Socio-Demographic Factors Driving the Choice of Alternative Safe Water Sources and Their Implications for Public Health: Lessons from Goalmari, Bangladesh

Abstract

This study examined how socio-economic driving forces influence households’ choice of water, ranging from a piped water supply provided by Veolia to untreated sources contaminated with high levels of arsenic and pathogens. Households fall into three cluster groups based on variations in socio-economic status and physical, infrastructure, and institutional elements. About 64% of the variations are related to differences in awareness, willingness, and ability to pay for safe water sources. Families with higher monthly income showed interest in accepting Veolia’s house connection options, resulting in the shutdown of six community tap points and ultimately affecting the low-income households’ accessibility to Veolia water. A causal loop diagram showed five feedback loops influencing the choice of drinking contaminated water. Bayesian Network models were sensitive to the ability, accessibility, and willingness to pay for safe water, deep tube well distributions, installation and maintenance costs, ownership of tube wells, household income level, and the level of awareness. Results suggest that the risks of drinking contaminated water can be minimized by raising awareness; accepting arsenic removal techniques; sharing expenses; training for deep tube well installations and maintenance; increasing Veolia pipeline coverage; and redesigning the tap point distributions. These results help identify the relative importance of such interventions to improve water security in safe water-poor areas.

1. Introduction

According to [1], about two billion people drink pathogen-contaminated water worldwide. Furthermore, previous studies reported that about 200 million people are exposed to toxic levels of arsenic (As) globally, which substantially increases the public health risks and health-related economic burden in the long run [2]. Diarrhea, a common gastrointestinal disease, ranks ninth among the ten leading causes of death worldwide [3].

In the rural areas of many developing countries in Southeast Asia, a large number of households do not have adequate access to safe water resources, consequently suffering from various waterborne diseases [4,5]. In Bangladesh, India, Nepal, and Pakistan, high As concentrations often restrict groundwater for drinking, cooking, and agricultural uses [6]. Studies have shown that high arsenic (As) exposure levels may increase the likelihood of lung, bladder, and skin cancers, cognitive impairment, obstructive pulmonary disease, black-foot disease, and various forms of cardiovascular disease [2,7,8,9].

To reduce the incidence of waterborne disease from drinking microbially contaminated surface water (SW), governmental and non-governmental organizations (NGOs) installed a large number of shallow tube wells (STW) in the 1970s and 1980s in the shallow aquifer in Bangladesh [10,11,12]. Later, researchers discovered that groundwater in shallow aquifers contains high As concentration, a known carcinogen, especially in Ganges–Meghna Flood plain areas [10,13]. Furthermore, groundwater in the shallow unconfined aquifer is not always free from fecal contamination, depending on the proximity and leaching rate from the pit latrines and soakage wells, which may cause diarrheal infection [14]. Bangladesh’s various types of available arsenic (As) mitigation options are relatively less viable considering their high cost and demanding maintenance requirements or the perennial availability of alternative safe water resources [6,15,16]. Deep tube well (DTW) water and treated piped water supply systems are considered a preferred choice, depending on the availability of infrastructures and accessibility to those resources [16]. Ref. [17] suggested affordable treatment techniques such as oxidation and co-precipitation could be used as an alternative means to reduce As concentration in drinking water.

Despite many studies in identifying and mapping potential contamination sources in drinking and domestic water, significant questions about the influences of key driving forces of drinking arsenic- and pathogen-contaminated water have remained beyond investigation. Therefore, this study aims to provide essential information to understand the socio-economic factors that drive households’ choice of water sources in rural areas in Bangladesh. This study sought to answer how the key socio-economic driving forces influence households’ choice of safe water for drinking and cooking in the Goalmari area. Possible water sources ranged from a piped water supply provided by Veolia to untreated sources contaminated with high levels of arsenic (As) and pathogens. A further aim of this study was to identify potential leverage points in the overall water supply system required for effective interventions to reduce the use of potentially contaminated water.

The study area, Goalmari, is located in the Meghna floodplain of south-central Bangladesh in Southeast Asia, which covers four adjoining villages of the Goalmari Union and Padua Union and lies between 23°29′30″ and 23°30′30″ latitude and 90°42′10″ and 90°43′20″ longitude (Figure 1). The Goalmari area was chosen as a suitable candidate site for this study, considering its reported high As contamination in the shallow aquifer and the large number of people affected by various waterborne diseases caused by the consumption of pathogen- and As-contaminated water. Another important reason is that a large-scale surface water (SW) treatment infrastructure and supply facilities have been operational in the area since 2009, providing an alternative means of safe water access.

2. Previous Studies

A prior survey by the Department of Public Health and Engineering (DPHE) in the study area showed that 83% of the tube-well water was contaminated with arsenic, exceeding the 50 ppb limit [18]. Research carried out by ref. [19] in Matlab, which is about 10–15 km south of Goalmari, reported that, out of a total of sixty-one shallow tube well samples, the As concentration in 70% of the wells exceeded the World Health Organization (WHO) drinking water quality standard (>10 µg/L). Several reports also indicated the presence of a high number of arsenic patients (201–500) in the study area [20,21].

The As content in sediments in the study area ranges from 0.27 to 13.26 mg/kg, whereas in surface soils, it ranged from 2.79 to 7.45 mg/kg [22]. The mineral composition of the upper part of the shallow aquifer consists mainly of quartz and feldspar grains, with a significant portion of biotite minerals. The middle part (43–106 m) contains a high organic carbon content of 0.67 w/w%. Elemental composition (%) of biotite, muscovite, and altered biotite in sediments contain Mg (0.41–6%), Al (4.15–18.3%), Si (10.7–21.24%), K (1.08–8.64%), Ti (0.21–1.59%), and Fe (1.01–21.53%). Ref. [22] reported the Fe and Mn oxyhydroxides in silty sand and clayey silt units and organic matter and phyllosilicate minerals are the predominant sources and sinks of As in the shallow aquifer in the study area. The concentrations of different chemical parameters in the shallow and deep aquifers are provided in

Table 1. Comparison of water quality parameters with refs [23,24] guidelines.

