Background Inflammation plays a pivotal role in the progression of diabetes and its cardiovascular complications, particularly acute myocardial infarction (AMI). Patients with AMI often face high mortality and morbidity, making accurate prognosis crucial for clinical decision-making and outcome improvement. Objective This study aims to develop and validate a stacked predictive model using inflammation-related indices to predict 3-year all-cause mortality among patients with severe diabetes who experience acute myocardial infarction (AMI), aiming to improve patient prognosis. Methods We included 833 patients with severe diabetes combined with AMI from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database (training/test cohort) and 166 cases from Zhongnan Hospital, China (external validation cohort with 3-year follow-up). A total of 5 inflammation-related indices (lymphocyte-to-monocyte ratio LMR, neutrophil-to-lymphocyte ratio NLR, neutrophil-to-platelet ratio NPR, platelet-to-lymphocyte ratio PLR, prognostic inflammatory value PIV) were analyzed for their association with mortality using Cox proportional hazards models. Kaplan-Meier curves and restricted cubic spline analysis explored survival probabilities and dose-response relationships. A stacked predictive model incorporating these indices was constructed, and its performance was evaluated using the area under the curve. Results LMR was a protective factor (hazard ratio HR 0.44, 95% CI 0.25-0.77; P<.001), while NLR (HR 1.78, 95% CI 1.19-2.65; P=.004) and PIV (HR 1.59, 95% CI 1.09-2.30; P=.01) were associated with increased mortality risk. Kaplan-Meier analysis showed mortality increased with decreasing LMR quartiles and increasing NLR, NPR, PLR, and PIV quartiles. Restricted cubic spline confirmed that decreasing LMR and increasing NLR, NPR, PLR, and PIV were associated with higher adverse event risk. The predictive model achieved an area under the curve of 0.803 (95% CI 0.736-0.871) in internal testing and 0.781 (95% CI 0.704-0.858) in the external validation cohort. Conclusions The stacked predictive model serves as a robust tool for predicting 3-year outcomes in patients with diabetes combined with AMI. Its reliance on routine, low-cost indices highlights its economic viability and potential for widespread clinical implementation, particularly for optimizing resource allocation in resource-limited settings. However, considering the single-center nature of the derivation data, future multicenter validation is essential to verify the model’s generalizability across different health care systems before establishing it as a standard policy for risk stratification.
Zhou et al. (Thu,) studied this question.