LOCALLY DERIVED SYSTEM FOR CARDIAC OPERATIVE RISK EVALUATION LOKALNI SISTEM ZA EVALUACIJU OPERATIVNOG RIZIKA U KARDIOHIRURGIJI

Summary Introduction. During the last two decades, many authors have found that European Systems for Cardiac Operative Risk Evalua- tion (additive and logistic models) overestimate the risk in cardiac surgery. The new European model has recently been introduced as an update to previous versions. The aim of the study was to in- vestigate the significance of locally derived system for cardiac operative risk evaluation and to compare its predictive power with the existing European systems. Material and Methods. For developing a local risk prediction model, data from 2681 patients submitted to cardiac surgery at the Institute of Cardiovascular Diseases Vojvodina have thoroughly been collected. Logistic re- gression analysis was used to construct a local model for predic- tion of outcome. The evaluation of the local model and three Eu- ropean systems was performed by comparing the observed and expected hospital mortality. Results. The difference between the predicted and observed mortality regardless of the type of sur- gery was statistically insignificant for the additive European sys- tem (p=0.073) and the local model (p=0.134). The logistic Euro- pean system overestimated the operative risk, while the new Eu- ropean model underestimated mortality. In coronary surgery, all models, except the logistic European system, performed well. In valvular surgery, the new European model and the local model underestimated mortality significantly, while the additive and lo- gistic European models performed well. In combined surgery, the new European system significantly underestimated mortality (p=0.029), while the local model performed well (p=0.252). Con- clusion. The locally derived model shows satisfactory results, with good calibration and discriminative power. The local model specifically outperforms all other European systems in terms of discriminatory power in combined surgery subset.