Helene R. Labonte, DO

The advent of work‐hour restrictions and admission limits for teaching services has led many academic hospitals to implement hospitalist‐run staff (ie, nonteaching) services.[1] Although this practice is not new,[2] it is growing in popularity[3] and has been endorsed as a way to protect resident teaching and prevent excessive workload.[4] One potential benefit is the assignment of more educational cases to teaching services, whereas the nonteaching services receive more patients whose care is presumably relatively mundane or routine.[5]

Despite the rapid growth of this system of educational triage,[6] little is known about the factors considered when teaching versus nonteaching decisions are made. Studies of clinical outcomes for patients assigned to teaching versus nonteaching services have understandably used random assignment,[7, 8] whereas a study finding that patients with unhealthy substance use were more likely to be on teaching services than nonteaching services relied on patient assignment based on the identity of the patient's primary care provider or insurer.[9] In 2009, O'Connor et al. reported that implementation of nonteaching services at 2 hospitals had led to unequal distribution of patients in terms of demographics, diagnosis, and illness severity.[10] Triage decisions were made by either a nurse coordinator or a medical chief resident, and sicker patients (and occasionally good teaching cases) were preferentially placed on the teaching services, reportedly out of respect for the comfort level of the midlevel providers who staffed the nonteaching services.

Our institution has used a system of hospitalist educational triage since 1998. Over that time, residents have often expressed concerns about the assignment of patients to the teaching services, reporting in particular that they receive a disproportionate number of complex cases and outside transfers. In 2006, the hospitalist group attempted to address these concerns by collecting real‐time admission data, but the application of the data was limited by suspicion on both sides of a Hawthorne effect (data not published).

If trainee and hospitalist expectations for what constitutes a great teaching case differ substantially, that difference can have significant implications for resident and medical student teaching, self‐perceived roles, and satisfaction. More significantly, an understanding of what faculty perceive as ideal teaching cases would provide valuable information about the strengths and weaknesses of the teachingnonteaching model, which may prove useful to other academic institutions considering such a system. In this study, we endeavored to understand what residents and hospitalists consider an educational admission and to compare these expectations to the actual triage decisions of hospitalists.

METHODS

Mayo Clinic Hospital (Phoenix, Arizona) has used separate teaching and nonteaching services since opening in 1998. At our institution, like many others,[11] a hospitalist is assigned to take all calls for emergency department (ED) admissions, admissions from outpatient clinics, and transfer requests; this physician directs patients to the teaching or nonteaching service. At the time of our study, the 2 teaching services alternated days in which they admitted up to 7 patients, and the 5 nonteaching services admitted all other patients and provided medicine consultative services for the hospital. Teaching services consisted of 1 hospitalist, 2 senior residents, 2 or 3 first‐year residents, and sometimes 1 third‐ or fourth‐year medical student. Nonteaching services consisted of a hospitalist with intermittent assistance from a physician assistant or nurse practitioner.

Although there are no formal guidelines for the hospitalist triage role, hospitalists are encouraged to assign more educational cases to the teaching services and to allow the residents enough time to address the acute needs of the prior admission before receiving the next admission. Residents are not assigned any patients between 4:00 _am and 7:00 _am. The goals and objectives for the resident rotation on the medicine teaching service include a list of diagnoses with which residents are expected to become familiar during their residency; triage hospitalists have on‐line access to these goals and objectives.

To assess resident and hospitalist opinions about what types of patients should or should not be admitted to teaching services and to compare those characteristics with those of the patients actually admitted to teaching services, we began by administering a simple, open‐ended survey and asked both groups: (1) In an ideal world, what kinds of patients should be admitted to the internal medicine teaching services at Mayo Clinic Hospital? (2) In the real world, what kinds of patients are admitted to the internal medicine teaching services at Mayo Clinic Hospital?

Ample space was provided for free‐text entries. Residents were additionally asked their postgraduate year level. The survey was administered in April 2011, at which time all residents would have rotated on the medicine teaching services several times. Survey responses were anonymous and were compiled and retyped by someone unfamiliar with the subjects' handwriting.

