Original Research

From the University of Michigan Medical School, Ann Arbor, MI.

Abstract

• Objective: Existing research has demonstrated overall low rates of residents establishing care with a primary care physician (PCP). We conducted a survey-based study to better understand chronic illness, PCP utilization, and prescription medication use patterns in resident physician populations.
• Methods: In 2017, we invited internal and family medicine trainees from a convenience sample of U.S. residency programs to participate in a survey. We compared the characteristics of residents who had established care with a PCP to those who had not.
• Results: The response rate was 45% (348/766 residents). The majority (n = 205, 59%) of respondents stated they had established care with a PCP primarily for routine preventative care (n = 159, 79%) and access in the event of an emergency (n = 132, 66%). However, 31% (n = 103) denied having had a wellness visit in over 3 years. Nearly a quarter of residents (n = 77, 23%) reported a chronic medical illness and 14% (n = 45) reported a preexisting mental health condition prior to residency. One-third (n = 111, 33%) reported taking a long-term prescription medication. Compared to residents who had not established care, those with a PCP (n = 205) more often reported a chronic condition (P < 0.001), seeing a subspecialist (P = 0.01), or taking long-term prescription medications (P < 0.001). One in 5 (n = 62,19%) respondents reported receiving prescriptions for an acute illness from an individual with whom they did not have a doctor-patient relationship.
• Conclusion: Medical residents have a substantial burden of chronic illness that may not be met through interactions with PCPs. Further understanding their medical needs and barriers to accessing care is necessary to ensure trainee well-being.

Keywords: Medical education-graduate, physician behavior, survey research, access to care.

Although internal medicine (IM) and family medicine (FM) residents must learn to provide high-quality primary care to their patients, little is known about whether they appropriately access such care themselves. Resident burnout and resilience has received attention [1,2], but there has been limited focus on understanding the burden of chronic medical and mental illness among residents. In particular, little is known about whether residents access primary care physicians (PCPs)—for either acute or chronic medical needs—and about resident self-medication practices.

Residency is often characterized by a life-changing geographic relocation. Even residents who do not relocate may still need to establish care with a new PCP due to health insurance or loss of access to a student clinic [3]. Establishing primary care with a new doctor typically requires scheduling a new patient visit, often with a wait time of several days to weeks [4,5]. Furthermore, lack of time, erratic schedules, and concerns about privacy and the stigma of being ill as a physician are barriers to establishing care [6-8]. Individuals who have not established primary care may experience delays in routine preventative health services, screening for chronic medical and mental health conditions, as well as access to care during acute illnesses [9,10]. Worse, they may engage in potentially unsafe practices, such as having colleagues write prescriptions for them, or even self-prescribing [8,11,12].

Existing research has demonstrated overall low rates of residents establishing care with a PCP [6–8,13]. However, these studies have either been limited to large academic centers or conducted outside the United States. Improving resident well-being may prove challenging without a clear understanding of current primary care utilization practices, the burden of chronic illness among residents, and patterns of prescription medication use and needs. Therefore, we conducted a survey-based study to understand primary care utilization and the burden of chronic illness among residents. We also assessed whether lack of primary care is associated with potentially risky behaviors, such as self-prescribing of medications.

Methods

Study Setting and Participants

The survey was distributed to current residents at IM and FM programs within the United States in 2017. Individual programs were recruited by directly contacting program directors or chief medical residents via email. Rather than contacting sites directly through standard templated emails, we identified programs both through personal contacts as well as the Electronic Residency Application Service list of accredited IM training programs. We elected to use this approach in order to increase response rates and to ensure that a sample representative of the trainee population was constructed. Programs were located in the Northeast, Midwest, South, and Pacific regions, and included small community-based programs and large academic centers.

Development of the Survey

The survey instrument was developed by the authors and reviewed by residents and PCPs at the University of Michigan to ensure relevance and comprehension of questions (The survey is available in the Appendix.). Once finalized, the survey was programmed into an online survey tool (Qualtrics, Provo, UT) and pilot-tested before being disseminated to the sampling frame. Data collected in the survey included: respondent’s utilization of a PCP, burden of chronic illness, long-term prescription medications, prescribing source, and demographic characteristics.

Each participating program distributed the survey to their residents through an email containing an anonymous hyperlink. The survey was available for completion for 4 weeks. We asked participating programs to send email reminders to encourage participation. Participants were given the option of receiving a $10 Amazon gift card after completion. All responses were recorded anonymously. The study received a “not regulated” status by the University of Michigan Institutional Review Board (HUM 00123888).

Statistical Analysis

Descriptive statistics were used to tabulate results. Respondents were encouraged, but not required, to answer all questions. Therefore, the response rate for each question was calculated using the total number of responses for that question as the denominator. Bivariable comparisons were made using Chi-squared or Fisher’s exact tests, as appropriate, for categorical data. A P value < 0.05, with 2-sided alpha, was considered statistically significant. All statistical analyses were conducted using Stata 13 SE (StataCorp, College Station, TX).

Results

Respondent Characteristics

Of the 29 programs contacted, 10 agreed to participate within the study timeframe. Of 766 potential respondents, 348 (45%) residents answered the survey (Table 1). The majority of respondents (n = 276, 82%) were from IM programs. Respondents were from all training years as follows: postgraduate year 1 residents (PGY-1, or interns; n = 130, 39%), PGY-2 residents (n = 98, 29%), PGY-3 residents (n = 93, 28%), and PGY-4 residents (n = 12, 4%). Most respondents were from the South (n = 130, 39%) and Midwest (n = 123, 37%) regions, and over half (n = 179, 54%) were female. Most respondents (n = 285, 86%) stated that they did not have children. The majority (n = 236, 71%) were completing residency in an area where they had not previously lived for more than 1 year.

Primary Care Utilization

Among the 348 respondents, 59% (n = 205) reported having established care with a PCP. An additional 6% (n = 21) had established care with an obstetrician/gynecologist for routine needs (Table 2). The 2 most common reasons for establishing care with a PCP were routine primary care needs, including contraception (n = 159, 79%), and access to a physician in the event of an acute medical need (n = 132, 66%).

Among respondents who had established care with a PCP, most (n = 188, 94%) had completed at least 1 appointment. However, among these 188 respondents, 68% (n = 127) stated that they had not made an acute visit in more than 12 months. When asked about wellness visits, almost one third of respondents (n = 103, 31%) stated that they had not been seen for a wellness visit in the past 3 years.

Burden of Chronic Illness

Most respondents (n = 223, 67%) stated that they did not have a chronic medical or mental health condition prior to residency (Table 3). However, 23% (n = 77) of respondents stated that they had been diagnosed with a chronic medical illness prior to residency, and 14% (n = 45) indicated they had been diagnosed with a mental health condition prior to residency. Almost one fifth of respondents (n = 60, 18%) reported seeing a subspecialist for a medical illness, and 33% (n = 111) reported taking a long-term prescription medication. With respect to major medical issues, the majority of residents (n = 239, 72%) denied experiencing events such as pregnancy, hospitalization, surgery, or an emergency department (ED) visit during training.

[polldaddy:10116940]

Inappropriate Prescriptions

While the majority of respondents denied writing a prescription for themselves for an acute or chronic medical condition, almost one fifth (n = 62, 19%) had received a prescription for an acute medical need from a provider outside of a clinical relationship (ie, from someone other than their PCP or specialty provider). Notably, 5% (n = 15) reported that this had occurred at least 2 or 3 times in the past 12 months (Table 4). Compared to respondents not taking long-term prescription medications, respondents who were already taking long-term prescription medications more frequently reported inappropriately receiving chronic prescriptions outside of an established clinical relationship (n = 14, 13% vs. n = 14, 6%; P = 0.05) and more often self-prescribed medications for acute needs (n = 12, 11% vs. n = 7, 3%; P = 0.005).

Important differences were noted between residents who had a PCP versus those who did not (Table 5). For example, a higher percentage of residents with a PCP indicated they had been diagnosed with a chronic medical illness (n = 55, 28% vs. n = 22, 16%; P = 0.01) or a chronic mental health condition (n = 34, 17% vs. n = 11, 8%; P = 0.02) before residency. Additionally, a higher percentage of residents with a PCP (n = 70, 35% vs. n = 25, 18%; P = 0.001) reported experiencing medical events such as pregnancy, hospitalization, surgery, ED visit, or new diagnosis of a chronic medical illness during residency. Finally, a higher percentage of respondents with a PCP stated that they had visited a subspecialist for a medical illness (n = 44, 22% vs. n = 16,12%; P = 0.01) or were taking long-term prescription medications (n = 86, 43% vs. n = 25; 18%; P < 0.001). When comparing PGY-1 to PGY-2–PGY-4 residents, the former reported having established a medical relationship with a PCP significantly less frequently (n = 56, 43% vs. n = 142, 70%; P < 0.001).

Discussion

This survey-based study of medical residents across the United States suggests that a substantial proportion do not establish relationships with PCPs. Additionally, our data suggest that despite establishing care, few residents subsequently visited their PCP during training for wellness visits or routine care. Self-reported rates of chronic medical and mental health conditions were substantial in our sample. Furthermore, inappropriate self-prescription and the receipt of prescriptions outside of a medical relationship were also reported. These findings suggest that future studies that focus on the unique medical and mental health needs of physicians in training, as well as interventions to encourage care in this vulnerable period, are necessary.

We observed that most respondents that established primary care were female trainees. Although it is impossible to know with certainty, one hypothesis behind this discrepancy is that women routinely need to access preventative care for gynecologic needs such as pap smears, contraception, and potentially pregnancy and preconception counseling [14,15]. Similarly, residents with a chronic medical or mental health condition prior to residency established care with a local PCP at a significantly greater frequency than those without such diagnoses. While selection bias cannot be excluded, this finding suggests that illness is a driving factor in establishing care. There also appears to be an association between accessing the medical system (either for prescription medications or subspecialist care) and having established care with a PCP. Collectively, these data suggest that individuals without a compelling reason to access medical services might have barriers to accessing care in the event of medical needs or may not receive routine preventative care [9,10].

In general, we found that rates of reported inappropriate prescriptions were lower than those reported in prior studies where a comparable resident population was surveyed [8,12,16]. Inclusion of multiple institutions, differences in temporality, social desirability bias, and reporting bias might have influenced our findings in this regard. Surprisingly, we found that having a PCP did not influence likelihood of inappropriate prescription receipt, perhaps suggesting that this behavior reflects some degree of universal difficulty in accessing care. Alternatively, this finding might relate to a cultural tendency to self-prescribe among resident physicians. The fact that individuals on chronic medications more often both received and wrote inappropriate prescriptions suggests this problem might be more pronounced in individuals who take medications more often, as these residents have specific needs [12]. Future studies targeting these individuals thus appear warranted.

Our study has several limitations. First, our sample size was modest and the response rate of 45% was low. However, to our knowledge, this remains among the largest survey on this topic, and our response rate is comparable to similar trainee studies [8,11,13]. Second, we designed and created a novel survey for this study. While the questions were pilot-tested with users prior to dissemination, validation of the instrument was not performed. Third, since the study population was restricted to residents in fields that participate in primary care, our findings may not be generalizable to patterns of PCP use in other specialties [6].

These limitations aside, our study has important strengths. This is the first national study of its kind with specific questions addressing primary care access and utilization, prescription medication use and related practices, and the prevalence of medical conditions among trainees. Important differences in the rates of establishing primary care between male and female respondents, first- year and senior residents, and those with and without chronic disease suggest a need to target specific resident groups (males, interns, those without pre-existing conditions) for wellness-related interventions. Such interventions could include distribution of a list of local providers to first year residents, advanced protected time for doctor’s appointments, and safeguards to ensure health information is protected from potential supervisors. Future studies should also include residents from non-primary care oriented specialties such as surgery, emergency medicine, and anesthesiology to obtain results that are more generalizable to the resident population as a whole. Additionally, the rates of inappropriate prescriptions were not insignificant and warrant further evaluation of the driving forces behind these behaviors.

Conclusion

Medical residents have a substantial burden of chronic illness that may not be met through interactions with PCPs. More research into barriers that residents face while accessing care and an assessment of interventions to facilitate their access to care is important to promote trainee well-being. Without such direction and initiative, it may prove harder for physicians to heal themselves or those for whom they provide care.

Acknowledgments: We thank Suzanne Winter, the study coordinator, for her support with editing and formatting the manuscript, Latoya Kuhn for performing the statistical analysis and creating data tables, and Dr. Namita Sachdev and Dr. Renuka Tipirneni for providing feedback on the survey instrument. We also thank the involved programs for their participation.

Corresponding author: Vineet Chopra, NCRC 2800 Plymouth Rd., Bldg 16, 432, Ann Arbor, MI 48109, [email protected].

Financial disclosures: None.

Previous presentations: Results were presented at the Annual Michigan Medicine 2017 Internal Medicine Research Symposium.

From the University of Michigan Medical School, Ann Arbor, MI.

Abstract

• Objective: Existing research has demonstrated overall low rates of residents establishing care with a primary care physician (PCP). We conducted a survey-based study to better understand chronic illness, PCP utilization, and prescription medication use patterns in resident physician populations.
• Methods: In 2017, we invited internal and family medicine trainees from a convenience sample of U.S. residency programs to participate in a survey. We compared the characteristics of residents who had established care with a PCP to those who had not.
• Results: The response rate was 45% (348/766 residents). The majority (n = 205, 59%) of respondents stated they had established care with a PCP primarily for routine preventative care (n = 159, 79%) and access in the event of an emergency (n = 132, 66%). However, 31% (n = 103) denied having had a wellness visit in over 3 years. Nearly a quarter of residents (n = 77, 23%) reported a chronic medical illness and 14% (n = 45) reported a preexisting mental health condition prior to residency. One-third (n = 111, 33%) reported taking a long-term prescription medication. Compared to residents who had not established care, those with a PCP (n = 205) more often reported a chronic condition (P < 0.001), seeing a subspecialist (P = 0.01), or taking long-term prescription medications (P < 0.001). One in 5 (n = 62,19%) respondents reported receiving prescriptions for an acute illness from an individual with whom they did not have a doctor-patient relationship.
• Conclusion: Medical residents have a substantial burden of chronic illness that may not be met through interactions with PCPs. Further understanding their medical needs and barriers to accessing care is necessary to ensure trainee well-being.

Keywords: Medical education-graduate, physician behavior, survey research, access to care.

Although internal medicine (IM) and family medicine (FM) residents must learn to provide high-quality primary care to their patients, little is known about whether they appropriately access such care themselves. Resident burnout and resilience has received attention [1,2], but there has been limited focus on understanding the burden of chronic medical and mental illness among residents. In particular, little is known about whether residents access primary care physicians (PCPs)—for either acute or chronic medical needs—and about resident self-medication practices.

Residency is often characterized by a life-changing geographic relocation. Even residents who do not relocate may still need to establish care with a new PCP due to health insurance or loss of access to a student clinic [3]. Establishing primary care with a new doctor typically requires scheduling a new patient visit, often with a wait time of several days to weeks [4,5]. Furthermore, lack of time, erratic schedules, and concerns about privacy and the stigma of being ill as a physician are barriers to establishing care [6-8]. Individuals who have not established primary care may experience delays in routine preventative health services, screening for chronic medical and mental health conditions, as well as access to care during acute illnesses [9,10]. Worse, they may engage in potentially unsafe practices, such as having colleagues write prescriptions for them, or even self-prescribing [8,11,12].

Existing research has demonstrated overall low rates of residents establishing care with a PCP [6–8,13]. However, these studies have either been limited to large academic centers or conducted outside the United States. Improving resident well-being may prove challenging without a clear understanding of current primary care utilization practices, the burden of chronic illness among residents, and patterns of prescription medication use and needs. Therefore, we conducted a survey-based study to understand primary care utilization and the burden of chronic illness among residents. We also assessed whether lack of primary care is associated with potentially risky behaviors, such as self-prescribing of medications.

Methods

Study Setting and Participants

The survey was distributed to current residents at IM and FM programs within the United States in 2017. Individual programs were recruited by directly contacting program directors or chief medical residents via email. Rather than contacting sites directly through standard templated emails, we identified programs both through personal contacts as well as the Electronic Residency Application Service list of accredited IM training programs. We elected to use this approach in order to increase response rates and to ensure that a sample representative of the trainee population was constructed. Programs were located in the Northeast, Midwest, South, and Pacific regions, and included small community-based programs and large academic centers.

Development of the Survey

The survey instrument was developed by the authors and reviewed by residents and PCPs at the University of Michigan to ensure relevance and comprehension of questions (The survey is available in the Appendix.). Once finalized, the survey was programmed into an online survey tool (Qualtrics, Provo, UT) and pilot-tested before being disseminated to the sampling frame. Data collected in the survey included: respondent’s utilization of a PCP, burden of chronic illness, long-term prescription medications, prescribing source, and demographic characteristics.

Each participating program distributed the survey to their residents through an email containing an anonymous hyperlink. The survey was available for completion for 4 weeks. We asked participating programs to send email reminders to encourage participation. Participants were given the option of receiving a $10 Amazon gift card after completion. All responses were recorded anonymously. The study received a “not regulated” status by the University of Michigan Institutional Review Board (HUM 00123888).

Statistical Analysis

Descriptive statistics were used to tabulate results. Respondents were encouraged, but not required, to answer all questions. Therefore, the response rate for each question was calculated using the total number of responses for that question as the denominator. Bivariable comparisons were made using Chi-squared or Fisher’s exact tests, as appropriate, for categorical data. A P value < 0.05, with 2-sided alpha, was considered statistically significant. All statistical analyses were conducted using Stata 13 SE (StataCorp, College Station, TX).

Results

Respondent Characteristics

Of the 29 programs contacted, 10 agreed to participate within the study timeframe. Of 766 potential respondents, 348 (45%) residents answered the survey (Table 1). The majority of respondents (n = 276, 82%) were from IM programs. Respondents were from all training years as follows: postgraduate year 1 residents (PGY-1, or interns; n = 130, 39%), PGY-2 residents (n = 98, 29%), PGY-3 residents (n = 93, 28%), and PGY-4 residents (n = 12, 4%). Most respondents were from the South (n = 130, 39%) and Midwest (n = 123, 37%) regions, and over half (n = 179, 54%) were female. Most respondents (n = 285, 86%) stated that they did not have children. The majority (n = 236, 71%) were completing residency in an area where they had not previously lived for more than 1 year.

Primary Care Utilization

Among the 348 respondents, 59% (n = 205) reported having established care with a PCP. An additional 6% (n = 21) had established care with an obstetrician/gynecologist for routine needs (Table 2). The 2 most common reasons for establishing care with a PCP were routine primary care needs, including contraception (n = 159, 79%), and access to a physician in the event of an acute medical need (n = 132, 66%).

Among respondents who had established care with a PCP, most (n = 188, 94%) had completed at least 1 appointment. However, among these 188 respondents, 68% (n = 127) stated that they had not made an acute visit in more than 12 months. When asked about wellness visits, almost one third of respondents (n = 103, 31%) stated that they had not been seen for a wellness visit in the past 3 years.

Burden of Chronic Illness

Most respondents (n = 223, 67%) stated that they did not have a chronic medical or mental health condition prior to residency (Table 3). However, 23% (n = 77) of respondents stated that they had been diagnosed with a chronic medical illness prior to residency, and 14% (n = 45) indicated they had been diagnosed with a mental health condition prior to residency. Almost one fifth of respondents (n = 60, 18%) reported seeing a subspecialist for a medical illness, and 33% (n = 111) reported taking a long-term prescription medication. With respect to major medical issues, the majority of residents (n = 239, 72%) denied experiencing events such as pregnancy, hospitalization, surgery, or an emergency department (ED) visit during training.

[polldaddy:10116940]

Inappropriate Prescriptions

While the majority of respondents denied writing a prescription for themselves for an acute or chronic medical condition, almost one fifth (n = 62, 19%) had received a prescription for an acute medical need from a provider outside of a clinical relationship (ie, from someone other than their PCP or specialty provider). Notably, 5% (n = 15) reported that this had occurred at least 2 or 3 times in the past 12 months (Table 4). Compared to respondents not taking long-term prescription medications, respondents who were already taking long-term prescription medications more frequently reported inappropriately receiving chronic prescriptions outside of an established clinical relationship (n = 14, 13% vs. n = 14, 6%; P = 0.05) and more often self-prescribed medications for acute needs (n = 12, 11% vs. n = 7, 3%; P = 0.005).

Important differences were noted between residents who had a PCP versus those who did not (Table 5). For example, a higher percentage of residents with a PCP indicated they had been diagnosed with a chronic medical illness (n = 55, 28% vs. n = 22, 16%; P = 0.01) or a chronic mental health condition (n = 34, 17% vs. n = 11, 8%; P = 0.02) before residency. Additionally, a higher percentage of residents with a PCP (n = 70, 35% vs. n = 25, 18%; P = 0.001) reported experiencing medical events such as pregnancy, hospitalization, surgery, ED visit, or new diagnosis of a chronic medical illness during residency. Finally, a higher percentage of respondents with a PCP stated that they had visited a subspecialist for a medical illness (n = 44, 22% vs. n = 16,12%; P = 0.01) or were taking long-term prescription medications (n = 86, 43% vs. n = 25; 18%; P < 0.001). When comparing PGY-1 to PGY-2–PGY-4 residents, the former reported having established a medical relationship with a PCP significantly less frequently (n = 56, 43% vs. n = 142, 70%; P < 0.001).

Discussion

This survey-based study of medical residents across the United States suggests that a substantial proportion do not establish relationships with PCPs. Additionally, our data suggest that despite establishing care, few residents subsequently visited their PCP during training for wellness visits or routine care. Self-reported rates of chronic medical and mental health conditions were substantial in our sample. Furthermore, inappropriate self-prescription and the receipt of prescriptions outside of a medical relationship were also reported. These findings suggest that future studies that focus on the unique medical and mental health needs of physicians in training, as well as interventions to encourage care in this vulnerable period, are necessary.

We observed that most respondents that established primary care were female trainees. Although it is impossible to know with certainty, one hypothesis behind this discrepancy is that women routinely need to access preventative care for gynecologic needs such as pap smears, contraception, and potentially pregnancy and preconception counseling [14,15]. Similarly, residents with a chronic medical or mental health condition prior to residency established care with a local PCP at a significantly greater frequency than those without such diagnoses. While selection bias cannot be excluded, this finding suggests that illness is a driving factor in establishing care. There also appears to be an association between accessing the medical system (either for prescription medications or subspecialist care) and having established care with a PCP. Collectively, these data suggest that individuals without a compelling reason to access medical services might have barriers to accessing care in the event of medical needs or may not receive routine preventative care [9,10].

In general, we found that rates of reported inappropriate prescriptions were lower than those reported in prior studies where a comparable resident population was surveyed [8,12,16]. Inclusion of multiple institutions, differences in temporality, social desirability bias, and reporting bias might have influenced our findings in this regard. Surprisingly, we found that having a PCP did not influence likelihood of inappropriate prescription receipt, perhaps suggesting that this behavior reflects some degree of universal difficulty in accessing care. Alternatively, this finding might relate to a cultural tendency to self-prescribe among resident physicians. The fact that individuals on chronic medications more often both received and wrote inappropriate prescriptions suggests this problem might be more pronounced in individuals who take medications more often, as these residents have specific needs [12]. Future studies targeting these individuals thus appear warranted.

Our study has several limitations. First, our sample size was modest and the response rate of 45% was low. However, to our knowledge, this remains among the largest survey on this topic, and our response rate is comparable to similar trainee studies [8,11,13]. Second, we designed and created a novel survey for this study. While the questions were pilot-tested with users prior to dissemination, validation of the instrument was not performed. Third, since the study population was restricted to residents in fields that participate in primary care, our findings may not be generalizable to patterns of PCP use in other specialties [6].

These limitations aside, our study has important strengths. This is the first national study of its kind with specific questions addressing primary care access and utilization, prescription medication use and related practices, and the prevalence of medical conditions among trainees. Important differences in the rates of establishing primary care between male and female respondents, first- year and senior residents, and those with and without chronic disease suggest a need to target specific resident groups (males, interns, those without pre-existing conditions) for wellness-related interventions. Such interventions could include distribution of a list of local providers to first year residents, advanced protected time for doctor’s appointments, and safeguards to ensure health information is protected from potential supervisors. Future studies should also include residents from non-primary care oriented specialties such as surgery, emergency medicine, and anesthesiology to obtain results that are more generalizable to the resident population as a whole. Additionally, the rates of inappropriate prescriptions were not insignificant and warrant further evaluation of the driving forces behind these behaviors.

Conclusion

Medical residents have a substantial burden of chronic illness that may not be met through interactions with PCPs. More research into barriers that residents face while accessing care and an assessment of interventions to facilitate their access to care is important to promote trainee well-being. Without such direction and initiative, it may prove harder for physicians to heal themselves or those for whom they provide care.

Acknowledgments: We thank Suzanne Winter, the study coordinator, for her support with editing and formatting the manuscript, Latoya Kuhn for performing the statistical analysis and creating data tables, and Dr. Namita Sachdev and Dr. Renuka Tipirneni for providing feedback on the survey instrument. We also thank the involved programs for their participation.

Corresponding author: Vineet Chopra, NCRC 2800 Plymouth Rd., Bldg 16, 432, Ann Arbor, MI 48109, [email protected].

Financial disclosures: None.

Previous presentations: Results were presented at the Annual Michigan Medicine 2017 Internal Medicine Research Symposium.

From the University of Michigan Medical School, Ann Arbor, MI.

Abstract

• Objective: Existing research has demonstrated overall low rates of residents establishing care with a primary care physician (PCP). We conducted a survey-based study to better understand chronic illness, PCP utilization, and prescription medication use patterns in resident physician populations.
• Methods: In 2017, we invited internal and family medicine trainees from a convenience sample of U.S. residency programs to participate in a survey. We compared the characteristics of residents who had established care with a PCP to those who had not.
• Results: The response rate was 45% (348/766 residents). The majority (n = 205, 59%) of respondents stated they had established care with a PCP primarily for routine preventative care (n = 159, 79%) and access in the event of an emergency (n = 132, 66%). However, 31% (n = 103) denied having had a wellness visit in over 3 years. Nearly a quarter of residents (n = 77, 23%) reported a chronic medical illness and 14% (n = 45) reported a preexisting mental health condition prior to residency. One-third (n = 111, 33%) reported taking a long-term prescription medication. Compared to residents who had not established care, those with a PCP (n = 205) more often reported a chronic condition (P < 0.001), seeing a subspecialist (P = 0.01), or taking long-term prescription medications (P < 0.001). One in 5 (n = 62,19%) respondents reported receiving prescriptions for an acute illness from an individual with whom they did not have a doctor-patient relationship.
• Conclusion: Medical residents have a substantial burden of chronic illness that may not be met through interactions with PCPs. Further understanding their medical needs and barriers to accessing care is necessary to ensure trainee well-being.

Keywords: Medical education-graduate, physician behavior, survey research, access to care.

Although internal medicine (IM) and family medicine (FM) residents must learn to provide high-quality primary care to their patients, little is known about whether they appropriately access such care themselves. Resident burnout and resilience has received attention [1,2], but there has been limited focus on understanding the burden of chronic medical and mental illness among residents. In particular, little is known about whether residents access primary care physicians (PCPs)—for either acute or chronic medical needs—and about resident self-medication practices.

Residency is often characterized by a life-changing geographic relocation. Even residents who do not relocate may still need to establish care with a new PCP due to health insurance or loss of access to a student clinic [3]. Establishing primary care with a new doctor typically requires scheduling a new patient visit, often with a wait time of several days to weeks [4,5]. Furthermore, lack of time, erratic schedules, and concerns about privacy and the stigma of being ill as a physician are barriers to establishing care [6-8]. Individuals who have not established primary care may experience delays in routine preventative health services, screening for chronic medical and mental health conditions, as well as access to care during acute illnesses [9,10]. Worse, they may engage in potentially unsafe practices, such as having colleagues write prescriptions for them, or even self-prescribing [8,11,12].

Existing research has demonstrated overall low rates of residents establishing care with a PCP [6–8,13]. However, these studies have either been limited to large academic centers or conducted outside the United States. Improving resident well-being may prove challenging without a clear understanding of current primary care utilization practices, the burden of chronic illness among residents, and patterns of prescription medication use and needs. Therefore, we conducted a survey-based study to understand primary care utilization and the burden of chronic illness among residents. We also assessed whether lack of primary care is associated with potentially risky behaviors, such as self-prescribing of medications.

Methods

Study Setting and Participants

The survey was distributed to current residents at IM and FM programs within the United States in 2017. Individual programs were recruited by directly contacting program directors or chief medical residents via email. Rather than contacting sites directly through standard templated emails, we identified programs both through personal contacts as well as the Electronic Residency Application Service list of accredited IM training programs. We elected to use this approach in order to increase response rates and to ensure that a sample representative of the trainee population was constructed. Programs were located in the Northeast, Midwest, South, and Pacific regions, and included small community-based programs and large academic centers.

Development of the Survey

The survey instrument was developed by the authors and reviewed by residents and PCPs at the University of Michigan to ensure relevance and comprehension of questions (The survey is available in the Appendix.). Once finalized, the survey was programmed into an online survey tool (Qualtrics, Provo, UT) and pilot-tested before being disseminated to the sampling frame. Data collected in the survey included: respondent’s utilization of a PCP, burden of chronic illness, long-term prescription medications, prescribing source, and demographic characteristics.

Each participating program distributed the survey to their residents through an email containing an anonymous hyperlink. The survey was available for completion for 4 weeks. We asked participating programs to send email reminders to encourage participation. Participants were given the option of receiving a $10 Amazon gift card after completion. All responses were recorded anonymously. The study received a “not regulated” status by the University of Michigan Institutional Review Board (HUM 00123888).

Statistical Analysis

Descriptive statistics were used to tabulate results. Respondents were encouraged, but not required, to answer all questions. Therefore, the response rate for each question was calculated using the total number of responses for that question as the denominator. Bivariable comparisons were made using Chi-squared or Fisher’s exact tests, as appropriate, for categorical data. A P value < 0.05, with 2-sided alpha, was considered statistically significant. All statistical analyses were conducted using Stata 13 SE (StataCorp, College Station, TX).

