Dicharge of a patient from the hospital is a complicated, interprofessional endeavor.1, 2 Several institutions report that discharge is one of the least satisfying elements of the patient's hospital experience.35 Recent evidence suggests that a poorly planned or disorganized discharge may compromise patient safety in the period soon after dismissal.6 Several initiatives have been aimed at improving patient satisfaction and safety related to discharge.710

In 2000 the Mayo Clinic (Rochester, Minnesota) Department of Internal Medicine leadership established a goal to improve patient satisfaction with the hospital dismissal process. Patient focus group data suggested that uncertainty about the anticipated date and time of discharge causes frustration to some patients and families.

We hypothesized that an appointment to leave the hospital might be practicable. We joined an Institute for Healthcare Improvement collaborative (Improving Flow Through Acute Care Settings, 1 of 6 Improvement Action Network [IMPACT] Learning and Innovation Communities) aimed at scheduling discharge appointments (DAs). The collaborating members deemed that, although the ideal DA is set at least a day in advance, a same‐day DA is also desirable for both patient satisfaction and staff task organization in pursuit of a high‐quality discharge.

METHODS

This project was approved by the Mayo Foundation Institutional Review Board. We tested the following hypotheses:

• 1

  It is possible to make and display DAs in various care units.

• 2

  Most DAs can be scheduled a day before dismissal.

• 3

  Most DA patients depart on time.

Preliminary Work: Design Pilot

We designed bedside dry‐erase wall displays and mounted them in the rooms in plain view of patients and their families and caregivers. Pilot testing of DA scheduling was done on a general medical care unit from August 2 to December 24, 2003. To optimize the process for scheduling a DA, our team developed 21 small tests to change the dismissal process through plan, do, study, and act cycles.11

The recommended process was that as soon as an organized discharge could be reasonably envisioned, the primary service provider would discuss with the patient, family, and primary nurse (and a social service worker, if involved) the anticipated discharge day. A member of the primary service was to handwrite (with a marker) the anticipated day on the specially designed bedside dry‐erase board (Fig. 1) in view of the patient. The same primary service prescribers could amend this anticipated day (or time) by repeating the process of consultation and discussion as needed. The time of the DA could be written on the DA board (or amended) by either a member of the primary service or the primary care nurse.

Figure 1

Each morning, the primary care nurse transmitted the DA board data to the admission, discharge, and transfer log kept at the unit secretary desk (in which the actual discharge time has always been routinely recorded by the unit secretary).

RESULTS

During the 4‐month study of discharges across 8 care units, 1256 of 2046 patients (61%) received a DA; 576 of the DAs (46%) were scheduled at least 1 day in advance (Table 1). Among patients with a DA, 752 were discharged on time (60%), and only 240 (19%) were tardy.

DISCUSSION

In response to patient focus group feedback, we designed a tool and a process by which a DA could be made and posted at bedside. Among 2046 patients discharged from 8 care units over 4 months, 61% (1256) had a posted, in‐room DA. Almost half the patients with DAs (46%) had a DA scheduled at least 1 calendar day ahead. Remarkably, among patients with a DA, fewer than 20% were discharged tardily. In‐room posting of DAs across a spectrum of care units appears to be practicable, even in the face of extant diagnostic or therapeutic uncertainty.

This was an initial test‐of‐concept project and an exploratory trial. The limitations are: (1) satisfaction (patient, family, nurse, and physician) was not tested with any validated survey instrument, (2) length of stay was not studied, (3) reasons for variable DA success among care units were not ascertained, and (4) resource use was not measured.

Anecdotal information from a postdischarge phone survey indicated that patients seemed appreciative of a DA. The survey data were not included in this article because the survey tool was not a validated instrument and the interviewer (a coauthor) was not blinded to the hypothesis and was therefore subject to bias. No negative comments were received through informal real‐time feedback from patients and family during the making and posting of DAs, and encouraging comments were common.

Physician participation in posting the DA appeared to be key, and the unavoidable dialogue about the clinical rationale for a chosen date seemed welcome. A telling anecdote came from a patient who did not have a DA board: I didn't get the same treatment as my roommate with the [DA] board. The other doctors talked with [him] more about discharge. I wish my team would have done this more with me.

We cannot be certain of the reasons for the care unit disparity in setting and meeting DAs. We speculate that the level of staff enthusiasm for DAs explains the variation rather than patient population characteristics. Further, we cannot explain why 39% of the patients did not receive a DA. Physician feedback was generally, but not uniformly, positive. Negative comments that might explain DA omissions include: (1) patients already are informed and awarethe tool is superfluous; (2) the day of discharge is unknowable in advance; and (3) patients or family members will hold us to it or be upset if the DA is changed.

We expected that diagnostic uncertainty might pose challenges to providing DAs. When primary service providers were reassured that DAs could be amended, this concern was reduced (but not eliminated). It seemed useful for providers to envision the earliest day of discharge by assuming that the results of a pending key test or consultation would be favorable. Frequency of DA modification was not studied. DAs were amended, however, and patients (to our knowledge) seemed unperturbedperhaps because of an almost unavoidable discussion of the clinical rationale because the act of posting the DA occurred in full view of (and in partnership with) the patient.

A trend toward discharge earlier in the day was observed (data not shown). Theoretically, such a trend offers the potential to improve inpatient flow, in part by discharging patients before morning surgical cases are completed.

Although we had many favorable comments about DAs from patients, family members, and nurses, satisfaction of patients, families, and staff members deserves formal study. Further, it is not known whether unused DA boards might contribute to patient dissatisfaction. Any effect that the display of DAs may have on the length of stay also may be a topic worthy of future study.

CONCLUSIONS

Patients and their families sometimes desire more communication about the anticipated day and time of hospital discharge. We designed a process for making a tentative DA and a tool by which the DA could be posted at the bedside. The results of this study suggest that (1) despite some uncertainty it is possible to schedule and post DAs in‐room in various care units and in various settings, (2) DAs were made at least a day ahead of time in almost half the DA discharges, and (3) most DA discharges were characterized by on‐time departure. In addition, patient, family, and nursing satisfaction (in relation to the DA) warrants further investigation.

Dicharge of a patient from the hospital is a complicated, interprofessional endeavor.1, 2 Several institutions report that discharge is one of the least satisfying elements of the patient's hospital experience.35 Recent evidence suggests that a poorly planned or disorganized discharge may compromise patient safety in the period soon after dismissal.6 Several initiatives have been aimed at improving patient satisfaction and safety related to discharge.710

In 2000 the Mayo Clinic (Rochester, Minnesota) Department of Internal Medicine leadership established a goal to improve patient satisfaction with the hospital dismissal process. Patient focus group data suggested that uncertainty about the anticipated date and time of discharge causes frustration to some patients and families.

We hypothesized that an appointment to leave the hospital might be practicable. We joined an Institute for Healthcare Improvement collaborative (Improving Flow Through Acute Care Settings, 1 of 6 Improvement Action Network [IMPACT] Learning and Innovation Communities) aimed at scheduling discharge appointments (DAs). The collaborating members deemed that, although the ideal DA is set at least a day in advance, a same‐day DA is also desirable for both patient satisfaction and staff task organization in pursuit of a high‐quality discharge.

METHODS

This project was approved by the Mayo Foundation Institutional Review Board. We tested the following hypotheses:

• 1

  It is possible to make and display DAs in various care units.

• 2

  Most DAs can be scheduled a day before dismissal.

• 3

  Most DA patients depart on time.

Preliminary Work: Design Pilot

We designed bedside dry‐erase wall displays and mounted them in the rooms in plain view of patients and their families and caregivers. Pilot testing of DA scheduling was done on a general medical care unit from August 2 to December 24, 2003. To optimize the process for scheduling a DA, our team developed 21 small tests to change the dismissal process through plan, do, study, and act cycles.11

The recommended process was that as soon as an organized discharge could be reasonably envisioned, the primary service provider would discuss with the patient, family, and primary nurse (and a social service worker, if involved) the anticipated discharge day. A member of the primary service was to handwrite (with a marker) the anticipated day on the specially designed bedside dry‐erase board (Fig. 1) in view of the patient. The same primary service prescribers could amend this anticipated day (or time) by repeating the process of consultation and discussion as needed. The time of the DA could be written on the DA board (or amended) by either a member of the primary service or the primary care nurse.

Figure 1

Each morning, the primary care nurse transmitted the DA board data to the admission, discharge, and transfer log kept at the unit secretary desk (in which the actual discharge time has always been routinely recorded by the unit secretary).

RESULTS

During the 4‐month study of discharges across 8 care units, 1256 of 2046 patients (61%) received a DA; 576 of the DAs (46%) were scheduled at least 1 day in advance (Table 1). Among patients with a DA, 752 were discharged on time (60%), and only 240 (19%) were tardy.

DISCUSSION

In response to patient focus group feedback, we designed a tool and a process by which a DA could be made and posted at bedside. Among 2046 patients discharged from 8 care units over 4 months, 61% (1256) had a posted, in‐room DA. Almost half the patients with DAs (46%) had a DA scheduled at least 1 calendar day ahead. Remarkably, among patients with a DA, fewer than 20% were discharged tardily. In‐room posting of DAs across a spectrum of care units appears to be practicable, even in the face of extant diagnostic or therapeutic uncertainty.

This was an initial test‐of‐concept project and an exploratory trial. The limitations are: (1) satisfaction (patient, family, nurse, and physician) was not tested with any validated survey instrument, (2) length of stay was not studied, (3) reasons for variable DA success among care units were not ascertained, and (4) resource use was not measured.

Anecdotal information from a postdischarge phone survey indicated that patients seemed appreciative of a DA. The survey data were not included in this article because the survey tool was not a validated instrument and the interviewer (a coauthor) was not blinded to the hypothesis and was therefore subject to bias. No negative comments were received through informal real‐time feedback from patients and family during the making and posting of DAs, and encouraging comments were common.

Physician participation in posting the DA appeared to be key, and the unavoidable dialogue about the clinical rationale for a chosen date seemed welcome. A telling anecdote came from a patient who did not have a DA board: I didn't get the same treatment as my roommate with the [DA] board. The other doctors talked with [him] more about discharge. I wish my team would have done this more with me.

We cannot be certain of the reasons for the care unit disparity in setting and meeting DAs. We speculate that the level of staff enthusiasm for DAs explains the variation rather than patient population characteristics. Further, we cannot explain why 39% of the patients did not receive a DA. Physician feedback was generally, but not uniformly, positive. Negative comments that might explain DA omissions include: (1) patients already are informed and awarethe tool is superfluous; (2) the day of discharge is unknowable in advance; and (3) patients or family members will hold us to it or be upset if the DA is changed.

We expected that diagnostic uncertainty might pose challenges to providing DAs. When primary service providers were reassured that DAs could be amended, this concern was reduced (but not eliminated). It seemed useful for providers to envision the earliest day of discharge by assuming that the results of a pending key test or consultation would be favorable. Frequency of DA modification was not studied. DAs were amended, however, and patients (to our knowledge) seemed unperturbedperhaps because of an almost unavoidable discussion of the clinical rationale because the act of posting the DA occurred in full view of (and in partnership with) the patient.

A trend toward discharge earlier in the day was observed (data not shown). Theoretically, such a trend offers the potential to improve inpatient flow, in part by discharging patients before morning surgical cases are completed.

Although we had many favorable comments about DAs from patients, family members, and nurses, satisfaction of patients, families, and staff members deserves formal study. Further, it is not known whether unused DA boards might contribute to patient dissatisfaction. Any effect that the display of DAs may have on the length of stay also may be a topic worthy of future study.

