OBJECTIVE. This study examined mechanisms of efficiency in a managed care hospitalist system on length of stay and total costs for common pediatric conditions.
PATIENTS AND METHODS. We conducted a retrospective cohort study (October 1993 to July 1998) of patients in a not-for-profit staff model (HMO 1) and a non–staff-model (HMO 2) managed care organization at a freestanding children's hospital. HMO 1 introduced a hospitalist system for patients in October 1996. Patients were included if they had 1 of 3 common diagnoses: asthma, dehydration, or viral illness. Linear regression models examining length-of-stay–specific costs for prehospitalist and posthospitalist systems were built. Distribution of length of stay for each diagnosis before and after the system change in both study groups was calculated. Interrupted time series analysis tested whether changes in the trends of length of stay and total costs occurred after implementation of the hospitalist system by HMO1 (HMO 2 as comparison group) for all 3 diagnoses combined.
RESULTS. A total of 1970 patients with 1 of the 3 study conditions were cared for in HMO 1, and 1001 in HMO 2. After the hospitalist system was introduced in HMO 1, length of stay was reduced by 0.23 days (13%) for asthma and 0.19 days (11%) for dehydration; there was no difference for patients with viral illness. The largest relative reduction in length of stay occurred in patients with a shorter length of stay whose hospitalizations were reduced from 2 days to 1 day. This shift resulted in an average cost-per-case reduction of $105.51 (9.3%) for patients with asthma and $86.22 (7.8%) for patients with dehydration. During the same period, length of stay and total cost rose in HMO 2.
CONCLUSIONS. Introduction of a hospitalist system in one health maintenance organization resulted in earlier discharges and reduced costs for children with asthma and dehydration compared with another one, with the largest reductions occurring in reducing some 2-day hospitalizations to 1 day. These findings suggest that hospitalists can increase efficiency and reduce costs for children with common pediatric conditions.
Hospitalist systems have been shown to reduce length of stay (LOS) and cost with no decrease in quality (usually measured by mortality and readmission rate) for children and adults in both community and academic environments.1,2 We demonstrated previously that a managed care pediatric hospitalist system reduced LOS and total costs when all diagnoses were included.3
However, little research has been conducted to understand which types of patients are better cared for in a hospitalist system and why. The majority of hospitalist studies to date have been conducted in tertiary care centers or teaching hospitals that include patients with complex conditions. This raises concern about generalizability to common conditions that predominate in nonteaching community hospitals, where about half of children's admissions occur.4 Most studies have focused on all diagnoses combined, with few examining specific diagnoses or groups of patients. Three studies (only 1 pediatric) that included analysis of the 10 or 11 most frequent diagnoses reported hospitalist reductions in LOS and cost for some or for overall diagnoses.5–7 With respect to specific diseases, 1 pediatric study found that hospitalists' patients with bronchiolitis experienced reductions in LOS, but this may have been attributable to the introduction of a clinical practice guideline coinciding with the hospitalist system.8 Another pediatric study found a reduction in LOS and costs for patients with asthma but not for other common conditions.9 In contrast, a pediatric study focused on patients with asthma and bronchiolitis found no differences in LOS or costs after a hospitalist system was introduced. The study reported that both the hospitalist and control group adhered to clinical practice guidelines, which may have minimized any differences.10 An adult study focusing on patients with community-acquired pneumonia found a lower LOS in the hospitalist group.11 Regarding specific demographic groups, 1 adult study found that a hospitalist system was more efficient for patients over the age of 65, compared with those under the age of 65, because of less frequent subspecialty consultation.12 Another adult hospitalist study demonstrated greater cost savings for routine admissions than for those patients admitted with severe illness.13 The potential for pediatric hospitalist systems to reduce LOS and costs for common pediatric conditions has not been sufficiently studied.
