PEDIATRICS Vol. 105 No. 3 March 2000, pp. 478-484

Evaluation of a Pediatric Hospitalist Service: Impact on Length of Stay and Hospital Charges

Paul S. Bellet, MD and Robert C. Whitaker, MD, MPH

From the Division of General and Community Pediatrics, Children's Hospital Medical Center, and University of Cincinnati College of Medicine, Cincinnati, Ohio.

	ABSTRACT

Top Abstract Methods Results Discussion References

Objectives.  Inpatient medical services supervised by pediatric hospitalist physicians are a new development in academic medical centers in the United States. In a large pediatric teaching hospital, we compared length of stay, readmission rates, and hospital charges for children admitted to medical services with and without a hospitalist system of care.

Design.  This retrospective observational study compared a baseline year of a traditional ward service (TS) with a subsequent year of a new hospitalist system of care called the Generalist Inpatient Service (GIS). Data were obtained from the hospital's clinical, demographic, and financial databases and from selected record review. All hospitalizations were at least 24 hours long and did not involve a stay in an intensive care unit.

Results.  The average length of stay was longer for the 627 TS hospitalizations than for the 813 GIS hospitalizations (2.7 ± 2.0 vs 2.4 ± 1.7 days). Total hospital charges were significantly lower on the GIS ($3002 ± $2160 vs $2720 ± $1933) because of lower room and respiratory therapy charges. Three readmissions to the TS and 8 to the GIS occurred within 24 hours of hospital discharge and were, therefore, considered potentially preventable by a longer initial hospital stay.

Conclusions.  In a large pediatric teaching hospital, a system of inpatient care provided by hospitalists can reduce length of stay. This model has the potential to control hospital charges in a period of increasing health care costs.  Key words:  health services, pediatrics, hospitalists, communication, hospital charges, length of stay, patient readmission.

Patients have traditionally been cared for in the hospital by their primary care physicians. Over the past few years, this system has been replaced in some areas of the country by one in which primary care physicians relinquish care of their hospitalized patients to inpatient specialists called hospitalists. One definition of a hospitalist physician is a physician that spends at least 25% of his/her time caring for hospitalized patients who have been referred by primary care physicians and who are referred back to their primary care physician at the time of discharge.¹

The same forces driving the hospitalist movement in adult medicinedemand for higher quality hospital care and more economical use of resourcesare also driving the hospitalist movement in pediatric medicine.² Some hospitalist programs in adult medicine have shown a decrease in average length of stay and hospital charges of 10% to 20%, while maintaining quality of care and patient satisfaction.^3,4 The development of pediatric hospitalist programs in academic medical centers in this country is new, and there are no published data concerning the performance of these programs. This retrospective observational study compares the traditional system of care with a hospitalist system of care at a large pediatric teaching hospital. We hypothesized that a hospitalist system, with a small number of academic general pediatricians who were focused on improving communication with referring physicians, could decrease length of stay and hospital charges.

	METHODS

Top Abstract Methods Results Discussion References

Overview

At Children's Hospital Medical Center (CHMC) in Cincinnati, Ohio, a hospitalist program called the Generalist Inpatient Service (GIS) began on September 1, 1996.² To develop the GIS, a questionnaire was mailed in June 1996 to all primary care physicians in 8 counties within a 50-mile radius of Cincinnati4 in Ohio, 3 in Kentucky, and 1 in Indiana. A total of 302 primary care physicians (240 family medicine and general practice physicians and 62 pediatricians) decided to refer their hospitalized patients to the GIS.

We analyzed all the hospitalizations at CHMC of children referred during a 2-year period by these 302 primary care physicians. The control hospitalizations occurred from September 1, 1995 to August 31, 1996 and were managed on the traditional ward service (TS) by hospital-based academic specialists and generalists and by volunteer community-based general pediatricians. The GIS hospitalizations occurred during the following year between September 1, 1996 and August 31, 1997, and they were managed only by hospital-based academic generalists. The CHMC institutional review board approved this study.