Results

Respondent Characteristics

Of the 29 programs contacted, 10 agreed to participate within the study timeframe. Of 766 potential respondents, 348 (45%) residents answered the survey (Table 1). The majority of respondents (n = 276, 82%) were from IM programs. Respondents were from all training years as follows: postgraduate year 1 residents (PGY-1, or interns; n = 130, 39%), PGY-2 residents (n = 98, 29%), PGY-3 residents (n = 93, 28%), and PGY-4 residents (n = 12, 4%). Most respondents were from the South (n = 130, 39%) and Midwest (n = 123, 37%) regions, and over half (n = 179, 54%) were female. Most respondents (n = 285, 86%) stated that they did not have children. The majority (n = 236, 71%) were completing residency in an area where they had not previously lived for more than 1 year.

Primary Care Utilization

Among the 348 respondents, 59% (n = 205) reported having established care with a PCP. An additional 6% (n = 21) had established care with an obstetrician/gynecologist for routine needs (Table 2). The 2 most common reasons for establishing care with a PCP were routine primary care needs, including contraception (n = 159, 79%), and access to a physician in the event of an acute medical need (n = 132, 66%).

Among respondents who had established care with a PCP, most (n = 188, 94%) had completed at least 1 appointment. However, among these 188 respondents, 68% (n = 127) stated that they had not made an acute visit in more than 12 months. When asked about wellness visits, almost one third of respondents (n = 103, 31%) stated that they had not been seen for a wellness visit in the past 3 years.

Burden of Chronic Illness

Most respondents (n = 223, 67%) stated that they did not have a chronic medical or mental health condition prior to residency (Table 3). However, 23% (n = 77) of respondents stated that they had been diagnosed with a chronic medical illness prior to residency, and 14% (n = 45) indicated they had been diagnosed with a mental health condition prior to residency. Almost one fifth of respondents (n = 60, 18%) reported seeing a subspecialist for a medical illness, and 33% (n = 111) reported taking a long-term prescription medication. With respect to major medical issues, the majority of residents (n = 239, 72%) denied experiencing events such as pregnancy, hospitalization, surgery, or an emergency department (ED) visit during training.

[polldaddy:10116940]

Inappropriate Prescriptions

While the majority of respondents denied writing a prescription for themselves for an acute or chronic medical condition, almost one fifth (n = 62, 19%) had received a prescription for an acute medical need from a provider outside of a clinical relationship (ie, from someone other than their PCP or specialty provider). Notably, 5% (n = 15) reported that this had occurred at least 2 or 3 times in the past 12 months (Table 4). Compared to respondents not taking long-term prescription medications, respondents who were already taking long-term prescription medications more frequently reported inappropriately receiving chronic prescriptions outside of an established clinical relationship (n = 14, 13% vs. n = 14, 6%; P = 0.05) and more often self-prescribed medications for acute needs (n = 12, 11% vs. n = 7, 3%; P = 0.005).

Important differences were noted between residents who had a PCP versus those who did not (Table 5). For example, a higher percentage of residents with a PCP indicated they had been diagnosed with a chronic medical illness (n = 55, 28% vs. n = 22, 16%; P = 0.01) or a chronic mental health condition (n = 34, 17% vs. n = 11, 8%; P = 0.02) before residency. Additionally, a higher percentage of residents with a PCP (n = 70, 35% vs. n = 25, 18%; P = 0.001) reported experiencing medical events such as pregnancy, hospitalization, surgery, ED visit, or new diagnosis of a chronic medical illness during residency. Finally, a higher percentage of respondents with a PCP stated that they had visited a subspecialist for a medical illness (n = 44, 22% vs. n = 16,12%; P = 0.01) or were taking long-term prescription medications (n = 86, 43% vs. n = 25; 18%; P < 0.001). When comparing PGY-1 to PGY-2–PGY-4 residents, the former reported having established a medical relationship with a PCP significantly less frequently (n = 56, 43% vs. n = 142, 70%; P < 0.001).

Discussion

This survey-based study of medical residents across the United States suggests that a substantial proportion do not establish relationships with PCPs. Additionally, our data suggest that despite establishing care, few residents subsequently visited their PCP during training for wellness visits or routine care. Self-reported rates of chronic medical and mental health conditions were substantial in our sample. Furthermore, inappropriate self-prescription and the receipt of prescriptions outside of a medical relationship were also reported. These findings suggest that future studies that focus on the unique medical and mental health needs of physicians in training, as well as interventions to encourage care in this vulnerable period, are necessary.

We observed that most respondents that established primary care were female trainees. Although it is impossible to know with certainty, one hypothesis behind this discrepancy is that women routinely need to access preventative care for gynecologic needs such as pap smears, contraception, and potentially pregnancy and preconception counseling [14,15]. Similarly, residents with a chronic medical or mental health condition prior to residency established care with a local PCP at a significantly greater frequency than those without such diagnoses. While selection bias cannot be excluded, this finding suggests that illness is a driving factor in establishing care. There also appears to be an association between accessing the medical system (either for prescription medications or subspecialist care) and having established care with a PCP. Collectively, these data suggest that individuals without a compelling reason to access medical services might have barriers to accessing care in the event of medical needs or may not receive routine preventative care [9,10].

In general, we found that rates of reported inappropriate prescriptions were lower than those reported in prior studies where a comparable resident population was surveyed [8,12,16]. Inclusion of multiple institutions, differences in temporality, social desirability bias, and reporting bias might have influenced our findings in this regard. Surprisingly, we found that having a PCP did not influence likelihood of inappropriate prescription receipt, perhaps suggesting that this behavior reflects some degree of universal difficulty in accessing care. Alternatively, this finding might relate to a cultural tendency to self-prescribe among resident physicians. The fact that individuals on chronic medications more often both received and wrote inappropriate prescriptions suggests this problem might be more pronounced in individuals who take medications more often, as these residents have specific needs [12]. Future studies targeting these individuals thus appear warranted.

Our study has several limitations. First, our sample size was modest and the response rate of 45% was low. However, to our knowledge, this remains among the largest survey on this topic, and our response rate is comparable to similar trainee studies [8,11,13]. Second, we designed and created a novel survey for this study. While the questions were pilot-tested with users prior to dissemination, validation of the instrument was not performed. Third, since the study population was restricted to residents in fields that participate in primary care, our findings may not be generalizable to patterns of PCP use in other specialties [6].

These limitations aside, our study has important strengths. This is the first national study of its kind with specific questions addressing primary care access and utilization, prescription medication use and related practices, and the prevalence of medical conditions among trainees. Important differences in the rates of establishing primary care between male and female respondents, first- year and senior residents, and those with and without chronic disease suggest a need to target specific resident groups (males, interns, those without pre-existing conditions) for wellness-related interventions. Such interventions could include distribution of a list of local providers to first year residents, advanced protected time for doctor’s appointments, and safeguards to ensure health information is protected from potential supervisors. Future studies should also include residents from non-primary care oriented specialties such as surgery, emergency medicine, and anesthesiology to obtain results that are more generalizable to the resident population as a whole. Additionally, the rates of inappropriate prescriptions were not insignificant and warrant further evaluation of the driving forces behind these behaviors.

Conclusion

Medical residents have a substantial burden of chronic illness that may not be met through interactions with PCPs. More research into barriers that residents face while accessing care and an assessment of interventions to facilitate their access to care is important to promote trainee well-being. Without such direction and initiative, it may prove harder for physicians to heal themselves or those for whom they provide care.

Acknowledgments: We thank Suzanne Winter, the study coordinator, for her support with editing and formatting the manuscript, Latoya Kuhn for performing the statistical analysis and creating data tables, and Dr. Namita Sachdev and Dr. Renuka Tipirneni for providing feedback on the survey instrument. We also thank the involved programs for their participation.

Corresponding author: Vineet Chopra, NCRC 2800 Plymouth Rd., Bldg 16, 432, Ann Arbor, MI 48109, [email protected].

Financial disclosures: None.

Previous presentations: Results were presented at the Annual Michigan Medicine 2017 Internal Medicine Research Symposium.

ABSTRACT

Cervical myelopathy is the most common cause of acquired spinal cord dysfunction in people aged >55 years. Advanced age and duration of symptoms have been implicated in the literature as negative prognostic indicators for postoperative functional improvement, but very few studies have evaluated the interaction of these factors. We retrospectively reviewed 125 patients who underwent surgery for cervical myelopathy. Patients were stratified according to age greater or less than 65 years and duration of symptoms of greater or less than 12 and 24 months. Functional outcomes were assessed using the Nurick score. Simple regression and multiple regression analyses were done, controlling for sex, preoperative Nurick score, surgical approach, smoking status, diabetes status, prior surgery, number of levels fused, ethanol use, and signal change on preoperative magnetic resonance imaging. The average change in Nurick score in all patients was 1.36, with a significant difference between patients with symptoms for <24 months and those with symptoms for >24 months (1.54 vs 0.98, P = .03). Multiple regression analysis revealed that older patients had a significant difference at 24 months (1.69 vs 1.25, P = .01), whereas younger patients showed slightly lower improvement overall and a change in Nurick score at both thresholds that was statistically nonsignificant.

Continue to: Cervical spondylotic myelopathy...

Cervical spondylotic myelopathy (CSM) is the most common acquired cause of spinal cord dysfunction in people aged >55 years.¹ It is a slowly progressive disorder usually caused by spinal cord compression and ischemia due to age-related changes in the spine and is characterized by neck pain, radicular arm pain, paresthesia, weakness, lower extremity hyperreflexia, and gait and balance abnormalities and may also present with bowel and bladder dysfunction. The majority of cases progress in a stepwise manner, but about 5% of cases decline rapidly, and the prognosis of nonoperative treatment is poor once the patient is truly myelopathic. The objective of surgery is to decompress the spinal cord before permanent damage has set in.^2-4

Several studies have attempted to describe the prognostic significance of duration of symptoms in surgical decompression of CSM. Some studies have found that there is no association with outcomes,^5-7 but most of the studies have concluded that there is an association. Several of these studies specify that duration of symptoms is significant beyond particular time points, typically of 12 months^8-12 or 24 months.^13,14 At least 2 review studies have found low evidence for the influence of symptom duration on postoperative outcomes.^15,16

Age has also been cited as an important prognostic factor in surgical decompression of CSM by some of these same studies. Only a few studies have concluded that age itself does not affect outcomes.^17-19 However, most of the studies conclude that advanced age is a significant factor. Most of these cite a cutoff of 60 years of age,^14,20 65 years of age,²¹ or 70 years of age,¹⁰ but at least 1 study has cited a cutoff as young as 40 years of age,⁹ and at least 1 other has cited 50 years of age.⁸

Most of the available literature has evaluated the effects of age and duration of symptoms separately. However, at least 2 studies have discussed the interplay between these variables, and both found that outcomes are associated with duration of symptoms only in the elderly, defined as above either 65 or 70 years of age.^5,19 This study is an attempt to clarify this relationship.

Continue to: MATERIALS AND METHODS...

MATERIALS AND METHODS

Institutional Review Board approval was obtained for this study. Informed consent was waived due to the retrospective nature of the work. The medical records of 212 patients who underwent surgery for CSM by the senior author were reviewed. All surgeries were performed at the University Hospital or the Veterans Administration (VA) between March 2005 and July 2012. CSM was diagnosed by magnetic resonance imaging (MRI) and based on the presence of upper motor signs, clonus, gait abnormalities, or difficulty with fine motor movements such as buttoning a shirt. Nurick score (Table 1) was assessed at presentation and at follow-up, and was the only outcome measure recorded in this cohort. Inclusion criteria were the diagnosis of CSM with a Nurick score, surgical intervention, and at least 2 years of follow-up. Age at presentation, sex, preoperative Nurick score, postoperative Nurick score, duration of symptoms preoperatively, duration of follow-up, procedure performed, approach (anterior vs posterior vs anterior and posterior), prior surgery, number of levels fused, diabetes status, cocaine use, ethanol use, tobacco use, signal change on preoperative MRI, and whether the patient belonged to the VA were recorded. Posterior cervical surgery was performed in patients who had ossification of the posterior longitudinal ligament, had multiple prior anterior cervical procedures, or had involvement of 3 or more levels with anatomy that would make an extensive exposure difficult. Surgeries were performed anteriorly for cases of 1- or 2-level stenosis in the absence of ossification of the posterior longitudinal ligament.

Anterior surgery was also considered in patients with 3-level disease who did not have anatomy that precluded a more extensive exposure.

Patients were stratified according to duration of symptoms by cutoffs of 12 or 24 months and according to age <65 years or >65 years. The age cutoff was chosen because this was the youngest cohort in which stratification revealed a significant difference in change in the Nurick score according to duration of symptoms, and because this age is consistent with the literature. Data were blinded, and outcomes according to duration of symptoms and age were analyzed. The analysis was conducted using simple linear regression and multiple regression.

SURGICAL TECHNIQUE

Patients were evaluated through a complete neurological examination and Nurick scores preoperatively and postoperatively at 6 weeks, 3 months, 6 months, 1 year, and annually thereafter. Decompression procedures performed included single or multilevel corpectomy, anterior decompression with strut grafting and instrumentation, posterior cervical laminoplasty, and posterior cervical laminectomy and fusion. Patients were placed in a Miami J collar (Össur) postoperatively and sent to physical and occupational therapy when able. All procedures were performed by the senior author with the assistance of residents and fellows.

RESULTS

Of the 125 patients who met all the inclusion criteria, 44 were females and 81 were males. The average follow-up duration was 30.9 months (standard deviation [SD], 13.23). The average age of all patients was 55.2 years (range, 27-89 years), and there was no difference in age according to gender (55.0 years for females vs 55.2 years for males). The average preoperative Nurick score was 2.61 (SD, 1.16), and there was no difference in preoperative Nurick score according to cutoff of duration of symptoms. Males had a higher preoperative Nurick score than females (2.73 vs 2.41, P = .12) and a longer but statistically nonsignificant duration of symptoms (25.7 vs 16.9 months, P = .1). There were 97 patients aged ≤65 years (average, 49.6 years) and 28 patients aged >65 years (average, 73.7 years). The younger cohort had a lower preoperative Nurick score than the older cohort, but this difference was not statistically significant (2.52 vs 3.0, P = .06). The younger cohort also had a longer duration of symptoms, but this difference was not significant (21.8 vs 26.2 months, P > .1). The initial analysis of the change in Nurick score in all patients according to duration of symptoms revealed an average change of 1.36 points (SD, 1.13) and a difference in postoperative change in Nurick score for the duration of symptom cutoffs of 12 and 24 months. This pattern was also present when patients were stratified according to age (Tables 2 and 3). The most common procedures performed were anterior cervical discectomy and fusion (ACDF) (58) and corpectomy (49). Data according to the procedure performed are recorded in Table 4. No significant complications were recorded. Simple regression and multiple regression analyses were undertaken to further evaluate these relationships.

Table 1. Nurick Score

Table 2. Change in Nurick According to Threshold of Duration of Symptoms

Abbreviation: SD, standard deviation.

Table 3. Change in Nurick According to Threshold of Duration of Symptoms, by Age

Abbreviation: SD, standard deviation.

Abbreviations: ACDF, anterior cervical discectomy and fusion; SD, standard deviation.

Continue to: Simple regression analysis of data...

Simple regression analysis of data of all patients revealed a statistically significant negative relationship between duration of symptoms and postoperative change in Nurick score (P = .044). There was no relationship between duration of symptoms and preoperative Nurick score (P = .482). When stratified according to duration of preoperative symptoms by 12 or 24 months, the relationship between duration of symptoms and change in Nurick score was statistically significant for cutoffs of 12 months (P = .03) and 24 months (P = .007). There was no relationship between duration of symptoms and preoperative Nurick score for any threshold of preoperative symptom duration. When these results were stratified according to age, patients aged ≤65 years showed a statistically significant association between duration of preoperative symptoms and change in Nurick score for cutoffs of 12 months (P = .016) and 24 months (P = .019). However, patients aged >65 years did not show a statistically significant association for cutoffs of 12 or 24 months (P = .85 and .29, respectively). There was also no relationship between duration of symptoms and preoperative Nurick score for any threshold of preoperative symptom duration in either age cohort.

Multiple regression analysis of the previously described findings was undertaken to assess the influence of potential confounding variables. These included age, gender, diabetes, cocaine use, alcohol use, tobacco use, signal change on preoperative MRI, severity of myelopathy, total levels fused, prior surgery, surgical approach (anterior vs posterior), and procedure performed (Table 4). Analysis of the relationship between duration of symptoms and change in Nurick score for all patients initially revealed a statistically nonsignificant correlation (P = .22). Significant factors in this model included diabetes status and tobacco use that correlated with decreasing change in Nurick score (P = .02 and .0001, respectively) and severity of myelopathy that correlated with increasing change in Nurick score (P = .0002). Notably, combined procedures also correlated with decreasing change in Nurick score (P = .03), but the performance of individual procedures did not correlate with change in Nurick score. There was no association between duration of symptoms and preoperative Nurick score (P = .76). When stratified according to duration of symptoms of 12 or 24 months, only 24 months was found to be statistically significant (P = .03). There was no relationship between duration of symptoms and preoperative Nurick score for any threshold of symptom duration. When further stratified according to age, the younger cohort did not show a statistically significant association between duration of preoperative symptoms and change in Nurick score for either threshold of symptom duration (P = .15 and .43, respectively). Diabetes status, tobacco use, number of levels fused, severity of myelopathy, and combined procedures remained significant predictors of change in Nurick score for both thresholds of symptom duration. In contrast, the older cohort showed a statistically significant association between duration of symptoms and postoperative change in Nurick score only for a threshold of 24 months (P = .01). In contrast to the younger cohort, the only other significant predictors in this group were preoperative severity of myelopathy, anterior approach (all ACDF procedures), and signal change on preoperative MRI (P = .02, .04, and .03, respectively). There was no relationship between duration of symptoms and preoperative Nurick score for any threshold of preoperative symptom duration in either age cohort.

DISCUSSION

Several studies have attempted to describe the prognostic influence of preoperative symptom duration on surgical outcomes for CSM. Few studies suggest that duration of symptoms does not correlate with functional outcomes. For example, Naderi and colleagues⁶ concluded in a retrospective study of 27 patients that there is no correlation as assessed by the modified Japanese Orthopedic Association scale. Handa and colleagues⁵ similarly concluded in a retrospective study of 61 patients that duration of symptoms was not significant, but only in patients aged <70 years. Furlan and colleagues⁷ conducted a prospective study of 81 patients with a mean follow-up of 10 months and concluded that there is no association as assessed using the modified Japanese Orthopedic Association (mJOA) and Nurick score. In contrast, the majority of studies support the notion that duration of symptoms adversely affects outcomes. Several of these studies do not provide a clear cutoff beyond which outcomes are significantly affected.^17-19,22

Of the studies that provide a cutoff, a fair number of studies suggest a limit of 12 months and a few suggest 24 months. In a retrospective study of 109 patients with cervical radiculopathy and 55 with cervical myelopathy, Bertalanffy and Eggert⁸ found that duration of symptoms beyond 12 months significantly correlated with worse outcomes as assessed by the evaluation criteria set forth by Roosen and Grote.²³ Using the more common European Myelopathy Score, Heidecke and colleagues⁹ arrived at the same conclusion from a retrospective review of 106 patients. In a large retrospective review of 248 patients, Pumberger and colleagues¹¹ found that patients who did not improve following surgical decompression for CSM, where improvement was defined as a reduction of at least 1 Nurick grade, had an average of 17.85 months of preoperative symptoms, whereas those who did improve had symptoms for an average of 11.21 months. In a prospective study of 98 patients, Suzuki and colleagues¹⁰ found that recovery rate of the JOA scale was significantly decreased in those with >1 year of preoperative symptoms. Both Chagas and colleagues¹⁴ and Suri and colleagues¹³ conducted prospective studies that revealed a significant difference in Nurick score improvement in patients with >2 years of symptoms. In reviews of the literature, both Holly and colleagues¹⁵ and Yoon and colleagues¹⁶ found a low level of evidence for the significance of symptom duration on outcomes. Similarly, Tetreault and colleagues²⁴ found that duration of symptoms was predictive of outcomes as assessed by both mJOA and Nurick score.

Continue to: Our results in all patients showed...

Our results in all patients showed a clear difference in outcomes at the 12-month cutoff as revealed by the simple regression and a trend that reached significance at the 24-month cutoff as assessed by the multiple regression. These results are consistent with those discussed, especially those that specifically used the Nurick score. We further showed that the influence of duration of symptoms on outcomes is dependent on age. Our simple regression analysis suggested that this dependence was evident for symptom durations of 12 and 24 months only in the younger cohort. However, our multiple regression analysis showed that the effect of symptom duration on outcomes is evident only in patients aged >65 years who have had symptoms for 24 months. The stark difference in results between the simple and multiple regressions is probably due to the several potentially confounding variables that were controlled for in the multiple regression analysis. Of course, it should be noted that a statistically nonsignificant difference does not necessarily translate into a clinically nonsignificant difference.  

Our results are consistent with the few studies that describe the influence of the interplay between age and duration of symptoms on postoperative outcomes in CSM. For example, Handa and colleagues⁵ retrospectively reviewed 61 patients who underwent expansive laminoplasty for CSM and stratified them according to age greater or less than 70 years. Compared with the younger patients, duration of symptoms in the 22 elderly patients correlated with a significant difference in outcomes as assessed by the mJOA, with a cutoff of 1 year.⁵ Similarly, Yamazaki and colleagues¹⁹ evaluated 64 patients who also underwent expansive laminoplasty for CSM and stratified them according to age greater or less than 65 years. Duration of symptoms in 35 elderly patients significantly correlated with outcomes as assessed by the JOA scale, such that those considered to have an excellent outcome had a mean duration of symptoms of 11.1 months compared to the 39 months of symptoms in those considered to have a fair outcome.¹⁹ In contrast to those studies, we found that 24 months rather than 12 months was significant. However, we also evaluated outcomes using the Nurick score rather than the JOA. The JOA is a more detailed instrument, and this may be the reason for the discrepancy. Nonetheless, our results are consistent with the extant literature and add to the limited number of studies that have commented on the combined interactions of symptom duration and age in postoperative outcomes for CSM.

There are several strengths and limitations to this study. One strength is the relatively large sample size of patients. However, there was an uneven distribution in the number of patients in each age cohort. Ideally, there would have been an equal number of patients in each age group. The fact that all patients were operated on by the same surgeon minimizes variability in outcomes due to surgeon skill. We also controlled for multiple variables that are known to affect CSM outcomes, but we did not have quantitative data with respect to degree of compression or cross-sectional area of the affected spinal cord, which have been described as significant variables in outcomes of CSM. Furthermore, we did not evaluate the results using several outcome measures such as the JOA in addition to the Nurick score, and this limits the comparability of our work to some of the existing literature. This study also suffers from the inherent biases and shortcomings of retrospective studies, and the fact that this was not a multicenter study may limit generalizability of the results. However, given the dearth of literature on this topic, our work adds to the literature. Further studies will be needed to more clearly elucidate this topic.

CONCLUSION

This study demonstrated that duration of symptoms may be a significant factor in the recovery of patients undergoing surgical decompression for CSM, but only in patients aged >65 years who have had symptoms for 24 months.

This paper will be judged for the Resident Writer’s Award.

ABSTRACT

Cervical myelopathy is the most common cause of acquired spinal cord dysfunction in people aged >55 years. Advanced age and duration of symptoms have been implicated in the literature as negative prognostic indicators for postoperative functional improvement, but very few studies have evaluated the interaction of these factors. We retrospectively reviewed 125 patients who underwent surgery for cervical myelopathy. Patients were stratified according to age greater or less than 65 years and duration of symptoms of greater or less than 12 and 24 months. Functional outcomes were assessed using the Nurick score. Simple regression and multiple regression analyses were done, controlling for sex, preoperative Nurick score, surgical approach, smoking status, diabetes status, prior surgery, number of levels fused, ethanol use, and signal change on preoperative magnetic resonance imaging. The average change in Nurick score in all patients was 1.36, with a significant difference between patients with symptoms for <24 months and those with symptoms for >24 months (1.54 vs 0.98, P = .03). Multiple regression analysis revealed that older patients had a significant difference at 24 months (1.69 vs 1.25, P = .01), whereas younger patients showed slightly lower improvement overall and a change in Nurick score at both thresholds that was statistically nonsignificant.

Continue to: Cervical spondylotic myelopathy...

Cervical spondylotic myelopathy (CSM) is the most common acquired cause of spinal cord dysfunction in people aged >55 years.¹ It is a slowly progressive disorder usually caused by spinal cord compression and ischemia due to age-related changes in the spine and is characterized by neck pain, radicular arm pain, paresthesia, weakness, lower extremity hyperreflexia, and gait and balance abnormalities and may also present with bowel and bladder dysfunction. The majority of cases progress in a stepwise manner, but about 5% of cases decline rapidly, and the prognosis of nonoperative treatment is poor once the patient is truly myelopathic. The objective of surgery is to decompress the spinal cord before permanent damage has set in.^2-4

Several studies have attempted to describe the prognostic significance of duration of symptoms in surgical decompression of CSM. Some studies have found that there is no association with outcomes,^5-7 but most of the studies have concluded that there is an association. Several of these studies specify that duration of symptoms is significant beyond particular time points, typically of 12 months^8-12 or 24 months.^13,14 At least 2 review studies have found low evidence for the influence of symptom duration on postoperative outcomes.^15,16

Age has also been cited as an important prognostic factor in surgical decompression of CSM by some of these same studies. Only a few studies have concluded that age itself does not affect outcomes.^17-19 However, most of the studies conclude that advanced age is a significant factor. Most of these cite a cutoff of 60 years of age,^14,20 65 years of age,²¹ or 70 years of age,¹⁰ but at least 1 study has cited a cutoff as young as 40 years of age,⁹ and at least 1 other has cited 50 years of age.⁸

Most of the available literature has evaluated the effects of age and duration of symptoms separately. However, at least 2 studies have discussed the interplay between these variables, and both found that outcomes are associated with duration of symptoms only in the elderly, defined as above either 65 or 70 years of age.^5,19 This study is an attempt to clarify this relationship.

Continue to: MATERIALS AND METHODS...

MATERIALS AND METHODS

Institutional Review Board approval was obtained for this study. Informed consent was waived due to the retrospective nature of the work. The medical records of 212 patients who underwent surgery for CSM by the senior author were reviewed. All surgeries were performed at the University Hospital or the Veterans Administration (VA) between March 2005 and July 2012. CSM was diagnosed by magnetic resonance imaging (MRI) and based on the presence of upper motor signs, clonus, gait abnormalities, or difficulty with fine motor movements such as buttoning a shirt. Nurick score (Table 1) was assessed at presentation and at follow-up, and was the only outcome measure recorded in this cohort. Inclusion criteria were the diagnosis of CSM with a Nurick score, surgical intervention, and at least 2 years of follow-up. Age at presentation, sex, preoperative Nurick score, postoperative Nurick score, duration of symptoms preoperatively, duration of follow-up, procedure performed, approach (anterior vs posterior vs anterior and posterior), prior surgery, number of levels fused, diabetes status, cocaine use, ethanol use, tobacco use, signal change on preoperative MRI, and whether the patient belonged to the VA were recorded. Posterior cervical surgery was performed in patients who had ossification of the posterior longitudinal ligament, had multiple prior anterior cervical procedures, or had involvement of 3 or more levels with anatomy that would make an extensive exposure difficult. Surgeries were performed anteriorly for cases of 1- or 2-level stenosis in the absence of ossification of the posterior longitudinal ligament.

Anterior surgery was also considered in patients with 3-level disease who did not have anatomy that precluded a more extensive exposure.

Patients were stratified according to duration of symptoms by cutoffs of 12 or 24 months and according to age <65 years or >65 years. The age cutoff was chosen because this was the youngest cohort in which stratification revealed a significant difference in change in the Nurick score according to duration of symptoms, and because this age is consistent with the literature. Data were blinded, and outcomes according to duration of symptoms and age were analyzed. The analysis was conducted using simple linear regression and multiple regression.

SURGICAL TECHNIQUE

Patients were evaluated through a complete neurological examination and Nurick scores preoperatively and postoperatively at 6 weeks, 3 months, 6 months, 1 year, and annually thereafter. Decompression procedures performed included single or multilevel corpectomy, anterior decompression with strut grafting and instrumentation, posterior cervical laminoplasty, and posterior cervical laminectomy and fusion. Patients were placed in a Miami J collar (Össur) postoperatively and sent to physical and occupational therapy when able. All procedures were performed by the senior author with the assistance of residents and fellows.

RESULTS

Of the 125 patients who met all the inclusion criteria, 44 were females and 81 were males. The average follow-up duration was 30.9 months (standard deviation [SD], 13.23). The average age of all patients was 55.2 years (range, 27-89 years), and there was no difference in age according to gender (55.0 years for females vs 55.2 years for males). The average preoperative Nurick score was 2.61 (SD, 1.16), and there was no difference in preoperative Nurick score according to cutoff of duration of symptoms. Males had a higher preoperative Nurick score than females (2.73 vs 2.41, P = .12) and a longer but statistically nonsignificant duration of symptoms (25.7 vs 16.9 months, P = .1). There were 97 patients aged ≤65 years (average, 49.6 years) and 28 patients aged >65 years (average, 73.7 years). The younger cohort had a lower preoperative Nurick score than the older cohort, but this difference was not statistically significant (2.52 vs 3.0, P = .06). The younger cohort also had a longer duration of symptoms, but this difference was not significant (21.8 vs 26.2 months, P > .1). The initial analysis of the change in Nurick score in all patients according to duration of symptoms revealed an average change of 1.36 points (SD, 1.13) and a difference in postoperative change in Nurick score for the duration of symptom cutoffs of 12 and 24 months. This pattern was also present when patients were stratified according to age (Tables 2 and 3). The most common procedures performed were anterior cervical discectomy and fusion (ACDF) (58) and corpectomy (49). Data according to the procedure performed are recorded in Table 4. No significant complications were recorded. Simple regression and multiple regression analyses were undertaken to further evaluate these relationships.

Table 1. Nurick Score

Table 2. Change in Nurick According to Threshold of Duration of Symptoms

Abbreviation: SD, standard deviation.