Parameters	Unit	WHO Standard (2004)	Bangladesh Standard (BS 1997)	Shallow Aquifer		Deep Aquifer
				Range	Median	Range	Median
pH	pH unit		6.5–8.5	6.6 to 7.4	7.1	6.8 to 7.1	7
EC	µS/cm			738–2420	1232	825–1217	933
Eh	mV			−170 to −145	−154	−13 to 11	−6
DO	mg/L			<0.1–5.2	0.2	0.7–1	0.8
Arsenic	µg/L	10	50	47–217	109	0.3–0.7	0.55
Iron	mg/L	0.3	0.3–1.0	0.85–5.83	3.45	0.12–0.39	0.14
Manganese	mg/L	0.1	0.1	0.02–2.2	0.15	0.05–3.18	0.11
Sodium	mg/L	200	200	9–583	102	26–165	56
Potassium	mg/l		12	0.9–10.58	4.15	1.57–6	3.26
Calcium	mg/L	75	75	6–133	55	25–94	50
Magnesium	mg/L	50	30–35	1.3–73.5	23.9	11–31	20
Chloride	mg/L	250	150–600	17–834	133	9–249	89
Bicarbonate	mg/L			190–688	292	161–278	220
Nitrate	mg/L	50	10	<0.1–22	0.62	0.1–12	2
Sulfate	mg/L	250	400	<0.1–39	0.14	0.1–13	0.14
Phosphate	mg/L		6	1.75–4.25	2.36	0.17–0.32	0.2
Here, EC: Electric conductivity, Eh: oxidation-reduction potential, EC: electrical conductivity, DO: dissolved oxygen.

Notes: (Source: Refs. [22,25] and data from the Japan International Cooperation Agency (JICA) and the Department of Public Health and Engineering (DPHE), 2018–2019).

below.

In many developing countries, a potable water supply through government initiatives in rural communities is rare [26]. Similarly, due to the high implementation costs and maintenance requirements, the piped water supply system is generally unavailable in rural areas in Bangladesh [27]. According to ref. [28], the piped water supply system is available to only 2% of rural communities in Bangladesh, with high As and salinity levels. Furthermore, the supplied water’s chemical, microbiological, and aesthetic quality often deteriorates due to improper operation and management, evidenced by intermittent water supply, disinfection failures, and water blending [29,30]. This results in various waterborne diseases and consumer dissatisfaction. The shallow aquifer is often contaminated with As in the study area, and the surface water contains biological pathogens causing different waterborne diseases [18,31]. Therefore, the Veolia authority, a French-based company, collaborated with Grameen Bangladesh and installed the treated river water distribution facilities for approximately 7000 residents in late 2009 [32].

Previous literature proposed that infrastructure, institutions, knowledge, and behavior are the key system components to improve effective As hazard mitigation in Bangladesh through identifying strategic intervention points [33]. Ref. [27] conducted a study in the Khulna and Satkhira districts in Bangladesh. They proposed “social acceptance”, i.e., switching from existing water sources to newly implemented safe water sources is a potential barrier to implementing improved water systems in rural areas. Ref. [34] proposed that a piped water system is generally well accepted by the community and can be a reliable means to increase accessibility to safe water for the community. However, installation and maintenance costs are high, making it challenging to implement in rural areas without the government’s support or external funds. Ref. [35] conducted a survey to learn about households’ preference for the piped water systems in Tala Upazila, Bangladesh. About 40% of the respondents in five out of six unions expressed willingness to pay (WTP) for piped water supply systems.

Elsewhere, ref. [26] investigated a community-driven small-scale water system in rural Northwest Cameroon. He concluded that most households accepted it well. Still, the need for a sufficient pipeline network and operation and maintenance costs are significant challenges to the long-term sustainability of this system. Ref. [36] reported that provisions for hand washing and drinking water stations, soap and chlorine solution in hand, and drinking water stations in rural Tanzania decreased the risks of pathogenic infections. Another study in four rural villages in southern Zambia showed a noticeable reduction in the time required to fetch water. It increased safe water accessibility and daily drinking water demands among households after installing a piped water system [37]. A three-year study in Ghana utilizing solar-powered water treatment reported that irrespective of the socio-economic class, unwillingness to pay for the paid piped scheme is the primary cause of not accepting private water systems in most households [38].

With regard to previously adopted analytical approaches reported in related studies, multivariate statistical analyses were applied for the Sustainability of Rural Water Systems, Water Service Continuity, and Rural Water Service studies [39,40]. One inherent limitation of the multivariate regression predictive models is that they include all causal factors in a single analysis [41]. In contrast, Bayesian inference can be used for variables in diverse and unrelated groups. It can efficiently capture the systematic variation in the outcomes for a wide range of variations of the causal factors [42]. A system thinking approach has been used in the past for Waterpoint failure to identify the impact of different interventions, to explore on-site factors influencing hand pump borehole functionality, cross-cutting analysis of water supply and sanitation, piped water continuity, and water system monitoring [43,44,45].

3. Data and Methods

Two semi-structured questionnaire survey data sets were considered, resulting from face-to-face interviews conducted by ICDDR’B, Bangladesh, in 2009 and 2019 before and after implementing the Veolia piped water system. Field enumerators and experts collected the household-level survey data for a project titled “Health effects of a large-scale drinking water intervention on arsenic levels in Goalmari, Bangladesh.” The baseline data consisted of 220 randomly chosen households (IDs 1–220) from the list of all the households in three villages in the Goalmari area. The sample size for the baseline survey was estimated considering prior study findings by the Department of Public Health and Engineering (DPHE). The DPHE noticed that 83% of the tube-well water was contaminated with high arsenic concentrations [18,46]. Then, assuming a confidence interval of 95% and a 5% margin of error, the estimated sample size is 219. The 2019 end-line survey data consisted of the same 220 households as the treatment group (IDs 1–220) and an additional 195 independent measurements as the control group (IDs 221–415) for the experiment (Figure 1).

In the rural Goalmari area, the household members acquire home ownership from their ancestors. Therefore, the household IDs remained the same during the ten-year time interval. Since the piped water intervention had been in operation for the previous ten years, the end-line survey was conducted assuming that improvements in literacy, economic status, awareness, and availability of alternative safe water technologies, including other physical, infrastructure, and institutional components, might bring significant positive changes to secure safe water sources over the years. Therefore, the social acceptance, accessibility, and efficacy of the existing Veolia piped water intervention, including other safe water options, can be estimated by comparing the baseline and end-line survey data and further by comparing the households in the intervention area (treatment group) with those in the adjacent areas, which did not receive the intervention (control group).

A comparison of baseline (2009) and end-line (2019) survey data was conducted using hypothesis testing and analysis of variances. End-line data analyses were conducted using multivariate statistical techniques. This study also assessed the state of the water supply system in the study area by developing a Casual Loop Diagram (CLD) to present a causal relationship with the factors of interest and then to develop Bayesian Network grids. Finally, we performed Bayesian Network (BN) analyses to explore factors influencing the probability of drinking As and pathogenically contaminated water.

At first, all questionnaire parameters were analyzed, and a comparison was made between baseline and end-line data using descriptive and inferential statistics methods, as described in the following sections.

