Rainfall and Temperature Modify Effects of On-Site Sanitation Intervention on E. coli Contamination in Bangladeshi Households

Introduction

Higher temperatures and rainfall associated with climate change have been linked to increased risk of infectious disease, potentially through increasing the dissemination and survival of pathogens in environmental compartments, such as water and soil^1–4. Water, sanitation and hygiene (WASH) interventions, such as improved access to clean water, sanitation facilities that isolate fecal waste from the environment, and adequate hygiene practices, are fundamental to preventing the spread of disease. However, the efficacy of these interventions can be influenced by environmental conditions, including weather fluctuations induced by climate change⁴. For example, heavy rainfall and/or flooding may impact service delivery or overwhelm sanitation systems, leading to increased fecal contamination of the environment, thereby undermining the effectiveness of sanitation interventions⁵. Alternatively, climate-resilient WASH interventions may deliver larger benefits during increased rainfall or temperature by protecting households from the effects of enhanced microbial dissemination and survival in the environment^3,6. Resilient interventions that remain effective under weather-related stressors can ensure sustained access to clean water and effective sanitation in the face of climate change. In low- and middle-income countries vulnerable to extreme weather events, it is important to understand how the efficacy of WASH interventions may vary under different weather conditions⁷.

Here, we use longitudinal data from a randomized controlled trial in rural Bangladesh (WASH Benefits) that implemented individual and combined WASH interventions to assess the effect of rainfall and temperature on intervention effectiveness against fecal contamination in the environment. The interventions reduced diarrhea and enteric infections among young children^8,9. However, a longitudinal analysis focused on the sanitation intervention demonstrated that the health benefits were confined to the rainy seasons; the intervention did not affect child diarrhea during dry seasons¹⁰. Similarly, a recent analysis of the trial data using daily rainfall and temperature values found that the WASH interventions reduced diarrhea prevalence by 51% following periods with heavy rainfall, compared to 13% following periods without heavy rainfall, and by 40% following periods with above-median temperature, compared to 9% following periods with below-median temperature¹¹.

Investigating the underlying environmental routes of fecal contamination under different weather conditions is important for interpreting these health effects and understanding how intervention effectiveness can be improved under adverse weather conditions. Previous evaluations of the effects of the WASH Benefits interventions on fecal contamination found that the water treatment, handwashing and combined WASH interventions reduced E. coli in stored drinking water and food, with evidence of greater reductions during the dry season^12,13. The sanitation intervention did not reduce E. coli in the domestic environment when measured approximately 4 months after intervention initiation¹². In a longitudinal study conducted between 1 and 3.5 years after intervention initiation, the sanitation intervention slightly reduced E. coli counts in stored drinking water and child hand rinses, with no effect modification by season¹⁴. However, two of these studies defined seasons based on calendar month, and one study defined the monsoon season based on spatiotemporal rainfall data. It is important to consider granular rainfall and temperature measurements rather than general/binary seasonality as this approach can capture discrete weather events as well as intra-annual variation in the onset and duration of dry and wet seasons, which are expected to shift with climate change.

This current study re-analyses data from environmental samples collected over 3.5 years in the sanitation and control arms of the WASH Benefit Bangladesh trial using daily weather data to investigate whether rainfall and temperature modify the effectiveness of the sanitation intervention on fecal contamination in the domestic environment. Understanding how weather patterns modify the efficacy of WASH interventions is essential for strengthening resilience, long-term sustainability, and adaptive capacity in vulnerable communities as well as inform intervention designs for WASH under climate change¹⁵.

Methods

Results

Between July 2013 and December 2016, we collected 5740 stored drinking water, 1629 stored food, 5397 mother hand rinse, 6494 child hand rinse, 1098 source water (tubewell), 1928 soil, 557 pond, and 395 fly samples from 1243 unique households in the control and sanitation arms of the WASH Benefits Bangladesh trial (Table S1). The 90^th percentile of daily rainfall (classified as extreme rainfall) corresponded to 28.20 mm and the 80^th percentile (classified as heavy rainfall) corresponded to 16.44 mm. The median of rolling average rainfall values for two- and seven-day periods were 0.27 mm and 1.10 mm, respectively. The 90^th percentile of daily mean temperature (classified as extreme temperature) corresponded to 30.21°C and the 80^th percentile (classified as elevated temperature) corresponded to 29.29°C. The median of rolling daily average temperature values for two- and seven-day periods were 27.46°C and 27.52°C, respectively. The highest rainfall occurred between the months of April–October and the highest temperatures occurred between the months of March–September (Fig S1, Fig S2). Of 23,238 total samples, 12.0% (2779) were collected within two days and 23% (5342) within seven days after extreme rainfall, and 14.1% (3282) were collected within two days and 20.5% (4765) within seven days after extreme temperature (Table S2).