Structure of the Two Services

In the year before the development of the GIS, our hospital had 3 TS ward teams. Forty-four physicians served as TS attending physicians for these teams. These included 31 hospital-based academic pediatricians (19 subspecialists, 9 generalists, 2 chief residents, and 1 physician who served part of the study period as a general pediatric fellow and part-time as a chief resident), and 13 community-based general pediatricians. Each TS team included the TS attending physician, a senior pediatric resident, 4 first-year residents, and 4 medical students on their third-year pediatric clerkship. The TS attending physician was responsible for all patients whose primary care physician did not come to CHMC to provide inpatient care. The majority of these primary care physicians later referred their patients to the GIS when it was developed. They had the option of using the TS at the time when the GIS was in operation, but they did not do so. These physicians were not concerned about the loss of patients to the hospital-based physicians because they had referred their patients under the TS system and their patients were referred back to them.

On the GIS, 9 academic generalists and 1 general pediatric fellow from the Division of General and Community Pediatrics served as attending physicians. Six of these 10 physicians also served as TS attending physicians during the previous year. When the GIS began, the patients referred to the service were admitted to the 3 hospital ward teams, and the GIS attending physician worked with the residents on all 3 teams. Because of its expanding role, in July 1997 the GIS was given its own ward team to replace 1 of the existing 3 ward teams. This GIS team was subsequently divided into 2 teams and each consisted of an attending physician, a senior resident, 2 first-year residents, and 2 third-year medical students.

Unlike the TS, the GIS involved a planned system of communication between a patient's primary care physician and the GIS team. The primary care physician was called by the attending physician or resident within 24 hours of admission, during the hospitalization as necessary, and at discharge. A brief summary letter was sent to the referring physician after hospital discharge.

None of the attending physicians on the TS or GIS received feedback about length of stay or hospital charges for their patients. When attending on the GIS, the generalist physicians worked 1 half-day per week in outpatient clinics compared with 1 to 2 half-days per week when they were on the TS.

Data Collection and Analysis

We identified all hospitalizations to both the TS and GIS for children who received their primary medical care from 1 of the 302 primary care physicians who referred their patients to the GIS once it was developed. Hospitalizations were excluded from the study if, during the hospitalization, a stay occurred in the intensive care unit, the psychiatric inpatient unit, the short-stay unit (admission <24 hours), or if the hospitalization involved transfer to or from a subspecialty service. One of us (P.S.B.) reviewed the medical records of all children readmitted to our hospital within 10 days of discharge. We did not collect data on readmissions to other hospitals. Because our hospital contains nearly 75% of the pediatric inpatient beds in the 8-county primary service area, readmission to another hospital was unlikely.

Clinical, demographic, and financial data were obtained from our hospital database systems. We compared TS and GIS hospitalizations in terms of length of stay, readmission rates to our hospital within 10 days of discharge, use of physician consultations, and hospital charges. No physician charges were included. All charges were adjusted to our hospital's 1995-1996 fiscal year charges (July 1, 1995 to June 30, 1996). Room charges for the entire study period were calculated by multiplying the length of stay by the standard 1995-1996 daily bed and nursing service charge. The other charges were adjusted to 1995-1996 charges by accounting for a 5% annual increase in hospital charges. Charges for oximetry were excluded from our analysis because the hospital stopped charging for this service in February 1997, in the middle of the GIS study period. However, oximetry use was found to be no different between the TS and the GIS during the time oximetry charges were in effect. We compared patient age, gender, and race plus the type of insurance, specialty of referring physician, and final diagnosis-related group (DRG) assignment to determine whether the populations hospitalized on the TS and the GIS were similar.

² tests were used to compare patient characteristics (Table 1). Outlying data on length of stay and charges (>3 standard deviations above the mean) were truncated to 3 standard deviations above the mean.³ We report the means and standard deviations after this data transformation. Student's t tests on the log-transformed data yielded the same results as when the data were truncated at 3 standard deviations above the mean. We performed linear regression analysis to compare differences in charges between the TS and the GIS after controlling for differences in patient characteristics. Independent variables in our regression models included patient age (1 day to 2 years, 3 to 11 years, and 12 years), gender, race (white vs non-white), insurance status (Medicaid and uninsured vs private), type of referring physician (family medicine and general practice physician vs pediatrician), number of years of experience as an attending physician, duration of attending physician time on service (2 weeks vs >2 weeks per year), and length of stay. We also added the DRG for each hospitalization to our model using the 7 highest frequency DRGs (Table 2) plus an other DRG category for the remaining DRGs.