Table 3. Change in Nurick According to Threshold of Duration of Symptoms, by Age

Abbreviation: SD, standard deviation.

Abbreviations: ACDF, anterior cervical discectomy and fusion; SD, standard deviation.

Continue to: Simple regression analysis of data...

Simple regression analysis of data of all patients revealed a statistically significant negative relationship between duration of symptoms and postoperative change in Nurick score (P = .044). There was no relationship between duration of symptoms and preoperative Nurick score (P = .482). When stratified according to duration of preoperative symptoms by 12 or 24 months, the relationship between duration of symptoms and change in Nurick score was statistically significant for cutoffs of 12 months (P = .03) and 24 months (P = .007). There was no relationship between duration of symptoms and preoperative Nurick score for any threshold of preoperative symptom duration. When these results were stratified according to age, patients aged ≤65 years showed a statistically significant association between duration of preoperative symptoms and change in Nurick score for cutoffs of 12 months (P = .016) and 24 months (P = .019). However, patients aged >65 years did not show a statistically significant association for cutoffs of 12 or 24 months (P = .85 and .29, respectively). There was also no relationship between duration of symptoms and preoperative Nurick score for any threshold of preoperative symptom duration in either age cohort.

Multiple regression analysis of the previously described findings was undertaken to assess the influence of potential confounding variables. These included age, gender, diabetes, cocaine use, alcohol use, tobacco use, signal change on preoperative MRI, severity of myelopathy, total levels fused, prior surgery, surgical approach (anterior vs posterior), and procedure performed (Table 4). Analysis of the relationship between duration of symptoms and change in Nurick score for all patients initially revealed a statistically nonsignificant correlation (P = .22). Significant factors in this model included diabetes status and tobacco use that correlated with decreasing change in Nurick score (P = .02 and .0001, respectively) and severity of myelopathy that correlated with increasing change in Nurick score (P = .0002). Notably, combined procedures also correlated with decreasing change in Nurick score (P = .03), but the performance of individual procedures did not correlate with change in Nurick score. There was no association between duration of symptoms and preoperative Nurick score (P = .76). When stratified according to duration of symptoms of 12 or 24 months, only 24 months was found to be statistically significant (P = .03). There was no relationship between duration of symptoms and preoperative Nurick score for any threshold of symptom duration. When further stratified according to age, the younger cohort did not show a statistically significant association between duration of preoperative symptoms and change in Nurick score for either threshold of symptom duration (P = .15 and .43, respectively). Diabetes status, tobacco use, number of levels fused, severity of myelopathy, and combined procedures remained significant predictors of change in Nurick score for both thresholds of symptom duration. In contrast, the older cohort showed a statistically significant association between duration of symptoms and postoperative change in Nurick score only for a threshold of 24 months (P = .01). In contrast to the younger cohort, the only other significant predictors in this group were preoperative severity of myelopathy, anterior approach (all ACDF procedures), and signal change on preoperative MRI (P = .02, .04, and .03, respectively). There was no relationship between duration of symptoms and preoperative Nurick score for any threshold of preoperative symptom duration in either age cohort.

DISCUSSION

Several studies have attempted to describe the prognostic influence of preoperative symptom duration on surgical outcomes for CSM. Few studies suggest that duration of symptoms does not correlate with functional outcomes. For example, Naderi and colleagues⁶ concluded in a retrospective study of 27 patients that there is no correlation as assessed by the modified Japanese Orthopedic Association scale. Handa and colleagues⁵ similarly concluded in a retrospective study of 61 patients that duration of symptoms was not significant, but only in patients aged <70 years. Furlan and colleagues⁷ conducted a prospective study of 81 patients with a mean follow-up of 10 months and concluded that there is no association as assessed using the modified Japanese Orthopedic Association (mJOA) and Nurick score. In contrast, the majority of studies support the notion that duration of symptoms adversely affects outcomes. Several of these studies do not provide a clear cutoff beyond which outcomes are significantly affected.^17-19,22

Of the studies that provide a cutoff, a fair number of studies suggest a limit of 12 months and a few suggest 24 months. In a retrospective study of 109 patients with cervical radiculopathy and 55 with cervical myelopathy, Bertalanffy and Eggert⁸ found that duration of symptoms beyond 12 months significantly correlated with worse outcomes as assessed by the evaluation criteria set forth by Roosen and Grote.²³ Using the more common European Myelopathy Score, Heidecke and colleagues⁹ arrived at the same conclusion from a retrospective review of 106 patients. In a large retrospective review of 248 patients, Pumberger and colleagues¹¹ found that patients who did not improve following surgical decompression for CSM, where improvement was defined as a reduction of at least 1 Nurick grade, had an average of 17.85 months of preoperative symptoms, whereas those who did improve had symptoms for an average of 11.21 months. In a prospective study of 98 patients, Suzuki and colleagues¹⁰ found that recovery rate of the JOA scale was significantly decreased in those with >1 year of preoperative symptoms. Both Chagas and colleagues¹⁴ and Suri and colleagues¹³ conducted prospective studies that revealed a significant difference in Nurick score improvement in patients with >2 years of symptoms. In reviews of the literature, both Holly and colleagues¹⁵ and Yoon and colleagues¹⁶ found a low level of evidence for the significance of symptom duration on outcomes. Similarly, Tetreault and colleagues²⁴ found that duration of symptoms was predictive of outcomes as assessed by both mJOA and Nurick score.

Continue to: Our results in all patients showed...

Our results in all patients showed a clear difference in outcomes at the 12-month cutoff as revealed by the simple regression and a trend that reached significance at the 24-month cutoff as assessed by the multiple regression. These results are consistent with those discussed, especially those that specifically used the Nurick score. We further showed that the influence of duration of symptoms on outcomes is dependent on age. Our simple regression analysis suggested that this dependence was evident for symptom durations of 12 and 24 months only in the younger cohort. However, our multiple regression analysis showed that the effect of symptom duration on outcomes is evident only in patients aged >65 years who have had symptoms for 24 months. The stark difference in results between the simple and multiple regressions is probably due to the several potentially confounding variables that were controlled for in the multiple regression analysis. Of course, it should be noted that a statistically nonsignificant difference does not necessarily translate into a clinically nonsignificant difference.  

Our results are consistent with the few studies that describe the influence of the interplay between age and duration of symptoms on postoperative outcomes in CSM. For example, Handa and colleagues⁵ retrospectively reviewed 61 patients who underwent expansive laminoplasty for CSM and stratified them according to age greater or less than 70 years. Compared with the younger patients, duration of symptoms in the 22 elderly patients correlated with a significant difference in outcomes as assessed by the mJOA, with a cutoff of 1 year.⁵ Similarly, Yamazaki and colleagues¹⁹ evaluated 64 patients who also underwent expansive laminoplasty for CSM and stratified them according to age greater or less than 65 years. Duration of symptoms in 35 elderly patients significantly correlated with outcomes as assessed by the JOA scale, such that those considered to have an excellent outcome had a mean duration of symptoms of 11.1 months compared to the 39 months of symptoms in those considered to have a fair outcome.¹⁹ In contrast to those studies, we found that 24 months rather than 12 months was significant. However, we also evaluated outcomes using the Nurick score rather than the JOA. The JOA is a more detailed instrument, and this may be the reason for the discrepancy. Nonetheless, our results are consistent with the extant literature and add to the limited number of studies that have commented on the combined interactions of symptom duration and age in postoperative outcomes for CSM.

There are several strengths and limitations to this study. One strength is the relatively large sample size of patients. However, there was an uneven distribution in the number of patients in each age cohort. Ideally, there would have been an equal number of patients in each age group. The fact that all patients were operated on by the same surgeon minimizes variability in outcomes due to surgeon skill. We also controlled for multiple variables that are known to affect CSM outcomes, but we did not have quantitative data with respect to degree of compression or cross-sectional area of the affected spinal cord, which have been described as significant variables in outcomes of CSM. Furthermore, we did not evaluate the results using several outcome measures such as the JOA in addition to the Nurick score, and this limits the comparability of our work to some of the existing literature. This study also suffers from the inherent biases and shortcomings of retrospective studies, and the fact that this was not a multicenter study may limit generalizability of the results. However, given the dearth of literature on this topic, our work adds to the literature. Further studies will be needed to more clearly elucidate this topic.

CONCLUSION

This study demonstrated that duration of symptoms may be a significant factor in the recovery of patients undergoing surgical decompression for CSM, but only in patients aged >65 years who have had symptoms for 24 months.

This paper will be judged for the Resident Writer’s Award.

ABSTRACT

Cervical myelopathy is the most common cause of acquired spinal cord dysfunction in people aged >55 years. Advanced age and duration of symptoms have been implicated in the literature as negative prognostic indicators for postoperative functional improvement, but very few studies have evaluated the interaction of these factors. We retrospectively reviewed 125 patients who underwent surgery for cervical myelopathy. Patients were stratified according to age greater or less than 65 years and duration of symptoms of greater or less than 12 and 24 months. Functional outcomes were assessed using the Nurick score. Simple regression and multiple regression analyses were done, controlling for sex, preoperative Nurick score, surgical approach, smoking status, diabetes status, prior surgery, number of levels fused, ethanol use, and signal change on preoperative magnetic resonance imaging. The average change in Nurick score in all patients was 1.36, with a significant difference between patients with symptoms for <24 months and those with symptoms for >24 months (1.54 vs 0.98, P = .03). Multiple regression analysis revealed that older patients had a significant difference at 24 months (1.69 vs 1.25, P = .01), whereas younger patients showed slightly lower improvement overall and a change in Nurick score at both thresholds that was statistically nonsignificant.

Continue to: Cervical spondylotic myelopathy...

Cervical spondylotic myelopathy (CSM) is the most common acquired cause of spinal cord dysfunction in people aged >55 years.¹ It is a slowly progressive disorder usually caused by spinal cord compression and ischemia due to age-related changes in the spine and is characterized by neck pain, radicular arm pain, paresthesia, weakness, lower extremity hyperreflexia, and gait and balance abnormalities and may also present with bowel and bladder dysfunction. The majority of cases progress in a stepwise manner, but about 5% of cases decline rapidly, and the prognosis of nonoperative treatment is poor once the patient is truly myelopathic. The objective of surgery is to decompress the spinal cord before permanent damage has set in.^2-4

Several studies have attempted to describe the prognostic significance of duration of symptoms in surgical decompression of CSM. Some studies have found that there is no association with outcomes,^5-7 but most of the studies have concluded that there is an association. Several of these studies specify that duration of symptoms is significant beyond particular time points, typically of 12 months^8-12 or 24 months.^13,14 At least 2 review studies have found low evidence for the influence of symptom duration on postoperative outcomes.^15,16

Age has also been cited as an important prognostic factor in surgical decompression of CSM by some of these same studies. Only a few studies have concluded that age itself does not affect outcomes.^17-19 However, most of the studies conclude that advanced age is a significant factor. Most of these cite a cutoff of 60 years of age,^14,20 65 years of age,²¹ or 70 years of age,¹⁰ but at least 1 study has cited a cutoff as young as 40 years of age,⁹ and at least 1 other has cited 50 years of age.⁸

Most of the available literature has evaluated the effects of age and duration of symptoms separately. However, at least 2 studies have discussed the interplay between these variables, and both found that outcomes are associated with duration of symptoms only in the elderly, defined as above either 65 or 70 years of age.^5,19 This study is an attempt to clarify this relationship.

Continue to: MATERIALS AND METHODS...

MATERIALS AND METHODS

Institutional Review Board approval was obtained for this study. Informed consent was waived due to the retrospective nature of the work. The medical records of 212 patients who underwent surgery for CSM by the senior author were reviewed. All surgeries were performed at the University Hospital or the Veterans Administration (VA) between March 2005 and July 2012. CSM was diagnosed by magnetic resonance imaging (MRI) and based on the presence of upper motor signs, clonus, gait abnormalities, or difficulty with fine motor movements such as buttoning a shirt. Nurick score (Table 1) was assessed at presentation and at follow-up, and was the only outcome measure recorded in this cohort. Inclusion criteria were the diagnosis of CSM with a Nurick score, surgical intervention, and at least 2 years of follow-up. Age at presentation, sex, preoperative Nurick score, postoperative Nurick score, duration of symptoms preoperatively, duration of follow-up, procedure performed, approach (anterior vs posterior vs anterior and posterior), prior surgery, number of levels fused, diabetes status, cocaine use, ethanol use, tobacco use, signal change on preoperative MRI, and whether the patient belonged to the VA were recorded. Posterior cervical surgery was performed in patients who had ossification of the posterior longitudinal ligament, had multiple prior anterior cervical procedures, or had involvement of 3 or more levels with anatomy that would make an extensive exposure difficult. Surgeries were performed anteriorly for cases of 1- or 2-level stenosis in the absence of ossification of the posterior longitudinal ligament.

Anterior surgery was also considered in patients with 3-level disease who did not have anatomy that precluded a more extensive exposure.

Patients were stratified according to duration of symptoms by cutoffs of 12 or 24 months and according to age <65 years or >65 years. The age cutoff was chosen because this was the youngest cohort in which stratification revealed a significant difference in change in the Nurick score according to duration of symptoms, and because this age is consistent with the literature. Data were blinded, and outcomes according to duration of symptoms and age were analyzed. The analysis was conducted using simple linear regression and multiple regression.

SURGICAL TECHNIQUE

Patients were evaluated through a complete neurological examination and Nurick scores preoperatively and postoperatively at 6 weeks, 3 months, 6 months, 1 year, and annually thereafter. Decompression procedures performed included single or multilevel corpectomy, anterior decompression with strut grafting and instrumentation, posterior cervical laminoplasty, and posterior cervical laminectomy and fusion. Patients were placed in a Miami J collar (Össur) postoperatively and sent to physical and occupational therapy when able. All procedures were performed by the senior author with the assistance of residents and fellows.

RESULTS

Of the 125 patients who met all the inclusion criteria, 44 were females and 81 were males. The average follow-up duration was 30.9 months (standard deviation [SD], 13.23). The average age of all patients was 55.2 years (range, 27-89 years), and there was no difference in age according to gender (55.0 years for females vs 55.2 years for males). The average preoperative Nurick score was 2.61 (SD, 1.16), and there was no difference in preoperative Nurick score according to cutoff of duration of symptoms. Males had a higher preoperative Nurick score than females (2.73 vs 2.41, P = .12) and a longer but statistically nonsignificant duration of symptoms (25.7 vs 16.9 months, P = .1). There were 97 patients aged ≤65 years (average, 49.6 years) and 28 patients aged >65 years (average, 73.7 years). The younger cohort had a lower preoperative Nurick score than the older cohort, but this difference was not statistically significant (2.52 vs 3.0, P = .06). The younger cohort also had a longer duration of symptoms, but this difference was not significant (21.8 vs 26.2 months, P > .1). The initial analysis of the change in Nurick score in all patients according to duration of symptoms revealed an average change of 1.36 points (SD, 1.13) and a difference in postoperative change in Nurick score for the duration of symptom cutoffs of 12 and 24 months. This pattern was also present when patients were stratified according to age (Tables 2 and 3). The most common procedures performed were anterior cervical discectomy and fusion (ACDF) (58) and corpectomy (49). Data according to the procedure performed are recorded in Table 4. No significant complications were recorded. Simple regression and multiple regression analyses were undertaken to further evaluate these relationships.

Table 1. Nurick Score

Table 2. Change in Nurick According to Threshold of Duration of Symptoms

Abbreviation: SD, standard deviation.

Table 3. Change in Nurick According to Threshold of Duration of Symptoms, by Age

Abbreviation: SD, standard deviation.

Abbreviations: ACDF, anterior cervical discectomy and fusion; SD, standard deviation.

Continue to: Simple regression analysis of data...

Simple regression analysis of data of all patients revealed a statistically significant negative relationship between duration of symptoms and postoperative change in Nurick score (P = .044). There was no relationship between duration of symptoms and preoperative Nurick score (P = .482). When stratified according to duration of preoperative symptoms by 12 or 24 months, the relationship between duration of symptoms and change in Nurick score was statistically significant for cutoffs of 12 months (P = .03) and 24 months (P = .007). There was no relationship between duration of symptoms and preoperative Nurick score for any threshold of preoperative symptom duration. When these results were stratified according to age, patients aged ≤65 years showed a statistically significant association between duration of preoperative symptoms and change in Nurick score for cutoffs of 12 months (P = .016) and 24 months (P = .019). However, patients aged >65 years did not show a statistically significant association for cutoffs of 12 or 24 months (P = .85 and .29, respectively). There was also no relationship between duration of symptoms and preoperative Nurick score for any threshold of preoperative symptom duration in either age cohort.

Multiple regression analysis of the previously described findings was undertaken to assess the influence of potential confounding variables. These included age, gender, diabetes, cocaine use, alcohol use, tobacco use, signal change on preoperative MRI, severity of myelopathy, total levels fused, prior surgery, surgical approach (anterior vs posterior), and procedure performed (Table 4). Analysis of the relationship between duration of symptoms and change in Nurick score for all patients initially revealed a statistically nonsignificant correlation (P = .22). Significant factors in this model included diabetes status and tobacco use that correlated with decreasing change in Nurick score (P = .02 and .0001, respectively) and severity of myelopathy that correlated with increasing change in Nurick score (P = .0002). Notably, combined procedures also correlated with decreasing change in Nurick score (P = .03), but the performance of individual procedures did not correlate with change in Nurick score. There was no association between duration of symptoms and preoperative Nurick score (P = .76). When stratified according to duration of symptoms of 12 or 24 months, only 24 months was found to be statistically significant (P = .03). There was no relationship between duration of symptoms and preoperative Nurick score for any threshold of symptom duration. When further stratified according to age, the younger cohort did not show a statistically significant association between duration of preoperative symptoms and change in Nurick score for either threshold of symptom duration (P = .15 and .43, respectively). Diabetes status, tobacco use, number of levels fused, severity of myelopathy, and combined procedures remained significant predictors of change in Nurick score for both thresholds of symptom duration. In contrast, the older cohort showed a statistically significant association between duration of symptoms and postoperative change in Nurick score only for a threshold of 24 months (P = .01). In contrast to the younger cohort, the only other significant predictors in this group were preoperative severity of myelopathy, anterior approach (all ACDF procedures), and signal change on preoperative MRI (P = .02, .04, and .03, respectively). There was no relationship between duration of symptoms and preoperative Nurick score for any threshold of preoperative symptom duration in either age cohort.

DISCUSSION

Several studies have attempted to describe the prognostic influence of preoperative symptom duration on surgical outcomes for CSM. Few studies suggest that duration of symptoms does not correlate with functional outcomes. For example, Naderi and colleagues⁶ concluded in a retrospective study of 27 patients that there is no correlation as assessed by the modified Japanese Orthopedic Association scale. Handa and colleagues⁵ similarly concluded in a retrospective study of 61 patients that duration of symptoms was not significant, but only in patients aged <70 years. Furlan and colleagues⁷ conducted a prospective study of 81 patients with a mean follow-up of 10 months and concluded that there is no association as assessed using the modified Japanese Orthopedic Association (mJOA) and Nurick score. In contrast, the majority of studies support the notion that duration of symptoms adversely affects outcomes. Several of these studies do not provide a clear cutoff beyond which outcomes are significantly affected.^17-19,22

Of the studies that provide a cutoff, a fair number of studies suggest a limit of 12 months and a few suggest 24 months. In a retrospective study of 109 patients with cervical radiculopathy and 55 with cervical myelopathy, Bertalanffy and Eggert⁸ found that duration of symptoms beyond 12 months significantly correlated with worse outcomes as assessed by the evaluation criteria set forth by Roosen and Grote.²³ Using the more common European Myelopathy Score, Heidecke and colleagues⁹ arrived at the same conclusion from a retrospective review of 106 patients. In a large retrospective review of 248 patients, Pumberger and colleagues¹¹ found that patients who did not improve following surgical decompression for CSM, where improvement was defined as a reduction of at least 1 Nurick grade, had an average of 17.85 months of preoperative symptoms, whereas those who did improve had symptoms for an average of 11.21 months. In a prospective study of 98 patients, Suzuki and colleagues¹⁰ found that recovery rate of the JOA scale was significantly decreased in those with >1 year of preoperative symptoms. Both Chagas and colleagues¹⁴ and Suri and colleagues¹³ conducted prospective studies that revealed a significant difference in Nurick score improvement in patients with >2 years of symptoms. In reviews of the literature, both Holly and colleagues¹⁵ and Yoon and colleagues¹⁶ found a low level of evidence for the significance of symptom duration on outcomes. Similarly, Tetreault and colleagues²⁴ found that duration of symptoms was predictive of outcomes as assessed by both mJOA and Nurick score.

Continue to: Our results in all patients showed...

Our results in all patients showed a clear difference in outcomes at the 12-month cutoff as revealed by the simple regression and a trend that reached significance at the 24-month cutoff as assessed by the multiple regression. These results are consistent with those discussed, especially those that specifically used the Nurick score. We further showed that the influence of duration of symptoms on outcomes is dependent on age. Our simple regression analysis suggested that this dependence was evident for symptom durations of 12 and 24 months only in the younger cohort. However, our multiple regression analysis showed that the effect of symptom duration on outcomes is evident only in patients aged >65 years who have had symptoms for 24 months. The stark difference in results between the simple and multiple regressions is probably due to the several potentially confounding variables that were controlled for in the multiple regression analysis. Of course, it should be noted that a statistically nonsignificant difference does not necessarily translate into a clinically nonsignificant difference.  

Our results are consistent with the few studies that describe the influence of the interplay between age and duration of symptoms on postoperative outcomes in CSM. For example, Handa and colleagues⁵ retrospectively reviewed 61 patients who underwent expansive laminoplasty for CSM and stratified them according to age greater or less than 70 years. Compared with the younger patients, duration of symptoms in the 22 elderly patients correlated with a significant difference in outcomes as assessed by the mJOA, with a cutoff of 1 year.⁵ Similarly, Yamazaki and colleagues¹⁹ evaluated 64 patients who also underwent expansive laminoplasty for CSM and stratified them according to age greater or less than 65 years. Duration of symptoms in 35 elderly patients significantly correlated with outcomes as assessed by the JOA scale, such that those considered to have an excellent outcome had a mean duration of symptoms of 11.1 months compared to the 39 months of symptoms in those considered to have a fair outcome.¹⁹ In contrast to those studies, we found that 24 months rather than 12 months was significant. However, we also evaluated outcomes using the Nurick score rather than the JOA. The JOA is a more detailed instrument, and this may be the reason for the discrepancy. Nonetheless, our results are consistent with the extant literature and add to the limited number of studies that have commented on the combined interactions of symptom duration and age in postoperative outcomes for CSM.

There are several strengths and limitations to this study. One strength is the relatively large sample size of patients. However, there was an uneven distribution in the number of patients in each age cohort. Ideally, there would have been an equal number of patients in each age group. The fact that all patients were operated on by the same surgeon minimizes variability in outcomes due to surgeon skill. We also controlled for multiple variables that are known to affect CSM outcomes, but we did not have quantitative data with respect to degree of compression or cross-sectional area of the affected spinal cord, which have been described as significant variables in outcomes of CSM. Furthermore, we did not evaluate the results using several outcome measures such as the JOA in addition to the Nurick score, and this limits the comparability of our work to some of the existing literature. This study also suffers from the inherent biases and shortcomings of retrospective studies, and the fact that this was not a multicenter study may limit generalizability of the results. However, given the dearth of literature on this topic, our work adds to the literature. Further studies will be needed to more clearly elucidate this topic.

CONCLUSION

This study demonstrated that duration of symptoms may be a significant factor in the recovery of patients undergoing surgical decompression for CSM, but only in patients aged >65 years who have had symptoms for 24 months.

This paper will be judged for the Resident Writer’s Award.

ABSTRACT

The purpose of this study is to assess the effect of insurance type (Medicaid, Medicare, private insurance) on the ability for patients with operative ankle fractures to access orthopedic traumatologists. The research team called 245 board-certified orthopedic surgeons specializing in orthopedic trauma within 8 representative states. The caller requested an appointment for their fictitious mother in order to be evaluated for an ankle fracture which was previously evaluated by her primary care physician and believed to require surgery. Each office was called 3 times to assess the response for each insurance type. For each call, information was documented regarding whether the patient was able to receive an appointment and the barriers the patient confronted to receive an appointment. Overall, 35.7% of offices scheduled an appointment for a patient with Medicaid, in comparison to 81.4%and 88.6% for Medicare and BlueCross, respectively (P < .0001). Medicaid patients confronted more barriers for receiving appointments. There was no statistically significant difference in access for Medicaid patients in states that had expanded Medicaid eligibility vs states that had not expanded Medicaid. Medicaid reimbursement for open reduction and internal fixation of an ankle fracture did not significantly correlate with appointment success rates or wait times. Despite the passage of the Affordable Care Act, patients with Medicaid have reduced access to orthopedic surgeons and more complex barriers to receiving appointments. A more robust strategy for increasing care-access for patients with Medicaid would be more equitable.

Continue to: In 2010, the Patient Protection and Affordable Care Act...

In 2010, the Patient Protection and Affordable Care Act (PPACA) expanded the eligibility criteria for Medicaid to all individuals with an income up to 138% of the poverty level.¹ A Supreme Court ruling stated that the decision to expand Medicaid was to be decided by individual states.² Currently, 31 states have chosen to expand Medicaid eligibility to their residents.² This expansion has allowed an additional 11.7 million people to enroll in Medicaid and the Children’s Health Insurance Program by May 2015.^3-5

Even with the passage of the PPACA, Medicaid patients seeking specialty orthopedic care have experienced more barriers to accessing care than Medicare or commercially-insured patients.^2,6-10 One major cited reason is Medicaid’s low reimbursement, which may discourage physicians from open panel participation in Medicaid.^11,12  

A common fundamental teaching for orthopedic traumatologists is the notion that they should be available to treat all injuries regardless of the patient’s ability to pay.¹³ This has resulted in both trauma centers and trauma surgeons becoming financially challenged due to the higher proportion of Medicaid and uninsured trauma patients and lower Medicaid reimbursement levels.^14,15

This study focuses on the effect of different types of insurance (Medicaid, Medicare, or commercial insurance) on the ability of patients to obtain care for operative ankle fractures. The purpose of this study is to evaluate, in the context of the PPACA, patient access to orthopedic surgeons for operative ankle fractures based on insurance-type. We hypothesized that patients with Medicaid would face a greater volume of obstacles when seeking appointments for an ankle fracture, even after the PPACA.

Continue to: MATERIALS AND METHODS...

MATERIALS AND METHODS

The study population included board-certified orthopedic surgeons who belonged to the Orthopaedic Trauma Association (OTA) from 8 representative states; 4 states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and 4 states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas). These states were selected due to their ability to represent diverse healthcare marketplaces throughout the country. Using the OTA website’s “Find a Surgeon” search tool,¹⁶ we created a list of surgeons for each state and matched each surgeon with a random number. The list of surgeons was ordered according to the value of the surgeon’s associated random number, and surgeons were called in ascending order. We excluded disconnected or inaccurate numbers from the calling list. Surgeons who did not manage ankle fractures were removed from the dataset. Approximately 30 orthopedic trauma surgeons per state were contacted.

Each office was called to make an appointment for the caller’s mother. Every surgeon’s office was specifically asked if the surgeon would accept the patient to be evaluated for an ankle fracture that occurred out-of-state. The caller had a standardized protocol to limit intra- and inter-office variations (Appendix). The scenario involved a request to be evaluated for an unstable ankle fracture, with the patient having Medicaid, Medicare, or BlueCross insurance. The scenario required 3 separate calls to the same surgeon in order to obtain data regarding each insurance-type. The calls were separated by at least 1 week to avoid caller recognition by the surgeon’s office.

Appendix

Scenario

1. Date of Birth: Medicaid–2/07/55; BlueCross PPO–2/09/55; Medicare–7/31/45.

2. Ankle fracture evaluated by primary care physician 1 or 2 days ago

3. Not seen previously by your clinic or hospital, she would be a new patient

4. Asked how early she could be scheduled for an appointment

5. Script:

“I’m calling for my mother who injured her ankle a few days ago. Her family doctor took an X-ray and believes she has a fracture and needs surgery. Is Dr. X accepting new patients for evaluation and treatment of ankle fractures?” If YES →

“I was wondering if you take Medicaid/Medicare/BlueCross plan?” If YES →

“When is your soonest available appointment?”

The date of each phone call and date of appointment, if provided, were recorded. If the office did not give an appointment, we asked for reasons why. If an appointment was denied for a patient with Medicaid, we asked for a referral to another office that accepted Medicaid. We considered barriers to obtaining an initial appointment, such as requiring a referral from a primary care physician (PCP), as an unsuccessful attempt at making an appointment. We determined the waiting period for an appointment by calculating the time between the date of the call and the date of the appointment. Appointments were not scheduled to ensure that actual patients were not disadvantaged. For both appointment success rates and waiting periods, we stratified the data into 2 groups: states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas).

We obtained Medicaid reimbursement rates for open reduction and internal fixation of an ankle fracture by querying each state’s reimbursement rate using Current Procedural Terminology code 27822.

Chi-square test or Fisher’s exact test was used to analyze acceptance rate differences based on the patient’s type of insurance. To compare the waiting periods for an appointment, we used an independent samples t-test after applying natural log-transformation, as the data was not normally distributed. We performed logistic regression analysis to detect whether reimbursement was a significant predictor of successfully making an appointment for patients, and a linear regression analysis was used to evaluate whether reimbursement predicted waiting periods. Unless otherwise stated, all statistical testing was performed two-tailed at an alpha-level of 0.05.

This study was approved by the Institutional Review Board of Yale University School of Medicine (HIC No. 1363).

Continue to: RESULTS...

RESULTS

In total, 350 offices were contacted across 8 states (4 states with and 4 states without expanded Medicaid eligibility) of which we identified 245 orthopedic surgeons who would surgically treat ankle fractures. The 245 surgeons’ offices were called 3 times for each separate insurance-type.

Table 1. Appointment Success Rate

^aComparison to Medicaid.