The mechanisms by which hospitalist systems reduce hospital costs or charges by an average of 14% are uncertain.2,14 LOS reductions may be driven by hospitalist, patient diagnostic group, hospitalist microsystem, or institutional factors. Relationships between reductions in LOS and cost savings are complex and depend in large measure on the economic environment and reimbursement arrangements under which individual hospitals operate. The impact on cost savings may vary depending on whether the LOS reduction occurs primarily among patients with the shortest stays (eg, 1 or 2 days) or those with longer stays (eg, 4 or 5 days) because of differential reimbursement formulas. Ways that hospitalists reduce LOS for common, short-stay conditions may be different from uncommon or longer-stay conditions, given the medical complexity of children in the latter group.
The specific aim of this study was to examine the impact of a managed care hospitalist system on LOS and total costs for each of 3 common pediatric conditions compared with a traditional nonhospitalist system. In addition, we sought to identify the specific foci of cost savings and LOS reductions.
After institutional review board approval by both the study hospital and a staff-model, not-for-profit managed care system (HMO 1), we conducted a retrospective cohort study (October 1993–July 1998) of all pediatric patients (0–17 years of age) admitted to the study hospital who received primary care from any of the 11 HMO 1 health centers (medical offices operated by the health maintenance organization [HMO] to provide outpatient services to its enrollees) that admitted patients exclusively to the study hospital. Three health centers that primarily admitted to other hospitals were excluded to avoid possible selection bias. The inclusion criterion for this study was admission with any 1 of the following 3 conditions as specified by all patient diagnosis-related groups (AP-DRGs, NY8 version): asthma, dehydration, or viral illness (Table 1). The AP-DRG classification system includes severity of illness and has been used for case-mix trending, utilization management, and improved patient identification for payment related issues in pediatrics.15–21
Beginning in October 1996, all HMO 1 patients admitted to the study hospital were cared for by a newly introduced pediatric hospitalist system. Throughout the study period, HMO 2 (our comparison group: a non–staff-model, not-for-profit HMO) cared for hospitalized patients by using a traditional rotating attending or private practice attending system without hospitalists.3 Planning and implementation of the hospitalist program took place over a 10-month period. In January 1996, a pediatric leadership group at HMO 1 met to discuss challenges in the existing traditional rounding system (eg, timeliness of access to attending physician for inpatients, amount of time required for pediatricians to travel to and from the hospital, and outpatient revenue lost during this time). In response to these issues, the possibility of establishing a hospitalist system for all HMO 1 patients was discussed. In March 1996, a survey was administered to all HMO 1 pediatricians to assess their views on establishing a hospitalist system at the study hospital. Over the remaining 7 months, plans were finalized, and hospitalists took over the care of all HMO 1 inpatients in October 1996. The hospitalists in HMO 1 did have access to outpatient information via electronic data systems. However, these data were previously available to any inpatient provider caring for a patient at the study hospital.
HMO 1 admits 700 to 1000 general pediatrics patients per year, and typically has a census of 10 to 20 patients (∼50% of these were 1 of the 3 common conditions in this study). Since October 1996, the patient census has been divided among 2 or 3 hospitalists, who spend 0.5 to 1.0 full-time equivalencies caring for inpatients. Housestaff and medical students are involved in the care of these patients and are supervised by the hospitalists. The nature of the relationship between staff physicians from HMO 1 and trainees did not change with the advent of the hospitalist system and was fundamentally similar to the relationship between trainees and providers from other HMOs and private practices. The use of clinical practice guidelines has been high for many years at the study hospital and is applied to all patients, regardless of attending type. On weekends, a pediatrician from the hospitalist group or from 1 of the local health centers cares for the patients. All of the hospitalists hired by HMO 1 were experienced general pediatricians from the community. Individuals in the group working 1.0 full-time equivalencies as hospitalists spend the remainder of their time on administrative committees. The hospitalist group employs nurse liaisons who aid in the care and discharge planning of all patients. There was little turnover in the hospitalist staff between 1993 and 1998; 2 of 3 hospitalists worked throughout the entire study period.