CONCLUSIONS

Patients and their families sometimes desire more communication about the anticipated day and time of hospital discharge. We designed a process for making a tentative DA and a tool by which the DA could be posted at the bedside. The results of this study suggest that (1) despite some uncertainty it is possible to schedule and post DAs in‐room in various care units and in various settings, (2) DAs were made at least a day ahead of time in almost half the DA discharges, and (3) most DA discharges were characterized by on‐time departure. In addition, patient, family, and nursing satisfaction (in relation to the DA) warrants further investigation.

Dicharge of a patient from the hospital is a complicated, interprofessional endeavor.1, 2 Several institutions report that discharge is one of the least satisfying elements of the patient's hospital experience.35 Recent evidence suggests that a poorly planned or disorganized discharge may compromise patient safety in the period soon after dismissal.6 Several initiatives have been aimed at improving patient satisfaction and safety related to discharge.710

In 2000 the Mayo Clinic (Rochester, Minnesota) Department of Internal Medicine leadership established a goal to improve patient satisfaction with the hospital dismissal process. Patient focus group data suggested that uncertainty about the anticipated date and time of discharge causes frustration to some patients and families.

We hypothesized that an appointment to leave the hospital might be practicable. We joined an Institute for Healthcare Improvement collaborative (Improving Flow Through Acute Care Settings, 1 of 6 Improvement Action Network [IMPACT] Learning and Innovation Communities) aimed at scheduling discharge appointments (DAs). The collaborating members deemed that, although the ideal DA is set at least a day in advance, a same‐day DA is also desirable for both patient satisfaction and staff task organization in pursuit of a high‐quality discharge.

METHODS

This project was approved by the Mayo Foundation Institutional Review Board. We tested the following hypotheses:

• 1

  It is possible to make and display DAs in various care units.

• 2

  Most DAs can be scheduled a day before dismissal.

• 3

  Most DA patients depart on time.

Preliminary Work: Design Pilot

We designed bedside dry‐erase wall displays and mounted them in the rooms in plain view of patients and their families and caregivers. Pilot testing of DA scheduling was done on a general medical care unit from August 2 to December 24, 2003. To optimize the process for scheduling a DA, our team developed 21 small tests to change the dismissal process through plan, do, study, and act cycles.11

The recommended process was that as soon as an organized discharge could be reasonably envisioned, the primary service provider would discuss with the patient, family, and primary nurse (and a social service worker, if involved) the anticipated discharge day. A member of the primary service was to handwrite (with a marker) the anticipated day on the specially designed bedside dry‐erase board (Fig. 1) in view of the patient. The same primary service prescribers could amend this anticipated day (or time) by repeating the process of consultation and discussion as needed. The time of the DA could be written on the DA board (or amended) by either a member of the primary service or the primary care nurse.

Figure 1

Each morning, the primary care nurse transmitted the DA board data to the admission, discharge, and transfer log kept at the unit secretary desk (in which the actual discharge time has always been routinely recorded by the unit secretary).

RESULTS

During the 4‐month study of discharges across 8 care units, 1256 of 2046 patients (61%) received a DA; 576 of the DAs (46%) were scheduled at least 1 day in advance (Table 1). Among patients with a DA, 752 were discharged on time (60%), and only 240 (19%) were tardy.

DISCUSSION

In response to patient focus group feedback, we designed a tool and a process by which a DA could be made and posted at bedside. Among 2046 patients discharged from 8 care units over 4 months, 61% (1256) had a posted, in‐room DA. Almost half the patients with DAs (46%) had a DA scheduled at least 1 calendar day ahead. Remarkably, among patients with a DA, fewer than 20% were discharged tardily. In‐room posting of DAs across a spectrum of care units appears to be practicable, even in the face of extant diagnostic or therapeutic uncertainty.

This was an initial test‐of‐concept project and an exploratory trial. The limitations are: (1) satisfaction (patient, family, nurse, and physician) was not tested with any validated survey instrument, (2) length of stay was not studied, (3) reasons for variable DA success among care units were not ascertained, and (4) resource use was not measured.

Anecdotal information from a postdischarge phone survey indicated that patients seemed appreciative of a DA. The survey data were not included in this article because the survey tool was not a validated instrument and the interviewer (a coauthor) was not blinded to the hypothesis and was therefore subject to bias. No negative comments were received through informal real‐time feedback from patients and family during the making and posting of DAs, and encouraging comments were common.

Physician participation in posting the DA appeared to be key, and the unavoidable dialogue about the clinical rationale for a chosen date seemed welcome. A telling anecdote came from a patient who did not have a DA board: I didn't get the same treatment as my roommate with the [DA] board. The other doctors talked with [him] more about discharge. I wish my team would have done this more with me.

We cannot be certain of the reasons for the care unit disparity in setting and meeting DAs. We speculate that the level of staff enthusiasm for DAs explains the variation rather than patient population characteristics. Further, we cannot explain why 39% of the patients did not receive a DA. Physician feedback was generally, but not uniformly, positive. Negative comments that might explain DA omissions include: (1) patients already are informed and awarethe tool is superfluous; (2) the day of discharge is unknowable in advance; and (3) patients or family members will hold us to it or be upset if the DA is changed.

We expected that diagnostic uncertainty might pose challenges to providing DAs. When primary service providers were reassured that DAs could be amended, this concern was reduced (but not eliminated). It seemed useful for providers to envision the earliest day of discharge by assuming that the results of a pending key test or consultation would be favorable. Frequency of DA modification was not studied. DAs were amended, however, and patients (to our knowledge) seemed unperturbedperhaps because of an almost unavoidable discussion of the clinical rationale because the act of posting the DA occurred in full view of (and in partnership with) the patient.

A trend toward discharge earlier in the day was observed (data not shown). Theoretically, such a trend offers the potential to improve inpatient flow, in part by discharging patients before morning surgical cases are completed.

Although we had many favorable comments about DAs from patients, family members, and nurses, satisfaction of patients, families, and staff members deserves formal study. Further, it is not known whether unused DA boards might contribute to patient dissatisfaction. Any effect that the display of DAs may have on the length of stay also may be a topic worthy of future study.

CONCLUSIONS

Patients and their families sometimes desire more communication about the anticipated day and time of hospital discharge. We designed a process for making a tentative DA and a tool by which the DA could be posted at the bedside. The results of this study suggest that (1) despite some uncertainty it is possible to schedule and post DAs in‐room in various care units and in various settings, (2) DAs were made at least a day ahead of time in almost half the DA discharges, and (3) most DA discharges were characterized by on‐time departure. In addition, patient, family, and nursing satisfaction (in relation to the DA) warrants further investigation.

In 1996 Wachter and Goldman anticipated the emergence of hospitalists,1 physicians who are responsible for the care of hospitalized patients in place of their primary care physicians. The number of physicians who identify themselves as hospitalists has grown rapidly since 1996 and is currently estimated to be 10,00012,000, with the potential to reach as high as 30,000 in the next decade.2 This growth includes academic medical centers. In surveys of chairs of internal medicine and pediatric departments, 50% have hospitalists employed at their institutions.3, 4

Hospitalists in academic institutions are playing an increasingly prominent role in the medical education of both residents and medical students. The implications of adopting a hospitalist model on medical education has been discussed.57 Despite such concerns as fragmented continuity of care; decreased exposure to primary care physicians, subspecialists and physician‐scientists; reduced autonomy; and fewer educational opportunities to observe the natural histories of illnesses because of improved efficiency,57 the overall impact of hospitalists on medical and resident education has generally been favorable.818 Internal medicine residents have rated the teaching skills of hospitalists comparable to traditional academic physicians,8, 9 and believe the addition of hospitalists has contributed to an improved educational experience.10, 11, 14 In addition, a survey of third‐year medical students at a single academic teaching hospital concluded that hospitalists were able to provide at least as positive an educational experience during their inpatient medicine rotations as highly rated nonhospitalist teaching faculty.13

The role of hospitalists as educators in pediatrics has been studied much less. Pediatric resident satisfaction has improved in institutions that have used a hospitalist model.1618 In another study, hospitalists were rated by pediatric residents as more effective teachers than nonhospitalists.15 Because we are unaware of any study that has evaluated hospitalists in the education of medical students during their inpatient pediatric rotation, the purpose of our study was to compare hospitalist and nonhospitalist faculty on the educational experience of third‐year medical students during the inpatient portion of their pediatric clerkships at a single university children's hospital.

METHODS

RESULTS

All 67 of the students who completed a pediatrics clerkship at PSCH returned evaluation forms, which were the data for further analysis. Thirty‐five students rotated with the hospitalist faculty, and 32 students rotated with the nonhospitalist faculty. There were no significant demographic differences between the hospitalist and nonhospitalist faculty in age, sex, academic rank, specialty, and years since completing training (Table 2). All the hospitalist faculty fulfilled the definition of a hospitalist,2 whereas none of the physicians in the nonhospitalist group did.

The hospitalists were rated significantly higher than the nonhospitalist faculty in all 4 of the attending characteristics measured (Table 1): teaching effectiveness (3.87 vs. 2.91; P < .0001), effectiveness as a pediatrician (3.94 vs. 3.25; P < .001), student advocacy effectiveness (3.76 vs. 2.97; P < .0001), and overall evaluation (3.93 vs. 3.06; P < .001).

Analysis of specific aspects of the rotation showed the only feature that hospitalists were rated significantly higher on was quality of ward rounds (3.15 vs. 2.58, P < .006). There was no significant difference between the hospitalists and nonhospitalists on features that were not specifically part of the inpatient rotation, including various conferences, outpatient clinics, and newborn care (Table 1).

DISCUSSION

Our study demonstrates that pediatric hospitalists had a positive impact on the overall educational experience of third‐year medical students during the inpatient portions of their pediatrics clerkships. Hospitalists were rated more favorably than nonhospitalists as teachers, as pediatricians, as student advocates, and overall. Medical students also rated the value of ward rounds more favorably when hospitalists conducted them. In addition, higher percentages of nonhospitalists than hospitalists were rated as adequate or inadequate for the above items. When other aspects of the clerkship were analyzed, there were no statistically significant differences between the students who rotated with hospitalists and the students who rotated with nonhospitalists. This suggests that the higher scores for hospitalists were specifically related to their interactions with students, rather than with an overall more positive view of the rotation.

It has been suggested that forces promoting the use of hospitalists in adult medicine are even more persuasive in the pediatric population, as the difference in severity of illness between the inpatient and outpatient setting is greater, and the average pediatrician has less experience than the average internist in managing hospitalized patients.19 In a recent systematic review of the literature, Landrigan et al.20 reported that 6 of 7 studies demonstrated hospitalist systems had decreased hospital length of stay compared to systems in which a primary pediatrician served as the physician of record. This improved efficiency, if combined with the pressure to see more patients while trying to balance teaching and research demands, may have a negative impact on the quality of medical education.

Several factors may have contributed to the students' satisfaction with hospitalists. Studies have demonstrated that students rate clinical teachers more favorably with whom they have greater involvement.21 Hospitalists may be more likely to spend time on the inpatient wards given that is the primary site of their clinical activity. This increased presence may have contributed to more favorable evaluations for the hospitalist faculty, whereas the additional outpatient workload for nonhospitalist faculty may have reduced inpatient teaching opportunities, accounting for their lower teaching score. Included in the pediatrician category was the attribute of being a role model. In a study by Wright et al.,22 spending more than 25% of the time or 25 or more hours per week teaching and conducting rounds was independently associated with being considered an excellent role model. Again, the increased availability of the hospitalists on the inpatient wards may have led to more teaching opportunities, contributing to their higher score.

