Application of Machine Learning in Hypertension Research and Management in Nigeria: A Systematic Review

Background Hypertension is prevalent in Nigeria and remains undetected and poorly controlled. Machine learning (ML) models provide tools for risk prediction and decision support, but the quality, applicability, and utility of Nigeria-specific ML work are unclear. This review was conducted to systematically synthesize primary studies that utilized ML for prediction, classification, management, or programme evaluation of hypertension across Nigerian populations. Methods We followed PRISMA 2020 guidelines. Searches were conducted in databases and institutional repositories (2010–Oct 2025). Eligible studies employed Nigerian data, described ML approaches, and presented at least one performance metric. Risk of bias was appraised using PROBAST. Results were narratively synthesized due to heterogeneity. Results Six primary studies were eligible for inclusion (single-site workplace, clinic, university datasets, and community surveys). The sample sizes ranged from n = 32 to n = 1,723. ML methods included artificial neural networks (ANN/MLP), decision trees (ID3/C4.5/CART), random forest, XGBoost, SVM, k-NN, and k-means clustering.. Reported internal test accuracies varied widely from ≈ 65% to 93%, with typical community-based models demonstrating an accuracy of ~ 70–75% and showed modest discrimination when AUC was reported (~ 0.7). The main limitations were very small, single-center datasets, varying outcome definitions, incomplete reporting of preprocessing and tuning, inappropriate metric use, and a lack of external validation. The overall risk of bias ranged from some concerns to high across studies. Conclusion Nigerian ML research on hypertension demonstrates feasibility using routine predictors (age, BMI, diabetes, family history); however, it is in its nascent stage and lacks methodological rigor. Actions to be prioritized are the establishment of centralized or federated Nigerian health data repositories, the adoption of standardized reporting and validation (including/external validation), capacity building in ML techniques, and investment in multisite, prospective studies to facilitate rigorous, generalizable models suitable for clinical or public-health deployment.