Original Research

The American Diabetes Association recommends that patients on intensive insulin regimens self-monitor blood glucose (SMBG) to assist in therapy optimization.¹ To be useful, SMBG data must be captured by patients, shared with care teams, and used and interpreted by patients and practitioners.^2,3 Communication of SMBG data from the patient to practitioner can be challenging. Although technology can help in this process, limitations exist, such as manual data entry into systems, patient and/or practitioner technological challenges (eg, accessing interface), and compatibility and integration between SMBG devices and electronic health record (EHR) systems.⁴

The Boise Veterans Affairs Medical Center (BVAMC) in Idaho serves more than 100,000 veterans. It includes a main site, community-based outpatient clinics, and a clinical resource hub that provides telehealth services to veterans residing in rural neighboring states. The BVAMC pharmacy department provides both inpatient and outpatient services. At the BVAMC, clinical pharmacist practitioners (CPPs) are independent practitioners who support their care teams in comprehensive medication management and have the ability to initiate, modify, and discontinue drug therapy for referred patients.⁵ A prominent role of CPPs in primary care teams is to manage patients with uncontrolled diabetes and intensive insulin regimens in which SMBG data are vital to therapy optimization. As collecting SMBG data from patients is seen anecdotally as time intensive, we determined the mean time spent by CPPs collecting patient SMBG data and its potential implications.

Methods

Pharmacists at BVAMC were asked to estimate and record the following: SMBG data collection method, time spent collecting data, extra time spent documenting or formatting SMBG readings, total patient visit time, and visit type. Time was collected in minutes. Extra time spent documenting or formatting SMBG readings included any additional time formatting or entering data in the clinical note after talking to the patient; if this was done while multitasking and talking to the patient, it was not considered extra time. For total patient visit time, pharmacists were asked to estimate only time spent discussing diabetes care and collecting SMBG data. Visit types were categorized as in-person/face-to-face, telephone, and telehealth using clinical video telehealth (CVT)/VA Video Connect (VVC). Data were collected using a standardized spreadsheet. The spreadsheet was pilot tested by a CPP before distribution to all pharmacists.

CPPs were educated about the project in March 2021 and were asked to record data for a 1-week period between April 5, 2021, and April 30, 2021. One CPP also provided delayed data collected from May 17 to 21, 2021, and these data were included in our analysis.

Descriptive statistics were used to determine the mean time spent by CPPs collecting SMBG data. Unpaired t tests were used to compare time spent collecting SMBG data by different collection methods and patient visit types. A P value of ≤ .05 was considered statistically significant. Data were organized in Microsoft Excel, and statistics were completed with JMP Pro v15.

Results

Eight CPPs provided data from 120 patient encounters. For all patient encounter types, the mean time spent collecting SMBG data was 3.3 minutes, and completing additional documentation/formatting was 1.3 minutes (Table 1). Total mean time for SMBG collection and documentation was 4.6 minutes in visits that had a mean length of 20.1 minutes. Twenty-three percent of the visit was devoted to SMBG data, 16% for data collection, and 6% for documentation. In 23 encounters, at least half the time was spent collecting and documenting/formatting data.

When compared by the SMBG collection method, the longest time spent collecting SMBG data was with patient report (3.7 minutes), and the longest time spent documenting/formatting time was with meter download/home telehealth (2 minutes). There was no statistically significant difference in the time to collect SMBG data between patient report and other methods (3.7 minutes vs 2.8 minutes; P = .07).

When compared by visit type, there was not a statistically significant difference between time spent collecting in person vs telephone or video SMBG data (3.8 minutes vs 3.2 minutes; P = .39) (Table 2). The most common SMBG collection method for in-person/face-to-face visits was continuous glucose monitor (CGM) (n = 10), followed by meter download/home telehealth (n = 5), patient report (n = 3), and directly from log/meter (n = 1). For telephone or video visits, the most common collection method was patient report (n = 72), followed by directly from log/meter (n = 18), CGM (n = 5), meter download/home telehealth (n = 4), and secure message (n = 2).

 

Discussion

We found that the mean amount of time spent collecting and documenting/formatting SMBG data was only 4.6 minutes; however, this still represented a substantial portion of visit time. For telephone and CVT/VVC appointments, this represented > 25% of total visit time. While CPPs make important contributions to interprofessional team management of patients with diabetes, their cost is not trivial.^6-8 It is worth exploring the most effective and efficient ways to use CPPs. Our results indicate that streamlining SMBG data collection may be beneficial.

Pharmacy technicians, licensed practical nurses/clinical associates, registered nurses/nurse care managers, or other team members could help improve SMBG data collection. Using other team members is also an opportunity for comanagement, for team collaboration, and for more patients to be seen. For example, if a CPP currently has 12 patient encounters that last 20 minutes each, this results in about 240 minutes of direct patient care. If patient encounters were 16 minutes, CPPs could have 15 patient encounters in 240 minutes. Saved time could be used for other clinical tasks involved in disease management or clinical reminder reviews. While there are benefits to CPPs collecting SMBG data, such as further inquiry about patient-reported values, other team members could be trained to ask appropriate follow-up questions for abnormal blood glucose readings. In addition, leveraging current team members and optimizing their roles could prevent the need to acquire additional full-time equivalent employees.

Another opportunity to increase efficiency in SMBG data collection is with SMBG devices and EHR integration.^4,9 However, integration can be difficult with different types of SMBG devices and EHR platforms. Education for patients and practitioners could help to ensure accurate and reliable data uploads; patient internet availability; data protection, privacy, and sharing; workflow management; and clear patient-practitioner expectations.¹⁰ For example, if patient SMBG data are automatically uploaded to practitioners, patients’ expectations for practitioner review of data and follow-up need to be determined.

We found a subset of 23 patient encounters where data collection and documenting/formatting represented more than half of the total visit time. In this subset, 13 SMBG reports were pulled from a log or meter, 8 were patient reported, and 3 were meter download or home telehealth.

Limitations

A potential reason for the lack of statistically significant differences in SMBG collection method or visit type in this study includes the small sample size. Participation in this work was voluntary, and all participating CPPs had ≥ 3 years of practice in their current setting, which includes a heavy workload of diabetes management. These pharmacists noted self-established procedures/systems for SMBG data collection, including the use of Excel spreadsheets with pregenerated formulas. For less experienced CPPs, SMBG data collection time may be even longer. Pharmacists also noted that they may limit time spent collecting SMBG data depending on the patient encounter and whether they have gathered sufficient data to guide clinical care. Other limitations of this work include data collection from a single institution and that the time documented represented estimates; there was no external monitor.

Conclusions

In this analysis, we found that CPPs spend about 3 minutes collecting SMBG data from patients and about an additional 1 minute documenting and formatting data. While 4 to 5 minutes may not represent a substantial amount of time for 1 patient, it can be when multiplied by several patient encounters. The time spent collecting SMBG data did not significantly differ by collection method or visit type. Opportunities to increase efficiency in SMBG data collection, such as the use of nonpharmacist team members, are worth exploring.

Acknowledgments

Thank you to the pharmacists at the Boise Veterans Affairs Medical Center for their time and support of this work: Danielle Ahlstrom, Paul Black, Robyn Cruz, Sarah Naidoo, Anthony Nelson, Laura Spoutz, Eileen Twomey, Donovan Victorine, and Michelle Wilkin.

The American Diabetes Association recommends that patients on intensive insulin regimens self-monitor blood glucose (SMBG) to assist in therapy optimization.¹ To be useful, SMBG data must be captured by patients, shared with care teams, and used and interpreted by patients and practitioners.^2,3 Communication of SMBG data from the patient to practitioner can be challenging. Although technology can help in this process, limitations exist, such as manual data entry into systems, patient and/or practitioner technological challenges (eg, accessing interface), and compatibility and integration between SMBG devices and electronic health record (EHR) systems.⁴

The Boise Veterans Affairs Medical Center (BVAMC) in Idaho serves more than 100,000 veterans. It includes a main site, community-based outpatient clinics, and a clinical resource hub that provides telehealth services to veterans residing in rural neighboring states. The BVAMC pharmacy department provides both inpatient and outpatient services. At the BVAMC, clinical pharmacist practitioners (CPPs) are independent practitioners who support their care teams in comprehensive medication management and have the ability to initiate, modify, and discontinue drug therapy for referred patients.⁵ A prominent role of CPPs in primary care teams is to manage patients with uncontrolled diabetes and intensive insulin regimens in which SMBG data are vital to therapy optimization. As collecting SMBG data from patients is seen anecdotally as time intensive, we determined the mean time spent by CPPs collecting patient SMBG data and its potential implications.

Methods

Pharmacists at BVAMC were asked to estimate and record the following: SMBG data collection method, time spent collecting data, extra time spent documenting or formatting SMBG readings, total patient visit time, and visit type. Time was collected in minutes. Extra time spent documenting or formatting SMBG readings included any additional time formatting or entering data in the clinical note after talking to the patient; if this was done while multitasking and talking to the patient, it was not considered extra time. For total patient visit time, pharmacists were asked to estimate only time spent discussing diabetes care and collecting SMBG data. Visit types were categorized as in-person/face-to-face, telephone, and telehealth using clinical video telehealth (CVT)/VA Video Connect (VVC). Data were collected using a standardized spreadsheet. The spreadsheet was pilot tested by a CPP before distribution to all pharmacists.

CPPs were educated about the project in March 2021 and were asked to record data for a 1-week period between April 5, 2021, and April 30, 2021. One CPP also provided delayed data collected from May 17 to 21, 2021, and these data were included in our analysis.

Descriptive statistics were used to determine the mean time spent by CPPs collecting SMBG data. Unpaired t tests were used to compare time spent collecting SMBG data by different collection methods and patient visit types. A P value of ≤ .05 was considered statistically significant. Data were organized in Microsoft Excel, and statistics were completed with JMP Pro v15.

Results

Eight CPPs provided data from 120 patient encounters. For all patient encounter types, the mean time spent collecting SMBG data was 3.3 minutes, and completing additional documentation/formatting was 1.3 minutes (Table 1). Total mean time for SMBG collection and documentation was 4.6 minutes in visits that had a mean length of 20.1 minutes. Twenty-three percent of the visit was devoted to SMBG data, 16% for data collection, and 6% for documentation. In 23 encounters, at least half the time was spent collecting and documenting/formatting data.

When compared by the SMBG collection method, the longest time spent collecting SMBG data was with patient report (3.7 minutes), and the longest time spent documenting/formatting time was with meter download/home telehealth (2 minutes). There was no statistically significant difference in the time to collect SMBG data between patient report and other methods (3.7 minutes vs 2.8 minutes; P = .07).

When compared by visit type, there was not a statistically significant difference between time spent collecting in person vs telephone or video SMBG data (3.8 minutes vs 3.2 minutes; P = .39) (Table 2). The most common SMBG collection method for in-person/face-to-face visits was continuous glucose monitor (CGM) (n = 10), followed by meter download/home telehealth (n = 5), patient report (n = 3), and directly from log/meter (n = 1). For telephone or video visits, the most common collection method was patient report (n = 72), followed by directly from log/meter (n = 18), CGM (n = 5), meter download/home telehealth (n = 4), and secure message (n = 2).

 

Discussion

We found that the mean amount of time spent collecting and documenting/formatting SMBG data was only 4.6 minutes; however, this still represented a substantial portion of visit time. For telephone and CVT/VVC appointments, this represented > 25% of total visit time. While CPPs make important contributions to interprofessional team management of patients with diabetes, their cost is not trivial.^6-8 It is worth exploring the most effective and efficient ways to use CPPs. Our results indicate that streamlining SMBG data collection may be beneficial.

Pharmacy technicians, licensed practical nurses/clinical associates, registered nurses/nurse care managers, or other team members could help improve SMBG data collection. Using other team members is also an opportunity for comanagement, for team collaboration, and for more patients to be seen. For example, if a CPP currently has 12 patient encounters that last 20 minutes each, this results in about 240 minutes of direct patient care. If patient encounters were 16 minutes, CPPs could have 15 patient encounters in 240 minutes. Saved time could be used for other clinical tasks involved in disease management or clinical reminder reviews. While there are benefits to CPPs collecting SMBG data, such as further inquiry about patient-reported values, other team members could be trained to ask appropriate follow-up questions for abnormal blood glucose readings. In addition, leveraging current team members and optimizing their roles could prevent the need to acquire additional full-time equivalent employees.

Another opportunity to increase efficiency in SMBG data collection is with SMBG devices and EHR integration.^4,9 However, integration can be difficult with different types of SMBG devices and EHR platforms. Education for patients and practitioners could help to ensure accurate and reliable data uploads; patient internet availability; data protection, privacy, and sharing; workflow management; and clear patient-practitioner expectations.¹⁰ For example, if patient SMBG data are automatically uploaded to practitioners, patients’ expectations for practitioner review of data and follow-up need to be determined.

We found a subset of 23 patient encounters where data collection and documenting/formatting represented more than half of the total visit time. In this subset, 13 SMBG reports were pulled from a log or meter, 8 were patient reported, and 3 were meter download or home telehealth.

Limitations

A potential reason for the lack of statistically significant differences in SMBG collection method or visit type in this study includes the small sample size. Participation in this work was voluntary, and all participating CPPs had ≥ 3 years of practice in their current setting, which includes a heavy workload of diabetes management. These pharmacists noted self-established procedures/systems for SMBG data collection, including the use of Excel spreadsheets with pregenerated formulas. For less experienced CPPs, SMBG data collection time may be even longer. Pharmacists also noted that they may limit time spent collecting SMBG data depending on the patient encounter and whether they have gathered sufficient data to guide clinical care. Other limitations of this work include data collection from a single institution and that the time documented represented estimates; there was no external monitor.

Conclusions

In this analysis, we found that CPPs spend about 3 minutes collecting SMBG data from patients and about an additional 1 minute documenting and formatting data. While 4 to 5 minutes may not represent a substantial amount of time for 1 patient, it can be when multiplied by several patient encounters. The time spent collecting SMBG data did not significantly differ by collection method or visit type. Opportunities to increase efficiency in SMBG data collection, such as the use of nonpharmacist team members, are worth exploring.

Acknowledgments

Thank you to the pharmacists at the Boise Veterans Affairs Medical Center for their time and support of this work: Danielle Ahlstrom, Paul Black, Robyn Cruz, Sarah Naidoo, Anthony Nelson, Laura Spoutz, Eileen Twomey, Donovan Victorine, and Michelle Wilkin.

The American Diabetes Association recommends that patients on intensive insulin regimens self-monitor blood glucose (SMBG) to assist in therapy optimization.¹ To be useful, SMBG data must be captured by patients, shared with care teams, and used and interpreted by patients and practitioners.^2,3 Communication of SMBG data from the patient to practitioner can be challenging. Although technology can help in this process, limitations exist, such as manual data entry into systems, patient and/or practitioner technological challenges (eg, accessing interface), and compatibility and integration between SMBG devices and electronic health record (EHR) systems.⁴

The Boise Veterans Affairs Medical Center (BVAMC) in Idaho serves more than 100,000 veterans. It includes a main site, community-based outpatient clinics, and a clinical resource hub that provides telehealth services to veterans residing in rural neighboring states. The BVAMC pharmacy department provides both inpatient and outpatient services. At the BVAMC, clinical pharmacist practitioners (CPPs) are independent practitioners who support their care teams in comprehensive medication management and have the ability to initiate, modify, and discontinue drug therapy for referred patients.⁵ A prominent role of CPPs in primary care teams is to manage patients with uncontrolled diabetes and intensive insulin regimens in which SMBG data are vital to therapy optimization. As collecting SMBG data from patients is seen anecdotally as time intensive, we determined the mean time spent by CPPs collecting patient SMBG data and its potential implications.

Methods

Pharmacists at BVAMC were asked to estimate and record the following: SMBG data collection method, time spent collecting data, extra time spent documenting or formatting SMBG readings, total patient visit time, and visit type. Time was collected in minutes. Extra time spent documenting or formatting SMBG readings included any additional time formatting or entering data in the clinical note after talking to the patient; if this was done while multitasking and talking to the patient, it was not considered extra time. For total patient visit time, pharmacists were asked to estimate only time spent discussing diabetes care and collecting SMBG data. Visit types were categorized as in-person/face-to-face, telephone, and telehealth using clinical video telehealth (CVT)/VA Video Connect (VVC). Data were collected using a standardized spreadsheet. The spreadsheet was pilot tested by a CPP before distribution to all pharmacists.

CPPs were educated about the project in March 2021 and were asked to record data for a 1-week period between April 5, 2021, and April 30, 2021. One CPP also provided delayed data collected from May 17 to 21, 2021, and these data were included in our analysis.

Descriptive statistics were used to determine the mean time spent by CPPs collecting SMBG data. Unpaired t tests were used to compare time spent collecting SMBG data by different collection methods and patient visit types. A P value of ≤ .05 was considered statistically significant. Data were organized in Microsoft Excel, and statistics were completed with JMP Pro v15.

Results

Eight CPPs provided data from 120 patient encounters. For all patient encounter types, the mean time spent collecting SMBG data was 3.3 minutes, and completing additional documentation/formatting was 1.3 minutes (Table 1). Total mean time for SMBG collection and documentation was 4.6 minutes in visits that had a mean length of 20.1 minutes. Twenty-three percent of the visit was devoted to SMBG data, 16% for data collection, and 6% for documentation. In 23 encounters, at least half the time was spent collecting and documenting/formatting data.

When compared by the SMBG collection method, the longest time spent collecting SMBG data was with patient report (3.7 minutes), and the longest time spent documenting/formatting time was with meter download/home telehealth (2 minutes). There was no statistically significant difference in the time to collect SMBG data between patient report and other methods (3.7 minutes vs 2.8 minutes; P = .07).

When compared by visit type, there was not a statistically significant difference between time spent collecting in person vs telephone or video SMBG data (3.8 minutes vs 3.2 minutes; P = .39) (Table 2). The most common SMBG collection method for in-person/face-to-face visits was continuous glucose monitor (CGM) (n = 10), followed by meter download/home telehealth (n = 5), patient report (n = 3), and directly from log/meter (n = 1). For telephone or video visits, the most common collection method was patient report (n = 72), followed by directly from log/meter (n = 18), CGM (n = 5), meter download/home telehealth (n = 4), and secure message (n = 2).

 

Discussion

We found that the mean amount of time spent collecting and documenting/formatting SMBG data was only 4.6 minutes; however, this still represented a substantial portion of visit time. For telephone and CVT/VVC appointments, this represented > 25% of total visit time. While CPPs make important contributions to interprofessional team management of patients with diabetes, their cost is not trivial.^6-8 It is worth exploring the most effective and efficient ways to use CPPs. Our results indicate that streamlining SMBG data collection may be beneficial.

Pharmacy technicians, licensed practical nurses/clinical associates, registered nurses/nurse care managers, or other team members could help improve SMBG data collection. Using other team members is also an opportunity for comanagement, for team collaboration, and for more patients to be seen. For example, if a CPP currently has 12 patient encounters that last 20 minutes each, this results in about 240 minutes of direct patient care. If patient encounters were 16 minutes, CPPs could have 15 patient encounters in 240 minutes. Saved time could be used for other clinical tasks involved in disease management or clinical reminder reviews. While there are benefits to CPPs collecting SMBG data, such as further inquiry about patient-reported values, other team members could be trained to ask appropriate follow-up questions for abnormal blood glucose readings. In addition, leveraging current team members and optimizing their roles could prevent the need to acquire additional full-time equivalent employees.

Another opportunity to increase efficiency in SMBG data collection is with SMBG devices and EHR integration.^4,9 However, integration can be difficult with different types of SMBG devices and EHR platforms. Education for patients and practitioners could help to ensure accurate and reliable data uploads; patient internet availability; data protection, privacy, and sharing; workflow management; and clear patient-practitioner expectations.¹⁰ For example, if patient SMBG data are automatically uploaded to practitioners, patients’ expectations for practitioner review of data and follow-up need to be determined.

We found a subset of 23 patient encounters where data collection and documenting/formatting represented more than half of the total visit time. In this subset, 13 SMBG reports were pulled from a log or meter, 8 were patient reported, and 3 were meter download or home telehealth.

Limitations

A potential reason for the lack of statistically significant differences in SMBG collection method or visit type in this study includes the small sample size. Participation in this work was voluntary, and all participating CPPs had ≥ 3 years of practice in their current setting, which includes a heavy workload of diabetes management. These pharmacists noted self-established procedures/systems for SMBG data collection, including the use of Excel spreadsheets with pregenerated formulas. For less experienced CPPs, SMBG data collection time may be even longer. Pharmacists also noted that they may limit time spent collecting SMBG data depending on the patient encounter and whether they have gathered sufficient data to guide clinical care. Other limitations of this work include data collection from a single institution and that the time documented represented estimates; there was no external monitor.

Conclusions

In this analysis, we found that CPPs spend about 3 minutes collecting SMBG data from patients and about an additional 1 minute documenting and formatting data. While 4 to 5 minutes may not represent a substantial amount of time for 1 patient, it can be when multiplied by several patient encounters. The time spent collecting SMBG data did not significantly differ by collection method or visit type. Opportunities to increase efficiency in SMBG data collection, such as the use of nonpharmacist team members, are worth exploring.

Acknowledgments

Thank you to the pharmacists at the Boise Veterans Affairs Medical Center for their time and support of this work: Danielle Ahlstrom, Paul Black, Robyn Cruz, Sarah Naidoo, Anthony Nelson, Laura Spoutz, Eileen Twomey, Donovan Victorine, and Michelle Wilkin.

Heart failure (HF) is a chronic, progressive condition that is characterized by the heart’s inability to effectively pump blood throughout the body. In 2018, approximately 6.2 million US adults had HF, and 13.4% of all death certificates noted HF as a precipitating factor.¹ Patients not receiving appropriate guideline-directed medical therapy (GDMT) face a 29% excess mortality risk over a 2-year period.² Each additional GDMT included in a patient’s regimen significantly reduces all-cause mortality.³

The Change the Management of Patients with Heart Failure (CHAMP) registry reports that only about 1% of patients with HF are prescribed 3 agents from contemporary GDMT at target doses, highlighting the need for optimizing clinicians’ approaches to GDMT.⁴ Similarly, The Get With The Guidelines Heart Failure Registry has noted that only 20.2% of patients with HF with reduced ejection fraction (HFrEF) are prescribed a sodium-glucose cotransporter 2 inhibitor (SGLT2i) following hospital discharge for HFrEF exacerbation.⁵ Overall, treatment rates with GDMT saw limited improvement between 2013 and 2019, with no significant difference between groups in mortality, indicating the need for optimized methods to encourage the initiation of GDMT.⁶

Remote monitoring and telecare are novel ways to improve GDMT rates in those with HFrEF. However, data are inconsistent regarding the impact of remote HF monitoring and improvements in GDMT or HF-related outcomes.^6-10 The modalities of remote monitoring for GDMT vary among studies, but the potential for telehealth monitoring to improve GDMT, thereby potentially reducing HF-related hospitalizations, is clear.

Telemonitoring has demonstrated improved participant adherence with weight monitoring, although the withdrawal rate was high, and has the potential to reduce all-cause mortality and HF-related hospitalizations.^11,12 Telemonitoring for GDMT optimization led to an increase in the proportion of patients who achieved optimal GDMT doses, a decrease in the time to dose optimization, and a reduction in the number of clinic visits.¹³ Remote GDMT titration was accomplished in the general patient population with HFrEF; however, in populations already followed by cardiologists or HF specialists, remote optimization strategies did not yield different proportions of GDMT use.¹⁴ The aim of this study was to assess the impact of the home telehealth (HT) monitoring program on the initiation and optimization of HF GDMT among veterans with HFrEF at the Veterans Affairs Ann Arbor Healthcare System (VAAAHS) in Michigan.

 

Methods

This was a single-center retrospective study of Computerized Patient Record System (CPRS)data. Patients at the VAAAHS were evaluated if they were diagnosed with HFrEF and were eligible for enrollment in the HT monitoring program. Eligibility criteria included a diagnosis of stage C HF, irrespective of EF, and a history of any HF-related hospitalization. We focused on patients with HFrEF due to stronger guideline-based recommendations for certain pharmacotherapies as compared with HF with mildly reduced ejection fraction (HFmrEF) and preserved ejection fraction (HFpEF). Initial patient data for HT enrolling were accessed using the Heart Failure Dashboard via the US Department of Veterans Affairs (VA) Academic Detailing Service. The target daily doses of typical agents used in HFrEF GDMT are listed in the Appendix.

The HT program is an embedded model in which HT nurses receive remote data from the patient and triage that with the VAAAHS cardiology team. Patients’ questions, concerns, and/or vital signs are recovered remotely. In this model, nurses are embedded in the cardiology team, working with the cardiologists, cardiology clinical pharmacist, and/or cardiology nurse practitioners to make medication interventions. Data are recorded with an HT device, including weight, blood pressure (BP), heart rate, and pulse oximetry. HT nurses are also available to the patient via phone or video. The program uses a 180-day disease management protocol for HF via remote device, enabling the patient to answer questions and receive education on their disease daily. Responses to questions and data are then reviewed by an HT nurse remotely during business hours and triaged as appropriate with the cardiology team. Data can be communicated to the cardiology team via the patient record, eliminating the need for the cardiology team to use the proprietary portal affiliated with the HT device.

Study Sample

Patient information was obtained from a list of 417 patients eligible for enrollment in the HT program; the list was sent to the HT program for review and enrollment. Patient data were extracted from the VAAAHS HF Dashboard and included all patients with HFrEF and available data on the platform. The sample for the retrospective chart review included 40 adults who had HFrEF, defined as a left ventricular EF (LVEF) of ≤ 40% as evidenced by a transthoracic echocardiogram or cardiac magnetic resonance imaging. These patients were contacted and agreed to enroll in the HT monitoring program. The HT program population was compared against a control group of 33 patients who were ineligible for the HT program. Patients were deemed ineligible for HT if they resided in a nursing home, lacked a VAAAHS primary care clinician, or declined participation in the HT program.

Procedures

Patients enrolled in the HT program were assigned a nurse who monitored and responded to changes in vital signs, such as an increase in weight of ≥ 3 lb in 24 hours or ≥ 5 lb in 7 days. Each participant in HT received a home BP machine to check BP and heart rates and a scale for daily weights. The patient also responded to questions on a tablet device with the BP machine and scale. The vital signs were put into the tablet device, and the patient answered questions about their symptoms for that day. If the patient endorsed any symptoms of HF, such as pedal edema, abdominal distention, orthopnea, paroxysmal nocturnal dyspnea, dyspnea on exertion, or orthopnea, they received a call from the nurse to discuss symptoms and make appropriate adjustments in the diuretic or HF regimen. The HF-trained pharmacist served as a resource for drug therapy management and provided recommendations for diuretic adjustments on an as-needed basis.

Patients who declined participation in the HT program followed the standard of care, which was limited to visits with primary care clinicians and/or cardiologists as per the follow-up plan. Patient data were collected over 12 months. The study was approved by the VAAAHS Institutional Review Board (reference number, 1703034), Research and Development Committee, and Research Administration.

 

 

Primary and Secondary Goals

The primary goal of the study was to assess the impact of the HT program on drug interventions, specifically initiating and titrating HFrEF pharmacotherapies. Interventions were based on GDMT with known mortality- and morbidity-reducing properties when used at their maximum tolerated doses, including angiotensin-converting enzyme inhibitors (ACEi), angiotensin receptor-neprilysin inhibitor (ARNi), or angiotensin receptor blockers (ARB), with a preference for ARNi, β-blockers for HFrEF (metoprolol succinate, bisoprolol, or carvedilol), aldosterone antagonists, and SGLT2is.

Secondary goals included HF-related hospitalizations, medication adherence, time to enrollment in HT, time to laboratory analysis after the initiation or titration of an ACEi/ARB/ARNi or aldosterone antagonist, and time enrolled in the HT program. Patients were considered adherent if their drug refill history showed consistent fills of their medications. The χ² test was used for total interventions made during the study period and Fisher exact test for all others.

Results

Patient data were collected between July 2022 and June 2023. All 73 patients were male, and the mean age in the HT group (n = 40) was 72.6 years and 75.2 years for the control group (n = 33). Overall, the baseline demographics were similar between the groups (Table 1). Of 40 patients screened for enrollment in the HT program, 33 were included in the analysis (Figure 1).

At baseline, the HT group included more individuals than the control group on ACEi/ARB/ARNi (24 vs 19, respectively), β-blocker (28 vs 24, respectively), SGLT2i (14 vs 11, respectively), and aldosterone antagonist (15 vs 9, respectively) (Figure 2). There were 20 interventions made in the HT group compared with 11 therapy changes in the control arm during the study (odds ratio, 1.43; P = .23) (Table 2). In the HT group, 1 patient achieved an ACEi target dose, 3 patients achieved a β-blocker target dose, and 7 achieved a target dose of spironolactone (titration is not required for SGLT2i therapy and is counted as target dose). In the HT group, 17 patients were on ≥ 3 recommended agents, while 9 patients were taking 4 agents. Seven of 20 HT group interventions resulted in titration to the target dose. In the control group, no patients achieved an ARNi target dose, 3 patients achieved a β-blocker target dose, and 2 patients achieved a spironolactone antagonist target dose. In the control arm, 7 patients were on ≥ 3 GDMTs, and 2 were taking 4 agents. No patient in either group achieved a target dose of 4 agents. Five of 11 control group interventions resulted in initiation or titration of GDMT to the target dose.

One HF-related hospitalization occurred in the HT group and 6 in the control arm (odds ratio, 0.12; P = .04). Most patients (90%) were adherent to their GDMT regimen, 3 patients not on any GDMT for HFrEF at the end of the review were excluded, and 1 patient was deemed not adherent to medication therapy (Table 3). The mean (SD) time-to-enrollment in the HT program was 5.5 (6.4) days after consultation for 34 patients who were assessed. Ten patients had laboratory tests within a mean (SD) of 17 (9.8) days after the intervention, 29 patients did not require routine laboratory monitoring, and 1 did not follow through.

Of the 40 HT group patients, 7 were excluded from analysis (3 failed to schedule HT, 1 was at a long-term care facility, 1 was nonadherent, 1 declined participation, and 1 died) and 33 remained in the program for a mean (SD) 5.3 (3.5) months. Death rates were tracked during the study: 1 patient died in the HT group and 3 in the control group.

We analyzed the overall percentage of VAAAHS patients with HFrEF who were on appropriate GDMT. Given the ongoing drive to improve HF-related outcomes, HT interventions could not be compared to a static population, so the HT and control patients were compared with the rates of GDMT at VAAAHS, which was available in the Academic Detailing Service Heart Failure Dashboard (Figure 3). Titration and optimization rates were also compared (Figure 4). From July 2022 to June 2023, ARNi use increased by 16.6%, aldosterone antagonist by 6.8%, and β-blockers by 2.4%. Target doses of GDMTs were more difficult to achieve in the hospital system. There was an increase in aldosterone antagonist target dose achievement by 4.7%, but overall there were decreases in target doses in other GDMTs: ACEi/ARB/ARNi target dose use decreased by 3.2%, ARNi target dose use decreased by 2.7% and target β-blocker use decreased by 0.9%.

 

Discussion

Telehealth yielded clinically important interventions, with some titrations bringing patients to their target doses of medications for HFrEF. The 20 interventions made in the HT group can be largely attributed to the nurses’ efforts to alert clinicians to drug titrations or ACEi/ARB to ARNi transitions. Although the findings were not statistically significant, the difference in the number of drug therapy changes supports the use of the HT program for a GDMT optimization strategy. Patients may be difficult to titrate secondary to adverse effects that make medication initiation or titration inappropriate, such as hypotension and hyperkalemia, although this was not observed in this small sample size. Considering a mean HT enrollment of 5.3 months, many patients had adequate disease assessment and medication titration. Given that patients are discharged from the service once deemed appropriate, this decreases the burden on the patient and increases the utility and implementation of the HT program for other patients.

A surprising finding of this study was the lower rate of HF-related hospitalizations in the HT group. Although not the primary subject of interest in the study, it reinforced the importance of close health care professional follow-up for patients living with HF. Telehealth may improve communication and shared decision making over medication use. Because the finding was unanticipated, the rate of diuretic adjustments was not tracked.

Patients were reevaluated every 6 months for willingness to continue the program, adherence, and clinical needs. These results are similar to those of other trials that demonstrated improved rates of GDMT in the setting of pharmacist- or nurse-led HF treatment optimization.^15,16 These positive results differ from other trials incorporating remote monitoring regarding patient continuation in HT programs. For example, in a study by Ding and colleagues, the withdrawal rate from their monitoring service was about 22%, while in our study only 1 patient withdrew from the HT program.¹¹

The HT program resulted in fewer hospitalizations than the control arm. There were 6 HF-related hospitalizations in the control group, although 5 involved a single patient. Typically, such a patient would be encouraged to follow HT monitoring after just 1 HF-related hospitalization; however, the patient declined to participate.

Early optimization of GDMT in patients who were recently discharged from the hospital for an HF-related hospitalization yields a reduction in hospital rehospitalization.¹⁷ GDMT optimization has unequivocal benefits in HF outcomes. Unfortunately, the issues surrounding methodologies on how to best optimize GDMT are lacking. While HT has been found to be feasible in the aid of optimizing medical therapy, the TIM-HF trial concluded that remote monitoring services had no significant benefit in reducing mortality.^7,8 On the other hand, the OptiLink HF Study showed that when clinicians respond to remote monitoring prompts from fluid index threshold crossing alerts, these interventions are associated with significantly improved clinical outcomes in patients with implantable cardioverter-defibrillators and advanced HF.⁹ In contrast to previous trials, the AMULET trial showed that remote monitoring compared with standard care significantly reduced the risk of HF hospitalization or cardiovascular death during the 12-month follow-up among patients with HF and LVEF ≤ 49% after an episode of acute exacerbation.¹⁰ Additionally, patients who received skilled home health services and participated in remote monitoring for their chronic HF experienced a reduction in all-cause 30-day readmission.¹⁸

Given the contrasting evidence regarding remote monitoring and variable modalities of implementing interventions, we investigated whether HT monitoring yields improvements in GDMT optimization. We found that HT nurses were able to nearly double the rate of interventions for patients with HFrEF. The HT program in providing expanded services will require adequate staffing responsibilities and support. The HT program is geared toward following a large, diverse patient population, such as those with chronic obstructive pulmonary disease, hypertension, and HF. We only evaluated services for patients with HFrEF, but the program also follows patients with HfmrEF and HfpEF. These patients were not included as GDMT optimization differs for patients with an LVEF > 40%.^19,20

The lower rates of achieving target doses of GDMTs were likely obstructed by continuous use of initial drug doses and further limited by lack of follow-up. When compared with the rest of the VAAAHS, there was a greater effort to increase ARNi use in the HT group as 7 of 33 patients (21%) were started on ARNi compared with a background increase of ARNi use of 17%. There was a lower mortality rate observed in the HT group compared with the control group. One patient in each group died of unrelated causes, while 2 deaths in the control group were due to worsening HF. The difference in mortality is likely multifactorial, possibly related to the control group’s greater disease burden or higher mean age (75.2 years vs 72.6 years).

Limitations

This was an observational cohort design, which is subject to bias. Thus, the findings of this study are entirely hypothesis-generating and a randomized controlled trial would be necessary for clearer results. Second, low numbers of participants may have skewed the data set. Given the observational nature of the study, this nonetheless is a positive finding to support the HT program for assisting with HF monitoring and prompting drug interventions. Due to the low number of participants, a single patient may have skewed the results with 5 hospitalizations.

Conclusions

This pilot study demonstrates the applicability of HT monitoring to optimize veteran HFrEF GDMT. The HT program yielded numerically relevant interventions and fewer HF-related hospitalizations compared with the control arm. The study shows the feasibility of the program to safely optimize GDMT toward their target doses and may serve as an additional catalyst to further develop HT programs specifically targeted toward HF monitoring and management. Cost-savings analyses would likely need to demonstrate the cost utility of such a service.

Acknowledgments

We thank the home telehealth nursing staff for their assistance in data collection and enrollment of patients into the monitoring program.

Heart failure (HF) is a chronic, progressive condition that is characterized by the heart’s inability to effectively pump blood throughout the body. In 2018, approximately 6.2 million US adults had HF, and 13.4% of all death certificates noted HF as a precipitating factor.¹ Patients not receiving appropriate guideline-directed medical therapy (GDMT) face a 29% excess mortality risk over a 2-year period.² Each additional GDMT included in a patient’s regimen significantly reduces all-cause mortality.³

The Change the Management of Patients with Heart Failure (CHAMP) registry reports that only about 1% of patients with HF are prescribed 3 agents from contemporary GDMT at target doses, highlighting the need for optimizing clinicians’ approaches to GDMT.⁴ Similarly, The Get With The Guidelines Heart Failure Registry has noted that only 20.2% of patients with HF with reduced ejection fraction (HFrEF) are prescribed a sodium-glucose cotransporter 2 inhibitor (SGLT2i) following hospital discharge for HFrEF exacerbation.⁵ Overall, treatment rates with GDMT saw limited improvement between 2013 and 2019, with no significant difference between groups in mortality, indicating the need for optimized methods to encourage the initiation of GDMT.⁶

Remote monitoring and telecare are novel ways to improve GDMT rates in those with HFrEF. However, data are inconsistent regarding the impact of remote HF monitoring and improvements in GDMT or HF-related outcomes.^6-10 The modalities of remote monitoring for GDMT vary among studies, but the potential for telehealth monitoring to improve GDMT, thereby potentially reducing HF-related hospitalizations, is clear.

