Background: Prior studies of hospitalist services have suggestedimproved efficiency and quality of care compared with traditionalinpatient services.
Objectives: To compare outcomes of patients on a new hospitalistservice with those on traditional inpatient services and todetermine the impact of hospitalists on particular patient subgroups.
Study Design: Prospective, quasiexperimental, observational.
Methods: The study was conducted on the general medicineservice at an academic teaching hospital, staffed by hospitalistphysicians (HP) and nonhospitalist physicians (NHP), and included1706 consecutive, directly admitted patients over 1 year.
Results: The 447 HP patients and 1259 NHP patients had similarrates of in-hospital mortality (1.3% vs 2.1%, respectively; =.29) and 30-day readmission (7.8% vs 8.7%, respectively; = .55).Mean hospital length of stay (LOS) was 1 day shorter for HPpatients in unadjusted analyses (5.5 vs 6.5 days, respectively; =.009) and in multivariable analyses adjusting for clustering andpatient factors. Physician experience was not correlated ( < .2)with LOS. In stratified analyses, differences in LOS between HPand NHP patients were greater for patients residing closer to thehospital. Mean total costs were $917 less for HP patients ( = .08)and 10% less ( = .04) in multivariable analyses. Decreases in costswere significant ( < .05) for nursing ($604; = .002) and laboratoryservices ($126; = .04). Nonetheless, mean costs per daywere $122 higher ( = .003) for HP patients.
Conclusions: Patients managed by hospitalists had shorter LOSand lower costs than patients managed by nonhospitalists, but hadhigher costs per day. These results suggest that hospitalists increasethe intensity of care and may have their greatest impact on specifictypes of patients and classes of hospital costs.
(Am J Manag Care. 2004;10:561-568)
Since the term hospitalist was coined 7 years ago1to describe physicians who provide inpatient carein place of primary care physicians or academic1-month-per-year attendings, several studies have comparedoutcomes of patients managed by hospitalist andnonhospitalist physicians. These comparisons have typicallyrevealed that hospitalist care reduced length ofstay (LOS) by approximately 1 day and costs by 10% to15%.1-5 In a recent review of the growing hospitalist movement,Wachter and Goldman concluded that "researchsupports the premise that hospitalists improve inpatientefficiency without harmful effects on quality or patientsatisfaction."6 Although multiple studies have shownreductions in hospital LOS and costs, the number ofstudies and hospitals upon which the above conclusionswere based are limited. In addition, hospitals are activelyrestructuring inpatient care services including theexpansion of hospitalist services, nurse case management,and other concurrent interventions that mayaffect LOS and costs. Evidence of the impact of hospitalistson individual hospitals is still needed.
An important determination to make is the factorsthat contribute to this shorter LOS and lower costs.Prior studies have not provided insight into the mechanismsby which hospitalist services achieved thisimproved efficiency. Some have argued that "practicemakes perfect," although the relationship betweenexperience and efficiency has not been proven.Reductions in laboratory and radiology testing or druguse may be another mechanism for reductions in costsof care. In addition, the relative impact of the cost associatedwith nursing care and excess LOS has not beenwell described. Finally, patient-specific factors have notbeen evaluated for associations with outcomes andresource utilization on hospitalist services.
Herein we report the findings from an observationalstudy of the first year of a hospitalist program at an academicmedical center, and compare outcomes andresource utilization for patients managed by hospitalistand nonhospitalist physicians. Whereas the primaryaim of the study was to add to the body of evidenceabout the effect of hospitalist programs, we also soughtto provide greater detail than prior studies on the natureof cost differences and whether differences were consistent across different cost categories (eg, nursing, laboratory).We further sought to examine whether particularpatient groups were somehow more or less affected bythe hospitalist model.
The study was conducted at the University of IowaHospitals and Clinics, an 831-bed academic teachinghospital and tertiary referral center. The general internalmedicine ward service has 4 teams consisting of 1attending physician, 1 senior resident, 1 intern, and 1or 2 medical students. Residents and students are randomlyassigned to the 4 teams. All teams admit patientsto the same hospital floors, have the same nursing staff,and work with the same social service and other hospitalpersonnel.
