User login
Hospitalists Effects on Outcomes, Costs, Point-of-Care HIV Testing, and More
Community Teaching
Halasyamani L, Valenstein P, Friedlander M. et al. A comparison of two hospitalist models with traditional care in a community teaching hospital. Am J Med. 2005;118:536-543.
Background: A growing body of literature has demonstrated the effects of hospitalists on reducing inpatient length of stay and cost of care, with some literature showing a decreased in-hospital and 30-day mortality. However, most prior studies were conducted in academic medical centers or health maintenance organizations where one group of hospitalists, employed by the institution within which they worked, was compared with traditional primary care physicians. Direct comparisons between different hospitalist models practicing within a single institution have not been published. As a result, the impact of different hospitalist characteristics, including employment status and reimbursement incentives, on inpatient resource utilization and patient care outcomes is unknown.
Methods: Halasyamani and colleagues conducted a retrospective cohort study of 10,595 patients in a tertiary care community-based teaching hospital in which private hospitalists, academic hospitalists, and community physicians all practice. They measured risk-adjusted length of stay, variable costs, 30-day readmission rates, and in-hospital and 30-day mortality for patients treated by each of these three groups, controlling for potentially confounding variables. Community physicians belonged to 21 rounding groups, most of which were private or solo. Two of the community physicians groups were hospital-owned practices reimbursed by a relative value unit system. The private hospitalist group was self-employed with no financial relationship to the hospital and worked an average of 40 weeks per year. Community physicians and private hospitalists worked Monday-Friday and covered weekends or holidays about 25% of the time. Academic hospitalists worked with internal medicine residents and students on a teaching service. They were employed by the hospital using a relative value unit system. They worked an average of 14 weeks per year as an inpatient attending in half-month rotations, which included weekend coverage.
Results: There was a 20% reduction (-0.72 days absolute difference) in length of stay on the academic hospitalist service (P<0.0001) and 8% (-0.28 days absolute difference) on the private hospitalist service (P=0.049) compared with community physicians. Case-mix adjusted relative total costs were 10% less ($173 absolute difference) on the academic (P<0.0001) and 6% less ($109 absolute difference) on the private hospitalist services (P=0.02) compared with community physicians. There were no differences in 30-day readmission, in-hospital and 30-day mortality between the three groups.
Discussion: This study is the first to look at the effects of two separate hospitalist models on resource utilization and patient outcomes within the same institution. Although both the academic and private hospitalist groups demonstrated improved resource utilization as compared with the community physicians, the magnitude of benefit was much greater for the academic hospitalist group.
As the authors point out, one major difference between the two groups was employment status, with the academic hospitalists employed directly by the hospital and the private hospitalists receiving all payment directly from payers. Previous studies have also focused on hospitalists, which were employed by the institution at which they worked, raising the question of whether alignment of employee and employer incentives is an important factor affecting resource utilization outcomes.
Results of this study highlight the need for more studies which seek to clarify specific physician-level, group-level, and organization-level characteristics of hospitalists that result in improved resource utilization and patient care outcomes.
The Last Few Hours
Bailey FA, Burgio KL, Woodby LL, et al. Improving the processes of hospital care during the last hours of life. Arch Int Med. 2005;165(15):1722-1727.
Background: End-of-life care in the acute care inpatient setting is often not initiated until very late in the dying process and may be related to inadequate early recognition of dying patients as well as difficulty transitioning from disease-modifying treatments to palliative measures. Additional barriers exist, including lack of familiarity of hospital staff with initiation and implementation of hospice care. Education about end-of-life care and introduction of a physician-led palliative care team available for consultation within acute care hospitals may help promote better recognition of the dying patient by staff and allow for a “good death.”
Methods: A single hospital within the Veterans Affairs (VA) medical system (Birmingham, Ala., VA Medical Center) was chosen as a pilot center for initiation of a physician-led Inpatient Comfort Care Program (ICCP). The study was framed as a “before-after intervention trial” and analyzed all inpatient deaths identified by the Computerized Patient Recognition System during a six-month period before and substantially after the introduction of the ICCP. A structured chart abstraction tool was used and data was obtained from the last seven days of hospitalization analyzing variables associated with recognition of the dying patient and initiation of palliative care. Education of hospital staff on both hospice care and case identification was initiated during the intervention phase of the study. Additionally, a flexible comfort care order set was introduced.
Results: Two hundred and three veterans were identified (98% men, average age 68) and no significant differences in clinical characteristics were noted between the two groups, pre-intervention and post-intervention. Post-intervention, 59.3% of patients had formal palliative care consultation. Significant findings (P<0.01) following implementation of ICCP were increased documentation of end-of-life symptoms, increased documentation of care plans, increased utilization of opioids (57.1% to 87.2%), increased initiation of do-not-resuscitate orders (61.9% to 85.1%) with a concurrent decrease in cardiopulmonary resuscitation at death (34.4% to 15.4%), and a surprising increase in restraint use (6.0% to 22.6%).
Discussion: Data on hospice care patients indicate that 10% to 30% die in an acute care hospital, identifying a need for increased education and training in palliative medicine. This study demonstrates the positive outcomes of implementation of an inpatient palliative care service both for heightened awareness of identifying the dying patient as well as initiation of end-of-life care. The increased use of opioid medications is an important marker given that many patients experience pain and dyspnea at the end of life. This study is limited by its single site and further validation at other centers implementing similar protocols and assessing similar outcomes is needed. While this intervention had important clinical benefits, additional studies examining the cost implications of this system would be helpful.
Education alone has not been shown to be entirely effective in creating change. This single-site implementation of a palliative care consultation service successfully integrated an education program with on-site consultants. Distributing pocket cards with clinical findings identifying the dying patient aided in recognition of those patients and pre-printed order sets facilitated initiation of end-of-life care. The intervention initiated is possible for many medical centers and promotes an environment allowing for a “good death” for dying patients.