Our study had several limitations. First, it was a retrospective study conducted at a single institution with only 2 hospitalists. Although there were not statistical significant demographic differences between the 2 groups, this may simply reflect the small size of the sample in our study; therefore, the results may not be applicable to other academic institutions. Second, we retrospectively analyzed an evaluation form that had not been validated or specifically designed to compare 2 physician groups. Third, there were multiple statements in each category that students were asked to consider before scoring each attending on the parameters measured. Although hospitalists were rated higher in each category, there may have been individual characteristics within each category for which the nonhospitalist faculty performed better. Fourth, although hospitalists received higher average ratings than nonhospitalist faculty from third‐year medical students, it is important to emphasize this study measured students' attitudes and beliefs not specific educational outcomes. However, even though we cannot rule out the possibility that potentially confounding factors such as the personality of an attending physician influenced the results, prior studies have demonstrated that medical students make sophisticated judgments about teaching in the clinical setting.23, 24 It is unlikely that hospitalists at our institution were specifically selected to attend more months on a new inpatient service because they had a history of having more favorable teaching qualities because 1 of the 2 hospitalists had just finished residency training, and there were no significant demographic differences between the 2 groups. In a study examining trainee satisfaction in an internal medicine rotation 4 years after adoption of a hospitalist model, where nonhospitalist faculty attended based on their own interest and inpatient skill rather than as a requirement, Hauer et al.14 reported that trainees experienced more effective teaching and a more satisfying inpatient rotation when supervised by hospitalists. This suggests that hospitalists may possess or develop a specific inpatient knowledge base and teaching acumen over time that distinguishes them from nonhospitalists. There is evidence of accumulated experience leading to improved outcomes in the clinical setting for HIV infection,25 various surgical procedures,26 and hospitalist systems.27

In conclusion, this is the first study to evaluate the performance of hospitalists in the setting of a third‐year medical student pediatrics clerkship. Although third‐year medical students rate hospitalists at least as highly as nonhospitalist faculty, further studies are needed to reproduce this finding. In addition to the increased time spent on the wards with students and increased experience in caring for hospitalized patients, further studies should also examine the role that communication plays in clinical teaching. Also, the recent development of core competencies in hospital medicine28 may lead to the development of educational outcomes that can be objectively measured.

In 1996 Wachter and Goldman anticipated the emergence of hospitalists,1 physicians who are responsible for the care of hospitalized patients in place of their primary care physicians. The number of physicians who identify themselves as hospitalists has grown rapidly since 1996 and is currently estimated to be 10,00012,000, with the potential to reach as high as 30,000 in the next decade.2 This growth includes academic medical centers. In surveys of chairs of internal medicine and pediatric departments, 50% have hospitalists employed at their institutions.3, 4

Hospitalists in academic institutions are playing an increasingly prominent role in the medical education of both residents and medical students. The implications of adopting a hospitalist model on medical education has been discussed.57 Despite such concerns as fragmented continuity of care; decreased exposure to primary care physicians, subspecialists and physician‐scientists; reduced autonomy; and fewer educational opportunities to observe the natural histories of illnesses because of improved efficiency,57 the overall impact of hospitalists on medical and resident education has generally been favorable.818 Internal medicine residents have rated the teaching skills of hospitalists comparable to traditional academic physicians,8, 9 and believe the addition of hospitalists has contributed to an improved educational experience.10, 11, 14 In addition, a survey of third‐year medical students at a single academic teaching hospital concluded that hospitalists were able to provide at least as positive an educational experience during their inpatient medicine rotations as highly rated nonhospitalist teaching faculty.13

The role of hospitalists as educators in pediatrics has been studied much less. Pediatric resident satisfaction has improved in institutions that have used a hospitalist model.1618 In another study, hospitalists were rated by pediatric residents as more effective teachers than nonhospitalists.15 Because we are unaware of any study that has evaluated hospitalists in the education of medical students during their inpatient pediatric rotation, the purpose of our study was to compare hospitalist and nonhospitalist faculty on the educational experience of third‐year medical students during the inpatient portion of their pediatric clerkships at a single university children's hospital.

METHODS

RESULTS

All 67 of the students who completed a pediatrics clerkship at PSCH returned evaluation forms, which were the data for further analysis. Thirty‐five students rotated with the hospitalist faculty, and 32 students rotated with the nonhospitalist faculty. There were no significant demographic differences between the hospitalist and nonhospitalist faculty in age, sex, academic rank, specialty, and years since completing training (Table 2). All the hospitalist faculty fulfilled the definition of a hospitalist,2 whereas none of the physicians in the nonhospitalist group did.

The hospitalists were rated significantly higher than the nonhospitalist faculty in all 4 of the attending characteristics measured (Table 1): teaching effectiveness (3.87 vs. 2.91; P < .0001), effectiveness as a pediatrician (3.94 vs. 3.25; P < .001), student advocacy effectiveness (3.76 vs. 2.97; P < .0001), and overall evaluation (3.93 vs. 3.06; P < .001).

Analysis of specific aspects of the rotation showed the only feature that hospitalists were rated significantly higher on was quality of ward rounds (3.15 vs. 2.58, P < .006). There was no significant difference between the hospitalists and nonhospitalists on features that were not specifically part of the inpatient rotation, including various conferences, outpatient clinics, and newborn care (Table 1).

DISCUSSION

Our study demonstrates that pediatric hospitalists had a positive impact on the overall educational experience of third‐year medical students during the inpatient portions of their pediatrics clerkships. Hospitalists were rated more favorably than nonhospitalists as teachers, as pediatricians, as student advocates, and overall. Medical students also rated the value of ward rounds more favorably when hospitalists conducted them. In addition, higher percentages of nonhospitalists than hospitalists were rated as adequate or inadequate for the above items. When other aspects of the clerkship were analyzed, there were no statistically significant differences between the students who rotated with hospitalists and the students who rotated with nonhospitalists. This suggests that the higher scores for hospitalists were specifically related to their interactions with students, rather than with an overall more positive view of the rotation.

It has been suggested that forces promoting the use of hospitalists in adult medicine are even more persuasive in the pediatric population, as the difference in severity of illness between the inpatient and outpatient setting is greater, and the average pediatrician has less experience than the average internist in managing hospitalized patients.19 In a recent systematic review of the literature, Landrigan et al.20 reported that 6 of 7 studies demonstrated hospitalist systems had decreased hospital length of stay compared to systems in which a primary pediatrician served as the physician of record. This improved efficiency, if combined with the pressure to see more patients while trying to balance teaching and research demands, may have a negative impact on the quality of medical education.

Several factors may have contributed to the students' satisfaction with hospitalists. Studies have demonstrated that students rate clinical teachers more favorably with whom they have greater involvement.21 Hospitalists may be more likely to spend time on the inpatient wards given that is the primary site of their clinical activity. This increased presence may have contributed to more favorable evaluations for the hospitalist faculty, whereas the additional outpatient workload for nonhospitalist faculty may have reduced inpatient teaching opportunities, accounting for their lower teaching score. Included in the pediatrician category was the attribute of being a role model. In a study by Wright et al.,22 spending more than 25% of the time or 25 or more hours per week teaching and conducting rounds was independently associated with being considered an excellent role model. Again, the increased availability of the hospitalists on the inpatient wards may have led to more teaching opportunities, contributing to their higher score.

Our study had several limitations. First, it was a retrospective study conducted at a single institution with only 2 hospitalists. Although there were not statistical significant demographic differences between the 2 groups, this may simply reflect the small size of the sample in our study; therefore, the results may not be applicable to other academic institutions. Second, we retrospectively analyzed an evaluation form that had not been validated or specifically designed to compare 2 physician groups. Third, there were multiple statements in each category that students were asked to consider before scoring each attending on the parameters measured. Although hospitalists were rated higher in each category, there may have been individual characteristics within each category for which the nonhospitalist faculty performed better. Fourth, although hospitalists received higher average ratings than nonhospitalist faculty from third‐year medical students, it is important to emphasize this study measured students' attitudes and beliefs not specific educational outcomes. However, even though we cannot rule out the possibility that potentially confounding factors such as the personality of an attending physician influenced the results, prior studies have demonstrated that medical students make sophisticated judgments about teaching in the clinical setting.23, 24 It is unlikely that hospitalists at our institution were specifically selected to attend more months on a new inpatient service because they had a history of having more favorable teaching qualities because 1 of the 2 hospitalists had just finished residency training, and there were no significant demographic differences between the 2 groups. In a study examining trainee satisfaction in an internal medicine rotation 4 years after adoption of a hospitalist model, where nonhospitalist faculty attended based on their own interest and inpatient skill rather than as a requirement, Hauer et al.14 reported that trainees experienced more effective teaching and a more satisfying inpatient rotation when supervised by hospitalists. This suggests that hospitalists may possess or develop a specific inpatient knowledge base and teaching acumen over time that distinguishes them from nonhospitalists. There is evidence of accumulated experience leading to improved outcomes in the clinical setting for HIV infection,25 various surgical procedures,26 and hospitalist systems.27

In conclusion, this is the first study to evaluate the performance of hospitalists in the setting of a third‐year medical student pediatrics clerkship. Although third‐year medical students rate hospitalists at least as highly as nonhospitalist faculty, further studies are needed to reproduce this finding. In addition to the increased time spent on the wards with students and increased experience in caring for hospitalized patients, further studies should also examine the role that communication plays in clinical teaching. Also, the recent development of core competencies in hospital medicine28 may lead to the development of educational outcomes that can be objectively measured.

In 1996 Wachter and Goldman anticipated the emergence of hospitalists,1 physicians who are responsible for the care of hospitalized patients in place of their primary care physicians. The number of physicians who identify themselves as hospitalists has grown rapidly since 1996 and is currently estimated to be 10,00012,000, with the potential to reach as high as 30,000 in the next decade.2 This growth includes academic medical centers. In surveys of chairs of internal medicine and pediatric departments, 50% have hospitalists employed at their institutions.3, 4

Hospitalists in academic institutions are playing an increasingly prominent role in the medical education of both residents and medical students. The implications of adopting a hospitalist model on medical education has been discussed.57 Despite such concerns as fragmented continuity of care; decreased exposure to primary care physicians, subspecialists and physician‐scientists; reduced autonomy; and fewer educational opportunities to observe the natural histories of illnesses because of improved efficiency,57 the overall impact of hospitalists on medical and resident education has generally been favorable.818 Internal medicine residents have rated the teaching skills of hospitalists comparable to traditional academic physicians,8, 9 and believe the addition of hospitalists has contributed to an improved educational experience.10, 11, 14 In addition, a survey of third‐year medical students at a single academic teaching hospital concluded that hospitalists were able to provide at least as positive an educational experience during their inpatient medicine rotations as highly rated nonhospitalist teaching faculty.13

The role of hospitalists as educators in pediatrics has been studied much less. Pediatric resident satisfaction has improved in institutions that have used a hospitalist model.1618 In another study, hospitalists were rated by pediatric residents as more effective teachers than nonhospitalists.15 Because we are unaware of any study that has evaluated hospitalists in the education of medical students during their inpatient pediatric rotation, the purpose of our study was to compare hospitalist and nonhospitalist faculty on the educational experience of third‐year medical students during the inpatient portion of their pediatric clerkships at a single university children's hospital.

