits with resident physicians for issues other than their referral to clinical pharmacists. The visits with pharmacists are appointment-based, occur independently from resident physician visits, and continue until the patient’s HbA_1c level or adherence is optimized. Clinical pharmacists continue to follow up with patients who may have reached their target HbA_1c level but still are deemed unstable due to inconsistency in their self-management and medication adherence.

After the desirable HbA_1c target is achieved along with full adherence to medications and self-management, clinical pharmacists will hand off patients back to resident physicians. At each visit, pharmacists perform a comprehensive medication assessment and reconciliation that includes adjusting medication therapy, placing orders for necessary laboratory tests and prescriptions, and assessing medication adherence. They also evaluate patients’ signs and symptoms for hyperglycemic complications, hypoglycemia, and other potential treatment-related adverse events. These are all within the pharmacist’s scope of practice in comprehensive medication management. Patient education is provided with the teach-back method and includes lifestyle modifications and medication counseling (Table 1). Pharmacists offer face-to-face visits as frequently as every 1 to 2 weeks to every 4 to 6 weeks, depending on the level of complexity and the severity of a patient’s conditions and medications. For patients whose HbA_1c has reached the target range but have not been deemed stable, pharmacists continue to check in with them every 2 months. Phone visits are also utilized as an additional care delivery method for patients having difficulty showing up for face-to-face visits or needing quick assessment of medication adherence and responses to changes in drug treatment in between the face-to-face visits. The maximal interval between pharmacist visits is offered no longer than every 8 weeks. Patients are contacted via phone or mail by the nursing staff to reschedule if they miss their appointments with pharmacists. Every pharmacy visit is documented in the patient’s electronic medical record.

 

Study design

This is a retrospective study describing the HbA_1c changes in a patient group that maintained pharmacist visits, with each interval less than 3 months, and in another group, who had a history of a 3-month or longer gap between pharmacist visits. The data were obtained from patients’ electronic medical records during the study period of October 1, 2018, and September 30, 2019, and collected using a HIPAA-compliant, electronic data storage website, REDCap. The institutional review board approval was obtained under HS-19-00929. Patients 18 years and older who were referred by primary care resident physicians for diabetes management, and had 2 or more visits with a pharmacist within the study period, were included. Patients were excluded if they had only 1 HbA_1c drawn during the study period, were referred to a pharmacist for reasons other than diabetes management, were concurrently managed by an endocrinologist, had only 1 visit with a pharmacist, or had no visits with their primary care resident physician for over a year. The patients were then divided into 2 groups: continuous care cohort (CCC) and gap in care cohort (GCC). Both face-to-face and phone visits were counted as pharmacist visits for each group.

Outcomes

The primary outcome was the change in HbA_1c from baseline between the 2 groups. Baseline HbA_1c was considered as the HbA_1c value obtained within 3 months prior to, or within 1 month, of the first visit with the pharmacist during the study period. The final HbA_1c was considered the value measured within 1 month of, or 3 months after, the patient’s last visit with the pharmacist during the study period.

Several subgroup analyses were conducted to examine the relationship between HbA_1c and each group. Among patients whose baseline HbA_1c was ≥ 8%, we looked at the percentage of patients reaching HbA_1c < 8%, the percentage of patients showing any level of improvement in HbA_1c, and the change in HbA_1c for each group. We also looked at the percentage of patients with baseline HbA_1c < 8% maintaining the level throughout the study period and the change in HbA_1c for each group. Additionally, we looked at health care utilization, which included pharmacist visits, primary care physician visits, emergency room and urgent care visits, and hospitalizations for each group. The latter 3 types of utilization were grouped as acute care utilization and further analyzed for visit reasons, which were subsequently categorized as diabetes related and non-diabetes related. The diabetes related reasons linking to acute care utilization were defined as any episodes related to hypoglycemia, diabetic ketoacidosis (DKA), hyperosmolar hyperglycemic state (HHS), foot ulcers, retinopathy, and osteomyelitis infection. All other reasons leading to acute care utilization were categorized as non-diabetes related.