Computers, Doctors, and Errors
Koppel R, Metlay JP, Cohen A, et al. Role of computerized physician order entry systems in facilitating medication errors. JAMA. 2005;293(10):1197-1203
For physicians, computerized physician order entry (CPOE) has become an important topic of discussion as many hospitals and health systems embark on the complex and lengthy process of implementing new enterprise clinical systems. Though there are undoubtedly benefits to such systems, practicing clinicians are apt to remain skeptical of the grandiose pictures the more vocal advocates of CPOE may paint. This is not to say that the promises of CPOE are empty; to the contrary, there have been substantial successes, notably in the realm of medication error prevention.
At the same time, CPOE is a mixture of complex technologies that interface in complicated ways with the culture of clinical medicine. The view that medical informatics is a technical problem that has been solved long ago is simplistic and naïve. The article by Koppel, et al, has two important implications: 1) It is critical to look at clinical information systems in the social milieu in which it functions, and 2) there are often unintended consequences that may not beneficial.
This article examines a widely used, commercial CPOE system in use at the University of Pennsylvania (Philadelphia) using both quantitative and qualitative methods. The researchers conducted focus groups and expert interviews in addition to field observations of physicians (house officers and attendings), nurses, and pharmacists in order to identify themes relating to work with the order entry system. This work helped to guide the creation of a survey instrument subsequently used to survey house staff about working conditions and sources of error and stress. There was an 85%-90% response rate that primarily included house staff who ordered more than nine medication orders per month.
Researchers found two broad categories of errors that were fostered in this environment. The first category, which they termed “information errors” were generated by fragmentation of data and the failure to integrate the hospitals various systems both electronic and paper. Examples of this type of error include antibiotic renewal failures. A common way this failure would occur is that renewal reminder stickers would be placed in the patients’ charts, but the house staff would overlook these because medication orders occurred electronically. Another example is assumed dose errors, where house staff would assume that the default dose displayed was the recommended starting dose, when in fact this was the smallest dose unit available. Physicians were assuming decision support was available when it was not.
The second type of error, human-machine interface flaws, occurred when machine rules did not correspond to work behaviors. An example of this is when patients were listed alphabetically rather than by service, making it easy to select the wrong patient. In another instance, many screens (up to 20) were required to view all of a patient’s medications, making it difficult to choose a correct medication for editing.
This study has been criticized by industry advocates for focusing on an older set of technologies or because a number of these issues related to training or “user factors.” At the other extreme, this study has been cited as a cautionary tale about the risks of CPOE. Both types of criticism miss the point. This study demonstrates that CPOE and the social environment in which it sits is a complex entity and that careful design, proper support, and maintenance are critical ingredients to the success of an incredibly complex but vital new component of hospital medicine.
Point-of-Care HIV Testing in Inpatients
Lubelchek R, Kroc K, Hota B, et al. The role of rapid vs conventional human immunodeficiency virus testing for inpatients: effects on quality of care. Arch Intern Med. 2005;165:1956-1960.
Despite advances in treatment, infection with HIV and AIDS remains a public health problem in the United States. According to the CDC the rate of new diagnosis of HIV infection has remained steady from 2000 to 2003 at about 20 per 100,000 people. (Centers for Disease Control and Prevention. Diagnosis of HIV/AIDS–32 states, 2009-2003. MMWR Morb Mortal Wkly Rep. 2004;53:1106-1110). Currently, about 850,000 to 950,000 people are believed to be living with HIV infection, and it is estimated that 180,000 to 280,000 are unaware of their diagnosis. (Fleming P, Byers RH, Sweeney PA, et al., HIV prevalence in the United States, 2000 [Abstract 11]. Presented at the Ninth Conference on Retroviruses and Opportunistic Infections, Seattle; February 24–28, 2002). These patients are not only at risk for disease progression, but can undermine efforts at disease prevention if they continue to engage in unsafe activities. Thus, increasing awareness of HIV status is an important aspect of disease prevention.
HIV testing remains a challenge. Conventional testing with enzyme immunoassay (EIA) and confirmatory Western blot requires patient follow-up for results, which approximately 25% of patients in various outpatient testing sites fail to do. (Centers for Disease Control and Prevention. Update: HIV counseling and testing using rapid tests, United States, 1995. MMWR Morb Mortal Wkly Rep. 1998;47:211-215). Given the difficulties inherent in the transition of care from the inpatient to outpatient setting, conventional testing in the inpatient setting presents additional barriers to appropriate notification. As various point-of-care HIV tests have been developed for commercial use, the possibility of rapid HIV testing presents an opportunity to reduce notification failure and improve patient care. While not replacing traditional testing, the CDC has endorsed rapid HIV testing as a means to initiate therapy and provide counseling with a particular focus on preventing further disease transmission. In this retrospective study, Lubelchek and colleagues present the effects of a rapid HIV test utilized in the emergency department on various inpatient quality of care measures for those patients who received a positive rapid HIV test later confirmed by Western blot as compared with those patients who were diagnosed after admission by traditional diagnostic methods. This study took place in the context of CDC-funded study of the use of OraQuick (OraSure Technologies, Bethlehem, Pa.) rapid HIV testing in the emergency department at Cook County Hospital in Chicago.
The manufacturer claims the product has a sensitivity of 99.6% and a specificity of 100% as compared with conventional testing. (OraQuick rapid HIV-1 antibody test summary of safety and effectiveness. November 7, 2002. Accessed October 1, 2005, at www.fda.gov/cber/pma/P010047.htm). In the initial study, two of the three emergency department’s treatment pods were equipped to provide HIV screening utilizing the point-of-care technology to consenting patients. Patients in the third pod could be referred to rapid testing based on symptoms or risk factors. All patients who received the rapid test also submitted specimens for conventional EIA and confirmatory Western blot testing. All positive rapid HIV tests were confirmed by Western blot.
In this study, patients who were not known to be infected and were subsequently admitted on non-obstetric or surgical services over 17 months from 2003 to 2004 and confirmed to be HIV positive by Western blot were identified utilizing administrative records. Where possible, charts were reviewed to confirm no prior diagnosis of HIV. Patients who received rapid HIV testing were compared with those who only received conventional testing. Endpoints included time to primary inpatient care service awareness of HIV diagnosis, time to admission or transfer to the inpatient HIV service, time to empiric treatment of diagnosis of opportunistic infection, length of stay, discharge with appropriate prophylactic medications, discharge with patient knowledge of HIV diagnosis, and initial engagement in outpatient care. Length of stay was adjusted by multivariate regression on co-morbid diagnoses (congestive heart failure, end-stage renal disease, cirrhosis, chronic obstructive pulmonary disease, and diabetes), opportunistic infections, ICU admission, need for mechanical ventilation, and CD4 count.