3.1. Multivariate Statistical Analysis

Principal Component Analysis (PCA) was run for the variables to assess the influence on the accessibility of safe drinking water and to see factors influencing pathogenic infections causing diarrhea. Cluster analysis was conducted using the Ward Hierarchical Clustering technique to group the households in the end-line data that showed similarities concerning socio-economic and demographic variables, including choice of water source for drinking and cooking.

3.2. Statistical Hypothesis Testing

The paired t-test and independent t-test were conducted at a 95% confidence interval to compare the statistical differences between baseline and end-line treatment groups and the treatment and control group datasets. If the normality and equality of variances assumptions were not satisfied, then the hypothesis tests were conducted using a permutation test or rank-based (Kruskal–Wallis) method depending on the differences in sample medians and the presence of any significant outliers that may affect the means.

3.3. Analysis of Variances (ANOVA)

Data for different categorical variables were assigned with a range of values for parametric or non-parametric tests, depending on the normality of distributions in the quantile-quantile (Q-Q) plot. Analysis of variances was conducted at a 95% confidence interval to assess the main and interaction effects of the causal factors on the observed outcomes related to accessibility, affordability, and willingness to pay for safe water sources, preference for the Veolia internal house connection option, and the prevalence of diarrhea at the household level in the study area.

3.4. Systems Analysis

3.4.1. Development of Causal Loop Diagram (CLD)

CLDs comprise nodes (variables in the system), edges (connections between variables indicating a relationship), and polarity notations (indicating a positive or negative relationship), which connect to show reinforcing and balancing loops, creating a simple, visually interpretable representation of a system [47].

At first, a CLD was prepared based on information from a literature review, a direct field visit in December 2019, and consultation with field enumerators, randomly selected household members, Veolia authorities, and relevant experts. The factors considered for developing the CLD are a combination of physical, socio-economic, infrastructure, and institutional elements, including various exogenous variables that influence drinking contaminated water.

3.4.2. Bayesian Network (BN) Analysis

BN analyses were conducted using standard BN software, Netica application 6.07 (64 bit version) [48]. The root factors and key driving forces in the CLD were updated in a BN grid to present a causal relationship with the factors of interest by linking various root factors for different functionality criteria. Several variables were discretized to generate two to three categories or levels to homogenize the dataset. The initial BN grid was refined by incorporating the significant factors identified based on the statistical relationship of the household-level survey datasets as mentioned above. This yielded the base scenario for establishing the probability of drinking both As and pathogen-contaminated water.

Bayesian analysis is based on Bayes’ Theorem of conditional probabilities as provided below:

$$\mathrm{P}\left(\mathrm{X}|\mathrm{Y}\right)=\frac{\left[\mathrm{P}\left(\mathrm{Y}|\mathrm{X}\right)\times \mathrm{P}\left(\mathrm{X}\right)\right]}{\mathrm{P}\left(\mathrm{Y}\right)}$$

(1)

where,

P (X) = probability of event X, P (Y) = probability of event Y;

P (X|Y = probability of observing X given Y is true;

P (Y|X) = probability of observing Y given X is true.

The inference was conducted using the law of total probability through a form of joint probability calculation (Equation (2)) and Bayes’ Theorem (Equation (1)) [45]:

$$\begin{array}{c}\mathrm{P}\left(\mathrm{C}1\right)=P\left(C1\right|A1,B1)\times P(\mathrm{A}1,\mathrm{B}1)+P(C1|A1,B2)\times P(\mathrm{A}1,\mathrm{B}2)+P\left(C1\right|A2,B1)\times P(\mathrm{A}2,\mathrm{B}1)\\ +P\left(C1\right|A2,B2)\times P(\mathrm{A}2,\mathrm{B}2)\end{array}$$

(2)

where A1 and A2 describe a factor A being True (1) or False (2), and P(C1) is the probability of the child factor C being True (1).

The sensitivity analyses of different root factors were conducted using Bayesian forward (predictive) and backward (diagnostic) propagation techniques. Finally, based on identifying the significant drivers, potential leverage points were assessed to propose intervention strategies to reduce potentially contaminated water use.

4. Results and Discussion

4.1. Site Specific Information

The household-level survey questionnaire consisted of sixty-two semi-structured questions, of which twenty-one variables were found relevant to this study. Those included family size, literacy, monthly income, drinking and cooking water demand (L/d), availability of safe drinking and cooking water sources, time required and distance to collect safe water from sources, willingness to pay, and ability to spend money for safe water, community preference for Veolia piped water alternative options (tap standpoints or internal house connection), shallow and deep tube well ownerships, water purification technologies, sanitation status, and number of diarrhea patients at households, including others. Summarized information on the key variables is provided below.

Family size, occupation group, and household income level: The median number of family members is between 4 and 5 in the baseline and end-line control and treatment groups. The household’s primary occupations are small local trades, fish cultivation, farming, agriculture practice, and foreign income. As of 2022, Bangladesh’s monthly household income per capita is about 9063 BDT [49]. About 34% of families in the study area have monthly income levels <5000–10,000 Bangladeshi Taka (BDT), whereas 37% have a monthly income in the 11,000–20,000 BDT range, and 29% of families have monthly income >25,000 BDT.

Maximum literacy level in households: The literacy level in the study area has significantly increased over the years. For example, in 2009, at least one family member in 72% of households completed grade six and above, whereas 84% and 93% of households in the treatment and control groups in 2019 attained the same level of education. Only seven households in the end-line population are unable to read, write, or do any basic calculations.

Daily drinking and cooking water demand: Daily household drinking and cooking water demands range between 3–60 and 3–80 L/day/household in the end-line group. In contrast, it was 5–100 and 10–200 L/day/household in the baseline population.

Existing infrastructure to access alternative water sources: The Department of Public Health and Engineering (DPHE) installed tube wells through a United Nations Children’s Fund (UNICEF) project earlier. Furthermore, some community people privately purchased deep tube wells to install in their neighborhoods. However, a large number of tube wells became non-functional over the years. Later, Veolia implemented the piped water supply system at Goalmari in 2009.

Drinking and cooking water collection from the types of water sources: People in the area collect water from a single or a combination of five different types of water sources, including Veolia piped water (treated river water), shallow aquifer, deep aquifer, rainwater, and untreated surface water (river, canal, and pond). During the end-line survey, it was noticed that about 87% of the survey households collect drinking water from a single source. About 163 out of the 415 households collect drinking water solely from shallow aquifers. On the other hand, about 28% use solely surface water such as river/canal or pond water for cooking or other domestic uses (e.g., cleaning utensils, hand washing, etc.). Therefore, a significant portion of the households are exposed to As-contaminated shallow aquifer water, and there is a potential risk of pathogenic infection from using surface water for cooking and other domestic purposes.

