Skip to main content
NCBI home page
PLOS Neglected Tropical Diseases logoLink to PLOS Neglected Tropical Diseases
. 2024 May 13;18(5):e0012157. doi: 10.1371/journal.pntd.0012157

Influence of hydrometeorological risk factors on child diarrhea and enteropathogens in rural Bangladesh

Jessica A Grembi 1,*, Anna T Nguyen 2, Marie Riviere 2, Gabriella Barratt Heitmann 2, Arusha Patil 2, Tejas S Athni 3, Stephanie Djajadi 4, Ayse Ercumen 5, Audrie Lin 6, Yoshika Crider 7, Andrew Mertens 3, Md Abdul Karim 8, Md Ohedul Islam 8, Rana Miah 8, Syeda L Famida 8, Md Saheen Hossen 8, Palash Mutsuddi 8, Shahjahan Ali 8, Md Ziaur Rahman 8, Zahir Hussain 8, Abul K Shoab 8, Rashidul Haque 8, Mahbubur Rahman 8, Leanne Unicomb 8, Stephen P Luby 1, Benjamin F Arnold 9, Adam Bennett 10,11, Jade Benjamin-Chung 2,12
Editor: Marilia Sá Carvalho13
PMCID: PMC11115220  PMID: 38739632

Abstract

Background

A number of studies have detected relationships between weather and diarrhea. Few have investigated associations with specific enteric pathogens. Understanding pathogen-specific relationships with weather is crucial to inform public health in low-resource settings that are especially vulnerable to climate change.

Objectives

Our objectives were to identify weather and environmental risk factors associated with diarrhea and enteropathogen prevalence in young children in rural Bangladesh, a population with high diarrheal disease burden and vulnerability to weather shifts under climate change.

Methods

We matched temperature, precipitation, surface water, and humidity data to observational longitudinal data from a cluster-randomized trial that measured diarrhea and enteropathogen prevalence in children 6 months-5.5 years from 2012–2016. We fit generalized additive mixed models with cubic regression splines and restricted maximum likelihood estimation for smoothing parameters.

Results

Comparing weeks with 30°C versus 15°C average temperature, prevalence was 3.5% higher for diarrhea, 7.3% higher for Shiga toxin-producing Escherichia coli (STEC), 17.3% higher for enterotoxigenic E. coli (ETEC), and 8.0% higher for Cryptosporidium. Above-median weekly precipitation (median: 13mm; range: 0-396mm) was associated with 29% higher diarrhea (adjusted prevalence ratio 1.29, 95% CI 1.07, 1.55); higher Cryptosporidium, ETEC, STEC, Shigella, Campylobacter, Aeromonas, and adenovirus 40/41; and lower Giardia, sapovirus, and norovirus prevalence. Other associations were weak or null.

Discussion

Higher temperatures and precipitation were associated with higher prevalence of diarrhea and multiple enteropathogens; higher precipitation was associated with lower prevalence of some enteric viruses. Our findings emphasize the heterogeneity of the relationships between hydrometeorological variables and specific enteropathogens, which can be masked when looking at composite measures like all-cause diarrhea. Our results suggest that preventive interventions targeted to reduce enteropathogens just before and during the rainy season may more effectively reduce child diarrhea and enteric pathogen carriage in rural Bangladesh and in settings with similar meteorological characteristics, infrastructure, and enteropathogen transmission.

Author summary

Location-specific weather factors influence the fate, transport, and transmission of enteropathogens that cause diarrhea. We sought to identify hydrometeorological risk factors associated with childhood diarrhea and specific enteropathogens in rural Bangladesh. In this setting, higher temperatures and precipitation were associated with higher diarrhea in children under 5 years old and with several enteropathogens including bacteria, viruses, and parasites. Higher precipitation was associated with the lower prevalence of some enteric viruses. Our data suggest that interventions targeted to reduce pathogen transmission just before and during the rainy season might be most effective to reduce diarrhea in this setting.

Introduction

Diarrhea is a leading cause of disability adjusted life years (DALYs) among children under 10 years old [1]. In low- and middle-income countries (LMICs), where the burden of diarrhea is high, asymptomatic enteropathogen carriage is also common [2] and is linked to child growth failure [2] and impaired child cognitive development [3]. While the global burden of diarrhea in diarrhea morbidity and mortality has steadily declined in LMICs [4], climate change poses a threat to further progress. Diarrhea risk is projected to increase by 15–20% due to climate change in the period 2040–2069 relative to the period 1961–1990 [5], and the World Health Organization estimates that additional deaths will be greatest among children from South Asia and Eastern Africa [6].

Because of the complex etiology of diarrhea, understanding relationships between weather and environmental risk factors and specific enteropathogen infection or asymptomatic carriage is critical to informing both current disease control efforts and climate change adaptation [5,7]. Enteric pathogens have differing environmental survival and transport, and relationships with weather (e.g. precipitation, temperature) and environmental factors (e.g. proximity of surface water to the household) are likely to be specific to particular pathogens [8]. Yet, there is limited evidence from LMICs linking these risk factors to a broad range of enteropathogens and diarrhea [9–12].

Prior studies have found increased diarrhea after heavy rainfall and floods, with at least some attribution given to the flow of fecal material into the environment from overwhelmed sewage treatment facilities in high-income contexts and latrine overflow/leakage in low- and middle-income contexts [13–15]. Parasitic and bacterial diarrhea are more common in the rainy season [12,16]. The concentration-dilution hypothesis posits that the greatest risk will occur at the beginning of a rainy season, after an antecedent period of no rainfall where pathogens can accumulate in the environment. Data have supported this hypothesis in most contexts [17]. Increased temperatures have been associated with higher incidence of bacterial diarrhea but lower incidence of viral diarrhea [16,18]. Higher temperatures are also associated with less Escherichia coli on children’s hands, more cases of food-borne salmonellosis, and higher E. coli concentrations on food, in source water, and in stored drinking water in the household [19–21]. Higher relative humidity can increase the viability of bacteria but increase viral inactivation, while also increasing the transfer efficiency of both from fomites [22–24]. In LMICs, higher temperatures have been associated with a decrease in rotavirus prevalence and small increase in adenovirus prevalence; higher humidity and temperature have been associated with higher enteric bacteria prevalence and lower virus prevalence; and associations with precipitation varied by pathogen [9]. Furthermore, human behaviors such as handwashing and water treatment practices have been shown to vary under differing weather conditions. For example people are less likely to wash their hands but more likely to boil their drinking water during colder weather, and the use of water storage and treatment has been shown to increase in dry periods [25].

In Bangladesh, the prevalence of childhood diarrhea remains high, despite laudable progress in recent decades [26,27]. In addition, due to its low altitude, seasonal monsoon from June-September with resulting flooding, and high population density, Bangladesh is among the top ten countries most vulnerable to extreme weather events due to climate change [28]. In the absence of weather-informed enteropathogen mitigation strategies, Bangladesh’s hard-earned reductions in diarrheal disease morbidity and mortality may be reversed. Yet, few studies have investigated relationships between weather, surface water, and enteric pathogen infection and carriage in a community setting in Bangladesh [29,30].

Here, we leverage community-level pathogen and diarrheal illness data from a large intervention trial in rural Bangladesh matched with fine-scale spatial resolution risk factors. Our objectives were to identify weather and environmental risk factors associated with diarrhea and enteropathogen prevalence in young children in order to inform targeted interventions for this vulnerable population. We hypothesize that there are pathogen-specific differences in the direction, magnitude, and significance of associations with hydrometeorological exposures.

Methods

Ethics statement

The trial was approved by ethical review committees at the International Centre for Diarrheal Disease Research, Bangladesh (icddr,b; PR- 11063), University of California, Berkeley (2011-09-3652), and Stanford University (25863). Participants provided written, informed consent before enrollment and before fecal specimen collection.

Study design

We analyzed data that was collected in a cluster-randomized trial of water, sanitation, handwashing, and nutrition interventions in rural Gazipur, Mymensingh, Tangail, and Kishoreganj districts of Bangladesh (Fig 1) [31]. From 2012–2013, the trial enrolled a total of 720 village clusters spanning a geographic area of 12,500 km2. The trial enrolled pregnant women and measured outcomes longitudinally in their children. On average, approximately 10% of the village was enrolled in each cluster [31] and the sample is generalizable to areas of mainland Bangladesh with similar ecological and socioeconomic features. This analysis was restricted to children enrolled in one of two cohorts of the WASH Benefits Bangladesh study: 1) the diarrhea cohort or 2) the enteropathogen cohort. These data span 3 years and were collected in one of six rounds (three rounds for the diarrhea cohort; and three rounds for the enteropathogen cohort which occurred ~6 months after each diarrhea cohort visit for children enrolled in this cohort). Each data collection round lasted roughly one year. Approximately 54 clusters were sampled per month from the diarrhea cohort, and 14 clusters were sampled per month from the enteropathogen cohort in a spatially dependent manner as the study team moved around the study area over the course of a year. Each round of household visits started in the same place and progressed along the same geographic path, thus compounds which were visited during the rainy season in the first year were more likely to be visited in the rainy season in subsequent years. Our pre-analysis plan is available at https://osf.io/f9cza/, and we note deviations from it in Appendix A in S1 Text.

Fig 1. Cluster-level diarrhea prevalence of study compounds in rural Bangladesh.

Fig 1

Open in a new tab

Panel A) situates the study area within Bangladesh. In panel B) points represent the centroid coordinates of a study cluster and are colored by cluster-level diarrhea prevalence within the past 7 days in the dry or rainy season. The blue star represents Dhaka, the capital of Bangladesh. Shapefiles were obtained from https://cran.r-project.org/web/packages/bangladesh/index.html.

Outcomes

Diarrhea. Field workers measured caregiver-reported diarrhea in the previous 7 days (three or more loose or watery stools in a 24-hour period or a single stool with blood) in children 6 months—5.5 years old (N = 4,478 children). The diarrhea cohort was restricted to the control arms of the original trial or data collected before intervention delivery (i.e. intervention arms at the baseline survey) (N = 7,320 measurements) because intervention effects on diarrhea are modified by weather [32]. Diarrhea prevalence was measured in three main rounds from the diarrhea cohort (N = 6,589 observations; May 2012—July 2013, September 2013—September 2014, December 2014—October 2015). This included index children (live births of women pregnant at the time of enrollment), other children living within the same compound that were younger than 3 years at study enrollment, and children within the same household born after the index children. Data on caregiver-reported diarrhea was collected in three additional rounds from children in the enteropathogen cohort (N = 731 observations from index children in the control arm; December 2012—January 2014, November 2013—November 2014, and March 2015—March 2016).

