Effects of water quality, sanitation, handwashing, and nutritional interventions on diarrhoea and child growth in rural Kenya: a cluster-randomised controlled trial

Background: Poor nutrition and exposure to faecal contamination are associated with diarrhoea and growth faltering, both of which have long-term consequences for child health. We aimed to assess whether water, sanitation, handwashing, and nutrition interventions reduced diarrhoea or growth faltering. Methods: The WASH Benefits cluster-randomised trial enrolled pregnant women from villages in rural Kenya and evaluated outcomes at 1 year and 2 years of follow-up. Geographically-adjacent clusters were block-randomised to active control (household visits to measure mid-upper-arm circumference), passive control (data collection only), or compound-level interventions including household visits to promote target behaviours: drinking chlorinated water (water); safe sanitation consisting of disposing faeces in an improved latrine (sanitation); handwashing with soap (handwashing); combined water, sanitation, and handwashing; counselling on appropriate maternal, infant, and young child feeding plus small-quantity lipid-based nutrient supplements from 6–24 months (nutrition); and combined water, sanitation, handwashing, and nutrition. Primary outcomes were caregiver-reported diarrhoea in the past 7 days and length-for-age Z score at year 2 in index children born to the enrolled pregnant women. Masking was not possible for data collection, but analyses were masked. Analysis was by intention to treat. This trial is registered with ClinicalTrials.gov, number NCT01704105. Findings: Between Nov 27, 2012, and May 21, 2014, 8246 women in 702 clusters were enrolled and randomly assigned an intervention or control group. 1919 women were assigned to the active control group; 938 to passive control; 904 to water; 892 to sanitation; 917 to handwashing; 912 to combined water, sanitation, and handwashing; 843 to nutrition; and 921 to combined water, sanitation, handwashing, and nutrition. Data on diarrhoea at year 1 or year 2 were available for 6494 children and data on length-for-age Z score in year 2 were available for 6583 children (86% of living children were measured at year 2). Adherence indicators for sanitation, handwashing, and nutrition were more than 70% at year 1, handwashing fell to less than 25% at year 2, and for water was less than 45% at year 1 and less than 25% at year 2; combined groups were comparable to single groups. None of the interventions reduced diarrhoea prevalence compared with the active control. Compared with active control (length-for-age Z score −1·54) children in nutrition and combined water, sanitation, handwashing, and nutrition were taller by year 2 (mean difference 0·13 [95% CI 0·01–0·25] in the nutrition group; 0·16 [0·05–0·27] in the combined water, sanitation, handwashing, and nutrition group). The individual water, sanitation, and handwashing groups, and combined water, sanitation, and handwashing group had no effect on linear growth. Interpretation: Behaviour change messaging combined with technologically simple interventions such as water treatment, household sanitation upgrades from unimproved to improved latrines, and handwashing stations did not reduce childhood diarrhoea or improve growth, even when adherence was at least as high as has been achieved by other programmes. Counselling and supplementation in the nutrition group and combined water, sanitation, handwashing, and nutrition interventions led to small growth benefits, but there was no advantage to integrating water, sanitation, and handwashing with nutrition. The interventions might have been more efficacious with higher adherence or in an environment with lower baseline sanitation coverage, especially in this context of high diarrhoea prevalence. Funding: Bill & Melinda Gates Foundation, United States Agency for International Development. The Kenya WASH Benefits study was a cluster-randomised trial done in rural villages in Bungoma, Kakamega, and Vihiga counties in Kenya’s western region (appendix p 11). We used a cluster design to facilitate the logistics of the behaviour change component of the interventions and minimise contamination between intervention and comparison households. We hypothesised that the interventions would improve the health of the index child in each household. We optimised the trial design to measure group-level differences in primary outcomes by including a large number of clusters, each comprising relatively few children (12 on average) with infrequent measurement. Each measurement round lasted roughly 1 year and was balanced across treatment groups and geography to minimise seasonal or geographic confounding when comparing outcomes across groups. With active and passive control groups and six intervention groups (water; sanitation; handwashing; combined water, sanitation, and handwashing; nutrition; and combined water, sanitation, handwashing, and nutrition), the design enabled 11 comparisons of each intervention group with the active control; combined water, sanitation, and handwashing with each intervention alone; and combined water, sanitation, handwashing, and