Climate, Spatial Clustering and Hotspots of Non-Communicable Disease Mortality in Sub-Saharan Africa: A Bayesian Spatial Epidemiology Study, 2000–2019

Climate, Spatial Clustering and Hotspots of Non-Communicable Disease Mortality in Sub-Saharan Africa: A Bayesian Spatial Epidemiology Study, 2000–2019 | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Climate, Spatial Clustering and Hotspots of Non-Communicable Disease Mortality in Sub-Saharan Africa: A Bayesian Spatial Epidemiology Study, 2000–2019 Tsikai Solomon Chinembiri¹, Godfrey Pachavo This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8250412/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 8 You are reading this latest preprint version Abstract Background Non-communicable diseases (NCDs) now account for a growing share of premature mortality in sub-Saharan Africa (SSA), yet little is known about how climate and geography shape spatial inequalities in NCD deaths. Methods We assembled a country–year panel for forty-one SSA countries from 2000–2019, combining World Health Organization mortality estimates for four major NCD groups namely cardiovascular diseases, diabetes mellitus, malignant neoplasms and chronic respiratory diseases, with mid-year population denominators, climate surfaces (mean temperature and precipitation from WorldClim v2.1) and macro-socioeconomic covariates. Expected deaths were derived from age-standardised mortality rates and used as offsets in disease-specific Bayesian Poisson spatio-temporal models with Besag–York–Mollié 2 (BYM2) spatial random effects, first-order random walk temporal effects, and country–year interaction terms. Models were fitted in INLA with penalised complexity priors. We mapped climate-adjusted spatial random effects, identified multi-disease hotspots using posterior relative risks averaged over 2015–2019, and quantified cross-disease spatial correlations at SSA and regional-bloc level. Predictive performance was assessed using conditional predictive ordinates, PIT histograms and observed-versus-fitted plots. Results After adjustment for socioeconomic, health-system and HIV indicators, substantial residual spatial clustering remained, with elevated NCD mortality in parts of southern and eastern Africa and lower risk in several Sahelian countries. Countries classified as hotspots for one NCD often exhibited raised risks for others. Spatial correlations were positive between cardiovascular disease and diabetes, but negative between cardiovascular and respiratory mortality. Higher long-term temperature was associated with lower diabetes and respiratory mortality, while precipitation effects were generally weak. Conclusions NCD mortality in SSA displays marked, climate-adjusted spatial heterogeneity and partially shared hotspot patterns across causes. These findings support geographically targeted, climate-sensitive NCD prevention and health-system strengthening strategies. Trial registration: Not applicable. Non-communicable diseases Spatial epidemiology Bayesian spatio-temporal modelling Climate variability Sub-Saharan Africa Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Non-communicable diseases (NCDs), principally cardiovascular diseases, cancers, diabetes and chronic respiratory diseases now account for roughly three-quarters of global deaths, with a rapidly rising contribution in low- and middle-income countries [ 1 ]. Sub-Saharan Africa (SSA) is undergoing a profound epidemiological transition, in which persistent infectious disease burdens increasingly coexist with growing NCD mortality, creating a “double burden” for already constrained health systems [ 1 ]. Recent Global Burden of Disease (GBD) assessments suggest that NCD mortality rates in SSA have stagnated or declined more slowly than in other regions, especially for cardiovascular disease, diabetes and some cancers, threatening progress towards Sustainable Development Goal (SDG) targets on premature NCD mortality [ 2 ]. Concurrently, SSA is experiencing rapid urbanisation, demographic growth and climatic change. Urban growth is reshaping exposure to behavioural and environmental NCD risk factors such as diet, physical inactivity, air pollution and heat stress [ 2 ]. Climate change is projected to increase the frequency and intensity of extreme heat events and alter rainfall regimes, with important implications for cardiovascular and respiratory morbidity and mortality [ 3 ]. Yet, despite increasing recognition that NCDs and climate are intertwined, large-scale, spatially explicit analyses of how NCD mortality patterns relate to climatic and socio-demographic gradients in SSA remain scarce. A growing body of work has quantified NCD burden in SSA using GBD estimates and national vital statistics, highlighting heterogeneities in levels and trends across countries and within regional economic blocs [ 4 ]. However, most analyses treat each NCD cause independently and rely on non-spatial statistical models, limiting the ability to identify shared high-risk areas or to quantify spatial dependence in mortality risk. Spatial epidemiology and Bayesian disease-mapping approaches provide tools to characterise geographic clustering of health outcomes and to borrow strength across neighbouring areas [ 5 ]. Conditional autoregressive (CAR) models and their extensions, such as the re-parameterised Besag–York–Mollié model (BYM2), allow decomposition of area-level risk into structured (spatially correlated) and unstructured components, while penalised complexity (PC) priors improve interpretability and regularisation [ 6 ]. The Integrated Nested Laplace Approximation (INLA) provides a computationally efficient framework for fitting such latent Gaussian models in large datasets. Evidence linking climatic factors to NCD outcomes has expanded rapidly. Multi-country studies show that non-optimal temperature contributes substantially to cardiovascular and respiratory mortality worldwide, with particularly high relative risks at extreme heat and cold [ 7 ]. Urban heat islands, air pollution and socio-economic vulnerability can amplify these effects. Nevertheless, empirical analyses in SSA remain limited by sparse ground data and incomplete civil registration, and few studies have examined whether the spatial structure of NCD mortality is shared across multiple disease groups or modified by climatic conditions. Three gaps are particularly relevant for SSA. First, there is limited understanding of whether major NCDs share common spatial risk patterns such as overlapping hotspots of high mortality, once differences in age structure and baseline rates are accounted for. Second, the extent to which these shared or divergent spatial patterns vary across regional economic communities (East African Community [EAC], Economic Community of West African States [ECOWAS] and Southern African Development Community [SADC]) remains under-explored. Third, while climatic variables such as long-term mean temperature and precipitation can now be characterised at high spatial resolution using global products like WorldClim 2.1 [ 8 ], they have rarely been integrated into Bayesian spatio-temporal NCD models for SSA. Addressing these gaps is essential for regional planning. If the same countries emerge as hotspots across multiple NCDs, this would support integrated prevention and health-system strengthening strategies. Conversely, weak or negative spatial correlations between causes would argue for disease-specific targeting. Understanding whether climate-adjusted residual risk clusters remain after controlling for socio-economic and health-system covariates could help identify areas where unobserved factors such as local health care quality, environmental exposures or social vulnerability drive excess mortality. This study aims to quantify and compare the spatial patterns of mortality from four major NCD groups namely cardiovascular diseases, diabetes mellitus, malignant neoplasms and chronic respiratory diseases, in forty-one countries of SSA between 2000 and 2019. Specifically, we address two research questions: To what extent do these four NCD groups share common spatial risk patterns across SSA and within the EAC, ECOWAS and SADC blocs after adjusting for demographic, socio-economic, health-system, HIV and climatic covariates? Where are persistent climate-adjusted hotspots of excess NCD mortality, and how do they align across disease groups? Methodologically, we develop a consistent Bayesian spatio-temporal modelling framework in which cause-specific mortality counts are modelled using shared covariates (urbanisation, GDP per capita, health expenditure, HIV prevalence, temperature and precipitation) and disease-specific spatial and temporal random effects. The models are fitted via INLA using BYM2 spatial priors and PC hyperpriors, enabling robust inference under data sparsity and avoiding over-fitting [ 9 ]. Substantively, this is, to our knowledge, one of the first multi-disease analyses to: (i) map climate-adjusted spatial random effects for four leading NCD causes across SSA; (ii) quantify SSA-wide and bloc-specific correlations between these spatial effects; and (iii) identify multi-disease hotspots based on posterior mean relative risks averaged over recent years. By linking region-specific NCD mortality patterns with climatic and socio-economic gradients, the study provides actionable evidence for regional NCD control and climate-resilient health planning. 2. Methodology 2.1 Study area and variable definitions The study covers forty-one countries in SSA (Fig. 1 ) that are members of at least one of three regional economic communities: EAC, ECOWAS and SADC. National territories are used as the areal units of analysis. A country-level shapefile was transformed to an equal-area projection, harmonised across years, and encoded with unique region identifiers (region_id) for spatial modelling. Annual cause specific deaths and age-standardised mortality rates (ASMR, per 100,000 population) for four broad NCD groups, cardiovascular diseases, diabetes mellitus, malignant neoplasms and chronic respiratory diseases, were extracted from the World health organisation (WHO) Global Health Observatory (GHO) and related country-level databases for 2000–2019 [ 2 ]. Observed deaths in country \(\:i\) , year \(\:t\) and disease group \(\:d\) are denoted \(\:{Y}_{itd}\) . Expected deaths \(\:{E}_{itd}\) were derived by combining mid-year population estimates \(\:{P}_{itd}\) with ASMR \(\:{R}_{itd}\) (per 100,000 population) for the same country, year and disease group according to $$\:Eᵢₜd\:=\:(Pᵢₜd\:/\:\text{100,000})\:\times\:\:Rᵢₜd$$ 1 , where \(\:Pᵢₜd\) reflects the population at risk and \(\:Rᵢₜd\) is based on WHO’s standard world population. The ratio \(\:{Y}_{itd}/{E}_{itd}\) defines a crude standardised mortality ratio (SMR). For modelling, we use the natural logarithm \(\:Eᵢₜd\) as an offset to ensure that fitted risks are interpreted as relative risks (RRs) compared to a reference with ASMR \(\:{R}_{itd}\) [ 10 ]. 2.2 Data sources and covariates Socio-economic and health-system covariates were compiled from the World Bank’s World Development Indicators (WDI) and related sources [ 11 ]. For each country and year, we extracted gross domestic product per capita (GDP_pc, constant US dollars), current health expenditure per capita (HExp), and the proportion of the population living in urban areas (Urban%). HIV prevalence among adults aged 15–49 years was obtained from UNAIDS and harmonised with the mortality series[ 2 ](World Bank Group, 2024). These indicators are expressed as log-transformed values (for GDP_pc and HExp) or standardised z-scores (for Urban% and HIV prevalence) to aid convergence and interpretation. Climatic covariates were derived from WorldClim v2.1, which provides 30-arc-second (~ 1 km) gridded surfaces of long-term monthly climate normals for 1970–2000 [ 8 ]. We extracted country-level averages of mean annual temperature (°C) and total annual precipitation (mm) by overlaying the national polygons on the gridded surfaces and computing area-weighted means. Although WorldClim normals pre-date the 2000–2019 mortality period, they capture persistent spatial gradients in climate that are relevant for chronic exposure profiles and long-term adaptation. Both temperature and precipitation were centred and scaled prior to analysis. The final analytic dataset therefore comprises, for each combination of country, year and disease category: observed and expected deaths, ASMR, bloc membership (EAC, ECOWAS, SADC), and covariates representing urbanisation, economic development, health expenditure, HIV prevalence, mean temperature and precipitation. 2.3 Spatio-temporal model formulation For each disease group d, we model the annual observed deaths \(\:{Y}_{itd}\) in country \(\:i\) and year \(\:t\) using a Poisson log-linear model: where \(\:{E}_{itd}\) is the expected death count from (1) and \(\:{\theta\:}_{itd}\) is the relative risk. The log-risk is decomposed as $$\:\text{log}{\theta\:}_{itd}={\alpha\:}_{d}+{\beta\:}^{T}{\varvec{X}}_{it}+{u}_{id}+{v}_{id}+{\gamma\:}_{td}+{\delta\:}_{itd}$$ 3 , Where \(\:{\alpha\:}_{d}\) is a disease-specific intercept; \(\:{\varvec{X}}_{it}\) is the vector of centred covariates with corresponding regression coefficients \(\:\beta\:\) shared across diseases; \(\:{u}_{id}\) and \(\:{v}_{id}\) are, respectively, structured and unstructured spatial random effects for disease \(\:d\) in country \(\:i\) ; \(\:{\gamma\:}_{td}\) is a disease-specific temporal random effect for year t; and \(\:{\delta\:}_{itd}\) is a space-time interaction term capturing residual heterogeneity. Spatially structured components \(\:{u}_{id}\) are assigned a BYM2 prior, which combines an intrinsic conditional autoregressive (ICAR) field and independent and identically distributed ( \(\:i.i.d\) ). Gaussian component with a mixing parameter \(\:\varphi\:\) that determines the proportion of spatially structured variance [ 12 ](Lee et al., 2018). Spatial adjacency is encoded via a symmetrical k-nearest-neighbours graph (k = 4) derived from country centroids. Temporal random effects \(\:{\gamma\:}_{td}\) follow a first-order random walk (RW1) process over calendar years to smooth trends [ 13 ](Banerjee et al., 2003). The interaction term \(\:{\delta\:}_{itd}\) is modelled as \(\:i.i.d\) Gaussian noise. To explore shared spatial structure across diseases, we fit two sets of models: Disease-specific climate-adjusted models. Each disease d is modelled separately using (2)-(3), allowing estimation of disease-specific spatial random effects \(\:{u}_{id}+{v}_{id}\) . These are then mapped and compared. Joint multivariate model. All four diseases are stacked into a single dataset with a disease index \(\:d=1,\:\dots\:,\:4\) . The model retains disease-specific intercepts and replicates the spatial and temporal random effect structures across diseases, enabling partial pooling while allowing patterns to differ by disease [ 14 ](Mahaki et al., 2011). From the fitted disease-specific models, we compute: (i) posterior mean spatial random effects for each disease and country; (ii) SSA-wide and bloc-specific correlation matrices of these effects to quantify shared spatial risk; and (iii) posterior mean relative risks averaged over 2015–2019 to identify recent hotspots. 2.4 Prior specification and computation We adopt penalised complexity (PC) priors for hyperparameters governing the variance of spatial, temporal and interaction random effects, following recommendations for disease mapping (Simpson et al., 2017). For each precision parameter \(\:\tau\:\) of a Gaussian randomeffect (variance \(\:1/\tau\:\) , we specify a PC prior such that $$\:\text{Pr}\left(\sigma\:>{\sigma\:}_{o}\right)={p}_{0},\:\sigma\:=1/\tau\:,$$ 4 With \(\:{\sigma\:}_{0}=1\) and \(\:{p}_{0}=0.01\) , implying string prior belief that the standard deviation is small but allowing heavier tails. For the BYM2 mixing parameter \(\:\varphi\:\) , we use a PC prior favouring a balanced contribution of structured and unstructured components while avoiding degeneracy at the boundaries [ 5 ] All models are fitted using the Integrated Nested Laplace Approximation (INLA) as implemented in the R–INLA package, which provides accurate and fast approximations to posterior marginals in latent Gaussian models [ 3 ]. 