Interventions to improve water quality for preventing diarrhoea

Summary of findings for the main comparison. Summary of findings table 1

Point‐of‐use water quality interventions for preventing diarrhoea in rural settings in low‐ and middle‐income countries
Patient or population: adults and children Settings: low‐ and middle‐income countries in rural areas Intervention: point of use water quality interventions Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (trials)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
Diarrhoea episodes	No intervention	Chlorination	RR 0.77 (0.65 to 0.91)	30,746 (14 trials)	⊕⊕⊝⊝ low^1,2,3,4
	3 episodes per person per year	2.3 episodes (2.0 to 2.7)
	No intervention	Flocculation/disinfection	RR 0.69 (0.58 to 0.82)	11,788 (4 trials)	⊕⊕⊕⊝ moderate^1,3,4,5,6
	3 episodes per person per year	2.1 episodes (1.7 to 2.5)
	No intervention	Filtration	RR 0.48 (0.38 to 0.59)	15,582 (18 trials)	⊕⊕⊕⊝ moderate^1,3,4,5
	3 episodes per person per year	1.4 episodes (1.1 to 1.8)
	No intervention	Solar disinfection (SODIS)	RR 0.62 (0.42 to 0.94)	3460 (4 trials)	⊕⊕⊕⊝ moderate^1,3,4,5
	3 episodes per person per year	1.9 episodes (1.3 to 2.8)
The assumed risk is taken from Fischer Walker 2012 and represents an estimated average for the incidence of diarrhoea in low‐ and middle‐income countries. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
¹Downgraded by 1 for serious risk of bias: the outcome was measured as self‐reported episodes of diarrhoea, and is susceptible to bias as most studies were unblinded. ²Downgraded by 1 for serious inconsistency: statistical heterogeneity was very high with six out of fourteen trials having point estimates close to no effect. A subgroup analysis by adherence with the intervention (assessed by measurements of residual chlorine in drinking water) found larger effects in the studies with better adherence but the results remained inconsistent. ³No serious indirectness: these studies are mainly from low‐ and middle‐income countries, in settings with both improved and unimproved water sources and sanitation. ⁴No serious imprecision: The analysis is adequately powered to detect this effect. ⁵No serious inconsistency: The evidence of benefit is consistent across trials, but there is substantial statistical heterogeneity in the size of the effect. ⁶ This analysis excludes one additional study which found a much larger effect than seen in the other four trials and was considered an outlier (Doocy 2006 LBR).

Background

Description of the condition

Diarrhoeal disease is the third leading cause of mortality in low‐income countries, causing an estimated 1.4 million deaths in 2012 (WHO 2014;GBD 2015). Young children are especially vulnerable, with diarrhoea accounting for more than a quarter of all deaths in children aged under five years in Africa and Southeast Asia (Murray 2012; Lanata 2013; Walker 2013).

The bacterial, viral, and protozoan pathogens causing diarrhoeal disease are primarily transmitted via the faecal‐oral route, through the consumption of faecally contaminated food and water (Byers 2001). Among the most important of these are rotavirus, Cryptosporidium sp.,Escherichia coli,Salmonella sp.,Shigella sp.,Campylobacter jejuni,Vibrio cholerae, norovirus, Giardia lamblia, and Entamoeba histolytica (Leclerc 2002; Kotloff 2013), though the relative importance of these varies among settings, seasons, and population groups.

An estimated 1.1 billion people worldwide rely on water supplies that are at high risk of faecal contamination (Bain 2014). Moreover, nearly half the world's population lack household water connections (WHO/UNICEF 2015), and are at increased risk of unsafe water due to contamination during collection, storage, and use in the home (Wright 2004).

Description of the intervention

Interventions to improve the microbiological quality of water can be grouped into four main categories:

Physical removal of pathogens (for example, filtration, adsorption, or sedimentation).
Chemical treatment to kill or deactivate pathogens (most commonly with chlorine).
Disinfection by heat (for example, boiling or pasturization) or ultraviolet (UV) radiation (for example, solar disinfection, or artificial UV lamps).
Combination of these approaches (for example, filtration or flocculation combined with disinfection).

In higher‐income countries, and in many urban settings worldwide, drinking water is treated centrally at the source of supply and distributed to consumers through a network of pipes and household taps. Alternatively, water may be treated at any point in the distribution network, or at the 'point‐of‐use' (POU) in people's homes, schools, or workplaces.

In remote and low‐income settings, source‐based water quality improvement may include providing protected groundwater (springs, wells, and bore holes) or harvested rainwater as an alternative to surface sources (rivers and lakes). These improvements frequently also improve both the quantity and access to water by increasing the volume or frequency of water delivery or reducing the time spent in collecting water. This may result in significant benefits not only in health but also in economic and social welfare (Hutton 2013; Stelmach 2015).

Potential and widely used POU interventions for remote or low‐income settings include boiling, filtration, chlorination, flocculation, and solar disinfection. These interventions have the potential to overcome both contaminated sources and recontamination of safe water in the home (Wright 2004). A review commissioned by the World Health Organization (WHO) identified a wide variety of options for household‐based water treatment and assessed the available evidence on their microbiological effectiveness, health impact, acceptability, affordability, sustainability, and scalability (Sobsey 2002).

How the intervention might work

Health authorities generally accept that microbiologically safe water plays an important role in preventing outbreaks of waterborne diseases (Reynolds 2008). Moreover, there is evidence that chlorination and filtration of municipal water supplies contributed to substantial health gains in the late 19^th and early 20^th century (Cutler 2005).

However, much of the epidemiological evidence for increased health benefits following improvements in the quality of drinking water has been equivocal, particularly in low‐income settings (Clasen 2006; Waddington 2009; Cairncross 2010).

This may be due to the variety of alternative transmission pathways, such as ingestion of contaminated food, person‐to‐person contact, or direct contact with infected faeces. In addition, interventions which only target the home may fail if unsafe water is consumed at work or school. Consequently, effective programmes may require combined interventions to address not only water quality, but also water quantity and access, the proper disposal of human faeces (sanitation), and the promotion of hand washing and hygiene practices within communities.

The effectiveness of individual water quality interventions may also vary between settings due to the varied prevalence of the organisms causing diarrhoea. For instance, ceramic filters are only marginally protective against viral illness, while chlorination may provide little protection against Cryptosporidium.

Why it is important to do this review

This is an update of a Cochrane Review that was first completed in 2006 (Clasen 2006). The review concluded that, in general, interventions to improve microbiological quality of drinking water are effective in preventing diarrhoea, and that interventions at the household level were more effective than those at the source.

New studies have been recently published, and other unpublished studies have been made available to us. In this Cochrane Review update, we have reapplied the inclusion criteria, repeated data extraction, added new studies, and used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach to assess the quality of the evidence. We were also able to apply statistical methods to unify the measures of effect and to apply additional criteria for subgrouping based on study design, setting, and length of follow‐up.

Objectives

To assess the effectiveness of interventions to improve water quality for preventing diarrhoea.

Methods

Criteria for considering studies for this review

Types of studies

Cluster‐randomized controlled trials (cluster‐RCTs), quasi‐randomized controlled trials (quasi‐RCTs) and controlled before‐and‐after studies (CBAs).

Types of participants

Children and adults.

Types of interventions

Intervention

Any intervention aimed at improving the microbiological quality of drinking water.

We included interventions that combined improvements in water quality with hygiene or health promotion, but excluded studies that combined water quality interventions with other water, sanitation, and hygiene (WASH) interventions, such as improvements in water quantity or sanitation. We also excluded studies where the water quality intervention was implemented away from the home, such as at schools, clinics, markets, or workplaces.

Control

No intervention, or a dummy intervention.

Types of outcome measures

Primary

Diarrhoea episodes among individuals, whether or not confirmed by microbiological examination.

The WHO's definition of diarrhoea is three or more loose or fluid stools (that take the shape of the container) in a 24‐hour period (WHO 1993). We defined diarrhoea and an episode in accordance with the case definitions used in each trial. In the 'Summary of findings' tables, we have converted the results to episodes per year from a baseline of three episodes/child year in 2010 (Fischer Walker 2012).

Secondary

Death.
Adverse events.

We excluded studies that had no clinical outcomes; for example, studies that only report on microbiological pathogens in the stool.

Search methods for identification of studies

We attempted to identify all relevant studies regardless of language or publication status (published, unpublished, in press, and in progress).

Electronic searches

We searched the following databases using the search terms and strategy described in Appendix 1: Cochrane Infectious Diseases Group Specialized Register (11 November 2014); Cochrane Central Register of Controlled Trials (CENTRAL), published in the Cochrane Library (7 November, 2014); MEDLINE (1966 to 10 November 2014); EMBASE (1974 to 10 November 2014); and LILACS (1982 to 7 November 2014).

Searching other resources

Conference proceedings

We searched the conference proceedings of the following organizations for relevant abstracts: International Water Association (IWA) (1990 to 11 November 2014); and Water, Engineering and Development Centre, Loughborough University, UK (WEDC) (1973 to 11 November 2014).

Researchers and organizations

We contacted individual researchers working in the field and the following organizations for unpublished and ongoing studies: Water, Sanitation and Health Programme of the WHO; World Bank Water and Sanitation Program; UNICEF Water, Sanitation and Hygiene; and IRC International Water and Sanitation Centre; Foodborne and Diarrhoeal Diseases Branch, Division of Bacterial and Mycotic Diseases, Centers for Disease Control and Prevention (CDC); US Agency for International Development (USAID), including its Environmental Health Project (EHP); and the UK Department for International Development (DFID).

Reference lists

We checked the reference lists of all studies identified by the above methods.

Data collection and analysis

Selection of studies

Two review authors (RP and SB) independently reviewed the titles and abstracts located in the searches and selected all potentially relevant studies. After obtaining the full‐text articles, they independently determined whether they met the inclusion criteria. Where they were unable to agree, they consulted a third review author (TFC) and arrived at a consensus. We have listed the potentially relevant studies that were ultimately excluded together with the reasons for exclusion in the 'Characteristics of excluded studies' section.

Data extraction and management

Two review authors (RP and SB) used a pre‐piloted form to extract and record the data described in Appendix 2. One review author entered the extracted data into Review Manager (RevMan) (KA).

Assessment of risk of bias in included studies

Two review authors (KA and FM) independently assessed the risk of bias of the included studies and resolved differences of opinion through discussion.

For cluster‐RCTs we used the Cochrane 'Risk of bias' assessment tool (Higgins 2011). We followed the guidance to assess whether adequate steps were taken to reduce the risk of bias across five domains: sequence generation; allocation concealment; blinding of participants and personnel; blinding of outcome assessors; and incomplete outcome data.

For sequence generation and allocation concealment, we reported the methods used. For blinding, we described who was blinded and the blinding method. For incomplete outcome data, we reported the percentage and proportion of participants lost to follow‐up. For selective outcome reporting, any discrepancies between the methods used and the results were stated in terms of the outcomes measured or the outcomes reported. For other biases, we described any other trial features that could have affected the trial result (for example, if the trial was stopped early).

We categorized our 'Risk of bias' judgements as 'low', 'high', or 'unclear'. Where risk of bias was unclear, we attempted to contact the study authors for clarification and we resolved any differences of opinion through discussion. We classified the inclusion of randomized participants in the analysis as 'low risk' if 90% or more of all participants randomized to the study were included in the analysis.

For quasi‐RCTs and CBA studies, we used two additional criteria:

Comparability of baseline characteristics: we classified studies as 'low risk' if there were no substantial differences between groups with respect to water quality, diarrhoeal morbidity, age, socioeconomic status, access to water, hygiene practices, and sanitation facilities.
Contemporaneous data collection: we classified studies as 'low risk' if data were collected at similar points in time, 'unclear' if the relative timing was not reported or not clear from trial, or 'high risk' if data were not collected at similar points in time.

Measures of treatment effect

Two review authors independently extracted and, where necessary, calculated the measure of effect of the intervention on diarrhoea. We extracted the measure of effect as reported by the authors of each study, whether it be risk ratios (RRs), rate ratios, odds ratios (ORs), longitudinal prevalence ratios, or means ratios. In using these various measures of effect, we noted the design effect in treating all such measures of effect as equivalent for common outcomes such as diarrhoea and the debate about methodologies for converting such measures of effect into a single measure (Zhang 1998; McNutt 2003).

For purposes of analysis, we transformed ORs into RRs using the assumed control risk and the formula prescribed in Higgins 2011 (Section 12.5.4.4).

Unit of analysis issues

A number of the included studies had multiple intervention arms (for example, treating water with bleach or with a flocculant and disinfectant) and compared two or more intervention groups against a single control group. In some analyses, we included multiple comparisons from the same study, which double counts the control group participants and yields results in the meta‐analysis that are artificially precise. Unfortunately, because of the way data was presented in included studies, it was not possible to correct for this error by dividing the control group participants between multiple groups.

Dealing with missing data

When data was missing or incomplete we attempted to contact the study authors.

Assessment of heterogeneity

We assessed the statistical heterogeneity between trials by visually examining the forest plots for overlapping confidence intervals (CIs), applying the Chi² test with a 10% level of statistical significance, and using the I² statistic with a value of 50% to denote moderate levels of heterogeneity.

Assessment of reporting biases

When there were sufficient studies, we assessed the possibility of publication bias by constructing funnel plots and looking for asymmetry.

Data synthesis

We entered the estimates of effect using the generic inverse variance method on the log scale (Higgins 2006), and analysed the data using Review Manager (RevMan).

We stratified our primary analysis by intervention type, and study design (cluster‐RCT, quasi‐RCT, or CBA). When appropriate we used meta‐analyses to derive pooled estimates of effect using a random‐effects model because of the substantial heterogeneity in study settings, interventions, and outcome measures.

We summarized the evidence using 'Summary of findings' tables that we created using the GRADE Guideline Development Tool (GRADEpro GDT). The quality of evidence was rated using the GRADE approach, which consists of five factors that are used to assess the quality of the evidence: study limitations (risk of bias), inconsistency, indirectness, imprecision, and publication bias (Guyatt 2008).

Subgroup analysis and investigation of heterogeneity

We investigated the potential causes of heterogeneity by conducting the following subgroup analyses: age (all ages versus children under five years old); adherence with intervention (< 50%, 50% to 85%, > 85%); water source; water access; water quantity; sanitation conditions; country income level; and length of follow‐up.

In the subgroup analyses based on water source, we followed terminology used by the WHO/UNICEF Joint Monitoring Programme (JMP) on Water and Sanitation (WHO/UNICEF 2015), using 'unimproved' to extend to unprotected wells or springs, vendor‐ or tanker‐provided water or bottled water, and 'improved' to extend to household connections, public standpipes, boreholes, protected dug wells or springs, or rainwater collection; we categorized studies as 'unclear' with respect to water supply if they contained insufficient information.

We used the same definitions from the WHO/UNICEF JMP criteria to classify sanitation conditions as 'improved' (connection to a public sewer or septic system, pour‐flush latrine, simple pit latrine, ventilated improved pit latrine) or 'unimproved' (service or bucket latrines, public latrines, open latrines); where the necessary information was unclear or unreported, we categorized the sanitation facilities as 'unclear'.

To subgroup studies based on access to water source, we used the classifications defined by the Sphere Project 2011, classifying access as 'sufficient' if a consistently available source was located within 500 m, with queuing no more than 15 minutes and filling time for a 20 L container no more than three minutes, 'insufficient' if any access failed any such criteria, and 'unclear' if such criteria was unreported or unclear.

The quantity of water available to study participants was considered 'sufficient' if consisting of a minimum of 15 L per person per day. For country income level, we used the World Bank classification of country income levels (high, upper middle, lower middle, low) (World Bank Country and Lending Groups).

Sensitivity analysis

We conducted a sensitivity analysis to investigate the robustness of the results to each of the 'Risk of bias' components by including only studies that were at low risk of bias. We used this information to guide our judgements on the quality of the evidence.

In addition, we explored the impact of non‐blinding of POU interventions using a Bayesian meta‐analysis with bias correction. For this purpose, we assumed the true log relative risks from non‐blinding studies are subject to a multiplicative bias that results in the observed relative risks being inflated in magnitude. We assumed the bias is normally distributed with a mean 1.48 or 1.65 and a corresponding standard deviation (SD) of 0.17 or 0.13. These values were derived from the additive bias correction employed in Wood 2008 and Savović 2012. While we believe an attempt to adjust for non‐blinding is appropriate, we urge caution in relying on these adjusted estimates since the basis for the adjustment is from clinical (mainly drug) studies that may not be transferable to field studies of environmental interventions and because methodology for the adjustment has not been validated.

Results

Description of studies

Results of the search

The search strategy identified 1088 titles and abstracts, 1076 from the databases and 12 from the other sources (Figure 1). We screened these titles and abstracts, and obtained the full‐text articles of 161 studies for further assessment.

Figure 1

Study flow diagram.

Included studies

Fifty‐five studies, including 84,023 participants, met the inclusion criteria (see Characteristics of included studies). Of these, six studies had two relevant intervention arms (Austin 1993; URL 1995; Luby 2004; Crump 2005; Brown 2008; Lindquist 2014), two had three arms (Luby 2006; Opryszko 2010), and one had four arms (Reller 2003), making a total of 65 discrete comparisons. Three included studies had inadequate information on disease morbidity to include in the quantitative analysis (Torun 1982 GTM; Kremer 2011 KEN; Patel 2012 KEN). We contacted the study authors for further information, but no data could be provided. Therefore we have only described these three studies and their results, but have not integrated these studies into the analysis.

Study design and length

Forty‐five studies were cluster‐RCTs, two were quasi‐RCTs, and eight were CBA studies. Most included cluster‐RCTs used households as the unit of randomization, though some used neighbourhoods, villages, or communities. Most CBA studies used villages or communities as the unit of allocation. The intervention period ranged from eight weeks to four years. The duration of the cluster‐RCTs (median seven months, range 9.5 weeks to 18 months) tended to be shorter than in the CBA studies (median 12 months, range two to 60 months). Studies of source‐based interventions were also longer (median 24 months, range eight weeks to two years) than those of POU interventions (median six months, range 9.5 weeks to 17 months).

Participants and settings

Nine studies included data only for children under five years of age, and three studies included data only on adults. The other studies enrolled and presented results for all ages of participants.

Most studies were undertaken in lower middle or low‐income countries based on World Bank criteria, but three studies were conducted in the USA (Colford 2002 USA; Colford 2005 USA; Colford 2009 USA), one in Australia (Rodrigo 2011 AUS), and one in Saudi Arabia (Mahfouz 1995 KSA). Five studies were conducted in urban settings (Semenza 1998 UZB; Colford 2002 USA; Colford 2005 USA; Colford 2009 USA; Rodrigo 2011 AUS), five in peri‐urban settings (Quick 1999 BOL; Quick 2002 ZMB; du Preez 2010 ZAF; Jain 2010 GHA; Peletz 2012 ZMB), two in urban informal or squatter settlements (Handzel 1998 BGD; Luby 2004), two in camps for refugees or displaced persons (Roberts 2001 MWI; Doocy 2006 LBR), five in multiple settings (URL 1995; Clasen 2005 COL; Stauber 2009 DOM; du Preez 2011 KEN; Boisson 2013 IND), and the others in villages or other rural settings.

Primary drinking water supply and sanitation facilities

The primary drinking water supply before the intervention was 'unimproved' in 30 studies, 'improved' in 15 studies, and 'unclear' or unreported in five studies. Sanitation facilities in trial settings were 'improved' in 12 studies, 'unimproved' in 15 studies, and 'unclear' or unreported in 19 studies. Access to a water source was deemed 'sufficient' in 14 studies, 'insufficient' in four studies, and 'unclear' or unreported in the remaining studies. The quantity of water available to study participants was considered 'sufficient' in eight studies, 'insufficient' in four studies, and 'unclear' in 43 studies.

Seventeen studies measured water quality before the introduction of the intervention as an indication of the ambient risk and the microbiological quality of the water consumed by the control group. Details on the indicators used varied among the studies (see Table 1). Thirty‐five studies measured colony‐forming units (CFUs) of thermotolerant coliforms, faecal coliforms, or E. coli, reporting geometric means, arithmetic means, number of CFUs/100 mL, mean faecal coliforms/100 mL, E. coli most probable number, median, or log₁₀CFUs/100 mL. Other studies measured the frequency of samples containing such bacteria, or the CFU of total coliforms or other indicators of microbial contamination. None continually measured the microbiological performance of their interventions against the full range of bacterial, viral, and protozoan pathogens known to cause diarrhoea.

