Predictors of length of hospital stay

The purpose of this study was to determine whether diagnostic, clinical, and demographic variables would adequately predict variation in length of hospital stay in an acute-care psychiatric unit. Three hundred and ten patients with first-ever admissions were recorded to have been admitted to the unit during the period from November 1993 to August 1995. The following independent factors were tested for their predictiveness in explaining variation in length of hospital stay: sex, marital status, nationality, ICD-10 Axis 1 diagnosis, ''receipt'' of electroconvulsive therapy, type of medication received, number of medications received, associated disabilities on admission, and employment status. Demographic variables were found to be poor predictors of length of hospital stay. The rest of the variables explained in total about 50% of variation in the dependent variable. Stepwise regression analysis showed that the best subset of the independent variables explained about 27% of the variation. The significance of the findings is discussed in the text.