Precision and accuracy of risk scores for in‐hospital death after percutaneous coronary intervention in the current era

Abstract

Various risk assessment scores were proposed in the last decade for prediction of in-hospital mortality in patients undergoing percutaneous coronary intervention (PCI). We sought to apply two validated scores, the Mayo Clinic Risk Score (MCRS) and the New York Risk Score (NYRS) to a contemporary cohort treated at a single institution and to simplify the NYRS, such that the parameters used in both scores are similar. Patients undergoing PCI in 2005-2007 were included. MCRS and NYRS were calculated for each patient. A simplified NYRS, similar to MCRS, was constructed by deleting two variables (gender and left main coronary stenosis). Model discrimination was assessed by the C statistic and goodness-of-fit (calibration) was measured with the Hosmer-Lemeshow test. There were 3,165 procedures. The in-hospital mortality was 0.56% (95% CI 0.31-0.83%). Mean MCRS was 2.7 +/- 2.4 (predicted mortality 0.3%). The C-statistic for MCRS was 0.82 (0.71-0.94) and the model was well calibrated (P = 0.79). Mean NYRS was 5.1 +/- 3.3, (predicted mortality 0.23%). The C-statistic for NYRS was 0.83 (0.74-0.95), not different from MCRS (P = 0.62) and the model was well calibrated (P = 0.29). The mean simplified NYRS was 4.6 +/- 3.1 among survivors and 10.9 +/- 5.8 among those who died, P < 0.001. The score had a C-statistic of 0.83 (0.71-0.95), not different from MCRS (P = 0.84) or NYRS (P = 0.27) and was well calibrated (P = 0.71). PCI risk scores utilizing easily collected variables are useful in discriminating risk and predicting death. NYRS might be simplified by removing the gender and left main coronary stenosis variables from its algorithm.