Across all observations (without stratification by weather), the sanitation intervention was associated with an approximately a 0.1-log reduction in E. coli counts in stored drinking water (Δlog10 = −0.08 (−0.14, −0.01), p-value= 0.03) and on child hands (Δlog10 = −0.06 (−0.13, 0.00), p-value=0.05) (Table 1). The intervention did not significantly reduce E. coli contamination in food, mother hands, soil, tubewell water, ponds, and flies.

Overall, rainfall modified the effect of the sanitation intervention on E. coli counts on mother hands, on child hands, in ponds, and in/on flies; the intervention consistently led to larger reductions in E. coli in these sample types following periods of higher rainfall. When extreme rainfall occurred within seven days prior to sampling, the intervention was associated with a 1-log reduction in E. coli in/on flies (Δlog10 = −0.91 (−1.65, −0.17)); there was no intervention effect in the absence of extreme rainfall (interaction p-value=0.03) (Fig 1, Table S3). Trends for flies were similar for extreme rain occurring within two days prior to sampling (with borderline significant interaction p-value), heavy rain occurring within two or seven days prior to sampling, and above-median rolling rainfall within two or seven days prior to sampling (interaction p-values<0.20) (Figs 1 and 2, Fig S3, Tables S3-S5). Extreme or heavy rainfall did not modify intervention effects on any other sample type during either antecedent period (interaction p-values >0.20) (Figs 1 and Tables S3 and S5). When the rolling average rainfall within two or seven days prior to sampling was above median, the intervention was associated with an approximately 0.1-log reduction in E. coli on child hands while there was no intervention effect when rolling average rainfall was below median, with a similar trend for E. coli on mother hands (interaction p-values<0.20) (Fig 2, Table S4). When the rolling average rainfall within seven days prior to sampling was above median, the intervention was also associated with a 0.25-log reduction in E. coli in ponds (Δlog10 = −0.25 (−0.48, −0.03)), with no effect when rolling average rainfall was below median (interaction p-value=0.10) (Fig 2, Table S4). Trends for ponds were qualitatively similar for all rainfall classifications and antecedent periods, with intervention effects appearing to be larger in magnitude following higher rainfall (interaction p-values ranging from 0.10 to 0.45). Rainfall of any classification over either antecedent period did not modify intervention effects on E. coli in stored drinking water, food, soil and tubewell water.

• Download figure
• Open in new tab

Figure 1. Forest plot of intervention effects from improved on-site sanitation on log10-transformed E. coli counts by sample type in strata of extreme rainfall and temperature ^a during 2- and 7-day antecedent periods ^b

^a Extreme rainfall defined as ≥ 28.20 mm, extreme temperature defined as ≥ 30.21°C. We could not estimate intervention effects on source water (tubewells), ponds and flies following periods of extreme temperature due to data sparsity.

^b Confidence intervals truncated for 2- and 7-day periods of extreme rainfall for flies, Δlog10 transformed E. coli counts and 95% CI: −0.90 (−2.30, 0.50) and −0.91 (−1.65, −0.17), respectively.

• Download figure
• Open in new tab

Figure 2. Forest plot of intervention effects from improved on-site sanitation on log10-transformed E. coli counts by sample type in strata of above and below median rolling average rainfall and temperature ^a during 2- and 7-day antecedent periods

^a Median 2-day rolling average rain defined as ≥ 0.27 mm, median 2-day rolling average temperature defined as ≥ 27.46°C, median 7-day rolling average rain defined as ≥ 1.10 mm and median 7-day rolling average temperature defined as ≥27.52°C.

Overall, temperature modified the effect of the sanitation intervention on E. coli counts in food, on mother hands, on child hands, in soil, and in ponds, with the intervention consistently leading to larger reductions in E. coli in these sample types following periods of higher temperature. When extreme temperature occurred within two days prior to sampling, the intervention was associated with a 0.22-log reduction in E. coli on child hands (Δlog10 = −0.22 (−0.36, −0.09), interaction p-value=0.01) and 0.40-log reduction in E. coli in soil (Δlog10 = −0.40 (−0.70, −0.15), interaction p-value=0.01) (Fig 1, Table S3). For E. coli in food and on mother hands, there were no significant intervention effects in either temperature stratum, but the point estimates indicated larger reductions when extreme temperature occurred within two days prior to sampling (interaction p-value=0.05 for food, 0.18 for mother hands) (Fig 1, Table S3). These trends were similar for extreme temperature during the seven-day antecedent period (except for mother hands, Fig 1, Table S3) and for elevated temperature during either antecedent period (Fig S3, Table S5). Additionally, when elevated temperature occurred within two days prior to sampling, the intervention was associated with a 0.40-log reduction in E. coli counts in ponds (Δlog10 = −0.38 (−0.59, −0.17), interaction p-value=0.01), with similar findings for the seven-day period (Fig S3, Table S5). When the rolling average temperature was above median within two or seven days prior to sampling, the sanitation intervention was associated with larger reductions in E. coli counts on mother hands (interaction p-value=0.17), child hands (interaction p-value=0.05) and soil (interaction p-value=0.05) (Fig 2, Table S4). No measure of temperature modified intervention effects on E. coli in stored drinking water, tubewell water, or flies.

