Development and validation of the INFLammation in Acute Myocardial diseases risk Estimation for Myocardial Infarction (INFLAME-MI) score

Abstract Introduction Elevated leucocytes, particularly neutrophils, are strongly associated with poor prognosis after acute myocardial infarction (AMI). However, inflammatory indices are not included in established post-AMI risk scores. We aimed to derive a novel, inflammation-based risk score with superior performance to contemporary scores to predict in-hospital mortality after AMI. Methods and results This was a retrospective, multicentre, longitudinal cohort study including a derivation cohort of consecutive patients presenting with AMI between 2016 and 2023. The primary outcome was in-hospital mortality. Primary outcome risk estimates were derived by regularized machine learning techniques, incorporating leucocyte subpopulations, which we termed the INFLammation in Acute Myocardial diseases risk Estimation for Myocardial Infarction (INFLAME-MI) score. The INFLAME-MI score was validated in an external international cohort and compared to the Global Registry of Acute Coronary Events (GRACE) score. The INFLAME-MI score was derived from a cohort of 3028 AMI patients and tested in an external validation cohort of 682 AMI patients. It demonstrated strong internal discrimination (area under the curve [AUC] 0.88) and maintained excellent external predictive performance for the primary outcome, achieving a non-inferior AUC of 0.92 relative to the GRACE score (AUC 0.94, P = 0.256). Furthermore, INFLAME-MI demonstrated superior calibration and improved reclassification compared to the GRACE score (Hosmer–Lemeshow χ²₈ = 6.2 (P = 0.102) vs. χ²₈ = 22.9 (P < 0.001), respectively). Conclusion INFLAME-MI is a novel, rapid, and easy-to-use risk score for in-hospital mortality after AMI, derived from common inflammatory indices. In external validation, it demonstrates comparable discrimination, improved calibration, and higher net benefit at ≤7% thresholds compared with the GRACE score, particularly within key decision-making ranges. Further work is needed to establish the INFLAME-MI score through validation of these findings in diverse, prospective patient cohorts.