2.5 Model diagnostics and predictive calibration Model adequacy is evaluated using a combination of global fit indices and posterior predictive checks. For each fitted model we report DIC and WAIC, with lower values indicating better trade-off between fit and complexity [ 13 ]. Conditional predictive ordinates are examined for numerical failures; counts of CPO failures close to zero indicate stable approximations. Predictive calibration is assessed using the probability integral transform (PIT). For each observed count \(\:{Y}_{itd}\) , the PIT value is computed from the posterior predictive distribution; histograms of PIT values should be approximately uniform under a well-calibrated model [ 13 ]. Furthermore, we compare observed deaths with the posterior mean of fitted values on a country-year basis. Scatter plots of observed versus predicted counts, overlaid with a 45° line and harmonised axes, provide a visual check for systematic under- or over-prediction, especially in countries with large populations and high mortality. 3. Results 3.1 Climate-adjusted spatial random effects and residual clustering of NCD mortality The spatial random effects as depicted in Fig. 2 , reveal substantial residual geographical structuring of NCD mortality risk that is not explained by bloc membership, socioeconomic indicators, HIV prevalence or climate covariates. For cardiovascular diseases, a pronounced high-risk cluster is visible in parts of south-eastern Africa, with neighbouring countries showing moderately elevated residual risk. Several western and central countries display neutral or slightly negative spatial effects, suggesting that once observed covariates are accounted for, their cardiovascular mortality is close to or below the regional average. Diabetes and respiratory diseases exhibit more moderate, but still heterogeneous, residual spatial patterns. For diabetes, a band of mildly elevated effects appears in selected central and southern countries, whereas much of West Africa shows near-null or slightly protective effects (Fig. 2 ). Chronic respiratory diseases show distinct high-risk foci in parts of the southern region, consistent with the hotspot patterns seen in the averaged relative risk maps, while many other countries have small or negative spatial effects, implying that observed covariates capture a larger share of the between-country variation. In contrast, malignant neoplasms display comparatively weak spatial structure, with most countries falling in the central categories of the legend. This suggests that, at the scale of national averages, residual cancer mortality risk is less spatially clustered than for cardiovascular or respiratory causes. Overall, the maps indicate that even after adjusting for climate and socioeconomic determinants, there remain geographically coherent pockets of excess NCD mortality, pointing to unmeasured local risk factors, differences in health-system performance, or data quality issues that warrant targeted investigation. 3.2 Persistent spatial hotspots of excess NCD mortality (posterior (RR), 2015–2019) Averaging model-based relative risks over the most recent five-year period highlights a pattern of widespread, persistent excess NCD mortality across the region rather than isolated, short-lived spikes (Fig. 3 ). For cardiovascular diseases, several southern and eastern African countries maintain consistently elevated RRs compared with the continental distribution, whereas parts of the western Sahel appear relatively less affected. Malignant neoplasm mortality similarly exhibits sustained excess risk in a band spanning southern Africa and pockets of West Africa, suggesting shared underlying determinants such as late presentation, limited diagnostic capacity and constrained access to cancer treatment. In contrast, as Fig. 3 clearly demonstrates, diabetes and chronic respiratory disease display more uniformly high relative risks across most of sub-Saharan Africa, indicating that metabolic and chronic respiratory hazards are now deeply entrenched regional problems rather than confined to a few high-income or rapidly urbanising settings. Taken together, these hotspot maps reinforce the view that NCD mortality in the late 2010s is both geographically clustered and simultaneously widespread, underscoring the need for region-wide strengthening of prevention and chronic care, with targeted intensification in countries that consistently occupy the highest risk categories. 3.3 SSA-wide and bloc-specific cross-disease spatial correlations of residual NCD risk As illustrated in Table 1 , the joint spatial correlation analysis shows that cardiovascular diseases and diabetes mellitus share a moderately similar residual spatial pattern across SSA (ρ = 0.49), with this correlation strengthening in SADC (ρ = 0.70) and ECOWAS (ρ = 0.71). This suggests that, after accounting for measured covariates, unobserved spatial determinants such as diet, metabolic risk factors or health-system performance cluster geographically in ways that simultaneously elevate cardiometabolic mortality in many countries within these blocs. In contrast, correlations between cardiovascular diseases and malignant neoplasms are weak (SSA-wide ρ = 0.13) (Table 1 ), indicating more distinct spatial structuring of cancer mortality, potentially reflecting differences in screening, diagnostic capacity and cancer-specific risk factors. Table 1 Pairwise spatial correlations between disease-specific spatial random effects from Bayesian spatio-temporal models, overall Sub-Saharan Africa (SSA-wide) and by regional economic bloc. Values are Pearson correlation coefficients between posterior mean spatial random effects for each disease pair, summarising the similarity of residual spatial patterns after adjusting for socioeconomic, health-system and climatic covariates. Positive values indicate that countries with elevated residual risk for one disease also tend to have elevated risk for the other; negative values indicate spatial divergence in residual risk. Disease pair SSA-wide SADC ECOWAS EAC Cardiovascular diseases vs. Diabetes mellitus 0.49 0.7 0.71 -0.13 Cardiovascular vs. Malignant neoplasms 0.13 -0.39 -0.23 -0.18 Cardiovascular vs. Respiratory diseases -0.56 0.54 0.12 -0.76 Diabetes mellitus vs. Malignant neoplasms 0.1 -0.35 0.23 -0.11 Diabetes mellitus vs. Respiratory diseases 0.19 0.88 0.12 0.54 Malignant neoplasms vs. Respiratory diseases -0.17 -0.23 0.08 0.03 The strongest positive association is observed between diabetes mellitus and respiratory diseases in SADC (ρ = 0.88), pointing to substantial overlap in residual spatial risk, which may relate to co-occurring urban air pollution, tobacco use or shared health-system constraints. Conversely, cardiovascular and respiratory diseases exhibit a moderate negative correlation at the SSA scale (ρ = − 0.56), driven largely by a strong negative correlation in the EAC (ρ = − 0.76), where countries with elevated residual respiratory risk tend to have comparatively lower residual cardiovascular risk (Table 1 ). Consequently, these patterns imply that while some NCDs share common unmeasured spatial determinants, others are governed by more disease-specific or regionally distinct processes, underscoring the need for bloc-tailored, disease-specific prevention and control strategies rather than a one-size-fits-all regional approach. 3.4 Association between climate variability and non-communicable disease mortality After adjustment for socioeconomic, health-system and HIV covariates, climate effects on NCD mortality were generally modest (Table 2 ). For cardiovascular diseases and malignant neoplasms, the posterior IRRs for both mean temperature and precipitation were close to the null with wide credible intervals that included 1. For example, a 1-SD increase in mean temperature was associated with an IRR of 1.13 (95% CrI 0.92–1.38) for cardiovascular deaths and 1.01 (95% CrI 0.87–1.18) for cancer mortality, indicating little evidence of a consistent climate signal for these outcomes. By contrast, mean temperature showed clear inverse associations with diabetes and respiratory disease mortality. For diabetes, each 1-SD increase in temperature was associated with a 13% lower mortality risk (IRR 0.87, 95% CrI 0.77–0.98). The association was stronger for chronic respiratory diseases, where the IRR was 0.74 (95% CrI 0.60–0.92), suggesting substantially higher respiratory mortality in cooler settings after accounting for other covariates and residual spatial and temporal structure. For precipitation, all IRR estimates lay close to unity with credible intervals spanning 1 across all four NCD groups (e.g. respiratory diseases IRR 0.97, 95% CrI 0.84–1.12), indicating no robust evidence that variation in average annual rainfall was independently associated with NCD mortality over the study period. Overall, these results suggest that temperature may play a selective role in shaping diabetes and respiratory mortality risks in sub-Saharan Africa, whereas long-term mean precipitation appears less influential once other determinants are controlled for. Table 2 Climate covariate effects from Bayesian spatio-temporal models for four major NCDs in sub-Saharan Africa, 2000–2019. Posterior fixed-effect estimates (β) and 95% credible intervals (CrI) are shown on the log relative-risk scale, together with corresponding incidence rate ratios (IRR) and 95% CrI. Temperature and precipitation are standardised (per 1-SD increase in annual mean temperature or precipitation), and all models adjust for GDP per capita, health expenditure, urbanisation, HIV prevalence and spatio-temporal random effects. Disease Covariate log RR β (95% CrI) IRR (95% CrI) Cardiovascular diseases Mean temperature (SD-scaled) 0.121 (-0.082, 0.323) 1.13 (0.9–1.38) Cardiovascular diseases Precipitation (SD-scaled) -0.014 (-0.152, 0.125) 0.99 (0.86–1.13) Diabetes mellitus Mean temperature (SD-scaled) -0.137 (-0.257, -0.018) 0.87 (0.77–0.98) Diabetes mellitus Precipitation (SD-scaled) -0.021 (-0.103, 0.061) 0.98 (0.90–1.06) Malignant neoplasms Mean temperature (SD-scaled) 0.011 (-0.144, 0.167) 1.01 (0.87–1.18) Malignant neoplasms Precipitation (SD-scaled) -0.044 (-0.152, 0.063) 0.96 (0.86–1.06) Respiratory diseases Mean temperature (SD-scaled) -0.297 (-0.508, -0.086) 0.74 (0.60–0.92) Respiratory diseases Precipitation (SD-scaled) -0.034 (-0.180, 0.111) 0.97 (0.84–1.12) In all climate-extended models, temperature and precipitation effects were estimated conditional on a common set of macro-level covariates: log GDP per capita, log health expenditure, urbanisation rate and HIV prevalence. Across the four disease groups, higher GDP per capita and health expenditure were generally associated with higher reported NCD mortality, consistent with a combination of more advanced epidemiological transition, improved case ascertainment and competing-risk patterns in better-resourced health systems, rather than genuinely protective effects of under-investment. Urbanisation and HIV prevalence showed disease-specific patterns, with positive associations particularly evident for diabetes and chronic respiratory disease. Against this background, the temperature and precipitation coefficients in Table 2 can be interpreted as climate associations adjusted for major socio-economic and health-system gradients, rather than simple ecological correlations. 3.4.1 Adjustment for socioeconomic, health-system and HIV covariates All climate coefficients reported above are estimated from multivariable Bayesian spatio-temporal Poisson models that simultaneously adjust for national income (log GDP per capita), per-capita health expenditure, urbanisation rate and HIV prevalence. Across the four disease groups, higher GDP per capita tended to be associated with modestly elevated NCD mortality, consistent with a shift towards more obesogenic and cardiometabolic risk profiles at higher income levels, whereas higher health expenditure showed weakly protective or null associations. Urbanisation exhibited heterogeneous effects by cause, with positive associations for diabetes and cancer and more mixed patterns for cardiovascular and respiratory mortality, suggesting that the balance between improved access to care and increased exposure to urban risk environments varies across conditions. HIV prevalence was positively associated with cardiovascular and diabetes mortality and less strongly related to cancer and chronic respiratory disease, reflecting known interactions between HIV, antiretroviral therapy and cardiometabolic risk. These adjustment variables are included primarily to reduce confounding of climate effects by broad socioeconomic and health-system gradients. 3.5 Model diagnostics and calibration CPO failure counts were zero for all disease-specific models, indicating numerically stable computation of conditional predictive ordinates. As illustrated in Fig. 4 , posterior predictive PIT histograms were approximately symmetric with no pronounced U-shaped or skewed patterns, suggesting adequate predictive calibration. While formal KS and χ² tests rejected exact uniformity (p < 0.001 across models), this likely reflects the very large number of observations rather than substantive miscalibration. Taken together, these diagnostics indicate that the models provide stable predictions with only mild deviations from perfect calibration. 4. Discussion policy implications and conclusions 4.1 Spatial random effects and residual clustering of NCD mortality We found clear evidence of residual spatial clustering in NCD mortality across sub-Saharan Africa (SSA) after adjusting for age structure, expected deaths and macro-level covariates. The BYM2 spatial random effects revealed coherent high- and low-risk regions for cardiovascular diseases, diabetes, cancers and chronic respiratory diseases, rather than purely random geographic noise. This is consistent with earlier work showing that NCD burdens in SSA and other low- and middle-income regions are shaped by spatially structured determinants such as access to care, health system capacity and built environment, which are not fully captured by national averages of income or health spending [ 16 , 17 ]. The persistence of spatial structure after covariate adjustment suggests that “where you live” still matters for NCD survival in SSA, even conditional on broad socio-economic context. Methodologically, the separation between structured and unstructured components in the BYM2 model, alongside reasonable values of the spatial mixing parameter, supports our use of penalised-complexity (PC) priors and latent Gaussian disease mapping for this setting [ 7 , 15 , 18 ]. 