Enteropathogens

The enteropathogen cohort included index children from a subsample of clusters that was evenly balanced across the control, WSH, nutrition, and N+WSH arms (allocation ratio 1:1:1:1). Clusters in each arm were selected based on logistical feasibility for specimen collection and transport to a central laboratory. Field workers collected stool samples from these children between November 2013 and November 2014 when they were approximately 14 months old and evaluated samples (N = 1,408) for 34 enteric viruses, bacteria and parasites using qPCR (details in Appendix B in S1 Text). Stool samples were collected as part of a routine household visit regardless of children’s symptom status (14% of children reported diarrheal symptoms in the previous 7 days at the time of stool collection). We chose to collect stool from both asymptomatic and symptomatic children because prior literature has shown the importance of asymptomatic pathogen carriage on growth faltering [2,33,34].

We measured the following pre-specified primary outcomes: prevalence of 1) caregiver-reported diarrhea in the past 7 days, 2) any enteric virus (adenovirus 40/41, astrovirus, norovirus GI/GII, rotavirus, sapovirus), and 3) any parasite (Cryptosporidium spp, Enterocytozoon bieneusi, or Giardia spp). We did not include prevalence of any bacteria because at least one bacterium was detected in over 95% of samples. Pathogen-specific prevalence for pathogens detected in >10% of samples were secondary outcomes.

To detect potential unmeasured confounding or outcome measurement bias in the diarrhea analysis, we repeated the analysis using caregiver-reported child bruising in the past 7 days as a negative control outcome [35,36].

Hydrometeorological variables

We matched hydrometeorological risk factors from remote sensing datasets to trial data by date of outcome measurements and geocoordinates of study compounds (groups of households in which patrilineal families share a common courtyard). Two of the exposures (temperature and precipitation) had daily resolution data and the other (surface water and humidity) had monthly resolution. Those with daily resolution were aggregated in different ways over a 7-day window (as well as a 30-day window for temperature) and lagged by 0, 1, 2, or 3 weeks (corresponding to the periods 1–7, 8–14, 15–21, and 21–28 days before outcome measurement). As diarrhea was measured at a weekly resolution (any diarrhea in the previous 7 days), we included 1- to 3-week lags for diarrhea and 0- to 3-week lags for pathogen outcomes per previous studies reporting these as relevant lag times between temperature or precipitation and either diarrhea or enteric pathogens [9,18,37,38]. The lag periods include expected time for microbial growth/death due to weather (e.g. temperature and humidity), mobilization and transport of enteropathogens persisting in the environment (e.g. after precipitation), and post-exposure enteropathogen-specific incubation times within the host to reach detectable levels and/or initiate a diarrheal response (Appendix C in S1 Text). Exposures with monthly resolution were linked to the outcomes by calendar month of outcome ascertainment without any lag period. Here we briefly summarize each variable; Appendix D in S1 Text includes additional details.

Temperature

Using 0.001 degree (~1 km) resolution daily near surface air temperature data from the National Aeronautics and Space Administration Famine Early Warning System Network Land Data Assimilation System (FLDAS) Central Asia dataset [39], we calculated the average, minimum and maximum temperature values for 7 and 30 days prior to outcome observation, with 1-3-week lags for diarrhea and 0-3-week lags for pathogen outcomes.

Precipitation

We obtained 0.1 degree (~10 km) resolution daily precipitation data from the Multi-Source Weighted-Ensemble Precipitation dataset from GloH2O which merges gauge, satellite, and reanalysis data and corrects for bias [40]. We created an indicator for heavy rainfall (any day in the prior week with total daily precipitation >80th percentile [17mm] on rainy days during the study period) and an indicator for whether the weekly sum of precipitation (over the past 7 days) was above or below the median (13mm) for the entire study period. Precipitation thresholds were based on the percentile of all daily totals (across all study years and not season-specific) in order to maximize the generalizability of our results. Each variable was calculated with 1-3-week lags for diarrhea and 0-3-week lags for pathogen outcomes. As a sensitivity analysis for the thresholds to define the categorical variables, we also include a heavy rainfall variable calculated using the 90th percentile (29mm), and weekly sum of precipitation above the 75th (58mm) and 90th (105mm) percentiles. Although prior studies have used different fixed periods between the months of June and October as the rainy season in Bangladesh [30,41,42], we observed variations in the start and end of the rainy period in different years of our study, with substantial rainfall before June in every year. Therefore, we determined the rainy season empirically [10], defined as the continuous period during which the 5-day rolling average of daily precipitation was ≥10mm/day.

Surface water

We obtained monthly 30 m resolution surface water data for the period 1984–2020 from the Global Surface Water Explorer [43]. Variables included: 1) seasonal surface water consistently present in a season, 2) ephemeral surface water present intermittently, and 3) any surface water (ephemeral, seasonal, or permanent) detected in the month of outcome measurement, including man-made and natural water bodies. We calculated tertiles of distance from each household to the nearest surface water and created an indicator for whether the proportion of pixels with surface water within 250m, 500m, or 750m of each household was above or below the median. The distance thresholds were determined to capture risks associated with the potential for flooding in the household environment during heavy rainfall.

Humidity

We obtained mean monthly vapor pressure deficit data with 4 km resolution from Terraclimate [44] and measured associations with a continuous measure in kilopascals (kPA).

Statistical analysis

We used generalized additive mixed models with a Poisson family with a log link to estimate the relationship between exposures (continuous exposures modeled with cubic splines, categorical exposures modeled as factors relative to a reference category) and outcomes [45–47]. To estimate simultaneous confidence intervals, we resampled from the variance-covariance matrix under a multivariate normal distribution [48]. To assess potential spatial autocorrelation, we used Moran’s I coefficient and included a bidimensional thin plate spline function of household latitude and longitude for models in which we detected spatial autocorrelation. Due to multiple measurements within a study compound in the diarrhea dataset (e.g. index children and siblings sampled within a single compound on the same day and data collected over multiple years), we included random intercepts for study compound for diarrhea models. In the pathogen dataset, only a single measurement was collected for each index child thus we included random intercepts for study cluster for pathogen outcomes to account for village-level clustering. For a small number of models evaluating categorical exposures, there was insufficient spatial variation leading to poor model fit for the generalized additive mixed models. In this case, we reran the models as generalized linear models with a Poisson family and log link without spatial terms and used robust standard errors to account for clustering. We adjusted for covariates that were associated with each outcome in our dataset using a likelihood ratio test (p-value < 0.1) (details in Appendix E in S1 Text). For pathogen carriage outcomes, we pooled across all study arms in which specimens were collected (N = 1,408 measurements) to maximize statistical power and included receipt of WASH or nutrition interventions as covariates; this was not necessary for the diarrhea outcome as only observations from the control arms were used. In cases when the number of covariates selected through screening was large and would have resulted in poor model fit due to data sparsity, we removed covariates from the adjustment set sequentially, starting with those with the weakest association to the outcome. We predicted prevalence from adjusted models with continuous covariates set at the median across all samples, intervention set at receiving both WASH and nutrition interventions, household wealth set at the lowest quartile, antibiotic use set at none in the previous 7 days, and sex set as male.

We assessed effect modification for the diarrhea outcome by child age (<1.5 vs. ≥1.5 years) because age is strongly associated with diarrhea risk [49,50]. Because pair-wise correlations between temperature and precipitation were strong (S1 Fig), we evaluated temperature exposure models with and without inclusion of precipitation as an additional covariate. We also assessed effect modification of temperature by precipitation using a model with both exposures and an interaction term. Neither vapor pressure deficit nor surface water displayed correlations with other hydrometeorological variables, so we did not adjust these exposure models for other exposures.

Results

Study participant characteristics

The diarrhea cohort included 4,478 children (mean age = 23 months; SD = 12m) measured in three rounds from 2012–2016, and the pathogen cohort included a subset of 1,408 children (mean age = 14 months; SD = 2) measured primarily in 2014 (Table 1). Other demographic characteristics of each cohort have previously been reported [31,51]. Overall, diarrhea prevalence was 7.1%, with 6.1% prevalence in the dry season and 8.3% in the rainy season (Fig 1B; prevalence ratio of rainy vs dry season 1.36, 95% CI 1.14, 1.62).

Table 1. Study participant and sample characteristics.

Diarrhea cohort Pathogens cohort
Children 4,478 1,408
Observations 7,320 1,408
Year of measurement
 2012 491 (6.7%) 0 (0.0%)
 2013 2,680 (36.6%) 19 (1.3%)
 2014 1,868 (25.5%) 1,389 (98.7%)
 2015 2,170 (29.6%) 0 (0.0%)
 2016 111 (1.5%) 0 (0.0%)
Mean age, months (SD) 22.6 (11.8) 14.0 (2.0)
Antibiotics consumed in the past week --a 233 (16.5%)
Season
 Rainyb 3,408 (46.6%) 620 (44.0%)
 Dry 3,912 (53.4%) 788 (56.0%)
Intervention arm in original trial
 Controlc 7,320 (100.0%) 328 (23.3%)
 WASHd 0 (0.0%) 368 (26.1%)
 Nutrition 0 (0.0%) 352 (25.0%)
 Nutrition + WASHd 0 (0.0%) 360 (25.6%)
Open in a new tab

a Not measured

b Defined as the period when the 5-day rolling average rainfall was ≥10mm/day.

c This includes 2066 baseline measurements (before intervention delivery) from children whose households were randomized to later receive an intervention.

d Water, sanitation, and handwashing

Temporal trends

The study period spanned three rainy seasons from 2013–2015. During the rainy season, average weekly temperature was higher (Fig 2). Distance to surface water was slightly lower during the peak of each rainy season. Diarrhea measurements were not collected in some weeks of the 2013–14 rainy seasons.

Fig 2. Time trends in risk factors, diarrhea prevalence, and number of observations.

Fig 2

Open in a new tab

Shaded bands indicate 95% confidence intervals using robust sandwich standard errors to account for correlation within study clusters. Estimates exclude biweekly periods when the number of observations was <10. Panel A) includes diarrhea measurements from the control arms of the original trial. Breaks in the line indicate periods when the study did not collect data on diarrhea status. In panels B) to E), we display the risk factor values that corresponded to the diarrhea measurements in our analysis. For periods when there was no diarrhea data collection, we display the risk factor values for the next biweekly period with at least 10 diarrhea measurements. Panel F) displays the number of diarrhea observations in each biweekly period. Rainy seasons (grey shaded regions) were determined empirically as periods where the 5-day rolling average rainfall was ≥10mm/day.

Temperature

During the study period, the average weekly temperature across the study area ranged from 15–33°C, the minimum weekly temperature ranged from 13–33°C, and the maximum weekly temperature ranged from 17–34°C. Enteropathogens were measured during a one-year period of the study, during which the minimum temperatures did not fall below 17°C and the average weekly temperature did not fall below 19°C.