A total of 103 patients were identified with complete chart review completed on 86 of them. All patients except one were admitted through the emergency department. Forty-eight patients were diagnosed initially with the rapid HIV test with 58% of these specifically referred for testing by the emergency department physician, and 55 were diagnosed with conventional testing. Overall, 78% were male, 62% African American, and 20% Hispanic. The two groups were comparable in terms of age, sex, ethnicity, history of substance abuse, HIV risk factors, psychiatric diagnoses, homelessness, CD4 count, presence of opportunistic infections, mechanical ventilation, and co-morbidities. However, conventionally tested patients were more likely to require an ICU stay (31% vs. 10%, P=.01).
Patients in the rapid test group were more quickly documented in the chart as having HIV (.8 vs. 6.4 days, p<.001), placed on an HIV service sooner (1.4 versus 6.9 days, P<.001), initiated outpatient follow-up sooner (21.5 versus 49.5 days, p=.05), and had less unawareness of their HIV status (0 vs. 16%, P=.002). There was no significant difference between the two groups in time from admission to empiric treatment or diagnosis of an opportunistic infection. Patients who received the rapid test did have a lower length of stay (6.4 versus 13.2 days, P<.001). Although much of this difference was due to higher number of ICU stays in the conventional group, in multivariate analysis conventional testing still increased length of stay significantly, OR 5.4 days (2.5, 8.3).
This study suggests that patients who are tested with rapid HIV testing can lead to more efficient inpatient treatment of the complications of HIV, improved patient awareness of HIV status, and quicker outpatient follow-up. These findings have ramifications not just to the inpatient management of patients with HIV but to general public health efforts to reduce the spread of HIV infection.
Nevertheless, these results must be interpreted with caution. They reflect the experience of one institution situated in an area with a high prevalence of HIV. Some degree of selection bias is suggested by the higher presence of ICU admissions in the conventional testing group. The multivariate analysis attempted to control for confounding factors, but the possibility remains that other unrecognized factors may have influenced results. The authors do note that an analysis of patients in the rapid test group stratified by whether the test was performed for screening or by referral of the physician did not demonstrate a statistically significant difference in length of stay. This finding provides further support that the sicker patients which triggered the rapid test had shorter lengths of stay on account of the rapid test and not simply because they were sicker.
As recognized by the authors, physicians in routine practice rely on surrogate markers of HIV infection, most notably a patient’s CD4 count, and thus it is not surprising that the rapid test did not affect time to empiric treatment or diagnosis of opportunistic infection. If treatment did not differ, then explaining the longer length of stay remains an unexplained puzzle. The fact that the two groups were equally matched socially and psychiatrically leaves open the possibility that it was actual knowledge of the HIV test result—and not its effect on treatment—that drove the longer length of stay.
One possibility not suggested by the authors is that definitive knowledge of HIV status helped to mobilize patient discharge. If there were legitimate concerns of follow-up, physicians may have delayed discharge in order to receive HIV test results. Alternatively, some patients may have resisted discharge until receiving test results and the development of a more concrete plan. It would be interesting to know if the time to follow-up for the two groups would be the same if the 16% who did not know their HIV status at discharge were excluded. This suggests that knowledge of HIV status drives follow-up time and would lend some support to the notion that patient discharge was delayed for test results and clarification of the follow-up treatment plan.
Even putting aside the difference in length of stay, the difference of rapid testing on improved knowledge of HIV status and quicker follow-up is likely real and meaningful. Although this study was not designed to assess the impact of this knowledge on patient behavior, immediate knowledge of HIV status during hospitalization may translate to decreased transmission as patients alter their behavior and lends further credibility to the utility of rapid HIV testing in conjunction with conventional methods in the management of inpatients. TH
Community Teaching
Halasyamani L, Valenstein P, Friedlander M. et al. A comparison of two hospitalist models with traditional care in a community teaching hospital. Am J Med. 2005;118:536-543.
Background: A growing body of literature has demonstrated the effects of hospitalists on reducing inpatient length of stay and cost of care, with some literature showing a decreased in-hospital and 30-day mortality. However, most prior studies were conducted in academic medical centers or health maintenance organizations where one group of hospitalists, employed by the institution within which they worked, was compared with traditional primary care physicians. Direct comparisons between different hospitalist models practicing within a single institution have not been published. As a result, the impact of different hospitalist characteristics, including employment status and reimbursement incentives, on inpatient resource utilization and patient care outcomes is unknown.
Methods: Halasyamani and colleagues conducted a retrospective cohort study of 10,595 patients in a tertiary care community-based teaching hospital in which private hospitalists, academic hospitalists, and community physicians all practice. They measured risk-adjusted length of stay, variable costs, 30-day readmission rates, and in-hospital and 30-day mortality for patients treated by each of these three groups, controlling for potentially confounding variables. Community physicians belonged to 21 rounding groups, most of which were private or solo. Two of the community physicians groups were hospital-owned practices reimbursed by a relative value unit system. The private hospitalist group was self-employed with no financial relationship to the hospital and worked an average of 40 weeks per year. Community physicians and private hospitalists worked Monday-Friday and covered weekends or holidays about 25% of the time. Academic hospitalists worked with internal medicine residents and students on a teaching service. They were employed by the hospital using a relative value unit system. They worked an average of 14 weeks per year as an inpatient attending in half-month rotations, which included weekend coverage.
Results: There was a 20% reduction (-0.72 days absolute difference) in length of stay on the academic hospitalist service (P<0.0001) and 8% (-0.28 days absolute difference) on the private hospitalist service (P=0.049) compared with community physicians. Case-mix adjusted relative total costs were 10% less ($173 absolute difference) on the academic (P<0.0001) and 6% less ($109 absolute difference) on the private hospitalist services (P=0.02) compared with community physicians. There were no differences in 30-day readmission, in-hospital and 30-day mortality between the three groups.