4.2 Persistent hotspots of excess mortality Averaging posterior relative risks over 2015–2019 highlighted persistent multi-year hotspots of excess mortality across all four NCD groups. Several countries in southern and western SSA consistently exhibited relative risks above one for multiple causes, whereas other countries tended to lie below the regional baseline. Because these hotspot estimates borrow strength in space and time, they are unlikely to reflect random annual fluctuations and instead point to structural differences in chronic care systems, risk factor environments and diagnostic capacity. For chronic respiratory diseases, for example, high-risk clusters may reflect overlapping burdens of household and occupational air pollution, tobacco use and post-tuberculosis lung damage, compounded by limited spirometry and chronic care services [ 19 , 20 ]. These multi-disease hotspots therefore provide a pragmatic starting point for regional targeting of hypertension and diabetes screening, essential medicines provision and integrated chronic care in the countries most consistently above the regional norm. 4.3 Cross-disease spatial correlations at SSA and bloc level The cross-disease correlation matrices showed that spatial patterns of NCD mortality are partly shared, but not uniform, across causes. At SSA level, we observed a moderate positive correlation between cardiovascular and diabetes spatial effects, and weaker positive correlations with cancer, consistent with a shared cardiometabolic and oncologic risk environment that includes obesity, diet, tobacco and delayed diagnosis [ 1 , 16 ]. In contrast, spatial effects for chronic respiratory disease were negatively or only weakly correlated with those of the other causes, suggesting that respiratory mortality is driven more by distinct exposures such as biomass smoke, occupational hazards and tuberculosis sequelae [ 19 , 20 ]. Bloc-specific correlation matrices revealed further heterogeneity: in some regional economic communities, diabetes and respiratory mortality patterns were strongly aligned, whereas elsewhere cardiovascular and respiratory patterns diverged. These differences are plausibly related to bloc-level variation in tobacco control, energy use, urban air quality, HIV burden and integration of chronic care services [ 21 ]. This heterogeneity underlines the importance of not assuming a single, continent-wide geography of NCD risk. 4.4 Climate variability and NCD mortality After adjustment for GDP per capita, health expenditure, urbanisation and HIV prevalence, climate variables showed modest and disease-specific associations with mortality. Higher mean temperature (per standard deviation) was associated with lower diabetes and chronic respiratory mortality, whereas effects on cardiovascular disease and cancer were small and imprecise; precipitation showed no consistent association with any cause. Taken at face value, these results suggest that long-term climatic gradients captured by country-level WorldClim normals are not the dominant drivers of cross-country NCD mortality differences in SSA during the study period. However, these findings must be interpreted in light of the broader climate–health literature. Multi-country time-series and case-crossover studies show that both heat and cold can increase short-term risks of cardiovascular and respiratory death, particularly among vulnerable groups [ 22 , 23 ]. Our analysis, by contrast, uses long-term climatic averages at the country scale and cannot capture heatwaves, intra-annual variability, or urban heat islands. The apparent protective association of higher temperature for diabetes and respiratory mortality is therefore more likely to reflect residual confounding by unmeasured factors that co-vary with climate (for example, altitude, urban form or service availability) than a true protective effect. Nonetheless, the absence of strong harmful associations after extensive adjustment suggests that, at present, social and health-system determinants remain the primary drivers of cross-country inequality in NCD mortality in SSA, with climate acting as a slower-moving background modifier. To situate these climate findings within the broader determinants, it is important to recall that the climate-extended models also included shared socio-economic and health-system covariates. Across the four disease groups, higher GDP per capita and health expenditure tended to be associated with higher observed mortality, which likely reflects a mix of more advanced epidemiological transition, better certification and coding of cause of death, and competing-risk structures in better-resourced systems, rather than a causal effect of under-investment being protective. Urbanisation and HIV prevalence showed disease-specific patterns, with particularly strong positive associations for diabetes and chronic respiratory disease, consistent with urban lifestyles, air pollution and chronic HIV-related co-morbidities [ 16 , 17 ]. Against this backdrop, the temperature and precipitation coefficients can be interpreted as climate associations that are already adjusted for major socio-economic and epidemiological gradients, not simple ecological correlations. 4.5 Model diagnostics and calibration Posterior predictive diagnostics indicated that the models were generally well calibrated. CPO failure counts were effectively zero for all disease-specific models, indicating numerical stability of the INLA approximations [ 7 ]. PIT histograms were close to uniform, suggesting that posterior predictive distributions were not systematically biased or over-/under-dispersed, and observed versus posterior mean plots showed tight clustering around the identity line for most of the mortality range, with greater dispersion only in the tails. These results support the adequacy of the chosen latent Gaussian structure, BYM2 spatial random effects, RW1 temporal trends and independent region–year interactions with PC priors that penalise excessive complexity [ 15 , 18 ]. While further refinements are possible, for example through non-linear effects or alternative temporal structures, there is no strong diagnostic evidence of major structural misspecification. This increases confidence that the residual spatial patterns and hotspots we report reflect genuine signal rather than artefacts of overfitting or numerical instability. 4.6 Policy implications Several policy-relevant messages emerge from these findings. First, the identification of persistent multi-disease hotspots after adjustment for expected deaths and macro-level covariates suggests that certain countries and subregions are systematically “left behind” in terms of NCD prevention, diagnosis and treatment. These areas should be prioritised for regional initiatives to scale up affordable antihypertensive and glucose-lowering therapies, strengthen continuity of care for chronic conditions and integrate NCD management into primary care and HIV platforms [ 23 ]. Second, the moderate spatial correlations between cardiovascular, diabetes and cancer mortality indicate that investments in core health-system functions such as primary care, essential diagnostics, and tobacco and alcohol control, are likely to generate benefits across multiple NCDs rather than in isolation [ 16 , 17 ]. By contrast, the weaker and sometimes divergent spatial patterns for chronic respiratory disease point to a need for targeted action on household fuel transitions, occupational health and post-tuberculosis lung disease in specific high-burden settings [ 20 ]. Finally, although long-term climate variables were not dominant drivers of cross-country differences in this analysis, they should not be neglected. Climate change is expected to amplify extreme heat events, alter air pollution patterns and interact with food and water systems, all of which may affect future NCD trajectories in SSA [ 22 ]. Strengthening NCD services and building climate-resilient health systems are therefore complementary priorities rather than competing agendas. 5. Conclusions and future directions In summary, this Bayesian spatial epidemiology study shows that NCD mortality in SSA is characterised by marked residual spatial clustering, persistent multi-disease hotspots and partially shared spatial structures across major causes, even after adjusting for demographic, socio-economic, health-system and climatic variables. Long-term temperature and precipitation patterns showed only modest associations with mortality at the country level, whereas geography, socio-economic conditions and health-system characteristics remained the dominant correlates of cross-country inequality in NCD outcomes. Future research should extend this framework by: (i) incorporating subnational mortality or hospitalisation data to identify within-country hotspots; (ii) combining long-term climatic normals with dynamic indicators of heatwaves, drought and air pollution; and (iii) developing multivariate models that jointly analyse NCDs and infectious diseases such as HIV and tuberculosis, to better capture syndemic interactions. By integrating robust Bayesian disease mapping with climate and socio-economic gradients, the approach presented here offers a transferable template for tracking spatial convergence or divergence in NCD risk and for guiding regional prioritisation in an era of epidemiological and climatic transition. Abbreviations ASMR – Age-standardised mortality rate BYM2 – Besag–York–Mollié 2 spatial model CPO – Conditional predictive ordinate CrI – Credible interval CVD – Cardiovascular disease DIC – Deviance Information Criterion GDP – Gross domestic product HIV – Human immunodeficiency virus INLA – Integrated nested Laplace approximation IRR – Incidence rate ratio NCD – Non-communicable disease PIT – Probability integral transform RR – Relative risk RW1 – First-order random walk SADC – Southern African Development Community EAC – East African Community ECOWAS – Economic Community of West African States SSA – Sub-Saharan Africa WAIC – Watanabe–Akaike Information Criterion Declarations Ethics approval and consent to participate Not applicable. This study did not involve human participants, human data, human tissue, or animals. Consent for publication Not applicable. This manuscript does not contain data from any individual person. Availability of data and materials The datasets analysed during the current study were derived from publicly available repositories, including the World Health Organisation (WHO) Global Health Observatory as follows: https://www.who.int/data/gho ; World Development Indicators as follows: https://databank.worldbank.org/source/world-development-indicators ; and WorldClim v2.1 data for climatic factors (precipitation and rainfall) as follows: https://www.worldclim.org/ Custom scripts and code developed NCDs modelling are available from the corresponding author upon reasonable request. Competing interests The authors declare that they have no competing interests. Funding This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. Authors' contributions TSC (Tsikai Solomon Chinembiri): Conceptualized the study, designed the research framework, conducted statistical modelling and Bayesian inference, and led the manuscript writing and interpretation of climate–health linkages. GP (Godfrey Pachavo): Contributed to data acquisition, pre-processing, and interpretation of socioeconomic indicators. Acknowledgements We gratefully acknowledge the World Health Organisation (WHO) Global Health Observatory for providing access to the datasets that enabled this analysis. We also extend our thanks to colleagues and peer reviewers whose insightful feedback enhanced the manuscript. Authors' information Optional — not included. References Abbafati C, Abbas KM, Abbasi M, Abbasifard M, Abbasi-Kangevari M, Abbastabar H, Global Burden of Disease Study 2019. Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the. Lancet. 2020;396:1204–22. https://doi.org/10.1016/S0140-6736(20)30925-9/ATTACHMENT/1802C2B8-7CCC-467E-B4DD-92F466CF5E15/MMC2E.PDF . World Bank Group. The Worldwide Governance Indicators : Methodology and 2024 Update (English). Washington DC; 2024. Avard Rue H˚, Martino S, Chopin N. Approximate Bayesian Inference for Latent Gaussian Models Using Integrated Nested Laplace Approxima-tions. Wang S, Ren Z, Liu X, Yin Q. Spatiotemporal trends in life expectancy and impacts of economic growth and air pollution in 134 countries: A Bayesian modeling study. Soc Sci Med. 2022;293:114660. https://doi.org/https://doi.org/10.1016/j.socscimed.2021.114660 . Blangiardo M, Cameletti M. Spatial and Spatio-temporal Bayesian Models with R - INLA. Spatial and Spatio-temporal Bayesian Models with R - INLA. 2015;:1–308. https://doi.org/10.1002/9781118950203 Besag J, York J, Molli´e. Bayesian image restoration, with two applications in spatial statistics. Ann Inst Stat Math. 1991;43:1–59. Rue H, Martino S, Chopin N. Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations. J R Stat Soc Ser B Stat Methodol. 2009;71:319–92. https://doi.org/https://doi.org/10.1111/j.1467-9868.2008.00700.x . Fick SE, Hijmans RJ. WorldClim 2: New 1-km spatial resolution climate surfaces for global land areas. Int J Climatol. 2017;37:4302–15. https://doi.org/https://doi.org/10.1002/joc.5086 . Demirhan H, Kalaylioglu Z. Joint prior distributions for variance parameters in Bayesian analysis of normal hierarchical models. J Multivar Anal. 2015;135:163–74. https://doi.org/https://doi.org/10.1016/j.jmva.2014.12.013 . Lawson A. Bayesian Disease Mapping: Hierarchical Modeling in Spatial Epidemiology. CRC; 2018. World Bank. World Development Indicators. Washington, D.C., USA; 2023. Lee D, Rushworth A, Napier G. Spatio-Temporal Areal Unit Modeling in R with Conditional Autoregressive Priors Using the CARBayesST Package. J Stat Softw. 2018. https://doi.org/10.18637/jss.v084.i09 . 84 9 SE-Articles:1–39. Banerjee S, Carlin BP, Gelfand AE. Hierarchical modeling and analysis for spatial data. Chapman and Hall/CRC; 2003. Mahaki B, Mehrabi Y, Kavousi A, Akbari ME, Waldhoer T, Schmid VJ, et al. Multivariate disease mapping of seven prevalent cancers in Iran using a shared component model. Asian Pac J Cancer Prev. 2011;12:2353–8. Simpson D, Rue H, Riebler A, Martins TG, Sørbye SH. Penalising model component complexity: A principled, practical approach to constructing priors. 2017. Miranda JJ, Kinra S, Casas JP, Davey Smith G, Ebrahim S. Non-communicable diseases in low- and middle-income countries: context, determinants and health policy. Trop Med Int Health. 2008;13:1225–34. https://doi.org/10.1111/J.1365-3156.2008.02116.X . Gouda HN, Charlson F, Sorsdahl K, Ahmadzada S, Ferrari AJ, Erskine H, et al. Burden of non-communicable diseases in sub-Saharan Africa, 1990–2017: results from the Global Burden of Disease Study 2017. Lancet Glob Health. 2019;7:e1375–87. Riebler Andrea H, Simpson, Daniel. Rue Håvard. An intuitive Bayesian spatial model for disease mapping that accounts for scaling. Stat Methods Med Res. 2016;25:1145–65. https://doi.org/10.1177/0962280216660421 Adeloye D, Basquill C, Aderemi AV, Thompson JY, Obi FA. An estimate of the prevalence of hypertension in Nigeria: a systematic review and meta-analysis. J Hypertens. 2015;33:230–42. https://doi.org/10.1097/HJH.0000000000000413 . Amegah AK, Jaakkola JJK. Household air pollution and the sustainable development goals. Bull World Health Organ. 2016;94:215–21. https://doi.org/10.2471/BLT.15.155812 . Feinstein MJ, Bogorodskaya M, Bloomfield GS, Vedanthan R, Siedner MJ, Kwan GF, et al. Cardiovascular Complications of HIV in Endemic Countries. Curr Cardiol Rep. 2016;18:113. https://doi.org/10.1007/s11886-016-0794-x . Gasparrini A, Guo Y, Hashizume M, Lavigne E, Zanobetti A, Schwartz J, et al. Mortality risk attributable to high and low ambient temperature: a multicountry observational study. Lancet. 2015;386:369–75. https://doi.org/10.1016/S0140-6736(14)62114-0 . Watts N, Amann M, Arnell N, Ayeb-Karlsson S, Beagley J, Belesova K, et al. The 2020 report of The Lancet Countdown on health and climate change: responding to converging crises. Lancet. 2021;397:129–70. https://doi.org/10.1016/S0140-6736(20)32290-X . Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 14 May, 2026 Reviewers agreed at journal 19 Feb, 2026 Reviewers agreed at journal 24 Dec, 2025 Reviewers agreed at journal 18 Dec, 2025 Reviewers invited by journal 16 Dec, 2025 Editor assigned by journal 02 Dec, 2025 Submission checks completed at journal 02 Dec, 2025 First submitted to journal 01 Dec, 2025 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8250412","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":561613745,"identity":"4f8e17ec-f83f-473c-aa09-5ebe9fbe58ba","order_by":0,"name":"Tsikai Solomon Chinembiri¹","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAABFklEQVRIie2QMUvEMBiGvyPQLvG6plTIX/hC4U5B9K+kFJyueOBycKCFQF0E/8q5nVtLwC4FV91E1xsKtwjeYFrPRUNnhzxk+Ah5eN8vAA7Hv2SkANAc0s05A8/PhwX6V6ElQDmoAOkH7BO7ACaHlbNAK/yYA5/6B9Xby/roYhxuxLaFUw6BfrWmsEQltwjiQY3TOGvYcRHNYlZCKnJ4RLsyUiVFkKjpJMoKhl40m5hiRJqadiWoVLX7VqafvRI2nXItgfitff1EpT8ppFcY7RQtzd/ZU54TFR8iEytN48jsgh49v2QN1qKgdG5T/Lv6PdzsTjg+NWKbra+Q3+j7drFY8iCoVzZlD/t9YSp5A+8dDofDMcwXxz1QRNIdb5QAAAAASUVORK5CYII=","orcid":"","institution":"University of Cape Town, African Climate \u0026 Development Initiative (ACDI)","correspondingAuthor":true,"prefix":"","firstName":"Tsikai","middleName":"Solomon","lastName":"Chinembiri¹","suffix":""},{"id":561613746,"identity":"91995374-9521-4dca-9941-698faf390161","order_by":1,"name":"Godfrey Pachavo","email":"","orcid":"","institution":"Marondera University of Agricultural Sciences and Technology","correspondingAuthor":false,"prefix":"","firstName":"Godfrey","middleName":"","lastName":"Pachavo","suffix":""}],"badges":[],"createdAt":"2025-12-01 12:38:22","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8250412/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8250412/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":98512967,"identity":"cae152ba-29f3-4b13-92de-3e1302e33f5f","added_by":"auto","created_at":"2025-12-18 12:05:05","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":917845,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript0032025submt.docx","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/184792411096092529ab2f81.docx"},{"id":98512966,"identity":"0d4976d3-6fca-4abe-aa1c-a646a7cc16f4","added_by":"auto","created_at":"2025-12-18 12:05:05","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":6267,"visible":true,"origin":"","legend":"","description":"","filename":"e229ba41489b448a8e067d44a1f3e362.json","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/62991b72cf51862644b75c3b.json"},{"id":98512973,"identity":"720cd75d-a490-4b41-96f0-1a0d5a8baf01","added_by":"auto","created_at":"2025-12-18 12:05:05","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":105405,"visible":true,"origin":"","legend":"","description":"","filename":"e229ba41489b448a8e067d44a1f3e3621enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/addfdbf12c1ea05b7bb52a87.xml"},{"id":98625622,"identity":"66041703-9d77-41ba-8efe-411295a16ac0","added_by":"auto","created_at":"2025-12-19 17:09:13","extension":"png","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":43145,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/b2a8ccf703f24a0e7405d623.png"},{"id":98512974,"identity":"6c93c9c7-5f9e-40ca-8ddb-954d1cedc0ad","added_by":"auto","created_at":"2025-12-18 12:05:05","extension":"png","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":48232,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/178156d70a1fe27bdc20b758.png"},{"id":98512975,"identity":"58e84806-264d-473e-8b66-9b4138a3f4c0","added_by":"auto","created_at":"2025-12-18 12:05:05","extension":"png","order_by":9,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":50949,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/535aed52ec3f4c83512242d6.png"},{"id":98625843,"identity":"7a8c4174-ab6d-4e99-b9a0-629c5bc551f9","added_by":"auto","created_at":"2025-12-19 17:09:22","extension":"png","order_by":10,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":40027,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/3d7ebc85ed406aeea022b90c.png"},{"id":98624521,"identity":"a7dce604-e56a-48a1-9624-a937ea1c02f6","added_by":"auto","created_at":"2025-12-19 17:08:29","extension":"xml","order_by":11,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":102159,"visible":true,"origin":"","legend":"","description":"","filename":"e229ba41489b448a8e067d44a1f3e3621structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/a6378a09a92cc085158cbbcc.xml"},{"id":98512977,"identity":"e60d3df5-51f2-4595-9c2f-10eaa2e5a44d","added_by":"auto","created_at":"2025-12-18 12:05:05","extension":"html","order_by":12,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":118193,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/170f1877b3b866d227c102e6.html"},{"id":98512965,"identity":"dcacb85a-34f9-487d-9458-5dfb6af51c54","added_by":"auto","created_at":"2025-12-18 12:05:05","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":115382,"visible":true,"origin":"","legend":"\u003cp\u003eSpatial extent of Sub-Saharan Africa (SSA) showing the locations of the three regional economic blocs namely ECOWAS, SADC, and EAC, highlighted using centroid indicator boxes. Detailed maps of each bloc are shown below for reference.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/f34dca93706ae8f41dec950a.png"},{"id":98624904,"identity":"712b3272-0040-426b-964b-e8b98bdc3cd7","added_by":"auto","created_at":"2025-12-19 17:08:48","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":178552,"visible":true,"origin":"","legend":"\u003cp\u003ePosterior mean spatial random effects (log relative risk (RR)) for four major NCD mortality groups in forty-one sub-Saharan African countries, 2000–2019. Maps show the posterior mean of the BYM2 spatial random effect for (a) cardiovascular diseases, (b) diabetes mellitus, (c) malignant neoplasms and (d) chronic respiratory diseases from fully adjusted Poisson models including bloc, time, socioeconomic indicators, HIV prevalence, temperature and precipitation. Colours represent departures from the regional mean on the log–relative risk scale, with darker red indicating higher-than-expected mortality after adjustment for covariates and darker blue indicating lower-than-expected mortality.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/1ee570e2bddf87185eac23c0.png"},{"id":98512971,"identity":"8966ea3e-37f0-4ad5-8906-719c5edbc5eb","added_by":"auto","created_at":"2025-12-18 12:05:05","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":169592,"visible":true,"origin":"","legend":"\u003cp\u003eHotspots of excess NCD mortality in sub-Saharan Africa, 2015–2019. Panels show the posterior mean RR of mortality, averaged over 2015–2019, for (a) cardiovascular diseases, (b) diabetes mellitus, (c) malignant neoplasms and (d) chronic respiratory diseases. Relative risks were obtained from disease-specific Bayesian spatio-temporal Poisson models with BYM2 spatial random effects and are standardised by expected deaths derived from WHO Global Health Observatory age-standardised mortality rates.\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/f31eb5b881759e790948594b.png"},{"id":98512968,"identity":"bd01198a-3b08-4f4d-9dd0-5fd57618ad96","added_by":"auto","created_at":"2025-12-18 12:05:05","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":46146,"visible":true,"origin":"","legend":"\u003cp\u003ePosterior predictive PIT histograms for the four disease-specific Bayesian spatio-temporal models of NCD mortality in sub-Saharan Africa, 2000–2019. Each panel shows the empirical distribution of PIT values for one cause of death; the dashed vertical line marks 0.5 (centre of a Uniform [0,1] distribution). The approximately symmetric, bell-shaped histograms without pronounced U-shaped patterns indicate reasonably well-calibrated posterior predictive distributions for all four models.\u003c/p\u003e","description":"","filename":"floatimage4.png","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/a3bbda687f6bf647f490f284.png"},{"id":98631752,"identity":"91587dfd-7f42-41ed-9862-fe84b0e206dc","added_by":"auto","created_at":"2025-12-19 17:20:25","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1566914,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8250412/v1/adf58366-6850-4efa-9599-b02361f1a23c.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Climate, Spatial Clustering and Hotspots of Non-Communicable Disease Mortality in Sub-Saharan Africa: A Bayesian Spatial Epidemiology Study, 2000–2019","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eNon-communicable diseases (NCDs), principally cardiovascular diseases, cancers, diabetes and chronic respiratory diseases now account for roughly three-quarters of global deaths, with a rapidly rising contribution in low- and middle-income countries [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Sub-Saharan Africa (SSA) is undergoing a profound epidemiological transition, in which persistent infectious disease burdens increasingly coexist with growing NCD mortality, creating a \u0026ldquo;double burden\u0026rdquo; for already constrained health systems [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e]. Recent Global Burden of Disease (GBD) assessments suggest that NCD mortality rates in SSA have stagnated or declined more slowly than in other regions, especially for cardiovascular disease, diabetes and some cancers, threatening progress towards Sustainable Development Goal (SDG) targets on premature NCD mortality [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eConcurrently, SSA is experiencing rapid urbanisation, demographic growth and climatic change. Urban growth is reshaping exposure to behavioural and environmental NCD risk factors such as diet, physical inactivity, air pollution and heat stress [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Climate change is projected to increase the frequency and intensity of extreme heat events and alter rainfall regimes, with important implications for cardiovascular and respiratory morbidity and mortality [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Yet, despite increasing recognition that NCDs and climate are intertwined, large-scale, spatially explicit analyses of how NCD mortality patterns relate to climatic and socio-demographic gradients in SSA remain scarce.\u003c/p\u003e \u003cp\u003eA growing body of work has quantified NCD burden in SSA using GBD estimates and national vital statistics, highlighting heterogeneities in levels and trends across countries and within regional economic blocs [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. However, most analyses treat each NCD cause independently and rely on non-spatial statistical models, limiting the ability to identify shared high-risk areas or to quantify spatial dependence in mortality risk.\u003c/p\u003e \u003cp\u003eSpatial epidemiology and Bayesian disease-mapping approaches provide tools to characterise geographic clustering of health outcomes and to borrow strength across neighbouring areas [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]. Conditional autoregressive (CAR) models and their extensions, such as the re-parameterised Besag\u0026ndash;York\u0026ndash;Molli\u0026eacute; model (BYM2), allow decomposition of area-level risk into structured (spatially correlated) and unstructured components, while penalised complexity (PC) priors improve interpretability and regularisation [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e]. The Integrated Nested Laplace Approximation (INLA) provides a computationally efficient framework for fitting such latent Gaussian models in large datasets.\u003c/p\u003e \u003cp\u003eEvidence linking climatic factors to NCD outcomes has expanded rapidly. Multi-country studies show that non-optimal temperature contributes substantially to cardiovascular and respiratory mortality worldwide, with particularly high relative risks at extreme heat and cold [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Urban heat islands, air pollution and socio-economic vulnerability can amplify these effects. Nevertheless, empirical analyses in SSA remain limited by sparse ground data and incomplete civil registration, and few studies have examined whether the spatial structure of NCD mortality is shared across multiple disease groups or modified by climatic conditions.\u003c/p\u003e \u003cp\u003eThree gaps are particularly relevant for SSA. First, there is limited understanding of whether major NCDs share common spatial risk patterns such as overlapping hotspots of high mortality, once differences in age structure and baseline rates are accounted for. Second, the extent to which these shared or divergent spatial patterns vary across regional economic communities (East African Community [EAC], Economic Community of West African States [ECOWAS] and Southern African Development Community [SADC]) remains under-explored. Third, while climatic variables such as long-term mean temperature and precipitation can now be characterised at high spatial resolution using global products like WorldClim 2.1 [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], they have rarely been integrated into Bayesian spatio-temporal NCD models for SSA.\u003c/p\u003e \u003cp\u003eAddressing these gaps is essential for regional planning. If the same countries emerge as hotspots across multiple NCDs, this would support integrated prevention and health-system strengthening strategies. Conversely, weak or negative spatial correlations between causes would argue for disease-specific targeting. Understanding whether climate-adjusted residual risk clusters remain after controlling for socio-economic and health-system covariates could help identify areas where unobserved factors such as local health care quality, environmental exposures or social vulnerability drive excess mortality.\u003c/p\u003e \u003cp\u003eThis study aims to quantify and compare the spatial patterns of mortality from four major NCD groups namely cardiovascular diseases, diabetes mellitus, malignant neoplasms and chronic respiratory diseases, in forty-one countries of SSA between 2000 and 2019. Specifically, we address two research questions:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eTo what extent do these four NCD groups share common spatial risk patterns across SSA and within the EAC, ECOWAS and SADC blocs after adjusting for demographic, socio-economic, health-system, HIV and climatic covariates?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eWhere are persistent climate-adjusted hotspots of excess NCD mortality, and how do they align across disease groups?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003eMethodologically, we develop a consistent Bayesian spatio-temporal modelling framework in which cause-specific mortality counts are modelled using shared covariates (urbanisation, GDP per capita, health expenditure, HIV prevalence, temperature and precipitation) and disease-specific spatial and temporal random effects. The models are fitted via INLA using BYM2 spatial priors and PC hyperpriors, enabling robust inference under data sparsity and avoiding over-fitting [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eSubstantively, this is, to our knowledge, one of the first multi-disease analyses to: (i) map climate-adjusted spatial random effects for four leading NCD causes across SSA; (ii) quantify SSA-wide and bloc-specific correlations between these spatial effects; and (iii) identify multi-disease hotspots based on posterior mean relative risks averaged over recent years. By linking region-specific NCD mortality patterns with climatic and socio-economic gradients, the study provides actionable evidence for regional NCD control and climate-resilient health planning.\u003c/p\u003e"},{"header":"2. Methodology","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Study area and variable definitions\u003c/h2\u003e \u003cp\u003eThe study covers forty-one countries in SSA (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e) that are members of at least one of three regional economic communities: EAC, ECOWAS and SADC. National territories are used as the areal units of analysis. A country-level shapefile was transformed to an equal-area projection, harmonised across years, and encoded with unique region identifiers (region_id) for spatial modelling.\u003c/p\u003e \u003cp\u003eAnnual cause specific deaths and age-standardised mortality rates (ASMR, per 100,000 population) for four broad NCD groups, cardiovascular diseases, diabetes mellitus, malignant neoplasms and chronic respiratory diseases, were extracted from the World health organisation (WHO) Global Health Observatory (GHO) and related country-level databases for 2000\u0026ndash;2019 [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Observed deaths in country \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i\\)\u003c/span\u003e\u003c/span\u003e, year \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:t\\)\u003c/span\u003e\u003c/span\u003e and disease group \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:d\\)\u003c/span\u003e\u003c/span\u003e are denoted \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{Y}_{itd}\\)\u003c/span\u003e\u003c/span\u003e.