The overall rate of successfully being offered an appointment with Medicaid was 35.7%, 81.4% for Medicare, and 88.6% for BlueCross (Table 1). For states with expanded Medicaid eligibility, the success rate for obtaining an appointment was 39.6%, 83.5%, and 89.2% for Medicaid, Medicare, and BlueCross, respectively. For states without expanded Medicaid eligibility, the success rate for obtaining an appointment was 31.9% for Medicaid, 79.4% for Medicare, and 87.9% for BlueCross. In all cases, the success rate for obtaining an appointment was significantly lower for Medicaid, compared to Medicare (P < .0001) or BlueCross (P < .0001). Medicaid appointment success rate was 39.6% in expanded states vs 31.9% in non-expanded states, however, the difference was not statistically significant (Table 2).

Table 2. Medicaid Appointment Success Rate in Expanded Vs Non-Expanded States

In 43.7% of occasions, patients with Medicaid did not have their insurance accepted, compared to 7.3% for Medicare and 0% for BlueCross. The majority of offices which did not accept Medicaid were not able to refer patients to another surgeon who would accept Medicaid. The requirement to have a primary care referral was the second most common reason for Medicaid patients not obtaining an appointment. No Medicare (10.4% vs 0.0%, P < .0001) or BlueCross (10.4% vs 0.0%, P < .0001) patients experienced this requirement (Table 3). There was no difference found between the percent of Medicaid patients who were required to have referrals in states with and without expanded Medicaid eligibility (Table 4).

Table 3. Referral Rate

^aComparison to Medicaid.

Table 4. Medicaid Referral Rates in Expanded Vs Non-Expanded States

Reimbursements for ankle fracture varied across states (Table 5). For Medicaid, Georgia paid the highest reimbursement ($1049.95) and Florida paid the lowest ($469.44). Logistic and linear regression analysis did not demonstrate a significant relationship between reimbursement and appointment success rate or waiting periods.

Table 5. Medicaid Reimbursements for Ankle Fracture Repair (CPT and HCPCS 27822) in 2014

^aStates with expanded Medicaid eligibility.

Abbreviations: CPT, Current Procedural Terminology; HCPCS, Healthcare Common Procedure Coding System.

Waiting periods (Table 6) varied significantly by the type of insurance (7.3 days for Medicaid, 6.0 days for Medicare, and 6.0 days for BlueCross; P = .002). For states with expanded Medicaid eligibility, waiting periods varied significantly by insurance (7.7 days for Medicaid, 6.2 days for Medicare, P = .003; and 6.1 days for BlueCross, P = .01). Waiting periods did not vary significantly for states without expanded Medicaid. Additionally, waiting periods did not differ significantly when comparing between states with and without Medicaid expansion.

Table 6. Waiting Period (Days) by Insurance Type.

Continue to: DISCUSSION...

DISCUSSION

This study assessed how insurance type (Medicaid, Medicare, and BlueCross) affects patient access to orthopedic trauma surgeons in 8 geographically representative states. We selected unstable ankle fractures as they are basic fractures treated by nearly all trauma surgeons and should often be surgically treated to prevent serious long-term consequences. Our hypothesis stated that despite the passage of the PPACA, patients with Medicaid would have reduced access to care. As the PPACA has changed the healthcare marketplace by increasing the number of Medicaid enrollees, it is important to ensure that patient access to care improves.

This nationwide survey of orthopedic trauma surgeons demonstrates that Medicaid patients experience added barriers to care that ultimately results in lower rates of successfully obtaining care. This is consistent with other investigations which have assessed Medicaid patient healthcare access.^6,8,10,17-19 This study did not demonstrate a statistically significant difference between Medicaid patients’ ability to obtain appointments in states with expanded Medicaid eligibility vs in states without expanded Medicaid eligibility (39.6% vs 31.9%, P < .18); this has been demonstrated in the literature.⁶  

A barrier that was unique to Medicaid patients was the requirement to have a PCP referral (Table 3). A PCP referral was not a barrier to receiving an appointment for patients with Medicare or BlueCross. One reason to explain why Medicaid patients may be required to have PCP referrals is due to their increased medical complexity, extra documentation requirements, and low reimbursement.⁴ Patients who have obtained a PCP referral may be characterized as being more medically compliant.

It is important to note that the Medicaid policies for 4 states included in this study (Massachusetts, North Carolina, Texas, and New York) required a PCP referral in order to see a specialist. However, we found that many orthopedic trauma practices in these states scheduled appointments for Medicaid patients without a PCP referral, suggesting that the decision depended on individual policy. In addition, the majority of offices within these states cited that they simply did not accept Medicaid as an insurance policy, and not that they required a referral.

Our regression analysis did not find a significant relationship between being able to successfully obtain an appointment to be evaluated for an ankle fracture and reimbursement rates for Medicaid. Although studies have stressed the importance of Medicaid reimbursements on physician participation, this result is consistent with previous studies regarding carpal tunnel release and total ankle replacements.^17,19 Long²⁰ suggested that although reimbursements may help, additional strategies for promoting Medicaid acceptance may be needed, including: lowering the costs of participating in Medicaid by simplifying administrative processes, speeding up reimbursement, and reducing the costs associated with caring for those patients.

Continue to: Previous studies have demonstrated...

Previous studies have demonstrated that more physicians may accept Medicaid if reimbursements increased.^4,12 Given the high percentage of trauma patients with Medicaid as their primary insurance or whom are emergently enrolled in Medicaid by hospital systems, it is concerning that the PPACA is reducing payments under the Medicare and Medicaid Disproportionate Share Hospital programs which provide hospitals for uncompensated care given to low-income and uninsured patients.²¹ Trauma centers generally operate at a deficit due to the higher proportion of Medicaid and uninsured patients.¹⁴ This is currently worsened by additional federal funding cuts for supporting trauma service’s humane mission.²¹

This study has several limitations. While the study evaluated access to care in 8 representative states, a thorough nationwide survey would be more representative. Some results may have become statistically significant if we had performed the study with a larger sample size. In addition, we were unable to control for many factors which could impact appointment wait times, such as physician call schedules and vacations. Socioeconomic factors can influence a patient’s ability to attend an appointment, such as transportation costs, time off from work, and childcare availability. In addition, this study did not assess access for the uninsured, who are predominantly the working poor who cannot afford health insurance, even with federal and state subsidies.

The authors apologize for inconveniencing these offices, however, data collection could not be achieved in a better manner. We hope that the value of this study compensates any inconvenience.

CONCLUSION

Overall, our results demonstrate that despite the ratification of the PPACA, Medicaid patients are confronted with more barriers to accessing care by comparison to patients with Medicare and BlueCross insurance. Medicaid patients have worse baseline health²² and are at an increased risk of complications. These disparities are thought to be due to decreased healthcare access,^23,24 as well as socioeconomic challenges. Interventions, such as increasing Medicaid’s reimbursement levels, reducing burdensome administrative responsibilities, and establishing partnerships between trauma centers and trauma surgeons, may enable underinsured patients to be appropriately cared for.

This paper will be judged for the Resident Writer’s Award.

ABSTRACT

The purpose of this study is to assess the effect of insurance type (Medicaid, Medicare, private insurance) on the ability for patients with operative ankle fractures to access orthopedic traumatologists. The research team called 245 board-certified orthopedic surgeons specializing in orthopedic trauma within 8 representative states. The caller requested an appointment for their fictitious mother in order to be evaluated for an ankle fracture which was previously evaluated by her primary care physician and believed to require surgery. Each office was called 3 times to assess the response for each insurance type. For each call, information was documented regarding whether the patient was able to receive an appointment and the barriers the patient confronted to receive an appointment. Overall, 35.7% of offices scheduled an appointment for a patient with Medicaid, in comparison to 81.4%and 88.6% for Medicare and BlueCross, respectively (P < .0001). Medicaid patients confronted more barriers for receiving appointments. There was no statistically significant difference in access for Medicaid patients in states that had expanded Medicaid eligibility vs states that had not expanded Medicaid. Medicaid reimbursement for open reduction and internal fixation of an ankle fracture did not significantly correlate with appointment success rates or wait times. Despite the passage of the Affordable Care Act, patients with Medicaid have reduced access to orthopedic surgeons and more complex barriers to receiving appointments. A more robust strategy for increasing care-access for patients with Medicaid would be more equitable.

Continue to: In 2010, the Patient Protection and Affordable Care Act...

In 2010, the Patient Protection and Affordable Care Act (PPACA) expanded the eligibility criteria for Medicaid to all individuals with an income up to 138% of the poverty level.¹ A Supreme Court ruling stated that the decision to expand Medicaid was to be decided by individual states.² Currently, 31 states have chosen to expand Medicaid eligibility to their residents.² This expansion has allowed an additional 11.7 million people to enroll in Medicaid and the Children’s Health Insurance Program by May 2015.^3-5

Even with the passage of the PPACA, Medicaid patients seeking specialty orthopedic care have experienced more barriers to accessing care than Medicare or commercially-insured patients.^2,6-10 One major cited reason is Medicaid’s low reimbursement, which may discourage physicians from open panel participation in Medicaid.^11,12  

A common fundamental teaching for orthopedic traumatologists is the notion that they should be available to treat all injuries regardless of the patient’s ability to pay.¹³ This has resulted in both trauma centers and trauma surgeons becoming financially challenged due to the higher proportion of Medicaid and uninsured trauma patients and lower Medicaid reimbursement levels.^14,15

This study focuses on the effect of different types of insurance (Medicaid, Medicare, or commercial insurance) on the ability of patients to obtain care for operative ankle fractures. The purpose of this study is to evaluate, in the context of the PPACA, patient access to orthopedic surgeons for operative ankle fractures based on insurance-type. We hypothesized that patients with Medicaid would face a greater volume of obstacles when seeking appointments for an ankle fracture, even after the PPACA.

Continue to: MATERIALS AND METHODS...

MATERIALS AND METHODS

The study population included board-certified orthopedic surgeons who belonged to the Orthopaedic Trauma Association (OTA) from 8 representative states; 4 states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and 4 states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas). These states were selected due to their ability to represent diverse healthcare marketplaces throughout the country. Using the OTA website’s “Find a Surgeon” search tool,¹⁶ we created a list of surgeons for each state and matched each surgeon with a random number. The list of surgeons was ordered according to the value of the surgeon’s associated random number, and surgeons were called in ascending order. We excluded disconnected or inaccurate numbers from the calling list. Surgeons who did not manage ankle fractures were removed from the dataset. Approximately 30 orthopedic trauma surgeons per state were contacted.

Each office was called to make an appointment for the caller’s mother. Every surgeon’s office was specifically asked if the surgeon would accept the patient to be evaluated for an ankle fracture that occurred out-of-state. The caller had a standardized protocol to limit intra- and inter-office variations (Appendix). The scenario involved a request to be evaluated for an unstable ankle fracture, with the patient having Medicaid, Medicare, or BlueCross insurance. The scenario required 3 separate calls to the same surgeon in order to obtain data regarding each insurance-type. The calls were separated by at least 1 week to avoid caller recognition by the surgeon’s office.

Appendix

Scenario

1. Date of Birth: Medicaid–2/07/55; BlueCross PPO–2/09/55; Medicare–7/31/45.

2. Ankle fracture evaluated by primary care physician 1 or 2 days ago

3. Not seen previously by your clinic or hospital, she would be a new patient

4. Asked how early she could be scheduled for an appointment

5. Script:

“I’m calling for my mother who injured her ankle a few days ago. Her family doctor took an X-ray and believes she has a fracture and needs surgery. Is Dr. X accepting new patients for evaluation and treatment of ankle fractures?” If YES →

“I was wondering if you take Medicaid/Medicare/BlueCross plan?” If YES →

“When is your soonest available appointment?”

The date of each phone call and date of appointment, if provided, were recorded. If the office did not give an appointment, we asked for reasons why. If an appointment was denied for a patient with Medicaid, we asked for a referral to another office that accepted Medicaid. We considered barriers to obtaining an initial appointment, such as requiring a referral from a primary care physician (PCP), as an unsuccessful attempt at making an appointment. We determined the waiting period for an appointment by calculating the time between the date of the call and the date of the appointment. Appointments were not scheduled to ensure that actual patients were not disadvantaged. For both appointment success rates and waiting periods, we stratified the data into 2 groups: states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas).

We obtained Medicaid reimbursement rates for open reduction and internal fixation of an ankle fracture by querying each state’s reimbursement rate using Current Procedural Terminology code 27822.

Chi-square test or Fisher’s exact test was used to analyze acceptance rate differences based on the patient’s type of insurance. To compare the waiting periods for an appointment, we used an independent samples t-test after applying natural log-transformation, as the data was not normally distributed. We performed logistic regression analysis to detect whether reimbursement was a significant predictor of successfully making an appointment for patients, and a linear regression analysis was used to evaluate whether reimbursement predicted waiting periods. Unless otherwise stated, all statistical testing was performed two-tailed at an alpha-level of 0.05.

This study was approved by the Institutional Review Board of Yale University School of Medicine (HIC No. 1363).

Continue to: RESULTS...

RESULTS

In total, 350 offices were contacted across 8 states (4 states with and 4 states without expanded Medicaid eligibility) of which we identified 245 orthopedic surgeons who would surgically treat ankle fractures. The 245 surgeons’ offices were called 3 times for each separate insurance-type.

Table 1. Appointment Success Rate

^aComparison to Medicaid.

The overall rate of successfully being offered an appointment with Medicaid was 35.7%, 81.4% for Medicare, and 88.6% for BlueCross (Table 1). For states with expanded Medicaid eligibility, the success rate for obtaining an appointment was 39.6%, 83.5%, and 89.2% for Medicaid, Medicare, and BlueCross, respectively. For states without expanded Medicaid eligibility, the success rate for obtaining an appointment was 31.9% for Medicaid, 79.4% for Medicare, and 87.9% for BlueCross. In all cases, the success rate for obtaining an appointment was significantly lower for Medicaid, compared to Medicare (P < .0001) or BlueCross (P < .0001). Medicaid appointment success rate was 39.6% in expanded states vs 31.9% in non-expanded states, however, the difference was not statistically significant (Table 2).

Table 2. Medicaid Appointment Success Rate in Expanded Vs Non-Expanded States

In 43.7% of occasions, patients with Medicaid did not have their insurance accepted, compared to 7.3% for Medicare and 0% for BlueCross. The majority of offices which did not accept Medicaid were not able to refer patients to another surgeon who would accept Medicaid. The requirement to have a primary care referral was the second most common reason for Medicaid patients not obtaining an appointment. No Medicare (10.4% vs 0.0%, P < .0001) or BlueCross (10.4% vs 0.0%, P < .0001) patients experienced this requirement (Table 3). There was no difference found between the percent of Medicaid patients who were required to have referrals in states with and without expanded Medicaid eligibility (Table 4).

Table 3. Referral Rate

^aComparison to Medicaid.

Table 4. Medicaid Referral Rates in Expanded Vs Non-Expanded States

Reimbursements for ankle fracture varied across states (Table 5). For Medicaid, Georgia paid the highest reimbursement ($1049.95) and Florida paid the lowest ($469.44). Logistic and linear regression analysis did not demonstrate a significant relationship between reimbursement and appointment success rate or waiting periods.

Table 5. Medicaid Reimbursements for Ankle Fracture Repair (CPT and HCPCS 27822) in 2014

^aStates with expanded Medicaid eligibility.

Abbreviations: CPT, Current Procedural Terminology; HCPCS, Healthcare Common Procedure Coding System.

Waiting periods (Table 6) varied significantly by the type of insurance (7.3 days for Medicaid, 6.0 days for Medicare, and 6.0 days for BlueCross; P = .002). For states with expanded Medicaid eligibility, waiting periods varied significantly by insurance (7.7 days for Medicaid, 6.2 days for Medicare, P = .003; and 6.1 days for BlueCross, P = .01). Waiting periods did not vary significantly for states without expanded Medicaid. Additionally, waiting periods did not differ significantly when comparing between states with and without Medicaid expansion.

Table 6. Waiting Period (Days) by Insurance Type.

Continue to: DISCUSSION...

DISCUSSION

This study assessed how insurance type (Medicaid, Medicare, and BlueCross) affects patient access to orthopedic trauma surgeons in 8 geographically representative states. We selected unstable ankle fractures as they are basic fractures treated by nearly all trauma surgeons and should often be surgically treated to prevent serious long-term consequences. Our hypothesis stated that despite the passage of the PPACA, patients with Medicaid would have reduced access to care. As the PPACA has changed the healthcare marketplace by increasing the number of Medicaid enrollees, it is important to ensure that patient access to care improves.

This nationwide survey of orthopedic trauma surgeons demonstrates that Medicaid patients experience added barriers to care that ultimately results in lower rates of successfully obtaining care. This is consistent with other investigations which have assessed Medicaid patient healthcare access.^6,8,10,17-19 This study did not demonstrate a statistically significant difference between Medicaid patients’ ability to obtain appointments in states with expanded Medicaid eligibility vs in states without expanded Medicaid eligibility (39.6% vs 31.9%, P < .18); this has been demonstrated in the literature.⁶  

A barrier that was unique to Medicaid patients was the requirement to have a PCP referral (Table 3). A PCP referral was not a barrier to receiving an appointment for patients with Medicare or BlueCross. One reason to explain why Medicaid patients may be required to have PCP referrals is due to their increased medical complexity, extra documentation requirements, and low reimbursement.⁴ Patients who have obtained a PCP referral may be characterized as being more medically compliant.

It is important to note that the Medicaid policies for 4 states included in this study (Massachusetts, North Carolina, Texas, and New York) required a PCP referral in order to see a specialist. However, we found that many orthopedic trauma practices in these states scheduled appointments for Medicaid patients without a PCP referral, suggesting that the decision depended on individual policy. In addition, the majority of offices within these states cited that they simply did not accept Medicaid as an insurance policy, and not that they required a referral.

Our regression analysis did not find a significant relationship between being able to successfully obtain an appointment to be evaluated for an ankle fracture and reimbursement rates for Medicaid. Although studies have stressed the importance of Medicaid reimbursements on physician participation, this result is consistent with previous studies regarding carpal tunnel release and total ankle replacements.^17,19 Long²⁰ suggested that although reimbursements may help, additional strategies for promoting Medicaid acceptance may be needed, including: lowering the costs of participating in Medicaid by simplifying administrative processes, speeding up reimbursement, and reducing the costs associated with caring for those patients.

Continue to: Previous studies have demonstrated...

Previous studies have demonstrated that more physicians may accept Medicaid if reimbursements increased.^4,12 Given the high percentage of trauma patients with Medicaid as their primary insurance or whom are emergently enrolled in Medicaid by hospital systems, it is concerning that the PPACA is reducing payments under the Medicare and Medicaid Disproportionate Share Hospital programs which provide hospitals for uncompensated care given to low-income and uninsured patients.²¹ Trauma centers generally operate at a deficit due to the higher proportion of Medicaid and uninsured patients.¹⁴ This is currently worsened by additional federal funding cuts for supporting trauma service’s humane mission.²¹

This study has several limitations. While the study evaluated access to care in 8 representative states, a thorough nationwide survey would be more representative. Some results may have become statistically significant if we had performed the study with a larger sample size. In addition, we were unable to control for many factors which could impact appointment wait times, such as physician call schedules and vacations. Socioeconomic factors can influence a patient’s ability to attend an appointment, such as transportation costs, time off from work, and childcare availability. In addition, this study did not assess access for the uninsured, who are predominantly the working poor who cannot afford health insurance, even with federal and state subsidies.

The authors apologize for inconveniencing these offices, however, data collection could not be achieved in a better manner. We hope that the value of this study compensates any inconvenience.

CONCLUSION

Overall, our results demonstrate that despite the ratification of the PPACA, Medicaid patients are confronted with more barriers to accessing care by comparison to patients with Medicare and BlueCross insurance. Medicaid patients have worse baseline health²² and are at an increased risk of complications. These disparities are thought to be due to decreased healthcare access,^23,24 as well as socioeconomic challenges. Interventions, such as increasing Medicaid’s reimbursement levels, reducing burdensome administrative responsibilities, and establishing partnerships between trauma centers and trauma surgeons, may enable underinsured patients to be appropriately cared for.

This paper will be judged for the Resident Writer’s Award.

ABSTRACT

The purpose of this study is to assess the effect of insurance type (Medicaid, Medicare, private insurance) on the ability for patients with operative ankle fractures to access orthopedic traumatologists. The research team called 245 board-certified orthopedic surgeons specializing in orthopedic trauma within 8 representative states. The caller requested an appointment for their fictitious mother in order to be evaluated for an ankle fracture which was previously evaluated by her primary care physician and believed to require surgery. Each office was called 3 times to assess the response for each insurance type. For each call, information was documented regarding whether the patient was able to receive an appointment and the barriers the patient confronted to receive an appointment. Overall, 35.7% of offices scheduled an appointment for a patient with Medicaid, in comparison to 81.4%and 88.6% for Medicare and BlueCross, respectively (P < .0001). Medicaid patients confronted more barriers for receiving appointments. There was no statistically significant difference in access for Medicaid patients in states that had expanded Medicaid eligibility vs states that had not expanded Medicaid. Medicaid reimbursement for open reduction and internal fixation of an ankle fracture did not significantly correlate with appointment success rates or wait times. Despite the passage of the Affordable Care Act, patients with Medicaid have reduced access to orthopedic surgeons and more complex barriers to receiving appointments. A more robust strategy for increasing care-access for patients with Medicaid would be more equitable.

Continue to: In 2010, the Patient Protection and Affordable Care Act...

In 2010, the Patient Protection and Affordable Care Act (PPACA) expanded the eligibility criteria for Medicaid to all individuals with an income up to 138% of the poverty level.¹ A Supreme Court ruling stated that the decision to expand Medicaid was to be decided by individual states.² Currently, 31 states have chosen to expand Medicaid eligibility to their residents.² This expansion has allowed an additional 11.7 million people to enroll in Medicaid and the Children’s Health Insurance Program by May 2015.^3-5

Even with the passage of the PPACA, Medicaid patients seeking specialty orthopedic care have experienced more barriers to accessing care than Medicare or commercially-insured patients.^2,6-10 One major cited reason is Medicaid’s low reimbursement, which may discourage physicians from open panel participation in Medicaid.^11,12  

A common fundamental teaching for orthopedic traumatologists is the notion that they should be available to treat all injuries regardless of the patient’s ability to pay.¹³ This has resulted in both trauma centers and trauma surgeons becoming financially challenged due to the higher proportion of Medicaid and uninsured trauma patients and lower Medicaid reimbursement levels.^14,15

This study focuses on the effect of different types of insurance (Medicaid, Medicare, or commercial insurance) on the ability of patients to obtain care for operative ankle fractures. The purpose of this study is to evaluate, in the context of the PPACA, patient access to orthopedic surgeons for operative ankle fractures based on insurance-type. We hypothesized that patients with Medicaid would face a greater volume of obstacles when seeking appointments for an ankle fracture, even after the PPACA.

Continue to: MATERIALS AND METHODS...

MATERIALS AND METHODS

The study population included board-certified orthopedic surgeons who belonged to the Orthopaedic Trauma Association (OTA) from 8 representative states; 4 states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and 4 states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas). These states were selected due to their ability to represent diverse healthcare marketplaces throughout the country. Using the OTA website’s “Find a Surgeon” search tool,¹⁶ we created a list of surgeons for each state and matched each surgeon with a random number. The list of surgeons was ordered according to the value of the surgeon’s associated random number, and surgeons were called in ascending order. We excluded disconnected or inaccurate numbers from the calling list. Surgeons who did not manage ankle fractures were removed from the dataset. Approximately 30 orthopedic trauma surgeons per state were contacted.

Each office was called to make an appointment for the caller’s mother. Every surgeon’s office was specifically asked if the surgeon would accept the patient to be evaluated for an ankle fracture that occurred out-of-state. The caller had a standardized protocol to limit intra- and inter-office variations (Appendix). The scenario involved a request to be evaluated for an unstable ankle fracture, with the patient having Medicaid, Medicare, or BlueCross insurance. The scenario required 3 separate calls to the same surgeon in order to obtain data regarding each insurance-type. The calls were separated by at least 1 week to avoid caller recognition by the surgeon’s office.

Appendix

Scenario

1. Date of Birth: Medicaid–2/07/55; BlueCross PPO–2/09/55; Medicare–7/31/45.

2. Ankle fracture evaluated by primary care physician 1 or 2 days ago

3. Not seen previously by your clinic or hospital, she would be a new patient

4. Asked how early she could be scheduled for an appointment

5. Script:

“I’m calling for my mother who injured her ankle a few days ago. Her family doctor took an X-ray and believes she has a fracture and needs surgery. Is Dr. X accepting new patients for evaluation and treatment of ankle fractures?” If YES →

“I was wondering if you take Medicaid/Medicare/BlueCross plan?” If YES →

“When is your soonest available appointment?”

The date of each phone call and date of appointment, if provided, were recorded. If the office did not give an appointment, we asked for reasons why. If an appointment was denied for a patient with Medicaid, we asked for a referral to another office that accepted Medicaid. We considered barriers to obtaining an initial appointment, such as requiring a referral from a primary care physician (PCP), as an unsuccessful attempt at making an appointment. We determined the waiting period for an appointment by calculating the time between the date of the call and the date of the appointment. Appointments were not scheduled to ensure that actual patients were not disadvantaged. For both appointment success rates and waiting periods, we stratified the data into 2 groups: states with expanded Medicaid eligibility (California, Massachusetts, New York, Ohio) and states without expanded Medicaid eligibility (Florida, North Carolina, Georgia, Texas).

We obtained Medicaid reimbursement rates for open reduction and internal fixation of an ankle fracture by querying each state’s reimbursement rate using Current Procedural Terminology code 27822.

Chi-square test or Fisher’s exact test was used to analyze acceptance rate differences based on the patient’s type of insurance. To compare the waiting periods for an appointment, we used an independent samples t-test after applying natural log-transformation, as the data was not normally distributed. We performed logistic regression analysis to detect whether reimbursement was a significant predictor of successfully making an appointment for patients, and a linear regression analysis was used to evaluate whether reimbursement predicted waiting periods. Unless otherwise stated, all statistical testing was performed two-tailed at an alpha-level of 0.05.

This study was approved by the Institutional Review Board of Yale University School of Medicine (HIC No. 1363).

Continue to: RESULTS...

RESULTS

In total, 350 offices were contacted across 8 states (4 states with and 4 states without expanded Medicaid eligibility) of which we identified 245 orthopedic surgeons who would surgically treat ankle fractures. The 245 surgeons’ offices were called 3 times for each separate insurance-type.

Table 1. Appointment Success Rate

^aComparison to Medicaid.

The overall rate of successfully being offered an appointment with Medicaid was 35.7%, 81.4% for Medicare, and 88.6% for BlueCross (Table 1). For states with expanded Medicaid eligibility, the success rate for obtaining an appointment was 39.6%, 83.5%, and 89.2% for Medicaid, Medicare, and BlueCross, respectively. For states without expanded Medicaid eligibility, the success rate for obtaining an appointment was 31.9% for Medicaid, 79.4% for Medicare, and 87.9% for BlueCross. In all cases, the success rate for obtaining an appointment was significantly lower for Medicaid, compared to Medicare (P < .0001) or BlueCross (P < .0001). Medicaid appointment success rate was 39.6% in expanded states vs 31.9% in non-expanded states, however, the difference was not statistically significant (Table 2).

Table 2. Medicaid Appointment Success Rate in Expanded Vs Non-Expanded States

In 43.7% of occasions, patients with Medicaid did not have their insurance accepted, compared to 7.3% for Medicare and 0% for BlueCross. The majority of offices which did not accept Medicaid were not able to refer patients to another surgeon who would accept Medicaid. The requirement to have a primary care referral was the second most common reason for Medicaid patients not obtaining an appointment. No Medicare (10.4% vs 0.0%, P < .0001) or BlueCross (10.4% vs 0.0%, P < .0001) patients experienced this requirement (Table 3). There was no difference found between the percent of Medicaid patients who were required to have referrals in states with and without expanded Medicaid eligibility (Table 4).

Table 3. Referral Rate

^aComparison to Medicaid.

Table 4. Medicaid Referral Rates in Expanded Vs Non-Expanded States

Reimbursements for ankle fracture varied across states (Table 5). For Medicaid, Georgia paid the highest reimbursement ($1049.95) and Florida paid the lowest ($469.44). Logistic and linear regression analysis did not demonstrate a significant relationship between reimbursement and appointment success rate or waiting periods.

Table 5. Medicaid Reimbursements for Ankle Fracture Repair (CPT and HCPCS 27822) in 2014

^aStates with expanded Medicaid eligibility.

Abbreviations: CPT, Current Procedural Terminology; HCPCS, Healthcare Common Procedure Coding System.

Waiting periods (Table 6) varied significantly by the type of insurance (7.3 days for Medicaid, 6.0 days for Medicare, and 6.0 days for BlueCross; P = .002). For states with expanded Medicaid eligibility, waiting periods varied significantly by insurance (7.7 days for Medicaid, 6.2 days for Medicare, P = .003; and 6.1 days for BlueCross, P = .01). Waiting periods did not vary significantly for states without expanded Medicaid. Additionally, waiting periods did not differ significantly when comparing between states with and without Medicaid expansion.

Table 6. Waiting Period (Days) by Insurance Type.

Continue to: DISCUSSION...

DISCUSSION

This study assessed how insurance type (Medicaid, Medicare, and BlueCross) affects patient access to orthopedic trauma surgeons in 8 geographically representative states. We selected unstable ankle fractures as they are basic fractures treated by nearly all trauma surgeons and should often be surgically treated to prevent serious long-term consequences. Our hypothesis stated that despite the passage of the PPACA, patients with Medicaid would have reduced access to care. As the PPACA has changed the healthcare marketplace by increasing the number of Medicaid enrollees, it is important to ensure that patient access to care improves.

