https://orcid.org/0000-0003-2726-4930
Figures
Abstract
Billions of people globally are at risk for severe water scarcity periodically throughout the year. Moreover, intensifying climate change and increasingly unpredictable weather patterns will increase the severity and scope of those affected by household water insecurity, with negative consequences on household health and wellbeing. Faced with water insecurity, households shift to alternative water sources, which may be more expensive or of lower quality and contribute to both financial and health burdens. The extent to which household water insecurity varies throughout the year, however, remains understudied. Using a panel dataset of 2014 households across 40 villages in Matlab, Bangladesh, we test for seasonal variation in household water insecurity and its determinants using a modified HWISE scale. We find that while water insecurity in Matlab was low both pre- and post-monsoon, household water experiences vary throughout the year. Households report significantly lower water insecurity post-monsoon, compared to pre-monsoon, suggesting an annual measure of household water insecurity is insufficient to fully characterize intra-annual household water experiences. Comparing determinants of household water insecurity, we find that geography, household water use behaviors, and household characteristics are significantly related to experiences of household water insecurity but vary in their extent depending on the season. Our results demonstrate that physical location, seasonality, water quality, and household-level factors contribute to the dynamic nature of intra-annual household water insecurity. Knowing when and to what extent such determinants influence household water experiences throughout the year is essential for guiding and adapting engineering and policy design to reduce the costs and consequences of household water insecurity.
Citation: Broyles LMT, Pakhtigian EL, Aziz S, Akanda AS, Mejia A (2023) Seasonal variation in household water insecurity in rural Bangladesh: A longitudinal analysis. PLOS Water 2(7): e0000157. https://doi.org/10.1371/journal.pwat.0000157
Editor: Sher Muhammad, ICIMOD: International Centre for Integrated Mountain Development, NEPAL
Received: March 15, 2023; Accepted: July 12, 2023; Published: July 31, 2023
Copyright: © 2023 Broyles et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data and analysis files have been uploaded along with this submission.
Funding: This research was supported through funding from the National Aeronautics and Space Administration cooperative agreement number NNX17AD26A with Resources for the Future to estimate the VOI obtained from satellite-based remote sensing (authors ELP, SA, and ASA received this award) and through the Institutes for Energy and the Environment at Pennsylvania State University (authors ELP and AM received this award). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
1 Introduction
It is increasingly important to understand the impacts of seasonality on household water security for human health and well-being, particularly given expectations that global climate change will influence the timing and intensity of temperature and precipitation events. This is a widespread challenge as an estimated four billion people globally face severe water scarcity throughout the year, and the world’s most vulnerable populations are expected to suffer most severely from the impacts of climate change [1, 2]. Understanding the effects of climate change at the household level requires an examination of the environmental seasonal influences on household water adequacy, reliability, safety, and affordability. Further, understanding the timing of water insecurity, as well as its determinants, will help to improve the efficiency and sustainability of policies designed to improve household water access [3–5].
Household water insecurity (HWI) characterizes when a household is unable to ‘access and benefit from affordable, adequate, reliable, and safe water to sustain well-being and a healthy life’ [6]. A growing body of HWI studies recognizes intra-annual variation in household water security and examines seasonal variation in household water reliability, adequacy, safety, and access [7–12]. Yet, this literature is limited by its reliance on locally developed scales that do not allow for global comparability; omission of certain ecosystems and geographies that overlooks some environmental seasonality conditions; and use of cross-sectional datasets, which prevents longitudinal comparison over time [6, 13]. In this paper, we use a modified version of the Household Water Insecurity Experiences Scale (HWISE) [14], a scale computed using data from 28 sites across 23 low- and middle-income countries for measuring HWI, to examine the effects of seasonal variation on HWI in rural Bangladesh. Using a two-wave panel collected prior to and following the 2021 monsoon, we examine HWI before and after the summer monsoon in Matlab, Bangladesh. Bangladesh is a distinctive and important context in which to study HWI, and, particularly, its seasonal variation. The country’s low elevation, dense population, and limited infrastructure contribute to Bangladesh’s climate vulnerability; indeed, globally Bangladesh ranks seventh in terms of greatest climate change impact in the past two decades [1]. In particular, the distinct seasonal nature of the regional hydroclimate and the monsoon strongly influence the availability of water resources, access to safe water and sanitation, and exposure to waterborne diseases [15]. Many Bangladeshi households find it difficult to cope with the effects of climate change; for example, 19 million children across Bangladesh are at risk of negative effects of climate change on health and well-being [16].
This study makes three main contributions to the growing literature of seasonal effects on HWI. First, it provides evidence of seasonal variation in household experiences with water using longitudinal data [6, 13]. By visiting the same households at two different points in the same year, our analysis provides insight into the ways households’ experiences with HWI do and do not vary within a calendar year. Second, to our knowledge, this study is among the first to use the HWISE scale to examine seasonal differences in HWI in a tropical monsoon climate, where climate change is expected to be most severe [1]. Third, we examine HWI in the context of other water-related risks in the region—microbial contamination of surface water, arsenic contamination of groundwater, and flooding. While the existing literature has examined the health effects of drinking water contaminated with arsenic [17] and microbial contaminants [18], this study contributes an examination of how seasonality in perceived arsenic exposure and cholera incidence relate to HWI. Additionally, this study builds on the few longitudinal studies in the HWI literature that examine effects of seasonal flooding [19–21]. While examining these water-related risks, we control for household-level water use and demographic characteristics as households within the same climate and region may experience HWI differently due to things such as water collection time [22–24], ability to pay for drinking water [25], water source access and ownership [26, 27], and the ability to store water during periods of scarcity [24, 28].
2 Materials and methods
2.1 Ethics statement
Sampling and data collection for this study were conducted by local survey enumerators who were recruited and trained by the field implementation team at icddr,b. Enumerators obtained free and informed written consent from all participant households. This study obtained ethical approval through Moravian University IRB (21–003) and the Responsible Research Conduct Board at icddr,b. Data collection for this analysis took place within a broader research effort examining household water quality and water use behaviors in the Matlab area; details of this work are available in Pakhtigian et al. [33]. Additional information regarding the ethical, cultural, and scientific considerations specific to inclusivity in global research is included in the Supporting Information (S2 Text).