In July 2000, the Division of General InternalMedicine implemented a hospitalist physician (HP)service for inpatient general internal medicine. Theservice was developed to meet the professional interestsof faculty. No implicit or explicit goals for resource use,clinical outcomes, or patient satisfaction were set.During the 2000-2001 academic year, 3 "hospitalist"faculty staffed 1 of the 4 general internal medicine servicesin month-long blocks and were provided with coveragefor 2 to 3 weekends per month by 12 other generalmedicine faculty in an effort to prevent burnout. Eachhospitalist faculty staffed for 3 to 6 months.
Thirty-four faculty from various divisions in internalmedicine staffed the nonhospitalist physician (NHP)services for the same year. Of the 3 NHP services, 1 wasstaffed by 7 endocrinologists; 1 by 12 nephrologists;and 1 by a combination of 2 rheumatologists, 6 infectiousdisease specialists, and 7 general internists. TheNHP services varied in their approach to weekend coverage.Some staff attended for the entire month (n = 9;26%), others split the month in half (n = 9; 26%), andall the others had variable coverage throughout themonth (n = 16; 47%). Of the 34 nonhospitalist faculty,1 spent 3 months on the service, 5 others spent a rangeof 5 to 8 weeks, and the remainder 1 month or less onthe service.
The eligible sample included all 1887 consecutivedischarges between July 1, 2000, and June 30, 2001,from the 4 general internal medicine services. Allpatients were sequentially admitted in a quasiexperimentalmanner to one of the teaching teams in an alternatingmanner without regard to diagnosis or to complexity.The teams took "long-call" every fourth day,during which teams admitted up to 10 patientsovernight. In addition, on weekdays teams took up to 4admissions during the day on "short-call," whichoccurred 2 days before and after long-call. All admissionsto general internal medicine from university- andcommunity-based physicians were assigned to the 4teaching services, including admissions from primarycare physicians. During the study period, there was no"nonteaching" service. Patients requiring an intensivecare unit (ICU) bed were admitted to a closed medicalor cardiovascular ICU. Separate cardiology and hematology/oncology teams also existed.
Of the 1887 consecutive admissions, 1706 (90%)patients were admitted directly to the general internalmedicine service (ie, direct admissions) from either theemergency room (n = 801) or from university clinics,referring physicians' offices, or other acute care hospitals(n = 905). The remaining 181 (10%) admissionswere transferred to general internal medicine (ie, transferadmissions) from the medical ICU (n = 105) or fromsurgical or other nonmedicine services (n = 76).Because costs incurred by transfer admissions prior totheir transfer to general internal medicine could not beseparated from costs after the transfer, our primaryanalyses were conducted in the 1706 direct admissions.
All study data, including outcome variables andmeasures of resource utilization, were obtained fromthe hospital's information systems. The principal outcomevariables were hospital mortality and 30-day readmissionrate. Resource utilization was measured usinghospital LOS and hospital costs. Because of the potentialbias introduced by outlier LOS values, all patientswith a LOS of longer than 60 days were truncated at 60days. This truncation included 1 patient in the hospitalistgroup and 5 in the nonhospitalist group.
Hospital costs were measured using the TSI costaccounting system (Transition Systems, Inc, Boston,Mass). This widely used methodology determines thefixed and variable costs of all billable hospital services.This system also allocates to these services indirectcosts associated with nonbillable services giving a totalcost by category (eg, nursing, physicians, medications)and a total cost for the hospital stay. Separate analyseswere performed on nursing, laboratory, pharmacy, andradiology services, as these cost categories representedthe top 4 categories in total cost and together more than70% of all inpatient costs in our study sample. Totalcosts were reported in the analyses because of the complexitiesof determining direct and indirect costs andwhat proportion may be attributed as fixed and variable.Total costs are more generalizable to other hospitalsand systems.
International Classificationof Diseases, Ninth Revision, Clinical Modification
Demographic and clinical variables available includedage, sex, race, health insurance status, admissionsource (eg, home, nursing home), discharge destination(ie, home, home healthcare, other hospital, nursinghome, died, against medical advice, or other care facility),admission and discharge date, discharge diagnosis-relatedgroup (DRG), and principal diagnosis (asmeasured by taxonomy of the [ICD-9-CM] In addition, for each patient, distance fromthe hospital was determined using a function thatmathematically estimates the "straight line" distancebetween 2 coordinates (longitude and latitude).7Coordinates used in the calculation included the longitudeand latitude of the hospital and the centroid of thezip code of a patient's residence.
