Climate change and antimicrobial resistance: a global-scale analysis

Abstract

Background

Antimicrobial resistance (AMR) represents a major global health threat. Although regional studies have explored the relationship between climate change and AMR, a comprehensive global analysis incorporating extreme climate events has not yet been conducted.

Methods

We analyzed global data from 2000 to 2023, encompassing over 28 million bacterial isolates from eight common pathogens and 14 antibiotic categories. Climate data were sourced from NOAA, and resistance data were obtained from ResistanceMap, ECDC, and PLISA databases. Linear mixed-effects models (LMMs) were applied to evaluate the associations between climate indices and resistance rates.

Results

Temperature was consistently positively correlated with resistance rates across most bacterial species. The mean temperature was significantly associated with resistance rates even after adjusting for covariates. Extreme temperature indicators, including intensity indices (TXx, TNx, TXn and TNn), absolute threshold indices (SU, TR and DTR), relative threshold indices (TN90p and TX90p), and duration indices (CSDI and WSDI) exhibited significant positive correlations with resistance rates. In contrast, cold-related indices (FD, ID, TN10p and TX10p) were negatively correlated with resistance rates. Among the precipitation indices, only CDD demonstrated a significant positive association with aggregated AMR after full adjustment; all the other precipitation metrics showed no statistically significant correlation. Furthermore, subgroup analyses of WHO priority pathogens confirmed the robust effect of temperature, but revealed that precipitation indices, particularly CDD, had opposing correlations with resistance across different pathogens.

Conclusions

This study provides robust global evidence that rising temperatures and extreme heat are consistent drivers of AMR, whereas the impact of precipitation is complex and pathogen dependent. These findings underscore the need for climate-informed public health strategies that integrate climate surveillance into AMR action plans to develop targeted interventions against these intertwined global threats.

Introduction

Antimicrobial resistance (AMR) is recognized as one of the top ten global health threats, representing a critical public health concern that requires novel strategies to combat it [1]. The emergence of multidrug-resistant bacterial pathogens, often referred to as “superbugs”, poses a continuously rising threat to public health worldwide [2]. This crisis contributes to significant mortality, with bacterial AMR being directly responsible for 1.27 million deaths in 2019 alone, as well as prolonged hospital stays, increased healthcare costs, and the failure of standard treatments for common infections [3]. The primary driver of AMR is widely acknowledged to be the selective pressure from the excessive and inappropriate use of antibiotics [4–6]. Beyond this primary driver, the landscape of resistance is shaped by a confluence of other critical factors. Socioeconomic conditions and population density, for example, significantly influence transmission dynamics [7]. Furthermore, public health infrastructure, particularly access to basic sanitation and clean water, has been identified as a crucial determinant in mitigating the spread of resistant pathogens [8, 9]. Moreover, emerging evidence suggests that environmental factors, including climate change, may also play a significant role in shaping resistance patterns [10–12]. A growing body of research highlights that temperature fluctuations can accelerate resistance in key pathogens, with warmer temperatures potentially enhancing the survival and transmission of resistant bacteria. Climate change, characterized by rising global temperatures and an increased frequency of extreme weather events, is increasingly recognized as a key environmental driver influencing the spread and distribution of resistant pathogens [13, 14]. Mechanistically, warmer temperatures may enhance the horizontal transfer of resistance genes, as well as promote the survival and transmission of resistant bacteria in the environment [15, 16].

The intricate links between climate change, industrialization, and AMR are gaining recognition [14, 17]. Recent studies have begun to correlate AMR gene prevalence with climate and industrial factors at the macro level [14]. However, a large-scale global analysis based on long-term, clinical phenotypic resistance data remains to be conducted. Furthermore, while the effects of average temperature changes have been a primary focus, the impacts of the increasing frequency and intensity of extreme climate events (such as heatwaves, droughts, and cold spells) on global AMR patterns have not yet been systematically quantified.

These extreme events may alter environmental conditions in ways that either directly facilitate the survival of resistant bacteria or disrupt ecosystems, potentially having more profound effects on the evolution and spread of resistance. Therefore, this study aims to address these gaps by analyzing global clinical resistance data in relation to a comprehensive suite of temperature, precipitation, and extreme climate indices from 2000 to 2023. While controlling for key socioeconomic and public health covariates, our goal is to provide the first global-scale evidence on how both average climate shifts and, crucially, extreme weather events influence AMR transmission, thereby offering robust evidence to inform public health policies and guide effective intervention strategies.

Methods

Data sources

This study analyzed global data from 2000 to 2023. This period was selected to ensure the availability of consistent and high-quality surveillance data for robust trend analysis. The year 2000 marked a point where many international surveillance networks began to provide more systematic data, establishing a reliable baseline. The year 2023 was the most recent full year for which comprehensive datasets were available at the time of our data collection.

