From behavioral to structural correlates: a multi-phase analysis of global reported mpox.

BACKGROUND: Mpox epidemiology has expanded from endemic zoonotic spillover to sustained human-to-human transmission. While behavioral risk factors have been well recognized, how macro-level correlates of mpox burden vary across epidemic phases remains insufficiently quantified, limiting targeted public health responses. METHODS: Using World Health Organization (WHO) global daily mpox surveillance data from January 1, 2022 to April 30, 2025, we calculated country-period reported mpox incidence and developed a phase-sensitive analytical framework integrating multi-domain macro-level variables across distinct epidemic phases. Five ensemble machine learning models (eXtreme Gradient Boosting, Gradient Boosting Decision Tree, Light Gradient Boosting Machine, Random Forest, Categorical Boosting) were optimized, and their outputs were interpreted using SHapley Additive exPlanations (SHAP) values to identify the changing correlates of mpox burden across phases. RESULTS: Reported mpox incidence exhibited four temporal-epidemiological phases, comprising localized sporadic emergence, accelerated global spread, sustained dissemination, and region-specific resurgence. SHAP-based interpretation revealed a phase-dependent shift in model-associated correlates, from behavioral factors in earlier phases, such as close-contact networks and population mobility, to later structural vulnerability indicators related to socioeconomic disparities and health-system capacity. The LGBT+ (lesbian, gay, bisexual, transgender, and other sexual and gender minorities) Legal Index was the strongest overall predictor, accounting for 13.5% of the overall SHAP importance. Its association with reported mpox incidence was context-dependent, with positive contributions occurring in legally inclusive settings and negative contributions observed in restrictive settings, a pattern consistent with differential case ascertainment and reporting visibility. CONCLUSIONS: Our findings highlight a transition from behavioral to structural indicators in global mpox burden, uncovering phase-specific and geographic heterogeneity. These results support phase-tailored interventions that combine early targeted prevention with later health-system strengthening, equitable surveillance, community support, and stigma reduction.