Study design
The study was conducted in August and September 2018 (during monsoon season) in two FDMN-camps in Kutupalong, Cox's Bazar. We selected Camp-1 and Camp-2, where both the longer-term and recently arrived FDMN population lived together (Figure S2). One of the investigators (NA), along with the field research team, conducted initial scoping visits to the selected camps and performed preliminary water quality assessments measuring iron, turbidity, and total dissolved solids (TDS) using digital field-test kits. They selected two shallow tubewells from each camp with different water physicochemical properties: tubewell ID-13 and tubewell ID-17 from Camp-1 and tubewell ID-2 and tubewell ID-5 from Camp-2.
The FDMN-camps in our study area consisted of multiple households with shared toilets, bathing spaces, and drinking water sources owned by the Government of Bangladesh and managed by humanitarian agencies and WASH sector coordinators Cox’s Bazar. Groundwater was extracted through shallow tubewells (< 76-meter depth) and maintained by Action Against Hunger (ACF). Each tubewell (water collection point) was located within an accessible distance (i.e., within 5-minute walk) for multiple households and was used for drinking and other household purposes. Water points were selected based on the following eligibility criteria: (1) water point shared by ~100 households (2) water extracted by a manual underground shallow tubewell, (3) the tubewells were the primary drinking water source for the families/users, and (4) had sufficient space around the platform of the tubewell for Zimba installation. There was no active chlorination project in place when enrolling the study participants. Although there was no active water treatment program in our selected neighborhoods, a small number of households had chlorine tablets [e.g., Aquatab (Medentech, Inc, Wexford, Ireland)] supplied by NGOs.
Field team
There were two assigned teams for this study, the field research team and the sample collection team. The field research team conducted surveys and community sessions. The team consisted of experienced staff who were fluent in Bengali and Rakhine language (the mother language spoken by the FDMN). The sample collection team consisted of trained staff responsible for water sample collection.
The survey team used convenience sampling to select ten households adjacent to each tubewell to participate in household surveys. Households who reported that they collected water from the intervention tubewells were prioritized for household surveys. The team visited and provided information on Zimba, described the study activity with household members using cue cards and other visual aids, and obtained their written informed consent in Bengali and Rakhine. They surveyed households at baseline, immediately after Zimba installation chlorine treatment of the primary tubewells, and at end-line, after four weeks of intervention (Figure S3). Mothers with at least one child less than five years of age were given enrollment preference in the surveys.
Baseline, follow-up and end-line data collection
We collected data on demographics, water collection and storage practice, drinking water treatment method, and perception of existing camp water supply.
The survey team conducted twice a week follow-up visits to the selected households. They collected water samples directly from the Zimba, and household storage containers after treatment. They performed hourly tests of immediately treated Zimba water every day for eight hours to ensure the source's chlorine dosing accuracy.
Within seven days of Zimba installation, the survey team conducted short surveys with the study participants to assess immediate acceptability, satisfaction, and problems they were facing with the Zimba. The team tried to address the problems (i.e., increased height of the hand pump, smell/taste of chlorine and queuing while collecting water) raised by the users in subsequent weeks. At the end of the 4th week of intervention, 20 intervention household members were interviewed during the end line survey. The survey team used convenience sampling to select one household member involved in regular water management activities for the end line. Household surveys were conducted to collect household demography, drinking water quality satisfaction and perception, water collection, storage practice, and treatment methods.
A total of 10 in-depth interviews (IDIs) were conducted with the study participants other than those participating in baseline surveys from among enrolled households (5 from each site) to understand their views and beliefs on Zimba water. One member of the survey team, along with a local Rakhine interpreter to conduct IDIs. Each of the interviews focused on Zimba's user perception, frequency of drinking from Zimba, perceptions (likes/dislikes and advantages/disadvantages/limitations) of treated water, and Zimba chlorination device, changes in taste and/or smell, and perception of treated water over the study period. Study participants were also asked their opinions on how the Zimba device could be improved.
Intervention components
Chlorine purchase, dilution, and dose adjustment
The sample collection team purchased household chlorine bleach (~5.25% NaOCl: Clorox®: the Clorox Company, Oakland, CA, USA) from the market near the Cox's Bazar field office. They diluted the raw chlorine in tap water (mean iron concentration <0.02 mg/L, mean turbidity 3.7= Nephelometric Turbidity Unit (NTU) and mean TDS=10 ppm) to 3.5% NaOCl to attain ~2 mg/L of free residual chlorine in household stored water within 24 hours of treatment, which they added to Zimba. This process of optimization was checked for each of the tubewells (Table S1) (23). The team tested the concentration of chlorine (NaOCl) solution with the colorimeter each time they refilled the Zimba. We subsequently reduced the NaOCl concentration to 2.85% for high iron content tubewell water and 2.25% for low iron content tubewell water to attain treated water with ~1.5 mg/L of free residual chlorine.
The Zimba device and its installation:
The descriptions of the Zimba and installation steps were reported previously for the urban Dhaka Zimba trial (22). In brief, the Zimba is made of fiberglass. It has three components:
- a dispenser containing diluted bleach (NaOCl)
- a dosing chamber containing a siphon
- an outer box that holds the siphon tank and the dispenser
The production cost of Zimba is estimated at around 200 USD, including installation. The Zimba does not have any mobile component and does not require any power source. The Zimba is mounted on a customized plastic stand approximately 30 cm in height (Figure S1). Two hand-pump mechanics from ACF installed the Zimba under the supervision of icddr,b. We also trained four ACF fieldworkers on regular chlorine refilling and maintenance of the Zimba so that it could be used after the intervention period.
