Study sites

A stakeholder meeting, which convened the research partners, local and international NGOs involved in WASH, and government officials responsible for WASH decision-making in Dhaka, was held in Dhaka on February 14, 2017 to inform the selection of study neighborhoods and to ensure the study’s relevancy to local WASH and health sector programming. Neighborhoods were selected based on socioeconomic status (low-income, high-income), housing conditions and stability (structured slum vs. unstructured slums, floating/transient vs. permanent populations), WASH infrastructure and services (unimproved WASH vs. improved WASH), and geographic location (Dhaka North City Corporation [DNCC], Dhaka South City Corporation [DSCC]) to help understand variation in risks of fecal exposure across neighborhood types (Table 1). Definitions of the neighborhood categories are provided in S1 Table. We utilised a modified JMP definition of improved and unimproved WASH conditions [25]. Unimproved WASH refers to the absence or inadequacy of safe water on premises (as seen in floating or low-income communities), lack of sanitation with proper fecal sludge management, and limited hygiene-often characterised by open drains and shared or non-functional latrines. In contrast, improved WASH is defined as conditions where communities have piped water supply on premises and toilets connected to a sewer or covered drainage network (i.e., high-income communities). Further detail on the neighborhood selection process and characteristics can be found in the previous publication by Amin et al. (2019).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Overview of behavioral surveys conducted in households, communities, and schools, Dhaka 2017.

https://doi.org/10.1371/journal.pone.0319883.t001

Data collection

From 25/04/2017 to 30/07/2017 (pre-monsoon to monsoon season), the SaniPath Exposure Assessment Tool [18,19] was used to conduct multi-pathway exposure assessment in the ten selected neighborhoods in Dhaka. Two separate teams of trained field staff conducted the data collection activities. Key informant interviews, mapping and environmental samples were collected by a team of fieldworkers from International Center for Diarrheal Disease Research, Bangladesh (icddr,b), led by the lead author (NA). Behavioral surveys were conducted by a local NGO, DATA, under the supervision of the Emory University team (SR). All data were collected using mobile phones programmed with ODK (Open Data Kit). The environmental sampling forms were administered in English, while the behavioral survey questionnaires were developed in both English and Bengali to ensure cultural and linguistic appropriateness. Both teams received standardized training conducted jointly by icddr,b and Emory teams in Dhaka. The training covered study objectives, ethical considerations, proper use of ODK tools, sample collection protocols, and quality control procedures to ensure consistency and reliability in data collection across all sites.

Key informant interviews with community leaders or local government officials and transect walks were conducted in each neighborhood to identify relevant environmental pathways of potential fecal exposure and contextualize the assessment [24]. Ten environmental pathways were identified, including municipal drinking water, non-municipal drinking water, bathing water, surface water, drain water, flood water, communal/shared latrine surfaces, raw produce, soil and street food (S2 Table). The pathways are described in further detail by Amin et al. (2019) [24]. A total of 1000 environmental samples (10 samples from each environmental pathway as well as 10 soil samples from each neighborhood) were collected according to SaniPath protocols and as described previously [18,24]. Environmental samples were transported from the field to the laboratory at icddr,b and tested for E. coli within six hours of sample collection using the IDEXX QuantiTray/2000 method and IDEXX-Colilert-24^® media (IDEXX Laboratories, Westbrook, ME) according to manufacturer’s guidelines [26]. For quality assurance, we included both a negative control and a field blank. The negative control consisted of PBS or sterile water processed in the lab using the same IDEXX methods, while the field blank involved PBS or sterile water taken to the field and “collected” using the same procedures as actual samples, to verify proper sample collection techniques [18]. Results were reported as the most probable number (MPN) of E. coli per sample unit. E. coli is used in the SaniPath Tool as the fecal indicator bacteria due to the availability of simple and widely-used testing methods and to allow comparability to national and international guidelines for water and food safety. The laboratory protocol is available in the SaniPath website [19] and has also been uploaded to the protocol.io [27].

