Submission ID 129792
| Issue/Objective | Type 2 Diabetes Mellitus (T2DM) poses a rapidly escalating public health challenge in Sub-Saharan Africa (SSA), including Cameroon, driven by urbanization, lifestyle transitions, and limited early detection capabilities. Traditional diagnostic approaches often fail to integrate multifaceted risk factors, leading to high rates of undiagnosed cases and poor glycemic control. Machine learning (ML) offers promise for enhanced prediction and personalized management by leveraging clinical, behavioural, demographic, and socioeconomic data, yet applications in resource-limited African settings remain underexplored. This review aims to synthesize recent evidence (2020-2025) on ML techniques for T2DM prediction and management in SSA, with particular emphasis on Cameroon, to identify effective models, methodological barriers, and future directions for contextually appropriate implementation. This work addresses the rising burden of Type 2 Diabetes Mellitus in Sub-Saharan Africa, with a focus on Cameroon, where health systems face significant resource constraints. By applying machine learning approaches for early prediction and management, the study demonstrates how innovative data-driven tools can strengthen prevention, improve equity in access to care, and guide policy for non-communicable diseases. The framework is scalable across the region, aligning with the conference's mission to advance global health through technology and artificial intelligence governance, evidence-based solutions, and context-specific strategies. |
|---|---|
| Methodology/Approach | Following PRISMA 2020 guidelines, we searched databases including Google Scholar, Web of Science, Springer, Scopus, and ScienceDirect for English-language studies published between 2020 and 2025. Keywords included "T2DM," "Machine Learning," "predictive models," and "Challenges," with Boolean operators. Inclusion focused on original research using ML algorithms for T2DM prediction in African populations or datasets. The study is expected to deliver validated machine learning models capable of accurately predicting early risk of Type 2 Diabetes Mellitus in diverse Cameroonian populations. Implementation in clinical and community settings will improve screening coverage, reduce late-stage diagnoses, and support targeted prevention strategies. By informing national policy and strengthening health systems, the project will contribute to equitable, scalable solutions for non-communicable disease management across Sub-Saharan Africa. |
| Results | Supervised ML models, particularly Random Forest and XGBoost, demonstrated strong performance in SSA cohorts (accuracies of 80-91%, AUC=0.96 in Kenyan data; 90% in Ethiopian hospital records). XGBoost frequently outperformed other models, and large-scale ML prediction studies in Cameroon were scarce. Key challenges included data scarcity, inconsistent records, infrastructural limitations, model generalizability, and lack of interpretability. |
| Discussion/Conclusion | ML holds substantial potential to improve early T2DM prediction and management in SSA by surpassing traditional methods, especially with locally trained models. The research contributes to the evidence base on non‑communicable disease surveillance in Sub‑Saharan Africa, highlighting the value of predictive analytics for guiding prevention and treatment strategies.It resonates with sub‑themes on health system strengthening, digital health innovations, and addressing the rising burden of non‑communicable diseases in low‑resource settings. |
| Presenters and Affiliations | EBONZILLE ELVIS EWANE Sharda University Dr. Sanam Preet Kour Sharda University |