Telemonitoring has demonstrated improved participant adherence with weight monitoring, although the withdrawal rate was high, and has the potential to reduce all-cause mortality and HF-related hospitalizations.^11,12 Telemonitoring for GDMT optimization led to an increase in the proportion of patients who achieved optimal GDMT doses, a decrease in the time to dose optimization, and a reduction in the number of clinic visits.¹³ Remote GDMT titration was accomplished in the general patient population with HFrEF; however, in populations already followed by cardiologists or HF specialists, remote optimization strategies did not yield different proportions of GDMT use.¹⁴ The aim of this study was to assess the impact of the home telehealth (HT) monitoring program on the initiation and optimization of HF GDMT among veterans with HFrEF at the Veterans Affairs Ann Arbor Healthcare System (VAAAHS) in Michigan.

 

Methods

This was a single-center retrospective study of Computerized Patient Record System (CPRS)data. Patients at the VAAAHS were evaluated if they were diagnosed with HFrEF and were eligible for enrollment in the HT monitoring program. Eligibility criteria included a diagnosis of stage C HF, irrespective of EF, and a history of any HF-related hospitalization. We focused on patients with HFrEF due to stronger guideline-based recommendations for certain pharmacotherapies as compared with HF with mildly reduced ejection fraction (HFmrEF) and preserved ejection fraction (HFpEF). Initial patient data for HT enrolling were accessed using the Heart Failure Dashboard via the US Department of Veterans Affairs (VA) Academic Detailing Service. The target daily doses of typical agents used in HFrEF GDMT are listed in the Appendix.

The HT program is an embedded model in which HT nurses receive remote data from the patient and triage that with the VAAAHS cardiology team. Patients’ questions, concerns, and/or vital signs are recovered remotely. In this model, nurses are embedded in the cardiology team, working with the cardiologists, cardiology clinical pharmacist, and/or cardiology nurse practitioners to make medication interventions. Data are recorded with an HT device, including weight, blood pressure (BP), heart rate, and pulse oximetry. HT nurses are also available to the patient via phone or video. The program uses a 180-day disease management protocol for HF via remote device, enabling the patient to answer questions and receive education on their disease daily. Responses to questions and data are then reviewed by an HT nurse remotely during business hours and triaged as appropriate with the cardiology team. Data can be communicated to the cardiology team via the patient record, eliminating the need for the cardiology team to use the proprietary portal affiliated with the HT device.

Study Sample

Patient information was obtained from a list of 417 patients eligible for enrollment in the HT program; the list was sent to the HT program for review and enrollment. Patient data were extracted from the VAAAHS HF Dashboard and included all patients with HFrEF and available data on the platform. The sample for the retrospective chart review included 40 adults who had HFrEF, defined as a left ventricular EF (LVEF) of ≤ 40% as evidenced by a transthoracic echocardiogram or cardiac magnetic resonance imaging. These patients were contacted and agreed to enroll in the HT monitoring program. The HT program population was compared against a control group of 33 patients who were ineligible for the HT program. Patients were deemed ineligible for HT if they resided in a nursing home, lacked a VAAAHS primary care clinician, or declined participation in the HT program.

Procedures

Patients enrolled in the HT program were assigned a nurse who monitored and responded to changes in vital signs, such as an increase in weight of ≥ 3 lb in 24 hours or ≥ 5 lb in 7 days. Each participant in HT received a home BP machine to check BP and heart rates and a scale for daily weights. The patient also responded to questions on a tablet device with the BP machine and scale. The vital signs were put into the tablet device, and the patient answered questions about their symptoms for that day. If the patient endorsed any symptoms of HF, such as pedal edema, abdominal distention, orthopnea, paroxysmal nocturnal dyspnea, dyspnea on exertion, or orthopnea, they received a call from the nurse to discuss symptoms and make appropriate adjustments in the diuretic or HF regimen. The HF-trained pharmacist served as a resource for drug therapy management and provided recommendations for diuretic adjustments on an as-needed basis.

Patients who declined participation in the HT program followed the standard of care, which was limited to visits with primary care clinicians and/or cardiologists as per the follow-up plan. Patient data were collected over 12 months. The study was approved by the VAAAHS Institutional Review Board (reference number, 1703034), Research and Development Committee, and Research Administration.

 

 

Primary and Secondary Goals

The primary goal of the study was to assess the impact of the HT program on drug interventions, specifically initiating and titrating HFrEF pharmacotherapies. Interventions were based on GDMT with known mortality- and morbidity-reducing properties when used at their maximum tolerated doses, including angiotensin-converting enzyme inhibitors (ACEi), angiotensin receptor-neprilysin inhibitor (ARNi), or angiotensin receptor blockers (ARB), with a preference for ARNi, β-blockers for HFrEF (metoprolol succinate, bisoprolol, or carvedilol), aldosterone antagonists, and SGLT2is.

Secondary goals included HF-related hospitalizations, medication adherence, time to enrollment in HT, time to laboratory analysis after the initiation or titration of an ACEi/ARB/ARNi or aldosterone antagonist, and time enrolled in the HT program. Patients were considered adherent if their drug refill history showed consistent fills of their medications. The χ² test was used for total interventions made during the study period and Fisher exact test for all others.

Results

Patient data were collected between July 2022 and June 2023. All 73 patients were male, and the mean age in the HT group (n = 40) was 72.6 years and 75.2 years for the control group (n = 33). Overall, the baseline demographics were similar between the groups (Table 1). Of 40 patients screened for enrollment in the HT program, 33 were included in the analysis (Figure 1).

At baseline, the HT group included more individuals than the control group on ACEi/ARB/ARNi (24 vs 19, respectively), β-blocker (28 vs 24, respectively), SGLT2i (14 vs 11, respectively), and aldosterone antagonist (15 vs 9, respectively) (Figure 2). There were 20 interventions made in the HT group compared with 11 therapy changes in the control arm during the study (odds ratio, 1.43; P = .23) (Table 2). In the HT group, 1 patient achieved an ACEi target dose, 3 patients achieved a β-blocker target dose, and 7 achieved a target dose of spironolactone (titration is not required for SGLT2i therapy and is counted as target dose). In the HT group, 17 patients were on ≥ 3 recommended agents, while 9 patients were taking 4 agents. Seven of 20 HT group interventions resulted in titration to the target dose. In the control group, no patients achieved an ARNi target dose, 3 patients achieved a β-blocker target dose, and 2 patients achieved a spironolactone antagonist target dose. In the control arm, 7 patients were on ≥ 3 GDMTs, and 2 were taking 4 agents. No patient in either group achieved a target dose of 4 agents. Five of 11 control group interventions resulted in initiation or titration of GDMT to the target dose.

One HF-related hospitalization occurred in the HT group and 6 in the control arm (odds ratio, 0.12; P = .04). Most patients (90%) were adherent to their GDMT regimen, 3 patients not on any GDMT for HFrEF at the end of the review were excluded, and 1 patient was deemed not adherent to medication therapy (Table 3). The mean (SD) time-to-enrollment in the HT program was 5.5 (6.4) days after consultation for 34 patients who were assessed. Ten patients had laboratory tests within a mean (SD) of 17 (9.8) days after the intervention, 29 patients did not require routine laboratory monitoring, and 1 did not follow through.

Of the 40 HT group patients, 7 were excluded from analysis (3 failed to schedule HT, 1 was at a long-term care facility, 1 was nonadherent, 1 declined participation, and 1 died) and 33 remained in the program for a mean (SD) 5.3 (3.5) months. Death rates were tracked during the study: 1 patient died in the HT group and 3 in the control group.

We analyzed the overall percentage of VAAAHS patients with HFrEF who were on appropriate GDMT. Given the ongoing drive to improve HF-related outcomes, HT interventions could not be compared to a static population, so the HT and control patients were compared with the rates of GDMT at VAAAHS, which was available in the Academic Detailing Service Heart Failure Dashboard (Figure 3). Titration and optimization rates were also compared (Figure 4). From July 2022 to June 2023, ARNi use increased by 16.6%, aldosterone antagonist by 6.8%, and β-blockers by 2.4%. Target doses of GDMTs were more difficult to achieve in the hospital system. There was an increase in aldosterone antagonist target dose achievement by 4.7%, but overall there were decreases in target doses in other GDMTs: ACEi/ARB/ARNi target dose use decreased by 3.2%, ARNi target dose use decreased by 2.7% and target β-blocker use decreased by 0.9%.

 

Discussion

Telehealth yielded clinically important interventions, with some titrations bringing patients to their target doses of medications for HFrEF. The 20 interventions made in the HT group can be largely attributed to the nurses’ efforts to alert clinicians to drug titrations or ACEi/ARB to ARNi transitions. Although the findings were not statistically significant, the difference in the number of drug therapy changes supports the use of the HT program for a GDMT optimization strategy. Patients may be difficult to titrate secondary to adverse effects that make medication initiation or titration inappropriate, such as hypotension and hyperkalemia, although this was not observed in this small sample size. Considering a mean HT enrollment of 5.3 months, many patients had adequate disease assessment and medication titration. Given that patients are discharged from the service once deemed appropriate, this decreases the burden on the patient and increases the utility and implementation of the HT program for other patients.

A surprising finding of this study was the lower rate of HF-related hospitalizations in the HT group. Although not the primary subject of interest in the study, it reinforced the importance of close health care professional follow-up for patients living with HF. Telehealth may improve communication and shared decision making over medication use. Because the finding was unanticipated, the rate of diuretic adjustments was not tracked.

Patients were reevaluated every 6 months for willingness to continue the program, adherence, and clinical needs. These results are similar to those of other trials that demonstrated improved rates of GDMT in the setting of pharmacist- or nurse-led HF treatment optimization.^15,16 These positive results differ from other trials incorporating remote monitoring regarding patient continuation in HT programs. For example, in a study by Ding and colleagues, the withdrawal rate from their monitoring service was about 22%, while in our study only 1 patient withdrew from the HT program.¹¹

The HT program resulted in fewer hospitalizations than the control arm. There were 6 HF-related hospitalizations in the control group, although 5 involved a single patient. Typically, such a patient would be encouraged to follow HT monitoring after just 1 HF-related hospitalization; however, the patient declined to participate.

Early optimization of GDMT in patients who were recently discharged from the hospital for an HF-related hospitalization yields a reduction in hospital rehospitalization.¹⁷ GDMT optimization has unequivocal benefits in HF outcomes. Unfortunately, the issues surrounding methodologies on how to best optimize GDMT are lacking. While HT has been found to be feasible in the aid of optimizing medical therapy, the TIM-HF trial concluded that remote monitoring services had no significant benefit in reducing mortality.^7,8 On the other hand, the OptiLink HF Study showed that when clinicians respond to remote monitoring prompts from fluid index threshold crossing alerts, these interventions are associated with significantly improved clinical outcomes in patients with implantable cardioverter-defibrillators and advanced HF.⁹ In contrast to previous trials, the AMULET trial showed that remote monitoring compared with standard care significantly reduced the risk of HF hospitalization or cardiovascular death during the 12-month follow-up among patients with HF and LVEF ≤ 49% after an episode of acute exacerbation.¹⁰ Additionally, patients who received skilled home health services and participated in remote monitoring for their chronic HF experienced a reduction in all-cause 30-day readmission.¹⁸

Given the contrasting evidence regarding remote monitoring and variable modalities of implementing interventions, we investigated whether HT monitoring yields improvements in GDMT optimization. We found that HT nurses were able to nearly double the rate of interventions for patients with HFrEF. The HT program in providing expanded services will require adequate staffing responsibilities and support. The HT program is geared toward following a large, diverse patient population, such as those with chronic obstructive pulmonary disease, hypertension, and HF. We only evaluated services for patients with HFrEF, but the program also follows patients with HfmrEF and HfpEF. These patients were not included as GDMT optimization differs for patients with an LVEF > 40%.^19,20

The lower rates of achieving target doses of GDMTs were likely obstructed by continuous use of initial drug doses and further limited by lack of follow-up. When compared with the rest of the VAAAHS, there was a greater effort to increase ARNi use in the HT group as 7 of 33 patients (21%) were started on ARNi compared with a background increase of ARNi use of 17%. There was a lower mortality rate observed in the HT group compared with the control group. One patient in each group died of unrelated causes, while 2 deaths in the control group were due to worsening HF. The difference in mortality is likely multifactorial, possibly related to the control group’s greater disease burden or higher mean age (75.2 years vs 72.6 years).

Limitations

This was an observational cohort design, which is subject to bias. Thus, the findings of this study are entirely hypothesis-generating and a randomized controlled trial would be necessary for clearer results. Second, low numbers of participants may have skewed the data set. Given the observational nature of the study, this nonetheless is a positive finding to support the HT program for assisting with HF monitoring and prompting drug interventions. Due to the low number of participants, a single patient may have skewed the results with 5 hospitalizations.

Conclusions

This pilot study demonstrates the applicability of HT monitoring to optimize veteran HFrEF GDMT. The HT program yielded numerically relevant interventions and fewer HF-related hospitalizations compared with the control arm. The study shows the feasibility of the program to safely optimize GDMT toward their target doses and may serve as an additional catalyst to further develop HT programs specifically targeted toward HF monitoring and management. Cost-savings analyses would likely need to demonstrate the cost utility of such a service.

Acknowledgments

We thank the home telehealth nursing staff for their assistance in data collection and enrollment of patients into the monitoring program.

Heart failure (HF) is a chronic, progressive condition that is characterized by the heart’s inability to effectively pump blood throughout the body. In 2018, approximately 6.2 million US adults had HF, and 13.4% of all death certificates noted HF as a precipitating factor.¹ Patients not receiving appropriate guideline-directed medical therapy (GDMT) face a 29% excess mortality risk over a 2-year period.² Each additional GDMT included in a patient’s regimen significantly reduces all-cause mortality.³

The Change the Management of Patients with Heart Failure (CHAMP) registry reports that only about 1% of patients with HF are prescribed 3 agents from contemporary GDMT at target doses, highlighting the need for optimizing clinicians’ approaches to GDMT.⁴ Similarly, The Get With The Guidelines Heart Failure Registry has noted that only 20.2% of patients with HF with reduced ejection fraction (HFrEF) are prescribed a sodium-glucose cotransporter 2 inhibitor (SGLT2i) following hospital discharge for HFrEF exacerbation.⁵ Overall, treatment rates with GDMT saw limited improvement between 2013 and 2019, with no significant difference between groups in mortality, indicating the need for optimized methods to encourage the initiation of GDMT.⁶

Remote monitoring and telecare are novel ways to improve GDMT rates in those with HFrEF. However, data are inconsistent regarding the impact of remote HF monitoring and improvements in GDMT or HF-related outcomes.^6-10 The modalities of remote monitoring for GDMT vary among studies, but the potential for telehealth monitoring to improve GDMT, thereby potentially reducing HF-related hospitalizations, is clear.

Telemonitoring has demonstrated improved participant adherence with weight monitoring, although the withdrawal rate was high, and has the potential to reduce all-cause mortality and HF-related hospitalizations.^11,12 Telemonitoring for GDMT optimization led to an increase in the proportion of patients who achieved optimal GDMT doses, a decrease in the time to dose optimization, and a reduction in the number of clinic visits.¹³ Remote GDMT titration was accomplished in the general patient population with HFrEF; however, in populations already followed by cardiologists or HF specialists, remote optimization strategies did not yield different proportions of GDMT use.¹⁴ The aim of this study was to assess the impact of the home telehealth (HT) monitoring program on the initiation and optimization of HF GDMT among veterans with HFrEF at the Veterans Affairs Ann Arbor Healthcare System (VAAAHS) in Michigan.

 

Methods

This was a single-center retrospective study of Computerized Patient Record System (CPRS)data. Patients at the VAAAHS were evaluated if they were diagnosed with HFrEF and were eligible for enrollment in the HT monitoring program. Eligibility criteria included a diagnosis of stage C HF, irrespective of EF, and a history of any HF-related hospitalization. We focused on patients with HFrEF due to stronger guideline-based recommendations for certain pharmacotherapies as compared with HF with mildly reduced ejection fraction (HFmrEF) and preserved ejection fraction (HFpEF). Initial patient data for HT enrolling were accessed using the Heart Failure Dashboard via the US Department of Veterans Affairs (VA) Academic Detailing Service. The target daily doses of typical agents used in HFrEF GDMT are listed in the Appendix.

The HT program is an embedded model in which HT nurses receive remote data from the patient and triage that with the VAAAHS cardiology team. Patients’ questions, concerns, and/or vital signs are recovered remotely. In this model, nurses are embedded in the cardiology team, working with the cardiologists, cardiology clinical pharmacist, and/or cardiology nurse practitioners to make medication interventions. Data are recorded with an HT device, including weight, blood pressure (BP), heart rate, and pulse oximetry. HT nurses are also available to the patient via phone or video. The program uses a 180-day disease management protocol for HF via remote device, enabling the patient to answer questions and receive education on their disease daily. Responses to questions and data are then reviewed by an HT nurse remotely during business hours and triaged as appropriate with the cardiology team. Data can be communicated to the cardiology team via the patient record, eliminating the need for the cardiology team to use the proprietary portal affiliated with the HT device.

Study Sample

Patient information was obtained from a list of 417 patients eligible for enrollment in the HT program; the list was sent to the HT program for review and enrollment. Patient data were extracted from the VAAAHS HF Dashboard and included all patients with HFrEF and available data on the platform. The sample for the retrospective chart review included 40 adults who had HFrEF, defined as a left ventricular EF (LVEF) of ≤ 40% as evidenced by a transthoracic echocardiogram or cardiac magnetic resonance imaging. These patients were contacted and agreed to enroll in the HT monitoring program. The HT program population was compared against a control group of 33 patients who were ineligible for the HT program. Patients were deemed ineligible for HT if they resided in a nursing home, lacked a VAAAHS primary care clinician, or declined participation in the HT program.

Procedures

Patients enrolled in the HT program were assigned a nurse who monitored and responded to changes in vital signs, such as an increase in weight of ≥ 3 lb in 24 hours or ≥ 5 lb in 7 days. Each participant in HT received a home BP machine to check BP and heart rates and a scale for daily weights. The patient also responded to questions on a tablet device with the BP machine and scale. The vital signs were put into the tablet device, and the patient answered questions about their symptoms for that day. If the patient endorsed any symptoms of HF, such as pedal edema, abdominal distention, orthopnea, paroxysmal nocturnal dyspnea, dyspnea on exertion, or orthopnea, they received a call from the nurse to discuss symptoms and make appropriate adjustments in the diuretic or HF regimen. The HF-trained pharmacist served as a resource for drug therapy management and provided recommendations for diuretic adjustments on an as-needed basis.

Patients who declined participation in the HT program followed the standard of care, which was limited to visits with primary care clinicians and/or cardiologists as per the follow-up plan. Patient data were collected over 12 months. The study was approved by the VAAAHS Institutional Review Board (reference number, 1703034), Research and Development Committee, and Research Administration.

 

 

Primary and Secondary Goals

The primary goal of the study was to assess the impact of the HT program on drug interventions, specifically initiating and titrating HFrEF pharmacotherapies. Interventions were based on GDMT with known mortality- and morbidity-reducing properties when used at their maximum tolerated doses, including angiotensin-converting enzyme inhibitors (ACEi), angiotensin receptor-neprilysin inhibitor (ARNi), or angiotensin receptor blockers (ARB), with a preference for ARNi, β-blockers for HFrEF (metoprolol succinate, bisoprolol, or carvedilol), aldosterone antagonists, and SGLT2is.

Secondary goals included HF-related hospitalizations, medication adherence, time to enrollment in HT, time to laboratory analysis after the initiation or titration of an ACEi/ARB/ARNi or aldosterone antagonist, and time enrolled in the HT program. Patients were considered adherent if their drug refill history showed consistent fills of their medications. The χ² test was used for total interventions made during the study period and Fisher exact test for all others.

Results

Patient data were collected between July 2022 and June 2023. All 73 patients were male, and the mean age in the HT group (n = 40) was 72.6 years and 75.2 years for the control group (n = 33). Overall, the baseline demographics were similar between the groups (Table 1). Of 40 patients screened for enrollment in the HT program, 33 were included in the analysis (Figure 1).

At baseline, the HT group included more individuals than the control group on ACEi/ARB/ARNi (24 vs 19, respectively), β-blocker (28 vs 24, respectively), SGLT2i (14 vs 11, respectively), and aldosterone antagonist (15 vs 9, respectively) (Figure 2). There were 20 interventions made in the HT group compared with 11 therapy changes in the control arm during the study (odds ratio, 1.43; P = .23) (Table 2). In the HT group, 1 patient achieved an ACEi target dose, 3 patients achieved a β-blocker target dose, and 7 achieved a target dose of spironolactone (titration is not required for SGLT2i therapy and is counted as target dose). In the HT group, 17 patients were on ≥ 3 recommended agents, while 9 patients were taking 4 agents. Seven of 20 HT group interventions resulted in titration to the target dose. In the control group, no patients achieved an ARNi target dose, 3 patients achieved a β-blocker target dose, and 2 patients achieved a spironolactone antagonist target dose. In the control arm, 7 patients were on ≥ 3 GDMTs, and 2 were taking 4 agents. No patient in either group achieved a target dose of 4 agents. Five of 11 control group interventions resulted in initiation or titration of GDMT to the target dose.

One HF-related hospitalization occurred in the HT group and 6 in the control arm (odds ratio, 0.12; P = .04). Most patients (90%) were adherent to their GDMT regimen, 3 patients not on any GDMT for HFrEF at the end of the review were excluded, and 1 patient was deemed not adherent to medication therapy (Table 3). The mean (SD) time-to-enrollment in the HT program was 5.5 (6.4) days after consultation for 34 patients who were assessed. Ten patients had laboratory tests within a mean (SD) of 17 (9.8) days after the intervention, 29 patients did not require routine laboratory monitoring, and 1 did not follow through.

Of the 40 HT group patients, 7 were excluded from analysis (3 failed to schedule HT, 1 was at a long-term care facility, 1 was nonadherent, 1 declined participation, and 1 died) and 33 remained in the program for a mean (SD) 5.3 (3.5) months. Death rates were tracked during the study: 1 patient died in the HT group and 3 in the control group.

We analyzed the overall percentage of VAAAHS patients with HFrEF who were on appropriate GDMT. Given the ongoing drive to improve HF-related outcomes, HT interventions could not be compared to a static population, so the HT and control patients were compared with the rates of GDMT at VAAAHS, which was available in the Academic Detailing Service Heart Failure Dashboard (Figure 3). Titration and optimization rates were also compared (Figure 4). From July 2022 to June 2023, ARNi use increased by 16.6%, aldosterone antagonist by 6.8%, and β-blockers by 2.4%. Target doses of GDMTs were more difficult to achieve in the hospital system. There was an increase in aldosterone antagonist target dose achievement by 4.7%, but overall there were decreases in target doses in other GDMTs: ACEi/ARB/ARNi target dose use decreased by 3.2%, ARNi target dose use decreased by 2.7% and target β-blocker use decreased by 0.9%.

 

Discussion

Telehealth yielded clinically important interventions, with some titrations bringing patients to their target doses of medications for HFrEF. The 20 interventions made in the HT group can be largely attributed to the nurses’ efforts to alert clinicians to drug titrations or ACEi/ARB to ARNi transitions. Although the findings were not statistically significant, the difference in the number of drug therapy changes supports the use of the HT program for a GDMT optimization strategy. Patients may be difficult to titrate secondary to adverse effects that make medication initiation or titration inappropriate, such as hypotension and hyperkalemia, although this was not observed in this small sample size. Considering a mean HT enrollment of 5.3 months, many patients had adequate disease assessment and medication titration. Given that patients are discharged from the service once deemed appropriate, this decreases the burden on the patient and increases the utility and implementation of the HT program for other patients.

A surprising finding of this study was the lower rate of HF-related hospitalizations in the HT group. Although not the primary subject of interest in the study, it reinforced the importance of close health care professional follow-up for patients living with HF. Telehealth may improve communication and shared decision making over medication use. Because the finding was unanticipated, the rate of diuretic adjustments was not tracked.

Patients were reevaluated every 6 months for willingness to continue the program, adherence, and clinical needs. These results are similar to those of other trials that demonstrated improved rates of GDMT in the setting of pharmacist- or nurse-led HF treatment optimization.^15,16 These positive results differ from other trials incorporating remote monitoring regarding patient continuation in HT programs. For example, in a study by Ding and colleagues, the withdrawal rate from their monitoring service was about 22%, while in our study only 1 patient withdrew from the HT program.¹¹

The HT program resulted in fewer hospitalizations than the control arm. There were 6 HF-related hospitalizations in the control group, although 5 involved a single patient. Typically, such a patient would be encouraged to follow HT monitoring after just 1 HF-related hospitalization; however, the patient declined to participate.

Early optimization of GDMT in patients who were recently discharged from the hospital for an HF-related hospitalization yields a reduction in hospital rehospitalization.¹⁷ GDMT optimization has unequivocal benefits in HF outcomes. Unfortunately, the issues surrounding methodologies on how to best optimize GDMT are lacking. While HT has been found to be feasible in the aid of optimizing medical therapy, the TIM-HF trial concluded that remote monitoring services had no significant benefit in reducing mortality.^7,8 On the other hand, the OptiLink HF Study showed that when clinicians respond to remote monitoring prompts from fluid index threshold crossing alerts, these interventions are associated with significantly improved clinical outcomes in patients with implantable cardioverter-defibrillators and advanced HF.⁹ In contrast to previous trials, the AMULET trial showed that remote monitoring compared with standard care significantly reduced the risk of HF hospitalization or cardiovascular death during the 12-month follow-up among patients with HF and LVEF ≤ 49% after an episode of acute exacerbation.¹⁰ Additionally, patients who received skilled home health services and participated in remote monitoring for their chronic HF experienced a reduction in all-cause 30-day readmission.¹⁸

Given the contrasting evidence regarding remote monitoring and variable modalities of implementing interventions, we investigated whether HT monitoring yields improvements in GDMT optimization. We found that HT nurses were able to nearly double the rate of interventions for patients with HFrEF. The HT program in providing expanded services will require adequate staffing responsibilities and support. The HT program is geared toward following a large, diverse patient population, such as those with chronic obstructive pulmonary disease, hypertension, and HF. We only evaluated services for patients with HFrEF, but the program also follows patients with HfmrEF and HfpEF. These patients were not included as GDMT optimization differs for patients with an LVEF > 40%.^19,20

The lower rates of achieving target doses of GDMTs were likely obstructed by continuous use of initial drug doses and further limited by lack of follow-up. When compared with the rest of the VAAAHS, there was a greater effort to increase ARNi use in the HT group as 7 of 33 patients (21%) were started on ARNi compared with a background increase of ARNi use of 17%. There was a lower mortality rate observed in the HT group compared with the control group. One patient in each group died of unrelated causes, while 2 deaths in the control group were due to worsening HF. The difference in mortality is likely multifactorial, possibly related to the control group’s greater disease burden or higher mean age (75.2 years vs 72.6 years).

Limitations

This was an observational cohort design, which is subject to bias. Thus, the findings of this study are entirely hypothesis-generating and a randomized controlled trial would be necessary for clearer results. Second, low numbers of participants may have skewed the data set. Given the observational nature of the study, this nonetheless is a positive finding to support the HT program for assisting with HF monitoring and prompting drug interventions. Due to the low number of participants, a single patient may have skewed the results with 5 hospitalizations.

Conclusions

This pilot study demonstrates the applicability of HT monitoring to optimize veteran HFrEF GDMT. The HT program yielded numerically relevant interventions and fewer HF-related hospitalizations compared with the control arm. The study shows the feasibility of the program to safely optimize GDMT toward their target doses and may serve as an additional catalyst to further develop HT programs specifically targeted toward HF monitoring and management. Cost-savings analyses would likely need to demonstrate the cost utility of such a service.

Acknowledgments

We thank the home telehealth nursing staff for their assistance in data collection and enrollment of patients into the monitoring program.

Opioid use disorder (OUD) is a significant cause of morbidity, mortality, and health care costs in the US.^1,2 Surgery can be the inciting cause for exposure to an opioid; as many as 23% of patients develop chronic OUD following surgery.^3,4 Patients with a history of substance use, mood disorders, anxiety, or previous chronic opioid use (COU) are at risk for relapse, dose escalation, and poor pain control after high-risk surgery, such as orthopedic joint procedures.⁵ Recently focus has been on identifying high-risk patients before orthopedic joint surgery and implementing evidence-based strategies that reduce the postoperative incidence of COU.

A transitional pain service (TPS) has been shown to reduce COU for high-risk surgical patients in different health care settings.^6-9 The TPS model bundles multiple interventions that can be applied to patients at high risk for COU within a health care system. This includes individually tailored programs for preoperative education or pain management planning, use of multimodal analgesia (including regional or neuraxial techniques or nonopioid systemic medications), application of nonpharmacologic modalities (such as cognitive-based intervention), and a coordinated approach to postdischarge instructions and transitions of care. These interventions are coordinated by a multidisciplinary clinical service consisting of anesthesiologists and advanced practice clinicians with specialization in acute pain management and opioid tapering, nurse care coordinators, and psychologists with expertise in cognitive behavioral therapy.

TPS has been shown to reduce the incidence of COU for patients undergoing orthopedic joint surgery, but its impact on health care use and costs is unknown.^6-9 The TPS intervention is resource intensive and increases the use of health care for preoperative education or pain management, which may increase the burden of costs. However, reducing long-term COU may reduce the use of health care for COU- and OUD-related complications, leading to cost savings. This study evaluated whether the TPS intervention influenced health care use and cost for inpatient, outpatient, or pharmacy services during the year following orthopedic joint surgery compared with that of the standard pain management care for procedures that place patients at high risk for COU. We used a difference-in-differences (DID) analysis to estimate this intervention effect, using multivariable regression models that controlled for unobserved time trends and cohort characteristics.

METHODS

This was a quasi-experimental study of patients who underwent orthopedic joint surgery and associated procedures at high risk for COU at the Veterans Affairs Salt Lake City Healthcare System (VASLCHS) between January 2016 through April 2020. The pre-TPS period between January 2016 through December 2017 was compared with the post-TPS period between January 2018 to September 2019. The control patient cohort was selected from 5 geographically diverse VA health care systems throughout the US: Eastern Colorado, Central Plains (Nebraska), White River Junction (Vermont), North Florida/South Georgia, and Portland (Oregon). By sampling health care costs from VA medical centers (VAMCs) across these different regions, our control group was generalizable to veterans receiving orthopedic joint surgery across the US. This study used data from the US Department of Veterans Affairs (VA) Corporate Data Warehouse, a repository of nearly all clinical and administrative data found in electronic health records for VA-provided care and fee-basis care paid for by the VA.¹⁰ All data were hosted and analyzed in the VA Informatics and Computing Infrastructure (VINCI) workspace. The University of Utah Institutional Review Board and the VASLCHS Office of Research and Development approved the protocol for this study.

TPS Intervention

The VASLCHS TPS has already been described in detail elsewhere.^6,7 Briefly, patients at high risk for COU at the VASLCHS were enrolled in the TPS program before surgery for total knee, hip, or shoulder arthroplasty or rotator cuff procedures. The TPS service consists of an anesthesiologist and advanced practice clinician with specialization in acute pain management and opioid tapering, a psychologist with expertise in cognitive behavioral therapy, and 3 nurse care coordinators. These TPS practitioners work together to provide preoperative education, including setting expectations regarding postoperative pain, recommending nonopioid pain management strategies, and providing guidance regarding the appropriate use of opioids for surgical pain. Individual pain plans were developed and implemented for the perioperative period. After surgery, the TPS provided recommendations and support for nonopioid pain therapies and opioid tapers. Patients were followed by the TPS team for at least 12 months after surgery. At a minimum, the goals set by TPS included cessation of all opioid use for prior nonopioid users (NOU) by 90 days after surgery and the return to baseline opioid use or lower for prior COU patients by 90 days after surgery. The TPS also encouraged and supported opioid tapering among COU patients to reduce or completely stop opioid use after surgery.

Patient Cohorts

Veterans having primary or revision total knee, hip, or shoulder arthroplasty or rotator cuff repair between January 1, 2016, and September 30, 2019, at the aforementioned VAMCs were included in the study. Patients who had any hospitalization within 90 days pre- or postindex surgery or who died within 8 months after surgery were excluded from analysis. Patients who had multiple surgeries during the index inpatient visit or within 90 days after the index surgery also were excluded. Comorbid conditions for mental health and substance use were identified using the International Classification of Diseases, 10th revision Clinical Modification (ICD-10) codes or 9th revision equivalent grouped by Clinical Classifications Software Refined (CCS-R).¹¹ Preoperative exposure to clinically relevant pharmacotherapy (ie, agents associated with prolonged opioid use and nonopioid adjuvants) was captured using VA outpatient prescription records (eAppendix 1).

 

The study patient cohort was stratified into either NOU or COU groups based on opioid use before surgery. Preoperative COU was defined as > 25% nonzero days (calculated using the cabinet supply method) in the 180 days before surgery admit date time (> 45 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released within 90 days before surgery date.^12,13 For NOU patients, the postoperative outcome of interest was new postoperative prolonged opioid use. This was defined by Page and colleagues as ≥ 1 opioid prescription released between surgery discharge date and postdischarge day 44, ≥ 1 opioid prescription released between postdischarge day 45 and 89, and ≥ 1 opioid prescription released between postdischarge day 90 and 180.¹⁴ For COU patients at the time of surgery, the postoperative outcome measure of interest was continued COU, defined as > 25% nonzero days between postdischarge day 90 and 240 (> 37 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released between postdischarge day 90 and 180.

Outcome Variables

Outcome variables included health care use and costs during 1-year pre- and postperiods from the date of surgery. VA health care costs for outpatient, inpatient, and pharmacy services for direct patient care were collected from the Managerial Cost Accounting System, an activity-based cost allocation system that generates estimates of the cost of individual VA hospital stays, health care encounters, and medications. Health care use was defined as the number of encounters for each visit type in the Managerial Cost Accounting System. All costs were adjusted to 2019 US dollars, using the Personal Consumption Expenditures price index for health care services.¹⁵

A set of sociodemographic variables including sex, age at surgery, race and ethnicity, rurality, military branch (Army, Air Force, Marine Corps, Navy, and other), and service connectivity were included as covariates in our regression models. In addition, ICD-9 and ICD-10 codes were used to define the indicators of whether veterans had a diagnosis of mental illness (including anxiety, bipolar disorder, depression, or trauma) or substance use (including alcohol, cannabis, opioids, or tobacco). Finally, pharmacy records were used to create indicators for prescribed opioid-relevant pharmacotherapy (including antidepressants, benzodiazepines [BZD], gabapentinoids, muscle relaxants, non-BZD sedative hypnotics) and active antidepressant drug use during the 1-year preindex period.

Statistical Analyses

Descriptive analyses were used to evaluate differences in baseline patient sociodemographic and clinical characteristics between pre- and postperiods for TPS intervention and control cohorts using 2-sample t tests for continuous variables and χ² tests for categorical variables. We summarized unadjusted health care use and costs for outpatient, inpatient, and pharmacy visits and compared the pre- and postintervention periods using the Mann-Whitney test. Both mean (SD) and median (IQR) were considered, reflecting the skewed distribution of the outcome variables.

We used a DID approach to assess the intervention effect while minimizing confounding from the nonrandom sample. The DID approach controls for unobserved differences between VAMCs that are related to both the intervention and outcomes while controlling for trends over time that could affect outcomes across clinics. To implement the DID approach, we included 3 key independent variables in our regression models: (1) an indicator for whether the observation occurred in the postintervention period; (2) an indicator for whether the patient was exposed to the TPS intervention; and (3) the interaction between these 2 variables.

For cost outcomes, we used multivariable generalized linear models with a log link and a Poisson or Υ family. We analyzed inpatient costs using a 2-part generalized linear model because only 17% to 20% of patients had ≥ 1 inpatient visit. We used multivariable negative binomial regression for health care use outcomes. Demographic and clinical covariates described earlier were included in the regression models to control for differences in the composition of patient groups and clinics that could lead to confounding bias.

 

RESULTS

Of the 4954 patients included in our study cohort, 3545 (71.6%) were in the NOU group and 1409 (28.4%) were in the COU group. Among the NOU cohort, 361 patients were in the intervention group and 3184 in the control group. Among the COU cohort, 149 patients were in the intervention group and 1260 in the control group (Table 1). Most patients were male, White race, with a mean (SD) age of 64 (11) years. The most common orthopedic procedure was total knee arthroplasty, followed by total hip arthroplasty. Among both NOU and COU cohorts, patients’ characteristics were similar between the pre- and postintervention period among either TPS or control cohort.

Figures 1 and 2 and eAppendix 2 depict unadjusted per-person average outpatient, inpatient, and pharmacy visits and costs incurred during the 1-year pre- and postintervention periods for the NOU and COU cohorts. Average total health care follow-up costs ranged from $40,000 to $53,000 for NOU and from $47,000 to $82,000 for COU cohort. Cost for outpatient visits accounted for about 70% of the average total costs, followed by costs for inpatient visits of about 20%, and costs for pharmacy for the remaining.