From 25/04/2017 to 30/01/2018, behavioral data for household, community, and school surveys were collected following SaniPath Tool protocols [18,19]. The household surveys were conducted with mothers about their behavior and that of their children aged 5–12 years, while community surveys were carried out with groups of approximately 12 males or females (separated by sex) about their behavior and that of children aged 5–12 years. School surveys were administered to students aged 10–12 years in local primary schools about their behaviors and that of their adult parents or caretakers. These surveys aimed to quantify the frequency of contact with various exposure pathways for adults (>18 years) and children (5–12 years). The SaniPath Tool does not assess exposure among children under five, as they typically spend more time in private spaces with their caregivers. Children aged 5–12 years are considered a sentinel population because they are active, independent, and likely to experience high environmental contact (e.g., soil, drains, water bodies, and food). This age group provides a conservative estimate of community exposure risk, reflecting behaviors characteristic of both children and adults in shared environments [17,20,21].

The SaniPath Tool focuses on exposure to fecal contamination in public, not private, spaces for adults and children ages 5–12. This focus was chosen because public areas: 1) can be influenced by sanitation policies, 2) often drive private contamination, 3) are where most environmental contact occurs in crowded urban settings, and 4) allow for feasible data collection. There was no difference in the content of the survey questionnaire across the three respondent groups. However, the focus of each survey type varied slightly based on the setting. Household surveys were conducted to understand individual-level contact behaviors. Community surveys focused on assessing various components of community-level exposures. School surveys aimed to explore exposure pathways specific to school-age children. No participants were compensated for their time in completing the surveys. All the surveys asked the same questions about frequency of: drinking municipal piped water and other drinking water, consuming street food, eating raw (uncooked) produce, contact with surface water, contact with open drains, contact with flood water, and using communal/shared latrines.

Questions about frequency of drinking water, ingestion of raw produce and street food, bathing, and using a communal/shared latrine were framed on a weekly timescale. Respondents were asked about consumption of municipal drinking water and non-municipal drinking water on a different frequency scale. The options given to respondents were “Every day in the past week”, “4 to 6 days in the past week”, “3 or less days in the past week”, “Never”, and “Do not know.” The reason for using different time scales is that drinking water is a common behavior that may be difficult to recall the number of times consumed in a week. Questions about frequency of contact with open drains, surface water, and flood water were framed on a monthly timescale as these contacts were likely to occur less often.

While collecting water, we also gathered information on household practices related to water treatment. However, these behavioral data were not included in the exposure modeling analysis. Evidence suggests that intermittent water treatment and produce washing alone, particularly without the use of sanitizing agents, has limited impact on E. coli removal, highlighting its potential relevance in exposure assessments [28].

We employed a systematic random sampling method to select households in low-income and floating/transient neighborhoods. The process began with fieldworkers identifying a central point within the community, such as a mosque, temple, or other prominent demarcation within the neighborhood. From this central point, fieldworkers systematically selected every third household that met the predetermined selection criteria. This process continued outward from the central point until the desired sample size was achieved. Recruiting participants for surveys was challenging in the four high-income neighborhoods. Door-to-door canvassing was not possible in high-income communities because many high-income communities have gated apartment complexes that are not easily accessible. Therefore, we were unable to randomly select households for surveys in these neighborhoods. Instead, paper surveys were sent home with students (with teacher permission) through schools in these neighborhoods, and their parents were able to complete the surveys and return them with their children. No community surveys were conducted in high-income neighborhoods due to difficulties with recruitment in gated communities and the lack of structured community leaders in high-income areas [18].

Data analysis

The primary objective of this study was to assess exposure to fecal contamination through multiple pathways for adults and children in ten Dhaka neighborhoods and identify dominant pathways. The analyses that are built into the SaniPath Exposure Assessment Tool were utilized to estimate the dose of fecal exposure (mean E. coli ingested per month) and the percentage of the population at risk for ingesting E. coli through a specific pathway (estimated from data reported in the household, community, and school surveys) to determine the dominant exposure pathway(s) for each neighborhood. By using a common metric for dose, the risk can be compared across different pathways.

