Objective. The staff/housestaff hospitalist system has been evaluated in 2 pediatric centers in the United States. In Canada, fewer residents and duty hour restrictions led to the development of a staff-only hospitalist system. The objective of this study was to compare the staff-only pediatric hospitalist system and the staff/housestaff hospitalist system with respect to traditional outcome measures.
Design. This cohort study (staff-only hospitalist system versus staff/housestaff system) used electronic health records data (July 1, 1996, to June 30, 1997) for all admissions (n = 3807) to the general pediatric inpatient unit of an urban, tertiary care, pediatric, teaching hospital in Toronto, Canada. Outcome measures included length of hospital stay, subspecialty consultations, readmission to the hospital, and death during the hospital stay.
Results. The median length of hospital stay was reduced by 14% for patients admitted to the staff-only hospitalist system, compared with the staff/housestaff hospitalist system (2.5 and 2.9 days, respectively). This difference remained statistically significant after adjustment for age, gender, and comorbidity. There were no significant differences between the 2 models of care with respect to subspecialty consultation, hospital readmission, or mortality rates. A stratified analysis showed similar findings for the 10 most frequent diagnostic groups.
Conclusions. The staff-only hospitalist system was associated with a significant reduction in the hospital length of stay, without evidence of adverse effects on mortality or readmission rates, compared with the staff/hospitalist system. In the context of recent restrictions on resident duty hours in the United States, these findings may be of interest to pediatric teaching hospitals considering the development of a similar staff-only hospitalist model.
Despite health care system differences, hospital costs constitute significant proportions of overall health care expenditures in both Canada and the United States. In 1996, Wachter and Goldman1 described the emerging role of “hospitalists” in the American health care system. In Canada, the hospitalist model might have been in existence for decades, although not named as such.2
A hospitalist has been defined as “… a physician who spends at least 25% of his or her professional time serving as the physician-of-record for inpatients, during which time he or she accepts ‘hand-offs’ of hospitalized patients from primary care providers, returning the patients to their primary care providers at the time of hospital discharge.”3(p338) The hospitalist system of care was evaluated at 2 tertiary care, pediatric, teaching hospitals in the United States.4,5 Using a before/after study design4 and an interrupted time-series analysis,5 the 2 studies demonstrated shorter hospital stays and lower hospital costs for the hospitalist system, compared with the traditional system. At both centers, pediatric residents and medical students were assigned to the hospitalist team.
The Hospital for Sick Children (HSC) in Toronto, Canada, is a tertiary care, pediatric, academic health sciences center affiliated with the University of Toronto. The division of general pediatrics operates a large inpatient unit, with ∼4000 admissions per year, representing ∼25% of all hospital admissions. Before the mid-1990s, pediatric residents played an integral role on the pediatric inpatient team. Reductions in medical manpower throughout Canada and negotiated union agreements limiting resident duty hours6–8 led the division of general pediatrics to undertake a reorganization of the inpatient pediatric teams in 1995. This resulted in 2 hospitalist models, in 2 separate inpatient units, ie, the staff/housestaff clinical teaching unit (CTU) and the staff-only clinical practice unit (CPU).
Although the literature supports the efficiency of the staff/housestaff pediatric hospitalist system, the staff-only pediatric hospitalist model has not been assessed. Evaluation of the staff-only pediatric hospitalist model may be of interest to pediatric academic centers in the United States, given the recent introduction of restrictions on residents' duty hours.9 The objective of this study was to compare the staff-only pediatric hospitalist system (CPU) with the staff/housestaff hospitalist system (CTU), with respect to traditional measures of patient outcomes and hospital resource use.
The Hospitalist Systems
After admission to the general pediatric inpatient unit, patients were assigned to the CPU or CTU according to bed availability and physician patient census. Approximately 90% of children were admitted via the emergency department. Allocation to the CPU or CTU was based on maintaining a daily patient census of 24 to 30 patients on the CTU, with the remainder allocated to the CPU. There was a divisional policy to allocate patients to the 2 units in an equitable manner, without regard for the admitting diagnosis. That is, it was the explicit intent of the division to allocate patients with acute illnesses (with anticipated short hospital stays) and patients with complex illnesses (with anticipated long hospital stays) equally to the 2 hospitalist systems.