The primary analyses compared outcomes ofpatients discharged from the HP service with outcomesof patients discharged from the 3 NHP services.Categorical variables were analyzed using the chi-squarestatistic; continuous variables were analyzedusing the Student test or Wilcoxon rank-sum, dependingon whether the data were normally distributed.Multiple linear regression was used to evaluate differencesin continuous outcome variables (ie, LOS, costs)and logistic regression for dichotomous variables (ie,death, 30-day readmission). To account for the skeweddistribution of costs and LOS, we used generalized linearmodels, assuming that the effects of the covariateswere proportional using a logarithmic link function.8 Tocontrol for potential differences in patient characteristics,the following independent variables were includedin the regression models: age, sex, type of health insurance,and admission month. Because of the potentialeffect of physician-level clustering9 on observed differences,we performed analyses using the PROC GENMODfunction with SAS software, which accounts forclustering by physician. Additional multivariable analysesof costs and LOS were performed that adjusted forprincipal diagnosis using the technique of absorption inthe PROC GLM function of SAS.10 This technique iscomputationally similar to creating indicator variablesfor individual ICD-9-CM codes, but does not allow forthe concurrent use of techniques to account for clustering.Results of these analyses yielded similar estimatesof the coefficients associated with hospitalist care, andwere not reported. No further case-mix adjustment wasperformed due to the quasirandom allocation ofpatients and the similar distribution of diagnoses basedon DRGs. No other indicators of differences in case-mixwere observed that were considered to potentially biasthe results.
Further stratified analyses were performed to determinewhether the impact of hospitalist care was similarin subgroups defined by distance from the hospital,need for postacute care nursing services, and admissionsource. Need for postacute care nursing services includedhome healthcare, intermediate nursing care, orskilled nursing care. Specific factors examined in theseanalyses were selected on the basis of discussions withhospitalists. Lastly, relationships between LOS and severalphysician characteristics, including the number ofyears since completion of residency and the number ofdays on the inpatient service during the academic year,were examined using the Pearson correlation coefficient.All analyses were performed using SAS forWindows, version 8.0 (SAS Institute, Cary, NC).
Patients admitted to HP and NHP services were nearlyidentical in mean age, sex, race, type of health insurance,admission source, discharge destination, anddistance from the home to hospital. (Table 1). Proportionsof patients in each group were similar for 9 of the10 most common DRGs (Table 2).
In-hospital mortality was similar for direct admissionpatients on HP and NHP services (1.3% vs 2.1%, respectively; = .29; Table 3), as were 30-day readmissionrates (7.8% vs 8.7%, respectively; = .55). Results weresimilar in analyses of transfer patients and in analysesof all admissions (ie, direct plus transfer admissions).Mean hospital LOS was roughly 1 day shorter inpatients on the hospitalist service for direct admissions(5.5 vs 6.5 days, respectively; = .009). In contrast,relationships between LOS and the number of yearssince attending physicians' completed residency ( =0.10; = .53) and number of days on service during theacademic year ( = −0.01; = .95) were not significant(data not shown).
Mean total costs for direct admissions were $917(11%) lower for patients in the HP group, although thisdifference was only of borderline statistical significance( = .08). The relative decline in LOS was greater thanthe relative declines in costs, such that the mean cost perhospital day was $122 higher in the HP group ( = .003).
In addition to total costs, nursing, laboratory, pharmacy,and radiology costs are reported for the HP andNHP groups in Table 3. For direct admissions, meannursing costs were $604 less for the HP patients ( =.002) and laboratory costs were $126 less ( = .04).Differences in pharmacy and radiology costs were notsignificant.
Additional analyses were performed on the entirecohort of "all admissions" (n = 1887) and "transfers" (n= 181), as reported in Table 3. The difference in LOS,nursing costs, and costs per day persisted in the "alladmissions" cohort, but no significant differences wereobserved among the "transfer" patients.
