Accuracy of actuarial and clinical predictions for length of stay and unauthorized absence

Multivariate-equation predictions of length of hospital stay and probability of unauthorized absence were compared with similar predictions made by clincians and actual outcome for 167 patients drawn from a random sample of admissions during a two-month period at five state hospitals. The computer proved to be about as accurate as clinicians in predicting length of stay. Neither computer nor clinician was able to usefully predict unauthorized absence because of its infrquent occurrance. Use and improvement of such actuarial predictions within the Missouri automated data base were discussed.