Comparison of APACHE III, APACHE IV, SAPS 3, and MPM₀III and influence of resuscitation status on model performance

Background: There are few comparisons among the most recent versions of the major adult ICU prognostic systems (APACHE [Acute Physiology and Chronic Health Evaluation] IV, Simplified Acute Physiology Score [SAPS] 3, Mortality Probability Model [MPM]₀III). Only MPM₀III includes resuscitation status as a predictor. Methods: We assessed the discrimination, calibration, and overall performance of the models in 2,596 patients in three ICUs at our tertiary referral center in 2006. For APACHE and SAPS, the analyses were repeated with and without inclusion of resuscitation status as a predictor variable. Results: Of the 2,596 patients studied, 283 (10.9%) died before hospital discharge. The areas under the curve (95% CI) of the models for prediction of hospital mortality were 0.868 (0.854-0.880), 0.861 (0.847-0.874), 0.801 (0.785-0.816), and 0.721 (0.704-0.738) for APACHE III, APACHE IV, SAPS 3, and MPM₀III, respectively. The Hosmer-Lemeshow statistics for the models were 33.7, 31.0, 36.6, and 21.8 for APACHE III, APACHE IV, SAPS 3, and MPM₀III, respectively. Each of the Hosmer-Lemeshow statistics generated P values < .05, indicating poor calibration. Brier scores for the models were 0.0771, 0.0749, 0.0890, and 0.0932, respectively. There were no significant differences between the discriminative ability or the calibration of APACHE or SAPS with and without "do not resuscitate" status. Conclusions: APACHE III and IV had similar discriminatory capability and both were better than SAPS 3, which was better than MPM₀III. The calibrations of the models studied were poor. Overall, models with more predictor variables performed better than those with fewer. The addition of resuscitation status did not improve APACHE III or IV or SAPS 3 prediction.