Discussion: This study is the first to look at the effects of two separate hospitalist models on resource utilization and patient outcomes within the same institution. Although both the academic and private hospitalist groups demonstrated improved resource utilization as compared with the community physicians, the magnitude of benefit was much greater for the academic hospitalist group.
As the authors point out, one major difference between the two groups was employment status, with the academic hospitalists employed directly by the hospital and the private hospitalists receiving all payment directly from payers. Previous studies have also focused on hospitalists, which were employed by the institution at which they worked, raising the question of whether alignment of employee and employer incentives is an important factor affecting resource utilization outcomes.
Results of this study highlight the need for more studies which seek to clarify specific physician-level, group-level, and organization-level characteristics of hospitalists that result in improved resource utilization and patient care outcomes.
The Last Few Hours
Bailey FA, Burgio KL, Woodby LL, et al. Improving the processes of hospital care during the last hours of life. Arch Int Med. 2005;165(15):1722-1727.
Background: End-of-life care in the acute care inpatient setting is often not initiated until very late in the dying process and may be related to inadequate early recognition of dying patients as well as difficulty transitioning from disease-modifying treatments to palliative measures. Additional barriers exist, including lack of familiarity of hospital staff with initiation and implementation of hospice care. Education about end-of-life care and introduction of a physician-led palliative care team available for consultation within acute care hospitals may help promote better recognition of the dying patient by staff and allow for a “good death.”
Methods: A single hospital within the Veterans Affairs (VA) medical system (Birmingham, Ala., VA Medical Center) was chosen as a pilot center for initiation of a physician-led Inpatient Comfort Care Program (ICCP). The study was framed as a “before-after intervention trial” and analyzed all inpatient deaths identified by the Computerized Patient Recognition System during a six-month period before and substantially after the introduction of the ICCP. A structured chart abstraction tool was used and data was obtained from the last seven days of hospitalization analyzing variables associated with recognition of the dying patient and initiation of palliative care. Education of hospital staff on both hospice care and case identification was initiated during the intervention phase of the study. Additionally, a flexible comfort care order set was introduced.
Results: Two hundred and three veterans were identified (98% men, average age 68) and no significant differences in clinical characteristics were noted between the two groups, pre-intervention and post-intervention. Post-intervention, 59.3% of patients had formal palliative care consultation. Significant findings (P<0.01) following implementation of ICCP were increased documentation of end-of-life symptoms, increased documentation of care plans, increased utilization of opioids (57.1% to 87.2%), increased initiation of do-not-resuscitate orders (61.9% to 85.1%) with a concurrent decrease in cardiopulmonary resuscitation at death (34.4% to 15.4%), and a surprising increase in restraint use (6.0% to 22.6%).
Discussion: Data on hospice care patients indicate that 10% to 30% die in an acute care hospital, identifying a need for increased education and training in palliative medicine. This study demonstrates the positive outcomes of implementation of an inpatient palliative care service both for heightened awareness of identifying the dying patient as well as initiation of end-of-life care. The increased use of opioid medications is an important marker given that many patients experience pain and dyspnea at the end of life. This study is limited by its single site and further validation at other centers implementing similar protocols and assessing similar outcomes is needed. While this intervention had important clinical benefits, additional studies examining the cost implications of this system would be helpful.
Education alone has not been shown to be entirely effective in creating change. This single-site implementation of a palliative care consultation service successfully integrated an education program with on-site consultants. Distributing pocket cards with clinical findings identifying the dying patient aided in recognition of those patients and pre-printed order sets facilitated initiation of end-of-life care. The intervention initiated is possible for many medical centers and promotes an environment allowing for a “good death” for dying patients.
Computers, Doctors, and Errors
Koppel R, Metlay JP, Cohen A, et al. Role of computerized physician order entry systems in facilitating medication errors. JAMA. 2005;293(10):1197-1203
For physicians, computerized physician order entry (CPOE) has become an important topic of discussion as many hospitals and health systems embark on the complex and lengthy process of implementing new enterprise clinical systems. Though there are undoubtedly benefits to such systems, practicing clinicians are apt to remain skeptical of the grandiose pictures the more vocal advocates of CPOE may paint. This is not to say that the promises of CPOE are empty; to the contrary, there have been substantial successes, notably in the realm of medication error prevention.
At the same time, CPOE is a mixture of complex technologies that interface in complicated ways with the culture of clinical medicine. The view that medical informatics is a technical problem that has been solved long ago is simplistic and naïve. The article by Koppel, et al, has two important implications: 1) It is critical to look at clinical information systems in the social milieu in which it functions, and 2) there are often unintended consequences that may not beneficial.
This article examines a widely used, commercial CPOE system in use at the University of Pennsylvania (Philadelphia) using both quantitative and qualitative methods. The researchers conducted focus groups and expert interviews in addition to field observations of physicians (house officers and attendings), nurses, and pharmacists in order to identify themes relating to work with the order entry system. This work helped to guide the creation of a survey instrument subsequently used to survey house staff about working conditions and sources of error and stress. There was an 85%-90% response rate that primarily included house staff who ordered more than nine medication orders per month.
Researchers found two broad categories of errors that were fostered in this environment. The first category, which they termed “information errors” were generated by fragmentation of data and the failure to integrate the hospitals various systems both electronic and paper. Examples of this type of error include antibiotic renewal failures. A common way this failure would occur is that renewal reminder stickers would be placed in the patients’ charts, but the house staff would overlook these because medication orders occurred electronically. Another example is assumed dose errors, where house staff would assume that the default dose displayed was the recommended starting dose, when in fact this was the smallest dose unit available. Physicians were assuming decision support was available when it was not.
The second type of error, human-machine interface flaws, occurred when machine rules did not correspond to work behaviors. An example of this is when patients were listed alphabetically rather than by service, making it easy to select the wrong patient. In another instance, many screens (up to 20) were required to view all of a patient’s medications, making it difficult to choose a correct medication for editing.