2.2 Study setting
Data for this study come from over 2000 households across 40 villages in the E, F, and G administrative blocks of Matlab, Bangladesh. These three administrative blocks are in the northern part of Matlab, with E being the eastern-most block; F central; and G western-most. Matlab is a rural part of the Chandpur district, approximately 50 km southeast of Bangladesh’s capital, Dhaka, and upstream from the confluence between the Padma and Meghna Rivers (see Fig A in S1 Text). The Dhonagoda River, a tributary of the Meghna, flows through the area. The region’s low elevation and proximity to the confluence of two major rivers puts households at risk of flooding. Both ground and surface water in Matlab have known health threats. Matlab has high, yet seasonally and geographically variable, naturally-occurring arsenic contamination in groundwater [29]. Microbial contamination of surface water also poses health risks; for example, cholera remains endemic to Matlab [18]. Environmental factors such as ambient and sea surface temperatures and seasonal flooding also contribute to seasonal cholera epidemics [30].
The region is characterized by a hot and dry spring (March through May), a wet summer monsoon (June through mid-October), and a cool and dry fall (late October through February) [31]. Most of the country’s annual rainfall—2000 mm in an average year—falls during the monsoon season. Local weather patterns, including temperature and rainfall variability, are largely influenced by El Niño Southern Oscillation [30]. Fig 1 plots average precipitation (Panel A) and temperature (Panel B) for the Chandpur district between 2001 and 2022, demonstrating the clear seasonality in both rainfall and temperatures in the area. The 2021 season—the year of data collection for this study—had temperature and precipitation patterns quite similar to these long term averages; annual rainfall was slightly below average and annual temperatures matched historic averages [32].
Precipitation (panel a) and temperature (panel b) in 2021 compared to climate averages in the Chandpur district. Data come from the National Bangladesh Drought Watch [32].
2.3 Sampling and data collection
Data collection took place pre-monsoon during the months of March-April 2021 and post-monsoon during the months of October-November 2021. Using a listing of each sample block of villages with at least 100 households, we used a stratified sampling frame to randomly select 40 villages for our sample. The final sample included 15 villages from block E, 15 villages from block G, and 10 villages from block F. To select households for the sample, we used a geographical systematic random sample to recruit approximately 50 households per sample village. To ensure geographical representation of each village, we recruited households in the following way. First, we obtained the number of households in the village (N). Next, enumerators selected a geographically central household. Finally, every kth household was recruited for the sample, where k = N/50. In this recruitment process, enumerators systematically visited every section of the village for sample recruitment. If a selected household was unable or unwilling to participate in the household survey, a neighboring household was approached instead. Households were recruited during the pre-monsoon season and remained enrolled in the study through the post-monsoon season.
The final sample consisted of 2142 households. Of these households, 2014 (95%) were re-surveyed post-monsoon; among re-surveyed households, the same respondent was re-interviewed in 98.4 percent of households (1981 out of 2014 households). Reasons for survey attrition include migration and the inability to find the intended respondent after multiple revisit attempts. Pre-monsoon, the total number of households from Block E was 824 (34.1%), from Block F was 546 (25.5%), and from Block G was 772 (36.1%). Post-monsoon, the total number of households from Block E was 769 (38.2%), from Block F was 513 (25.5%), and from Block G was 732 (36.3%). In addition to the pre- and post-monsoon household surveys, sample households from blocks E and G (n = 1596) were interviewed by phone every month between May and August. Households from Block F were omitted from monthly data collection due to resources constraints for data collection and the prioritization of households from blocks E and G to the design of the broader research effort from which these data come [33].
2.4 Household survey
As this was a within-subjects research design, the same household survey was used to collect household data pre- and post-monsoon. The survey was back translated for accuracy and pilot-tested in March 2021 with a small group of households (n = 67). The survey was administered in Bangla, the local language. Enumerators targeted an adult household member responsible for water collection and knowledgeable about household activities as the respondent. Additionally, due to the use of smartphone technology for broader research objectives [33], only households that owned a smartphone were eligible for the survey. A recent nationally representative survey found that 55 percent of Bangladeshi households have smartphones [34], suggesting that our sample contains households that are socioeconomically advantaged.
The household survey included sections on household cholera incidence, treatment costs, and knowledge; household water treatment and storage; sanitation and hygiene; routine water-related behaviors of household members; risk preferences; environmental concerns; known arsenic exposure; socioeconomic characteristics; and HWI—specifically the Household Water Insecurity Experiences Scale (HWISE) [14]. During piloting, we tested the 12 cross-country validated HWISE questions as well as two additional HWISE-related questions related to water demands associated with raising crops and livestock [14] (see Table A in S1 Text). After piloting the survey, we modified the HWISE-scale to reduce five questions where there was no within variation, or, in other words, where there were no respondents that indicated they ever experienced the HWI dimension raised in the question. These five questions were combined into three, such that the final pre-monsoon survey module consisted of 10 questions (see Table B in S1 Text). We combined the questions about handwashing and washing one’s body into one question about washing; questions about feeling angry and feeling ashamed/excluded/stigmatized into one question about feeling embarrassed about the water situation; and questions about going to sleep thirsty and having no useable or drinkable water into one question about having as much water to drink as they would like. While modifications to the HWISE scale are generally discouraged as they disallow for one-to-one comparisons across contexts in which it is implemented, other studies have also made modifications to better match their implementation contexts. For example, Rosinger et al. [35] and Stoler et al. [25] reduce the scale from twelve to eleven questions to allow for comparisons across a broader set of countries in their respective analyses. At endline, the post-monsoon survey module contained the same 10 questions from the pre-monsoon module in addition to two HWI questions, which were added to more carefully explore the influence of household water quality on HWI [36] (see Table B in S1 Text).
In addition to pre- and post-monsoon household surveys, monthly phone interviews were conducted with a subset of sample households in May, June, July, and August 2021. The short monthly survey collected data on household water treatment, health, water stress, and adequacy of available water in the previous month (see Table C in S1 Text). To reconnect with households and respondents throughout the study, identifiable data were collected as part of the household survey. These identifiers were used only for the purposes of data collection and were removed from the datasets used for analysis.
2.5 Statistical analysis
Descriptive statistics were determined for pre-monsoon and post-monsoon samples. Wilcoxon signed rank two-tailed test was used to assess the difference between HWISE score before and after the monsoon. Wilcoxon signed rank was chosen due to the within-subjects, repeated measurements nature of our design. Two-tailed Chi-Square tests were used to test seasonal differences between other categorical variables. Significance was determined at alpha values of 0.05 and 0.01, and marginal significance was determined at an alpha value of 0.10.
2.5.1 Outcomes.
Our main outcome was HWI, measured using our modified HWISE scale [14]. For the HWISE scale, respondents were asked how frequently in the last month they experienced an element of HWI such as insufficient water for cooking, drinking, or cleaning or stress or concern about sufficient water for the household. Using the responses to the 10 HWISE questions used in our household survey, we calculated a cumulative HWISE score on a scale from 0 (no HWI) to 30 (highest possible HWI). Additionally, following previous HWI studies [37], we constructed an alternative outcome—a binary indicator for any HWI that took a value of 1 for HWISE scores greater than zero, 0 otherwise.