METHODS

RESULTS

All 67 of the students who completed a pediatrics clerkship at PSCH returned evaluation forms, which were the data for further analysis. Thirty‐five students rotated with the hospitalist faculty, and 32 students rotated with the nonhospitalist faculty. There were no significant demographic differences between the hospitalist and nonhospitalist faculty in age, sex, academic rank, specialty, and years since completing training (Table 2). All the hospitalist faculty fulfilled the definition of a hospitalist,2 whereas none of the physicians in the nonhospitalist group did.

The hospitalists were rated significantly higher than the nonhospitalist faculty in all 4 of the attending characteristics measured (Table 1): teaching effectiveness (3.87 vs. 2.91; P < .0001), effectiveness as a pediatrician (3.94 vs. 3.25; P < .001), student advocacy effectiveness (3.76 vs. 2.97; P < .0001), and overall evaluation (3.93 vs. 3.06; P < .001).

Analysis of specific aspects of the rotation showed the only feature that hospitalists were rated significantly higher on was quality of ward rounds (3.15 vs. 2.58, P < .006). There was no significant difference between the hospitalists and nonhospitalists on features that were not specifically part of the inpatient rotation, including various conferences, outpatient clinics, and newborn care (Table 1).

DISCUSSION

Our study demonstrates that pediatric hospitalists had a positive impact on the overall educational experience of third‐year medical students during the inpatient portions of their pediatrics clerkships. Hospitalists were rated more favorably than nonhospitalists as teachers, as pediatricians, as student advocates, and overall. Medical students also rated the value of ward rounds more favorably when hospitalists conducted them. In addition, higher percentages of nonhospitalists than hospitalists were rated as adequate or inadequate for the above items. When other aspects of the clerkship were analyzed, there were no statistically significant differences between the students who rotated with hospitalists and the students who rotated with nonhospitalists. This suggests that the higher scores for hospitalists were specifically related to their interactions with students, rather than with an overall more positive view of the rotation.

It has been suggested that forces promoting the use of hospitalists in adult medicine are even more persuasive in the pediatric population, as the difference in severity of illness between the inpatient and outpatient setting is greater, and the average pediatrician has less experience than the average internist in managing hospitalized patients.19 In a recent systematic review of the literature, Landrigan et al.20 reported that 6 of 7 studies demonstrated hospitalist systems had decreased hospital length of stay compared to systems in which a primary pediatrician served as the physician of record. This improved efficiency, if combined with the pressure to see more patients while trying to balance teaching and research demands, may have a negative impact on the quality of medical education.

Several factors may have contributed to the students' satisfaction with hospitalists. Studies have demonstrated that students rate clinical teachers more favorably with whom they have greater involvement.21 Hospitalists may be more likely to spend time on the inpatient wards given that is the primary site of their clinical activity. This increased presence may have contributed to more favorable evaluations for the hospitalist faculty, whereas the additional outpatient workload for nonhospitalist faculty may have reduced inpatient teaching opportunities, accounting for their lower teaching score. Included in the pediatrician category was the attribute of being a role model. In a study by Wright et al.,22 spending more than 25% of the time or 25 or more hours per week teaching and conducting rounds was independently associated with being considered an excellent role model. Again, the increased availability of the hospitalists on the inpatient wards may have led to more teaching opportunities, contributing to their higher score.

Our study had several limitations. First, it was a retrospective study conducted at a single institution with only 2 hospitalists. Although there were not statistical significant demographic differences between the 2 groups, this may simply reflect the small size of the sample in our study; therefore, the results may not be applicable to other academic institutions. Second, we retrospectively analyzed an evaluation form that had not been validated or specifically designed to compare 2 physician groups. Third, there were multiple statements in each category that students were asked to consider before scoring each attending on the parameters measured. Although hospitalists were rated higher in each category, there may have been individual characteristics within each category for which the nonhospitalist faculty performed better. Fourth, although hospitalists received higher average ratings than nonhospitalist faculty from third‐year medical students, it is important to emphasize this study measured students' attitudes and beliefs not specific educational outcomes. However, even though we cannot rule out the possibility that potentially confounding factors such as the personality of an attending physician influenced the results, prior studies have demonstrated that medical students make sophisticated judgments about teaching in the clinical setting.23, 24 It is unlikely that hospitalists at our institution were specifically selected to attend more months on a new inpatient service because they had a history of having more favorable teaching qualities because 1 of the 2 hospitalists had just finished residency training, and there were no significant demographic differences between the 2 groups. In a study examining trainee satisfaction in an internal medicine rotation 4 years after adoption of a hospitalist model, where nonhospitalist faculty attended based on their own interest and inpatient skill rather than as a requirement, Hauer et al.14 reported that trainees experienced more effective teaching and a more satisfying inpatient rotation when supervised by hospitalists. This suggests that hospitalists may possess or develop a specific inpatient knowledge base and teaching acumen over time that distinguishes them from nonhospitalists. There is evidence of accumulated experience leading to improved outcomes in the clinical setting for HIV infection,25 various surgical procedures,26 and hospitalist systems.27

In conclusion, this is the first study to evaluate the performance of hospitalists in the setting of a third‐year medical student pediatrics clerkship. Although third‐year medical students rate hospitalists at least as highly as nonhospitalist faculty, further studies are needed to reproduce this finding. In addition to the increased time spent on the wards with students and increased experience in caring for hospitalized patients, further studies should also examine the role that communication plays in clinical teaching. Also, the recent development of core competencies in hospital medicine28 may lead to the development of educational outcomes that can be objectively measured.

In this issue of the Journal of Hospital Medicine, Simon et al.1 provide the first report of pediatric hospitalist comanagement of patients undergoing spinal fusion surgery. In this retrospective cohort study, 14 of 115 patients were comanaged by a pediatric hospitalist. The primary outcomes of the study were length of stay and variability in length of stay. Prior to the initiation of hospitalist comanagement service, all patients were managed preoperatively by a spine surgery nurse and aided by medical subspecialists and other allied health professionals (nutritionists, respiratory therapists, physical therapists, social workers). After the intervention, patients with the most complex medical disease were assigned to comanagement by a pediatric hospitalist. When compared to a historical control of patients with similar medical complexity but not comanaged by hospitalists, the length of stay was reduced by 2.4 days (8.6 vs. 6.2 days). The variability in mean length of stay was also reduced.

This study follows on the heels of 3 important studies addressing the utility of hospitalists in the comanagement of surgical patients. The HOT (Hospitalist Orthopedic Team) trial was a randomized controlled trial assessing the effect of hospitalists on the management of patients undergoing elective hip and knee arthroplasty.2 There was no effect on length of stay or patient outcomes, though the comanagement model did decrease minor postoperative medical complications and improve physician and nurse satisfaction. Macpherson et al. conducted a retrospective trial where an internist joined a cardiothoracic surgery service at a tertiary‐care center.3 They found a decrease in overall mortality and resource utilization such as labs testing and consultations. There was significant reduction in the length of stay and number of x‐rays performed. The third study, by Jaffer et al.,4 showed that an outpatient, preoperative evaluation clinic staffed by hospitalists at a large tertiary‐care center provides a practical model for managing preoperative patients and may be associated with a low rate of postoperative pulmonary complications.

The study by Simon et al. in this issue of the journal has limitations. It is a retrospective cohort trial, and like all such study designs, the validity of the results is subject to confounding. Severity of patient medical disease, intraoperative complications, and advances in surgical technique are examples. While the authors did everything possible to minimize the effect of confounding, it remains a limitation of the study. The study also enrolled only 14 patients in the comanagement group, and this limited any stratification or subgroup analysis to offset known confounders. Patients assigned to the hospitalist comanagement service were by design more medically complex than other spinal fusion patients, and generalizing the results of this trial to all spinal fusion patients may not be possible.

From the study's limitations, however, comes great insight into the role of the hospitalist in surgical comanagement. It is clear from the aforementioned studies that there is a role for the hospitalist in comanagement of surgical patients. While the evidence is conflicting, there are scenarios in which comanagement improves efficiency and quality of care. Yet it is also possible that hospitalist comanagement is not ideal for all surgical patients. The HOT trial did not show benefits in length‐of‐stay reduction or patient mortality because the patients were homogenous in their complexity and pre‐ and postoperative care was protocol driven. Length of stay was limited by accessibility of rehabilitation facilities after discharge and not the efficiency of medical care in the hospital. The study in this issue of the Journal of Hospital Medicine selectively included patients with the highest complexity of medical disease, and there was a reduction in length of stay. Both trials suggest that the greatest potential benefit for augmenting efficiency and outcomes with hospitalist comanagement may be predicated on the complexity of the patients involved and the surgical system through which they will receive care.

The next step in assessing hospitalist comanagement should not be a hunt‐and‐peck exercise to stumble on the surgical procedures that show benefit from comanagement. Rather, the prudent next step is to follow the lead of Simon et al. and others3, 4 in trying to identify those surgical patients who represent the greatest medical complexity or have the most variability in their preoperative and postoperative medical care. These are the patients for whom the hospitalist can effect the greatest benefit and the services for which the hospitalist can best augment efficiency. High‐risk procedures, patients with multiple comorbitities or elevated preoperative risks, and surgical procedures without defined pre‐ and postoperative protocols would appear to be the ideal candidates for hospitalist comanagement.

As the discussion of hospital comanagement progresses, it is important to recognize comanagement as a paradigm shift. Surgical comanagement is not merely medical consultation. To be successful, the role of the hospitalist in comanaging surgical patients must be clearly defined as advancing postoperative care as much as it is in assessing preoperative risk. As a comanager, a hospitalist must actively manage preexisting and newly developed medical issues rather than just make recommendations for the surgical team.

The hospitalist must also be more than a discharge coordinator postoperatively; investing in hospitalists merely for discharge planning is a poor use of resources both from a financial and an opportunity‐cost perspective. The paradigm of comanagement is not foreign, however, and hospitalists are likely to prosper by learning from the experience of our nephrology and hepatology colleagues, who have successfully found collaborative roles in improving patient care on renal and liver transplant services. The success of these services is due to the precisely defined roles for the internist and the surgeon and because the complexity of the patient being managed warrants continuity of expert consultation.

There is great potential for the hospitalist in surgical comanagement. In less than a decade, the focus of hospital‐based medical care has shifted from staffing a shift to improving the quality of the system through which patients traverse the hospital. The lessons hospitalists have learned in quality improvement and in augmenting systems of care are perfectly suited for application to surgical services. Hospitalist comanagement is right not only because it may offer improvement in a surgical patient's medical care, but also for the augmentation of quality improvement in surgical services that have yet to reap the benefits that have defined the excellence of hospitalist medicine. The next step is to embark on the road of prudent prospective research: identifying the patients, and the procedures, that have the greatest opportunity for improvement by hospitalist comanagement. And at the end of that road will be a new home for the hospitalist, assuming the role of the quality‐advocate for all aspects of hospital care: pediatric, medical, and surgical patients.

In this issue of the Journal of Hospital Medicine, Simon et al.1 provide the first report of pediatric hospitalist comanagement of patients undergoing spinal fusion surgery. In this retrospective cohort study, 14 of 115 patients were comanaged by a pediatric hospitalist. The primary outcomes of the study were length of stay and variability in length of stay. Prior to the initiation of hospitalist comanagement service, all patients were managed preoperatively by a spine surgery nurse and aided by medical subspecialists and other allied health professionals (nutritionists, respiratory therapists, physical therapists, social workers). After the intervention, patients with the most complex medical disease were assigned to comanagement by a pediatric hospitalist. When compared to a historical control of patients with similar medical complexity but not comanaged by hospitalists, the length of stay was reduced by 2.4 days (8.6 vs. 6.2 days). The variability in mean length of stay was also reduced.

