Geospatial determinants of diabetes risk in Thailand: socioeconomic and health service factors

Diabetes prevalence is increasing in Thailand, creating growing demands on the health system. Understanding the spatial distribution of diabetes risk and its association with socioeconomic and healthcare system factors among the diabetes risk population is critical for designing targeted prevention and intervention strategies. We examined the distribution of diabetes risk groups across provinces in Thailand with reference to the spatial association between economic, social and public health service factors based on data from the Ministry of Public Health's Health Data Center (HDC) for the year 2021. The dataset included 22,491,934 individuals across the 76 provinces as well as social, economic and public health services. The methods included Local Indicators of Spatial Association (LISA), Ordinary Least Squares (OLS), Spatial Lag Model (SLM) and Spatial Error Model (SEM). Explanatory variables included average night-time light intensity, average monthly income, hospital-to-population ratio and proportion of the population with health insurance. Major clusters of High-High (HH) diabetes risk were identified by LISA mainly located in the North of Thailand. In all models, the direction and significance of the associations were consistent (p<0.001 for all variables investigated and p<0.01). R2=0.47. The SLM gave the best fit, capturing spatial spill-over effects. Higher night-time light intensity (coefficient = -85.70, p<0.05) and higher monthly income (coefficient = -0.079, p<0.001) were negatively associated with diabetes risk. These inverse relationships implied that greater urbanization and higher socio-economic standing may protect against diabetes risk, possibly through improved access to health infrastructure, improved health education and preventive services. Conversely, the higher hospital-to-population ratios (coefficient = 572.28, p<0.001) and the larger proportions of Civil Servant Medical Benefits Scheme (CSMBS) coverage (coefficient = 226.46, p<0.001) the higher diabetes risk. These counterintuitive findings likely reflect reverse causation, in which provinces with higher disease burden or poor health attract more resources of health care and have increased insurance coverage, a pattern consistent with healthcare service distribution responding to existing health needs rather than preventing occurrence of disease.