SMABS 2004 Jena University
SMABS 2004 Home Organization About Jena Sponsors Links Imprint / Contact SMABS Home

European Association of Methodology

Department of methodology and evaluation research

Jena University

Contributions: Abstract

Patient length of stay: an application of multi-level modeling in health services research

Elizabeth Gifford E. Michael Foster Allison Olchowski
Penn State University

OBJECTIVE: Previous research on inpatient psychiatric care for children with emotional problems indicates that patient characteristics only weakly predict length of stay (LOS). One explanation is that LOS is largely determined by facility characteristics. While infrequently used in health services research, multilevel modeling is well suited to addressing this question.

STUDY DESIGN: This study uses insurance claims data from 1996 to 2001 representing 30,584 admissions involving 17,656 youth in 263 facilities. The analyses reflect the cross-classified nature of the data - admissions are nested both within patients and facilities. Patients, however, are not nested within facilities as they may be treated in multiple facilities. Covariates include both patient-level characteristics and facility-level characteristics. We assume an improper uniform prior for the fixed parameters and diffuse gamma priors for the variance parameters.

PRINCIPAL FINDINGS: Our results suggest that roughly 8 percent of the variation in LOS is explained at the patient-level while 47 percent is explained at the facility-level. Our analyses also demonstrate the benefits of the cross-classified data structure; having repeated observations on individuals in different facilities improves the precision of variance estimates by 80 percent. By presenting shrinkage estimates of the residuals, we demonstrate a method for identifying facilities whose average LOS is exceptionally long or short and compare those results with frequentist results.

CONCLUSIONS: Nearly 50 percent of the variation in LOS is explained at the facility-level. Given the vulnerable nature of youth who are in need of inpatient psychiatric care, it may be particularly important to monitor provider-level processes and outcomes. Measuring facility or provider level quality is complicated because of difficulties in directly comparing the illness levels of the patients that each treats. The methodology presented here improves our ability to identify high performing or low performing facilities.