Photo Credit: EyeEm
The following is a summary of “Machine learning to identify a composite indicator to predict cardiac death in ischemic heart disease,” published in the June 2024 issue of Cardiology by Pingitore et al.
Machine learning (ML) uses algorithms to learn from data, creating models that predict events by analyzing many variables and interactions.
Researchers conducted a retrospective study evaluating how ML can identify patients with high-risk ischemic heart disease (IHD) for cardiac death (CD) prediction.
They enrolled 3,987 hospitalized patients with IHD (mean age 68 ± 11). Various ML models and ensembles were implemented and compared. The models’ outputs created a new indicator for patient stratification, and critical variables were evaluated using ablation tests to determine their importance.
The results showed an ensemble classifier of 3 ML models predicted cardiac death with AUROC 0.830 and F1-macro 0.726. Using the ML indicator in Cox survival analysis improved stratification by approximately 20% compared to standard methods. Patients classified as low risk by the ML indicator had significantly higher survival rates (88.8% vs. 29.1%). Key variables included Dyslipidemia, LVEF, Previous CABG, Diabetes, Myocardial Infarction, Smoke, Documented resting or exertional ischemia during rest or exertion, with AUROC 0.791 and F1-score 0.674.
Investigators concluded that ML offered a quicker, cost-effective, and dependable way to assess the prognosis of patients with IHD. They integrated numerous predictors and highlighted the most impactful ones, enhancing outcome prediction for advancing precision medicine in this area.
Source: sciencedirect.com/science/article/pii/S016752732400531X
