In this systematic review and meta-analysis, we found that hospitalists reduce hospital length of stay without increasing costs.
Published Online: January 16, 2012
Jean-Sebastien Rachoin, MD; Jad Skaf, MD; Elizabeth Cerceo, MD; Erin Fitzpatrick, MD; Barry Milcarek, PhD; Eric Kupersmith, MD; and Danielle Bowen Scheurer, MD, MSCR
Background: Hospital medicine has undergone remarkable growth since its creation. Most (but not all) of the published literature demonstrates better outcomes for patients cared for by hospitalists.
Purpose: We performed a systematic review and meta-analysis to estimate the magnitude of the impact of hospitalists on length of stay (LOS) and cost.
Data Sources: Medline/PubMed.
Study Selection: Articles published through February 2011 comparing outcomes (LOS and/or cost) of hospitalists with others.
Data Extraction: Two reviewers independently searched for abstracted information. We also contacted individual authors to provide us with missing data.
Data Synthesis: We used a random-effects model.
Results: A total of 502 abstracts were initially identified and 17 studies of 137,561 patients were included in the final analysis. LOS was significantly shorter in the hospitalist group compared with the non-hospitalist group, with a mean difference of –0.44 days (95% confidence interval [CI] –0.68 to –0.20, P <.001). In studies that compared a (non-resident) hospitalist service with a (non-resident) non-hospitalist service, LOS was also significantly shorter in the hospitalist group (mean difference –0.69 days [95% CI –0.93 to –0.46, P <.001]). Cost was not found to be significantly different (11 studies). There was significant heterogeneity between studies and we found no evidence of publication bias.
Conclusions: Despite its limitations, our analysis supports the conclusion that hospitalists significantly reduce LOS without increasing costs. These findings can be used to define and measure expectations of performance for hospital medicine groups.
(Am J Manag Care. 2012;18(1):e23-e30)
In this comprehensive meta-analysis of published studies to date, overall we found a significantly shorter length of stay of –0.44 days among hospitalists compared with nonhospitalists, with no significant difference in cost of care.
This reduction in LOS was most pronounced when non-resident hospitalists were compared with non-resident non-hospitalist services (–0.69 days).
There was significant heterogeneity between studies but no evidence of publication bias.
Our study findings can be used to serve as a benchmark for expectations for medical centers employing hospitalists.
The specialty of hospital medicine has undergone remarkable growth since the term “hospitalist” was introduced.1 The primary catalyst for the creation of many hospital medicine programs in the 1990s was operational; many hospitals were faced with the need to reduce length of stay (LOS) and total cost in order to maintain robust operating margins under a diagnosis-related group (DRG)-based reimbursement system. During this time, many hospitals found themselves maintaining a mixture of attending structures under their roofs, with some of their patients being attended by hospitalists, and others being attended by a “traditional” non-hospitalist (NH). Concurrently, due to the operating expenses involved in maintaining hospitalist programs, many hospital medicine program leaders were being asked to demonstrate their “value” to the hospital in which they were serving. As the value equation at the time was primarily related to cost and efficiency, a multitude of observational studies were published comparing these metrics between hospitalists and NHs.2-21
Although the catalysts for continued growth of the field have changed over time (including resident work-hour restrictions and the need for quality and safety physicians), hospital medicine programs are still expected to yield a return on investment for LOS and cost. For this purpose, it is vital to have a robust “benchmark” from which a hospitalist program’s efficiency and cost-reducing abilities can be evaluated. The purpose of this study is to perform a meta-analysis of published manuscripts to date, comparing overall mean differences in LOS and cost between hospitalists and NHs.
Search Methodology and Inclusion/Exclusion Criteria
The Medline/PubMed database was searched for all relevant articles published in English through February 28, 2011, using the following key words: “hospitalists” AND “length of stay” OR “cost” OR “outcomes.” Inclusion criteria comprised all adult prospective or retrospective studies that directly compared mean LOS or cost between a hospitalist group and an NH comparator group. Study definitions for hospitalist and NH comparator groups are further detailed in Table 1. Exclusion criteria included pediatric hospitalist studies, review articles, letters to the editors, and case reports.
Different definitions of hospitalist were used in the included studies, ranging from membership in the Society of Hospital Medicine, to physicians (family medicine physicians working a specific percentage of time caring for inpatients, ranging from 25% to 100%). Comparator group included primary care physicians, family medicine teaching, NHs, and private physicians.
The studies were categorized according to whether the attending physicians did or did not supervise residents. Resident services (either hospitalist or NH) were those patient care services in which the attending physician cared for the patient in conjunction with a resident in training.
