• Center on Health Equity and Access
  • Clinical
  • Health Care Cost
  • Health Care Delivery
  • Insurance
  • Policy
  • Technology
  • Value-Based Care

Comparison of Hospital Costs and Length of Stay Associated With General Internists and Hospitalist Physicians at a Community Hospital

The American Journal of Managed CareSeptember 2004
Volume 10
Issue 9

Objective: To compare initial and long-term hospital resourceutilization of hospitalists and general internists at an urban communityhospital.

Study Design: A retrospective cohort analysis.

Methods: Data were collected from 11 750 admissions associatedwith 28 diagnosis-related groups for 6 years (October 1, 1994,to September 30, 2000). Hospital cost, length of stay, mortality,and 30-day readmission rates of general and hospitalist internistswere analyzed for comparison.

Results: Hospitalists had generally lower utilization comparedwith general internists. Overall, length of stay and hospital costwere 16.1% and 8.3% lower, respectively, for hospitalists. Hospitalmortality and 30-day readmissions were similar for both physiciangroups.

Conclusions: Hospital resource utilization is moderately lowerfor hospitalists compared with private practicing general internists.

(Am J Manag Care. 2004;10:626-630)

The cost of medical care and especially hospitalcare in the United States has been rapidlyincreasing after several years of stability. Oneproposed answer to the high cost of hospital care hasbeen to introduce hospitalists or hospital-based facultyin teaching and nonteaching hospitals.1-9 Evidence isgrowing that hospitalists may be more efficient in hospitalresource utilization than traditional practicingphysicians. However, introduction of the hospitalist systemhas not always led to reduced costs.3,8 Studies ofhospitalist or hospital-based faculty introductions havebeen limited by narrow diagnostic focus,1 short observationperiod,2-9 small numbers of hospitalist physicians,4-7 or relatively few admissions.1,7,9 Any or all ofthese factors could affect the outcome of a hospitalistsystem.

We decided to study the natural history of hospitalresource utilization trends of physicians in a communityhospital setting after hospitalists began practicing.We wanted to compare long-term utilization patterns ofhospitalist internists and general internists in privatepractice. We included many more admissions over alonger time, more diagnostic categories, and more hospitalistsin our study than previously reported.



The study was conducted at a 500-bed urban, non-university,community hospital from October 1, 1994,to September 30, 2000. The study population consistedof all patients admitted to the internal medicine physiciansof the hospital medical staff who were in privatepractice.

Private practice internists included 131 generalinternists (generalists) and 27 hospitalist internists(hospitalists). All members of the Internal MedicineDepartment were board certified or board eligible ininternal medicine.

The generalists consisted of a mixture of privatepractice models including solo practitioners and smalland large group practices. Hospital coverage arrangementsvaried from cross-coverage among solo practitionersand small group practices to self-containedcoverage in larger groups.

The hospitalists consisted of 1 practice initially andgrew to 6 separate practices. Each hospitalist practicewas responsible for its own hospital coverage. All hospitalistpractices covered more than 1 hospital and noneprovided 24-hour in-hospital physician coverage. Only1 hospitalist group provided outpatient facilities for follow-up care. No hospital policies or financial incentiveswere provided to the hospitalists to encourage or discouragepractice at the hospital. The decision of thehospital administration to use hospitalists was completelyvoluntary.

Most of the patients admitted to private practicephysicians came from the Emergency Department ofthe hospital. Patients were admitted to any availablehospital bed without regard to socioeconomic or physicianfactors. The protocol for patient assignment fromthe Emergency Department was that patients with anestablished physician were admitted to the physician.Hospitalist patients were assigned by prior arrangementwith other community physicians.

The Internal Medicine Department also contained aresidency-training program. The medical house staffwas supervised by full-time teaching physicians employedby the hospital. Uninsured and half of otherunassigned patients coming to the Emergency Departmentwere cared for by the residency-faculty teams.The other half of the unassigned patients was admittedto a rotating schedule of physicians includingthe hospitalists and private practice physicians.Medical residents did not participate in the care ofpatients assigned to generalists or hospitalists. The patientsof the medical residency teams were not includedin the study.


The study had a retrospective cohort design.Resource utilization (cost and length of stay), 30-dayreadmission rates, and mortality were measuredsequentially during 6 years of the study for both physiciangroups.

