Mortality risk prediction in infective endocarditis surgery: reliability analysis of specific scores

OBJECTIVES: We assessed the prognostic utility of risk scores in surgery for infective endocarditis (IE) to evaluate their reliability in mortality risk prediction. METHODS: An observational retrospective study was developed to include all patients who underwent surgery for active IE from 2002 to 2016. Classical and endocarditis-specific risk scores were calculated. RESULTS: A total of 180 patients were included in the study. The 30-day mortality rate was 26.82% [95% confidence interval (CI) 20.26-33.20%]. Classical risk scores were confirmed to have a suboptimal prognostic ability. Therefore, 4 IE-specific risk scores were calculated. Discrimination was evaluated using the area under the receiver operating characteristic curve. It was 0.76 (95% CI 0.68-0.82) for the Society of Thoracic Surgeons-IE (STS-IE) score; 0.68 (95% CI 0.58-0.76) for the De Feo-Cotrufo score; 0.73 (95% CI 0.66-0.79) for the PALSUSE score and 0.65 (95% CI 0.57-0.72) for the Costa score. The STS-IE score had higher discrimination when compared with the De Feo-Cotrufo score (P = 0.055) and the Costa score (P = 0.024); however, there was no significant difference when we compared the STS-IE score with the PALSUSE score (P = 0.58). Calibration was assessed using the Hosmer-Lemeshow test; an adequate calibration was confirmed in all 4 scores. CONCLUSIONS: Specific risk scores had better prognostic performance than classical risk scores. The STS-IE score had the highest discrimination and was adequately calibrated. The PALSUSE score also showed optimal discrimination and calibration. The De Feo-Cotrufo score had a lower discrimination in our sample; however, the De Feo-Cotrufo score is recommended in the current guidelines. The Costa score had the lowest discrimination.