See: Summary of findings for the main comparison Summary of findings table 1

Table 1. Water quality indicators post‐intervention

Trial	Water quality indicator	Water quality post‐intervention: Intervention group	Water quality post intervention: Control group
Abebe 2014 ZAF	CFUs/100 mL	0	80% of control HHs had 10 to 10000
Austin 1993a GMB	Geometric mean CFUs/100 mL	178	3020
Austin 1993b GMB	Geometric mean CFUs/100 mL	42	3020
Boisson 2009 ETH	Arithmetic mean TTC/100 mL (95% CI)	0	725.7 (621.0 to 830.4)
Boisson 2010 DRC	Geometric mean TTC/100 mL (95% CI)	1.3 (0.9 to 1.7)	173.7 (136.6 to 220.9)
Boisson 2013 IND	Geometric mean TTC/100 mL (95% CI)	50 (44 to 57)	122 (107 to 139)
Brown 2008a KHM	Geometric mean E. coli /100 mL	17	600
Brown 2008b KHM	Geometric mean E. coli /100 mL	15	600
Clasen 2004b BOL	Mean TTC/100 mL	0.13	108
Clasen 2004c BOL	Arithmetic mean TTC/100 mL	100% of intervention households: 0	16% of control households: 0 66% > 10, 34% > 100, and 11% > 1000
Clasen 2005 COL	Arithmetic mean TTC/100 mL (95% CI)	37.3 (6.3 to 48.3)	150.6 (34.8 to 166.4)
Colford 2002 USA ; Colford 2005 USA ; Colford 2009 USA	All water met FDA requirements	Not measured because of high water quality	Not measured because of high water quality
Crump 2005a KEN	Samples met WHO guidelines for water quality	82%	14%
Crump 2005b KEN	Samples met WHO guidelines for water quality	78%	14%
du Preez 2008 ZAF/ZWE	Samples met WHO guidelines for water quality	57%	30%
du Preez 2010 ZAF	E. coli in concentrations/100 mL	62%	"No significant difference between intervention and control groups"
du Preez 2011 KEN	E. coli ln concentrations/100 mL	Storage containers: 0.723 SODIS bottles: ‐0.727	Not reported
Fabiszewski 2012 HND	Geometric mean E. coli counts per 100 mL (95% CI)	23.4 (20.2 to 27.0)	45.4 (38.6 to 53.4)
Gasana 2002 RWA	Total coliforms/100 mL	Range: 3 to 43	Range: 4 to 1100
Gruber 2013 MEX	Samples with detectableE. coli	43%	59%
Günther 2013 BEN	E. coli contamination > 1000 CFU/100 mL	Not reported specifically; findings imply a 70% reduction in E. coli incidence for intervention households
Handzel 1998 BGD	Stored water samples with E. coli 100 MPN/100 mL	3%	16%
Jain 2010 GHA	Samples with E. coli	8%	54%
Jensen 2003 PAK	Geometric mean E. coli /100 mL	3	49
Kirchhoff 1985 BRA	Mean number of faecal coliforms/dL in the samples	70	16000
Kremer 2011 KEN	Average reduction in log E. coli	‐1.07, corresponding to a 66% reduction
Lule 2005 UGA	Median E. coli CFU/100 mL	23	59
McGuigan 2011 KHM	Geometric mean CFU/100 mL	6.8	48
Mengistie 2013 ETH	Mean E. coli	0	60
Peletz 2012 ZMB	Geometric mean TTC/100 mL	Stored water: 3	Stored water: 181
Quick 1999 BOL	Median E. coli /100 mL	0	6400
Quick 2002 ZMB	Median E. coli /100 mL	0	3
Reller 2003a GTM	Samples with < 1 E. coli /100 mL (flocculant/disinfectant)	40%	7%
Reller 2003b GTM	Samples with < 1 E. coli /100 mL (flocculant/disinfectant+ vessel)	57%	7%
Reller 2003c GTM	Samples with < 1 E. coli /100 mL (bleach)	51%	7%
Reller 2003d GTM	Samples with < 1 E. coli /100 mL (bleach + vessel)	61%	7%
Semenza 1998 UZB	Faecal colonies/100 mL	47	52
Stauber 2009 DOM	E. coli MPN/100 mL	11	19
Stauber 2012a KHM	E. coli CFU/100 mL	2.9	19.7
Stauber 2012b GHA	Geometric mean E. coli MPN/100 mL (95% CI)	Direct filtrate 16 (13 to 20) Stored filtrate: 76 (62 to 91)	490 (426 to 549)
Tiwari 2009 KEN	Geometric mean faecal coliforms/100 mL (95% CI)	30.0 (21.3 to 42.1)	88.9 (58.7 to 135)
URL 1995a GTM	Samples with fecal coliforms	91% had 0 fecal coliforms	Not reported
URL 1995b GTM	Samples with fecal coliforms	91% had 0 fecal coliforms	Not reported

Abbreviations: E. coli: Escherichia coli; FC: faecal coliform.

Eight studies did not report actually having measured microbiological water quality at all (Alam 1989 BGD; Xiao 1997 CHN; Luby 2006; Mäusezhal 2009 BOL; Opryszko 2010; Majuru 2011 ZAF; Rodrigo 2011 AUS; Lindquist 2014). Thus, it cannot be concluded definitively that the interventions investigated in these studies actually resulted in an improvement in drinking water quality.

Among the eight studies investigating interventions to improve water quality at the point of distribution, only four tested microbiological water quality (Torun 1982 GTM; Gasana 2002 RWA; Jensen 2003 PAK; Kremer 2011 KEN). As these tests were at the source or point of distribution and not the POU, their results do not reflect possible post‐collection contamination.

Interventions

Eight studies evaluated source‐based interventions: improved wells or boreholes (Alam 1989 BGD; Xiao 1997 CHN; Opryszko 2010b AFG; Opryszko 2010c AFG) or improved community sources and distribution to public tap stands (Torun 1982 GTM; Gasana 2002 RWA; Jensen 2003 PAK; Kremer 2011 KEN; Majuru 2011 ZAF); none evaluated reliable piped‐in household connections.

Fourty‐seven studies evaluated POU interventions: chlorination (17 studies), filtration (20 studies), combined flocculation and disinfection (five studies), SODIS solar disinfection (six studies), combination UV disinfection and filtration (one study), and improved storage (two studies). Significantly, there were no eligible studies that investigated the impact of boiling, even though that is by far the most common type of POU water treatment (Rosa 2010).

Many studies provided a supplementary hygiene education or instruction beyond the use of the intervention itself, and among POU interventions the primary intervention was often combined with some form of improved storage. In only three multiple‐intervention arm studies did study authors establish different intervention groups with and without hygiene or other non‐water improvement steps in order to isolate the impact of water quality (URL 1995; Opryszko 2010; Lindquist 2014).

Except in blinded trials involving placebos, control arms generally continued to use their pre‐trial water supply and treatment practices. In one trial of POU chlorination plus a safe storage container, however, control households also received the container (Jain 2010 GHA). In two of the solar disinfection studies (Conroy 1996 KEN; Conroy 1999 KEN) both intervention and control households received plastic bottles for storing their drinking water. The intervention group was instructed to place the bottles on roofs to expose them to the sun, while the control group was told to keep the filled bottles indoors. It is important to note that since improved storage even in the absence of treatment has been shown to improve microbial water quality (Wright 2004), the comparison between the intervention and control in these studies may understate the effectiveness of the intervention when compared to the controls following customary water handling practices.

Adherence with the intervention

Studies of source water interventions tended to assume adherence based on the fact that the primary water supply had been improved. Some studies of POU water treatment undertook indirect assessments of adherence by measuring residual chlorine levels in stored water, comparing microbiological water quality of intervention and control groups, conducting periodic or post‐study surveys, or counting the amount of intervention product used. Most other studies measured adherence only by occasional observation, while eight cluster‐RCTs did not report on adherence.

The studies of chlorine residuals reported adherence ranging from a high of 95% (Doocy 2006 LBR) to a low of 11% (Opryszko 2010a AFG). Even among these studies, however, investigators acknowledged that it was not possible to know to what extent intervention group participants actually consumed treated water or avoided consuming untreated water. For those studies that reported on adherence, three took the additional step of investigating and reporting on continued consumption of untreated water (Boisson 2010 DRC; Peletz 2012 ZMB; Boisson 2013 IND) a source of exposure that could be masked by less direct metrics of adherence.

Outcome measures

The studies' main outcome measure was diarrhoeal disease, but different methods were used to define, assess, and report this. Thirty‐six studies used the WHO's definition of diarrhoea, while other studies used the following definitions: the mother's or respondent's definition (Austin 1993; Gasana 2002 RWA; Reller 2003; Crump 2005; Chiller 2006 GTM); 'watery diarrhoea as a component of gastroenteritis' (Colford 2002 USA; Colford 2005 USA; Colford 2009 USA; Rodrigo 2011 AUS); the local term (Conroy 1996 KEN; Conroy 1999 KEN; Boisson 2009 ETH); "significant change in bowel habits towards decreased consistency or increased frequency" (Kirchhoff 1985 BRA); or dysentery (du Preez 2010 ZAF; du Preez 2011 KEN). Four studies did not report the case definition used for diarrhoea (Torun 1982 GTM; Xiao 1997 CHN; Günther 2013 BEN; Lindquist 2014).

The method of diarrhoea surveillance and assessment also varied. In most cases, participants were visited on a periodic basis, either weekly (19 studies), fortnightly (16 studies), or more infrequently (14 studies). Participants were asked to recall and report on cases of diarrhoea during a previous period, usually seven days (30 studies) or 14 days (six studies), with four studies having recall periods of one to four days and one trial having a recall period of four weeks (Günther 2013 BEN). Twelve studies asked each participant or a designated householder to keep a log or record to indicate days with or without diarrhoea, one procured data on diarrhoea from family records and disease registries (Mahfouz 1995 KSA), or used paediatricians to assess the participants during regular medical checkups (Gasana 2002 RWA). Only one trial did not report the method (Xiao 1997 CHN).

Using these data, study authors reported diarrhoeal disease using one or more of the following epidemiological measures of disease frequency: incidence (34 studies); period prevalence (12 studies); and longitudinal prevalence (nine studies). The studies also reported other measures of disease, including incidence of persistent diarrhoea, gastrointestinal illness, including specific symptoms thereof, incidence or prevalence of bloody diarrhoea, and days of work or school lost due to diarrhoea (Lule 2005 UGA). Seven studies also reported on mortality (Crump 2005; Colford 2009 USA; Boisson 2010 DRC; du Preez 2011 KEN; Kremer 2011 KEN; Peletz 2012 ZMB; Boisson 2013 IND). None reported adverse events.

None of these studies were primarily designed to investigate the impact of the intervention on death, and as such most were underpowered to evaluate this outcome.

Data presentation

Forty‐three studies presented results both for children aged under five years (or a subgroup thereof) and for all ages or older age groups, three presented results only for adults, and nine presented results only for children under five years (or a subgroup thereof). Most of the studies adjusted raw data to account for possible covariates, including age, sex, sanitation or hygiene practices, area of residence, household income or proxies thereof, education or maternal literacy, age and occupation of the head of household, number of participants in the household or absent there from, baseline diarrhoea or conditions at baseline, or other variables associated with the household environment and participant behaviour.

Most studies of interventions at the POU also used statistical methods to adjust their results, either for repeated episodes of diarrhoea by the same participant or for clustering within the household, village or both. The studies that did not adjust for clustering may receive excess weight in meta‐analysis due to artificial precision (Kirchhoff 1985 BRA; Austin 1993; Mahfouz 1995 KSA; URL 1995).

Excluded studies

We excluded 108 studies for the reasons given in the Characteristics of excluded studies table. Two studies that appear to meet this review's inclusion criteria are currently ongoing (see Characteristics of ongoing studies).

Risk of bias in included studies

We have summarized our judgements about the risk of bias of included studies in Figure 2.

Figure 2

Risk of bias graph: summary of authors' judgements about each 'Risk of bias' item presented as percentages across all included studies.

Allocation

The allocation sequence was generated using an adequate method and classified as 'low risk' in 36 of the 45 cluster‐RCTs, 'high risk' in two, and 'unclear' in seven Figure 2. The method of allocation concealment was 'low risk' in 34 trials and 'high risk' in two and 'unclear' in nine.

Comparability of baseline characteristics (confounding bias)

All the quasi‐RCTs and CBA studies were judged to be at low risk of bias for this criteria except Gasana 2002 RWA, which was at 'unclear' risk.

Contemporaneous data collection

We judged all the quasi‐RCTs and CBA studies to be at low risk of bias for this criteria except Gasana 2002 RWA, which was at 'unclear' risk.

Blinding

Nine trials were blinded at the participant level (Kirchhoff 1985 BRA; Austin 1993; Colford 2002 USA; Colford 2005 USA; Colford 2009 USA; Boisson 2010 DRC; Jain 2010 GHA; Rodrigo 2011 AUS; Boisson 2013 IND); all but two of these were blinded at the assessor level as well (Kirchhoff 1985 BRA; Austin 1993). The others followed an open design, classified as 'high risk' of bias. One of the principal objectives of Colford 2002 USA was to assess the effectiveness of its blinding methodology; it therefore provides the most comprehensive analysis of these issues. Colford 2002 USA, Colford 2005 USA, Boisson 2010 DRC and Rodrigo 2011 AUS all used household sham water filters. Austin 1993, Kirchhoff 1985 BRA, Jain 2010 GHA and Boisson 2013 IND, which were assessing the effectiveness of home‐based chlorination, provided placebos to control households.

Incomplete outcome data

Twenty four studies were at 'low risk' of bias, 18 at 'high risk', and three studies were unclear.

Effects of interventions

Analysis 1: Any water quality intervention versus no intervention

Diarrhoea episodes

An overall pooled analysis, across different trial designs, interventions and settings, finds the risk of diarrhoea to be lower with any water quality intervention compared to no intervention, both among all ages (RR 0.59, 95% CI 0.51 to 0.69, 81215 participants; 52 studies Analysis 1.1), and under fives (RR 0.61, 95% CI 0.49 to 0.75 Analysis 1.2). However, as would be expected given the diverse nature of the trials, statistical heterogeneity between trials is very high (I² statistic = 98% and 97%, respectively). Our primary analysis is therefore stratified by the specific intervention type (for example, interventions at water source, POU chlorination, POU filtration), and by study design (for example, cluster‐RCT, quasi‐RCT, CBAs).

Mortality

Only nine studies reported any deaths among study participants. Five reported the number of deaths in each study arm without differences evident (see Table 2). Two studies reported the total number of deaths without stating how many occurred in each group (du Preez 2010 ZAF; Boisson 2013 IND), and two reported recording deaths but the numbers were not presented in the papers (Boisson 2009 ETH; Kremer 2011 KEN).

Table 2. Studies reporting deaths

Study ID	Intervention		Control		P value	Comment
Study ID	Deaths	Participants	Deaths	Participants	P value	Comment
Boisson 2010 DRC	12	546	8	598	0.27	—
Colford 2009 USA	7	385	6	385	> 0.05	—
Crump 2005a KEN	17	2249	28	2277	0.108	—
Crump 2005b KEN	14	2124	28	2277	0.052	—
du Preez 2011 KEN	3	555	3	534	> 0.05	—
Peletz 2012 ZMB	3	300	6	299	0.28	—
Boisson 2013 IND	?	6119	?	5965	—	Only reports total deaths (46)
du Preez 2010 ZAF	?	383	?	335	—	Only reports total deaths (7)
Kremer 2011 KEN	?	—	?	—	—	Reports recording deaths but does not state how many
Boisson 2009 ETH	?	731	?	785	—	Reports recording deaths but does not state how many

None of these studies were primarily designed to investigate the impact of the intervention on mortality, and all were underpowered to investigate these effects.

Adverse events

No trial reported adverse events from the interventions.

Analysis 2: Interventions at the water source

One cluster‐RCT and five CBA studies evaluated interventions at the water source (Table 3). All but one study were from settings with 'unimproved' water sources (unprotected wells or surface water), and all had unclear levels of sanitation. Three studies evaluated improved wells or boreholes, two evaluated chlorination or filtration of community water sources, and one evaluated an improved community piped supply. No studies evaluated reliable household connections to a clean water source (see Table 4 and Table 5 for a description of study settings and interventions).

Table 3. Summary of findings: improved water source

Improved water source compared with no intervention for preventing diarrhoea in rural settings in low‐ and middle‐income countries
Patient or population: adults and children Settings: low‐ and middle‐income countries in rural areas Intervention: water source improvement Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	Water source improvement
Diarrhoea episodes Cluster‐RCTs	3 episodes per person per year	3.7 episodes per person per year (2.9 to 4.7)	RR 1.24 (0.98 to 1.57)	3266 (1 trial)	⊕⊝⊝⊝ very low^1,2,3
Diarrhoea episodes CBA studies	—	—	—	5895 (5 studies)	⊕⊝⊝⊝ very low^1,4,5
The basis for the assumed risk (for example, the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012).

¹Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. None of these studies blinded participants and outcome assessors to the treatment allocation.
²No serious inconsistency.
³Downgraded by 2 for serious indirectness: this single RCT from Afghanistan evaluated the provision of protected wells. It is not possible to make broad generalizations to other settings.
⁴Downgraded by 1 for serious inconsistency: statistical heterogeneity was very high (I² statistic = 98%), such that the data could not be pooled. Some large and statistically significant effects were seen in some individual trials, but not others.
⁵Downgraded by 1 for serious indirectness: these studies are from a variety of low‐ and middle‐income countries (Bangladesh, Rwanda, Pakistan, South Africa, China). However, as only single trials evaluated each intervention it is not possible to make broad generalizations.

Table 4. Improved water source: description of the interventions

Study ID	Study design	Setting	Incidence of diarrhoea in the control group	Intervention areas		Control areas
Study ID	Study design	Setting	Incidence of diarrhoea in the control group	Water source intervention	Health promotion activities	Water source	Health promotion activities
Opryszko 2010b AFG	Cluster‐RCT	Rural villages	3.1 episodes per person per year	One well per 25 households providing 25 litres/person/day	None	35% used unprotected hand dug wells	None
Alam 1989 BGD	CBA	Rural villages	4.1 episodes per child per year	Provision of one hand pump per 4‐6 households (3 times as many as control areas)	Female health visitors visited peoples homes and organised group discussion and demonstrations to promote hygienic practices for hand pump use, water storage, child faeces disposal, hand washing.	Shallow, hand‐dug wells; some hand pumps	None described
Gasana 2002 RWA	CBA	Rural villages	3 episodes per child per year	Site A: Sedimentation tank/Katadyn filter with communal tap Site B: Gravel‐sand‐charcoal filter on existing water spring Site C: Protective fence around an existing water spring	None described	An existing water spring	None described
Jensen 2003 PAK	CBA	Rural villages	2.8 episodes per person per year	Chlorination of public water supply	None described	Unchlorinated poorly functioning sand filter system	None described
Majuru 2011 ZAF	CBA	Rural villages	0.6 episodes per person per year	Provision of intermittently operated small community water systems distributing potable water to multiple taps throughout the community	None described	Untreated water from a river and its tributaries	None described
Xiao 1997 CHN	CBA	Rural villages	Not reported	Improved water supply through structural improvements to wells	Hygiene education	Not reported	None described

Table 5. Improved water source: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient water quality	Sanitation⁴
Alam 1989 BGD	Shallow, hand‐dug wells; some hand pumps	Unimproved	Unclear	Unclear	Not tested	Unclear
Gasana 2002 RWA	Spring	Unimproved	Unclear	Unclear	Baseline range 4 to 1100 total coliforms/100 mL	Unimproved
Jensen 2003 PAK	Some slow sand filters in poor condition; some household taps; majority used ground water	Improved	Unclear	Unclear	Baseline geometric mean in intervention village: 13.3 E. coli CFU/100 mL; control villages: 137/100 mL	Unclear
Majuru 2011 ZAF	Surface water, boreholes, water tankers	Improved and unimproved	Unclear	Unclear	Not tested	Unclear
Opryszko 2010	35% use unprotected dug wells	Unimproved	Sufficient	Sufficient	Not tested	Unclear
Xiao 1997 CHN	Well water	Unimproved	Unclear	Unclear	Not tested	Unclear

¹'Improved' includes household connection, public standpipe, borehole, protected dug well, protected spring, rainwater collection; 'unimproved' includes unprotected well, unprotected spring, vendor‐provided water, bottled water; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015.
²'Sufficient' means located within 500 m, queuing no more than 15 minutes, no more than three minutes to fill 20 L container, and maintained so available consistently; 'insufficient' means that it does not meet any of above; and 'unclear' means unclear or not reported; definition based minimum standards established by The Sphere Project 2011.
³'Sufficient' means a minimum of 15 L/day/person; 'insufficient' means less than 15 L/day/person; and 'unclear' means unclear or not reported; definition based on minimum standards established by The Sphere Project 2011.
⁴'Improved' means connection to a public sewer or septic system, pour flush latrine, simple pit latrine, or ventilated improved pit latrine; 'unimproved' means service or bucket latrine, public latrines, open latrines; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015.

