In-hospital mortality risk prediction after percutaneous coronary interventions: Validating and updating the Toronto score in Brazil.

Abstract

We aimed to assess the accuracy of the simple, contemporary and well-designed Toronto PCI mortality risk score in ICP-BR registry, the first Brazilian PCI multicenter registry with follow-up information. Estimating percutaneous coronary intervention (PCI) mortality risk by a clinical prediction model is imperative to help physicians, patients and family members make informed clinical decisions and optimize participation in the consent process, reducing anxiety and improving quality of care. At a healthcare system level, risk prediction scores are essential to measure and benchmark performance. Between 2009 and 2013, a cohort of 4,806 patients from the ICP-BR registry, treated with PCI in eight tertiary referral medical centers, was included in the analysis. This population was compared to 10,694 patients of the derivation dataset from the Toronto study. To assess predictive performance, an update of the model was performed by three different methods, which were compared by discrimination, calculating the area under the receiver operating characteristic curve (AUC), and by calibration, assessed through Hosmer-Lemeshow (H-L) test and graphical analysis. Death occurred in 2.6% of patients in the ICP-BR registry and in 1.3% in the Toronto cohort. The median age was 64 and 63 years, 23.8 and 32.8% were female, 28.6 and 32.3% were diabetics, respectively. Through recalibration of intercept and slope (AUC = 0.8790; H-L P value = 0.3132), we achieved a well-calibrated and well-discriminative model. After updating to our dataset, we demonstrated that the Toronto PCI in-hospital mortality risk score performed well in Brazilian hospitals.