The medical team on the CTU consisted of 1 attending pediatrician, 3 or 4 pediatric residents, and 2 medical students. Attending pediatricians served for 4 to 8 weeks per year, with the remainder of the year being allocated to outpatient clinical activity (hospital or community-based), research, education, and administrative activities. Attending physicians assessed all patients the first morning after admission, led daily ward rounds, were available by telephone for the residents, and assessed all patients before discharge. The daily patient census was maintained at 24 to 30 patients for the CTU team.
On the CPU, 3 pediatricians were responsible for all aspects of patient care from admission to discharge, including admission and daily clinical assessment and documentation, computerized order entry, review of results of laboratory tests and diagnostic imaging, dialogue with subspecialty consultants, communication with community primary care providers, and preparation of discharge documents. (In comparison, the majority of these activities were undertaken by pediatric residents on the CTU.) The CPU physicians worked in the hospital Monday through Friday from 8 am to 5 pm and on weekends on the basis of a rotating schedule. CPU physicians spent ∼11 months of the year providing inpatient care. Medical students but not pediatric residents were assigned to the CPU. Each physician maintained a daily census of 8 to 10 patients.
Overnight and on weekends, patients on the CTU were cared for by pediatric residents on a rotating schedule, including some of the residents assigned to the CTU in the daytime. Therefore, there was adequate continuity of care. On the CPU, subspecialty clinical fellows not assigned to the CPU in the daytime provided night and weekend coverage as part of a separate initiative.6 Therefore, there was limited continuity of care. Neither the CPU nor CTU attending pediatricians stayed in the hospital overnight, but they were available at all times by pager. On weekends, both CPU and CTU attending pediatricians provided in-hospital care for their respective patients, according to a rotating schedule. As on weekdays, CTU attending physicians made ward rounds with housestaff members. On the CPU, attending physicians provided direct patient care to newly admitted patients, and housestaff members provided daily care for the remaining patients.
In accordance with hospital policy, all discharge decisions were made by the attending physician, either the CPU or CTU attending physician. That is, trainees were not responsible for these decisions, and patient length of hospital stay was determined by the attending pediatricians only. Although medical students were assigned to both the CPU and CTU teams, their role was largely educational, with limited responsibility for direct patient care.
All attending physicians had a minimum of 4 years of pediatric specialty training and had completed all qualifications for the Canadian specialist certificate in pediatrics. The mean number of years from medical school graduation for CPU attending physicians was 8 years (range: 6–11 years), and that for CTU attending physicians was 22 years (range: 8–35 years).
Data Collection and Analyses
Data for the study were obtained from the HSC electronic health records database. The study period was from July 1, 1996, through June 30, 1997, the first full academic year with the new system at HSC. All patients admitted to the general pediatric inpatient unit (in either the CTU or the CPU) during this period were identified. Demographic and clinical information gathered for each patient included age, gender, mode of entry (through the emergency department or through transfer from another hospital), stay in a special care unit, and most responsible diagnosis.
Each patient was categorized as having comorbidity or no comorbidity with a variable developed by the Canadian Institute for Health Information. This variable reflects the presence of diagnoses other than the most responsible diagnosis. This standardized variable (coded by trained hospital records technicians) has 4 categories, ie, (1) uncomplicated hospital stay, (2) complicated hospital stay related to chronic conditions, (3) complicated hospital stay related to serious/important conditions, and (4) complicated hospital stay related to potentially life-threatening conditions. A priori, we developed a dichotomous variable for the purposes of analysis. Children assigned to category 1 were analyzed as having no comorbidity. Children assigned to categories 2, 3, and 4 were analyzed as having comorbidity.
