Bayesian spatio-temporal modelling of tuberculosis in Vietnam: Insights from a local-area analysis

Spatial analysis and disease mapping have the potential to enhance understanding of tuberculosis (TB) dynamics, whose spatial dynamics may be complicated by the mix of short and long-range transmission and long latency periods. TB notifications in Nam Dinh Province for individuals aged 15 and older from 2013 to 2022 were analyzed with a variety of spatio-temporal methods. The study commenced with an analysis of spatial autocorrelation to identify clustering patterns, followed by the evaluation of several candidate Bayesian spatio-temporal models. These models varied from simple assessments of spatial heterogeneity to more complex configurations incorporating covariates and interactions. The findings highlighted a peak in the TB notification rate in 2017, with 98 cases per 100,000 population, followed by a sharp decline in 2021. Significant spatial autocorrelation at the commune level was detected over most of the 10-year period. The Bayesian model that best balanced goodness-of-fit and complexity indicated that TB trends were associated with poverty: each percentage point increase in the proportion of poor households was associated with a 1.3% increase in TB notifications, emphasizing a significant socioeconomic factor in TB transmission dynamics. The integration of local socioeconomic data with spatio-temporal analysis could further enhance our understanding of TB epidemiology.