Climate data were obtained from the National Oceanic and Atmospheric Administration (NOAA), covering the period from 2000 to 2023. Extreme climate indices were calculated using the expert team on climate change detection and indices (ETCCDI) framework, a globally recognized system of 26 climate indices designed to quantify temperature and precipitation extremes. The extreme climate indices were categorized into four groups: intensity indices, absolute threshold indices, relative threshold indices, and duration indices.

Temperature-related indices

Intensity indices include monthly maximum value of daily maximum temperature (TXx); monthly maximum value of daily minimum temperature (TNx); monthly minimum value of daily maximum temperature (TXn); monthly minimum value of daily minimum temperature (TNn). Absolute threshold indices include the number of frost days (FD): annual count of days when TN (daily minimum temperature) < 0 °C; the number of icing days (ID): annual count of days when TX (daily maximum temperature) < 0 °C; the number of summer days (SU): annual count of days when TX (daily maximum temperature) > 25 °C; the number of tropical nights (TR): annual count of days when TN (daily minimum temperature) > 20 °C; daily temperature range (DTR): monthly mean difference between TX and TN. Relative threshold indices include the percentage of days when TN < 10th percentile (TN10p); the percentage of days when TX < 10th percentile (TX10p); the percentage of days when TN > 90th percentile (TN90p); the percentage of days when TX > 90th percentile (TX90p). Duration indices include growing season length (GSL): annual count between the first span of at least 6 days with daily mean temperature TG > 5 °C and the first span after July 1 st (or January 1 st in the Southern Hemisphere) of 6 days with TG < 5 °C; cold spell duration index (CSDI): annual count of days with at least 6 consecutive days when TN < 10th percentile; warm spell duration index (WSDI): annual count of days with at least 6 consecutive days when TX > 90th percentile.

Precipitation-related indices

Intensity indices include monthly maximum 1-day precipitation (Rx1day); monthly maximum consecutive 5-day precipitation (Rx5day); simple precipitation intensity index (SDII). Absolute threshold indices include annual count of days when PRCP ≥ 10 mm (R10mm); annual count of days when PRCP ≥ 20 mm (R20mm); annual total precipitation on wet days (PRCPTOT). Relative threshold indices include annual total PRCP when RR > 95th percentile (R95p); annual total PRCP when RR > 99th percentile (R99p). Duration indices include the maximum length of dry spell (CDD): the maximum number of consecutive days with RR < 1 mm; the maximum length of wet spell (CWD): the maximum number of consecutive days with RR ≥ 1mm.

Detailed definitions of these indices can be found in the ETCCDI climate indices literature [18].

AMR data were collected from several authoritative databases and institutions: ResistanceMap, a global database of AMR [19]; the European Centre for Disease Prevention and Control (ECDC) Surveillance Atlas, which provides AMR data for European countries [20]; and the Platform for Health Information of the Americas (PLISA), which covers AMR data for countries in the Americas [21].

Data on antibiotic usage were obtained from IQVIA, a leading source of global pharmaceutical data [22], as well as from the WHO Antimicrobial Consumption Surveillance Reports, which provide information on antibiotic consumption patterns across various countries [23]. Additionally, data from the European Surveillance of Antimicrobial Consumption Network (ESAC-Net) were used, offering detailed information on antibiotic use across 27 European countries [24]. Key socioeconomic, demographic, and public health infrastructure data were obtained from the World Bank Open Data Platform [25]. This included gross domestic product (GDP) per capita, population density, access to basic drinking water services, and access to basic sanitation services.

A detailed summary of all the data sources used in this study, along with a critical appraisal of their respective advantages and potential limitations, is provided in Table 1.

Table 1.

