Predicting length of hospital stay of patients admitted to the psychiatric ward of a general hospital

Recent emphasis on cutting the costs of psychiatric care and the increasing possibility that reimbursement for psychiatric services will be based on diagnosis related groups (DRGs) has prompted research on the relationship between patient and environmental variables and length of hospital stay. A review of the literature revealed that diagnosis alone is not an accurate prediction of hospital length of stay. However, if DRGs are combined with other patient and environmental variables its predictive ability may improve. The current study used discharge data from 234 psychiatric patient charts from Evanston Hospital's general psychiatric inpatient unit to determine what patient characteristics and environmental factors are useful in predicting hospital length of stay. Eight forecasting models were replicated. A new model was also presented using 31 independent variables and the log transformation of length of stay as the dependent variable. The findings support the previous conclusion that DRGs are not good predictors for length of stay. Only one of the replicated studies produced significant findings (Gruber, 1982). The new model proposed in this study produced the most significant findings and explained 39% of the variance in length of stay. A potentially more useful approach for future decisions on length of stay is proposed.