	RESULTS

Top Abstract Methods Results Discussion References

Hospitalization Demographics

Of the 44 TS attending physicians, 15 served for 1 to 2 weeks, 27 for 3 to 4 weeks, and 2 for 6 weeks. The mean number of years of experience as an attending physician for TS physicians was 9.1 ± 7.3 years. On the GIS, 1 physician served for 2 weeks, 4 for 4 weeks, 4 for 5 to 6 weeks, and 1 for 16 weeks. The mean number of years of experience as an attending physician for GIS physicians was 11.9 ± 9.4 years. On average, the TS physicians were on service a shorter time than the GIS physicians (3.3 vs 3.6 weeks/year).

Of the 721 TS hospitalizations, 94 were excluded73 involved a stay in the intensive care or short stay units and 21 involved a transfer to or from a specialty service. Of the 923 GIS hospitalizations, 110 were excluded105 involved a stay in the intensive care or short stay units and 5 involved transfer to or from a specialty service. Our final study sample contained 1440 hospitalizations627 on the TS and 813 on the GIS. These represented 9% and 13%, respectively, of the general medical admissions to the hospital during each study year.

The number of TS hospitalizations referred by family medicine and general practice physicians was 71% compared with 29% by pediatricians, whereas the number of GIS hospitalizations referred by family medicine and general practice physicians was 54% compared with 46% by pediatricians. Patients hospitalized on the TS and the GIS were similar in age, gender, race, and DRG assignment (Tables 1 and 2). Slightly more of the patients on the GIS were white and more had private insurance.

Length of Stay

The mean length of stay on the TS was longer than on the GIS (2.7 ± 2.0 vs 2.4 ± 1.7 days; P = .005). Mean length of stay was different between the services for 2 of the 7 most frequent DRGsDRG 98 (bronchitis and asthma) and DRG 298 (nutritional and miscellaneous metabolic disorders). Within DRG 98, approximately half the diagnoses were bronchiolitis and half were asthma. There was a longer mean length of stay for bronchiolitis hospitalizations on the TS than on the GIS (2.9 ± 2.0 vs 2.3 ± 1.4 days; P = .02), but there was no difference in the mean length of stay for asthma hospitalizations on these 2 services (2.0 ± 1.0 vs 2.0 ± 1.4 days; P = .95). Hospitalizations with DRG 298 (60% of which were for dehydration and 30% were for failure to thrive) also had a longer mean length of stay on the TS (3.8 ± 3.3 vs 3.2 ± 1.5 days; P = .019). The mean length of stay was the same (2.7 ± 2.0 days) among TS physicians that attended 2 weeks or less and those that attended >2 weeks. A similar analysis could not be performed for the GIS because only 1 physician served for 2 weeks or less.

Readmission Rates

The rate of readmission to our hospital within 10 days of discharge was lower on the TS than on the GIS (1% vs 3%; P = .006). The reasons for readmission included worsening of symptoms, unresponsiveness to prescribed outpatient treatment, new diagnoses, an intravenous line complication, and a drug reaction. Of the 6 TS and 25 GIS readmissions, 3 readmissions to the TS and 8 to the GIS occurred within 24 hours and were, therefore, considered potentially preventable if the initial hospital stay were longer (Table 3)

Consultation Rates

The percentage of hospitalizations with at least 1 physician consultation was similar on the 2 services (33% for TS vs 31% for GIS; P = .37) with both services having an average of 1.5 consultations per hospitalization among hospitalizations with any consultations. The pattern of consultation was similar on both services, with psychiatry and neurology being the 2 services most frequently consulted.

Hospital Charges

The mean total hospital charges were lower on the GIS and this resulted primarily from lower room and respiratory therapy charges (Table 4). Total charges were higher on the TS for bronchiolitis hospitalizations ($3234 ± $1875 vs $2402 ± $1322; P = .001) with this difference accounted for by both higher room charges ($1905 ± $1300 vs $1517 ± $931; P = .02) and respiratory therapy charges ($669 ± $537 vs $290 ± $364; P < .001). Total charges were similar on both services for asthma hospitalizations and for DRG 91 (simple pneumonia and pleurisy). However, in DRG 91, the TS mean respiratory therapy charges were higher ($446 ± $485 vs $264 ± $401; P = .019).