For the NOU cohort, the number of health care encounters remained fairly stable between periods except for the outpatient visits among the TPS group. The TPS group experienced an increase in mean outpatient visits in the postperiod: 30 vs 37 visits (23%) (P < .001). Mean outpatient and inpatient costs in the pre- and postintervention periods were not significantly different for either the TPS or control groups. Similarly, within the COU cohort, the TPS group experienced a 27% increase in the mean number of outpatient visits (41 vs 52 visits; P = .02) and a 39% increase in mean outpatient costs in the postintervention compared with the preintervention period ($44,682 vs $61,890; P = .02). Although the mean number of outpatient visits for control group remained at a similar level, average outpatient costs increased roughly 13% ($31,068 vs $35,148; P = .01) between the pre- and postintervention periods.

Table 2 summarizes the results from the multivariable DID analyses for the outpatient, inpatient, and pharmacy visit and cost outcomes. Here, the estimated effect of the TPS intervention is the coefficient from the interaction between the postintervention and TPS exposure indicator variables. This coefficient was calculated as the difference in the outcome before and after the TPS intervention among the TPS group minus the difference in the outcome before and after the TPS intervention among the control group. For the NOU cohort, TPS was associated with an increase in the use of outpatient health care (mean [SD] increase of 6.9 [2] visits; P < .001) after the surgery with no statistically significant effect on outpatient costs (mean [SD] increase of $2787 [$3749]; P = .55). There was no statistically significant effect of TPS on the use of inpatient visits or pharmacy, but a decrease in costs for inpatient visits among those who had at least 1 inpatient visit (mean [SD] decrease of $12,170 [$6100]; P = .02). For the COU cohort, TPS had no statistically significant impact on the use of outpatient, inpatient, or pharmacy or the corresponding costs.

 

DISCUSSION

TPS is a multidisciplinary approach to perioperative pain management that has been shown to reduce both the quantity and duration of opioid use among orthopedic surgery patients.^6,7 Although the cost burden of providing TPS services to prevent COU is borne by the individual health care system, it is unclear whether this expense is offset by lower long-term medical costs and health care use for COU- and OUD-related complications. In this study focused on a veteran population undergoing orthopedic joint procedures, a DID analysis of cost and health care use showed that TPS, which has been shown to reduce COU for high-risk surgical patients, can be implemented without increasing the overall costs to the VA health care system during the 1 year following surgery, even with increased outpatient visits. For NOU patients, there was no difference in outpatient visit costs or pharmacy costs over 12 months after surgery, although there was a significant reduction in subsequent inpatient costs over the same period. Further, there was no difference in outpatient, inpatient, or pharmacy costs after surgery for COU patients. These findings suggest that TPS can be a cost-effective approach to reduce opioid use among patients undergoing orthopedic joint surgery in VAMCs.

The costs of managing COU after surgery are substantial. Prior reports have shown that adjusted total health care costs are 1.6 to 2.5 times higher for previously NOU patients with new COU after major surgery than those for such patients without persistent use.¹⁶ The 1-year costs associated with new COU in this prior study ranged between $7944 and $17,702 after inpatient surgery and between $5598 and $12,834 after outpatient index surgery, depending on the payer, which are in line with the cost differences found in our current study. Another report among patients with COU following orthopedic joint replacement showed that they had higher use of inpatient, emergency department, and ambulance/paramedic services in the 12 months following their surgery than did those without persistent use.¹⁷ Although these results highlight the impact that COU plays in driving increased costs after major surgery, there have been limited studies focused on interventions that can neutralize the costs associated with opioid misuse after surgery. To our knowledge, our study is the first analysis to show the impact of using an intervention such as TPS to reduce postoperative opioid use on health care use and cost.

Although a rigorous and comprehensive return on investment analysis was beyond the scope of this analysis, these results may have several implications for other health care systems and hospitals that wish to invest in a multidisciplinary perioperative pain management program such as TPS but may be reluctant due to the upfront investment. First, the increased number of patient follow-up visits needed during TPS seems to be more than offset by the reduction in opioid use and associated complications that may occur after surgery. Second, TPS did not seem to be associated with an increase in overall health care costs during the 1-year follow-up period. Together, these results indicate that the return on investment for a TPS approach to perioperative pain management in which optimal patient-centered outcomes are achieved without increasing long-term costs to a health care system may be positive.

Limitations

This study has several limitations. First, this was a quasi-experimental observational study, and the associations we identified between intervention and outcomes should not be assumed to demonstrate causality. Although our DID analysis controlled for an array of demographic and clinical characteristics, differences in medical costs and health care use between the 2 cohorts might be driven by unobserved confounding variables.

Our study also was limited to veterans who received medical care at the VA, and results may not be generalizable to other non-VA health care systems or to veterans with Medicare insurance who have dual benefits. While our finding on health care use and costs may be incomplete because of the uncaptured health care use outside the VA, our DID analysis helped reduce unobserved bias because the absence of data outside of VA care applies to both TPS and control groups. Further, the total costs of operating a TPS program at any given institution will depend on the size of the hospital and volume of surgical patients who meet criteria for enrollment. However, the relative differences in health care use and costs may be extrapolated to patients undergoing orthopedic surgery in other types of academic and community-based health care systems.

Furthermore, this analysis focused primarily on COU and NOU patients undergoing orthopedic joint surgery. While this represents a high-risk population for OUD, the costs and health care use associated with delivering the TPS intervention to other types of surgical procedures may be significantly different. All costs in this analysis were based on 2019 estimates and do not account for the potential inflation over the past several years. Nonmonetary costs to the patient and per-person average total intervention costs were not included in the study. However, we assumed that costs associated with TPS and standard of care would have increased to an equivalent degree over the same period. Further, the average cost of TPS per patient (approximately $900) is relatively small compared with the average annual costs during 1-year pre- and postoperative periods and was not expected to have a significant effect on the analysis.

Conclusions

We found that the significant reduction in COU seen in previous studies following the implementation of TPS was not accompanied by increased health care costs.^6,7 When considering the other costs of long-term opioid use, such as abuse potential, overdose, death, and increased disability, implementation of a TPS service has the potential to improve patient quality of life while reducing other health-related costs. Health care systems should consider the implementation of similar multidisciplinary approaches to perioperative pain management to improve outcomes after orthopedic joint surgery and other high-risk procedures.

Opioid use disorder (OUD) is a significant cause of morbidity, mortality, and health care costs in the US.^1,2 Surgery can be the inciting cause for exposure to an opioid; as many as 23% of patients develop chronic OUD following surgery.^3,4 Patients with a history of substance use, mood disorders, anxiety, or previous chronic opioid use (COU) are at risk for relapse, dose escalation, and poor pain control after high-risk surgery, such as orthopedic joint procedures.⁵ Recently focus has been on identifying high-risk patients before orthopedic joint surgery and implementing evidence-based strategies that reduce the postoperative incidence of COU.

A transitional pain service (TPS) has been shown to reduce COU for high-risk surgical patients in different health care settings.^6-9 The TPS model bundles multiple interventions that can be applied to patients at high risk for COU within a health care system. This includes individually tailored programs for preoperative education or pain management planning, use of multimodal analgesia (including regional or neuraxial techniques or nonopioid systemic medications), application of nonpharmacologic modalities (such as cognitive-based intervention), and a coordinated approach to postdischarge instructions and transitions of care. These interventions are coordinated by a multidisciplinary clinical service consisting of anesthesiologists and advanced practice clinicians with specialization in acute pain management and opioid tapering, nurse care coordinators, and psychologists with expertise in cognitive behavioral therapy.

TPS has been shown to reduce the incidence of COU for patients undergoing orthopedic joint surgery, but its impact on health care use and costs is unknown.^6-9 The TPS intervention is resource intensive and increases the use of health care for preoperative education or pain management, which may increase the burden of costs. However, reducing long-term COU may reduce the use of health care for COU- and OUD-related complications, leading to cost savings. This study evaluated whether the TPS intervention influenced health care use and cost for inpatient, outpatient, or pharmacy services during the year following orthopedic joint surgery compared with that of the standard pain management care for procedures that place patients at high risk for COU. We used a difference-in-differences (DID) analysis to estimate this intervention effect, using multivariable regression models that controlled for unobserved time trends and cohort characteristics.

METHODS

This was a quasi-experimental study of patients who underwent orthopedic joint surgery and associated procedures at high risk for COU at the Veterans Affairs Salt Lake City Healthcare System (VASLCHS) between January 2016 through April 2020. The pre-TPS period between January 2016 through December 2017 was compared with the post-TPS period between January 2018 to September 2019. The control patient cohort was selected from 5 geographically diverse VA health care systems throughout the US: Eastern Colorado, Central Plains (Nebraska), White River Junction (Vermont), North Florida/South Georgia, and Portland (Oregon). By sampling health care costs from VA medical centers (VAMCs) across these different regions, our control group was generalizable to veterans receiving orthopedic joint surgery across the US. This study used data from the US Department of Veterans Affairs (VA) Corporate Data Warehouse, a repository of nearly all clinical and administrative data found in electronic health records for VA-provided care and fee-basis care paid for by the VA.¹⁰ All data were hosted and analyzed in the VA Informatics and Computing Infrastructure (VINCI) workspace. The University of Utah Institutional Review Board and the VASLCHS Office of Research and Development approved the protocol for this study.

TPS Intervention

The VASLCHS TPS has already been described in detail elsewhere.^6,7 Briefly, patients at high risk for COU at the VASLCHS were enrolled in the TPS program before surgery for total knee, hip, or shoulder arthroplasty or rotator cuff procedures. The TPS service consists of an anesthesiologist and advanced practice clinician with specialization in acute pain management and opioid tapering, a psychologist with expertise in cognitive behavioral therapy, and 3 nurse care coordinators. These TPS practitioners work together to provide preoperative education, including setting expectations regarding postoperative pain, recommending nonopioid pain management strategies, and providing guidance regarding the appropriate use of opioids for surgical pain. Individual pain plans were developed and implemented for the perioperative period. After surgery, the TPS provided recommendations and support for nonopioid pain therapies and opioid tapers. Patients were followed by the TPS team for at least 12 months after surgery. At a minimum, the goals set by TPS included cessation of all opioid use for prior nonopioid users (NOU) by 90 days after surgery and the return to baseline opioid use or lower for prior COU patients by 90 days after surgery. The TPS also encouraged and supported opioid tapering among COU patients to reduce or completely stop opioid use after surgery.

Patient Cohorts

Veterans having primary or revision total knee, hip, or shoulder arthroplasty or rotator cuff repair between January 1, 2016, and September 30, 2019, at the aforementioned VAMCs were included in the study. Patients who had any hospitalization within 90 days pre- or postindex surgery or who died within 8 months after surgery were excluded from analysis. Patients who had multiple surgeries during the index inpatient visit or within 90 days after the index surgery also were excluded. Comorbid conditions for mental health and substance use were identified using the International Classification of Diseases, 10th revision Clinical Modification (ICD-10) codes or 9th revision equivalent grouped by Clinical Classifications Software Refined (CCS-R).¹¹ Preoperative exposure to clinically relevant pharmacotherapy (ie, agents associated with prolonged opioid use and nonopioid adjuvants) was captured using VA outpatient prescription records (eAppendix 1).

 

The study patient cohort was stratified into either NOU or COU groups based on opioid use before surgery. Preoperative COU was defined as > 25% nonzero days (calculated using the cabinet supply method) in the 180 days before surgery admit date time (> 45 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released within 90 days before surgery date.^12,13 For NOU patients, the postoperative outcome of interest was new postoperative prolonged opioid use. This was defined by Page and colleagues as ≥ 1 opioid prescription released between surgery discharge date and postdischarge day 44, ≥ 1 opioid prescription released between postdischarge day 45 and 89, and ≥ 1 opioid prescription released between postdischarge day 90 and 180.¹⁴ For COU patients at the time of surgery, the postoperative outcome measure of interest was continued COU, defined as > 25% nonzero days between postdischarge day 90 and 240 (> 37 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released between postdischarge day 90 and 180.

Outcome Variables

Outcome variables included health care use and costs during 1-year pre- and postperiods from the date of surgery. VA health care costs for outpatient, inpatient, and pharmacy services for direct patient care were collected from the Managerial Cost Accounting System, an activity-based cost allocation system that generates estimates of the cost of individual VA hospital stays, health care encounters, and medications. Health care use was defined as the number of encounters for each visit type in the Managerial Cost Accounting System. All costs were adjusted to 2019 US dollars, using the Personal Consumption Expenditures price index for health care services.¹⁵

A set of sociodemographic variables including sex, age at surgery, race and ethnicity, rurality, military branch (Army, Air Force, Marine Corps, Navy, and other), and service connectivity were included as covariates in our regression models. In addition, ICD-9 and ICD-10 codes were used to define the indicators of whether veterans had a diagnosis of mental illness (including anxiety, bipolar disorder, depression, or trauma) or substance use (including alcohol, cannabis, opioids, or tobacco). Finally, pharmacy records were used to create indicators for prescribed opioid-relevant pharmacotherapy (including antidepressants, benzodiazepines [BZD], gabapentinoids, muscle relaxants, non-BZD sedative hypnotics) and active antidepressant drug use during the 1-year preindex period.

Statistical Analyses

Descriptive analyses were used to evaluate differences in baseline patient sociodemographic and clinical characteristics between pre- and postperiods for TPS intervention and control cohorts using 2-sample t tests for continuous variables and χ² tests for categorical variables. We summarized unadjusted health care use and costs for outpatient, inpatient, and pharmacy visits and compared the pre- and postintervention periods using the Mann-Whitney test. Both mean (SD) and median (IQR) were considered, reflecting the skewed distribution of the outcome variables.

We used a DID approach to assess the intervention effect while minimizing confounding from the nonrandom sample. The DID approach controls for unobserved differences between VAMCs that are related to both the intervention and outcomes while controlling for trends over time that could affect outcomes across clinics. To implement the DID approach, we included 3 key independent variables in our regression models: (1) an indicator for whether the observation occurred in the postintervention period; (2) an indicator for whether the patient was exposed to the TPS intervention; and (3) the interaction between these 2 variables.

For cost outcomes, we used multivariable generalized linear models with a log link and a Poisson or Υ family. We analyzed inpatient costs using a 2-part generalized linear model because only 17% to 20% of patients had ≥ 1 inpatient visit. We used multivariable negative binomial regression for health care use outcomes. Demographic and clinical covariates described earlier were included in the regression models to control for differences in the composition of patient groups and clinics that could lead to confounding bias.

 

RESULTS

Of the 4954 patients included in our study cohort, 3545 (71.6%) were in the NOU group and 1409 (28.4%) were in the COU group. Among the NOU cohort, 361 patients were in the intervention group and 3184 in the control group. Among the COU cohort, 149 patients were in the intervention group and 1260 in the control group (Table 1). Most patients were male, White race, with a mean (SD) age of 64 (11) years. The most common orthopedic procedure was total knee arthroplasty, followed by total hip arthroplasty. Among both NOU and COU cohorts, patients’ characteristics were similar between the pre- and postintervention period among either TPS or control cohort.

Figures 1 and 2 and eAppendix 2 depict unadjusted per-person average outpatient, inpatient, and pharmacy visits and costs incurred during the 1-year pre- and postintervention periods for the NOU and COU cohorts. Average total health care follow-up costs ranged from $40,000 to $53,000 for NOU and from $47,000 to $82,000 for COU cohort. Cost for outpatient visits accounted for about 70% of the average total costs, followed by costs for inpatient visits of about 20%, and costs for pharmacy for the remaining.

For the NOU cohort, the number of health care encounters remained fairly stable between periods except for the outpatient visits among the TPS group. The TPS group experienced an increase in mean outpatient visits in the postperiod: 30 vs 37 visits (23%) (P < .001). Mean outpatient and inpatient costs in the pre- and postintervention periods were not significantly different for either the TPS or control groups. Similarly, within the COU cohort, the TPS group experienced a 27% increase in the mean number of outpatient visits (41 vs 52 visits; P = .02) and a 39% increase in mean outpatient costs in the postintervention compared with the preintervention period ($44,682 vs $61,890; P = .02). Although the mean number of outpatient visits for control group remained at a similar level, average outpatient costs increased roughly 13% ($31,068 vs $35,148; P = .01) between the pre- and postintervention periods.

Table 2 summarizes the results from the multivariable DID analyses for the outpatient, inpatient, and pharmacy visit and cost outcomes. Here, the estimated effect of the TPS intervention is the coefficient from the interaction between the postintervention and TPS exposure indicator variables. This coefficient was calculated as the difference in the outcome before and after the TPS intervention among the TPS group minus the difference in the outcome before and after the TPS intervention among the control group. For the NOU cohort, TPS was associated with an increase in the use of outpatient health care (mean [SD] increase of 6.9 [2] visits; P < .001) after the surgery with no statistically significant effect on outpatient costs (mean [SD] increase of $2787 [$3749]; P = .55). There was no statistically significant effect of TPS on the use of inpatient visits or pharmacy, but a decrease in costs for inpatient visits among those who had at least 1 inpatient visit (mean [SD] decrease of $12,170 [$6100]; P = .02). For the COU cohort, TPS had no statistically significant impact on the use of outpatient, inpatient, or pharmacy or the corresponding costs.

 

DISCUSSION

TPS is a multidisciplinary approach to perioperative pain management that has been shown to reduce both the quantity and duration of opioid use among orthopedic surgery patients.^6,7 Although the cost burden of providing TPS services to prevent COU is borne by the individual health care system, it is unclear whether this expense is offset by lower long-term medical costs and health care use for COU- and OUD-related complications. In this study focused on a veteran population undergoing orthopedic joint procedures, a DID analysis of cost and health care use showed that TPS, which has been shown to reduce COU for high-risk surgical patients, can be implemented without increasing the overall costs to the VA health care system during the 1 year following surgery, even with increased outpatient visits. For NOU patients, there was no difference in outpatient visit costs or pharmacy costs over 12 months after surgery, although there was a significant reduction in subsequent inpatient costs over the same period. Further, there was no difference in outpatient, inpatient, or pharmacy costs after surgery for COU patients. These findings suggest that TPS can be a cost-effective approach to reduce opioid use among patients undergoing orthopedic joint surgery in VAMCs.

The costs of managing COU after surgery are substantial. Prior reports have shown that adjusted total health care costs are 1.6 to 2.5 times higher for previously NOU patients with new COU after major surgery than those for such patients without persistent use.¹⁶ The 1-year costs associated with new COU in this prior study ranged between $7944 and $17,702 after inpatient surgery and between $5598 and $12,834 after outpatient index surgery, depending on the payer, which are in line with the cost differences found in our current study. Another report among patients with COU following orthopedic joint replacement showed that they had higher use of inpatient, emergency department, and ambulance/paramedic services in the 12 months following their surgery than did those without persistent use.¹⁷ Although these results highlight the impact that COU plays in driving increased costs after major surgery, there have been limited studies focused on interventions that can neutralize the costs associated with opioid misuse after surgery. To our knowledge, our study is the first analysis to show the impact of using an intervention such as TPS to reduce postoperative opioid use on health care use and cost.

Although a rigorous and comprehensive return on investment analysis was beyond the scope of this analysis, these results may have several implications for other health care systems and hospitals that wish to invest in a multidisciplinary perioperative pain management program such as TPS but may be reluctant due to the upfront investment. First, the increased number of patient follow-up visits needed during TPS seems to be more than offset by the reduction in opioid use and associated complications that may occur after surgery. Second, TPS did not seem to be associated with an increase in overall health care costs during the 1-year follow-up period. Together, these results indicate that the return on investment for a TPS approach to perioperative pain management in which optimal patient-centered outcomes are achieved without increasing long-term costs to a health care system may be positive.

Limitations

This study has several limitations. First, this was a quasi-experimental observational study, and the associations we identified between intervention and outcomes should not be assumed to demonstrate causality. Although our DID analysis controlled for an array of demographic and clinical characteristics, differences in medical costs and health care use between the 2 cohorts might be driven by unobserved confounding variables.

Our study also was limited to veterans who received medical care at the VA, and results may not be generalizable to other non-VA health care systems or to veterans with Medicare insurance who have dual benefits. While our finding on health care use and costs may be incomplete because of the uncaptured health care use outside the VA, our DID analysis helped reduce unobserved bias because the absence of data outside of VA care applies to both TPS and control groups. Further, the total costs of operating a TPS program at any given institution will depend on the size of the hospital and volume of surgical patients who meet criteria for enrollment. However, the relative differences in health care use and costs may be extrapolated to patients undergoing orthopedic surgery in other types of academic and community-based health care systems.

Furthermore, this analysis focused primarily on COU and NOU patients undergoing orthopedic joint surgery. While this represents a high-risk population for OUD, the costs and health care use associated with delivering the TPS intervention to other types of surgical procedures may be significantly different. All costs in this analysis were based on 2019 estimates and do not account for the potential inflation over the past several years. Nonmonetary costs to the patient and per-person average total intervention costs were not included in the study. However, we assumed that costs associated with TPS and standard of care would have increased to an equivalent degree over the same period. Further, the average cost of TPS per patient (approximately $900) is relatively small compared with the average annual costs during 1-year pre- and postoperative periods and was not expected to have a significant effect on the analysis.

Conclusions

We found that the significant reduction in COU seen in previous studies following the implementation of TPS was not accompanied by increased health care costs.^6,7 When considering the other costs of long-term opioid use, such as abuse potential, overdose, death, and increased disability, implementation of a TPS service has the potential to improve patient quality of life while reducing other health-related costs. Health care systems should consider the implementation of similar multidisciplinary approaches to perioperative pain management to improve outcomes after orthopedic joint surgery and other high-risk procedures.

Opioid use disorder (OUD) is a significant cause of morbidity, mortality, and health care costs in the US.^1,2 Surgery can be the inciting cause for exposure to an opioid; as many as 23% of patients develop chronic OUD following surgery.^3,4 Patients with a history of substance use, mood disorders, anxiety, or previous chronic opioid use (COU) are at risk for relapse, dose escalation, and poor pain control after high-risk surgery, such as orthopedic joint procedures.⁵ Recently focus has been on identifying high-risk patients before orthopedic joint surgery and implementing evidence-based strategies that reduce the postoperative incidence of COU.

A transitional pain service (TPS) has been shown to reduce COU for high-risk surgical patients in different health care settings.^6-9 The TPS model bundles multiple interventions that can be applied to patients at high risk for COU within a health care system. This includes individually tailored programs for preoperative education or pain management planning, use of multimodal analgesia (including regional or neuraxial techniques or nonopioid systemic medications), application of nonpharmacologic modalities (such as cognitive-based intervention), and a coordinated approach to postdischarge instructions and transitions of care. These interventions are coordinated by a multidisciplinary clinical service consisting of anesthesiologists and advanced practice clinicians with specialization in acute pain management and opioid tapering, nurse care coordinators, and psychologists with expertise in cognitive behavioral therapy.

TPS has been shown to reduce the incidence of COU for patients undergoing orthopedic joint surgery, but its impact on health care use and costs is unknown.^6-9 The TPS intervention is resource intensive and increases the use of health care for preoperative education or pain management, which may increase the burden of costs. However, reducing long-term COU may reduce the use of health care for COU- and OUD-related complications, leading to cost savings. This study evaluated whether the TPS intervention influenced health care use and cost for inpatient, outpatient, or pharmacy services during the year following orthopedic joint surgery compared with that of the standard pain management care for procedures that place patients at high risk for COU. We used a difference-in-differences (DID) analysis to estimate this intervention effect, using multivariable regression models that controlled for unobserved time trends and cohort characteristics.

METHODS

This was a quasi-experimental study of patients who underwent orthopedic joint surgery and associated procedures at high risk for COU at the Veterans Affairs Salt Lake City Healthcare System (VASLCHS) between January 2016 through April 2020. The pre-TPS period between January 2016 through December 2017 was compared with the post-TPS period between January 2018 to September 2019. The control patient cohort was selected from 5 geographically diverse VA health care systems throughout the US: Eastern Colorado, Central Plains (Nebraska), White River Junction (Vermont), North Florida/South Georgia, and Portland (Oregon). By sampling health care costs from VA medical centers (VAMCs) across these different regions, our control group was generalizable to veterans receiving orthopedic joint surgery across the US. This study used data from the US Department of Veterans Affairs (VA) Corporate Data Warehouse, a repository of nearly all clinical and administrative data found in electronic health records for VA-provided care and fee-basis care paid for by the VA.¹⁰ All data were hosted and analyzed in the VA Informatics and Computing Infrastructure (VINCI) workspace. The University of Utah Institutional Review Board and the VASLCHS Office of Research and Development approved the protocol for this study.

TPS Intervention

The VASLCHS TPS has already been described in detail elsewhere.^6,7 Briefly, patients at high risk for COU at the VASLCHS were enrolled in the TPS program before surgery for total knee, hip, or shoulder arthroplasty or rotator cuff procedures. The TPS service consists of an anesthesiologist and advanced practice clinician with specialization in acute pain management and opioid tapering, a psychologist with expertise in cognitive behavioral therapy, and 3 nurse care coordinators. These TPS practitioners work together to provide preoperative education, including setting expectations regarding postoperative pain, recommending nonopioid pain management strategies, and providing guidance regarding the appropriate use of opioids for surgical pain. Individual pain plans were developed and implemented for the perioperative period. After surgery, the TPS provided recommendations and support for nonopioid pain therapies and opioid tapers. Patients were followed by the TPS team for at least 12 months after surgery. At a minimum, the goals set by TPS included cessation of all opioid use for prior nonopioid users (NOU) by 90 days after surgery and the return to baseline opioid use or lower for prior COU patients by 90 days after surgery. The TPS also encouraged and supported opioid tapering among COU patients to reduce or completely stop opioid use after surgery.

Patient Cohorts

Veterans having primary or revision total knee, hip, or shoulder arthroplasty or rotator cuff repair between January 1, 2016, and September 30, 2019, at the aforementioned VAMCs were included in the study. Patients who had any hospitalization within 90 days pre- or postindex surgery or who died within 8 months after surgery were excluded from analysis. Patients who had multiple surgeries during the index inpatient visit or within 90 days after the index surgery also were excluded. Comorbid conditions for mental health and substance use were identified using the International Classification of Diseases, 10th revision Clinical Modification (ICD-10) codes or 9th revision equivalent grouped by Clinical Classifications Software Refined (CCS-R).¹¹ Preoperative exposure to clinically relevant pharmacotherapy (ie, agents associated with prolonged opioid use and nonopioid adjuvants) was captured using VA outpatient prescription records (eAppendix 1).

 

The study patient cohort was stratified into either NOU or COU groups based on opioid use before surgery. Preoperative COU was defined as > 25% nonzero days (calculated using the cabinet supply method) in the 180 days before surgery admit date time (> 45 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released within 90 days before surgery date.^12,13 For NOU patients, the postoperative outcome of interest was new postoperative prolonged opioid use. This was defined by Page and colleagues as ≥ 1 opioid prescription released between surgery discharge date and postdischarge day 44, ≥ 1 opioid prescription released between postdischarge day 45 and 89, and ≥ 1 opioid prescription released between postdischarge day 90 and 180.¹⁴ For COU patients at the time of surgery, the postoperative outcome measure of interest was continued COU, defined as > 25% nonzero days between postdischarge day 90 and 240 (> 37 nonzero days) or ≥ 1 opioid prescription for ≥ 28-day supply released between postdischarge day 90 and 180.

Outcome Variables

Outcome variables included health care use and costs during 1-year pre- and postperiods from the date of surgery. VA health care costs for outpatient, inpatient, and pharmacy services for direct patient care were collected from the Managerial Cost Accounting System, an activity-based cost allocation system that generates estimates of the cost of individual VA hospital stays, health care encounters, and medications. Health care use was defined as the number of encounters for each visit type in the Managerial Cost Accounting System. All costs were adjusted to 2019 US dollars, using the Personal Consumption Expenditures price index for health care services.¹⁵

A set of sociodemographic variables including sex, age at surgery, race and ethnicity, rurality, military branch (Army, Air Force, Marine Corps, Navy, and other), and service connectivity were included as covariates in our regression models. In addition, ICD-9 and ICD-10 codes were used to define the indicators of whether veterans had a diagnosis of mental illness (including anxiety, bipolar disorder, depression, or trauma) or substance use (including alcohol, cannabis, opioids, or tobacco). Finally, pharmacy records were used to create indicators for prescribed opioid-relevant pharmacotherapy (including antidepressants, benzodiazepines [BZD], gabapentinoids, muscle relaxants, non-BZD sedative hypnotics) and active antidepressant drug use during the 1-year preindex period.

Statistical Analyses

Descriptive analyses were used to evaluate differences in baseline patient sociodemographic and clinical characteristics between pre- and postperiods for TPS intervention and control cohorts using 2-sample t tests for continuous variables and χ² tests for categorical variables. We summarized unadjusted health care use and costs for outpatient, inpatient, and pharmacy visits and compared the pre- and postintervention periods using the Mann-Whitney test. Both mean (SD) and median (IQR) were considered, reflecting the skewed distribution of the outcome variables.

We used a DID approach to assess the intervention effect while minimizing confounding from the nonrandom sample. The DID approach controls for unobserved differences between VAMCs that are related to both the intervention and outcomes while controlling for trends over time that could affect outcomes across clinics. To implement the DID approach, we included 3 key independent variables in our regression models: (1) an indicator for whether the observation occurred in the postintervention period; (2) an indicator for whether the patient was exposed to the TPS intervention; and (3) the interaction between these 2 variables.

For cost outcomes, we used multivariable generalized linear models with a log link and a Poisson or Υ family. We analyzed inpatient costs using a 2-part generalized linear model because only 17% to 20% of patients had ≥ 1 inpatient visit. We used multivariable negative binomial regression for health care use outcomes. Demographic and clinical covariates described earlier were included in the regression models to control for differences in the composition of patient groups and clinics that could lead to confounding bias.

 

RESULTS

Of the 4954 patients included in our study cohort, 3545 (71.6%) were in the NOU group and 1409 (28.4%) were in the COU group. Among the NOU cohort, 361 patients were in the intervention group and 3184 in the control group. Among the COU cohort, 149 patients were in the intervention group and 1260 in the control group (Table 1). Most patients were male, White race, with a mean (SD) age of 64 (11) years. The most common orthopedic procedure was total knee arthroplasty, followed by total hip arthroplasty. Among both NOU and COU cohorts, patients’ characteristics were similar between the pre- and postintervention period among either TPS or control cohort.

Figures 1 and 2 and eAppendix 2 depict unadjusted per-person average outpatient, inpatient, and pharmacy visits and costs incurred during the 1-year pre- and postintervention periods for the NOU and COU cohorts. Average total health care follow-up costs ranged from $40,000 to $53,000 for NOU and from $47,000 to $82,000 for COU cohort. Cost for outpatient visits accounted for about 70% of the average total costs, followed by costs for inpatient visits of about 20%, and costs for pharmacy for the remaining.

For the NOU cohort, the number of health care encounters remained fairly stable between periods except for the outpatient visits among the TPS group. The TPS group experienced an increase in mean outpatient visits in the postperiod: 30 vs 37 visits (23%) (P < .001). Mean outpatient and inpatient costs in the pre- and postintervention periods were not significantly different for either the TPS or control groups. Similarly, within the COU cohort, the TPS group experienced a 27% increase in the mean number of outpatient visits (41 vs 52 visits; P = .02) and a 39% increase in mean outpatient costs in the postintervention compared with the preintervention period ($44,682 vs $61,890; P = .02). Although the mean number of outpatient visits for control group remained at a similar level, average outpatient costs increased roughly 13% ($31,068 vs $35,148; P = .01) between the pre- and postintervention periods.

Table 2 summarizes the results from the multivariable DID analyses for the outpatient, inpatient, and pharmacy visit and cost outcomes. Here, the estimated effect of the TPS intervention is the coefficient from the interaction between the postintervention and TPS exposure indicator variables. This coefficient was calculated as the difference in the outcome before and after the TPS intervention among the TPS group minus the difference in the outcome before and after the TPS intervention among the control group. For the NOU cohort, TPS was associated with an increase in the use of outpatient health care (mean [SD] increase of 6.9 [2] visits; P < .001) after the surgery with no statistically significant effect on outpatient costs (mean [SD] increase of $2787 [$3749]; P = .55). There was no statistically significant effect of TPS on the use of inpatient visits or pharmacy, but a decrease in costs for inpatient visits among those who had at least 1 inpatient visit (mean [SD] decrease of $12,170 [$6100]; P = .02). For the COU cohort, TPS had no statistically significant impact on the use of outpatient, inpatient, or pharmacy or the corresponding costs.

 

DISCUSSION

TPS is a multidisciplinary approach to perioperative pain management that has been shown to reduce both the quantity and duration of opioid use among orthopedic surgery patients.^6,7 Although the cost burden of providing TPS services to prevent COU is borne by the individual health care system, it is unclear whether this expense is offset by lower long-term medical costs and health care use for COU- and OUD-related complications. In this study focused on a veteran population undergoing orthopedic joint procedures, a DID analysis of cost and health care use showed that TPS, which has been shown to reduce COU for high-risk surgical patients, can be implemented without increasing the overall costs to the VA health care system during the 1 year following surgery, even with increased outpatient visits. For NOU patients, there was no difference in outpatient visit costs or pharmacy costs over 12 months after surgery, although there was a significant reduction in subsequent inpatient costs over the same period. Further, there was no difference in outpatient, inpatient, or pharmacy costs after surgery for COU patients. These findings suggest that TPS can be a cost-effective approach to reduce opioid use among patients undergoing orthopedic joint surgery in VAMCs.

The costs of managing COU after surgery are substantial. Prior reports have shown that adjusted total health care costs are 1.6 to 2.5 times higher for previously NOU patients with new COU after major surgery than those for such patients without persistent use.¹⁶ The 1-year costs associated with new COU in this prior study ranged between $7944 and $17,702 after inpatient surgery and between $5598 and $12,834 after outpatient index surgery, depending on the payer, which are in line with the cost differences found in our current study. Another report among patients with COU following orthopedic joint replacement showed that they had higher use of inpatient, emergency department, and ambulance/paramedic services in the 12 months following their surgery than did those without persistent use.¹⁷ Although these results highlight the impact that COU plays in driving increased costs after major surgery, there have been limited studies focused on interventions that can neutralize the costs associated with opioid misuse after surgery. To our knowledge, our study is the first analysis to show the impact of using an intervention such as TPS to reduce postoperative opioid use on health care use and cost.

Although a rigorous and comprehensive return on investment analysis was beyond the scope of this analysis, these results may have several implications for other health care systems and hospitals that wish to invest in a multidisciplinary perioperative pain management program such as TPS but may be reluctant due to the upfront investment. First, the increased number of patient follow-up visits needed during TPS seems to be more than offset by the reduction in opioid use and associated complications that may occur after surgery. Second, TPS did not seem to be associated with an increase in overall health care costs during the 1-year follow-up period. Together, these results indicate that the return on investment for a TPS approach to perioperative pain management in which optimal patient-centered outcomes are achieved without increasing long-term costs to a health care system may be positive.

Limitations

This study has several limitations. First, this was a quasi-experimental observational study, and the associations we identified between intervention and outcomes should not be assumed to demonstrate causality. Although our DID analysis controlled for an array of demographic and clinical characteristics, differences in medical costs and health care use between the 2 cohorts might be driven by unobserved confounding variables.

Our study also was limited to veterans who received medical care at the VA, and results may not be generalizable to other non-VA health care systems or to veterans with Medicare insurance who have dual benefits. While our finding on health care use and costs may be incomplete because of the uncaptured health care use outside the VA, our DID analysis helped reduce unobserved bias because the absence of data outside of VA care applies to both TPS and control groups. Further, the total costs of operating a TPS program at any given institution will depend on the size of the hospital and volume of surgical patients who meet criteria for enrollment. However, the relative differences in health care use and costs may be extrapolated to patients undergoing orthopedic surgery in other types of academic and community-based health care systems.

Furthermore, this analysis focused primarily on COU and NOU patients undergoing orthopedic joint surgery. While this represents a high-risk population for OUD, the costs and health care use associated with delivering the TPS intervention to other types of surgical procedures may be significantly different. All costs in this analysis were based on 2019 estimates and do not account for the potential inflation over the past several years. Nonmonetary costs to the patient and per-person average total intervention costs were not included in the study. However, we assumed that costs associated with TPS and standard of care would have increased to an equivalent degree over the same period. Further, the average cost of TPS per patient (approximately $900) is relatively small compared with the average annual costs during 1-year pre- and postoperative periods and was not expected to have a significant effect on the analysis.

Conclusions

We found that the significant reduction in COU seen in previous studies following the implementation of TPS was not accompanied by increased health care costs.^6,7 When considering the other costs of long-term opioid use, such as abuse potential, overdose, death, and increased disability, implementation of a TPS service has the potential to improve patient quality of life while reducing other health-related costs. Health care systems should consider the implementation of similar multidisciplinary approaches to perioperative pain management to improve outcomes after orthopedic joint surgery and other high-risk procedures.

