ACTION and NCDR risk models outperform GRACE for prediction of in-hospital mortality in acute coronary syndrome – a retrospective analysis from a German Heart Center

Abstract

Background Risk prediction with the GRACE risk model is guideline-recommended clinical practice in acute coronary syndrome (ACS). However, more modern risk models such as ACTION (Acute Coronary Treatment and Intervention Outcomes Network) Registry–GWTG (Get With the Guidelines) and National Cardiovascular Data Registry (NCDR) risk models are available. We aimed to compare these models to the established GRACE risk model in ACS. Methods and results In-hospital mortality was retrospectively assessed in 1,138 patients undergoing cardiac catheterization for Non-ST-Elevation Myocardial Infarction (NSTEMI, 566 patients, 70.7% male) or ST-Elevation Myocardial Infarction (STEMI, 572 patients, 69.1% male) at a German University Hospital from 2014 to 2017. In-hospital mortality was 14.7% for STEMI and 3.7% for NSTEMI, respectively. GRACE, ACTION and NCDR risk models for prediction of in-hospital mortality were calculated for individual patients, 0.75% missing data were imputed. ACTION risk model showed a good discrimination of risk (c-index 0.85, 95% confidence interval (CI) 0.83–0.87) with a slight numerical advantage in NSTEMI (c-index 0.92, 95% CI 0.86–0.98) over STEMI patients (c-index 0.83, 95% CI 0.79–0.88). The NCDR risk model showed comparable performance in the overall cohort (c-index 0.86, 95% CI 0.84–0.88; NCDR vs. ACTION p=0.4097), also with superior performance in NSTEMI (c-index 0.89, 95% CI 0.86–0.91) vs. STEMI (c-index 0.81, 95% CI 0.78–0.84). The GRACE risk model showed significantly worse performance in the overall cohort (c-index 0.76, 95% CI 0.74–0.79; vs ACTION p<0.0001; vs. NCDR p<0.0001) and in STEMI patients (c-index 0.72, 95% CI 0.69–0.76; vs ACTION p<0.0001; vs. NCDR p=0.0018). In NSTEMI patients, GRACE discrimination performance was comparable to NCDR (c-index 0.87, 95% CI 0.84–0.90, p=0.73), but still inferior to ACTION (p=0.04). The ACTION risk model showed a good calibration whereas NCDR and GRACE models lacked accuracy in our cohort. Conclusion In a contemporary German patient population with acute coronary syndrome, ACTION and NCDR risk models outperform the established GRACE risk model for prediction of in-hospital mortality. This performance difference was more pronounced in STEMI than in NSTEMI. Funding Acknowledgement Type of funding source: None