The single cluster‐RCT from Afghanistan reported no statistically significant difference in diarrhoea with improved wells compared to no intervention (one trial, 3266 participants; Analysis 2.1; very low quality evidence).

The CBA studies evaluated different interventions, had variable findings, and were all at unclear risk of multiple sources of bias (see Figure 3). Three of the five studies reported statistically significant effects on diarrhoea (Analysis 2.1; Analysis 2.2): in Bangladesh, provision of one hand pump per four to six households (three times as many as control areas) was associated with a small reduction in diarrhoea over three‐years follow‐up (RR 0.83, 95% CI 0.71 to 0.97); in remote areas in South Africa a new community piped water supply was associated with around a 50% reduction in diarrhoea compared to untreated river water (RR 0.43, 95% CI 0.24 to 0.77); and in China structural well improvements were also associated with around a 50% reduction in diarrhoea (RR 0.45, 95% CI 0.43 to 0.47). In contrast, chlorination and filtration of community water supplies were not associated with positive benefits in Rwanda and Pakistan respectively. Overall, the body of evidence is judged to be of very low quality (Table 3). Given the variability in interventions, further subgroup analyses to try to understand the heterogeneity were not useful.

Figure 3

Forest plot of comparison: 2 Source: water supply improvement versus control, outcome: 2.1 Diarrhoea: CBA studies subgrouped by age.

Analysis 3. POU chlorination

Fourteen cluster‐RCTs, with 16 comparisons, evaluated POU chlorination versus control. Chlorine was delivered to households free of charge every one to four weeks, with instructions on how to use it, and in eight trials a water storage container was also provided (see Table 6 and Table 7 for a description of study settings and interventions).

Table 6. POU chlorination: description of the intervention

Trial	Study design	Chlorination product?	Distributed free?	Frequency of distribution?	Storage container also distributed?	Compliance	Additional hygiene promotion
Austin 1993a GMB	Cluster‐RCT	Sodium hypochlorite solution	Yes	Fortnightly	No	40% compliance measured by residual chlorine	None
Austin 1993b GMB	Cluster‐RCT	Sodium hypochlorite solution	Yes	Fortnightly	No	59% compliance measured by residual chlorine	None
Boisson 2013 IND	Cluster‐RCT	Sodim dichloro‐isocyanurate tablets	Yes	Bimonthly	No	32% compliance measured by residual chlorine	None
Crump 2005a KEN	Cluster‐RCT	1% sodium hypochlorite	Yes	Weekly	No	61% compliance during unannounced weekly visits measured by residual chlorine	Use of ORS, treatment seeking for diarrhoea
Handzel 1998 BGD	Cluster‐RCT	0.25% to 0.3% chlorine solution	Yes	Weekly	Yes	90% compliance based on residual chlorine measurements	Hygiene and sanitation messages
Jain 2010 GHA	Cluster‐RCT	Sodim dichloro‐isocyanurate tablets	Yes	Twice weekly	Yes	74% to 89% compliance measured by chlorine residual	ORS provided to those with diarrhoea
Kirchhoff 1985 BRA	Cluster‐RCT	10% sodium hypochlorite	Yes	Daily	No	Not reported	Chlorination preformed by study staff
Luby 2006a PAK	Cluster‐RCT	Sodium hypochlorite solution	Yes	Unclear	Yes	Yes, though rate unclear	Encouraged to only drink treated water
Lule 2005 UGA	Cluster‐RCT	0.5% sodium hypochlorite	Yes	Weekly	Yes	Not reported	hygiene education
Mahfouz 1995 KSA	Cluster‐RCT	Packets of 50 g calcium hypochloride 70%.	Yes	Unclear	No	Some residual chlorine in all intervention samples	None
Mengistie 2013 ETH	Cluster‐RCT	1.25% sodium hypochlorite solution	Yes	Weekly	No	80% compliance measured by chlorine residual	None
Opryszko 2010c AFG	Cluster‐RCT	0.05% sodium hypochlorite solution	Yes	Monthly	Yes	78% compliance measured by previous 2 weeks self‐report use of chlorine	None
Quick 1999 BOL	Cluster‐RCT	MIOX unit electrolytically produced disinfectant with 3% brine solution, hypochlorite, chlorine dioxide, ozone, peroxide and other oxidants.	Yes	Weekly	Yes	63% compliance measured by water in vessel with chlorine residual, average across six rounds	Community health volunteers reinforced messages about proper use of the disinfectant and vessels and of different applications for treated water.
Reller 2003b GTM	Cluster‐RCT	Sodium hypochlorite solution (50,000 ppm)	Yes	Monthly	No	36% compliance measure by residual chlorine > 0.1 mg/L on unannounced visits.	Motivational and educational messages about chlorination, use of ORS, care seeking for diarrhoea
Reller 2003c GTM	Cluster‐RCT	Sodium hypochlorite solution (50,000 ppm)	Yes	Monthly	Yes	44% compliance measure by residual chlorine > 0.1 mg/L on unannounced visits.	Motivational and educational messages about chlorination, use of ORS, care seeking for diarrhoea
Semenza 1998 UZB	Cluster‐RCT	1.5% chlorine solution	Yes	Unclear but households were visited twice weekly	Yes	73% based on residual chlorine levels at time of visit	Only drink chlorinated water and wash all fruit and vegetables with chlorinated water
Luby 2004a PAK	CBA	Bleach (sodium hypochlorite)	Yes	Study workers visited weekly and re‐supplied the households with dilute bleach.	Yes	Not reported	Encouraged regular treatment of drinking water
Luby 2004b PAK	CBA	Bleach (sodium hypochlorite)	Yes	Study workers visited weekly and re‐supplied the households with dilute bleach.	Yes	Not reported	Encouraged regular treatment of drinking water
Quick 2002 ZMB	CBA	0.5% sodium hypochlorite	Yes	Unclear but households were visited once every two weeks	HHs paid for vessel	72% compliance measured by water in vessel with chlorine residual	Community volunteers, gave education about causes and prevention of diarrhoea and safe storage of water and motivated households about the intervention.

Table 7. POU chlorination: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient water quality	Sanitation⁴
Austin 1993	Open wells	Unimproved	Sufficient	Unclear	Mean 1871 FC/100 mL in wells; among stored water samples: mean 3358 FC/100 mL in rainy season, 1014 FC/100 mL in dry season	Unclear
Boisson 2013 IND	62% unprotected dug well, 17% tubewell, 14% tap, 5% surface water	Unimproved	Unlcear	Unclear	Baseline not reported. Control households: Geometric mean 122 TTC/100 mL	Unimproved
Crump 2005	50% ponds, 49% rivers	Unimproved	Unclear	Insufficient	Baseline mean 98 E. coli /100 mL	Unclear; 33% defecate on ground
Handzel 1998 BGD	48% tap, 52% tubewell; 61% paid for drinking water	Improved	Sufficient	Sufficient	Baseline geometric mean 138.1 faecal coloform counts/100 mL	Unimproved
Jain 2010 GHA	95% of households use tap, 84% surface water, 46% wells, 35% rainwater, 25% borehole	Improved and unimproved	Unclear	Unclear	Baseline: median E. coli MPN 93/100 mL	Unimproved
Kirchhoff 1985 BRA	Pond water stored in clay pots after filtering with cloth	Unimproved	Unclear	Insufficient	Source water: mean 970 faecal coliforms/100 mL	Unimproved
Luby 2004	Tanker trucks, municipal taps (household and community level)	Mostly unimproved	Unclear	Unclear	Baseline: approximately 60% of stored drinking water samples were free of E. coli	Improved
Luby 2006	Tanker trucks, municipal taps (household and community level), water bearer, boreholes	Mostly improved	Unclear	Unclear	Not tested	Improved
Lule 2005 UGA	16% surface or shallow wells, 50% protected springs, 49% boreholes or taps	Unimproved	Sufficient	Sufficient	Source mean E. coli counts: 11/100 mL	Improved
Mahfouz 1995 KSA	Shallow wells	Unimproved	Unclear	Unclear	Source: 92% positive with E. coli; precise level not reported	Improved
Mengistie 2013 ETH	50% well, 41% spring, 9% river	Unimproved	Unclear	Unclear	Baseline: E. coli MPN 70/100 mL	Unimproved
Opryszko 2010	35% use unprotected dug wells	Unimproved	Sufficient	Sufficient	Not tested	Unclear
Quick 1999 BOL	Shallow uncovered wells; 38% treated water	Unimproved	Unclear	Unclear	Source water: median colony count E. coli: 57,050/100 mL	Unimproved, but 47% used latrine
Quick 2002 ZMB	Shallow wells; some boiling	Unimproved	Unclear	Unclear	Source water: median colony count E. coli: 34/100 mL	Unclear
Reller 2003	Surface water from shallow wells, rivers and springs	Unimproved	Unclear	Unclear	Baseline drinking water: median colony count E. coli 63/100 mL	Unclear
Semenza 1998 UZB	Households without piped water (procured from street tap, neighbour tap, well, vendor, or river)	Unimproved	Unclear	Unclear	Source water: 54 coliform colonies/100 mL	Unclear

On average, POU chlorination in cluster RCTs reduced the risk of diarrhoea episodes by around a quarter, both for all ages (RR 0.77, 95% CI 0.65 to 0.91; 14 trials, 30,746 participants; Analysis 3.2) and for children under five years of age (RR 0.77, 95% CI 0.64 to 0.92; Analysis 3.2). However, there was substantial heterogeneity in the size of the effect which was not well explained by a series of subgroup analyses (Analysis 3.2 to Analysis 3.9).

As might be expected from an effective intervention, the trials finding larger effects from chlorination tended to be those where adherence with the intervention was higher (as measured by residual chlorine) (Analysis 3.3; Figure 4), but in the four trials which had adequate blinding no effects of water chlorination were seen (Analysis 3.4). A subgroup analysis looking at interventions with and without the provision of water storage containers did not find statistical evidence of subgroup differences (Analysis 3.5). Effects were seen in trials with 3, 6, and 12 months of follow‐up, but no effect was demonstrated in the two trials with follow‐up longer than 12 months (Analysis 3.9). The funnel plot for this comparison has some asymmetry which may be the result of publication bias (see Figure 5). The overall quality of the evidence was therefore judged to be low (Table 8).

Figure 4

Forest plot of comparison: 3 POU: water chlorination versus control, outcome: 3.3 Diarrhoea: cluster‐RCTs; subgrouped by adherence.

Figure 5

Funnel plot of comparison: 3 POU: water chlorination versus control, outcome: 3.1 Diarrhoea: subgrouped by study design.

Table 8. Summary of findings: POU chlorination

POU chlorination compared with no intervention for preventing diarrhoea
Patient or population: adults and children Settings: low‐ and middle‐income countries Intervention: distribution of chlorine for POU water treatment and instruction on use Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	POU Chlorination
Diarrhoea episodes cluster‐RCTs	3 episodes per person per year	2.3 episodes per year (2.0 to 2.7)	RR 0.77 (0.65 to 0.91)	30,746 (14 trials)	⊕⊕⊝⊝ low^1,2,3,4
Diarrhoea episodes CBA studies	3 episodes per person per year	1.5 episodes per year (1.0 to 2.3)	RR 0.51 (0.34 to 0.75)	3948 (2 studies)	⊕⊝⊝⊝ very low^5,6,7,8
The basis for the assumed risk is provided in the footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012).

¹Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. Only two of these studies blinded participants and outcome assessors to the treatment allocation, and these two studies found no evidence of an effect with chlorination.
²Downgraded by 1 for serious inconsistency: statistical heterogeneity was very high (I² statistic = 91%). In a subgroup analysis by compliance with the intervention (assessed by measurements of residual chlorine in drinking water) found larger effects in the studies with better compliance.
³No serious indirectness: these studies are mainly from low‐ and middle‐income countries (the Gambia, India, Kenya, Bangladesh, Ghana, Brazil, Pakistan,Uganda, Saudi Arabia, Ethiopia, Afghanistan, Bolivia, Guatemala, and Uzbekistan). The interventions consisted of free distribution of chlorine (every one to four weeks) plus instructions on how to use it. In some cases, the intervention included hygiene education and storage containers in which to treat and store water.
⁴No serious imprecision: the average effect suggests POU chlorination may reduce diarrhoea episodes by about a quarter. The analysis is adequately powered to detect this effect.
⁵Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. None of these studies blinded participants and outcome assessors to the treatment allocation.
⁶Downgraded by 1 for serious inconsistency: statistical heterogeneity was very high (I² statistic = 63%).
⁷Downgraded by 1 for serious indirectness: there are only two studies (three comparisons) from Pakistan and Zambia.
⁸No serious imprecision.

An additional two CBA studies evaluated POU chlorination but only provide very low quality evidence of any effect (Analysis 3.1, Table 8).

Analysis 4. POU combined flocculation and disinfection

Five cluster‐RCTs from low‐income settings evaluated interventions where sachets of flocculant and disinfectant were distributed to households to treat water from unimproved sources (three trials), improved sources (one trial), and unclear sources (one trial). Four trials also provided water containers and mixing equipment (see Table 9 and Table 10 for a description of study settings and interventions). None of the trials blinded the outcome assessment.

Table 9. POU flocculation/disinfection: description of the interventions

Study ID	Study design	Setting	Intervention areas			Control areas
Study ID	Study design	Setting	Water quality intervention	Health promotion activities	Compliance	Water source	Health promotion activities
Chiller 2006 GTM	Cluster‐RCT	Rural villages	Provided households with a large spoon and a wide‐mouthed bucket for mixing, a narrow‐topped vessel with a lid for storing treated water and provided households with sachets of the flocculant–disinfectant every week	None	44% compliance measured by residual chlorine at week 10 of study	31% tap, 40% river or spring and 25% well.	None
Crump 2005b KEN	Cluster‐RCT	Rural villages	Each week households were given sachets of the flocculant–disinfectant	None	44% compliance during unannounced weekly visits measured by residual chlorine	50% pond, 49% river and 2% spring	None
Doocy 2006 LBR	Cluster‐RCT	Liberian camps for displaced persons	Households received a bucket and large mixing spoon for preparation, a decanting cloth, a funnel and a storage container with a narrow opening and lid. Each household received a maximum of 21 flocculation–disinfectant packets per week	None	85% compliance based on residual chlorine sampling	Received a funnel and an identical storage container	None
Luby 2006b PAK	Cluster‐RCT	Squatter settlements	Provided households with flocculant‐disinfectant sachets, a water vessel and soap. Weekly distributions of sachets	Field workers educated neighbourhoods about health problems resulting from hand and water contamination and instructed households on how and when to wash hands	Yes, though rate unclear	Municipal supply at household (33%), at community tap (37%), tanker truck (12%), water bearer (13%) and tube well (5%)	None
Luby 2006c PAK	Cluster‐RCT	Squatter settlements	Flocculant‐disinfectant and vessel. Weekly distributions of sachets	Field workers educated neighbourhoods about health problems resulting from hand and water contamination	Yes, though rate unclear	Municipal supply at household (33%), at community tap (37%), tanker truck (12%), water bearer (13%) and tube well (5%)	None
Reller 2003a GTM	Cluster‐RCT	Rural villages	Weekly distribution of flocculant‐disinfectant and gave 2 cloths initially, which could be exchanged	Field workers discussed the importance of water treatment and demonstrated the water preparation process	27% compliance measure by residual chlorine > 0.1 mg/L on unannounced visits.	33% tap, 46% river or spring, 21% well.	None
Reller 2003d GTM	Cluster‐RCT	Rural villages	Weekly distribution of flocculant‐disinfectant and gave 2 cloths initially, which could be exchanged and received a large plastic spoon for stirring, a large‐mouthed bucket for mixing, and a vessel with a secure lid and a spigot for storing treated water	Field workers discussed the importance of water treatment and demonstrated the water preparation process	34% compliance measure by residual chlorine > 0.1 mg/L on unannounced visits.	33% tap, 46% river or spring, 21% well.	None

Table 10. POU flocculation/disinfection: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient H₂O quality	Sanitation⁴
Chiller 2006 GTM	Rivers, springs, taps, and wells	Unclear	Unclear	Sufficient	98% of source samples contained E. coli; precise level not reported	Mostly unimproved
Crump 2005b KEN	50% ponds, 49% rivers	Unimproved	Unclear	Insufficient	Baseline mean 98 E. coli /100 mL	Unclear; 33% defecate on ground
Doocy 2006 LBR	Surface sources and some tap stands	Unimproved	Unclear	Insufficient	Source water: 88% samples tested positive for faecal contamination; precise level not reported	Unimproved
Luby 2006b PAK	Tanker trucks, municipal taps (household and community level), water bearer, boreholes	Mostly improved	Unclear	Unclear	Not tested	Improved
Reller 2003a GTM	Surface water from shallow wells, rivers and springs	Unimproved	Unclear	Unclear	Baseline drinking water: median colony count E. coli 63/100 mL	Unclear

Four of the five trials found statistically significant reductions in diarrhoea with the intervention (Table 11), but statistical heterogeneity in the size of this effect made pooling the data difficult (I² statistic = 99%; Analysis 4.1). This heterogeneity relates to one trial from Liberia IDP camps, Doocy 2006 LBR, where the flocculation and disinfection kits reduced diarrhoea by 88% (RR 0.12, 95% CI 0.11 to 0.13; one trial, 2191 participants). Exclusion of this potential outlier finds a more modest effect with the other four trials both for all ages (RR 0.69, 95% CI 0.58 to 0.82; four trials, 11788 participants; Analysis 4.2) and for children under five years of age (RR 0.71, 95% CI 0.61 to 0.84; Analysis 4.2).

Table 11. Summary of findings: POU flocculation and disinfection

POU water flocculation and disinfection compared with no intervention for preventing diarrhoea
Patient or population: adults and children Settings: low‐ and middle‐income countries Intervention: distribution of sachets combining water flocculation and disinfection and instructions on use Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	Water flocculation and disinfection
Diarrhoea episodes Cluster‐RCTs	3 episodes per person per year	2.1 episodes per person per year (1.7 to 2.5)	RR 0.69 (0.58 to 0.82)	11,788 (4 trials)	⊕⊕⊕⊝ moderate^1,2,3,4
The basis for the assumed risk is provided in the footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012).

¹Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. None of these studies blinded participants and outcome assessors to the treatment allocation.
²No serious inconsistency: In the complete analysis of five trials statistical heterogeneity was very high (I² statistic = 99%). However, this heterogeneity was related to a single trial showing very large effects conducted in an emergency setting in Liberia possibly due to epidemic diarrhoea. When this trial was removed as an outlier, there was a smaller, but more consistent effect.
³No serious indirectness: the studies were conducted in rural areas in Guatemala (two studies), and Kenya (one study), one trial was from a camp for displaced persons in Liberia and one from squatter settlements in Pakistan. Sanitation was improved in only one of these studies.
⁴No serious imprecision: all five studies found benefits with flocculation. The 95% CI of the pooled effect includes the possibility of no effect, but this imprecision is a result of the heterogeneity between studies.

Adherence with the intervention, as measured by residual chlorine, was generally low (< 50%), but higher in the trial from Liberia showing large effects (Analysis 4.3). Larger effects tended to also be seen in the trials also providing water storage containers (Analysis 4.4). The effects were present in trials with both improved and unimproved water source and sanitation (Analysis 4.5; Analysis 4.6; Analysis 4.7). None of the trials had follow‐up longer than 12 months (Analysis 4.8).

Analysis 5. POU filtration

Overall 20 cluster‐RCTs evaluated POU filtration: ceramic filtration (nine trials), biosand filtration (five trials), LifeStraw® filters (three trials), and plumbed‐in filtration (three trials) (see Table 12 and Table 13 for a description of study settings and interventions).