2.5.2 Empirical framework.
Using multivariable regression, we examined how seasonality and other household characteristics affect HWI among households in Matlab. First, using our balanced, two-wave panel data collected pre- and post-monsoon, we estimated a linear regression with household fixed effects. This analysis tests for changes in HWI between seasons within households. Second, using each cross-sectional wave individually, we estimated linear regressions to estimate the relationships between key household water use behaviors and characteristics and experienced HWI. We used a linear regression model to allow for the inclusion of the whole range of HWISE score outcomes [36, 38, 39] and to facilitate the panel fixed-effects regression analysis. As a robustness check, we re-estimated our empirical specifications using logit regression models for our binary HWI outcome. As further model specification checks, we used tobit and Poisson regression models; these results, which are descriptively similar to our main specification, are available in S1 and S2 Tables.
Specifically, for our main specifications we estimate the following linear models
(1)
and
(2)
Eq (1) is the fixed-effects model and Eq (2) is the cross-sectional regression model (estimated individually for the pre-monsoon and post-monsoon waves). The outcome variable HWIit is the HWISE score for household i in wave t; wt is an indicator for the post-monsoon wave; νi is a household fixed effect; Xit is a vector of household water use and demographic characteristics of household i in wave t; and εit is the error term. In Eq (2), all components are defined in the same way; however, no time varying components are included in the analysis. We calculate robust standard errors for our main specifications. All statistical analyses were performed via Python using Jupyter Notebooks 6.4.12 and confirmed using Stata 17. Variables were assessed for collinearity, and we report statistical significance at alpha values of 0.05, 0.01, and 0.001 as well as marginal significance at an alpha value of 0.10. During the post-monsoon wave of data collection 128 households were not re-interviewed due to either household migration or the respondent not being home at the time of the interview (and multiple revisits). While these households were included for our pre-monsoon cross-sectional analysis, they were dropped from the panel analysis to allow for the construction of a balanced panel of households interviewed both before and after the monsoon.
Household water use and demographic characteristics include the following three categories. First, water-related risks—if a household member had cholera in the last month, if the nearest tubewell is known to have arsenic contamination, and if flooding is an environmental concern. Second, water use characteristics—if household water source is improved, if time to collect water is greater than 30 minutes, if the household pays for drinking water, if household has long-term water storage vessels, and if household treats water. Third, household characteristics including block location, household size, presence of children under 5, age, gender, and education of the household head, and an asset index constructed using principal component analysis. Table D in S1 Text contains further definitions and cited literature relevant to covariate choices. All time-varying covariates were included in the fixed effects model; time-varying and time invariant covariates were included in the cross-sectional models.
We report our main results in Section 3 of the manuscript. Full results of regression analyses are included in S1 and S2 Tables, which detail all estimated regression coefficients, standard errors, exact p-values, confidence intervals, and measures of goodness of fit.
3 Results
3.1 Descriptive results
We describe households in our sample both overall and within the different administrative blocks pre- and post-monsoon (Table 1). The descriptive statistics show similarities across the blocks in terms of the size and demographics of the household; the average household has between 4 and 5 members, and nearly 40% of households have young children. Further, the majority of households have male household heads who are just over 50 years old. Across the entire sample, approximately 40 percent of household heads have a secondary education level or higher. In terms of socioeconomic status (as measured by an asset index), we find that households in block G (the western-most block) are the most economically secure and households in block F (the central block) the least economically secure. Given some differences in household characteristics by block, we include block fixed effects and use other household-level controls in our empirical analysis.
3.1.1 Household water insecurity.
We first describe HWI pre- and post- monsoon in our sample as a whole and across the three administrative blocks (Fig 2, Table 2). Fig 2 was generated using ArcGIS with base map shapefiles from [40, 41]. Geographically, we found the villages with higher HWI (HWISE score > 0) were in block E, which is located in the northwestern part of Matlab. Despite some apparent geographical variation, average HWISE scores were low across all three blocks in both the pre- and post-monsoon seasons, indicating that most households in Matlab have affordable, reliable, adequate, and safe access to water.
Households that indicate any HWI (HWISE score > 0): Block E: 154 households (18.7%, num. = 154, denom. = 824) pre-monsoon; 50 households (6.5%, num. = 50, denom. = 769) post-monsoon. Block F: 69 (11.7%, num. = 69, denom. = 546) pre-monsoon; 29 (5.7%, num. = 29, denom. = 513) post-monsoon. Bock G: 86 (11.1%, num. = 86, denom. = 772) pre-monsoon; 53 (7.2%, num. = 53, denom. = 732) post-monsoon, where num. = numerator and denom. = denominator. Fig 2 generated using ArcGIS with base map shapefiles from [40, 41].
Table 2 reports descriptive statistics related to HWI and household water interactions, which show that average pre-monsoon HWI was higher than average post-monsoon HWI. Pre-monsoon, the average HWISE score was 0.32, with scores ranging from 0 to 10 and 85% of respondents reporting no water insecurity. Post-monsoon, the average HWISE score was 0.13, with scores ranging from 0–12 and 95% of respondents reported no water insecurity (see also Fig B in S1 Text for a histogram of these HWI values). This seasonal difference in mean HWI was statistically significant for the sample overall as well as in all block subsamples (p < 0.01), suggesting that there is a meaningful seasonal difference in HWI despite the low HWI observed in Matlab across seasons. This pattern of decreasing HWI between pre- and post-monsoon is further evidenced by data collected monthly during the monsoon season. Fig C in S1 Text depicts the percentage of households reporting sufficient water to meet their needs (Panel A) and percentage of households reporting stress related to their water supply (Panel B). Trends in these data suggest greater HWI experienced earlier compared to later in the monsoon season.
3.1.2 Water-related risks and household water use.
We also report descriptive statistics for seasonal differences in perceived water-related risks and household water use behaviors (Table 2). Across all survey respondents, we found that reported cholera incidence was 5.1 percentage points lower (p < 0.01) post-monsoon compared to pre-monsoon (Table 2). The seasonal difference held true across all blocks, as we found lower self-reported cholera incidence ranging from 3.5 percentage points (p < 0.05) to 7.3 percentage points (p < 0.01) between pre- and post-monsoon across the three administrative blocks. This finding corroborates recent trends of higher cholera incidence in the pre-monsoon compared to the post-monsoon months in Bangladesh [42].