nutrition with nutrition alone, and combined water, sanitation, and handwashing. A double-sized active control group was used to increase power because there were six separate intervention comparisons against control.10 Households in the active control and all intervention groups were visited by community-based health promoters monthly to measure the child’s mid-upper arm circumference. Health promoters did not visit households in passive control clusters. Measurement of outcomes, as well as water, sanitation, handwashing, and nutrition characteristics were measured in the passive control group at the same times as in other groups. The study design and rationale have been published previously.10 The study protocol was approved by the Committee for the Protection of Human Subjects at the University of California, Berkeley (protocol number 2011-09-3654), the institutional review board at Stanford University (IRB-23310), and the scientific and ethics review unit at the Kenya Medical Research Institute (protocol number SSC-2271). Under direction of the study investigators, Innovations for Poverty Action (IPA) was responsible for intervention delivery and data collection. Villages were eligible for selection into the study if they were rural, most of the population relied on communal water sources and had unimproved sanitation facilities, and there were no other ongoing water, sanitation, handwashing, or nutrition programmes. Participants were identified through a complete census of eligible villages. Within selected villages, women were eligible to participate if they reported that they were in their second or third trimester of pregnancy, planned to continue to live at their current residence for the next 2 years, and could speak Kiswahili, Luhya, or English well enough to respond to an interviewer administered survey. IPA staff formed clusters from one to three neighbouring villages to have six or more pregnant women per cluster after the enrolment survey. Outcomes were assessed in the children born from these pregnancies (index children), including twins. Although the study area is one of the areas with the highest HIV prevalence in Kenya, according to the 2012 Kenya AIDS Indicator Survey, the prevalence in women aged 15–64 years in the study area was below 6% (that survey did not include testing of children). Because there would not have been sufficient sample size to allow for subgroup analysis by HIV status, no attempt was made to identify participants who were HIV positive. Participants gave written informed consent before enrolment. Clusters were randomly allocated to treatment using a random number generator with reproducible seed at the University of California, Berkeley. Groups of nine geographically-adjacent clusters were block-randomised into a double-sized active control; passive control; water; sanitation; handwashing; water, sanitation, and handwashing; nutrition; or water, sanitation, handwashing, and nutrition. Allocation by cluster identification number was communicated directly to the field team; investigators remained blinded to treatment assignments. Blinding of participants was not possible. Participants were informed of their treatment assignment after baseline data collection and might have known the treatment assignment of nearby villages. The health promoters and staff who delivered the interventions were not involved in data collection, but the data collection team could have inferred treatment status if they saw intervention materials in study communities. The interventions were designed to maximise adherence to behaviours that could protect children from exposure to pathogens in their environment and improve diet quality. Formative research in the study area concluded that the health benefits of target behaviours were already well understood, but this knowledge was not sufficient to lead to action. As such, the behaviour change strategy and intervention materials were selected to create enabling environments, build supportive social norms, and target emotional drivers of decision making. The messages and delivery modes for the behaviour change strategy drew from existing information, education, and communication materials from organisations such as WHO, the Kenyan Government, UNICEF, and the Alive and Thrive network, and extensive previous qualitative work on the drivers of handwashing behaviours. Monthly visit modules were developed and pilot-tested to provide behavioural recommendations to mothers and other caregivers using key thematic constructs of convenience, nurturing care, and aspiration. We did a pilot randomised controlled trial11 to test the feasibility and acceptability of all the interventions and to collect data that allowed us to optimise the ratio of community-based promoters to study participants. To identify and correct systematic problems with adherence, staff confirmed that intervention materials were delivered to all study participants at the outset of the trial, and collected monitoring data on availability of intervention materials and recommended behaviours during unannounced visits to a random sample of at least 20% of participants in intervention groups 2, 6, 10, and 