AI-based cardiovascular risk stratification using population health data: An intelligent risk assessment agent (IRAA)
Year: 2026
Type: article
Abstract: Cardiovascular disease remains one of the leading causes of death worldwide, and identifying Artificial Intelligence paradigms that can support early diagnosis and preventive measures is therefore of great importance. Although many machine learning (ML) studies report high classification accuracy using cardiovascular risk predictors, these results can be misleading because of strong class imbalance and the population-screening nature of the data. In this study, we developed an explainable AI-based IRAA that focuses on cardiovascular risk categorization rather than binary diagnosis. To systematically evaluate different ML models using an imbalance- and risk-sensitive assessment framework, we employed a large population-based health investigation dataset. The proposed system achieved a stable ROC-AUC of approximately 0.83 and a PR-AUC of around 0.31, identifying more than 63% of heart disease cases within the top 25% of risk groups and nearly 78% within the top 30%. These results demonstrate the potential of the model for early screening and case prioritization rather than final clinical decision-making. To improve transparency and user trust, SHAP-based explanations were integrated into a conversational IRAA interface, enabling doctors and users to understand how demographic, lifestyle, and comorbidity factors contribute to an individual’s risk assessment. This functionality helps bridge the gap between the interpretability of complex predictive models and user understanding. The findings highlight the limitations of accuracy-focused evaluation methods and support a shift toward explainable and risk-aware AI-based cardiovascular screening at the population level.
Author: Oualid Ali
Institution: Applied Science University
Language: English
Subfield: Health Information Management
Field: Health Professions
Domain: Health Sciences
Open Access status: gold
Estimated APC paid: $469
OpenAlex ID: W7162995161