The primary outcome variable was length of hospital stay. Secondary outcome variables were subspecialty consultations, readmission rate, and mortality rate. Length of hospital stay was measured in hours, converted to days, and rounded to 1 decimal place. Subspecialty consultations were categorized as no consultations or ≥1 consultation, and readmission was defined as readmission to HSC with the same diagnosis or a related diagnosis within 7 days after discharge.
Univariate analyses, with the Mann-Whitney U test for continuous variables and χ2 or Fisher's exact test for categorical variables, were used to test for differences between the CTU and CPU hospitalizations with respect to baseline characteristics and outcome measures. Multivariate linear regression analysis was also performed, with length of stay as the dependent variable, CPU/CTU classification as the independent variable, and age, gender, and comorbidity as potential confounding variables. For the binary outcomes (subspecialty consultation and readmission), multivariate logistic regression analysis was performed with CPU/CTU classification as the independent variable and age, gender, and comorbidity as potential confounding variables. Lastly, a stratified exploratory analysis examined differences between the CPU and CTU with respect to length of stay, consultations, and readmission for each of the 10 most common diagnostic groups. The purpose of this exploratory analysis was to ensure that differences in length of stay were not attributable to a single diagnosis that was distributed unequally between the 2 groups.
During the study period, there were 3807 admissions to the general pediatric inpatient unit. Of these, 1274 (33%) were to the CPU and 2533 (67%) were to the CTU. Table 1 presents the baseline characteristics of patients admitted to the 2 units. Patients admitted to the CPU, compared with patients admitted to the CTU, were significantly older (median age: 95 weeks and 69 weeks, respectively; P < .01), less likely to be male (54% and 57%, respectively; P = .06), and less likely to have a comorbid condition (24% and 30%, respectively; P < .01) The distributions of all 4 categories of comorbidity were similar in the CPU and CTU groups. Patients admitted to the 2 units were equally likely to have been admitted through the emergency department (90%) and to have had a stay in a special care unit during their hospitalization (6%).
Table 2 describes the clinical outcomes for the CPU and CTU patients. The median length of hospital stay was 2.5 days (interquartile range [IQR]: 1.6-4.4 days) for the CPU, compared with 2.9 days (IQR: 1.8-4.9 days) for the CTU (P < .01); values expressed in hours were 60 hours (IQR: 31-90 hours) for the CPU and 68 hours (IQR: 30-106 hours) for the CTU (P < .01). Multivariate linear regression analysis showed a significant difference in the length of hospital stay between the CPU and CTU (P = .04) after adjustment for age, gender, and comorbidity.
The frequency of ≥1 subspecialty consultation did not differ between the CPU and CTU (28% and 30%, respectively; P = .20). Multivariate logistic regression analysis showed no difference in consultation frequency between the 2 units after adjustment for age, gender, and comorbidity.
The frequency of readmission to the hospital within 7 days because of the same diagnosis or a related diagnosis was not significantly different between the CPU and CTU (2% and 2%, respectively; P = .70). Multivariate logistic regression analysis also showed no difference in readmission frequency between the 2 units after adjustment for age, gender, and comorbidity. During the study period, there were 7 deaths (1 in the CPU and 6 in the CTU, P = .40).
The 10 most frequent diagnoses were asthma (n = 515), bronchiolitis (n = 282), pneumonia (n = 219), upper respiratory tract infection (n = 142), sickle cell disease (n = 130), urinary tract infection (n = 128), convulsions (n = 119), viral infection (n = 113), rotavirus gastroenteritis (n = 101), and other gastroenteritis or colitis (n = 91). These diagnoses (n = 1840) constituted 48% of all hospitalizations. Table 3 presents the distribution and length of stay for CPU and CTU patients according to diagnostic group. The distributions of cases in the 10 most frequent diagnostic groups were similar for the CPU and the CTU. For all 10 diagnostic groups, the median length of hospital stay for the CPU was shorter than the median length of stay for the CTU (2.1 days vs 2.6 days, P < .01). There were no differences between the CPU and the CTU with respect to subspecialty consultations, hospital readmission, or mortality rates for the 10 most common diagnostic groups (data not shown).