According to adjusted model results using a 1-week lag, diarrhea prevalence was 3.3% (95% CI 2.0%, 5.6%) at average weekly temperature of 15°C and 6.8% (95% CI 5.2%, 8.9%) at 30°C; unadjusted models produced similar results (prevalence of 2.9% [95% CI 1.8%, 4.7%] at 15°C and 5.8% [95% CI 4.9%, 7.0%] at 30°C). Trends remained evident at longer lags, and the rate of increase in diarrhea prevalence was higher at weekly average temperatures above 30°C (Fig 3A). Higher average weekly temperatures were also associated with higher prevalence of Shiga toxin-producing E. coli (STEC) (S1 Table), enterotoxigenic E. coli (ETEC) (S1 Table), and higher Cryptosporidium (8.0% [95% CI 3.8%, 16.7%] at 19°C; 16.0% [95% CI 11.5%, 22.1%] at 30°C; 2-week lag using adjusted models) (Fig 3C and 3E). Sapovirus prevalence displayed an inverse relationship, with 15.9% prevalence (95% CI 6.4%, 39.3%) at average weekly temperatures of 19°C and 7.1% (95% CI 4.6%, 11.2%) at 30°C (1-week lag using adjusted models). These models all appeared linear, thus we can infer that each 1° C increase in temperature was associated with approximate prevalence increases of 0.23% for diarrhea, 0.66% for STEC, 1.57% for ETEC, and 0.73% for Cryptosporidium, and a 0.80% decrease for sapovirus. The relationship between temperature and the prevalence of Enterocytozoon bieneusi, Giardia, and norovirus was nonlinear (Fig 3D and 3E), but the confidence intervals were wide. There was no association between temperature and the prevalence of other bacteria (Fig 3B) or other viruses (Fig 3D). Associations between temperature and the combined metrics of ‘any virus’ or ‘any parasite’ masked the heterogeneous effects of individual viral or parasitic enteropathogens (Fig 3D and 3E). Overall, results for diarrhea and pathogen prevalence were similar using alternative lag periods and with stronger effects at 2- and 3-week lags for bacterial pathogens (Figs 3A and S2) and using minimum or maximum weekly temperature (S3–S5 Figs).

Fig 3. Diarrhea and enteropathogen prevalence by weekly average temperature.

Fig 3

Open in a new tab

All panels present adjusted models (see Appendix E in S1 Text for details); shaded bands indicate simultaneous 95% confidence intervals (CIs) accounting for clustering. Panel A) includes diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels B-E) include measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. We present results for the lag periods that best aligned with the incubation period for each outcome (Appendix C in S1 Text); results with alternative lag periods are in S2 Fig. S1 Table provides point estimates and 95% CIs for bacteria (panels C-D) at 19°C and 30°C. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

Precipitation

Median total weekly precipitation during the study period was 13mm (range: 0-396mm), with medians of 0mm (range: 0-198mm) in the dry season and 63mm (range: 0-396mm) in the rainy season. Above-median weekly total precipitation (>13mm) was associated with higher diarrhea prevalence (PR = 1.21; 95% CI 0.98, 1.51; adjusted PR [aPR] = 1.29, 95% CI 1.07, 1.55) using a 2-week lag, but not using 1- or 3-week lags (Fig 4A) nor when using higher threshold cutoffs (75th and 90th percentile, S6 Fig). For parasites, above-median weekly precipitation was associated with higher Cryptosporidium prevalence (e.g., aPR = 2.07; 95% CI 1.39, 3.10 for a 3-week lag), but lower Giardia prevalence (e.g., aPR = 0.66; 95% CI 0.49, 0.90 for a 2-week lag) (Fig 4B). For viruses, above-median weekly precipitation was associated with higher prevalence of adenovirus 40/41, but lower prevalence of sapovirus and norovirus (Fig 4C). For bacteria, above-median weekly precipitation was associated with higher prevalence of enterotoxigenic E. coli (ETEC), STEC, Shigella/EIEC, Campylobacter, and Aeromonas (Fig 4D). With higher percentile cutoffs, results were similar for viruses, and similar but attenuated for bacteria and parasites except for E. bieneusi, for which increased precipitation was associated with a lower prevalence using both the 75th and 90th percentile cutoffs with a 1-week lag (S6 Fig). Heavy rainfall (>80th percentile, 17mm) on a single day was associated with higher prevalence of diarrhea (aPR = 1.31, 95% CI 1.09, 1.57; 2-week lag), Cryptosporidium, adenovirus 40/41, Aeromonas, Campylobacter, Shigella/EIEC, ETEC and lower prevalence of norovirus and sapovirus (S7 Fig). Results were similar with a 90th percentile cutoff threshold for all outcomes except diarrhea where the association was lost (>29mm; S7 Fig). Results varied by lag period, and the lag period expected to be most important based on enteropathogen-specific incubation times (Appendix C in S1 Text) were often not the most strongly associated with above-median precipitation or heavy rainfall (Figs 4B–4D and S6 and S7).

Fig 4. Association between diarrhea and enteropathogen prevalence and above- vs. below-median average weekly precipitation.

Fig 4

Open in a new tab

All panels present adjusted models including an indicator for above median (13mm) average weekly precipitation as the independent variable; unadjusted models produced similar results. Error bars present 95% confidence intervals adjusted for clustering. The x-axis is on the log scale. Panel A) includes diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels B-D) include measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Closed circles in panels B-D indicate the expected most important lag based on enteropathogen-specific incubation times (Appendix C in S1 Text).

Surface water

The mean distance from study households to the closest surface water was 277m (range: 11-1818m) for any surface water, 446m (range: 11–1852) for ephemeral surface water, and 396 (range: 11–1839) for seasonal surface water. Compared to households in the highest tertile of distance to any surface water (> 315m), Aeromonas prevalence was lower for those in the middle tertile (165-315m) (aPR = 0.53; 95% CI 0.29, 0.96) and in the lowest tertile (<165m) (aPR = 0.43; 95% CI 0.23, 0.81) (S8 Fig). Aeromonas results were similar for ephemeral and seasonal surface water. Distance to surface water was not associated with other outcomes.

Around each study household, the median proportion of land that contained any surface water was 0.4% (range: 0–76%) within 250m, 4% (range: 0–76%) within 500m, and 6% (range: 0–76%) within 750m. Most outcomes were not associated with the proportion of land that contained surface water. For certain radii and surface water types, an above-median proportion of surface water was associated with lower diarrhea, Aeromonas, adenovirus 40/41, norovirus, and E. bieneusi prevalence and higher prevalence of Shigella/EIEC; however, most confidence intervals were close to or spanned the null (S9 Fig).

Humidity

Vapor pressure deficit (VPD) ranged from 0.49 to 1.71 kPa (mean 0.86 kPa) during the study period. Higher values of VPD, corresponding to lower humidity, were associated with decreases in ETEC (S10 Fig); prevalence was 41.6% (95% CI 26.8%, 64.5%) at VPD of 0.66 kPa and 16.1% (95% CI 8.1%, 32.0%) at VPD of 1.56 kPa according to the adjusted model. Higher VPD was also associated with lower prevalence of Shigella/EIEC, sapovirus, and any virus (S10 Fig). No associations with other outcomes were observed.

Other analyses

We detected meaningful effect modification in our pre-specified subgroup analyses evaluating child age category (6 months-1.5 years, 1.5–5 years) for the diarrhea outcome. Diarrhea in children aged 1.5–5 years old tended to be impacted by precipitation to a greater extent than younger children (S11 Fig). Heavy rain significantly increased diarrhea prevalence in older children (aPR 1.54, 95% CI 1.21, 1.97) but not those in the younger group (aPR 1.07, 95% CI 0.83, 1.37) in adjusted models with a 3-week lag (results were similar with a 2-week lag).

Due to the strong correlation between temperature and precipitation, we included models with both temperature and precipitation variables. Models which included both variables often had improved model fit (lower Akaike Information Criterion) over models with only temperature. There were few instances in which models were improved by the inclusion of an interaction term (S2 Table).

Our negative control analysis using caregiver-reported child bruising found null associations with most risk factors and a small association in the opposite direction than we observed for diarrhea for a few of the temperature risk factors, suggesting that potential misclassification of the outcome did not substantially influence our results (S3 Table and S12 Fig).

Discussion

In this analysis of the association between weather and environmental variables and enteric pathogen carriage in children in rural Bangladesh, we found that higher weekly average temperatures and above-median precipitation were associated with higher prevalence of diarrhea and certain enteric pathogens. Surface water presence near the household was also associated with carriage of certain enteric bacteria and parasites. Consistent with prior studies, our analysis found that the effects between hydrometeorological variables and enteropathogens are heterogeneous [9,14,16]. We emphasize the heterogeneity even within pathogen class, such that effects can be masked when looking at composite variables (e.g. any viral enteropathogen) or even at the level of all-cause diarrhea.

Precipitation and diarrhea

Above-median weekly precipitation with a 2-week lag was associated with up to 29% higher diarrhea prevalence and heavy rainfall with a 2-week lag was associated with 31% higher prevalence of diarrhea. A prior meta-analysis found both positive and negative associations between precipitation and diarrhea risk [17]. Generally, heavy rainfall preceded by a dry period has been associated with higher diarrhea risk, while heavy rainfall preceded by a wet period has been associated with lower diarrhea risk. This may be because pathogens concentrate during dry periods and then are flushed into the environment when heavy rainfall occurs (the concentration-dilution hypothesis) [17]. We were not able to investigate associations with heavy rainfall preceded by a dry vs. wet period because almost all heavy rainfall periods were preceded by rainy days. Taken together, our findings suggest that in our study setting, even moderately higher precipitation was associated with meaningful increases in diarrhea. However, these results obscure important heterogeneous impacts of precipitation on specific enteropathogens, which we discuss below.

Precipitation and enteropathogens

We found that above-median (>13mm) total weekly precipitation was associated with a higher prevalence of Cryptosporidium, adenovirus 40/41, Aeromonas, Shigella/EIEC, Campylobacter, ETEC, and STEC and lower prevalence of norovirus, sapovirus, and Giardia. Thus, five of the top ten diarrhea-attributable pathogens for children under 24 months old displayed positive associations with above-median total weekly precipitation [49]. A recent meta-analysis of data from children up to 6 years old in 19 LMICs with tropical climates found that higher precipitation was associated with a small decrease in ETEC and Campylobacter prevalence and no difference in Cryptosporidium, Shigella, Giardia, or enteric virus prevalence [9]. Pathogen prevalence was similar for viruses, slightly higher for Campylobacter and Cryptosporidium, and lower for Giardia and Shigella/EIEC in our study compared to the prior meta-analysis. Some of these differences are likely due to the narrower age range of children included in our pathogen cohort (approximately 14 months old) compared to those in the meta-analysis, particularly for Giardia which is typically seen in older children. Additionally, our pathogen cohort captured primarily asymptomatic enteropathogen carriage within the community (only 14% of children had reported diarrhea) while the previous meta-analysis relied heavily on hospital-based surveillance and case-control studies of overt diarrheal illness. Our findings may also differ due to varying urbanicity and WASH infrastructure [52] compared to the locations included in the meta-analysis.