Water quality: Surface water is often contaminated with pathogens, whereas groundwater in shallow aquifers contains high As concentrations. The color, taste, and odor of untreated surface water also concern many family members. A deep, confined aquifer is relatively free from contamination, but a slightly high Fe content creates color and taste problems. On the other hand, the Veolia piped water is a preferred choice for many households.

Water quality monitoring infrastructure: The DPHE funded two local NGOs, Tribedi and VARK, 3–4 years ago to identify and mark the tube wells with red color, indicating wells extracting As-contaminated water. However, over the years, the majority of those marks have been washed out from rainwater.

Ownership of tube wells: About 37% of households in the sample population have access to shallow wells, whereas 43% of households have access to deep wells. Of those who have access, 43% of the shallow tube wells and 8% of the deep tube wells are privately owned by immediate family members. Due to the high cost of installing and maintaining deep tube wells, most are jointly owned through cost-sharing among extended family members, neighbors, or communities.

Distance and accessibility to safe water sources: More than 54% of the households have access to water sources within less than 10 m (m). The average distance from tap points to households ranges between 6.1 m and 91.5 m. Data show that about 45% of households are interested in drinking Veolia water but do not have access to that water within a reasonable distance.

Impacts of Veolia alternative options: Community preference for Veolia alternative house connection options resulted in shutting six community tap points. During the end-line survey period (2019), 79 households switched to internal house connections; therefore, only 18 tap points remained operational.

Economic disparity: Households willing to receive Veolia internal house connections need to pay 5000 BDT in advance to manage the expenses for pipeline connection and accessories. Low-income families could not afford that cost, and due to shutting down the nearest tap points to maintain Veolia’s operational cost recovery, several least-income households switched to unsafe water sources, making them vulnerable to diseases.

Water purification technologies: among the 415 households in 2019, twenty-two used different home-based indigenous water purification techniques to remove colloids, odor, Fe, and pathogens from drinking water (e.g., alum, boiling, and Fe removal filters).

Installation cost, maintenance, and service requirements: The costs associated with the DTW installation are about 10,000 BDT, which is a barrier to individual households’ installing DTWs. High Fe content reduces lifespan and efficiency and also clogs the filters of the DTWs. Another common form of maintenance required is to change the well screen once every two to three years, which generally costs more than 2000 BDT. Centralized rainfall harvesting infrastructure is not feasible considering its installation cost (approx. 10,000 BDT) and seasonal water unavailability. On the other hand, individual households spend between 75 and 200 BDT per month to access Veolia water from the tap points.

The installation cost for sanitary latrines: the average cost for installing sanitary latrines varies between 8000 and 12,000 BDT, with an annual operation cost ranging from 200 to 500 BDT.

Sanitation status: About 91% of the households have access to safe sanitary latrines. Out of 88 toilets, 85 are used by sharing among less than five households, and more than five households manage three.

Willingness to pay for Veolia water: Out of 97 households that do not have ownership of any tube wells, 91 are interested in getting access to Veolia water. In contrast, only 32 of the rest of the households buy treated surface water from Veolia.

Performance of piped water distribution facilities: The Veolia authority provides uninterrupted treated surface water to local households. A few households complained about the occasional water odor, color problem, and maintenance requirements of the internal house connection system. However, the Veolia support team was quite efficient in promptly fixing those problems.

Awareness campaign regarding contaminated water and health implications: The Veolia piped water authority has been contributing to sharing information regarding As contamination and health effects through focus group discussions, supporting the annual rally of school-going students on World Water Day, open space loud announcements to the community using microphones (mikes), video shows, placing posters, and leaflet distribution since its inception.

4.2. Principal Component Analysis (PCA) of the Socio-Demographic Variables

Causal factors that influenced water collection from safe water sources and were found statistically significant based on p-values (<0.05) were considered for the cluster analysis. Households mainly fall into three main cluster groups based on variations in socio-economic status and other causal factors, including physical, infrastructure, and institutional elements (Figure 2).

The PCA shows that 64% of the variations in the dataset are related to differences in awareness, willingness to pay, and the ability to pay to access safe water sources (Equations (3)–(5)).

$$\mathrm{P}\mathrm{C}1=0.32\mathrm{N}\mathrm{D}\mathrm{S}-0.19\mathrm{I}\mathrm{n}\mathrm{c}-0.49\mathrm{S}\mathrm{p}\mathrm{e}\mathrm{n}\mathrm{d}-0.22\mathrm{D}\mathrm{W}1-0.14\mathrm{D}\mathrm{W}2+0.53\mathrm{S}\mathrm{W}1-0.52\mathrm{V}\mathrm{e}\mathrm{o}\mathrm{l}$$

(3)

$$\mathrm{P}\mathrm{C}2=-0.34\mathrm{N}\mathrm{D}\mathrm{S}+0.13\mathrm{I}\mathrm{n}\mathrm{c}+0.34\mathrm{S}\mathrm{p}\mathrm{e}\mathrm{n}\mathrm{d}-0.57\mathrm{D}\mathrm{W}1-0.48\mathrm{D}\mathrm{W}2+0.18\mathrm{S}\mathrm{W}1+0.39\mathrm{V}\mathrm{e}\mathrm{o}\mathrm{l}$$

(4)

$$\mathrm{P}\mathrm{C}3=-0.63\mathrm{I}\mathrm{n}\mathrm{c}-0.16\mathrm{S}\mathrm{p}\mathrm{e}\mathrm{n}\mathrm{d}-0.68\mathrm{C}\mathrm{D}-0.18\mathrm{D}\mathrm{W}1-0.15\mathrm{S}\mathrm{T}\mathrm{W}+0.23\mathrm{V}\mathrm{e}\mathrm{o}\mathrm{l}$$

(5)

where, NDS—number of drinking water sources; Inc—average monthly household income level; Spend—willingness to spend money for Veolia Piped water; DW1 and DW2—upper and lower parts of the deep aquifer; STW—shallow aquifer water; SW1—surface water source; Veol—households collecting Veolia piped water; and CD—household cooking water demands (daily).

The first principal component shows a contrast between variable drinking surface water with drinking deep aquifer water, the willingness to pay for Veolia water, and the collection of Veolia-supplied water, which indicates the level of awareness among the households. Households unaware of drinking potentially contaminated water are reluctant to spend money on drinking from alternative safe water sources. Previous studies carried out in Northwest Cameroon, Ghana, and several villages in Tala Upazila in the Satkhira district, Bangladesh, showed similar results [26,35,38].

