MON-644 Predictors of In-Hospital Mortality in Type 1 Diabetes Following Acute Myocardial Infarction using machine learning: Insights from the Myocardial Ischaemia National Audit Project (MINAP)

Abstract Disclosure: M.A. Shaikh: None. H. Deshmukh: None. T. Sathyapalan: None. S. Chattopadhyay: None. J. john: None. Objective: This study aimed to identify key predictors of in-hospital mortality in individuals with type 1 diabetes following acute myocardial infarction (AMI) using data from the Myocardial Ischaemia National Audit Project (MINAP). Methods: A total of 28,374 patients with type 1 diabetes were included. Baseline variables were assessed using a gradient boosting machine (GBM) model to determine relative importance (RI). The top five variables identified were subsequently included in a logistic regression model to quantify their impact on mortality. Model performance was assessed using the area under the receiver operating characteristic curve (AUC). Results: The mean age at procedure was 68.85 years (SD 12.76), with 38.1% being female. In-hospital mortality occurred in 8.4% of patients with type 1 diabetes. The most important predictors identified via GBM were Killip classification (RI=35.88), LVEF category (RI=12.15), systolic blood pressure (RI=11.87), age at procedure (RI=10.19), and creatinine level (RI=6.55). In logistic regression, cardiogenic shock (OR 8.63, 95% CI 7.00-10.65, p<0.0001) and poor LVEF (OR 3.84, 95% CI 3.20-4.62, p<0.0001) were the strongest predictors of mortality. Higher systolic blood pressure was protective (OR 0.98, 95% CI 0.98-0.99, p<0.0001), whereas increasing age (OR 1.04, 95% CI 1.03-1.04, p<0.0001) and elevated creatinine levels (OR 1.001, 95% CI 0.999-1.003, p=0.297) were associated with higher mortality risk. The model achieved an AUC of 80.1%, indicating strong predictive performance. Conclusion: Killip classification, LVEF category, systolic blood pressure, age, and creatinine are key predictors of in-hospital mortality in type 1 diabetes patients post-AMI. These findings may help guide risk stratification and inform clinical decision-making in this high-risk population. Presentation: Monday, July 14, 2025