Interestingly, some enteropathogen associations with precipitation were stronger for lag periods that differed from the lag period we expected to be most relevant based on within-host incubation times. Particularly for bacteria, analyses using 2- and 3-week lag periods had stronger associations than those with a 1-week lag; this may reflect higher environmental growth, survival, and/or delayed transport (e.g. from flooding that occurs days after a rain event) following higher precipitation. Adenovirus 40/41, a double-stranded DNA (dsDNA) virus, displayed an inverse relationship with precipitation compared to the other viruses evaluated, which were both single-stranded RNA (ssRNA) viruses. A prior study showed that ultraviolet irradiation led to a 4-log10 inactivation of ssRNA viruses but only a 2-log10 inactivation of dsDNA viruses in water [53]. It is possible that increased water turbidity associated with precipitation events intensifies this relationship. Estimating environmental survival of enteric pathogens under different weather conditions is an important area for future research.

Temperature

For temperature, our estimated associations were more modest overall than those from previous studies. One meta-analysis–encompassing populations of all ages from various climate regions–estimated that a 1° C increase in mean temperature was associated with relative increases of 7% for diarrhea and bacterial diarrhea and no association with viral diarrhea [16]. Another study focusing on children from LMICs estimated that increasing weekly average temperatures in the range between 10–40° C was associated with higher risk of Campylobacter, ETEC, Shigella, Cryptosporidium, Giardia, and adenovirus and lower risk of sapovirus and rotavirus, and generally associations were stronger [9]. Our more modest estimates may reflect the fine-scale spatial resolution of our study or differing background levels of enteropathogen transmission. Our findings reinforce that relationships between temperature and enteropathogen carriage are heterogeneous, even within pathogen class (i.e. virus, bacteria, parasite). The null results for the composite measures of ‘any virus’ or ‘any parasite’ in the background of significant effects for individual viruses/parasites cautions against evaluating higher-level groupings that may mask pathogen-specific effects.

Limitations

This study was subject to several limitations. First, in 2014, the original trial did not collect data during peak rainy season (Fig 2), limiting our ability to make inferences about associations with heavy rainfall in that season. Second, we only measured enteropathogens in a subsample, so statistical power was limited for certain analyses; for this reason, we did not consider it feasible to estimate associations with class-specific diarrhea (e.g., bacterial diarrhea) as prior studies have done. However, measuring enteropathogen carriage in a community-based cohort (including asymptomatic and symptomatic infections) sheds light on weather-specific transmission patterns that may not be discernable when restricting to symptomatic cases and individuals seeking care at health facilities. Third, the enteropathogen sample included a narrow age range (approximately 14 months); results may not generalize to other ages. Rotavirus, a highly climate-sensitive pathogen, was not included in our pathogen-specific analysis because it did not meet our 10% prevalence threshold (although detection of rotavirus is accounted for in our composite “Any virus” variable). Rotavirus was detected in only 2.2% of our community-based cohort, which is similar to other cross-sectional studies where the majority of samples are not associated with diarrheal symptoms. Our effect modification analyses may also have had limited statistical power. Finally, because relationships between hydrometeorological factors, diarrhea, and enteropathogen infection vary by climate zone [54], our findings might not generalize to settings with differing enteropathogen transmission levels, WASH infrastructure, or hydrometeorological characteristics.

Climate and policy implications

Climate change models predict increases in temperature (mean increase of 1° C) and precipitation (mean increase of 6%) in Bangladesh from 2021–2050 [55]. Our findings suggest that under these projections, there would be small increases in the prevalence of diarrhea and certain enteric pathogen infections.

One potential implication of our findings is that interventions to reduce enteropathogen infections in Bangladesh and similar settings may be most beneficial prior to or during the rainy season. A recent analysis found that WASH interventions were more effective at reducing diarrhea during periods of heavy precipitation [32,56]. Seasonal targeting of interventions is a common and cost-effective strategy for other diseases with seasonal transmission patterns, such as malaria [57], but has yet to be used for enteric illness. Of note, the enteropathogens that had higher prevalence under higher temperature and precipitation levels are not currently vaccine-preventable. Many can be prevented by household WASH interventions [51], but sustained use is critical to realizing health benefits [58,59] and has proven difficult to maintain in the long term [60–62]. Given the high cost and user burden of household WASH interventions [60], seasonal targeting via the direct provision of resources and/or increased promotion may improve both public health impact and cost-effectiveness.

Conclusions

We observed heterogeneous impacts of weather on community-level enteropathogen carriage by pathogen class and species for young children in rural Bangladesh. However, the prevalence of a majority of the enteropathogens, as well as diarrheal illness, displayed a positive association with total weekly precipitation above 13mm. In similar settings, preventive interventions targeted at the beginning of the rainy season may be an effective strategy for reducing enteric pathogen infections and carriage, particularly under climate change.

Supporting information

S1 Text. Appendices.

Includes: A. Deviations from pre-analysis plan. Appendix B. Detection of enteropathogens in stool. Appendix C. Incubation periods for each pathogen outcome. Appendix D. Description of risk factors and data sources. Appendix E. Covariate sets tested for each risk factor.

(PDF)

pntd.0012157.s001.pdf (477.1KB, pdf)
S1 Fig. Pairwise relationships between weather variables.

Bivariate scatter plot of continuous climatic risk factors in the diarrhea cohort. Correlation ellipses depict the strength of the association on the basis of the Spearman rank correlation, color of the ellipse indicates the direction of the correlation, and the correlation coefficient is printed inside each ellipse.

(PDF)

pntd.0012157.s002.pdf (408.1KB, pdf)
S2 Fig. Predicted enteropathogen prevalence by weekly average temperature (C) with different lags.

All panels present adjusted models for children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

(PDF)

pntd.0012157.s003.pdf (444.2KB, pdf)
S3 Fig. Predicted diarrhea prevalence by temperature minimum, mean and maximum.

All panels present adjusted models for temperature with a 1-week lag period for diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

(PDF)

pntd.0012157.s004.pdf (254.4KB, pdf)
S4 Fig. Predicted enteropathogen prevalence by temperature minimum.

All panels present adjusted models for children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

(PDF)

pntd.0012157.s005.pdf (377.5KB, pdf)
S5 Fig. Predicted enteropathogen prevalence by temperature maximum.

All panels present adjusted models for children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

(PDF)

pntd.0012157.s006.pdf (357KB, pdf)
S6 Fig. Diarrhea and enteropathogen prevalence by above- vs. below-cutoff for total weekly precipitation at 75th and 90th percentile cutoffs.

All panels present adjusted models including an indicator for above 75th (58mm) or 90th (105mm) percentile average weekly precipitation as the independent variable; unadjusted models produced similar results. Error bars present 95% confidence intervals adjusted for clustering. The x-axis is on the log scale. Panel A) includes diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels B-D) include measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Closed circles in panels B-D indicate the expected most important lag based on enteropathogen-specific incubation times (Appendix C in S1 Text).

(PDF)

pntd.0012157.s007.pdf (335KB, pdf)
S7 Fig. Diarrhea and enteropathogen prevalence by heavy rainfall.

All panels present adjusted models including an indicator variable for heavy rainfall (total weekly precipitation > 80th (17mm) or 90th (29mm) percentile during the study period); unadjusted models produced similar results. Error bars present 95% confidence intervals adjusted for clustering. The x-axis is on the log scale. Panel A) includes diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels B-D) include measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Closed circles in panels B-D indicate the expected most important lag based on enteropathogen-specific incubation times (Appendix C in S1 Text).

(PDF)

pntd.0012157.s008.pdf (333.9KB, pdf)
S8 Fig. Prevalence ratios for diarrhea and enteropathogen carriage and distance from study households to surface water.

All panels present adjusted models; unadjusted models produced similar results. The independent variable was a categorical variable for tertiles of distance from each household to the nearest surface water (<165m; 165m to <316m, 3316m). Error bars present 95% confidence intervals adjusted for clustering. The x-axis is on the log scale. Panel A) includes diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels B-D) include measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial.

(PDF)

pntd.0012157.s009.pdf (361.3KB, pdf)
S9 Fig. Prevalence ratios for diarrhea and enteropathogen carriage associated with the proportion of land around study households that contained surface water.

All panels present adjusted models; unadjusted models produced similar results. The independent variable was an indicator for whether the proportion of pixels with surface water within 250m, 500m, 750m of each household was above or below the median. Error bars present 95% confidence intervals adjusted for clustering. The x-axis is on the log scale. Panel A) includes diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels B-D) include measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial.

(PDF)

pntd.0012157.s010.pdf (339.2KB, pdf)
S10 Fig. Predicted diarrhea and enteropathogen prevalence by vapor pressure deficit.

All panels present adjusted models; shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Pathogen models included measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Diarrhea model includes measurements in children aged 6 months—5.5 years in the control arms in the original trial. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

(PDF)

pntd.0012157.s011.pdf (338.4KB, pdf)
S11 Fig. Interaction between age category and hydrometeorological risk factors on childhood diarrhea prevalence.

Data are from adjusted models and include measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels A) and B) Error bars present 95% confidence intervals adjusted for clustering and the x-axis is on the log scale. Panel A) shows an indicator variable for heavy rainfall (total weekly precipitation > 80th (17mm) or 90th (29mm) percentile during the study period or for above median (13mm), 75th (58mm) or 90th (105mm) percentile average weekly precipitation as the independent variable. Panel B) The independent variable was an indicator for whether the proportion of pixels with surface water within 250m, 500m, 750m of each household was above or below the median or a categorical variable for tertiles of distance from each household to the nearest surface water (<165m; 165m to <316m, ≥316m). Panel C) Shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details). No significant results were observed for other temperature variables or vapor pressure deficit.

(PDF)

pntd.0012157.s012.pdf (378.6KB, pdf)
S12 Fig. Negative control analysis–bruising and temperature.

All panels present adjusted models for weekly minimum, mean, or maximum temperature with the indicated lag period; shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Includes measurements in children aged 6 months—5.5 years in the control arms in the original trial. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

(PDF)

pntd.0012157.s013.pdf (258.2KB, pdf)
S1 Table. Bacterial enteropathogen prevalence estimates at weekly average temperatures of 19°C and 30°C.

(PDF)

pntd.0012157.s014.pdf (202.4KB, pdf)
S2 Table. Akaike Information Criterion (AIC) from adjusted models including temperature and precipitation variables and with or without interaction terms for diarrhea in the previous 7 days or detection of an enteropathogen.

Blue boxes indicate where models without the interaction term had lower AIC while orange boxes highlight models where including the interaction term results in lower AIC. Bolded values emphasize the model with the lowest AIC for the given temperature variable.

(XLSX)

pntd.0012157.s015.xlsx (25.5KB, xlsx)
S3 Table. Adjusted prevalence ratios for caregiver-reported child bruising in the prior 7 days.

(PDF)

pntd.0012157.s016.pdf (172.2KB, pdf)

Acknowledgments

We thank the study participants for their time and sample donation, and hospitality during household sample collection. We also acknowledge the Stanford Research Computing Center for computational resources at the Sherlock high-performance cluster and the International Centre for Diarrhoeal Disease Research, Bangladesh for institutional and administrative support for the field data collection.

