Construction and Validation of a Risk Prediction Model for Sepsis-Induced Myocardial Injury

Background: Sepsis patients face a high risk of myocardial injury, which increases the risk of death. Therefore, the rapid and accurate assessment of myocardial injury risk is crucial for improving prognosis. Objective: To construct and validate a risk prediction model for sepsis-induced myocardial injury (SMCI). Methods: Patients were randomly assigned to a training cohort and an internal validation cohort in a 7:3 ratio. Least Absolute Shrinkage and Selection Operator (LASSO) regression and multivariate logistic regression were used to identify independent predictors for the construction of a nomogram. The model's discrimination, calibration, and clinical applicability were evaluated using area under curve (AUC), Hosmer-Lemeshow tests, decision curve analysis (DCA) and clinical impact curve (CIC). Meanwhile, internal validation was conducted. Results: The study included 370 patients, with 262 in the training cohort and 108 in the validation cohort. 3 independent risk factors were identified, including Log myoglobin (Myo), Log B-type natriuretic peptide (BNP), and Log interleukin-6 (IL-6) and a nomogram incorporating these factors was constructed. The AUC in the training and validation cohorts was 0.856 and 0.853, respectively. The Hosmer-Lemeshow test indicated good calibration in both cohorts, while DCA and CIC demonstrated strong clinical applicability. Conclusion: The nomogram based on Log Myo, Log BNP, and Log IL-6 may serve as a practical tool for the early identification of high-risk patients by facilitating the rapid calculation of SMCI risk.