Table 12. POU filtration: description of interventions

Study ID	Intervention sub‐group	Study design	Setting	Intervention areas			Control areas
Study ID	Intervention sub‐group	Study design	Setting	Water quality intervention	Health promotion activities	Compliance	Water source	Health promotion activities
Abebe 2014 ZAF	Ceramic filter	Cluster‐RCT	Rural	Ceramic water filter impregnated with silver nanoparticles with safe storage containers	Education about safe water and hygiene and information on how to use the filter and maintain it.	Not reported	Personal tap in home (44%), community tap (44%) and river (3%)	Received usual clinical care including education about safe water and hygiene at the clinic
Brown 2008a KHM	Ceramic filter	Cluster‐RCT	Rural	CWP (Cambodian Ceramic Water Purifier) including safe storage container.	None	98% compliance measured by self‐report	Surface water (55%) and ground water (48%) during the dry season and surface water (45%), ground water (48%) and rain water (73%) during the rainy season	None
Brown 2008b KHM	Ceramic filter	Cluster‐RCT	Rural	CWP‐Fe (iron‐rich ceramic water purifier) including safe storage container.	None	98% compliance measured by self‐report	Surface water (55%) and ground water (48%) during the dry season and surface water (45%), ground water (48%) and rain water (73%) during the rainy season	None
Clasen 2004b BOL	Ceramic filter	Cluster‐RCT	Rural	Ceramic filters including improved storage	None	67% of households had filters in regular use	68% had taps and 11% boiled water.	None
Clasen 2004c BOL	Ceramic filter	Cluster‐RCT	Rural	Ceramic filters including improved storage	None	100% of intervention households' water free of TTC	Water from canal (52%), river (35%) or rainwater (4%)	None
Clasen 2005 COL	Ceramic filter	Cluster‐RCT	Rural and urban affected by conflict	Ceramic water filter system including improved storage	None	Not reported	River (27.6%), rainwater(12.1%), yard tap (67.2%). 70.7% claimed to treat water.	None
du Preez 2008 ZAF/ZWE	Ceramic filter	Cluster‐RCT	Rural	Ceramic filters including improved storage	None	55% compliance measured by water quality (approximate compliance across intervention households in Zimbabwe and South Africa).	Protected water source (53.8%) and unprotected water source (46.2%)	None
Lindquist 2014a BOL	Ceramic filter	Cluster‐RCT	Peri‐urban	Received a PointONE Filter and a 30 L bucket (with lid)	Participants were instructed on diarrhoeal transmission (biological versus cultural beliefs‐based), prevention and treatment.	97% compliance based on reported use	83% used water from tanker trucks and 12% from water coolers.	Received weekly messages on life skills and attitudes. Also were instructed on diarrhoeal transmission, prevention and treatment.
Lindquist 2014b BOL	Ceramic filter	Cluster‐RCT	Peri‐urban	Received a PointONE Filter and a 30‐L bucket (with lid) and WASH education	Participants received weekly WASH messages on personal and family hygiene, sanitation, boiling and chlorine‐based water treatments (excluding filtration),vitamin A, hygienic food preparation and cleaning, and parasite prevention.	90% compliance based on reported use	83% used water from tanker trucks and 12% from water coolers.	Received weekly messages on life skills and attitudes. Also were instructed on diarrhoeal transmission, prevention and treatment.
URL 1995a GTM	Ceramic filter	Cluster‐RCT	Rural	Handmade ceramic water filter	None	87% to 93% use of filter by children	Majority of households collected water from household tap (not chlorinated)	None
URL 1995b GTM	Ceramic filter	Cluster‐RCT	Rural	Handmade ceramic water filter	Education on nutrition (ORS, basic nutrition and maternal and child nutrition), health (hygiene) and family values.	As above	Majority of households collected water from household tap (not chlorinated)	None
Fabiszewski 2012 HND	Sand filtration	Cluster‐RCT	Rural	Hydraid plastic‐housing BioSand filter (BSF) + 20 L water jug	Training for the use and maintenance of the BSF and general education about hygiene and sanitation.	Not reported	Among all study participants‐ the main source of drinking water were: protected water sources (49% to 69% households per month), protected sources (24% to 50% per month), piped water (1% to 11% per month), and rainwater (0% to 2% per month).	Training for the use and maintenance of the BSF and general education about hygiene and sanitation.
Stauber 2009 DOM	Sand filtration	Cluster‐RCT	Semi‐rural and urban	Received a biosand filter and safe storage container	Nothing	Water quality testing, however no intervention household level compliance reported	42% reported treating drinking water.	None
Stauber 2012a KHM	Sand filtration	Cluster‐RCT	Rural	Plastic biosand filter. HHs were asked to pay USD 10 for the filter.	Health and hygiene education sessions	89% compliance measured by household‐reported use at least 3 times per week	Improved water sources during the dry season (7.1%) and during the rainy season (88.9%). 49.5% reported boiling drinking water.	Health and hygiene education sessions
Stauber 2012b GHA	Sand filtration	Cluster‐RCT	Rural	Plastic biosand filter	Not specified	97% compliance measured by household‐reported use	Use surface water during dry season (95%) and use surface water during rainy season (70.6%). 96.5% reported sieving drinking water through cloth.	nothing
Tiwari 2009 KEN	Sand filtration	Cluster‐RCT	Rural	Provided with the concrete BioSand Filter	At each visit, three oral rehydration packets and instructions were provided.	Not reported	All control houses reported drinking river or unprotected spring water; drink rainwater (96.6%), drink improved source (24.1%). 34.5% reported boiling drinking water.	At each visit, three oral rehydration packets and instructions were provided.
Boisson 2009 ETH	LifeStraw® Personal	Cluster‐RCT	Rural	A LifeStraw® personal pipe‐style water treatment device was given to each member of the household >6 months and encouraged to use it at home and away from home.	None	13% report use today	The primary drinking water source for 84% was from spring, 12% from rivers, 2.5% from hand dug wells and 4% from communal taps.	None
Boisson 2010 DRC	LifeStraw® Family	Cluster‐RCT	Rural	Households received a LifeStraw® Family filters	None	76% compliance measured by self‐report use today or yesterday (at 14 month follow‐up)	Received a placebo filter.	None
Peletz 2012 ZMB	LifeStraw® Family	Cluster‐RCT	Peri‐urban	Households received a LifeStraw® Family filter and two 5 L safe storage containers.	None	87% compliance measured by improved water quality	46% use unprotected dug wells, 19% boreholes, 17% public standpipes, 12% protected dug well, 5% piped into home or yard and 2% surface water.	None
Colford 2002 USA	Plumbed in filter	Cluster‐RCT	Urban	Installation of water treatment devices to 1 tap in HH that include: a 1‐micron absolute prefilter cartridge and a UV lamp.	None	96% compliance measured by not dropping out of study (plumbed‐in unit)	Sham device	None
Colford 2005 USA	Plumbed in filter	Cluster‐RCT	Urban	Installation of filter (1‐micron filter and a UV lamp) to main faucet of household	All participants received the current CDC safe drinking water guidelines for immuno‐compromised persons	90% compliance measured by not dropping out of study (filter attached to kitchen sink)	Sham device	All participants received the current CDC safe drinking water guidelines for immuno‐compromised persons
Colford 2009 USA	Plumbed in filter	Cluster‐RCT	Urban	Installation of filter (1‐micron filter and a UV lamp) to main faucet of household	None	83% compliance measured by not dropping out of study (filter attached to kitchen sink)	Sham device	None
Rodrigo 2011 AUS	Ceramic filter/plumbed in	Cluster‐RCT	Urban	Bench‐top silver impregnated ceramic water treatment units, which required participants to use fill it but then households that had rainwater piped into kitchen were offered an under sink unit	None	Not reported	Sham water treatment unit	None

Table 13. POU filtration: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient H₂O quality	Sanitation⁴
Abebe 2014 ZAF	In‐home taps or community taps	Improved	Sufficient	Unclear	80% of households had contamination between 10 to 10000 CFUs/100 mL	Unclear
Brown 2008	62% households rely on surface water during dry season and 55% rely on surface water during rainy season	Unimproved	Unlcear	Unclear	Baseline not reported. Control households: Geometric mean 600 E. coli /100 mL	Improved
Clasen 2004b BOL	80% yard taps supplied by untreated surface source, 20% directly from untreated surface sources	80% improved, 20% unimproved	Sufficient	Sufficient	Baseline arithmetic mean 86 TTC/100 mL	Unimproved
Clasen 2004c BOL	Irrigation canals and other surface sources	Unimproved	Sufficient	Sufficient	Baseline arithmetic mean 797 TTC/100 mL	Unimproved
Clasen 2005 COL	67% yard tap from municipality (not treated), 28% river, 12% rainwater	Unimproved	Unclear	Unclear	Baseline not reported. Control households: arithmetic mean 151 TTC/100 mL	Mostly improved
du Preez 2008 ZAF/ZWE	Protected wells	Improved	Sufficient	Unclear	Baseline not reported. Control households: 30% samples post‐intervention met WHO guidelines for water quality	Improved
Lindquist 2014	Municipal supply	Improved	Sufficient	Unclear	Not tested	Unimproved
URL 1995	Household tap (27%), public tap (21%), well (23%)	Improved	Unclear	Unclear	Range 5 to 260; average 106 faecal coliforms/100 mL across three sites.	Improved
Fabiszewski 2012 HND	49% to 69% households use unprotected sources, 24% to 50% use protected sources, 1% to 11% piped water, 0% to 2 % rainwater	Improved and unimproved	Unclear	Unclear	Geometric mean E. coli concentrations of both unprotected and protected sources were > 100 MPN/100 mL	Unimproved
Stauber 2009 DOM	Unclear	Unclear	Unclear	Unclear	Baseline: geometric mean 21 MPN E. coli /100 mL	Improved
Stauber 2012a KHM	77% used improved water source during dry season, 89% during rainy season	Improved	Unclear	Unclear	Baseline: geometric mean 27.5 CFU/100 mL	Unimproved
Stauber 2012b GHA	Surface water 70% in dry season, 95% in rainy season	Unimproved	Unclear	Unclear	Baseline: geometric mean 792 or 832 E. coli /100 mL for control and intervention households, respectively	Unimproved
Tiwari 2009 KEN	Primarily river water; 27% drink protected sources	Unimproved	Unclear	Unclear	Baseline not reported. Control households: 88.9 faecal coliforms/100 mL	Unclear
Boisson 2009 ETH	84% springs, 12% river, 2% handdug well, 4% communal tap	Unimproved	Unclear	Unclear	Baseline arithmetic mean 449 TTC/100 mL	Unimproved
Boisson 2010 DRC	97% surface water, 38% rainwater, 16% springs	Unimproved	Unclear	Unclear	Source drinking water: 75% of household samples > 1000 TTC/100 mL	Unimproved
Peletz 2012 ZMB	46% unprotected dug wells, 22% taps, 16% borehole or protected dug well, 2% surface water	Improved and unimproved	Unclear	Unclear	Unfiltered water: Geometric mean 190 TTC/100 mL	Unimproved
Colford 2002 USA	Household taps supplied by municipal water treatment	Improved	Sufficient	Sufficient	Data from water treatment plant: met US federal and California drinking water standards	Improved
Colford 2005 USA	Household taps supplied by municipal water treatment	Improved	Sufficient	Sufficent	Data from water treatment plant: met US federal drinking water standards	Improved
Colford 2009 USA	Household taps supplied by municipal water treatment	Improved	Sufficient	Sufficient	Data from water treatment plant: met US federal drinking water standards	Improved
Rodrigo 2011 AUS	Untreated rainwater	Improved	Sufficient	Sufficient	Not tested	Improved

Abbreviations: TTC: thermotolerant coliforms, MPN: most probable number, CFU: colony‐forming units

On average, POU filtration technologies reduced diarrhoea by around a half, both for all ages (RR 0.48, 95% CI 0.38 to 0.59; 18 trials, 15,582 participants; Analysis 5.1) and for children under five years of age (RR 0.49, 95% CI 0.38 to 0.62; Analysis 5.1). However, the number of trials and the quality of evidence was different for each specific intervention (Analysis 5.2; Figure 6). The lack of blinding in these studies is a major concern: of the five trials with adequate blinding only one found a statistically significant effect (Analysis 5.3). The quality of evidence was therefore downgraded for all types of filters due to risk of bias (Table 14).

Figure 6

Forest plot of comparison: 4 POU: filtration versus control, outcome: 4.2 Diarrhoea: cluster‐RCTs: subgrouped by type of filtration.

Table 14. Summary of findings: POU filtration

POU filtration compared with no intervention for preventing diarrhoea
Patient or population: adults and children Settings: low‐, middle‐ and high‐income countries Intervention: distribution of water filters and instructions on use Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	Water filtration
Diarrhoea episodes Cluster‐RCTs	3 episodes per person per year	All filters	RR 0.48 (0.38 to 0.59)	15,582 (18 trials)	⊕⊕⊕⊝ moderate^1,2,3,4
		1.4 episodes per person per year (1.1 to 1.8)
	3 episodes per person per year	Ceramic filters	RR 0.39 (0.29 to 0.53)	5763 (8 trials)	⊕⊕⊕⊝ moderate^2,4,5,6
		1.1 episodes per person per year (0.8 to 1.5)
		Biosand filters	RR 0.47 (0.39 to 0.57)	5504 (4 trials)	⊕⊕⊕⊝ moderate^4,7,8,9
		1.4 episodes per person per year (1.2 to 1.7)
		LifeStraw®filters	RR 0.69 (0.51 to 0.93)	3259 (3 trials)	⊕⊕⊝⊝ low^2,4,10,11
		2.1 episodes per person per year (1.5 to 2.8)
		Plumbed filters	RR 0.73 (0.52 to 1.03)	1056 (3 trials)	⊕⊕⊕⊝ moderate^2,4,12,13
		2.2 episodes per person per year (1.6 to 3.1)
The basis for the assumed risk is provided in the footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012).

¹Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants, this outcome is susceptible to bias from lack of blinding. Only five studies blinded participants and outcome assessors to the treatment allocation and only one found an effect of the intervention.
²No serious inconsistency: statistical heterogeneity was very high, however there is consistency in the direction of the effect.
³No serious indirectness: these studies are from a variety of low‐, middle‐, and high‐income countries (South Africa, Ethiopia, Democratic Republic of Congo, Cambodia, Bolivia, Colombia, USA, Australia, Honduras, Zimbabwe, Zambia, Dominican Republic, Ghana, Kenya and Guatemala).
⁴No serious imprecision.
⁵Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants, this outcome is susceptible to bias from lack of blinding. Only one of these studies, Rodrigo 2011 AUS, blinded participants and outcome assessors to the treatment allocation.
⁶No serious indirectness: these studies are from a variety of low‐, middle‐, and high‐income countries (South Africa, Cambodia, Bolivia, Colombia, Zimbabwe, Guatemala and Australia). The interventions consisted of distribution of water filters (which included a safe storage chamber) plus instructions on how to use them. In some cases, the intervention included hygiene education.
⁷Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants, this outcome is susceptible to bias from lack of blinding. None these studies blinded participants and outcome assessors to the treatment allocation.
⁸No serious inconsistency: there was no statistical heterogeneity between studies, I² statistic = 0%.
⁹No serious indirectness: the studies were conducted in a variety of rural and urban settings in a variety of low‐ and middle‐income countries (Honduras, Dominican Republic, Cambodia, Ghana and Kenya). The interventions consisted of distribution of water filters plus instructions on how to use them. In some cases, the intervention included hygiene education and a separate storage vessel.
¹⁰Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. Only one of these studies, Boisson 2010 DRC, blinded participants and outcome assessors to the treatment allocation and found no evidence of effect of the filter.
¹¹Downgraded by 1 for some indirectness, the studies were only performed in three sub‐Saharan African countries (Ethiopia, Democratic Republic of Congo, and Zambia).
¹²No serious risk of bias: the three studies blinded participants and outcome assessors to the treatment allocation.
¹³Downgraded by 1 for some indirectness, the three studies were only performed in the USA in water conditions that presumed to meet US EPA standards.

POU ceramic filters reduced diarrhoea by around 60% in nine trials mainly from low‐ or middle‐income countries, regardless of whether the water source or sanitation was classified as improved or unimproved (RR 0.39, 95% CI 0.29 to 0.53, eight trials, 5763 participants; Analysis 5.3; Analysis 5.4; moderate quality evidence).

Similarly, biosand filtration reduced diarrhoea by around a half consistently across five trials from low‐ or middle‐income settings, again regardless of whether the water source or sanitation was improved or unimproved (RR 0.47, 95% CI 0.39 to 0.57, four trials, 5504 participants; Analysis 5.6; Analysis 5.7; moderate quality evidence).

On average, the use of LifeStraw® filters reduced diarrhoea by around a third in three trials from low‐income settings with unimproved water sources (RR 0.69, 95% CI 0.51 to 0.93; three trials, 3259 participants; Analysis 5.2; low quality evidence).

Plumbed‐in filtration has only been evaluated in high‐income settings (USA). There is a modest effect in all three trials, although only one reaches standard levels of statistical significance. The overall meta‐analysis has similar effect sizes with both fixed effects and random effects models, but wider confidence intervals with random effects (Fixed‐effects: RR 0.81, 95% CI 0.70 to 0.94; Random‐effects: RR 0.73, 95% CI 0.52 to 1.03; three trials, 1056 participants; Analysis 5.2; moderate quality evidence).

Adherence with the filtration systems was reported by 14 trials, of which eight assessed this by self‐reported use which is at high risk of bias due to the lack of blinding. Adherence was generally reported as high, and larger effects were apparent in trials with higher adherence (Analysis 5.8). A subgroup analysis looking at filtration interventions with and without the provision of water storage containers (excluding the trials evaluating plumbed in filtration), found larger effects in the nine trials providing containers (Analysis 5.9). Effects were seen in trials with 3, 6, and 12 months of follow‐up, but no effect was demonstrated in the one trial with follow‐up longer than 12 months (Analysis 5.10).

Analysis 6. POU solar disinfection (SODIS)

Four cluster‐RCTs and two quasi‐RCTs evaluated solar disinfection of water from improved sources (one study) and unimproved sources (five studies) in low‐income settings. Plastic bottles were distributed to households with instructions to leave filled bottles in direct sunlight for at least six hours before drinking (see Table 15 and Table 16 for a description of study settings and interventions).

Table 15. POU solar disinfection (SODIS): description of the interventions

Study ID	Study design	Setting	Intervention areas			Control areas
Study ID	Study design	Setting	Water quality intervention	Health promotion activities	Compliance	Water source	Health promotion activities
Conroy 1996 KEN	Quasi‐RCT	Rural	Children were given two 1.5 L plastic bottles and told to keep the bottles on the roof of the hut throughout the day in full sunlight	None	100%‐ random checks by project workers uncovered no evidence of non‐compliance	Children were given two 1.5 L plastic bottles and told to keep the bottles indoors	None
Conroy 1999 KEN	Quasi‐RCT	Rural	Mothers were given plastic bottles and told to keep the bottles on the roof of the hut throughout the day in full sunlight	None	Not reported	Mothers were given plastic bottles and told to keep the bottles indoors	None
du Preez 2010 ZAF	Cluster‐RCT	Peri urban	Received two 2 L polyethylene terephtalate (PET) bottles for each child. Carers were instructed to fill one bottle and place it in full, unobscured sunlight for a minimum of 6 h every day.	None	25% compliance measured by participants filling out diarrhoeal diaries at least 75% of the time	No SODIS bottles and maintain their usual practices	None
du Preez 2011 KEN	Cluster‐RCT	Peri urban and rural	Received two 2 L PET bottles for each child. Carers were instructed to fill one bottle and place it in full, unobscured sunlight for a minimum of 6 h every day.	None	Not specified.	No SODIS bottles and maintain their usual practices	None
Mäusezhal 2009 BOL	Cluster‐RCT	Rural	Households were supplied regularly with clean, PET bottles. They were instructed to expose the waterfilled bottles for at least 6 h to the sun.	Households were taught about the importance and benefits of drinking only treated water, the germ–disease concept, and promoted hygiene measures such as safe drinking water storage and hand washing.	32% compliance measured by observation	Drinking water from spring (48.1%), tap (51.9%), river (22.1%), rain (14.9%) and dug well (14.9%)	None
McGuigan 2011 KHM	Cluster‐RCT	Rural	Households were provided with two transparent 2 L plastic bottles for each child and a sheet of corrugated iron on which to place the bottles to expose them to sunlight. Carers were instructed to fill one bottle and place it in full, unobscured sunlight for a minimum of 6 h every day.	The parents or carers were given verbal and written information on the disease concept and a simple explanation of the solar disinfection process and its effect on the microbial quality of their drinking water and subsequently the health of their children	90% (5% of children having < 10 months of follow‐up and 2.3% having < 6 months)	Almost all of the households (97%) obtained water from unprotected boreholes. An important subgroup of these, 25%, drew water from shallow tube wells fitted with hand pumps. The remainder used unprotected wells or surface ponds	None

Table 16. POU solar disinfection (SODIS): primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient H₂O quality	Sanitation⁴
Conroy 1996 KEN	Open water holes, tank fed by untreated piped water supply.	Unimproved	Unclear	Unclear	Source water: 10³ CFU/100 mL	Unclear
Conroy 1999 KEN	Open water holes, tank fed by untreated piped water supply.	Unimproved	Unclear	Unclear	Source water: 10³ CFU/100 mL	Unclear
du Preez 2010 ZAF	39% standpipes, 28% protected borehole, 10% unprotected boreholes, protected springs	Mostly improved	Sufficient	Sufficient	Baseline not reported. Intervention households: 62% of samples met WHO guidelines for water quality; no significant difference from control households	Unclear
du Preez 2011 KEN	Spring, protected and unprotected dug wells protected, canals, other	Mostly unimproved	Unclear	Unclear	50% of samples from stored water had 10 CFU/100 mL or less; no significant difference for intervention and controls	Unclear
Mäusezhal 2009 BOL	48% spring, 52% tap, 22% river, 15% rain, 15% dug well	Improved and unimproved	Sufficient	Sufficient	Not tested	Unimproved
McGuigan 2011 KHM	97% households use unprotected sources: unprotected wells, surface ponds	Unimproved	Unclear	Unclear	Baseline not reported. Control households: geometric mean 48 CFU/100 mL	Unimproved

Overall in the cluster‐RCTs, solar disinfection reduced diarrhoea by around a third for all ages (RR 0.62, 95% CI 0.42 to 0.94; four trials, 3460 participants; Analysis 6.1), and almost a half in children under five years of age (RR 0.55, 95% CI 0.34 to 0.91; Analysis 6.2). The largest effect was seen in the trial with the highest adherence (Analysis 6.3). The quality of evidence was downgraded to moderate due to the lack of blinding and the inherent risk of bias (Table 17).

