Associations between long-term drought and diarrhea among children under five in low- and middle-income countries

The study addresses how climate change–induced drought affects diarrheal disease among children under five in LMICs. Diarrhea is a leading cause of child mortality and morbidity, with the greatest burden in regions lacking adequate water, sanitation, and hygiene (WASH). Prior evidence links ambient temperature and rainfall variability to diarrheal outcomes, but findings for rainfall are mixed, reflecting complex pathways whereby both low and high rainfall can elevate risk. Drought is more complex than low precipitation alone, involving water demand (evapotranspiration), yet epidemiologic studies using advanced drought metrics are scarce. The authors aim to quantify associations between drought measured by the standardized precipitation evapotranspiration index (SPEI) and childhood diarrhea risk across LMICs, and to assess whether WASH practices mediate or modify this relationship.

Previous research has shown positive associations between temperature and bacterial diarrhea and negative associations with some viral diarrheas (e.g., rotavirus). Studies examining rainfall and diarrhea report conflicting results: increased risk with both low and high rainfall, with some studies indicating risk at moderate, highest, or lowest rainfall levels. Drought definitions vary and often neglect water demand components. Limited epidemiological work has used complex drought indices; one Afghanistan study reported positive associations between aridity and diarrhea. WASH interventions generally reduce diarrheal risk, but drought may undermine WASH access and effectiveness. This study builds on limited literature by employing SPEI, integrating supply and demand aspects of hydrology, across multiple LMICs.

Design and data sources: The authors used 141 Demographic and Health Surveys (DHS) from 51 LMICs (1990–2019), encompassing 1,379,566 children under 5, with geocoded cluster locations. Outcome: Maternal report of whether the child had diarrhea in the two weeks preceding the interview. Exposure: Drought measured using the Standardized Precipitation Evapotranspiration Index (SPEI) at 0.1° (~10 km) resolution using ERA5-Land reanalysis (1988–2019) inputs (precipitation, temperatures, evaporation, wind speed, solar radiation, pressure), with PET estimated via FAO-56 Penman–Monteith. SPEI was computed at multiple aggregation timescales (6, 12, 18, 24 months). Drought severity categories were defined as no drought, mild drought, and severe drought (e.g., severe approximately SPEI ≤ −1.3; extreme/exceptional defined in secondary analyses as SPEI ≤ −1.6). Statistical analysis: Log-binomial generalized linear mixed models estimated risk ratios (RRs) for diarrhea associated with drought categories at each timescale. Models included random effects for survey cluster/site and fixed effects for individual and household covariates: child age, sex, residence (urban/rural), maternal education, household wealth quintile. Meteorological covariates (2-month averaged mean temperature and total precipitation) were modeled with natural cubic splines (3 df) to adjust for concurrent weather, along with seasonality and long-term trends. Crude models adjusted for meteorology, seasonality, and locality trend; main models additionally included baseline characteristics. Effect modification was tested by climate zone (tropical, dry, temperate), round-trip time to collect water, water availability at handwashing site, and availability of soap/detergent for handwashing, among others. Mediation by WASH was assessed by comparing total effects (main model) and direct effects (models additionally adjusting for WASH), with indirect effects derived and 95% CIs obtained via 1,000 bootstrap samples. Multiple imputation by chained equations was used in sensitivity analyses for missing data. Secondary analyses disaggregated drought into finer categories (e.g., moderate, severe, extreme/exceptional) and added child nutrition indicators (stunting, wasting, underweight) in separate models. Analyses were conducted in R 4.0.2.