\u003c/p\u003e \u003cp\u003eExpected deaths \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{E}_{itd}\\)\u003c/span\u003e\u003c/span\u003e were derived by combining mid-year population estimates \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{P}_{itd}\\)\u003c/span\u003e\u003c/span\u003e with ASMR \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{itd}\\)\u003c/span\u003e\u003c/span\u003e (per 100,000 population) for the same country, year and disease group according to\u003cdiv id=\"Equ1\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ1\" name=\"EquationSource\"\u003e\n$$\\:Eᵢₜd\\:=\\:(Pᵢₜd\\:/\\:\\text{100,000})\\:\\times\\:\\:Rᵢₜd$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e1\u003c/div\u003e\u003c/div\u003e,\u003c/p\u003e \u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Pᵢₜd\\)\u003c/span\u003e\u003c/span\u003e reflects the population at risk and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Rᵢₜd\\)\u003c/span\u003e\u003c/span\u003e is based on WHO\u0026rsquo;s standard world population. The ratio\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{Y}_{itd}/{E}_{itd}\\)\u003c/span\u003e\u003c/span\u003e defines a crude standardised mortality ratio (SMR). For modelling, we use the natural logarithm \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:Eᵢₜd\\)\u003c/span\u003e\u003c/span\u003e as an offset to ensure that fitted risks are interpreted as relative risks (RRs) compared to a reference with ASMR \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{R}_{itd}\\)\u003c/span\u003e\u003c/span\u003e [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Data sources and covariates\u003c/h2\u003e \u003cp\u003eSocio-economic and health-system covariates were compiled from the World Bank\u0026rsquo;s World Development Indicators (WDI) and related sources [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. For each country and year, we extracted gross domestic product per capita (GDP_pc, constant US dollars), current health expenditure per capita (HExp), and the proportion of the population living in urban areas (Urban%). HIV prevalence among adults aged 15\u0026ndash;49 years was obtained from UNAIDS and harmonised with the mortality series[\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e](World Bank Group, 2024). These indicators are expressed as log-transformed values (for GDP_pc and HExp) or standardised z-scores (for Urban% and HIV prevalence) to aid convergence and interpretation.\u003c/p\u003e \u003cp\u003eClimatic covariates were derived from WorldClim v2.1, which provides 30-arc-second (~\u0026thinsp;1 km) gridded surfaces of long-term monthly climate normals for 1970\u0026ndash;2000 [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e]. We extracted country-level averages of mean annual temperature (\u0026deg;C) and total annual precipitation (mm) by overlaying the national polygons on the gridded surfaces and computing area-weighted means. Although WorldClim normals pre-date the 2000\u0026ndash;2019 mortality period, they capture persistent spatial gradients in climate that are relevant for chronic exposure profiles and long-term adaptation. Both temperature and precipitation were centred and scaled prior to analysis.\u003c/p\u003e \u003cp\u003eThe final analytic dataset therefore comprises, for each combination of country, year and disease category: observed and expected deaths, ASMR, bloc membership (EAC, ECOWAS, SADC), and covariates representing urbanisation, economic development, health expenditure, HIV prevalence, mean temperature and precipitation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Spatio-temporal model formulation\u003c/h2\u003e \u003cp\u003eFor each disease group d, we model the annual observed deaths \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{Y}_{itd}\\)\u003c/span\u003e\u003c/span\u003e in country \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i\\)\u003c/span\u003e\u003c/span\u003e and year \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:t\\)\u003c/span\u003e\u003c/span\u003e using a Poisson log-linear model:\u003c/p\u003e\u003cp\u003e\u003cimg src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAvQAAAAiCAYAAAAwJE0NAAAAAXNSR0IArs4c6QAAAARnQU1BAACxjwv8YQUAAAAJcEhZcwAAFiUAABYlAUlSJPAAABDySURBVHhe7d3PaxtH2Afwr94/oGHjW0lL0bqH0IBLsgolOAcH6hVtDoW2yMccAkEyFPrLpZuEHupA5IYEQho7IYYcklqiPVZu5EIClQhNrASJprSHrgj9ddJGTfsHzHt49ey7Ozsr7SqKLUXPB3TwjLw7Ws3sPJqdnU0IIQQYY4wxxhhjI+l/5ATGGGOMMcbY6OCAnjHGGGOMsRHGAT1jjDHGGGMjjAN6xhhjjDHGRtjIBPTVahWJRALValXOchUKBUxOTiKRSCCdTsvZbMwtLS1hcnISjuP40h3HwdLSEnbu3IlEIoGlpSVf/tPWaDQwOTmJQqEgZ7EObtuMMcZYuK4BfTqdRiKRcF/r6+u+fG9eIpHw5W21arWKH374AXfu3EGlUkG5XA6Ud1QUCoXAsaVXKpXa8oBz1DUaDaRSKdy/fx8bGxuYmJjw5a+urmLHjh149OgR8vk8Pvnkk0DQ/zRNTU1hY2MDV69eRSqV2tJ9j4JnqW33o9v5wPu6dOmS/K9PXbPZRC6Xc38Mp1IpNBoN+W2MMTaWms0mEolE3wN2c3Nz0QexRBf1el1omiYAiHw+L2eLUqkkAAhN00S9XpezB6pSqQgAolKpyFlKmqYpyzwqLMsSAEQmk3HTbNsWpmmGfh9x1ev1wD6eNbZtC03TRDablbOU6JhErWeDls1mhWEYotVqyVmsY9Tbdj9U5wOSz+cFgKd+DpZR/2CaprBtW7RaLWGaptA0TX4rY4yNpWw2K3Rdl5NFqVQSmUxGAHDj6Gw2K2zb9r3Ptm0BQKytrfnSVbqO0E9NTWFubg4A8OKLL8rZePDgAQDg1q1bmJqakrO31f79++WkkbJjxw4AwMzMjJuWTCaRz+cBAJcvX3bT+/Xjjz8C0j6eNTR6ePHiRTlLabvr8eeff452u435+Xk5i3WMetvuB50P9u7dK2fhwIED0DRtS+tus9nEzMwMdF3H9evXkUwmMTExgSNHjqDdbnedGskYY+Og2WxieXkZH3/8sS+9UCjgzTffRLFYhGmasCwLuq5jeXkZhmH4rtInk0mYpokTJ074tqHSNaAHgMOHDwMAfvrpJ1+64zg4ffo0VlZWtrQjGRcPHz4EALz22mu+9F27dgEAbNv2pffj2LFjEELg2LFjctYzoVqtolwuY3FxUc4aWhMTE1hcXESxWOSgiLkeP34MdIJ32fT0NB49eiQnP1VnzpxBu93GuXPnfFPY6Pz0888/e97NGGPjZ3V1FQDw+uuv+9JzuRw0TUO9XseNGzdw6tQpbG5uIp/Po91u4+zZs773HzlyBLZt94wJegb0b7zxBgCgWCz60ufn5zE7OzuUwaDjOLh7966cPFI2NzeVo2537twBAJim6UtnQRcuXICmae5Vpih6NZitQI3/woULctbYexbadj82NjYAALt375azthyNOhmGgenpaV/e7du3Ac8VBcYYG1fFYhGGYSCZTLppjUYD7XYbs7OzgfjunXfeAQDcu3fPl04xwVdffeVLl/UM6AEgk8nAtm00m00AwKVLl2DbNr788kv5rUNhfn4e7XZbTh4ZjuOgVqthdnZWzsL58+eBzi82lUKhgFQq5buJVr5JbWlpyc1X3Wwh3+hG28nlcr730eowtPpIIpHA5OQk5ubm3Lri1atsjuO4+925c6cbXK+vr/tWOIly06jjOCgWi7GmZziOE3pct9LExARM0wz8iGaj37b7VavVoOu6bzQ8l8ttyw3yNOp09OhROctFI/WMMTaOms0mbNsOxHFTU1MQQihvkg0bCJmYmIBhGO7ATphIAT3Nsb579y4ajQa++OILXLlyJbBayDBYX18f+UCIRuG9c9ur1SrS6TTK5TKy2Wxg1NlxHDfo/vDDDyGEcKflzMzM+ALshYUFWJYFADh06JCbTmZnZ7G5uYlarQYhBCqViu8HHTl58iROnz6Nq1evuvvTNA3FYtH3izRq2U6ePInDhw+jVquh3W7j9u3byOVy+OOPP/Dbb7/BsiyUy2V888037rbD/PLLL0DI5wtz9uzZgUxlGoR9+/YBQ3LFYFgMa9uWVwOL8ooTiFMdMAzDl7a8vKycgvO0UaciX0YGgJs3bwIAnn/+eTmLMcbGxq+//gqE3H8a5vvvvwc8/b9XKpWCbdvdBzTlu2RV6C7bTCYjDMOIdLftoEVZ5abVaglN04RhGMI0zW1bCaNSqQhd1wUAYZpmaJlt2xalUklOdle0kF+ZTEb5ftG5kxpAIH9tbU0AECsrK750WhlDLhsdZ/nY5fN5YVmWL40+n1elUhGGYfjS4paN6ptpmr48WlVJLpsKfT5522Hoc9Oxl4/LVqPyb0dbG0bD0ra3A9UF1WurV0NqtVqBMsgvXuWGMTbuwmKsMNTHaZoWWOlGRNxepBH6ZDIJXddRLBaRSqUCo8Mk7siTFz046km2QZfjz507J2f50JSTfnmnrMgjqI1GA++//z42NjbQarVw6NAhHDx4UPmZLMtyVwryovlTQgjfq1AouPc0eHnntMr5zz33HOC5qY7cv38f6NxQ50Uja19//bXvl+DCwgJOnTrleSeg67p71YZMT09jc3PT/bufsnnnSHvv0aBjFWdU8pVXXpGTAmiqjaZp+OCDD+Rsn3Q6rZymFJV3NDeK33//XU4aS1HbdpQH0HXTrW1vFxr1rlQq7rkgn88HpuB4PWk9pWMgb4OufGWz2cD5qVKpAJ0rfLInOa+jS3kYY2zUOY6DdDqNdruNa9eu+WY4xBEpoAcATdOAzrJ6KtT57dmzx5d+/PjxSCfh6elplEolIGbARuhyvGVZgSBVdvPmTd/laxL1oU3eKSvyvo4ePYorV664y7gtLCygUqng9OnTSKfTgfsQFhYWfP8PAOVyWVm+MDQFRTWnVfWDAZ3L5qp9JJNJWJaFWq2Gl19+WTnPi5w/fx7tdhuvvvoqlpaWlJeC+ikbraj03nvv+dIpsBn0jYE01ebatWuhARIpl8uBy2Hr6+uRA/QbN27AMAy+qTmGOG2bbsqU68gg2vZ2oR+43vIcOHBA2X7Jk9bTVqsFKKas0eo1L730ki8dAL777jsAwFtvveVLD+sbBlEexhgbZRTM12o1rK2tBQY+Y5GH7FXoMqs8laIXuoQQ9fI4TcFQ6Tblxnupgi5BG4YReb+ix/ZVstlsYLqJ6FwWUbFt2516AkBks1nl5XIqR9QHIQkh3IdNqcpOed6HztDDk+QpNF6lUsk3bUhVVtHZFu1D1/XAw23ilk10vju5HsStg1EuTwnP8abvkvbT6/+8TNNU1oUwuq6H1hNC5e/1vmfdMLXtMFSP47yilp/aapzyhIlTT2nam3zMurUrXdd931MvgygPY4wNo27nSkIP59M0LTAlWRZle+roWUJzl8MCQOrQvCfntbU1oXWeMut9eZVKJTd4o/mxYQFbt06ZnrblnW+s6jRpG/J2VHPW5Y4mn8+7Aa6u60LX9dDj8SToS4s691t4jr+MglP5KWUrKysCijntslar5f4I6VUe2qb8JMu4ZaN0+fj3qoOyKAFxq9Vyv1OasxZWz2h73s9Sr9fd//e+vP9Lx5DaAh2PXsee9jfuc+ijtm1Kl/OGqW33gwY5VOWhHzve+tZvPa3X676nFlI9lYPzsE6Fyil/L6q+YZDlYYyxYRR2riTeYF4e1FShWEw1v54EIy0F2lBYcEFBmNzpUJCnKoB8Q2SvUeOwQIsCPbmThqJzEZ4OXu4YMplMILgkFJDRvqMGxP2gTkz+nN1QZycfZ/qs8vdG+1C9Xz5mdNy96aVSKXC8ReeYy+lxyxZ24ys1jqjHXFVuGZXB+56weiY6I8PyD06qt/LnEJ12Qf9Dn5+OvVz/ZGE351Kgo9rfsyZu26b3y3VkWNp2P8JuKBeeNiGLW0+pY6GrhvRDV96G8Jy3vfWSfljouh6o12F9w6DKIzrHIZPJBPbNGGPbhWIJ1WAonUe9sUEvpmkKrceCA8HeQGLbtju6qCqY8Jy05U4nm80qO1LqAORpJWEneBESaNF2oJi2YZqm2yGYnikjlmUpOwZd1wPlEZ79estFaVG/iKiok5P31wsFMt4OkDp71WfyjnaVSiX3PaZpCl3X3WPc8ozQe487bdtbRgqE5EArbtnCAndvekWxko6M6qQcDBL6DlVBCH02Cp6JYRiBwIQapqouUJm9efnOaHAvdOVKRt+dKu9Z0k/bpromf5/D0Lb7RSPZcnmo3qnqd9x6Su1eTlMdM+oPqF3Yti0MwwgdZaJ2KLfnQZWHvi8ozj2MMbZd6Nwnn7eon0JnoCOfzytf8rkRihkQsq5RARXI+1KdNKmAcgHCOlLVyH2vjlQV0NNoZ7d9mKbp62gMwwi83+7Mz1QF0arRvahBWRyqYy13gt2srKy4nT997rD/L5VK7qUey7LcAMiyLF/AGLadtbU1YXZ+LdL7jC7zmuOUjfYvB2U0aofO9x1WT7zCtiWEcLcV9p3Tfuh/6fuR358N+dEqOvuQ61pYYOJF+1I1XgqE8IwH9P20bcuyAt/FMLTtftEx6PaS21zceho2Uq5KIyXP/TXekXSVsL5hUOVpda6CqY4FY4xtp0wmExhVp4G+Xi9vrKsafFIZSFQwiI7UsqzAB/dSBfRxhXV23UaLVEGZYRjKYIsNF/pe5e+7H2GBSdiPVlVgQvUv7EoXoYA1rK5nY960OS5UP9bHrW3HraeqwRXahmrEPS5V3yAGXB7qa8LaC2OMbQfqf8IGMKOi+Dhs4IREXraym3v37gWWUPv777+BkEeA//PPP5icnHT/prXK9+/f73vfoNEayvKSdrTWt2rtz3a77VuibX19HbVaDXv37vW9jw2fubk5mKaJEydOKJfUjOPBgwfQNC1QR2zbVi7h999//wFS/T979izQY218x3Hw6aefwjRN5bKJS0tLKBQKyOfzctbYq9VqmJqa8qWNW9uOW0/pGRDe93/22WdA5xHlT0rVN2CA5Wk0GshkMshms8r2whhj22Vubg66ruPbb7+Vs2JZXl7G3Nxcz2W1BxLQl8tl6LqORqOB48ePAwD+/PNPAMC///6LZrOJdDrtBlXJZBJ3796F4zhoNBo4c+YM9u/fj3379vm2MWi0RvWuXbuQy+XcNeEfPnwIdIKparXqe3CWruvu+ufr6+vu2ul79uxBoVDouk47234XL17Eo0ePcPLkSTkrlps3b0LXdTiO49YPqj+PHz+G4zjI5XLumtv0gC6qc9410Hfv3u2rf17z8/PYuXMnrl+/LmcBnbp669YtZXAzzui4v/DCC752OW5tO249pceSV6tVNy+VSsEwDN82+qXqGwZZnr/++gvvvvsuLl686EtnjLFhsLi4iOXlZWV/HwX1Qx999JGcFSQP2fcj27lx0rvSgHduozzXte5ZtozmX9KqHmbImueDmHJD87B1XfdNhZDnZ3v3v9ZZflPzrKdPn0u+8YwNp3q9LgzDEKZpBqYiREWXzgzD8NVlqvu6rgcuq1Gdpjy7c0OhpmmBaUB25+ZCI8Zd7+z/0flG69wXQsatbcetp63OjcXe8zRNcdEVz5WIS9U3eNO3ujyMMbbV8vl837HryspKILYIkxD/d8PR0KtWqzh48CAqlQpfWmWxOY6D1dVVXL58GXfu3Ol56WorNRoNvP3221hcXAyMQDLGGGOM9TIyAT1jjDHGGGMsaCBz6BljjDHGGGPbgwN6xhhjjDHGRhgH9IwxxhhjjI0wDugZY4wxxhgbYRzQM8YYY4wxNsI4oGeMMcYYY2yEcUDPGGOMMcbYCPtfCqcdPQTMhd4AAAAASUVORK5CYII=\" width=\"609\" height=\"34\"\u003e\u003c/p\u003e\u003cp\u003ewhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{E}_{itd}\\)\u003c/span\u003e\u003c/span\u003e is the expected death count from (1) and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\theta\\:}_{itd}\\)\u003c/span\u003e\u003c/span\u003e is the relative risk. The log-risk is decomposed as\u003cdiv id=\"Equ3\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ3\" name=\"EquationSource\"\u003e\n$$\\:\\text{log}{\\theta\\:}_{itd}={\\alpha\\:}_{d}+{\\beta\\:}^{T}{\\varvec{X}}_{it}+{u}_{id}+{v}_{id}+{\\gamma\\:}_{td}+{\\delta\\:}_{itd}$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e3\u003c/div\u003e\u003c/div\u003e,\u003c/p\u003e \u003cp\u003eWhere \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\alpha\\:}_{d}\\)\u003c/span\u003e\u003c/span\u003e is a disease-specific intercept; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varvec{X}}_{it}\\)\u003c/span\u003e\u003c/span\u003e is the vector of centred covariates with corresponding regression coefficients \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\beta\\:\\)\u003c/span\u003e\u003c/span\u003e shared across diseases; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{u}_{id}\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{v}_{id}\\)\u003c/span\u003e\u003c/span\u003e are, respectively, structured and unstructured spatial random effects for disease \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:d\\)\u003c/span\u003e\u003c/span\u003e in country \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i\\)\u003c/span\u003e\u003c/span\u003e; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\gamma\\:}_{td}\\)\u003c/span\u003e\u003c/span\u003e is a disease-specific temporal random effect for year t; and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\delta\\:}_{itd}\\)\u003c/span\u003e\u003c/span\u003e is a space-time interaction term capturing residual heterogeneity.\u003c/p\u003e \u003cp\u003eSpatially structured components \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{u}_{id}\\)\u003c/span\u003e\u003c/span\u003e are assigned a BYM2 prior, which combines an intrinsic conditional autoregressive (ICAR) field and independent and identically distributed (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i.i.d\\)\u003c/span\u003e\u003c/span\u003e). Gaussian component with a mixing parameter \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\varphi\\:\\)\u003c/span\u003e\u003c/span\u003e that determines the proportion of spatially structured variance [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e](Lee et al., 2018). Spatial adjacency is encoded via a symmetrical k-nearest-neighbours graph (k\u0026thinsp;=\u0026thinsp;4) derived from country centroids. Temporal random effects \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\gamma\\:}_{td}\\)\u003c/span\u003e\u003c/span\u003e follow a first-order random walk (RW1) process over calendar years to smooth trends [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e](Banerjee et al., 2003). The interaction term \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\delta\\:}_{itd}\\)\u003c/span\u003e\u003c/span\u003e is modelled as \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:i.i.d\\)\u003c/span\u003e\u003c/span\u003e Gaussian noise.\u003c/p\u003e \u003cp\u003eTo explore shared spatial structure across diseases, we fit two sets of models:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eDisease-specific climate-adjusted models. Each disease d is modelled separately using (2)-(3), allowing estimation of disease-specific spatial random effects \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{u}_{id}+{v}_{id}\\)\u003c/span\u003e\u003c/span\u003e. These are then mapped and compared.\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eJoint multivariate model. All four diseases are stacked into a single dataset with a disease index \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:d=1,\\:\\dots\\:,\\:4\\)\u003c/span\u003e\u003c/span\u003e. The model retains disease-specific intercepts and replicates the spatial and temporal random effect structures across diseases, enabling partial pooling while allowing patterns to differ by disease [\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e](Mahaki et al., 2011).\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e \u003cp\u003eFrom the fitted disease-specific models, we compute: (i) posterior mean spatial random effects for each disease and country; (ii) SSA-wide and bloc-specific correlation matrices of these effects to quantify shared spatial risk; and (iii) posterior mean relative risks averaged over 2015\u0026ndash;2019 to identify recent hotspots.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Prior specification and computation\u003c/h2\u003e \u003cp\u003eWe adopt penalised complexity (PC) priors for hyperparameters governing the variance of spatial, temporal and interaction random effects, following recommendations for disease mapping (Simpson et al., 2017). For each precision parameter \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\tau\\:\\)\u003c/span\u003e\u003c/span\u003e of a Gaussian randomeffect (variance \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:1/\\tau\\:\\)\u003c/span\u003e\u003c/span\u003e, we specify a PC prior such that\u003cdiv id=\"Equ4\" class=\"Equation\"\u003e\u003cdiv format=\"TEX\" class=\"mathdisplay\" id=\"FileID_Equ4\" name=\"EquationSource\"\u003e\n$$\\:\\text{Pr}\\left(\\sigma\\:\u0026gt;{\\sigma\\:}_{o}\\right)={p}_{0},\\:\\sigma\\:=1/\\tau\\:,$$\u003c/div\u003e\u003cdiv class=\"EquationNumber\"\u003e4\u003c/div\u003e\u003c/div\u003e\u003c/p\u003e \u003cp\u003eWith \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\sigma\\:}_{0}=1\\)\u003c/span\u003e\u003c/span\u003e and \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{p}_{0}=0.01\\)\u003c/span\u003e\u003c/span\u003e, implying string prior belief that the standard deviation is small but allowing heavier tails. For the BYM2 mixing parameter \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:\\varphi\\:\\)\u003c/span\u003e\u003c/span\u003e, we use a PC prior favouring a balanced contribution of structured and unstructured components while avoiding degeneracy at the boundaries [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e]\u003c/p\u003e \u003cp\u003eAll models are fitted using the Integrated Nested Laplace Approximation (INLA) as implemented in the R\u0026ndash;INLA package, which provides accurate and fast approximations to posterior marginals in latent Gaussian models [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5 Model diagnostics and predictive calibration\u003c/h2\u003e \u003cp\u003eModel adequacy is evaluated using a combination of global fit indices and posterior predictive checks. For each fitted model we report DIC and WAIC, with lower values indicating better trade-off between fit and complexity [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Conditional predictive ordinates are examined for numerical failures; counts of CPO failures close to zero indicate stable approximations.