We also found lower flooding concern post-monsoon (p < 0.01). Overall, 21.1% of households were concerned about flooding pre-monsoon, whereas only 12.3% were concerned with flooding post-monsoon. This reduction in flooding concern for Block G (4.1 percentage points) was lower than in Blocks E (11.4 percentage points) and F (12.2 percentage points). The smaller reduction in flooding concern in Block G suggests more consistent flood risk concerns among households in this part of Matlab throughout the year, perhaps given the block’s flood-prone location near the confluence between the Meghna and Dhonagoda rivers.
We found that fewer households treat their drinking water post-monsoon compared to pre-monsoon (p < 0.01). Overall, 15.6% of respondents treated their drinking water at home pre-monsoon, and 12.6% post-monsoon. Reductions are seen across all blocks, although the difference is only significant for households in Block F. These results suggest a perception of improved water quality across the study site in the post-monsoon season and may follow the general expectation of improved water quality in seasons of greater water availability [43]. Water treatment methods included boiling water, adding chlorination or purification tablets to the water, and using water filters. Where some households pursue multiple treatment options, the most common ways to improve drinking water safety at home include water filtration (pre-monsoon: 8.1%; post-monsoon: 7.8%) and chlorination (pre-monsoon: 5.6%; post-monsoon: 4.2%).
Finally, we found little seasonal variation in access to improved water (using the UNICEF/WHO definition [44]), time to collect household water, and long-term water storage. Improved water sources consisted of piped water, purchased water, rainwater, shared tubewells, privately owned tubewells, or motorized water—deep groundwater wells requiring a motorized pump. Tubewells were the most common improved water source, and the average time it took a household member to walk to the nearest tubewell pre-monsoon was just over 4 minutes, compared to just under 4 minutes post-monsoon. Long-term water storage consisted of rooftop water storage or large water storage containers, while short-term storage consisted of small clay pots or plastic buckets.
3.1.3 Descriptive statistics among water insecure households.
For households that experienced HWI, we found that the specific water-related concerns most frequently experienced were similar pre- and post-monsoon. During both seasons, households most frequently cited not having sufficient water for household needs and interruptions in main water supply as the biggest threats to water security. This phenomenon is observed across rural Bangladesh as both prolonged dry season-induced water scarcity and monsoon-induced flooding affect available water quantity and quality and, thus, affect water supply to vulnerable households [18]. Other important HWI dimensions cited by impacted households included feelings of embarassment related to the household water situation, insufficient drinking water, and negative impacts on crops (see Fig D in S1 Text for additional information related to frequency of HWI dimensions experienced by sample households).
Table 3 reports household water experiences by HWI dimension experienced and season. We find that most households that experinced some level of HWI used improved water sources; the majority of these households relied on boreholes. This finding suggests that although improved water sources are widespread in Matlab, they may not always provide safe and reliable access to water—that is, households that use improved water sources may still experience water insecurity. For example, in Bangladesh, tubewell service may be interrupted or inaccessible at certain times of the year, especially pre-monsoon when lower groundwater tables render many tubewells ineffecitve and water demand is greater at functioning wells [27].
We also found that, compared to the overall sample, fewer water insecure households were able to collect household water within 30 minutes (Table 3). Among households experiencing HWI, less than two thirds could access daily water in less than 30 minutes pre-monsoon and approximately 81 percent could access daily water in less than 30 minutes post-monsoon. In the entire household sample, just over three quarters could collected household water within 30 minutes pre-monsoon. This comparison is particularly aparent for households that identified insufficient drinking water as a dimension of HWI; for these households, only 42 percent and 52 percent could collect household water within 30 minutes pre- and post-monsoon, repectively. The general trend of higher water collection time burdens among water insecure households may indicate that time to collect household water may be an important predictor of HWI in Matlab.
3.2 Regression results
Table 4 presents our regression results including our fixed effects analysis (Model 1) and our cross- sectional, seasonal analyses (Models 2 and 3). Each of these models was estimated using linear regression, and estimates indicate the change in outcome associated with a one unit change in the predictor variable. Turning first to our estimation of seasonality using our panel regression, we found that HWI was lower post-monsoon compared to pre-monsoon; after controlling for water-related risks, household water use behaviors, and household characteristics, we found the average HWISE score was approximately -0.18 points lower post-monsoon (p < 0.001). This could be explained by hotter temperatures pre-monsoon, which have been found to contribute to higher water demand and dehydration in other locations [45]. Additionally, due to limited pre-monsoon rainfall, households may run out of stored rainwater or may need to seek water sources that are farther away to meet their water needs [23, 27, 46].
Time to collect water was significantly associated with higher HWI. Our panel regression shows that HWISE scores are, on average, 0.25 higher for households that take more than 30 minutes to collect daily household water (p < 0.001). This result appears to be driven by pre-monsoon differences: In the separate pre- and post-monsoon cross-sectional analyses, we found that the average HWISE score is 0.36 higher pre-monsoon among households that needed more than 30 minutes to collect water (p < 0.001) and only 0.045 higher in the post-monsoon season, although the latter estimate is not statistically distinguishable from zero. This result can be attributed to the fact that surface water sources often dry up across rural Bangladesh in the prolonged heat of the pre-monsoon season and many tubewells become dry, forcing villagers to travel greater distances in search of available water. These results also support other literature that has found decreased satisfaction with water sources farther from the home [23].
We also find seasonality in the way payment for drinking water influences HWI. Post-monsoon, households that pay for drinking water have average HWISE scores that are 0.078 lower than households that do not pay for drinking water (p < 0.05). This could suggest that households that pay for drinking water may be able to acquire higher quality or quantity of water compared to households that do not. Additionally payment for water may be an effective coping strategy that decreases prevalence of experiences measured by the HWISE scale. We only see this in the post-monsoon model, which could be reflective of fewer pre-monsoon water sources.
Finally, we found geographical differences in HWI pre-monsoon. In the pre-monsoon model, average HWISE scores are 0.14 lower in Block F compared to those in in Block E (p < 0.05); for Block G, 0.20 lower (p < 0.001) (Table 4). The post-monsoon model, however, showed little difference between blocks, none of which were significant. These regression models confirm our descriptive findings and also suggest the factors which contribute to higher HWI pre-monsoon may be more severe for the households in Block E compared to those in Blocks F and G. For example, households in Block E may experience drier conditions in the months leading up to the monsoon, such as insufficient surface water sources and lower groundwater tables, due to being located farther from the Meghna River than those in Blocks F and G.