This nationwide survey of orthopedic trauma surgeons demonstrates that Medicaid patients experience added barriers to care that ultimately results in lower rates of successfully obtaining care. This is consistent with other investigations which have assessed Medicaid patient healthcare access.^6,8,10,17-19 This study did not demonstrate a statistically significant difference between Medicaid patients’ ability to obtain appointments in states with expanded Medicaid eligibility vs in states without expanded Medicaid eligibility (39.6% vs 31.9%, P < .18); this has been demonstrated in the literature.⁶  

A barrier that was unique to Medicaid patients was the requirement to have a PCP referral (Table 3). A PCP referral was not a barrier to receiving an appointment for patients with Medicare or BlueCross. One reason to explain why Medicaid patients may be required to have PCP referrals is due to their increased medical complexity, extra documentation requirements, and low reimbursement.⁴ Patients who have obtained a PCP referral may be characterized as being more medically compliant.

It is important to note that the Medicaid policies for 4 states included in this study (Massachusetts, North Carolina, Texas, and New York) required a PCP referral in order to see a specialist. However, we found that many orthopedic trauma practices in these states scheduled appointments for Medicaid patients without a PCP referral, suggesting that the decision depended on individual policy. In addition, the majority of offices within these states cited that they simply did not accept Medicaid as an insurance policy, and not that they required a referral.

Our regression analysis did not find a significant relationship between being able to successfully obtain an appointment to be evaluated for an ankle fracture and reimbursement rates for Medicaid. Although studies have stressed the importance of Medicaid reimbursements on physician participation, this result is consistent with previous studies regarding carpal tunnel release and total ankle replacements.^17,19 Long²⁰ suggested that although reimbursements may help, additional strategies for promoting Medicaid acceptance may be needed, including: lowering the costs of participating in Medicaid by simplifying administrative processes, speeding up reimbursement, and reducing the costs associated with caring for those patients.

Continue to: Previous studies have demonstrated...

Previous studies have demonstrated that more physicians may accept Medicaid if reimbursements increased.^4,12 Given the high percentage of trauma patients with Medicaid as their primary insurance or whom are emergently enrolled in Medicaid by hospital systems, it is concerning that the PPACA is reducing payments under the Medicare and Medicaid Disproportionate Share Hospital programs which provide hospitals for uncompensated care given to low-income and uninsured patients.²¹ Trauma centers generally operate at a deficit due to the higher proportion of Medicaid and uninsured patients.¹⁴ This is currently worsened by additional federal funding cuts for supporting trauma service’s humane mission.²¹

This study has several limitations. While the study evaluated access to care in 8 representative states, a thorough nationwide survey would be more representative. Some results may have become statistically significant if we had performed the study with a larger sample size. In addition, we were unable to control for many factors which could impact appointment wait times, such as physician call schedules and vacations. Socioeconomic factors can influence a patient’s ability to attend an appointment, such as transportation costs, time off from work, and childcare availability. In addition, this study did not assess access for the uninsured, who are predominantly the working poor who cannot afford health insurance, even with federal and state subsidies.

The authors apologize for inconveniencing these offices, however, data collection could not be achieved in a better manner. We hope that the value of this study compensates any inconvenience.

CONCLUSION

Overall, our results demonstrate that despite the ratification of the PPACA, Medicaid patients are confronted with more barriers to accessing care by comparison to patients with Medicare and BlueCross insurance. Medicaid patients have worse baseline health²² and are at an increased risk of complications. These disparities are thought to be due to decreased healthcare access,^23,24 as well as socioeconomic challenges. Interventions, such as increasing Medicaid’s reimbursement levels, reducing burdensome administrative responsibilities, and establishing partnerships between trauma centers and trauma surgeons, may enable underinsured patients to be appropriately cared for.

This paper will be judged for the Resident Writer’s Award.

Care transitions between hospitals and skilled nursing facilities (SNFs) are a vulnerable time for patients. The current health care climate of decreasing hospital length of stay, readmission penalties, and increasing patient complexity has made hospital care transitions an important safety concern. Suboptimal transitions across clinical settings can result in adverse events, inadequately controlled comorbidities, deficient patient and caregiver preparation for discharge, medication errors, relocation stress, and overall increased morbidity and mortality.^1,2 Such care transitions also may generate unnecessary spending, including avoidable readmissions, emergency department utilization, and duplicative laboratory and imaging studies. Approximately 23% of patients admitted to SNFs are readmitted to acute care hospitals within 30 days, and these patients have increased mortality rates in risk-adjusted analyses. ^3,4

Compounding the magnitude of this risk and vulnerability is the significant growth in the number of patients discharged to SNFs over the past 30 years. In 2013, more than 20% of Medicare patients discharged from acute care hospitals were destined for SNFs.^5,6 Paradoxically, despite the increasing need for SNF providers, there is a shortage of clinicians with training in geriatrics or nursing home care.⁷ The result is a growing need to identify organizational systems to optimize physician practice in these settings, enhance quality of care, especially around transitions, and increase educational training opportunities in SNFs for future practitioners.

Many SNFs today are staffed by physicians and other licensed clinicians whose exclusive practice location is the nursing facility or possibly several such facilities. This prevailing model of care can isolate the physicians, depriving them of interaction with clinicians in other specialties, and can contribute to burnout.⁸ This model does not lend itself to academic scholarship, quality improvement (QI), and student or resident training, as each of these endeavors depends on interprofessional collaboration as well as access to an academic medical center with additional resources.⁹

Few studies have described innovative hospitalist rotation models from acute to subacute care. The Cleveland Clinic implemented the Connected Care model where hospital-employed physicians and advanced practice professionals integrated into postacute care and reduced the 30-day hospital readmission rate from SNFs from 28% to 22%.¹⁰ Goth and colleagues performed a comparative effectiveness trial between a postacute care hospitalist (PACH) model and a community-based physician model of nursing home care. They found that the institution of a PACH model in a nursing home was associated with a significant increase in laboratory costs, nonsignificant reduction in medication errors and pharmacy costs, and no improvement in fall rates.¹¹ The conclusion was that the PACH model may lead to greater clinician involvement and that the potential decrease in pharmacy costs and medications errors may offset the costs associated with additional laboratory testing. Overall, there has been a lack of studies on the impact of these hospitalist rotation models from acute to subacute care on educational programs, QI activities, and the interprofessional environment.

To achieve a system in which physicians in a SNF can excel in these areas, Veterans Affairs Boston Healthcare System (VABHS) adopted a staffing model in which academic hospitalist physicians rotate between the inpatient hospital and subacute settings. This report describes the model structure, the varying roles of the physicians, and early indicators of its positive effects on educational programs, QI activities, and the interprofessional environment.

Methods

The VABHS consists of a 159-bed acute care hospital in West Roxbury, Massachusetts; and a 110-bed SNF in Brockton, Massachusetts, with 3 units: a 65-bed transitional care unit (TCU), a 30-bed long-term care unit, and a 15-bed palliative care/hospice unit. The majority of patients admitted to the SNF are transferred from the acute care hospital in West Roxbury and other regional hospitals. Prior to 2015, the TCU was staffed with full-time clinicians who exclusively practiced in the SNF.

In the new staffing model, 6 hospitalist physicians divide their clinical time between the acute care hospital’s inpatient medical service and the TCU. The hospitalists come from varied backgrounds in terms of years in practice and advanced training (Table 1). 

The amount of nonclinical (protected) time and clinical time on the acute inpatient service and the TCU varies for each physician. For example, a physician serves as principal investigator for several major research grants and has a hospital-wide administrative leadership role; as a result, the principal investigator has fewer months of clinical responsibility. Physicians are expected to use the protected time for scholarship, educational program development and teaching, QI, and administrative responsibilities. The VABHS leadership determines the amount of protected time based on individualized benchmarks for research, education, and administrative responsibilities that follow VA national and local institutional guidelines. These metrics and time allocations are negotiated at the time of recruitment and then are reviewed annually.

The TCU also is staffed with 4 full-time clinicians (2 physicians and 2 physician assistants) who provide additional continuity of care. The new hospitalist staffing model only required an approximate 10% increase in TCU clinical staffing full-time equivalents. Patients and admissions are divided equally among clinicians on service (census per clinician 12-15 patients), with redistribution of patients at times of transition from clinical to nonclinical time. Blocks of clinical time are scheduled for greater than 2 weeks at a time to preserve continuity. In addition, the new staffing model allocates assignment of clinical responsibilities that allows for clinicians to take leave without resultant shortages in clinical coverage.

To facilitate communication among physicians serving in the acute inpatient facility and the TCU, leaders of both of these programs meet monthly and ad hoc to review the transitions of care between the 2 settings. The description of this model and its assessment have been reviewed and deemed exempt from oversight by the VA Boston Healthcare System Research and Development Committee.

Results

Since the implementation of this staffing model in 2015, the system has grown considerably in the breadth and depth of educational programming, QI, and systems redesign in the TCU and, more broadly, in the SNF. The TCU, which previously had limited training opportunities, has experienced marked expansion of educational offerings. It is now a site for core general medicine rotations for first-year psychiatry residents and physician assistant students. The TCU also has expanded as a clinical site for transitions-in-care internal medicine resident curricula and electives, as well as a clinical site for a geriatrics fellowship.

A hospitalist developed and implemented a 4-week interprofessional curriculum for all clinical trainees and students, which occurs continuously. The curriculum includes a monthly academic conference and 12 didactic lectures and is taught by 16 interprofessional faculty from the TCU and the Palliative Care/Hospice Unit, including medicine, geriatric and palliative care physicians, physician assistants, social workers, physical and occupational therapists, pharmacists, and a geriatric psychologist. The goal of the curriculum is to provide learners the knowledge, attitudes, and skills necessary to perform effective, efficient, and safe transfers between clinical settings as well as education in transitional care. In addition, using a team of interprofessional faculty, the curriculum develops the interprofessional competencies of teamwork and communication. The curriculum also has provided a significant opportunity for interprofessional collaboration among faculty who have volunteered their teaching time in the development and teaching of the curriculum, with potential for improved clinical staff knowledge of other disciplines.

Quality improvement and system redesign projects in care transitions also have expanded (Table 2). 

Early assessment indicates that the new staffing model is having positive effects on the clinical environment of the TCU. A survey was conducted of a convenience sample of all physicians, nurse managers, social workers, and other members of the clinical team in the TCU (N=24)(Table 3), with response categories ranging on a Likert scale from 1 (very negative) to 5 (very positive). 

Although not rigorously analyzed using qualitative research methods, comments from respondents have consistently indicated that this staffing model increases the transfer of clinical and logistical knowledge among staff members working in the acute inpatient facility and the TCU. 

Discussion

With greater numbers of increasingly complex patients transitioning from the hospital to SNF, health care systems need to expand the capacity of their skilled nursing systems, not only to provide clinical care, but also to support QI and medical education. The VABHS developed a physician staffing model with the goal of enriching physician practice and enhancing QI and educational opportunities in its SNF. The model offers an opportunity to improve transitions in care as physicians gain a greater knowledge of both the hospital and subacute clinical settings. This hospitalist rotation model may improve the knowledge necessary for caring for patients moving across care settings, as well as improve communication between settings. It also has served as a foundation for systematic innovation in QI and education at this institution. Clinical staff in the transitional care setting have reported positive effects of this model on clinical skills and patient care, educational opportunities, as well as a desire for replication in other health care systems.

The potential generalizability of this model requires careful consideration. The VABHS is a tertiary care integrated health care system, enabling physicians to work in multiple clinical settings. Other settings may not have the staffing or clinical volume to sustain such a model. In addition, this model may increase discontinuity in patient care as hospitalists move between acute and subacute settings and nonclinical roles. This loss of continuity may be a greater concern in the SNF setting, as the inpatient hospitalist model generally involves high provider turnover as shift work. Our survey included nurse managers, and not floor nurses due to survey administration limitations, and feedback may not have captured a comprehensive view from CLC staff. Moreover, some of the perceived positive impacts also may be related to professional and personal attributes of the physicians rather than the actual model of care. In addition, the survey response rate was 86%. However, the nature of the improvement work (focused on care transitions) and educational opportunities (interprofessional care) would likely not occur had the physicians been based in one clinical setting.

Other new physician staffing models have been designed to improve the continuity between the hospital, subacute, and outpatient settings. For example, the University of Chicago Comprehensive Care model pairs patients with trained hospitalists who provide both inpatient and outpatient care, thereby optimizing continuity between these settings.¹⁴ At CareMore Health System, high-risk patients also are paired with hospitalists, referred to as “extensivists,” who lead care teams that follow patients between settings and provide acute, postacute, and outpatient care.¹⁵ In these models, a single physician takes responsibility for the patient throughout transitions of care and through various care settings. Both models have shown reduction in hospital readmissions. One concern with such models is that the treatment teams need to coexist in the various settings of care, and the ability to impact and create systematic change within each environment is limited. This may limit QI, educational opportunities, and system level impact within each environment of care.

In comparison, the “transitionalist” model proposed here features hospitalist physicians rotating between the acute inpatient hospital and subacute care with dedicated time in each environment. This innovative organizational structure may enhance physician practice and enrich QI and educational opportunities in SNFs. Further evaluation will include the impact on quality metrics of patient care and patient satisfaction, as this model has the potential to influence quality, cost, and overall health outcomes.

Acknowledgments
We would like to thank Shivani Jindal, Matthew Russell, Matthew Ronan, Juman Hijab, Wei Shen, Sandra Vilbrun-Bruno, and Jack Earnshaw for their significant contributions to this staffing model. We would also like to thank Paul Conlin, Jay Orlander, and the leadership team of Veterans Affairs Boston Healthcare System for supporting this staffing model.

Care transitions between hospitals and skilled nursing facilities (SNFs) are a vulnerable time for patients. The current health care climate of decreasing hospital length of stay, readmission penalties, and increasing patient complexity has made hospital care transitions an important safety concern. Suboptimal transitions across clinical settings can result in adverse events, inadequately controlled comorbidities, deficient patient and caregiver preparation for discharge, medication errors, relocation stress, and overall increased morbidity and mortality.^1,2 Such care transitions also may generate unnecessary spending, including avoidable readmissions, emergency department utilization, and duplicative laboratory and imaging studies. Approximately 23% of patients admitted to SNFs are readmitted to acute care hospitals within 30 days, and these patients have increased mortality rates in risk-adjusted analyses. ^3,4

Compounding the magnitude of this risk and vulnerability is the significant growth in the number of patients discharged to SNFs over the past 30 years. In 2013, more than 20% of Medicare patients discharged from acute care hospitals were destined for SNFs.^5,6 Paradoxically, despite the increasing need for SNF providers, there is a shortage of clinicians with training in geriatrics or nursing home care.⁷ The result is a growing need to identify organizational systems to optimize physician practice in these settings, enhance quality of care, especially around transitions, and increase educational training opportunities in SNFs for future practitioners.

Many SNFs today are staffed by physicians and other licensed clinicians whose exclusive practice location is the nursing facility or possibly several such facilities. This prevailing model of care can isolate the physicians, depriving them of interaction with clinicians in other specialties, and can contribute to burnout.⁸ This model does not lend itself to academic scholarship, quality improvement (QI), and student or resident training, as each of these endeavors depends on interprofessional collaboration as well as access to an academic medical center with additional resources.⁹

Few studies have described innovative hospitalist rotation models from acute to subacute care. The Cleveland Clinic implemented the Connected Care model where hospital-employed physicians and advanced practice professionals integrated into postacute care and reduced the 30-day hospital readmission rate from SNFs from 28% to 22%.¹⁰ Goth and colleagues performed a comparative effectiveness trial between a postacute care hospitalist (PACH) model and a community-based physician model of nursing home care. They found that the institution of a PACH model in a nursing home was associated with a significant increase in laboratory costs, nonsignificant reduction in medication errors and pharmacy costs, and no improvement in fall rates.¹¹ The conclusion was that the PACH model may lead to greater clinician involvement and that the potential decrease in pharmacy costs and medications errors may offset the costs associated with additional laboratory testing. Overall, there has been a lack of studies on the impact of these hospitalist rotation models from acute to subacute care on educational programs, QI activities, and the interprofessional environment.

To achieve a system in which physicians in a SNF can excel in these areas, Veterans Affairs Boston Healthcare System (VABHS) adopted a staffing model in which academic hospitalist physicians rotate between the inpatient hospital and subacute settings. This report describes the model structure, the varying roles of the physicians, and early indicators of its positive effects on educational programs, QI activities, and the interprofessional environment.

Methods

The VABHS consists of a 159-bed acute care hospital in West Roxbury, Massachusetts; and a 110-bed SNF in Brockton, Massachusetts, with 3 units: a 65-bed transitional care unit (TCU), a 30-bed long-term care unit, and a 15-bed palliative care/hospice unit. The majority of patients admitted to the SNF are transferred from the acute care hospital in West Roxbury and other regional hospitals. Prior to 2015, the TCU was staffed with full-time clinicians who exclusively practiced in the SNF.

In the new staffing model, 6 hospitalist physicians divide their clinical time between the acute care hospital’s inpatient medical service and the TCU. The hospitalists come from varied backgrounds in terms of years in practice and advanced training (Table 1). 

The amount of nonclinical (protected) time and clinical time on the acute inpatient service and the TCU varies for each physician. For example, a physician serves as principal investigator for several major research grants and has a hospital-wide administrative leadership role; as a result, the principal investigator has fewer months of clinical responsibility. Physicians are expected to use the protected time for scholarship, educational program development and teaching, QI, and administrative responsibilities. The VABHS leadership determines the amount of protected time based on individualized benchmarks for research, education, and administrative responsibilities that follow VA national and local institutional guidelines. These metrics and time allocations are negotiated at the time of recruitment and then are reviewed annually.

The TCU also is staffed with 4 full-time clinicians (2 physicians and 2 physician assistants) who provide additional continuity of care. The new hospitalist staffing model only required an approximate 10% increase in TCU clinical staffing full-time equivalents. Patients and admissions are divided equally among clinicians on service (census per clinician 12-15 patients), with redistribution of patients at times of transition from clinical to nonclinical time. Blocks of clinical time are scheduled for greater than 2 weeks at a time to preserve continuity. In addition, the new staffing model allocates assignment of clinical responsibilities that allows for clinicians to take leave without resultant shortages in clinical coverage.

To facilitate communication among physicians serving in the acute inpatient facility and the TCU, leaders of both of these programs meet monthly and ad hoc to review the transitions of care between the 2 settings. The description of this model and its assessment have been reviewed and deemed exempt from oversight by the VA Boston Healthcare System Research and Development Committee.

Results

Since the implementation of this staffing model in 2015, the system has grown considerably in the breadth and depth of educational programming, QI, and systems redesign in the TCU and, more broadly, in the SNF. The TCU, which previously had limited training opportunities, has experienced marked expansion of educational offerings. It is now a site for core general medicine rotations for first-year psychiatry residents and physician assistant students. The TCU also has expanded as a clinical site for transitions-in-care internal medicine resident curricula and electives, as well as a clinical site for a geriatrics fellowship.

A hospitalist developed and implemented a 4-week interprofessional curriculum for all clinical trainees and students, which occurs continuously. The curriculum includes a monthly academic conference and 12 didactic lectures and is taught by 16 interprofessional faculty from the TCU and the Palliative Care/Hospice Unit, including medicine, geriatric and palliative care physicians, physician assistants, social workers, physical and occupational therapists, pharmacists, and a geriatric psychologist. The goal of the curriculum is to provide learners the knowledge, attitudes, and skills necessary to perform effective, efficient, and safe transfers between clinical settings as well as education in transitional care. In addition, using a team of interprofessional faculty, the curriculum develops the interprofessional competencies of teamwork and communication. The curriculum also has provided a significant opportunity for interprofessional collaboration among faculty who have volunteered their teaching time in the development and teaching of the curriculum, with potential for improved clinical staff knowledge of other disciplines.

Quality improvement and system redesign projects in care transitions also have expanded (Table 2). 

Early assessment indicates that the new staffing model is having positive effects on the clinical environment of the TCU. A survey was conducted of a convenience sample of all physicians, nurse managers, social workers, and other members of the clinical team in the TCU (N=24)(Table 3), with response categories ranging on a Likert scale from 1 (very negative) to 5 (very positive). 

Although not rigorously analyzed using qualitative research methods, comments from respondents have consistently indicated that this staffing model increases the transfer of clinical and logistical knowledge among staff members working in the acute inpatient facility and the TCU. 

Discussion

With greater numbers of increasingly complex patients transitioning from the hospital to SNF, health care systems need to expand the capacity of their skilled nursing systems, not only to provide clinical care, but also to support QI and medical education. The VABHS developed a physician staffing model with the goal of enriching physician practice and enhancing QI and educational opportunities in its SNF. The model offers an opportunity to improve transitions in care as physicians gain a greater knowledge of both the hospital and subacute clinical settings. This hospitalist rotation model may improve the knowledge necessary for caring for patients moving across care settings, as well as improve communication between settings. It also has served as a foundation for systematic innovation in QI and education at this institution. Clinical staff in the transitional care setting have reported positive effects of this model on clinical skills and patient care, educational opportunities, as well as a desire for replication in other health care systems.

The potential generalizability of this model requires careful consideration. The VABHS is a tertiary care integrated health care system, enabling physicians to work in multiple clinical settings. Other settings may not have the staffing or clinical volume to sustain such a model. In addition, this model may increase discontinuity in patient care as hospitalists move between acute and subacute settings and nonclinical roles. This loss of continuity may be a greater concern in the SNF setting, as the inpatient hospitalist model generally involves high provider turnover as shift work. Our survey included nurse managers, and not floor nurses due to survey administration limitations, and feedback may not have captured a comprehensive view from CLC staff. Moreover, some of the perceived positive impacts also may be related to professional and personal attributes of the physicians rather than the actual model of care. In addition, the survey response rate was 86%. However, the nature of the improvement work (focused on care transitions) and educational opportunities (interprofessional care) would likely not occur had the physicians been based in one clinical setting.

Other new physician staffing models have been designed to improve the continuity between the hospital, subacute, and outpatient settings. For example, the University of Chicago Comprehensive Care model pairs patients with trained hospitalists who provide both inpatient and outpatient care, thereby optimizing continuity between these settings.¹⁴ At CareMore Health System, high-risk patients also are paired with hospitalists, referred to as “extensivists,” who lead care teams that follow patients between settings and provide acute, postacute, and outpatient care.¹⁵ In these models, a single physician takes responsibility for the patient throughout transitions of care and through various care settings. Both models have shown reduction in hospital readmissions. One concern with such models is that the treatment teams need to coexist in the various settings of care, and the ability to impact and create systematic change within each environment is limited. This may limit QI, educational opportunities, and system level impact within each environment of care.

In comparison, the “transitionalist” model proposed here features hospitalist physicians rotating between the acute inpatient hospital and subacute care with dedicated time in each environment. This innovative organizational structure may enhance physician practice and enrich QI and educational opportunities in SNFs. Further evaluation will include the impact on quality metrics of patient care and patient satisfaction, as this model has the potential to influence quality, cost, and overall health outcomes.

Acknowledgments
We would like to thank Shivani Jindal, Matthew Russell, Matthew Ronan, Juman Hijab, Wei Shen, Sandra Vilbrun-Bruno, and Jack Earnshaw for their significant contributions to this staffing model. We would also like to thank Paul Conlin, Jay Orlander, and the leadership team of Veterans Affairs Boston Healthcare System for supporting this staffing model.

Care transitions between hospitals and skilled nursing facilities (SNFs) are a vulnerable time for patients. The current health care climate of decreasing hospital length of stay, readmission penalties, and increasing patient complexity has made hospital care transitions an important safety concern. Suboptimal transitions across clinical settings can result in adverse events, inadequately controlled comorbidities, deficient patient and caregiver preparation for discharge, medication errors, relocation stress, and overall increased morbidity and mortality.^1,2 Such care transitions also may generate unnecessary spending, including avoidable readmissions, emergency department utilization, and duplicative laboratory and imaging studies. Approximately 23% of patients admitted to SNFs are readmitted to acute care hospitals within 30 days, and these patients have increased mortality rates in risk-adjusted analyses. ^3,4

Compounding the magnitude of this risk and vulnerability is the significant growth in the number of patients discharged to SNFs over the past 30 years. In 2013, more than 20% of Medicare patients discharged from acute care hospitals were destined for SNFs.^5,6 Paradoxically, despite the increasing need for SNF providers, there is a shortage of clinicians with training in geriatrics or nursing home care.⁷ The result is a growing need to identify organizational systems to optimize physician practice in these settings, enhance quality of care, especially around transitions, and increase educational training opportunities in SNFs for future practitioners.

Many SNFs today are staffed by physicians and other licensed clinicians whose exclusive practice location is the nursing facility or possibly several such facilities. This prevailing model of care can isolate the physicians, depriving them of interaction with clinicians in other specialties, and can contribute to burnout.⁸ This model does not lend itself to academic scholarship, quality improvement (QI), and student or resident training, as each of these endeavors depends on interprofessional collaboration as well as access to an academic medical center with additional resources.⁹

Few studies have described innovative hospitalist rotation models from acute to subacute care. The Cleveland Clinic implemented the Connected Care model where hospital-employed physicians and advanced practice professionals integrated into postacute care and reduced the 30-day hospital readmission rate from SNFs from 28% to 22%.¹⁰ Goth and colleagues performed a comparative effectiveness trial between a postacute care hospitalist (PACH) model and a community-based physician model of nursing home care. They found that the institution of a PACH model in a nursing home was associated with a significant increase in laboratory costs, nonsignificant reduction in medication errors and pharmacy costs, and no improvement in fall rates.¹¹ The conclusion was that the PACH model may lead to greater clinician involvement and that the potential decrease in pharmacy costs and medications errors may offset the costs associated with additional laboratory testing. Overall, there has been a lack of studies on the impact of these hospitalist rotation models from acute to subacute care on educational programs, QI activities, and the interprofessional environment.

To achieve a system in which physicians in a SNF can excel in these areas, Veterans Affairs Boston Healthcare System (VABHS) adopted a staffing model in which academic hospitalist physicians rotate between the inpatient hospital and subacute settings. This report describes the model structure, the varying roles of the physicians, and early indicators of its positive effects on educational programs, QI activities, and the interprofessional environment.

Methods

The VABHS consists of a 159-bed acute care hospital in West Roxbury, Massachusetts; and a 110-bed SNF in Brockton, Massachusetts, with 3 units: a 65-bed transitional care unit (TCU), a 30-bed long-term care unit, and a 15-bed palliative care/hospice unit. The majority of patients admitted to the SNF are transferred from the acute care hospital in West Roxbury and other regional hospitals. Prior to 2015, the TCU was staffed with full-time clinicians who exclusively practiced in the SNF.

In the new staffing model, 6 hospitalist physicians divide their clinical time between the acute care hospital’s inpatient medical service and the TCU. The hospitalists come from varied backgrounds in terms of years in practice and advanced training (Table 1). 

The amount of nonclinical (protected) time and clinical time on the acute inpatient service and the TCU varies for each physician. For example, a physician serves as principal investigator for several major research grants and has a hospital-wide administrative leadership role; as a result, the principal investigator has fewer months of clinical responsibility. Physicians are expected to use the protected time for scholarship, educational program development and teaching, QI, and administrative responsibilities. The VABHS leadership determines the amount of protected time based on individualized benchmarks for research, education, and administrative responsibilities that follow VA national and local institutional guidelines. These metrics and time allocations are negotiated at the time of recruitment and then are reviewed annually.

The TCU also is staffed with 4 full-time clinicians (2 physicians and 2 physician assistants) who provide additional continuity of care. The new hospitalist staffing model only required an approximate 10% increase in TCU clinical staffing full-time equivalents. Patients and admissions are divided equally among clinicians on service (census per clinician 12-15 patients), with redistribution of patients at times of transition from clinical to nonclinical time. Blocks of clinical time are scheduled for greater than 2 weeks at a time to preserve continuity. In addition, the new staffing model allocates assignment of clinical responsibilities that allows for clinicians to take leave without resultant shortages in clinical coverage.

To facilitate communication among physicians serving in the acute inpatient facility and the TCU, leaders of both of these programs meet monthly and ad hoc to review the transitions of care between the 2 settings. The description of this model and its assessment have been reviewed and deemed exempt from oversight by the VA Boston Healthcare System Research and Development Committee.

Results

Since the implementation of this staffing model in 2015, the system has grown considerably in the breadth and depth of educational programming, QI, and systems redesign in the TCU and, more broadly, in the SNF. The TCU, which previously had limited training opportunities, has experienced marked expansion of educational offerings. It is now a site for core general medicine rotations for first-year psychiatry residents and physician assistant students. The TCU also has expanded as a clinical site for transitions-in-care internal medicine resident curricula and electives, as well as a clinical site for a geriatrics fellowship.

A hospitalist developed and implemented a 4-week interprofessional curriculum for all clinical trainees and students, which occurs continuously. The curriculum includes a monthly academic conference and 12 didactic lectures and is taught by 16 interprofessional faculty from the TCU and the Palliative Care/Hospice Unit, including medicine, geriatric and palliative care physicians, physician assistants, social workers, physical and occupational therapists, pharmacists, and a geriatric psychologist. The goal of the curriculum is to provide learners the knowledge, attitudes, and skills necessary to perform effective, efficient, and safe transfers between clinical settings as well as education in transitional care. In addition, using a team of interprofessional faculty, the curriculum develops the interprofessional competencies of teamwork and communication. The curriculum also has provided a significant opportunity for interprofessional collaboration among faculty who have volunteered their teaching time in the development and teaching of the curriculum, with potential for improved clinical staff knowledge of other disciplines.

Quality improvement and system redesign projects in care transitions also have expanded (Table 2). 

Early assessment indicates that the new staffing model is having positive effects on the clinical environment of the TCU. A survey was conducted of a convenience sample of all physicians, nurse managers, social workers, and other members of the clinical team in the TCU (N=24)(Table 3), with response categories ranging on a Likert scale from 1 (very negative) to 5 (very positive). 

Although not rigorously analyzed using qualitative research methods, comments from respondents have consistently indicated that this staffing model increases the transfer of clinical and logistical knowledge among staff members working in the acute inpatient facility and the TCU. 

Discussion

With greater numbers of increasingly complex patients transitioning from the hospital to SNF, health care systems need to expand the capacity of their skilled nursing systems, not only to provide clinical care, but also to support QI and medical education. The VABHS developed a physician staffing model with the goal of enriching physician practice and enhancing QI and educational opportunities in its SNF. The model offers an opportunity to improve transitions in care as physicians gain a greater knowledge of both the hospital and subacute clinical settings. This hospitalist rotation model may improve the knowledge necessary for caring for patients moving across care settings, as well as improve communication between settings. It also has served as a foundation for systematic innovation in QI and education at this institution. Clinical staff in the transitional care setting have reported positive effects of this model on clinical skills and patient care, educational opportunities, as well as a desire for replication in other health care systems.