Statistical analysis

Descriptive analyses were conducted using the Mann-Whitney test for continuous data and χ² (or Fisher exact) test for categorical data. A basic difference-in-differences (D-I-D) method was used to compare the changes of HbA_1c between the CCC and GCC over 2 time points: baseline and final measurements. The repeated measures ANOVA was used for analyzing D-I-D. P < .05 was considered significant. Statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC).

Results

Baseline data

A total of 1272 patients were identified within the study period, and 189 met the study inclusion criteria. The CCC included 132 patients, the GCC 57. The mean age of patients in both groups was similar at 57 years old (P = .39). Most patients had Medicaid as their primary insurance. About one-third of patients in each group experienced clinical atherosclerotic cardiovascular disease, and about 12% overall had chronic kidney disease stage 3 and higher. The average number of days that patients were under pharmacist care during the study period was longer in the GCC compared to the CCC, and it was statistically significant (P < .001) (Table 2). The mean ± SD baseline HbA_1c for the CCC and GCC was 10.0% ± 2.0% and 9.9% ± 1.7%, respectively, and the difference was not statistically significant (P = .93). About 86% of patients in the CCC and 90% in the GCC had a baseline HbA_1c of ≥ 8%.

 

 

HbA_1c

The mean change in HbA_1c between the 2 groups was not statistically significant (-1.5% ± 2.0% in the CCC vs -1.0% ± 2.1% in the GCC, P = .36) (Table 3). However, an absolute mean HbA_1c reduction of 1.3% was observed in both groups combined at the end of the study. Figure 1 shows a D-I-D model of the 2 groups. Based on the output, the P value of .11 on the interaction term (time*group) indicates that the D-I-D in HbA_1c change from baseline to final between the CCC and GCC is not statistically different. However, the magnitude of the difference calculated from the LSMEANS results showed a trend. The HbA_1c from baseline to final measurement of patients in the GCC declined by 0.97 percentage points (from 9.94% to 8.97%), while those in the CCC saw their HbA_1c decline by 1.48 percentage points (from 9.96% to 8.48%), for a D-I-D of 0.51. In other words, those in the GCC had an HbA_1c that decreased by 0.51% less than that of patients in the CCC, suggesting that the CCC shows a steeper line declining from baseline to final HbA_1c compared to the GCC, whose line declines less sharply.

In the subgroup analysis of patients whose baseline HbA_1c was ≥ 8%, about 42% in the CCC and 37% in the GCC achieved an HbA_1c < 8% (P = .56) (Table 4). Approximately 83% of patients in the CCC had some degree of HbA_1c improvement—the final HbA_1c was lower than their baseline HbA_1c—whereas this was observed in about 75% of patients in the GCC (P = .19). Of patients whose baseline HbA_1c was < 8%, there was no significant difference in proportion of patients maintaining an HbA_1c < 8% between the groups (P = .57), although some increases in HbA_1c and HbA_1c changes were observed in the GCC (Table 5).

Health care utilization

Patients in the CCC visited pharmacists 5 times on average over 12 months, whereas patients in the GCC had an average of 6 visits (5 ± 2.6 in the CCC vs 6 ± 2.6 in the GCC, P = .01) (Table 6). The mean length between any 2 adjacent visits was significantly different, averaging about 33 days in the CCC compared to 64 days in the GCC (33.2 ± 10 in the CCC vs 63.7 ± 39.4 in the GCC, P < .001). As shown in Figure 2, the GCC shows wider ranges between any adjacent pharmacy visits throughout until the 10th visit. Both groups had a similar number of visits with primary care physicians during the same time period (4.6 ± 1.86 in the CCC vs 4.3 ± 2.51 in the GCC, P = .44). About 30% of patients in the CCC and 47% in the GCC had at least 1 visit to the emergency room or urgent care or had at least 1 hospital admission, for a total of 124 acute care utilizations between the 2 groups combined. Only a small fraction of acute care visits with or without hospitalizations were related to diabetes and its complications (23.1% in the CCC vs 22.0% in the GCC).