Data Availability

Not all data can be shared publicly due to Personally Identifiable Information (household GPS coordinates). However, de-identified data is publicly available on the Open Science Framework (https://osf.io/yt67k/) and all code used for the analysis is available on GitHub at https://github.com/jgrembi/wbb-weather-diarrhea-enteropathogens. Analyses used R statistical software (version 4.2.0).

Funding Statement

Research reported in this publication was supported in part by the Bill and Melinda Gates Foundation (https://www.gatesfoundation.org; to the University of California, Berkeley, under grant number OPPGD759 to SPL and to Stanford University, under grant number OPP1161946 to SPL). This work was supported by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health (https://www.niaid.nih.gov) under Award Numbers K01AI141616 to JBC and R01AI166671 to BFA. JBC is a Chan Zuckerberg Biohub (https://www.czbiohub.org) Investigator. This work was also supported by National Heart, Lung, And Blood Institute of the National Institutes of Health (https://www.nhlbi.nih.gov) under award number T32HL151323 to ATN; by a Stanford University School of Medicine Dean’s Postdoctoral Fellowship (https://med.stanford.edu) and by the National Institute of Diabetes and Digestive and Kidney Diseases of the National Institute of Health (https://www.niddk.nih.gov) under award number F32DK130574 to JAG; and by the National Institute of General Medical Sciences under grant number T32GM144273 to TSA. Funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

