INTRODUCTION:
Sub-saharan africa faces a growing burden of non-communicable diseases alongside persistent human immunodeficiency virus epidemics. We quantified shared and outcome-specific spatiotemporal structure linking human immunodeficiency virus prevalence and non-communicable disease mortality across the economic community of west african states, east african community and southern african development community 2000–2019.

METHODS:
We fitted a joint bayesian spatiotemporal model to four cause-specific mortality counts (cardiovascular diseases, diabetes mellitus, respiratory diseases and malignant neoplasms) and human immunodeficiency virus prevalence. Mortality was modelled with negative binomial likelihoods and exposure offsets; prevalence with a beta likelihood and logit link. Covariate effects were outcome-specific via interactions with standardised predictors. Spatial structure used a shared besag–york–mollié reparameterisation (BYM2) field plus outcome-specific BYM2 fields; temporal structure used a shared first-order random walk (RW1) plus bloc-specific outcome RW1 terms. Inference used integrated nested laplace approximation. Residual co-clustering was summarised using bloc-level correlations between human immunodeficiency virus and outcome-specific spatial and temporal effects. Predictive performance was assessed with watanabe–akaike information criterion, conditional predictive ordinates and probability integral transform diagnostics compared with univariate benchmarks.

RESULTS:
Shared spatial structure persisted across blocs after covariate adjustment. Disease-specific residual patterns aligned with human immunodeficiency virus differently by disease and bloc including both positive and negative correlations. Temporal co-movement also varied by bloc and outcome. Joint modelling improved or matched out-of-sample performance for several outcomes and showed generally acceptable calibration with outcome–bloc-specific deviations.

CONCLUSIONS:
Joint poisson–beta spatiotemporal modelling separates common geographic drivers from disease-specific residual risks and reveals bloc-dependent co-clustering of human immunodeficiency virus and non-communicable disease burden supporting targeted geography-specific integration of services.