Conference Abstract | Volume 8, Abstract NACNDC/19JASH057 (Poster) | Published: 07 Dec 2025
&Corresponding author: Dafala Kezimbira, Department of Immunology & Molecular Biology, School of Biomedical Sciences,
Makerere University College of Health Sciences, Email: dafala.kezimbira@mak.ac.ug ORCID: https://orcid.org/0009-0005-0717-6150
Received: 11 Sep 2025, Accepted: 20 Oct 2025, Published: 07 Dec 2025
Domain: Non-Communicable Disease Epidemiology
Keywords: Hypertension risk prediction, machine learning, explainable AI, lifestyle factors, WHO STEPS survey 2023
©Dafala Kezimbira et al. Journal of Interventional Epidemiology and Public Health (ISSN: 2664-2824). This is an Open Access article distributed under the terms of the Creative Commons Attribution International 4.0 License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Cite this article: Dafala Kezimbira et al., Using machine learning and explainable AI to predict lifestyle-driven hypertension risk in Uganda’s WHO STEPS 2023 survey. Journal of Interventional Epidemiology and Public Health. 2025;8(ConfProc6):057.https://doi.org/10.37432/JIEPH-CONFPRO6-00057
Hypertension is a leading driver of cardiovascular disease globally, accounting for over 10.8 million deaths each year. In sub-Saharan Africa, more than 46% of adults are hypertensive, yet diagnosis and control rates remain critically low. In Uganda, the 2023 WHO STEPS survey revealed that 27.7% of adults aged 18–69 years are hypertensive, with fewer than 9% achieving adequate control. This rising burden is linked to modifiable lifestyle risks: over 86% of Ugandans consume inadequate fruits and vegetables, 31% engage in harmful alcohol use, and physical inactivity is widespread, exacerbated by urbanisation and dietary shifts. Despite this, traditional tools like the Framingham Risk Score are ill-suited for local contexts, as they lack behavioural depth and cultural adaptation. This study applies Machine Learning (ML) and Explainable AI (XAI) to develop a context-aware, interpretable hypertension risk prediction model tailored to Uganda.
Using nationally representative data from 3,694 adults in the 2023 WHO STEPS Uganda Survey, we developed three ML models: Logistic Regression, Random Forest, and XGBoost to predict hypertension based on lifestyle, physiological, demographic, and biochemical features. Model performance was assessed using ROC–AUC, PR–AUC, F1-score, and Brier score. SHAP and LIME were applied for global and individual-level explainability. Counterfactual simulations examined how modest lifestyle improvements, such as reducing BMI, lowering salt intake, or incorporating 30 minutes of daily walking (in line with WHO recommendations) could affect risk outcomes.
The Random Forest model achieved the best performance (ROC–AUC = 0.93, F1 = 0.86, Brier score = 0.151). Top predictors included systolic/diastolic blood pressure, age, BMI, salt intake, and physical activity. XAI tools revealed that many individuals’ risk profiles were driven by modifiable lifestyle, not fixed demographics. Simulation scenarios showed that introducing 30 minutes of walking daily, equivalent to the 150 minutes/week WHO minimum, could significantly reduce predicted hypertension risk, alongside 2 kg/m² BMI reduction or lower salt consumption. Model performance remained robust across sex, age, and region.
Interpretable ML combined with explainable AI offers a scalable, equitable, and clinically useful tool for hypertension risk prediction in Uganda. By embedding lifestyle insights and simulating real-life interventions, the model empowers both clinicians and citizens with actionable knowledge. It supports the Ugandan National NCD Strategy (2021–2026) and the WHO Global Digital Health Framework, offering a vital step toward AI-enabled, preventive public health in low- and middle-income countries.
Menu