The importance of formalized antimicrobial stewardship programs (ASPs) has gained recognition over the past 2 decades. The increasing requirements for ASP programs from national entities often outpace the staffing, technology, and analytic support needed to meet these demands.^1,2 A multimodal approach to stewardship that includes education initiatives, audit-and-feedback methodology, and system support is effective in producing sustained change.³ However, this approach is resource intensive, and many ASPs must look outward for additional support.

Centralized ASP collaboratives and stewardship networks have been effective in positively impacting initiatives and outcomes through resource sharing.^3-5 These collaboratives can take on multiple forms ranging from centralized education distribution to individual sites coming together to set goals and develop strategies to address common issues.^5-8 Collaboratives can provide enhanced data analysis through data pooling, which may lead to shared dashboards or antibiotic use (AU) reports, allowing for robust benchmarking.^5-7 Productivity at individual centers is often measured by AU and antimicrobial resistance (AMR) rates, but these measures alone do not fully capture the benefits of collaborative participation.

The US Department of Veterans Affairs (VA), similar to other large health care systems, is uniquely positioned to promote the development of ASP collaboratives due to the use of the same electronic health record system and infrastructure for data. This centralized data lends itself more readily to data dashboards and interfacility comparison. In turn, the identification of facilities that have outlying data for specific measures can lead to a collaborative effort to identify aberrant processes or facility-specific problems and identify, implement, and track the progress of appropriate solutions with less effort and resources.⁷ The VA has a national stewardship group, the Antimicrobial Stewardship Task Force (ASTF), that identifies and disseminates best practices and advocates for stewardship resources.

VA facilities are heterogeneous with regard to patient population, services, availability of specialists, and antibiotic resistance patterns.⁹ Therefore, clinical practice and needs vary. The ASTF has spearheaded the development of regional collaboratives, recognizing the potential benefit of smaller groups with shared leadership. The Veterans Integrated Services Networks (VISNs) are geographically demarcated regions that lend themselves well to coordination among member facilities due to similar populations, challenges, and opportunities. The Veterans Affairs Midsouth Healthcare Network (VISN 9) includes 5 facilities across Tennessee, Kentucky, Mississippi, Arkansas, Georgia, Virginia, and Indiana and serves about 293,000 veterans, ranging from 35,000 to 105,000 per facility.

A VISN 9 stewardship collaborative (as described by Buckel and colleagues in 2022) was established to enhance member facility ASPs through shared goal setting.⁶ Initially, the collaborative met quarterly; however, with increased participation and the onset of COVID-19, the collaborative evolved to meet burgeoning ASP needs. While intrafacility multidisciplinary ASP collaboration has been previously published, few publications on interfacility collaborations exist.^3-6 To our knowledge, no previous publications have reported the impact of a VA ASP collaborative on the productivity and effectiveness of participating ASP facilities and the region. We aim to share the structure and processes of this ASP collaborative, demonstrate its impact through quantification of productivity, and aid others in developing similar collaboratives to further ASPs’ impact.

 

Methods

The regional VISN 9 ASP collaborative was formed in January 2020 to address common issues across facilities and optimize human capital and resources. The initial collaborative included ASP pharmacists but quickly expanded to include physicians and nurse practitioners. The collaborative is co-led by 2 members from different facilities that rotate.

In April 2021, clinical guidance and research/quality improvement (QI) subcommittees were created. The monthly research/QI subcommittee discusses current initiatives and barriers to ongoing research, adapt and disseminate successful interventions to other facilities, and develop new collaborative initiatives. The clinical guidance subcommittee creates and disseminates clinical expert recommendations regarding common issues or emerging needs.

Data Plan and Collection

To measure success and growth, we evaluated annual facility reports that convey the state of each facility’s ASP, outline its current initiatives and progress, highlight areas of need, and set a programmatic goal and strategy for the upcoming year. These reports, required by a VA directive, are submitted annually by each facility to local and VISN leadership and must address the following 7 areas: (1) ASP structure and fulfillment of national VA policy for ASP; (2) fulfillment of the Joint Commission ASP standards; (3) ASP metrics; (4) ASP activities and interventions; (5) ASP QI and research initiatives; (6) education; and (7) goals and priorities.

To standardize evaluation and accurately reflect ASP effort across heterogeneous reports, 4 core areas were identified from areas 1, 3, 4 and 5 listed previously. Area 2 was excluded for its similarity among all facilities, and areas 6 and 7 were excluded for significant differences in definitions and reporting across facilities.

The project team consisted of 5 members from the collaborative who initially discussed definitions and annual report review methodology. A subgroup was assigned to area 1 and another to areas 3, 4, and 5 for initial review and data extraction. Results were later reviewed to address discrepancies and finalize collation and presentation. 
The impact of the collaborative on individual facilities was measured by both quantitative and qualitative measures. Quantitative measures included: (1) designated ASP pharmacy, physician, or advanced practice provider (APP) full-time equivalents (FTE) at each facility compared with the recommended FTE for facility size; (2) the number of inpatient and outpatient ASP AU metrics for each facility and the VISN total; (3) reported improvement in annual ASP metrics calculated as frequency of improved metrics for each facility and the VISN; (4) the number of QI or research initiatives for each facility and the VISN, which included clinical pathways and order sets; and (5) the number of initiatives published as either abstract or manuscript.¹⁰ Additionally, the number of collaborative efforts involving more than 1 facility was tracked. Qualitative data included categories of metrics and QI and research initiatives. Data were collected by year and facility. Facilities are labeled A to E throughout this article.

Along with facility annual ASP reports, facility and VISN AU trends for fiscal years (FY) 2019-2022 were collected from existing VA dashboards tracking AU in both acute respiratory infections (ARI) and in patients with COVID-19. Quantitative data included facility and VISN quarterly AU rates for ARI, extracted from the national VA dashboard. Facility and VISN AU rates in patients with COVID-19 were extracted from a dashboard developed by the VISN 9 ASP collaborative. The VISN 9 Institutional Review Board deemed this work QI and approval was waived.

 

Results

In 2019, only 2 sites (A and C) reported dedicated FTE compared with recommended minimum staffing; neither met minimum requirements. In 2020, 1 facility (B) met the physician FTE recommendation, and 2 facilities met the pharmacy minimum FTE (D and E). In 2021 and 2022, 2 of 5 facilities (B and E) met the physician minimum FTE, and 2 of 5 (D and E) met the minimum pharmacy FTE recommendations. For the study years 2019 to 2022, 1 facility (E) met both pharmacy and physician FTE recommendations in 2021 and 2022, and 2 facilities (A and C) never met minimum FTE recommendations.

Regarding ASP metrics, all facilities tracked and reported inpatient AU; however, facility A did not document inpatient metrics for FY 2021. The number of individual inpatient metrics varied annually; however, FY 2022 saw the highest reported for the VISN (n = 40), with a more even distribution across facilities (Figure 1). Common metrics in 2022 included total AU, broad-spectrum gram-negative AU, anti–methicillin-resistant Staphylococcus aureus (MRSA) agent use, antibiotics with high risk for Clostridioides difficile infection (CDI), and AU in patients with COVID-19. The percentage of improved metrics for VISN 9 was consistent, ranging from 26.5% to 34.8%, throughout the study period.

From 2019 to 2022, facilities reporting outpatient AU increased from 3 to 5 and included fluoroquinolone use and AU in ARI. VISN 9 outpatient metrics increased every year except in 2021 with improved distribution across facilities. The number of total metrics with reported improvement in the outpatient setting overall increased from 3 of 11 (27%) in 2019 to 20 of 33 (60%) in 2022.

Antimicrobial Stewardship Initiatives

Quantitative and qualitative data regarding initiatives are reported in Figure 2 and the eAppendix respectively. Since the formation of the collaborative, total initiatives increased from 33 in 2019 to 41 in 2022. In 2019, before the collaborative, individual facilities were working on similar projects in parallel, which included MRSA decolonization (A and C), surgical prophylaxis (A and E), asymptomatic bacteriuria (A and C), and CDI (B, C, D, and E). The development of clinical pathways and order sets remained consistent, ranging from 15 to 19 throughout the study period except for 2020, when 33 clinical pathways and/or order sets were developed. Collaboration between sites also remained consistent, with 1 shared clinical pathway and/or order menu between at least 1 site reported yearly for 2020, 2021, and 2022. The number of publications from VISN 9 grew from 2 in 2019 to 17 in 2022. In 2019, there were no collaborative research or QI publications, but in 2022 there were 2 joint publications, 1 between 2 facilities (A and C) and 1 including all facilities.

ARI and COVID-19 were identified by the collaborative as VISN priorities, leading to shared metrics and benchmarking across facilities. From 2019 to 2022, increased collaboration on these initiatives was noted at all facilities. The ARI goal was established to reduce inappropriate prescribing for ARI/bronchitis to under 20% across VISN 9. Rates dropped from 50.3% (range, 35.4%-77.6%) in FY 2019 quarter (Q) 1 to 15% (range, 8%-18.3%) in FY 2022 Q4. The clinical guidance subcommittee developed a guideline for AU in patients with COVID-19 that was approved by the VISN 9 Pharmacy & Therapeutics Committee. A VISN 9 dashboard was developed to track inpatient and outpatient AU for COVID-19. Antibiotic prescribing in the first 4 days of hospitalization decreased from 62.2% at the start of the COVID-19 pandemic to 48.7% after dissemination of COVID-19 guidance.

 

Discussion

This study demonstrates the benefit of participating in a regional ASP collaborative for individual facilities and the region. Some products from the collaborative include the development of regionwide guidance for the use of antimicrobials in COVID-19, interfacility collaborative initiatives, a COVID-19 dashboard, improvement in metrics, and several publications. Importantly, this expansion occurred during the COVID-19 pandemic when many ASP members were spread thin. Moreover, despite 4 sites not meeting VA-recommended ASP staffing requirements for both pharmacists and physicians, productivity increased within the VISN as facilities worked together sharing local challenges and successful paths in removing ASP barriers. The collaborative shared QI strategies, advocated for technological support (ie, Theradoc and dashboards) to maximize available ASP human capital, standardized metric reporting, and made continued efforts sustainable. VA ASTF disseminates evidence-based practice but is not designed to develop tailored site-specific interventions, which has led to the support of VISN-level collaboratives to serve local facilities’ needs. We postulate the use of a collaborative as a beneficial strategy to increase productivity and achieve local goals with limited resources.

Previous reports in the literature have found ASP collaboratives to be an effective model for long-term program growth.³ Two collaboratives found improved adherence to the Centers for Disease Control and Prevention core elements for ASP.^4,5 Similar to our findings, other collaboratives noted a reduction in AU after implementation, although statistical analysis of improvement over time was not performed to verify significance.^3-5,7 One VA study reviewed the use of dashboards with a monthly learning collaborative and identified a reduction in AU.⁷ However, the structure of our ASP collaborative was through joint meetings and projects, as defined by Buckel and colleagues.⁶

Our findings highlight that ASP collaboratives can help answer the recent call to action from McGregor, Fitzpatrick, and Suda who advocated for ASPs to take the next steps in stewardship, which include standardization of evaluating metrics and the use of robust QI frameworks.¹¹ Moving forward, an area for research could include a comparison of ASP collaborative infrastructures and productivity to identify optimal fit dependent on facility structure and setting. Parallel to our experience, other reports cite heterogeneous ASP metrics and a lack of benchmarking, spotlighting the need for standardization.^8,11,12 The VA and other health care facilities would benefit from national benchmarking of AU metrics to make comparisons across sites beneficial.

Limitations

Using annual reports was a limitation for analyzing and reporting the full impact of the collaborative. Local facility-level discretion of content inclusion led to many facilities only reporting on the forefront of new initiatives that they had developed and may have led to the omission of other ongoing work. Further, time invested into the ASP regional collaborative was not captured within annual reports; therefore, the opportunity cost cannot be determined.

Conclusions

The VA has an advantage that many private health care facilities do not: the ability to work across systems to ease the burden of duplicative work and more readily disseminate effective strategies. The regional ASP collaborative bred innovation and the tearing down of silos. The implementation of the collaborative aided in robust QI infrastructure, standardization of reporting and metrics, and greater support through facility alignments with regional guidance. ASP interfacility collaboratives provide a sustainable solution in a resource-limited landscape.

Acknowledgments

This work was made possible by the resources provided through the Antimicrobial Stewardship Programs in the Veterans Integrated Services Network (VISN) 9.

The importance of formalized antimicrobial stewardship programs (ASPs) has gained recognition over the past 2 decades. The increasing requirements for ASP programs from national entities often outpace the staffing, technology, and analytic support needed to meet these demands.^1,2 A multimodal approach to stewardship that includes education initiatives, audit-and-feedback methodology, and system support is effective in producing sustained change.³ However, this approach is resource intensive, and many ASPs must look outward for additional support.

Centralized ASP collaboratives and stewardship networks have been effective in positively impacting initiatives and outcomes through resource sharing.^3-5 These collaboratives can take on multiple forms ranging from centralized education distribution to individual sites coming together to set goals and develop strategies to address common issues.^5-8 Collaboratives can provide enhanced data analysis through data pooling, which may lead to shared dashboards or antibiotic use (AU) reports, allowing for robust benchmarking.^5-7 Productivity at individual centers is often measured by AU and antimicrobial resistance (AMR) rates, but these measures alone do not fully capture the benefits of collaborative participation.

The US Department of Veterans Affairs (VA), similar to other large health care systems, is uniquely positioned to promote the development of ASP collaboratives due to the use of the same electronic health record system and infrastructure for data. This centralized data lends itself more readily to data dashboards and interfacility comparison. In turn, the identification of facilities that have outlying data for specific measures can lead to a collaborative effort to identify aberrant processes or facility-specific problems and identify, implement, and track the progress of appropriate solutions with less effort and resources.⁷ The VA has a national stewardship group, the Antimicrobial Stewardship Task Force (ASTF), that identifies and disseminates best practices and advocates for stewardship resources.

VA facilities are heterogeneous with regard to patient population, services, availability of specialists, and antibiotic resistance patterns.⁹ Therefore, clinical practice and needs vary. The ASTF has spearheaded the development of regional collaboratives, recognizing the potential benefit of smaller groups with shared leadership. The Veterans Integrated Services Networks (VISNs) are geographically demarcated regions that lend themselves well to coordination among member facilities due to similar populations, challenges, and opportunities. The Veterans Affairs Midsouth Healthcare Network (VISN 9) includes 5 facilities across Tennessee, Kentucky, Mississippi, Arkansas, Georgia, Virginia, and Indiana and serves about 293,000 veterans, ranging from 35,000 to 105,000 per facility.

A VISN 9 stewardship collaborative (as described by Buckel and colleagues in 2022) was established to enhance member facility ASPs through shared goal setting.⁶ Initially, the collaborative met quarterly; however, with increased participation and the onset of COVID-19, the collaborative evolved to meet burgeoning ASP needs. While intrafacility multidisciplinary ASP collaboration has been previously published, few publications on interfacility collaborations exist.^3-6 To our knowledge, no previous publications have reported the impact of a VA ASP collaborative on the productivity and effectiveness of participating ASP facilities and the region. We aim to share the structure and processes of this ASP collaborative, demonstrate its impact through quantification of productivity, and aid others in developing similar collaboratives to further ASPs’ impact.

 

Methods

The regional VISN 9 ASP collaborative was formed in January 2020 to address common issues across facilities and optimize human capital and resources. The initial collaborative included ASP pharmacists but quickly expanded to include physicians and nurse practitioners. The collaborative is co-led by 2 members from different facilities that rotate.

In April 2021, clinical guidance and research/quality improvement (QI) subcommittees were created. The monthly research/QI subcommittee discusses current initiatives and barriers to ongoing research, adapt and disseminate successful interventions to other facilities, and develop new collaborative initiatives. The clinical guidance subcommittee creates and disseminates clinical expert recommendations regarding common issues or emerging needs.

Data Plan and Collection

To measure success and growth, we evaluated annual facility reports that convey the state of each facility’s ASP, outline its current initiatives and progress, highlight areas of need, and set a programmatic goal and strategy for the upcoming year. These reports, required by a VA directive, are submitted annually by each facility to local and VISN leadership and must address the following 7 areas: (1) ASP structure and fulfillment of national VA policy for ASP; (2) fulfillment of the Joint Commission ASP standards; (3) ASP metrics; (4) ASP activities and interventions; (5) ASP QI and research initiatives; (6) education; and (7) goals and priorities.

To standardize evaluation and accurately reflect ASP effort across heterogeneous reports, 4 core areas were identified from areas 1, 3, 4 and 5 listed previously. Area 2 was excluded for its similarity among all facilities, and areas 6 and 7 were excluded for significant differences in definitions and reporting across facilities.

The project team consisted of 5 members from the collaborative who initially discussed definitions and annual report review methodology. A subgroup was assigned to area 1 and another to areas 3, 4, and 5 for initial review and data extraction. Results were later reviewed to address discrepancies and finalize collation and presentation. 
The impact of the collaborative on individual facilities was measured by both quantitative and qualitative measures. Quantitative measures included: (1) designated ASP pharmacy, physician, or advanced practice provider (APP) full-time equivalents (FTE) at each facility compared with the recommended FTE for facility size; (2) the number of inpatient and outpatient ASP AU metrics for each facility and the VISN total; (3) reported improvement in annual ASP metrics calculated as frequency of improved metrics for each facility and the VISN; (4) the number of QI or research initiatives for each facility and the VISN, which included clinical pathways and order sets; and (5) the number of initiatives published as either abstract or manuscript.¹⁰ Additionally, the number of collaborative efforts involving more than 1 facility was tracked. Qualitative data included categories of metrics and QI and research initiatives. Data were collected by year and facility. Facilities are labeled A to E throughout this article.

Along with facility annual ASP reports, facility and VISN AU trends for fiscal years (FY) 2019-2022 were collected from existing VA dashboards tracking AU in both acute respiratory infections (ARI) and in patients with COVID-19. Quantitative data included facility and VISN quarterly AU rates for ARI, extracted from the national VA dashboard. Facility and VISN AU rates in patients with COVID-19 were extracted from a dashboard developed by the VISN 9 ASP collaborative. The VISN 9 Institutional Review Board deemed this work QI and approval was waived.

 

Results

In 2019, only 2 sites (A and C) reported dedicated FTE compared with recommended minimum staffing; neither met minimum requirements. In 2020, 1 facility (B) met the physician FTE recommendation, and 2 facilities met the pharmacy minimum FTE (D and E). In 2021 and 2022, 2 of 5 facilities (B and E) met the physician minimum FTE, and 2 of 5 (D and E) met the minimum pharmacy FTE recommendations. For the study years 2019 to 2022, 1 facility (E) met both pharmacy and physician FTE recommendations in 2021 and 2022, and 2 facilities (A and C) never met minimum FTE recommendations.

Regarding ASP metrics, all facilities tracked and reported inpatient AU; however, facility A did not document inpatient metrics for FY 2021. The number of individual inpatient metrics varied annually; however, FY 2022 saw the highest reported for the VISN (n = 40), with a more even distribution across facilities (Figure 1). Common metrics in 2022 included total AU, broad-spectrum gram-negative AU, anti–methicillin-resistant Staphylococcus aureus (MRSA) agent use, antibiotics with high risk for Clostridioides difficile infection (CDI), and AU in patients with COVID-19. The percentage of improved metrics for VISN 9 was consistent, ranging from 26.5% to 34.8%, throughout the study period.

From 2019 to 2022, facilities reporting outpatient AU increased from 3 to 5 and included fluoroquinolone use and AU in ARI. VISN 9 outpatient metrics increased every year except in 2021 with improved distribution across facilities. The number of total metrics with reported improvement in the outpatient setting overall increased from 3 of 11 (27%) in 2019 to 20 of 33 (60%) in 2022.

Antimicrobial Stewardship Initiatives

Quantitative and qualitative data regarding initiatives are reported in Figure 2 and the eAppendix respectively. Since the formation of the collaborative, total initiatives increased from 33 in 2019 to 41 in 2022. In 2019, before the collaborative, individual facilities were working on similar projects in parallel, which included MRSA decolonization (A and C), surgical prophylaxis (A and E), asymptomatic bacteriuria (A and C), and CDI (B, C, D, and E). The development of clinical pathways and order sets remained consistent, ranging from 15 to 19 throughout the study period except for 2020, when 33 clinical pathways and/or order sets were developed. Collaboration between sites also remained consistent, with 1 shared clinical pathway and/or order menu between at least 1 site reported yearly for 2020, 2021, and 2022. The number of publications from VISN 9 grew from 2 in 2019 to 17 in 2022. In 2019, there were no collaborative research or QI publications, but in 2022 there were 2 joint publications, 1 between 2 facilities (A and C) and 1 including all facilities.

ARI and COVID-19 were identified by the collaborative as VISN priorities, leading to shared metrics and benchmarking across facilities. From 2019 to 2022, increased collaboration on these initiatives was noted at all facilities. The ARI goal was established to reduce inappropriate prescribing for ARI/bronchitis to under 20% across VISN 9. Rates dropped from 50.3% (range, 35.4%-77.6%) in FY 2019 quarter (Q) 1 to 15% (range, 8%-18.3%) in FY 2022 Q4. The clinical guidance subcommittee developed a guideline for AU in patients with COVID-19 that was approved by the VISN 9 Pharmacy & Therapeutics Committee. A VISN 9 dashboard was developed to track inpatient and outpatient AU for COVID-19. Antibiotic prescribing in the first 4 days of hospitalization decreased from 62.2% at the start of the COVID-19 pandemic to 48.7% after dissemination of COVID-19 guidance.

 

Discussion

This study demonstrates the benefit of participating in a regional ASP collaborative for individual facilities and the region. Some products from the collaborative include the development of regionwide guidance for the use of antimicrobials in COVID-19, interfacility collaborative initiatives, a COVID-19 dashboard, improvement in metrics, and several publications. Importantly, this expansion occurred during the COVID-19 pandemic when many ASP members were spread thin. Moreover, despite 4 sites not meeting VA-recommended ASP staffing requirements for both pharmacists and physicians, productivity increased within the VISN as facilities worked together sharing local challenges and successful paths in removing ASP barriers. The collaborative shared QI strategies, advocated for technological support (ie, Theradoc and dashboards) to maximize available ASP human capital, standardized metric reporting, and made continued efforts sustainable. VA ASTF disseminates evidence-based practice but is not designed to develop tailored site-specific interventions, which has led to the support of VISN-level collaboratives to serve local facilities’ needs. We postulate the use of a collaborative as a beneficial strategy to increase productivity and achieve local goals with limited resources.

Previous reports in the literature have found ASP collaboratives to be an effective model for long-term program growth.³ Two collaboratives found improved adherence to the Centers for Disease Control and Prevention core elements for ASP.^4,5 Similar to our findings, other collaboratives noted a reduction in AU after implementation, although statistical analysis of improvement over time was not performed to verify significance.^3-5,7 One VA study reviewed the use of dashboards with a monthly learning collaborative and identified a reduction in AU.⁷ However, the structure of our ASP collaborative was through joint meetings and projects, as defined by Buckel and colleagues.⁶

Our findings highlight that ASP collaboratives can help answer the recent call to action from McGregor, Fitzpatrick, and Suda who advocated for ASPs to take the next steps in stewardship, which include standardization of evaluating metrics and the use of robust QI frameworks.¹¹ Moving forward, an area for research could include a comparison of ASP collaborative infrastructures and productivity to identify optimal fit dependent on facility structure and setting. Parallel to our experience, other reports cite heterogeneous ASP metrics and a lack of benchmarking, spotlighting the need for standardization.^8,11,12 The VA and other health care facilities would benefit from national benchmarking of AU metrics to make comparisons across sites beneficial.

Limitations

Using annual reports was a limitation for analyzing and reporting the full impact of the collaborative. Local facility-level discretion of content inclusion led to many facilities only reporting on the forefront of new initiatives that they had developed and may have led to the omission of other ongoing work. Further, time invested into the ASP regional collaborative was not captured within annual reports; therefore, the opportunity cost cannot be determined.

Conclusions

The VA has an advantage that many private health care facilities do not: the ability to work across systems to ease the burden of duplicative work and more readily disseminate effective strategies. The regional ASP collaborative bred innovation and the tearing down of silos. The implementation of the collaborative aided in robust QI infrastructure, standardization of reporting and metrics, and greater support through facility alignments with regional guidance. ASP interfacility collaboratives provide a sustainable solution in a resource-limited landscape.

Acknowledgments

This work was made possible by the resources provided through the Antimicrobial Stewardship Programs in the Veterans Integrated Services Network (VISN) 9.

The importance of formalized antimicrobial stewardship programs (ASPs) has gained recognition over the past 2 decades. The increasing requirements for ASP programs from national entities often outpace the staffing, technology, and analytic support needed to meet these demands.^1,2 A multimodal approach to stewardship that includes education initiatives, audit-and-feedback methodology, and system support is effective in producing sustained change.³ However, this approach is resource intensive, and many ASPs must look outward for additional support.

Centralized ASP collaboratives and stewardship networks have been effective in positively impacting initiatives and outcomes through resource sharing.^3-5 These collaboratives can take on multiple forms ranging from centralized education distribution to individual sites coming together to set goals and develop strategies to address common issues.^5-8 Collaboratives can provide enhanced data analysis through data pooling, which may lead to shared dashboards or antibiotic use (AU) reports, allowing for robust benchmarking.^5-7 Productivity at individual centers is often measured by AU and antimicrobial resistance (AMR) rates, but these measures alone do not fully capture the benefits of collaborative participation.

The US Department of Veterans Affairs (VA), similar to other large health care systems, is uniquely positioned to promote the development of ASP collaboratives due to the use of the same electronic health record system and infrastructure for data. This centralized data lends itself more readily to data dashboards and interfacility comparison. In turn, the identification of facilities that have outlying data for specific measures can lead to a collaborative effort to identify aberrant processes or facility-specific problems and identify, implement, and track the progress of appropriate solutions with less effort and resources.⁷ The VA has a national stewardship group, the Antimicrobial Stewardship Task Force (ASTF), that identifies and disseminates best practices and advocates for stewardship resources.

VA facilities are heterogeneous with regard to patient population, services, availability of specialists, and antibiotic resistance patterns.⁹ Therefore, clinical practice and needs vary. The ASTF has spearheaded the development of regional collaboratives, recognizing the potential benefit of smaller groups with shared leadership. The Veterans Integrated Services Networks (VISNs) are geographically demarcated regions that lend themselves well to coordination among member facilities due to similar populations, challenges, and opportunities. The Veterans Affairs Midsouth Healthcare Network (VISN 9) includes 5 facilities across Tennessee, Kentucky, Mississippi, Arkansas, Georgia, Virginia, and Indiana and serves about 293,000 veterans, ranging from 35,000 to 105,000 per facility.

A VISN 9 stewardship collaborative (as described by Buckel and colleagues in 2022) was established to enhance member facility ASPs through shared goal setting.⁶ Initially, the collaborative met quarterly; however, with increased participation and the onset of COVID-19, the collaborative evolved to meet burgeoning ASP needs. While intrafacility multidisciplinary ASP collaboration has been previously published, few publications on interfacility collaborations exist.^3-6 To our knowledge, no previous publications have reported the impact of a VA ASP collaborative on the productivity and effectiveness of participating ASP facilities and the region. We aim to share the structure and processes of this ASP collaborative, demonstrate its impact through quantification of productivity, and aid others in developing similar collaboratives to further ASPs’ impact.

 

Methods

The regional VISN 9 ASP collaborative was formed in January 2020 to address common issues across facilities and optimize human capital and resources. The initial collaborative included ASP pharmacists but quickly expanded to include physicians and nurse practitioners. The collaborative is co-led by 2 members from different facilities that rotate.

In April 2021, clinical guidance and research/quality improvement (QI) subcommittees were created. The monthly research/QI subcommittee discusses current initiatives and barriers to ongoing research, adapt and disseminate successful interventions to other facilities, and develop new collaborative initiatives. The clinical guidance subcommittee creates and disseminates clinical expert recommendations regarding common issues or emerging needs.

Data Plan and Collection

To measure success and growth, we evaluated annual facility reports that convey the state of each facility’s ASP, outline its current initiatives and progress, highlight areas of need, and set a programmatic goal and strategy for the upcoming year. These reports, required by a VA directive, are submitted annually by each facility to local and VISN leadership and must address the following 7 areas: (1) ASP structure and fulfillment of national VA policy for ASP; (2) fulfillment of the Joint Commission ASP standards; (3) ASP metrics; (4) ASP activities and interventions; (5) ASP QI and research initiatives; (6) education; and (7) goals and priorities.

To standardize evaluation and accurately reflect ASP effort across heterogeneous reports, 4 core areas were identified from areas 1, 3, 4 and 5 listed previously. Area 2 was excluded for its similarity among all facilities, and areas 6 and 7 were excluded for significant differences in definitions and reporting across facilities.

The project team consisted of 5 members from the collaborative who initially discussed definitions and annual report review methodology. A subgroup was assigned to area 1 and another to areas 3, 4, and 5 for initial review and data extraction. Results were later reviewed to address discrepancies and finalize collation and presentation. 
The impact of the collaborative on individual facilities was measured by both quantitative and qualitative measures. Quantitative measures included: (1) designated ASP pharmacy, physician, or advanced practice provider (APP) full-time equivalents (FTE) at each facility compared with the recommended FTE for facility size; (2) the number of inpatient and outpatient ASP AU metrics for each facility and the VISN total; (3) reported improvement in annual ASP metrics calculated as frequency of improved metrics for each facility and the VISN; (4) the number of QI or research initiatives for each facility and the VISN, which included clinical pathways and order sets; and (5) the number of initiatives published as either abstract or manuscript.¹⁰ Additionally, the number of collaborative efforts involving more than 1 facility was tracked. Qualitative data included categories of metrics and QI and research initiatives. Data were collected by year and facility. Facilities are labeled A to E throughout this article.

Along with facility annual ASP reports, facility and VISN AU trends for fiscal years (FY) 2019-2022 were collected from existing VA dashboards tracking AU in both acute respiratory infections (ARI) and in patients with COVID-19. Quantitative data included facility and VISN quarterly AU rates for ARI, extracted from the national VA dashboard. Facility and VISN AU rates in patients with COVID-19 were extracted from a dashboard developed by the VISN 9 ASP collaborative. The VISN 9 Institutional Review Board deemed this work QI and approval was waived.

 

Results

In 2019, only 2 sites (A and C) reported dedicated FTE compared with recommended minimum staffing; neither met minimum requirements. In 2020, 1 facility (B) met the physician FTE recommendation, and 2 facilities met the pharmacy minimum FTE (D and E). In 2021 and 2022, 2 of 5 facilities (B and E) met the physician minimum FTE, and 2 of 5 (D and E) met the minimum pharmacy FTE recommendations. For the study years 2019 to 2022, 1 facility (E) met both pharmacy and physician FTE recommendations in 2021 and 2022, and 2 facilities (A and C) never met minimum FTE recommendations.

Regarding ASP metrics, all facilities tracked and reported inpatient AU; however, facility A did not document inpatient metrics for FY 2021. The number of individual inpatient metrics varied annually; however, FY 2022 saw the highest reported for the VISN (n = 40), with a more even distribution across facilities (Figure 1). Common metrics in 2022 included total AU, broad-spectrum gram-negative AU, anti–methicillin-resistant Staphylococcus aureus (MRSA) agent use, antibiotics with high risk for Clostridioides difficile infection (CDI), and AU in patients with COVID-19. The percentage of improved metrics for VISN 9 was consistent, ranging from 26.5% to 34.8%, throughout the study period.

From 2019 to 2022, facilities reporting outpatient AU increased from 3 to 5 and included fluoroquinolone use and AU in ARI. VISN 9 outpatient metrics increased every year except in 2021 with improved distribution across facilities. The number of total metrics with reported improvement in the outpatient setting overall increased from 3 of 11 (27%) in 2019 to 20 of 33 (60%) in 2022.

Antimicrobial Stewardship Initiatives

Quantitative and qualitative data regarding initiatives are reported in Figure 2 and the eAppendix respectively. Since the formation of the collaborative, total initiatives increased from 33 in 2019 to 41 in 2022. In 2019, before the collaborative, individual facilities were working on similar projects in parallel, which included MRSA decolonization (A and C), surgical prophylaxis (A and E), asymptomatic bacteriuria (A and C), and CDI (B, C, D, and E). The development of clinical pathways and order sets remained consistent, ranging from 15 to 19 throughout the study period except for 2020, when 33 clinical pathways and/or order sets were developed. Collaboration between sites also remained consistent, with 1 shared clinical pathway and/or order menu between at least 1 site reported yearly for 2020, 2021, and 2022. The number of publications from VISN 9 grew from 2 in 2019 to 17 in 2022. In 2019, there were no collaborative research or QI publications, but in 2022 there were 2 joint publications, 1 between 2 facilities (A and C) and 1 including all facilities.

ARI and COVID-19 were identified by the collaborative as VISN priorities, leading to shared metrics and benchmarking across facilities. From 2019 to 2022, increased collaboration on these initiatives was noted at all facilities. The ARI goal was established to reduce inappropriate prescribing for ARI/bronchitis to under 20% across VISN 9. Rates dropped from 50.3% (range, 35.4%-77.6%) in FY 2019 quarter (Q) 1 to 15% (range, 8%-18.3%) in FY 2022 Q4. The clinical guidance subcommittee developed a guideline for AU in patients with COVID-19 that was approved by the VISN 9 Pharmacy & Therapeutics Committee. A VISN 9 dashboard was developed to track inpatient and outpatient AU for COVID-19. Antibiotic prescribing in the first 4 days of hospitalization decreased from 62.2% at the start of the COVID-19 pandemic to 48.7% after dissemination of COVID-19 guidance.

 

Discussion

This study demonstrates the benefit of participating in a regional ASP collaborative for individual facilities and the region. Some products from the collaborative include the development of regionwide guidance for the use of antimicrobials in COVID-19, interfacility collaborative initiatives, a COVID-19 dashboard, improvement in metrics, and several publications. Importantly, this expansion occurred during the COVID-19 pandemic when many ASP members were spread thin. Moreover, despite 4 sites not meeting VA-recommended ASP staffing requirements for both pharmacists and physicians, productivity increased within the VISN as facilities worked together sharing local challenges and successful paths in removing ASP barriers. The collaborative shared QI strategies, advocated for technological support (ie, Theradoc and dashboards) to maximize available ASP human capital, standardized metric reporting, and made continued efforts sustainable. VA ASTF disseminates evidence-based practice but is not designed to develop tailored site-specific interventions, which has led to the support of VISN-level collaboratives to serve local facilities’ needs. We postulate the use of a collaborative as a beneficial strategy to increase productivity and achieve local goals with limited resources.

Previous reports in the literature have found ASP collaboratives to be an effective model for long-term program growth.³ Two collaboratives found improved adherence to the Centers for Disease Control and Prevention core elements for ASP.^4,5 Similar to our findings, other collaboratives noted a reduction in AU after implementation, although statistical analysis of improvement over time was not performed to verify significance.^3-5,7 One VA study reviewed the use of dashboards with a monthly learning collaborative and identified a reduction in AU.⁷ However, the structure of our ASP collaborative was through joint meetings and projects, as defined by Buckel and colleagues.⁶

Our findings highlight that ASP collaboratives can help answer the recent call to action from McGregor, Fitzpatrick, and Suda who advocated for ASPs to take the next steps in stewardship, which include standardization of evaluating metrics and the use of robust QI frameworks.¹¹ Moving forward, an area for research could include a comparison of ASP collaborative infrastructures and productivity to identify optimal fit dependent on facility structure and setting. Parallel to our experience, other reports cite heterogeneous ASP metrics and a lack of benchmarking, spotlighting the need for standardization.^8,11,12 The VA and other health care facilities would benefit from national benchmarking of AU metrics to make comparisons across sites beneficial.

Limitations

Using annual reports was a limitation for analyzing and reporting the full impact of the collaborative. Local facility-level discretion of content inclusion led to many facilities only reporting on the forefront of new initiatives that they had developed and may have led to the omission of other ongoing work. Further, time invested into the ASP regional collaborative was not captured within annual reports; therefore, the opportunity cost cannot be determined.

Conclusions

The VA has an advantage that many private health care facilities do not: the ability to work across systems to ease the burden of duplicative work and more readily disseminate effective strategies. The regional ASP collaborative bred innovation and the tearing down of silos. The implementation of the collaborative aided in robust QI infrastructure, standardization of reporting and metrics, and greater support through facility alignments with regional guidance. ASP interfacility collaboratives provide a sustainable solution in a resource-limited landscape.

Acknowledgments

This work was made possible by the resources provided through the Antimicrobial Stewardship Programs in the Veterans Integrated Services Network (VISN) 9.

Chronic kidney disease (CKD) affects nearly 37 million people (11%) in the US and is a leading cause of death and morbidity. Due to their older age and higher prevalence of comorbid conditions, the prevalence of CKD among veterans is approximately 34% higher than in the general population and the fourth most common chronic disease diagnosed among US veterans.^1,2 US veterans and those with prior military service (MS) may be at a particularly high risk for CKD and associated health care outcomes including increased hospitalization and death. The observed excess burden of CKD is not mirrored in the general population, and it is unclear whether prior MS confers a unique risk profile for CKD.