Summary of data sources, strengths, and limitations

Data Category	Indicator	URL	Strengths	Limitations
Antimicrobial Resistance (AMR) Data	AMR Data	https://resistancemap.onehealthtrust.org https://www.ecdc.europa.eu/en https://opendata.paho.org/en	1. Large Scale & Scope: Provides extensive global and regional coverage with millions of clinical isolates. 2. Standardization: Data are largely standardized according to international guidelines (e.g., CLSI/EUCAST), ensuring high comparability. 3. Clinical Relevance: Directly reflects the phenotypic resistance encountered in patient care, which is the ultimate public health concern.	1. Hospital-centric Bias: Data are primarily from clinical (hospital) surveillance, omitting AMR dynamics in community and environmental/agricultural settings. 2. Geographical Bias: Coverage is skewed towards high- and middle-income countries, limiting generalizability to low-income regions with the highest AMR burden.
Climate Data	ETCCDI Indices	https://www.noaa.gov/ https://etccdi.pacificclimate.org/indices_def.shtml	1. Comprehensiveness: The ETCCDI framework provides standardized, globally recognized indices for extreme climate events. 2. High Resolution & Consistency: NOAA’s gridded data offers consistent, long-term spatial and temporal coverage across the globe. 3. Nuance: Allows analysis to move beyond simple mean temperature to assess the impact of specific events like heatwaves, droughts, and cold spells.	1. Ecological Fallacy Risk: National-level aggregation of climate data may mask significant sub-national and local variations where climate-AMR interactions actually occur. 2. Indirect Measurement: Some indices (e.g., precipitation) may be poor proxies for complex events like floods or their impact on human behavior.
Covariate Data	GDP per capita, Population density, Basic water access, Basic sanitation	https://data.worldbank.org.cn	1. Authority & Standardization: Provides globally standardized, widely accepted indicators for socioeconomic and public health infrastructure. 2. Broad Coverage: Offers data for a vast majority of countries over a long time period.	1. Reporting Lags & Gaps: Data for some indicators, particularly in low-income countries, can have reporting lags or significant gaps, necessitating missing data handling (in our case, complete-case analysis). 2. Macro-level Proxy: National-level indicators (e.g., % access to sanitation) are macro-proxies and may not fully capture local-level realities and disparities.
	Antimicrobial Consumption	https://www.iqvia.com https://www.who.int/publications/i/item/who-report-on-surveillance-of-antibiotic-consumption https://www.ecdc.europa.eu/en/about-us/partnerships-and-networks/disease-and-laboratory-networks/esac-net	1. Best Available Data: Represents the most comprehensive and widely used sources for estimating national-level antimicrobial consumption.	1. Geographical Bias: Data quality and availability are significantly higher in high-income countries, potentially underestimating consumption in regions with large informal pharmaceutical sectors. 2. Usage vs. Appropriateness: Measures overall consumption (volume) but cannot distinguish between appropriate and inappropriate use.

Statistical analysis

To simplify the visualization of different AMR trends, the data were standardized via min–max normalization to eliminate dimensional differences between variables. The aggregated resistance, which represents the overall resistance level centered on the mean, was used for analysis.

LMMs were utilized to examine the relationships between climate indices and overall AMR rates. In the initial analysis, climate indices were treated as the primary independent variables, whereas overall AMR rates served as the dependent variable. To ensure robust results, the models were subsequently adjusted for potential covariates, including GDP per capita, population density, Antimicrobial Consumption (AMC), basic water access and basic sanitation, thereby accounting for their potential impact on the outcomes.

The entire workflow for data collection, processing, and statistical analysis is summarized in Fig. 1.

Fig. 1.

Workflow of the study design and data analysis. Abbreviations: AMR, Antimicrobial resistance; GDP, Gross Domestic Product; AMC, Antimicrobial Consumption; LMMs, linear mixed-effects models

Results

The final dataset included over 28 million bacteria from eight common hospital pathogens, Acinetobacter baumannii, Enterobacter aerogenes/cloacae, Enterococcus faecalis, Enterococcus faecium, Escherichia coli, Klebsiella pneumoniae, Pseudomonas aeruginosa, and Staphylococcus aureus, across 14 antibiotic categories.

Baseline characteristics, including mean values and quartiles for all indicators, are shown in Table 2. The average temperature has a mean of 14.17 °C with a large range from −1.22 °C to 31.6 °C. The mean precipitation is 1.77 mm, with a range from 0 to 9.83 mm. The table also includes economic and demographic data, such as GDP per capita and population density, antibiotic use, access to basic water, and access to basic sanitation. These data were incorporated as covariates into the model. Maps depicting aggregated resistance, average temperature, and average precipitation across countries were generated (Fig. 2). The analysis revealed a potential correlation between aggregated resistance and mean temperature, with regions exhibiting relatively high resistance rates, such as certain African countries, which generally present relatively high average temperatures. However, the relationship between the aggregated resistance and average precipitation was less clear, with no prominent or consistent patterns observed globally. Scatter plots of log-transformed aggregated resistance rates against temperature and precipitation (Figs. 3A, 4A) revealed a positive correlation between temperature and resistance rates, with a clearer trend in earlier years, although this weakened over time (Fig. 3C, D). In contrast, precipitation had a minimal effect on resistance rates. Spearman correlation analysis revealed consistently significant correlations between temperature and resistance across most bacterial species and antibiotic classes globally (Fig. 3B), whereas the effect of precipitation was inconsistent and pathogen-dependent.

Table 2.

