Asian Journal of Research in Nursing and Health

Background: Diabetes is spreading fast in West Africa, but the available evidence has been limited due to the lack of proper treatment of outliers, nonlinear dynamics, and uncertainty in epidemiological estimates.

Purpose: To present a sound, uncertainty-conscious, multilevel study of the prevalence, inequality, and treatment disparity of diabetes in West Africa (1990-2022).

Methods: A longitudinal panel dataset (country x year x age x sex) was processed by a strong and reproducible framework. The sensitivity to outliers was measured by Z-score, IQR, and MAD with winsorisation and model performance was measured by cross-validation and information criteria (AIC/BIC). Linear mixed-effects models, generalised additive models, and beta mixed regression were used to model prevalence dynamics. Cross-country heterogeneity was captured using random effects. Sigma dispersion, coefficient of variation, Gini, and Theil indices were used to measure inequality and convergence. We modelled the treatment gaps and uncertainty using a two-part approach and simulation-based inference.

Findings: MAD approach detected more outliers (11.55) and improved model fit, and cross-validation confirmed the stability of estimates. The prevalence of diabetes has increased with time (β = 0.019), in women and in older age groups. There were nonlinear dynamics in time and a large country-wide heterogeneity. There was partial convergence of inequality, with decreasing relative but level absolute inequality. There were no improvements in treatment coverage relative to prevalence, which means that there were growing gaps.

Conclusion: Diabetes burden in West Africa is growing nonlinearly, with its inequalities and lack of treatment growth. Policy should be reinforced with early detection, fair access to treatment, and age- and gender-sensitive approaches to reach SDG 3 and minimise inequalities (SDG 10).