In ﬂ uence of container cleanliness, container disinfection with chlorine, and container handling on recontamination of water collected from a water kiosk in a Kenyan slum

The study assessed whether using clean containers that had been disinfected with chlorine at a water kiosk in the Kangemi slum in Nairobi reduced recontamination of treated water during drinking transport and storage. At the same time, the impacts of container handling and hygiene conditions at the household level on water quality changes during storage were evaluated. Data were collected during interviews with 135 households using either new, clean Maji Sa ﬁ containers (MSCs) that had been disinfected with chlorine or normal uncleaned jerrycans (NJCs). Bacteriological water quality and free chlorine levels in both types of containers were measured after container ﬁ lling at the kiosk and in the same containers after 24 h storage in households. The use of MSCs signi ﬁ cantly reduced the risk of recontaminating the treated water. After water ﬁ lling at the kiosk, none of the MSCs contained Escherichia coli bacteria, and 2.8% were contaminated after 24 h storage. In contrast, 6.2% of NJCs were contaminated after ﬁ lling, and 15.2% after 24 h storage. Multivariate logistic regression indicated that the use of a clean water container and suf ﬁ cient chlorine and the frequency of cleaning the container in the household mitigated recontamination. We suggest further investigation of water container designs that facilitate cleaning.


INTRODUCTION
) but is likely to be subject to recontamination and regrowth of pathogens during transport and storage. In their study on water kiosks in Ghana, Opryszko et al. () found that even though 91% of water samples at the tap of the kiosk met WHO Guidelines for drinking water quality, only 40% of samples collected at households had no detectable levels of Escherichia coli per 100 mL sample.
As documented by Wright et al. () in their metaanalysis, microbiological water quality often deteriorates during transport and storage. A number of other studies have investigated potential mechanisms and sources of recontamination and regrowth during transport and storage of drinking water: In Sub-Saharan Africa, Harris et al. () found that levels of fecal indicator bacteria increased immediately after storage containers were filled and water extracted from the container in the home. Certain extraction methods, such as decanting from the container and using a cup or ladle, were related to higher levels of fecal bacteria.
Deterioration immediately after filling collection containers has also been observed by other authors (Trevett et al. ; Meierhofer et al. ). Contact of hands and utensils with drinking water has been identified as an important source of contamination (Trevett et al. ; Pickering et al. ).
The design of the container has been found to have a significant impact on risks of recontaminating water during storage (Mintz et al. ; Reed et al. ). In observational studies, Mintz et al. () identified a wide container opening and water extraction with utensils and hands as contamination sources. They found that water in containers with a narrow neck, tightly fitting lid, and faucet was less contaminated during one month of use. Mellor et al. () similarly found higher levels of regrowth of total coliforms in containers with wide openings than in narrow-neck containers. In contrast, Levy et al. () did not find a statistically significant difference in water safety between containers with small openings (<8 cm) and containers with large openings (>8 cm). Roberts et al. () evaluated the impact of water containers with covers and spouts on recontamination during transport and storage in a refugee camp in Malawi and found an average of 53.3% fewer fecal coliforms in the improved buckets than in unimproved buckets without a cover and spout. The greatest difference between buckets was found at the time of water collection.
Children below the age of five in families using improved buckets had 31% less diarrhea, but the difference was statistically not significant. The cleanliness of containers used for the transport and storage of drinking water may also impact recontamination. This can be due to the amount of pathogens attached to container walls and the formation of biofilm, including AOC. Jagals et al. () analyzed the influence of biofilm attached to the walls of plastic water containers on water quality and found that counts of total coliforms and spores of Clostridium perfringens were significantly higher in water from containers containing biofilm. Murphy et al.
() quantified the biofilm in the storage containers of ceramic water filters and did not find a significant difference between containers that had or had not been cleaned in households, probably due to a small sample size, but the difference between containers undergoing controlled cleaning practice and those undergoing improper cleaning practice was significant. Mellor et al. () looked at contamination mechanisms during water storage at household level and found that hands and biofilm layers on containers' inner walls were important contamination sources.
The material of water containers has been suggested to have an impact on water quality changes during storage.
Several studies have found a higher chlorine demand for households using clay pots than those using plastic The goal of our study was to assess whether using clean containers that had been disinfected with chlorine at the water kiosk reduced recontamination risks during drinking water transport and storage in households in the Kangemi slum in Nairobi. The study also assessed the impact of container handling and hygiene conditions in households on water quality changes during storage.