Descriptive statistics of climate indices and covariates

Variable	Mean	SD	Percentile
			25th	50th	75th	Range
Aggregated AMR	0.28	0.16	0.17	0.24	0.35	0–1
Temperature
Average temperature (°C)	14.17	7.21	9.13	11.69	19.35	−1.22–31.6
Intensity Indices
TXx (°C)	33.22	4.93	30.64	33.99	36.37	10.95–49.11
TNx (°C)	21.54	4.46	18.6	20.74	24.93	5.92–36
TXn (°C)	2.43	12.33	−6.54	−0.43	10.65	−29.56–35.54
TNn (°C)	−5.55	12.13	−14.36	−6.88	2.36	−37.84–22.3
Absolute threshold indices
FD (d)	52.04	49.08	3.19	38.35	94.09	0–193.12
ID (d)	15.18	21.76	0	3.72	24.79	0–128.11
SU (d)	102.49	100.66	27.67	71.35	137.92	0–366
TR (d)	53.8	86.39	0.47	6.59	75.27	0–366
DTR (°C)	8.71	1.89	7.5	8.5	10.02	4.04–16.67
Relative threshold indices
TN10p (%)	13.71	6.45	9.16	12.63	17.28	0–41.12
TX10p (%)	12.92	6.37	8.47	11.83	16.35	0–40.76
TN90p (%)	21.46	10.96	14.12	18.1	25.37	1.3–80.23
TX90p (%)	20.66	11.04	13.19	18.18	25.41	0.56–95.61
Duration indices
GSL (d)	264.31	66.45	215.17	270	318.49	62.38–366
CSDI (d)	13.1	9.45	6.62	11.22	16.8	0–58.8
WSDI (d)	29.25	30.72	11.02	19	34.5	0–251
Precipitation
Average precipitation(mm)	1.77	1.17	1.1	1.59	2.26	0–9.83
Intensity indices
Rx1day (mm)	45.28	30.98	26.46	39.37	55.76	0–320.04
Rx5day (mm)	75.83	50.2	45.75	68.27	93.17	0–362.97
SDII (mm)	9.24	7.3	6.04	7.76	10.76	2.33–152.4
Absolute threshold indices
R10mm (d)	15.91	11.13	8.67	14.28	20.84	0–89
R20mm (d)	6.57	6.51	2.42	4.91	8.26	0–61
PRCPTOT (mm)	570.06	391.92	335.66	510.1	722.38	0–3415.28
Relative threshold indices
R95p (mm)	369.08	276.54	202.7	320.89	457.96	0–2137.41
R99p (mm)	232.45	202.79	109.71	178.75	296.42	0–1484.04
Duration indices
CDD (d)	49.48	40.52	23.46	35.46	60.87	8.61–298.88
CWD (d)	7.39	3.7	5	6.27	9.23	1–24.64
Other indicators
GDP per capita (US$)	26,255.35	23,273.61	8507.1	19,056	40,872.36	371.27–135,682.79
Population density (people/km2)	170.71	429.66	45.9	97.46	142.09	2.81–7965.88
AMC (DDD/1,000/day)	19.31	11.5	14.3	18.3	23	5–186
Basic water access (%)	97.49	6.12	97.53	99.79	100	47.21—100
Basic sanitation (%)	75.72	22.36	64.89	82.53	92.33	6.93—100

Large standard deviations for certain temperature indices (e.g., TNn, TXn) are due to the valid climatic heterogeneity inherent in the global-scale dataset and are accounted for in the mixed-effects models

AMR Antimicrobial resistance, GDP Gross Domestic Product, AMC Antimicrobial Consumption, DDD Daily Dose

Fig. 2.

The maps of temperature, precipitation and aggregated AMR. A A heatmap of mean normalized AMR for major global pathogens across all antibiotics. B A heatmap of global mean temperatures. C A heatmap of global mean precipitation. These maps provide a descriptive visualization of global spatial patterns; formal statistical associations are evaluated in the subsequent multivariable models

Fig. 3.

Analysis of the relationship between AMR and temperature. A Scatter plot showing the relationship between log-transformed aggregate AMR and temperature (°C). The black line represents the weighted regression trend, with the gray shading indicating the 95% CI. Points are sized by the number of tested isolates and colored by continent. B Circular bar chart illustrating the percentage change in AMR per degree Celsius increase across various pathogens and antibiotics, with distinct colors representing pathogens. C AMR plotted against temperature, stratified by time periods (2000–2007, 2008–2015, and 2016–2023). Regression lines in different colors represent temporal trends. D Density plots of resistance–temperature slopes for each time period, with vertical dashed lines indicating median slopes

Fig. 4.

