Abstract 16250: Comparison of Risk Models to Predict In-Hospital Mortality for Patients With Acute Coronary Syndrome in India: The CSI-Kerala Risk Score
Introduction: Acute coronary syndrome (ACS) risk models such as the Global Registry of Acute Coronary Events (GRACE) risk score are commonly used to predict in-hospital death. However, no India-specific risk models have been derived to date.
Methods: The Cardiological Society of India - Kerala Chapter (CSI-K) prospectively collected data on 25,748 ACS admissions from 2007-9. Demographic, medical history, clinical, and outcome data were collected, including in-hospital mortality. We compared the utility of 3 logistic regression risk models to predict in-hospital death across deciles of risk: 1) original GRACE score, 2) calibrated GRACE score, and 3) internally derived CSI-Kerala risk score.
Results: The predicted in-hospital mortality rate using the GRACE model was lower than the observed in-hospital mortality rate across all deciles risk (range of underestimation=2.5-11.1%). The following independent risk factors (with respective beta coefficients) were included in the CSI-Kerala risk model to optimize prediction of in-hospital mortality in the Kerala ACS Registry: age (0.0001), sex (0.0914), BMI (0.03201), symptom-to-door time >6 hours (0.9587), Killip classification>1 (0.4648), STEMI diagnosis (1.3316), abnormal cardiac enzyme (0.9364), serum creatinine (0.0945), hemoglobin (-0.0827), and in-hospital heparin use (-0.2815). The Figure shows the differences in observed in-hospital mortality rates of the 3 models across deciles of predicted risk. The discrimination of the CSI-Kerala risk score (C statistic=0.78, goodness-of-fit chi2>0.99) was greater than both the GRACE score (C statistic=0.65) and calibrated GRACE score (C statistic=0.73).
Conclusions: The GRACE score underestimated in-hospital mortality in the Kerala ACS Registry. The internally derived CSI-Kerala risk model is the first Indian-specific ACS risk model to predict in-hospital death and has greater discrimination than the GRACE risk score, even after calibration.
- © 2011 by American Heart Association, Inc.