The remaining diagnoses (n = 1967) constituted 52% of all hospitalizations (each diagnosis contributed <2% of all hospitalizations). For these conditions also, the length of stay for patients admitted to the CPU was shorter than that for patients admitted to the CTU (2.8 days vs 3.1 days, P < .01).
This study showed that patients admitted to a staff-only hospitalist system in a large, urban, tertiary care, pediatric center in Toronto, Canada, experienced a 14% reduction in length of hospital stay, compared with patients admitted to the staff/housestaff-model CTU. This finding remained statistically significant even after adjustment for age, gender, and comorbidity. The staff-only hospitalist system of care demonstrated similar rates of subspecialty consultations, hospital readmission, and death, compared with the staff/housestaff system of care. A stratified analysis showed similar findings for the 10 most frequent diagnostic groups. The difference in length of stay of 8 hours may be important enough to reduce the time to inpatient admission for patients waiting for transfer from the emergency department or from special care units.
Although the study was observational in nature, several features lend strength to the findings. Selection bias was considered minimal, because patients were assigned to the CPU or the CTU on the basis of bed and physician availability. For example, the distributions of the 10 most frequent diagnoses (accounting for 48% of all admissions) were similar for the 2 units. Information bias was reduced because the same standardized data source was used for collection of patient information. Potential confounding variables (age, gender, and comorbidity) were taken into account in the multivariate regression analysis. In addition, a stratified exploratory analysis examined clinical outcomes for the 10 most frequent diagnostic groups. The primary outcome variable, length of stay, was determined only by the attending physicians (either CPU or CTU). Finally, the use of concurrent control cases, selected from the same target population within the same hospital, limits the biases associated with a before/after study design.
There were several limitations to the study. The study was conducted at a single center; therefore, the results should be generalized with caution. Some baseline characteristics were found to be different, likely because of the large sample size. The clinical meaning of these differences is unclear, and the final results remained significant after adjustment for those characteristics. We were unable to examine hospital costs, because such data were not collected by the hospital at the time of the study. However, previous investigators found that hospital costs are attributable largely to the length of stay.4 Finally, the outcome measures used in this study (length of hospital stay and rates of subspecialty consultations, hospital readmission, and death), although similar to those used by previous investigators,4,5 may be of limited value in demonstrating the more subtle strengths and weaknesses of the various systems. Although we were unable to measure the satisfaction of patients, physicians, and other members of the health care team, we recognize that these measures are important indicators of the success of an integrated health care system.
In 2002, Wachter and Goldman10 reviewed the published literature on the impact of hospitalist systems on resource use, quality of care, satisfaction, and teaching. They identified 19 published studies (12 articles and 7 abstracts) from a variety of health care delivery settings (teaching and nonteaching, urban and rural, adult and pediatric, and managed care and nonmanaged care). All except 2 of the 19 studies identified a reduction in the average length of stay (range: 7–27% shorter length of stay). Most studies found no change in clinical outcomes such as mortality or hospital readmission rates. Those authors concluded that the hospitalist system of care improved inpatient efficiency without harmful effects on the quality of care.
Two studies evaluated staff/housestaff hospitalist systems of care in the pediatric setting.4,5 With a before/after study design, Bellet and Whitaker4 compared a traditional ward service with a subsequent 1 year of a new system of care called the generalist inpatient service. The traditional ward service teams were lead by 44 physicians (academic subspecialists, academic generalists, and community-based general pediatricians), whose attendance on the inpatient unit averaged 3.3 weeks per year (range: 1–6 weeks). The generalist inpatient service teams were lead by 10 academic generalists, whose attendance on the inpatient unit averaged 3.6 weeks per year (range: 2–16 weeks). Residents participated in both teams. The generalist inpatient service system resulted in a reduction in the average length of stay of 11%. Hospital charges were reduced by 9%, with the reduced length of stay being the primary reason. The authors reported no differences in consultation rates but noted a higher rate of readmission for patients discharged by the generalist inpatient service team, compared with the traditional ward service team.