Analysis of the relationship between AMR and precipitation. A Scatter plot showing the relationship between the log-transformed aggregate AMR and precipitation (mm). B AMR plotted against precipitation, stratified by time periods (2000–2007, 2008–2015, and 2016–2023)

LMMs revealed that both average temperature and high-temperature-related extreme climate indices (e.g., TXx, TNx, TXn, TNn, SU, TR, DTR, TN90p, TX90p and WSDI) were significantly positively correlated with AMR rates, and these associations remained significant after controlling for GDP per capita, population density, and AMC (Model 2), as well as further adjusting for basic water access and basic sanitation (Model 3) (Table 3). Cold-related indices (e.g., TN10p and CSDI) were negatively correlated with resistance rates. Other duration indices (e.g., GSL) had no significant effect. Among precipitation indices, only CDD exhibited a positive correlation with resistance after controlling for covariates.

Table 3.

Effects of climate change on aggregated AMR

Indices	Model 1		Model 2		Model 3
	Est	P	Est	P	Est	P
Temperature
Average temperature (°C)	0.05277	< 0.0001	0.0564	< 0.0001	0.054	0.0003
Intensity Indices
TXx (°C)	0.02718	0.0002	0.02138	0.0055	0.01774	0.0281
TNx (°C)	0.02956	0.0007	0.02805	0.0022	0.02446	0.0118
TXn (°C)	0.03264	0.0019	0.03697	0.0005	0.03831	0.0008
TNn (°C)	0.02806	0.0046	0.03448	0.0005	0.03455	0.001
Absolute threshold indices
FD (d)	−0.02385	0.0108	−0.02662	0.0042	−0.01804	0.0647
ID (d)	−0.01323	0.0559	−0.01292	0.0574	−0.00971	0.1532
SU (d)	0.03031	0.0043	0.04587	0.0001	0.04928	0.0001
TR (d)	0.02279	0.0303	0.03828	0.0025	0.04589	0.0011
DTR (°C)	0.03116	0.0003	0.03296	0.0004	0.02359	0.0156
Relative threshold indices
TN10p (%)	−0.01325	0.0072	−0.01106	0.0335	−0.01224	0.0204
TX10p (%)	−0.01092	0.0207	−0.00777	0.1236	−0.00922	0.0769
TN90p (%)	0.01334	0.0846	0.01252	0.1428	0.01348	0.1225
TX90p (%)	0.02189	0.0031	0.02205	0.0056	0.02107	0.0088
Duration indices
GSL (d)	0.00183	0.7792	0.0067	0.3457	0.01188	0.1128
CSDI (d)	−0.00999	0.0648	−0.01032	0.0614	−0.01161	0.0417
WSDI (d)	0.01598	0.0635	0.01512	0.0896	0.02037	0.0226
Precipitation
Average precipitation (mm)	−0.00385	0.5629	0.00148	0.8503	−0.00362	0.6594
Intensity indices
Rx1day (mm)	−0.0088	0.1337	−0.0049	0.4744	−0.00512	0.4759
Rx5day (mm)	−0.00797	0.2485	−0.00375	0.5734	−0.00434	0.5537
SDII (mm)	−0.00212	0.6506	0.00003	0.9947	0.003	0.6472
Absolute threshold indices
R10mm (d)	−0.00754	0.2354	−0.00537	0.4698	−0.00954	0.2183
R20mm (d)	−0.00069	0.9224	0.0034	0.6912	0.00034	0.97
PRCPTOT (mm)	−0.00783	0.2246	−0.00614	0.4247	−0.00878	0.2744
Relative threshold indices
R95p (mm)	−0.0018	0.7677	0.00182	0.7976	−0.00196	0.7925
R99p (mm)	0.00011	0.9849	0.00309	0.6379	−0.00001	0.9987
Duration indices
CDD (d)	0.00639	0.2051	0.01273	0.0187	0.01526	0.0124
CWD (d)	0.00758	0.5088	0.01361	0.0658	0.00905	0.1594

The models were linear mixed-effects models with random intercepts for country and year. Model 1 was an unadjusted model. Model 2 was adjusted for GDP per capita, population density, and antimicrobial consumption. Model 3 represents the fully adjusted model, which included all covariates from Model 2 plus basic water access and basic sanitation