In multivariate analyses, adjusting for age, sex,admission month, and type of insurance, the odds ofdeath for patients on the HP service were similar to thatfor patients on NHP services (odds ratio [OR] 0.65, 95%confidence interval [CI] 0.26 to 1.63; = .37). AdjustedLOS was 16.2% lower (95% CI —31.2% to —1.6%; = .03)for patients on the HP service,while adjusted total costs were9.7% lower (95% CI —19.3% to—0.5%; = .04) (data notshown).
Exploration of Associationswith Length of Stay
In stratified analyses, differencesin LOS between patientson HP and NHP services tendedto be largest for patients residing25 miles or less from thehospital (Table 4). Absolutedifferences in LOS were alsogreater among patients dischargedwith home nursingcare or to skilled nursing facilities(1.7 days) than those whodid not require nursing services(0.8 days). Absolute differencesin LOS between HP andNHP patients were also similarin patients admitted fromhome (1.0 days) and fromother acute care hospitals (1.2days), although only the differencesin patients admittedfrom home were statisticallysignificant. In contrast, LOSwas nearly identical in HP andNHP patients among patientsadmitted from other sources.
During the first year of anacademic hospitalist program at a large university teachinghospital, mean LOS was 1 day (16%) shorter amongpatients on the hospitalist service. Based on theseresults, we could extrapolate that the hospitalist serviceresulted in approximately 450 fewer days of care duringfiscal 2001, compared with the nonhospitalist services,before and after adjustment for potential differences incase-mix.
Similarly, mean costs were approximately $900lower for patients on the hospitalist service. Roughlytwo thirds of the cost differences were attributable toreductions in nursing costs. Cost savings were alsoobserved for laboratory service, although differences inpharmacy and radiology services were not significant.The lower nursing cost was likely driven primarily bythe shorter hospital LOS. Lower laboratory costs mighthave been due to more prudent use of laboratory testing,or simply a function of the shorter LOS. The lack ofdifference in pharmacy and radiology costs may be thatthese services are less discretionary for inpatients andless dependent on physician practice style.
Interestingly, the mean cost per day of care on theHP service was $122 more than on the NHP service.This result may suggest that the greater availability ofhospitalists or their style of care may lead to quicker ormore intense evaluation of patients.
The cost data suggest that our hospitalist serviceresulted in savings to our institution of more than$370 000 ($835 per patient Ã— 447 patients) during thefirst year of its implementation. The results also suggestthat if patients managed by nonhospitalist physicianswere cared for with the same efficiency as the hospitalistphysicians, our hospital could have appreciated anadditional savings of more than $1 million. This reductionin cost on the HP service was attained with similarquality of care, as measured by hospital mortality andreadmission rate. Of note, no additional resources wereexpended in the development of this hospitalist program.Although dedicated hospitalists were hired, thisendeavor was part of the expansion of the GeneralInternal Medicine Division and allowed other departmentfaculty to spend less time on the inpatient service.In addition, no additional support staff were hired forthe hospitalist program.
The near-random allocation of patients allowed forthis retrospective analysis to mimic a prospectivelydesigned clinical trial.In some ways, ourobservational analysismay be superior to aprospectively designedtrial because none ofthe physicians knewthey were being measuredor compared.Therefore our resultsmay be more reflectiveof the potential effectivenessof a hospitalistservice rather than theefficacy found in a randomizedcontrolledtrial. Of further importanceis that our hospitalistphysicians werenot directly chargedwith reducing resourceutilization. Had thisobjective been a stated goal, the differences may havebeen greater.
The significant findings were limited to the "direct"admissions, excluding the 10% of "transfer" patients.This result may be due to the fact that patients transferredin to the general medicine service received a largepercentage of their care in another setting (ie, surgeryservice, intensive care unit) in which the general medicineteam was not involved in their care, and thereforecould not affect outcomes or resource utilization.
Our findings of an approximately 16% reduction inLOS and 10% reduction in hospital costs are consistentwith the results of other prior studies of the impact ofhospitalists in other academic medical centers. In aseminal study by Wachter et al,11 the first year of a hospitalistservice resulted in a 12% reduction in hospitalLOS and a 10% reduction in hospital costs. Mortality,readmission rates, and satisfaction of patients, residents,and students were unaffected. More recently,Meltzer et al5 evaluated an academic hospitalist serviceover 2 years and found no difference in HP and NHPservices in the first year, but by the second year founda 0.5-day shorter LOS (10%) and $740 lower cost (8%)( < .01). Similar results were found by Auerbach andcolleagues4 in a community hospital where hospitalistshad a 0.61-day shorter LOS and $822 less cost of carein the second year of the program. In their recently publishedreview of studies of hospitalist services, Wachterand Goldman6 reported that of 19 studies that reportedhospital cost and LOS, they found an average decrease incost of 12.4% and average decrease in LOS of 16.6%.