References

  • 1.Vos T, Lim SS, Abbafati C, Abbas KM, Abbasi M, Abbasifard M, et al. Global burden of 369 diseases and injuries in 204 countries and territories, 1990–2019: a systematic analysis for the Global Burden of Disease Study 2019. The Lancet. 2020. Oct 17;396(10258):1204–22. doi: 10.1016/S0140-6736(20)30925-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Rogawski ET, Liu J, Platts-Mills JA, Kabir F, Lertsethtakarn P, Siguas M, et al. Use of quantitative molecular diagnostic methods to investigate the effect of enteropathogen infections on linear growth in children in low-resource settings: longitudinal analysis of results from the MAL-ED cohort study. The Lancet Global Health. 2018. Dec 1;6(12):e1319–28. doi: 10.1016/S2214-109X(18)30351-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.MAL-ED Network Investigators. Early childhood cognitive development is affected by interactions among illness, diet, enteropathogens and the home environment: findings from the MAL-ED birth cohort study. BMJ Global Health. 2018. Jul 1;3(4):e000752. doi: 10.1136/bmjgh-2018-000752 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Institute for Health Metrics and Evaluation (IHME). GBD Compare Data Visualization. [Internet]. 2020. [cited 2023 May 25]. Available from: http://vizhub.healthdata.org/gbd-compare. [Google Scholar]
  • 5.Kolstad EW, Johansson KA. Uncertainties Associated with Quantifying Climate Change Impacts on Human Health: A Case Study for Diarrhea. Environ Health Perspect. 2011. Mar;119(3):299–305. doi: 10.1289/ehp.1002060 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Hales S, Kovats R, Lloyd S, Campbell-Lendrum D. Quantitative risk assessment of the effects of climate change on selected causes of death, 2030s and 2050s [Internet]. World Health Organization; 2014. Available from: https://apps.who.int/iris/bitstream/handle/10665/134014/9789241507691_eng.pdf [Google Scholar]
  • 7.Levy K, Smith SM, Carlton EJ. Climate Change Impacts on Waterborne Diseases: Moving Toward Designing Interventions. Curr Environ Health Rep. 2018. Jun;5(2):272–82. doi: 10.1007/s40572-018-0199-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Sterk A, Schijven J, de Nijs T, de Roda Husman AM. Direct and Indirect Effects of Climate Change on the Risk of Infection by Water-Transmitted Pathogens. Environ Sci Technol. 2013. Nov 19;47(22):12648–60. doi: 10.1021/es403549s [DOI] [PubMed] [Google Scholar]
  • 9.Colston JM, Zaitchik BF, Badr HS, Burnett E, Ali SA, Rayamajhi A, et al. Associations Between Eight Earth Observation-Derived Climate Variables and Enteropathogen Infection: An Independent Participant Data Meta-Analysis of Surveillance Studies With Broad Spectrum Nucleic Acid Diagnostics. GeoHealth. 2022;6(1):e2021GH000452. doi: 10.1029/2021GH000452 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Chao DL, Roose A, Roh M, Kotloff KL, Proctor JL. The seasonality of diarrheal pathogens: A retrospective study of seven sites over three years. PLOS Neglected Tropical Diseases. 2019. Aug 15;13(8):e0007211. doi: 10.1371/journal.pntd.0007211 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Chua PLC, Ng CFS, Tobias A, Seposo XT, Hashizume M. Associations between ambient temperature and enteric infections by pathogen: a systematic review and meta-analysis. The Lancet Planetary Health. 2022. Mar 1;6(3):e202–18. doi: 10.1016/S2542-5196(22)00003-1 [DOI] [PubMed] [Google Scholar]
  • 12.Mero S, Lääveri T, Ursing J, Rombo L, Kofoed PE, Kantele A. Seasonal variation of diarrhoeal pathogens among Guinea-Bissauan children under five years of age. PLOS Neglected Tropical Diseases. 2023. Mar 13;17(3):e0011179. doi: 10.1371/journal.pntd.0011179 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Mertens A, Balakrishnan K, Ramaswamy P, Rajkumar P, Ramaprabha P, Durairaj N, et al. Associations between High Temperature, Heavy Rainfall, and Diarrhea among Young Children in Rural Tamil Nadu, India: A Prospective Cohort Study. Environmental Health Perspectives. 2019. Apr;127(4):047004–047004. doi: 10.1289/EHP3711 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Levy K, Woster AP, Goldstein RS, Carlton EJ. Untangling the Impacts of Climate Change on Waterborne Diseases: a Systematic Review of Relationships between Diarrheal Diseases and Temperature, Rainfall, Flooding, and Drought. Environ Sci Technol. 2016. May 17;50(10):4905–22. doi: 10.1021/acs.est.5b06186 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Wang P, Asare EO, Pitzer VE, Dubrow R, Chen K. Floods and Diarrhea Risk in Young Children in Low- and Middle-Income Countries. JAMA Pediatrics. 2023. Nov 1;177(11):1206–14. doi: 10.1001/jamapediatrics.2023.3964 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Carlton EJ, Woster AP, DeWitt P, Goldstein RS, Levy K. A systematic review and meta-analysis of ambient temperature and diarrhoeal diseases. International Journal of Epidemiology. 2016. Feb 1;45(1):117–30. doi: 10.1093/ije/dyv296 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Kraay ANM, Man O, Levy MC, Levy K, Ionides E, Eisenberg JNS. Understanding the Impact of Rainfall on Diarrhea: Testing the Concentration-Dilution Hypothesis Using a Systematic Review and Meta-Analysis. Environmental Health Perspectives. 2020. Dec 7;128(12):126001. doi: 10.1289/EHP6181 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Wang P, Goggins WB, Chan EYY. A time-series study of the association of rainfall, relative humidity and ambient temperature with hospitalizations for rotavirus and norovirus infection among children in Hong Kong. Science of The Total Environment. 2018. Dec 1;643:414–22. doi: 10.1016/j.scitotenv.2018.06.189 [DOI] [PubMed] [Google Scholar]
  • 19.Powers JE, Mureithi M, Mboya J, Campolo J, Swarthout JM, Pajka J, et al. Effects of High Temperature and Heavy Precipitation on Drinking Water Quality and Child Hand Contamination Levels in Rural Kenya. Environ Sci Technol. 2023. May 2;57(17):6975–88. doi: 10.1021/acs.est.2c07284 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Kovats RS, Edwards SJ, Hajat S, Armstrong BG, Ebi KL, Menne B. The effect of temperature on food poisoning: a time-series analysis of salmonellosis in ten European countries. Epidemiology and infection. 2004. Jun;132(3):443–53. doi: 10.1017/s0950268804001992 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Niven C, Islam M, Nguyen A, Mertens A, Pickering AJ, Kwong LH, et al. Effects of weather extremes on fecal contamination along pathogen transmission pathways in rural Bangladeshi households [Internet]. medRxiv; 2023. [cited 2024 Feb 8]. p. 2023.12.27.23300582. Available from: https://www.medrxiv.org/content/10.1101/2023.12.27.23300582v1 [Google Scholar]
  • 22.Lin K, Marr LC. Humidity-Dependent Decay of Viruses, but Not Bacteria, in Aerosols and Droplets Follows Disinfection Kinetics. Environ Sci Technol. 2020;54:44–44. doi: 10.1021/acs.est.9b04959 [DOI] [PubMed] [Google Scholar]
  • 23.Kim SJ, Si J, Lee JE, Ko G. Temperature and Humidity Influences on Inactivation Kinetics of Enteric Viruses on Surfaces. 2012; Available from: https://pubs.acs.org/sharingguidelines [DOI] [PubMed] [Google Scholar]
  • 24.Lopez GU, Gerba CP, Tamimi AH, Kitajima M, Maxwell SL, Rose JB. Transfer Efficiency of Bacteria and Viruses from Porous and Nonporous Fomites to Fingers under Different Relative Humidity Conditions. Applied and Environmental Microbiology. 2013;79:5728–34. doi: 10.1128/AEM.01030-13 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Charles KJ, Howard G, Villalobos Prats E, Gruber J, Alam S, Alamgir ASM, et al. Infrastructure alone cannot ensure resilience to weather events in drinking water supplies. Science of The Total Environment. 2022. Mar 20;813:151876. doi: 10.1016/j.scitotenv.2021.151876 [DOI] [PubMed] [Google Scholar]
  • 26.Troeger C, Blacker BF, Khalil IA, Rao PC, Cao S, Zimsen SR, et al. Estimates of the global, regional, and national morbidity, mortality, and aetiologies of diarrhoea in 195 countries: a systematic analysis for the Global Burden of Disease Study 2016. The Lancet Infectious Diseases. 2018. Nov;18(11):1211–28. doi: 10.1016/S1473-3099(18)30362-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Karambizi NU, McMahan CS, Blue CN, Temesvari LA. Global estimated Disability-Adjusted Life-Years (DALYs) of diarrheal diseases: A systematic analysis of data from 28 years of the global burden of disease study. PLOS ONE. 2021. Oct 27;16(10):e0259077. doi: 10.1371/journal.pone.0259077 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Eckstein D, Künzel V, Schäfer L. The Global Climate Risk Index 2021 [Internet]. Bonn: Germanwatch; 2021. [cited 2023 Feb 17]. Available from: https://bvearmb.do/handle/123456789/1306 [Google Scholar]
  • 29.Dewan AM, Corner R, Hashizume M, Ongee ET. Typhoid Fever and Its Association with Environmental Factors in the Dhaka Metropolitan Area of Bangladesh: A Spatial and Time-Series Approach. PLOS Neglected Tropical Diseases. 2013. Jan 24;7(1):e1998. doi: 10.1371/journal.pntd.0001998 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Ali M, Emch M, Donnay JP, Yunus M, Sack RB. Identifying environmental risk factors for endemic cholera: a raster GIS approach. Health & Place. 2002. Sep;8(3):201–10. doi: 10.1016/s1353-8292(01)00043-0 [DOI] [PubMed] [Google Scholar]
  • 31.Luby S, Rahman M, Arnold B, Unicomb L, Ashraf S, Winch P, et al. Effects of water quality, sanitation, handwashing and nutritional interventions on diarrhoea and child growth in rural Bangladesh: A cluster randomized trial. Lancet Global Health. 2018;6(3):e302–15. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Nguyen AT, Grembi JA, Riviere M, Heitmann GB, Hutson WD, Athni TS, et al. Influence of Temperature and Precipitation on the Effectiveness of Water, Sanitation, and Handwashing Interventions against Childhood Diarrheal Disease in Rural Bangladesh: A Reanalysis of the WASH Benefits Bangladesh Trial. Environ Health Perspect. 2024. Apr 11;132(4):047006. doi: 10.1289/EHP13807 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.MAL-ED Network Investigators. Relationship between growth and illness, enteropathogens and dietary intakes in the first 2 years of life: findings from the MAL-ED birth cohort study. BMJ Global Health. 2017. Dec 1;2(4):e000370. doi: 10.1136/bmjgh-2017-000370 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Kabir F, Iqbal J, Jamil Z, Iqbal NT, Mallawaarachchi I, Aziz F, et al. Impact of enteropathogens on faltering growth in a resource-limited setting. Frontiers in Nutrition. 2023. Jan 10;9:1081833. doi: 10.3389/fnut.2022.1081833 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Lipsitch M, Tchetgen ET, Cohen T. Negative Controls: A Tool for Detecting Confounding and Bias in Observational Studies. Epidemiology. 2010. May;21(3):383–8. doi: 10.1097/EDE.0b013e3181d61eeb [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Arnold BF, Ercumen A, Benjamin-Chung J, Colford JM. Brief Report: Negative Controls to Detect Selection Bias and Measurement Bias in Epidemiologic Studies. Epidemiology. 2016. Sep;27(5):637–41. doi: 10.1097/EDE.0000000000000504 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Chen MJ, Lin CY, Wu YT, Wu PC, Lung SC, Su HJ. Effects of extreme precipitation to the distribution of infectious diseases in Taiwan, 1994–2008. PloS one. 2012;7(6):e34651–e34651. doi: 10.1371/journal.pone.0034651 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Andhikaputra G, Sapkota A, Lin YK, Chan TC, Gao C, Deng LW, et al. The impact of temperature and precipitation on all-infectious-, bacterial-, and viral-diarrheal disease in Taiwan. Science of The Total Environment. 2023. Mar 1;862:160850. doi: 10.1016/j.scitotenv.2022.160850 [DOI] [PubMed] [Google Scholar]
  • 39.McNally A, Jacob J, Arsenault K, Slinski K, Sarmiento DP, Hoell A, et al. A Central Asia hydrologic monitoring dataset for food and water security applications in Afghanistan. Earth System Science Data. 2022. Jul 8;14(7):3115–35. [Google Scholar]
  • 40.Beck HE, Wood EF, Pan M, Fisher CK, Miralles DG, Dijk AIJM van, et al. MSWEP V2 Global 3-Hourly 0.1° Precipitation: Methodology and Quantitative Assessment. Bulletin of the American Meteorological Society. 2019. Mar 1;100(3):473–500. [Google Scholar]
  • 41.Chowdhury FR, Ibrahim QSU, Bari MS, Alam MMJ, Dunachie SJ, Rodriguez-Morales AJ, et al. The association between temperature, rainfall and humidity with common climate-sensitive infectious diseases in Bangladesh. PLOS ONE. 2018. Jun 21;13(6):e0199579. doi: 10.1371/journal.pone.0199579 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Hashizume M, Armstrong B, Wagatsuma Y, Faruque ASG, Hayashi T, Sack DA. Rotavirus infections and climate variability in Dhaka, Bangladesh: a time-series analysis. Epidemiology & Infection. 2008. Sep;136(9):1281–9. doi: 10.1017/S0950268807009776 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Pekel JF, Cottam A, Gorelick N, Belward AS. High-resolution mapping of global surface water and its long-term changes. Nature. 2016. Dec;540(7633):418–22. doi: 10.1038/nature20584 [DOI] [PubMed] [Google Scholar]
  • 44.Abatzoglou JT, Dobrowski SZ, Parks SA, Hegewisch KC. TerraClimate, a high-resolution global dataset of monthly climate and climatic water balance from 1958–2015. Sci Data. 2018. Dec;5(1):170191. doi: 10.1038/sdata.2017.191 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Hastie T, Tibshirani R. Generalized Additive Models. Statist Sci. 1986. Aug;1(3):297–310. [DOI] [PubMed] [Google Scholar]
  • 46.Wood SN. Generalized additive models: an introduction with R. Boca Raton, FL: Chapman & Hall/CRC; 2006. 391 p. (Texts in statistical science). [Google Scholar]
  • 47.Zou G. A Modified Poisson Regression Approach to Prospective Studies with Binary Data. American Journal of Epidemiology. 2004. Apr 1;159(7):702–6. doi: 10.1093/aje/kwh090 [DOI] [PubMed] [Google Scholar]
  • 48.Marra G, Wood SN. Coverage Properties of Confidence Intervals for Generalized Additive Model Components: Coverage properties of GAM intervals. Scandinavian Journal of Statistics. 2012. Mar;39(1):53–74. [Google Scholar]
  • 49.Platts-Mills JA, Liu J, Rogawski ET, Kabir F, Lertsethtakarn P, Siguas M, et al. Use of quantitative molecular diagnostic methods to assess the aetiology, burden, and clinical characteristics of diarrhoea in children in low-resource settings: a reanalysis of the MAL-ED cohort study. The Lancet Global Health. 2018. Dec;6(12):e1309–18. doi: 10.1016/S2214-109X(18)30349-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Platts-Mills JA, Babji S, Bodhidatta L, Gratz J, Haque R, Havt A, et al. Pathogen-specific burdens of community diarrhoea in developing countries: A multisite birth cohort study (MAL-ED). The Lancet Global Health. 2015;3(9):564–75. doi: 10.1016/S2214-109X(15)00151-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Grembi JA, Lin A, Karim MA, Islam MO, Miah R, Arnold BF, et al. Effect of Water, Sanitation, Handwashing, and Nutrition Interventions on Enteropathogens in Children 14 Months Old: A Cluster-Randomized Controlled Trial in Rural Bangladesh. The Journal of Infectious Diseases. 2023. Feb 1;227(3):434–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Dimitrova A, Gershunov A, Levy MC, Benmarhnia T. Uncovering social and environmental factors that increase the burden of climate-sensitive diarrheal infections on children. Proceedings of the National Academy of Sciences. 2023. Jan 17;120(3):e2119409120. doi: 10.1073/pnas.2119409120 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Calgua B, Carratalà A, Guerrero-Latorre L, De Abreu Corrêa A, Kohn T, Sommer R, et al. UVC Inactivation of dsDNA and ssRNA Viruses in Water: UV Fluences and a qPCR-Based Approach to Evaluate Decay on Viral Infectivity. Food Environ Virol. 2014. Dec;6(4):260–8. doi: 10.1007/s12560-014-9157-1 [DOI] [PubMed] [Google Scholar]
  • 54.Dimitrova A, McElroy S, Levy M, Gershunov A, Benmarhnia T. Precipitation variability and risk of infectious disease in children under 5 years for 32 countries: a global analysis using Demographic and Health Survey data. The Lancet Planetary Health. 2022. Feb 1;6(2):e147–55. doi: 10.1016/S2542-5196(21)00325-9 [DOI] [PubMed] [Google Scholar]
  • 55.Choi YW, Campbell DJ, Aldridge JC, Eltahir EAB. Near-term regional climate change over Bangladesh. Clim Dyn. 2021. Dec 1;57(11):3055–73. [Google Scholar]
  • 56.Ante-Testard PA, Rerolle F, Nguyen AT, Ashraf S, Parvez SM, Naser AM, et al. WASH interventions and child diarrhea at the interface of climate and socioeconomic position in Bangladesh. Nat Commun. 2024. Feb 20;15(1):1556. doi: 10.1038/s41467-024-45624-1 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Chandramohan D, Zongo I, Sagara I, Cairns M, Yerbanga RS, Diarra M, et al. Seasonal Malaria Vaccination with or without Seasonal Malaria Chemoprevention. New England Journal of Medicine. 2021. Sep 9;385(11):1005–17. doi: 10.1056/NEJMoa2026330 [DOI] [PubMed] [Google Scholar]
  • 58.Brown J, Clasen T. High Adherence Is Necessary to Realize Health Gains from Water Quality Interventions. PLOS ONE. 2012. May 7;7(5):e36735. doi: 10.1371/journal.pone.0036735 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Garn JV, Trinies V, Toubkiss J, Freeman MC. The Role of Adherence on the Impact of a School-Based Water, Sanitation, and Hygiene Intervention in Mali. Am J Trop Med Hyg. 2017. Apr 5;96(4):984–93. doi: 10.4269/ajtmh.16-0558 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Humphrey JH. Reducing the user burden in WASH interventions for low-income countries. The Lancet Global Health. 2019. Sep 1;7(9):e1158–9. doi: 10.1016/S2214-109X(19)30340-7 [DOI] [PubMed] [Google Scholar]
  • 61.Luoto J, Najnin N, Mahmud M, Albert J, Islam MS, Luby S, et al. What Point-of-Use Water Treatment Products Do Consumers Use? Evidence from a Randomized Controlled Trial among the Urban Poor in Bangladesh. PLOS ONE. 2011. Oct 20;6(10):e26132. doi: 10.1371/journal.pone.0026132 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.Curtis VA, Danquah LO, Aunger RV. Planned, motivated and habitual hygiene behaviour: an eleven country review. Health Education Research. 2009. Aug 1;24(4):655–73. doi: 10.1093/her/cyp002 [DOI] [PMC free article] [PubMed] [Google Scholar]
PLoS Negl Trop Dis. doi: 10.1371/journal.pntd.0012157.r001

Decision Letter 0

Stuart D Blacksell, Marilia Sá Carvalho

30 Oct 2023

Dear Dr. Grembi,

Thank you very much for submitting your manuscript "Influence of weather and environmental risk factors on child diarrhea and enteropathogens in rural Bangladesh" for consideration at PLOS Neglected Tropical Diseases. As with all papers reviewed by the journal, your manuscript was reviewed by members of the editorial board and by several independent reviewers. In light of the reviews (below this email), we would like to invite the resubmission of a significantly-revised version that takes into account the reviewers' comments.

We cannot make any decision about publication until we have seen the revised manuscript and your response to the reviewers' comments. Your revised manuscript is also likely to be sent to reviewers for further evaluation.