This study follows on the heels of 3 important studies addressing the utility of hospitalists in the comanagement of surgical patients. The HOT (Hospitalist Orthopedic Team) trial was a randomized controlled trial assessing the effect of hospitalists on the management of patients undergoing elective hip and knee arthroplasty.2 There was no effect on length of stay or patient outcomes, though the comanagement model did decrease minor postoperative medical complications and improve physician and nurse satisfaction. Macpherson et al. conducted a retrospective trial where an internist joined a cardiothoracic surgery service at a tertiary‐care center.3 They found a decrease in overall mortality and resource utilization such as labs testing and consultations. There was significant reduction in the length of stay and number of x‐rays performed. The third study, by Jaffer et al.,4 showed that an outpatient, preoperative evaluation clinic staffed by hospitalists at a large tertiary‐care center provides a practical model for managing preoperative patients and may be associated with a low rate of postoperative pulmonary complications.

The study by Simon et al. in this issue of the journal has limitations. It is a retrospective cohort trial, and like all such study designs, the validity of the results is subject to confounding. Severity of patient medical disease, intraoperative complications, and advances in surgical technique are examples. While the authors did everything possible to minimize the effect of confounding, it remains a limitation of the study. The study also enrolled only 14 patients in the comanagement group, and this limited any stratification or subgroup analysis to offset known confounders. Patients assigned to the hospitalist comanagement service were by design more medically complex than other spinal fusion patients, and generalizing the results of this trial to all spinal fusion patients may not be possible.

From the study's limitations, however, comes great insight into the role of the hospitalist in surgical comanagement. It is clear from the aforementioned studies that there is a role for the hospitalist in comanagement of surgical patients. While the evidence is conflicting, there are scenarios in which comanagement improves efficiency and quality of care. Yet it is also possible that hospitalist comanagement is not ideal for all surgical patients. The HOT trial did not show benefits in length‐of‐stay reduction or patient mortality because the patients were homogenous in their complexity and pre‐ and postoperative care was protocol driven. Length of stay was limited by accessibility of rehabilitation facilities after discharge and not the efficiency of medical care in the hospital. The study in this issue of the Journal of Hospital Medicine selectively included patients with the highest complexity of medical disease, and there was a reduction in length of stay. Both trials suggest that the greatest potential benefit for augmenting efficiency and outcomes with hospitalist comanagement may be predicated on the complexity of the patients involved and the surgical system through which they will receive care.

The next step in assessing hospitalist comanagement should not be a hunt‐and‐peck exercise to stumble on the surgical procedures that show benefit from comanagement. Rather, the prudent next step is to follow the lead of Simon et al. and others3, 4 in trying to identify those surgical patients who represent the greatest medical complexity or have the most variability in their preoperative and postoperative medical care. These are the patients for whom the hospitalist can effect the greatest benefit and the services for which the hospitalist can best augment efficiency. High‐risk procedures, patients with multiple comorbitities or elevated preoperative risks, and surgical procedures without defined pre‐ and postoperative protocols would appear to be the ideal candidates for hospitalist comanagement.

As the discussion of hospital comanagement progresses, it is important to recognize comanagement as a paradigm shift. Surgical comanagement is not merely medical consultation. To be successful, the role of the hospitalist in comanaging surgical patients must be clearly defined as advancing postoperative care as much as it is in assessing preoperative risk. As a comanager, a hospitalist must actively manage preexisting and newly developed medical issues rather than just make recommendations for the surgical team.

The hospitalist must also be more than a discharge coordinator postoperatively; investing in hospitalists merely for discharge planning is a poor use of resources both from a financial and an opportunity‐cost perspective. The paradigm of comanagement is not foreign, however, and hospitalists are likely to prosper by learning from the experience of our nephrology and hepatology colleagues, who have successfully found collaborative roles in improving patient care on renal and liver transplant services. The success of these services is due to the precisely defined roles for the internist and the surgeon and because the complexity of the patient being managed warrants continuity of expert consultation.

There is great potential for the hospitalist in surgical comanagement. In less than a decade, the focus of hospital‐based medical care has shifted from staffing a shift to improving the quality of the system through which patients traverse the hospital. The lessons hospitalists have learned in quality improvement and in augmenting systems of care are perfectly suited for application to surgical services. Hospitalist comanagement is right not only because it may offer improvement in a surgical patient's medical care, but also for the augmentation of quality improvement in surgical services that have yet to reap the benefits that have defined the excellence of hospitalist medicine. The next step is to embark on the road of prudent prospective research: identifying the patients, and the procedures, that have the greatest opportunity for improvement by hospitalist comanagement. And at the end of that road will be a new home for the hospitalist, assuming the role of the quality‐advocate for all aspects of hospital care: pediatric, medical, and surgical patients.

In this issue of the Journal of Hospital Medicine, Simon et al.1 provide the first report of pediatric hospitalist comanagement of patients undergoing spinal fusion surgery. In this retrospective cohort study, 14 of 115 patients were comanaged by a pediatric hospitalist. The primary outcomes of the study were length of stay and variability in length of stay. Prior to the initiation of hospitalist comanagement service, all patients were managed preoperatively by a spine surgery nurse and aided by medical subspecialists and other allied health professionals (nutritionists, respiratory therapists, physical therapists, social workers). After the intervention, patients with the most complex medical disease were assigned to comanagement by a pediatric hospitalist. When compared to a historical control of patients with similar medical complexity but not comanaged by hospitalists, the length of stay was reduced by 2.4 days (8.6 vs. 6.2 days). The variability in mean length of stay was also reduced.

This study follows on the heels of 3 important studies addressing the utility of hospitalists in the comanagement of surgical patients. The HOT (Hospitalist Orthopedic Team) trial was a randomized controlled trial assessing the effect of hospitalists on the management of patients undergoing elective hip and knee arthroplasty.2 There was no effect on length of stay or patient outcomes, though the comanagement model did decrease minor postoperative medical complications and improve physician and nurse satisfaction. Macpherson et al. conducted a retrospective trial where an internist joined a cardiothoracic surgery service at a tertiary‐care center.3 They found a decrease in overall mortality and resource utilization such as labs testing and consultations. There was significant reduction in the length of stay and number of x‐rays performed. The third study, by Jaffer et al.,4 showed that an outpatient, preoperative evaluation clinic staffed by hospitalists at a large tertiary‐care center provides a practical model for managing preoperative patients and may be associated with a low rate of postoperative pulmonary complications.

The study by Simon et al. in this issue of the journal has limitations. It is a retrospective cohort trial, and like all such study designs, the validity of the results is subject to confounding. Severity of patient medical disease, intraoperative complications, and advances in surgical technique are examples. While the authors did everything possible to minimize the effect of confounding, it remains a limitation of the study. The study also enrolled only 14 patients in the comanagement group, and this limited any stratification or subgroup analysis to offset known confounders. Patients assigned to the hospitalist comanagement service were by design more medically complex than other spinal fusion patients, and generalizing the results of this trial to all spinal fusion patients may not be possible.

From the study's limitations, however, comes great insight into the role of the hospitalist in surgical comanagement. It is clear from the aforementioned studies that there is a role for the hospitalist in comanagement of surgical patients. While the evidence is conflicting, there are scenarios in which comanagement improves efficiency and quality of care. Yet it is also possible that hospitalist comanagement is not ideal for all surgical patients. The HOT trial did not show benefits in length‐of‐stay reduction or patient mortality because the patients were homogenous in their complexity and pre‐ and postoperative care was protocol driven. Length of stay was limited by accessibility of rehabilitation facilities after discharge and not the efficiency of medical care in the hospital. The study in this issue of the Journal of Hospital Medicine selectively included patients with the highest complexity of medical disease, and there was a reduction in length of stay. Both trials suggest that the greatest potential benefit for augmenting efficiency and outcomes with hospitalist comanagement may be predicated on the complexity of the patients involved and the surgical system through which they will receive care.

The next step in assessing hospitalist comanagement should not be a hunt‐and‐peck exercise to stumble on the surgical procedures that show benefit from comanagement. Rather, the prudent next step is to follow the lead of Simon et al. and others3, 4 in trying to identify those surgical patients who represent the greatest medical complexity or have the most variability in their preoperative and postoperative medical care. These are the patients for whom the hospitalist can effect the greatest benefit and the services for which the hospitalist can best augment efficiency. High‐risk procedures, patients with multiple comorbitities or elevated preoperative risks, and surgical procedures without defined pre‐ and postoperative protocols would appear to be the ideal candidates for hospitalist comanagement.

As the discussion of hospital comanagement progresses, it is important to recognize comanagement as a paradigm shift. Surgical comanagement is not merely medical consultation. To be successful, the role of the hospitalist in comanaging surgical patients must be clearly defined as advancing postoperative care as much as it is in assessing preoperative risk. As a comanager, a hospitalist must actively manage preexisting and newly developed medical issues rather than just make recommendations for the surgical team.

The hospitalist must also be more than a discharge coordinator postoperatively; investing in hospitalists merely for discharge planning is a poor use of resources both from a financial and an opportunity‐cost perspective. The paradigm of comanagement is not foreign, however, and hospitalists are likely to prosper by learning from the experience of our nephrology and hepatology colleagues, who have successfully found collaborative roles in improving patient care on renal and liver transplant services. The success of these services is due to the precisely defined roles for the internist and the surgeon and because the complexity of the patient being managed warrants continuity of expert consultation.

There is great potential for the hospitalist in surgical comanagement. In less than a decade, the focus of hospital‐based medical care has shifted from staffing a shift to improving the quality of the system through which patients traverse the hospital. The lessons hospitalists have learned in quality improvement and in augmenting systems of care are perfectly suited for application to surgical services. Hospitalist comanagement is right not only because it may offer improvement in a surgical patient's medical care, but also for the augmentation of quality improvement in surgical services that have yet to reap the benefits that have defined the excellence of hospitalist medicine. The next step is to embark on the road of prudent prospective research: identifying the patients, and the procedures, that have the greatest opportunity for improvement by hospitalist comanagement. And at the end of that road will be a new home for the hospitalist, assuming the role of the quality‐advocate for all aspects of hospital care: pediatric, medical, and surgical patients.

One year of the Journal of Hospital Medicine is done, and we now embark on our second with this first issue of volume 2. Before moving on, I heartily thank all the authors who contributed their manuscripts to the Journal of Hospital Medicine (JHM), bravely investing in this new academic periodical. A remarkable 284 manuscripts have been submitted since we first opened the JHM Web site, 197 of them during 2006. This clearly reflects the robust demand by hospitalists and their colleagues for original research and relevant clinical reviews about our evolving specialty of hospital medicine. I probably should not be surprised that this demand exists among the 15,000‐plus hospitalists in America and the 6000‐plus members of the Society of Hospital Medicine. Regardless, I am ineffably humbled by the enthusiasm and energy of all the contributors.

Understandably, this volume of submissions, exceeding our projections by nearly 50%, required yeoman's work by our associate editors and reviewers. On page 55 we list the 203 reviewers who donated their time and acumen to assure the quality of our publication. Many reviewed more than 4 articles during the year. Our associate editors deserve particular appreciation and gratitude for their willingness to donate extraordinary amounts of time and effort to ensure the success of JHMVincent Chang from Boston Children's Hospital, Scott Flanders from the University of Michigan, Karen Hauer from the University of California, San Francisco, Jean Kutner from the University of Colorado, James Pile from Cleveland MetroHealth, and Kaveh Shojania from the University of Ottawa. Additionally, the energetic assistant editors have supported them and me with frequent reviews, article submissions, and creative ideas for improving the journal. Finally, our auspicious editorial board has proffered sage guidance, and many of its members have also submitted manuscripts and participated in reviewing articles.