\u003c/p\u003e \u003cp\u003ePredictive calibration is assessed using the probability integral transform (PIT). For each observed count \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{Y}_{itd}\\)\u003c/span\u003e\u003c/span\u003e, the PIT value is computed from the posterior predictive distribution; histograms of PIT values should be approximately uniform under a well-calibrated model [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e]. Furthermore, we compare observed deaths with the posterior mean of fitted values on a country-year basis. Scatter plots of observed versus predicted counts, overlaid with a 45\u0026deg; line and harmonised axes, provide a visual check for systematic under- or over-prediction, especially in countries with large populations and high mortality.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Climate-adjusted spatial random effects and residual clustering of NCD mortality\u003c/h2\u003e \u003cp\u003eThe spatial random effects as depicted in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, reveal substantial residual geographical structuring of NCD mortality risk that is not explained by bloc membership, socioeconomic indicators, HIV prevalence or climate covariates. For cardiovascular diseases, a pronounced high-risk cluster is visible in parts of south-eastern Africa, with neighbouring countries showing moderately elevated residual risk. Several western and central countries display neutral or slightly negative spatial effects, suggesting that once observed covariates are accounted for, their cardiovascular mortality is close to or below the regional average.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eDiabetes and respiratory diseases exhibit more moderate, but still heterogeneous, residual spatial patterns. For diabetes, a band of mildly elevated effects appears in selected central and southern countries, whereas much of West Africa shows near-null or slightly protective effects (Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). Chronic respiratory diseases show distinct high-risk foci in parts of the southern region, consistent with the hotspot patterns seen in the averaged relative risk maps, while many other countries have small or negative spatial effects, implying that observed covariates capture a larger share of the between-country variation.\u003c/p\u003e \u003cp\u003eIn contrast, malignant neoplasms display comparatively weak spatial structure, with most countries falling in the central categories of the legend. This suggests that, at the scale of national averages, residual cancer mortality risk is less spatially clustered than for cardiovascular or respiratory causes. Overall, the maps indicate that even after adjusting for climate and socioeconomic determinants, there remain geographically coherent pockets of excess NCD mortality, pointing to unmeasured local risk factors, differences in health-system performance, or data quality issues that warrant targeted investigation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Persistent spatial hotspots of excess NCD mortality (posterior (RR), 2015\u0026ndash;2019)\u003c/h2\u003e \u003cp\u003eAveraging model-based relative risks over the most recent five-year period highlights a pattern of widespread, persistent excess NCD mortality across the region rather than isolated, short-lived spikes (Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). For cardiovascular diseases, several southern and eastern African countries maintain consistently elevated RRs compared with the continental distribution, whereas parts of the western Sahel appear relatively less affected. Malignant neoplasm mortality similarly exhibits sustained excess risk in a band spanning southern Africa and pockets of West Africa, suggesting shared underlying determinants such as late presentation, limited diagnostic capacity and constrained access to cancer treatment.\u003c/p\u003e \u003cp\u003eIn contrast, as Fig.\u0026nbsp;\u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e clearly demonstrates, diabetes and chronic respiratory disease display more uniformly high relative risks across most of sub-Saharan Africa, indicating that metabolic and chronic respiratory hazards are now deeply entrenched regional problems rather than confined to a few high-income or rapidly urbanising settings. Taken together, these hotspot maps reinforce the view that NCD mortality in the late 2010s is both geographically clustered and simultaneously widespread, underscoring the need for region-wide strengthening of prevention and chronic care, with targeted intensification in countries that consistently occupy the highest risk categories.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.3 SSA-wide and bloc-specific cross-disease spatial correlations of residual NCD risk\u003c/h2\u003e \u003cp\u003eAs illustrated in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, the joint spatial correlation analysis shows that cardiovascular diseases and diabetes mellitus share a moderately similar residual spatial pattern across SSA (ρ\u0026thinsp;=\u0026thinsp;0.49), with this correlation strengthening in SADC (ρ\u0026thinsp;=\u0026thinsp;0.70) and ECOWAS (ρ\u0026thinsp;=\u0026thinsp;0.71). This suggests that, after accounting for measured covariates, unobserved spatial determinants such as diet, metabolic risk factors or health-system performance cluster geographically in ways that simultaneously elevate cardiometabolic mortality in many countries within these blocs. In contrast, correlations between cardiovascular diseases and malignant neoplasms are weak (SSA-wide ρ\u0026thinsp;=\u0026thinsp;0.13) (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e), indicating more distinct spatial structuring of cancer mortality, potentially reflecting differences in screening, diagnostic capacity and cancer-specific risk factors.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePairwise spatial correlations between disease-specific spatial random effects from Bayesian spatio-temporal models, overall Sub-Saharan Africa (SSA-wide) and by regional economic bloc. Values are Pearson correlation coefficients between posterior mean spatial random effects for each disease pair, summarising the similarity of residual spatial patterns after adjusting for socioeconomic, health-system and climatic covariates. Positive values indicate that countries with elevated residual risk for one disease also tend to have elevated risk for the other; negative values indicate spatial divergence in residual risk.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDisease pair\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSSA-wide\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSADC\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eECOWAS\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eEAC\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCardiovascular diseases vs. Diabetes mellitus\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.49\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.7\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.71\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-0.13\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCardiovascular vs. Malignant neoplasms\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.13\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.39\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e-0.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-0.18\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCardiovascular vs. Respiratory diseases\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-0.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-0.76\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiabetes mellitus vs. Malignant neoplasms\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e-0.11\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiabetes mellitus vs. Respiratory diseases\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.19\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.88\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.54\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMalignant neoplasms vs. Respiratory diseases\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-0.17\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e-0.23\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e \u003cp\u003e0.03\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe strongest positive association is observed between diabetes mellitus and respiratory diseases in SADC (ρ\u0026thinsp;=\u0026thinsp;0.88), pointing to substantial overlap in residual spatial risk, which may relate to co-occurring urban air pollution, tobacco use or shared health-system constraints. Conversely, cardiovascular and respiratory diseases exhibit a moderate negative correlation at the SSA scale (ρ = \u0026minus;\u0026thinsp;0.56), driven largely by a strong negative correlation in the EAC (ρ = \u0026minus;\u0026thinsp;0.76), where countries with elevated residual respiratory risk tend to have comparatively lower residual cardiovascular risk (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e). Consequently, these patterns imply that while some NCDs share common unmeasured spatial determinants, others are governed by more disease-specific or regionally distinct processes, underscoring the need for bloc-tailored, disease-specific prevention and control strategies rather than a one-size-fits-all regional approach.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Association between climate variability and non-communicable disease mortality\u003c/h2\u003e \u003cp\u003eAfter adjustment for socioeconomic, health-system and HIV covariates, climate effects on NCD mortality were generally modest (Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e). For cardiovascular diseases and malignant neoplasms, the posterior IRRs for both mean temperature and precipitation were close to the null with wide credible intervals that included 1. For example, a 1-SD increase in mean temperature was associated with an IRR of 1.13 (95% CrI 0.92\u0026ndash;1.38) for cardiovascular deaths and 1.01 (95% CrI 0.87\u0026ndash;1.18) for cancer mortality, indicating little evidence of a consistent climate signal for these outcomes.\u003c/p\u003e \u003cp\u003eBy contrast, mean temperature showed clear inverse associations with diabetes and respiratory disease mortality. For diabetes, each 1-SD increase in temperature was associated with a 13% lower mortality risk (IRR 0.87, 95% CrI 0.77\u0026ndash;0.98). The association was stronger for chronic respiratory diseases, where the IRR was 0.74 (95% CrI 0.60\u0026ndash;0.92), suggesting substantially higher respiratory mortality in cooler settings after accounting for other covariates and residual spatial and temporal structure.\u003c/p\u003e \u003cp\u003eFor precipitation, all IRR estimates lay close to unity with credible intervals spanning 1 across all four NCD groups (e.g. respiratory diseases IRR 0.97, 95% CrI 0.84\u0026ndash;1.12), indicating no robust evidence that variation in average annual rainfall was independently associated with NCD mortality over the study period. Overall, these results suggest that temperature may play a selective role in shaping diabetes and respiratory mortality risks in sub-Saharan Africa, whereas long-term mean precipitation appears less influential once other determinants are controlled for.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eClimate covariate effects from Bayesian spatio-temporal models for four major NCDs in sub-Saharan Africa, 2000\u0026ndash;2019. Posterior fixed-effect estimates (β) and 95% credible intervals (CrI) are shown on the log relative-risk scale, together with corresponding incidence rate ratios (IRR) and 95% CrI. Temperature and precipitation are standardised (per 1-SD increase in annual mean temperature or precipitation), and all models adjust for GDP per capita, health expenditure, urbanisation, HIV prevalence and spatio-temporal random effects.\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\"\u0026minus;\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDisease\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCovariate\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003elog RR β (95% CrI)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eIRR (95% CrI)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCardiovascular diseases\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean temperature\u003c/p\u003e \u003cp\u003e(SD-scaled)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e0.121 (-0.082, 0.323)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.13 (0.9\u0026ndash;1.38)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCardiovascular diseases\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrecipitation\u003c/p\u003e \u003cp\u003e(SD-scaled)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e-0.014 (-0.152, 0.125)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.99 (0.86\u0026ndash;1.13)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiabetes mellitus\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean temperature\u003c/p\u003e \u003cp\u003e(SD-scaled)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e-0.137 (-0.257, -0.018)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.87 (0.77\u0026ndash;0.98)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiabetes mellitus\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrecipitation\u003c/p\u003e \u003cp\u003e(SD-scaled)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e-0.021 (-0.103, 0.061)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.98 (0.90\u0026ndash;1.06)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMalignant neoplasms\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean temperature\u003c/p\u003e \u003cp\u003e(SD-scaled)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e0.011 (-0.144, 0.167)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.01 (0.87\u0026ndash;1.18)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMalignant neoplasms\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrecipitation\u003c/p\u003e \u003cp\u003e(SD-scaled)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e-0.044 (-0.152, 0.063)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.96 (0.86\u0026ndash;1.06)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRespiratory diseases\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean temperature\u003c/p\u003e \u003cp\u003e(SD-scaled)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e-0.297 (-0.508, -0.086)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.74 (0.60\u0026ndash;0.92)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRespiratory diseases\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrecipitation\u003c/p\u003e \u003cp\u003e(SD-scaled)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\"\u0026minus;\" colname=\"c3\"\u003e \u003cp\u003e-0.034 (-0.180, 0.111)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.97 (0.84\u0026ndash;1.12)\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eIn all climate-extended models, temperature and precipitation effects were estimated conditional on a common set of macro-level covariates: log GDP per capita, log health expenditure, urbanisation rate and HIV prevalence. Across the four disease groups, higher GDP per capita and health expenditure were generally associated with higher reported NCD mortality, consistent with a combination of more advanced epidemiological transition, improved case ascertainment and competing-risk patterns in better-resourced health systems, rather than genuinely protective effects of under-investment. Urbanisation and HIV prevalence showed disease-specific patterns, with positive associations particularly evident for diabetes and chronic respiratory disease. Against this background, the temperature and precipitation coefficients in Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e can be interpreted as climate associations adjusted for major socio-economic and health-system gradients, rather than simple ecological correlations.