This study has been criticized by industry advocates for focusing on an older set of technologies or because a number of these issues related to training or “user factors.” At the other extreme, this study has been cited as a cautionary tale about the risks of CPOE. Both types of criticism miss the point. This study demonstrates that CPOE and the social environment in which it sits is a complex entity and that careful design, proper support, and maintenance are critical ingredients to the success of an incredibly complex but vital new component of hospital medicine.
Point-of-Care HIV Testing in Inpatients
Lubelchek R, Kroc K, Hota B, et al. The role of rapid vs conventional human immunodeficiency virus testing for inpatients: effects on quality of care. Arch Intern Med. 2005;165:1956-1960.
Despite advances in treatment, infection with HIV and AIDS remains a public health problem in the United States. According to the CDC the rate of new diagnosis of HIV infection has remained steady from 2000 to 2003 at about 20 per 100,000 people. (Centers for Disease Control and Prevention. Diagnosis of HIV/AIDS–32 states, 2009-2003. MMWR Morb Mortal Wkly Rep. 2004;53:1106-1110). Currently, about 850,000 to 950,000 people are believed to be living with HIV infection, and it is estimated that 180,000 to 280,000 are unaware of their diagnosis. (Fleming P, Byers RH, Sweeney PA, et al., HIV prevalence in the United States, 2000 [Abstract 11]. Presented at the Ninth Conference on Retroviruses and Opportunistic Infections, Seattle; February 24–28, 2002). These patients are not only at risk for disease progression, but can undermine efforts at disease prevention if they continue to engage in unsafe activities. Thus, increasing awareness of HIV status is an important aspect of disease prevention.
HIV testing remains a challenge. Conventional testing with enzyme immunoassay (EIA) and confirmatory Western blot requires patient follow-up for results, which approximately 25% of patients in various outpatient testing sites fail to do. (Centers for Disease Control and Prevention. Update: HIV counseling and testing using rapid tests, United States, 1995. MMWR Morb Mortal Wkly Rep. 1998;47:211-215). Given the difficulties inherent in the transition of care from the inpatient to outpatient setting, conventional testing in the inpatient setting presents additional barriers to appropriate notification. As various point-of-care HIV tests have been developed for commercial use, the possibility of rapid HIV testing presents an opportunity to reduce notification failure and improve patient care. While not replacing traditional testing, the CDC has endorsed rapid HIV testing as a means to initiate therapy and provide counseling with a particular focus on preventing further disease transmission. In this retrospective study, Lubelchek and colleagues present the effects of a rapid HIV test utilized in the emergency department on various inpatient quality of care measures for those patients who received a positive rapid HIV test later confirmed by Western blot as compared with those patients who were diagnosed after admission by traditional diagnostic methods. This study took place in the context of CDC-funded study of the use of OraQuick (OraSure Technologies, Bethlehem, Pa.) rapid HIV testing in the emergency department at Cook County Hospital in Chicago.
The manufacturer claims the product has a sensitivity of 99.6% and a specificity of 100% as compared with conventional testing. (OraQuick rapid HIV-1 antibody test summary of safety and effectiveness. November 7, 2002. Accessed October 1, 2005, at www.fda.gov/cber/pma/P010047.htm). In the initial study, two of the three emergency department’s treatment pods were equipped to provide HIV screening utilizing the point-of-care technology to consenting patients. Patients in the third pod could be referred to rapid testing based on symptoms or risk factors. All patients who received the rapid test also submitted specimens for conventional EIA and confirmatory Western blot testing. All positive rapid HIV tests were confirmed by Western blot.
In this study, patients who were not known to be infected and were subsequently admitted on non-obstetric or surgical services over 17 months from 2003 to 2004 and confirmed to be HIV positive by Western blot were identified utilizing administrative records. Where possible, charts were reviewed to confirm no prior diagnosis of HIV. Patients who received rapid HIV testing were compared with those who only received conventional testing. Endpoints included time to primary inpatient care service awareness of HIV diagnosis, time to admission or transfer to the inpatient HIV service, time to empiric treatment of diagnosis of opportunistic infection, length of stay, discharge with appropriate prophylactic medications, discharge with patient knowledge of HIV diagnosis, and initial engagement in outpatient care. Length of stay was adjusted by multivariate regression on co-morbid diagnoses (congestive heart failure, end-stage renal disease, cirrhosis, chronic obstructive pulmonary disease, and diabetes), opportunistic infections, ICU admission, need for mechanical ventilation, and CD4 count.
A total of 103 patients were identified with complete chart review completed on 86 of them. All patients except one were admitted through the emergency department. Forty-eight patients were diagnosed initially with the rapid HIV test with 58% of these specifically referred for testing by the emergency department physician, and 55 were diagnosed with conventional testing. Overall, 78% were male, 62% African American, and 20% Hispanic. The two groups were comparable in terms of age, sex, ethnicity, history of substance abuse, HIV risk factors, psychiatric diagnoses, homelessness, CD4 count, presence of opportunistic infections, mechanical ventilation, and co-morbidities. However, conventionally tested patients were more likely to require an ICU stay (31% vs. 10%, P=.01).
Patients in the rapid test group were more quickly documented in the chart as having HIV (.8 vs. 6.4 days, p<.001), placed on an HIV service sooner (1.4 versus 6.9 days, P<.001), initiated outpatient follow-up sooner (21.5 versus 49.5 days, p=.05), and had less unawareness of their HIV status (0 vs. 16%, P=.002). There was no significant difference between the two groups in time from admission to empiric treatment or diagnosis of an opportunistic infection. Patients who received the rapid test did have a lower length of stay (6.4 versus 13.2 days, P<.001). Although much of this difference was due to higher number of ICU stays in the conventional group, in multivariate analysis conventional testing still increased length of stay significantly, OR 5.4 days (2.5, 8.3).
This study suggests that patients who are tested with rapid HIV testing can lead to more efficient inpatient treatment of the complications of HIV, improved patient awareness of HIV status, and quicker outpatient follow-up. These findings have ramifications not just to the inpatient management of patients with HIV but to general public health efforts to reduce the spread of HIV infection.