Discussion

The WASH Benefits sanitation intervention, including double-pit latrines, child potties and child and animal feces management tools, reduced fecal contamination in rural Bangladeshi households more effectively following periods of higher rainfall and temperature. Across different classifications and antecedent periods, rain modified intervention effects on E. coli counts for four out of the eight sample types (mother and child hands, ponds, and flies) while temperature modified intervention effects on E. coli counts for five out of the eight sample types (food, mother and child hands, soil, and ponds). Reductions in E. coli counts ranged from 0.10 log on child hands to 0.91 log in flies following increased rainfall, and from 0.11 log on child hands to 0.40 log in soil following increased temperature. In contrast, analyses without stratifying by daily weather found small (<0.10 log) reductions in E. coli counts only for stored drinking water and child hands associated with the intervention, indicating that averaging across the study period concealed intervention effects taking place following increased rainfall and temperature. These findings can shed light on the mechanism behind the larger reductions in child diarrhea observed in the parent trial during the rainy seasons and following episodes of higher rainfall and temperature²³. A re-analysis of diarrhea outcomes from the parent trial also found that the largest health benefits occurred among the lowest socioeconomic strata during the monsoon season, indicating joint effects from socioeconomic position and vulnerability to climate-sensitive pathogen transmission²⁴. Our findings are consistent with a recent meta-analysis that found that when households lack access to improved latrines and piped drinking water, increasing temperature and rainfall are associated with increased incidence of diarrhea in children under the age of five⁴.

Higher rainfall can lead to increased initial dissemination (e.g. overflowing latrine pits) and subsequent dilution of fecal contamination in the environment^25,26; these patterns, in return, can influence how effectively a sanitation intervention contains fecal contamination. For example, in our previous analysis, we found that increased rainfall was associated with increased E. coli contamination in ponds but decreased contamination in tubewell water and courtyard soil²¹. These patterns may indicate that, during heavy rain, latrine pits overflow into the surroundings and/or leak into the subsurface but fecal waste is then flushed out of the soil matrix into ponds with run-off ²⁷. The larger reduction from the sanitation intervention in pond E. coli counts following increased rainfall in the current analysis suggests that the sanitation upgrades may have reduced rainfall-associated release of fecal waste from latrines into ponds. Possible mechanisms include that the concrete lining of the improved pits reduced leakage or that the double-pit design prevented overly full pits that are more likely to overflow with rain. Larger reductions from the sanitation intervention in E. coli counts in/on flies and on mother and child hands following increased rainfall also support the possibility that the intervention reduced rainfall-associated dissemination of fecal waste from latrines into the environment and subsequent vector/hand contact with fecal contamination.

Our previous findings also indicated increased contamination of stored drinking water and food following extreme rainfall²¹. However, the sanitation intervention was not any more or less effective against contamination of these pathways following increased rainfall. These findings suggest that the mechanisms for rainfall-associated contamination of stored drinking water and food (e.g. handling and storage conditions) were not influenced by sanitation improvements, despite the observed reductions in mother and child hand contamination after increased rain, which would be expected to reduce contamination of both stored water and food²⁸. Drinking water treatment and safe storage of water and food following increased rainfall may protect against rainfall-associated contamination. A recent study in Kenya found that household water treatment mitigated the effects of rainfall on drinking water quality²⁹.

Warmer temperatures can stimulate microbiological activity, potentially creating ideal incubation environments for vectors and fecal pathogens, which could be mitigated by sanitation interventions that isolate feces from the environment^3,30. Conversely, warmer temperatures can also be associated with more sun exposure, which can inactivate microorganisms^3,6. Our previous analysis found that higher temperatures were associated with increased contamination of stored drinking water and food in this setting²¹. In our current analysis, the sanitation intervention appeared to more effectively reduce food contamination following higher temperatures, but the reductions could not be distinguished from the null in any temperature stratum. There was a similar pattern for mother hands. The sanitation intervention was associated with a larger reduction in child hand contamination following increased temperatures, as well as larger reductions in contamination of soil and ponds. These findings may indicate that improved containment of fecal waste can differentially reduce proliferation of fecal organisms in the environment during warmer conditions, which may translate to less contaminated child hands.