Table 17. Summary of findings: POU solar disinfection (SODIS)

POU solar disinfection (SODIS) of water compared with no intervention for preventing diarrhoea
Patient or population: adults and children Settings: low‐ and middle‐income countries Intervention: distribution of plastic bottles with instructions on using them to treat water using the SODIS method. Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	SODIS
Diarrhoea episodes Cluster‐RCTs	3 episodes per person per year	1.9 episodes per person per year (1.3 to 2.8)	RR 0.62 (0.42 to 0.94)	3460 (4 trials)	⊕⊕⊕⊝ moderate^1,2,3,4
Diarrhoea episodes Quasi‐RCTs	3 episodes per person per year	2.5 episodes per person per year (2.1 to 2.9)	RR 0.82 (0.69 to 0.97)	555 (2 studies)	⊕⊕⊝⊝ low^1,5,6,7
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012).

¹Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. None of these studies blinded participants and outcome assessors to the treatment allocation.
²No serious inconsistency: statistical heterogeneity was very high (I² statistic = 89%), however there is consistency in the direction of the effect. This heterogeneity may relate to differences in compliance across the studies, however compliance was not measured in the same way across studies.
³No serious indirectness: the studies were conducted in peri‐urban South Africa (one study), peri‐urban and rural Kenya (one study), rural Bolivia (one study) and rural Cambodia (one study).
⁴No serious imprecision: the average effect suggests that the intervention may reduce diarrhoea episodes by about one third.
⁵No serious inconsistency: statistical heterogeneity was low (I² statistic = 0%).
⁶Downgraded by 1 for serious indirectness: there are only two studies and both were conducted in the same province in Kenya (one study included children five to 16 years old and the other included children younger than six years old).
⁷No serious imprecision.

In the quasi‐RCTs the observed effect was lower (RR 0.82, 95% CI 0.69 to 0.97; two trials, 555 participants; Analysis 6.1).

Analysis 7. POU UV disinfection

One cluster‐RCT from Mexico evaluated an UV tube disinfection technology (Gruber 2013 MEX; see Table 18 and Table 19 for a description of the study setting and intervention).

Table 18. POU UV: description of the interventions

Study ID	Study design	Setting	Intervention areas			Control areas
Study ID	Study design	Setting	Water quality intervention	Health promotion activities	Compliance	Water source	Health promotion activities
Gruber 2013 MEX	Cluster‐RCT	Rural	Promotion of the UV Tube disinfection technology and safe storage	Unclear	51% compliance measured by access to treatment device	Unclear	None

Table 19. POU UV: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient H₂O quality	Sanitation⁴
Gruber 2013 MEX	Unclear	Unclear	Unclear	Unclear	Baseline: 60% of samples with detectable E. coli	Improved

The effect on diarrhoea among all age populations did not reach standard levels of statistical significance (RR 0.79, 95% CI 0.49 to 1.27; one trial, 1913 participants; Analysis 7.1), and did not report separately for children under five years of age.

Analysis 8. POU improved storage

Two trials from Malawi and Benin evaluated the distribution of improved water storage containers in settings with improved water sources (see Table 20 and Table 21 for a description of the study setting and intervention).

Table 20. POU Improved storage: description of the interventions

Study ID	Study design	Setting	Intervention areas			Control areas
Study ID	Study design	Setting	Water quality intervention	Health promotion activities	Compliance	Water source	Health promotion activities
Günther 2013 BEN	Cluster‐RCT	Rural	Provided households with a new 30 L household water storage with a tap at the bottom, a new plastic container to transport water from the water source to the household and a sign attached to the transport and storage containers which emphasized the importance of avoiding hand‐contact with the water and to only use water from an improved water source.	None	After 7 months, 88% of households were still using the improved storage containers	68% only consume improved water source	None
Roberts 2001 MWI	Cluster‐RCT	Refugee camp	All of the participating household's water collection vessels were exchanged for improved buckets (20 L with a narrow opening to limit hand entry). Households were offered 1 improved bucket in exchange for 1 vessel, 2 for 2, and 3 improved buckets for any number of containers > 2. Households were asked never to put their hands in the improved buckets and were shown how to rinse the bucket without hand entry.	None	Intervention householders received buckets; actual use was not reported	Provided with 20 L standard ration bucket	None

Table 21. POU Improved storage: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient H₂O quality	Sanitation⁴
Günther 2013 BEN	Public tap or pump	Improved	Sufficient	Unclear	12% source water contaminated (≥ 1000 CFU per 100 mL)	Unclear
Roberts 2001 MWI	Traditional pots or standard ration buckets filled at refugee camp water point	Improved	Unclear	Unclear	Source water: 71% of samples had ≤ 1 faecal coliform/100 mL	Unclear

Overall, there was no statistically significant effect on diarrhoea for all ages (RR 0.91, 95% CI 0.74 to 1.11; two trials, 1871 participants; Analysis 8.1), or children under five years of age (RR 0.69, 95% CI 0.47 to 1.01; Analysis 8.1). Both studies were at high risk of bias due to being non‐blinded, and the overall quality of the evidence was judged to be low (Table 22).

Table 22. Summary of findings: POU improved water storage

Improved water storage compared with no intervention for preventing diarrhoea
Patient or population: adults and children in sub‐Saharan Africa Settings: areas with improved water sources Intervention: distribution of improved water containers Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	Water storage
Diarrhoea episodes Cluster‐RCTs	3 episodes per person per year	2.7 episodes per person per year (2.2 to 3.3 )	RR 0.91 (0.74 to 1.11)	1871 (2 trials)	⊕⊕⊝⊝ low^1,2,3,4
The basis for the assumed risk is the median control group risk across studies. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.

The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012).

Analyses adjusted for non‐blinding

In Table 23 we have presented meta‐analysis results adjusted for non‐blinding using an approach described in the Methods section and based in part on those employed by other researchers (Hunter 2009; Wolf 2014). In these analyses, the effects of POU chlorination and filtration are smaller but remain statistically significant; the effect of POU solar disinfection becomes borderline non‐significant.

Table 23. Estimates of household‐level interventions after adjustment for non‐blinding

POU intervention	Number of comparisons	Not adjusted for non‐blinding		Adjusted for non‐blinding
POU intervention	Number of comparisons	RR	95% CI	RR	95% CI
All	55	0.56	(0.46 to 0.68)	0.70	(0.64 to 0.77)
Chlorination	19	0.72	(0.61 to 0.84)	0.80	(0.69 to 0.92)
Filtration	23	0.48	(0.38 to 0.59)	0.62	(0.55 to 0.70)
Flocculation and disinfection	7	0.48	(0.20 to 1.16)	0.65	(0.40 to 1.09)
SODIS	6	0.68	(0.53 to 0.89)	0.80	(0.60 to 1.01)

Abbreviation: SODIS: solar disinfection; CI: confidence interval.

Discussion

Summary of main results

The distribution and promotion of point‐of‐use water chlorination products may reduce diarrhoea by around one quarter (low quality evidence). Similarly, distribution and promotion of flocculation and disinfection sachets probably reduces diarrhoea but had highly variable effects (moderate quality evidence).

Point‐of‐use filtration systems probably reduce diarrhoea by around a half (moderate quality evidence). This reduction was apparent for ceramic filters, biosand systems and LifeStraw® filters, but plumbed in filtration has only been evaluated in high‐ income settings and a statistically significant effect has not been demonstrated.

In low‐income settings, distribution of plastic bottles with instructions to leave filled bottles in direct sunlight for at least six hours before drinking (SODIS) probably reduces diarrhoea by around a third (moderate quality evidence).

In subgroup analyses, larger effects were seen in trials with higher adherence, and trials that provided a safe storage container.

Overall completeness and applicability of evidence

Fifty‐five studies met the inclusion criteria, of which most studies were conducted in low‐ or middle‐income countries (50 studies), with unimproved water sources (30 studies), and unimproved or unclear sanitation (34 studies).

For water source interventions, there are simply too few studies to make conclusions about what may or may not be effective in different settings. While protective effects were seen in some individual trials, it is unclear whether these effects could be expected to be reproducible in other settings, and all of the trials had multiple potential sources of bias. Significantly, we found no studies evaluating reliable, piped‐in water supplies.

In contrast, some POU interventions do appear to be broadly protective against diarrhoea across many settings regardless of whether water sources and sanitation are 'improved' or 'unimproved'. This finding affirms the current strategy of the WHO and UNICEF to promote POU water treatment and safe storage, even though this will not increase the number of households with access to improved water supplies and therefore will not contribute towards achieving current international water targets (WHO 2011). The effectiveness of POU interventions in settings without improved sanitation contradicts earlier findings that interventions to improve water quality are effective only where sanitation has already been addressed (Esrey 1986; VanDerslice 1995), or that environmental interventions to prevent diarrhoea are effective only by employing an integrated approach (Eisenberg 2007).

Although we provide average estimates of effect for each individual POU intervention, we recommend caution in using these estimates to conclude the superiority of one intervention over another. Such an observational analysis would be highly susceptible to confounding by study setting and population, and may not represent true differences in the size of the effects. Head‐to‐head trials would be necessary to reliably conclude superiority and these were not the focus of this review.

As few studies continued follow‐up beyond 12 months, we are unable to comment reliably on the long‐term sustainability of these effects. While pooled estimates of studies with follow‐up periods under 12 months were generally protective, those with follow‐up periods in excess of 12 months were not.

Quality of the evidence

The quality of evidence for the effects of the individual interventions on diarrhoea ranged from moderate (for ceramic filters and biosand filtration), to low (for distribution of chlorination kits, flocculation and disinfection sachets, and LifeStraw® filters), to very low (for water source improvements).

The primary reason for downgrading the quality of evidence was the risk of bias inherent in unblinded studies evaluating the efficacy of an intervention on a self‐reported outcome. Notably, only one of the nine blinded trials reported a statistically significant protective effect, but this observation may be explained by other confounding factors present in these nine trials (see Table 24):

Table 24. Potential reasons for finding of no‐effect in trials with adequate blinding

Study	Risk from ambient water quality	Compliance	Other issues
Colford 2002 USA	Very low (USA)	High (Sham filter)	None
Colford 2005 USA	Very low (USA)	High (Sham filter)	None
Colford 2009 USA	Very low (USA)	High (Sham filter)	None
Rodrigo 2011 AUS	Very low (Australia)	Not reported	None
Jain 2010 GHA	Low (11 CFU/100 mL)	High (RFC)	Control group received jerry can; 13 week follow‐up
Kirchhoff 1985 BRA	Very high (mean 16000 FC/dL)	Not reported	Only 112 persons from 16 households; 18 week trial
Austin 1993	High (1871 FC/100 mL)	Low ("50% to 60%")	No test of blinding; not peer reviewed
Boisson 2010 DRC	High (75% of samples > 1000 TTC/100 mL)	High, but 73% of adults and 95% of children drank from untreated sources	"Placebo" removed > 90% of TTC in control arm
Boisson 2013 IND	Moderate (mean 122 TTC/100 mL)	Low and inconsistent (32% of samples positive for RFC)	None

Abbreviations: TTC: thermotolerant coliforms, CFU: colony‐forming units, FC: faecal coliforms, RFC: residual free chlorine.

Four studies were conducted in high‐income countries where the water was of good microbiological quality even in the control groups (Colford 2002 USA; Colford 2005 USA; Colford 2009 USA; Rodrigo 2011 AUS).
One further trial from Ghana found very low levels of faecal contamination of water supplies in the control group which were likely to present only minimal risk (Jain 2010 GHA).
Three studies had either low adherence with the intervention (Austin 1993; Boisson 2013 IND), or very high reported use of drinking untreated water from other sources (Boisson 2010 DRC).
Two studies employed control interventions which may have improved water quality: Boisson 2010 DRC employed a "placebo" that actually removed one log (90%) of faecal indicator bacteria and Jain 2010 GHA provided control households with safe storage.

The second common reason for downgrading the quality of evidence was unexplained heterogeneity. For some of the POU interventions, the protective effect varied considerably across studies. Some of this variability could be explained by adherence with the intervention, with larger effects in studies with higher adherence, but some variability remained which we were unable to explain despite multiple subgroup analyses. This is likely to reflect important underlying clinical heterogeneity: the aetiology and epidemiology of diarrhoea is complex and variable, transmission pathways are multiple, and even the portion of diarrhoea that is waterborne is not well understood (Eisenberg 2012).

There was also some evidence of possible publication bias in the trials evaluating home chlorination but this was not strong enough to further downgrade the quality of evidence.

Potential biases in the review process

A number of the included studies had multiple intervention arms comparing two or more intervention groups against a single control group. In some analyses, we included multiple comparisons from the same trial which double counts the control group participants and yields results in the meta‐analysis that are artificially precise. However, this bias is unlikely to have significantly impacted the overall quality of evidence or conclusions.

Agreements and disagreements with other studies or reviews

Our results are generally consistent with the prior version of this Cochrane Review (Clasen 2006) and with other reviews of water quality interventions (Fewtrell 2005; Arnold 2007; Waddington 2009; Cairncross 2010; Wolf 2014).

One additional review of water quality interventions reports no effect with POU interventions once blinding is taken into account (Engell 2013). While we share the concerns about the lack of blinding in many of these trials (and have downgraded the quality of evidence accordingly), and also found no effect in any of the trials with adequate blinding, we have identified several possible confounders in this observation (discussed above), and retain low to moderate confidence that these interventions are effective.

Although we found no controlled trials evaluating piped‐in water supplies, a recent review that also included some observational studies reported some evidence of a protective effect with this intervention (Wolf 2014).

The finding of larger effects with increased adherence is consistent with modelling data based on quantitative microbial risk assessment which suggest a dose‐response association between water quality and diarrhoea (Brown 2012; Enger 2013).

Figure 1

Study flow diagram.

Figure 2

Risk of bias graph: summary of authors' judgements about each 'Risk of bias' item presented as percentages across all included studies.

Figure 3

Forest plot of comparison: 2 Source: water supply improvement versus control, outcome: 2.1 Diarrhoea: CBA studies subgrouped by age.

Figure 4

Forest plot of comparison: 3 POU: water chlorination versus control, outcome: 3.3 Diarrhoea: cluster‐RCTs; subgrouped by adherence.

Figure 5

Funnel plot of comparison: 3 POU: water chlorination versus control, outcome: 3.1 Diarrhoea: subgrouped by study design.

Figure 6

Forest plot of comparison: 4 POU: filtration versus control, outcome: 4.2 Diarrhoea: cluster‐RCTs: subgrouped by type of filtration.

Analysis 1.1

Comparison 1 Water quality intervention versus control, Outcome 1 Diarrhoea: all ages.

Analysis 1.2

Comparison 1 Water quality intervention versus control, Outcome 2 Diarrhoea: children < 5 years.

Analysis 2.1

Comparison 2 Source: water supply improvement versus control, Outcome 1 Diarrhoea: CBA studies subgrouped by age.

Analysis 2.2

Comparison 2 Source: water supply improvement versus control, Outcome 2 Diarrhoea: CBA studies subgrouped by age.

Analysis 3.1

Comparison 3 POU: water chlorination versus control, Outcome 1 Diarrhoea: subgrouped by study design.

Analysis 3.2

Comparison 3 POU: water chlorination versus control, Outcome 2 Diarrhoea: cluster‐RCTs: subgrouped by age.

Analysis 3.3

Comparison 3 POU: water chlorination versus control, Outcome 3 Diarrhoea: cluster‐RCTs; subgrouped by adherence.

Analysis 3.4

Comparison 3 POU: water chlorination versus control, Outcome 4 Diarrhoea: cluster‐RCTs by risk of bias by blinding of participants.

Analysis 3.5

Comparison 3 POU: water chlorination versus control, Outcome 5 Diarrhoea: cluster‐RCTs; subgrouped by additional water storage intervention.

Analysis 3.6

Comparison 3 POU: water chlorination versus control, Outcome 6 Diarrhoea: cluster‐RCTs: subgrouped by sufficiency of water quantity.

Analysis 3.7

Comparison 3 POU: water chlorination versus control, Outcome 7 Diarrhoea: cluster‐RCTs: subgrouped by water source.

Analysis 3.8

Comparison 3 POU: water chlorination versus control, Outcome 8 Diarrhoea: cluster‐RCTs: subgrouped by sanitation level.

Analysis 3.9

Comparison 3 POU: water chlorination versus control, Outcome 9 Diarrhoea: cluster‐RCTs; subgrouped by length of follow‐up.

Analysis 4.1

Comparison 4 POU: flocculation and disinfection versus control, Outcome 1 Diarrhoea: cluster‐RCTs.

Analysis 4.2

Comparison 4 POU: flocculation and disinfection versus control, Outcome 2 Diarrhoea: cluster‐RCTs: subgrouped by age; excluding Doocy 2006 LBR.

Analysis 4.3

Comparison 4 POU: flocculation and disinfection versus control, Outcome 3 Diarrhoea: cluster‐RCTs: subgrouped by adherence.

Analysis 4.4

Comparison 4 POU: flocculation and disinfection versus control, Outcome 4 Diarrhoea: cluster‐RCTs: subgrouped by additional storage container.

Analysis 4.5

Comparison 4 POU: flocculation and disinfection versus control, Outcome 5 Diarrhoea: cluster‐RCTs: subgrouped by sufficiency of water quantity.

Analysis 4.6

Comparison 4 POU: flocculation and disinfection versus control, Outcome 6 Diarrhoea: cluster‐RCTs: subgrouped by water source.

Analysis 4.7

Comparison 4 POU: flocculation and disinfection versus control, Outcome 7 Diarrhoea: cluster‐RCTs: subgrouped by sanitation level.

Analysis 4.8

Comparison 4 POU: flocculation and disinfection versus control, Outcome 8 Diarrhoea: cluster‐RCTs: subgrouped by length of follow‐up.

Analysis 5.1

Comparison 5 POU: filtration versus control, Outcome 1 Diarrhoea: cluster‐RCTs: subgrouped by age.

Analysis 5.2

Comparison 5 POU: filtration versus control, Outcome 2 Diarrhoea: cluster‐RCTs: subgrouped by type of filtration.

Analysis 5.3

Comparison 5 POU: filtration versus control, Outcome 3 Diarrhoea: cluster‐RCTs: subgrouped by blinding of participants.

Analysis 5.4

Comparison 5 POU: filtration versus control, Outcome 4 Diarrhoea: ceramic filter studies subgrouped by water source.

Analysis 5.5

Comparison 5 POU: filtration versus control, Outcome 5 Diarrhoea: ceramic filter studies subgrouped by sanitation level.

Analysis 5.6

Comparison 5 POU: filtration versus control, Outcome 6 Diarrhoea: sand filter studies: subgrouped by water source.

Analysis 5.7

Comparison 5 POU: filtration versus control, Outcome 7 Diarrhoea: sand filter studies: subgrouped by sanitation level.

Analysis 5.8

Comparison 5 POU: filtration versus control, Outcome 8 Diarrhoea: cluster‐RCTs: subgrouped by adherence.

Analysis 5.9

Comparison 5 POU: filtration versus control, Outcome 9 Diarrhoea: cluster‐RCTs: subgrouped by additional water storage intervention.

Analysis 5.10

Comparison 5 POU: filtration versus control, Outcome 10 Diarrhoea: cluster‐RCTs; subgrouped by length of follow‐up.

Analysis 6.1

Comparison 6 POU: solar disinfection versus control, Outcome 1 Diarrhoea: subgrouped by study design.

Analysis 6.2

Comparison 6 POU: solar disinfection versus control, Outcome 2 Diarrhoea: cluster‐RCTs; subgrouped by age.

Analysis 6.3

Comparison 6 POU: solar disinfection versus control, Outcome 3 Diarrhoea: cluster‐RCTs; subgrouped by adherence.

Analysis 6.4

Comparison 6 POU: solar disinfection versus control, Outcome 4 Diarrhoea: cluster‐RCTs; subgrouped by sufficiency of water supply level.

Analysis 6.5

Comparison 6 POU: solar disinfection versus control, Outcome 5 Diarrhoea: cluster‐RCTs; subgrouped by water source.