The exposure assessment approach focuses on estimating distribution parameters, rather than relying solely on a single point estimate, to capture behavior frequencies and fecal contamination concentrations in the environment. Bayesian methods, implemented through JAGS (Just Another Gibbs Sampler) [29], were utilized to estimate these distribution parameters [18]. Using these estimates, Monte Carlo simulations were conducted to assess exposure to fecal contamination for both adults and children across various pathways, incorporating intake volumes derived from existing literature and formative study data [18,24,30] (Fig 1).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. SaniPath tool analysis diagram with monte carlo simulation (1000 iteration).

https://doi.org/10.1371/journal.pone.0319883.g001

An exposure metric (E) was calculated for each pathway as the log-transformed product of dose and the percentage of the popoulation exopsed. Pathways are classified as dominant if they have an E value greater than 10 (indicating high risk) or fall within one log unit of the maximum E value observed (e.g., if Eₘₐₓ = 5, dominant pathways are those with E ∈ 4 ≤ E ≤ 5). If all pathways have E values below 1, no dominant pathway is identified. In the absence of established exposure thresholds for many environmental pathways, these cutoffs were informed by formative and pilot testing conducted in Accra [18]. We defined exposure risk or risk of exposure“ as likelihood of being exposed to fecal contamination given someone’s behavior and the E. coli contamination measured in a specific environmental compartment (Fig 1).

To assess the relationship between the frequency of contact with various environmental compartments and types of neighborhoods, we stratified neighborhoods according to socio-economic status and WASH infrastructure [low-income (considered unimproved WASH conditions), high-income with unimproved WASH, and high-income with improved WASH]. All statistical analyses were conducted using R (version 3.4.1) [31]. All the data used for Dhaka SaniPath deployment is available at SaniPath data portals [19].

The key assumptions for ingestion amounts from different environmental pathways are that intake volumes and durations of exposure are fixed values rather than distributions due to limited literature on their variation. The SaniPath tool uses age-specific parameters (adults vs. children) and pathway-specific multiplication factors to calculate the volume or percentage consumed per exposure event. For water-related pathways, ingestion is calculated using the formula: (Intake Volume/Intake Time) × Duration of Event, while food and surface contact pathways use fixed percentages of consumption per event. The details of those assumptions can be found in the S2 Appendix of Raj et al. (2020) [18].

Ethics

The study protocol was reviewed and approved by the human subject research committees at the icddr,b (PR-17027) and at Emory University (IRB00051584). Written informed consent was obtained from all respondents during surveys, and for sample collection we obtained consent from the household head. All data were anonymized prior to analysis.

General behavior frequency

A total of 823 household surveys, 28 community surveys, and 35 school surveys were conducted among the ten neighborhoods. Sample sizes for each neighborhood can be found in Table 1. Household behavioral surveys indicated that children in floating and low-income neighborhoods reported more frequent contact (i.e., at least one contact or ingestion event) with nearly all environmental pathways (bathing water = 99% per week, raw produce, street food, and shared latrines >90% per week, floodwater = 71–91% per month, drain water = 67–81% per month, municipal drinking water = 43–67% per week, and non-municipal drinking water = 33–65% per week), except for surface water (34–43% per month), compared to children living in high-income neighborhoods. In high-income neighborhoods, more children reported weekly contact with or ingestion of bathing water (75%), raw produce (70%), municipal water (66%), and street food (58%) compared to the shared latrines (42%), flood water (18%), and monthly contact with drain water (10%) (Table 2). Results from the community surveys in floating and low-income communities showed similar results for children as what was reported in the household surveys, i.e., a high (>50%) percentage of this population reported contact/ingestion with all pathways, except non-municipal water (34–47% per week) (S3 Table). Similar results were found from school surveys of children. School surveys indicated that children from high-income neighborhoods were likely to come into contact with or consume (54%–94%) most environmental pathways except for shared latrines (42% per week) and surface water (44% per month) (S3 Table). The frequencies of adult behaviors across nine environmental pathways are presented in S4 Table.