3.2.1 Post-monsoon HWISE water quality analysis.
One of the prevailing challenges in assessing HWI is disentangling water security issues related to water quality and water quantity. In Matlab, we found that water quality presents a greater threat to HWI than does water quantity. Accordingly, during the post-monsoon data collection, we incorporated additional water quality questions based on the work of Rosinger et al. [36]. We added the responses related to water quality concerns to our cumulative HWISE score to construct a new measure of HWI. Table 5 reports the results of this analysis, which are consistent for models showing robust standard errors (Model 1) and standard errors clustered at the village level (Model 2). Overall, the results are consistent with our primary findings.
3.3 Robustness checks
We conducted two robustness checks of our regression results. First, we re-estimated the pre- and post-monsoon cross-sectional linear regression models with standard errors clustered at the village level in place of robust standard errors. Second, we re-estimated all models using logit regression with a binary any HWI outcome with both robust and clustered standard errors.
3.3.1 Alternative standard errors: Village clustering.
Given the sampling strategy which was based on village-level units, clustered standard errors present an alternative option to the robust standard errors presented with our primary analysis. To assess whether our results are sensitive to our standard error calculations, we re-estimated our cross-sectional linear regression models with standard errors clustered at the village level. We report these results in Table E in S1 Text. Overall, we found these estimates to be substantively similar to our primary specification; if anything, our estimates using clustered standard errors are more precise, making our primary model results conservative estimates. In particular, we find that during the post monsoon season, payment for drinking water is associated with lower HWI.
3.3.2 Logit specification.
In cases with limited HWI—such as our context in Matlab—researchers also characterize HWI using a binary outcome and model its determinants using logit regression [37]. Accordingly, as a robustness check, we used logit regression models to replicate our main results (Table F in S1 Text), our results with alternative standard errors clustered at the village level (Table G in S1 Text), and our analysis using the water quality adjusted HWISE outcome for the post-monsoon period (Table H in S1 Text). For each of these analyses, we report the results as odds ratios.
Comparing our linear and logit models, we find consistent trends, yet somewhat more precise estimates with our logit specifications. First, we found that households that expressed flooding concerns experienced lower HWI both in the fixed-effects and pre-monsoon models. This provides further evidence that water insecurity in Matlab is driven by concerns about water quality rather than water quantity. Second, we found that, pre-monsoon, households with more highly educated household heads experienced lower HWI and, in our panel regression, households with older household heads experienced higher HWI. These results suggest that, on the extensive margin, household characteristics may factor into household water experiences. Finally, we found that HWI is increasing in water collection time (found across the fixed-effect and pre-monsoon models) and drinking water payment (found during the pre-monsoon season). The substantive consistency between our primary results using the linear regression analysis and our robustness checks using logit regression suggests that our results are not overly sensitive to our choice of modeling framework.
4 Discussion
Using a two-wave panel collected pre- and post-monsoon in Matlab, Bangladesh, we found that HWI changes throughout the year, with households experiencing greater water insecurity pre-monsoon compared to post-monsoon. We observed this descriptively by demonstrating that households scored lower on the HWISE scale post-monsoon and by documenting decreased instances of water stress and concern across monthly phone-call during monsoon months. Using cross-sectional and panel multivariable regression models, we examined intra-annual changes in HWI and its determinants. We found that, after controlling for household characteristics and water use behaviors and risks, HWI was higher during the pre-monsoon season. Further, we found that water collection time and payment for drinking water are important determinants of household experiences with water insecurity.
The dynamic nature of HWI across the monsoon season is further characterized by comparing determinants of HWI pre- and post-monsoon. For example, even within a geographically constrained area such as Matlab, there are geographical differences in HWI. Pre-monsoon, households in the northern part of Matlab experienced higher water insecurity than did households farther south. Households in the north live further from the Meghna River, which runs through the Matlab region; accordingly, these households may perceive greater water stress due to their distance from surface water sources—especially during the pre-monsoon season when rainfall is scarce. Drier and hotter pre-monsoon conditions may contribute to water sources farther from the main river drying up before monsoon rains replenish surface and groundwater storages. We also find more pronounced HWI among households whose water sources were farther away, and that water insecurity was driven by worries of sufficient water and supply interruption. This phenomenon can be seen across Bangladesh, where northern and western regions receive less rainfall on average and have fewer water resources—and, thus, experience higher levels of water insecurity—than do the more rain-prone and riverine eastern and southern parts of the country [47]. Taken together, these results suggest that households experience more insecurity in water supply pre-monsoon, especially those households located further away from key water sources in the area.
We also find that perceptions of water insecurity do not always align with experienced water-related risks. Although flooding, cholera incidence, and arsenic exposure are known water-related risks in rural Bangladesh, we find little evidence of relationships between these three variables and HWI. We do find a statistically significant difference in flooding concern and household cholera incidence pre- and post-monsoon, with households expressing greater concern of these water-related risks pre-monsoon. Yet, we do not find evidence of relationships between these water-related risks and experienced HWI in our main regression specification. This may indicate that households in Matlab are well equipped to cope with certain water-related risks such as seasonal flooding. Thus, a recognized threat of flooding does not relate to their experiences with water insecurity. Our study was also conducted during a typical year in terms of temperature and precipitation. To fully gauge the impact of flooding on HWI, future work should examine seasonality in HWI in years with above and below average monsoon flooding. Future studies could also use more objective measures of flooding, such as flooding incidence as defined by exposure to floods in past year or flood risk as defined by proximity to river or vulnerability to flash floods or tidal surges.
With regard to water-related health risks, it is possible that our imprecise findings related to cholera and arsenic risks reflect recent improvements in water quality in the Matlab area. For example, recent evidence shows decreased cholera transmission in Matlab [48]. There are many other parts of Bangladesh and regions globally, however, that experience high rates of cholera transmission and other water-related diseases [18, 49]. Further examining seasonal HWI in the context of water-related health risks, contamination, and safe drinking water interventions are other important areas of future study [24]. For example, western Bangladesh experiences substantially higher water scarcity during the pre-monsoon season and exhibits cholera transmission during this season only; these areas, thus, may reflect much different perceptions of water-related health risks and associated water insecurity.
Similarly with regard to arsenic exposure, while there is well-known arsenic contamination in Matlab [29], it is possible that households have adjusted to this risk in their water-use behavior or that, due to aging infrastructure, knowledge of tubewell-specific contamination may be limited [50]. Given these limitations to our proxy for arsenic contamination, future work should include water quality testing for arsenic and other contaminants to more fully characterize both environmental risk and objective HWI. This is important given the health consequences of arsenic contamination in drinking water and seasonal variability of arsenic in groundwater sources. For each of these water-related risks—flooding, arsenic, and cholera—households in Matlab may have developed coping and adaptation strategies, which are not well captured by the current design of the HWISE scale. Indeed, in light of climatic change, Stoler et al. [51] argue that adaptation and resiliency to HWI should be incorporated into future HWI measurement. Our findings suggest that further evaluation of coping and mitigation strategies could provide a more holistic understanding of the water-related challenges households face.