Our findings for in-hospital mortality and 30-dayreadmission rates are similar to those in previously publishedstudies.3,11-15 Two studies have shown a decreasein mortality in hospitalistgroups,4,5 whereas11 have shown no difference.6 Similarly, of12 studies reporting30-day readmissionrate, 1 showed adecrease in readmissionrate,2 1 anincrease,16 and theother 10 showed nodifference.3-5,11-15,17,18In-hospital mortality and 30-day readmissionrates were lower in our hospitalist cohort,but the differences were not significant, and thedata were likely not powered to detect a difference,due to the low frequency of these events.
Our analyses may also provide furtherinsights into factors underlying the greaterefficiency of hospitalists.We observed nocorrelation betweenLOS and measures ofphysician experience,which does notappear to support thehypothesis that "practicemakes perfect" orthat simply havingmore time on serviceimproves efficiency,per se. What may bemore noteworthy is thefact that hospitalistsare self-selected tofocus their careers oninpatient medicine andmay be more satisfiedin this role and moreinvested in working collaborativelywith nursing andother personnel.
In addition, we found that LOS differences tended tobe greater among patients who resided closer to the hospital.This finding may indicate that the effect of hospitalists is greater in settings inwhich it is easier to ensure follow-up outpatient care afterdischarge. Although our a priori hypothesis was that hospitalistswould have a greater effect on patients whorequired nursing care after discharge because of greater easein working with nurses and social workers, relative differencesin LOS were similar among patients who were dischargedwith and without nursing care. It is possible thatany greater effect on the patients discharged to nursingcare was obscured by general delays in waiting for skillednursing beds for patients with certain conditions that madeplacement difficult.
In interpreting our results itis important to consider several potential limitations. First,the study was limited to a single academic medical centerand may not be generalizable to all medical centersand patient populations. Second, although our analysisaccounted for clustering of patients among physiciansand variation across physicians, the study examinedonly 3 hospitalist physicians. However, our analysiswas conducted during the first year of our program.Based on prior studies, cost savings associated withhospitalist programs often increase over time.5 Thus,the differences we observed may increase as our hospitalistsgain more experience and familiarity with hospitalpersonnel. Another possibility is that themanagement practice of hospitalists that lead toimproved efficiency will diffuse to nonhospitalist servicesthrough interactions with house staff and nonhospitalists.Third, although we believe that our patientallocation approximated randomization, we nonethelesscannot exclude unmeasured selection bias inpatients admitted to the hospitalist and nonhospitalistservices. Fourth, the hospitalists received consistentweekend coverage for 2 to 3 weekends per month,whereas the nonhospitalists had inconsistent coveragefor weekends, and many attended for less than 1-month blocks of time. This inconsistent coverage onthe nonhospitalist teams may have lead to inefficienciesof care. Also, patients admitted on weekends thathospitalists had coverage were counted as hospitalistpatients, in case the different case-mix admitted onweekends biased outcomes. Fifth, 27 of the 34 nonhospitalistphysicians were subspecialists. It is unknownwhether this would bias the results in favor of subspecialistsbecause they had more formal training, oragainst them because they may not have the samebreadth of experience as a general internist. Finally, wewere not able to adjust for severity of illness usingadministrative data, but due to the quasirandom allocationof patients and similar case-mix based on DRG,there should be no systematic bias against 1 groupbased on severity of illness. Future studies are neededto better understand these and other limitations.