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Frequency of reported children’s (age 5-12 years) behaviors (contact/ingestion) across nine environmental pathways based on household surveys in Dhaka, 2017.

https://doi.org/10.1371/journal.pone.0319883.t002

Key dominant pathways in Dhaka

Combining the results of the behavior surveys and E. coli concentrations in the environmental samples using Bayesian analyses, we examined the dominant pathways of exposure to fecal contamination using the same metric of log₁₀ E. coli MPN ingestion per month. The dominant pathways of fecal exposure varied across neighborhoods. For some neighborhoods, there was one very clear dominant fecal exposure pathway (e.g., Gulshan, Fig 2), and in other neighborhoods, there were multiple important exposure pathways (e.g., Ganderia, Fig 2). Despite these variations, several overall patterns were evident.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Total fecal exposure (estimated as log10 MPN E. coli ingested per month) for children in ten neighborhoods of Dhaka, 2017.

[The tree plots are arranged left to right according to the neighborhood's socioeconomic status, and the top row is from the North City Corporation while the bottom is from the South City Corporation. The size of the boxes and the darkness of the color reflect the contribution to the total exposure for different pathways within the neighborhood].

https://doi.org/10.1371/journal.pone.0319883.g002

Dhaka North City Corporation (DNCC) vs. Dhaka South City Corporation (DSCC)

First, comparing the dominant pathways in the north (DNCC) to those in the south part of the city (DSCC), we observed that exposure to fecal contamination through municipal water was a dominant pathway in three of the five DSCC neighborhoods compared to one of the five DNCC neighborhoods (S1 Fig). This pattern was seen for both child exposure (Fig 2) and adult exposure (Fig 3).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Total fecal exposure (estimated as log10 MPN E. coli ingested per month) for adults in ten neighborhoods of Dhaka, 2017.

[The tree plots are arranged left to right according to the neighborhood’s socioeconomic status, and the top row is from the North City Corporation while the bottom is from the South City Corporation. The size of the boxes and the darkness of the color reflect the contribution to the total exposure for different pathways within the neighborhood.].

https://doi.org/10.1371/journal.pone.0319883.g003

Low-income vs. high-income neighborhoods

We then examined differences in fecal exposure by the socioeconomic status of the neighborhoods. There were two higher income neighborhoods, four slum communities, and two low-income floating communities included in the study. In high-income neighborhoods (Gulshan and Dhanmondi), the dominant exposure pathways were mainly via produce and street food (Fig 3), and the overall fecal exposure was lower (orange shading) for adults in high income neighborhoods compared to low-income neighborhoods (Fig 3). In low-income neighborhoods, there were multiple important fecal exposure pathways including contaminated municipal drinking water, contaminated surface waters and flood waters, as well as exposure to contaminated food (e.g., Gabtoli, Kalshi Mirpure, Baddah, Fig 3). However, in DSCC, exposure to contaminated municipal water eclipsed all the other exposure pathways in low-income neighborhoods (e.g., Kamalapur, Shampur, Hazaribagh, Fig 3). While residents in low-income neighborhoods tended to come into contact with open drains, flood water, and surface water more often, our study showed that the residents in high-income neighborhoods, particularly those with unimproved WASH conditions, still ingested high doses of E. coli per month due to elevated concentrations in these pathways (Fig 4).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Average Log10 E. coli/month population exposure to fecal contamination through nine different environmental pathways for both adults and children in ten study neighborhoods, Dhaka 2017.

Note: The y-axis for bar charts does not show the negative values for Log₁₀ Population Exposure (log₁₀ E. coli MPN/month) for communal and shared latrines, where estimated exposure was negative on a log₁₀ scale.

https://doi.org/10.1371/journal.pone.0319883.g004

Improved WASH vs. unimproved WASH

Two higher income neighborhoods (Gulshan and Dhanmondi) had improved WASH infrastructure compared to the eight neighborhoods with unimproved WASH facilities. For the neighborhoods with improved WASH, food-related pathways were primarily responsible for fecal exposure (Fig 3). In the neighborhoods with unimproved WASH facilities, exposure to fecal contamination occurred through ingestion of contaminated municipal water and contact with open drains, in addition to fecal exposure through food-related pathways (Figs 2 and 3).

