Global tuberculosis incidence and its macrostructural determinants: A hybrid Explainable Boosting Machine–Bayesian Structural Equation Modeling analysis of socioeconomic, health system, and population-risk pathways

Abstract

Tuberculosis (TB) remains a leading infectious cause of mortality worldwide, with marked cross-country disparities shaped by socioeconomic conditions, health system capacity, and population-level risk factors. The aim of this study was to investigate the macrostructural determinants of global TB incidence and to clarify the pathways through which socioeconomic, health system, and population-risk domains are associated with TB incidence. Country-level secondary data were obtained from WHO, World Bank Open Data, and UNDP. A hybrid analytical framework was applied in two stages. First, EBM was used to identify non-linear predictor signals and rank the relative importance of macrostructural determinants of TB incidence. Second, the empirically derived predictors were translated into BSEM to estimate latent measurement structures and structural pathways toward TB burden. A total of 172 countries were included in the final analytical dataset. In the EBM phase, the strongest predictor signal was observed for Education Index, followed by diabetes prevalence, HIV incidence, TB case detection rate, BCG coverage, and physician density. These signals were subsequently mapped into three latent constructs: socioeconomic conditions, health system capacity, and population-level health risk. In the final BSEM model, socioeconomic showed the strongest association with health system capacity (β=0.62), indicating that stronger socioeconomic conditions were associated with greater TB-relevant service capacity. Health system was subsequently associated with lower TB incidence (β=-0.26). Direct associations with TB incidence were also observed for socioeconomic (β=-0.46) and health risk (β=-0.33), although the health risk pathway required cautious interpretation because of limited latent coherence at the country level. The final model explained a substantial proportion of cross-country variation in TB incidence (posterior R^²=0.53). These findings indicate that the global TB burden is shaped by interconnected structural pathways rather than by a single macroeconomic gradient. Socioeconomic conditions appeared to influence TB burden primarily through health system capacity, particularly diagnostic reach, workforce availability, case detection, and immunization coverage. The hybrid EBM–BSEM framework provides an interpretable approach for identifying upstream determinants of TB burden and may support evidence-based prioritization of global TB prevention strategies.