For the critical pathogens listed in the World Health Organization’s “2024 Bacterial Priority Pathogens List”: carbapenem-resistant Acinetobacter baumannii (CRAB), carbapenem-resistant Enterobacteriaceae (CRE), and third-generation cephalosporin-resistant Enterobacteriaceae (3GCRE) [26], LMMs (fully adjusted in Model 3 for GDP per capita, population density, AMC, and water/sanitation access) were used to analyze the associations between indices and AMR rates. The heat-related extreme climate indices, such as TNx, TXn, TNn, SU and TX90p, presented consistent positive correlations with resistance rates, suggesting that warmer climates are associated with greater antimicrobial resistance. Moreover, consistent negative correlations were observed between the cold-related indices (FD, ID, and CSDI) and resistance rates. However, for certain pathogen subgroups, this trend did not reach statistical significance (Fig. 5A). In contrast to the patterns observed for the temperature-related indices, the subgroup analysis for the precipitation indices revealed a more complex and generally nonsignificant landscape (Fig. 5B). Among all three pathogen groups, the analysis of associations between 3GCRE and precipitation indices (R10mm, R20mm, PRCPTOT, R95pp, R99p) revealed relatively consistent results, indicating a negative correlation between increased precipitation and resistance rates, whereas CDD presented a positive correlation with increased resistance rates, with these findings being fairly uniform. However, for CRAB and CRE, only a few indices yielded statistically significant results, and the direction of association was opposite to that of 3GCRE, with a positive correlation between bacterial resistance rates and increased precipitation and a negative correlation with drought.

Fig. 5.

Results of correlation analysis between climate indices and key priority bacteria. A Temperature-related indices; B precipitation-related indices. Forest plot of associations between climate indices and AMR in specific pathogen groups. The estimates are derived from LMMs adjusted for five key covariates (GDP per capita, population density, AMC, and access to basic water and sanitation). Points represent the estimated coefficients (Est), and horizontal bars represent their 95% confidence intervals (95% CI). Associations are considered statistically significant (P < 0.05) if their 95% CI does not cross the vertical zero line and are highlighted in red. Abbreviations: CRAB, carbapenem-resistant Acinetobacter baumannii, CRE, carbapenem-resistant Enterobacteriaceae, 3GCRE, third-generation cephalosporin-resistant Enterobacteriaceae

To synthesize the key findings from our multi-faceted analysis, we developed a flowchart summarizing the main outcomes (Fig. 6). This visual summary outlines the consistent positive association between temperature indices and aggregated AMR, the more complex, pathogen-dependent effects of precipitation, and the contrasting relationships revealed in the subgroup analyses of WHO priority pathogens.

Fig. 6.

Flowchart summarizing the main outcomes of the association analysis between climate indices and AMR

Discussion

This study presents a global-scale analysis over a 23-year period to examine the association between climate change, specifically extreme climate events, and clinical AMR. To our knowledge, this is the first study to systematically apply the ETCCDI framework to a large, multi-decade dataset of clinical phenotypic resistance. After adjusting for key covariates, including antibiotic use, socioeconomic factors, and sanitary conditions, our findings indicate that rising average temperature and a higher frequency of extreme heat events are consistently and positively associated with increased AMR rates. Conversely, the role of precipitation was more nuanced; while most precipitation metrics were not significantly associated with AMR, we identified prolonged drought (measured by CDD) as having a significant, albeit pathogen-dependent, relationship. These findings provide evidence that climate change is an important environmental factor shaping AMR patterns and underscore the need for integrating climate considerations into public health strategies to combat AMR.

Our findings provide robust, global-scale support for mean ambient temperature as a significant driver of AMR, aligning with and extending the conclusions of previous regional studies. For instance, research conducted in Europe and China has already demonstrated positive correlations between mean temperature and resistance rates for specific pathogens like E. coli and K. pneumoniae [10–12]. By encompassing multiple climatic zones and a broader spectrum of pathogens, our study substantiates earlier findings, suggesting that the facilitative effect of mean temperature on AMR may be a globally generalizable phenomenon. This association is mechanistically plausible, as warmer temperatures can directly accelerate the growth and replication rates of many pathogenic bacteria and may also increase the frequency of horizontal gene transfer of resistance genes through various pathways such as conjugation, thereby creating more favorable conditions for the evolution and spread of resistance [16]. However, a noteworthy finding was the apparent attenuation of this temperature–AMR correlation after 2015 (Fig. 3C, D). This temporal shift does not necessarily negate the impact of temperature but rather suggests a confluence of complex, countervailing factors. First, the launch of the WHO Global Action Plan on AMR in 2015 may have intensified antimicrobial stewardship and infection control programs globally [27], potentially counteracting some of the environmental pressures from climate warming. Second, the profound confounding effects of the COVID-19 pandemic, which dramatically altered antibiotic prescribing patterns and disrupted routine AMR surveillance, cannot be overlooked in this period [28, 29]. Finally, this trend may also signify a shift in the dominant drivers of AMR, such as the global dissemination of high-risk, multidrug-resistant clones (e.g., CRE), whose spread may be more dependent on healthcare networks and patient mobility than on direct regulation by mean ambient temperature [30].