In sum, our findings add to the growing body of literaturethat support the increased efficiency of inpatientcare delivered by hospitalists and that suchimprovement in efficiency may be observed during thefirst year of a hospitalist program with newly recruitedfaculty. Our findings also indicate that hospitalists mayincrease the intensity of care by focusing similar evaluationand treatment into a shorter LOS. Moreover, tothe extent that hospitalists have a greater impact inpatients with particular characteristics (eg, patientswho reside closer to the hospital), our findings provideinsight into possible explanatory factors for thegreater efficiency of hospitalists. Conversely, thesefindings may help identify subgroups in which theopportunities are greatest for further improving theefficiency of hospital care.
From the Research Service, Iowa City Veterans Affairs Medical Center, Iowa City, Iowa(PJK, MJB, GER); and the Division of General Internal Medicine, Department of InternalMedicine, University of Iowa College of Medicine and University of Iowa Hospital andClinics, Iowa City, Iowa (PJK, GER).
Dr Kaboli is supported, in part, by a Research Associate Award, Health ServicesResearch and Development Service, Department of Veterans Affairs.This work was presented at the 25th Annual Meeting of the Society of General InternalMedicine, Atlanta, Georgia, May 2-4, 2002.
Address correspondence to: Peter J. Kaboli, MD, MS, Division of General InternalMedicine, University of Iowa Hospitals and Clinics SE615GH, 200 Hawkins Drive, IowaCity, IA 52242. E-mail: email@example.com.
N Engl J Med
1. Wachter RM, Goldman L. The emerging role of "hospitalists" in the Americanhealth care system. . 1996;335:514-517.
Ann Intern Med
2. Diamond HS, Goldberg E, Janosky JE. The effect of full-time faculty hospitalistson the efficiency of care at a community teaching hospital. .1998;129:197-203.
3. Palmer HC, Armistead NS, Elnicki DM, et al. The effect of a hospitalist servicewith nurse discharge planner on patient care in an academic teaching hospital. . 2001;111:627-632.
Ann Intern Med
4. Auerbach AD, Wachter RM, Katz P, Showstack J, Baron RB, Goldman L.Implementation of a voluntary hospitalist service at a community teaching hospital:improved clinical efficiency and patient outcomes. . 2002;137:859-865.
Ann Intern Med
5. Meltzer D, Manning WG, Morrison J, et al. Effects of physician experience oncosts and outcomes on an academic general medicine service: results of a trial ofhospitalists. . 2002;137:866-874.
6. Wachter RM, Goldman L. The hospitalist movement 5 years later. .2002;287:487-494.
SAS Notes SN-005325: Calculating the Distance Between ZIP Codes
7. . Cary, NC:SAS Institute Inc. http://support.sas.com/techsup/unotes/SN/005/005325.html.Accessed July 22, 2004.
J Health Econ
8. Manning WG, Mullahy J. Estimating log models: to transform or not to transform?. 2001;20:461-494.
Ann Intern Med
9. Greenfield S, Kaplan SH, Kahn R, Ninomiya J, Griffith JL. Profiling care providedby different groups of physicians: effects of patient case-mix (bias) and physician-level clustering on quality assessment results. . 2002;136:111-121.
The GLM procedure. SAS/STAT User's Guide
10. SAS. . Cary, NC: SAS Institute Inc;1999:1481, 1532-1535.
11. 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.
J Gen Intern Med
12. Stein MD, Hanson S, Tammaro D, Hanna L, Most AS. Economic effects ofcommunity versus hospital-based faculty pneumonia care. .1998;13:774-777.
13. 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.
Am J Manag Care
14. Halpert AP, Pearson SD, LeWine HE, McKean SC. The impact of an inpatientphysician program on quality, utilization, and satisfaction. .2000;6:549-555.
15. Hackner D, Tu G, Braunstein GD, Ault M, Weingarten S, Mohsenifar Z. Thevalue of a hospitalist service: efficient care for the aging population? .2001;119:580-589.
16. Bellet PS, Whitaker RC. Evaluation of a pediatric hospitalist service: impact onlength of stay and hospital charges. . 2000;105:478-484.
Arch Intern Med
17. Kearns PJ, Wang CC, Morris WJ, et al. Hospital care by hospital-based andclinic-based faculty: a prospective, controlled trial. .2001;161:235-241.
18. Landrigan C, Srivastava R, Muret-Wagstaff S, et al. Outcomes of hospitalizationin pediatric patients insured by HMOs: comparison of care by hospitalists andtraditional academic providers [abstract]. . 2000;47:204A.