Current estimates of CKD burden among veterans or those with prior MS are widely variable and have been limited by unique regions, specific exposure profiles, or to single health care systems. As such, there remains a paucity of data examining CKD burden more broadly. We performed a study in the adult population of the US to quantify associations with the extent of CKD, enumerate temporal trends of CKD among those with prior MS, describe risk within subgroups, and compare heterogeneity of risk factors for CKD by MS.

Methods

The National Health and Nutrition Examination Survey (NHANES) is a suite of nationally representative, cross-sectional surveys of the noninstitutionalized US population. It is conducted by the National Center for Health Statistics and uses a stratified, clustered probability design, with surveys carried out without interruption, collated, and made accessible to the public at 2-year intervals.³ The survey consists of a questionnaire, physical examination, and laboratory data.

The inclusion criteria for our study were age ≥ 20 years along with serum creatinine and urinary albumin-creatinine measurements. The following definitions were used for the study:

• CKD: Estimated glomerular filtration rate < 60 mL/min/1.73 m² calibrated to isotope dilution mass spectrometry (IDMS).

• Traceable: Creatinine-based CKD Epidemiology Collaboration formula or urinary albumin-creatine ratio ≥ 30 mg/g.

• MS: Positive response to the questions “Did you ever serve in the Armed Forces of the United States?” (1999 to 2010) or “Have you ever served on active duty in the US Armed Forces, military Reserves, or National Guard?” (2011 to 2018).

• Diabetes: Self-reported history, medication for diabetes, or glycated hemoglobin ≥ 7%.

• Hypertension: Blood pressure ≥ 140/90 or ≥ 130/40 mm Hg in the presence of diabetes, medication for hypertension, cardiovascular disease, or CKD, myocardial infarction, cardiac failure, or cerebrovascular disease by self-report.^2,3

Analysis

Primary sampling unit, stratum, and weight variables were employed throughout to generate parameter estimates that are generalizable to the US population.^4,5 The χ² test and logistic regression, respectively, were employed for comparison of proportions and estimation of odds ratios. R Version 4.1.2 was employed for data analysis.

 

 

Results

In the overall sample, the frequencies (95% standard error [SE]) of CKD and prior MS were 15.2% (0.3) and 11.5% (0.3) (Table 1). The proportion (SE) with CKD was significantly higher among those with prior MS vs the overall population: 22.7% (0.7) vs 15.2% (0.3) (P < .001). Significant associations with CKD were observed (P < .05) by age, sex, race and ethnicity, family poverty, school education, health insurance, smoking, body mass index, diabetes, hypertension, cardiovascular disease, and malignancy. Within those reporting prior MS, the proportion (SE) with CKD differed by era: 1999 to 2002, 18.9% (1.1); 2003 to 2006, 24.9% (1.5); 2007 to 2010, 22.3% (1.5); 2011 to 2014, 24.3% (1.7); and 2015 to 2018, 24.0% (1.8) (P = .02) (Figure 1).

Without covariate adjustment, prior MS was significantly associated with an increased risk of CKD (unadjusted odds ratio [OR], 1.78; 95% CI, 1.64-1.93; P < .05) (Table 2). Prior MS was significantly associated with CKD in the following subgroups: 2003 to 2006, 2011 to 2014, 2015 to 2018, age groups of 40 to 64 years and ≥ 65 years, male sex, non-Hispanic White and Hispanic ethnicity, school education of grade 0 to 11, and private or other health insurance. Additional comorbidities strongly associated with CKD included hypertension (OR, 6.37; 95% CI, 5.37-7.55), diabetes (OR, 4.16; 95% CI, 3.45-5.03), and cardiovascular disease (OR, 4.20; 95% CI, 3.57-4.95).

In the population reporting prior MS, the unadjusted OR of CKD vs 1999 to 2002 was greater for all other examined eras; with the greatest likelihood observed for the 2003 to 2006 era. Unadjusted ORs of CKD differed in groups with and without prior MS (P value for interaction < .05) for 2003 to 2006, those aged 40 to 64 years and ≥ 65 years, female sex, non-Hispanic African American and Hispanic race and ethnicity, family poverty, high school education, private health insurance, any smoking history, diabetes, hypertension, and cardiovascular disease (Figure 2A).

Following adjustment for age, sex, and race and ethnicity, MS was associated with a 17% higher likelihood of CKD (adjusted odds ratio [AOR], 1.17; 95% CI, 1.06-1.28; P < .01) (Table 3). Prior MS was significantly associated (P < .05) with CKD in the subgroups: age groups 40 to 64 years and ≥ 65 years, non-Hispanic African American, and body mass index ≥ 30. Among those with prior MS, comorbidities strongly associated with CKD in adjusted models included hypertension (AOR, 3.86; 95% CI, 3.18-4.69), diabetes (AOR, 3.05; 95% CI, 2.44-3.82), and cardiovascular disease (AOR, 2.51; 95% CI, 2.09-3.01). In the population with prior MS, the adjusted likelihood of CKD vs 1999 to 2002 was similar across all eras. Adjusted associations of CKD differed in groups with and without prior MS for age groups 40 to 64 years and ≥ 65 years, female sex, and family poverty (P < .05) (Figure 2B).

Discussion

We observed that prior MS was associated with CKD, all eras were associated with CKD in the subgroup with MS, and risk factors for CKD differed among many subgroups both with and without MS history, a finding that remained present in adjusted models. In addition, the finding of CKD was relatively common among those with prior MS (approximately 15%) and was most strongly associated with increasing age and comorbidities frequently associated with CKD.

Although many studies have demonstrated associations of US veteran status with various comorbidities, including hypertension, obesity, and diabetes, these studies often are limited to those both qualifying and receiving care within the US Department of Veterans Affairs (VA) health care system.^6-9 The crude proportion of individuals reporting multiple chronic conditions, which included hypertension, diabetes, and weak or failing kidneys, was 49.7% for US veterans compared with 24.1% for nonveterans.² Large-scale, nationally representative cohorts for use in this context have been limited by the heterogeneity of definitions of CKD applied with limited timeframes yielding variable estimates.^1,10 Moreover, few studies have examined the clinical epidemiology of CKD more broadly in the US among those with prior MS. For example, a PubMed search on March 3, 2022, with the terms “epidemiology”, “military service”, and “chronic kidney disease” produced only 9 citations, one of which examined trends among a non-US cohort and quantifying disease burden another among adolescents.

Whether or not prior MS confers a unique risk profile for CKD is unknown. While our findings of an increased CKD burden among those reporting MS may partially reflect observed increases in baseline comorbidities, the observed excess CKD among those with MS remained across multiple categories even after adjustment for baseline demography. As several studies have demonstrated, enlistment into MS may select for a more diverse population; however those enlisted personnel may be of lower socioeconomic status and possibly at higher risk of CKD.^11,12 Our findings of important differences in baseline determinants of health mirror this. The proportion of MS respondents with CKD vs CKD alone reporting a high school education or lower was higher (36.0% vs 21.8%) as well as among those with a history of family poverty (21.1% vs 18.0%).

 

 

Limitations

Our study has several limitations, including its cross-sectional study design, a lack of longitudinal data within individuals, and exclusion of institutionalized individuals. Limitations notwithstanding this study has several important aspects. As prior MS is highly variable, we were limited in our inability to stratify by service type or length of service. For example, veteran status is conferred to a “Reservist or member of the National Guard called to federal active duty or disabled from a disease or injury incurred or aggravated in line of duty or while in training status also qualify as a veteran” (13 CFR § 125.11). For the purposes of our study, prior MS would include all active-duty service (veterans) as well as reservists and National Guard members who have not been activated. This may be more representative of the overall effect of MS, as limitation to those receiving care within the VA may select for an older, more multimorbid population of patients, limiting generalizability.

In addition, more detailed information regarding service-related exposures and other service-connected conditions would allow for a more granular risk assessment by service type, era, and military conflict. Our finding of excess CKD burden among those with prior MS compared with the overall population is timely given the recent passage of the Promise to Address Comprehensive Toxics (PACT) Act. Exposure to and injury from Agent Orange—a known service-connected exposure associated with incident hypertension and diabetes—may be a significant contributor to CKD that may have a significant era effect. In addition, water contamination among those stationed in Camp Lejeune in North Carolina has notable genitourinary associations. Finally, burn pit exposures in more recent military conflicts may also have important associations with chronic disease, possibly including CKD. While similar attempts at the creation of large-scale US veteran cohorts have been limited by incomplete capture of creatinine, the large proportion of missing race data, and limited inclusion of additional markers of kidney disease, our use of a well-described, nationally representative survey along with standardized capture of clinical and laboratory elements mitigate the use of various societal or other codified definitions.¹

Conclusions

Prior MS is associated with an increased risk of CKD overall and across several important subgroups. This finding was observed in various unadjusted and adjusted models and may constitute a unique risk profile of risk.

Chronic kidney disease (CKD) affects nearly 37 million people (11%) in the US and is a leading cause of death and morbidity. Due to their older age and higher prevalence of comorbid conditions, the prevalence of CKD among veterans is approximately 34% higher than in the general population and the fourth most common chronic disease diagnosed among US veterans.^1,2 US veterans and those with prior military service (MS) may be at a particularly high risk for CKD and associated health care outcomes including increased hospitalization and death. The observed excess burden of CKD is not mirrored in the general population, and it is unclear whether prior MS confers a unique risk profile for CKD.

Current estimates of CKD burden among veterans or those with prior MS are widely variable and have been limited by unique regions, specific exposure profiles, or to single health care systems. As such, there remains a paucity of data examining CKD burden more broadly. We performed a study in the adult population of the US to quantify associations with the extent of CKD, enumerate temporal trends of CKD among those with prior MS, describe risk within subgroups, and compare heterogeneity of risk factors for CKD by MS.

Methods

The National Health and Nutrition Examination Survey (NHANES) is a suite of nationally representative, cross-sectional surveys of the noninstitutionalized US population. It is conducted by the National Center for Health Statistics and uses a stratified, clustered probability design, with surveys carried out without interruption, collated, and made accessible to the public at 2-year intervals.³ The survey consists of a questionnaire, physical examination, and laboratory data.

The inclusion criteria for our study were age ≥ 20 years along with serum creatinine and urinary albumin-creatinine measurements. The following definitions were used for the study:

• CKD: Estimated glomerular filtration rate < 60 mL/min/1.73 m² calibrated to isotope dilution mass spectrometry (IDMS).

• Traceable: Creatinine-based CKD Epidemiology Collaboration formula or urinary albumin-creatine ratio ≥ 30 mg/g.

• MS: Positive response to the questions “Did you ever serve in the Armed Forces of the United States?” (1999 to 2010) or “Have you ever served on active duty in the US Armed Forces, military Reserves, or National Guard?” (2011 to 2018).

• Diabetes: Self-reported history, medication for diabetes, or glycated hemoglobin ≥ 7%.

• Hypertension: Blood pressure ≥ 140/90 or ≥ 130/40 mm Hg in the presence of diabetes, medication for hypertension, cardiovascular disease, or CKD, myocardial infarction, cardiac failure, or cerebrovascular disease by self-report.^2,3

Analysis

Primary sampling unit, stratum, and weight variables were employed throughout to generate parameter estimates that are generalizable to the US population.^4,5 The χ² test and logistic regression, respectively, were employed for comparison of proportions and estimation of odds ratios. R Version 4.1.2 was employed for data analysis.

 

 

Results

In the overall sample, the frequencies (95% standard error [SE]) of CKD and prior MS were 15.2% (0.3) and 11.5% (0.3) (Table 1). The proportion (SE) with CKD was significantly higher among those with prior MS vs the overall population: 22.7% (0.7) vs 15.2% (0.3) (P < .001). Significant associations with CKD were observed (P < .05) by age, sex, race and ethnicity, family poverty, school education, health insurance, smoking, body mass index, diabetes, hypertension, cardiovascular disease, and malignancy. Within those reporting prior MS, the proportion (SE) with CKD differed by era: 1999 to 2002, 18.9% (1.1); 2003 to 2006, 24.9% (1.5); 2007 to 2010, 22.3% (1.5); 2011 to 2014, 24.3% (1.7); and 2015 to 2018, 24.0% (1.8) (P = .02) (Figure 1).

Without covariate adjustment, prior MS was significantly associated with an increased risk of CKD (unadjusted odds ratio [OR], 1.78; 95% CI, 1.64-1.93; P < .05) (Table 2). Prior MS was significantly associated with CKD in the following subgroups: 2003 to 2006, 2011 to 2014, 2015 to 2018, age groups of 40 to 64 years and ≥ 65 years, male sex, non-Hispanic White and Hispanic ethnicity, school education of grade 0 to 11, and private or other health insurance. Additional comorbidities strongly associated with CKD included hypertension (OR, 6.37; 95% CI, 5.37-7.55), diabetes (OR, 4.16; 95% CI, 3.45-5.03), and cardiovascular disease (OR, 4.20; 95% CI, 3.57-4.95).

In the population reporting prior MS, the unadjusted OR of CKD vs 1999 to 2002 was greater for all other examined eras; with the greatest likelihood observed for the 2003 to 2006 era. Unadjusted ORs of CKD differed in groups with and without prior MS (P value for interaction < .05) for 2003 to 2006, those aged 40 to 64 years and ≥ 65 years, female sex, non-Hispanic African American and Hispanic race and ethnicity, family poverty, high school education, private health insurance, any smoking history, diabetes, hypertension, and cardiovascular disease (Figure 2A).

Following adjustment for age, sex, and race and ethnicity, MS was associated with a 17% higher likelihood of CKD (adjusted odds ratio [AOR], 1.17; 95% CI, 1.06-1.28; P < .01) (Table 3). Prior MS was significantly associated (P < .05) with CKD in the subgroups: age groups 40 to 64 years and ≥ 65 years, non-Hispanic African American, and body mass index ≥ 30. Among those with prior MS, comorbidities strongly associated with CKD in adjusted models included hypertension (AOR, 3.86; 95% CI, 3.18-4.69), diabetes (AOR, 3.05; 95% CI, 2.44-3.82), and cardiovascular disease (AOR, 2.51; 95% CI, 2.09-3.01). In the population with prior MS, the adjusted likelihood of CKD vs 1999 to 2002 was similar across all eras. Adjusted associations of CKD differed in groups with and without prior MS for age groups 40 to 64 years and ≥ 65 years, female sex, and family poverty (P < .05) (Figure 2B).

Discussion

We observed that prior MS was associated with CKD, all eras were associated with CKD in the subgroup with MS, and risk factors for CKD differed among many subgroups both with and without MS history, a finding that remained present in adjusted models. In addition, the finding of CKD was relatively common among those with prior MS (approximately 15%) and was most strongly associated with increasing age and comorbidities frequently associated with CKD.

Although many studies have demonstrated associations of US veteran status with various comorbidities, including hypertension, obesity, and diabetes, these studies often are limited to those both qualifying and receiving care within the US Department of Veterans Affairs (VA) health care system.^6-9 The crude proportion of individuals reporting multiple chronic conditions, which included hypertension, diabetes, and weak or failing kidneys, was 49.7% for US veterans compared with 24.1% for nonveterans.² Large-scale, nationally representative cohorts for use in this context have been limited by the heterogeneity of definitions of CKD applied with limited timeframes yielding variable estimates.^1,10 Moreover, few studies have examined the clinical epidemiology of CKD more broadly in the US among those with prior MS. For example, a PubMed search on March 3, 2022, with the terms “epidemiology”, “military service”, and “chronic kidney disease” produced only 9 citations, one of which examined trends among a non-US cohort and quantifying disease burden another among adolescents.

Whether or not prior MS confers a unique risk profile for CKD is unknown. While our findings of an increased CKD burden among those reporting MS may partially reflect observed increases in baseline comorbidities, the observed excess CKD among those with MS remained across multiple categories even after adjustment for baseline demography. As several studies have demonstrated, enlistment into MS may select for a more diverse population; however those enlisted personnel may be of lower socioeconomic status and possibly at higher risk of CKD.^11,12 Our findings of important differences in baseline determinants of health mirror this. The proportion of MS respondents with CKD vs CKD alone reporting a high school education or lower was higher (36.0% vs 21.8%) as well as among those with a history of family poverty (21.1% vs 18.0%).

 

 

Limitations

Our study has several limitations, including its cross-sectional study design, a lack of longitudinal data within individuals, and exclusion of institutionalized individuals. Limitations notwithstanding this study has several important aspects. As prior MS is highly variable, we were limited in our inability to stratify by service type or length of service. For example, veteran status is conferred to a “Reservist or member of the National Guard called to federal active duty or disabled from a disease or injury incurred or aggravated in line of duty or while in training status also qualify as a veteran” (13 CFR § 125.11). For the purposes of our study, prior MS would include all active-duty service (veterans) as well as reservists and National Guard members who have not been activated. This may be more representative of the overall effect of MS, as limitation to those receiving care within the VA may select for an older, more multimorbid population of patients, limiting generalizability.

In addition, more detailed information regarding service-related exposures and other service-connected conditions would allow for a more granular risk assessment by service type, era, and military conflict. Our finding of excess CKD burden among those with prior MS compared with the overall population is timely given the recent passage of the Promise to Address Comprehensive Toxics (PACT) Act. Exposure to and injury from Agent Orange—a known service-connected exposure associated with incident hypertension and diabetes—may be a significant contributor to CKD that may have a significant era effect. In addition, water contamination among those stationed in Camp Lejeune in North Carolina has notable genitourinary associations. Finally, burn pit exposures in more recent military conflicts may also have important associations with chronic disease, possibly including CKD. While similar attempts at the creation of large-scale US veteran cohorts have been limited by incomplete capture of creatinine, the large proportion of missing race data, and limited inclusion of additional markers of kidney disease, our use of a well-described, nationally representative survey along with standardized capture of clinical and laboratory elements mitigate the use of various societal or other codified definitions.¹

Conclusions

Prior MS is associated with an increased risk of CKD overall and across several important subgroups. This finding was observed in various unadjusted and adjusted models and may constitute a unique risk profile of risk.

Chronic kidney disease (CKD) affects nearly 37 million people (11%) in the US and is a leading cause of death and morbidity. Due to their older age and higher prevalence of comorbid conditions, the prevalence of CKD among veterans is approximately 34% higher than in the general population and the fourth most common chronic disease diagnosed among US veterans.^1,2 US veterans and those with prior military service (MS) may be at a particularly high risk for CKD and associated health care outcomes including increased hospitalization and death. The observed excess burden of CKD is not mirrored in the general population, and it is unclear whether prior MS confers a unique risk profile for CKD.

Current estimates of CKD burden among veterans or those with prior MS are widely variable and have been limited by unique regions, specific exposure profiles, or to single health care systems. As such, there remains a paucity of data examining CKD burden more broadly. We performed a study in the adult population of the US to quantify associations with the extent of CKD, enumerate temporal trends of CKD among those with prior MS, describe risk within subgroups, and compare heterogeneity of risk factors for CKD by MS.

Methods

The National Health and Nutrition Examination Survey (NHANES) is a suite of nationally representative, cross-sectional surveys of the noninstitutionalized US population. It is conducted by the National Center for Health Statistics and uses a stratified, clustered probability design, with surveys carried out without interruption, collated, and made accessible to the public at 2-year intervals.³ The survey consists of a questionnaire, physical examination, and laboratory data.

The inclusion criteria for our study were age ≥ 20 years along with serum creatinine and urinary albumin-creatinine measurements. The following definitions were used for the study:

• CKD: Estimated glomerular filtration rate < 60 mL/min/1.73 m² calibrated to isotope dilution mass spectrometry (IDMS).

• Traceable: Creatinine-based CKD Epidemiology Collaboration formula or urinary albumin-creatine ratio ≥ 30 mg/g.

• MS: Positive response to the questions “Did you ever serve in the Armed Forces of the United States?” (1999 to 2010) or “Have you ever served on active duty in the US Armed Forces, military Reserves, or National Guard?” (2011 to 2018).

• Diabetes: Self-reported history, medication for diabetes, or glycated hemoglobin ≥ 7%.

• Hypertension: Blood pressure ≥ 140/90 or ≥ 130/40 mm Hg in the presence of diabetes, medication for hypertension, cardiovascular disease, or CKD, myocardial infarction, cardiac failure, or cerebrovascular disease by self-report.^2,3

Analysis

Primary sampling unit, stratum, and weight variables were employed throughout to generate parameter estimates that are generalizable to the US population.^4,5 The χ² test and logistic regression, respectively, were employed for comparison of proportions and estimation of odds ratios. R Version 4.1.2 was employed for data analysis.

 

 

Results

In the overall sample, the frequencies (95% standard error [SE]) of CKD and prior MS were 15.2% (0.3) and 11.5% (0.3) (Table 1). The proportion (SE) with CKD was significantly higher among those with prior MS vs the overall population: 22.7% (0.7) vs 15.2% (0.3) (P < .001). Significant associations with CKD were observed (P < .05) by age, sex, race and ethnicity, family poverty, school education, health insurance, smoking, body mass index, diabetes, hypertension, cardiovascular disease, and malignancy. Within those reporting prior MS, the proportion (SE) with CKD differed by era: 1999 to 2002, 18.9% (1.1); 2003 to 2006, 24.9% (1.5); 2007 to 2010, 22.3% (1.5); 2011 to 2014, 24.3% (1.7); and 2015 to 2018, 24.0% (1.8) (P = .02) (Figure 1).

Without covariate adjustment, prior MS was significantly associated with an increased risk of CKD (unadjusted odds ratio [OR], 1.78; 95% CI, 1.64-1.93; P < .05) (Table 2). Prior MS was significantly associated with CKD in the following subgroups: 2003 to 2006, 2011 to 2014, 2015 to 2018, age groups of 40 to 64 years and ≥ 65 years, male sex, non-Hispanic White and Hispanic ethnicity, school education of grade 0 to 11, and private or other health insurance. Additional comorbidities strongly associated with CKD included hypertension (OR, 6.37; 95% CI, 5.37-7.55), diabetes (OR, 4.16; 95% CI, 3.45-5.03), and cardiovascular disease (OR, 4.20; 95% CI, 3.57-4.95).

In the population reporting prior MS, the unadjusted OR of CKD vs 1999 to 2002 was greater for all other examined eras; with the greatest likelihood observed for the 2003 to 2006 era. Unadjusted ORs of CKD differed in groups with and without prior MS (P value for interaction < .05) for 2003 to 2006, those aged 40 to 64 years and ≥ 65 years, female sex, non-Hispanic African American and Hispanic race and ethnicity, family poverty, high school education, private health insurance, any smoking history, diabetes, hypertension, and cardiovascular disease (Figure 2A).

Following adjustment for age, sex, and race and ethnicity, MS was associated with a 17% higher likelihood of CKD (adjusted odds ratio [AOR], 1.17; 95% CI, 1.06-1.28; P < .01) (Table 3). Prior MS was significantly associated (P < .05) with CKD in the subgroups: age groups 40 to 64 years and ≥ 65 years, non-Hispanic African American, and body mass index ≥ 30. Among those with prior MS, comorbidities strongly associated with CKD in adjusted models included hypertension (AOR, 3.86; 95% CI, 3.18-4.69), diabetes (AOR, 3.05; 95% CI, 2.44-3.82), and cardiovascular disease (AOR, 2.51; 95% CI, 2.09-3.01). In the population with prior MS, the adjusted likelihood of CKD vs 1999 to 2002 was similar across all eras. Adjusted associations of CKD differed in groups with and without prior MS for age groups 40 to 64 years and ≥ 65 years, female sex, and family poverty (P < .05) (Figure 2B).

Discussion

We observed that prior MS was associated with CKD, all eras were associated with CKD in the subgroup with MS, and risk factors for CKD differed among many subgroups both with and without MS history, a finding that remained present in adjusted models. In addition, the finding of CKD was relatively common among those with prior MS (approximately 15%) and was most strongly associated with increasing age and comorbidities frequently associated with CKD.

Although many studies have demonstrated associations of US veteran status with various comorbidities, including hypertension, obesity, and diabetes, these studies often are limited to those both qualifying and receiving care within the US Department of Veterans Affairs (VA) health care system.^6-9 The crude proportion of individuals reporting multiple chronic conditions, which included hypertension, diabetes, and weak or failing kidneys, was 49.7% for US veterans compared with 24.1% for nonveterans.² Large-scale, nationally representative cohorts for use in this context have been limited by the heterogeneity of definitions of CKD applied with limited timeframes yielding variable estimates.^1,10 Moreover, few studies have examined the clinical epidemiology of CKD more broadly in the US among those with prior MS. For example, a PubMed search on March 3, 2022, with the terms “epidemiology”, “military service”, and “chronic kidney disease” produced only 9 citations, one of which examined trends among a non-US cohort and quantifying disease burden another among adolescents.

Whether or not prior MS confers a unique risk profile for CKD is unknown. While our findings of an increased CKD burden among those reporting MS may partially reflect observed increases in baseline comorbidities, the observed excess CKD among those with MS remained across multiple categories even after adjustment for baseline demography. As several studies have demonstrated, enlistment into MS may select for a more diverse population; however those enlisted personnel may be of lower socioeconomic status and possibly at higher risk of CKD.^11,12 Our findings of important differences in baseline determinants of health mirror this. The proportion of MS respondents with CKD vs CKD alone reporting a high school education or lower was higher (36.0% vs 21.8%) as well as among those with a history of family poverty (21.1% vs 18.0%).

 

 

Limitations

Our study has several limitations, including its cross-sectional study design, a lack of longitudinal data within individuals, and exclusion of institutionalized individuals. Limitations notwithstanding this study has several important aspects. As prior MS is highly variable, we were limited in our inability to stratify by service type or length of service. For example, veteran status is conferred to a “Reservist or member of the National Guard called to federal active duty or disabled from a disease or injury incurred or aggravated in line of duty or while in training status also qualify as a veteran” (13 CFR § 125.11). For the purposes of our study, prior MS would include all active-duty service (veterans) as well as reservists and National Guard members who have not been activated. This may be more representative of the overall effect of MS, as limitation to those receiving care within the VA may select for an older, more multimorbid population of patients, limiting generalizability.

In addition, more detailed information regarding service-related exposures and other service-connected conditions would allow for a more granular risk assessment by service type, era, and military conflict. Our finding of excess CKD burden among those with prior MS compared with the overall population is timely given the recent passage of the Promise to Address Comprehensive Toxics (PACT) Act. Exposure to and injury from Agent Orange—a known service-connected exposure associated with incident hypertension and diabetes—may be a significant contributor to CKD that may have a significant era effect. In addition, water contamination among those stationed in Camp Lejeune in North Carolina has notable genitourinary associations. Finally, burn pit exposures in more recent military conflicts may also have important associations with chronic disease, possibly including CKD. While similar attempts at the creation of large-scale US veteran cohorts have been limited by incomplete capture of creatinine, the large proportion of missing race data, and limited inclusion of additional markers of kidney disease, our use of a well-described, nationally representative survey along with standardized capture of clinical and laboratory elements mitigate the use of various societal or other codified definitions.¹

Conclusions

Prior MS is associated with an increased risk of CKD overall and across several important subgroups. This finding was observed in various unadjusted and adjusted models and may constitute a unique risk profile of risk.

Inhaled corticosteroids (ICSs) are frequently prescribed for the treatment of chronic obstructive pulmonary disease (COPD) to reduce exacerbations in a specific subset of patients. The long-term use of ICSs, however, is associated with several potential systemic adverse effects, including adrenal suppression, decreased bone mineral density, and immunosuppression.¹ The concern for immunosuppression is particularly notable and leads to a known increased risk for developing pneumonia in patients with COPD. These patients frequently have other concurrent risk factors for pneumonia (eg, history of tobacco use, older age, and severe airway limitations) and are at higher risk for more severe outcomes in the setting of pneumonia.^2,3

Primarily due to the concern of pneumonia risks, the Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines have recommended ICS discontinuation in patients who are less likely to receive significant benefits from therapy.⁴ Likely due to an anti-inflammatory mechanism of action, ICSs have been shown to reduce COPD exacerbation rates in patients with comorbid asthma or who have evidence of a strong inflammatory component to their COPD. The strongest indicator of an inflammatory component is an elevated blood eosinophil (EOS) count; those with EOS > 300 cells/µL are most likely to benefit from ICSs, whereas those with a count < 100 cells/µL are unlikely to have a significant response. In addition to the inflammatory component consideration, prior studies have shown improvements in lung function and reduction of exacerbations with ICS use in patients with frequent moderate-to-severe COPD exacerbations.⁵ Although the GOLD guidelines provide recommendations about who is appropriate to discontinue ICS use, clinicians have no clear guidance on the risks or the best discontinuation strategy.

Based primarily on data from a prior randomized controlled trial, the Veterans Integrated Services Network (VISN) 17, which includes the Veterans Affairs North Texas Health Care System (VANTHCS) in Dallas, established a recommended ICS de-escalation strategy.^6,7 The strategy included a 12-week stepwise taper using a mometasone inhaler for all patients discontinuing a moderate or high dose ICS. The lack of substantial clinical trial data or expert consensus guideline recommendations has left open the question of whether a taper is necessary. To answer that question, this study was conducted to evaluate whether there is a difference in the rate of COPD exacerbations following abrupt discontinuation vs gradual taper of ICS therapy.

Methods

This single-center, retrospective cohort study was conducted at VANTHCS. Patient electronic health records between January 10, 2021, and September 1, 2021, were reviewed for the last documented fill date of any inhaler containing a steroid component. This time frame was chosen to coincide with a VANTHCS initiative to follow GOLD guidelines for ICS discontinuation. Patients were followed for outcomes until November 1, 2022.

To be included in this study, patients had to have active prescriptions at VANTHCS, have a documented diagnosis of COPD in their chart, and be prescribed a stable dose of ICS for ≥ 1 year prior to their latest refill. The inhaler used could contain an ICS as monotherapy, in combination with a long-acting β-agonist (LABA), or as part of triple therapy with an additional long-acting muscarinic antagonist (LAMA). The inhaler needed to be discontinued during the study period of interest.

Patients were excluded if they had a diagnosis of asthma, were aged < 40 years, had active prescriptions for multiple ICS inhalers or nebulizers, or had significant oral steroid use (≥ 5 mg/d prednisone or an equivalent steroid for > 6 weeks) within 1 year of their ICS discontinuation date. In addition, to reduce the risk of future events being misclassified as COPD exacerbations, patients were excluded if they had a congestive heart failure exacerbation up to 2 years before ICS discontinuation or a diagnosis of COVID-19 infection up to 1 year before or 6 months after ICS discontinuation. Patients with a COPD exacerbation requiring an emergency department or hospital visit within 2 years prior to ICS discontinuation were also excluded, as de-escalation of ICS therapy was likely inappropriate in these cases. Finally, patients were excluded if they were started on a different ICS immediately following the discontinuation of their first ICS.

The primary outcome for this study was COPD exacerbations requiring an emergency department visit or hospitalization within 6 months of ICS discontinuation. A secondary outcome examining the rates of COPD exacerbations within 12 months also was used. The original study design called for the use of inferential statistics to compare the rates of primary and secondary outcomes in patients whose ICS was abruptly discontinued with those who were tapered slowly. After data collection, however, the small sample size and low event rate meant that the planned statistical tests were no longer appropriate. Instead, we decided to analyze the planned outcomes using descriptive statistics and look at an additional number of post hoc outcomes to provide deeper insight into clinical practice. We examined the association between relevant demographic factors, such as age, comorbidity burden, ICS potency, duration of ICS therapy, and EOS count and the clinician decision whether to taper the ICS. These same factors were also evaluated for potential association with the increased risk of COPD exacerbations following ICS discontinuation.

 

 

Results

A total of 75 patients were included. Most patients were White race and male with a mean (SD) age of 71.6 (7.4) years. Charlson Comorbidity Index scores were calculated for all included patients with a mean (SD) score of 5.4 (2.0). Of note, scores > 5 are considered a severe comorbidity burden and have an estimated mean 10-year survival rate < 21%. The overwhelming majority of patients were receiving budesonide/formoterol as their ICS inhaler with 1 receiving mometasone monotherapy. When evaluating the steroid dose, 18 (24%) patients received a low dose ICS (200-400 µg of budesonide or 110-220 µg of mometasone), while 57 (76%) received a medium dose (400-800 µg of budesonide or 440 µg of mometasone). No patients received a high ICS dose. The mean (SD) duration of therapy before discontinuation was 4.0 (2.7) years (Table 1).

Nine (12%) patients had their ICS slowly tapered, while therapy was abruptly discontinued in the other 66 (88%) patients. A variety of taper types were used (Figure) without a strong preference for a particular dosing strategy. The primary outcome of COPD exacerbation requiring emergency department visit or hospitalization within 6 months occurred in 2 patients. When the time frame was extended to 12 months for the secondary outcome, an additional 3 patients experienced an event. The mean time to event was 172 days following ICS discontinuation. All the events occurred in patients whose ICS was discontinued without any type of taper.

In a post hoc analysis, we examined the relationship between specific variables and the clinician choice whether to taper an ICS. There was no discernable impact of age, race and ethnicity, comorbidity score, or ICS dose on whether an ICS was tapered. We observed a slight association between shorter duration of therapy and lower EOS count and use of a taper. When evaluating the relationship between these same factors and exacerbation occurrence, we saw comparable trends (Table 2). Patients with an exacerbation had a slightly longer mean duration of ICS therapy and lower mean EOS count.

Discussion

Despite facility guidance recommending tapering of therapy when discontinuing a moderate- or high-dose ICS, most patients in this study discontinued the ICS abruptly. The clinician may have been concerned with patients being able to adhere to a taper regimen, skeptical of the actual need to taper, or unaware of the VANTHCS recommendations for a specific taper method. Shared decision making with patients may have also played a role in prescribing patterns. Currently, there is not sufficient data to support the use of any one particular type of taper over another, which accounts for the variability seen in practice.

The decision to taper ICSs did not seem to be strongly associated with any specific demographic factor, although the ability to examine the impact of factors (eg, race and ethnicity) was limited due to the largely homogenous population. One may have expected a taper to be more common in older patients or in those with more comorbidities; however, this was not observed in this study. The only discernible trends seen were a lower frequency of tapering in patients who had a shorter duration of ICS therapy and those with lower EOS counts. These patients were at lower risk of repeat COPD exacerbations compared with those with longer ICS therapy duration and higher EOS counts; therefore, this finding was unexpected. This suggests that patient-specific factors may not be the primary driving force in the ICS tapering decision; instead it may be based on general clinician preferences or shared decision making with individual patients.

Overall, we noted very low rates of COPD exacerbations. As ICS discontinuation was occurring in stable patients without any recent exacerbations, lower rates of future exacerbations were expected compared with the population of patients with COPD as a whole. This suggests that ICS therapy can be safely stopped in stable patients with COPD who are not likely to receive significant benefits as defined in the GOLD guidelines. All of the exacerbations that occurred were in patients whose ICS was abruptly discontinued; however, given the small number of patients who had a taper, it is difficult to draw conclusions. The low overall rate of exacerbations suggests that a taper may not be necessary to ensure safety while stopping a low- or moderate-intensity ICS.

Several randomized controlled trials have attempted to evaluate the need for an ICS taper; however, results remain mixed. The COSMIC study showed a decline in lung function following ICS discontinuation in patients with ≥ 2 COPD exacerbations in the previous year.⁸ Similar results were seen in the SUNSET study with increased exacerbation rates after ICS discontinuation in patients with elevated EOS counts.⁹ However, these studies included patients for whom ICS discontinuation is currently not recommended. Alternatively, the INSTEAD trial looked at patients without frequent recent exacerbations and found no difference in lung function, exacerbation rates, or rescue inhaler use in patients that continued combination ICS plus bronchodilator use vs those de-escalated to bronchodilator monotherapy.¹⁰

All 3 studies chose to abruptly stop the ICS when discontinuing therapy; however, using a slow, stepwise taper similar to that used after long periods of oral steroid use may reduce the risk of worsening exacerbations. The WISDOM trial is the only major randomized trial to date that stopped ICS therapy using a stepwise withdrawal of therapy.⁷ In patients who were continued on triple inhaled therapy (2 bronchodilators plus ICS) vs those who were de-escalated to dual bronchodilator therapy, de-escalation was noninferior to continuation of therapy in time to first COPD exacerbation. Both the WISDOM and INSTEAD trials were consistent with the results found in our real-world retrospective evaluation.

There did not seem to be an increased exacerbation risk following ICS discontinuation in any patient subpopulation based on sex, age, race and ethnicity, or comorbidity burden. We noted a trend toward more exacerbations in patients with a longer duration of ICS therapy, suggesting that additional caution may be needed when stopping ICS therapy for these patients. We also noted a trend toward more exacerbations in patients with a lower mean EOS count; however, given the low event rate and wide variability in observed patient EOS counts, this is likely a spurious finding.

 

Limitations

The small sample size, resulting from the strict exclusion criteria, limits the generalizability of the results. Although the low number of events seen in this study supports safety in ICS discontinuation, there may have been higher rates observed in a larger population. The most common reason for patient exclusion was the initiation of another ICS immediately following discontinuation of the original ICS. During the study period, VANTHCS underwent a change to its formulary: Fluticasone/salmeterol replaced budesonide/formoterol as the preferred ICS/LABA combination. As a result, many patients had their budesonide/formoterol discontinued during the study period solely to initiate fluticasone/salmeterol therapy. As these patients did not truly have their ICS discontinued or have a significant period without ICS therapy, they were not included in the results, and the total patient population available to analyze was relatively limited.