Beyond mean temperature, a primary contribution of this study is the systematic assessment of the impact of extreme climate events on AMR. Our analysis consistently demonstrates that, among the suite of extreme climate indices, heat-related indicators are robust predictors of AMR rates. For instance, indices such as TNx, TXn, TNn, SU and TX90p were significantly and positively associated with overall AMR rates, an association that remained robust in subgroup analyses of key WHO priority pathogens. In stark contrast, cold-related indicators exhibited consistent negative correlations, with metrics like FD, ID and CSDI being associated with lower AMR rates. This dichotomous pattern strongly suggests that warmer environments, and particularly conditions of extreme warmth, systematically facilitate the prevalence of AMR.

This finding lends strong support to an “acute selection pressure” hypothesis: compared with the gradual shifts in annual mean temperature, extreme events such as heatwaves may represent more intense, acute selection pressures that more effectively drive the evolution of resistance. Such acute thermal stress events can trigger bacterial stress responses, increasing mutation rates by activating error-prone repair enzymes and downregulating mismatch-repair enzymes [16, 31]. Furthermore, bacteria leverage the immense plasticity of their pan-genome to adapt to environmental stressors, with mobile genetic elements (MGEs) playing a pivotal role in the rapid acquisition and dissemination of genetic material, including AMR genes [32, 33]. It is likely during the acute thermal stress created by industrialization and urbanization—processes known to amplify extreme heat via the “Urban Heat Island” effect [34], that intense selection pressure is most likely to facilitate the fixation of resistance mutations and the transfer of MGEs. As recent genotypic studies have linked higher levels of industrialization to a greater prevalence of AMR genes [14], our work provides critical phenotypic evidence for this link, pointing to extreme events as a key triggering mechanism.

In contrast to the clear and consistent signal for temperature, the impact of precipitation on AMR was remarkably complex and context dependent. Subgroup analysis revealed a negative correlation between increased rainfall and resistance rates in 3GCRE. This seemingly counterintuitive finding was not isolated; a recent global study on PM_2.5 and AMR also reported a significant negative association between rainfall and AMR in its multivariable model [9]. A multi-layered hypothesis may explain this phenomenon. First, frequent or regular rainfall may be dominated by a “dilution effect”. ESBL-producing Enterobacterales are widespread in community and agricultural settings, constituting diffuse “non-point source” pollution [35], and studies have shown that frequent rainfall can mitigate the accumulation of surface contaminants [36]. Second, rainfall may reduce selection pressure by diluting environmental antibiotic residues. The carriage of large resistance plasmids often imposes a “fitness cost”, and in an environment with diminished selection pressure, sensitive strains may outcompete their resistant counterparts [37].

However, the role of precipitation was not always protective. A substantial body of evidence has demonstrated that heavy or extreme rainfall events are critical drivers of the dissemination of fecal-source contaminants and AMR. For instance, studies have found that extreme rainfall can flush fecal indicator bacteria (e.g., E.coli) accumulated during dry periods into water bodies via an “accumulation and flush” mechanism [36]. More directly, another study showed that a rainfall event significantly increased the abundance of key ESBL resistance genes (e.g., bla_CTX-M) in river water [38]. This dichotomy between “dilution” and “dissemination” highlights the complexity of precipitation’s role, where the net effect depends on rainfall type, antecedent climate conditions, and pollution source type (non-point vs. point) [39]. This pattern may also explain the opposite trend observed for carbapenem-resistant strains, whose environmental origins are more concentrated in healthcare settings [40].

Finally, it is noteworthy that, among all precipitation-related indices, CDD revealed the clearest signal, albeit one that was also highly pathogen specific. The positive correlation of CDD with resistance in 3GCRE likely reflects the concentration of pollutants in shrinking water bodies during droughts [41]. Conversely, its negative correlation with CRAB strongly supports a niche disruption effect on pathogens dependent on moist environments [42]. This reinforces the necessity for a context-specific, pathogen-aware analysis when assessing the climatic drivers of AMR. Furthermore, these direct environmental mechanisms are compounded by indirect, climate-driven human behaviors: extreme events such as floods can trigger increased prophylactic antibiotic use in affected communities, while prolonged droughts may compromise sanitation and hygiene standards, both of which can independently amplify selection pressure for resistance.

