Water, sanitation and hygiene interventions for acute childhood diarrhea: a systematic review to provide estimates for the Lives Saved Tool

We systematically reviewed the published literature up to September 2016. We relied on a search that was previously conducted by our team for a broader evaluation of WASH interventions in September 2014 and updated that search in September 2016 to incorporate relevant new evidence. The search was conducted in Medline, CINAHL, EMBASE, CAB Abstracts, Cochrane, BLDS, EconLit, IDEAS, SIGLE, WHOLIS and JOLIS. Further articles from secondary sources were retrieved by screening the reference list of a Gapmap by Waddington and colleagues [11] and the reference lists of relevant reviews and reports [3–9]. A search strategy was designed including Medical Subject Heading Terms (MeSH) and keywords using various combinations. No language or date restrictions were employed in the electronic searches.

We initially screened, in duplicate, the titles and abstracts of retrieved articles to determine whether they met our inclusion and exclusion criteria. The full-texts of all selected studies were then retrieved and assessed by two reviewers for eligibility. In duplicate, we abstracted descriptive and quantitative data from included studies into a standardized form.

Two authors independently assessed study eligibility using pre-defined inclusion and exclusion criteria. Discrepancies between the reviewers in the decision to include or exclude studies were resolved by discussion aimed at reaching consensus or by consulting with a third author.

We limited included studies to randomized controlled trials (RCTs), cluster randomized controlled trials (cRCTs) and quasi-experimental (QE) trials where the following WASH interventions were evaluated in community settings in children 0–5 years old:

We included studies published in English that evaluated the impact of these interventions on acute childhood diarrhea in children 0–5 years old. Our outcomes of interest included diarrhea-related mortality, diarrhea-related morbidity and risk of diarrhea. We excluded studies reporting only behavioral outcomes. We excluded studies comparing the effect of different interventions without a control group; studies conducted in specific settings such as schools, daycares, and hospitals; studies where the intervention was the use of hand scrubs or disinfectants; studies measuring the impact on dysentery only, specific pathogens such as cholera or soil-transmitted helminths (STHs), or general gastrointestinal outcomes like highly-credible gastrointestinal illness (HCGI); studies conducted in emergency settings or refugee camps; or studies conducted only with specific populations such as HIV-infected persons. We also excluded studies where multiple interventions were evaluated together and the impact of a single intervention could not be inferred, or where the data were not reported sufficiently to be included in a meta-analysis.

The quality of studies was assessed using methods adapted from the Cochrane ‘Risk of bias’ assessment tool [12] and the Child Health Epidemiology Reference Group (CHERG) guidelines [13]. For each study, two reviewers independently assessed the quality of included studies for the following domains; allocation concealment, sequence generation; blinding of outcome assessors, blinding of participants and personnel, and incomplete outcome data. During quality assessment, RCTs and cRCTs started at a ‘high’ rating and quasi-experimental (QE) studies started at a ‘low’ rating with each study’s rating adjusted accordingly and given either high, moderate, low or very low scores. Where a study reported multiple outcomes, we assigned a separate overall study score for each, depending on how the outcome was measured.

We entered the abstracted effect estimates into Review Manager (RevMan) 5.3 and made calculations where necessary [12, 14]. In duplicate, the effect of the interventions on diarrheal outcomes was extracted, and calculated when necessary. These included risk ratios (RRs), odds ratios (ORs), rate ratios, means ratios, and longitudinal prevalence ratios, depending on how the individual study authors chose to display the effect. For treating all effect measures as equivalent, the design effect was considered for the various effect measures for common outcomes like diarrhea. The different measures of effect were then converted to a single measure for such outcomes [15]. In our analysis, ORs were transformed into RRs using an assumed control risk and formula recommended by Higgins et al. [12].