Finally, the addition of the water quality questions to the HWISE scale that we implemented in the post-monsoon data collection and subsequent analysis suggests a potential methodological extension for the HWISE scale. Inclusion of these questions may be important and relevant to certain contexts, especially those investigating environmental seasonal effects on HWI as water quality has been shown to vary seasonally and influence household experiences with water and well-being differentially depending on the season [43]. These inclusions may also help in characterizing both the intensive and extensive margins of HWI in future research.
While our analysis sheds light on the ways in which seasonality impacts household water experiences, there are several limitations to our work. First, as the inclusion criteria for the study required that sample households own a smartphone, our sample was relatively socioeconomically advantaged compared to the average household in Matlab and, indeed, in Bangladesh. Accordingly, some of our findings, including lower HWI among our sample than has been found in other sites in Bangladesh [52, 53] could reflect this sampling reality. Nevertheless, our findings reflect experiences with water insecurity among our sample and provide key insights into the intra-year variation in HWI in Matlab. Second, as has been raised in the HWI literature [54], water insecurity scales implemented in the field have not been robustly validated for test-retest reliability. While such tests were beyond the scope of this work, we acknowledge the importance of testing and validating the HWISE, and other scales of HWI, in the field and suggest that this is an important direction for future research. Finally, after piloting our household survey, we modified the HWISE for the local context by reducing questions where all respondents indicated they never experienced a particular dimension of HWI that was included in HWISE. While modifications to the HWISE scale are generally discouraged as they disallow for one-to-one comparisons across contexts in which it is implemented, other studies have also made modifications to better match their implementation contexts. Notwithstanding, however, our results may not be directly comparable to other HWISE studies.
5 Conclusions
Our study demonstrates that there is important intra-annual variation in HWI that annual and cross-sectional analyses of HWI will miss. This dynamic understanding of HWI is important to design technological and policy solutions to the critical issue of HWI. These dynamics are all the more important from a policy perspective given that water scarcity and unpredictability is expected to increase with climate change. We also examined intersections of HWI with household water use characteristics and water-related risks, finding that geography, water source type, water collection time, household payment for drinking water, and flooding concern are significant predictors of HWI in Matlab, Bangladesh. This provides important implications for engineering and policy as it elucidates the importance of timing for the allocation of resources to improve household water access and increase wellbeing. This study provides evidence related to intra-annual household water needs, which is important for guiding where and when adjustments are needed in water provisioning, treatment, and infrastructure.
Supporting information
S1 Text. Figs A-D, Tables A-H. Fig A.
Map of research study site of Matlab, Bangladesh. Map generated using ArcGIS Pro with base map shapefiles from National Geographic Style (available at https://www.arcgis.com/sharing/rest/content/items/3d1a30626bbc46c582f148b9252676ce/resources/styles/root.json and https://tiles.arcgis.com/tiles/P3ePLMYs2RVChkJx/arcgis/rest/services/NatGeoStyleBase/MapServer), World Ocean Base (available at https://services.arcgisonline.com/arcgis/rest/services/Ocean/World_Ocean_Base/MapServer) and World Street Map (available at https://cdn.arcgis.com/sharing/rest/content/items/de26a3cf4cc9451298ea173c4b324736/resources/styles/root.json). Fig B. Histogram of HWISE scores collected during the pre- and post-monsoon seasons. Fig C. Panel (A) Percentage of households that felt they had sufficient water to meet household needs in the past month. Panel (B) Percentage of households that felt unable to deal with stress related to drinking water quality in past month. For both (A) and (B), the numerators for percentage calculations were the number of respondents who responded affirmatively and the denominator is the total number of respondents who received phone-call follow-up surveys (n = 1596). Fig D. Dimensions of HWI identified by households in Matlab, Bangladesh. Table A. HWISE questions included in pilot study. Table B. HWISE questions included in final survey instrument. Table C. Questions included in monthly phone interviews. Table D. Independent variables included in regression analysis. Table E. Linear regression results with standard errors clustered at the village level. Table F. Logit regression results for binary any HWI outcome. Table G. Logit regression results for standard error robustness check. Table H. Logit regression results for post-monsoon robustness checks.
https://doi.org/10.1371/journal.pwat.0000157.s001
(DOCX)
S2 Text. Inclusivity in global research checklist.
https://doi.org/10.1371/journal.pwat.0000157.s002
(DOCX)
S3 Text. Do file for (Stata) analysis that includes all code for generating tables in main text and supplementary analysis.
https://doi.org/10.1371/journal.pwat.0000157.s003
(DO)
S7 Data. Data for generating Fig B in S1 Text.
https://doi.org/10.1371/journal.pwat.0000157.s010
(XLSX)
S8 Data. Data for generating Fig C in S1 Text.
https://doi.org/10.1371/journal.pwat.0000157.s011
(XLSX)
S9 Data. Data for generating Fig D in S1 Text.
https://doi.org/10.1371/journal.pwat.0000157.s012
(XLSX)
S1 Table. Poisson and logit regression results.
https://doi.org/10.1371/journal.pwat.0000157.s013
(XLSX)
Acknowledgments
We are grateful to our field team at icddr,b—especially M.A. Hanifi, Ammatul Fardousi, Srizan Chowdhury, and Mehedi Hasan—for expert data collection. Participants at the 8th Annual Environmental Politics and Governance Conference provided valuable comments. Any remaining errors are our own.
References
- 1.
Eckstein D, Kunzel V, Schafer L. Global Climate Risk Index 2021: Who suffers most from extreme weather events? Weather-related loss events in 2019 and 200–2019 [Internet]. Berlin: Germanwatch; 2021. Available from: https://germanwatch.org/sites/default/files/Global%20Climate%20Risk%20Index%202021_2.pdf
- 2. Mekonnen MM, Hoekstra AY. Four billion people facing severe water scarcity. Sci Adv. 2016 Feb 5;2(2):e1500323. pmid:26933676
- 3. Di Baldassarre G, Sivapalan M, Rusca M, Cudennec C, Garcia M, Kreibich H, et al. Sociohydrology: scientific challenges in addressing the sustainable development goals. Water Resour Res. 2019;55(8):6327–55. pmid:32742038
- 4. Rhodes-Dicker L, Brown NJ, Currell M. Unpacking intersecting complexities for WASH in challenging contexts: A review. WAter Res. 2022;209:117909.