Study design
The water quality in clean and disinfected MSCs was compared with the water quality in uncleaned, nondisinfected NJCs at the moment of filling the container and after 24 h of storage at household level. In addition to water quality tests, quantitative, structured household interviews took place with water buyers to determine the influence of water and container handling and hygiene conditions in the households on water contamination during storage.
Data were collected from 66 households using NJCs to collect drinking water from the water kiosk and from 69 households using MSCs. All customers visiting the kiosk during the data collection period were informed about the study. All households that provided informed consent and purchased water from the kiosk during the study period were included.
The study protocol was approved by the Ethics and Scientific Review Committee of the African Medical and

Research Foundation in Kenya and the Kenya National
Council of Science and Technology on 5 December 2014.
Households were identified by name, a mobile phone number, and their GPS location. The registration was crosschecked with the kiosk operators' customer list. Each household was involved only once. Group attribution was not randomized, but was formed on the basis of purchasing an MSC. An MSC promotion campaign was implemented at the onset of the study. The MSCs assessed during the study were new. They were washed and disinfected at the kiosk in accordance with the procedure described above and handed out to customers immediately after disinfection. Normal jerry cans were not cleaned at the kiosk.
Water quality at the kiosk was measured over 12 days.
Daily water samples were taken from the kiosk's tap after letting the water run for 3 seconds. A water sample was taken from each study participant's jerry can after it was filled at the kiosk. Interviewers filled the containers halfway and shook them for 10 seconds before taking the samples.
The container was then marked, and interviewers accompanied participants to their households. They instructed the people not to completely empty the water container but to leave a little water in it until the interviewer's visit the next day. After 24 hours, the households were revisited and interviewed and another water sample was taken from the marked container. All water samples were put in Nasco Whirl-Pak Thiobags with sodium thiosulfate.
The water samples were kept inside cooler bags for transport to the field lab, which was located at the water kiosk. Water quality analysis of E. coli, total coliforms, and the measurement of FRC was conducted immediately after the samples collected at the households arrived at the field lab (the average walking time between the households and the lab was 4.5 minutes).
The contamination levels of total coliforms and E. coli Control experiments were conducted with five NJCs and MSCs to evaluate recontamination and potential regrowth in both types of containers without handling by local households. New MSCs containers were obtained from the water kiosk, while used, uncleaned NJCs were purchased from local households.
Quantitative information was collected from households through face-to-face interviews with the person in the household responsible for drinking water management.
A structured questionnaire was used that incorporated closed, multiple choice questions mostly in categorical variables, Likert-scale answer categories, and some scale variables. The interviews were complemented by structured observations. The questionnaires were coded on tablets and contained questions on drinking water purchases (the time required to collect water from the kiosk, the frequency and volume of purchases per week), the use of containers for collecting water from different sources, the use and maintenance of containers for transporting and storing water, water treatment practices, the handling of water in the household, and hygiene indicators (type and cleanliness of handwashing station, type and cleanliness of the toilet, and frequency of handwashing).

Data analysis
Data were imported into SPSS for statistical analysis.
General drinking water purchase and use patterns, the use and maintenance of containers for drinking water transport and storage, contamination levels with E. coli and total coliforms, and FRC levels were analyzed using descriptive statistics. The distribution of log-transformed coliform counts was not normal; therefore, the significance of differences between the two groups was assessed with a Mann-Whitney test. The difference between other variables was assessed using the t-test for variables with equal variances or the Mann-Whitney test for variables that violated the assumption of normality. Counts of zero E. coli or zero total coliforms were replaced by 0.5 to allow logarithmic transformations. Effect sizes of the differences between groups were calculated using the formulas proposed by Rosenthal & Roow () and Rosnow et al. ().
To analyze the impact of water handling and hygiene conditions at household level on recontamination, a binary variable was formed with households with recontamination (E. coli or total coliforms 1 CFU/100 mL) and households without recontamination (E. coli or total coliforms ¼ 0 CFU/100 mL) between container filling and 24 hours of storage. Bivariate analysis between the outcome variable and various water handling and hygiene factors were calculated using Chi-square. The factors considered were the number of people in the household, the number of school-age children, the time required to collect water from the kiosk, the number of water purchases per week, the amount purchased, the use of additional sources of drinking water, whether the same container was used to collect water from different sources, whether the same container was used for the transport and storage of the drinking water, various materials used for the cleaning of containers, the number of times hands were washed per day with soap, the type of handwashing station used by the household, and the type and condition of the toilet used by the household.
Variables that had a significant relation with the outcome variable were further analyzed using a binary multivariate logistic regression model. Power calculations using G*Power 3.1 revealed that a sample size of 137 households detects an odds ratio of 0.1 with a power of 100% and an odds ratio of 0.4 with a power of 98% at a two-tailed alpha of 0.05 in logistic regression (Faul et al.

