The value of anthropometric and biochemical indices in predicting short-term mortality among patients in general medical wards was assessed in 294 patients admitted consecutively to a district hospital over a one month period. Using a stepwise logistic regression model and supported by the linear discriminant analysis method, mortality within 3 months could be predicted with sensitivity of 83% and specificity of 84% using the following variables: sex, functional ability, urea, total protein, alkaline phosphatase and albumin-adjusted calcium. Addition of anthropometric values and biochemical nutritional indices did little to improve the accuracy of the prediction, contrary to previous findings among surgical patients and elderly residents of long-term care institutions.