Two authors (D.L.R. and H.R.L.) reviewed the results of the first survey and used conventional content analysis to group responses into categories and tally them.[12] Responses from hospitalists and residents were used to determine the content for a second, quantitative survey that asked respondents to rate specific possible factors that affected triage decisions on a Likert scale from 1 (Argues against teaching admission) to 5 (Argues for teaching admission). The second survey, administered to the same residents and hospitalists in May 2011, asked: (1) In an ideal world, how do these factors contribute to the decision about which patients should be admitted to the internal medicine teaching services at Mayo Clinic Hospital? (2) In the real world, how do these factors contribute to the decision about which patients are admitted to the internal medicine teaching services at Mayo Clinic Hospital?

Assuming a 3:1 ratio of nonteaching to teaching admissions, we calculated that we would need to analyze 1028 admissions to detect a 10% difference in the proportion of a specific trait present in 50% of patients admitted to the nonteaching service, with the use of a 2‐sided test with 80% statistical power and a significance level of 0.05.

We collected data on patient assignment via retrospective chart review to avoid the possibility of a Hawthorne effect. We studied all admissions to the internal medicine services for a 3‐month period before the administration of the first survey (January 1, 2011 through March 31, 2011). The following patient data were collected: service assignment (teaching vs nonteaching), age, sex, source of admission (ED, direct from clinic, outside transfer, internal transfer from another hospital service), first visit to our institution, prior hematology or oncology visit at our institution (as a surrogate for cancer), prior psychiatry visit at our institution (as a surrogate for psychiatric disease), transplantation history, human immunodeficiency virus (HIV) or acquired immune deficiency syndrome (AIDS) history, chronic or functional pain mentioned in ED or admission note, need for translator, and benefactor status. Additionally, an online calculator was used to determine the Charlson Comorbidity Index score for each patient.[13] We collected actual patient data corresponding to factors reported by survey respondents whenever possible and practical, but not every factor reported by survey respondents was amenable to rigorous analysis; for example, no unbiased method could be devised to rigorously categorize patients whose admissions are likely to take more time or difficult patients and families.

Responses to the second (quantitative) survey and patient data were compared using the Pearson [2] and Fisher exact test for categorical variables and the Student t test or Wilcoxon rank sum test for continuous variables. Categorical variables that achieved statistical significance for overall difference were analyzed on a post hoc basis using the Bonferroni method to control for the overall type I error rate. We also examined the differences between actual and ideal triage decisions using the Wilcoxon signed rank test. Data were analyzed using SAS 9.3 (SAS Institute, Inc., Cary, NC). Statistical significance was defined as P<0.05.

The project was deemed exempt by the Mayo Clinic institutional review board.

RESULTS

We surveyed all categorical internal medicine residents (n=30, 10 each from postgraduate year [PGY]‐1, PGY‐2, and PGY‐3) and hospitalists except the authors (n=21; average years since completing training=13.3; range, 129 years). For both surveys, responses were collected from 29 (96.7%) residents. The nonresponding resident was a PGY‐2. The response rate for hospitalists was 20/21 (95.2%) for the first survey and 16/21 (76.2%) for the second survey.

DISCUSSION

The results of our qualitative and quantitative surveys showed significant differences between resident and staff perceptions of the faculty triage role. Although both groups similarly valued many features, residents expressed a clear preference for more bread‐and‐butter admissions, whereas the staff prioritized selecting the most complex, challenging, and rare cases from among the day's admissions to give to the residents. (Residents were also very interested in rare cases, suggesting that they saw benefit to admitting patients with a variety of degrees of rarity and complexity.) Residents and faculty seemed to agree that the number of social admissions and outside transfers admitted to teaching services was not ideal.

These perceptions have substantial implications. If the current triage process is to continue, there may be benefit to designing a faculty development project focused on the triage process, which previously has been largely unexamined. Efforts to remove or limit time barriers that prevent perceived educational cases from being admitted to teaching services is also a worthy endeavor (eg, structuring the 2 teams to admit simultaneously so that teaching teams can admit patients back to back without exceeding capacity). In addition, residents may benefit from teaching hospitalists who concentrate educational efforts on the learning that can be extracted from the care of any patient, including admissions that initially seem mundane or purely social.[14] A concerted effort to divert more traditional medicine admissions and fewer unusual cases to the teaching service might improve resident perceptions of the triage process. Further, although the care of any patient can have education benefit, the fact that both groups perceived excessive social admissions in the teaching service suggests that a potential benefit of a nonteaching service (ie, absorbing the most mundane admissions) may not yet be fully realized.