The second principal component showed positive loadings for variables such as the willingness to spend money for Veolia water and the collection of Veolia-supplied water and negative loadings for the number of drinking water sources and drinking deep aquifer water, which indicates accessibility to existing resources. People who generally do not have access to deep tube wells are more inclined to spend money on treated Veolia water. On the other hand, the third principal component indicates the affordability of the households. The larger households with high daily water demands are more inclined to rely on the existing shared shallow and deep aquifer water sources instead of spending to purchase treated piped water, which conforms to the research findings in several rural villages in coastal Bangladesh and Ghana [35,38].

Overall, the result indicates that households with low monthly incomes also negatively influence willingness to pay for safe water sources, which is consistent with studies carried out in Ghana, Northwest Cameroon, and coastal districts in Bangladesh [26,35,38].

4.3. Differences in Socio-Economic, Water Availability, and Sanitation Indicators Using Statistical Hypothesis Test

4.3.1. Baseline Group (2009) Versus End-Line Treatment Group (2019)

A comparison of the baseline survey samples and the end-line survey population from the same households shows that the small (1–3 members) and medium (4–6 members) family size has increased from 19.5% to 24% and 60.5% to 65%, respectively. Statistically, significant differences were observed in household maximum literacy level (p-value: 2.2 × 10⁻¹⁶ < 0.05), daily drinking and cooking water demands ((p-value ~2.2 × 10⁻¹⁶); 95% CI: (2.285, 20.71) and (4.785, 34.57)), and the number of households using sanitary latrines for defecation (p-value ~0.002). The household literacy level (grade 6 and above) has increased from 72% to 95%. A significant improvement was noticed regarding sanitation facilities, where the access to sanitary latrines increased from 11% to 96%. The households that needed a relatively smaller amount of water (<20 L/day) for drinking have increased from 47% to 86% and from 14% to 45% for cooking. Due to the availability of multiple alternative water sources in 2019, the average maximum distance (>10 m) for the households to collect water has decreased from 69% to 51% compared to the baseline survey period.

The number of households who accepted the Veolia piped water option (p-value ~2.2 × 10⁻¹⁶) significantly increased over ten years. However, a relatively higher percentage of the people (about 41%) were found to entirely rely on shallow aquifer water, surface water, or a combination of those available sources for consumption, which may increase the likelihood of As disease and pathogenic infection. The major causes of the adoption of multiple water sources, including potentially unsafe water sources in the study area, might result from the simultaneous influence of several factors. These include water service continuity, since a lot of tube wells are non-functional, as well as the seasonal unavailability of untreated surface water during dry months. There is also a lack of ownership of tube wells and Veolia pipeline coverage at a commutable distance. Further issues include the affordability of paying for a piped water supply and deep tube well installation and maintenance costs, together with a lack of awareness regarding drinking safe water and, hence, an unwillingness to pay for alternative safe water sources. Similar results were reported in several rural areas in developing countries such as Pakistan, Ethiopia, and Mozambique [50,51].

4.3.2. End-Line (2019) Treatment Group Versus the Control Group

Statistically, no significant differences were noticed in income level, literacy level, family size, ownership of tube wells, daily drinking and cooking water demands, acceptance of home-based water purification technologies, and ability to pay for Veolia water between the treatment and control groups (p values > 0.05). A comparison between the treatment and control groups revealed that almost similar numbers of households (49%) are willing to pay for Veolia water in the control group households. However, statistically, a higher proportion of the treatment group households are collecting Veolia water (p-value: 2.2 × 10⁻¹⁶ < 0.05), indicating relatively better accessibility to piped water systems, including the willingness to pay that resulted from awareness campaigns. Even though both groups are equally concerned about child sanitation, the treatment group samples were more inclined to use adult sanitary latrines than the control group (p-value: 0.0051 < 0.05).

4.3.3. Influence of the Main and Interaction Effects on the Outcomes of Response Variables in the End-Line (2019) Dataset

ANOVA test results showed household income level and family size influence their affordability to access safe water for drinking and cooking purposes (p-value: 2.441 × 10⁻⁷ < 0.05). Previous studies proposed similar results in the Lagos metropolis and several low- and middle-income countries in Central and South America, Africa, and Asia [52,53]. Despite piped water intervention, a preference for a Veolia internal house connection (p-value: 0.0004057 < 0.05) among the relatively affluent households further restricted the safe water accessibility to the poor households due to the shutdown of several Veolia tap standpoints. Similar findings were obtained in a separate study in a different rural region of the country. Results showed that although the water supply system is acceptable, the existing water governance cannot provide safe water security to marginalized communities and negatively affects the water index in Singair Upazila of Bangladesh [54]. Statistically significant interactions (p-value: 0.02215 < 0.05) were noticed for the preference for the Veolia internal house connection option with respect to income level, literacy, and DTW ownership (Figure 3). On the other hand, the willingness to pay for Veolia water does not significantly differ based on individual causal factors such as household income levels (p-value: 0.2215 > 0.05), indicating a lack of awareness as the possible reason. In [55], the authors reported similar findings in Nigeria. However, the interaction effects of literacy, income level, and the number of diarrhea patients affect the willingness factor, which is consistent with the study findings from the Chila Union of the Bagerhat district, Bangladesh [56]. Along with the various causal factors mentioned, ownership of deep tube wells influenced the choice to adopt Veolia piped water (p-value: 0.0053 < 0.05). Overall, these results indicate affordability, a lack of awareness, the availability of existing water sources, and a lack of interest in switching to a new water source, individually or simultaneously influencing the adoption of Veolia piped water in the study area. These findings are consistent with the PCA results mentioned in Section 3.2. Ref. [35] reported similar results in their study at Tala Upazila in a Coastal District of Bangladesh.

Statistically, significant interactions were observed in a number of diarrhea cases at the household level, considering drinking and cooking water sources (p values < 0.05). Filter and water purification technologies are generally not very common in the study area. Interactions were observed for households using relatively inferior quality drinking water (e.g., untreated surface water) and no water purification and, in some cases, using alum and filters (p-value: 0.00057 < 0.05). Households using settling and straining, including boiling water, showed no presence of diarrhea patients. This result contradicts a study’s findings in a rural area of the Pisco Province, where no associations were reported between households adopting boiling and chlorination for water purification with pathogenic infection among children below five years of age [57].