- 5. Douville H, Allan RP, Arias PA, Betts RA, Caretta MA, Cherchi A, et al. Water remains a blind spot in climate change policies. Pradhanang SM, editor. PLOS Water. 2022 Dec 15;1(12):e0000058.
- 6. Jepson W, Wutich A, Colllins SM, Boateng GO, Young SL. Progress in household water insecurity metrics: a cross-disciplinary approach: Progress in household water insecurity metrics. Wiley Interdiscip Rev Water. 2017 May;4(3):e1214.
- 7. Hadley C, Wutich A. Experience-based Measures of Food and Water Security: Biocultural Approaches to Grounded Measures of Insecurity. Hum Organ. 2009;68(4):451–60.
- 8. Mason LR. Gender and Asset Dimensions of Seasonal Water Insecurity in Urban Philippines. Weather Clim Soc. 2012 Jan 1;4(1):20–33.
- 9. Mason LR. Beyond Improved Access: Seasonal and Multidimensional Water Security in Urban Philippines. Glob Soc Welf. 2015 Sep;2(3):119–28.
- 10. Mason LR, Agan TC. Weather variability in urban Philippines: a gender analysis of household impacts. Clim Change. 2015 Oct;132(4):589–99.
- 11. Wutich A. Intrahousehold disparities in women and men’s experiences of water insecurity and emotional distress in urban Bolivia. Med Anthropol Q. 2009;23(4):436–54. pmid:20092053
- 12. Wutich A, Ragsdale K. Water insecurity and emotional distress: Coping with supply, access, and seasonal variability of water in a Bolivian squatter settlement. Soc Sci. 2008;10.
- 13. Broyles LMT, Pakhtigian EL, Rosinger AY, Mejia A. Climate and hydrological seasonal effects on household water insecurity: A systematic review. WIREs Water [Internet]. 2022 Apr 5 [cited 2022 Apr 5]; Available from: https://onlinelibrary.wiley.com/doi/10.1002/wat2.1593
- 14. Young SL, Boateng GO, Jamaluddine Z, Miller JD, Frongillo EA, Neilands TB, et al. The Household Water InSecurity Experiences (HWISE) Scale: development and validation of a household water insecurity measure for low-income and middle-income countries. BMJ Glob Health. 2019 Sep;4(5):e001750. pmid:31637027
- 15. Akanda AS, Jutla AS, Alam M, De Magny GC, Siddique AK, Sack, R. B., et al. Hydroclimatic influences on seasonal and spatial cholera transmission cycles: implications for public health intervention in the Bengal Delta. Water Resour Res. 2011;47(3).
- 16.
Ingram S. A gathering storm: Climate change clouds the future of children in Bangladesh [Internet]. Geneva: UNICEF; 2019. Available from: https://www.unicef.org/rosa/media/3001/file/Report%20ClimateChange-Embargoed%20-%20UNICEF.pdf
- 17. Abedin MA, Collins AE, Habiba U, Shaw R. Climate Change, Water Scarcity, and Health Adaptation in Southwestern Coastal Bangladesh. Int J Disaster Risk Sci. 2019;10(1):28–42.
- 18. Akanda AS, Jutla AS, Gute DM, Sack RB, Alam M, Huq A, et al. Population Vulnerability to Biannual Cholera Outbreaks and Associated Macro-Scale Drivers in the Bengal Delta. Am J Trop Med Hyg. 2013 Nov 6;89(5):950–9. pmid:24019441
- 19. Few R, Lake I, Hunter PR, Tran PG. Seasonality, disease and behavior: Using multiple methods to explore socio-environmental health risks in the Mekong Delta. Soc Sci Med. 2013;80:1–9. pmid:23415585
- 20. Mondal NC, Biswas R, Manna A. Risk factors of diarrhoea among flood victims: a controlled epidemiological study. Indian J Public Health. 2001 Dec;45(4):122–7. pmid:11917333
- 21. Tubatsi G, Bonyongo MC, Gondwe M. Water use practices, water quality, and households’ diarrheal encounters in communities along the Boro-Thamalakane-Boteti river system, Northern Botswana. J Health Popul Nutr. 2015 Dec;33(1):21. pmid:26825320
- 22. Mushavi RC, Burns BFO, Kakuhikire B, Owembabazi M, Vořechovská D, McDonough AQ, et al. “When you have no water, it means you have no peace”: A mixed-methods, whole-population study of water insecurity and depression in rural Uganda. Soc Sci Med. 2020 Jan;245:112561. pmid:31790879
- 23. Victor C, Ocasio DV, Cumbe ZA, Garn JV, Hubbard S, Mangamela M, et al. Spatial Heterogeneity of neighborhood-level water and sanitation access in informal urban settlements: A cross-sectional case study in Beira, Mozambique. PLOS Water. 2022 Jan 27;1(6):e0000022. pmid:36258753
- 24. Nounkeu CD, Metapi YD, Ouabo FK, Kamguem AST, Nono B, Azza N, et al. Assessment of drinking water access and household water insecurity: A cross sectional study in three rural communities of the Menoua division, West Cameroon. Hubbart J, editor. PLOS Water. 2022 Aug 18;1(8):e0000029.
- 25. Stoler J, Pearson AL, Staddon C, Wutich A, Mack E, Brewis A, et al. Cash water expenditures are associated with household water insecurity, food insecurity, and perceived stress in study sites across 20 low- and middle-income countries. Sci Total Environ. 2020 May;716:135881. pmid:31874751
- 26. Boateng GO, Collins SM, Mbullo P, Wekesa P, Onono M, Neilands TB, et al. A novel household water insecurity scale: Procedures and psychometric analysis among postpartum women in western Kenya. Gnambs T, editor. PLOS ONE. 2018 Jun 8;13(6):e0198591. pmid:29883462
- 27. Hoque SF. Socio-spatial and seasonal dynamics of small, private water service providers in Khulna district, Bangladesh. Int J Water Resour Dev. 2021 Aug 6;1–24.
- 28. Eakin H, Lerner AM, Manuel-Navarrete D, Hernández Aguilar B, Martínez-Canedo A, Tellman B, et al. Adapting to risk and perpetuating poverty: Household’s strategies for managing flood risk and water scarcity in Mexico City. Environ Sci Policy. 2016 Dec;66:324–33.