The low event rate also limits the ability to compare various factors influencing exacerbation risk, particularly taper vs abrupt ICS discontinuation. This is further compounded by the small number of patients who had a taper performed and the lack of consistency in the method of tapering used. Statistical significance could not be determined for any outcome, and all findings were purely hypothesis generating. Finally, data were only collected for moderate or severe COPD exacerbations that resulted in an emergency department visit or hospitalization, so there may have been mild exacerbations treated in the outpatient setting that were not captured.

Despite these limitations, this study adds data to an area of COPD management that currently lacks strong clinical guidance. Since investigators had access to clinician notes, we were able to capture ICS tapers even if patients did not receive a prescription with specific taper instructions. The extended follow-up period of 12 months evaluated a longer potential time to impact of ICS discontinuation than is done in most COPD clinical trials.

Conclusions

Overall, very low rates of COPD exacerbations occurred following ICS discontinuation, regardless of whether a taper was used. The results suggest that there may be several appropriate ways to discontinue ICS therapy. However, there is insufficient evidence to support a particular taper or the need to taper at all. It seems to be safe to discontinue ICS therapy in patients who are unlikely to benefit from continued use; however, patient-specific factors should be considered as part of clinical decision making.

Inhaled corticosteroids (ICSs) are frequently prescribed for the treatment of chronic obstructive pulmonary disease (COPD) to reduce exacerbations in a specific subset of patients. The long-term use of ICSs, however, is associated with several potential systemic adverse effects, including adrenal suppression, decreased bone mineral density, and immunosuppression.¹ The concern for immunosuppression is particularly notable and leads to a known increased risk for developing pneumonia in patients with COPD. These patients frequently have other concurrent risk factors for pneumonia (eg, history of tobacco use, older age, and severe airway limitations) and are at higher risk for more severe outcomes in the setting of pneumonia.^2,3

Primarily due to the concern of pneumonia risks, the Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines have recommended ICS discontinuation in patients who are less likely to receive significant benefits from therapy.⁴ Likely due to an anti-inflammatory mechanism of action, ICSs have been shown to reduce COPD exacerbation rates in patients with comorbid asthma or who have evidence of a strong inflammatory component to their COPD. The strongest indicator of an inflammatory component is an elevated blood eosinophil (EOS) count; those with EOS > 300 cells/µL are most likely to benefit from ICSs, whereas those with a count < 100 cells/µL are unlikely to have a significant response. In addition to the inflammatory component consideration, prior studies have shown improvements in lung function and reduction of exacerbations with ICS use in patients with frequent moderate-to-severe COPD exacerbations.⁵ Although the GOLD guidelines provide recommendations about who is appropriate to discontinue ICS use, clinicians have no clear guidance on the risks or the best discontinuation strategy.

Based primarily on data from a prior randomized controlled trial, the Veterans Integrated Services Network (VISN) 17, which includes the Veterans Affairs North Texas Health Care System (VANTHCS) in Dallas, established a recommended ICS de-escalation strategy.^6,7 The strategy included a 12-week stepwise taper using a mometasone inhaler for all patients discontinuing a moderate or high dose ICS. The lack of substantial clinical trial data or expert consensus guideline recommendations has left open the question of whether a taper is necessary. To answer that question, this study was conducted to evaluate whether there is a difference in the rate of COPD exacerbations following abrupt discontinuation vs gradual taper of ICS therapy.

Methods

This single-center, retrospective cohort study was conducted at VANTHCS. Patient electronic health records between January 10, 2021, and September 1, 2021, were reviewed for the last documented fill date of any inhaler containing a steroid component. This time frame was chosen to coincide with a VANTHCS initiative to follow GOLD guidelines for ICS discontinuation. Patients were followed for outcomes until November 1, 2022.

To be included in this study, patients had to have active prescriptions at VANTHCS, have a documented diagnosis of COPD in their chart, and be prescribed a stable dose of ICS for ≥ 1 year prior to their latest refill. The inhaler used could contain an ICS as monotherapy, in combination with a long-acting β-agonist (LABA), or as part of triple therapy with an additional long-acting muscarinic antagonist (LAMA). The inhaler needed to be discontinued during the study period of interest.

Patients were excluded if they had a diagnosis of asthma, were aged < 40 years, had active prescriptions for multiple ICS inhalers or nebulizers, or had significant oral steroid use (≥ 5 mg/d prednisone or an equivalent steroid for > 6 weeks) within 1 year of their ICS discontinuation date. In addition, to reduce the risk of future events being misclassified as COPD exacerbations, patients were excluded if they had a congestive heart failure exacerbation up to 2 years before ICS discontinuation or a diagnosis of COVID-19 infection up to 1 year before or 6 months after ICS discontinuation. Patients with a COPD exacerbation requiring an emergency department or hospital visit within 2 years prior to ICS discontinuation were also excluded, as de-escalation of ICS therapy was likely inappropriate in these cases. Finally, patients were excluded if they were started on a different ICS immediately following the discontinuation of their first ICS.

The primary outcome for this study was COPD exacerbations requiring an emergency department visit or hospitalization within 6 months of ICS discontinuation. A secondary outcome examining the rates of COPD exacerbations within 12 months also was used. The original study design called for the use of inferential statistics to compare the rates of primary and secondary outcomes in patients whose ICS was abruptly discontinued with those who were tapered slowly. After data collection, however, the small sample size and low event rate meant that the planned statistical tests were no longer appropriate. Instead, we decided to analyze the planned outcomes using descriptive statistics and look at an additional number of post hoc outcomes to provide deeper insight into clinical practice. We examined the association between relevant demographic factors, such as age, comorbidity burden, ICS potency, duration of ICS therapy, and EOS count and the clinician decision whether to taper the ICS. These same factors were also evaluated for potential association with the increased risk of COPD exacerbations following ICS discontinuation.

 

 

Results

A total of 75 patients were included. Most patients were White race and male with a mean (SD) age of 71.6 (7.4) years. Charlson Comorbidity Index scores were calculated for all included patients with a mean (SD) score of 5.4 (2.0). Of note, scores > 5 are considered a severe comorbidity burden and have an estimated mean 10-year survival rate < 21%. The overwhelming majority of patients were receiving budesonide/formoterol as their ICS inhaler with 1 receiving mometasone monotherapy. When evaluating the steroid dose, 18 (24%) patients received a low dose ICS (200-400 µg of budesonide or 110-220 µg of mometasone), while 57 (76%) received a medium dose (400-800 µg of budesonide or 440 µg of mometasone). No patients received a high ICS dose. The mean (SD) duration of therapy before discontinuation was 4.0 (2.7) years (Table 1).

Nine (12%) patients had their ICS slowly tapered, while therapy was abruptly discontinued in the other 66 (88%) patients. A variety of taper types were used (Figure) without a strong preference for a particular dosing strategy. The primary outcome of COPD exacerbation requiring emergency department visit or hospitalization within 6 months occurred in 2 patients. When the time frame was extended to 12 months for the secondary outcome, an additional 3 patients experienced an event. The mean time to event was 172 days following ICS discontinuation. All the events occurred in patients whose ICS was discontinued without any type of taper.

In a post hoc analysis, we examined the relationship between specific variables and the clinician choice whether to taper an ICS. There was no discernable impact of age, race and ethnicity, comorbidity score, or ICS dose on whether an ICS was tapered. We observed a slight association between shorter duration of therapy and lower EOS count and use of a taper. When evaluating the relationship between these same factors and exacerbation occurrence, we saw comparable trends (Table 2). Patients with an exacerbation had a slightly longer mean duration of ICS therapy and lower mean EOS count.

Discussion

Despite facility guidance recommending tapering of therapy when discontinuing a moderate- or high-dose ICS, most patients in this study discontinued the ICS abruptly. The clinician may have been concerned with patients being able to adhere to a taper regimen, skeptical of the actual need to taper, or unaware of the VANTHCS recommendations for a specific taper method. Shared decision making with patients may have also played a role in prescribing patterns. Currently, there is not sufficient data to support the use of any one particular type of taper over another, which accounts for the variability seen in practice.

The decision to taper ICSs did not seem to be strongly associated with any specific demographic factor, although the ability to examine the impact of factors (eg, race and ethnicity) was limited due to the largely homogenous population. One may have expected a taper to be more common in older patients or in those with more comorbidities; however, this was not observed in this study. The only discernible trends seen were a lower frequency of tapering in patients who had a shorter duration of ICS therapy and those with lower EOS counts. These patients were at lower risk of repeat COPD exacerbations compared with those with longer ICS therapy duration and higher EOS counts; therefore, this finding was unexpected. This suggests that patient-specific factors may not be the primary driving force in the ICS tapering decision; instead it may be based on general clinician preferences or shared decision making with individual patients.

Overall, we noted very low rates of COPD exacerbations. As ICS discontinuation was occurring in stable patients without any recent exacerbations, lower rates of future exacerbations were expected compared with the population of patients with COPD as a whole. This suggests that ICS therapy can be safely stopped in stable patients with COPD who are not likely to receive significant benefits as defined in the GOLD guidelines. All of the exacerbations that occurred were in patients whose ICS was abruptly discontinued; however, given the small number of patients who had a taper, it is difficult to draw conclusions. The low overall rate of exacerbations suggests that a taper may not be necessary to ensure safety while stopping a low- or moderate-intensity ICS.

Several randomized controlled trials have attempted to evaluate the need for an ICS taper; however, results remain mixed. The COSMIC study showed a decline in lung function following ICS discontinuation in patients with ≥ 2 COPD exacerbations in the previous year.⁸ Similar results were seen in the SUNSET study with increased exacerbation rates after ICS discontinuation in patients with elevated EOS counts.⁹ However, these studies included patients for whom ICS discontinuation is currently not recommended. Alternatively, the INSTEAD trial looked at patients without frequent recent exacerbations and found no difference in lung function, exacerbation rates, or rescue inhaler use in patients that continued combination ICS plus bronchodilator use vs those de-escalated to bronchodilator monotherapy.¹⁰

All 3 studies chose to abruptly stop the ICS when discontinuing therapy; however, using a slow, stepwise taper similar to that used after long periods of oral steroid use may reduce the risk of worsening exacerbations. The WISDOM trial is the only major randomized trial to date that stopped ICS therapy using a stepwise withdrawal of therapy.⁷ In patients who were continued on triple inhaled therapy (2 bronchodilators plus ICS) vs those who were de-escalated to dual bronchodilator therapy, de-escalation was noninferior to continuation of therapy in time to first COPD exacerbation. Both the WISDOM and INSTEAD trials were consistent with the results found in our real-world retrospective evaluation.

There did not seem to be an increased exacerbation risk following ICS discontinuation in any patient subpopulation based on sex, age, race and ethnicity, or comorbidity burden. We noted a trend toward more exacerbations in patients with a longer duration of ICS therapy, suggesting that additional caution may be needed when stopping ICS therapy for these patients. We also noted a trend toward more exacerbations in patients with a lower mean EOS count; however, given the low event rate and wide variability in observed patient EOS counts, this is likely a spurious finding.

 

Limitations

The small sample size, resulting from the strict exclusion criteria, limits the generalizability of the results. Although the low number of events seen in this study supports safety in ICS discontinuation, there may have been higher rates observed in a larger population. The most common reason for patient exclusion was the initiation of another ICS immediately following discontinuation of the original ICS. During the study period, VANTHCS underwent a change to its formulary: Fluticasone/salmeterol replaced budesonide/formoterol as the preferred ICS/LABA combination. As a result, many patients had their budesonide/formoterol discontinued during the study period solely to initiate fluticasone/salmeterol therapy. As these patients did not truly have their ICS discontinued or have a significant period without ICS therapy, they were not included in the results, and the total patient population available to analyze was relatively limited.

The low event rate also limits the ability to compare various factors influencing exacerbation risk, particularly taper vs abrupt ICS discontinuation. This is further compounded by the small number of patients who had a taper performed and the lack of consistency in the method of tapering used. Statistical significance could not be determined for any outcome, and all findings were purely hypothesis generating. Finally, data were only collected for moderate or severe COPD exacerbations that resulted in an emergency department visit or hospitalization, so there may have been mild exacerbations treated in the outpatient setting that were not captured.

Despite these limitations, this study adds data to an area of COPD management that currently lacks strong clinical guidance. Since investigators had access to clinician notes, we were able to capture ICS tapers even if patients did not receive a prescription with specific taper instructions. The extended follow-up period of 12 months evaluated a longer potential time to impact of ICS discontinuation than is done in most COPD clinical trials.

Conclusions

Overall, very low rates of COPD exacerbations occurred following ICS discontinuation, regardless of whether a taper was used. The results suggest that there may be several appropriate ways to discontinue ICS therapy. However, there is insufficient evidence to support a particular taper or the need to taper at all. It seems to be safe to discontinue ICS therapy in patients who are unlikely to benefit from continued use; however, patient-specific factors should be considered as part of clinical decision making.

Inhaled corticosteroids (ICSs) are frequently prescribed for the treatment of chronic obstructive pulmonary disease (COPD) to reduce exacerbations in a specific subset of patients. The long-term use of ICSs, however, is associated with several potential systemic adverse effects, including adrenal suppression, decreased bone mineral density, and immunosuppression.¹ The concern for immunosuppression is particularly notable and leads to a known increased risk for developing pneumonia in patients with COPD. These patients frequently have other concurrent risk factors for pneumonia (eg, history of tobacco use, older age, and severe airway limitations) and are at higher risk for more severe outcomes in the setting of pneumonia.^2,3

Primarily due to the concern of pneumonia risks, the Global Initiative for Chronic Obstructive Lung Disease (GOLD) guidelines have recommended ICS discontinuation in patients who are less likely to receive significant benefits from therapy.⁴ Likely due to an anti-inflammatory mechanism of action, ICSs have been shown to reduce COPD exacerbation rates in patients with comorbid asthma or who have evidence of a strong inflammatory component to their COPD. The strongest indicator of an inflammatory component is an elevated blood eosinophil (EOS) count; those with EOS > 300 cells/µL are most likely to benefit from ICSs, whereas those with a count < 100 cells/µL are unlikely to have a significant response. In addition to the inflammatory component consideration, prior studies have shown improvements in lung function and reduction of exacerbations with ICS use in patients with frequent moderate-to-severe COPD exacerbations.⁵ Although the GOLD guidelines provide recommendations about who is appropriate to discontinue ICS use, clinicians have no clear guidance on the risks or the best discontinuation strategy.

Based primarily on data from a prior randomized controlled trial, the Veterans Integrated Services Network (VISN) 17, which includes the Veterans Affairs North Texas Health Care System (VANTHCS) in Dallas, established a recommended ICS de-escalation strategy.^6,7 The strategy included a 12-week stepwise taper using a mometasone inhaler for all patients discontinuing a moderate or high dose ICS. The lack of substantial clinical trial data or expert consensus guideline recommendations has left open the question of whether a taper is necessary. To answer that question, this study was conducted to evaluate whether there is a difference in the rate of COPD exacerbations following abrupt discontinuation vs gradual taper of ICS therapy.

Methods

This single-center, retrospective cohort study was conducted at VANTHCS. Patient electronic health records between January 10, 2021, and September 1, 2021, were reviewed for the last documented fill date of any inhaler containing a steroid component. This time frame was chosen to coincide with a VANTHCS initiative to follow GOLD guidelines for ICS discontinuation. Patients were followed for outcomes until November 1, 2022.

To be included in this study, patients had to have active prescriptions at VANTHCS, have a documented diagnosis of COPD in their chart, and be prescribed a stable dose of ICS for ≥ 1 year prior to their latest refill. The inhaler used could contain an ICS as monotherapy, in combination with a long-acting β-agonist (LABA), or as part of triple therapy with an additional long-acting muscarinic antagonist (LAMA). The inhaler needed to be discontinued during the study period of interest.

Patients were excluded if they had a diagnosis of asthma, were aged < 40 years, had active prescriptions for multiple ICS inhalers or nebulizers, or had significant oral steroid use (≥ 5 mg/d prednisone or an equivalent steroid for > 6 weeks) within 1 year of their ICS discontinuation date. In addition, to reduce the risk of future events being misclassified as COPD exacerbations, patients were excluded if they had a congestive heart failure exacerbation up to 2 years before ICS discontinuation or a diagnosis of COVID-19 infection up to 1 year before or 6 months after ICS discontinuation. Patients with a COPD exacerbation requiring an emergency department or hospital visit within 2 years prior to ICS discontinuation were also excluded, as de-escalation of ICS therapy was likely inappropriate in these cases. Finally, patients were excluded if they were started on a different ICS immediately following the discontinuation of their first ICS.

The primary outcome for this study was COPD exacerbations requiring an emergency department visit or hospitalization within 6 months of ICS discontinuation. A secondary outcome examining the rates of COPD exacerbations within 12 months also was used. The original study design called for the use of inferential statistics to compare the rates of primary and secondary outcomes in patients whose ICS was abruptly discontinued with those who were tapered slowly. After data collection, however, the small sample size and low event rate meant that the planned statistical tests were no longer appropriate. Instead, we decided to analyze the planned outcomes using descriptive statistics and look at an additional number of post hoc outcomes to provide deeper insight into clinical practice. We examined the association between relevant demographic factors, such as age, comorbidity burden, ICS potency, duration of ICS therapy, and EOS count and the clinician decision whether to taper the ICS. These same factors were also evaluated for potential association with the increased risk of COPD exacerbations following ICS discontinuation.

 

 

Results

A total of 75 patients were included. Most patients were White race and male with a mean (SD) age of 71.6 (7.4) years. Charlson Comorbidity Index scores were calculated for all included patients with a mean (SD) score of 5.4 (2.0). Of note, scores > 5 are considered a severe comorbidity burden and have an estimated mean 10-year survival rate < 21%. The overwhelming majority of patients were receiving budesonide/formoterol as their ICS inhaler with 1 receiving mometasone monotherapy. When evaluating the steroid dose, 18 (24%) patients received a low dose ICS (200-400 µg of budesonide or 110-220 µg of mometasone), while 57 (76%) received a medium dose (400-800 µg of budesonide or 440 µg of mometasone). No patients received a high ICS dose. The mean (SD) duration of therapy before discontinuation was 4.0 (2.7) years (Table 1).

Nine (12%) patients had their ICS slowly tapered, while therapy was abruptly discontinued in the other 66 (88%) patients. A variety of taper types were used (Figure) without a strong preference for a particular dosing strategy. The primary outcome of COPD exacerbation requiring emergency department visit or hospitalization within 6 months occurred in 2 patients. When the time frame was extended to 12 months for the secondary outcome, an additional 3 patients experienced an event. The mean time to event was 172 days following ICS discontinuation. All the events occurred in patients whose ICS was discontinued without any type of taper.

In a post hoc analysis, we examined the relationship between specific variables and the clinician choice whether to taper an ICS. There was no discernable impact of age, race and ethnicity, comorbidity score, or ICS dose on whether an ICS was tapered. We observed a slight association between shorter duration of therapy and lower EOS count and use of a taper. When evaluating the relationship between these same factors and exacerbation occurrence, we saw comparable trends (Table 2). Patients with an exacerbation had a slightly longer mean duration of ICS therapy and lower mean EOS count.

Discussion

Despite facility guidance recommending tapering of therapy when discontinuing a moderate- or high-dose ICS, most patients in this study discontinued the ICS abruptly. The clinician may have been concerned with patients being able to adhere to a taper regimen, skeptical of the actual need to taper, or unaware of the VANTHCS recommendations for a specific taper method. Shared decision making with patients may have also played a role in prescribing patterns. Currently, there is not sufficient data to support the use of any one particular type of taper over another, which accounts for the variability seen in practice.

The decision to taper ICSs did not seem to be strongly associated with any specific demographic factor, although the ability to examine the impact of factors (eg, race and ethnicity) was limited due to the largely homogenous population. One may have expected a taper to be more common in older patients or in those with more comorbidities; however, this was not observed in this study. The only discernible trends seen were a lower frequency of tapering in patients who had a shorter duration of ICS therapy and those with lower EOS counts. These patients were at lower risk of repeat COPD exacerbations compared with those with longer ICS therapy duration and higher EOS counts; therefore, this finding was unexpected. This suggests that patient-specific factors may not be the primary driving force in the ICS tapering decision; instead it may be based on general clinician preferences or shared decision making with individual patients.

Overall, we noted very low rates of COPD exacerbations. As ICS discontinuation was occurring in stable patients without any recent exacerbations, lower rates of future exacerbations were expected compared with the population of patients with COPD as a whole. This suggests that ICS therapy can be safely stopped in stable patients with COPD who are not likely to receive significant benefits as defined in the GOLD guidelines. All of the exacerbations that occurred were in patients whose ICS was abruptly discontinued; however, given the small number of patients who had a taper, it is difficult to draw conclusions. The low overall rate of exacerbations suggests that a taper may not be necessary to ensure safety while stopping a low- or moderate-intensity ICS.

Several randomized controlled trials have attempted to evaluate the need for an ICS taper; however, results remain mixed. The COSMIC study showed a decline in lung function following ICS discontinuation in patients with ≥ 2 COPD exacerbations in the previous year.⁸ Similar results were seen in the SUNSET study with increased exacerbation rates after ICS discontinuation in patients with elevated EOS counts.⁹ However, these studies included patients for whom ICS discontinuation is currently not recommended. Alternatively, the INSTEAD trial looked at patients without frequent recent exacerbations and found no difference in lung function, exacerbation rates, or rescue inhaler use in patients that continued combination ICS plus bronchodilator use vs those de-escalated to bronchodilator monotherapy.¹⁰

All 3 studies chose to abruptly stop the ICS when discontinuing therapy; however, using a slow, stepwise taper similar to that used after long periods of oral steroid use may reduce the risk of worsening exacerbations. The WISDOM trial is the only major randomized trial to date that stopped ICS therapy using a stepwise withdrawal of therapy.⁷ In patients who were continued on triple inhaled therapy (2 bronchodilators plus ICS) vs those who were de-escalated to dual bronchodilator therapy, de-escalation was noninferior to continuation of therapy in time to first COPD exacerbation. Both the WISDOM and INSTEAD trials were consistent with the results found in our real-world retrospective evaluation.

There did not seem to be an increased exacerbation risk following ICS discontinuation in any patient subpopulation based on sex, age, race and ethnicity, or comorbidity burden. We noted a trend toward more exacerbations in patients with a longer duration of ICS therapy, suggesting that additional caution may be needed when stopping ICS therapy for these patients. We also noted a trend toward more exacerbations in patients with a lower mean EOS count; however, given the low event rate and wide variability in observed patient EOS counts, this is likely a spurious finding.

 

Limitations

The small sample size, resulting from the strict exclusion criteria, limits the generalizability of the results. Although the low number of events seen in this study supports safety in ICS discontinuation, there may have been higher rates observed in a larger population. The most common reason for patient exclusion was the initiation of another ICS immediately following discontinuation of the original ICS. During the study period, VANTHCS underwent a change to its formulary: Fluticasone/salmeterol replaced budesonide/formoterol as the preferred ICS/LABA combination. As a result, many patients had their budesonide/formoterol discontinued during the study period solely to initiate fluticasone/salmeterol therapy. As these patients did not truly have their ICS discontinued or have a significant period without ICS therapy, they were not included in the results, and the total patient population available to analyze was relatively limited.

The low event rate also limits the ability to compare various factors influencing exacerbation risk, particularly taper vs abrupt ICS discontinuation. This is further compounded by the small number of patients who had a taper performed and the lack of consistency in the method of tapering used. Statistical significance could not be determined for any outcome, and all findings were purely hypothesis generating. Finally, data were only collected for moderate or severe COPD exacerbations that resulted in an emergency department visit or hospitalization, so there may have been mild exacerbations treated in the outpatient setting that were not captured.

Despite these limitations, this study adds data to an area of COPD management that currently lacks strong clinical guidance. Since investigators had access to clinician notes, we were able to capture ICS tapers even if patients did not receive a prescription with specific taper instructions. The extended follow-up period of 12 months evaluated a longer potential time to impact of ICS discontinuation than is done in most COPD clinical trials.

Conclusions

Overall, very low rates of COPD exacerbations occurred following ICS discontinuation, regardless of whether a taper was used. The results suggest that there may be several appropriate ways to discontinue ICS therapy. However, there is insufficient evidence to support a particular taper or the need to taper at all. It seems to be safe to discontinue ICS therapy in patients who are unlikely to benefit from continued use; however, patient-specific factors should be considered as part of clinical decision making.

When thyroid nodules are found clinically or incidentally on imaging, the patient’s thyrotropin level should be measured.¹ Ultrasound is the first-line imaging recommended to assess thyroid nodules.^1,2 Nodules can then be evaluated by a fine-needle aspiration (FNA) biopsy, which provides cytological information to determine whether the nodule is benign or malignant.^1,3,4 Most thyroid nodules pose a low risk of malignancy.¹

The American Thyroid Association guidelines on thyroid nodule management do not specify any recommendations for follow-up thyrotropin testing in patients who do not have any history that is known to affect thyroid function.¹ Therefore, clinicians have to make decisions regarding follow-up testing in these patients without any evidence-based guidelines. There is a lack of data in the literature on whether thyrotropin levels change over time in this patient population. If thyrotropin levels do not become abnormal over time, then patients would not need thyrotropin monitoring or treatment for hypo- or hyperthyroidism.

The aim of this study was to determine whether thyrotropin levels change over time in patients with thyroid nodules and determine whether repeat thyrotropin testing was required after initial testing. The authors hypothesized that thyrotropin values do not change substantially over time in patients with thyroid nodules, except in patients with a history of hot nodules, autoimmune thyroid disease, thyroid or pituitary surgery, radioactive iodine ablation, neck radiation, or use of medications affecting thyroid function. This study may be able to contribute to the clinical guidelines for thyrotropin testing in patients with thyroid nodules so that clinicians can make evidence-based decisions.

METHODS

This retrospective chart review was conducted using the Computerized Patient Record System at the Veterans Affairs Dayton Healthcare System (VADHS) in Ohio. Patients aged ≥ 18 years who were diagnosed with ≥ 1 thyroid nodule from January 2010 to December 2016 and had a normal thyrotropin level at the time of diagnosis were included in the study. Patients who were found to have thyroid nodules multiple times were included only once from the time of the initial diagnosis. Patients were excluded if they had a medical history known to affect thyroid function. Exclusion criteria included a history of hot thyroid nodules; autoimmune thyroid disease on imaging or blood work; history of thyroid surgery, including pituitary surgery; history of radioactive iodine treatment; history of neck radiation; use of thyroxine before nodule diagnosis; use of amiodarone, programmed cell death-1 inhibitors, programmed cell death ligand-1 inhibitors, or cytotoxic T-lymphocyte-associated protein-4 inhibitors; or 3 consecutive months of steroid use.

Age at nodule diagnosis, sex, race, thyrotropin values at and after the time of nodule diagnosis, and duration from nodule diagnosis to most recent thyrotropin value were retrospectively collected until 100 patients met inclusion criteria for the study. Of note, from 2010 to 2016, the assays used at the VADHS to measure thyrotropin values changed over time, as did the normal reference ranges and the type of sample used for the assays. Normal thyrotropin range at time of diagnosis based on serum or plasma samples and for repeat thyrotropin levels are provided in Table 1, also based on serum or plasma samples. All collected data in the study was de-identified for analysis.

Statistical Analysis

Patients were divided into 2 groups: those who had an abnormal most recent thyrotropin value and those who did not. Mean (SD) of both groups was calculated for continuous variables of age at diagnosis, initial thyrotropin value and most recent thyrotropin value, and time from diagnosis to most recent thyrotropin value. Percentages for both groups were calculated for categorical variables of sex, race, and whether initial and most recent thyrotropin values were based on serum or plasma samples and old or new reference ranges. A 95% CI was determined for the true population rate of patients with an abnormal thyrotropin value at most recent testing. Independent sample t tests were used to compare the continuous variables between the abnormal and normal most recent thyrotropin groups. Categorical variables between the 2 groups were compared using χ² tests. P < .05 was considered statistically significant. Statistical analyses were completed using IBM SPSS Statistics 27. This study was approved by the Wright State University Institutional Review Board and the VADHS Research and Development Committee.

 

 

RESULTS

Of 557 patient charts studied, 100 patients were included; the mean (SD) age at nodule diagnosis was 62.4 (11.1) years, and the mean (SD) initial thyrotropin level at nodule diagnosis was 1.51 (0.87) μIU/mL. The mean (SD) most recent thyrotropin level was 1.60 (1.03) μIU/mL after a mean duration of 5.7 (2.5) years postnodule diagnosis (Table 2).

Six patients (6%; 95% CI, 2.5%-12.7%) who had a normal thyrotropin level at nodule diagnosis developed an abnormal thyrotropin level in a mean (SD) of 6.9 (3.1) years. These 6 patients had a mean age at nodule diagnosis of 69.2 (11.4) years. Five of the 6 were male, and all were White patients. One patient’s thyrotropin level rose from an initial thyrotropin of 3.38 μIU/mL at nodule diagnosis to a high of 7.76 μIU/mL after 8.5 years. This patient was diagnosed with subclinical hypothyroidism and did not require treatment.

Five patients’ thyrotropin levels dropped below normal in a mean 7 years, with levels ranging from 0.25 to 0.52 μIU/mL. Of these patients, 2 became symptomatic from the nodules, experiencing dysphagia or hoarseness, with 1 diagnosed with hyperthyroidism. This patient was treated with methimazole and radioactive iodine ablation 9 years after diagnosis. The other 3 patients who developed low thyrotropin had no nodule symptoms or treatment. Ninety-four patients maintained thyrotropin values in the normal range for a mean (SD) of 5.7 (2.5) years and had a mean (SD) age at nodule diagnosis of 61.9 (11.0) years.

Both thyrotropin groups were compared. For categorical variables, there were no significant differences for sex (P = .99) or race (P = .55). For continuous variables, there were no significant differences in age at diagnosis (P = .12), initial thyrotropin (P = .24), most recent thyrotropin (P = .98), or time from diagnosis to most recent thyrotropin level (P = .23) (Table 3).

Of note, the VADHS changed the type of blood sample used to generate thyrotropin values from serum to plasma and had 3 different normal reference ranges during the 2010 to 2016 period studied. The thyrotropin values fell into 4 categories: serum sample with normal range 0.4 to 5.5 μIU/mL, serum sample with normal range 0.4 to 4.0 μIU/mL, plasma sample with normal range 0.4 to 4.0 μIU/mL, and plasma sample with normal range 0.6 to 4.8 μIU/mL. There were no significant differences between the abnormal and normal most recent thyrotropin groups in sample type for initial or most recent thyrotropin (P = .44 and P = .99, respectively) or in normal range for initial or most recent thyrotropin level (P = .99 and P = .09, respectively).

DISCUSSION

We found no statistically significant change in blood thyrotropin levels over time among patients with thyroid nodules with no history of medical conditions or medications known to affect thyroid hormone levels. Six of 100 patients developed abnormal thyrotropin, but only 2 eventually were treated for thyroid dysfunction: 1 for hypothyroidism and 1 for hyperthyroidism. The other 4 patients who did not receive treatment developed low thyrotropin but had no official diagnosis of hyperthyroidism in their health records, seemingly due to lack of multiple, consistently low thyrotropin values or due to lack of follow-up. Based on these data, monitoring thyrotropin over time may not be necessary in patients without any medical history known to affect thyroid function. The results provide support for the original hypothesis.

 

Although only thyrotropin values at the time of nodule diagnosis and most recent thyrotropin values were analyzed, thyrotropin trends over time were considered. Some patients did have transient abnormal thyrotropin values; however, a search of the patients’ records showed that these transient abnormalities did not lead to any initiation of hypothyroidism or hyperthyroidism treatment.

Another consideration is that changes in the sample type processed and in the normal thyrotropin ranges over time could have been confounding variables. However, statistical analyses showed that the abnormal and normal most recent thyrotropin groups did not show any significant differences in sample type or reference range for either the initial or most recent thyrotropin values. Hospitals change the laboratory assays they use for clinical tests over time, but these changes likely did not affect the results of this study.

The data from this study showed similar results to previously reported research. This study found that 6% of patients developed abnormal thyrotropin levels over time. A study of 157 patients with nonfunctioning benign thyroid nodules found that 8.3% of patients developed thyroid dysfunction.⁵ In another follow-up study on patients with thyroid nodules who were otherwise euthyroid, 2 of 118 patients eventually received treatment for hyperthyroidism.⁶ In the current study, we report that just 1 of 100 included patients had to begin treatment for hyperthyroidism.

The literature also includes research on using thyrotropin and age to predict malignancy in patients with thyroid nodules. One study suggested that a thyrotropin cutoff point of ≥ 2.1 mU/I and an age cutoff point of ≥ 47 years were significantly associated with a diagnosis of malignancy.⁷ Although the current study did not study malignancy, the results showed that the mean age at nodule diagnosis was higher in patients who had abnormal vs normal most recent thyrotropin levels: 69 vs 62 years, respectively. Future studies could determine whether a certain initial thyrotropin value or age could be used as a cutoff for requiring further thyrotropin monitoring to check for development of hyperthyroidism or hypothyroidism.

Limitations

This study was limited by its small size of 100 subjects. Most patients had to be excluded to focus on the aim of determining whether thyrotropin monitoring is needed in the specific group of patients without medical history that would be expected to affect thyroid function. Another limitation was that 83% of the patients included in the study were male, which does not reflect the general population. Future studies should include a greater number of patients and aim for a balance of 50% male and 50% female patients.

Additionally, it is important to note that the changing definition of the normal thyrotropin range was a limitation. It is possible that some patients who were considered normal at the time of a particular thyrotropin measurement may have had an abnormal reading if measured at a different time. Another consideration is that the VADHS changed the type of blood sample used to generate thyrotropin values from serum to plasma during the time that analyzed thyrotropin values were measured. This could have led to different thyrotropin values and, therefore, different results of this study compared with if the sample type had stayed the same. However, a previous study showed very similar thyrotropin values generated from serum and plasma samples in 17 patients.⁸ Therefore, possibly the change in sample type in the current study only minimally affected the results.

CONCLUSIONS

Current American Thyroid Association guidelines do not specify recommendations for follow-up thyrotropin testing in patients with thyroid nodules who do not have a history of conditions or medications known to affect thyroid hormone levels.¹ This study suggests that repeat thyrotropin monitoring may not be necessary for this group of patients. Individuals who had an abnormal most recent thyrotropin had an older age at thyroid nodule diagnosis compared with patients who had a normal most recent thyrotropin, so it is possible that thyrotropin monitoring may be recommended for people with nodules who are above a certain age. The results of this study as well as future studies could help create new clinical recommendations for thyrotropin monitoring in patients with thyroid nodules that clinicians can use to make evidence-based clinical decisions. There would also be a decreased financial, physical, and time burden on the patients if guidelines specify that they are not required to get continued blood thyrotropin testing.

Acknowledgments

The authors acknowledge Ronald J. Markert, PhD, formerly of Wright State University Boonshoft School of Medicine, for his contributions to the statistical analysis of this research.

When thyroid nodules are found clinically or incidentally on imaging, the patient’s thyrotropin level should be measured.¹ Ultrasound is the first-line imaging recommended to assess thyroid nodules.^1,2 Nodules can then be evaluated by a fine-needle aspiration (FNA) biopsy, which provides cytological information to determine whether the nodule is benign or malignant.^1,3,4 Most thyroid nodules pose a low risk of malignancy.¹

The American Thyroid Association guidelines on thyroid nodule management do not specify any recommendations for follow-up thyrotropin testing in patients who do not have any history that is known to affect thyroid function.¹ Therefore, clinicians have to make decisions regarding follow-up testing in these patients without any evidence-based guidelines. There is a lack of data in the literature on whether thyrotropin levels change over time in this patient population. If thyrotropin levels do not become abnormal over time, then patients would not need thyrotropin monitoring or treatment for hypo- or hyperthyroidism.

The aim of this study was to determine whether thyrotropin levels change over time in patients with thyroid nodules and determine whether repeat thyrotropin testing was required after initial testing. The authors hypothesized that thyrotropin values do not change substantially over time in patients with thyroid nodules, except in patients with a history of hot nodules, autoimmune thyroid disease, thyroid or pituitary surgery, radioactive iodine ablation, neck radiation, or use of medications affecting thyroid function. This study may be able to contribute to the clinical guidelines for thyrotropin testing in patients with thyroid nodules so that clinicians can make evidence-based decisions.

METHODS

This retrospective chart review was conducted using the Computerized Patient Record System at the Veterans Affairs Dayton Healthcare System (VADHS) in Ohio. Patients aged ≥ 18 years who were diagnosed with ≥ 1 thyroid nodule from January 2010 to December 2016 and had a normal thyrotropin level at the time of diagnosis were included in the study. Patients who were found to have thyroid nodules multiple times were included only once from the time of the initial diagnosis. Patients were excluded if they had a medical history known to affect thyroid function. Exclusion criteria included a history of hot thyroid nodules; autoimmune thyroid disease on imaging or blood work; history of thyroid surgery, including pituitary surgery; history of radioactive iodine treatment; history of neck radiation; use of thyroxine before nodule diagnosis; use of amiodarone, programmed cell death-1 inhibitors, programmed cell death ligand-1 inhibitors, or cytotoxic T-lymphocyte-associated protein-4 inhibitors; or 3 consecutive months of steroid use.