It is also possible that rainfall and temperature affect adherence to sanitation practices, including the use of latrines, location of defecation, and proper disposal of child and animal feces³¹. Understanding the user convenience of sanitation hardware and any behavior changes during different weather conditions will be helpful to improve the implementation and adoption of sanitation practices and intervention strategies³². During increased rainfall or higher temperatures, individuals may change their latrine usage behavior, due to damage or resource conservation, or modify their defecation practices to avoid discomfort or inconvenience, potentially affecting levels of fecal contamination in the environment^26,33. For example, during heavy rain, young children may be more likely to defecate in a potty indoors than openly defecate outdoors and similarly, older children and adults may be more likely to use latrines than open defecate. Domestic animals might be confined to different spaces during heavy rain and/or higher temperatures, thereby impacting where their feces are disposed of and subsequent levels of environmental contamination.

Our findings differ from previous assessments of the WASH Benefits interventions, which found that the water treatment, handwashing and combined WASH interventions more effectively reduced E. coli in stored water and food during the dry season compared to the wet season^12,13 and no effect modification by season in sanitation intervention effects on environmental contamination¹⁴. The calendar-based (June-October) and/or binary definitions of wet season in these studies do not capture rain-free episodes during the wet season or occurrences of rain during the dry season. They also do not capture temperature fluctuations that occur within the same season and cannot differentiate between joint variations in rain and temperature (e.g., generally warmer temperatures coinciding with the wet season in Bangladesh). These factors can potentially explain the different findings from our current analysis that used daily weather data. As weather patterns shift under climate change and longstanding definitions of season no longer hold empirically, using finer grained weather data can provide improved insights over relying on season definitions. However, in regions like Bangladesh, where a distinct monsoon season can be reliably identified using rainfall data, using seasonality as an effect modifier could still be valuable²⁴. This approach can distill a complex set of environmental/temporal factors into a single variable definition and help capture the combined influence of rainfall and temperature which might be missed if each factor is modelled separately. Operationally defining a monsoon season can also simplify program implementation and messaging (e.g. deliver and emphasize interventions only or more heavily during the monsoon season).

Our findings also offer nuance to previously published findings that only indicated small reductions in stored drinking water and child hand contamination from the sanitation intervention but no other effects on the broader domestic environment. Our current analysis indicates additional significant sanitation intervention effects – including reductions in E. coli contamination of ponds, courtyard soil, and flies – that are confined to periods of increased rainfall and/or temperature. Future research should use fine-grained and location-specific rainfall and temperature data to quantify the efficacy of sanitation and WASH interventions during different weather conditions to detect effects that may be concealed when averaged over the study period²⁹. Longitudinal assessments spanning multiple seasons and a range of weather conditions can provide comprehensive insights into intervention effectiveness and inform strategies for maximizing impact under fluctuating weather conditions. Future WASH intervention studies should also consider targeted environmental sampling following periods of heavy rain and elevated temperatures. However, we also note that even when the sanitation intervention resulted in significant reductions in contamination under some weather conditions, E. coli counts in the environment remained relatively high among intervention recipients¹².

This study leveraged an extensive longitudinal dataset to understand how daily weather variability affects the impact of multi-faceted sanitation improvements on several fecal-oral transmission pathways. However, the study also has limitations. We could not assess effect modification by extreme temperature for source water (tubewells), ponds, and flies because only a small number of samples coincided with periods of extreme temperature for these sample types. However, our sensitivity analysis using elevated instead of extreme temperatures supported our primary conclusions. Additionally, generalizing our findings to other settings may be limited by variations in environmental factors, socioeconomics, infrastructure, and cultural practices. While this study focused on rural households in Bangladesh, the specific underlying mechanisms driving intervention effects during different rainfall/temperature patterns may not apply to other settings.

Our findings indicate increased protection from sanitation improvements following periods of increased rainfall and temperature, suggesting that sanitation and potentially other WASH approaches can support efforts to mitigate adverse health impacts from climate change. This study contributes to the growing body of literature on climate change adaptation and incorporating climate concerns in the context of public health, providing insights for policymakers and practitioners to design context-specific sanitation interventions that remain effective in the face of evolving climatic conditions. Our findings hold significant implications for policy and public health practice, particularly in low- and middle-income countries facing climate-related health risks. Adaptive management strategies capable of addressing changing environmental conditions and robust interventions that are resilient to weather fluctuations are essential to mitigate impacts from climate variability. By integrating climate resilience into sanitation policy and practice, governments and organizations can better safeguard public health and foster sustainable development in vulnerable communities.