Although several households use sanitary latrines (a ring slab pit and ventilated pit), they have a higher prevalence of diarrhea due to poor drinking and cooking water. Overall, results showed that the prevalence of diarrhea in the study area was affected by the microbial water quality of the drinking and cooking water and the choice of water purification technology, as well as the extent of toilet sharing and the distance of shallow tube wells from the open pit or hanging latrines. These results conform with the findings from the previous studies about the associations of microbial water quality for drinking and cooking, the adoption of water purification, and safe sanitation practices with the prevalence of diarrhea in Ethiopia, Nepal, Bangladesh, South India, Pisco Province, and West Cameroon [57,58,59,60,61].

4.4. A Systems Approach

4.4.1. Construction of a Causal Loop Diagram

A Causal Loop Diagram (CLD) was prepared to represent the interactions between the key factors and driving forces for identifying effective leverage points related to drinking As-contaminated water (Figure 4). A brief description of the key components has been provided in Section 3.1 above.

Feedbackloops

Five feedback loops were found in the model that directly or indirectly influence the high As contamination in drinking water. Loops I through IV show a reinforcing relationship to drinking As-contaminated water, whereas loop V is balancing to reduce drinking As-contaminated water. Arrows in the feedback loops indicate the direction of causality, where the positive signs (+) and negative signs (−) indicate direct and inverse relationships between the connected factors. Feedback loop I mainly depicts how household-level awareness can lead to the avoidance of contaminated shallow tube wells and the adoption of alternative safe water sources and water purification technologies for drinking water. The second feedback loop presents how drinking contaminated water influences household health and economic status and restricts their affordability and willingness to pay for safe water in the long run. The feedback loop III further extends the idea that financial hardship adversely affects their accessibility to safe water sources. The feedback loop IV highlights the involvement of the private sector and the central government in providing infrastructural support, strengthening routine As monitoring measures, knowledge-sharing with the local communities, and raising awareness that might ease the household cost burden and their willingness to pay for safe water. To maintain equitable, safe water access, proper monitoring of the public fund disbursement (e.g., DTW distributions at subsidized costs) among the community people is vital. The feedback loop V depicts how the local government and media can play roles in disbursing local facilities, impartially considering household needs. This, and a willingness to pay, can increase accessibility to safe water resources irrespective of socio-economic class.

Driving forces, exogenous variables, and significant delays

The feedback loops represent several driving forces operating in the system continuously with varying intensity, potentially leading to drinking As-contaminated water. These are the availability of and accessibility to alternative safe water sources, awareness and perception of As hazards, local government initiative and social media participation, willingness to pay for As-safe drinking water, and environmental hazards. Earlier studies reported those driving forces to drink potentially contaminated water in many developing nations [26,35,55,62].

Several exogenous variables were identified, each influencing the system at variable strengths, affecting the household’s choice to drink As-safe water. For instance, different environmental hazards such as monsoon floods and hurricanes are common in the study area, affecting household income levels and also deteriorating microbial and chemical water quality in the surface and shallow aquifer water through the mixing of contaminated water from the unsanitary latrines and irrigation lands. A gradually increasing water demand for various industrial, commercial, and farming operations, as well as water quality deterioration from environmental hazards, negatively influences the availability of alternative safe water sources. Ref. [63] also reported that increased water demand due to the growing population and economic activities may further hamper the accessibility to safe water resources in rural areas. The literacy level in the study area is gradually increasing, which influences the household income level to some extent, makes people aware of water-borne diseases, and motivates them to pay for safe water sources. The gradual increase in installation and maintenance costs of different safe water sources (e.g., deep tube wells, rainwater), including standard water purification technologies due to inflation, negatively affects the accessibility of safe water for low-income households.

Several bottlenecks significantly delay the system’s progress. For example, it can take a significant amount of time to process official paperwork and conduct planning and field logistics after funding approval from the central government. Sometimes, rural people do not accept switching from existing water resources to new ones. Poverty and socio-economic disparity often negatively affect affordability and cause inequality and even inaccessibility to safe drinking water facilities.

The driving forces and exogenous variables, including the causes of significant delays, are not entirely independent and act individually or simultaneously to influence the household’s choice to drink safe water in the study area. The driving forces are linked to more than one feedback mechanism shown in the CLD. Therefore, potential leverages can be identified, and interventions can be undertaken to minimize the dominance of some of these forces, aiming to improve safe water accessibility in households. Several potential areas of intervention are discussed in the following sections.

4.4.2. Bayesian Network (BN) Analysis

The inference of the Bayesian Network analysis was conducted using the law of total probability through joint probability calculation and Bayes’ Theorem. A brief description of the various root factors and the functionality criteria are provided in the previous sections. The initial base model scenarios depicting the sequential cause and effect relationships for “Drinking As-contaminated water” and “Infection from drinking and domestic uses of pathogen-contaminated water” are shown in Figure 5a,b, respectively.

BN Analysis: Probability (%) of Drinking As-Contaminated Water

The BN grid tested for five different functionality criteria:

• Willingness to pay for Veolia water;
• Willingness to pay for DTW installation and maintenance;
• Spending money for As mitigation;
• STW users (no alternative sources);
• Infrastructural access to safe water sources.

These are influenced by various behavioral, infrastructural, demographic, and economic root factors, ultimately contributing to the household’s choice of drinking As-free water. In the backward propagation, the probability estimation of the factor of interest, i.e., drinking As-bearing water, was set to zero to determine the consequent changes in the root factors needed to achieve this. The sensitivity analysis results, using backward propagation, identified the level of improvement of the root factors needed to confirm a 100% probability of drinking As-safe water (Figure 6a below).

The results indicated that accessibility to alternative safe water sources, including the Veolia tap point distance, level of awareness, adherence to existing shallow tube well sources, and ability and willingness to pay for Veolia water and deep tube well installations, are the most significant root factors that need further improvement.

In the forward propagation, the sensitivities of each of the root factors to the probability of drinking As-safe water were assessed, in turn, by changing them individually or by changing multiple root factors simultaneously to their ideal condition state. The sensitivity analysis results using forward propagation, in

Table 2. Summary of the single- and multi-factor scenario test results using Bayesian forward propagation to assess influence on drinking As-contaminated water.

