Spatiotemporal heterogeneity and the shifting epicenter of pulmonary tuberculosis notifications in thailand over a 10-year period: Evidence from national surveillance data

Pulmonary tuberculosis (PTB) remains a critical public health challenge in Thailand. While national incidence rates have shown a gradual decline, aggregate statistics often conceal localized epidemic dynamics. This study aimed to identify statistically significant spatiotemporal clusters of PTB and analyze shifting patterns of disease epicenters in Thailand from 2014 to 2023 using global and local spatial statistics. A retrospective ecological study was conducted using national PTB surveillance data from the Bureau of Epidemiology, Ministry of Public Health, Thailand, covering all 77 provinces for 2014–2023. Three complementary spatial methods were employed at different analytical scales, all at the province level, the finest spatial resolution available in the national notification system. Global spatial autocorrelation was assessed using Moran’s I with inverse distance weighting (IDW). Local Indicators of Spatial Association (LISA) were used to identify local clustering patterns and spatial outliers. Discrete Poisson space-time scan statistics (SaTScan™) were employed to detect high-risk clusters. The completeness of provincial notification was assessed for every province-year, and the spatial distribution of incomplete reporting was compared across the detected clusters. Global Moran’s I indicated no significant spatial clustering nationally (Z-scores: -1.08 to 1.33). However, LISA analysis revealed significant local heterogeneity with the temporal evolution of spatial patterns. Space-time scan statistics identified seven clusters. The most likely cluster during 2014–2019 was in the Eastern industrial region (RR = 3.64, LLRatio = 1,463.28), whereas during 2020–2023 it was in the Lower Northeastern border region (Ubon Ratchathani, Si Sa Ket, Amnat Charoen, Yasothon, Roi Et; RR = 3.22, LLRatio = 3,145.62, P < 0.001). Assessment of notification completeness showed that this change in cluster location cannot be separated from a change in reporting coverage: 154 of 770 province-years (20.0%) recorded zero notified cases, concentrated in 2014–2016 (27, 30 and 37 of 77 provinces respectively), and the gap was not spatially uniform. The Eastern cluster notified cases in every year of the study, whereas 47% of province-years in the Lower Northeastern cluster and 100% of those in the Prachuap Khiri Khan cluster recorded zero notifications during 2014–2016. This 10-year analysis revealed substantial spatiotemporal heterogeneity in Thailand’s notified pulmonary tuberculosis distribution, with the most likely cluster changing from Eastern industrial zones to the Lower Northeastern border region. Because the early-period reporting gap was concentrated in the provinces that later formed the Lower Northeastern cluster, and absent from the provinces forming the Eastern cluster, this change in cluster location is reported as a change in the geography of notification and not as a demonstrated change in transmission. While global spatial statistics indicated no nationwide clustering, local-scale analysis successfully identified significant hotspots and their temporal evolution, underscoring the necessity of multi-scale analytical approaches in tuberculosis epidemiology. These findings indicate that notification completeness must be assessed before space-time cluster detection is applied to passive surveillance data, and support geographically-targeted strengthening of reporting as a prerequisite for interpreting apparent spatial shifts in tuberculosis burden.