Temperature-related mortality burden and future climate projections across 88 counties in a plateau region of southwestern China

Background We aimed to quantify temperature-attributable mortality across population subgroups and project future cold- and heat-attributable mortality burdens in 88 counties of Guizhou Province, southwestern China. Methods We analysed daily non-accidental mortality, 2-m air temperature, relative humidity and air-pollution data across 88 counties during 2020–2024. County-specific quasi-Poisson distributed lag non-linear models with lags of 0–28 days were pooled using multivariate meta-regression. County-specific best linear unbiased prediction exposure–response functions and minimum mortality temperatures were used to estimate daily attributable fractions and numbers, with uncertainty quantified by Monte Carlo simulation. Sex- and age-group differences in attributable fractions were assessed using two-sided Z tests. Projections for 2030–2099 combined bias-corrected daily temperatures from five CMIP6 models, SSP-specific county population totals and scenario- and year-specific provincial age and sex composition applied to each county under SSP2-4.5 and SSP5-8.5. Temperature–mortality functions, minimum mortality temperatures, baseline mortality rates and seasonal mortality patterns were held constant. Results Among 1,103,188 non-accidental deaths, the overall temperature-mortality association was inverse J-shaped, with a pooled minimum mortality temperature of 20.58 °C. The overall cold- and heat-attributable fractions were 8.66% (95% eCI, 7.37–9.45%) and 2.95% (2.32–3.40%), respectively. Cold-attributable fractions were highest for cardiovascular mortality at 11.08%, whereas heat-attributable fractions were highest among females at 4.03%. Under SSP2-4.5, total attributable deaths increased from 349,515 in the 2030s to 580,063 in the 2090s; and cold- and heat-attributable fractions converged to 6.48% and 6.26%, respectively, by the 2090s. Under SSP5-8.5, heat-attributable deaths exceeded cold-attributable deaths from the 2060s onward, and the heat-attributable fraction increased from 3.68% to 14.10% by the 2090s. Meta-regression identified exploratory associations between county-level characteristics and heterogeneity in temperature-mortality associations. Conclusions Temperature-attributable mortality in Guizhou is currently dominated by cold exposure, but future warming is projected to reshape both its relative composition and absolute burden. SSP2-4.5 is projected to produce coexisting cold- and heat-related burdens, whereas SSP5-8.5 is projected to shift towards heat-dominated burden. Climate-health strategies should retain cold-risk protection while strengthening heat-health adaptation, with particular attention to potentially vulnerable groups, including older adults and women in relation to heat.