Despite the perceived differences noted on the surveys, we found remarkably few differences between patients admitted to the teaching and nonteaching services. Although both groups rated complexity; outside transfers; being seen at the institution for the first time; and histories of transplantation, cancer, chronic or functional pain, and psychiatric disease as increasing the likelihood of admission to a teaching service, no differences were observed for these factors or their quantifiable surrogates. (Although the overall test for admission type achieved marginal statistical significance, none of the individual admission types were significantly different in post hoc analysis.) Residents, but not faculty, thought that older patients were over‐represented on the teaching service, but their assigned patients were significantly younger than those on the nonteaching service.

These findings have several possible explanations. First, although most hospitalists spend time on teaching and nonteaching services (and therefore are familiar with the patient composition of each), residents get very little exposure to the nonteaching services (until they are senior residents with a rotation on a consulting service). Their impression of inequity may be due to misunderstanding the patient composition of the nonteaching services. Second, the mere existence of a triage role may create false expectations about patient composition; that is, simply by knowing that every admission was chosen for its educational merit, residents may have disproportionate perceptions about those cases judged to have less educational value, even ifas our data suggestassignments to teaching versus nonteaching services are occurring fairly equitably.

CONCLUSION

Triage hospitalists distributed patients fairly evenly between teaching and nonteaching services, but residents and faculty alike perceived that residents would benefit from more bread‐and‐butter cases. Hospitals considering the addition of a nonteaching service may want to incorporate a faculty development project focused on the triage process to ensure that these traditional medicine cases are assigned to resident services and to ensure that the great teaching case is not considered such because of complexity and acuity alone.

The advent of work‐hour restrictions and admission limits for teaching services has led many academic hospitals to implement hospitalist‐run staff (ie, nonteaching) services.[1] Although this practice is not new,[2] it is growing in popularity[3] and has been endorsed as a way to protect resident teaching and prevent excessive workload.[4] One potential benefit is the assignment of more educational cases to teaching services, whereas the nonteaching services receive more patients whose care is presumably relatively mundane or routine.[5]

Despite the rapid growth of this system of educational triage,[6] little is known about the factors considered when teaching versus nonteaching decisions are made. Studies of clinical outcomes for patients assigned to teaching versus nonteaching services have understandably used random assignment,[7, 8] whereas a study finding that patients with unhealthy substance use were more likely to be on teaching services than nonteaching services relied on patient assignment based on the identity of the patient's primary care provider or insurer.[9] In 2009, O'Connor et al. reported that implementation of nonteaching services at 2 hospitals had led to unequal distribution of patients in terms of demographics, diagnosis, and illness severity.[10] Triage decisions were made by either a nurse coordinator or a medical chief resident, and sicker patients (and occasionally good teaching cases) were preferentially placed on the teaching services, reportedly out of respect for the comfort level of the midlevel providers who staffed the nonteaching services.

Our institution has used a system of hospitalist educational triage since 1998. Over that time, residents have often expressed concerns about the assignment of patients to the teaching services, reporting in particular that they receive a disproportionate number of complex cases and outside transfers. In 2006, the hospitalist group attempted to address these concerns by collecting real‐time admission data, but the application of the data was limited by suspicion on both sides of a Hawthorne effect (data not published).

If trainee and hospitalist expectations for what constitutes a great teaching case differ substantially, that difference can have significant implications for resident and medical student teaching, self‐perceived roles, and satisfaction. More significantly, an understanding of what faculty perceive as ideal teaching cases would provide valuable information about the strengths and weaknesses of the teachingnonteaching model, which may prove useful to other academic institutions considering such a system. In this study, we endeavored to understand what residents and hospitalists consider an educational admission and to compare these expectations to the actual triage decisions of hospitalists.

METHODS

Mayo Clinic Hospital (Phoenix, Arizona) has used separate teaching and nonteaching services since opening in 1998. At our institution, like many others,[11] a hospitalist is assigned to take all calls for emergency department (ED) admissions, admissions from outpatient clinics, and transfer requests; this physician directs patients to the teaching or nonteaching service. At the time of our study, the 2 teaching services alternated days in which they admitted up to 7 patients, and the 5 nonteaching services admitted all other patients and provided medicine consultative services for the hospital. Teaching services consisted of 1 hospitalist, 2 senior residents, 2 or 3 first‐year residents, and sometimes 1 third‐ or fourth‐year medical student. Nonteaching services consisted of a hospitalist with intermittent assistance from a physician assistant or nurse practitioner.