Data Source and Collection

Trendstar Clinical Costing software (McKessenHBOC Inc, San Francisco, Calif) was used to collectand report information on resource utilization andclinical information for the entire 6 years of the study.To determine costs, Trendstar used an activity-basedcost accounting system.10,11 Briefly, the hospital'sledger provided all hospital costs (direct, indirect,fixed, and variable), which were then assigned to adirect cost center. The hospital's procedure codes providedthe frequencies of each procedure. The proceduresand their frequencies were totaled to obtaintotal costs.

All 17 diagnosis-related groups (DRGs) with 250 ormore admissions during the 6 years of the study wereselected for serial comparison over the study intervals.The 11 DRGs closely associated with the 17 primaryDRGs were included as well. For instance, DRG 80 (respiratoryinfections without complications) was includedalong with the primary DRG 79 (respiratory infectionswith complications). The DRGs represented a widerange of medical disorders. Diagnosis-related groupsthat pertained only to a specific medical or surgical procedure(such as DRGs 112, 115, and 116 for percutaneouscardiac procedures) were not included in theanalysis.

End points in the study included resource utilizationand quality-of-care markers. The resource utilizationend points were total hospital costs (cost) and length ofstay (LOS) per discharged case. Physician fees were notincluded in the costs. Quality-of-care end points weremortality (death during the hospital stay) and 30-dayreadmission rates (calculated for each physician in thestudy and averaged for each physician category).Readmissions were attributed to the first physician whodischarged the patient within the 30-day intervalregardless of the second or subsequent physicianassignment. Costs were contemporary and not discountedfor inflation.

Demographic information that was collected on eachcase included age, race (Caucasian, black, other), sex,and health insurance coverage (Medicare, Medicaid,commercial, other).

The Institutional Review Board of Orlando RegionalHealthcare approved the study methods and analyticprocedure.

Statistical Analysis

Cost, LOS, mortality, and readmission rates of the 28(17 primary and 11 secondary) DRGs were comparedbetween the physician categories. The highest and lowest0.5% cost and LOS admissions were removed as outliersprior to analysis.

Because of skewness and non-normality, cost andLOS underwent log base 10 transformation prior to statisticalanalysis. Geometric means from these analyseswere reported. Cost and LOS were compared over thestudy periods using general linear modeling regressionanalysis, adjusted for differences in DRG relative weight,and demographic variables for each of the physician categories.The dependent variables were cost and LOS.The independent variables were age, sex, race, insurance,DRG relative weight, physician category, and year.Year was placed in the model as a dummy variable toadjust for patient frequency changes over time.

Mortality and readmission rates were comparedusing binary logistic regression analysis. In these analyses,mortality and readmission were the dependent variablesand age, sex, race, insurance, DRG relative weight,and physician category were independent variables.


All of the independent variables were includedbecause they were each statistically significantly ( <.01) associated with the dependent variable in univariateand multivariable regressions.

SAS software (SAS Institute, Cary, NC) was used forall statistical analyses.


Table 1 depicts the basic demographic and DRGinformation for the 11 750 patients in the 2 physiciancategories. Statistically significant differences werefound between the physician categories for each of thepatient characteristics. Because of these differences, allend points were statistically adjusted during the analysisfor the demographic and case-mix factors.

Table 2 displays the number of physicians, numberof admissions, and average DRG relative weight byyear and physician category. The number of hospitalistsand hospitalist admissions increased rapidly duringthe study, whereas those of the generalists wererelatively constant over time. The average DRG relativeweight of hospitalist patients was lower than thatof generalist patients until the last 2 years of thestudy.

Cost and LOS trends for all DRGs during the 6 yearsof the study are depicted in the Figure for both physiciancategories. The cost and LOS were adjusted for age,sex, race, insurance, and DRG relative weight.Hospitalists had lower adjusted cost and LOS than generaliststhroughout most of the study years. Only 1 hospitalistaccounted for the cases in year 1 and 2, but thenumber of hospitalists increased over time.

The resource and utilization gap narrowed as thenumber of physicians and admissions increased. Table3 shows resource utilization for the physician categories.Both unadjusted and adjusted geometric meansare displayed. We found statistically significant differencesin adjusted cost and LOS between physiciancategories after adjusting for the effect of differencesin demographic and case-mix variables. hospitalists had16.1% lower LOS and 8.3% lower cost than generalists.