Analysis 6.6

Comparison 6 POU: solar disinfection versus control, Outcome 6 Diarrhoea: cluster‐RCTs; subgrouped by sanitation level.

Analysis 6.7

Comparison 6 POU: solar disinfection versus control, Outcome 7 Diarrhoea: cluster‐RCTs; subgrouped by length of follow‐up.

Analysis 7.1

Comparison 7 POU: UV disinfection versus control, Outcome 1 Diarrhoea: cluster‐RCT.

Analysis 8.1

Comparison 8 POU: improved storage versus control, Outcome 1 Diarrhoea: cluster‐RCTs: subgrouped by age.

Summary of findings for the main comparison. Summary of findings table 1

Point‐of‐use water quality interventions for preventing diarrhoea in rural settings in low‐ and middle‐income countries
Patient or population: adults and children Settings: low‐ and middle‐income countries in rural areas Intervention: point of use water quality interventions Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (trials)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
Diarrhoea episodes	No intervention	Chlorination	RR 0.77 (0.65 to 0.91)	30,746 (14 trials)	⊕⊕⊝⊝ low^1,2,3,4
	3 episodes per person per year	2.3 episodes (2.0 to 2.7)
	No intervention	Flocculation/disinfection	RR 0.69 (0.58 to 0.82)	11,788 (4 trials)	⊕⊕⊕⊝ moderate^1,3,4,5,6
	3 episodes per person per year	2.1 episodes (1.7 to 2.5)
	No intervention	Filtration	RR 0.48 (0.38 to 0.59)	15,582 (18 trials)	⊕⊕⊕⊝ moderate^1,3,4,5
	3 episodes per person per year	1.4 episodes (1.1 to 1.8)
	No intervention	Solar disinfection (SODIS)	RR 0.62 (0.42 to 0.94)	3460 (4 trials)	⊕⊕⊕⊝ moderate^1,3,4,5
	3 episodes per person per year	1.9 episodes (1.3 to 2.8)
The assumed risk is taken from Fischer Walker 2012 and represents an estimated average for the incidence of diarrhoea in low‐ and middle‐income countries. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
¹Downgraded by 1 for serious risk of bias: the outcome was measured as self‐reported episodes of diarrhoea, and is susceptible to bias as most studies were unblinded. ²Downgraded by 1 for serious inconsistency: statistical heterogeneity was very high with six out of fourteen trials having point estimates close to no effect. A subgroup analysis by adherence with the intervention (assessed by measurements of residual chlorine in drinking water) found larger effects in the studies with better adherence but the results remained inconsistent. ³No serious indirectness: these studies are mainly from low‐ and middle‐income countries, in settings with both improved and unimproved water sources and sanitation. ⁴No serious imprecision: The analysis is adequately powered to detect this effect. ⁵No serious inconsistency: The evidence of benefit is consistent across trials, but there is substantial statistical heterogeneity in the size of the effect. ⁶ This analysis excludes one additional study which found a much larger effect than seen in the other four trials and was considered an outlier (Doocy 2006 LBR).

Summary of findings for the main comparison. Summary of findings table 1

Table 1. Water quality indicators post‐intervention

Trial	Water quality indicator	Water quality post‐intervention: Intervention group	Water quality post intervention: Control group
Abebe 2014 ZAF	CFUs/100 mL	0	80% of control HHs had 10 to 10000
Austin 1993a GMB	Geometric mean CFUs/100 mL	178	3020
Austin 1993b GMB	Geometric mean CFUs/100 mL	42	3020
Boisson 2009 ETH	Arithmetic mean TTC/100 mL (95% CI)	0	725.7 (621.0 to 830.4)
Boisson 2010 DRC	Geometric mean TTC/100 mL (95% CI)	1.3 (0.9 to 1.7)	173.7 (136.6 to 220.9)
Boisson 2013 IND	Geometric mean TTC/100 mL (95% CI)	50 (44 to 57)	122 (107 to 139)
Brown 2008a KHM	Geometric mean E. coli /100 mL	17	600
Brown 2008b KHM	Geometric mean E. coli /100 mL	15	600
Clasen 2004b BOL	Mean TTC/100 mL	0.13	108
Clasen 2004c BOL	Arithmetic mean TTC/100 mL	100% of intervention households: 0	16% of control households: 0 66% > 10, 34% > 100, and 11% > 1000
Clasen 2005 COL	Arithmetic mean TTC/100 mL (95% CI)	37.3 (6.3 to 48.3)	150.6 (34.8 to 166.4)
Colford 2002 USA ; Colford 2005 USA ; Colford 2009 USA	All water met FDA requirements	Not measured because of high water quality	Not measured because of high water quality
Crump 2005a KEN	Samples met WHO guidelines for water quality	82%	14%
Crump 2005b KEN	Samples met WHO guidelines for water quality	78%	14%
du Preez 2008 ZAF/ZWE	Samples met WHO guidelines for water quality	57%	30%
du Preez 2010 ZAF	E. coli in concentrations/100 mL	62%	"No significant difference between intervention and control groups"
du Preez 2011 KEN	E. coli ln concentrations/100 mL	Storage containers: 0.723 SODIS bottles: ‐0.727	Not reported
Fabiszewski 2012 HND	Geometric mean E. coli counts per 100 mL (95% CI)	23.4 (20.2 to 27.0)	45.4 (38.6 to 53.4)
Gasana 2002 RWA	Total coliforms/100 mL	Range: 3 to 43	Range: 4 to 1100
Gruber 2013 MEX	Samples with detectableE. coli	43%	59%
Günther 2013 BEN	E. coli contamination > 1000 CFU/100 mL	Not reported specifically; findings imply a 70% reduction in E. coli incidence for intervention households
Handzel 1998 BGD	Stored water samples with E. coli 100 MPN/100 mL	3%	16%
Jain 2010 GHA	Samples with E. coli	8%	54%
Jensen 2003 PAK	Geometric mean E. coli /100 mL	3	49
Kirchhoff 1985 BRA	Mean number of faecal coliforms/dL in the samples	70	16000
Kremer 2011 KEN	Average reduction in log E. coli	‐1.07, corresponding to a 66% reduction
Lule 2005 UGA	Median E. coli CFU/100 mL	23	59
McGuigan 2011 KHM	Geometric mean CFU/100 mL	6.8	48
Mengistie 2013 ETH	Mean E. coli	0	60
Peletz 2012 ZMB	Geometric mean TTC/100 mL	Stored water: 3	Stored water: 181
Quick 1999 BOL	Median E. coli /100 mL	0	6400
Quick 2002 ZMB	Median E. coli /100 mL	0	3
Reller 2003a GTM	Samples with < 1 E. coli /100 mL (flocculant/disinfectant)	40%	7%
Reller 2003b GTM	Samples with < 1 E. coli /100 mL (flocculant/disinfectant+ vessel)	57%	7%
Reller 2003c GTM	Samples with < 1 E. coli /100 mL (bleach)	51%	7%
Reller 2003d GTM	Samples with < 1 E. coli /100 mL (bleach + vessel)	61%	7%
Semenza 1998 UZB	Faecal colonies/100 mL	47	52
Stauber 2009 DOM	E. coli MPN/100 mL	11	19
Stauber 2012a KHM	E. coli CFU/100 mL	2.9	19.7
Stauber 2012b GHA	Geometric mean E. coli MPN/100 mL (95% CI)	Direct filtrate 16 (13 to 20) Stored filtrate: 76 (62 to 91)	490 (426 to 549)
Tiwari 2009 KEN	Geometric mean faecal coliforms/100 mL (95% CI)	30.0 (21.3 to 42.1)	88.9 (58.7 to 135)
URL 1995a GTM	Samples with fecal coliforms	91% had 0 fecal coliforms	Not reported
URL 1995b GTM	Samples with fecal coliforms	91% had 0 fecal coliforms	Not reported
Abbreviations: E. coli: Escherichia coli; FC: faecal coliform.

Table 1. Water quality indicators post‐intervention

Table 2. Studies reporting deaths

Study ID	Intervention		Control		P value	Comment
Study ID	Deaths	Participants	Deaths	Participants	P value	Comment
Boisson 2010 DRC	12	546	8	598	0.27	—
Colford 2009 USA	7	385	6	385	> 0.05	—
Crump 2005a KEN	17	2249	28	2277	0.108	—
Crump 2005b KEN	14	2124	28	2277	0.052	—
du Preez 2011 KEN	3	555	3	534	> 0.05	—
Peletz 2012 ZMB	3	300	6	299	0.28	—
Boisson 2013 IND	?	6119	?	5965	—	Only reports total deaths (46)
du Preez 2010 ZAF	?	383	?	335	—	Only reports total deaths (7)
Kremer 2011 KEN	?	—	?	—	—	Reports recording deaths but does not state how many
Boisson 2009 ETH	?	731	?	785	—	Reports recording deaths but does not state how many

Table 2. Studies reporting deaths

Table 3. Summary of findings: improved water source

Improved water source compared with no intervention for preventing diarrhoea in rural settings in low‐ and middle‐income countries
Patient or population: adults and children Settings: low‐ and middle‐income countries in rural areas Intervention: water source improvement Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	Water source improvement
Diarrhoea episodes Cluster‐RCTs	3 episodes per person per year	3.7 episodes per person per year (2.9 to 4.7)	RR 1.24 (0.98 to 1.57)	3266 (1 trial)	⊕⊝⊝⊝ very low^1,2,3
Diarrhoea episodes CBA studies	—	—	—	5895 (5 studies)	⊕⊝⊝⊝ very low^1,4,5
The basis for the assumed risk (for example, the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012). ¹Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. None of these studies blinded participants and outcome assessors to the treatment allocation. ²No serious inconsistency. ³Downgraded by 2 for serious indirectness: this single RCT from Afghanistan evaluated the provision of protected wells. It is not possible to make broad generalizations to other settings. ⁴Downgraded by 1 for serious inconsistency: statistical heterogeneity was very high (I² statistic = 98%), such that the data could not be pooled. Some large and statistically significant effects were seen in some individual trials, but not others. ⁵Downgraded by 1 for serious indirectness: these studies are from a variety of low‐ and middle‐income countries (Bangladesh, Rwanda, Pakistan, South Africa, China). However, as only single trials evaluated each intervention it is not possible to make broad generalizations.

Table 3. Summary of findings: improved water source

Table 4. Improved water source: description of the interventions

Study ID	Study design	Setting	Incidence of diarrhoea in the control group	Intervention areas		Control areas
Study ID	Study design	Setting	Incidence of diarrhoea in the control group	Water source intervention	Health promotion activities	Water source	Health promotion activities
Opryszko 2010b AFG	Cluster‐RCT	Rural villages	3.1 episodes per person per year	One well per 25 households providing 25 litres/person/day	None	35% used unprotected hand dug wells	None
Alam 1989 BGD	CBA	Rural villages	4.1 episodes per child per year	Provision of one hand pump per 4‐6 households (3 times as many as control areas)	Female health visitors visited peoples homes and organised group discussion and demonstrations to promote hygienic practices for hand pump use, water storage, child faeces disposal, hand washing.	Shallow, hand‐dug wells; some hand pumps	None described
Gasana 2002 RWA	CBA	Rural villages	3 episodes per child per year	Site A: Sedimentation tank/Katadyn filter with communal tap Site B: Gravel‐sand‐charcoal filter on existing water spring Site C: Protective fence around an existing water spring	None described	An existing water spring	None described
Jensen 2003 PAK	CBA	Rural villages	2.8 episodes per person per year	Chlorination of public water supply	None described	Unchlorinated poorly functioning sand filter system	None described
Majuru 2011 ZAF	CBA	Rural villages	0.6 episodes per person per year	Provision of intermittently operated small community water systems distributing potable water to multiple taps throughout the community	None described	Untreated water from a river and its tributaries	None described
Xiao 1997 CHN	CBA	Rural villages	Not reported	Improved water supply through structural improvements to wells	Hygiene education	Not reported	None described

Table 4. Improved water source: description of the interventions

Table 5. Improved water source: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient water quality	Sanitation⁴
Alam 1989 BGD	Shallow, hand‐dug wells; some hand pumps	Unimproved	Unclear	Unclear	Not tested	Unclear
Gasana 2002 RWA	Spring	Unimproved	Unclear	Unclear	Baseline range 4 to 1100 total coliforms/100 mL	Unimproved
Jensen 2003 PAK	Some slow sand filters in poor condition; some household taps; majority used ground water	Improved	Unclear	Unclear	Baseline geometric mean in intervention village: 13.3 E. coli CFU/100 mL; control villages: 137/100 mL	Unclear
Majuru 2011 ZAF	Surface water, boreholes, water tankers	Improved and unimproved	Unclear	Unclear	Not tested	Unclear
Opryszko 2010	35% use unprotected dug wells	Unimproved	Sufficient	Sufficient	Not tested	Unclear
Xiao 1997 CHN	Well water	Unimproved	Unclear	Unclear	Not tested	Unclear
¹'Improved' includes household connection, public standpipe, borehole, protected dug well, protected spring, rainwater collection; 'unimproved' includes unprotected well, unprotected spring, vendor‐provided water, bottled water; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015. ²'Sufficient' means located within 500 m, queuing no more than 15 minutes, no more than three minutes to fill 20 L container, and maintained so available consistently; 'insufficient' means that it does not meet any of above; and 'unclear' means unclear or not reported; definition based minimum standards established by The Sphere Project 2011. ³'Sufficient' means a minimum of 15 L/day/person; 'insufficient' means less than 15 L/day/person; and 'unclear' means unclear or not reported; definition based on minimum standards established by The Sphere Project 2011. ⁴'Improved' means connection to a public sewer or septic system, pour flush latrine, simple pit latrine, or ventilated improved pit latrine; 'unimproved' means service or bucket latrine, public latrines, open latrines; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015.

Table 5. Improved water source: primary drinking water supply and sanitation facilities

Table 6. POU chlorination: description of the intervention

Trial	Study design	Chlorination product?	Distributed free?	Frequency of distribution?	Storage container also distributed?	Compliance	Additional hygiene promotion
Austin 1993a GMB	Cluster‐RCT	Sodium hypochlorite solution	Yes	Fortnightly	No	40% compliance measured by residual chlorine	None
Austin 1993b GMB	Cluster‐RCT	Sodium hypochlorite solution	Yes	Fortnightly	No	59% compliance measured by residual chlorine	None
Boisson 2013 IND	Cluster‐RCT	Sodim dichloro‐isocyanurate tablets	Yes	Bimonthly	No	32% compliance measured by residual chlorine	None
Crump 2005a KEN	Cluster‐RCT	1% sodium hypochlorite	Yes	Weekly	No	61% compliance during unannounced weekly visits measured by residual chlorine	Use of ORS, treatment seeking for diarrhoea
Handzel 1998 BGD	Cluster‐RCT	0.25% to 0.3% chlorine solution	Yes	Weekly	Yes	90% compliance based on residual chlorine measurements	Hygiene and sanitation messages
Jain 2010 GHA	Cluster‐RCT	Sodim dichloro‐isocyanurate tablets	Yes	Twice weekly	Yes	74% to 89% compliance measured by chlorine residual	ORS provided to those with diarrhoea
Kirchhoff 1985 BRA	Cluster‐RCT	10% sodium hypochlorite	Yes	Daily	No	Not reported	Chlorination preformed by study staff
Luby 2006a PAK	Cluster‐RCT	Sodium hypochlorite solution	Yes	Unclear	Yes	Yes, though rate unclear	Encouraged to only drink treated water
Lule 2005 UGA	Cluster‐RCT	0.5% sodium hypochlorite	Yes	Weekly	Yes	Not reported	hygiene education
Mahfouz 1995 KSA	Cluster‐RCT	Packets of 50 g calcium hypochloride 70%.	Yes	Unclear	No	Some residual chlorine in all intervention samples	None
Mengistie 2013 ETH	Cluster‐RCT	1.25% sodium hypochlorite solution	Yes	Weekly	No	80% compliance measured by chlorine residual	None
Opryszko 2010c AFG	Cluster‐RCT	0.05% sodium hypochlorite solution	Yes	Monthly	Yes	78% compliance measured by previous 2 weeks self‐report use of chlorine	None
Quick 1999 BOL	Cluster‐RCT	MIOX unit electrolytically produced disinfectant with 3% brine solution, hypochlorite, chlorine dioxide, ozone, peroxide and other oxidants.	Yes	Weekly	Yes	63% compliance measured by water in vessel with chlorine residual, average across six rounds	Community health volunteers reinforced messages about proper use of the disinfectant and vessels and of different applications for treated water.
Reller 2003b GTM	Cluster‐RCT	Sodium hypochlorite solution (50,000 ppm)	Yes	Monthly	No	36% compliance measure by residual chlorine > 0.1 mg/L on unannounced visits.	Motivational and educational messages about chlorination, use of ORS, care seeking for diarrhoea
Reller 2003c GTM	Cluster‐RCT	Sodium hypochlorite solution (50,000 ppm)	Yes	Monthly	Yes	44% compliance measure by residual chlorine > 0.1 mg/L on unannounced visits.	Motivational and educational messages about chlorination, use of ORS, care seeking for diarrhoea
Semenza 1998 UZB	Cluster‐RCT	1.5% chlorine solution	Yes	Unclear but households were visited twice weekly	Yes	73% based on residual chlorine levels at time of visit	Only drink chlorinated water and wash all fruit and vegetables with chlorinated water
Luby 2004a PAK	CBA	Bleach (sodium hypochlorite)	Yes	Study workers visited weekly and re‐supplied the households with dilute bleach.	Yes	Not reported	Encouraged regular treatment of drinking water
Luby 2004b PAK	CBA	Bleach (sodium hypochlorite)	Yes	Study workers visited weekly and re‐supplied the households with dilute bleach.	Yes	Not reported	Encouraged regular treatment of drinking water
Quick 2002 ZMB	CBA	0.5% sodium hypochlorite	Yes	Unclear but households were visited once every two weeks	HHs paid for vessel	72% compliance measured by water in vessel with chlorine residual	Community volunteers, gave education about causes and prevention of diarrhoea and safe storage of water and motivated households about the intervention.

Table 6. POU chlorination: description of the intervention

Table 7. POU chlorination: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient water quality	Sanitation⁴
Austin 1993	Open wells	Unimproved	Sufficient	Unclear	Mean 1871 FC/100 mL in wells; among stored water samples: mean 3358 FC/100 mL in rainy season, 1014 FC/100 mL in dry season	Unclear
Boisson 2013 IND	62% unprotected dug well, 17% tubewell, 14% tap, 5% surface water	Unimproved	Unlcear	Unclear	Baseline not reported. Control households: Geometric mean 122 TTC/100 mL	Unimproved
Crump 2005	50% ponds, 49% rivers	Unimproved	Unclear	Insufficient	Baseline mean 98 E. coli /100 mL	Unclear; 33% defecate on ground
Handzel 1998 BGD	48% tap, 52% tubewell; 61% paid for drinking water	Improved	Sufficient	Sufficient	Baseline geometric mean 138.1 faecal coloform counts/100 mL	Unimproved
Jain 2010 GHA	95% of households use tap, 84% surface water, 46% wells, 35% rainwater, 25% borehole	Improved and unimproved	Unclear	Unclear	Baseline: median E. coli MPN 93/100 mL	Unimproved
Kirchhoff 1985 BRA	Pond water stored in clay pots after filtering with cloth	Unimproved	Unclear	Insufficient	Source water: mean 970 faecal coliforms/100 mL	Unimproved
Luby 2004	Tanker trucks, municipal taps (household and community level)	Mostly unimproved	Unclear	Unclear	Baseline: approximately 60% of stored drinking water samples were free of E. coli	Improved
Luby 2006	Tanker trucks, municipal taps (household and community level), water bearer, boreholes	Mostly improved	Unclear	Unclear	Not tested	Improved
Lule 2005 UGA	16% surface or shallow wells, 50% protected springs, 49% boreholes or taps	Unimproved	Sufficient	Sufficient	Source mean E. coli counts: 11/100 mL	Improved
Mahfouz 1995 KSA	Shallow wells	Unimproved	Unclear	Unclear	Source: 92% positive with E. coli; precise level not reported	Improved
Mengistie 2013 ETH	50% well, 41% spring, 9% river	Unimproved	Unclear	Unclear	Baseline: E. coli MPN 70/100 mL	Unimproved
Opryszko 2010	35% use unprotected dug wells	Unimproved	Sufficient	Sufficient	Not tested	Unclear
Quick 1999 BOL	Shallow uncovered wells; 38% treated water	Unimproved	Unclear	Unclear	Source water: median colony count E. coli: 57,050/100 mL	Unimproved, but 47% used latrine
Quick 2002 ZMB	Shallow wells; some boiling	Unimproved	Unclear	Unclear	Source water: median colony count E. coli: 34/100 mL	Unclear
Reller 2003	Surface water from shallow wells, rivers and springs	Unimproved	Unclear	Unclear	Baseline drinking water: median colony count E. coli 63/100 mL	Unclear
Semenza 1998 UZB	Households without piped water (procured from street tap, neighbour tap, well, vendor, or river)	Unimproved	Unclear	Unclear	Source water: 54 coliform colonies/100 mL	Unclear
¹'Improved' includes household connection, public standpipe, borehole, protected dug well, protected spring, rainwater collection; 'unimproved' includes unprotected well, unprotected spring, vendor‐provided water, bottled water; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015. ²'Sufficient' means located within 500 m, queuing no more than 15 minutes, no more than three minutes to fill 20 L container, and maintained so available consistently; 'insufficient' means that it does not meet any of above; and 'unclear' means unclear or not reported; definition based minimum standards established by The Sphere Project 2011. ³'Sufficient' means a minimum of 15 L/day/person; 'insufficient' means less than 15 L/day/person; and 'unclear' means unclear or not reported; definition based on minimum standards established by The Sphere Project 2011. ⁴'Improved' means connection to a public sewer or septic system, pour flush latrine, simple pit latrine, or ventilated improved pit latrine; 'unimproved' means service or bucket latrine, public latrines, open latrines; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015.