Root Factors	Scenario (Base Condition, 2019)	Functionality Criteria					Factor of Interest
		WTP for Veolia Water	Spend for As Mitigation	WTP for DTW Installation and Maintenance	STW Users Only	Infrastructural Access to Safe Water	Probability (%) of Drinking As-Contaminated Water
Awareness	100% Yes	40.4	2.56	53.8	-	-	41.7
	100% No	31.9	2.16	35.1	-	-	45.2
Non-ownership of Tube wells	100% Yes	51		42.2	-	-	41.9
	100% No	32.4		47.1	-	-	43.6
Household income level	100% Low	26.3	2.21	49	29.4	-	44.7
	100% Medium	-	-	-	-	-	43.4
	100% High	50.9	2.63	41.8	25.1	-	41.2
Family size	100% Small	37.3	2.4	45.8	27.5	-	43.1
	100% Medium	-	-	-	-	-	43.2
	100% Big	36.4	2.38	46.1	27.6	-	43.2
Households Ability to pay	100% Yes	82.2	3.16	32.6	19.7	-	36.7
	100% No	19.3	2.09	51.1	30.6	-	45.6
Average DTW Installation cost	100% Public	-	-	57.1	-	-	41.7
	100% Collective	-	-	47.7	-	-	42.9
	100% Private	-	-	38.3	-	-	44.2
Community preference (Veolia)	100% Tap Point	-	-	-	25.9	63.4	41.2
	100% House connection	-	-	-	28.5	52.2	44.3
Veolia water Access	100% Yes	-	-	-	23.2	74.4	38.2
	100% No	-	-	-	32.7	34.4	49.1
DTW Distribution	100% Even	-	-	-	23.3	74.1	38.3
	100% Moderate	-	-	-	-	-	42.7
	100% Poor	-	-	-	31.2	40.6	47.4

Note: “-” missing data.

below, show the relative sensitivities of different root factors to drinking As-free water based on Shannon’s entropy reduction. According to ref. [64], a greater entropy reduction score indicates a stronger influence on the outcomes. Following this procedure, the ranking below shows the degree to which altering one factor in the grid will likely contribute to the probability outcomes of others.

Households’ ability to pay > Veolia water access > DTW distributions > Average monthly income > Non-ownership of tube wells > Community preference to Veolia house connections > Awareness = Average DTW installation cost.

The model results indicate that several potential root causes are significant barriers to drinking As-safe water. Among the various root factors in the backward propagation, the model was susceptible to the root factor “adherence to accessing shallow aquifer water”. About 15% of households solely drink shallow aquifer water, and another large fraction of the total households access drinking water from multiple sources where shallow tube well water is common. It is inversely influenced by two other significant root factors: the household’s ability to pay and inadequate infrastructural accessibility to safe drinking water sources. Another reason might be a lack of social acceptance of the newly introduced piped water system. Ref. [27] also reported that users’ adherence to contaminated shallow aquifer water is a potential barrier to switching to available safe water sources in Khulna and Satkhira districts, Bangladesh.

Both forward and backward models predict that the functionality criteria “Willingness to pay for Vaolia water” and “Willingness to pay for DTW installation and maintenance” are predominantly controlled by their ability to pay for safe water sources and partially controlled by their awareness level and ownership of shallow tube wells. The forward propagation results also indicate an increase in the market price for DTW installation materials and maintenance costs, and inequalities in the public fund disbursement (e.g., DTW installation at subsidized costs) negatively influence the household’s willingness to pay for deep tube well installations. Both models suggest that the infrastructural accessibility is largely influenced by the availability of the Veolia pipeline coverage within a reasonable distance. It is also somehow influenced by the preference of the relatively wealthy residents for internal house connection, which resulted in the shutdown of several tap points and further affected the accessibility of piped water by the poor households. In addition, the distribution of the DTWs was one significant component contributing to the overall infrastructural accessibility to safe water sources in the forward propagation model.

BN Analysis: Probability (%) of Infection by Microbial Pathogens Causing Diarrhea

The backward propagation predicted the best-case scenario for the “probability (%) of Infection from drinking and domestic uses of pathogen-contaminated water”, shown in Figure 6b below. The BN grid tested for four different functionality criteria:

• Potentially contaminated water source (drinking and domestic);
• Pay for safe sanitation;
• Water purification;
• Infrastructure accessibility.

These are influenced by a range of root factors, including behavioral, infrastructural, sanitation status, economic, and accessibility to resources, ultimately contributing to the household’s choice of adopting microbially contaminated water. Similar to the BN model in the previous section, this model shows the level of change needed in different root factors to confirm a 100% probability of not taking pathogen-contaminated water and, therefore, zero diarrhea cases. The results indicated that the level of awareness, ability to pay, surface water uses, adoption of water purification technologies, willingness to pay, and infrastructural accessibility to safe water and sanitation were this model’s most significant root factors. This model indicates that household-level awareness, tube well ownership, distance from water sources, and their ability to pay influence the willingness to pay for safe water sources such as Veolia water and DTW installations, the adoption of water purification, and safe sanitation. This result is consistent with the research findings from a study at Tala Upazila in coastal Bangladesh [35]. On the other hand, infrastructural accessibility (e.g., availability of DTW or Veolia pipeline coverage) and their ability to pay to encourage them to adopt untreated surface water for drinking and cooking purposes.

In the forward propagation, the sensitivities of each of those root factors to the probability of pathogen-free sources were assessed using the same procedure as mentioned earlier. The sensitivity analysis using forward propagation in

Table 3. Summary of single- and multi-factor scenario test results using Bayesian forward propagation to assess the influence of drinking pathogen-contaminated water.

Root Factors	Scenario (Base Condition, 2019)	Functionality Criteria				Factor of Interest
		Potential Contaminated Water Sources (Drinking and Domestic)	Pay for Safe Sanitation	Water Purification	Accessibility to Infrastructure	Perceived (%) Probability of Pathogenic Infection
Awareness	100% Yes	43.7	100	7.53	-	18
	100% No	51.2	42.3	4.3	-	21
Non-ownership of tube wells	100% Yes	46.5	-	-	-	19.3
	100% No	47	-	-	-	19.3
Installation (sanitary latrines)	100% High	46.3	75.4	-	-	19.3
	100% Low	46.9	83.1	-	-	19
Households’ ability to pay	100% Yes	33.1	84.2	17.3	-	16.6
	100% No	52.2	72.4	1.87	-	20.3
Avg. DTW Installation cost	100% Public	43	-	-	-	18.9
	100% Collective	46.3	-	-	-	19.2
	100% Private	49.5	-	-	-	19.6
Veolia water Access	100% Yes	46.1	-	-	77	17
	100% No	47.8	-	-	27	22.1
DTW distribution	100% Even	46.1	-	-	77.2	17
	100% Moderate	-	-	-	-	19
	100% Poor	47.6	-	-	32.2	21.5

Note: “-” missing data.

below shows the relative sensitivities of different root factors influencing contaminated water use that may increase the likelihood of diarrhea infection. The below ranking shows the relative sensitivities of the different root factors in decreasing order.

Households’ ability to pay > DTW distribution = Veolia water access > Awareness = Water Purification > Avg. DTW Installation cost> Installation (sanitary latrines) > Non-ownership of tube wells.