The robust temperature–AMR association is likely underpinned by multi-layered biological mechanisms. At the microscale, elevated temperatures act directly on bacterial genetics and physiology. First, heat is a potent selective pressure that can favor mutations conferring thermal adaptation, some of which may pleiotropically confer antibiotic resistance through “cross-protection” (e.g., rpoB mutations) [16]. It can also induce stress responses that elevate mutation rates [31]. Second, bacteria leverage the immense plasticity of their pan-genome to adapt to environmental stressors. Warmer temperatures may accelerate the horizontal gene transfer (HGT) of MGEs, thereby facilitating the rapid dissemination of AMR genes within microbial communities [32, 33]. At the macroecological scale, warmer conditions create a favorable environment for the proliferation and persistence of resistant bacteria. Higher temperatures can accelerate bacterial growth and promote the formation of biofilms, structures that inherently confer increased tolerance to antibiotics [43, 44]. Furthermore, sustained high temperatures can help mitigate the “fitness cost” often associated with AMR, allowing resistant strains to compete more effectively with their susceptible counterparts and persist in the environment [45]. It is the cumulative effect of these multiple mechanisms, from the molecular to the ecosystem level, that likely manifests as the strong, positive correlation between temperature and clinical AMR rates observed at the global scale.

Our findings, which establish a clear link between climate change and AMR, have profound policy implications, pointing toward a dual strategy of climate mitigation and public health adaptation. First, this study adds a novel public health rationale for accelerating global climate mitigation efforts. By demonstrating that climate extremes drive AMR, we frame policies such as reducing greenhouse gas emissions as a fundamental, upstream public health intervention to curb future AMR growth, highlighting a critical AMR-related co-benefit of climate action [46]. Second, on the adaptation front, our findings provide a scientific basis for shifting from a reactive to a proactive posture by implementing a “precision public health” strategy. First, given the strong positive correlation between extreme heat indices and AMR rates, we strongly advocate for the integration of climate surveillance into existing AMR monitoring systems. Specifically, climate-informed early warning systems, based on heatwave forecasts, can be developed to alert hospitals and public health agencies of impending periods of heightened AMR risk, guiding the intensification of infection control measures and antimicrobial stewardship efforts [47]. Furthermore, our findings provide a theoretical basis for a potential, highly differentiated “precision environmental hygiene” strategy. Our discovery of opposing effects of precipitation and drought on different pathogens challenges a “one-size-fits-all” approach to environmental intervention. For example, for pathogens dependent on moist environments such as CRAB, local control strategies focused on reducing water stagnation and maintaining dry surfaces could be effective, implying that climate-resilient designs for hospitals should prioritize rapid drainage and humidity control. In stark contrast, for 3GCRE, periodic flushing and dilution with clean water could, counterintuitively, be a potential intervention to suppress their environmental colonization. We must emphasize that these seemingly contradictory strategies are currently theoretical derivations requiring further investigation. Nevertheless, our findings point to a critical future direction: AMR control strategies may need to be tailored and even dynamically adjusted based on the predominant circulating pathogen and local climatic characteristics to achieve true precision prevention.

Strengths and limitations

The primary strength of this study lies in its global scale, a 23-year temporal scope, and its novel application of the ETCCDI framework to quantify the multidimensional impacts of extreme climate events on clinical phenotypic AMR. By applying LMMs to a multi-decade dataset, this study provides robust evidence for these associations across broad geographical and temporal scales.

However, this study has several inherent limitations. First, the analysis is constrained by data source biases. Reliance on hospital-based surveillance networks means our findings primarily reflect clinical AMR, omitting crucial data from community and environmental settings within the “One Health” continuum. Second, the dataset is skewed toward high-income nations, which limits the generalizability of our findings to low- and middle-income countries where the AMR burden is highest. Finally, our macroscale ecological approach could not fully capture the complex, local-level mechanisms linking precipitation with AMR, such as interactions with sanitation infrastructure and human behavior, as highlighted by the complexity of our precipitation findings.

Conclusion

This study provides robust global evidence that climate change, particularly through extreme temperature events, is a significant but highly pathogen-specific driver of AMR. While extreme heat was consistently associated with higher AMR for several key pathogens, the impact of precipitation was more complex, with prolonged drought showing significant but contradictory effects across different bacteria. These findings demand a shift towards more tailored, climate-informed public health strategies.

Authors’ contributions

YN was responsible for data collection, part of the data organization, and drafting the initial manuscript. JZ was responsible for data analysis. YY participated in the review and revision of the manuscript. BF was responsible for overseeing the progress of the entire project and revising the manuscript.

Funding

This research received no specific grant from any funding agency, commercial, or not-for-profit sectors.

Data availability

All data used in this study were obtained from publicly available sources, and the specific sources and access paths are cited in the references of this article. The datasets analyzed during the current study are not further distributed or made publicly available due to restrictions on reuse of the original data, but they can be accessed directly via the sources cited in the references.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Clinical trials

Not applicable