Age at nodule diagnosis, sex, race, thyrotropin values at and after the time of nodule diagnosis, and duration from nodule diagnosis to most recent thyrotropin value were retrospectively collected until 100 patients met inclusion criteria for the study. Of note, from 2010 to 2016, the assays used at the VADHS to measure thyrotropin values changed over time, as did the normal reference ranges and the type of sample used for the assays. Normal thyrotropin range at time of diagnosis based on serum or plasma samples and for repeat thyrotropin levels are provided in Table 1, also based on serum or plasma samples. All collected data in the study was de-identified for analysis.

Statistical Analysis

Patients were divided into 2 groups: those who had an abnormal most recent thyrotropin value and those who did not. Mean (SD) of both groups was calculated for continuous variables of age at diagnosis, initial thyrotropin value and most recent thyrotropin value, and time from diagnosis to most recent thyrotropin value. Percentages for both groups were calculated for categorical variables of sex, race, and whether initial and most recent thyrotropin values were based on serum or plasma samples and old or new reference ranges. A 95% CI was determined for the true population rate of patients with an abnormal thyrotropin value at most recent testing. Independent sample t tests were used to compare the continuous variables between the abnormal and normal most recent thyrotropin groups. Categorical variables between the 2 groups were compared using χ² tests. P < .05 was considered statistically significant. Statistical analyses were completed using IBM SPSS Statistics 27. This study was approved by the Wright State University Institutional Review Board and the VADHS Research and Development Committee.

 

 

RESULTS

Of 557 patient charts studied, 100 patients were included; the mean (SD) age at nodule diagnosis was 62.4 (11.1) years, and the mean (SD) initial thyrotropin level at nodule diagnosis was 1.51 (0.87) μIU/mL. The mean (SD) most recent thyrotropin level was 1.60 (1.03) μIU/mL after a mean duration of 5.7 (2.5) years postnodule diagnosis (Table 2).

Six patients (6%; 95% CI, 2.5%-12.7%) who had a normal thyrotropin level at nodule diagnosis developed an abnormal thyrotropin level in a mean (SD) of 6.9 (3.1) years. These 6 patients had a mean age at nodule diagnosis of 69.2 (11.4) years. Five of the 6 were male, and all were White patients. One patient’s thyrotropin level rose from an initial thyrotropin of 3.38 μIU/mL at nodule diagnosis to a high of 7.76 μIU/mL after 8.5 years. This patient was diagnosed with subclinical hypothyroidism and did not require treatment.

Five patients’ thyrotropin levels dropped below normal in a mean 7 years, with levels ranging from 0.25 to 0.52 μIU/mL. Of these patients, 2 became symptomatic from the nodules, experiencing dysphagia or hoarseness, with 1 diagnosed with hyperthyroidism. This patient was treated with methimazole and radioactive iodine ablation 9 years after diagnosis. The other 3 patients who developed low thyrotropin had no nodule symptoms or treatment. Ninety-four patients maintained thyrotropin values in the normal range for a mean (SD) of 5.7 (2.5) years and had a mean (SD) age at nodule diagnosis of 61.9 (11.0) years.

Both thyrotropin groups were compared. For categorical variables, there were no significant differences for sex (P = .99) or race (P = .55). For continuous variables, there were no significant differences in age at diagnosis (P = .12), initial thyrotropin (P = .24), most recent thyrotropin (P = .98), or time from diagnosis to most recent thyrotropin level (P = .23) (Table 3).

Of note, the VADHS changed the type of blood sample used to generate thyrotropin values from serum to plasma and had 3 different normal reference ranges during the 2010 to 2016 period studied. The thyrotropin values fell into 4 categories: serum sample with normal range 0.4 to 5.5 μIU/mL, serum sample with normal range 0.4 to 4.0 μIU/mL, plasma sample with normal range 0.4 to 4.0 μIU/mL, and plasma sample with normal range 0.6 to 4.8 μIU/mL. There were no significant differences between the abnormal and normal most recent thyrotropin groups in sample type for initial or most recent thyrotropin (P = .44 and P = .99, respectively) or in normal range for initial or most recent thyrotropin level (P = .99 and P = .09, respectively).

DISCUSSION

We found no statistically significant change in blood thyrotropin levels over time among patients with thyroid nodules with no history of medical conditions or medications known to affect thyroid hormone levels. Six of 100 patients developed abnormal thyrotropin, but only 2 eventually were treated for thyroid dysfunction: 1 for hypothyroidism and 1 for hyperthyroidism. The other 4 patients who did not receive treatment developed low thyrotropin but had no official diagnosis of hyperthyroidism in their health records, seemingly due to lack of multiple, consistently low thyrotropin values or due to lack of follow-up. Based on these data, monitoring thyrotropin over time may not be necessary in patients without any medical history known to affect thyroid function. The results provide support for the original hypothesis.

 

Although only thyrotropin values at the time of nodule diagnosis and most recent thyrotropin values were analyzed, thyrotropin trends over time were considered. Some patients did have transient abnormal thyrotropin values; however, a search of the patients’ records showed that these transient abnormalities did not lead to any initiation of hypothyroidism or hyperthyroidism treatment.

Another consideration is that changes in the sample type processed and in the normal thyrotropin ranges over time could have been confounding variables. However, statistical analyses showed that the abnormal and normal most recent thyrotropin groups did not show any significant differences in sample type or reference range for either the initial or most recent thyrotropin values. Hospitals change the laboratory assays they use for clinical tests over time, but these changes likely did not affect the results of this study.

The data from this study showed similar results to previously reported research. This study found that 6% of patients developed abnormal thyrotropin levels over time. A study of 157 patients with nonfunctioning benign thyroid nodules found that 8.3% of patients developed thyroid dysfunction.⁵ In another follow-up study on patients with thyroid nodules who were otherwise euthyroid, 2 of 118 patients eventually received treatment for hyperthyroidism.⁶ In the current study, we report that just 1 of 100 included patients had to begin treatment for hyperthyroidism.

The literature also includes research on using thyrotropin and age to predict malignancy in patients with thyroid nodules. One study suggested that a thyrotropin cutoff point of ≥ 2.1 mU/I and an age cutoff point of ≥ 47 years were significantly associated with a diagnosis of malignancy.⁷ Although the current study did not study malignancy, the results showed that the mean age at nodule diagnosis was higher in patients who had abnormal vs normal most recent thyrotropin levels: 69 vs 62 years, respectively. Future studies could determine whether a certain initial thyrotropin value or age could be used as a cutoff for requiring further thyrotropin monitoring to check for development of hyperthyroidism or hypothyroidism.

Limitations

This study was limited by its small size of 100 subjects. Most patients had to be excluded to focus on the aim of determining whether thyrotropin monitoring is needed in the specific group of patients without medical history that would be expected to affect thyroid function. Another limitation was that 83% of the patients included in the study were male, which does not reflect the general population. Future studies should include a greater number of patients and aim for a balance of 50% male and 50% female patients.

Additionally, it is important to note that the changing definition of the normal thyrotropin range was a limitation. It is possible that some patients who were considered normal at the time of a particular thyrotropin measurement may have had an abnormal reading if measured at a different time. Another consideration is that the VADHS changed the type of blood sample used to generate thyrotropin values from serum to plasma during the time that analyzed thyrotropin values were measured. This could have led to different thyrotropin values and, therefore, different results of this study compared with if the sample type had stayed the same. However, a previous study showed very similar thyrotropin values generated from serum and plasma samples in 17 patients.⁸ Therefore, possibly the change in sample type in the current study only minimally affected the results.

CONCLUSIONS

Current American Thyroid Association guidelines do not specify recommendations for follow-up thyrotropin testing in patients with thyroid nodules who do not have a history of conditions or medications known to affect thyroid hormone levels.¹ This study suggests that repeat thyrotropin monitoring may not be necessary for this group of patients. Individuals who had an abnormal most recent thyrotropin had an older age at thyroid nodule diagnosis compared with patients who had a normal most recent thyrotropin, so it is possible that thyrotropin monitoring may be recommended for people with nodules who are above a certain age. The results of this study as well as future studies could help create new clinical recommendations for thyrotropin monitoring in patients with thyroid nodules that clinicians can use to make evidence-based clinical decisions. There would also be a decreased financial, physical, and time burden on the patients if guidelines specify that they are not required to get continued blood thyrotropin testing.

Acknowledgments

The authors acknowledge Ronald J. Markert, PhD, formerly of Wright State University Boonshoft School of Medicine, for his contributions to the statistical analysis of this research.

When thyroid nodules are found clinically or incidentally on imaging, the patient’s thyrotropin level should be measured.¹ Ultrasound is the first-line imaging recommended to assess thyroid nodules.^1,2 Nodules can then be evaluated by a fine-needle aspiration (FNA) biopsy, which provides cytological information to determine whether the nodule is benign or malignant.^1,3,4 Most thyroid nodules pose a low risk of malignancy.¹

The American Thyroid Association guidelines on thyroid nodule management do not specify any recommendations for follow-up thyrotropin testing in patients who do not have any history that is known to affect thyroid function.¹ Therefore, clinicians have to make decisions regarding follow-up testing in these patients without any evidence-based guidelines. There is a lack of data in the literature on whether thyrotropin levels change over time in this patient population. If thyrotropin levels do not become abnormal over time, then patients would not need thyrotropin monitoring or treatment for hypo- or hyperthyroidism.

The aim of this study was to determine whether thyrotropin levels change over time in patients with thyroid nodules and determine whether repeat thyrotropin testing was required after initial testing. The authors hypothesized that thyrotropin values do not change substantially over time in patients with thyroid nodules, except in patients with a history of hot nodules, autoimmune thyroid disease, thyroid or pituitary surgery, radioactive iodine ablation, neck radiation, or use of medications affecting thyroid function. This study may be able to contribute to the clinical guidelines for thyrotropin testing in patients with thyroid nodules so that clinicians can make evidence-based decisions.

METHODS

This retrospective chart review was conducted using the Computerized Patient Record System at the Veterans Affairs Dayton Healthcare System (VADHS) in Ohio. Patients aged ≥ 18 years who were diagnosed with ≥ 1 thyroid nodule from January 2010 to December 2016 and had a normal thyrotropin level at the time of diagnosis were included in the study. Patients who were found to have thyroid nodules multiple times were included only once from the time of the initial diagnosis. Patients were excluded if they had a medical history known to affect thyroid function. Exclusion criteria included a history of hot thyroid nodules; autoimmune thyroid disease on imaging or blood work; history of thyroid surgery, including pituitary surgery; history of radioactive iodine treatment; history of neck radiation; use of thyroxine before nodule diagnosis; use of amiodarone, programmed cell death-1 inhibitors, programmed cell death ligand-1 inhibitors, or cytotoxic T-lymphocyte-associated protein-4 inhibitors; or 3 consecutive months of steroid use.

Age at nodule diagnosis, sex, race, thyrotropin values at and after the time of nodule diagnosis, and duration from nodule diagnosis to most recent thyrotropin value were retrospectively collected until 100 patients met inclusion criteria for the study. Of note, from 2010 to 2016, the assays used at the VADHS to measure thyrotropin values changed over time, as did the normal reference ranges and the type of sample used for the assays. Normal thyrotropin range at time of diagnosis based on serum or plasma samples and for repeat thyrotropin levels are provided in Table 1, also based on serum or plasma samples. All collected data in the study was de-identified for analysis.

Statistical Analysis

Patients were divided into 2 groups: those who had an abnormal most recent thyrotropin value and those who did not. Mean (SD) of both groups was calculated for continuous variables of age at diagnosis, initial thyrotropin value and most recent thyrotropin value, and time from diagnosis to most recent thyrotropin value. Percentages for both groups were calculated for categorical variables of sex, race, and whether initial and most recent thyrotropin values were based on serum or plasma samples and old or new reference ranges. A 95% CI was determined for the true population rate of patients with an abnormal thyrotropin value at most recent testing. Independent sample t tests were used to compare the continuous variables between the abnormal and normal most recent thyrotropin groups. Categorical variables between the 2 groups were compared using χ² tests. P < .05 was considered statistically significant. Statistical analyses were completed using IBM SPSS Statistics 27. This study was approved by the Wright State University Institutional Review Board and the VADHS Research and Development Committee.

 

 

RESULTS

Of 557 patient charts studied, 100 patients were included; the mean (SD) age at nodule diagnosis was 62.4 (11.1) years, and the mean (SD) initial thyrotropin level at nodule diagnosis was 1.51 (0.87) μIU/mL. The mean (SD) most recent thyrotropin level was 1.60 (1.03) μIU/mL after a mean duration of 5.7 (2.5) years postnodule diagnosis (Table 2).

Six patients (6%; 95% CI, 2.5%-12.7%) who had a normal thyrotropin level at nodule diagnosis developed an abnormal thyrotropin level in a mean (SD) of 6.9 (3.1) years. These 6 patients had a mean age at nodule diagnosis of 69.2 (11.4) years. Five of the 6 were male, and all were White patients. One patient’s thyrotropin level rose from an initial thyrotropin of 3.38 μIU/mL at nodule diagnosis to a high of 7.76 μIU/mL after 8.5 years. This patient was diagnosed with subclinical hypothyroidism and did not require treatment.

Five patients’ thyrotropin levels dropped below normal in a mean 7 years, with levels ranging from 0.25 to 0.52 μIU/mL. Of these patients, 2 became symptomatic from the nodules, experiencing dysphagia or hoarseness, with 1 diagnosed with hyperthyroidism. This patient was treated with methimazole and radioactive iodine ablation 9 years after diagnosis. The other 3 patients who developed low thyrotropin had no nodule symptoms or treatment. Ninety-four patients maintained thyrotropin values in the normal range for a mean (SD) of 5.7 (2.5) years and had a mean (SD) age at nodule diagnosis of 61.9 (11.0) years.

Both thyrotropin groups were compared. For categorical variables, there were no significant differences for sex (P = .99) or race (P = .55). For continuous variables, there were no significant differences in age at diagnosis (P = .12), initial thyrotropin (P = .24), most recent thyrotropin (P = .98), or time from diagnosis to most recent thyrotropin level (P = .23) (Table 3).

Of note, the VADHS changed the type of blood sample used to generate thyrotropin values from serum to plasma and had 3 different normal reference ranges during the 2010 to 2016 period studied. The thyrotropin values fell into 4 categories: serum sample with normal range 0.4 to 5.5 μIU/mL, serum sample with normal range 0.4 to 4.0 μIU/mL, plasma sample with normal range 0.4 to 4.0 μIU/mL, and plasma sample with normal range 0.6 to 4.8 μIU/mL. There were no significant differences between the abnormal and normal most recent thyrotropin groups in sample type for initial or most recent thyrotropin (P = .44 and P = .99, respectively) or in normal range for initial or most recent thyrotropin level (P = .99 and P = .09, respectively).

DISCUSSION

We found no statistically significant change in blood thyrotropin levels over time among patients with thyroid nodules with no history of medical conditions or medications known to affect thyroid hormone levels. Six of 100 patients developed abnormal thyrotropin, but only 2 eventually were treated for thyroid dysfunction: 1 for hypothyroidism and 1 for hyperthyroidism. The other 4 patients who did not receive treatment developed low thyrotropin but had no official diagnosis of hyperthyroidism in their health records, seemingly due to lack of multiple, consistently low thyrotropin values or due to lack of follow-up. Based on these data, monitoring thyrotropin over time may not be necessary in patients without any medical history known to affect thyroid function. The results provide support for the original hypothesis.

 

Although only thyrotropin values at the time of nodule diagnosis and most recent thyrotropin values were analyzed, thyrotropin trends over time were considered. Some patients did have transient abnormal thyrotropin values; however, a search of the patients’ records showed that these transient abnormalities did not lead to any initiation of hypothyroidism or hyperthyroidism treatment.

Another consideration is that changes in the sample type processed and in the normal thyrotropin ranges over time could have been confounding variables. However, statistical analyses showed that the abnormal and normal most recent thyrotropin groups did not show any significant differences in sample type or reference range for either the initial or most recent thyrotropin values. Hospitals change the laboratory assays they use for clinical tests over time, but these changes likely did not affect the results of this study.

The data from this study showed similar results to previously reported research. This study found that 6% of patients developed abnormal thyrotropin levels over time. A study of 157 patients with nonfunctioning benign thyroid nodules found that 8.3% of patients developed thyroid dysfunction.⁵ In another follow-up study on patients with thyroid nodules who were otherwise euthyroid, 2 of 118 patients eventually received treatment for hyperthyroidism.⁶ In the current study, we report that just 1 of 100 included patients had to begin treatment for hyperthyroidism.

The literature also includes research on using thyrotropin and age to predict malignancy in patients with thyroid nodules. One study suggested that a thyrotropin cutoff point of ≥ 2.1 mU/I and an age cutoff point of ≥ 47 years were significantly associated with a diagnosis of malignancy.⁷ Although the current study did not study malignancy, the results showed that the mean age at nodule diagnosis was higher in patients who had abnormal vs normal most recent thyrotropin levels: 69 vs 62 years, respectively. Future studies could determine whether a certain initial thyrotropin value or age could be used as a cutoff for requiring further thyrotropin monitoring to check for development of hyperthyroidism or hypothyroidism.

Limitations

This study was limited by its small size of 100 subjects. Most patients had to be excluded to focus on the aim of determining whether thyrotropin monitoring is needed in the specific group of patients without medical history that would be expected to affect thyroid function. Another limitation was that 83% of the patients included in the study were male, which does not reflect the general population. Future studies should include a greater number of patients and aim for a balance of 50% male and 50% female patients.

Additionally, it is important to note that the changing definition of the normal thyrotropin range was a limitation. It is possible that some patients who were considered normal at the time of a particular thyrotropin measurement may have had an abnormal reading if measured at a different time. Another consideration is that the VADHS changed the type of blood sample used to generate thyrotropin values from serum to plasma during the time that analyzed thyrotropin values were measured. This could have led to different thyrotropin values and, therefore, different results of this study compared with if the sample type had stayed the same. However, a previous study showed very similar thyrotropin values generated from serum and plasma samples in 17 patients.⁸ Therefore, possibly the change in sample type in the current study only minimally affected the results.

CONCLUSIONS

Current American Thyroid Association guidelines do not specify recommendations for follow-up thyrotropin testing in patients with thyroid nodules who do not have a history of conditions or medications known to affect thyroid hormone levels.¹ This study suggests that repeat thyrotropin monitoring may not be necessary for this group of patients. Individuals who had an abnormal most recent thyrotropin had an older age at thyroid nodule diagnosis compared with patients who had a normal most recent thyrotropin, so it is possible that thyrotropin monitoring may be recommended for people with nodules who are above a certain age. The results of this study as well as future studies could help create new clinical recommendations for thyrotropin monitoring in patients with thyroid nodules that clinicians can use to make evidence-based clinical decisions. There would also be a decreased financial, physical, and time burden on the patients if guidelines specify that they are not required to get continued blood thyrotropin testing.

Acknowledgments

The authors acknowledge Ronald J. Markert, PhD, formerly of Wright State University Boonshoft School of Medicine, for his contributions to the statistical analysis of this research.

Patients planning plastic surgery traditionally were instructed to stop anticoagulants and antiplatelet medications during the perioperative period to avoid bleeding, which could result in flap loss, pain, skin necrosis, and blood transfusions. In the veteran patient population, anticoagulants are prescribed for the prevention of limb- and life-threatening embolic and thrombotic events.^1-3 As of June 2021, > 332,000 veterans were prescribed direct oral anticoagulants.¹

In 2015, the Malcom Randall Veterans Affairs Medical Center (MRVAMC) in Gainesville, Florida, Plastic Surgery Service began instructing patients planning elective hand surgery to continue their prescription anticoagulants and antiplatelets during the perioperative period. This decision was prompted by a patient who needed carpal tunnel release surgery and was prescribed coumadin for repeated thrombosis of his dialysis grafts. Hand surgery literature at the time suggested allowing patients to continue their anticoagulants and antiplatelets through the perioperative period to avoid life- and limb-threatening events and wide fluctuations in blood anticoagulant levels.^4-6 The MRVAMC Plastic Surgery Service chose to accept the risk of perioperative bleeding after shared decision making with the patients rather than risk a cardiac stent obstruction, pulmonary embolism, or embolic stroke in the at-risk patients.

The objective of this study was to determine the postoperative bleeding complication rate over a 7.5-year period in the veteran patients who did not interrupt their prescription blood thinners. This would assist the MRVAMC Plastic Surgery Service with providing data-driven informed consent and determine whether this protocol should continue.

Methods

The North Florida/South Georgia Veterans Health System Research Committee and the University of Florida Institutional Review Board approved a retrospective chart review of elective hand cases performed by the MRVAMC Plastic Surgery Service from January 1, 2015, through June 30, 2022. Elective hand cases were identified based on the operation description and included nerve decompressions, tendon releases, trapeziectomy, small-joint fusion, neurectomy, elective amputations, and benign neoplasm removals (Table). Hand surgery included cubital tunnel releases (decompression of the ulnar nerve at the level of the elbow) because hand surgery fellowships, hand surgery training, and hand surgery practices traditionally include a high volume of cubital tunnel releases. We wanted this study to have real-world applications.

Patients’ histories and physicals were reviewed for prescription antithrombotics and for instructions not to interrupt these medications. Postoperative notes were reviewed for 30 days for evidence of postoperative bleeding complications.

The following prescription anticoagulants were included in the study: dabigatran, rivaroxaban, warfarin, edoxaban, and apixaban. In addition, the following prescription antiplatelets were included in the study: clopidogrel, aspirin/dipyridamole, prasugrel, cilostazol, and ticagrelor. Indications for the medications included a history of thromboembolic events, cardiac stents, cerebrovascular disease, atrial fibrillation, hypercoagulable states, and mechanical valves. Over-the-counter antiplatelet medications, such as aspirin and ibuprofen, were not included as a standalone medication for accuracy because patients taking those medications may not be captured in the electronic health record review.

Results

One hundred seventy-eight patients were identified for maintaining prescription blood thinners during their elective hand surgery. There was 1 major complication (0.6%) and 4 minor bleeding complications (2.2%). The major complication occurred when a patient had to return to surgery from the recovery room for emergent control of bleeding. The surgery was for an in situ cubital tunnel release. The patient, aged 48 years, was taking clopidogrel and aspirin and had a personal and family history of cardiovascular disease. The bleeding was controlled with bipolar cautery and Floseal, a topical haemostatic matrix made of bovine gelatin and human thrombin. The minor bleeding complications were treated in the clinic with compression, wound care, or expedited follow-up for reassurance. These included an in situ cubital tunnel release for a patient taking warfarin and aspirin, a digital inclusion cyst for a patient taking apixaban, an endoscopic carpal tunnel for a patient taking aspirin and clopidogrel, and an open carpal tunnel and ulnar tunnel release for a patient taking aspirin and clopidogrel. There were no thrombotic events during the study.

Discussion

Higher utilization of anticoagulation has been evidenced by a 30% increase in Medicare claims and a 277% increase in Medicaid anticoagulation claims between 2014 and 2019, driven by more prescriptions for direct oral anticoagulants such as apixaban and rivaroxaban.⁷ The MRVAMC Plastic Surgery Service began a protocol for managing perioperative anticoagulation in 2015 to avoid the risk of perioperative thrombotic events in veteran patients. Patients who choose elective hand surgery were instructed to continue their prescription blood thinners. Exceptions to this protocol were patients scheduled for a partial fasciectomy (for Dupuytren contracture) or cubital tunnel release with anterior ulnar nerve transposition. A hematoma would increase the risk for skin necrosis in the patients receiving a fasciectomy, resulting from the thin skin flaps and meticulous dissection to identify and protect the digital nerves. Worsening nerve dysfunction could result from hematoma compression and scarring in the ulnar nerve cases. If the risk of holding the blood thinner was felt to be unreasonably high, based on recommendations from the patients’ cardiologist or primary care doctor, we offered an in situ cubital tunnel release for the ulnar nerve patients.

 

Concerns regarding interrupting chronic anticoagulation involve the increased risk of thromboembolism and the theoretical risk of a rebound hypercoagulable effect.⁸ Patients prescribed warfarin have been found to unintentionally discontinue this medication after outpatient surgery at more than 1.5 times the rate of the general population.⁹

A systematic review of 9 published studies looking specifically at elective hand and wrist surgeries demonstrated no significant increase in perioperative bleeding risk with the continuation of anticoagulation and antiplatelet medications.¹⁰ Sardenberg and colleagues reviewed 7 studies in which 410 hand and wrist surgeries were performed in patients prescribed warfarin or aspirin and clopidogrel. These patients had a 0.7% serious complication rate, requiring surgical treatment only in patients having complex wrist surgeries (wrist arthrodesis with tenosynovectomy, resection of the distal ulna with tenosynovectomy and tendon transfer, and proximal row carpectomy).¹¹ Bogunovic and colleagues compared 50 hand and wrist patients who were on uninterrupted warfarin with those who were not. They required patients to have an international normalized ratio (INR) < 3.5, but 1 patient required a return to the operating room for a bleeding complication due to an INR of 5.4 on postoperative day 4. They caution vigilant monitoring of INR.¹²

These and our study are consistent with other disciplines, such as facial plastic surgery, dermatology, and ophthalmology, which do not support routine suspension of anticoagulants.^13-16 A review of 30 cutaneous surgery studies involving > 14,000 patients recommended meticulous hemostasis over cessation of blood thinners.¹⁵ The University of Massachusetts Dermatology Clinic found a 40 times higher rate of bleeding complications in patients on clopidogrel and warfarin but still recommended continuation of these medications to avoid thrombotic events.¹⁶

Limitations

This study is a retrospective chart review and limited by what is already documented in the electronic health record. We can verify that the patients were given instructions to continue their medications up to the day of surgery but cannot be certain whether the instructions were followed. No control group was told to hold their anticoagulants for the same surgery. Once we decided on a protocol, we applied it to all patients. The study approval was for the specific time frame when the protocol was in place.

Our study was designed for elective hand cases because those surgeries can be anticipated, predicted, and patients can be given instructions during the preoperative appointments. We did incidentally find several nonelective hand cases (traumas, infections, and cancers) during the review of patients taking prescription blood thinners that had to be expedited to the operating room. Based on morbidity data during that time period, there were no additional postoperative hand surgery bleeding complications that had to return to the operating room. Future studies are indicated, but we believe our protocol can be applied to urgent and emergent hand surgeries as well as elective cases.

Conclusions

Our study supports continuing prescription anticoagulant and antiplatelet medications during the perioperative period for elective hand surgery. We found this is a safe practice in our veteran population with an acceptably low local bleeding complication rate.

Acknowledgments

This manuscript is the result of work supported with the resources and the use of facilities at the North Florida/South Georgia Veterans Health System in Gainesville, Florida.

Patients planning plastic surgery traditionally were instructed to stop anticoagulants and antiplatelet medications during the perioperative period to avoid bleeding, which could result in flap loss, pain, skin necrosis, and blood transfusions. In the veteran patient population, anticoagulants are prescribed for the prevention of limb- and life-threatening embolic and thrombotic events.^1-3 As of June 2021, > 332,000 veterans were prescribed direct oral anticoagulants.¹

In 2015, the Malcom Randall Veterans Affairs Medical Center (MRVAMC) in Gainesville, Florida, Plastic Surgery Service began instructing patients planning elective hand surgery to continue their prescription anticoagulants and antiplatelets during the perioperative period. This decision was prompted by a patient who needed carpal tunnel release surgery and was prescribed coumadin for repeated thrombosis of his dialysis grafts. Hand surgery literature at the time suggested allowing patients to continue their anticoagulants and antiplatelets through the perioperative period to avoid life- and limb-threatening events and wide fluctuations in blood anticoagulant levels.^4-6 The MRVAMC Plastic Surgery Service chose to accept the risk of perioperative bleeding after shared decision making with the patients rather than risk a cardiac stent obstruction, pulmonary embolism, or embolic stroke in the at-risk patients.

The objective of this study was to determine the postoperative bleeding complication rate over a 7.5-year period in the veteran patients who did not interrupt their prescription blood thinners. This would assist the MRVAMC Plastic Surgery Service with providing data-driven informed consent and determine whether this protocol should continue.

Methods

The North Florida/South Georgia Veterans Health System Research Committee and the University of Florida Institutional Review Board approved a retrospective chart review of elective hand cases performed by the MRVAMC Plastic Surgery Service from January 1, 2015, through June 30, 2022. Elective hand cases were identified based on the operation description and included nerve decompressions, tendon releases, trapeziectomy, small-joint fusion, neurectomy, elective amputations, and benign neoplasm removals (Table). Hand surgery included cubital tunnel releases (decompression of the ulnar nerve at the level of the elbow) because hand surgery fellowships, hand surgery training, and hand surgery practices traditionally include a high volume of cubital tunnel releases. We wanted this study to have real-world applications.

Patients’ histories and physicals were reviewed for prescription antithrombotics and for instructions not to interrupt these medications. Postoperative notes were reviewed for 30 days for evidence of postoperative bleeding complications.

The following prescription anticoagulants were included in the study: dabigatran, rivaroxaban, warfarin, edoxaban, and apixaban. In addition, the following prescription antiplatelets were included in the study: clopidogrel, aspirin/dipyridamole, prasugrel, cilostazol, and ticagrelor. Indications for the medications included a history of thromboembolic events, cardiac stents, cerebrovascular disease, atrial fibrillation, hypercoagulable states, and mechanical valves. Over-the-counter antiplatelet medications, such as aspirin and ibuprofen, were not included as a standalone medication for accuracy because patients taking those medications may not be captured in the electronic health record review.

Results

One hundred seventy-eight patients were identified for maintaining prescription blood thinners during their elective hand surgery. There was 1 major complication (0.6%) and 4 minor bleeding complications (2.2%). The major complication occurred when a patient had to return to surgery from the recovery room for emergent control of bleeding. The surgery was for an in situ cubital tunnel release. The patient, aged 48 years, was taking clopidogrel and aspirin and had a personal and family history of cardiovascular disease. The bleeding was controlled with bipolar cautery and Floseal, a topical haemostatic matrix made of bovine gelatin and human thrombin. The minor bleeding complications were treated in the clinic with compression, wound care, or expedited follow-up for reassurance. These included an in situ cubital tunnel release for a patient taking warfarin and aspirin, a digital inclusion cyst for a patient taking apixaban, an endoscopic carpal tunnel for a patient taking aspirin and clopidogrel, and an open carpal tunnel and ulnar tunnel release for a patient taking aspirin and clopidogrel. There were no thrombotic events during the study.

Discussion

Higher utilization of anticoagulation has been evidenced by a 30% increase in Medicare claims and a 277% increase in Medicaid anticoagulation claims between 2014 and 2019, driven by more prescriptions for direct oral anticoagulants such as apixaban and rivaroxaban.⁷ The MRVAMC Plastic Surgery Service began a protocol for managing perioperative anticoagulation in 2015 to avoid the risk of perioperative thrombotic events in veteran patients. Patients who choose elective hand surgery were instructed to continue their prescription blood thinners. Exceptions to this protocol were patients scheduled for a partial fasciectomy (for Dupuytren contracture) or cubital tunnel release with anterior ulnar nerve transposition. A hematoma would increase the risk for skin necrosis in the patients receiving a fasciectomy, resulting from the thin skin flaps and meticulous dissection to identify and protect the digital nerves. Worsening nerve dysfunction could result from hematoma compression and scarring in the ulnar nerve cases. If the risk of holding the blood thinner was felt to be unreasonably high, based on recommendations from the patients’ cardiologist or primary care doctor, we offered an in situ cubital tunnel release for the ulnar nerve patients.

 

Concerns regarding interrupting chronic anticoagulation involve the increased risk of thromboembolism and the theoretical risk of a rebound hypercoagulable effect.⁸ Patients prescribed warfarin have been found to unintentionally discontinue this medication after outpatient surgery at more than 1.5 times the rate of the general population.⁹

A systematic review of 9 published studies looking specifically at elective hand and wrist surgeries demonstrated no significant increase in perioperative bleeding risk with the continuation of anticoagulation and antiplatelet medications.¹⁰ Sardenberg and colleagues reviewed 7 studies in which 410 hand and wrist surgeries were performed in patients prescribed warfarin or aspirin and clopidogrel. These patients had a 0.7% serious complication rate, requiring surgical treatment only in patients having complex wrist surgeries (wrist arthrodesis with tenosynovectomy, resection of the distal ulna with tenosynovectomy and tendon transfer, and proximal row carpectomy).¹¹ Bogunovic and colleagues compared 50 hand and wrist patients who were on uninterrupted warfarin with those who were not. They required patients to have an international normalized ratio (INR) < 3.5, but 1 patient required a return to the operating room for a bleeding complication due to an INR of 5.4 on postoperative day 4. They caution vigilant monitoring of INR.¹²

These and our study are consistent with other disciplines, such as facial plastic surgery, dermatology, and ophthalmology, which do not support routine suspension of anticoagulants.^13-16 A review of 30 cutaneous surgery studies involving > 14,000 patients recommended meticulous hemostasis over cessation of blood thinners.¹⁵ The University of Massachusetts Dermatology Clinic found a 40 times higher rate of bleeding complications in patients on clopidogrel and warfarin but still recommended continuation of these medications to avoid thrombotic events.¹⁶

Limitations

This study is a retrospective chart review and limited by what is already documented in the electronic health record. We can verify that the patients were given instructions to continue their medications up to the day of surgery but cannot be certain whether the instructions were followed. No control group was told to hold their anticoagulants for the same surgery. Once we decided on a protocol, we applied it to all patients. The study approval was for the specific time frame when the protocol was in place.

Our study was designed for elective hand cases because those surgeries can be anticipated, predicted, and patients can be given instructions during the preoperative appointments. We did incidentally find several nonelective hand cases (traumas, infections, and cancers) during the review of patients taking prescription blood thinners that had to be expedited to the operating room. Based on morbidity data during that time period, there were no additional postoperative hand surgery bleeding complications that had to return to the operating room. Future studies are indicated, but we believe our protocol can be applied to urgent and emergent hand surgeries as well as elective cases.

Conclusions

Our study supports continuing prescription anticoagulant and antiplatelet medications during the perioperative period for elective hand surgery. We found this is a safe practice in our veteran population with an acceptably low local bleeding complication rate.

Acknowledgments

This manuscript is the result of work supported with the resources and the use of facilities at the North Florida/South Georgia Veterans Health System in Gainesville, Florida.

Patients planning plastic surgery traditionally were instructed to stop anticoagulants and antiplatelet medications during the perioperative period to avoid bleeding, which could result in flap loss, pain, skin necrosis, and blood transfusions. In the veteran patient population, anticoagulants are prescribed for the prevention of limb- and life-threatening embolic and thrombotic events.^1-3 As of June 2021, > 332,000 veterans were prescribed direct oral anticoagulants.¹

In 2015, the Malcom Randall Veterans Affairs Medical Center (MRVAMC) in Gainesville, Florida, Plastic Surgery Service began instructing patients planning elective hand surgery to continue their prescription anticoagulants and antiplatelets during the perioperative period. This decision was prompted by a patient who needed carpal tunnel release surgery and was prescribed coumadin for repeated thrombosis of his dialysis grafts. Hand surgery literature at the time suggested allowing patients to continue their anticoagulants and antiplatelets through the perioperative period to avoid life- and limb-threatening events and wide fluctuations in blood anticoagulant levels.^4-6 The MRVAMC Plastic Surgery Service chose to accept the risk of perioperative bleeding after shared decision making with the patients rather than risk a cardiac stent obstruction, pulmonary embolism, or embolic stroke in the at-risk patients.

The objective of this study was to determine the postoperative bleeding complication rate over a 7.5-year period in the veteran patients who did not interrupt their prescription blood thinners. This would assist the MRVAMC Plastic Surgery Service with providing data-driven informed consent and determine whether this protocol should continue.

Methods

The North Florida/South Georgia Veterans Health System Research Committee and the University of Florida Institutional Review Board approved a retrospective chart review of elective hand cases performed by the MRVAMC Plastic Surgery Service from January 1, 2015, through June 30, 2022. Elective hand cases were identified based on the operation description and included nerve decompressions, tendon releases, trapeziectomy, small-joint fusion, neurectomy, elective amputations, and benign neoplasm removals (Table). Hand surgery included cubital tunnel releases (decompression of the ulnar nerve at the level of the elbow) because hand surgery fellowships, hand surgery training, and hand surgery practices traditionally include a high volume of cubital tunnel releases. We wanted this study to have real-world applications.

Patients’ histories and physicals were reviewed for prescription antithrombotics and for instructions not to interrupt these medications. Postoperative notes were reviewed for 30 days for evidence of postoperative bleeding complications.

The following prescription anticoagulants were included in the study: dabigatran, rivaroxaban, warfarin, edoxaban, and apixaban. In addition, the following prescription antiplatelets were included in the study: clopidogrel, aspirin/dipyridamole, prasugrel, cilostazol, and ticagrelor. Indications for the medications included a history of thromboembolic events, cardiac stents, cerebrovascular disease, atrial fibrillation, hypercoagulable states, and mechanical valves. Over-the-counter antiplatelet medications, such as aspirin and ibuprofen, were not included as a standalone medication for accuracy because patients taking those medications may not be captured in the electronic health record review.