Table 7. POU chlorination: primary drinking water supply and sanitation facilities

Table 8. Summary of findings: POU chlorination

POU chlorination compared with no intervention for preventing diarrhoea
Patient or population: adults and children Settings: low‐ and middle‐income countries Intervention: distribution of chlorine for POU water treatment and instruction on use Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	POU Chlorination
Diarrhoea episodes cluster‐RCTs	3 episodes per person per year	2.3 episodes per year (2.0 to 2.7)	RR 0.77 (0.65 to 0.91)	30,746 (14 trials)	⊕⊕⊝⊝ low^1,2,3,4
Diarrhoea episodes CBA studies	3 episodes per person per year	1.5 episodes per year (1.0 to 2.3)	RR 0.51 (0.34 to 0.75)	3948 (2 studies)	⊕⊝⊝⊝ very low^5,6,7,8
The basis for the assumed risk is provided in the footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012). ¹Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. Only two of these studies blinded participants and outcome assessors to the treatment allocation, and these two studies found no evidence of an effect with chlorination. ²Downgraded by 1 for serious inconsistency: statistical heterogeneity was very high (I² statistic = 91%). In a subgroup analysis by compliance with the intervention (assessed by measurements of residual chlorine in drinking water) found larger effects in the studies with better compliance. ³No serious indirectness: these studies are mainly from low‐ and middle‐income countries (the Gambia, India, Kenya, Bangladesh, Ghana, Brazil, Pakistan,Uganda, Saudi Arabia, Ethiopia, Afghanistan, Bolivia, Guatemala, and Uzbekistan). The interventions consisted of free distribution of chlorine (every one to four weeks) plus instructions on how to use it. In some cases, the intervention included hygiene education and storage containers in which to treat and store water. ⁴No serious imprecision: the average effect suggests POU chlorination may reduce diarrhoea episodes by about a quarter. The analysis is adequately powered to detect this effect. ⁵Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. None of these studies blinded participants and outcome assessors to the treatment allocation. ⁶Downgraded by 1 for serious inconsistency: statistical heterogeneity was very high (I² statistic = 63%). ⁷Downgraded by 1 for serious indirectness: there are only two studies (three comparisons) from Pakistan and Zambia. ⁸No serious imprecision.

Table 8. Summary of findings: POU chlorination

Table 9. POU flocculation/disinfection: description of the interventions

Study ID	Study design	Setting	Intervention areas			Control areas
Study ID	Study design	Setting	Water quality intervention	Health promotion activities	Compliance	Water source	Health promotion activities
Chiller 2006 GTM	Cluster‐RCT	Rural villages	Provided households with a large spoon and a wide‐mouthed bucket for mixing, a narrow‐topped vessel with a lid for storing treated water and provided households with sachets of the flocculant–disinfectant every week	None	44% compliance measured by residual chlorine at week 10 of study	31% tap, 40% river or spring and 25% well.	None
Crump 2005b KEN	Cluster‐RCT	Rural villages	Each week households were given sachets of the flocculant–disinfectant	None	44% compliance during unannounced weekly visits measured by residual chlorine	50% pond, 49% river and 2% spring	None
Doocy 2006 LBR	Cluster‐RCT	Liberian camps for displaced persons	Households received a bucket and large mixing spoon for preparation, a decanting cloth, a funnel and a storage container with a narrow opening and lid. Each household received a maximum of 21 flocculation–disinfectant packets per week	None	85% compliance based on residual chlorine sampling	Received a funnel and an identical storage container	None
Luby 2006b PAK	Cluster‐RCT	Squatter settlements	Provided households with flocculant‐disinfectant sachets, a water vessel and soap. Weekly distributions of sachets	Field workers educated neighbourhoods about health problems resulting from hand and water contamination and instructed households on how and when to wash hands	Yes, though rate unclear	Municipal supply at household (33%), at community tap (37%), tanker truck (12%), water bearer (13%) and tube well (5%)	None
Luby 2006c PAK	Cluster‐RCT	Squatter settlements	Flocculant‐disinfectant and vessel. Weekly distributions of sachets	Field workers educated neighbourhoods about health problems resulting from hand and water contamination	Yes, though rate unclear	Municipal supply at household (33%), at community tap (37%), tanker truck (12%), water bearer (13%) and tube well (5%)	None
Reller 2003a GTM	Cluster‐RCT	Rural villages	Weekly distribution of flocculant‐disinfectant and gave 2 cloths initially, which could be exchanged	Field workers discussed the importance of water treatment and demonstrated the water preparation process	27% compliance measure by residual chlorine > 0.1 mg/L on unannounced visits.	33% tap, 46% river or spring, 21% well.	None
Reller 2003d GTM	Cluster‐RCT	Rural villages	Weekly distribution of flocculant‐disinfectant and gave 2 cloths initially, which could be exchanged and received a large plastic spoon for stirring, a large‐mouthed bucket for mixing, and a vessel with a secure lid and a spigot for storing treated water	Field workers discussed the importance of water treatment and demonstrated the water preparation process	34% compliance measure by residual chlorine > 0.1 mg/L on unannounced visits.	33% tap, 46% river or spring, 21% well.	None

Table 9. POU flocculation/disinfection: description of the interventions

Table 10. POU flocculation/disinfection: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient H₂O quality	Sanitation⁴
Chiller 2006 GTM	Rivers, springs, taps, and wells	Unclear	Unclear	Sufficient	98% of source samples contained E. coli; precise level not reported	Mostly unimproved
Crump 2005b KEN	50% ponds, 49% rivers	Unimproved	Unclear	Insufficient	Baseline mean 98 E. coli /100 mL	Unclear; 33% defecate on ground
Doocy 2006 LBR	Surface sources and some tap stands	Unimproved	Unclear	Insufficient	Source water: 88% samples tested positive for faecal contamination; precise level not reported	Unimproved
Luby 2006b PAK	Tanker trucks, municipal taps (household and community level), water bearer, boreholes	Mostly improved	Unclear	Unclear	Not tested	Improved
Reller 2003a GTM	Surface water from shallow wells, rivers and springs	Unimproved	Unclear	Unclear	Baseline drinking water: median colony count E. coli 63/100 mL	Unclear
¹'Improved' includes household connection, public standpipe, borehole, protected dug well, protected spring, rainwater collection; 'unimproved' includes unprotected well, unprotected spring, vendor‐provided water, bottled water; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015. ²'Sufficient' means located within 500 m, queuing no more than 15 minutes, no more than three minutes to fill 20 L container, and maintained so available consistently; 'insufficient' means that it does not meet any of above; and 'unclear' means unclear or not reported; definition based minimum standards established by The Sphere Project 2011. ³'Sufficient' means a minimum of 15 L/day/person; 'insufficient' means less than 15 L/day/person; and 'unclear' means unclear or not reported; definition based on minimum standards established by The Sphere Project 2011. ⁴'Improved' means connection to a public sewer or septic system, pour flush latrine, simple pit latrine, or ventilated improved pit latrine; 'unimproved' means service or bucket latrine, public latrines, open latrines; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015.

Table 10. POU flocculation/disinfection: primary drinking water supply and sanitation facilities

Table 11. Summary of findings: POU flocculation and disinfection

POU water flocculation and disinfection compared with no intervention for preventing diarrhoea
Patient or population: adults and children Settings: low‐ and middle‐income countries Intervention: distribution of sachets combining water flocculation and disinfection and instructions on use Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	Water flocculation and disinfection
Diarrhoea episodes Cluster‐RCTs	3 episodes per person per year	2.1 episodes per person per year (1.7 to 2.5)	RR 0.69 (0.58 to 0.82)	11,788 (4 trials)	⊕⊕⊕⊝ moderate^1,2,3,4
The basis for the assumed risk is provided in the footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012). ¹Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. None of these studies blinded participants and outcome assessors to the treatment allocation. ²No serious inconsistency: In the complete analysis of five trials statistical heterogeneity was very high (I² statistic = 99%). However, this heterogeneity was related to a single trial showing very large effects conducted in an emergency setting in Liberia possibly due to epidemic diarrhoea. When this trial was removed as an outlier, there was a smaller, but more consistent effect. ³No serious indirectness: the studies were conducted in rural areas in Guatemala (two studies), and Kenya (one study), one trial was from a camp for displaced persons in Liberia and one from squatter settlements in Pakistan. Sanitation was improved in only one of these studies. ⁴No serious imprecision: all five studies found benefits with flocculation. The 95% CI of the pooled effect includes the possibility of no effect, but this imprecision is a result of the heterogeneity between studies.

Table 11. Summary of findings: POU flocculation and disinfection

Table 12. POU filtration: description of interventions

Study ID	Intervention sub‐group	Study design	Setting	Intervention areas			Control areas
Study ID	Intervention sub‐group	Study design	Setting	Water quality intervention	Health promotion activities	Compliance	Water source	Health promotion activities
Abebe 2014 ZAF	Ceramic filter	Cluster‐RCT	Rural	Ceramic water filter impregnated with silver nanoparticles with safe storage containers	Education about safe water and hygiene and information on how to use the filter and maintain it.	Not reported	Personal tap in home (44%), community tap (44%) and river (3%)	Received usual clinical care including education about safe water and hygiene at the clinic
Brown 2008a KHM	Ceramic filter	Cluster‐RCT	Rural	CWP (Cambodian Ceramic Water Purifier) including safe storage container.	None	98% compliance measured by self‐report	Surface water (55%) and ground water (48%) during the dry season and surface water (45%), ground water (48%) and rain water (73%) during the rainy season	None
Brown 2008b KHM	Ceramic filter	Cluster‐RCT	Rural	CWP‐Fe (iron‐rich ceramic water purifier) including safe storage container.	None	98% compliance measured by self‐report	Surface water (55%) and ground water (48%) during the dry season and surface water (45%), ground water (48%) and rain water (73%) during the rainy season	None
Clasen 2004b BOL	Ceramic filter	Cluster‐RCT	Rural	Ceramic filters including improved storage	None	67% of households had filters in regular use	68% had taps and 11% boiled water.	None
Clasen 2004c BOL	Ceramic filter	Cluster‐RCT	Rural	Ceramic filters including improved storage	None	100% of intervention households' water free of TTC	Water from canal (52%), river (35%) or rainwater (4%)	None
Clasen 2005 COL	Ceramic filter	Cluster‐RCT	Rural and urban affected by conflict	Ceramic water filter system including improved storage	None	Not reported	River (27.6%), rainwater(12.1%), yard tap (67.2%). 70.7% claimed to treat water.	None
du Preez 2008 ZAF/ZWE	Ceramic filter	Cluster‐RCT	Rural	Ceramic filters including improved storage	None	55% compliance measured by water quality (approximate compliance across intervention households in Zimbabwe and South Africa).	Protected water source (53.8%) and unprotected water source (46.2%)	None
Lindquist 2014a BOL	Ceramic filter	Cluster‐RCT	Peri‐urban	Received a PointONE Filter and a 30 L bucket (with lid)	Participants were instructed on diarrhoeal transmission (biological versus cultural beliefs‐based), prevention and treatment.	97% compliance based on reported use	83% used water from tanker trucks and 12% from water coolers.	Received weekly messages on life skills and attitudes. Also were instructed on diarrhoeal transmission, prevention and treatment.
Lindquist 2014b BOL	Ceramic filter	Cluster‐RCT	Peri‐urban	Received a PointONE Filter and a 30‐L bucket (with lid) and WASH education	Participants received weekly WASH messages on personal and family hygiene, sanitation, boiling and chlorine‐based water treatments (excluding filtration),vitamin A, hygienic food preparation and cleaning, and parasite prevention.	90% compliance based on reported use	83% used water from tanker trucks and 12% from water coolers.	Received weekly messages on life skills and attitudes. Also were instructed on diarrhoeal transmission, prevention and treatment.
URL 1995a GTM	Ceramic filter	Cluster‐RCT	Rural	Handmade ceramic water filter	None	87% to 93% use of filter by children	Majority of households collected water from household tap (not chlorinated)	None
URL 1995b GTM	Ceramic filter	Cluster‐RCT	Rural	Handmade ceramic water filter	Education on nutrition (ORS, basic nutrition and maternal and child nutrition), health (hygiene) and family values.	As above	Majority of households collected water from household tap (not chlorinated)	None
Fabiszewski 2012 HND	Sand filtration	Cluster‐RCT	Rural	Hydraid plastic‐housing BioSand filter (BSF) + 20 L water jug	Training for the use and maintenance of the BSF and general education about hygiene and sanitation.	Not reported	Among all study participants‐ the main source of drinking water were: protected water sources (49% to 69% households per month), protected sources (24% to 50% per month), piped water (1% to 11% per month), and rainwater (0% to 2% per month).	Training for the use and maintenance of the BSF and general education about hygiene and sanitation.
Stauber 2009 DOM	Sand filtration	Cluster‐RCT	Semi‐rural and urban	Received a biosand filter and safe storage container	Nothing	Water quality testing, however no intervention household level compliance reported	42% reported treating drinking water.	None
Stauber 2012a KHM	Sand filtration	Cluster‐RCT	Rural	Plastic biosand filter. HHs were asked to pay USD 10 for the filter.	Health and hygiene education sessions	89% compliance measured by household‐reported use at least 3 times per week	Improved water sources during the dry season (7.1%) and during the rainy season (88.9%). 49.5% reported boiling drinking water.	Health and hygiene education sessions
Stauber 2012b GHA	Sand filtration	Cluster‐RCT	Rural	Plastic biosand filter	Not specified	97% compliance measured by household‐reported use	Use surface water during dry season (95%) and use surface water during rainy season (70.6%). 96.5% reported sieving drinking water through cloth.	nothing
Tiwari 2009 KEN	Sand filtration	Cluster‐RCT	Rural	Provided with the concrete BioSand Filter	At each visit, three oral rehydration packets and instructions were provided.	Not reported	All control houses reported drinking river or unprotected spring water; drink rainwater (96.6%), drink improved source (24.1%). 34.5% reported boiling drinking water.	At each visit, three oral rehydration packets and instructions were provided.
Boisson 2009 ETH	LifeStraw® Personal	Cluster‐RCT	Rural	A LifeStraw® personal pipe‐style water treatment device was given to each member of the household >6 months and encouraged to use it at home and away from home.	None	13% report use today	The primary drinking water source for 84% was from spring, 12% from rivers, 2.5% from hand dug wells and 4% from communal taps.	None
Boisson 2010 DRC	LifeStraw® Family	Cluster‐RCT	Rural	Households received a LifeStraw® Family filters	None	76% compliance measured by self‐report use today or yesterday (at 14 month follow‐up)	Received a placebo filter.	None
Peletz 2012 ZMB	LifeStraw® Family	Cluster‐RCT	Peri‐urban	Households received a LifeStraw® Family filter and two 5 L safe storage containers.	None	87% compliance measured by improved water quality	46% use unprotected dug wells, 19% boreholes, 17% public standpipes, 12% protected dug well, 5% piped into home or yard and 2% surface water.	None
Colford 2002 USA	Plumbed in filter	Cluster‐RCT	Urban	Installation of water treatment devices to 1 tap in HH that include: a 1‐micron absolute prefilter cartridge and a UV lamp.	None	96% compliance measured by not dropping out of study (plumbed‐in unit)	Sham device	None
Colford 2005 USA	Plumbed in filter	Cluster‐RCT	Urban	Installation of filter (1‐micron filter and a UV lamp) to main faucet of household	All participants received the current CDC safe drinking water guidelines for immuno‐compromised persons	90% compliance measured by not dropping out of study (filter attached to kitchen sink)	Sham device	All participants received the current CDC safe drinking water guidelines for immuno‐compromised persons
Colford 2009 USA	Plumbed in filter	Cluster‐RCT	Urban	Installation of filter (1‐micron filter and a UV lamp) to main faucet of household	None	83% compliance measured by not dropping out of study (filter attached to kitchen sink)	Sham device	None
Rodrigo 2011 AUS	Ceramic filter/plumbed in	Cluster‐RCT	Urban	Bench‐top silver impregnated ceramic water treatment units, which required participants to use fill it but then households that had rainwater piped into kitchen were offered an under sink unit	None	Not reported	Sham water treatment unit	None

Table 12. POU filtration: description of interventions

Table 13. POU filtration: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient H₂O quality	Sanitation⁴
Abebe 2014 ZAF	In‐home taps or community taps	Improved	Sufficient	Unclear	80% of households had contamination between 10 to 10000 CFUs/100 mL	Unclear
Brown 2008	62% households rely on surface water during dry season and 55% rely on surface water during rainy season	Unimproved	Unlcear	Unclear	Baseline not reported. Control households: Geometric mean 600 E. coli /100 mL	Improved
Clasen 2004b BOL	80% yard taps supplied by untreated surface source, 20% directly from untreated surface sources	80% improved, 20% unimproved	Sufficient	Sufficient	Baseline arithmetic mean 86 TTC/100 mL	Unimproved
Clasen 2004c BOL	Irrigation canals and other surface sources	Unimproved	Sufficient	Sufficient	Baseline arithmetic mean 797 TTC/100 mL	Unimproved
Clasen 2005 COL	67% yard tap from municipality (not treated), 28% river, 12% rainwater	Unimproved	Unclear	Unclear	Baseline not reported. Control households: arithmetic mean 151 TTC/100 mL	Mostly improved
du Preez 2008 ZAF/ZWE	Protected wells	Improved	Sufficient	Unclear	Baseline not reported. Control households: 30% samples post‐intervention met WHO guidelines for water quality	Improved
Lindquist 2014	Municipal supply	Improved	Sufficient	Unclear	Not tested	Unimproved
URL 1995	Household tap (27%), public tap (21%), well (23%)	Improved	Unclear	Unclear	Range 5 to 260; average 106 faecal coliforms/100 mL across three sites.	Improved
Fabiszewski 2012 HND	49% to 69% households use unprotected sources, 24% to 50% use protected sources, 1% to 11% piped water, 0% to 2 % rainwater	Improved and unimproved	Unclear	Unclear	Geometric mean E. coli concentrations of both unprotected and protected sources were > 100 MPN/100 mL	Unimproved
Stauber 2009 DOM	Unclear	Unclear	Unclear	Unclear	Baseline: geometric mean 21 MPN E. coli /100 mL	Improved
Stauber 2012a KHM	77% used improved water source during dry season, 89% during rainy season	Improved	Unclear	Unclear	Baseline: geometric mean 27.5 CFU/100 mL	Unimproved
Stauber 2012b GHA	Surface water 70% in dry season, 95% in rainy season	Unimproved	Unclear	Unclear	Baseline: geometric mean 792 or 832 E. coli /100 mL for control and intervention households, respectively	Unimproved
Tiwari 2009 KEN	Primarily river water; 27% drink protected sources	Unimproved	Unclear	Unclear	Baseline not reported. Control households: 88.9 faecal coliforms/100 mL	Unclear
Boisson 2009 ETH	84% springs, 12% river, 2% handdug well, 4% communal tap	Unimproved	Unclear	Unclear	Baseline arithmetic mean 449 TTC/100 mL	Unimproved
Boisson 2010 DRC	97% surface water, 38% rainwater, 16% springs	Unimproved	Unclear	Unclear	Source drinking water: 75% of household samples > 1000 TTC/100 mL	Unimproved
Peletz 2012 ZMB	46% unprotected dug wells, 22% taps, 16% borehole or protected dug well, 2% surface water	Improved and unimproved	Unclear	Unclear	Unfiltered water: Geometric mean 190 TTC/100 mL	Unimproved
Colford 2002 USA	Household taps supplied by municipal water treatment	Improved	Sufficient	Sufficient	Data from water treatment plant: met US federal and California drinking water standards	Improved
Colford 2005 USA	Household taps supplied by municipal water treatment	Improved	Sufficient	Sufficent	Data from water treatment plant: met US federal drinking water standards	Improved
Colford 2009 USA	Household taps supplied by municipal water treatment	Improved	Sufficient	Sufficient	Data from water treatment plant: met US federal drinking water standards	Improved
Rodrigo 2011 AUS	Untreated rainwater	Improved	Sufficient	Sufficient	Not tested	Improved
Abbreviations: TTC: thermotolerant coliforms, MPN: most probable number, CFU: colony‐forming units ¹'Improved' includes household connection, public standpipe, borehole, protected dug well, protected spring, rainwater collection; 'unimproved' includes unprotected well, unprotected spring, vendor‐provided water, bottled water; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015. ²'Sufficient' means located within 500 m, queuing no more than 15 minutes, no more than three minutes to fill 20 L container, and maintained so available consistently; 'insufficient' means that it does not meet any of above; and 'unclear' means unclear or not reported; definition based minimum standards established by The Sphere Project 2011. ³'Sufficient' means a minimum of 15 L/day/person; 'insufficient' means less than 15 L/day/person; and 'unclear' means unclear or not reported; definition based on minimum standards established by The Sphere Project 2011. ⁴'Improved' means connection to a public sewer or septic system, pour flush latrine, simple pit latrine, or ventilated improved pit latrine; 'unimproved' means service or bucket latrine, public latrines, open latrines; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015.

