In Ghana, disparities in health risks and child mortality, mainly attributed to socioeconomic and environmental factors, exhibit significant geographic characteristics. We employed a comprehensive approach that integrated advanced spatiotemporal modeling techniques to address the intricate interplay between spatial and temporal dynamics underlying the risk of under-five pneumonia (U-5P) in Ghana. Pneumonia cases for children under five from 2018 to 2022 were obtained from the District Health Information Management System-2 (DHIMS-2). We coupled an integrated nested Laplace approximation (INLA) with an intrinsic conditional autoregressive (ICAR) model based on the Besag-York-Mollie (BYM2) model to estimate the relative risk of U-5P between 2018 and 2022. Temporal dependences were handled using a random walk order 1. Furthermore, we used Local Indicators of the Spatial Association (LISA) to detect spatial clusters of the relative risk estimates calculated with INLA. Overall, there was statistically significant variation in cases across years (K2 = 115.63, p-value < 0.0001). The spatial distribution of the relative risk of U-5P was not random across all the years studied: 2018 (z-score = 2.7917; p = 0.00523), 2019 (z-score = 3.48; p = 0.0048), 2021 (z-score = 3.83; p = 0.000126), and 2022 (z-score = 5.50; p < 0.0001), except for 2020 (z-score = 0.38; p = 0.699). Specifically, there was high spatial variability in risk, with districts in the Greater Accra region consistently experiencing lower-than-expected risk over the years studied. In contrast, those in the Volta region had a consistently higher-than-expected risk. Our findings have two main takeaways. First, there is spatial heterogeneity in U-5P risk levels, with statistically significant clusters. Second, the spatial pattern may persist or even intensify with time. Therefore, a real-time surveillance system that integrates time and geography must be developed.
Asori et al. (Wed,) studied this question.