Where studies presented outcomes at different time points, we selected the effect estimate from the longest follow-up time. When studies provided effect estimates separated into different age strata of children 0–5 years old, we combined the point estimates from each stratum in RevMan using fixed effects models and then added the resulting pooled effect estimate into our main meta-analysis [16]. To quantitatively synthesize the available evidence, we grouped together similar intervention and outcome types and conducted meta-analyses using the generic inverse variance method. Random effects models were used to estimate the average effect of the intervention under the assumption that the intervention effects from individual studies were drawn from a distribution of effects rather than indicating the same fixed effect. For each intervention-outcome pair, the pooled RR was reported with a 95% confidence interval (CI). Subgroup analysis was conducted for the difference in the intervention.

After each study was assessed for methodological quality and assigned a rating according to the CHERG adaptation of the GRADE technique [13], the quality of the overall evidence for each intervention and outcome combination was assessed on a four-level scale (high, moderate, low, very low).

We identified five studies that provided water quality improvement intervention at the water supply [17‐21]; two studies were cRCTs and three were QE. All of these studies were conducted in low and middle-income (LMIC) settings and the interventions included improved supply systems, hand pumps, and water disinfection (chlorination). The combined analysis suggested no effect of water quality interventions at source on risk of diarrhea (pooled RR: 0.98 95%CI: 0.73, 1.32) and the subgroup analyses for the various interventions also suggested no effects (Fig. 2).

We identified 32 studies for inclusion in analysis that had a water quality improvement intervention at point-of-use [18, 21‐51]; 27 of these were RCTs or cRCTs while five were QE study designs. Studies were from Africa (Kenya, Ghana, Democratic Republic of the Congo, Ethiopia, Zimbabwe, South Africa), Asia (Bangladesh, Pakistan, India, Afghanistan, Saudi Arabia, Uzbekistan, Cambodia), South America (Bolivia, Brazil, Colombia), Central America (Honduras, Guatemala), and the Caribbean (Haiti, Dominican Republic). There were a range of interventions delivered which were broadly categorized into ‘water filtration’ [22‐34] and ‘water disinfection’ [18, 21, 35‐51] interventions. Water filtration interventions included biosand filters, ceramic filters, lifestraws, and hollow water filters while disinfection interventions included chlorination, use of flocculent-disinfectant, and solar disinfection (SODIS). One study reported the impact of flocculent-disinfectant on all-cause mortality in children under the age of two years and reported a 65% reduction (RR: 0.35, 95%CI: 0.13, 0.94) [37]. Overall, ‘water quality interventions at the point-of-use’ showed a significant decrease in risk of diarrhea by 40% (RR: 0.60, 95%CI: 0.53, 0.68), while the subgroup analyses suggested a 53% decrease (pooled RR: 0.47, 95% CI: 0.36, 0.62) with respect to water filtration and a 31% decrease (pooled RR: 0.69, 95% CI: 0.60, 0.79) with respect to water disinfection (Fig. 3). A further subgroup analysis suggested a significant effect for each of the specific interventions except for lifestraw (Fig. 4).

We identified six studies which evaluated the effect of handwashing with soap [52‐57]; four were cRCTs and two were QE study designs. All studies were conducted in South Asian countries. Study participants were provided soap with education about handwashing before eating or food handling, after defecation or handling of child stools, or a combination of these. No study reported on mortality and the analysis suggests that handwashing with soap leads to a 27% decrease in risk of diarrhea (pooled RR: 0.73, 95% CI: 0.57, 0.94) (Fig. 5).

The search for studies for excreta disposal interventions resulted in few studies with study designs that met our inclusion criteria, and some studies had other interventions including water supply interventions or multiple interventions evaluated together, hence the impact of excreta disposal alone could not be ascertained [58‐63]. One study was included which showed that latrine construction in India increased mean village-level latrine coverage from 9% of households to 63% in the intervention group, but there was no impact on the risk of diarrhea in children younger than 5 years (RR: 0.97, 95% CI: 0.83–1.12)[64].