Moving forward we expect continued growth, as both the submitted articles and demand for the journal are being recognized. At 7:29 a.m. on November 30, 2006, Vickie Thaw (Vice President and Publisher, John Wiley & Sons, Inc.) called me to report that the National Library of Medicine validated all our efforts. The Journal of Hospital Medicine had been selected for indexing and inclusion in the National Library of Medicine's MEDLINE (Medical Literature Analysis and Retrieval System Online). The primary component of PubMed, MEDLINE is a bibliographic database containing approximately 13 million references to journal articles on medicine, nursing, dentistry, veterinary medicine, health care systems, and preclinical sciences dating to the mid‐1960s. With this approval, hospital medicine has achieved another milestone in its evolution into a new specialty.

We now hope to respond to the robust interest in clinical materials as well as to continue publication of original research. To achieve our aim of increasing the amount of clinically relevant content for practicing hospitalists, authors are encouraged to submit to JHM case reports, clinical updates, and clinical images that convey novel or underappreciated teaching points. Teaching points may be purely clinical and may focus on clinical pearls or unusual presentations of well‐known diseases, although submission of straightforward presentations of rare diseases is discouraged. Alternatively, manuscripts may involve succinct case‐based descriptions of innovations, quality improvementrelated issues, or medical errors. Submitted case reports should be less than 800 words and should contain a maximum of 5 references and no more than 1 table or figure. Case reports should not include an abstract. Submission of the case report and review type should be avoided. Instead, we seek formal clinical updates of no more than 2000 words that present important aspects of a case along with new research findings and citations from the literature that change what has historically been the standard of delivery of care. Finally, we continue to seek cases most appropriate for the Hospital Images Dx section, edited by Paul Aronowitz. They should be submitted with that designation and have fewer than 150 words. These 3 categories are identified on our Manuscript Central website (http://mc.manuscriptcentral.com/jhm).

Again, thanks to all of you for making the launch of the Journal of Hospital Medicine an unqualified success. We look forward to your continued participation as we grow as the premier journal for the specialty of hospital medicine.

P.S. Sadly, one of our superstar associate editors, Kaveh Shojania, is stepping aside, and we sincerely express thanks for his terrific contributions. We welcome suggestions for an alternative to fulfill his responsibilities.

One year of the Journal of Hospital Medicine is done, and we now embark on our second with this first issue of volume 2. Before moving on, I heartily thank all the authors who contributed their manuscripts to the Journal of Hospital Medicine (JHM), bravely investing in this new academic periodical. A remarkable 284 manuscripts have been submitted since we first opened the JHM Web site, 197 of them during 2006. This clearly reflects the robust demand by hospitalists and their colleagues for original research and relevant clinical reviews about our evolving specialty of hospital medicine. I probably should not be surprised that this demand exists among the 15,000‐plus hospitalists in America and the 6000‐plus members of the Society of Hospital Medicine. Regardless, I am ineffably humbled by the enthusiasm and energy of all the contributors.

Understandably, this volume of submissions, exceeding our projections by nearly 50%, required yeoman's work by our associate editors and reviewers. On page 55 we list the 203 reviewers who donated their time and acumen to assure the quality of our publication. Many reviewed more than 4 articles during the year. Our associate editors deserve particular appreciation and gratitude for their willingness to donate extraordinary amounts of time and effort to ensure the success of JHMVincent Chang from Boston Children's Hospital, Scott Flanders from the University of Michigan, Karen Hauer from the University of California, San Francisco, Jean Kutner from the University of Colorado, James Pile from Cleveland MetroHealth, and Kaveh Shojania from the University of Ottawa. Additionally, the energetic assistant editors have supported them and me with frequent reviews, article submissions, and creative ideas for improving the journal. Finally, our auspicious editorial board has proffered sage guidance, and many of its members have also submitted manuscripts and participated in reviewing articles.

Moving forward we expect continued growth, as both the submitted articles and demand for the journal are being recognized. At 7:29 a.m. on November 30, 2006, Vickie Thaw (Vice President and Publisher, John Wiley & Sons, Inc.) called me to report that the National Library of Medicine validated all our efforts. The Journal of Hospital Medicine had been selected for indexing and inclusion in the National Library of Medicine's MEDLINE (Medical Literature Analysis and Retrieval System Online). The primary component of PubMed, MEDLINE is a bibliographic database containing approximately 13 million references to journal articles on medicine, nursing, dentistry, veterinary medicine, health care systems, and preclinical sciences dating to the mid‐1960s. With this approval, hospital medicine has achieved another milestone in its evolution into a new specialty.

We now hope to respond to the robust interest in clinical materials as well as to continue publication of original research. To achieve our aim of increasing the amount of clinically relevant content for practicing hospitalists, authors are encouraged to submit to JHM case reports, clinical updates, and clinical images that convey novel or underappreciated teaching points. Teaching points may be purely clinical and may focus on clinical pearls or unusual presentations of well‐known diseases, although submission of straightforward presentations of rare diseases is discouraged. Alternatively, manuscripts may involve succinct case‐based descriptions of innovations, quality improvementrelated issues, or medical errors. Submitted case reports should be less than 800 words and should contain a maximum of 5 references and no more than 1 table or figure. Case reports should not include an abstract. Submission of the case report and review type should be avoided. Instead, we seek formal clinical updates of no more than 2000 words that present important aspects of a case along with new research findings and citations from the literature that change what has historically been the standard of delivery of care. Finally, we continue to seek cases most appropriate for the Hospital Images Dx section, edited by Paul Aronowitz. They should be submitted with that designation and have fewer than 150 words. These 3 categories are identified on our Manuscript Central website (http://mc.manuscriptcentral.com/jhm).

Again, thanks to all of you for making the launch of the Journal of Hospital Medicine an unqualified success. We look forward to your continued participation as we grow as the premier journal for the specialty of hospital medicine.

P.S. Sadly, one of our superstar associate editors, Kaveh Shojania, is stepping aside, and we sincerely express thanks for his terrific contributions. We welcome suggestions for an alternative to fulfill his responsibilities.

One year of the Journal of Hospital Medicine is done, and we now embark on our second with this first issue of volume 2. Before moving on, I heartily thank all the authors who contributed their manuscripts to the Journal of Hospital Medicine (JHM), bravely investing in this new academic periodical. A remarkable 284 manuscripts have been submitted since we first opened the JHM Web site, 197 of them during 2006. This clearly reflects the robust demand by hospitalists and their colleagues for original research and relevant clinical reviews about our evolving specialty of hospital medicine. I probably should not be surprised that this demand exists among the 15,000‐plus hospitalists in America and the 6000‐plus members of the Society of Hospital Medicine. Regardless, I am ineffably humbled by the enthusiasm and energy of all the contributors.

Understandably, this volume of submissions, exceeding our projections by nearly 50%, required yeoman's work by our associate editors and reviewers. On page 55 we list the 203 reviewers who donated their time and acumen to assure the quality of our publication. Many reviewed more than 4 articles during the year. Our associate editors deserve particular appreciation and gratitude for their willingness to donate extraordinary amounts of time and effort to ensure the success of JHMVincent Chang from Boston Children's Hospital, Scott Flanders from the University of Michigan, Karen Hauer from the University of California, San Francisco, Jean Kutner from the University of Colorado, James Pile from Cleveland MetroHealth, and Kaveh Shojania from the University of Ottawa. Additionally, the energetic assistant editors have supported them and me with frequent reviews, article submissions, and creative ideas for improving the journal. Finally, our auspicious editorial board has proffered sage guidance, and many of its members have also submitted manuscripts and participated in reviewing articles.

Moving forward we expect continued growth, as both the submitted articles and demand for the journal are being recognized. At 7:29 a.m. on November 30, 2006, Vickie Thaw (Vice President and Publisher, John Wiley & Sons, Inc.) called me to report that the National Library of Medicine validated all our efforts. The Journal of Hospital Medicine had been selected for indexing and inclusion in the National Library of Medicine's MEDLINE (Medical Literature Analysis and Retrieval System Online). The primary component of PubMed, MEDLINE is a bibliographic database containing approximately 13 million references to journal articles on medicine, nursing, dentistry, veterinary medicine, health care systems, and preclinical sciences dating to the mid‐1960s. With this approval, hospital medicine has achieved another milestone in its evolution into a new specialty.

We now hope to respond to the robust interest in clinical materials as well as to continue publication of original research. To achieve our aim of increasing the amount of clinically relevant content for practicing hospitalists, authors are encouraged to submit to JHM case reports, clinical updates, and clinical images that convey novel or underappreciated teaching points. Teaching points may be purely clinical and may focus on clinical pearls or unusual presentations of well‐known diseases, although submission of straightforward presentations of rare diseases is discouraged. Alternatively, manuscripts may involve succinct case‐based descriptions of innovations, quality improvementrelated issues, or medical errors. Submitted case reports should be less than 800 words and should contain a maximum of 5 references and no more than 1 table or figure. Case reports should not include an abstract. Submission of the case report and review type should be avoided. Instead, we seek formal clinical updates of no more than 2000 words that present important aspects of a case along with new research findings and citations from the literature that change what has historically been the standard of delivery of care. Finally, we continue to seek cases most appropriate for the Hospital Images Dx section, edited by Paul Aronowitz. They should be submitted with that designation and have fewer than 150 words. These 3 categories are identified on our Manuscript Central website (http://mc.manuscriptcentral.com/jhm).

Again, thanks to all of you for making the launch of the Journal of Hospital Medicine an unqualified success. We look forward to your continued participation as we grow as the premier journal for the specialty of hospital medicine.

P.S. Sadly, one of our superstar associate editors, Kaveh Shojania, is stepping aside, and we sincerely express thanks for his terrific contributions. We welcome suggestions for an alternative to fulfill his responsibilities.

The complete blood count (CBC) bundles the automated hemogram, an automated differential count of 5 types of cells, and a reflex manual differential count (when required by protocol) and is one of the most frequently ordered laboratory tests on admission to the hospital. In practice, it is a routine ingredient of all hospital admission ordersphysicians order a hemogram either alone or as part of a complete blood count for 98% of our medical/surgical admissions, and the same is true at most institutions.1 We know that the white blood cell count and hematocrit from the automated hemogram predict disease severity and mortality risk.25 For example, elevated WBC counts predict a worse prognosis in patients with cancer or coronary artery disease,6, 7 and anemia predicts increased risk of death of patients with heart failure.8, 9 Further, these two tests provide direct management guidance in common circumstances, for example, bleeding and infection.

The CBC describes the number and morphology of more than 40 cell types, from acanthocytosis to vacuolated white blood cells. Disagreement exists about the clinical significance of many of these observations.1013 And only a few components of the manual differential, for example, nucleated red blood cells (NRBCs) and lymphocytosis, have been quantitatively evaluated to determine their prognostic significance.1417 But these two observations have not been examined to determine their independent contributions to predictions of mortality when taken in conjunction with their accompanying CBC observations. Which of the numerous cell types and cell counts in the commonly ordered CBC, indicate that a patient is at high risk of death? In this article we report an inpatient study that used univariate and multivariate analyses of admission CBCs to predict 30‐day mortality in order to answer that question.