Nevertheless, these results must be interpreted with caution. They reflect the experience of one institution situated in an area with a high prevalence of HIV. Some degree of selection bias is suggested by the higher presence of ICU admissions in the conventional testing group. The multivariate analysis attempted to control for confounding factors, but the possibility remains that other unrecognized factors may have influenced results. The authors do note that an analysis of patients in the rapid test group stratified by whether the test was performed for screening or by referral of the physician did not demonstrate a statistically significant difference in length of stay. This finding provides further support that the sicker patients which triggered the rapid test had shorter lengths of stay on account of the rapid test and not simply because they were sicker.
As recognized by the authors, physicians in routine practice rely on surrogate markers of HIV infection, most notably a patient’s CD4 count, and thus it is not surprising that the rapid test did not affect time to empiric treatment or diagnosis of opportunistic infection. If treatment did not differ, then explaining the longer length of stay remains an unexplained puzzle. The fact that the two groups were equally matched socially and psychiatrically leaves open the possibility that it was actual knowledge of the HIV test result—and not its effect on treatment—that drove the longer length of stay.
One possibility not suggested by the authors is that definitive knowledge of HIV status helped to mobilize patient discharge. If there were legitimate concerns of follow-up, physicians may have delayed discharge in order to receive HIV test results. Alternatively, some patients may have resisted discharge until receiving test results and the development of a more concrete plan. It would be interesting to know if the time to follow-up for the two groups would be the same if the 16% who did not know their HIV status at discharge were excluded. This suggests that knowledge of HIV status drives follow-up time and would lend some support to the notion that patient discharge was delayed for test results and clarification of the follow-up treatment plan.
Even putting aside the difference in length of stay, the difference of rapid testing on improved knowledge of HIV status and quicker follow-up is likely real and meaningful. Although this study was not designed to assess the impact of this knowledge on patient behavior, immediate knowledge of HIV status during hospitalization may translate to decreased transmission as patients alter their behavior and lends further credibility to the utility of rapid HIV testing in conjunction with conventional methods in the management of inpatients. TH
Community Teaching
Halasyamani L, Valenstein P, Friedlander M. et al. A comparison of two hospitalist models with traditional care in a community teaching hospital. Am J Med. 2005;118:536-543.
Background: A growing body of literature has demonstrated the effects of hospitalists on reducing inpatient length of stay and cost of care, with some literature showing a decreased in-hospital and 30-day mortality. However, most prior studies were conducted in academic medical centers or health maintenance organizations where one group of hospitalists, employed by the institution within which they worked, was compared with traditional primary care physicians. Direct comparisons between different hospitalist models practicing within a single institution have not been published. As a result, the impact of different hospitalist characteristics, including employment status and reimbursement incentives, on inpatient resource utilization and patient care outcomes is unknown.
Methods: Halasyamani and colleagues conducted a retrospective cohort study of 10,595 patients in a tertiary care community-based teaching hospital in which private hospitalists, academic hospitalists, and community physicians all practice. They measured risk-adjusted length of stay, variable costs, 30-day readmission rates, and in-hospital and 30-day mortality for patients treated by each of these three groups, controlling for potentially confounding variables. Community physicians belonged to 21 rounding groups, most of which were private or solo. Two of the community physicians groups were hospital-owned practices reimbursed by a relative value unit system. The private hospitalist group was self-employed with no financial relationship to the hospital and worked an average of 40 weeks per year. Community physicians and private hospitalists worked Monday-Friday and covered weekends or holidays about 25% of the time. Academic hospitalists worked with internal medicine residents and students on a teaching service. They were employed by the hospital using a relative value unit system. They worked an average of 14 weeks per year as an inpatient attending in half-month rotations, which included weekend coverage.
Results: There was a 20% reduction (-0.72 days absolute difference) in length of stay on the academic hospitalist service (P<0.0001) and 8% (-0.28 days absolute difference) on the private hospitalist service (P=0.049) compared with community physicians. Case-mix adjusted relative total costs were 10% less ($173 absolute difference) on the academic (P<0.0001) and 6% less ($109 absolute difference) on the private hospitalist services (P=0.02) compared with community physicians. There were no differences in 30-day readmission, in-hospital and 30-day mortality between the three groups.
Discussion: This study is the first to look at the effects of two separate hospitalist models on resource utilization and patient outcomes within the same institution. Although both the academic and private hospitalist groups demonstrated improved resource utilization as compared with the community physicians, the magnitude of benefit was much greater for the academic hospitalist group.
As the authors point out, one major difference between the two groups was employment status, with the academic hospitalists employed directly by the hospital and the private hospitalists receiving all payment directly from payers. Previous studies have also focused on hospitalists, which were employed by the institution at which they worked, raising the question of whether alignment of employee and employer incentives is an important factor affecting resource utilization outcomes.
Results of this study highlight the need for more studies which seek to clarify specific physician-level, group-level, and organization-level characteristics of hospitalists that result in improved resource utilization and patient care outcomes.
The Last Few Hours
Bailey FA, Burgio KL, Woodby LL, et al. Improving the processes of hospital care during the last hours of life. Arch Int Med. 2005;165(15):1722-1727.
Background: End-of-life care in the acute care inpatient setting is often not initiated until very late in the dying process and may be related to inadequate early recognition of dying patients as well as difficulty transitioning from disease-modifying treatments to palliative measures. Additional barriers exist, including lack of familiarity of hospital staff with initiation and implementation of hospice care. Education about end-of-life care and introduction of a physician-led palliative care team available for consultation within acute care hospitals may help promote better recognition of the dying patient by staff and allow for a “good death.”
Methods: A single hospital within the Veterans Affairs (VA) medical system (Birmingham, Ala., VA Medical Center) was chosen as a pilot center for initiation of a physician-led Inpatient Comfort Care Program (ICCP). The study was framed as a “before-after intervention trial” and analyzed all inpatient deaths identified by the Computerized Patient Recognition System during a six-month period before and substantially after the introduction of the ICCP. A structured chart abstraction tool was used and data was obtained from the last seven days of hospitalization analyzing variables associated with recognition of the dying patient and initiation of palliative care. Education of hospital staff on both hospice care and case identification was initiated during the intervention phase of the study. Additionally, a flexible comfort care order set was introduced.
