Predictive value of EuroSCORE and Parsonnet scoring in Saudi population

There is no previously published report on the applicability of risk stratification systems to adult cardiac patients in Saudi Arabia.

We have tried to verify the predictive value of Parsonnet and EuroSCORE in Saudi patients.

Data of 194 consecutive adult patients operated in our unit over the period of last few years, was obtained retrospectively. There were no omissions. The EuroSCORE and Parsonnet scores were estimated for all the patients and compared with the observed mortality, morbidity and hospital length of stay (LOS). Calibration was determined by Hosmer-Lemeshow goodness of fit statistics (HL) while discrimination was assessed by using receiver operating characteristic (ROC) curve.

For prediction of "mortality" the area under ROC curve was 0.766+/-0.077 S.E. for EuroSCORE and 0.685+/-0.087 S.E. for Parsonnet score. For "morbidity" the area under ROC curve was 0.628+/-0.048 S.E. for EuroSCORE and 0.660+/-0.048 S.E. for Parsonnet score while for the LOS, ROC curves were 0.632+/-0.056 S.E. and 0.671+/-0.053 S.E. accordingly. Both scoring systems were adequately calibrated in our data.

Parsonnet and EuroSCORE are equally applicable as mortality predictors in Saudi population. EuroSCORE was slightly better than Parsonnet scoring in predicting the mortality in medium and high risks categories. Both the systems have less discrimination for prediction of "morbidity" and LOS.