For all study and comparison group patients, we extracted demographic, case-mix, financial, and clinical data from the hospital database. To validate insurance status, HMO 1 patients who were cared for in any of the 11 HMO 1 centers were identified through the hospital's computerized database and cross-referenced with patient lists maintained by their insurer. Resource utilization outcomes included LOS and total inflation-adjusted direct costs. Costs were determined by using the hospital's utilization-based cost-allocation system, which controls for relative shifts in charges over time and for differences in negotiated payment scales by different payers. Total direct costs were adjusted for inflation using the Medical Care Index of the Consumer Price Index.22 Costs are reported in 1993 dollars.
We analyzed data for each of the 3 diagnoses using time (prepost hospitalist system) and the interaction between time and study group (our primary variable of interest) as predictors in diagnosis-specific, patient-level linear regression models to test the significance of differences between study and comparison groups in the prepost changes in LOS and costs.23 These models also included terms to adjust for disease severity and patient demographics. Age was classified in 5 groups; 0 to 2, 3 to 5, 6 to 10, 11 to14, and 15 to 18 years of age.24 Race was coded as white, black, or other. Severity weights attached to the diagnosis of each AP-DRG were used in the analyses. We logged the values of total cost and LOS in all multivariate regression analyses to normalize their distributions; for clarity of interpretation, results reported in the article are reconverted to normal scale.
Next, by using the adjusted LOS data from these regression models, LOS-specific costs for prehospitalist and posthospitalist system were calculated and compared in the 2 HMOs to identify which durations of hospitalization (with each patient categorized according to number of days of hospitalization) were associated with the greatest savings. For diagnoses for which the prepost changes in LOS in the hospitalist system were significantly different from the comparison group, we calculated the changes in average cost per case in the hospitalist system for admissions during that period. Average cost per case for each LOS category was calculated by using adjusted data from regression models in the prehospitalist period (eg, a patient with asthma who stayed 1 day cost the hospital $592, a patient with asthma who stayed 2 days cost the hospital $978). These LOS-specific costs were then assigned to the number and proportion of patients hospitalized in each period, and average daily costs for the sample as a whole were calculated by summing across LOS categories.
Finally, we used time-series analyses to examine trends in LOS for HMO 1 and HMO 2 for the total sample of patients for the 3 diagnoses combined.3
A total of 1970 patients with 1 of the 3 study conditions were cared for in HMO 1 and 1001 patients were cared for in HMO 2 during the study period. These accounted for 54% and 52% of total pediatric admissions for HMO 1 and HMO 2 to the hospitals' general medical services, respectively. Among patients admitted with all diagnoses during the study period, HMO 1 patients were more likely to be nonwhite and had somewhat more serious illness than those in the HMO 2 group (Table 2). The HMO 1 group included proportionately more patients with a diagnosis of asthma (32.1% vs 25%; P < .001) and fewer patients with a diagnosis of dehydration (12.8% vs 16.7%; P < .001).
After the hospitalist system was introduced in HMO 1, LOS was significantly reduced compared with HMO 2 for asthma and dehydration (Fig 1). LOS was reduced by 0.23 days for asthma and 0.19 days for dehydration in HMO 1, compared with increases of 0.13 and 0.26 days in HMO 2 during the same time period. No significant LOS differences were detected for patients with viral illness.
The proportions of hospitalizations that exceeded 1 day declined in HMO 1 for patients with dehydration and asthma after the hospitalist system was implemented (see Table 3). The largest relative reduction in LOS was a shift from 2-day to 1-day hospitalizations for both diagnoses. For patients with asthma, the average cost per case declined from $1131 to $1025, an estimated cost savings of $106 (9.3%) per case. For patients with dehydration, the average cost per case in the prehospitalist period was $1110 compared with $1024 after the system was implemented, an estimated cost savings of $86 (7.8%) per case.