When you are ready to resubmit, please upload the following:

[1] A letter containing a detailed list of your responses to the review comments and a description of the changes you have made in the manuscript. Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

[2] Two versions of the revised manuscript: one with either highlights or tracked changes denoting where the text has been changed; the other a clean version (uploaded as the manuscript file).

Important additional instructions are given below your reviewer comments.

Please prepare and submit your revised manuscript within 60 days. If you anticipate any delay, please let us know the expected resubmission date by replying to this email. Please note that revised manuscripts received after the 60-day due date may require evaluation and peer review similar to newly submitted manuscripts.

Thank you again for your submission. We hope that our editorial process has been constructive so far, and we welcome your feedback at any time. Please don't hesitate to contact us if you have any questions or comments.

Sincerely,

Marilia Sá Carvalho

Academic Editor

PLOS Neglected Tropical Diseases

Stuart Blacksell

Section Editor

PLOS Neglected Tropical Diseases

***********************

Reviewer's Responses to Questions

Key Review Criteria Required for Acceptance?

As you describe the new analyses required for acceptance, please consider the following:

Methods

-Are the objectives of the study clearly articulated with a clear testable hypothesis stated?

-Is the study design appropriate to address the stated objectives?

-Is the population clearly described and appropriate for the hypothesis being tested?

-Is the sample size sufficient to ensure adequate power to address the hypothesis being tested?

-Were correct statistical analysis used to support conclusions?

-Are there concerns about ethical or regulatory requirements being met?

Reviewer #1: -The study objectives are clear and the methods are appropriate and well-presented.

-Line 202-203: I am not sure I agree with your conclusion that pair-wise correlations between weather exposures were not strong and thus not necessary to adjust individual exposure models for other weather exposures. This makes sense for surface water distance and perhaps also for vapor pressure deficit, but I don’t think it makes sense for temperature and precipitation. I would consider correlation coefficients around 0.6-0.7 (as you show in Fig S1) to be moderate or strong, and this makes sense - temperature and precipitation are typically correlated. I would consider looking at the effects of temperature and precipitation in combined models, or address this further and provide a justification for not doing so.

- If you do include temp and precip in combined models, you might also consider testing whether the inclusion of an interaction term between temp and precip improves the model.

-How did you select these exposure variables? E.g. duration of lag periods, heavy rainfall as >80th percentile, weekly precip over the median. Recommend including a justification. Are the results consistent if these specifications change? It might be valuable to include sensitivity analyses in the supplemental. I see you analyzed 1, 2, and 3 week lags. What about the choice of the 80th percentile and the median precipitation?

-Line 323: How were 250m, 500m, 750m distance thresholds selected?

-Line 232: Could you provide a citation or justification for this method of determining the rainy season?

-The authors have clearly made an effort to make their research process transparent (publicly available deidentified data, code, and pre-specified analysis plan). This is commendable. For the code - you state that code is available at https://github.com/jgrembi/WBB-weather-diarrhea-pathogens, but I get an error message when I try to access this. I am assuming you intend to make it available post publication?

Reviewer #2: The manuscript reports an analysis of secondary data that was originally collected to test the efficacy of WASH and nutrition interventions on childhood diarrhea and growth. The objective is stated clearly at the end of the introduction, though it is not restated as a testable hypothesis, perhaps because of the complexity of the analysis and the multiple pairwise exposure/outcome associations tested. If stating a hypothesis is a requirement, perhaps it could be something like “there are pathogen-specific differences in the direction, magnitude, and significance of associations between environmental exposures and enteropathy outcomes”.

Broadly speaking, the study design is appropriate and follows approaches that have precedents in prior literature. The authors match publicly available hydrometeorological exposure data spatiotemporally with epidemiological outcome data collected by the trial. Unadjusted and adjusted associations are characterized by fitting GAM models, while giving appropriate consideration to lagged effects, spatial autocorrelation, and potential confounding. Notably, they link to their pre-analysis plan and document deviations from it in an appendix, a level of transparency that is rare.

I think that the structure of the dataset and its implication for the model specification could be better explained. It seems to me that the data is hierarchical, with stool samples/diarrhea episodes from multiple rounds nested within subjects (children), subjects nested within compounds, and compounds within villages (clusters). Random intercepts were used in the models to account for this clustering but at different levels depending on the outcome (compound-level for diarrhea, village-level for enteropathogens). I’m sure there were good reasons for treating the outcomes differently, but the justification is not clear as currently written. Were subject-level random effects specified in all models?

Were the stool samples in which enteropathogens were diagnosed all collected during diarrheal episodes, or were there samples from asymptomatic children?

The use of child bruising as a negative control outcome is appropriate since it can’t plausibly be linked with the exposures or outcomes, however I disagree that the reason for this is “to detect potential misclassification of diarrhea”. The purpose of a negative control outcome is to infer causality with greater confidence (by demonstrating the specificity of the association) and to rule out measured and unmeasured confounding as an explanation for any statistically significant main associations observed (i.e., if an association between an exposure and an NCO is significant, then the main association may be spurious).

I think that the authors’ choice of climate data sources requires some justification. FLDAS was generated for Central Asia, specifically Afghanistan, and not for Bangladesh. The resolution of the NOAA precipitation dataset used is extremely coarse at 55km2. It’s difficult to tell from figure 1 (perhaps include a scalebar) but probably all the clusters are covered by a handful of grid squares, while precipitation varies on a very small scale. The CHIRPS precipitation dataset (https://doi.org/10.1038/sdata.2015.66) has much higher resolution, but also consider the ERA5 dataset (https://doi.org/10.1002/qj.3803), which has been used in epidemiological studies of infectious syndromes (https://doi.org/10.1038/s41467-021-25914-8) and includes many of the variables relevant to this analysis (https://doi.org/10.1038/s41597-023-02276-y), with the advantage that they are from a single source and mutually consistent.

The explanation for the aggregation and lagging of the meteorological exposures is a little unclear. Was the resolution of both the enteropathogen and diarrhea variables daily? At points it sounds like for diarrhea it was weekly. In any case, I think you should state in the main text that two of the main exposures had daily resolution (temperature and precipitation) and the other had monthly resolution (surface water and VPD). Those with daily resolution were aggregated in different ways over a 7-day window (as well as a 30-day window for temperature), and lagged by 7, 14, and 21 days. So, if t0 is the day of outcome ascertainment, then one of the exposures was the average temperature from t-7 to t-14 (see notation used in 10.1029/2021GH000452). The monthly exposures were linked to the outcomes by calendar month and year of outcome ascertainment (without lagging). If I’ve understood this correctly, then I think it might be possible to tighten this explanation. This is especially important for the surface water data as I expect it will not be familiar to readers.

Some additional details of the modeling approach appear to be missing. Please state the link function and family of the model used. From what I can infer, the outcomes were binary (presence or absence of a diarrhea episode/enteropathogen), so would that have been a log link and binomial family? Were prevalence and prevalence ratios approximated from transformations of odds ratios or risk ratios?

If I’ve understood appendix 5 correctly, there were a base set of non-environmental exposures which were included in all models to adjust for their potential confounding effects. This is appropriate. Then for each environmental exposure, a single specific other confounder was considered (but not in the case of precipitation and temperature). Since monthly precipitation was considered as a confounder for two exposures, why not fit a fully adjusted model, testing the associations of all 5 risk factors in table S5 in the presence of each other? (The answer to this question may be that you would need a way of choosing which lag and aggregation window length to use for each exposure/outcome pair and this would overcomplicate the analysis).

Has the distribution of animal ownership by compound been reported anywhere? The use of this variable seems to me to be a bit of a buried lead, and I think a lot of readers would be interested to see the main effect of that variable on the outcomes, or its interaction with the environmental exposures. I do understand, however, that this is beyond the scope of this analysis.

Lines 194 – 195: I’m a little unsure about the justification for dealing with study arm differently for diarrhea and pathogens. Are the results different if you include study arm as a covariate for the diarrhea models, or exclude the intervention arms for pathogens? It seems like a common approach to both outcomes would be fine and would make results more comparable.

Lines 199 – 200: I suggest “(continuous exposures modeled with cubic splines, categorical exposures modeled as factors relative to a reference category)”.

Reviewer #3: The methodology lacks clarity and structure.

First I suggest describing at the beginning of the methodology the study design used in order to correctly interpret the association measures evaluated in the study.

I suggest adding information on the country, what percentage of the population is represented in the sample of this research, as well as the representativeness of each of the provinces sampled.

I suggest incorporating more information about the climate of the country and the provinces of the study. For example, the monsoon from June to September that often causing flooding, Hurricane season can occur from May to June and October to November, month with high temperatures and humidity and the period of time with major storms.

Why were trials conducted on pregnant women?

I consider that Figure 1 should show the provinces sampled with the continental location of Bangladesh and the left panel I suggest to put it in a separate figure later when the associations with climate are presented.

Why was the subsample taken in that year (2014)?

Although the sampling information is detailed in supplementary material, I suggest clarifying some issues in the body of the manuscript:

1º At what scale was the climate data worked on?

2ºHow many weather stations do you work with?

3º At what time of the year were the samples taken, given the intra-annual climatic variability?

4º In the case of the environmental variable: only natural or artificial water bodies were recorded (for example, ponds in parks, drains or containment pools).

Proximity to sources of garbage accumulation could have been another interesting environmental variable to take into account and open dumps can be recorded through spectral signatures as well.

Line 212: “We assessed effect modification for the diarrhea outcome by child age”. Why?

--------------------

Results

-Does the analysis presented match the analysis plan?

-Are the results clearly and completely presented?

-Are the figures (Tables, Images) of sufficient quality for clarity?

Reviewer #1: -The results are clearly presented and the figures are effective.

-Lines 192-195: In your pre-specified analysis plan, you state that “We will pool across all intervention arms from the original study to maximize statistical power and will include study arm as a covariate to adjust for any effects of study interventions on each outcome, except for the diarrhea outcome where we will conduct subgroup analyses separately within the control arm or intervention arms (all grouped together) due to the significant influence of the interventions on diarrhea in a seasonally-dependent manner.” However, in the manuscript, it seems like you exclude diarrhea data for the intervention arms. Please describe if and why this was changed.

-Line 337: Where are these results?

-S1 Appendix 2: I don’t see these additional results at the link included.

Reviewer #2: Line 221: Was the difference in diarrhea incidence in the rainy vs the dry season statistically significant?

Line 258: When you go from describing the distribution of the variables to reporting predictions of the model results, I would start a new paragraph and start it with “According to model results…” or something like that. If I’ve understood correctly, the prevalences reported in table S1 and lines 258 – 264 are predictions from GAMs at two values for temperature at either extreme of the range (and statistically significant results are in bold typeface in table S1). This is not clearly stated, however.

I assume that rotavirus was excluded as a pathogen-specific outcome because it was detected in fewer than 10% of samples (same for astrovirus). Since rotavirus is a highly climate sensitive pathogen and readers will be looking for that result, I think that this should be stated somewhere (e.g., in the limitations paragraph).