19 months after the interventions began. Community-based promoters for intervention and active control groups were nominated by study mothers and other mothers of children younger than 3 years in the community. A second promoter was added if there were more than ten participants (single groups) or more than eight participants (combined groups) in the cluster, giving a total of 1031 promoters. Promoters attended 2 days (active control), 6 days (single groups), or 7 days (combined groups) of initial training led by study staff on how to measure mid-upper-arm circumference, communication skills, intervention-specific behaviour change messages and intervention materials, and the information they were expected to report to IPA. Refresher trainings were done 6, 12, and 18 months after the initial training. At 2, 4, 9, 15, and 21 months, study staff met with promoters in their clusters to observe visits and offer supportive supervision. Study staff called promoters monthly to collect information on their activities, intervention adherence in the households they visited, referrals to health centres, and births or deaths of study children. Promoters received a branded T-shirt, a mobile phone, job aids and intervention materials, and compensation of approximately US$15 per month for the first 6 months when they had more intensive engagement with the study participants, and $9 per month thereafter (the prevailing daily wage for unskilled labour in the study area is $1–2). Promoters were instructed to visit all participants in their cluster monthly and measure the child’s arm circumference or the pregnant mother’s abdomen. In intervention groups, promoters engaged study participants and other compound members through interactive activities such as guided discussions using visual aids, song, and storytelling; resupplied consumable intervention materials; encouraged consistent practice of targeted behaviours; and helped troubleshoot barriers to adherence, including problems with intervention hardware and behavioural barriers. Promoters were provided with detailed plans for every visit, including key messages, scripts for discussing visual aids, and instructions for activities that emphasised the learning objectives. Visits lasted about 10 min in the active control group and 45–60 min in intervention groups during the first year when the key messages were conveyed. In the second year, promoters reinforced messages to maintain habits. All groups used messages on themes of nurture, aspiration, and self-efficacy, particularly in the context of a new birth. Interventions used convenience and social norms to encourage target behaviours. In the three intervention groups that included water, promoters advocated treatment of drinking water with sodium hypochlorite. Chlorine dispensers for convenient water treatment at the point of collection were installed at an average of five communal water sources in the cluster and refilled as needed. Every 6 months, households in study compounds were given a 1 L bottle of chlorine for point-of-use water treatment in case households collected rainwater or used a source without a dispenser. Promoters used chlorine test strips during their regular visits to determine if the household was using chlorine, and negative results stimulated conversation about addressing barriers to chlorination. In the three intervention groups that included sanitation, promoters advocated using latrines for defecation and safe disposal of children’s and animals’ faeces into a latrine. Existing unimproved latrines in study households were upgraded to improved latrines by installing a plastic slab, which also had a tight-fitting lid over the hole. New latrines were constructed for study households that did not have a latrine or whose latrine was unlikely to last for 2 years. All households in study compounds received a sani-scoop with a paddle as a dedicated faeces-removal tool. Finally, all households with children younger than 3 years in study compounds received plastic potties to facilitate toilet training and transfer of child faeces to the latrine. In the three intervention groups that included handwashing, promoters advocated handwashing with soap before handling food and after defecation (including assisting a child). Study compounds were given two permanent, water-frugal handwashing stations intended to be installed near the food preparation area and the latrine. Handwashing stations were constructed of painted metal, with two foot-pedal-operated jerry-cans that dispensed a light flow of rinse water and soapy water. Promoters added chunks of bar soap to the soapy water container quarterly. In the two intervention groups that included nutrition, a set of ten age-targeted modules were developed to enable promoters to advocate for best practices in maternal, infant, and young child feeding: recommendations for dietary diversity during pregnancy and lactation, early initiation of breastfeeding, exclusive breastfeeding until 6 months, introduction of appropriate and diverse complementary foods at 6 months, and continued breastfeeding through 24 months. Facilitators and barriers to behaviour change were elicited using formative