The Computerized Psychiatric Severity Index (CPSI) and 22 patient variables were used to predict length of hospitalization for 304 psychiatric patients in DRG 430 who were diagnosed with schizophrenia or affective disorder and had no secondary diagnoses. Length of stay, which correlated .96 with total charges, was used as the dependent variable (measure of resource use). The patient variables and CPSI score explained 32.5 percent of the variation in length of stay for all of DRG 430 (27.5 percent for affective disorder patients and 70.3 percent for schizophrenia patients). Addition of the treatment variable "receipt of ECT" (electroconvulsive therapy) permitted the regression models to explain 40.9 percent of the variation in length of stay (36.24 percent for affective disorder and 71.22 percent for schizophrenia). In each regression model, maximum CPSI score was significant, indicating that much heterogeneity in DRG 430 can be explained by CPSI. Using one payment for such a diverse group places health care institutions at great risk of financial loss. Our study indicates that a continuing need exists for research in the area of case-mix measures for psychiatric inpatients.