Table 4 displays mortality and readmission data forthe physician categories. Logistic regression analysis ofmortality and readmission adjusted for age, race, sex,insurance, and DRG relative weight revealed no statisticallysignificant differences between the groups for mortalityor 30-day readmissions.


We have demonstrated that a voluntary system composedof multiple private hospitalist practices can providemodest reductions in hospital costs and LOScompared with general internists in a variety of traditionalpractice styles. The LOS and cost differentials wefound (16.1% and 8.3%, respectively) were comparableto those of other recently published studies.1-3,5,7,8 Wehave also demonstrated that erosion of the reductionsin utilization can occur as the number of hospitalistsand hospital cases. Caution is, therefore, indicated incoming to early conclusions about the success of hospitalistsin any one setting.

As in all epidemiologic investigations, statisticaladjustment for inequality of groups is always a potentialproblem. By using a very large database with 100% casesampling of the most frequent DRGs of all hospitalistsand private practice general internists, selection biasesshould be minimized.

Hospital mortality and readmission rates have limitationsas quality-of-care markers. Mortality may betransferred to other settingssuch as nursing homes, andreadmissions may occur inother regional hospitals thatwould not be captured.Despite these limitations, hospitalmortality and readmissionrates were similar for the2 physicians groups and wereless than the rates found inother contemporary studies.2,3 Both physician groupshad similar rates of dischargeto home and skilled nursingfacilities.

From the Departments of Internal Medicine (GDE) and Managed Care (BKJ), OrlandoRegional Healthcare, Orlando, Fla; the Department of Healthcare Administration, MercyHospital Miami, Miami, Fla (MPA); and the Department of Statistics (NU), University ofCentral Florida, Orlando, Fla (CS).

Financial support for this study was provided by Orlando Regional Healthcare.

Address correspondence to: George D. Everett, MD, MS, Associate Program Directorof the Department of Internal Medicine, Orlando Regional Healthcare, 86 WestUnderwood Street, Orlando, FL 32806. E-mail: georgee@orhs.org.

J Gen Intern Med

1. Stein MD, Hanson S, Tammaro D, HannaL, Most AS. Economic effects of communityversus hospital-based faculty pneumoniacare. . 1998;13:774-777.

Ann Intern Med

2. Diamond HS, Goldberg E, Janosky JE.The effect of full-time faculty hospitalists onthe efficiency of care at a community teachinghospital. . 1998;129:197-203.

Arch Intern Med

3. Kearns PJ, Wang CC, Morris WJ, et al.Hospital care by hospital-based and clinic-basedfaculty: a prospective, controlled trial.. 2001;161:235-241.

Ann Intern Med

4. Meltzer D, Manning WG, Morrison J, etal. Effects of physician experience on costsand outcomes on an academic general medicineservice: results of a trial of hospitalists.. 2002;137:866-874.

AnnIntern Med

5. Auerbach AD, Wachter RM, Katz P,Showstack J, Baron RB, Goldman L. Implementationof a voluntary hospitalist serviceat a community teaching hospital: improvedclinical efficiency and patient outcomes. . 2002;137:859-865.

Am J Manag Care

6. Molinari C, Short R. Effects of an HMOhospitalist program on inpatient utilization. . 2001;7:1051-1057.

Am JMed

7. Davis KM, Koch KE, Harvey JK, Wilson R, Englert J, Gerard PD. Effects of hospitalistson cost, outcomes, and patient satisfaction in a rural health system. . 2000;108:621-626.

Ann Intern Med

8. Craig DE, Hartka L, Likosky WH, Caplan WM, Litsky P, Smithey J.Implementation of a hospitalist system in a large health maintenance organization:the Kaiser Permanente experience. . 1999;130:355-359.


9. Wachter RM, Katz P, Showstack J, Bindman AB, Goldman L. Reorganizing anacademic medical service: impact on cost, quality, patient satisfaction, and education.. 1998;279:1560-1565.

Health Inf Manag

10. Priddis D. The implementation of TRENDSTAR in NSW. .1996-1997;26:186-188.


11. Whiting JF, Martin J, Zavala E, Hanto D. The influence of clinical variables onhospital costs after orthotopic liver transplantation. . 1999;125:217-222.

Related Videos
Related Content
© 2024 MJH Life Sciences
All rights reserved.