Effect of Bronchiolitis Clinical Guidelines

DRG 98 (bronchiolitis and asthma) was responsible for the highest percentage of hospitalizations on both the TS and GIS (Table 2). Therefore, our findings on length of stay and hospital charges for this DRG were potentially affected by the hospital's new clinical guidelines for bronchiolitis management, which were established on January 17, 1997, during the GIS period.⁵ The guidelines did not recommend chest physiotherapy, cool mist, supervised cough and suction, saline aerosol treatments, and the routine use of bronchodilator aerosol therapies in infants <1 year of age with a first episode of uncomplicated bronchiolitis. However, epinephrine inhalations could be considered as a trial therapy in selected patients. None of the GIS attending physicians were responsible for the development of the guidelines.

When bronchiolitis hospitalizations on the TS (n = 83) were compared with those on the GIS (n = 41) before implementation of the bronchiolitis guidelines, the charges were still higher on the TS: total ($3234 ± $1875 vs $2476 ± $1203; P = .008), room ($1905 ± $1300 vs $1474 ± $866; P = .031), and respiratory therapy ($669 ± $537 vs $381 ± $402; P = .001). Length of stay was also still longer on the TS for these hospitalizations (2.9 ± 2.0 vs 2.2 ± 1.3 days; P = .031).

When the bronchiolitis hospitalizations on the GIS alone were compared before (n = 41) and after (n = 85) the implementation of the bronchiolitis guidelines, there was a trend toward lower respiratory therapy charges after the guidelines ($381 ± $402 before vs $247 ± $338 after; P = .052), but no change in mean length of stay (2.2 ± 1.3 days before vs 2.3 ± 1.5 days after; P = .71). These lower respiratory therapy charges were attributable to less use of both chest physiotherapy and inhaled -agonist therapy after the guidelines were implemented (data not shown).

Effect of Generalist Practice

We performed 2 secondary analyses to assess the possible effect of generalist practice on the differences we observed between the TS and the GIS. When the TS hospitalizations not supervised by the 9 academic generalists (n = 483) were compared with the GIS hospitalizations (n = 813), we found the same significant reductions in both length of stay and hospital charges as in our primary analysis. We also compared the 87 TS and 605 GIS hospitalizations supervised only by the 6 hospital-based generalists who served on both the TS and GIS. Although the smaller sample sizes did not produce statistically significant differences, the reduced length of stay and hospital charges found on the GIS were in the same direction and magnitude as in our primary analysis.

Secular Trend

For the 2-year study period, we examined length of stay for hospitalized children who were not receiving primary care from any of the 302 community physicians referring patients in our study. To increase comparability of these hospitalizations with our study sample, we examined only hospitalizations over these 2 years that were supervised by the 35 physicians who had served as attendings on the TS but not on the GIS. As in our study sample, we also excluded hospitalizations involving a stay in the intensive care unit, short-stay unit, psychiatric unit, or on a specialty service. Among these 1923 hospitalizations, there was no significant decrease in mean length of stay between the TS study period and the GIS study period (2.7 days in both periods).

Regression Analysis

In regression analyses using using either length of stay or charges in Table 4 as dependent variables, the same significant length of stay and charge reductions (total, room, and respiratory) on the GIS were found after controlling for patient age, gender, race, insurance status, type of referring physician, years of experience as an attending physician, duration of time on service, and DRG. In a regression model for total charges, when length of stay was entered as the first independent variable, it accounted for 95% of the variance in total charges, and there was no longer a significant effect of service (TS vs GIS). This model suggests that charge reductions on the GIS were explained almost entirely by reductions in length of stay.

	DISCUSSION

Top Abstract Methods Results Discussion References

We studied hospitalizations under 2 systems of care at a large pediatric teaching hospital. On the GIS, a smaller group of academic general pediatricians accepted the responsibility of caring for the hospitalized patients of 302 referring physicians and referred the patients back to the primary care physician at discharge. A planned system of communication was established between the GIS physicians and referring physicians. Although only 1 of the 10 GIS attending physicians met the definition of Wachter and Goldman¹ of a hospitalist, this was the first year of a developing hospitalist service. Compared with the TS physicians, the smaller group of GIS physicians had more years of experience as attending physicians and attended more weeks per year.