Discussion

This is a real-world study that describes HbA_1c changes in patients who maintained pharmacy visits regularly and in those who had a history of a 3-month or longer gap in pharmacy visits. Although the study did not show statistically significant differences in HbA_1c reduction between the 2 groups, pharmacists’ care, overall, provided mean HbA_1c reductions of 1.3%. This result is consistent with those from multiple previous studies.^10-13 It is worth noting that the final HbA_1c was numerically lower in patients who followed up with pharmacists regularly than in patients with gaps in visits, with a difference of about 0.5 percentage points. This difference is considered clinically significant,¹⁷ and potentially could be even greater if the study duration was longer, as depicted by the slope of HbA_1c reductions in the D-I-D model (Figure 1).

Previous studies have shown that pharmacist visits are conducted in shorter intervals than primary care physician visits to provide closer follow-up and to resolve any medication-related problems that may hinder therapeutic outcome improvements.^3-4,7-9 Increasing access via pharmacists is particularly important in this clinic, where resident physician continuity and access is challenging. The pharmacist-driven program described in this study does not deviate from the norm, and this study confirms that pharmacist care, regardless of gaps in pharmacist visits, may still be beneficial.

 

Another notable finding from this study was that although the average number of pharmacist visits per patient was significantly different, this difference of 1 visit did not result in a statistically significant improvement in HbA_1c. In fact, the average number of pharmacist visits per patient seemed to be within the reported range by Choe et al in a similar setting.⁷ Conversely, patients with a history of a gap in pharmacist visits spent longer durations under pharmacist care compared to those who had continuous follow-up. This could mean that it may take longer times or 1 additional visit to achieve similar HbA_1c results with continuous pharmacist care. Higher number of visits with pharmacists in the group with the history of gaps between pharmacist visits could have been facilitated by resident physicians, as both groups had a similar number of visits with them. Although this is not conclusive, identifying the optimal number of visits with pharmacists in this underserved population could be beneficial in strategizing pharmacist visits. Acute care utilization was not different between the 2 groups, and most cases that led to acute care utilization were not directly related to diabetes or its complications.

The average HbA_1c at the end of the study did not measure < 8%, a target that was reached by less than half of patients from each group; however, this study is a snapshot of a series of ongoing clinical pharmacy services. About 25% of our patients started their first visit with a pharmacist less than 6 months from the study end date, and these patients may not have had enough time with pharmacists for their HbA_1c to reach below the target goal. In addition, most patients in this clinic were enrolled in public health plans and may carry a significant burden of social and behavioral factors that can affect diabetes management.^18,19 These patients may need longer care by pharmacists along with other integrated services, such as behavioral health and social work, to achieve optimal HbA_1c levels.²⁰

There are several limitations to this study, including the lack of a propensity matched control group of patients who only had resident physician visits; thus, it is hard to test the true impact of continuous or intermittent pharmacist visits on the therapeutic outcomes. The study also does not address potential social, economic, and physical environment factors that might have contributed to pharmacist visits and to overall diabetes care. These factors can negatively impact diabetes control and addressing them could help with an individualized diabetes management approach.^17,18 Additionally, by nature of being a descriptive study, the results may be subject to undetermined confounding factors.

Conclusion

Patients maintaining continuous pharmacist visits do not have statistically significant differences in change in HbA_1c compared to patients who had a history of 3-month or longer gaps in pharmacist visits at a resident physician primary care safety-net clinic. However, patients with diabetes will likely derive a benefit in HbA_1c reduction regardless of regularity of pharmacist care. This finding still holds true in collaboration with resident physicians who also regularly meet with patients.

The study highlights that it is important to integrate clinical pharmacists into primary care teams for improved therapeutic outcomes. It is our hope that regular visits to pharmacists can be a gateway for behavioral health and social work referrals, thereby addressing pharmacist-identified social barriers. Furthermore, exploration of socioeconomic and behavioral barriers to pharmacist visits is necessary to address and improve the patient experience, health care delivery, and health outcomes.

Acknowledgments: The authors thank Roxanna Perez, PharmD, Amy Li, and Julie Dopheide, PharmD, BCPP, FASHP for their contributions to this project.

Corresponding author: Michelle Koun Lee Chu, PharmD, BCACP, APh, Titus Family Department of Clinical Pharmacy, School of Pharmacy, University of Southern California, 1985 Zonal Ave, Los Angeles, CA 90089-9121; lee118@usc.edu.

Financial disclosures: None.