METHODS

RESULTS

During the 10 years from January 1993 through December 2002, physicians admitted 46,522 unique eligible patients to Wishard Memorial Hospital. Each patient averaged 2 admissions during the study period, for a total of 94,582 admissions. The overall 30‐day mortality of these admissions was 3.4%. Automated hemograms (white blood cell count, hemoglobin, red cell count, and red blood cell indices) were performed on blood samples from 45,709 of these patients (98%) within one day of admission. Seventy‐seven percent (35,692) had a complete blood count that included an automated differential count plus a reflex manual count and smear when required by the CAP protocol, as well as an automated hemogram. The patients with an admission CBC with differential count had a 30‐day mortality rate of 4%, slightly higher than that of patients who had only a hemogram. The patients' mean Charlson score for the CBC with differential count was 0.83, which was lower than the national average, which is closer to 1.22 Table 1 shows the demographics of this study population.

DISCUSSION

Some investigators have incorporated selected CBC measures, for example, white blood cell count and hemoglobin/hematocrit, into multivariable models that predict mortality or rehospitalizations.6, 7, 9, 23 However, CBC reports can include a spectrum of more than 40 distinct counts and morphologic findings. Our study was the first to take into account all the different variables in the complete blood count and differential to determine elements that independently predict a high risk of mortality.

In addition to age and sex, our multivariable analysis of the 45 CBC variables found 13 independent predictors of mortality. Five were observations about white blood cells: absolute leukocytosis, high band form cell count, the presence of metamyelocytes, the presence of toxic granules, and absolute lymphocytosis. Eight were observations about red blood cells: high hematocrit, low hematocrit, high MCV and the presence of macrocytes, high red cell distribution width, the presence of NRBCs, the presence of burr cells, and the presence of sickle cells. Because controlling for severity of illness by Charlson comorbidity scores did not significantly change the model, the CBC abnormalities among the predictors of mortality did not simply reflect how sick the patients were. Including the 10 most common admission diagnoses did not significantly attenuate our reported odds ratios, suggesting the CBC predictors did not merely reflect the primary reason for admission. Interestingly, however, admission for chest pain did correlate with a greatly reduced risk of 30‐day mortality, which may reflect the low threshold that physicians have for admitting patients with this complaint. Admission for acute but ill‐defined cerebrovascular disease independently predicted a 2‐fold increased risk of 30‐day mortality.

What is the message to physicians from this analysis? Physicians commonly order CBCs and may rely on quick heuristics to sift through the myriad findings in CBC reports. Our analysis focuses physician attention on high‐impact findings in the CBC. We assume that physicians already consider low hematocrit, high hematocrit (a sign of fluid loss and/or chronic hypoxia), high WBC count, high band cell count, and the presence of metamyeloctes (left shift) as important prognostic indicators. These abnormal findings are routinely mentioned at morning report and in a physician's notes.

Physicians, however, may not appreciate the importance of other CBC findings that our analysis found are predictive of mortality. Macrocytosis and a high RDW count (indicating an abnormally wide distribution of red blood cell size) have not previously been reported as predictors of mortality. And although other studies have suggested that bands are not predictors of mortality,11 our study found they were an important prognostic indicator, with an OR =1.59, approaching those of leukocytosis and anemia.

The most impressive predictors of mortality were burr cells, NRBCs, and absolute lymphocytosis. The multivariate ORs of these 3, ranging from 2.8 to 3.2, were the highest of any CBC finding. In univariate analysis, the first 2 were associated with mortality rates 8 to 10 times higher than that of the average admitted patient. There are anecdotal reports in the literature of burr cells being associated with ominous prognoses2426 and more robust statistical analyses showing NRBCs to be associated with increased mortality.14 Lymphocytosis has also been reported as a mortality risk in patients with trauma and emergency medical conditions.15, 16 Our analysis has shown that, indeed, all 3 of these findings are strong independent predictors of mortality.

The presence of sickle cells was also a strong predictor, but of decreased mortality. Patients with sickle cells in their smear had a risk of death one third that of patients without sickle cells. This does not indicate a protective effect. Rather, patients with sickle‐cell disease typically are young and admitted for pain control and other non‐life‐threatening conditions. The presence of NRBCs in patients with sickle‐cell disease appears to be intrinsic to the disease itself and did not have the same implications for mortality as it did for other patients in our study.

The overall logistic model including age, sex, and admission CBC variables had a respectable c statistic for predicting 30‐day mortality of 0.80. This compares well with findings in other multivariable models. For example, the APACHE II score used to predict the mortality of hospitalized critical care patients has a c statistic that ranges from 0.78 to 0.86.3, 27, 28 The APACHE score uses the worst value from the first 2 days after admission for some of its predictors so it cannot provide as early a warning as the admission CBC, and it requires collection of significantly more data. The inclusion of more CBC findings in the APACHE model might increase its predictive accuracy.

Our multivariate analysis was based on a very large number of patient samples using data collected through routine clinical care. However, our study has a number of limitations. The analysis was done at only a single institution, and the exact logistic regression model may not apply to other institutions that have different case mixes and laboratory procedures. Our institution's reported 30‐day mortality rate of 3.4% was lower than the 4.6%11.9% reported in studies of patients admitted to general ward services,2931 but this may be accounted for by the lower‐than‐average Charlson comorbidity scores in our study population. Our risk adjustment by Charlson comorbidity scores may not be as precise as a risk adjustment tailored for our particular institution.32 Our 30‐day mortality rate was calculated using state death tapes, which means we would have missed patients who died outside the state, although we believe this rarely happens. We developed predictive equations on the basis of 30‐day mortality, so we cannot comment on whether the CBC elements predict mortality beyond 30 days. We analyzed most variables as either high or low or as present or absent. Increasing degrees of abnormality may further increase the predictive power of some variables. Finally, the CBC is only one of many tests and clinical findings; it may be that some of these other variables would displace some CBC variables and/or improve the overall predictive power at the time the admission laboratory tests were performed. In this initial study, we have described the prognostic implication of the CBC across a wide range of diagnoses. Future work will focus on the predictive power of commonly gathered variables in more specific conditions (eg, low white blood cell count in sepsis).

Physicians generally have an intuitive ability to identify patients who are seriously ill and at high risk of dying33 and adjust their diagnostic and therapeutic efforts accordingly. Our analysis highlights the value that certain observations in the CBC, notably burr cells, NRBCs, and absolute lymphocytosis, add to physicians' assessments of mortality risk. Even after adjustment for age, sex, comorbidities, common admission diagnoses, and other variables in the CBC, the presence of these findings predicted a 3‐fold increase in 30‐day mortality. Identifying the red flags within this ubiquitously performed test can make the difference in premature discharge or inappropriate triage of patients. Busy physicians can choose from a wide selection of ever‐improving diagnostic tests, yet the workhorse CBC can serve as a simple and early identifier of patients with a poor prognosis.

The complete blood count (CBC) bundles the automated hemogram, an automated differential count of 5 types of cells, and a reflex manual differential count (when required by protocol) and is one of the most frequently ordered laboratory tests on admission to the hospital. In practice, it is a routine ingredient of all hospital admission ordersphysicians order a hemogram either alone or as part of a complete blood count for 98% of our medical/surgical admissions, and the same is true at most institutions.1 We know that the white blood cell count and hematocrit from the automated hemogram predict disease severity and mortality risk.25 For example, elevated WBC counts predict a worse prognosis in patients with cancer or coronary artery disease,6, 7 and anemia predicts increased risk of death of patients with heart failure.8, 9 Further, these two tests provide direct management guidance in common circumstances, for example, bleeding and infection.

The CBC describes the number and morphology of more than 40 cell types, from acanthocytosis to vacuolated white blood cells. Disagreement exists about the clinical significance of many of these observations.1013 And only a few components of the manual differential, for example, nucleated red blood cells (NRBCs) and lymphocytosis, have been quantitatively evaluated to determine their prognostic significance.1417 But these two observations have not been examined to determine their independent contributions to predictions of mortality when taken in conjunction with their accompanying CBC observations. Which of the numerous cell types and cell counts in the commonly ordered CBC, indicate that a patient is at high risk of death? In this article we report an inpatient study that used univariate and multivariate analyses of admission CBCs to predict 30‐day mortality in order to answer that question.

METHODS

RESULTS

During the 10 years from January 1993 through December 2002, physicians admitted 46,522 unique eligible patients to Wishard Memorial Hospital. Each patient averaged 2 admissions during the study period, for a total of 94,582 admissions. The overall 30‐day mortality of these admissions was 3.4%. Automated hemograms (white blood cell count, hemoglobin, red cell count, and red blood cell indices) were performed on blood samples from 45,709 of these patients (98%) within one day of admission. Seventy‐seven percent (35,692) had a complete blood count that included an automated differential count plus a reflex manual count and smear when required by the CAP protocol, as well as an automated hemogram. The patients with an admission CBC with differential count had a 30‐day mortality rate of 4%, slightly higher than that of patients who had only a hemogram. The patients' mean Charlson score for the CBC with differential count was 0.83, which was lower than the national average, which is closer to 1.22 Table 1 shows the demographics of this study population.

DISCUSSION

Some investigators have incorporated selected CBC measures, for example, white blood cell count and hemoglobin/hematocrit, into multivariable models that predict mortality or rehospitalizations.6, 7, 9, 23 However, CBC reports can include a spectrum of more than 40 distinct counts and morphologic findings. Our study was the first to take into account all the different variables in the complete blood count and differential to determine elements that independently predict a high risk of mortality.

In addition to age and sex, our multivariable analysis of the 45 CBC variables found 13 independent predictors of mortality. Five were observations about white blood cells: absolute leukocytosis, high band form cell count, the presence of metamyelocytes, the presence of toxic granules, and absolute lymphocytosis. Eight were observations about red blood cells: high hematocrit, low hematocrit, high MCV and the presence of macrocytes, high red cell distribution width, the presence of NRBCs, the presence of burr cells, and the presence of sickle cells. Because controlling for severity of illness by Charlson comorbidity scores did not significantly change the model, the CBC abnormalities among the predictors of mortality did not simply reflect how sick the patients were. Including the 10 most common admission diagnoses did not significantly attenuate our reported odds ratios, suggesting the CBC predictors did not merely reflect the primary reason for admission. Interestingly, however, admission for chest pain did correlate with a greatly reduced risk of 30‐day mortality, which may reflect the low threshold that physicians have for admitting patients with this complaint. Admission for acute but ill‐defined cerebrovascular disease independently predicted a 2‐fold increased risk of 30‐day mortality.

What is the message to physicians from this analysis? Physicians commonly order CBCs and may rely on quick heuristics to sift through the myriad findings in CBC reports. Our analysis focuses physician attention on high‐impact findings in the CBC. We assume that physicians already consider low hematocrit, high hematocrit (a sign of fluid loss and/or chronic hypoxia), high WBC count, high band cell count, and the presence of metamyeloctes (left shift) as important prognostic indicators. These abnormal findings are routinely mentioned at morning report and in a physician's notes.

Physicians, however, may not appreciate the importance of other CBC findings that our analysis found are predictive of mortality. Macrocytosis and a high RDW count (indicating an abnormally wide distribution of red blood cell size) have not previously been reported as predictors of mortality. And although other studies have suggested that bands are not predictors of mortality,11 our study found they were an important prognostic indicator, with an OR =1.59, approaching those of leukocytosis and anemia.