- Sample: 141 surveys in 51 LMICs; total N=1,379,566 children under 5. Main analytic sample without missing covariates: N=713,918 (80 surveys, 43 countries). Mediation sample: N=365,975 (52 surveys, 35 countries). - Overall diarrhea incidence: 14.4% across LMICs, with substantial regional variability. - Primary associations: At the 6-month timescale, mild drought was associated with a 5% increased diarrhea risk (RR≈1.05, 95% CI 1.03–1.07) and severe drought with an 8% increase (RR≈1.08, 95% CI 1.05–1.11). - Climate zone modification: For 24-month droughts, associations were strong in dry zones (e.g., mild RR≈1.22, 95% CI 1.16–1.28; severe RR≈1.18, 95% CI 1.09–1.27) but weak or null in tropical/temperate zones. For 6-month droughts, elevated risks were observed in tropical and temperate zones but not in dry zones. - WASH mediation: For 6-month drought, an estimated 11.6% of the total effect of mild drought on diarrhea risk was mediated by WASH variables (combined). Alternative mediation analyses of individual WASH components suggested mediation ranges of ~0.27–20.7% (mild) and ~1.0–30.6% (severe). - WASH effect modification: Elevated RRs were observed among children in households requiring ≥30 minutes round-trip to collect water, without water available at the handwashing site, or lacking soap/detergent; associations were attenuated or null where water was on premises. - Drought severity disaggregation: The strongest association was for extreme/exceptional drought at 6 months (RR 1.14; 95% CI 1.09–1.18). Less extreme drought categories showed significant associations at multiple timescales, with notably strong associations for 18- and 24-month moderate drought. - Sensitivity analyses: Results using multiple imputation and alternative specifications were broadly consistent, though crude models on the full dataset showed somewhat weaker estimates.

The study demonstrates that long-term drought, measured by SPEI, is associated with higher diarrhea risk among children under five in LMICs, with timescale- and climate zone–specific patterns. Mechanistically, drought may reduce surface and groundwater availability, concentrate pathogens, increase groundwater pumping and depletion, contribute to salinity and turbidity issues, and increase interruptions in piped water supply and treatment efficacy, all of which can elevate enteric pathogen exposure via water and food pathways. WASH practices only partially mediate the drought–diarrhea relationship, indicating that while improved WASH can mitigate some risk, it cannot fully offset climate-driven hydrological stressors. Effect modification suggests that households with constrained water access or lacking handwashing resources are particularly vulnerable during drought. Climate zone differences imply that dry regions are more affected by prolonged (12–24 month) droughts, potentially reflecting limited capacity to adapt to persistent aridity, whereas tropical/temperate regions show stronger associations with shorter-term (6-month) droughts. These findings underscore the urgency of climate adaptation strategies that enhance water security and WASH infrastructure, particularly as climate change is expected to increase drought frequency, duration, and intensity.

This large, multi-country analysis quantifies associations between long-term drought and increased diarrhea risk in children under five across LMICs, using an advanced drought metric (SPEI). Risk increases were observed across drought severities and timescales, with pronounced vulnerabilities linked to limited water access and inadequate hand hygiene resources, and climate zone–specific patterns. Findings suggest that strengthening WASH access and reliability, improving water storage and treatment infrastructure, and developing climate-resilient water supplies could reduce drought-associated diarrheal burdens, a pressing need given projected increases in drought under climate change. Future research should incorporate pathogen-specific diagnostics, better characterize water system interruptions and slum contexts, evaluate food safety and security pathways, and examine the roles of migration and socioeconomic adaptation in modulating drought–health relationships.

- Exposure misclassification due to DHS cluster coordinate displacement (2–10 km) relative to 10-km SPEI grids, potentially mismatching exposure near grid borders. - Lack of a universal drought definition and variability in drought indices and thresholds; SPEI-based severity classifications are somewhat arbitrary and may limit generalizability. - Reduced analytic samples due to missing baseline and WASH data (main model N=713,918 vs. original N=1,379,566; mediation N=365,975), introducing potential selection bias; some discrepancies in effect sizes across samples. - Outcome based on maternal recall of diarrhea in the prior two weeks, subject to recall error and misclassification; no laboratory confirmation or pathogen-specific data. - Multiple comparisons in effect modification analyses raise potential for chance findings; these analyses were exploratory. - Unmeasured or unaccounted contextual factors (e.g., detailed water supply interruption patterns, food safety/security dynamics, and migration) may confound or modify associations and could not be disentangled due to data limitations.