Although there are no formal guidelines for the hospitalist triage role, hospitalists are encouraged to assign more educational cases to the teaching services and to allow the residents enough time to address the acute needs of the prior admission before receiving the next admission. Residents are not assigned any patients between 4:00 _am and 7:00 _am. The goals and objectives for the resident rotation on the medicine teaching service include a list of diagnoses with which residents are expected to become familiar during their residency; triage hospitalists have on‐line access to these goals and objectives.

To assess resident and hospitalist opinions about what types of patients should or should not be admitted to teaching services and to compare those characteristics with those of the patients actually admitted to teaching services, we began by administering a simple, open‐ended survey and asked both groups: (1) In an ideal world, what kinds of patients should be admitted to the internal medicine teaching services at Mayo Clinic Hospital? (2) In the real world, what kinds of patients are admitted to the internal medicine teaching services at Mayo Clinic Hospital?

Ample space was provided for free‐text entries. Residents were additionally asked their postgraduate year level. The survey was administered in April 2011, at which time all residents would have rotated on the medicine teaching services several times. Survey responses were anonymous and were compiled and retyped by someone unfamiliar with the subjects' handwriting.

Two authors (D.L.R. and H.R.L.) reviewed the results of the first survey and used conventional content analysis to group responses into categories and tally them.[12] Responses from hospitalists and residents were used to determine the content for a second, quantitative survey that asked respondents to rate specific possible factors that affected triage decisions on a Likert scale from 1 (Argues against teaching admission) to 5 (Argues for teaching admission). The second survey, administered to the same residents and hospitalists in May 2011, asked: (1) In an ideal world, how do these factors contribute to the decision about which patients should be admitted to the internal medicine teaching services at Mayo Clinic Hospital? (2) In the real world, how do these factors contribute to the decision about which patients are admitted to the internal medicine teaching services at Mayo Clinic Hospital?

Assuming a 3:1 ratio of nonteaching to teaching admissions, we calculated that we would need to analyze 1028 admissions to detect a 10% difference in the proportion of a specific trait present in 50% of patients admitted to the nonteaching service, with the use of a 2‐sided test with 80% statistical power and a significance level of 0.05.

We collected data on patient assignment via retrospective chart review to avoid the possibility of a Hawthorne effect. We studied all admissions to the internal medicine services for a 3‐month period before the administration of the first survey (January 1, 2011 through March 31, 2011). The following patient data were collected: service assignment (teaching vs nonteaching), age, sex, source of admission (ED, direct from clinic, outside transfer, internal transfer from another hospital service), first visit to our institution, prior hematology or oncology visit at our institution (as a surrogate for cancer), prior psychiatry visit at our institution (as a surrogate for psychiatric disease), transplantation history, human immunodeficiency virus (HIV) or acquired immune deficiency syndrome (AIDS) history, chronic or functional pain mentioned in ED or admission note, need for translator, and benefactor status. Additionally, an online calculator was used to determine the Charlson Comorbidity Index score for each patient.[13] We collected actual patient data corresponding to factors reported by survey respondents whenever possible and practical, but not every factor reported by survey respondents was amenable to rigorous analysis; for example, no unbiased method could be devised to rigorously categorize patients whose admissions are likely to take more time or difficult patients and families.

Responses to the second (quantitative) survey and patient data were compared using the Pearson [2] and Fisher exact test for categorical variables and the Student t test or Wilcoxon rank sum test for continuous variables. Categorical variables that achieved statistical significance for overall difference were analyzed on a post hoc basis using the Bonferroni method to control for the overall type I error rate. We also examined the differences between actual and ideal triage decisions using the Wilcoxon signed rank test. Data were analyzed using SAS 9.3 (SAS Institute, Inc., Cary, NC). Statistical significance was defined as P<0.05.

The project was deemed exempt by the Mayo Clinic institutional review board.

RESULTS

We surveyed all categorical internal medicine residents (n=30, 10 each from postgraduate year [PGY]‐1, PGY‐2, and PGY‐3) and hospitalists except the authors (n=21; average years since completing training=13.3; range, 129 years). For both surveys, responses were collected from 29 (96.7%) residents. The nonresponding resident was a PGY‐2. The response rate for hospitalists was 20/21 (95.2%) for the first survey and 16/21 (76.2%) for the second survey.