Child exposure vs. Adult exposure

Differences in behavior between children and adults affected their exposure to fecal contamination in the environment. Children had higher overall fecal exposure compared to adults (predominance of dark red in Fig 2 compared to Fig 3). Children had more exposure to environmental waters, such as open drains, surface water, and flood water, as dominant pathways in both DNCC and DSCC (six neighborhoods) compared to adults. However, exposure to contaminated produce and streetfood in both DNCC and DSCC was a greater risk for adults (Fig 3) compared to children (Fig 2). Pathways that involved direct ingestion, such as uncooked produce, street food, and municipal water,were the dominant exposure pathways for adults in all neighborhoods and for children in high-income neighborhoods of DNCC (Fig 2).

Estimated monthly dose of E. coli

Overall monthly exposure to E. coli ranged from 7.56 to 9.59 log₁₀ MPN/month for adults, with the highest exposure in the Shampur neighborhood, and for children the monthly exposure to E. coli ranged from 7.21 to 9.65 log₁₀ MPN/month) with the greatest exposure in the Gabtoli neighborhood. Examining exposure to each specific pathway, it is clear that the greatest risks of fecal exposure were associated with raw produce, street food, drain water, and flood water in most of the study neighborhoods (Fig 4 and S5 and S6 Table). Risks of fecal exposure from municipal drinking water varied by neighborhood with the greatest risks from municipal drinking water for adults in DSCC with very high estimated monthly E. coli doses for adults [e.g., Hazaribagh 8.52, Kamalapur 8.23, Motijhil 6.68, and Shampur 8.57 log₁₀ MPN; S5 Table]. Exposure to contaminated surface waters was high in many communities in DSCC and DNCC, depending on their proximity to surface water, and was always higher for children than adults [e.g., Badda 8.35, Gabtoli 9.11, Hazaribagh 7.06, Kalshi 6.58, Kamalapur 8.03, Shampur 7.25, and Uttarkhan 6.49 log₁₀ MPN E. coli per month estimated doses for children; S6 Table]. Generally, lower risks of fecal exposure were associated with communal/shared latrines, bathing water (with the exception of the Shampur neighborhood) and non-municipal drinking water (except for some low income neighborhoods).

For children, there was difference in exposure between low-income/floating communities and high-income communities in DNCC for both flood water and surface waters (Fig 4, S7 and S8 Table). There was less variation in exposure to drain water (10⁶–10⁸ MPN E. coli/month) in all ten neighborhoods, except Gulshan – a high-income neighborhood (Fig 4). In contrast, in low-income/floating neighborhoods of DSCC, municipal water was the overwhelmingly dominant pathway of exposure for children, with monthly doses of E. coli exceeding 10⁸ MPN. In low-income neighborhoods in DNCC, exposure was much lower, ranging from 10² to 10⁴ MPN E. coli/month (Fig 4). Further examination of the municipal drinking water pathway results indicates that the percentage of children exposed to contaminated municipal drinking water was above 75% in all ten neighborhoods except Gabtoli, where the percent of children exposed to contaminated drinking water was only around 25% (Fig 5).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. The proportion of children exposed to fecal contamination and log 10 dose (avg log₁₀ E. coli MPN ingested/month) for municipal drinking water across all ten-study neighborhoods.

https://doi.org/10.1371/journal.pone.0319883.g005

Examining the dose for children exposed to contamination in municipal drinking water shows that among DNCC neighborhoods, the estimated average dose was approximately 10³MPN E. coli/month for all neighborhoods except Uttarkhan (10⁷ MPN E. coli/month). However, in DSCC, the estimated average dose of E. coli ingested by children through municipal drinking water ranged from 10⁵ to 10⁸ MPN E. coli/month, with greater variation within neighborhoods (Fig 5 and S6 Table).