).
The concentration of FRC was normally distributed.
Linear regression was used to analyze the relation between FRC at the kiosk and FRC after filling the MSCs and NJCs.
Once more, t-tests were used to analyze the difference between the two groups. Spearman's rho was used to calculate the correlation between FRC and the log-transformed counts of E. coli and total coliforms after container filling and after 24 h of storage.

Water handling and hygiene conditions
Observations showed that the 69 households in the MSC group had slightly better sanitary infrastructure than the 66 households in the NJC group: 6% more households in the MSC group had a handwashing station, 11.5% more had a private latrine, and 15% more had access to toilets that looked clean.
According to the answers received during the interviews, the groups did not differ in the frequency of handwashing. While 70% of the NJC group used their container to collect water from other, potentially unsafe drinking water sources, only 40% of the MSC group did so. The practice of storing safe and unsafe water in the same container may lead to higher contamination of the insides of the containers used by NJC households.
Interviews also showed that the NJC group purchased significantly more water from the kiosk, but the MSC group perceived the quality of water to be better. Both groups cleaned their containers about once per week in the household. There was no difference between the groups in the frequency of cleaning their drinking water containers in the household, but in contrast to the MSC group, the NJC group did not have their containers disinfected at the kiosk.
Further details of water handling and hygiene conditions in both groups are presented in Table A Our findings suggest that the use of clean and disinfected MSCs reduced the risk of recontaminating treated water during transport and storage.
Bacterial contamination is presented in Table 1 as the percentage of water samples found in correspondence with WHO risk categories for E. coli in drinking water (WHO ).

The influence of water handling and hygiene conditions on water quality
Bivariate analysis revealed that most water handling and household hygiene factors were not significantly correlated with recontamination in the containers after 24 h of storage. Table 2 shows that the risk of recontaminating treated water during transport and storage was significantly reduced by sufficiently high levels of FRC after 24 hours of storage (OR ¼ 0.001, p ¼ 0.028), the use of a clean container (OR ¼ 0.14, p ¼ 0.035), and the frequency of cleaning the inside of the transport container in the household Except for the frequency of cleaning the inside of the water container in the household, none of the water handling, sanitation, and hygiene conditions in the households had a significant impact on recontamination during water storage. These findings are similar to results from a study conducted in an urban slum in Hyderabad, India, which found a 36% increase in fecal contamination in drinking water containers between point of supply and stored drinking water but failed to find a significant correlation between contamination and any household practice of water handling, hygiene, sanitation, or handwashing (Eshcol et al. ). This may indicate that bacterial transmission paths between the household environment and stored drinking water can vary greatly between households.
Container's influence on free residual chlorine FRC levels at the water kiosk tap varied between 0.1 and 0.6 mg/L with an average level of 0.4 mg/L (SD ¼ 0.17 mg/L).  In our study, which used water treated by ultrafiltration with a turbidity of less than 10 NTUs, mean FRC levels of

CONCLUSIONS
Our study showed that significantly lower levels of recontamination with E. coli and total coliforms were observed if clean containers that had been disinfected with chlorine at the kiosk before filling with treated water were used for water transport and storage. The use of contaminated containers led to a higher consumption of FRC than the use of clean MSCs. This suggests that contamination present in uncleaned containers, including biofilms attached to container walls, may have contributed to a higher chlorine demand in these containers.
The use of chlorine during the cleaning of water containers led to high variations in concentrations of FRC in water storage containers. To achieve a more consistent level of free chlorine in the containers we recommend chlorinating water at the point of distribution, instead of using it during the cleaning process.
In addition to using a clean container and providing residual disinfection, more frequent container cleaning in cleaning. We suggest the investigation of water storage containers that have a small opening or tap, preventing contact with contaminated utensils or hands when extracting water, but that also have an opening large enough to enable effective container cleaning.
Except for the cleaning of the container, none of the water handling or hygiene behaviors in households influenced stored water quality in this study. This indicates that the use of a clean water storage container, together with sufficiently high levels of FRC, may protect water from recontamination during 24 h of storage even if the water is stored in an environment with unfavorable hygienic conditions.