Abstract
BACKGROUND: Myocardial infarction (MI) can lead to significant psychological distress, adversely affecting recovery and prognosis. This study aimed to investigate post-MI depression and anxiety, identify key predictors including heart rate variability (HRV), and assess the predictive performance of machine learning (ML) models. METHODS: This prospective, single-center observational cohort study collected demographic, clinical, and psychological data from 70 post-MI patients. The main outcomes included the Hospital Anxiety and Depression Scale (HADS) and heart rate variability (HRV) metrics. Methods involved data processing, feature ranking and selection to identify key predictors, and machine learning (ML) to predict depression and anxiety. RESULTS: High prevalence rates of HADS-defined depression (41.43%) and anxiety (44.29%) were observed post-MI, particularly in women. For depression prediction, Logistic Regression yielded a recall of 0.80, balanced accuracy of 0.71, an F1 Score of 0.67, and an AUC-ROC of 0.75. For anxiety prediction, GaussianNB (F1 Score 0.63, Recall 1.00, AUC-ROC 0.67) and Random Forest Classifier (F1 Score 0.55, Recall 0.60, AUC-ROC 0.63) demonstrated the strongest performance. CONCLUSION: This study underscores the significant prevalence of depression and anxiety in post-MI patients. Machine learning models incorporating clinical and HRV measures show potential for improved early risk stratification of these conditions. Health literacy challenges encountered during HADS administration highlight the need for objective measures to complement psychological assessment in this vulnerable population, potentially improving patient care.