Municipal water

Our results indicate that a high percentage of adults and children were exposed to contaminated municipal water, and higher concentrations of E. coli were detected in municipal water samples from DSCC, particularly from Hazaribagh (a low-income neighborhood) compared to DNCC [24]. As discussed by Amin, et al. (2019), potential reasons for the higher water contamination levels in the southern portion of the city include older infrastructure, lack of adequate drainage system, high population density, older, inadequate sanitation facilities, and proximity to the polluted Buriganga river. In Dhaka, municipal water, supplied by Dhaka Water Supply and Sewerage Authority (DWASA), serves as the primary source of drinking and bathing water for an estimated 10.2 million residents. This makes its quality critical to public health, addressing contamination in this system is critical to reducing fecal exposure. Indeed, our study showed that 86% of adults in high-income neighborhoods with improved WASH reported drinking municipal water every day in the prior week (compared to 34% in low-income and 24% in high-income neighborhoods with unimproved WASH). The Dhaka municipal water distribution system faces major challenges, including power outages that reduce pipe pressure and allow contaminants to infiltrate from the surrounding environment. Additional issues include leaking pipes, illegal connections, over-reliance on groundwater extraction, uncontrolled water demand, and the lack of effective metering systems [34,35]. Despite the complexity of these two municipal water distribution systems, this may be the most feasible pathway to target interventions to reduce exposure to fecal contamination because these systems are under direct municipal government control [34] in contrast to food supplies and sanitation systems.

Raw produce and street food

Our results indicated that the levels of fecal contamination in raw produce were high and eating raw produce and street-vended foods was common in most of the study neighborhoods with little variation between neighborhoods with different income levels, WASH conditions, or geographic location. Fecal contamination of produce and street food in Bangladesh and other LMICs has been reported in numerous SaniPath deployments [21] and studies by others [13,36,37]. It is well documented that the consumption of contaminated foods, which are raw or undercooked, can lead to the spread of foodborne pathogens and cause gastrointestinal infections and disease [13,38,39]. However, food supplies are inherently complex and involve a wide range of sources, stakeholders, and regulatory agencies from “farm to fork”, making this exposure pathway challenging to target for intervention.

Consumption of raw produce is common in Bangladesh, where many dishes include ingredients such as raw herbs, vegetables, and spiced water [40]. Contamination of produce may occur at various points along the supply chain from farm to market, and other studies in Bangladesh have reported varying levels of fecal contamination at different points along the chain due to wastewater irrigation, application of fecal material as fertilizer, and unhygienic handling and transport practices [13,41]. While there are risk-mitigating behaviors that can be practiced by farmers, market merchants, and consumers, such as the use of clean water for irrigation, handwashing before and during food preparation, cleaning food preparation surfaces and utensils, washing produce before consumption, these behaviors are not always practiced or are not always effective at eliminating the exposure to produce contamination [40,42].

Street foods are widely consumed across income levels, age groups, and geographic areas in Dhaka and provide convenient and affordable options when people are outside of the home. With the urbanization of Bangladesh continuing to rise, the demand for street food in Dhaka is expected to grow. Consumer surveys have shown that many people report street foods as a major part of their diet and that diverse populations regularly eat street foods, including students, children, wage laborers, housewives, and office workers [38]. With such pervasive consumption of street foods, it is especially concerning that over 90% of street food samples collected during this assessment were contaminated with E. coli [24]. Contamination of street foods in Dhaka generally occurs due to poor personal and food hygiene practices among food vendors [43].

There have been significant efforts to improve the safety of street foods in Dhaka. The UN’s Food and Agriculture Organization (FAO) partnered with the Dhaka North City and South City Corporations (DNCC), and the government of Netherlands to implement the “Improving Food Safety in Bangladesh” program. Through this initiative, participating food vendors are licensed and provided with a new identifiable food cart, receive training on proper hygiene practices, and their food quality and practices are closely monitored by trained citizens [44]. While this is a positive step, licensing and training of vendors are not required by city authority, and contamination of street foods remains a substantial public health concern. Beyond the FAO program, the national government of Bangladesh has made strides to address food safety concerns by enacting regulations that address food production, importation, processing, stockpiling, and supplying, and to ensure people’s rights to access safe foods [45]. However, this authority has not yet created or enforced regulations specifically aimed at street food vendors to improve food safety.