\u003c/p\u003e \u003cdiv id=\"Sec13\" class=\"Section3\"\u003e \u003ch2\u003e3.4.1 Adjustment for socioeconomic, health-system and HIV covariates\u003c/h2\u003e \u003cp\u003eAll climate coefficients reported above are estimated from multivariable Bayesian spatio-temporal Poisson models that simultaneously adjust for national income (log GDP per capita), per-capita health expenditure, urbanisation rate and HIV prevalence. Across the four disease groups, higher GDP per capita tended to be associated with modestly elevated NCD mortality, consistent with a shift towards more obesogenic and cardiometabolic risk profiles at higher income levels, whereas higher health expenditure showed weakly protective or null associations.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cp\u003eUrbanisation exhibited heterogeneous effects by cause, with positive associations for diabetes and cancer and more mixed patterns for cardiovascular and respiratory mortality, suggesting that the balance between improved access to care and increased exposure to urban risk environments varies across conditions. HIV prevalence was positively associated with cardiovascular and diabetes mortality and less strongly related to cancer and chronic respiratory disease, reflecting known interactions between HIV, antiretroviral therapy and cardiometabolic risk. These adjustment variables are included primarily to reduce confounding of climate effects by broad socioeconomic and health-system gradients.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e3.5 Model diagnostics and calibration\u003c/h2\u003e \u003cp\u003eCPO failure counts were zero for all disease-specific models, indicating numerically stable computation of conditional predictive ordinates. As illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e4\u003c/span\u003e, posterior predictive PIT histograms were approximately symmetric with no pronounced U-shaped or skewed patterns, suggesting adequate predictive calibration. While formal KS and χ\u0026sup2; tests rejected exact uniformity (p\u0026thinsp;\u0026lt;\u0026thinsp;0.001 across models), this likely reflects the very large number of observations rather than substantive miscalibration. Taken together, these diagnostics indicate that the models provide stable predictions with only mild deviations from perfect calibration.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion policy implications and conclusions","content":"\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Spatial random effects and residual clustering of NCD mortality\u003c/h2\u003e \u003cp\u003eWe found clear evidence of residual spatial clustering in NCD mortality across sub-Saharan Africa (SSA) after adjusting for age structure, expected deaths and macro-level covariates. The BYM2 spatial random effects revealed coherent high- and low-risk regions for cardiovascular diseases, diabetes, cancers and chronic respiratory diseases, rather than purely random geographic noise. This is consistent with earlier work showing that NCD burdens in SSA and other low- and middle-income regions are shaped by spatially structured determinants such as access to care, health system capacity and built environment, which are not fully captured by national averages of income or health spending [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThe persistence of spatial structure after covariate adjustment suggests that \u0026ldquo;where you live\u0026rdquo; still matters for NCD survival in SSA, even conditional on broad socio-economic context. Methodologically, the separation between structured and unstructured components in the BYM2 model, alongside reasonable values of the spatial mixing parameter, supports our use of penalised-complexity (PC) priors and latent Gaussian disease mapping for this setting [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Persistent hotspots of excess mortality\u003c/h2\u003e \u003cp\u003eAveraging posterior relative risks over 2015\u0026ndash;2019 highlighted persistent multi-year hotspots of excess mortality across all four NCD groups. Several countries in southern and western SSA consistently exhibited relative risks above one for multiple causes, whereas other countries tended to lie below the regional baseline. Because these hotspot estimates borrow strength in space and time, they are unlikely to reflect random annual fluctuations and instead point to structural differences in chronic care systems, risk factor environments and diagnostic capacity.\u003c/p\u003e \u003cp\u003eFor chronic respiratory diseases, for example, high-risk clusters may reflect overlapping burdens of household and occupational air pollution, tobacco use and post-tuberculosis lung damage, compounded by limited spirometry and chronic care services [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. These multi-disease hotspots therefore provide a pragmatic starting point for regional targeting of hypertension and diabetes screening, essential medicines provision and integrated chronic care in the countries most consistently above the regional norm.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Cross-disease spatial correlations at SSA and bloc level\u003c/h2\u003e \u003cp\u003eThe cross-disease correlation matrices showed that spatial patterns of NCD mortality are partly shared, but not uniform, across causes. At SSA level, we observed a moderate positive correlation between cardiovascular and diabetes spatial effects, and weaker positive correlations with cancer, consistent with a shared cardiometabolic and oncologic risk environment that includes obesity, diet, tobacco and delayed diagnosis [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e]. In contrast, spatial effects for chronic respiratory disease were negatively or only weakly correlated with those of the other causes, suggesting that respiratory mortality is driven more by distinct exposures such as biomass smoke, occupational hazards and tuberculosis sequelae [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eBloc-specific correlation matrices revealed further heterogeneity: in some regional economic communities, diabetes and respiratory mortality patterns were strongly aligned, whereas elsewhere cardiovascular and respiratory patterns diverged. These differences are plausibly related to bloc-level variation in tobacco control, energy use, urban air quality, HIV burden and integration of chronic care services [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e]. This heterogeneity underlines the importance of not assuming a single, continent-wide geography of NCD risk.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Climate variability and NCD mortality\u003c/h2\u003e \u003cp\u003eAfter adjustment for GDP per capita, health expenditure, urbanisation and HIV prevalence, climate variables showed modest and disease-specific associations with mortality. Higher mean temperature (per standard deviation) was associated with lower diabetes and chronic respiratory mortality, whereas effects on cardiovascular disease and cancer were small and imprecise; precipitation showed no consistent association with any cause. Taken at face value, these results suggest that long-term climatic gradients captured by country-level WorldClim normals are not the dominant drivers of cross-country NCD mortality differences in SSA during the study period.\u003c/p\u003e \u003cp\u003eHowever, these findings must be interpreted in light of the broader climate\u0026ndash;health literature. Multi-country time-series and case-crossover studies show that both heat and cold can increase short-term risks of cardiovascular and respiratory death, particularly among vulnerable groups [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e]. Our analysis, by contrast, uses long-term climatic averages at the country scale and cannot capture heatwaves, intra-annual variability, or urban heat islands. The apparent protective association of higher temperature for diabetes and respiratory mortality is therefore more likely to reflect residual confounding by unmeasured factors that co-vary with climate (for example, altitude, urban form or service availability) than a true protective effect. Nonetheless, the absence of strong harmful associations after extensive adjustment suggests that, at present, social and health-system determinants remain the primary drivers of cross-country inequality in NCD mortality in SSA, with climate acting as a slower-moving background modifier.\u003c/p\u003e \u003cp\u003eTo situate these climate findings within the broader determinants, it is important to recall that the climate-extended models also included shared socio-economic and health-system covariates. Across the four disease groups, higher GDP per capita and health expenditure tended to be associated with higher observed mortality, which likely reflects a mix of more advanced epidemiological transition, better certification and coding of cause of death, and competing-risk structures in better-resourced systems, rather than a causal effect of under-investment being protective. Urbanisation and HIV prevalence showed disease-specific patterns, with particularly strong positive associations for diabetes and chronic respiratory disease, consistent with urban lifestyles, air pollution and chronic HIV-related co-morbidities [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. Against this backdrop, the temperature and precipitation coefficients can be interpreted as climate associations that are already adjusted for major socio-economic and epidemiological gradients, not simple ecological correlations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e4.5 Model diagnostics and calibration\u003c/h2\u003e \u003cp\u003ePosterior predictive diagnostics indicated that the models were generally well calibrated. CPO failure counts were effectively zero for all disease-specific models, indicating numerical stability of the INLA approximations [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. PIT histograms were close to uniform, suggesting that posterior predictive distributions were not systematically biased or over-/under-dispersed, and observed versus posterior mean plots showed tight clustering around the identity line for most of the mortality range, with greater dispersion only in the tails.\u003c/p\u003e \u003cp\u003eThese results support the adequacy of the chosen latent Gaussian structure, BYM2 spatial random effects, RW1 temporal trends and independent region\u0026ndash;year interactions with PC priors that penalise excessive complexity [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e]. While further refinements are possible, for example through non-linear effects or alternative temporal structures, there is no strong diagnostic evidence of major structural misspecification. This increases confidence that the residual spatial patterns and hotspots we report reflect genuine signal rather than artefacts of overfitting or numerical instability.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e4.6 Policy implications\u003c/h2\u003e \u003cp\u003eSeveral policy-relevant messages emerge from these findings. First, the identification of persistent multi-disease hotspots after adjustment for expected deaths and macro-level covariates suggests that certain countries and subregions are systematically \u0026ldquo;left behind\u0026rdquo; in terms of NCD prevention, diagnosis and treatment. These areas should be prioritised for regional initiatives to scale up affordable antihypertensive and glucose-lowering therapies, strengthen continuity of care for chronic conditions and integrate NCD management into primary care and HIV platforms [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eSecond, the moderate spatial correlations between cardiovascular, diabetes and cancer mortality indicate that investments in core health-system functions such as primary care, essential diagnostics, and tobacco and alcohol control, are likely to generate benefits across multiple NCDs rather than in isolation [\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e, \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e]. By contrast, the weaker and sometimes divergent spatial patterns for chronic respiratory disease point to a need for targeted action on household fuel transitions, occupational health and post-tuberculosis lung disease in specific high-burden settings [\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eFinally, although long-term climate variables were not dominant drivers of cross-country differences in this analysis, they should not be neglected. Climate change is expected to amplify extreme heat events, alter air pollution patterns and interact with food and water systems, all of which may affect future NCD trajectories in SSA [\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Strengthening NCD services and building climate-resilient health systems are therefore complementary priorities rather than competing agendas.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Conclusions and future directions","content":"\u003cp\u003eIn summary, this Bayesian spatial epidemiology study shows that NCD mortality in SSA is characterised by marked residual spatial clustering, persistent multi-disease hotspots and partially shared spatial structures across major causes, even after adjusting for demographic, socio-economic, health-system and climatic variables. Long-term temperature and precipitation patterns showed only modest associations with mortality at the country level, whereas geography, socio-economic conditions and health-system characteristics remained the dominant correlates of cross-country inequality in NCD outcomes.\u003c/p\u003e \u003cp\u003eFuture research should extend this framework by: (i) incorporating subnational mortality or hospitalisation data to identify within-country hotspots; (ii) combining long-term climatic normals with dynamic indicators of heatwaves, drought and air pollution; and (iii) developing multivariate models that jointly analyse NCDs and infectious diseases such as HIV and tuberculosis, to better capture syndemic interactions. By integrating robust Bayesian disease mapping with climate and socio-economic gradients, the approach presented here offers a transferable template for tracking spatial convergence or divergence in NCD risk and for guiding regional prioritisation in an era of epidemiological and climatic transition.