The most impressive predictors of mortality were burr cells, NRBCs, and absolute lymphocytosis. The multivariate ORs of these 3, ranging from 2.8 to 3.2, were the highest of any CBC finding. In univariate analysis, the first 2 were associated with mortality rates 8 to 10 times higher than that of the average admitted patient. There are anecdotal reports in the literature of burr cells being associated with ominous prognoses2426 and more robust statistical analyses showing NRBCs to be associated with increased mortality.14 Lymphocytosis has also been reported as a mortality risk in patients with trauma and emergency medical conditions.15, 16 Our analysis has shown that, indeed, all 3 of these findings are strong independent predictors of mortality.

The presence of sickle cells was also a strong predictor, but of decreased mortality. Patients with sickle cells in their smear had a risk of death one third that of patients without sickle cells. This does not indicate a protective effect. Rather, patients with sickle‐cell disease typically are young and admitted for pain control and other non‐life‐threatening conditions. The presence of NRBCs in patients with sickle‐cell disease appears to be intrinsic to the disease itself and did not have the same implications for mortality as it did for other patients in our study.

The overall logistic model including age, sex, and admission CBC variables had a respectable c statistic for predicting 30‐day mortality of 0.80. This compares well with findings in other multivariable models. For example, the APACHE II score used to predict the mortality of hospitalized critical care patients has a c statistic that ranges from 0.78 to 0.86.3, 27, 28 The APACHE score uses the worst value from the first 2 days after admission for some of its predictors so it cannot provide as early a warning as the admission CBC, and it requires collection of significantly more data. The inclusion of more CBC findings in the APACHE model might increase its predictive accuracy.

Our multivariate analysis was based on a very large number of patient samples using data collected through routine clinical care. However, our study has a number of limitations. The analysis was done at only a single institution, and the exact logistic regression model may not apply to other institutions that have different case mixes and laboratory procedures. Our institution's reported 30‐day mortality rate of 3.4% was lower than the 4.6%11.9% reported in studies of patients admitted to general ward services,2931 but this may be accounted for by the lower‐than‐average Charlson comorbidity scores in our study population. Our risk adjustment by Charlson comorbidity scores may not be as precise as a risk adjustment tailored for our particular institution.32 Our 30‐day mortality rate was calculated using state death tapes, which means we would have missed patients who died outside the state, although we believe this rarely happens. We developed predictive equations on the basis of 30‐day mortality, so we cannot comment on whether the CBC elements predict mortality beyond 30 days. We analyzed most variables as either high or low or as present or absent. Increasing degrees of abnormality may further increase the predictive power of some variables. Finally, the CBC is only one of many tests and clinical findings; it may be that some of these other variables would displace some CBC variables and/or improve the overall predictive power at the time the admission laboratory tests were performed. In this initial study, we have described the prognostic implication of the CBC across a wide range of diagnoses. Future work will focus on the predictive power of commonly gathered variables in more specific conditions (eg, low white blood cell count in sepsis).

Physicians generally have an intuitive ability to identify patients who are seriously ill and at high risk of dying33 and adjust their diagnostic and therapeutic efforts accordingly. Our analysis highlights the value that certain observations in the CBC, notably burr cells, NRBCs, and absolute lymphocytosis, add to physicians' assessments of mortality risk. Even after adjustment for age, sex, comorbidities, common admission diagnoses, and other variables in the CBC, the presence of these findings predicted a 3‐fold increase in 30‐day mortality. Identifying the red flags within this ubiquitously performed test can make the difference in premature discharge or inappropriate triage of patients. Busy physicians can choose from a wide selection of ever‐improving diagnostic tests, yet the workhorse CBC can serve as a simple and early identifier of patients with a poor prognosis.

The complete blood count (CBC) bundles the automated hemogram, an automated differential count of 5 types of cells, and a reflex manual differential count (when required by protocol) and is one of the most frequently ordered laboratory tests on admission to the hospital. In practice, it is a routine ingredient of all hospital admission ordersphysicians order a hemogram either alone or as part of a complete blood count for 98% of our medical/surgical admissions, and the same is true at most institutions.1 We know that the white blood cell count and hematocrit from the automated hemogram predict disease severity and mortality risk.25 For example, elevated WBC counts predict a worse prognosis in patients with cancer or coronary artery disease,6, 7 and anemia predicts increased risk of death of patients with heart failure.8, 9 Further, these two tests provide direct management guidance in common circumstances, for example, bleeding and infection.

The CBC describes the number and morphology of more than 40 cell types, from acanthocytosis to vacuolated white blood cells. Disagreement exists about the clinical significance of many of these observations.1013 And only a few components of the manual differential, for example, nucleated red blood cells (NRBCs) and lymphocytosis, have been quantitatively evaluated to determine their prognostic significance.1417 But these two observations have not been examined to determine their independent contributions to predictions of mortality when taken in conjunction with their accompanying CBC observations. Which of the numerous cell types and cell counts in the commonly ordered CBC, indicate that a patient is at high risk of death? In this article we report an inpatient study that used univariate and multivariate analyses of admission CBCs to predict 30‐day mortality in order to answer that question.

METHODS

RESULTS

During the 10 years from January 1993 through December 2002, physicians admitted 46,522 unique eligible patients to Wishard Memorial Hospital. Each patient averaged 2 admissions during the study period, for a total of 94,582 admissions. The overall 30‐day mortality of these admissions was 3.4%. Automated hemograms (white blood cell count, hemoglobin, red cell count, and red blood cell indices) were performed on blood samples from 45,709 of these patients (98%) within one day of admission. Seventy‐seven percent (35,692) had a complete blood count that included an automated differential count plus a reflex manual count and smear when required by the CAP protocol, as well as an automated hemogram. The patients with an admission CBC with differential count had a 30‐day mortality rate of 4%, slightly higher than that of patients who had only a hemogram. The patients' mean Charlson score for the CBC with differential count was 0.83, which was lower than the national average, which is closer to 1.22 Table 1 shows the demographics of this study population.

DISCUSSION

Some investigators have incorporated selected CBC measures, for example, white blood cell count and hemoglobin/hematocrit, into multivariable models that predict mortality or rehospitalizations.6, 7, 9, 23 However, CBC reports can include a spectrum of more than 40 distinct counts and morphologic findings. Our study was the first to take into account all the different variables in the complete blood count and differential to determine elements that independently predict a high risk of mortality.

In addition to age and sex, our multivariable analysis of the 45 CBC variables found 13 independent predictors of mortality. Five were observations about white blood cells: absolute leukocytosis, high band form cell count, the presence of metamyelocytes, the presence of toxic granules, and absolute lymphocytosis. Eight were observations about red blood cells: high hematocrit, low hematocrit, high MCV and the presence of macrocytes, high red cell distribution width, the presence of NRBCs, the presence of burr cells, and the presence of sickle cells. Because controlling for severity of illness by Charlson comorbidity scores did not significantly change the model, the CBC abnormalities among the predictors of mortality did not simply reflect how sick the patients were. Including the 10 most common admission diagnoses did not significantly attenuate our reported odds ratios, suggesting the CBC predictors did not merely reflect the primary reason for admission. Interestingly, however, admission for chest pain did correlate with a greatly reduced risk of 30‐day mortality, which may reflect the low threshold that physicians have for admitting patients with this complaint. Admission for acute but ill‐defined cerebrovascular disease independently predicted a 2‐fold increased risk of 30‐day mortality.

What is the message to physicians from this analysis? Physicians commonly order CBCs and may rely on quick heuristics to sift through the myriad findings in CBC reports. Our analysis focuses physician attention on high‐impact findings in the CBC. We assume that physicians already consider low hematocrit, high hematocrit (a sign of fluid loss and/or chronic hypoxia), high WBC count, high band cell count, and the presence of metamyeloctes (left shift) as important prognostic indicators. These abnormal findings are routinely mentioned at morning report and in a physician's notes.

Physicians, however, may not appreciate the importance of other CBC findings that our analysis found are predictive of mortality. Macrocytosis and a high RDW count (indicating an abnormally wide distribution of red blood cell size) have not previously been reported as predictors of mortality. And although other studies have suggested that bands are not predictors of mortality,11 our study found they were an important prognostic indicator, with an OR =1.59, approaching those of leukocytosis and anemia.

The most impressive predictors of mortality were burr cells, NRBCs, and absolute lymphocytosis. The multivariate ORs of these 3, ranging from 2.8 to 3.2, were the highest of any CBC finding. In univariate analysis, the first 2 were associated with mortality rates 8 to 10 times higher than that of the average admitted patient. There are anecdotal reports in the literature of burr cells being associated with ominous prognoses2426 and more robust statistical analyses showing NRBCs to be associated with increased mortality.14 Lymphocytosis has also been reported as a mortality risk in patients with trauma and emergency medical conditions.15, 16 Our analysis has shown that, indeed, all 3 of these findings are strong independent predictors of mortality.

The presence of sickle cells was also a strong predictor, but of decreased mortality. Patients with sickle cells in their smear had a risk of death one third that of patients without sickle cells. This does not indicate a protective effect. Rather, patients with sickle‐cell disease typically are young and admitted for pain control and other non‐life‐threatening conditions. The presence of NRBCs in patients with sickle‐cell disease appears to be intrinsic to the disease itself and did not have the same implications for mortality as it did for other patients in our study.

The overall logistic model including age, sex, and admission CBC variables had a respectable c statistic for predicting 30‐day mortality of 0.80. This compares well with findings in other multivariable models. For example, the APACHE II score used to predict the mortality of hospitalized critical care patients has a c statistic that ranges from 0.78 to 0.86.3, 27, 28 The APACHE score uses the worst value from the first 2 days after admission for some of its predictors so it cannot provide as early a warning as the admission CBC, and it requires collection of significantly more data. The inclusion of more CBC findings in the APACHE model might increase its predictive accuracy.

Our multivariate analysis was based on a very large number of patient samples using data collected through routine clinical care. However, our study has a number of limitations. The analysis was done at only a single institution, and the exact logistic regression model may not apply to other institutions that have different case mixes and laboratory procedures. Our institution's reported 30‐day mortality rate of 3.4% was lower than the 4.6%11.9% reported in studies of patients admitted to general ward services,2931 but this may be accounted for by the lower‐than‐average Charlson comorbidity scores in our study population. Our risk adjustment by Charlson comorbidity scores may not be as precise as a risk adjustment tailored for our particular institution.32 Our 30‐day mortality rate was calculated using state death tapes, which means we would have missed patients who died outside the state, although we believe this rarely happens. We developed predictive equations on the basis of 30‐day mortality, so we cannot comment on whether the CBC elements predict mortality beyond 30 days. We analyzed most variables as either high or low or as present or absent. Increasing degrees of abnormality may further increase the predictive power of some variables. Finally, the CBC is only one of many tests and clinical findings; it may be that some of these other variables would displace some CBC variables and/or improve the overall predictive power at the time the admission laboratory tests were performed. In this initial study, we have described the prognostic implication of the CBC across a wide range of diagnoses. Future work will focus on the predictive power of commonly gathered variables in more specific conditions (eg, low white blood cell count in sepsis).

Physicians generally have an intuitive ability to identify patients who are seriously ill and at high risk of dying33 and adjust their diagnostic and therapeutic efforts accordingly. Our analysis highlights the value that certain observations in the CBC, notably burr cells, NRBCs, and absolute lymphocytosis, add to physicians' assessments of mortality risk. Even after adjustment for age, sex, comorbidities, common admission diagnoses, and other variables in the CBC, the presence of these findings predicted a 3‐fold increase in 30‐day mortality. Identifying the red flags within this ubiquitously performed test can make the difference in premature discharge or inappropriate triage of patients. Busy physicians can choose from a wide selection of ever‐improving diagnostic tests, yet the workhorse CBC can serve as a simple and early identifier of patients with a poor prognosis.