Mettananda, C.Solangaarachchige, M. B.Haddela, P. S.Dassanayake, A. S.Kasturiratne, A.Wickramasinghe, A. R.Kato, N.De Silva, H. J.2025-09-102024Mettananda, C., Solangaarachchige, M. B., Haddela, P. S., Dassanayake, A. S., Kasturiratne, A., Wickramasinghe, A. R., Kato, N., & De Silva, H. J. (2024). IMPROVING CARDIOVASCULAR RISK PREDICTION OF SRI LANKANS USING ARTIFICIAL INTELLIGENCE (pp. 15–26). Desk Research Conference – DRC 2024, The Library, University of Kelaniya, Sri Lanka.http://repository.kln.ac.lk/handle/123456789/29878There are no CV risk prediction models derived from Sri Lankan cohorts. Therefore, the World Health Organization(WHO) risk charts developed for the Southeast Asia Region are being used to risk stratify Sri Lankans. However, Sri Lankans are quite different to some Southeast Asian countries and may not agree with Sri Lankans. Therefore, we aimed to develop a CV risk prediction model specific to a cohort of Sri Lankans. Using supervised machine learning of 10-year follow-up data of a randomly selected, population-based cohort of Sri Lankans, we developed a model to predict the 10-year risk of developing a cardiovascular event. We compared predictions of the new model at baseline in 2007 with the observed events in 2017 following a 10-year follow-up using receiver operating characteristic curves(ROC) to find the predictive performance. We compared the predictions of the new model and the currently used WHO risk charts. We selected 2596 Sri Lankans between 40 and 65 years old with no history of previous CV diseases (CVD) at recruitment and who had completed 10-year follow-ups. There were 179 hard CVDs recorded over the ten years. CVD included all cardiovascular deaths confirmed or presumed cases as mentioned in death certificates, non-fatal strokes, and physician-diagnosed non-fatal acute coronary syndromes, including elective percutaneous coronary interventions and coronary artery bypass grafts done on patients with symptomatic unstable angina. Any cardiac presentation except those mentioned here was excluded. Of 179 events, the ML-based model predicted 124; only 33 were predicted by the new model, while only 33 were predicted by 2019 WHO risk charts. The new ML-based model had 0.93 accuracy with an AUC-ROC of 0.74 ± 0.06. Machine learning of individual data of a Sri Lankan cohort improved CV risk prediction of Sri Lankans than using risk charts developed for an epidemiological region using a modelling approach.Cardiovascular RiskPredictionMachine LearningArtificial IntelligenceSri LankaArticle