Results

One hundred seventy-eight patients were identified for maintaining prescription blood thinners during their elective hand surgery. There was 1 major complication (0.6%) and 4 minor bleeding complications (2.2%). The major complication occurred when a patient had to return to surgery from the recovery room for emergent control of bleeding. The surgery was for an in situ cubital tunnel release. The patient, aged 48 years, was taking clopidogrel and aspirin and had a personal and family history of cardiovascular disease. The bleeding was controlled with bipolar cautery and Floseal, a topical haemostatic matrix made of bovine gelatin and human thrombin. The minor bleeding complications were treated in the clinic with compression, wound care, or expedited follow-up for reassurance. These included an in situ cubital tunnel release for a patient taking warfarin and aspirin, a digital inclusion cyst for a patient taking apixaban, an endoscopic carpal tunnel for a patient taking aspirin and clopidogrel, and an open carpal tunnel and ulnar tunnel release for a patient taking aspirin and clopidogrel. There were no thrombotic events during the study.

Discussion

Higher utilization of anticoagulation has been evidenced by a 30% increase in Medicare claims and a 277% increase in Medicaid anticoagulation claims between 2014 and 2019, driven by more prescriptions for direct oral anticoagulants such as apixaban and rivaroxaban.⁷ The MRVAMC Plastic Surgery Service began a protocol for managing perioperative anticoagulation in 2015 to avoid the risk of perioperative thrombotic events in veteran patients. Patients who choose elective hand surgery were instructed to continue their prescription blood thinners. Exceptions to this protocol were patients scheduled for a partial fasciectomy (for Dupuytren contracture) or cubital tunnel release with anterior ulnar nerve transposition. A hematoma would increase the risk for skin necrosis in the patients receiving a fasciectomy, resulting from the thin skin flaps and meticulous dissection to identify and protect the digital nerves. Worsening nerve dysfunction could result from hematoma compression and scarring in the ulnar nerve cases. If the risk of holding the blood thinner was felt to be unreasonably high, based on recommendations from the patients’ cardiologist or primary care doctor, we offered an in situ cubital tunnel release for the ulnar nerve patients.

 

Concerns regarding interrupting chronic anticoagulation involve the increased risk of thromboembolism and the theoretical risk of a rebound hypercoagulable effect.⁸ Patients prescribed warfarin have been found to unintentionally discontinue this medication after outpatient surgery at more than 1.5 times the rate of the general population.⁹

A systematic review of 9 published studies looking specifically at elective hand and wrist surgeries demonstrated no significant increase in perioperative bleeding risk with the continuation of anticoagulation and antiplatelet medications.¹⁰ Sardenberg and colleagues reviewed 7 studies in which 410 hand and wrist surgeries were performed in patients prescribed warfarin or aspirin and clopidogrel. These patients had a 0.7% serious complication rate, requiring surgical treatment only in patients having complex wrist surgeries (wrist arthrodesis with tenosynovectomy, resection of the distal ulna with tenosynovectomy and tendon transfer, and proximal row carpectomy).¹¹ Bogunovic and colleagues compared 50 hand and wrist patients who were on uninterrupted warfarin with those who were not. They required patients to have an international normalized ratio (INR) < 3.5, but 1 patient required a return to the operating room for a bleeding complication due to an INR of 5.4 on postoperative day 4. They caution vigilant monitoring of INR.¹²

These and our study are consistent with other disciplines, such as facial plastic surgery, dermatology, and ophthalmology, which do not support routine suspension of anticoagulants.^13-16 A review of 30 cutaneous surgery studies involving > 14,000 patients recommended meticulous hemostasis over cessation of blood thinners.¹⁵ The University of Massachusetts Dermatology Clinic found a 40 times higher rate of bleeding complications in patients on clopidogrel and warfarin but still recommended continuation of these medications to avoid thrombotic events.¹⁶

Limitations

This study is a retrospective chart review and limited by what is already documented in the electronic health record. We can verify that the patients were given instructions to continue their medications up to the day of surgery but cannot be certain whether the instructions were followed. No control group was told to hold their anticoagulants for the same surgery. Once we decided on a protocol, we applied it to all patients. The study approval was for the specific time frame when the protocol was in place.

Our study was designed for elective hand cases because those surgeries can be anticipated, predicted, and patients can be given instructions during the preoperative appointments. We did incidentally find several nonelective hand cases (traumas, infections, and cancers) during the review of patients taking prescription blood thinners that had to be expedited to the operating room. Based on morbidity data during that time period, there were no additional postoperative hand surgery bleeding complications that had to return to the operating room. Future studies are indicated, but we believe our protocol can be applied to urgent and emergent hand surgeries as well as elective cases.

Conclusions

Our study supports continuing prescription anticoagulant and antiplatelet medications during the perioperative period for elective hand surgery. We found this is a safe practice in our veteran population with an acceptably low local bleeding complication rate.

Acknowledgments

This manuscript is the result of work supported with the resources and the use of facilities at the North Florida/South Georgia Veterans Health System in Gainesville, Florida.

Regularly scheduled educational conferences, such as case-based morning reports, have been a standard part of internal medicine residencies for decades.^1-4 In addition to better patient care from the knowledge gained at educational conferences, attendance by interns and residents (collectively called house staff) may be associated with higher in-service examination scores.⁵ Unfortunately, competing priorities, including patient care and trainee supervision, may contribute to an action-intention gap among house staff that reduces attendance.^6-8 Low attendance at morning reports represents wasted effort and lost educational opportunities; therefore, strategies to increase attendance are needed. Of several methods studied, more resource-intensive interventions (eg, providing food) were the most successful.^6,9-12

Using the behavioral economics framework of nudge strategies, we hypothesized that a less intensive intervention of a daily reminder text page would encourage medical students, interns, and residents (collectively called learners) to attend the morning report conference.^8,13 However, given the high cognitive load created by frequent task switching, a reminder text page could disrupt workflow and patient care without promoting the intended behavior change.^14-17 Because of this uncertainty, our objective was to determine whether a preconference text page increased learner attendance at morning report conferences.

Methods

This study was a single-center, multiple-crossover cluster randomized controlled trial conducted at the Veteran Affairs Boston Healthcare System (VABHS) in Massachusetts. Study participants included house staff rotating on daytime inpatient rotations from 4 residency programs and students from 2 medical schools. The setting was the morning report, an in-person, interactive, case-based conference held Monday through Thursday, from 8:00 am to 8:45 am. On Friday mornings, the morning report was replaced with a medical Jeopardy game-style conference. Historically, attendance has not been recorded for these conferences.

Learners assigned to rotate on the inpatient medicine, cardiology, medicine consultation, and patient safety rotations were eligible to attend these conferences and for inclusion in the study. Learners rotating in the medical intensive care unit, on night float, or on day float (an admitting shift for which residents are not on-site until late afternoon) were excluded. Additional details of the study population are available in the supplement (eAppendix). The study period was originally planned for September 30, 2019, to March 31, 2020, but data collection was stopped on March 12, 2020, due to the COVID-19 pandemic and suspension of in-person conferences. We chose the study period, which determined our sample size, to exclude the first 3 months of the academic year (July-September) because during that time learners acclimate to the inpatient workflow. We also chose not to include the last 3 months of the academic year to provide time for data analysis and preparation of the manuscript within the academic year.

Intervention and Outcome Assessment

Each intervention and control period was 3 weeks long; the first period was randomly determined by coin flip and alternated thereafter. Additional details of randomization are available in the supplement (Appendix 1). During intervention periods, all house staff received a page at 7:55 am that listed the time and location of the upcoming morning report or Jeopardy conference. Medical students do not carry pagers and did not receive reminder pages; however, we included these learners because changes in their conference attendance behavior would indicate an extension of the effect of reminder pages beyond the individual learner who received the page.

A daily facesheet (a roster of house staff names and photos) was used to identify learners for conference attendance. This facesheet was already used for other purposes at VABHS. At 8:00 am and 8:10 am, a chief medical resident who was not blinded to the intervention or control period recorded the attendance of each eligible learner as present or absent; learners were unaware that their attendance was being recorded. This approach to data collection was selected to minimize the likelihood that the behavior of the study participants would be influenced.

During control periods, no text page reminder of upcoming conferences was sent, but the attendance of total learners at 8:00 am and 8:10 am was recorded by a chief medical resident who used the same method as during the intervention periods. Attendance at 8:10 am was chosen as the primary outcome to account for the possibility that learners may arrive after a conference begins. Attendance at 8:00 am also was recorded to assess the effect of reminder pages on attendance at the start of morning reports.

Statistical Analysis

The primary outcome was the proportion of eligible learners present at 8:10 am at the morning report, expressed as the risk difference for attendance between intervention and control periods. Secondary outcomes included the proportion of learners present at 8:00 am (on-time attendance), the proportion of learners present by type (student vs house staff), and the proportion of learners present at the Friday Jeopardy conference. Two preplanned subgroup analyses were performed: one assessing the impact of rotating on clinical services with lighter workloads, and the other assessing the impact of the number of overnight admissions received on the relationship between receipt of a reminder page and conference attendance.

To estimate the primary outcome, we modeled the risk difference adjusted for covariates using a generalized estimating equation accounting for the clustering of attendance behavior within individuals and controlling for date and team. Secondary outcomes were estimated similarly. To evaluate the robustness of the primary outcome, we performed a sensitivity analysis using a multilevel generalized linear model with clustering by individual learner and team. Additional details on our statistical analysis plan, including accessing our raw data and analysis code, are available in Appendices 2 and 3. Categorical variables were compared using the χ² or Fisher exact test. Continuous variables were compared using the t test or Wilcoxon rank-sum tests. All P values were 2-sided, and a significance level of ≤ .05 was considered statistically significant. Analysis was performed in Stata v16.1. Our study was deemed exempt by the VABHS Institutional Review Board, and this article was prepared following the CONSORT reporting guidelines. The trial protocol has been registered with the International Standard Randomized Controlled Trial Number registry (ISRCTN14675095).

 

 

Results

Over the study period, 329 unique learners rotated on inpatient medical services at the VABHS and 211 were eligible to attend 85 morning report conferences and 22 Jeopardy conferences (Figure). Outcomes data were available for 100% of eligible participants. Forty-seven (55%) of the morning report conferences occurred during the intervention period (Table 1).

Morning report attendance observed at 8:10 am was 5.5% higher during the intervention period compared with the control period (49.9% vs 44.4%, P = .007). Accounting for clustering within individuals, the unadjusted risk difference in morning report attendance associated with sending a reminder page was 3.6% (95% CI, 0.09%-7.2%; P = .04) compared with no reminder page. When adding date and team to our model, the adjusted risk difference in conference attendance increased to 4.0% (95% CI, 0.5%-7.6%; P = .03) (Table 2). Results were similar in a sensitivity analysis using a multilevel generalized linear model accounting for clustering by both individual and team (adjusted risk difference, 4.0% [95% CI, 0.4%-7.6%; P = .03]).

On-time attendance was lower than at 8:10 am in both groups, with no difference in the observed attendance at 8:00 am between the control and intervention groups (22.4% vs 25.0%, P = .14). Regarding Jeopardy-like conferences, on-time attendance differed between the control and intervention groups at 8:00 am (15.3% vs 23.6%, P = .01), but not at 8:10 am (42.9% vs 42.8%, P > .99). We found no evidence of an interaction between receipt of a reminder page and learner type (student vs house staff, P = .33).

To estimate the impact of rotating on teams with lighter clinical workloads on the association between receipt of a reminder page and conference attendance, we repeated our primary analysis with a test of interaction between team assignment and the intervention, which was not significant (P = .90). To estimate the impact of morning workload on the association between receipt of a reminder page and conference attendance, we performed a subgroup analysis limited to learners rotating on teams eligible to receive overnight admissions and included the number of overnight admissions as a covariate in our regression model. A test of interaction between the intervention and the number of overnight admissions on conference attendance was not significant (P = .73).

In a subgroup analysis limited to learners on teams eligible to receive overnight admissions and controlling for the number of overnight admissions (a proxy for morning workload), no significant interaction between the intervention and admissions was observed. We also assessed for interaction between learner type and receipt of a reminder page on conference attendance and found no evidence of such an effect.

Discussion

Among a diverse population of learners from multiple academic institutions rotating at a single, large, urban VA medical center, a nudge strategy of sending a reminder text page before morning report conferences was associated with a 4.0% absolute increase in attendance measured 10 minutes after the conference started compared with not sending a reminder page. Overall, only one-quarter of learners attended the morning report at the start at 8:00 am, with no difference in on-time attendance between the intervention and control periods.

We designed our analysis to overcome several limitations of prior studies on the effect of reminder text pages on conference attendance. First, to account for differences in conference attendance behavior of individual learners, we used a generalized estimating equation model that allowed clustering of outcomes by individual. Second, we controlled for the date to account for secular trends in conference attendance over the academic year. Finally, we controlled for the team to account for the possibility that the conference attendance behavior of one learner on a team influences the behavior of other learners on the same team.

We also evaluated the effect of a reminder page on attendance at a weekly Jeopardy conference. Interestingly, reminder pages seemed to increase on-time Jeopardy attendance, although this effect was no longer statistically significant at 8:10 am. A possible explanation for this is that the fun and collegial nature of Jeopardy conferences entices learners to attend independent of a reminder page.

We also assessed the interaction between sending a reminder page and learner type and its effect on conference attendance and found no evidence to support such an effect. Because medical students do not receive reminder pages, their conference attendance behavior can be thought of as indicative of clustering within teams. Though there was no evidence of a significant interaction, given the small number of students, our study may be underpowered to find a benefit for this group.

The results of this study differ from Smith and colleagues, who found that reminder pages had no overall effect on conference attendance for fellows; however, no sample size justification was provided in that study, making it difficult to evaluate the likelihood of a false-negative finding.⁷ Our study differs in several ways: the timing of the reminder page (5 minutes vs 30 minutes prior to the conference), the method by which attendance was recorded (by an independent observer vs learner sign-in), and the time that attendance was recorded (2 prespecified times vs continuously). As far as we know, our study is the first to evaluate the nudge effect of reminder text pages on internal medicine resident attendance at conferences, with attendance taken by an observer.

 

 

Limitations

This study has some limitations. First, it was conducted at a single VA medical center. An additional limitation was our decision to classify learners who arrived after 8:10 am as absent, which likely underestimated total conference attendance. Further, we did not record whether learners stayed until the end of the conference. Additionally, many hospitals are transitioning away from pagers in favor of mobile phones; however, we have no reason to expect that the device on which a reminder is received (pager or phone) should affect the generalizability of these results.

Unfortunately, due to the COVID-19 pandemic and the suspension of in-person conferences, our study ended earlier than anticipated. This resulted in an imbalance of morning report conferences that occurred during each period: 55% during the intervention period, and 45% during the control period. However, because we accounted for the clustering of conference attendance behavior within individuals in our model, this imbalance is unlikely to introduce bias in our estimation of the effect of the intervention.

Another limitation relates to the evolving landscape of educational conferences in the postpandemic era.¹⁸ Whether our results can be generalized to increase virtual conference attendance is unknown. Finally, it is not clear whether a 4% absolute increase in conference attendance is educationally meaningful or justifies the effort of sending a reminder page.

Conclusions

In this cluster randomized controlled trial conducted at a single VA medical center, reminder pages sent 5 minutes before the start of morning report conferences resulted in a 4% increase in conference attendance. Our results suggest that reminder pages are one strategy that may result in a small increase in conference attendance, but whether this small increase is educationally significant will vary across training programs applying this strategy.

Acknowledgments

The authors are indebted to Kenneth J. Mukamal and Katharine A. Robb, who provided invaluable guidance in data analysis. Todd Reese assisted in data organization and presentation of data, and Mark Tuttle designed the facesheet. None of these individuals received compensation for their assistance.

Regularly scheduled educational conferences, such as case-based morning reports, have been a standard part of internal medicine residencies for decades.^1-4 In addition to better patient care from the knowledge gained at educational conferences, attendance by interns and residents (collectively called house staff) may be associated with higher in-service examination scores.⁵ Unfortunately, competing priorities, including patient care and trainee supervision, may contribute to an action-intention gap among house staff that reduces attendance.^6-8 Low attendance at morning reports represents wasted effort and lost educational opportunities; therefore, strategies to increase attendance are needed. Of several methods studied, more resource-intensive interventions (eg, providing food) were the most successful.^6,9-12

Using the behavioral economics framework of nudge strategies, we hypothesized that a less intensive intervention of a daily reminder text page would encourage medical students, interns, and residents (collectively called learners) to attend the morning report conference.^8,13 However, given the high cognitive load created by frequent task switching, a reminder text page could disrupt workflow and patient care without promoting the intended behavior change.^14-17 Because of this uncertainty, our objective was to determine whether a preconference text page increased learner attendance at morning report conferences.

Methods

This study was a single-center, multiple-crossover cluster randomized controlled trial conducted at the Veteran Affairs Boston Healthcare System (VABHS) in Massachusetts. Study participants included house staff rotating on daytime inpatient rotations from 4 residency programs and students from 2 medical schools. The setting was the morning report, an in-person, interactive, case-based conference held Monday through Thursday, from 8:00 am to 8:45 am. On Friday mornings, the morning report was replaced with a medical Jeopardy game-style conference. Historically, attendance has not been recorded for these conferences.

Learners assigned to rotate on the inpatient medicine, cardiology, medicine consultation, and patient safety rotations were eligible to attend these conferences and for inclusion in the study. Learners rotating in the medical intensive care unit, on night float, or on day float (an admitting shift for which residents are not on-site until late afternoon) were excluded. Additional details of the study population are available in the supplement (eAppendix). The study period was originally planned for September 30, 2019, to March 31, 2020, but data collection was stopped on March 12, 2020, due to the COVID-19 pandemic and suspension of in-person conferences. We chose the study period, which determined our sample size, to exclude the first 3 months of the academic year (July-September) because during that time learners acclimate to the inpatient workflow. We also chose not to include the last 3 months of the academic year to provide time for data analysis and preparation of the manuscript within the academic year.

Intervention and Outcome Assessment

Each intervention and control period was 3 weeks long; the first period was randomly determined by coin flip and alternated thereafter. Additional details of randomization are available in the supplement (Appendix 1). During intervention periods, all house staff received a page at 7:55 am that listed the time and location of the upcoming morning report or Jeopardy conference. Medical students do not carry pagers and did not receive reminder pages; however, we included these learners because changes in their conference attendance behavior would indicate an extension of the effect of reminder pages beyond the individual learner who received the page.

A daily facesheet (a roster of house staff names and photos) was used to identify learners for conference attendance. This facesheet was already used for other purposes at VABHS. At 8:00 am and 8:10 am, a chief medical resident who was not blinded to the intervention or control period recorded the attendance of each eligible learner as present or absent; learners were unaware that their attendance was being recorded. This approach to data collection was selected to minimize the likelihood that the behavior of the study participants would be influenced.

During control periods, no text page reminder of upcoming conferences was sent, but the attendance of total learners at 8:00 am and 8:10 am was recorded by a chief medical resident who used the same method as during the intervention periods. Attendance at 8:10 am was chosen as the primary outcome to account for the possibility that learners may arrive after a conference begins. Attendance at 8:00 am also was recorded to assess the effect of reminder pages on attendance at the start of morning reports.

Statistical Analysis

The primary outcome was the proportion of eligible learners present at 8:10 am at the morning report, expressed as the risk difference for attendance between intervention and control periods. Secondary outcomes included the proportion of learners present at 8:00 am (on-time attendance), the proportion of learners present by type (student vs house staff), and the proportion of learners present at the Friday Jeopardy conference. Two preplanned subgroup analyses were performed: one assessing the impact of rotating on clinical services with lighter workloads, and the other assessing the impact of the number of overnight admissions received on the relationship between receipt of a reminder page and conference attendance.

To estimate the primary outcome, we modeled the risk difference adjusted for covariates using a generalized estimating equation accounting for the clustering of attendance behavior within individuals and controlling for date and team. Secondary outcomes were estimated similarly. To evaluate the robustness of the primary outcome, we performed a sensitivity analysis using a multilevel generalized linear model with clustering by individual learner and team. Additional details on our statistical analysis plan, including accessing our raw data and analysis code, are available in Appendices 2 and 3. Categorical variables were compared using the χ² or Fisher exact test. Continuous variables were compared using the t test or Wilcoxon rank-sum tests. All P values were 2-sided, and a significance level of ≤ .05 was considered statistically significant. Analysis was performed in Stata v16.1. Our study was deemed exempt by the VABHS Institutional Review Board, and this article was prepared following the CONSORT reporting guidelines. The trial protocol has been registered with the International Standard Randomized Controlled Trial Number registry (ISRCTN14675095).

 

 

Results

Over the study period, 329 unique learners rotated on inpatient medical services at the VABHS and 211 were eligible to attend 85 morning report conferences and 22 Jeopardy conferences (Figure). Outcomes data were available for 100% of eligible participants. Forty-seven (55%) of the morning report conferences occurred during the intervention period (Table 1).

Morning report attendance observed at 8:10 am was 5.5% higher during the intervention period compared with the control period (49.9% vs 44.4%, P = .007). Accounting for clustering within individuals, the unadjusted risk difference in morning report attendance associated with sending a reminder page was 3.6% (95% CI, 0.09%-7.2%; P = .04) compared with no reminder page. When adding date and team to our model, the adjusted risk difference in conference attendance increased to 4.0% (95% CI, 0.5%-7.6%; P = .03) (Table 2). Results were similar in a sensitivity analysis using a multilevel generalized linear model accounting for clustering by both individual and team (adjusted risk difference, 4.0% [95% CI, 0.4%-7.6%; P = .03]).

On-time attendance was lower than at 8:10 am in both groups, with no difference in the observed attendance at 8:00 am between the control and intervention groups (22.4% vs 25.0%, P = .14). Regarding Jeopardy-like conferences, on-time attendance differed between the control and intervention groups at 8:00 am (15.3% vs 23.6%, P = .01), but not at 8:10 am (42.9% vs 42.8%, P > .99). We found no evidence of an interaction between receipt of a reminder page and learner type (student vs house staff, P = .33).

To estimate the impact of rotating on teams with lighter clinical workloads on the association between receipt of a reminder page and conference attendance, we repeated our primary analysis with a test of interaction between team assignment and the intervention, which was not significant (P = .90). To estimate the impact of morning workload on the association between receipt of a reminder page and conference attendance, we performed a subgroup analysis limited to learners rotating on teams eligible to receive overnight admissions and included the number of overnight admissions as a covariate in our regression model. A test of interaction between the intervention and the number of overnight admissions on conference attendance was not significant (P = .73).

In a subgroup analysis limited to learners on teams eligible to receive overnight admissions and controlling for the number of overnight admissions (a proxy for morning workload), no significant interaction between the intervention and admissions was observed. We also assessed for interaction between learner type and receipt of a reminder page on conference attendance and found no evidence of such an effect.

Discussion

Among a diverse population of learners from multiple academic institutions rotating at a single, large, urban VA medical center, a nudge strategy of sending a reminder text page before morning report conferences was associated with a 4.0% absolute increase in attendance measured 10 minutes after the conference started compared with not sending a reminder page. Overall, only one-quarter of learners attended the morning report at the start at 8:00 am, with no difference in on-time attendance between the intervention and control periods.

We designed our analysis to overcome several limitations of prior studies on the effect of reminder text pages on conference attendance. First, to account for differences in conference attendance behavior of individual learners, we used a generalized estimating equation model that allowed clustering of outcomes by individual. Second, we controlled for the date to account for secular trends in conference attendance over the academic year. Finally, we controlled for the team to account for the possibility that the conference attendance behavior of one learner on a team influences the behavior of other learners on the same team.

We also evaluated the effect of a reminder page on attendance at a weekly Jeopardy conference. Interestingly, reminder pages seemed to increase on-time Jeopardy attendance, although this effect was no longer statistically significant at 8:10 am. A possible explanation for this is that the fun and collegial nature of Jeopardy conferences entices learners to attend independent of a reminder page.

We also assessed the interaction between sending a reminder page and learner type and its effect on conference attendance and found no evidence to support such an effect. Because medical students do not receive reminder pages, their conference attendance behavior can be thought of as indicative of clustering within teams. Though there was no evidence of a significant interaction, given the small number of students, our study may be underpowered to find a benefit for this group.

The results of this study differ from Smith and colleagues, who found that reminder pages had no overall effect on conference attendance for fellows; however, no sample size justification was provided in that study, making it difficult to evaluate the likelihood of a false-negative finding.⁷ Our study differs in several ways: the timing of the reminder page (5 minutes vs 30 minutes prior to the conference), the method by which attendance was recorded (by an independent observer vs learner sign-in), and the time that attendance was recorded (2 prespecified times vs continuously). As far as we know, our study is the first to evaluate the nudge effect of reminder text pages on internal medicine resident attendance at conferences, with attendance taken by an observer.

 

 

Limitations

This study has some limitations. First, it was conducted at a single VA medical center. An additional limitation was our decision to classify learners who arrived after 8:10 am as absent, which likely underestimated total conference attendance. Further, we did not record whether learners stayed until the end of the conference. Additionally, many hospitals are transitioning away from pagers in favor of mobile phones; however, we have no reason to expect that the device on which a reminder is received (pager or phone) should affect the generalizability of these results.

Unfortunately, due to the COVID-19 pandemic and the suspension of in-person conferences, our study ended earlier than anticipated. This resulted in an imbalance of morning report conferences that occurred during each period: 55% during the intervention period, and 45% during the control period. However, because we accounted for the clustering of conference attendance behavior within individuals in our model, this imbalance is unlikely to introduce bias in our estimation of the effect of the intervention.

Another limitation relates to the evolving landscape of educational conferences in the postpandemic era.¹⁸ Whether our results can be generalized to increase virtual conference attendance is unknown. Finally, it is not clear whether a 4% absolute increase in conference attendance is educationally meaningful or justifies the effort of sending a reminder page.

Conclusions

In this cluster randomized controlled trial conducted at a single VA medical center, reminder pages sent 5 minutes before the start of morning report conferences resulted in a 4% increase in conference attendance. Our results suggest that reminder pages are one strategy that may result in a small increase in conference attendance, but whether this small increase is educationally significant will vary across training programs applying this strategy.

Acknowledgments

The authors are indebted to Kenneth J. Mukamal and Katharine A. Robb, who provided invaluable guidance in data analysis. Todd Reese assisted in data organization and presentation of data, and Mark Tuttle designed the facesheet. None of these individuals received compensation for their assistance.

Regularly scheduled educational conferences, such as case-based morning reports, have been a standard part of internal medicine residencies for decades.^1-4 In addition to better patient care from the knowledge gained at educational conferences, attendance by interns and residents (collectively called house staff) may be associated with higher in-service examination scores.⁵ Unfortunately, competing priorities, including patient care and trainee supervision, may contribute to an action-intention gap among house staff that reduces attendance.^6-8 Low attendance at morning reports represents wasted effort and lost educational opportunities; therefore, strategies to increase attendance are needed. Of several methods studied, more resource-intensive interventions (eg, providing food) were the most successful.^6,9-12

Using the behavioral economics framework of nudge strategies, we hypothesized that a less intensive intervention of a daily reminder text page would encourage medical students, interns, and residents (collectively called learners) to attend the morning report conference.^8,13 However, given the high cognitive load created by frequent task switching, a reminder text page could disrupt workflow and patient care without promoting the intended behavior change.^14-17 Because of this uncertainty, our objective was to determine whether a preconference text page increased learner attendance at morning report conferences.

Methods

This study was a single-center, multiple-crossover cluster randomized controlled trial conducted at the Veteran Affairs Boston Healthcare System (VABHS) in Massachusetts. Study participants included house staff rotating on daytime inpatient rotations from 4 residency programs and students from 2 medical schools. The setting was the morning report, an in-person, interactive, case-based conference held Monday through Thursday, from 8:00 am to 8:45 am. On Friday mornings, the morning report was replaced with a medical Jeopardy game-style conference. Historically, attendance has not been recorded for these conferences.

Learners assigned to rotate on the inpatient medicine, cardiology, medicine consultation, and patient safety rotations were eligible to attend these conferences and for inclusion in the study. Learners rotating in the medical intensive care unit, on night float, or on day float (an admitting shift for which residents are not on-site until late afternoon) were excluded. Additional details of the study population are available in the supplement (eAppendix). The study period was originally planned for September 30, 2019, to March 31, 2020, but data collection was stopped on March 12, 2020, due to the COVID-19 pandemic and suspension of in-person conferences. We chose the study period, which determined our sample size, to exclude the first 3 months of the academic year (July-September) because during that time learners acclimate to the inpatient workflow. We also chose not to include the last 3 months of the academic year to provide time for data analysis and preparation of the manuscript within the academic year.

Intervention and Outcome Assessment

Each intervention and control period was 3 weeks long; the first period was randomly determined by coin flip and alternated thereafter. Additional details of randomization are available in the supplement (Appendix 1). During intervention periods, all house staff received a page at 7:55 am that listed the time and location of the upcoming morning report or Jeopardy conference. Medical students do not carry pagers and did not receive reminder pages; however, we included these learners because changes in their conference attendance behavior would indicate an extension of the effect of reminder pages beyond the individual learner who received the page.

A daily facesheet (a roster of house staff names and photos) was used to identify learners for conference attendance. This facesheet was already used for other purposes at VABHS. At 8:00 am and 8:10 am, a chief medical resident who was not blinded to the intervention or control period recorded the attendance of each eligible learner as present or absent; learners were unaware that their attendance was being recorded. This approach to data collection was selected to minimize the likelihood that the behavior of the study participants would be influenced.

During control periods, no text page reminder of upcoming conferences was sent, but the attendance of total learners at 8:00 am and 8:10 am was recorded by a chief medical resident who used the same method as during the intervention periods. Attendance at 8:10 am was chosen as the primary outcome to account for the possibility that learners may arrive after a conference begins. Attendance at 8:00 am also was recorded to assess the effect of reminder pages on attendance at the start of morning reports.

Statistical Analysis

The primary outcome was the proportion of eligible learners present at 8:10 am at the morning report, expressed as the risk difference for attendance between intervention and control periods. Secondary outcomes included the proportion of learners present at 8:00 am (on-time attendance), the proportion of learners present by type (student vs house staff), and the proportion of learners present at the Friday Jeopardy conference. Two preplanned subgroup analyses were performed: one assessing the impact of rotating on clinical services with lighter workloads, and the other assessing the impact of the number of overnight admissions received on the relationship between receipt of a reminder page and conference attendance.

To estimate the primary outcome, we modeled the risk difference adjusted for covariates using a generalized estimating equation accounting for the clustering of attendance behavior within individuals and controlling for date and team. Secondary outcomes were estimated similarly. To evaluate the robustness of the primary outcome, we performed a sensitivity analysis using a multilevel generalized linear model with clustering by individual learner and team. Additional details on our statistical analysis plan, including accessing our raw data and analysis code, are available in Appendices 2 and 3. Categorical variables were compared using the χ² or Fisher exact test. Continuous variables were compared using the t test or Wilcoxon rank-sum tests. All P values were 2-sided, and a significance level of ≤ .05 was considered statistically significant. Analysis was performed in Stata v16.1. Our study was deemed exempt by the VABHS Institutional Review Board, and this article was prepared following the CONSORT reporting guidelines. The trial protocol has been registered with the International Standard Randomized Controlled Trial Number registry (ISRCTN14675095).

 

 

Results

Over the study period, 329 unique learners rotated on inpatient medical services at the VABHS and 211 were eligible to attend 85 morning report conferences and 22 Jeopardy conferences (Figure). Outcomes data were available for 100% of eligible participants. Forty-seven (55%) of the morning report conferences occurred during the intervention period (Table 1).

Morning report attendance observed at 8:10 am was 5.5% higher during the intervention period compared with the control period (49.9% vs 44.4%, P = .007). Accounting for clustering within individuals, the unadjusted risk difference in morning report attendance associated with sending a reminder page was 3.6% (95% CI, 0.09%-7.2%; P = .04) compared with no reminder page. When adding date and team to our model, the adjusted risk difference in conference attendance increased to 4.0% (95% CI, 0.5%-7.6%; P = .03) (Table 2). Results were similar in a sensitivity analysis using a multilevel generalized linear model accounting for clustering by both individual and team (adjusted risk difference, 4.0% [95% CI, 0.4%-7.6%; P = .03]).

On-time attendance was lower than at 8:10 am in both groups, with no difference in the observed attendance at 8:00 am between the control and intervention groups (22.4% vs 25.0%, P = .14). Regarding Jeopardy-like conferences, on-time attendance differed between the control and intervention groups at 8:00 am (15.3% vs 23.6%, P = .01), but not at 8:10 am (42.9% vs 42.8%, P > .99). We found no evidence of an interaction between receipt of a reminder page and learner type (student vs house staff, P = .33).

To estimate the impact of rotating on teams with lighter clinical workloads on the association between receipt of a reminder page and conference attendance, we repeated our primary analysis with a test of interaction between team assignment and the intervention, which was not significant (P = .90). To estimate the impact of morning workload on the association between receipt of a reminder page and conference attendance, we performed a subgroup analysis limited to learners rotating on teams eligible to receive overnight admissions and included the number of overnight admissions as a covariate in our regression model. A test of interaction between the intervention and the number of overnight admissions on conference attendance was not significant (P = .73).

In a subgroup analysis limited to learners on teams eligible to receive overnight admissions and controlling for the number of overnight admissions (a proxy for morning workload), no significant interaction between the intervention and admissions was observed. We also assessed for interaction between learner type and receipt of a reminder page on conference attendance and found no evidence of such an effect.

Discussion

Among a diverse population of learners from multiple academic institutions rotating at a single, large, urban VA medical center, a nudge strategy of sending a reminder text page before morning report conferences was associated with a 4.0% absolute increase in attendance measured 10 minutes after the conference started compared with not sending a reminder page. Overall, only one-quarter of learners attended the morning report at the start at 8:00 am, with no difference in on-time attendance between the intervention and control periods.

We designed our analysis to overcome several limitations of prior studies on the effect of reminder text pages on conference attendance. First, to account for differences in conference attendance behavior of individual learners, we used a generalized estimating equation model that allowed clustering of outcomes by individual. Second, we controlled for the date to account for secular trends in conference attendance over the academic year. Finally, we controlled for the team to account for the possibility that the conference attendance behavior of one learner on a team influences the behavior of other learners on the same team.

We also evaluated the effect of a reminder page on attendance at a weekly Jeopardy conference. Interestingly, reminder pages seemed to increase on-time Jeopardy attendance, although this effect was no longer statistically significant at 8:10 am. A possible explanation for this is that the fun and collegial nature of Jeopardy conferences entices learners to attend independent of a reminder page.

We also assessed the interaction between sending a reminder page and learner type and its effect on conference attendance and found no evidence to support such an effect. Because medical students do not receive reminder pages, their conference attendance behavior can be thought of as indicative of clustering within teams. Though there was no evidence of a significant interaction, given the small number of students, our study may be underpowered to find a benefit for this group.

The results of this study differ from Smith and colleagues, who found that reminder pages had no overall effect on conference attendance for fellows; however, no sample size justification was provided in that study, making it difficult to evaluate the likelihood of a false-negative finding.⁷ Our study differs in several ways: the timing of the reminder page (5 minutes vs 30 minutes prior to the conference), the method by which attendance was recorded (by an independent observer vs learner sign-in), and the time that attendance was recorded (2 prespecified times vs continuously). As far as we know, our study is the first to evaluate the nudge effect of reminder text pages on internal medicine resident attendance at conferences, with attendance taken by an observer.

 

 

Limitations

This study has some limitations. First, it was conducted at a single VA medical center. An additional limitation was our decision to classify learners who arrived after 8:10 am as absent, which likely underestimated total conference attendance. Further, we did not record whether learners stayed until the end of the conference. Additionally, many hospitals are transitioning away from pagers in favor of mobile phones; however, we have no reason to expect that the device on which a reminder is received (pager or phone) should affect the generalizability of these results.

Unfortunately, due to the COVID-19 pandemic and the suspension of in-person conferences, our study ended earlier than anticipated. This resulted in an imbalance of morning report conferences that occurred during each period: 55% during the intervention period, and 45% during the control period. However, because we accounted for the clustering of conference attendance behavior within individuals in our model, this imbalance is unlikely to introduce bias in our estimation of the effect of the intervention.

Another limitation relates to the evolving landscape of educational conferences in the postpandemic era.¹⁸ Whether our results can be generalized to increase virtual conference attendance is unknown. Finally, it is not clear whether a 4% absolute increase in conference attendance is educationally meaningful or justifies the effort of sending a reminder page.

Conclusions

In this cluster randomized controlled trial conducted at a single VA medical center, reminder pages sent 5 minutes before the start of morning report conferences resulted in a 4% increase in conference attendance. Our results suggest that reminder pages are one strategy that may result in a small increase in conference attendance, but whether this small increase is educationally significant will vary across training programs applying this strategy.

Acknowledgments

The authors are indebted to Kenneth J. Mukamal and Katharine A. Robb, who provided invaluable guidance in data analysis. Todd Reese assisted in data organization and presentation of data, and Mark Tuttle designed the facesheet. None of these individuals received compensation for their assistance.