Weekly extremes of PM2.5-attributable cardiovascular mortality in Northeastern Thailand: Extreme-value return levels and 26-week forecasts

In Northeastern (NE) Thailand, cardiovascular mortality can increase during brief PM 2.5 surges; however, subnational health risks associated with extreme events remain poorly characterized. We constructed a province-week panel (2017–2023) by linking de-identified vital statistics for cardiovascular disease (CVD) and ischemic heart disease (IHD) with routine PM 2.5 monitoring data. The analysis was restricted to 10 of 20 NE provinces with continuously operating monitors meeting predefined completeness criteria, enabling consistent weekly time-series inference. We estimated weekly attributable deaths (AD wk ) using short-term concentration–response functions with lag 01 exposure over a WHO-consistent counterfactual (b 15 = 15 µg/m 3 ). We then quantified extreme mortality surges using Extreme Value Theory (EVT) via generalized extreme-value (GEV) models for annual maxima and peaks-over-threshold generalized Pareto (POT–GPD) models for weekly exceedances, yielding return levels (RL T , T = 2–50 years). Return-level results indicated that rare haze-related weeks can substantially exceed typical weekly burdens, with risks clustering in the dry season (November-April; peaking December-March). The highest return levels occurred in Khon Kaen, Nakhon Ratchasima, and Ubon Ratchathani, where RL 5 exceeded baseline AD wk by more than 2.5-fold. Nine province-cause series (CVD = 4; IHD = 5) met quality standards and showed acceptable cross-framework RL 5 concordance (Set B). Complementary 26-week province-specific forecasts provided near-term situational awareness consistent with seasonal haze dynamics. Although the analysis relies on ambient PM 2.5 observations without source apportionment, the findings underscore the public health significance of episodic haze extremes and support early-warning and preparedness strategies alongside long-term air-quality mitigation.