Role of patient severity in predicting length of hospital stay

grantor: University of TorontoThis study assessed the role of the Computerized Severity Index (CSI$\sp\circler$) in predicting length of stay (LOS) for discharges from psychiatric inpatient units. Using the CSI, raters retrospectively reviewed medical charts, producing three continuous severity ratings for each patient-admission, maximum and discharge. Hierarchical and stepwise multiple regression analyses were conducted to predict LOS using CSI ratings and other patient data. Subgroups with psychotic disorders and major depressive disorders were modeled separately. In both groups severity measures predicted 8-10% variation in LOS controlling for discharge abstract variables, and final models predicted 14-19% variation in LOS. While severity improves capacity to predict LOS, prediction is still inadequate for funding purposes. Future work should concentrate on incorporating measures of personal and social functioning into severity ratings, and on assessing provider and site influences on LOS. In mental health, funding models based on program rather than individual performance may be more appropriate.M.Sc