DISCUSSION

The results of our qualitative and quantitative surveys showed significant differences between resident and staff perceptions of the faculty triage role. Although both groups similarly valued many features, residents expressed a clear preference for more bread‐and‐butter admissions, whereas the staff prioritized selecting the most complex, challenging, and rare cases from among the day's admissions to give to the residents. (Residents were also very interested in rare cases, suggesting that they saw benefit to admitting patients with a variety of degrees of rarity and complexity.) Residents and faculty seemed to agree that the number of social admissions and outside transfers admitted to teaching services was not ideal.

These perceptions have substantial implications. If the current triage process is to continue, there may be benefit to designing a faculty development project focused on the triage process, which previously has been largely unexamined. Efforts to remove or limit time barriers that prevent perceived educational cases from being admitted to teaching services is also a worthy endeavor (eg, structuring the 2 teams to admit simultaneously so that teaching teams can admit patients back to back without exceeding capacity). In addition, residents may benefit from teaching hospitalists who concentrate educational efforts on the learning that can be extracted from the care of any patient, including admissions that initially seem mundane or purely social.[14] A concerted effort to divert more traditional medicine admissions and fewer unusual cases to the teaching service might improve resident perceptions of the triage process. Further, although the care of any patient can have education benefit, the fact that both groups perceived excessive social admissions in the teaching service suggests that a potential benefit of a nonteaching service (ie, absorbing the most mundane admissions) may not yet be fully realized.

Despite the perceived differences noted on the surveys, we found remarkably few differences between patients admitted to the teaching and nonteaching services. Although both groups rated complexity; outside transfers; being seen at the institution for the first time; and histories of transplantation, cancer, chronic or functional pain, and psychiatric disease as increasing the likelihood of admission to a teaching service, no differences were observed for these factors or their quantifiable surrogates. (Although the overall test for admission type achieved marginal statistical significance, none of the individual admission types were significantly different in post hoc analysis.) Residents, but not faculty, thought that older patients were over‐represented on the teaching service, but their assigned patients were significantly younger than those on the nonteaching service.

These findings have several possible explanations. First, although most hospitalists spend time on teaching and nonteaching services (and therefore are familiar with the patient composition of each), residents get very little exposure to the nonteaching services (until they are senior residents with a rotation on a consulting service). Their impression of inequity may be due to misunderstanding the patient composition of the nonteaching services. Second, the mere existence of a triage role may create false expectations about patient composition; that is, simply by knowing that every admission was chosen for its educational merit, residents may have disproportionate perceptions about those cases judged to have less educational value, even ifas our data suggestassignments to teaching versus nonteaching services are occurring fairly equitably.

CONCLUSION

Triage hospitalists distributed patients fairly evenly between teaching and nonteaching services, but residents and faculty alike perceived that residents would benefit from more bread‐and‐butter cases. Hospitals considering the addition of a nonteaching service may want to incorporate a faculty development project focused on the triage process to ensure that these traditional medicine cases are assigned to resident services and to ensure that the great teaching case is not considered such because of complexity and acuity alone.

The advent of work‐hour restrictions and admission limits for teaching services has led many academic hospitals to implement hospitalist‐run staff (ie, nonteaching) services.[1] Although this practice is not new,[2] it is growing in popularity[3] and has been endorsed as a way to protect resident teaching and prevent excessive workload.[4] One potential benefit is the assignment of more educational cases to teaching services, whereas the nonteaching services receive more patients whose care is presumably relatively mundane or routine.[5]

Despite the rapid growth of this system of educational triage,[6] little is known about the factors considered when teaching versus nonteaching decisions are made. Studies of clinical outcomes for patients assigned to teaching versus nonteaching services have understandably used random assignment,[7, 8] whereas a study finding that patients with unhealthy substance use were more likely to be on teaching services than nonteaching services relied on patient assignment based on the identity of the patient's primary care provider or insurer.[9] In 2009, O'Connor et al. reported that implementation of nonteaching services at 2 hospitals had led to unequal distribution of patients in terms of demographics, diagnosis, and illness severity.[10] Triage decisions were made by either a nurse coordinator or a medical chief resident, and sicker patients (and occasionally good teaching cases) were preferentially placed on the teaching services, reportedly out of respect for the comfort level of the midlevel providers who staffed the nonteaching services.

Our institution has used a system of hospitalist educational triage since 1998. Over that time, residents have often expressed concerns about the assignment of patients to the teaching services, reporting in particular that they receive a disproportionate number of complex cases and outside transfers. In 2006, the hospitalist group attempted to address these concerns by collecting real‐time admission data, but the application of the data was limited by suspicion on both sides of a Hawthorne effect (data not published).