research and health promoter guides were developed to address common barriers and questions. Study mothers with children between 6–24 months were provided with two 10 g sachets per day of a small quantity of lipid-based nutrient supplement (LNS; Nutriset; Malauny, France) that could be mixed into the child’s food. LNS provided 118 kcal per day and 12 essential vitamins and ten minerals. Promoters explained that LNS was not to replace breastfeeding or complementary foods. Promoters and intervention materials were introduced at community meetings roughly 6 weeks after enrolment. All interventions were delivered within 3 months of enrolment (appendix p 1). LNS was introduced to each child when they turned 6 months old. All handwashing stations and latrines were inspected within a month of construction, and a subset of households was periodically visited to observe group-specific indicators of intervention adherence. These data alerted study investigators to any issues with intervention implementation so they could be addressed consistently across all clusters and groups. The enrolment survey included baseline demographics; assets; water, sanitation, and handwashing infrastructure; and target behaviours. Follow-up at 1 year and 2 years after intervention delivery consisted of an unannounced visit to study compounds to observe objective indicators of target behaviours (in all groups other than the passive control) and, on the following day, growth and health outcome measurements at a central location in the cluster (eg, a church or school). Children identified as possibly malnourished (mid-upper-arm circumference <11·5 cm), either by the promoter during routine visits or by study staff during follow-up measurements, were referred to health facilities for treatment. Adherence to the interventions was assessed using objective, observable indicators where possible (appendix pp 2, 3). We calculated Z scores for length for age, weight for length, weight for age, and head circumference for age using the WHO 2006 child growth standards. All child deaths reported by the health promoters were confirmed by a staff nurse who visited households. All outcomes were prespecified. Primary outcomes were caregiver-reported diarrhoea in the past 7 days (based on all data from year 1 and year 2) and length-for-age Z score at year 2 in index children. Secondary and tertiary outcomes reported in this paper are length-for-age Z score at year 1; weight-for-length Z score, weight-for-age Z score, head circumference-for-age Z score at year 1 and year 2; prevalence of stunting (length-for-age Z score less than −2), severe stunting (length-for-age Z score less than −3), wasting (weight-for-length Z score less than −2), and underweight (weight-for-age Z score less than −2); and all-cause mortality. We excluded children from Z-score analyses if their measurements were outside biologically plausible ranges following WHO recommendations. More details on exclusion criteria, measurement protocols, and outcome definitions are in the appendix (p 1). Sample size calculations for the two primary outcomes were based on a minimum detectable effect of 0·15 in length-for-age Z score (intracluster correlation of 0·02 in our pilot study) and a relative risk of diarrhoea of 0·7 or smaller (assuming a 7-day prevalence of 12% in the active control group based on a pilot study to inform this trial) for a comparison of any intervention with the double-sized control group, assuming a type I error (α) of 0·05 and power (1–β) of 0·8, a one-sided test for a two-sample comparison of means, and 10% loss to follow-up.10, 11 Sample size calculations indicated 80 clusters per group, each with ten children. Two biostatisticians, blinded to treatment assignment, independently replicated the analyses following the prespecified analysis plan with minor updates.10 We analysed participants according to their randomised assignment (intention to treat), regardless of adherence to the intervention, using the active control group as the comparator. We used paired t tests for unadjusted length-for-age Z score comparisons and the Mantel-Haenszel prevalence ratio and difference for unadjusted diarrhoea and stunting comparisons, with randomisation block defining matched pairs or stratification. In secondary analyses, we estimated prevalence ratios and differences, adjusting for baseline covariates using targeted maximum likelihood estimation.12 Analyses were done in R (version 3.2.3). We tested for the presence of between-cluster spillover effects using a non-parametric method described in the prespecified analysis plan, which tested whether primary outcomes were the same in control households with more versus fewer households receiving interventions within a 2 km radius. In an analysis that was not prespecified, we tested for intervention effects on diarrhoea using only year 1 data. The trial is registered at ClinicalTrials.gov, number {"type":"clinical-trial","attrs":{"text":"NCT01704105","term_id":"NCT01704105"}}NCT01704105. IPA convened a data and safety monitoring board. The funders of the study approved the study design, but had no role in data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.