The potential generalizability of this model requires careful consideration. The VABHS is a tertiary care integrated health care system, enabling physicians to work in multiple clinical settings. Other settings may not have the staffing or clinical volume to sustain such a model. In addition, this model may increase discontinuity in patient care as hospitalists move between acute and subacute settings and nonclinical roles. This loss of continuity may be a greater concern in the SNF setting, as the inpatient hospitalist model generally involves high provider turnover as shift work. Our survey included nurse managers, and not floor nurses due to survey administration limitations, and feedback may not have captured a comprehensive view from CLC staff. Moreover, some of the perceived positive impacts also may be related to professional and personal attributes of the physicians rather than the actual model of care. In addition, the survey response rate was 86%. However, the nature of the improvement work (focused on care transitions) and educational opportunities (interprofessional care) would likely not occur had the physicians been based in one clinical setting.

Other new physician staffing models have been designed to improve the continuity between the hospital, subacute, and outpatient settings. For example, the University of Chicago Comprehensive Care model pairs patients with trained hospitalists who provide both inpatient and outpatient care, thereby optimizing continuity between these settings.¹⁴ At CareMore Health System, high-risk patients also are paired with hospitalists, referred to as “extensivists,” who lead care teams that follow patients between settings and provide acute, postacute, and outpatient care.¹⁵ In these models, a single physician takes responsibility for the patient throughout transitions of care and through various care settings. Both models have shown reduction in hospital readmissions. One concern with such models is that the treatment teams need to coexist in the various settings of care, and the ability to impact and create systematic change within each environment is limited. This may limit QI, educational opportunities, and system level impact within each environment of care.

In comparison, the “transitionalist” model proposed here features hospitalist physicians rotating between the acute inpatient hospital and subacute care with dedicated time in each environment. This innovative organizational structure may enhance physician practice and enrich QI and educational opportunities in SNFs. Further evaluation will include the impact on quality metrics of patient care and patient satisfaction, as this model has the potential to influence quality, cost, and overall health outcomes.

Acknowledgments
We would like to thank Shivani Jindal, Matthew Russell, Matthew Ronan, Juman Hijab, Wei Shen, Sandra Vilbrun-Bruno, and Jack Earnshaw for their significant contributions to this staffing model. We would also like to thank Paul Conlin, Jay Orlander, and the leadership team of Veterans Affairs Boston Healthcare System for supporting this staffing model.

ABSTRACT

The morbidity and mortality after hip fracture in the elderly are influenced by non-modifiable comorbidities. Time-to-surgery is a modifiable factor that may play a role in postoperative morbidity. This study investigates the outcomes and complications in the elderly hip fracture surgery as a function of time-to-surgery.

Using the American College of Surgeons-National Surgical Quality Improvement Program data from 2011 to 2012, a study population was generated using the Current Procedural Terminology codes for percutaneous or open treatment of femoral neck fractures (27235, 27236) and fixation with a screw and side plate or intramedullary fixation (27244, 27245) for peritrochanteric fractures. Three time-to-surgery groups (<24 hours to surgical intervention, 24-48 hours, and >48 hours) were created and matched for surgery type, sex, age, and American Society of Anesthesiologists class. Time-to-surgery was then studied for its effect on the post-surgical outcomes using the adjusted regression modeling.

A study population of 6036 hip fractures was created, and 2012 patients were assigned to each matched time-to-surgery group. The unadjusted models showed that the earlier surgical intervention groups (<24 hours and 24-48 hours) exhibited a lower overall complication rate (P = .034) compared with the group waiting for surgery >48 hours. The unadjusted mortality rates increased with delay to surgical intervention (P = .039). Time-to-surgery caused no effect on the return to the operating room rate (P = .554) nor readmission rate (P = .285). Compared with other time-to-surgeries, the time-to-surgery of >48 hours was associated with prolonged total hospital length of stay (10.9 days) (P < .001) and a longer surgery-to-discharge time (hazard ratio, 95% confidence interval: 0.74, 0.69-0.79) (P < .001). Adjusted analyses showed no time-to-surgery related difference in complications (P = .143) but presented an increase in the total length of stay (P < .001) and surgery-to-discharge time (P < .001).

Timeliness of surgical intervention in a comorbidity-adjusted population of elderly hip fracture patients causes no effect on the overall complications, readmissions, nor 30-day mortality. However, time-to-surgery of >48 hours is associated with costly increase in the total length of stay, including an increased post-surgery-to-discharge time.

Continue to: Despite the best efforts to optimize surgical care...

Despite the best efforts to optimize surgical care and postoperative rehabilitation following hip fracture, elderly patients feature alarmingly high in-hospital and 1-year mortality rates of 4.35% to 9.2%^1-4 and 36%,⁵ respectively. Those who survive are unlikely to return to independent living, with only 17% of the patients following hip fracture being able to walk independently 6 months postoperatively, and 12% being able to climb stairs⁶. Possibly, these poor outcomes reflect a preoperative medical comorbidity burden rather than a measure of medical or surgical quality. Given the absence of consensus regarding optimal time-to-surgery, treating physicians often opt to delay surgical intervention for the purposes of medically optimizing highly comorbid patients without significant data to suggest clinical benefit of such practice.

Numerous investigators have attempted to identify the modifiable risk factors for complication after surgical care of elderly hip fracture patients. However, consensus guidelines of care are missing. This condition is largely due to the difficulties in effectively modifying preoperative demographic and medical comorbidities on a semi-urgent basis. However, timeliness to surgery is one area for study that the care team can affect. Although time-to-surgery is dependent on multiple factors, including time of presentation, day of week of admission, difficulties with scheduling, and administrative delays, the care team plays a role in hastening or retarding time-to-surgery. Several studies have considered various time cut-offs (24, 48, 72, and 120 hours) to define early intervention, but none have defined a specific role for early or delayed surgery. Several investigators have discovered a positive association between delayed time-to-surgery and mortality;^4,8-14 however, the most rigorously conducted studies that stringently control for preoperative comorbidities and demographics conclude that variance in time-to-surgery causes no effect on the in-hospital or 1-year mortality risk.^1-3,15-18  

Other investigators have shown that with early surgical intervention for hip fracture, patients experience shorter hospital stays,^{1,3,16,17,19-22} less days in pain,¹⁹ decreased risk of decubitus ulcers,^15,17,19,22 and an increased likelihood of independence following fracture,^22-25 regardless of preoperative medical status. Despite this evidence of improved outcomes with early surgery, 40% to 54% of hip fracture patients in the United States experience surgical delays of more than 24 to 48 hours. Additionally, with the recent (2013) national estimates of cost per day spent in the hospital falling between $1791 to $2289,²⁶ minimizing the days spent in the hospital would likely lead to significant cost-savings, presuming no adverse effect on health-related outcomes. To this end, we hypothesize that the value (outcomes per associated cost)⁷ of hip fracture surgical care can be positively influenced by minimizing surgical wait-times. We assessed the effect of early surgical intervention, within 24 or 48 hours of presentation, on 30-day mortality, postoperative morbidity, hospital length of stay, and readmission rates in a comorbidity-adjusted population from a nationally representative cohort.

Continue to: METHODS AND MATERIALS...

METHODS AND MATERIALS 

This study used the data from the American College of Surgeon-National Surgical Quality Improvement Program (ACS-NSQIP) database. With over 258 participating hospitals, this database has been widely used to identify national trends in various surgical specialties.^27-34 The database includes information from participants in 43 states with hospitals ranging from rural community hospitals to large academic centers. Each site employs surgical clinical reviewers who are rigorously trained to collect data through chart review and discussion with the treating surgeon and/or patient,³⁵ allowing for the use of robust and quality data with proven inter-rater reliability.^36,37

Using the 2011 to 2012 NSQIP database, we used primary Current Procedural Terminology codes to identify all patients who underwent percutaneous (27235) or open (27236) fixation of femoral neck fractures; and fixation with a screw and side plate (27244) or intramedullary fixation (27245) for peritrochanteric fractures. The sample was divided into 3 time-to-surgery groups (<24 hours from presentation to surgery, 24-48 hours, and >48 hours) which were matched for fracture type (femoral neck or peritrochanteric), sex, age (under 75 years or ≥75 years), and American Society of Anesthesiologists (ASA) class used as a surrogate for severity of medical infirmary. The subjects were randomly matched 1:1:1 to create 3 statistically equivalent time-to-surgery groups using Proc SurveySelect (SAS version 9.2, SAS Institute).

Generalized linear models using logit link function for binary variables and identity link function for normally distributed characteristics were used to compare the 3 time-to-surgery groups. Descriptive statistics are presented as counts and percentages or least-square means with standard deviations. Preoperative lab values that were not normally distributed were log transformed and presented in their original scales with median values and 25th to 75th percentiles. Outcomes were similarly modeled.  

Total hospital stay was modeled with a negative binomial distribution. Proportional hazards models were used to model the time from operating room (OR) to discharge, censoring patients who died before discharge, with results presented as hazard ratios (HR) and 95% confidence intervals (CI) (Figure). The assumption of the proportional hazards was tested using a Wald test. Using this model, a HR of <1 denotes a longer postoperative hospital stay, as a longer hospital stay decreases the “risk” for discharge.

All models were adjusted for confounders, including race, body mass index (BMI), hypertension, chronic obstructive pulmonary disease, cancer, bleeding disorders, transfusion within 72 hours before surgery, preoperative levels of creatinine, platelet count, white blood cells (WBCs), hematocrit anesthesia type, and wound infection. These covariates were selected based upon their observed relationship to the studied outcomes and time-to-surgery groups, and were evaluated across the models for all outcomes for consistency and clarity. All statistical analyses were run at a type I error rate of 5% and performed in SAS version 9.2 software.

Continue to: RESULTS...

RESULTS

A study population of 6036 hip fractures was identified and divided into 3 groups of 2012 subjects each based upon time-to-surgery. The groups were successfully matched for surgery type, age (≥75 years old), gender, and ASA class. In each group, 594 of the 2012 (29.5%) patients were male, 1525 (75.8%) were ≥75 years of age, 9 (.5%) were ASA Class I, 269 (13.4%) were ASA Class II, 1424 (70.8%) were ASA class III, and 309 (15.4%) were ASA class IV.

Significant differences in preoperative comorbidity burden and preoperative lab values were identified between the 3 cohorts. Increased time-to-surgery was associated with differences in race (P < .001), elevated BMI (P = .010), higher rates of congestive heart failure (P < .001), hypertension medication (P = .020), bleeding disorders (P < .001), blood transfusion within 72 hours of surgery (P < .001), and systemic sepsis (P = .001). Delay to surgery was also associated with lower preoperative sodium (P = .005), blood urea nitrogen (P = .013), serum WBC (P < .001), hematocrit (P < .001), and platelets (P < .001) (Table 1).

The unadjusted analyses revealed no association between time-to-surgery and return to OR (P = .554) nor readmission (P = .285). However, increasing time-to-surgery was associated with an increase in overall complications (P = .034), total length of hospital stay (P < .001), and 30-day mortality (P = .039) (Table 2).

Table 2. Estimated Event Rates from Matched Cohorts (Unadjusted)

Abbreviation: OR, operating room.

The adjusted analysis controlling for preoperative demographic and comorbidity variables revealed trends toward the increased overall complications and 30-day mortality with increased time-to-surgery; these trends showed no statistical significance (P = .143 and P = .08). No statistical relationship was observed between return to OR nor readmission and time-to-surgery. Increasing time-to-surgery remained significantly associated with the increased total length of hospital stay (P < .001). The adjusted analysis also revealed that the delay of >48 hours in time-to-surgery resulted in a longer surgery-to-discharge time (P < .001) (Table 3). No evidence of violation of the proportional hazards assumption was observed in the unadjusted nor adjusted clustered proportional hazards models (Wald test, P = .27 and P = .25, respectively).

Table 3. Estimated Event Rates from Matched Cohorts (Adjusted^a)

^aModel adjusted for race, hypertension medication, cancer, bleeding disorders, transfusion within 72 hours before surgery, emergency status, wound infection, anesthesia type (general), body mass index (18.5-25), history of chronic obstructive pulmonary disease, and preoperative levels of creatinine, platelet count, white blood cell count, and hematocrit.

Continue to: DISCUSSION...

DISCUSSION

Previous research has demonstrated an association between age,^3,4,25 comorbidity burden,^1,3,25 gender,^3,4 and ASA class^4,18,21 with outcomes following hip fractures and serves as the basis of our matched analysis statistical methodology in assessing the effect of time-to-surgery on the outcome following hip fracture surgery. Prior investigators have also established the positive correlation between increased preoperative comorbidity burden and delay in time-to-surgery.^10,15 This finding was confirmed in our unadjusted comparison of 3 time-to-surgery groups. However, prior investigations have not established a clear association between time-to-surgical intervention and postoperative morbidity and mortality.^{1,15,16,18,20,38} This study utilized a nationally representative dataset known for its data integrity and from which 6036 patients with surgically treated hip fractures, matched for surgery type, age, gender, and ASA class (a surrogate for severity of medical infirmary), were studied using adjusted regression modeling to afford an isolated statistical assessment of the effect of time-to-surgery on outcomes following hip fracture surgery.

Despite a large sample size and rigorous statistical methodology, for many outcome measures, our results show no support for the early or late operative intervention following hip fracture. We found no difference in 30-day mortality, readmission rate, nor total complication rate between the 3 time-to-surgery cohorts. This result indicates that the care of elderly patients following hip fracture is inherently complicated and that perioperative complication risk is probably only modestly modifiable by best medical practices, including optimizing time from clinical presentation to surgery.

As expected, patients who experienced longer delays from presentation to surgery were on average, more comorbid and more likely to yield abnormal preoperative lab values. However, in the adjusted analysis, delay in time-to-surgery, presumably for medical management, was not found to be associated with improved outcomes. In the same adjusted analysis, we uniquely identified that in the patients whose surgeries were delayed for more than 48 hours, the time from surgery-to-discharge was significantly increased. As a result, these patients spent extra days in the hospital both preoperatively and postoperatively, but without any corollary improvement in the outcomes.

Continue to: Recent estimates of the cost of hospital admission...

Recent estimates of the cost of hospital admission is approximated nationally at $2000/day.²⁶ Although our data fail to support the formal cost-analysis of the effect of time-to-surgery in hip fracture care, a simple value-based analysis indicates that quality is preserved (no difference in outcome), whereas costly hospital days are eliminated with earlier surgery. The value in elderly hip fracture care. defined as the outcomes relative to the costs,⁷ is ultimately optimized by earlier time-to-surgery.

Although using a large, multi-institutional database is advantageous for finding population-based trends that are representative of a large cohort, using the ACS-NSQIP database features its limitations. Our analysis was limited to the defined scope of NSQIP and nature of the injury, whereas root cause for delay was not available for study. We were unable to identify which patients were delayed for administrative reasons or surgical convenience and which were delayed for medical optimization. Participation in the ACS-NSQIP database is voluntary, and no randomized hospital sampling was conducted. Participating hospitals were de-identified in the database. As expected, we were unable to identify the specific institution-based hip fracture protocols that may affect the outcomes following treatment for these fractures. Further, socioeconomic information and payer-status are unavailable for the study. Additionally, observations are limited to 30 days postoperative, and we cannot comment on longer-term outcomes. Finally, discharge disposition and functional outcome data are not represented, and we were unable to correlate time-to-surgery and functional recovery. However, previous studies have established that delay in time-to-surgery following hip fractures is negatively correlated with functional outcomes.^22-25

Nevertheless, the ACS-NSQIP database remains one of the largest American surgical databases available, and includes care centers from nearly every state with variable demographics including rural, urban, and academic centers. The ACS performs broad-based inter-rater reliability audits on every participating site and has found an overall disagreement rate of only 1.8%. As such, although discrepancies exist between the complete patient chart and the data entered, the data found in the ACS-NSQIP database are reliable and considered a valid source of study.^34,35 The large sample size, quality of data collection, wide geographic representation, and varied hospital types within the dataset possibly make our findings relevant in the majority of American healthcare settings.

CONCLUSION

This study demonstrates an associated increased length of hospital stay, including the increased time from surgery-to-discharge, in patients with hip fractures whose surgical intervention is delayed for >48 hours after presentation. Given the prior evidence that early surgical intervention improves the functional outcomes and the current evidence that surgical delay for any cause increases costly hospital length of stay without corollary improvement in the outcomes, a value-based assessment of hip fracture care argues for early surgical intervention whenever possible. Our findings should inform physician, institution, and policy maker value-based decision making regarding the best practices in geriatric hip fracture care.

ABSTRACT

The morbidity and mortality after hip fracture in the elderly are influenced by non-modifiable comorbidities. Time-to-surgery is a modifiable factor that may play a role in postoperative morbidity. This study investigates the outcomes and complications in the elderly hip fracture surgery as a function of time-to-surgery.

Using the American College of Surgeons-National Surgical Quality Improvement Program data from 2011 to 2012, a study population was generated using the Current Procedural Terminology codes for percutaneous or open treatment of femoral neck fractures (27235, 27236) and fixation with a screw and side plate or intramedullary fixation (27244, 27245) for peritrochanteric fractures. Three time-to-surgery groups (<24 hours to surgical intervention, 24-48 hours, and >48 hours) were created and matched for surgery type, sex, age, and American Society of Anesthesiologists class. Time-to-surgery was then studied for its effect on the post-surgical outcomes using the adjusted regression modeling.

A study population of 6036 hip fractures was created, and 2012 patients were assigned to each matched time-to-surgery group. The unadjusted models showed that the earlier surgical intervention groups (<24 hours and 24-48 hours) exhibited a lower overall complication rate (P = .034) compared with the group waiting for surgery >48 hours. The unadjusted mortality rates increased with delay to surgical intervention (P = .039). Time-to-surgery caused no effect on the return to the operating room rate (P = .554) nor readmission rate (P = .285). Compared with other time-to-surgeries, the time-to-surgery of >48 hours was associated with prolonged total hospital length of stay (10.9 days) (P < .001) and a longer surgery-to-discharge time (hazard ratio, 95% confidence interval: 0.74, 0.69-0.79) (P < .001). Adjusted analyses showed no time-to-surgery related difference in complications (P = .143) but presented an increase in the total length of stay (P < .001) and surgery-to-discharge time (P < .001).

Timeliness of surgical intervention in a comorbidity-adjusted population of elderly hip fracture patients causes no effect on the overall complications, readmissions, nor 30-day mortality. However, time-to-surgery of >48 hours is associated with costly increase in the total length of stay, including an increased post-surgery-to-discharge time.

Continue to: Despite the best efforts to optimize surgical care...

Despite the best efforts to optimize surgical care and postoperative rehabilitation following hip fracture, elderly patients feature alarmingly high in-hospital and 1-year mortality rates of 4.35% to 9.2%^1-4 and 36%,⁵ respectively. Those who survive are unlikely to return to independent living, with only 17% of the patients following hip fracture being able to walk independently 6 months postoperatively, and 12% being able to climb stairs⁶. Possibly, these poor outcomes reflect a preoperative medical comorbidity burden rather than a measure of medical or surgical quality. Given the absence of consensus regarding optimal time-to-surgery, treating physicians often opt to delay surgical intervention for the purposes of medically optimizing highly comorbid patients without significant data to suggest clinical benefit of such practice.

Numerous investigators have attempted to identify the modifiable risk factors for complication after surgical care of elderly hip fracture patients. However, consensus guidelines of care are missing. This condition is largely due to the difficulties in effectively modifying preoperative demographic and medical comorbidities on a semi-urgent basis. However, timeliness to surgery is one area for study that the care team can affect. Although time-to-surgery is dependent on multiple factors, including time of presentation, day of week of admission, difficulties with scheduling, and administrative delays, the care team plays a role in hastening or retarding time-to-surgery. Several studies have considered various time cut-offs (24, 48, 72, and 120 hours) to define early intervention, but none have defined a specific role for early or delayed surgery. Several investigators have discovered a positive association between delayed time-to-surgery and mortality;^4,8-14 however, the most rigorously conducted studies that stringently control for preoperative comorbidities and demographics conclude that variance in time-to-surgery causes no effect on the in-hospital or 1-year mortality risk.^1-3,15-18  

Other investigators have shown that with early surgical intervention for hip fracture, patients experience shorter hospital stays,^{1,3,16,17,19-22} less days in pain,¹⁹ decreased risk of decubitus ulcers,^15,17,19,22 and an increased likelihood of independence following fracture,^22-25 regardless of preoperative medical status. Despite this evidence of improved outcomes with early surgery, 40% to 54% of hip fracture patients in the United States experience surgical delays of more than 24 to 48 hours. Additionally, with the recent (2013) national estimates of cost per day spent in the hospital falling between $1791 to $2289,²⁶ minimizing the days spent in the hospital would likely lead to significant cost-savings, presuming no adverse effect on health-related outcomes. To this end, we hypothesize that the value (outcomes per associated cost)⁷ of hip fracture surgical care can be positively influenced by minimizing surgical wait-times. We assessed the effect of early surgical intervention, within 24 or 48 hours of presentation, on 30-day mortality, postoperative morbidity, hospital length of stay, and readmission rates in a comorbidity-adjusted population from a nationally representative cohort.

Continue to: METHODS AND MATERIALS...

METHODS AND MATERIALS 

This study used the data from the American College of Surgeon-National Surgical Quality Improvement Program (ACS-NSQIP) database. With over 258 participating hospitals, this database has been widely used to identify national trends in various surgical specialties.^27-34 The database includes information from participants in 43 states with hospitals ranging from rural community hospitals to large academic centers. Each site employs surgical clinical reviewers who are rigorously trained to collect data through chart review and discussion with the treating surgeon and/or patient,³⁵ allowing for the use of robust and quality data with proven inter-rater reliability.^36,37

Using the 2011 to 2012 NSQIP database, we used primary Current Procedural Terminology codes to identify all patients who underwent percutaneous (27235) or open (27236) fixation of femoral neck fractures; and fixation with a screw and side plate (27244) or intramedullary fixation (27245) for peritrochanteric fractures. The sample was divided into 3 time-to-surgery groups (<24 hours from presentation to surgery, 24-48 hours, and >48 hours) which were matched for fracture type (femoral neck or peritrochanteric), sex, age (under 75 years or ≥75 years), and American Society of Anesthesiologists (ASA) class used as a surrogate for severity of medical infirmary. The subjects were randomly matched 1:1:1 to create 3 statistically equivalent time-to-surgery groups using Proc SurveySelect (SAS version 9.2, SAS Institute).

Generalized linear models using logit link function for binary variables and identity link function for normally distributed characteristics were used to compare the 3 time-to-surgery groups. Descriptive statistics are presented as counts and percentages or least-square means with standard deviations. Preoperative lab values that were not normally distributed were log transformed and presented in their original scales with median values and 25th to 75th percentiles. Outcomes were similarly modeled.  

Total hospital stay was modeled with a negative binomial distribution. Proportional hazards models were used to model the time from operating room (OR) to discharge, censoring patients who died before discharge, with results presented as hazard ratios (HR) and 95% confidence intervals (CI) (Figure). The assumption of the proportional hazards was tested using a Wald test. Using this model, a HR of <1 denotes a longer postoperative hospital stay, as a longer hospital stay decreases the “risk” for discharge.

All models were adjusted for confounders, including race, body mass index (BMI), hypertension, chronic obstructive pulmonary disease, cancer, bleeding disorders, transfusion within 72 hours before surgery, preoperative levels of creatinine, platelet count, white blood cells (WBCs), hematocrit anesthesia type, and wound infection. These covariates were selected based upon their observed relationship to the studied outcomes and time-to-surgery groups, and were evaluated across the models for all outcomes for consistency and clarity. All statistical analyses were run at a type I error rate of 5% and performed in SAS version 9.2 software.

Continue to: RESULTS...

RESULTS

A study population of 6036 hip fractures was identified and divided into 3 groups of 2012 subjects each based upon time-to-surgery. The groups were successfully matched for surgery type, age (≥75 years old), gender, and ASA class. In each group, 594 of the 2012 (29.5%) patients were male, 1525 (75.8%) were ≥75 years of age, 9 (.5%) were ASA Class I, 269 (13.4%) were ASA Class II, 1424 (70.8%) were ASA class III, and 309 (15.4%) were ASA class IV.

Significant differences in preoperative comorbidity burden and preoperative lab values were identified between the 3 cohorts. Increased time-to-surgery was associated with differences in race (P < .001), elevated BMI (P = .010), higher rates of congestive heart failure (P < .001), hypertension medication (P = .020), bleeding disorders (P < .001), blood transfusion within 72 hours of surgery (P < .001), and systemic sepsis (P = .001). Delay to surgery was also associated with lower preoperative sodium (P = .005), blood urea nitrogen (P = .013), serum WBC (P < .001), hematocrit (P < .001), and platelets (P < .001) (Table 1).

The unadjusted analyses revealed no association between time-to-surgery and return to OR (P = .554) nor readmission (P = .285). However, increasing time-to-surgery was associated with an increase in overall complications (P = .034), total length of hospital stay (P < .001), and 30-day mortality (P = .039) (Table 2).

Table 2. Estimated Event Rates from Matched Cohorts (Unadjusted)

Abbreviation: OR, operating room.

The adjusted analysis controlling for preoperative demographic and comorbidity variables revealed trends toward the increased overall complications and 30-day mortality with increased time-to-surgery; these trends showed no statistical significance (P = .143 and P = .08). No statistical relationship was observed between return to OR nor readmission and time-to-surgery. Increasing time-to-surgery remained significantly associated with the increased total length of hospital stay (P < .001). The adjusted analysis also revealed that the delay of >48 hours in time-to-surgery resulted in a longer surgery-to-discharge time (P < .001) (Table 3). No evidence of violation of the proportional hazards assumption was observed in the unadjusted nor adjusted clustered proportional hazards models (Wald test, P = .27 and P = .25, respectively).

Table 3. Estimated Event Rates from Matched Cohorts (Adjusted^a)

^aModel adjusted for race, hypertension medication, cancer, bleeding disorders, transfusion within 72 hours before surgery, emergency status, wound infection, anesthesia type (general), body mass index (18.5-25), history of chronic obstructive pulmonary disease, and preoperative levels of creatinine, platelet count, white blood cell count, and hematocrit.

Continue to: DISCUSSION...

DISCUSSION

Previous research has demonstrated an association between age,^3,4,25 comorbidity burden,^1,3,25 gender,^3,4 and ASA class^4,18,21 with outcomes following hip fractures and serves as the basis of our matched analysis statistical methodology in assessing the effect of time-to-surgery on the outcome following hip fracture surgery. Prior investigators have also established the positive correlation between increased preoperative comorbidity burden and delay in time-to-surgery.^10,15 This finding was confirmed in our unadjusted comparison of 3 time-to-surgery groups. However, prior investigations have not established a clear association between time-to-surgical intervention and postoperative morbidity and mortality.^{1,15,16,18,20,38} This study utilized a nationally representative dataset known for its data integrity and from which 6036 patients with surgically treated hip fractures, matched for surgery type, age, gender, and ASA class (a surrogate for severity of medical infirmary), were studied using adjusted regression modeling to afford an isolated statistical assessment of the effect of time-to-surgery on outcomes following hip fracture surgery.

Despite a large sample size and rigorous statistical methodology, for many outcome measures, our results show no support for the early or late operative intervention following hip fracture. We found no difference in 30-day mortality, readmission rate, nor total complication rate between the 3 time-to-surgery cohorts. This result indicates that the care of elderly patients following hip fracture is inherently complicated and that perioperative complication risk is probably only modestly modifiable by best medical practices, including optimizing time from clinical presentation to surgery.

As expected, patients who experienced longer delays from presentation to surgery were on average, more comorbid and more likely to yield abnormal preoperative lab values. However, in the adjusted analysis, delay in time-to-surgery, presumably for medical management, was not found to be associated with improved outcomes. In the same adjusted analysis, we uniquely identified that in the patients whose surgeries were delayed for more than 48 hours, the time from surgery-to-discharge was significantly increased. As a result, these patients spent extra days in the hospital both preoperatively and postoperatively, but without any corollary improvement in the outcomes.

Continue to: Recent estimates of the cost of hospital admission...

Recent estimates of the cost of hospital admission is approximated nationally at $2000/day.²⁶ Although our data fail to support the formal cost-analysis of the effect of time-to-surgery in hip fracture care, a simple value-based analysis indicates that quality is preserved (no difference in outcome), whereas costly hospital days are eliminated with earlier surgery. The value in elderly hip fracture care. defined as the outcomes relative to the costs,⁷ is ultimately optimized by earlier time-to-surgery.

Although using a large, multi-institutional database is advantageous for finding population-based trends that are representative of a large cohort, using the ACS-NSQIP database features its limitations. Our analysis was limited to the defined scope of NSQIP and nature of the injury, whereas root cause for delay was not available for study. We were unable to identify which patients were delayed for administrative reasons or surgical convenience and which were delayed for medical optimization. Participation in the ACS-NSQIP database is voluntary, and no randomized hospital sampling was conducted. Participating hospitals were de-identified in the database. As expected, we were unable to identify the specific institution-based hip fracture protocols that may affect the outcomes following treatment for these fractures. Further, socioeconomic information and payer-status are unavailable for the study. Additionally, observations are limited to 30 days postoperative, and we cannot comment on longer-term outcomes. Finally, discharge disposition and functional outcome data are not represented, and we were unable to correlate time-to-surgery and functional recovery. However, previous studies have established that delay in time-to-surgery following hip fractures is negatively correlated with functional outcomes.^22-25

Nevertheless, the ACS-NSQIP database remains one of the largest American surgical databases available, and includes care centers from nearly every state with variable demographics including rural, urban, and academic centers. The ACS performs broad-based inter-rater reliability audits on every participating site and has found an overall disagreement rate of only 1.8%. As such, although discrepancies exist between the complete patient chart and the data entered, the data found in the ACS-NSQIP database are reliable and considered a valid source of study.^34,35 The large sample size, quality of data collection, wide geographic representation, and varied hospital types within the dataset possibly make our findings relevant in the majority of American healthcare settings.