When the 3 conditions were combined, mean LOS for HMO 1 was 2.36 days before the introduction of the hospitalist system, and mean cost per admission was $1235. After the introduction of the hospitalist system, time-series analyses indicated that there were no significant changes in either level or trend in these parameters (P ≥ .15 for all estimates). Mean LOS for HMO 2, which was 2.35 days before October 1996 when the hospitalist system was introduced in HMO 1, also exhibited no significant changes in level (P = .93) or trend (P = .37) after that date. In contrast, mean per admission costs for HMO 2, which were $1320 before October 1996, showed a borderline significant increase in trend of $56 per quarter (P = .06) after that date.
For 2 of the most common causes of hospitalization of children, asthma and dehydration, we found reductions in LOS and average cost per case in a tertiary children's hospital after the introduction of a hospitalist system. The greatest impact occurred among children who, under the care of a hospitalist, were hospitalized for 1 rather than 2 days for these conditions. No change was observed for the category of children admitted for viral illnesses. During the same time period in the same hospital, LOS and costs for asthma and dehydration increased for a nonhospitalist comparison HMO. Time-series analysis demonstrated no significant overall differences in LOS or costs for the 3 conditions combined, but differences for asthma and dehydration may have been masked by inclusion of patients with viral syndrome, for whom there was no difference.
The estimated hospital net savings calculated in 1993 dollars were approximately $41446 and $18748 per year for study group patients with asthma and dehydration, respectively. This is a conservative estimate of savings for several reasons. First, it does not take into account the increases in hospital costs that were noted in HMO 2 during the same time period. Second, we were unable to include possible differences in resource utilization (radiology, laboratory test, and pharmacy). Several studies have suggested a lower use of such resources and fewer consultations requested by hospitalists.3,9,25 Third, we did not calculate or account for the increased capacity for throughput related to shortened LOS, a potential source of substantial revenue benefit.26
The cost reduction of 9.3% and 7.8% for patients with asthma and dehydration is modest on a per case basis (not accounting for the increase in cost by the comparison group during the same time period), but because a large number of patients are admitted each year with these diagnoses, the total cost savings generated are substantial. The value of these cost savings from a hospitalist system depends on the perspective of the involved parties, the reimbursement methods of the payers to the hospital, and the market forces. These perspectives include the payer, the hospital, the patient, the physicians (hospitalist group), and the health care system as a whole.27 For the purposes of this article, we have restricted our discussion primarily to the perspective of the hospital.
Two reimbursement systems28,29 were prevalent in our study hospital: per-diem charges and discounted billed charges (1 form of this is the case rate from Medicaid, which remained constant during the study period at 15% of total reimbursement to the hospital from payers). Diagnosis Related Groups (DRG) reimbursement has the most potential for favoring an efficient hospitalist system from the hospital's perspective; however, this did not apply to our study population of HMO 1 patients.30 At the study hospital, there was a shift in the reimbursement systems for HMO 1 patients over the study period from 1993 through 1998. In 1993, the reimbursement system was mostly discounted billed charges. In early 1995, contractual relationships were renegotiated between the hospital and the HMO 1, and HMO 1 moved to a per-diem system. HMO 2 (comparison group) remained on an unchanged contractual DRG system throughout the study period (D. Kirschner, MBA, CPA, verbal communication, 2004).28,31
In the discounted billed charges system of reimbursement, the impact of an efficient hospitalist system had the potential effect of an overall reduction in total revenue to the hospital (by reducing the total charges to the payer), leading to an advantage to the payer and a disadvantage to the hospital. This reduction in total revenue may have been partially offset by an increase in intensity of care (the number of services delivered per period of time).
