Assessment of Euroscore and SAPS III as hospital mortality predicted in cardiac surgery.

OBJECTIVES: To perform an external validation of Euroscore I, Euroscore II and SAPS III.

PATIENTS AND METHOD: Retrospective cohort study over three years on all adult patients who underwent cardiac surgery. We reviewed the clinical data, following the patient until outcome or discharge from hospital (dead, alive). We computed the predicted mortality by Euroscore I (EI), II (EII) and SAPS III. The model validation was assessed by discrimination: area under curve ROC; and calibration (Hosmer-Lemeshow test).

RESULTS: 866 patients were included. 62.5% of them male, with a median age of 69 years, 6.1% died during hospitalization. Predicted mortality: E I 7.94%, E II 3.54, SAPS III 12.1%. Area under curve (95% IC): E I 0.862 (0.812-0.912); E II 0.861 (0.806-0.915); SAPS III 0.692 (0.601-0.784). Hosmer-Lemeshow test: E I 14.0046 (P=.08164); E II 33.67 (P=.00004660); SAPS III 11.57 (P=.171).

CONCLUSIONS: EII had good discrimination, but the calibration was not good with predicted mortality lower than the real mortality. E I showed the best discrimination with good calibration and a tendency to overestimate the mortality. SAPS III showed poor discrimination with good calibration and a tendency to greatly overestimate the predicted mortality. We saw no improvement in the predictive performance of EII over I and we reject the use of SAPS III in this kind of patient.