CONCLUSION

This study demonstrates an associated increased length of hospital stay, including the increased time from surgery-to-discharge, in patients with hip fractures whose surgical intervention is delayed for >48 hours after presentation. Given the prior evidence that early surgical intervention improves the functional outcomes and the current evidence that surgical delay for any cause increases costly hospital length of stay without corollary improvement in the outcomes, a value-based assessment of hip fracture care argues for early surgical intervention whenever possible. Our findings should inform physician, institution, and policy maker value-based decision making regarding the best practices in geriatric hip fracture care.

ABSTRACT

The morbidity and mortality after hip fracture in the elderly are influenced by non-modifiable comorbidities. Time-to-surgery is a modifiable factor that may play a role in postoperative morbidity. This study investigates the outcomes and complications in the elderly hip fracture surgery as a function of time-to-surgery.

Using the American College of Surgeons-National Surgical Quality Improvement Program data from 2011 to 2012, a study population was generated using the Current Procedural Terminology codes for percutaneous or open treatment of femoral neck fractures (27235, 27236) and fixation with a screw and side plate or intramedullary fixation (27244, 27245) for peritrochanteric fractures. Three time-to-surgery groups (<24 hours to surgical intervention, 24-48 hours, and >48 hours) were created and matched for surgery type, sex, age, and American Society of Anesthesiologists class. Time-to-surgery was then studied for its effect on the post-surgical outcomes using the adjusted regression modeling.

A study population of 6036 hip fractures was created, and 2012 patients were assigned to each matched time-to-surgery group. The unadjusted models showed that the earlier surgical intervention groups (<24 hours and 24-48 hours) exhibited a lower overall complication rate (P = .034) compared with the group waiting for surgery >48 hours. The unadjusted mortality rates increased with delay to surgical intervention (P = .039). Time-to-surgery caused no effect on the return to the operating room rate (P = .554) nor readmission rate (P = .285). Compared with other time-to-surgeries, the time-to-surgery of >48 hours was associated with prolonged total hospital length of stay (10.9 days) (P < .001) and a longer surgery-to-discharge time (hazard ratio, 95% confidence interval: 0.74, 0.69-0.79) (P < .001). Adjusted analyses showed no time-to-surgery related difference in complications (P = .143) but presented an increase in the total length of stay (P < .001) and surgery-to-discharge time (P < .001).

Timeliness of surgical intervention in a comorbidity-adjusted population of elderly hip fracture patients causes no effect on the overall complications, readmissions, nor 30-day mortality. However, time-to-surgery of >48 hours is associated with costly increase in the total length of stay, including an increased post-surgery-to-discharge time.

Continue to: Despite the best efforts to optimize surgical care...

Despite the best efforts to optimize surgical care and postoperative rehabilitation following hip fracture, elderly patients feature alarmingly high in-hospital and 1-year mortality rates of 4.35% to 9.2%^1-4 and 36%,⁵ respectively. Those who survive are unlikely to return to independent living, with only 17% of the patients following hip fracture being able to walk independently 6 months postoperatively, and 12% being able to climb stairs⁶. Possibly, these poor outcomes reflect a preoperative medical comorbidity burden rather than a measure of medical or surgical quality. Given the absence of consensus regarding optimal time-to-surgery, treating physicians often opt to delay surgical intervention for the purposes of medically optimizing highly comorbid patients without significant data to suggest clinical benefit of such practice.

Numerous investigators have attempted to identify the modifiable risk factors for complication after surgical care of elderly hip fracture patients. However, consensus guidelines of care are missing. This condition is largely due to the difficulties in effectively modifying preoperative demographic and medical comorbidities on a semi-urgent basis. However, timeliness to surgery is one area for study that the care team can affect. Although time-to-surgery is dependent on multiple factors, including time of presentation, day of week of admission, difficulties with scheduling, and administrative delays, the care team plays a role in hastening or retarding time-to-surgery. Several studies have considered various time cut-offs (24, 48, 72, and 120 hours) to define early intervention, but none have defined a specific role for early or delayed surgery. Several investigators have discovered a positive association between delayed time-to-surgery and mortality;^4,8-14 however, the most rigorously conducted studies that stringently control for preoperative comorbidities and demographics conclude that variance in time-to-surgery causes no effect on the in-hospital or 1-year mortality risk.^1-3,15-18  

Other investigators have shown that with early surgical intervention for hip fracture, patients experience shorter hospital stays,^{1,3,16,17,19-22} less days in pain,¹⁹ decreased risk of decubitus ulcers,^15,17,19,22 and an increased likelihood of independence following fracture,^22-25 regardless of preoperative medical status. Despite this evidence of improved outcomes with early surgery, 40% to 54% of hip fracture patients in the United States experience surgical delays of more than 24 to 48 hours. Additionally, with the recent (2013) national estimates of cost per day spent in the hospital falling between $1791 to $2289,²⁶ minimizing the days spent in the hospital would likely lead to significant cost-savings, presuming no adverse effect on health-related outcomes. To this end, we hypothesize that the value (outcomes per associated cost)⁷ of hip fracture surgical care can be positively influenced by minimizing surgical wait-times. We assessed the effect of early surgical intervention, within 24 or 48 hours of presentation, on 30-day mortality, postoperative morbidity, hospital length of stay, and readmission rates in a comorbidity-adjusted population from a nationally representative cohort.

Continue to: METHODS AND MATERIALS...

METHODS AND MATERIALS 

This study used the data from the American College of Surgeon-National Surgical Quality Improvement Program (ACS-NSQIP) database. With over 258 participating hospitals, this database has been widely used to identify national trends in various surgical specialties.^27-34 The database includes information from participants in 43 states with hospitals ranging from rural community hospitals to large academic centers. Each site employs surgical clinical reviewers who are rigorously trained to collect data through chart review and discussion with the treating surgeon and/or patient,³⁵ allowing for the use of robust and quality data with proven inter-rater reliability.^36,37

Using the 2011 to 2012 NSQIP database, we used primary Current Procedural Terminology codes to identify all patients who underwent percutaneous (27235) or open (27236) fixation of femoral neck fractures; and fixation with a screw and side plate (27244) or intramedullary fixation (27245) for peritrochanteric fractures. The sample was divided into 3 time-to-surgery groups (<24 hours from presentation to surgery, 24-48 hours, and >48 hours) which were matched for fracture type (femoral neck or peritrochanteric), sex, age (under 75 years or ≥75 years), and American Society of Anesthesiologists (ASA) class used as a surrogate for severity of medical infirmary. The subjects were randomly matched 1:1:1 to create 3 statistically equivalent time-to-surgery groups using Proc SurveySelect (SAS version 9.2, SAS Institute).

Generalized linear models using logit link function for binary variables and identity link function for normally distributed characteristics were used to compare the 3 time-to-surgery groups. Descriptive statistics are presented as counts and percentages or least-square means with standard deviations. Preoperative lab values that were not normally distributed were log transformed and presented in their original scales with median values and 25th to 75th percentiles. Outcomes were similarly modeled.  

Total hospital stay was modeled with a negative binomial distribution. Proportional hazards models were used to model the time from operating room (OR) to discharge, censoring patients who died before discharge, with results presented as hazard ratios (HR) and 95% confidence intervals (CI) (Figure). The assumption of the proportional hazards was tested using a Wald test. Using this model, a HR of <1 denotes a longer postoperative hospital stay, as a longer hospital stay decreases the “risk” for discharge.

All models were adjusted for confounders, including race, body mass index (BMI), hypertension, chronic obstructive pulmonary disease, cancer, bleeding disorders, transfusion within 72 hours before surgery, preoperative levels of creatinine, platelet count, white blood cells (WBCs), hematocrit anesthesia type, and wound infection. These covariates were selected based upon their observed relationship to the studied outcomes and time-to-surgery groups, and were evaluated across the models for all outcomes for consistency and clarity. All statistical analyses were run at a type I error rate of 5% and performed in SAS version 9.2 software.

Continue to: RESULTS...

RESULTS

A study population of 6036 hip fractures was identified and divided into 3 groups of 2012 subjects each based upon time-to-surgery. The groups were successfully matched for surgery type, age (≥75 years old), gender, and ASA class. In each group, 594 of the 2012 (29.5%) patients were male, 1525 (75.8%) were ≥75 years of age, 9 (.5%) were ASA Class I, 269 (13.4%) were ASA Class II, 1424 (70.8%) were ASA class III, and 309 (15.4%) were ASA class IV.

Significant differences in preoperative comorbidity burden and preoperative lab values were identified between the 3 cohorts. Increased time-to-surgery was associated with differences in race (P < .001), elevated BMI (P = .010), higher rates of congestive heart failure (P < .001), hypertension medication (P = .020), bleeding disorders (P < .001), blood transfusion within 72 hours of surgery (P < .001), and systemic sepsis (P = .001). Delay to surgery was also associated with lower preoperative sodium (P = .005), blood urea nitrogen (P = .013), serum WBC (P < .001), hematocrit (P < .001), and platelets (P < .001) (Table 1).

The unadjusted analyses revealed no association between time-to-surgery and return to OR (P = .554) nor readmission (P = .285). However, increasing time-to-surgery was associated with an increase in overall complications (P = .034), total length of hospital stay (P < .001), and 30-day mortality (P = .039) (Table 2).

Table 2. Estimated Event Rates from Matched Cohorts (Unadjusted)

Abbreviation: OR, operating room.

The adjusted analysis controlling for preoperative demographic and comorbidity variables revealed trends toward the increased overall complications and 30-day mortality with increased time-to-surgery; these trends showed no statistical significance (P = .143 and P = .08). No statistical relationship was observed between return to OR nor readmission and time-to-surgery. Increasing time-to-surgery remained significantly associated with the increased total length of hospital stay (P < .001). The adjusted analysis also revealed that the delay of >48 hours in time-to-surgery resulted in a longer surgery-to-discharge time (P < .001) (Table 3). No evidence of violation of the proportional hazards assumption was observed in the unadjusted nor adjusted clustered proportional hazards models (Wald test, P = .27 and P = .25, respectively).

Table 3. Estimated Event Rates from Matched Cohorts (Adjusted^a)

^aModel adjusted for race, hypertension medication, cancer, bleeding disorders, transfusion within 72 hours before surgery, emergency status, wound infection, anesthesia type (general), body mass index (18.5-25), history of chronic obstructive pulmonary disease, and preoperative levels of creatinine, platelet count, white blood cell count, and hematocrit.

Continue to: DISCUSSION...

DISCUSSION

Previous research has demonstrated an association between age,^3,4,25 comorbidity burden,^1,3,25 gender,^3,4 and ASA class^4,18,21 with outcomes following hip fractures and serves as the basis of our matched analysis statistical methodology in assessing the effect of time-to-surgery on the outcome following hip fracture surgery. Prior investigators have also established the positive correlation between increased preoperative comorbidity burden and delay in time-to-surgery.^10,15 This finding was confirmed in our unadjusted comparison of 3 time-to-surgery groups. However, prior investigations have not established a clear association between time-to-surgical intervention and postoperative morbidity and mortality.^{1,15,16,18,20,38} This study utilized a nationally representative dataset known for its data integrity and from which 6036 patients with surgically treated hip fractures, matched for surgery type, age, gender, and ASA class (a surrogate for severity of medical infirmary), were studied using adjusted regression modeling to afford an isolated statistical assessment of the effect of time-to-surgery on outcomes following hip fracture surgery.

Despite a large sample size and rigorous statistical methodology, for many outcome measures, our results show no support for the early or late operative intervention following hip fracture. We found no difference in 30-day mortality, readmission rate, nor total complication rate between the 3 time-to-surgery cohorts. This result indicates that the care of elderly patients following hip fracture is inherently complicated and that perioperative complication risk is probably only modestly modifiable by best medical practices, including optimizing time from clinical presentation to surgery.

As expected, patients who experienced longer delays from presentation to surgery were on average, more comorbid and more likely to yield abnormal preoperative lab values. However, in the adjusted analysis, delay in time-to-surgery, presumably for medical management, was not found to be associated with improved outcomes. In the same adjusted analysis, we uniquely identified that in the patients whose surgeries were delayed for more than 48 hours, the time from surgery-to-discharge was significantly increased. As a result, these patients spent extra days in the hospital both preoperatively and postoperatively, but without any corollary improvement in the outcomes.

Continue to: Recent estimates of the cost of hospital admission...

Recent estimates of the cost of hospital admission is approximated nationally at $2000/day.²⁶ Although our data fail to support the formal cost-analysis of the effect of time-to-surgery in hip fracture care, a simple value-based analysis indicates that quality is preserved (no difference in outcome), whereas costly hospital days are eliminated with earlier surgery. The value in elderly hip fracture care. defined as the outcomes relative to the costs,⁷ is ultimately optimized by earlier time-to-surgery.

Although using a large, multi-institutional database is advantageous for finding population-based trends that are representative of a large cohort, using the ACS-NSQIP database features its limitations. Our analysis was limited to the defined scope of NSQIP and nature of the injury, whereas root cause for delay was not available for study. We were unable to identify which patients were delayed for administrative reasons or surgical convenience and which were delayed for medical optimization. Participation in the ACS-NSQIP database is voluntary, and no randomized hospital sampling was conducted. Participating hospitals were de-identified in the database. As expected, we were unable to identify the specific institution-based hip fracture protocols that may affect the outcomes following treatment for these fractures. Further, socioeconomic information and payer-status are unavailable for the study. Additionally, observations are limited to 30 days postoperative, and we cannot comment on longer-term outcomes. Finally, discharge disposition and functional outcome data are not represented, and we were unable to correlate time-to-surgery and functional recovery. However, previous studies have established that delay in time-to-surgery following hip fractures is negatively correlated with functional outcomes.^22-25

Nevertheless, the ACS-NSQIP database remains one of the largest American surgical databases available, and includes care centers from nearly every state with variable demographics including rural, urban, and academic centers. The ACS performs broad-based inter-rater reliability audits on every participating site and has found an overall disagreement rate of only 1.8%. As such, although discrepancies exist between the complete patient chart and the data entered, the data found in the ACS-NSQIP database are reliable and considered a valid source of study.^34,35 The large sample size, quality of data collection, wide geographic representation, and varied hospital types within the dataset possibly make our findings relevant in the majority of American healthcare settings.

CONCLUSION

This study demonstrates an associated increased length of hospital stay, including the increased time from surgery-to-discharge, in patients with hip fractures whose surgical intervention is delayed for >48 hours after presentation. Given the prior evidence that early surgical intervention improves the functional outcomes and the current evidence that surgical delay for any cause increases costly hospital length of stay without corollary improvement in the outcomes, a value-based assessment of hip fracture care argues for early surgical intervention whenever possible. Our findings should inform physician, institution, and policy maker value-based decision making regarding the best practices in geriatric hip fracture care.

Benzodiazepines (BZDs) are among the most commonly prescribed medications. A recent study found that in 2008, more than 5% of Americans used a BZD, and the percentage was almost 9% among Americans aged ≥ 65 years.^1,2 Among veterans, BZD use is even higher, in part because of the high prevalence of posttraumatic stress disorder (PTSD). One study found that more than 30% of veterans with PTSD received at least 1 BZD prescription.³ The risks associated with BZD treatment for PTSD are compounded by concurrent use of other sedatives and opioids prescribed for co-occurring chronic pain and insomnia.³

Older adults metabolize long-acting BZDs more slowly and generally have an increased sensitivity to the adverse effects (AEs) of all BZDs.⁴ In older adults, BZD use has been associated with cognitive decline, dementia, falls and consequent fractures, and adverse respiratory outcomes.^5-12 The risk of most but not all of these AEs was increased with higher BZD dose or long-term BZD use, which this quality improvement project (QIP) defines as having at least a 60-day supply of BZD prescriptions dispensed within the past year.

Long-term BZD use increases with age. One study found that, among patients receiving a BZD, the rate of long-term BZD use was more than double in older adults (31.4%) than it was in adults aged between 18 and 35 years (14.7%).² For these reasons, the 2012 Beers criteria of the American Geriatrics Society recommend avoiding all types of BZDs in the treatment of insomnia, agitation, or delirium in patients aged > 65 years.¹³ Despite this recommendation, the prevalence of BZD use in older adults remains high.¹⁴

Some innovative approaches have been developed to address the inappropriate use, including overuse and misuse, of BZDs in older adults.¹⁵ In one approach, direct-to-consumer (DTC) information is used to empower patients to collaborate with their physician to manage their health. Results from several studies suggest that providing older patients with information on BZD risks and benefits increases patient–physician interaction and thereby decreases inappropriate BZD use and improves health outcomes.^4,16,17 One study found that perceptions of BZD risks increased 1 week after exposure to a DTC educational brochure (EB), with intention to discuss BZD discontinuation with their physician higher for patients who received the EB than it was for those who did not (83.1% vs 44.3%; P < .0001).¹⁶ The EMPOWER (Eliminating Medications Through Patient Ownership of End Results) cluster randomized controlled trial assessed the effectiveness of a DTC EB focused on BZD risks in older adults.¹⁷ In that seminal study, patients who received a DTC EB were more likely than were comparison patients to discontinue BZD within 6 months (27% vs 5%; risk difference, 23%; 95% CI, 14%-32%).

The Veterans Integrated Systems Network (VISN) 22 Academic Detailing Program is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20 With BZD use among older veterans remaining high, the VISN 22 program initiated a clinical QIP modeled on the EMPOWER trial. Veterans in VISN 22 received the DTC EB, which included information on BZD risks and encouraged them to discuss their BZD treatment with their health care provider. VISN 22 was the first VISN in the VHA to implement the EMPOWER protocol.

As this was a QIP, all eligible veterans in VISN 22 were mailed the DTC EB, thus making it difficult to estimate the impact of the EB on BZD discontinuation in this VISN. Therefore, DTC EB efficacy was estimated by comparing BZD discontinuation between VISN 22 and VISN 21, an adjacent VISN that did not mail the DTC EB. To reduce selection bias associated with different controls in the 2 VISNs, the authors performed propensity score matching (PSM) to balance the covariates and provide an unbiased estimate of the mean treatment effect of the DTC EB in VISN 22 veterans who were included in the initial descriptive QIP and received the EB; these veterans were compared with VISN 21 veterans who did not receive the EB.

Methods

Two QIPs were undertaken to determine the impact of DTC EB on BZD use in older veterans in the VHA.

Quality Improvement Project 1

Design. A retrospective cohort analysis was performed. The VISN 22 catchment area, which encompasses VA facilities and clinics in southern California and southern Nevada, serves about 500,000 veterans, a substantial proportion of whom are aged ≥ 65 years. Among these older veterans are active long-term BZD users, who were defined as having ≥ 60-day supply of BZD prescriptions dispensed within the past year. Each active long-term user with a BZD prescription released within 200 days before the index date (the date the user was to meet with the prescribing physician) was mailed an EB 2 to 8 weeks in advance of the visit. Excluded from analysis were veterans with a schizophrenia, spinal cord injury, or seizure disorder diagnosis recorded in both their inpatient and outpatient medical records; veterans seen by Palliative Care within the past year; and veterans who died before analysis was completed.

Education Brochure. The EB for VISN 22 (Figure 1, see

Patients. The sample consisted of all veterans identified as meeting the inclusion criteria and being enrolled in VISN 22. The EB was mailed once to veterans on a rolling basis from December 2014 to February 2016. Change in BZD use was analyzed only after 9 to 24 months had passed since the index appointment with the prescribing physician. This period included 12 weeks for BZD taper and then 6 months after taper.

Analysis. For each veteran, monthly mean lorazepam equivalent (LE) was calculated using as many as 12 fills before the index date. Average daily dose of LE was calculated by dividing the sum of LE from all included prescriptions by total number of days between the first fill and the index date. The BZD prescription fills were evaluated after the index date. Veterans who received at least 1 prescription after the index date but then had no BZD prescription activity in VA clinics for 3 consecutive months during the 9-month observation period were recorded as having tapered and then discontinued BZD. Veterans who had no BZD prescription activity in VA clinics after the index date and during the 9-month observation period were recorded as having discontinued BZD without tapering. For veterans who had BZD prescription activity in VA clinics after the index date and during the 9-month observation period, mean LE was calculated by dividing the total LE for BZD prescriptions after the index date by number of days from the first fill after the index date to the date of analysis.

Quality Improvement Project 2

Design. A retrospective cohort analysis using PSM was performed on a subgroup of the QIP-1 sample to evaluate the impact of EB on BZD prescribing in the VA during 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. Veterans in the analysis were active long-term BZD users, had at least 1 BZD prescription released within 200 days before the index date, were aged ≥ 65 years, and had an appointment scheduled with their BZD prescriber within 2 to 8 weeks (Figure 2). 

Patients. VISN 22 implemented QIP-2, a real-world application of a modified EMPOWER program, by identifying eligible veterans on a rolling basis from December 2014 to August 2015. All veterans who were identified and sent an EB during this period were included in the case group. The index date was defined as the first of the month the EB was mailed. Veterans with a pending appointment were chosen because the lead time would allow them to receive the EB and prepare to discuss it with the physician during the visit.

A comparator group was drawn from the adjacent VISN 21 catchment area, which encompasses VA facilities and clinics in Hawaii, northern California, and northern Nevada. During the observation period, VISN 21 did not mail any EBs specifically addressing BZD risks. Veterans in the comparator group had an appointment scheduled with their BZD prescribing physician within 4 weeks, were aged ≥ 65 years on the index date (first of the month before the next appointment, coinciding with the date EBs were sent to VISN 22 veterans), were active long-term BZD users, and had at least 1 BZD prescription released within 200 days before the index date. All patients were followed for up to 12 months after the index date, with BZD discontinuation recorded 9 and 12 months after the index date.

Propensity Score Matching

Propensity score (PS) was estimated with logistic regression analysis with treatment as the dependent variable and baseline characteristics as the independent variables.^21,22 One-to-one matching on the PS was performed using the nearest neighbor approach without replacements. Independent variables related to outcome but unrelated to EB exposure were selected for PS development.²² These variables included year of birth; male sex; Hispanic ethnicity; annual income; service connection status; region; body mass index; Charlson Comorbidity Index category; total baseline BZD dose; and diagnosis of AIDS, nonmetastatic cancer, metastatic cancer, chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), dementia, diabetes mellitus (DM), DM with complications, gastroesophageal reflux disease (GERD), general anxiety disorder (GAD), hemiparaplegia, liver disease (mild), liver disease (moderate to severe), myocardial infarction (MI), Parkinson disease, peptic ulcer disease (PUD), psychosis, renal disease, rheumatoid arthritis (RA), or substance use disorder (SUD).

The EMPOWER cluster randomized controlled trial (RCT) demonstrated the effectiveness of EB exposure in a Canadian population of elderly patients who were long-term BZD users.¹⁷ Randomized controlled trials are the gold standard for clinical trials because they can establish causal inference.^23-25 Given ethical and practical concerns, however, RCTs cannot be applied to all clinical scenarios. Although EMPOWER is reported to be an effective tool in reducing BZD use in older adults, its application in a real-world, large, integrated health care system remains untested. Observational studies are often conducted as an alternative to RCTs but are subject to selection bias because of their lack of randomization.²⁶ Therefore, robust research methods are needed to generate unbiased estimates of the impact of an intervention on an outcome. Propensity score matching simulates an RCT by balancing the covariates across treatment groups.^21,22,27 Observed patient characteristics are used to estimate PS, the probability that treatment will be received. Logistic or probit regression is used to balance the potential confounding covariates between the treatment groups.Once PSs are known, mean treatment effect can be estimated without the mean model.²⁸ In other words, PSM methods can be used to generate an unbiased estimate of the treatment.

Propensity Score Analysis

Baseline characteristics were compared using Student t test (continuous variables) and χ² test (discrete variables). Results are presented as means and standard deviations (continuous variables) and frequency and percentage (discrete variables).

The main outcome was BZD discontinuation 9 and 12 months after the index date. A postindex lag of 6 months was used to capture any tapering (Figure 2). Discontinuation, defined as 3 consecutive months of no BZD prescription on hand, was measured for 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. An estimate was made of the difference in the proportions of BZD discontinuers who received the EB and BZD discontinuers who did not receive the EB, where mean treatment (risk difference) was presented as the absolute risk difference with a 95% CI. Standard errors and 95% CIs for the risk differences were generated with biased-corrected CIs from 1,000 bootstrap samples.

Sensitivity Analyses

Naïve multivariate logistic regression analysis was performed to evaluate the association between EB exposure and BZD discontinuation while controlling for potential confounders. Results are presented as odds ratios (ORs) and 95% CIs. Confounders identified were the same covariates used to generate the PSs.

Several analyses were performed to test the sensitivity of the methods applied using PSM by changing caliber size while maintaining the nearest neighbor approach without replacement. Linear regression analysis was performed with robust standard errors to estimate the risk difference of BZD discontinuation between EB-exposed and EB-unexposed veterans.

Statistical significance was set at P < .05. All statistical analyses were performed with Stata/SE Version 13 (College Station, TX).

Results

Quality Improvement Project 1

On a rolling basis from December 2014 to February 2016, the EB was mailed once to 3,896 VISN 22 veterans 2 to 8 weeks before a clinic appointment with their BZD prescribing physician. 

Quality Improvement Project 2

Of all the VISN 22 and VISN 21 veterans, 24,420 met the inclusion and exclusion criteria. Of these 24,420 veterans, 2,020 (8.3%) were in VISN 22 and received the EB between December 2014 and August 2015 (QIP-1), and 22,400 (91.7%) were in VISN 21 and did not receive the EB.

Naïve Results Before PS Matching. In the naïve analyses, a larger proportion of EB-exposed vs unexposed veterans discontinued BZD; in addition, reductions were 6.6%, 7.4%, and 9.5% larger for 6 to 9 months, 6 to 12 months, and 1 to 12 months after the index date, respectively (P < .0001 for all comparisons; Table 2).

Propensity Score Matching. Before matching, there were significant differences in baseline characteristics of veterans who met the inclusion and exclusion criteria, with few exceptions (eAppendices 2 and 3, ).

Propensity Score Matching Results. Inspection of PSs revealed good coverage across treatment groups on a histogram plot and a kernel density plot (eAppendices 5 and 6).

Discussion

This QIP was the first to evaluate the impact of an EMPOWER-modeled DTC EB in a large, integrated health care system in the U.S. It was also the first to demonstrate potential benefits of a DTC EB designed for older veterans who are long-term BZD users. In this QIP, which mailed the EB to 3,896 veterans, 1,847 (47.4%) decreased their BZD dose, 458 (11.7%) tapered and then discontinued BZD, and 455 (11.7%) immediately discontinued BZD. The total percentage of veterans who discontinued BZD (23.4%; 913/3,896) was similar to the 27% reported in the EMPOWER trial.¹⁷ However, the risk difference between the 1,316 EB-exposed VISN 22 veterans (QIP-1) and the 1,316 EB-unexposed VISN 21 veterans in this QIP was significantly lower than the 23% risk difference in EMPOWER (though it still demonstrated a significantly larger reduction for EB-exposed veterans).¹⁷

Given this inclusion of all qualifying veterans from the catchment area studied in this QIP, and given the ethical and practical concerns, an RCT was not possible. Therefore, PSM methods were used to balance the covariates across treatment groups and thereby simulate an RCT.^21,22,27 With use of the PSM approach, findings from the descriptive analysis were confirmed and potential selection bias reduced.

Study Limitations

The less robust risk difference found in this QIP has several possible explanations. The authors’ use of a DTC EB coincided with a national VA effort to reduce older veterans’ use of BZDs and other inappropriate medications. For instance, during the study period, academic detailing was being implemented to reduce use of BZDs, particularly in combination with opioids, across VHA facilities and clinics. (Academic detailing is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20) However, QIP-2 results and PS analysis of a subgroup of the original sample suggest that EB-exposed veterans were significantly more likely than were their unexposed counterparts were to discontinue BZD. To an extent, this analysis controlled for these other efforts to reduce BZD use in VHA clinics and can be considered a study strength.

Another limitation is the study design, which lacked a control group and did not consider the possibility that some facility or clinic physicians might influence others. Although the region variable was controlled for in PSM, the authors did not capture facility characteristics, including frequency of prescribing BZD and use of a protocol for enforcing the Beers criteria. Such confounders might have influenced outcomes. Unlike the EMPOWER trial,¹⁷ this QIP did not assess or exclude cognitively impaired veterans. It is reasonable to assume that these veterans might not understand some EB messages and consequently might fail to engage their physicians. Failure to initiate discussion with a physician would attenuate the impact of the EB.

Study Strengths

A strength of this QIP was its use of a DTC EB in a large, regional sample of older veterans in a real-world clinical setting. In addition, the study group (EB-exposed veterans) and the comparator group (EB-unexposed veterans) were from similar geographic areas (primarily California and Nevada).

Conclusion

Results of this study suggest that a DTC EB, designed to reduce BZD use among older veterans, was effective in helping patients lower their BZD dose and discontinue BZD. The likelihood of discontinuing BZD 9 and 12 months after the index date was significantly higher for veterans who received an EB modeled on the EMPOWER educational brochure than for a comparator group of veterans who did not receive the EB and were receiving care during the same observation period. In the future, it would be beneficial to use a design that controls for physician exposure to academic detailing focused on BZD reduction and that accounts for the cluster effects of facility practice. Despite these limitations, this QIP is the first real-world empirical example of using an EMPOWER-modeled DTC EB to decrease BZD use among older veterans. Furthermore, these results suggest that a DTC EB can be used to target other high-risk prescription drugs, such as opioids, particularly if alternative treatment options can be provided.

Acknowledgments
Dr. Hauser thanks Cathy, Anika, Katia, and Max Hauser, and Alba and Kevin Quinlan, for their support. In memory of Jirina Hauser, who died on Mother’s Day, May 14, 2017, at the age of 100.