The hospitalist system (GIS) reduced mean length of stay and hospital charges, when compared with the traditional system of care (TS). Shortened length of stay on the GIS was the primary determinant of the lower total charges because room charges, calculated as a function of length of stay, accounted for 58% of total charges. The reduced charges were primarily from hospitalizations for bronchiolitis.

Although mean length of stay on the TS was already just 2.7 days, length of stay on the GIS was 11% shorter (2.4 days). This decrease is consistent with that seen in 2 adult hospitalist programs.^3,4 Our study did not include process measures to explore the mechanism by which the GIS may have resulted in reduced length of stay. However, we suspect that the GIS reduced length of stay by focusing on improved communication between the GIS and the primary care physicians. It is possible that both the hospitalists and the primary care physicians may have been more comfortable discharging children earlier from the GIS because there had been better communication about follow-up plans by the smaller number of GIS physicians who were more familiar with the referring physicians. Just because hospitalists care for patients does not mean that communication improves, but improved communication was a specific focus of the system we implemented. We cannot exclude that other factors unrelated to the implementation of the hospitalist system of care caused the shorter length of stay. However, a separate analysis showed no significant secular trend at CHMC in decreasing length of stay during our study period.

Our secondary analyses support the idea that the structure of the GISnot the particular attending physicians involvedwas responsible for the decreased length of stay. In 1 secondary analysis, which eliminated TS hospitalizations supervised by the 9 academic generalists, the results were the same as the primary analysis (Table 4). Furthermore, in another secondary analysis, which compared hospitalizations supervised only by the 6 hospital-based generalists who served on both the TS and the GIS, our findings were also similar to the primary analysis.

On the GIS, respiratory therapy charges were lower for bronchiolitis and pneumonia hospitalizations but not for asthma hospitalizations. For bronchiolitis hospitalizations, the lower respiratory charges on the GIS were attributable to decreased length of stay both before and after the bronchiolitis guidelines and less intensive use of some respiratory services after the bronchiolitis guidelines. For the pneumonia hospitalizations (DRG 91), the trend toward shorter length of stay on the GIS (2.7 ± 1.7 vs 2.4 ± 1.9 days; P = .34) explained the lower respiratory therapy charges with length of stay explaining 67% of the variance in respiratory charges in this DRG.

Hospital charges can serve as a proxy for resource utilization for hospitalized patients.⁶ One way in which a hospitalist service might reduce use of hospital resources is by being more efficient and cost-effective in both the diagnosis and treatment of common childhood illnesses requiring hospitalization. Supporting the idea that practice makes perfect is evidence from adult medicine in the care of patients with illnesses such as acute myocardial infarction, stroke, and AIDS.^7-9 However, aside from the impact of the bronchiolitis guidelines on respiratory therapy charges, we had no evidence that the GIS physicians used fewer resources for diagnostic testing (radiology or laboratory) or therapy (respiratory or pharmacy). This may be because only 1 GIS physician met the definition of Wachter and Goldman¹ of a hospitalist or because the TS physicians were a self- selected group who had a high level of knowledge and interest concerning general pediatric inpatient care.

There was a higher readmission rate on the GIS compared with the TS, but many of the readmissions may not have been preventable (Table 3). Improved communication with the primary care physician may have caused the GIS physicians to discharge children sooner, and perhaps, sicker, because proper follow-up was ensured. In our experience, there was no undue pressure to discharge earlier. Rather, because proper follow-up was assured, patients could be discharged earlier. Proper communication occurred and management plans were in place, but even this does not always prevent readmission. Even when we reanalyzed our data including the initial admission and the readmission as a single hospitalization, all the major findings shown in Table 4 were still the same, including a shorter length of stay on the GIS (2.7 vs 2.5 days; P = .027).

