Enhancing Cardiovascular Risk Prediction through Age-Based and Behavioral Health Indicator Machine Learning Models

Cardiovascular diseases (CVDs) are among the leading causes of death globally, causing around 17.9 million deaths every year, as reported by the WHO. Patient outcomes would also be maximally improved if predictions and management were to be made available early, especially in regions that are resource-scarce. The prevailing conventional diagnosis has mainly relied on relatively costly tests, such as angiography and electrocardiograms (ECG). However, the developing field of machine learning now provides a cost-effective and efficient way of analyzing health data to predict the risk of heart disease. This paper introduces a new approach that integrates, in the heart disease predictive models, factors concerning psychological states, such as stress, hours worked, and sleep patterns. Using the Cleveland Heart Disease dataset enriched with synthetic features representing mental health, we track the amount each of these factors contributes to predictions. We also show personalized modeling based directly on training within demographic subgroups- age under 40 versus 40 years of age or older, as well as gender - for determining whether such personal models are better than global models. We use many ML classifiers, namely Random Forest, XGBoost, KNN, SVM, Logistic Regression, Decision Tree, and Naive Bayes, to develop and compare the prediction models. All the models are assessed both on the original dataset as well as the augmented dataset using metrics such as accuracy, precision, recall, F1 score. Incorporating mental health indicators significantly improves the power of the models and lends importance to the balanced monitoring of health in cardiovascular care. When specific groups of patients are considered for the demographic-based models, the performance is good, and it underlines possible applications in personalized healthcare solutions.