All abstracts identified by the search were independently reviewed by 2 investigators (J-SR and JS), who determined the list of included articles by consensus. The full manuscripts of all included articles were then reviewed by a third independent investigator (EF) to confirm they met the inclusion criteria. Bibliographies of each eligible study were examined to identify any additional relevant references that might not have been captured in the Medline search. After the final selection, data were extracted by 2 independent reviewers (J-SR and JR).
For studies with incompletely published data (eg, did not publish the mean AND standard deviation for LOS or cost in BOTH the hospitalist and NH groups), the authors were contacted to determine if the required information was available. For those able to be contacted and with available information, they were included. Otherwise the data could not be included in the meta-analysis, but the study information is qualitatively discussed separately.
One study (Carek et al) had 3 distinct subgroups: 1) nonresident hospitalist group, 2) resident NH group, and 3) nonresident NH group. We therefore included the appropriate subgroups in the following analyses: group 1 versus group 3, and group 1 versus group 2.
The outcomes LOS and cost from each study are presented as unadjusted mean differences between hospitalist and NH groups. To account for significant differences between studies (clinical settings, patient characteristics, dates, definition of hospitalist), we used a random-effects model with inverse variance tau method for weighting. To assess for heterogeneity we used a funnel plot, and I2 index (derived for the standard Q index), with a value of 25% or less, suggesting low heterogeneity.2 Publication bias was assessed using weighted least squares Egger's test, and the Trim and Fill method.22,23 Data was analyzed using the MIX software 2.0.24
Study Inclusion and Description
The Medline search yielded 502 abstracts, all of which were independently reviewed (by J-SR and JS); 44 articles were deemed eligible by consensus. After full manuscript reviews (by EF), there were 12 additional articles excluded (no outcomes of interest or no comparisons between hospitalists and NHs), resulting in 32 studies for the final analysis (Appendix A). Of the 32 studies, 15 had complete data available (eg, mean and standard deviation for LOS or cost, for both groups.3-6,14,16-19,21,25-29 For 2 additional studies, the additional required information was available from the first author.15,30 For the remaining 15, the authors were either unable to be contacted, or did not have the required information available.2,7-13,20,31-36
The 17 included articles are outlined in Table 1. Study year ranged from 1998 to 2010, all but 4 were of retrospective design,6,21,26,27 and 11 were from academic medical centers. Study duration varied from 6 months to 5 years. Most patient groups analyzed were general hospitalized inpatients, although some studies focused on specific diagnoses, including heart failure, pneumonia, hip fracture, and gastrointestinal bleeding (Table 1).
Outcome Length of Stay
A total of 137,561 patients from 17 studies were included in the meta-analysis for mean LOS. LOS was significantly shorter in the hospitalist group compared with the NH group, with a mean difference of –0.44 days (95% confidence interval [CI] –0.68 to –0.20, P <.001) (Figure 1). LOS was further analyzed between resident and non-resident services. In the 8 studies that compared a (non-resident) hospitalist service with a (non-resident) NH service (102,684 patients), mean difference in LOS was also significantly shorter in the hospitalist group (mean difference –0.69 days (95% CI –0.93 to –0.46, P <.001) (Figure 2; Table 2).6,16,17,21,27,28,30 In the 5 studies that compared a resident hospitalist service with a resident NH service, and the 5 studies that compared a non-resident hospitalist service with a resident NH service, the mean LOS was not statistically different, with smaller sample sizes and wider confidence intervals (Table 2).
A total of 117,721 patients from 11 studies were included in the meta-analysis for mean difference in cost. There was no significant difference in cost between the hospitalist and NH groups, with a mean difference of $12 (95% CI –374 to 399, P = .95). When cost was further analyzed between resident and non-resident services, there were also no significant differences in cost between hospitalist and NH groups, except hospitalists with residents had costs significantly higher than NHs with residents (mean difference $890) (Table 2).
Studies With Incomplete Published Data
Due to incomplete data, 15 studies (including 75,493 patients) could not be entered in the meta-analysis. After average weighting by study sample size, the mean difference in LOS was –0.35 days shorter in the hospitalist group compared with the NH group (N = 62,122), and the mean difference in cost was –$51 less in the hospitalist group compared with the NH group (N = 55,220). Of the individual studies, 13 of the 15 studies found a shorter mean LOS for hospitalists, and 11 of the 15 studies found a lower mean cost for hospitalists, compared with NHs.
Assessment of Heterogeneity and Dissemination Bias
PDF is available on the last page.