Profiling hospitals for length of stay for treatment of psychiatric disorders.

Abstract

Managed behavioral health care organizations (MBHOs) often profile hospitals on length of stay (LOS) and other performance measures. However, previous research has suggested that most of the variation in utilization for general medical conditions is attributable to case-mix indicators and random sources rather than individual providers. Hospital discharge data are used to estimate hierarchical linear models, where hospitals and physicians within hospitals are treated as a random effect. The goal was to determine the intraclass correlation coefficient (ICC) for psychiatric LOS for hospitals and for physicians before and after making case-mix adjustments. After controlling for case-mix, the hospital ICCs for depression, schizophrenia, and bipolar disorder show that 32%, 36%, and 11% of the variation in LOS, respectively, can be attributed to hospitals, while 7%, 5%, and 6% of the variation in LOS, respectively, can be attributed to physicians or provider practice. Unlike health services for other conditions, the variation in LOS for inpatient psychiatric treatment of depression and schizophrenia is quite dependent upon hospitals.