\u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cp\u003eASMR – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Age-standardised mortality rate\u003c/p\u003e\n\u003cp\u003eBYM2 –\u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Besag–York–Mollié 2 spatial model\u003c/p\u003e\n\u003cp\u003eCPO – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Conditional predictive ordinate\u003c/p\u003e\n\u003cp\u003eCrI – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Credible interval\u003c/p\u003e\n\u003cp\u003eCVD – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Cardiovascular disease\u003c/p\u003e\n\u003cp\u003eDIC – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Deviance Information Criterion\u003c/p\u003e\n\u003cp\u003eGDP – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Gross domestic product\u003c/p\u003e\n\u003cp\u003eHIV – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Human immunodeficiency virus\u003c/p\u003e\n\u003cp\u003eINLA – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Integrated nested Laplace approximation\u003c/p\u003e\n\u003cp\u003eIRR – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Incidence rate ratio\u003c/p\u003e\n\u003cp\u003eNCD – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Non-communicable disease\u003c/p\u003e\n\u003cp\u003ePIT – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Probability integral transform\u003c/p\u003e\n\u003cp\u003eRR – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Relative risk\u003c/p\u003e\n\u003cp\u003eRW1 – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;First-order random walk\u003c/p\u003e\n\u003cp\u003eSADC – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Southern African Development Community\u003c/p\u003e\n\u003cp\u003eEAC – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;East African Community\u003c/p\u003e\n\u003cp\u003eECOWAS – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Economic Community of West African States\u003c/p\u003e\n\u003cp\u003eSSA – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;\u0026nbsp;Sub-Saharan Africa\u003c/p\u003e\n\u003cp\u003eWAIC – \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp; \u0026nbsp;Watanabe–Akaike Information Criterion\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable. This study did not involve human participants, human data, human tissue, or animals.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable. This manuscript does not contain data from any individual person.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets analysed during the current study were derived from publicly available repositories, including the World Health Organisation (WHO) Global Health Observatory as follows:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ehttps://www.who.int/data/gho \u0026nbsp; ;\u003c/p\u003e\n\u003cp\u003eWorld Development Indicators as follows:\u003c/p\u003e\n\u003cp\u003ehttps://databank.worldbank.org/source/world-development-indicators ;\u003c/p\u003e\n\u003cp\u003eand WorldClim v2.1 data for climatic factors (precipitation and rainfall) as follows:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003ehttps://www.worldclim.org/\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCustom scripts and code developed NCDs modelling are available from the corresponding author upon reasonable request.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.\u003cstrong\u003e\u003cbr\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTSC\u003c/strong\u003e (Tsikai Solomon Chinembiri): Conceptualized the study, designed the research framework, conducted statistical modelling and Bayesian inference, and led the manuscript writing and interpretation of climate\u0026ndash;health linkages.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGP\u003c/strong\u003e (Godfrey Pachavo): Contributed to data acquisition, pre-processing, and interpretation of socioeconomic indicators.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe gratefully acknowledge the World Health Organisation (WHO) Global Health Observatory for providing access to the datasets that enabled this analysis. We also extend our thanks to colleagues and peer reviewers whose insightful feedback enhanced the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors\u0026apos; information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eOptional \u0026mdash; not included.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAbbafati C, Abbas KM, Abbasi M, Abbasifard M, Abbasi-Kangevari M, Abbastabar H, Global Burden of Disease Study 2019. Global burden of 369 diseases and injuries in 204 countries and territories, 1990\u0026ndash;2019: a systematic analysis for the. Lancet. 2020;396:1204\u0026ndash;22. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S0140-6736(20)30925-9/ATTACHMENT/1802C2B8-7CCC-467E-B4DD-92F466CF5E15/MMC2E.PDF\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(20)30925-9/ATTACHMENT/1802C2B8-7CCC-467E-B4DD-92F466CF5E15/MMC2E.PDF\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Bank Group. The Worldwide Governance Indicators : Methodology and 2024 Update (English). Washington DC; 2024.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAvard Rue H˚, Martino S, Chopin N. Approximate Bayesian Inference for Latent Gaussian Models Using Integrated Nested Laplace Approxima-tions.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWang S, Ren Z, Liu X, Yin Q. Spatiotemporal trends in life expectancy and impacts of economic growth and air pollution in 134 countries: A Bayesian modeling study. Soc Sci Med. 2022;293:114660. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/https://doi.org/10.1016/j.socscimed.2021.114660\u003c/span\u003e\u003cspan address=\"10.1016/j.socscimed.2021.114660\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlangiardo M, Cameletti M. Spatial and Spatio-temporal Bayesian Models with R - INLA. Spatial and Spatio-temporal Bayesian Models with R - INLA. 2015;:1\u0026ndash;308. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/9781118950203\u003c/span\u003e\u003cspan address=\"10.1002/9781118950203\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBesag J, York J, Molli\u0026acute;e. Bayesian image restoration, with two applications in spatial statistics. Ann Inst Stat Math. 1991;43:1\u0026ndash;59.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRue H, Martino S, Chopin N. Approximate Bayesian inference for latent Gaussian models by using integrated nested Laplace approximations. J R Stat Soc Ser B Stat Methodol. 2009;71:319\u0026ndash;92. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/https://doi.org/10.1111/j.1467-9868.2008.00700.x\u003c/span\u003e\u003cspan address=\"10.1111/j.1467-9868.2008.00700.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFick SE, Hijmans RJ. WorldClim 2: New 1-km spatial resolution climate surfaces for global land areas. Int J Climatol. 2017;37:4302\u0026ndash;15. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/https://doi.org/10.1002/joc.5086\u003c/span\u003e\u003cspan address=\"10.1002/joc.5086\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDemirhan H, Kalaylioglu Z. Joint prior distributions for variance parameters in Bayesian analysis of normal hierarchical models. J Multivar Anal. 2015;135:163\u0026ndash;74. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/https://doi.org/10.1016/j.jmva.2014.12.013\u003c/span\u003e\u003cspan address=\"10.1016/j.jmva.2014.12.013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLawson A. Bayesian Disease Mapping: Hierarchical Modeling in Spatial Epidemiology. CRC; 2018.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Bank. World Development Indicators. Washington, D.C., USA; 2023.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee D, Rushworth A, Napier G. Spatio-Temporal Areal Unit Modeling in R with Conditional Autoregressive Priors Using the CARBayesST Package. J Stat Softw. 2018. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.18637/jss.v084.i09\u003c/span\u003e\u003cspan address=\"10.18637/jss.v084.i09\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. 84 9 SE-Articles:1\u0026ndash;39.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBanerjee S, Carlin BP, Gelfand AE. Hierarchical modeling and analysis for spatial data. Chapman and Hall/CRC; 2003.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMahaki B, Mehrabi Y, Kavousi A, Akbari ME, Waldhoer T, Schmid VJ, et al. Multivariate disease mapping of seven prevalent cancers in Iran using a shared component model. Asian Pac J Cancer Prev. 2011;12:2353\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSimpson D, Rue H, Riebler A, Martins TG, S\u0026oslash;rbye SH. Penalising model component complexity: A principled, practical approach to constructing priors. 2017.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMiranda JJ, Kinra S, Casas JP, Davey Smith G, Ebrahim S. Non-communicable diseases in low- and middle-income countries: context, determinants and health policy. Trop Med Int Health. 2008;13:1225\u0026ndash;34. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/J.1365-3156.2008.02116.X\u003c/span\u003e\u003cspan address=\"10.1111/J.1365-3156.2008.02116.X\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGouda HN, Charlson F, Sorsdahl K, Ahmadzada S, Ferrari AJ, Erskine H, et al. Burden of non-communicable diseases in sub-Saharan Africa, 1990\u0026ndash;2017: results from the Global Burden of Disease Study 2017. Lancet Glob Health. 2019;7:e1375\u0026ndash;87.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRiebler Andrea H, Simpson, Daniel. Rue H\u0026aring;vard. An intuitive Bayesian spatial model for disease mapping that accounts for scaling. Stat Methods Med Res. 2016;25:1145\u0026ndash;65. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/0962280216660421\u003c/span\u003e\u003cspan address=\"10.1177/0962280216660421\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdeloye D, Basquill C, Aderemi AV, Thompson JY, Obi FA. An estimate of the prevalence of hypertension in Nigeria: a systematic review and meta-analysis. J Hypertens. 2015;33:230\u0026ndash;42. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1097/HJH.0000000000000413\u003c/span\u003e\u003cspan address=\"10.1097/HJH.0000000000000413\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAmegah AK, Jaakkola JJK. Household air pollution and the sustainable development goals. Bull World Health Organ. 2016;94:215\u0026ndash;21. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2471/BLT.15.155812\u003c/span\u003e\u003cspan address=\"10.2471/BLT.15.155812\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFeinstein MJ, Bogorodskaya M, Bloomfield GS, Vedanthan R, Siedner MJ, Kwan GF, et al. Cardiovascular Complications of HIV in Endemic Countries. Curr Cardiol Rep. 2016;18:113. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11886-016-0794-x\u003c/span\u003e\u003cspan address=\"10.1007/s11886-016-0794-x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGasparrini A, Guo Y, Hashizume M, Lavigne E, Zanobetti A, Schwartz J, et al. Mortality risk attributable to high and low ambient temperature: a multicountry observational study. Lancet. 2015;386:369\u0026ndash;75. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S0140-6736(14)62114-0\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(14)62114-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWatts N, Amann M, Arnell N, Ayeb-Karlsson S, Beagley J, Belesova K, et al. The 2020 report of The \u0026lt;\u0026thinsp;em\u0026thinsp;\u0026gt;\u0026thinsp;Lancet\u0026thinsp;Countdown on health and climate change: responding to converging crises. Lancet. 2021;397:129\u0026ndash;70. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/S0140-6736(20)32290-X\u003c/span\u003e\u003cspan address=\"10.1016/S0140-6736(20)32290-X\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-health-geographics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijhg","sideBox":"Learn more about [International Journal of Health Geographics](http://ij-healthgeographics.biomedcentral.com/)","snPcode":"12942","submissionUrl":"https://submission.nature.com/new-submission/12942/3","title":"International Journal of Health Geographics","twitterHandle":"@IJHGeo","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"Non-communicable diseases, Spatial epidemiology, Bayesian spatio-temporal modelling, Climate variability, Sub-Saharan Africa","lastPublishedDoi":"10.21203/rs.3.rs-8250412/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8250412/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eBackground\u003c/h2\u003e \u003cp\u003eNon-communicable diseases (NCDs) now account for a growing share of premature mortality in sub-Saharan Africa (SSA), yet little is known about how climate and geography shape spatial inequalities in NCD deaths.\u003c/p\u003e\u003ch2\u003eMethods\u003c/h2\u003e \u003cp\u003eWe assembled a country\u0026ndash;year panel for forty-one SSA countries from 2000\u0026ndash;2019, combining World Health Organization mortality estimates for four major NCD groups namely cardiovascular diseases, diabetes mellitus, malignant neoplasms and chronic respiratory diseases, with mid-year population denominators, climate surfaces (mean temperature and precipitation from WorldClim v2.1) and macro-socioeconomic covariates. Expected deaths were derived from age-standardised mortality rates and used as offsets in disease-specific Bayesian Poisson spatio-temporal models with Besag\u0026ndash;York\u0026ndash;Molli\u0026eacute; 2 (BYM2) spatial random effects, first-order random walk temporal effects, and country\u0026ndash;year interaction terms. Models were fitted in INLA with penalised complexity priors. We mapped climate-adjusted spatial random effects, identified multi-disease hotspots using posterior relative risks averaged over 2015\u0026ndash;2019, and quantified cross-disease spatial correlations at SSA and regional-bloc level. Predictive performance was assessed using conditional predictive ordinates, PIT histograms and observed-versus-fitted plots.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eAfter adjustment for socioeconomic, health-system and HIV indicators, substantial residual spatial clustering remained, with elevated NCD mortality in parts of southern and eastern Africa and lower risk in several Sahelian countries. Countries classified as hotspots for one NCD often exhibited raised risks for others. Spatial correlations were positive between cardiovascular disease and diabetes, but negative between cardiovascular and respiratory mortality. Higher long-term temperature was associated with lower diabetes and respiratory mortality, while precipitation effects were generally weak.\u003c/p\u003e\u003ch2\u003eConclusions\u003c/h2\u003e \u003cp\u003eNCD mortality in SSA displays marked, climate-adjusted spatial heterogeneity and partially shared hotspot patterns across causes. These findings support geographically targeted, climate-sensitive NCD prevention and health-system strengthening strategies.\u003c/p\u003e\u003ch2\u003eTrial registration:\u003c/h2\u003e \u003cp\u003eNot applicable.\u003c/p\u003e","manuscriptTitle":"Climate, Spatial Clustering and Hotspots of Non-Communicable Disease Mortality in Sub-Saharan Africa: A Bayesian Spatial Epidemiology Study, 2000–2019","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-18 12:05:00","doi":"10.21203/rs.3.rs-8250412/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2026-05-14T22:21:49+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"192822186885831866682505989047484556514","date":"2026-02-19T15:49:28+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"152506463165878235346598179397127192051","date":"2025-12-24T11:12:33+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"135778048774734627038292205711570815457","date":"2025-12-18T14:53:24+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-12-16T19:58:12+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-12-02T09:11:18+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-12-02T09:10:50+00:00","index":"","fulltext":""},{"type":"submitted","content":"International Journal of Health Geographics","date":"2025-12-01T12:21:58+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"international-journal-of-health-geographics","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ijhg","sideBox":"Learn more about [International Journal of Health Geographics](http://ij-healthgeographics.biomedcentral.com/)","snPcode":"12942","submissionUrl":"https://submission.nature.com/new-submission/12942/3","title":"International Journal of Health Geographics","twitterHandle":"@IJHGeo","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"em","reportingPortfolio":"BMC/SO AJ","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"97dc9161-bedd-4d9c-8c85-09c387262c55","owner":[],"postedDate":"December 18th, 2025","published":true,"recentEditorialEvents":[{"type":"editorInvitedReview","content":"","date":"2026-05-14T22:21:49+00:00","index":58,"fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-12-18T12:05:00+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-18 12:05:00","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8250412","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8250412","identity":"rs-8250412","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}