In 1995, the impact of a change for the HMO 1 patients to a per-diem payment system to the hospital may have led to an increase in reimbursement for the hospital after the introduction of the managed care hospitalist system. This may have occurred for several reasons. First, the economic situation changed to a high demand (patients waiting for hospital beds) and a low supply (hospital beds) market. Second, the efficient managed care hospitalist system was able to reduce LOS by 0.23 days for asthma and 0.19 days for dehydration (accounting for ∼30% of their inpatient volume). Third, there was an increase in the opportunity to gain from increased throughput, thereby increasing total revenues for the hospital.32
The primary reason for the creation of the hospitalist system was to improve the quality of care delivered to children admitted to the inpatient pediatric service by relieving the majority of physicians of their inpatient responsibilities and creating a core group of dedicated inpatient physicians, easily accessed by housestaff. As a secondary goal, the efficiencies achieved were hoped to be sufficient to pay for the program although this was found not be true, given specifics of the reimbursement system (J. Graef, MD, verbal communication, 2002).
In our previous time-series study, we found that introduction of this hospitalist system led to reductions in LOS and costs overall in a large tertiary care center, including children with complex conditions.3 In considering only the subgroup of patients with asthma, dehydration, and viral illness in this study, we also found significant differences in LOS or costs with 2 of the 3 conditions. These combined findings suggest that hospitalist system effects on LOS and costs were distributed among children with both complex and common conditions.
These findings address 1 of several key questions dominating the current hospitalist debate: How ironclad is the case that the hospitalist model improves quality and efficiency?33 Although this study was not designed to assess quality improvements, it demonstrates that a hospitalist model can significantly increase efficiency for 2 high-frequency diagnoses in children, asthma and dehydration, particularly for the more common short stays. This opens the opportunity to focus efforts to understand, enhance, and replicate the mechanisms of hospitalist efficiency through condition-specific process and outcome studies. Hospitalist systems may also provide substantial noneconomic benefits, which have been summarized recently in the literature.14,34
Numerous mechanisms have been postulated to explain hospitalist efficiency, but findings are inconsistent. These include differences at the levels of hospitalist and patient characteristics, various hospitalist models and teams, and hospital systems. Although “practice makes perfect” has been suggested, and beneficial volume–outcome relationships for physicians and hospitals frequently hold true in health care, the magnitude of association varies widely, and effects often are modest.35 Multilevel modeling and other rigorous methodologic approaches can strengthen hospitalist research in this regard.36 A new generation of studies is needed to advance from “whether” hospitalist systems work to “for whom” and “how” hospitalist systems work.14 Only then will be we able to meet the design principles and targets for hospitalist systems recently recommended by pediatric leaders.37–39
As in all studies, our findings are limited in generalizability to the types of hospitals and systems in which the study was conducted. However, the discovery of efficiencies of care specific to common diagnoses increases the likelihood that hospitalist systems can improve care not only in tertiary centers but also in community hospitals where these diagnoses predominate. This study's limitation in estimating cost savings noted above yields a conservative estimate of efficiency and lends additional strength to the overall finding. The lack of change noted for the children with viral illnesses may not be applicable to other hospitals. In the study hospital, this AP-DRG is comprised of disparate groups of patients with croup, viral meningitis, and fever of unknown origin, and proportions of children with these varying conditions may differ in other institutions. The comparison group HMO 2 was not identical to HMO 1 in all aspects; however, it was the closest analog to HMO 1 that existed at the study hospital. Both managed care organizations were large, not-for-profit HMOs that had a large number of patients admitted to the study hospital each year. Both have been consistently ranked highly in national studies of HMO quality. Both have comprehensive preventive and hospital care plans. A causal relationship between the divergence in costs and LOS for patients with asthma and dehydration after October 1996 and the introduction of hospitalists is an implied one, and not proven. However, no other significant changes occurred in the care of inpatients from either of these 2 HMOs during the intervention period.
The introduction of a hospitalist system in 1 HMO was associated with earlier discharges for children with asthma and dehydration, with the greatest reductions related to changing 2-day into 1-day hospitalizations. This resulted in significant estimated cost savings for these 2 diagnoses at a time when costs for the comparison group were increasing. This new finding of hospitalist efficiencies in care of children with common diagnoses extends our understanding of the scope of hospitalist impact. The results allow us to focus efforts in uncovering mechanisms that can inform decision-making by clinicians and administrators and the design of emerging pediatric hospitalist models.
