Ambient temperature variations and antimicrobial resistance: a multicountry analysis of non-linear relationships and...

Antimicrobial resistance (AMR) is an escalating global health crisis being worsened by climate change. Studies of temperature-AMR associations remain limited by geographic scope, time frames and linear approaches. We aimed to identify temperature thresholds where AMR dynamics shift across pathogen-drug combinations, periods and socioeconomic contexts.

We analysed data from 56 countries and territories over 24 years (1999-2022), focusing on six WHO-designated 'critical' antibiotic-resistant pathogens. Using segmented regression models with mean ambient temperature as the primary independent variable, we divided the study period into three intervals (1999-2006, 2007-2014, 2015-2022) and incorporated 16 socioeconomic and environmental covariates. Model robustness was validated through bootstrap cross-validation with 1000 resamples.

We identified distinct temperature association thresholds for each pathogen-drug combination (5.7°C-18.4°C). Below these thresholds, resistance rates consistently decreased with rising temperatures; above thresholds, responses varied by phenotype and time period. For third-generation cephalosporin-resistant Escherichia coli, each 1°C rise above 16.5°C corresponded to a 0.46% increase (95% CI 0.28% to 0.65%; p<0.001) during 1999-2006, a 0.11% decrease (95% CI -0.17% to -0.06%; p<0.001) during 2007-2014 and a 0.18% increase (95% CI 0.11% to 0.25%; p<0.001) during 2015-2022. The Corruption Perception Index showed consistent negative associations with resistance rates, particularly for carbapenem-resistant Acinetobacter baumannii (β=-1.40, p<0.001).

Phenotype-specific temperature association thresholds provide descriptive, hypothesis-generating reference points for understanding how AMR burden varies along the global temperature gradient. The observed temporal heterogeneity suggests complex patterns requiring long-term monitoring and climate-adaptive AMR control strategies considering both phenotype-specific temperature sensitivities and socioeconomic contexts.