Benzodiazepines (BZDs) are among the most commonly prescribed medications. A recent study found that in 2008, more than 5% of Americans used a BZD, and the percentage was almost 9% among Americans aged ≥ 65 years.^1,2 Among veterans, BZD use is even higher, in part because of the high prevalence of posttraumatic stress disorder (PTSD). One study found that more than 30% of veterans with PTSD received at least 1 BZD prescription.³ The risks associated with BZD treatment for PTSD are compounded by concurrent use of other sedatives and opioids prescribed for co-occurring chronic pain and insomnia.³

Older adults metabolize long-acting BZDs more slowly and generally have an increased sensitivity to the adverse effects (AEs) of all BZDs.⁴ In older adults, BZD use has been associated with cognitive decline, dementia, falls and consequent fractures, and adverse respiratory outcomes.^5-12 The risk of most but not all of these AEs was increased with higher BZD dose or long-term BZD use, which this quality improvement project (QIP) defines as having at least a 60-day supply of BZD prescriptions dispensed within the past year.

Long-term BZD use increases with age. One study found that, among patients receiving a BZD, the rate of long-term BZD use was more than double in older adults (31.4%) than it was in adults aged between 18 and 35 years (14.7%).² For these reasons, the 2012 Beers criteria of the American Geriatrics Society recommend avoiding all types of BZDs in the treatment of insomnia, agitation, or delirium in patients aged > 65 years.¹³ Despite this recommendation, the prevalence of BZD use in older adults remains high.¹⁴

Some innovative approaches have been developed to address the inappropriate use, including overuse and misuse, of BZDs in older adults.¹⁵ In one approach, direct-to-consumer (DTC) information is used to empower patients to collaborate with their physician to manage their health. Results from several studies suggest that providing older patients with information on BZD risks and benefits increases patient–physician interaction and thereby decreases inappropriate BZD use and improves health outcomes.^4,16,17 One study found that perceptions of BZD risks increased 1 week after exposure to a DTC educational brochure (EB), with intention to discuss BZD discontinuation with their physician higher for patients who received the EB than it was for those who did not (83.1% vs 44.3%; P < .0001).¹⁶ The EMPOWER (Eliminating Medications Through Patient Ownership of End Results) cluster randomized controlled trial assessed the effectiveness of a DTC EB focused on BZD risks in older adults.¹⁷ In that seminal study, patients who received a DTC EB were more likely than were comparison patients to discontinue BZD within 6 months (27% vs 5%; risk difference, 23%; 95% CI, 14%-32%).

The Veterans Integrated Systems Network (VISN) 22 Academic Detailing Program is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20 With BZD use among older veterans remaining high, the VISN 22 program initiated a clinical QIP modeled on the EMPOWER trial. Veterans in VISN 22 received the DTC EB, which included information on BZD risks and encouraged them to discuss their BZD treatment with their health care provider. VISN 22 was the first VISN in the VHA to implement the EMPOWER protocol.

As this was a QIP, all eligible veterans in VISN 22 were mailed the DTC EB, thus making it difficult to estimate the impact of the EB on BZD discontinuation in this VISN. Therefore, DTC EB efficacy was estimated by comparing BZD discontinuation between VISN 22 and VISN 21, an adjacent VISN that did not mail the DTC EB. To reduce selection bias associated with different controls in the 2 VISNs, the authors performed propensity score matching (PSM) to balance the covariates and provide an unbiased estimate of the mean treatment effect of the DTC EB in VISN 22 veterans who were included in the initial descriptive QIP and received the EB; these veterans were compared with VISN 21 veterans who did not receive the EB.

Methods

Two QIPs were undertaken to determine the impact of DTC EB on BZD use in older veterans in the VHA.

Quality Improvement Project 1

Design. A retrospective cohort analysis was performed. The VISN 22 catchment area, which encompasses VA facilities and clinics in southern California and southern Nevada, serves about 500,000 veterans, a substantial proportion of whom are aged ≥ 65 years. Among these older veterans are active long-term BZD users, who were defined as having ≥ 60-day supply of BZD prescriptions dispensed within the past year. Each active long-term user with a BZD prescription released within 200 days before the index date (the date the user was to meet with the prescribing physician) was mailed an EB 2 to 8 weeks in advance of the visit. Excluded from analysis were veterans with a schizophrenia, spinal cord injury, or seizure disorder diagnosis recorded in both their inpatient and outpatient medical records; veterans seen by Palliative Care within the past year; and veterans who died before analysis was completed.

Education Brochure. The EB for VISN 22 (Figure 1, see

Patients. The sample consisted of all veterans identified as meeting the inclusion criteria and being enrolled in VISN 22. The EB was mailed once to veterans on a rolling basis from December 2014 to February 2016. Change in BZD use was analyzed only after 9 to 24 months had passed since the index appointment with the prescribing physician. This period included 12 weeks for BZD taper and then 6 months after taper.

Analysis. For each veteran, monthly mean lorazepam equivalent (LE) was calculated using as many as 12 fills before the index date. Average daily dose of LE was calculated by dividing the sum of LE from all included prescriptions by total number of days between the first fill and the index date. The BZD prescription fills were evaluated after the index date. Veterans who received at least 1 prescription after the index date but then had no BZD prescription activity in VA clinics for 3 consecutive months during the 9-month observation period were recorded as having tapered and then discontinued BZD. Veterans who had no BZD prescription activity in VA clinics after the index date and during the 9-month observation period were recorded as having discontinued BZD without tapering. For veterans who had BZD prescription activity in VA clinics after the index date and during the 9-month observation period, mean LE was calculated by dividing the total LE for BZD prescriptions after the index date by number of days from the first fill after the index date to the date of analysis.

Quality Improvement Project 2

Design. A retrospective cohort analysis using PSM was performed on a subgroup of the QIP-1 sample to evaluate the impact of EB on BZD prescribing in the VA during 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. Veterans in the analysis were active long-term BZD users, had at least 1 BZD prescription released within 200 days before the index date, were aged ≥ 65 years, and had an appointment scheduled with their BZD prescriber within 2 to 8 weeks (Figure 2). 

Patients. VISN 22 implemented QIP-2, a real-world application of a modified EMPOWER program, by identifying eligible veterans on a rolling basis from December 2014 to August 2015. All veterans who were identified and sent an EB during this period were included in the case group. The index date was defined as the first of the month the EB was mailed. Veterans with a pending appointment were chosen because the lead time would allow them to receive the EB and prepare to discuss it with the physician during the visit.

A comparator group was drawn from the adjacent VISN 21 catchment area, which encompasses VA facilities and clinics in Hawaii, northern California, and northern Nevada. During the observation period, VISN 21 did not mail any EBs specifically addressing BZD risks. Veterans in the comparator group had an appointment scheduled with their BZD prescribing physician within 4 weeks, were aged ≥ 65 years on the index date (first of the month before the next appointment, coinciding with the date EBs were sent to VISN 22 veterans), were active long-term BZD users, and had at least 1 BZD prescription released within 200 days before the index date. All patients were followed for up to 12 months after the index date, with BZD discontinuation recorded 9 and 12 months after the index date.

Propensity Score Matching

Propensity score (PS) was estimated with logistic regression analysis with treatment as the dependent variable and baseline characteristics as the independent variables.^21,22 One-to-one matching on the PS was performed using the nearest neighbor approach without replacements. Independent variables related to outcome but unrelated to EB exposure were selected for PS development.²² These variables included year of birth; male sex; Hispanic ethnicity; annual income; service connection status; region; body mass index; Charlson Comorbidity Index category; total baseline BZD dose; and diagnosis of AIDS, nonmetastatic cancer, metastatic cancer, chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), dementia, diabetes mellitus (DM), DM with complications, gastroesophageal reflux disease (GERD), general anxiety disorder (GAD), hemiparaplegia, liver disease (mild), liver disease (moderate to severe), myocardial infarction (MI), Parkinson disease, peptic ulcer disease (PUD), psychosis, renal disease, rheumatoid arthritis (RA), or substance use disorder (SUD).

The EMPOWER cluster randomized controlled trial (RCT) demonstrated the effectiveness of EB exposure in a Canadian population of elderly patients who were long-term BZD users.¹⁷ Randomized controlled trials are the gold standard for clinical trials because they can establish causal inference.^23-25 Given ethical and practical concerns, however, RCTs cannot be applied to all clinical scenarios. Although EMPOWER is reported to be an effective tool in reducing BZD use in older adults, its application in a real-world, large, integrated health care system remains untested. Observational studies are often conducted as an alternative to RCTs but are subject to selection bias because of their lack of randomization.²⁶ Therefore, robust research methods are needed to generate unbiased estimates of the impact of an intervention on an outcome. Propensity score matching simulates an RCT by balancing the covariates across treatment groups.^21,22,27 Observed patient characteristics are used to estimate PS, the probability that treatment will be received. Logistic or probit regression is used to balance the potential confounding covariates between the treatment groups.Once PSs are known, mean treatment effect can be estimated without the mean model.²⁸ In other words, PSM methods can be used to generate an unbiased estimate of the treatment.

Propensity Score Analysis

Baseline characteristics were compared using Student t test (continuous variables) and χ² test (discrete variables). Results are presented as means and standard deviations (continuous variables) and frequency and percentage (discrete variables).

The main outcome was BZD discontinuation 9 and 12 months after the index date. A postindex lag of 6 months was used to capture any tapering (Figure 2). Discontinuation, defined as 3 consecutive months of no BZD prescription on hand, was measured for 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. An estimate was made of the difference in the proportions of BZD discontinuers who received the EB and BZD discontinuers who did not receive the EB, where mean treatment (risk difference) was presented as the absolute risk difference with a 95% CI. Standard errors and 95% CIs for the risk differences were generated with biased-corrected CIs from 1,000 bootstrap samples.

Sensitivity Analyses

Naïve multivariate logistic regression analysis was performed to evaluate the association between EB exposure and BZD discontinuation while controlling for potential confounders. Results are presented as odds ratios (ORs) and 95% CIs. Confounders identified were the same covariates used to generate the PSs.

Several analyses were performed to test the sensitivity of the methods applied using PSM by changing caliber size while maintaining the nearest neighbor approach without replacement. Linear regression analysis was performed with robust standard errors to estimate the risk difference of BZD discontinuation between EB-exposed and EB-unexposed veterans.

Statistical significance was set at P < .05. All statistical analyses were performed with Stata/SE Version 13 (College Station, TX).

Results

Quality Improvement Project 1

On a rolling basis from December 2014 to February 2016, the EB was mailed once to 3,896 VISN 22 veterans 2 to 8 weeks before a clinic appointment with their BZD prescribing physician. 

Quality Improvement Project 2

Of all the VISN 22 and VISN 21 veterans, 24,420 met the inclusion and exclusion criteria. Of these 24,420 veterans, 2,020 (8.3%) were in VISN 22 and received the EB between December 2014 and August 2015 (QIP-1), and 22,400 (91.7%) were in VISN 21 and did not receive the EB.

Naïve Results Before PS Matching. In the naïve analyses, a larger proportion of EB-exposed vs unexposed veterans discontinued BZD; in addition, reductions were 6.6%, 7.4%, and 9.5% larger for 6 to 9 months, 6 to 12 months, and 1 to 12 months after the index date, respectively (P < .0001 for all comparisons; Table 2).

Propensity Score Matching. Before matching, there were significant differences in baseline characteristics of veterans who met the inclusion and exclusion criteria, with few exceptions (eAppendices 2 and 3, ).

Propensity Score Matching Results. Inspection of PSs revealed good coverage across treatment groups on a histogram plot and a kernel density plot (eAppendices 5 and 6).

Discussion

This QIP was the first to evaluate the impact of an EMPOWER-modeled DTC EB in a large, integrated health care system in the U.S. It was also the first to demonstrate potential benefits of a DTC EB designed for older veterans who are long-term BZD users. In this QIP, which mailed the EB to 3,896 veterans, 1,847 (47.4%) decreased their BZD dose, 458 (11.7%) tapered and then discontinued BZD, and 455 (11.7%) immediately discontinued BZD. The total percentage of veterans who discontinued BZD (23.4%; 913/3,896) was similar to the 27% reported in the EMPOWER trial.¹⁷ However, the risk difference between the 1,316 EB-exposed VISN 22 veterans (QIP-1) and the 1,316 EB-unexposed VISN 21 veterans in this QIP was significantly lower than the 23% risk difference in EMPOWER (though it still demonstrated a significantly larger reduction for EB-exposed veterans).¹⁷

Given this inclusion of all qualifying veterans from the catchment area studied in this QIP, and given the ethical and practical concerns, an RCT was not possible. Therefore, PSM methods were used to balance the covariates across treatment groups and thereby simulate an RCT.^21,22,27 With use of the PSM approach, findings from the descriptive analysis were confirmed and potential selection bias reduced.

Study Limitations

The less robust risk difference found in this QIP has several possible explanations. The authors’ use of a DTC EB coincided with a national VA effort to reduce older veterans’ use of BZDs and other inappropriate medications. For instance, during the study period, academic detailing was being implemented to reduce use of BZDs, particularly in combination with opioids, across VHA facilities and clinics. (Academic detailing is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20) However, QIP-2 results and PS analysis of a subgroup of the original sample suggest that EB-exposed veterans were significantly more likely than were their unexposed counterparts were to discontinue BZD. To an extent, this analysis controlled for these other efforts to reduce BZD use in VHA clinics and can be considered a study strength.

Another limitation is the study design, which lacked a control group and did not consider the possibility that some facility or clinic physicians might influence others. Although the region variable was controlled for in PSM, the authors did not capture facility characteristics, including frequency of prescribing BZD and use of a protocol for enforcing the Beers criteria. Such confounders might have influenced outcomes. Unlike the EMPOWER trial,¹⁷ this QIP did not assess or exclude cognitively impaired veterans. It is reasonable to assume that these veterans might not understand some EB messages and consequently might fail to engage their physicians. Failure to initiate discussion with a physician would attenuate the impact of the EB.

Study Strengths

A strength of this QIP was its use of a DTC EB in a large, regional sample of older veterans in a real-world clinical setting. In addition, the study group (EB-exposed veterans) and the comparator group (EB-unexposed veterans) were from similar geographic areas (primarily California and Nevada).

Conclusion

Results of this study suggest that a DTC EB, designed to reduce BZD use among older veterans, was effective in helping patients lower their BZD dose and discontinue BZD. The likelihood of discontinuing BZD 9 and 12 months after the index date was significantly higher for veterans who received an EB modeled on the EMPOWER educational brochure than for a comparator group of veterans who did not receive the EB and were receiving care during the same observation period. In the future, it would be beneficial to use a design that controls for physician exposure to academic detailing focused on BZD reduction and that accounts for the cluster effects of facility practice. Despite these limitations, this QIP is the first real-world empirical example of using an EMPOWER-modeled DTC EB to decrease BZD use among older veterans. Furthermore, these results suggest that a DTC EB can be used to target other high-risk prescription drugs, such as opioids, particularly if alternative treatment options can be provided.

Acknowledgments
Dr. Hauser thanks Cathy, Anika, Katia, and Max Hauser, and Alba and Kevin Quinlan, for their support. In memory of Jirina Hauser, who died on Mother’s Day, May 14, 2017, at the age of 100.

Benzodiazepines (BZDs) are among the most commonly prescribed medications. A recent study found that in 2008, more than 5% of Americans used a BZD, and the percentage was almost 9% among Americans aged ≥ 65 years.^1,2 Among veterans, BZD use is even higher, in part because of the high prevalence of posttraumatic stress disorder (PTSD). One study found that more than 30% of veterans with PTSD received at least 1 BZD prescription.³ The risks associated with BZD treatment for PTSD are compounded by concurrent use of other sedatives and opioids prescribed for co-occurring chronic pain and insomnia.³

Older adults metabolize long-acting BZDs more slowly and generally have an increased sensitivity to the adverse effects (AEs) of all BZDs.⁴ In older adults, BZD use has been associated with cognitive decline, dementia, falls and consequent fractures, and adverse respiratory outcomes.^5-12 The risk of most but not all of these AEs was increased with higher BZD dose or long-term BZD use, which this quality improvement project (QIP) defines as having at least a 60-day supply of BZD prescriptions dispensed within the past year.

Long-term BZD use increases with age. One study found that, among patients receiving a BZD, the rate of long-term BZD use was more than double in older adults (31.4%) than it was in adults aged between 18 and 35 years (14.7%).² For these reasons, the 2012 Beers criteria of the American Geriatrics Society recommend avoiding all types of BZDs in the treatment of insomnia, agitation, or delirium in patients aged > 65 years.¹³ Despite this recommendation, the prevalence of BZD use in older adults remains high.¹⁴

Some innovative approaches have been developed to address the inappropriate use, including overuse and misuse, of BZDs in older adults.¹⁵ In one approach, direct-to-consumer (DTC) information is used to empower patients to collaborate with their physician to manage their health. Results from several studies suggest that providing older patients with information on BZD risks and benefits increases patient–physician interaction and thereby decreases inappropriate BZD use and improves health outcomes.^4,16,17 One study found that perceptions of BZD risks increased 1 week after exposure to a DTC educational brochure (EB), with intention to discuss BZD discontinuation with their physician higher for patients who received the EB than it was for those who did not (83.1% vs 44.3%; P < .0001).¹⁶ The EMPOWER (Eliminating Medications Through Patient Ownership of End Results) cluster randomized controlled trial assessed the effectiveness of a DTC EB focused on BZD risks in older adults.¹⁷ In that seminal study, patients who received a DTC EB were more likely than were comparison patients to discontinue BZD within 6 months (27% vs 5%; risk difference, 23%; 95% CI, 14%-32%).

The Veterans Integrated Systems Network (VISN) 22 Academic Detailing Program is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20 With BZD use among older veterans remaining high, the VISN 22 program initiated a clinical QIP modeled on the EMPOWER trial. Veterans in VISN 22 received the DTC EB, which included information on BZD risks and encouraged them to discuss their BZD treatment with their health care provider. VISN 22 was the first VISN in the VHA to implement the EMPOWER protocol.

As this was a QIP, all eligible veterans in VISN 22 were mailed the DTC EB, thus making it difficult to estimate the impact of the EB on BZD discontinuation in this VISN. Therefore, DTC EB efficacy was estimated by comparing BZD discontinuation between VISN 22 and VISN 21, an adjacent VISN that did not mail the DTC EB. To reduce selection bias associated with different controls in the 2 VISNs, the authors performed propensity score matching (PSM) to balance the covariates and provide an unbiased estimate of the mean treatment effect of the DTC EB in VISN 22 veterans who were included in the initial descriptive QIP and received the EB; these veterans were compared with VISN 21 veterans who did not receive the EB.

Methods

Two QIPs were undertaken to determine the impact of DTC EB on BZD use in older veterans in the VHA.

Quality Improvement Project 1

Design. A retrospective cohort analysis was performed. The VISN 22 catchment area, which encompasses VA facilities and clinics in southern California and southern Nevada, serves about 500,000 veterans, a substantial proportion of whom are aged ≥ 65 years. Among these older veterans are active long-term BZD users, who were defined as having ≥ 60-day supply of BZD prescriptions dispensed within the past year. Each active long-term user with a BZD prescription released within 200 days before the index date (the date the user was to meet with the prescribing physician) was mailed an EB 2 to 8 weeks in advance of the visit. Excluded from analysis were veterans with a schizophrenia, spinal cord injury, or seizure disorder diagnosis recorded in both their inpatient and outpatient medical records; veterans seen by Palliative Care within the past year; and veterans who died before analysis was completed.

Education Brochure. The EB for VISN 22 (Figure 1, see

Patients. The sample consisted of all veterans identified as meeting the inclusion criteria and being enrolled in VISN 22. The EB was mailed once to veterans on a rolling basis from December 2014 to February 2016. Change in BZD use was analyzed only after 9 to 24 months had passed since the index appointment with the prescribing physician. This period included 12 weeks for BZD taper and then 6 months after taper.

Analysis. For each veteran, monthly mean lorazepam equivalent (LE) was calculated using as many as 12 fills before the index date. Average daily dose of LE was calculated by dividing the sum of LE from all included prescriptions by total number of days between the first fill and the index date. The BZD prescription fills were evaluated after the index date. Veterans who received at least 1 prescription after the index date but then had no BZD prescription activity in VA clinics for 3 consecutive months during the 9-month observation period were recorded as having tapered and then discontinued BZD. Veterans who had no BZD prescription activity in VA clinics after the index date and during the 9-month observation period were recorded as having discontinued BZD without tapering. For veterans who had BZD prescription activity in VA clinics after the index date and during the 9-month observation period, mean LE was calculated by dividing the total LE for BZD prescriptions after the index date by number of days from the first fill after the index date to the date of analysis.

Quality Improvement Project 2

Design. A retrospective cohort analysis using PSM was performed on a subgroup of the QIP-1 sample to evaluate the impact of EB on BZD prescribing in the VA during 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. Veterans in the analysis were active long-term BZD users, had at least 1 BZD prescription released within 200 days before the index date, were aged ≥ 65 years, and had an appointment scheduled with their BZD prescriber within 2 to 8 weeks (Figure 2). 

Patients. VISN 22 implemented QIP-2, a real-world application of a modified EMPOWER program, by identifying eligible veterans on a rolling basis from December 2014 to August 2015. All veterans who were identified and sent an EB during this period were included in the case group. The index date was defined as the first of the month the EB was mailed. Veterans with a pending appointment were chosen because the lead time would allow them to receive the EB and prepare to discuss it with the physician during the visit.

A comparator group was drawn from the adjacent VISN 21 catchment area, which encompasses VA facilities and clinics in Hawaii, northern California, and northern Nevada. During the observation period, VISN 21 did not mail any EBs specifically addressing BZD risks. Veterans in the comparator group had an appointment scheduled with their BZD prescribing physician within 4 weeks, were aged ≥ 65 years on the index date (first of the month before the next appointment, coinciding with the date EBs were sent to VISN 22 veterans), were active long-term BZD users, and had at least 1 BZD prescription released within 200 days before the index date. All patients were followed for up to 12 months after the index date, with BZD discontinuation recorded 9 and 12 months after the index date.

Propensity Score Matching

Propensity score (PS) was estimated with logistic regression analysis with treatment as the dependent variable and baseline characteristics as the independent variables.^21,22 One-to-one matching on the PS was performed using the nearest neighbor approach without replacements. Independent variables related to outcome but unrelated to EB exposure were selected for PS development.²² These variables included year of birth; male sex; Hispanic ethnicity; annual income; service connection status; region; body mass index; Charlson Comorbidity Index category; total baseline BZD dose; and diagnosis of AIDS, nonmetastatic cancer, metastatic cancer, chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF), dementia, diabetes mellitus (DM), DM with complications, gastroesophageal reflux disease (GERD), general anxiety disorder (GAD), hemiparaplegia, liver disease (mild), liver disease (moderate to severe), myocardial infarction (MI), Parkinson disease, peptic ulcer disease (PUD), psychosis, renal disease, rheumatoid arthritis (RA), or substance use disorder (SUD).

The EMPOWER cluster randomized controlled trial (RCT) demonstrated the effectiveness of EB exposure in a Canadian population of elderly patients who were long-term BZD users.¹⁷ Randomized controlled trials are the gold standard for clinical trials because they can establish causal inference.^23-25 Given ethical and practical concerns, however, RCTs cannot be applied to all clinical scenarios. Although EMPOWER is reported to be an effective tool in reducing BZD use in older adults, its application in a real-world, large, integrated health care system remains untested. Observational studies are often conducted as an alternative to RCTs but are subject to selection bias because of their lack of randomization.²⁶ Therefore, robust research methods are needed to generate unbiased estimates of the impact of an intervention on an outcome. Propensity score matching simulates an RCT by balancing the covariates across treatment groups.^21,22,27 Observed patient characteristics are used to estimate PS, the probability that treatment will be received. Logistic or probit regression is used to balance the potential confounding covariates between the treatment groups.Once PSs are known, mean treatment effect can be estimated without the mean model.²⁸ In other words, PSM methods can be used to generate an unbiased estimate of the treatment.

Propensity Score Analysis

Baseline characteristics were compared using Student t test (continuous variables) and χ² test (discrete variables). Results are presented as means and standard deviations (continuous variables) and frequency and percentage (discrete variables).

The main outcome was BZD discontinuation 9 and 12 months after the index date. A postindex lag of 6 months was used to capture any tapering (Figure 2). Discontinuation, defined as 3 consecutive months of no BZD prescription on hand, was measured for 2 periods: 6 to 9 months and 6 to 12 months after the index date. A secondary outcome was discontinuation 1 to 12 months after the index date. An estimate was made of the difference in the proportions of BZD discontinuers who received the EB and BZD discontinuers who did not receive the EB, where mean treatment (risk difference) was presented as the absolute risk difference with a 95% CI. Standard errors and 95% CIs for the risk differences were generated with biased-corrected CIs from 1,000 bootstrap samples.

Sensitivity Analyses

Naïve multivariate logistic regression analysis was performed to evaluate the association between EB exposure and BZD discontinuation while controlling for potential confounders. Results are presented as odds ratios (ORs) and 95% CIs. Confounders identified were the same covariates used to generate the PSs.

Several analyses were performed to test the sensitivity of the methods applied using PSM by changing caliber size while maintaining the nearest neighbor approach without replacement. Linear regression analysis was performed with robust standard errors to estimate the risk difference of BZD discontinuation between EB-exposed and EB-unexposed veterans.

Statistical significance was set at P < .05. All statistical analyses were performed with Stata/SE Version 13 (College Station, TX).

Results

Quality Improvement Project 1

On a rolling basis from December 2014 to February 2016, the EB was mailed once to 3,896 VISN 22 veterans 2 to 8 weeks before a clinic appointment with their BZD prescribing physician. 

Quality Improvement Project 2

Of all the VISN 22 and VISN 21 veterans, 24,420 met the inclusion and exclusion criteria. Of these 24,420 veterans, 2,020 (8.3%) were in VISN 22 and received the EB between December 2014 and August 2015 (QIP-1), and 22,400 (91.7%) were in VISN 21 and did not receive the EB.

Naïve Results Before PS Matching. In the naïve analyses, a larger proportion of EB-exposed vs unexposed veterans discontinued BZD; in addition, reductions were 6.6%, 7.4%, and 9.5% larger for 6 to 9 months, 6 to 12 months, and 1 to 12 months after the index date, respectively (P < .0001 for all comparisons; Table 2).

Propensity Score Matching. Before matching, there were significant differences in baseline characteristics of veterans who met the inclusion and exclusion criteria, with few exceptions (eAppendices 2 and 3, ).

Propensity Score Matching Results. Inspection of PSs revealed good coverage across treatment groups on a histogram plot and a kernel density plot (eAppendices 5 and 6).

Discussion

This QIP was the first to evaluate the impact of an EMPOWER-modeled DTC EB in a large, integrated health care system in the U.S. It was also the first to demonstrate potential benefits of a DTC EB designed for older veterans who are long-term BZD users. In this QIP, which mailed the EB to 3,896 veterans, 1,847 (47.4%) decreased their BZD dose, 458 (11.7%) tapered and then discontinued BZD, and 455 (11.7%) immediately discontinued BZD. The total percentage of veterans who discontinued BZD (23.4%; 913/3,896) was similar to the 27% reported in the EMPOWER trial.¹⁷ However, the risk difference between the 1,316 EB-exposed VISN 22 veterans (QIP-1) and the 1,316 EB-unexposed VISN 21 veterans in this QIP was significantly lower than the 23% risk difference in EMPOWER (though it still demonstrated a significantly larger reduction for EB-exposed veterans).¹⁷

Given this inclusion of all qualifying veterans from the catchment area studied in this QIP, and given the ethical and practical concerns, an RCT was not possible. Therefore, PSM methods were used to balance the covariates across treatment groups and thereby simulate an RCT.^21,22,27 With use of the PSM approach, findings from the descriptive analysis were confirmed and potential selection bias reduced.

Study Limitations

The less robust risk difference found in this QIP has several possible explanations. The authors’ use of a DTC EB coincided with a national VA effort to reduce older veterans’ use of BZDs and other inappropriate medications. For instance, during the study period, academic detailing was being implemented to reduce use of BZDs, particularly in combination with opioids, across VHA facilities and clinics. (Academic detailing is a pharmacy educational outreach program that uses unbiased clinical guidelines to promote physicians’ safety initiatives and align prescribing behavior with best practices.^18-20) However, QIP-2 results and PS analysis of a subgroup of the original sample suggest that EB-exposed veterans were significantly more likely than were their unexposed counterparts were to discontinue BZD. To an extent, this analysis controlled for these other efforts to reduce BZD use in VHA clinics and can be considered a study strength.

Another limitation is the study design, which lacked a control group and did not consider the possibility that some facility or clinic physicians might influence others. Although the region variable was controlled for in PSM, the authors did not capture facility characteristics, including frequency of prescribing BZD and use of a protocol for enforcing the Beers criteria. Such confounders might have influenced outcomes. Unlike the EMPOWER trial,¹⁷ this QIP did not assess or exclude cognitively impaired veterans. It is reasonable to assume that these veterans might not understand some EB messages and consequently might fail to engage their physicians. Failure to initiate discussion with a physician would attenuate the impact of the EB.

Study Strengths

A strength of this QIP was its use of a DTC EB in a large, regional sample of older veterans in a real-world clinical setting. In addition, the study group (EB-exposed veterans) and the comparator group (EB-unexposed veterans) were from similar geographic areas (primarily California and Nevada).

Conclusion

Results of this study suggest that a DTC EB, designed to reduce BZD use among older veterans, was effective in helping patients lower their BZD dose and discontinue BZD. The likelihood of discontinuing BZD 9 and 12 months after the index date was significantly higher for veterans who received an EB modeled on the EMPOWER educational brochure than for a comparator group of veterans who did not receive the EB and were receiving care during the same observation period. In the future, it would be beneficial to use a design that controls for physician exposure to academic detailing focused on BZD reduction and that accounts for the cluster effects of facility practice. Despite these limitations, this QIP is the first real-world empirical example of using an EMPOWER-modeled DTC EB to decrease BZD use among older veterans. Furthermore, these results suggest that a DTC EB can be used to target other high-risk prescription drugs, such as opioids, particularly if alternative treatment options can be provided.

Acknowledgments
Dr. Hauser thanks Cathy, Anika, Katia, and Max Hauser, and Alba and Kevin Quinlan, for their support. In memory of Jirina Hauser, who died on Mother’s Day, May 14, 2017, at the age of 100.