Open drains, surface water, and flood water

The concentration of fecal contamination in samples from open drains was similar across all study neighborhoods, while contamination levels of surface water and flood water were substantially higher in low-income neighborhoods compared to high-income neighborhoods likely due to higher population density and relatively poorer fecal sludge management (FSM) practices in low-income areas [24]. The extensive fecal contamination in environmental water samples in Dhaka is likely a consequence of unsafe FSM in the city. Indeed, The World Bank’s review of FSM in Dhaka in 2014 indicated that only 2% of fecal sludge in Dhaka was effectively treated while the majority of waste was released into the environment without treatment via abandoned onsite sanitation facilities, unsafe collection and emptying of sludge, ineffective treatment of sewage, and sewer leakages [5].

The behavioral data collected in the assessment shows variation in the contact behavior for all three of these pathways across the neighborhood categorizations, with low-income neighborhood residents more likely to report contact with these environmental compartments than high-income neighborhood residents. Low-income “slum” neighborhoods in Dhaka are exceedingly crowded, with over 95% of households measuring less than 14 square meters and a population density of over 220,000 people per km² compared to non-slums with just over 29,000 people per km² [46,47]. Furthermore, slum communities in Dhaka, like those included in this assessment, lack access to clean water and basic sanitation services, have broken drainage systems, and experience frequent flooding and waterlogging [46].

Unsurprisingly, those living in low-income communities come into contact with contaminated open drains and flood water more often than those in high-income neighborhoods. A case study conducted by Ahmed, et al. found that adults in urban slums in Bangladesh engaged in activities that exposed them to surface waters, including bathing, ablution, and water collection, which is consistent with our results [48]. Because these waters have been shown to be highly contaminated, any contact may increase the risks of ingestion of fecal contamination and enteric pathogens. It is therefore important to understand the nature and frequency of such behaviors and how to provide safe alternatives to mitigate this exposure.

Strengths and limitations

This is the largest multi-pathway fecal exposure assessment of its kind conducted in any city. The study collected adult and child behavioral data and environmental data across ten neighborhoods that represented a diverse range of characteristics such as income levels, slum types, WASH quality and access, and geographic regions. Thus, we assessed how risks of exposure from different environmental pathways compared across populations in Dhaka, particularly between high- and low-income communities.

The study has several limitations as well. The assessment was cross-sectional and therefore did not account for any seasonal variation in exposure. We purposely collected all data during the rainy season in Dhaka to represent a “worst case scenario,” but climate change and annual variation in rainfall could create further variation in exposure within a year. Assessments of exposure during both rainy and dry seasons are recommended to better understand seasonal variation. Also, the contamination data was limited to E. coli as a fecal indicator. However, a recent study conducted in similar neighborhoods of Dhaka suggested an association between the magnitude of E. coli present and the likelihood of enteric pathogens (especially at high levels of E. coli) [6].

The behavioral studies posed a particular challenge to conduct in the four high-income neighborhoods of Dhaka. We were not able to conduct any community surveys in these four neighborhoods because there tended to be less cohesion among these neighborhood residents than in the low-income neighborhoods, and opportunities were not available to schedule group surveys in these communities. Collecting household surveys in these neighborhoods was also a challenge because most households were in gated or secured buildings that were not accessible to the enumerators, whereas enumerators in low-income neighborhoods were able to select households randomly and approach them directly. These challenges, along with inherent differences between the different survey methodologies, may have introduced some bias in the surveys. Future studies in high-income communities in Dhaka should consider employing alternative survey methods, for example, telephone or online surveys, as these approaches may be more appropriate for those communities. Detailed strengths and limitations of The SaniPath Tool can be found in Raj et al. (2020).

Although we collected data on household water treatment practices and whether raw produce was washed before consumption, these behavioral factors were not incorporated into the exposure modeling. This exclusion was intentional, as prior evidence indicates that produce washing, especially in the absence of disinfectants, has limited efficacy in removing E. coli and may not significantly mitigate exposure [49,50]. Nonetheless, we recognize that such practices can influence actual exposure levels. We have clarified this methodological decision in the revised Methods section. Future modeling studies should consider integrating these behavioral variables to improve the accuracy and contextual relevance of exposure and risk estimates.