In line 285, an adjusted prevalence ratio (aPR) is reported, but after that, the abbreviation “PR” is used suggesting that the rest of the results are unadjusted. Similarly, in the coefficient plots, unadjusted PRs are plotted and the figure legends state that adjusted results are similar. I think it makes more sense to consistently report the adjusted results.

What possible causal explanation could there be for why aEPEC has a negative association with rainfall, but tEPEC has a positive one? I strongly suspect that this is the result of confounding. aEPEC, tEPEC a

Attachment

Submitted filename: Revision PNTD-D-23-01115.docx

pntd.0012157.s017.docx (14.7KB, docx)
PLoS Negl Trop Dis. doi: 10.1371/journal.pntd.0012157.r003

Decision Letter 1

Stuart D Blacksell, Marilia Sá Carvalho

18 Apr 2024

Dear Dr. Grembi,

We are pleased to inform you that your manuscript 'Influence of hydrometeorological risk factors on child diarrhea and enteropathogens in rural Bangladesh' has been provisionally accepted for publication in PLOS Neglected Tropical Diseases.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. A member of our team will be in touch with a set of requests.

Please note that your manuscript will not be scheduled for publication until you have made the required changes, so a swift response is appreciated.

IMPORTANT: The editorial review process is now complete. PLOS will only permit corrections to spelling, formatting or significant scientific errors from this point onwards. Requests for major changes, or any which affect the scientific understanding of your work, will cause delays to the publication date of your manuscript.

Should you, your institution's press office or the journal office choose to press release your paper, you will automatically be opted out of early publication. We ask that you notify us now if you or your institution is planning to press release the article. All press must be co-ordinated with PLOS.

Thank you again for supporting Open Access publishing; we are looking forward to publishing your work in PLOS Neglected Tropical Diseases.

Best regards,

Marilia Sá Carvalho

Academic Editor

PLOS Neglected Tropical Diseases

Stuart Blacksell

Section Editor

PLOS Neglected Tropical Diseases

***********************************************************

PLoS Negl Trop Dis. doi: 10.1371/journal.pntd.0012157.r004

Acceptance letter

Stuart D Blacksell, Marilia Sá Carvalho

6 May 2024

Dear Dr. Grembi,

We are delighted to inform you that your manuscript, "Influence of hydrometeorological risk factors on child diarrhea and enteropathogens in rural Bangladesh," has been formally accepted for publication in PLOS Neglected Tropical Diseases.

We have now passed your article onto the PLOS Production Department who will complete the rest of the publication process. All authors will receive a confirmation email upon publication.

The corresponding author will soon be receiving a typeset proof for review, to ensure errors have not been introduced during production. Please review the PDF proof of your manuscript carefully, as this is the last chance to correct any scientific or type-setting errors. Please note that major changes, or those which affect the scientific understanding of the work, will likely cause delays to the publication date of your manuscript. Note: Proofs for Front Matter articles (Editorial, Viewpoint, Symposium, Review, etc...) are generated on a different schedule and may not be made available as quickly.

Soon after your final files are uploaded, the early version of your manuscript will be published online unless you opted out of this process. The date of the early version will be your article's publication date. The final article will be published to the same URL, and all versions of the paper will be accessible to readers.

Thank you again for supporting open-access publishing; we are looking forward to publishing your work in PLOS Neglected Tropical Diseases.

Best regards,

Shaden Kamhawi

co-Editor-in-Chief

PLOS Neglected Tropical Diseases

Paul Brindley

co-Editor-in-Chief

PLOS Neglected Tropical Diseases

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Supplementary Materials

    S1 Text. Appendices.

    Includes: A. Deviations from pre-analysis plan. Appendix B. Detection of enteropathogens in stool. Appendix C. Incubation periods for each pathogen outcome. Appendix D. Description of risk factors and data sources. Appendix E. Covariate sets tested for each risk factor.

    (PDF)

    pntd.0012157.s001.pdf (477.1KB, pdf)
    S1 Fig. Pairwise relationships between weather variables.

    Bivariate scatter plot of continuous climatic risk factors in the diarrhea cohort. Correlation ellipses depict the strength of the association on the basis of the Spearman rank correlation, color of the ellipse indicates the direction of the correlation, and the correlation coefficient is printed inside each ellipse.

    (PDF)

    pntd.0012157.s002.pdf (408.1KB, pdf)
    S2 Fig. Predicted enteropathogen prevalence by weekly average temperature (C) with different lags.

    All panels present adjusted models for children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

    (PDF)

    pntd.0012157.s003.pdf (444.2KB, pdf)
    S3 Fig. Predicted diarrhea prevalence by temperature minimum, mean and maximum.

    All panels present adjusted models for temperature with a 1-week lag period for diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

    (PDF)

    pntd.0012157.s004.pdf (254.4KB, pdf)
    S4 Fig. Predicted enteropathogen prevalence by temperature minimum.

    All panels present adjusted models for children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

    (PDF)

    pntd.0012157.s005.pdf (377.5KB, pdf)
    S5 Fig. Predicted enteropathogen prevalence by temperature maximum.

    All panels present adjusted models for children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

    (PDF)

    pntd.0012157.s006.pdf (357KB, pdf)
    S6 Fig. Diarrhea and enteropathogen prevalence by above- vs. below-cutoff for total weekly precipitation at 75th and 90th percentile cutoffs.

    All panels present adjusted models including an indicator for above 75th (58mm) or 90th (105mm) percentile average weekly precipitation as the independent variable; unadjusted models produced similar results. Error bars present 95% confidence intervals adjusted for clustering. The x-axis is on the log scale. Panel A) includes diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels B-D) include measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Closed circles in panels B-D indicate the expected most important lag based on enteropathogen-specific incubation times (Appendix C in S1 Text).

    (PDF)

    pntd.0012157.s007.pdf (335KB, pdf)
    S7 Fig. Diarrhea and enteropathogen prevalence by heavy rainfall.

    All panels present adjusted models including an indicator variable for heavy rainfall (total weekly precipitation > 80th (17mm) or 90th (29mm) percentile during the study period); unadjusted models produced similar results. Error bars present 95% confidence intervals adjusted for clustering. The x-axis is on the log scale. Panel A) includes diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels B-D) include measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Closed circles in panels B-D indicate the expected most important lag based on enteropathogen-specific incubation times (Appendix C in S1 Text).

    (PDF)

    pntd.0012157.s008.pdf (333.9KB, pdf)
    S8 Fig. Prevalence ratios for diarrhea and enteropathogen carriage and distance from study households to surface water.

    All panels present adjusted models; unadjusted models produced similar results. The independent variable was a categorical variable for tertiles of distance from each household to the nearest surface water (<165m; 165m to <316m, 3316m). Error bars present 95% confidence intervals adjusted for clustering. The x-axis is on the log scale. Panel A) includes diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels B-D) include measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial.

    (PDF)

    pntd.0012157.s009.pdf (361.3KB, pdf)
    S9 Fig. Prevalence ratios for diarrhea and enteropathogen carriage associated with the proportion of land around study households that contained surface water.

    All panels present adjusted models; unadjusted models produced similar results. The independent variable was an indicator for whether the proportion of pixels with surface water within 250m, 500m, 750m of each household was above or below the median. Error bars present 95% confidence intervals adjusted for clustering. The x-axis is on the log scale. Panel A) includes diarrhea measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels B-D) include measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial.

    (PDF)

    pntd.0012157.s010.pdf (339.2KB, pdf)
    S10 Fig. Predicted diarrhea and enteropathogen prevalence by vapor pressure deficit.

    All panels present adjusted models; shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Pathogen models included measurements in children approximately 14 months of age in the control, combined water + sanitation + handwashing (WASH), nutrition, and combined nutrition + WASH arms of the original trial. Diarrhea model includes measurements in children aged 6 months—5.5 years in the control arms in the original trial. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

    (PDF)

    pntd.0012157.s011.pdf (338.4KB, pdf)
    S11 Fig. Interaction between age category and hydrometeorological risk factors on childhood diarrhea prevalence.

    Data are from adjusted models and include measurements in children aged 6 months—5.5 years in the control arms in the original trial. Panels A) and B) Error bars present 95% confidence intervals adjusted for clustering and the x-axis is on the log scale. Panel A) shows an indicator variable for heavy rainfall (total weekly precipitation > 80th (17mm) or 90th (29mm) percentile during the study period or for above median (13mm), 75th (58mm) or 90th (105mm) percentile average weekly precipitation as the independent variable. Panel B) The independent variable was an indicator for whether the proportion of pixels with surface water within 250m, 500m, 750m of each household was above or below the median or a categorical variable for tertiles of distance from each household to the nearest surface water (<165m; 165m to <316m, ≥316m). Panel C) Shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details). No significant results were observed for other temperature variables or vapor pressure deficit.

    (PDF)

    pntd.0012157.s012.pdf (378.6KB, pdf)
    S12 Fig. Negative control analysis–bruising and temperature.

    All panels present adjusted models for weekly minimum, mean, or maximum temperature with the indicated lag period; shaded bands indicate simultaneous 95% confidence intervals accounting for clustering. Includes measurements in children aged 6 months—5.5 years in the control arms in the original trial. Prevalence estimates were predicted under conditions which held all adjustment covariates at fixed representative values (see Methods for details).

    (PDF)

    pntd.0012157.s013.pdf (258.2KB, pdf)
    S1 Table. Bacterial enteropathogen prevalence estimates at weekly average temperatures of 19°C and 30°C.

    (PDF)

    pntd.0012157.s014.pdf (202.4KB, pdf)
    S2 Table. Akaike Information Criterion (AIC) from adjusted models including temperature and precipitation variables and with or without interaction terms for diarrhea in the previous 7 days or detection of an enteropathogen.

    Blue boxes indicate where models without the interaction term had lower AIC while orange boxes highlight models where including the interaction term results in lower AIC. Bolded values emphasize the model with the lowest AIC for the given temperature variable.

    (XLSX)

    pntd.0012157.s015.xlsx (25.5KB, xlsx)
    S3 Table. Adjusted prevalence ratios for caregiver-reported child bruising in the prior 7 days.

    (PDF)

    pntd.0012157.s016.pdf (172.2KB, pdf)
    Attachment

    Submitted filename: Revision PNTD-D-23-01115.docx

    pntd.0012157.s017.docx (14.7KB, docx)
    Attachment

    Submitted filename: PNTD-D-23-01115_Grembi_ReviewerResponse.docx

    pntd.0012157.s018.docx (249.5KB, docx)

    Data Availability Statement

    Not all data can be shared publicly due to Personally Identifiable Information (household GPS coordinates). However, de-identified data is publicly available on the Open Science Framework (https://osf.io/yt67k/) and all code used for the analysis is available on GitHub at https://github.com/jgrembi/wbb-weather-diarrhea-enteropathogens. Analyses used R statistical software (version 4.2.0).

    Articles from PLOS Neglected Tropical Diseases are provided here courtesy of PLOS

    RESOURCES