This work was supported, in part, by grants from the Health Resources and Services Administration (T32 PE10018), Rockville, Maryland, to the Harvard Pediatric Health Services Research Fellowship Program, and from the Harvard Pilgrim Health Care Foundation, Boston, Massachusetts. Dr Srivastava is the recipient of National Institute for Child Health and Human Development career development award K23 HD052553-01A1. Dr Landrigan is the recipient of Agency for Healthcare Research and Quality career development award K08 HS13333.
- Accepted March 19, 2007.
- Address correspondence to Rajendu Srivastava, MD, FRCP(C), MPH, Division of Inpatient Medicine, University of Utah Health Sciences Center, 100 N Medical Dr, Primary Children's Medical Center, Salt Lake City, UT 84113. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Coffman J, Rundall TG. The impact of hospitalists on the cost and quality of inpatient care in the United States: a research synthesis. Med Care Res Rev.2005;62 :379– 406
- ↵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
- ↵HCUPnet. Healthcare Cost and Utilization Project. Agency for Healthcare Research and Quality. Available at: hcupnet.ahrq.gov. Accessed October 3, 2005
- ↵Dwight P, MacArthur C, Friedman JN, Parkin PC. Evaluation of a staff-only hospitalist system in a tertiary care, academic children's hospital. Pediatrics.2004;114 :1545– 1549
- ↵Bellet PS, Whitaker RC. Evaluation of a pediatric hospitalist service: impact on length of stay and hospital charges. Pediatrics.2000;105 :478– 484
- ↵Wells RD, Dahl B, Wilson SD. Pediatric hospitalists: quality care for the underserved? Am J Med Qual.2001;16 :174– 180
- ↵Seid M, Quinn K, Kurtin PS. Hospital-based and community pediatricians: comparing outcomes for asthma and bronchiolitis. J Clin Outcomes Manage.1997;4 :21– 24
- ↵Meltzer D, Morrison J, Guth T, et al. Effects of hospitalist physicians on an academic general medicine service: results of a randomized trail [abstract]. J Gen Intern Med.1999;14(suppl 2) :112 . Available at: www.blackwell-synergy.com/doi/abs/10.1046/j.1525-1497.1999.1420083.x. Accessed June 13, 2007
- ↵Landrigan CP, Conway PH, Edwards S, Srivastava R. Pediatric hospitalists: a systematic review of the literature. Pediatrics.2006;117 :1736– 1744
- Muldoon JH. Structure and performance of different DRG classification systems for neonatal medicine. Pediatrics.1999;103 :302– 318
- Boyd J, 3rd, Samaddar K, Parra-Roide L, Allen EP, White B. Comparison of outcome measures for a traditional pediatric faculty service and nonfaculty hospitalist services in a community teaching hospital. Pediatrics.2006;118 :1327– 1331
- ↵US Bureau of Labor Statistics. Boston-Brockton-Nashua medical care CPI. Available at: http://stats.bls.gov/cpi. Accessed July 11, 2006
- ↵Kleinbaum DG. Applied Regression Analysis and Other Multivariable Methods. 3rd ed. Pacific Grove, CA: Duxbury Press; 1998
- ↵Kongstvedt PK. Negotiating and contracting with hospitals and institutions. In: The Managed Health Care Handbook. 4th ed. Gaithersburg, MD: Aspen; 2001:191– 205
- ↵Percelay JM. Physicians' roles in coordinating care of hospitalized children. Pediatrics.2003;111 :707– 709
- Percelay JM, Strong GB. Guiding principles for pediatric hospitalist programs. Pediatrics.2005;115 :1101– 1102
- ↵Lye PS, Rauch DA, Ottolini MC, et al. Pediatric hospitalists: report of a leadership conference. Pediatrics.2006;117 :1122– 1130
- Copyright © 2007 by the American Academy of Pediatrics