- 29. Bhattacharya P, Hossain M, Rahman SN, Robinson C, Nath B, Rahman M, et al. Temporal and seasonal variability of arsenic in drinking water wells in Matlab, southeastern Bangladesh: A preliminary evaluation on the basis of a 4 year study. J Environ Sci Health Part A. 2011 Sep;46(11):1177–84. pmid:21879850
- 30. Islam MS, Sharker MAY, Rheman S, Hossain S, Mahmud ZH, Islam MS, et al. Effects of local climate variability on transmission dynamics of cholera in Matlab, Bangladesh. Trans R Soc Trop Med Hyg. 2009 Nov;103(11):1165–70. pmid:19477477
- 31. Bangladesh Meteorological Department (BMD). Seasonal Climate Analysis [Internet]. Bangladesh Meteorlogical Department. 2017 [cited 2022 Mar 14]. Available from: http://live4.bmd.gov.bd//
- 32.
ICIMOD. National agricultural drought watch-Bangladesh [Internet]. 2022. Available from: http://tethys.icimod.org/apps/droughtbd/?d=l2Dhaka&p=mm&i=rain,evap,soilMoist,tempExtreme
- 33. Pakhtigian EL, Aziz S, Boyle K, Akanda AS, Hanifi MA. Early Warning Systems, Mobile Technology, and Cholera Aversion: Evidence from Rural Bangladesh. Work Pap. 2022.
- 34.
Alliance for Affordable Internet and Access to Information Programme (AfAI and A2I). Bangladesh National ICT Household Survey: Report and Overview of Main Results. 2018.
- 35. Rosinger AY, Brewis A, Wutich A, Jepson W, Staddon C, Stoler J, et al. Water borrowing is consistently practiced globally and is associated with water-related system failures across diverse environments. Glob Environ Change. 2020 Sep;64:102148. pmid:33071475
- 36. Rosinger AY, Bethancourt HJ, Young SL, Schultz AF. The embodiment of water insecurity: Injuries and chronic stress in lowland Bolivia. Soc Sci Med. 2021;291:11490. pmid:34662760
- 37. Rosinger AY. Household water insecurity after a historic flood: Diarrhea and dehydration in the Bolivian Amazon. Soc Sci Med. 2018 Jan;197:192–202. pmid:29253721
- 38. Bethancourt HJ, Swanson ZS, Nzunza R, Young SL, Lomeiku L, Douglass MJ, et al. The co-occurrence of water insecurity and food insecurity among Daasanach pastoralists in northern Kenya. Public Health Nutr. 2023 Mar;26(3):693–703.
- 39. Libey A, Kebede A, Ibrahim J, Hutchings P, Mekonta L, Butterworth J, et al. Surveyed from Afar: Household water security, emotional well-being, and the reliability of water supply in the Ethiopian lowlands. Int J Hyg Environ Health. 2022;246(114059). pmid:36323173
- 40.
Esri, Garmin, GEBCO, NOAA NGDC, and other contributors. Ocean/World_Ocean_Base (Map Server) [Internet]. ArcGIS REST Services Directory. Available from: https://services.arcgisonline.com/arcgis/rest/services/Ocean/World_Ocean_Base/MapServer
- 41. OCHA. Bangladesh-Subnational Administrative Boundaries [Internet]. The Humanitarian Data Exchange. 2020. Available from: https://data.humdata.org/dataset/cod-ab-bgd?
- 42. Daisy SS, Saiful Islam AK, Akanda AS, Faruque ASG, Amin N, Jensen PKM. Developing a forecasting model for cholera incidence in Dhaka megacity through time series climate data. J Water Health. 2020;18(2):207–23. pmid:32300093
- 43. Collins SM, Mbullo Owuor P, Miller JD, Boateng GO, Wekesa P, Onono M, et al. ‘I know how stressful it is to lack water!’ Exploring the lived experiences of household water insecurity among pregnant and postpartum women in western Kenya. Glob Public Health. 2019 May 4;14(5):649–62.
- 44.
UNICEF and WHO. Progress on household drinking water, sanitation and hygiene: 2000–2017: Special Focus on Inequalities. New York: UNICEF and WHO; 2019.
- 45. Rosinger AY, Bethancourt HJ, Swanson ZS, Lopez K, Kenney WL, Huanca T, et al. Cross-cultural variation in thirst perception in hot-humid and hot-arid environments: Evidence from two small-scale populations. Am J Hum Biol. 2021;e23715. pmid:34942040
- 46. Haque SS, Yanez-Pagans M, Arias-Granada Y, Joseph G. Water and sanitation in Dhaka slums: access, quality, and informality in service provision. Water Int. 2020 Nov 16;45(7–8):791–811.
- 47. Ahammed SJ, Chung ES, Shahid S. Parametric assessment of pre-monsoon agricultural water scarcity in Bangladesh. Sustainability. 2018;10(3):819.
- 48. Zaman K, Ryun Kim D, Ali M, Ahmmed F, Im J, Islam MT, et al. Can cholera ‘hotspots’ be converted to cholera ‘coldspots’ in cholera endemic countries? The Matlab, Bangladesh experience. Int J Infect Dis. 2020 Jun;95:28–31. pmid:32126325
- 49. Ali M, Nelson AR, Lopez AL, Sack DA. Updated Global Burden of Cholera in Endemic Countries. Remais JV, editor. PLoS Negl Trop Dis. 2015 Jun 4;9(6):e0003832. pmid:26043000
- 50. Fischer A, Hope R, Manandhar A, Hoque S, Foster T, Hakim A, et al. Risky responsibilities for rural drinking water institutions: The case of unregulated self-supply in Bangladesh. Glob Environ Change. 2020 Nov;65:102152.
- 51. Stoler J, Jepson WE, Brewis A, Wutich A. Frontiers of household water insecurity metrics: severity, adaptation and resilience. BMJ Glob Health. 2023;8(5):e011756. pmid:37137537
- 52. Stoler J, Miller JD, Adams EA, Ahmed F, Alexander M, Asiki G, et al. The Household Water Insecurity Experiences (HWISE) Scale: comparison scores from 27 sites in 22 countries. J Water Sanit Hyg Dev. 2021 Nov 1;11(6):1102–10.
- 53. Young SL, Bethancourt HJ, Ritter ZR, Frongillo EA. Estimating national, demographic, and socioeconomic disparities in water insecurity experiences in low-income and middle-income countries in 2020–21: a cross-sectional, observational study using nationally representative survey data. Lancet Planet Health. 2022 Nov;6(11):e880–91. pmid:36370726
- 54. Sidote MN, Goodman ZT, Paraggio CL, Tutu RA, Stoler J. Measurement invariance of a household water insecurity metric in Greater Accra, Ghana: Implications for test-retest reliability. Int J Hyg Environ Health. 2022;240(113922). pmid:35026673