This model shows that the functionality criterion “Potentially contaminated water source (drinking and domestic)” depends on several root factors that are considered. However, the greatest sensitivities were noticed for household awareness, their ability to pay, and the cost-sharing (individual households, cost-sharing with neighbors/extended family members, and the public) for the DTW installation. The functionality criterion “Pay for safe sanitation” largely depends on household awareness and is also partially influenced by sanitary latrines’ installation cost and ability to pay.

In short, reliability, ability, and habit are the significant barriers influencing the household’s choice to connect to potentially arsenic-contaminated shallow aquifer sources. On the other hand, households’ preference to adhere to pathogen-contaminated surface water sources is governed by the norm factor, i.e., the type of water sources their neighbors depend on. Two additional factors, i.e., people’s perceived risk of becoming sick from drinking pathogen-contaminated water and attitude factors relating to the taste/odor of surface water and the distance of the water source, can facilitate switching to an alternative safe water source.

Both models for drinking or domestic uses of As- and pathogen-contaminated water were found to be sensitive to the household’s ability to pay, accessibility to Veolia piped water, deep tube well (DTW) installation and maintenance costs, DTW distributions, ownership of tube wells, and level of awareness. The As model was also found sensitive to the community preference for Veolia house connections, resulting in the shutdown of community tap points. On the other hand, accessibility to sanitary latrines and water purification technologies are two significant root factors influencing exposure to pathogen contamination.

Considering the results from the systems view approach, recommendations are provided for potential areas of intervention in Section 4.4.3 below.

4.4.3. Recommendations Based on Systems Analysis

Several interventions can be considered to reduce the drinking of potentially contaminated water. They are as follows:

Household-level initiatives: Households need to know the types and sources of contaminants and be able to identify the appropriate water purification techniques required. Least-income households, drinking shallow aquifer water, can utilize affordable home-based As removal techniques.

Community participation: community members can share expenditures among several neighborhood households to install deep tube wells and water purification filters.

Initiatives from the Veolia authority: The deeper aquifer in the study area is confined in nature. As a result, mining deep aquifer water can be a short-term solution, considering its limited recharge potential to avoid groundwater depletion. Therefore, the Veolia authority may conduct ongoing technical and economic feasibility assessments to increase its pipeline coverage. They can also effectively redesign the tap point distributions considering the current demands or make the Veolia house connection options affordable to low-income households depending on the availability of resources.

Private sector, local, and central government initiatives: The central government can also assist Veolia in managing the expenditures to increase its pipeline coverage and the number of tap points. Furthermore, the government and NGOs can put joint efforts into reducing installation and maintenance costs and making materials more available to local households. A portion of government funds can be efficiently allotted to regularly monitor As and pathogen contamination, recruit skilled personnel to monitor the piped water management system, engage in active awareness-building campaigns, and provide necessary technical training to maintain and repair existing deep tube wells.

The various components that may affect the adoption and implementation of the suggested interventions include the following:

i.

Cost barriers to improving the pipeline infrastructures.

ii.

There is a lack of collaboration among the stakeholders, which include community members, the Veolia authority, local business people selling resources, and the central government.

iii.

Technical, institutional, and administrative barriers to maintaining equality in the distribution of resources, disregarding their socio-economic status.

iv.

Learning and adopting new technologies, such as accepting home-based As removal filters and cost-sharing for deep tube well installations, represents behavioral challenges.

5. Summary and Conclusions

The novelty of this study is that it is the first time data have been available on water point selection before and after the implementation of a large piped infrastructure in rural Bangladesh. Among the various causal factors, awareness, affordability, willingness to pay, and infrastructural accessibility to safe water sources, including the inclination to adopt water purification technologies, were crucial in driving households to consume As- and pathogen-free water. The availability of the centralized and treated Veolia piped water system has opened a reliable alternative safe water source at a reasonable price to achieve long-term solutions for safe water scarcity in one of the most As-prone rural areas in Bangladesh. However, due to a lack of adequate financing, the piped water networks could not be evenly distributed in every community in the study area. Furthermore, the uptake of house connections caused the shutdown of affordable community tap points due to a relatively low-cost recovery potential for the limited number of tap point users at specific locations. As a result, wealth inequalities indirectly restricted the poor households’ accessibility to treated piped water systems and forced many of them to choose contaminated surface and shallow aquifer water. Equitable access to safe water can be achieved by increasing the Veolia pipeline coverage and tap point distributions and limiting the choice of Veolia house connections. Sharing expenditures among the group of households to install DTWs and As removal filters might be a potential short-term option for low-income groups. Additionally, portions of public funds should be disbursed for regular water quality monitoring and technical training to maintain and repair existing technologies.

Like many other developing nations, the major constraints for any sustainable large-scale water supply infrastructure projects in the rural areas of Bangladesh are (i) the risks of low-cost recovery and, consequently, the unwillingness of private partners in the financing, (ii) the lack of evidence-based intervention design, and (iii) poor management and operations. The Government of Bangladesh should identify priority areas considering the level of water contamination, available water resources, existing infrastructures, and socio-economic characteristics. The Government of Bangladesh can support the initial establishment cost, and an evidence-based, affordable tariff structure can be designed to preliminarily assess the cost recovery potentials for an equitable and sustainable water distribution system. Site-specific preliminary evidence should be gathered through a pilot program to establish the demands, ability, and willingness to pay for the alternative safe water options, prevailing facilitators and societal barriers to adopting the newly implemented safe water sources, and the level of government funding support requirements for the marginal communities.

For instance, in the rural Goalmari area, the costs borne by the marginal communities may be subsidized through the National Strategy for Water and Sanitation 2011 (amendment 2021) program funds for hard-to-reach areas of Bangladesh. This might ultimately reduce the impacts of wealth inequalities and, therefore, improve the water security index in the study area. Such initiatives can eventually support the operational costs of Veolia and, thus, can contribute to increasing the tap point distributions. It can facilitate the households’ accessibility within a reasonable distance and improve attitudinal factors to accept the treated surface water for consumption. Institutional coordination can be strengthened by engaging community representatives, including the local government and stakeholders, who can indirectly support the improvement of normative and reliability factors in the community’s switch from arsenic- and pathogen-contaminated water to alternative safe water options. The Veolia authority and local government should ensure careful investigation and monitoring practices regarding piped system operation and management, specifically consistency of water supply, water quality, disinfection failures, and public satisfaction. These results would be helpful to the regulators and policymakers in planning interventions in relatively water-poor areas in the country. Furthermore, the analysis and results in this research might serve as a guide for sustainably planning future piped water installation projects and integrated water resource management practices in rural settings in other developing nations.