Results: Two hundred and three veterans were identified (98% men, average age 68) and no significant differences in clinical characteristics were noted between the two groups, pre-intervention and post-intervention. Post-intervention, 59.3% of patients had formal palliative care consultation. Significant findings (P<0.01) following implementation of ICCP were increased documentation of end-of-life symptoms, increased documentation of care plans, increased utilization of opioids (57.1% to 87.2%), increased initiation of do-not-resuscitate orders (61.9% to 85.1%) with a concurrent decrease in cardiopulmonary resuscitation at death (34.4% to 15.4%), and a surprising increase in restraint use (6.0% to 22.6%).
Discussion: Data on hospice care patients indicate that 10% to 30% die in an acute care hospital, identifying a need for increased education and training in palliative medicine. This study demonstrates the positive outcomes of implementation of an inpatient palliative care service both for heightened awareness of identifying the dying patient as well as initiation of end-of-life care. The increased use of opioid medications is an important marker given that many patients experience pain and dyspnea at the end of life. This study is limited by its single site and further validation at other centers implementing similar protocols and assessing similar outcomes is needed. While this intervention had important clinical benefits, additional studies examining the cost implications of this system would be helpful.
Education alone has not been shown to be entirely effective in creating change. This single-site implementation of a palliative care consultation service successfully integrated an education program with on-site consultants. Distributing pocket cards with clinical findings identifying the dying patient aided in recognition of those patients and pre-printed order sets facilitated initiation of end-of-life care. The intervention initiated is possible for many medical centers and promotes an environment allowing for a “good death” for dying patients.
Computers, Doctors, and Errors
Koppel R, Metlay JP, Cohen A, et al. Role of computerized physician order entry systems in facilitating medication errors. JAMA. 2005;293(10):1197-1203
For physicians, computerized physician order entry (CPOE) has become an important topic of discussion as many hospitals and health systems embark on the complex and lengthy process of implementing new enterprise clinical systems. Though there are undoubtedly benefits to such systems, practicing clinicians are apt to remain skeptical of the grandiose pictures the more vocal advocates of CPOE may paint. This is not to say that the promises of CPOE are empty; to the contrary, there have been substantial successes, notably in the realm of medication error prevention.
At the same time, CPOE is a mixture of complex technologies that interface in complicated ways with the culture of clinical medicine. The view that medical informatics is a technical problem that has been solved long ago is simplistic and naïve. The article by Koppel, et al, has two important implications: 1) It is critical to look at clinical information systems in the social milieu in which it functions, and 2) there are often unintended consequences that may not beneficial.
This article examines a widely used, commercial CPOE system in use at the University of Pennsylvania (Philadelphia) using both quantitative and qualitative methods. The researchers conducted focus groups and expert interviews in addition to field observations of physicians (house officers and attendings), nurses, and pharmacists in order to identify themes relating to work with the order entry system. This work helped to guide the creation of a survey instrument subsequently used to survey house staff about working conditions and sources of error and stress. There was an 85%-90% response rate that primarily included house staff who ordered more than nine medication orders per month.
Researchers found two broad categories of errors that were fostered in this environment. The first category, which they termed “information errors” were generated by fragmentation of data and the failure to integrate the hospitals various systems both electronic and paper. Examples of this type of error include antibiotic renewal failures. A common way this failure would occur is that renewal reminder stickers would be placed in the patients’ charts, but the house staff would overlook these because medication orders occurred electronically. Another example is assumed dose errors, where house staff would assume that the default dose displayed was the recommended starting dose, when in fact this was the smallest dose unit available. Physicians were assuming decision support was available when it was not.
The second type of error, human-machine interface flaws, occurred when machine rules did not correspond to work behaviors. An example of this is when patients were listed alphabetically rather than by service, making it easy to select the wrong patient. In another instance, many screens (up to 20) were required to view all of a patient’s medications, making it difficult to choose a correct medication for editing.
This study has been criticized by industry advocates for focusing on an older set of technologies or because a number of these issues related to training or “user factors.” At the other extreme, this study has been cited as a cautionary tale about the risks of CPOE. Both types of criticism miss the point. This study demonstrates that CPOE and the social environment in which it sits is a complex entity and that careful design, proper support, and maintenance are critical ingredients to the success of an incredibly complex but vital new component of hospital medicine.
Point-of-Care HIV Testing in Inpatients
Lubelchek R, Kroc K, Hota B, et al. The role of rapid vs conventional human immunodeficiency virus testing for inpatients: effects on quality of care. Arch Intern Med. 2005;165:1956-1960.
Despite advances in treatment, infection with HIV and AIDS remains a public health problem in the United States. According to the CDC the rate of new diagnosis of HIV infection has remained steady from 2000 to 2003 at about 20 per 100,000 people. (Centers for Disease Control and Prevention. Diagnosis of HIV/AIDS–32 states, 2009-2003. MMWR Morb Mortal Wkly Rep. 2004;53:1106-1110). Currently, about 850,000 to 950,000 people are believed to be living with HIV infection, and it is estimated that 180,000 to 280,000 are unaware of their diagnosis. (Fleming P, Byers RH, Sweeney PA, et al., HIV prevalence in the United States, 2000 [Abstract 11]. Presented at the Ninth Conference on Retroviruses and Opportunistic Infections, Seattle; February 24–28, 2002). These patients are not only at risk for disease progression, but can undermine efforts at disease prevention if they continue to engage in unsafe activities. Thus, increasing awareness of HIV status is an important aspect of disease prevention.