Despite the limitations of using the TS hospitalizations as historical controls, the TS patient demographics and final DRG assignments were remarkably similar to those of the GIS hospitalizations because all the patients were referred from the same group of primary care physicians. Our study did not address the issues of patient satisfaction and resident teaching, but these subjects have been addressed in adult medicine. Two studies showed no significant difference in patient satisfaction between the hospitalist system and the previous system,^3,4 and 1 study also demonstrated that resident and faculty satisfaction were equivalent in the 2 systems.³

We have shown that the implementation of a hospitalist service in a large pediatric teaching hospital is associated with a decreased length of stay for medical hospitalizations. This system of care was also associated with a decrease in charges for hospitalization at a time when health care charges were increasing faster than the general rate of inflation. By implementing a pediatric hospitalist system of care in hospitals where the average length of stay is even longer than in our hospital, greater reductions in length of stay and hospital charges may be possible. In our hospital, however, where the mean length of stay on GIS is now 2.4 days, we believe that future research should focus on the effect of the hospitalist system on quality of care. In particular, we wish to understand how improved communication among the hospitalist, the primary care physician, and a child's family may allow shorter hospital stay and also improve health outcomes and satisfaction with care.

	ACKNOWLEDGMENTS

This work was supported by the Generalist Physician Faculty Scholars Award from the Robert Wood Johnson Foundation, Princeton, NJ (to R.C.W).

We are indebted to the following individuals from Children's Hospital Medical Center, Cincinnati, OH: Dr Thomas DeWitt, Division of General and Community Pediatrics, for encouragement and critical review of the manuscript; Elisa Immerman, Division of Health Policy and Clinical Effectiveness, and Wayne Geers, Information Services, for help with data collection; Deborah Kaiser, Division of Patient Services, for assistance with financial analysis; and Dr Judy Bean, Biostatistics Program, for assistance with statistical analysis. We are also indebted to Dr Jonathan Showstack from the Institute for Health Policy Studies, University of California, San Francisco, CA, for advice on conceptualization and performance of the data analysis.

	FOOTNOTES

Received for publication Mar 29, 1999; accepted Aug 25, 1999.

Reprint requests to (P.S.B.) Division of General and Community Pediatrics, Children's Hospital Medical Center, 3333 Burnet Ave, Cincinnati, OH 45229-3039. E-mail: paul.bellet{at}chmcc.org

	ABBREVIATIONS

CHMC, Children's Hospital Medical Center; GIS, Generalist Inpatient Service; TS, traditional ward service; DRG, diagnosis-related group.

	REFERENCES

Top Abstract Methods Results Discussion References

1. Wachter RM, Goldman L The emerging role of "hospitalists" in the American health care system. N Engl J Med 1996; 335:514-517 [Free Full Text]
2. Bellet PS, Wachter RM The hospitalist movement and its implications for the care of hospitalized children. Pediatrics 1999; 103:473-477 [Free Full Text]
3. Wachter RM, Katz P, Showstack J, Bindman AB, Goldman L Reorganizing an academic medical service: impact on cost, quality, patient satisfaction, and education. JAMA 1998; 279:1560-1565 [Abstract/Free Full Text]
4. Freese RB The Park Nicollet experience in establishing a hospitalist system. Ann Intern Med 1999; 130:350-354 [Abstract/Free Full Text]
5. Perlstein PH, Kotagal UR, Bolling C, Evaluation of an evidence-based guideline for bronchiolitis. Pediatrics. 1999; 104:1334-1341 [Abstract/Free Full Text]
6. Garber AM, Fuchs VR, Silverman JF Case mix, costs, and outcomes: differences between faculty and community services in a university hospital. N Engl J Med 1984; 310:1231-1237 [Abstract]
7. Jollis JG, DeLong ER, Peterson ED, Outcome of acute myocardial infarction according to the specialty of the admitting physician. N Engl J Med 1996; 335:1880-1887 [Abstract/Free Full Text]
8. Stroke Unit Trialists' Collaboration Collaborative systematic review of the randomised trials of organized inpatient (stroke unit) care after stroke. Br Med J 1997; 314:1151-1159 [Abstract/Free Full Text]
9. Kitahata MM, Koepsell TD, Deyo RA, Maxwell CL, Dodge WT, Wagner EH Physicians' experience with the acquired immunodeficiency syndrome as a factor in patients' survival. N Engl J Med 1996; 334:701-706 [Abstract/Free Full Text]