Water storage containers:
During the first follow-up visit, the survey team provided a new 5 L plastic jerrycan of Zimba chlorinated water (mean residual chlorine at time of distribution=1.4 mg/L) with an airtight lid as a control. The day prior to collecting household stored treated water, the field research team refilled the jerrycan and provided it at the respondent's household. The field research team encouraged the household members to drink water from the jerrycan and to keep the lid close after use. The field research team also requested the respondents to retain some water (~one glass) in the jerrycan to be collected the next day. This was done to determine if the current household water storage practices impacted the chlorine treatment programme.
Intervention delivery
We used a similar approach for promoting the Zimba as the urban Dhaka trial (22). The day before Zimba installation, the survey team held a meeting with the camp Majhi (local FDMN leader), a local volunteer, and all study participants from the household and introduced water treatment with chlorine and potential health impacts. Zimba usage was explained using flip charts. The survey team requested that the Majhi and study participants share and discuss this information with other household members.
Tubewell and household stored water testing
The sample collection team tested the tubewell and stored water samples from all enrolled households for iron, using an iron meter: (Hach color disk test kit, Model IR-18B, Hack Company, Loveland, Colorado), TDS using a pocket TDS meter (Hanna instruments, Business Park Dr. Vista, California), turbidity, free and total chlorine level in collected water using a digital turbidity meter (LaMotte Model 2020i, LaMotte Company, Chestertown, Maryland) and Colorimeter (LaMotte Model 1200, LaMotte Company, Chestertown, Maryland). Using 100 mL sterile sample collection bags, the sample collection team collected tubewell and water samples stored in all selected households (Nasco Whirl-Pak®, 19 x 38 cm, Fort Atkinson, Wisconsin). Collected samples maintained at a temperature of <10°C through primary storage in cold boxes with ice packs were sent to the ACF field laboratory at Cox's Bazar for E. coli contamination assessment.
Water sample laboratory analysis
After measuring the physiochemical (iron, TDS, water turbidity, and free and total residual chlorine) property of each sample, the sample collection team collected four types of water samples: untreated tubewell water, pre-intervention water stored in the household vessel, treated Zimba water directly from the Zimba storage tank (immediately after treatment), post-treated water stored in the household vessel and project provided safe storage containers at households. Post-treated water samples were collected at two-time points: twice-weekly follow-up visits and during end-line surveys.
All samples were received within 6 hours of collection by the laboratory technician and quantified the most probable number (MPN) of E. coli per 100 mL of a water sample using the IDEXX- Quanti-tray® 2000 technique with Colilert-24 media (IDEXX Laboratories, Westbrook, Seattle, WA) (24). In recent water treatment efficacy trials, E. coli was most commonly used as an indicator of faecal microbial contamination in drinking water samples (21, 25, 26). All water samples were processed on the same day; the laboratory supervisor pre-tested two different dilutions [undiluted (1:1) and 1:10] of water samples to determine the ultimate dilution factor to minimize undetectable samples with E. coli or E. coli concentrations exceeding the upper detection limit (27). Due to the low water contamination level found during the initial laboratory analysis, all drinking water samples (both tubewell and household stored water) were analyzed without dilution. We used >1 to <=2419.6 MPN detectable range per tray to detect positive E. coli wells within the IDEXX Quanti-Tray (27).
During each day of sample collection, the fieldworkers collected one field blank of distilled water from the ACF laboratory and then tested on the same day with other water samples in the laboratory for E. coli, and if we found any growth in the field blank, we reinforced aseptic precautions for subsequent sample collection. The laboratory technician proceeds one laboratory blank per day and one negative control (distilled water) per batch of Colilert per day for quality control. Only 3% (1/30) of the tested blanks and 3% (1/30) negative control had E. coli growth. Finally, the laboratory technician processed and sealed the samples in a Quanti-Tray and incubated them at 37°C for 24 hours. The laboratory technician determined the MPN of E. coli by counting the number of fluorescing wells and calculating according to the manufacturer's instructions. We reported all water samples as MPN of E. coli/100 mL
Quantitative data analysis
We estimated the value of 0.5 MPN for samples below the detection limit and 2419.6 MPN for samples above the detection limit. We compared E. coli contamination and water parameters (iron, TDS, water turbidity, and free and total chlorine) between pre-treated water to post-treated household storage containers, using generalized linear regression models. We also compared levels of iron, free and total chlorine for baseline stored water vs. post-treated household storage containers, baseline stored water vs. post-treated project provided safe storage container, and post-treated household storage containers vs. post-treated project provided safe storage container and using generalized linear regression model, adjusted for clustering at the households with a robust standard error of the mean difference.
Qualitative data analysis
The survey team and local FDMN interpreters recorded interviews, downloaded, translated, and transcribed them into Bengali, followed by performed thematic content analysis. The investigator N.A. went through the translations and coded the transcripts according to our research objectives manually. After coding, the investigator thematically categorized the data and matched themes to factors influencing acceptability, feasibility, and chlorine smell and/or taste.