HIV testing remains a challenge. Conventional testing with enzyme immunoassay (EIA) and confirmatory Western blot requires patient follow-up for results, which approximately 25% of patients in various outpatient testing sites fail to do. (Centers for Disease Control and Prevention. Update: HIV counseling and testing using rapid tests, United States, 1995. MMWR Morb Mortal Wkly Rep. 1998;47:211-215). Given the difficulties inherent in the transition of care from the inpatient to outpatient setting, conventional testing in the inpatient setting presents additional barriers to appropriate notification. As various point-of-care HIV tests have been developed for commercial use, the possibility of rapid HIV testing presents an opportunity to reduce notification failure and improve patient care. While not replacing traditional testing, the CDC has endorsed rapid HIV testing as a means to initiate therapy and provide counseling with a particular focus on preventing further disease transmission. In this retrospective study, Lubelchek and colleagues present the effects of a rapid HIV test utilized in the emergency department on various inpatient quality of care measures for those patients who received a positive rapid HIV test later confirmed by Western blot as compared with those patients who were diagnosed after admission by traditional diagnostic methods. This study took place in the context of CDC-funded study of the use of OraQuick (OraSure Technologies, Bethlehem, Pa.) rapid HIV testing in the emergency department at Cook County Hospital in Chicago.
The manufacturer claims the product has a sensitivity of 99.6% and a specificity of 100% as compared with conventional testing. (OraQuick rapid HIV-1 antibody test summary of safety and effectiveness. November 7, 2002. Accessed October 1, 2005, at www.fda.gov/cber/pma/P010047.htm). In the initial study, two of the three emergency department’s treatment pods were equipped to provide HIV screening utilizing the point-of-care technology to consenting patients. Patients in the third pod could be referred to rapid testing based on symptoms or risk factors. All patients who received the rapid test also submitted specimens for conventional EIA and confirmatory Western blot testing. All positive rapid HIV tests were confirmed by Western blot.
In this study, patients who were not known to be infected and were subsequently admitted on non-obstetric or surgical services over 17 months from 2003 to 2004 and confirmed to be HIV positive by Western blot were identified utilizing administrative records. Where possible, charts were reviewed to confirm no prior diagnosis of HIV. Patients who received rapid HIV testing were compared with those who only received conventional testing. Endpoints included time to primary inpatient care service awareness of HIV diagnosis, time to admission or transfer to the inpatient HIV service, time to empiric treatment of diagnosis of opportunistic infection, length of stay, discharge with appropriate prophylactic medications, discharge with patient knowledge of HIV diagnosis, and initial engagement in outpatient care. Length of stay was adjusted by multivariate regression on co-morbid diagnoses (congestive heart failure, end-stage renal disease, cirrhosis, chronic obstructive pulmonary disease, and diabetes), opportunistic infections, ICU admission, need for mechanical ventilation, and CD4 count.
A total of 103 patients were identified with complete chart review completed on 86 of them. All patients except one were admitted through the emergency department. Forty-eight patients were diagnosed initially with the rapid HIV test with 58% of these specifically referred for testing by the emergency department physician, and 55 were diagnosed with conventional testing. Overall, 78% were male, 62% African American, and 20% Hispanic. The two groups were comparable in terms of age, sex, ethnicity, history of substance abuse, HIV risk factors, psychiatric diagnoses, homelessness, CD4 count, presence of opportunistic infections, mechanical ventilation, and co-morbidities. However, conventionally tested patients were more likely to require an ICU stay (31% vs. 10%, P=.01).
Patients in the rapid test group were more quickly documented in the chart as having HIV (.8 vs. 6.4 days, p<.001), placed on an HIV service sooner (1.4 versus 6.9 days, P<.001), initiated outpatient follow-up sooner (21.5 versus 49.5 days, p=.05), and had less unawareness of their HIV status (0 vs. 16%, P=.002). There was no significant difference between the two groups in time from admission to empiric treatment or diagnosis of an opportunistic infection. Patients who received the rapid test did have a lower length of stay (6.4 versus 13.2 days, P<.001). Although much of this difference was due to higher number of ICU stays in the conventional group, in multivariate analysis conventional testing still increased length of stay significantly, OR 5.4 days (2.5, 8.3).
This study suggests that patients who are tested with rapid HIV testing can lead to more efficient inpatient treatment of the complications of HIV, improved patient awareness of HIV status, and quicker outpatient follow-up. These findings have ramifications not just to the inpatient management of patients with HIV but to general public health efforts to reduce the spread of HIV infection.
Nevertheless, these results must be interpreted with caution. They reflect the experience of one institution situated in an area with a high prevalence of HIV. Some degree of selection bias is suggested by the higher presence of ICU admissions in the conventional testing group. The multivariate analysis attempted to control for confounding factors, but the possibility remains that other unrecognized factors may have influenced results. The authors do note that an analysis of patients in the rapid test group stratified by whether the test was performed for screening or by referral of the physician did not demonstrate a statistically significant difference in length of stay. This finding provides further support that the sicker patients which triggered the rapid test had shorter lengths of stay on account of the rapid test and not simply because they were sicker.
As recognized by the authors, physicians in routine practice rely on surrogate markers of HIV infection, most notably a patient’s CD4 count, and thus it is not surprising that the rapid test did not affect time to empiric treatment or diagnosis of opportunistic infection. If treatment did not differ, then explaining the longer length of stay remains an unexplained puzzle. The fact that the two groups were equally matched socially and psychiatrically leaves open the possibility that it was actual knowledge of the HIV test result—and not its effect on treatment—that drove the longer length of stay.
One possibility not suggested by the authors is that definitive knowledge of HIV status helped to mobilize patient discharge. If there were legitimate concerns of follow-up, physicians may have delayed discharge in order to receive HIV test results. Alternatively, some patients may have resisted discharge until receiving test results and the development of a more concrete plan. It would be interesting to know if the time to follow-up for the two groups would be the same if the 16% who did not know their HIV status at discharge were excluded. This suggests that knowledge of HIV status drives follow-up time and would lend some support to the notion that patient discharge was delayed for test results and clarification of the follow-up treatment plan.
Even putting aside the difference in length of stay, the difference of rapid testing on improved knowledge of HIV status and quicker follow-up is likely real and meaningful. Although this study was not designed to assess the impact of this knowledge on patient behavior, immediate knowledge of HIV status during hospitalization may translate to decreased transmission as patients alter their behavior and lends further credibility to the utility of rapid HIV testing in conjunction with conventional methods in the management of inpatients. TH