If trainee and hospitalist expectations for what constitutes a great teaching case differ substantially, that difference can have significant implications for resident and medical student teaching, self‐perceived roles, and satisfaction. More significantly, an understanding of what faculty perceive as ideal teaching cases would provide valuable information about the strengths and weaknesses of the teachingnonteaching model, which may prove useful to other academic institutions considering such a system. In this study, we endeavored to understand what residents and hospitalists consider an educational admission and to compare these expectations to the actual triage decisions of hospitalists.

METHODS

Mayo Clinic Hospital (Phoenix, Arizona) has used separate teaching and nonteaching services since opening in 1998. At our institution, like many others,[11] a hospitalist is assigned to take all calls for emergency department (ED) admissions, admissions from outpatient clinics, and transfer requests; this physician directs patients to the teaching or nonteaching service. At the time of our study, the 2 teaching services alternated days in which they admitted up to 7 patients, and the 5 nonteaching services admitted all other patients and provided medicine consultative services for the hospital. Teaching services consisted of 1 hospitalist, 2 senior residents, 2 or 3 first‐year residents, and sometimes 1 third‐ or fourth‐year medical student. Nonteaching services consisted of a hospitalist with intermittent assistance from a physician assistant or nurse practitioner.

Although there are no formal guidelines for the hospitalist triage role, hospitalists are encouraged to assign more educational cases to the teaching services and to allow the residents enough time to address the acute needs of the prior admission before receiving the next admission. Residents are not assigned any patients between 4:00 _am and 7:00 _am. The goals and objectives for the resident rotation on the medicine teaching service include a list of diagnoses with which residents are expected to become familiar during their residency; triage hospitalists have on‐line access to these goals and objectives.

To assess resident and hospitalist opinions about what types of patients should or should not be admitted to teaching services and to compare those characteristics with those of the patients actually admitted to teaching services, we began by administering a simple, open‐ended survey and asked both groups: (1) In an ideal world, what kinds of patients should be admitted to the internal medicine teaching services at Mayo Clinic Hospital? (2) In the real world, what kinds of patients are admitted to the internal medicine teaching services at Mayo Clinic Hospital?

Ample space was provided for free‐text entries. Residents were additionally asked their postgraduate year level. The survey was administered in April 2011, at which time all residents would have rotated on the medicine teaching services several times. Survey responses were anonymous and were compiled and retyped by someone unfamiliar with the subjects' handwriting.

Two authors (D.L.R. and H.R.L.) reviewed the results of the first survey and used conventional content analysis to group responses into categories and tally them.[12] Responses from hospitalists and residents were used to determine the content for a second, quantitative survey that asked respondents to rate specific possible factors that affected triage decisions on a Likert scale from 1 (Argues against teaching admission) to 5 (Argues for teaching admission). The second survey, administered to the same residents and hospitalists in May 2011, asked: (1) In an ideal world, how do these factors contribute to the decision about which patients should be admitted to the internal medicine teaching services at Mayo Clinic Hospital? (2) In the real world, how do these factors contribute to the decision about which patients are admitted to the internal medicine teaching services at Mayo Clinic Hospital?

Assuming a 3:1 ratio of nonteaching to teaching admissions, we calculated that we would need to analyze 1028 admissions to detect a 10% difference in the proportion of a specific trait present in 50% of patients admitted to the nonteaching service, with the use of a 2‐sided test with 80% statistical power and a significance level of 0.05.

We collected data on patient assignment via retrospective chart review to avoid the possibility of a Hawthorne effect. We studied all admissions to the internal medicine services for a 3‐month period before the administration of the first survey (January 1, 2011 through March 31, 2011). The following patient data were collected: service assignment (teaching vs nonteaching), age, sex, source of admission (ED, direct from clinic, outside transfer, internal transfer from another hospital service), first visit to our institution, prior hematology or oncology visit at our institution (as a surrogate for cancer), prior psychiatry visit at our institution (as a surrogate for psychiatric disease), transplantation history, human immunodeficiency virus (HIV) or acquired immune deficiency syndrome (AIDS) history, chronic or functional pain mentioned in ED or admission note, need for translator, and benefactor status. Additionally, an online calculator was used to determine the Charlson Comorbidity Index score for each patient.[13] We collected actual patient data corresponding to factors reported by survey respondents whenever possible and practical, but not every factor reported by survey respondents was amenable to rigorous analysis; for example, no unbiased method could be devised to rigorously categorize patients whose admissions are likely to take more time or difficult patients and families.