A program in Boston, Massachusetts, was evaluated by Landrigan et al.5 With an interrupted time-series analysis, the authors compared outcomes during the 3 years preceding and the 2 years following the introduction of a hospitalist system. The hospitalists spent 50% to 100% of their time caring for inpatients. Housestaff members and medical students were also involved in the care of the patients. Similar time-series analyses were undertaken for several comparison groups of hospitalized patients during the same time period. The before/after analysis showed a reduction in the mean length of stay of 12% (from 2.5 days to 2.2 days) and a reduction in the mean cost of hospitalization of 16%. Parental ratings of care improved. There were no changes in follow-up, mortality, or readmission rates. In contrast, the comparison groups did not show changes in any of the outcomes.
The staff/housestaff CTU model at HSC in Toronto most closely resembles the hospitalist models described for Cincinnati and Boston. Our comparison of a staff/housestaff model (CTU) with a staff-only model (CPU) suggests that additional hospital efficiencies may be achieved. In addition, the relative inexperience of the CPU physicians, compared with the CTU physicians, suggests that the efficiencies may be sustainable. Davis et al11 suggested that the main benefit of a hospitalist system is physician availability. The presence of an attending physician in the inpatient unit throughout the day allows for repeat patient encounters, ongoing evaluation of patient status, and timely interpretation of diagnostic tests. It has also been suggested that physicians who are in the hospital throughout the day may be more aware of the hospital system and may be better able to manage system difficulties or disruptions.2 Furthermore, physicians who spend large amounts of time on the inpatient unit form relationships with other caregivers, such as nurses, administrators, and discharge planners, which leads to greater efficiency of care.3 It is likely that these factors play an even larger role in the staff-only model, compared with the staff/housestaff model.
New models of health care delivery often emerge as a result of changing economic and political realities. In the United States, the hospitalist model emerged in the 1990s, whereas a similar model might have existed for decades in Canada.2 In the Canadian setting, hospitals and academic departments have been driven to develop new models as a result of reductions in postgraduate training positions and resident duty hours in the past decade.6–8 At our hospital, the development of the staff-only hospitalist model was one such innovation. In the United States, the Accreditation Council for Graduate Medical Education introduced duty hour standards with which all residency programs must comply (effective July 1, 2003).9 These standards, which limit resident duty hours to a maximum of 80 hours per week and set other restrictions, may result in the consideration of staff-only hospitalist models in other pediatric teaching hospitals.
The implementation of a staff-only hospitalist system of care at a large, urban, pediatric, academic health sciences center was associated with a reduction in hospital length of stay, without adverse effects on mortality or hospital readmission rates, compared with the staff/hospitalist system.
The Pediatric Outcomes Research Team is supported by a grant from the Hospital for Sick Children Foundation (Toronto, Canada).
We gratefully acknowledge Arran Shemmans and Betty Longlade, Department of Health Records, Hospital for Sick Children, who assisted with data collection from the hospital electronic health database.
- Accepted June 16, 2004.
- Address correspondence to Patricia C. Parkin, MD, FRCPC, Division of Pediatric Medicine, Hospital for Sick Children, 555 University Ave, Toronto, Ontario, Canada M5G 1X8. E-mail:
No conflict of interest declared.
- ↵Bellet PS, Whitaker RC. Evaluation of a pediatric hospitalist service: impact on length of stay and hospital charges. Pediatrics.2000;105 :478– 484
- ↵Landrigan CP, Srivastava R, Muret-Wagstaff S, et al. Impact of a health maintenance organization hospitalist system in academic pediatrics. Pediatrics.2002;110 :720– 728
- ↵Friedman J, Laxer RM. Providing after-hours on-call clinical coverage in academic health sciences centres: the Hospital for Sick Children experience. CMAJ.2000;163 :298– 299
- ↵Professional Association of Internes and Residents of Ontario. Available at: www.pairo.org
- ↵Accreditation Council for Graduate Medical Education. Available at: www.acgme.org
- Copyright © 2004 by the American Academy of Pediatrics