One hundred and eighteen patients consecutively submitted to mechanical ventilation during a one year period, and admitted to the Respiratory Intensive Care Unit of the Hospital Clinic of Barcelona, were prospectively studied in order to define the importance of hypoxaemia as a predictor of mortality. Using a discriminant multivariate analysis, the following variables were selected as the best predictors of outcome: 1) the number of associated complications (NAC) on admission; 2) the simplified acute physiologic score (SAPS); 3) oxygenation index (PAO2-AaPO2)/PAO2 + 0.014 positive end-expiratory pressure (PEEP); and 4) the age of the patients. Using these predictors, 84% of the patients were accurately classified as survivors or as nonsurvivors. The partial contribution of each predictor to the model was also assessed using a logistic regression, by eliminating each single predictor and each possible pair of predictors. Using this means of analysis, the NAC and SAPS were the only predictors of mortality. The inclusion of short-term mechanically-ventilated patients did not bias the accuracy of prediction.