Responses to the second (quantitative) survey and patient data were compared using the Pearson [2] and Fisher exact test for categorical variables and the Student t test or Wilcoxon rank sum test for continuous variables. Categorical variables that achieved statistical significance for overall difference were analyzed on a post hoc basis using the Bonferroni method to control for the overall type I error rate. We also examined the differences between actual and ideal triage decisions using the Wilcoxon signed rank test. Data were analyzed using SAS 9.3 (SAS Institute, Inc., Cary, NC). Statistical significance was defined as P<0.05.

The project was deemed exempt by the Mayo Clinic institutional review board.

RESULTS

We surveyed all categorical internal medicine residents (n=30, 10 each from postgraduate year [PGY]‐1, PGY‐2, and PGY‐3) and hospitalists except the authors (n=21; average years since completing training=13.3; range, 129 years). For both surveys, responses were collected from 29 (96.7%) residents. The nonresponding resident was a PGY‐2. The response rate for hospitalists was 20/21 (95.2%) for the first survey and 16/21 (76.2%) for the second survey.

DISCUSSION

The results of our qualitative and quantitative surveys showed significant differences between resident and staff perceptions of the faculty triage role. Although both groups similarly valued many features, residents expressed a clear preference for more bread‐and‐butter admissions, whereas the staff prioritized selecting the most complex, challenging, and rare cases from among the day's admissions to give to the residents. (Residents were also very interested in rare cases, suggesting that they saw benefit to admitting patients with a variety of degrees of rarity and complexity.) Residents and faculty seemed to agree that the number of social admissions and outside transfers admitted to teaching services was not ideal.

These perceptions have substantial implications. If the current triage process is to continue, there may be benefit to designing a faculty development project focused on the triage process, which previously has been largely unexamined. Efforts to remove or limit time barriers that prevent perceived educational cases from being admitted to teaching services is also a worthy endeavor (eg, structuring the 2 teams to admit simultaneously so that teaching teams can admit patients back to back without exceeding capacity). In addition, residents may benefit from teaching hospitalists who concentrate educational efforts on the learning that can be extracted from the care of any patient, including admissions that initially seem mundane or purely social.[14] A concerted effort to divert more traditional medicine admissions and fewer unusual cases to the teaching service might improve resident perceptions of the triage process. Further, although the care of any patient can have education benefit, the fact that both groups perceived excessive social admissions in the teaching service suggests that a potential benefit of a nonteaching service (ie, absorbing the most mundane admissions) may not yet be fully realized.

Despite the perceived differences noted on the surveys, we found remarkably few differences between patients admitted to the teaching and nonteaching services. Although both groups rated complexity; outside transfers; being seen at the institution for the first time; and histories of transplantation, cancer, chronic or functional pain, and psychiatric disease as increasing the likelihood of admission to a teaching service, no differences were observed for these factors or their quantifiable surrogates. (Although the overall test for admission type achieved marginal statistical significance, none of the individual admission types were significantly different in post hoc analysis.) Residents, but not faculty, thought that older patients were over‐represented on the teaching service, but their assigned patients were significantly younger than those on the nonteaching service.

These findings have several possible explanations. First, although most hospitalists spend time on teaching and nonteaching services (and therefore are familiar with the patient composition of each), residents get very little exposure to the nonteaching services (until they are senior residents with a rotation on a consulting service). Their impression of inequity may be due to misunderstanding the patient composition of the nonteaching services. Second, the mere existence of a triage role may create false expectations about patient composition; that is, simply by knowing that every admission was chosen for its educational merit, residents may have disproportionate perceptions about those cases judged to have less educational value, even ifas our data suggestassignments to teaching versus nonteaching services are occurring fairly equitably.

CONCLUSION

Triage hospitalists distributed patients fairly evenly between teaching and nonteaching services, but residents and faculty alike perceived that residents would benefit from more bread‐and‐butter cases. Hospitals considering the addition of a nonteaching service may want to incorporate a faculty development project focused on the triage process to ensure that these traditional medicine cases are assigned to resident services and to ensure that the great teaching case is not considered such because of complexity and acuity alone.