Table 13. POU filtration: primary drinking water supply and sanitation facilities

Table 14. Summary of findings: POU filtration

POU filtration compared with no intervention for preventing diarrhoea
Patient or population: adults and children Settings: low‐, middle‐ and high‐income countries Intervention: distribution of water filters and instructions on use Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	Water filtration
Diarrhoea episodes Cluster‐RCTs	3 episodes per person per year	All filters	RR 0.48 (0.38 to 0.59)	15,582 (18 trials)	⊕⊕⊕⊝ moderate^1,2,3,4
		1.4 episodes per person per year (1.1 to 1.8)
	3 episodes per person per year	Ceramic filters	RR 0.39 (0.29 to 0.53)	5763 (8 trials)	⊕⊕⊕⊝ moderate^2,4,5,6
		1.1 episodes per person per year (0.8 to 1.5)
		Biosand filters	RR 0.47 (0.39 to 0.57)	5504 (4 trials)	⊕⊕⊕⊝ moderate^4,7,8,9
		1.4 episodes per person per year (1.2 to 1.7)
		LifeStraw®filters	RR 0.69 (0.51 to 0.93)	3259 (3 trials)	⊕⊕⊝⊝ low^2,4,10,11
		2.1 episodes per person per year (1.5 to 2.8)
		Plumbed filters	RR 0.73 (0.52 to 1.03)	1056 (3 trials)	⊕⊕⊕⊝ moderate^2,4,12,13
		2.2 episodes per person per year (1.6 to 3.1)
The basis for the assumed risk is provided in the footnotes. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012). ¹Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants, this outcome is susceptible to bias from lack of blinding. Only five studies blinded participants and outcome assessors to the treatment allocation and only one found an effect of the intervention. ²No serious inconsistency: statistical heterogeneity was very high, however there is consistency in the direction of the effect. ³No serious indirectness: these studies are from a variety of low‐, middle‐, and high‐income countries (South Africa, Ethiopia, Democratic Republic of Congo, Cambodia, Bolivia, Colombia, USA, Australia, Honduras, Zimbabwe, Zambia, Dominican Republic, Ghana, Kenya and Guatemala). ⁴No serious imprecision. ⁵Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants, this outcome is susceptible to bias from lack of blinding. Only one of these studies, Rodrigo 2011 AUS, blinded participants and outcome assessors to the treatment allocation. ⁶No serious indirectness: these studies are from a variety of low‐, middle‐, and high‐income countries (South Africa, Cambodia, Bolivia, Colombia, Zimbabwe, Guatemala and Australia). The interventions consisted of distribution of water filters (which included a safe storage chamber) plus instructions on how to use them. In some cases, the intervention included hygiene education. ⁷Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants, this outcome is susceptible to bias from lack of blinding. None these studies blinded participants and outcome assessors to the treatment allocation. ⁸No serious inconsistency: there was no statistical heterogeneity between studies, I² statistic = 0%. ⁹No serious indirectness: the studies were conducted in a variety of rural and urban settings in a variety of low‐ and middle‐income countries (Honduras, Dominican Republic, Cambodia, Ghana and Kenya). The interventions consisted of distribution of water filters plus instructions on how to use them. In some cases, the intervention included hygiene education and a separate storage vessel. ¹⁰Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. Only one of these studies, Boisson 2010 DRC, blinded participants and outcome assessors to the treatment allocation and found no evidence of effect of the filter. ¹¹Downgraded by 1 for some indirectness, the studies were only performed in three sub‐Saharan African countries (Ethiopia, Democratic Republic of Congo, and Zambia). ¹²No serious risk of bias: the three studies blinded participants and outcome assessors to the treatment allocation. ¹³Downgraded by 1 for some indirectness, the three studies were only performed in the USA in water conditions that presumed to meet US EPA standards.

Table 14. Summary of findings: POU filtration

Table 15. POU solar disinfection (SODIS): description of the interventions

Study ID	Study design	Setting	Intervention areas			Control areas
Study ID	Study design	Setting	Water quality intervention	Health promotion activities	Compliance	Water source	Health promotion activities
Conroy 1996 KEN	Quasi‐RCT	Rural	Children were given two 1.5 L plastic bottles and told to keep the bottles on the roof of the hut throughout the day in full sunlight	None	100%‐ random checks by project workers uncovered no evidence of non‐compliance	Children were given two 1.5 L plastic bottles and told to keep the bottles indoors	None
Conroy 1999 KEN	Quasi‐RCT	Rural	Mothers were given plastic bottles and told to keep the bottles on the roof of the hut throughout the day in full sunlight	None	Not reported	Mothers were given plastic bottles and told to keep the bottles indoors	None
du Preez 2010 ZAF	Cluster‐RCT	Peri urban	Received two 2 L polyethylene terephtalate (PET) bottles for each child. Carers were instructed to fill one bottle and place it in full, unobscured sunlight for a minimum of 6 h every day.	None	25% compliance measured by participants filling out diarrhoeal diaries at least 75% of the time	No SODIS bottles and maintain their usual practices	None
du Preez 2011 KEN	Cluster‐RCT	Peri urban and rural	Received two 2 L PET bottles for each child. Carers were instructed to fill one bottle and place it in full, unobscured sunlight for a minimum of 6 h every day.	None	Not specified.	No SODIS bottles and maintain their usual practices	None
Mäusezhal 2009 BOL	Cluster‐RCT	Rural	Households were supplied regularly with clean, PET bottles. They were instructed to expose the waterfilled bottles for at least 6 h to the sun.	Households were taught about the importance and benefits of drinking only treated water, the germ–disease concept, and promoted hygiene measures such as safe drinking water storage and hand washing.	32% compliance measured by observation	Drinking water from spring (48.1%), tap (51.9%), river (22.1%), rain (14.9%) and dug well (14.9%)	None
McGuigan 2011 KHM	Cluster‐RCT	Rural	Households were provided with two transparent 2 L plastic bottles for each child and a sheet of corrugated iron on which to place the bottles to expose them to sunlight. Carers were instructed to fill one bottle and place it in full, unobscured sunlight for a minimum of 6 h every day.	The parents or carers were given verbal and written information on the disease concept and a simple explanation of the solar disinfection process and its effect on the microbial quality of their drinking water and subsequently the health of their children	90% (5% of children having < 10 months of follow‐up and 2.3% having < 6 months)	Almost all of the households (97%) obtained water from unprotected boreholes. An important subgroup of these, 25%, drew water from shallow tube wells fitted with hand pumps. The remainder used unprotected wells or surface ponds	None

Table 15. POU solar disinfection (SODIS): description of the interventions

Table 16. POU solar disinfection (SODIS): primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient H₂O quality	Sanitation⁴
Conroy 1996 KEN	Open water holes, tank fed by untreated piped water supply.	Unimproved	Unclear	Unclear	Source water: 10³ CFU/100 mL	Unclear
Conroy 1999 KEN	Open water holes, tank fed by untreated piped water supply.	Unimproved	Unclear	Unclear	Source water: 10³ CFU/100 mL	Unclear
du Preez 2010 ZAF	39% standpipes, 28% protected borehole, 10% unprotected boreholes, protected springs	Mostly improved	Sufficient	Sufficient	Baseline not reported. Intervention households: 62% of samples met WHO guidelines for water quality; no significant difference from control households	Unclear
du Preez 2011 KEN	Spring, protected and unprotected dug wells protected, canals, other	Mostly unimproved	Unclear	Unclear	50% of samples from stored water had 10 CFU/100 mL or less; no significant difference for intervention and controls	Unclear
Mäusezhal 2009 BOL	48% spring, 52% tap, 22% river, 15% rain, 15% dug well	Improved and unimproved	Sufficient	Sufficient	Not tested	Unimproved
McGuigan 2011 KHM	97% households use unprotected sources: unprotected wells, surface ponds	Unimproved	Unclear	Unclear	Baseline not reported. Control households: geometric mean 48 CFU/100 mL	Unimproved
¹'Improved' includes household connection, public standpipe, borehole, protected dug well, protected spring, rainwater collection; 'unimproved' includes unprotected well, unprotected spring, vendor‐provided water, bottled water; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015. ²'Sufficient' means located within 500 m, queuing no more than 15 minutes, no more than three minutes to fill 20 L container, and maintained so available consistently; 'insufficient' means that it does not meet any of above; and 'unclear' means unclear or not reported; definition based minimum standards established by The Sphere Project 2011. ³'Sufficient' means a minimum of 15 L/day/person; 'insufficient' means less than 15 L/day/person; and 'unclear' means unclear or not reported; definition based on minimum standards established by The Sphere Project 2011. ⁴'Improved' means connection to a public sewer or septic system, pour flush latrine, simple pit latrine, or ventilated improved pit latrine; 'unimproved' means service or bucket latrine, public latrines, open latrines; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015.

Table 16. POU solar disinfection (SODIS): primary drinking water supply and sanitation facilities

Table 17. Summary of findings: POU solar disinfection (SODIS)

POU solar disinfection (SODIS) of water compared with no intervention for preventing diarrhoea
Patient or population: adults and children Settings: low‐ and middle‐income countries Intervention: distribution of plastic bottles with instructions on using them to treat water using the SODIS method. Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	SODIS
Diarrhoea episodes Cluster‐RCTs	3 episodes per person per year	1.9 episodes per person per year (1.3 to 2.8)	RR 0.62 (0.42 to 0.94)	3460 (4 trials)	⊕⊕⊕⊝ moderate^1,2,3,4
Diarrhoea episodes Quasi‐RCTs	3 episodes per person per year	2.5 episodes per person per year (2.1 to 2.9)	RR 0.82 (0.69 to 0.97)	555 (2 studies)	⊕⊕⊝⊝ low^1,5,6,7
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012). ¹Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. None of these studies blinded participants and outcome assessors to the treatment allocation. ²No serious inconsistency: statistical heterogeneity was very high (I² statistic = 89%), however there is consistency in the direction of the effect. This heterogeneity may relate to differences in compliance across the studies, however compliance was not measured in the same way across studies. ³No serious indirectness: the studies were conducted in peri‐urban South Africa (one study), peri‐urban and rural Kenya (one study), rural Bolivia (one study) and rural Cambodia (one study). ⁴No serious imprecision: the average effect suggests that the intervention may reduce diarrhoea episodes by about one third. ⁵No serious inconsistency: statistical heterogeneity was low (I² statistic = 0%). ⁶Downgraded by 1 for serious indirectness: there are only two studies and both were conducted in the same province in Kenya (one study included children five to 16 years old and the other included children younger than six years old). ⁷No serious imprecision.

Table 17. Summary of findings: POU solar disinfection (SODIS)

Table 18. POU UV: description of the interventions

Study ID	Study design	Setting	Intervention areas			Control areas
Study ID	Study design	Setting	Water quality intervention	Health promotion activities	Compliance	Water source	Health promotion activities
Gruber 2013 MEX	Cluster‐RCT	Rural	Promotion of the UV Tube disinfection technology and safe storage	Unclear	51% compliance measured by access to treatment device	Unclear	None

Table 18. POU UV: description of the interventions

Table 19. POU UV: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient H₂O quality	Sanitation⁴
Gruber 2013 MEX	Unclear	Unclear	Unclear	Unclear	Baseline: 60% of samples with detectable E. coli	Improved
¹'Improved' includes household connection, public standpipe, borehole, protected dug well, protected spring, rainwater collection; 'unimproved' includes unprotected well, unprotected spring, vendor‐provided water, bottled water; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015. ²'Sufficient' means located within 500 m, queuing no more than 15 minutes, no more than three minutes to fill 20 L container, and maintained so available consistently; 'insufficient' means that it does not meet any of above; and 'unclear' means unclear or not reported; definition based minimum standards established by The Sphere Project 2011. ³'Sufficient' means a minimum of 15 L/day/person; 'insufficient' means less than 15 L/day/person; and 'unclear' means unclear or not reported; definition based on minimum standards established by The Sphere Project 2011. ⁴'Improved' means connection to a public sewer or septic system, pour flush latrine, simple pit latrine, or ventilated improved pit latrine; 'unimproved' means service or bucket latrine, public latrines, open latrines; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015.

Table 19. POU UV: primary drinking water supply and sanitation facilities

Table 20. POU Improved storage: description of the interventions

Study ID	Study design	Setting	Intervention areas			Control areas
Study ID	Study design	Setting	Water quality intervention	Health promotion activities	Compliance	Water source	Health promotion activities
Günther 2013 BEN	Cluster‐RCT	Rural	Provided households with a new 30 L household water storage with a tap at the bottom, a new plastic container to transport water from the water source to the household and a sign attached to the transport and storage containers which emphasized the importance of avoiding hand‐contact with the water and to only use water from an improved water source.	None	After 7 months, 88% of households were still using the improved storage containers	68% only consume improved water source	None
Roberts 2001 MWI	Cluster‐RCT	Refugee camp	All of the participating household's water collection vessels were exchanged for improved buckets (20 L with a narrow opening to limit hand entry). Households were offered 1 improved bucket in exchange for 1 vessel, 2 for 2, and 3 improved buckets for any number of containers > 2. Households were asked never to put their hands in the improved buckets and were shown how to rinse the bucket without hand entry.	None	Intervention householders received buckets; actual use was not reported	Provided with 20 L standard ration bucket	None

Table 20. POU Improved storage: description of the interventions

Table 21. POU Improved storage: primary drinking water supply and sanitation facilities

Trial	Description	Source¹	Access to source²	Quantity available³	Ambient H₂O quality	Sanitation⁴
Günther 2013 BEN	Public tap or pump	Improved	Sufficient	Unclear	12% source water contaminated (≥ 1000 CFU per 100 mL)	Unclear
Roberts 2001 MWI	Traditional pots or standard ration buckets filled at refugee camp water point	Improved	Unclear	Unclear	Source water: 71% of samples had ≤ 1 faecal coliform/100 mL	Unclear
¹'Improved' includes household connection, public standpipe, borehole, protected dug well, protected spring, rainwater collection; 'unimproved' includes unprotected well, unprotected spring, vendor‐provided water, bottled water; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015. ²'Sufficient' means located within 500 m, queuing no more than 15 minutes, no more than three minutes to fill 20 L container, and maintained so available consistently; 'insufficient' means that it does not meet any of above; and 'unclear' means unclear or not reported; definition based minimum standards established by The Sphere Project 2011. ³'Sufficient' means a minimum of 15 L/day/person; 'insufficient' means less than 15 L/day/person; and 'unclear' means unclear or not reported; definition based on minimum standards established by The Sphere Project 2011. ⁴'Improved' means connection to a public sewer or septic system, pour flush latrine, simple pit latrine, or ventilated improved pit latrine; 'unimproved' means service or bucket latrine, public latrines, open latrines; and 'unclear' means unclear or not reported; definition based on WHO/UNICEF 2015.

Table 21. POU Improved storage: primary drinking water supply and sanitation facilities

Table 22. Summary of findings: POU improved water storage

Improved water storage compared with no intervention for preventing diarrhoea
Patient or population: adults and children in sub‐Saharan Africa Settings: areas with improved water sources Intervention: distribution of improved water containers Comparison: no intervention
Outcomes	*Illustrative comparative risks (95% CI)**		Relative effect (95% CI)	Number of participants (studies)	Quality of the evidence (GRADE)
	Assumed risk	Corresponding risk
	No intervention	Water storage
Diarrhoea episodes Cluster‐RCTs	3 episodes per person per year	2.7 episodes per person per year (2.2 to 3.3 )	RR 0.91 (0.74 to 1.11)	1871 (2 trials)	⊕⊕⊝⊝ low^1,2,3,4
The basis for the assumed risk is the median control group risk across studies. The corresponding risk (and its 95% CI) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: confidence interval; RR: risk ratio.
GRADE Working Group grades of evidence High quality: further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: we are very uncertain about the estimate.
The assumed risk is based on 2.9 episodes/child year in 2010 (Fischer Walker 2012). ¹Downgraded by 1 for serious risk of bias: as diarrhoea episodes were reported by participants this outcome is susceptible to bias from lack of blinding. None of these studies blinded participants and outcome assessors to the treatment allocation. ²No serious inconsistency. ³Downgraded by 1 for indirectness: only 2 studies, from rural Benin and a refugee camp in Malawi, have been conducted to assess improved water storage. ⁴No serious imprecision.

Table 22. Summary of findings: POU improved water storage

Table 23. Estimates of household‐level interventions after adjustment for non‐blinding

POU intervention	Number of comparisons	Not adjusted for non‐blinding		Adjusted for non‐blinding
POU intervention	Number of comparisons	RR	95% CI	RR	95% CI
All	55	0.56	(0.46 to 0.68)	0.70	(0.64 to 0.77)
Chlorination	19	0.72	(0.61 to 0.84)	0.80	(0.69 to 0.92)
Filtration	23	0.48	(0.38 to 0.59)	0.62	(0.55 to 0.70)
Flocculation and disinfection	7	0.48	(0.20 to 1.16)	0.65	(0.40 to 1.09)
SODIS	6	0.68	(0.53 to 0.89)	0.80	(0.60 to 1.01)
Abbreviation: SODIS: solar disinfection; CI: confidence interval.

Table 23. Estimates of household‐level interventions after adjustment for non‐blinding

Table 24. Potential reasons for finding of no‐effect in trials with adequate blinding

Study	Risk from ambient water quality	Compliance	Other issues
Colford 2002 USA	Very low (USA)	High (Sham filter)	None
Colford 2005 USA	Very low (USA)	High (Sham filter)	None
Colford 2009 USA	Very low (USA)	High (Sham filter)	None
Rodrigo 2011 AUS	Very low (Australia)	Not reported	None
Jain 2010 GHA	Low (11 CFU/100 mL)	High (RFC)	Control group received jerry can; 13 week follow‐up
Kirchhoff 1985 BRA	Very high (mean 16000 FC/dL)	Not reported	Only 112 persons from 16 households; 18 week trial
Austin 1993	High (1871 FC/100 mL)	Low ("50% to 60%")	No test of blinding; not peer reviewed
Boisson 2010 DRC	High (75% of samples > 1000 TTC/100 mL)	High, but 73% of adults and 95% of children drank from untreated sources	"Placebo" removed > 90% of TTC in control arm
Boisson 2013 IND	Moderate (mean 122 TTC/100 mL)	Low and inconsistent (32% of samples positive for RFC)	None
Abbreviations: TTC: thermotolerant coliforms, CFU: colony‐forming units, FC: faecal coliforms, RFC: residual free chlorine.

Table 24. Potential reasons for finding of no‐effect in trials with adequate blinding

Comparison 1. Water quality intervention versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Diarrhoea: all ages Show forest plot	64	81215	Risk Ratio (Random, 95% CI)	0.59 [0.51, 0.69]

1.1 Source water improvement	6	9161	Risk Ratio (Random, 95% CI)	0.76 [0.48, 1.19]
1.2 POU treatment	58	72054	Risk Ratio (Random, 95% CI)	0.58 [0.48, 0.69]
2 Diarrhoea: children < 5 years Show forest plot	49		Risk Ratio (Random, 95% CI)	0.61 [0.49, 0.75]

2.1 Source water improvement	4		Risk Ratio (Random, 95% CI)	0.96 [0.82, 1.12]
2.2 POU treatment	45		Risk Ratio (Random, 95% CI)	0.58 [0.46, 0.73]

Comparison 1. Water quality intervention versus control

Comparison 2. Source: water supply improvement versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Diarrhoea: CBA studies subgrouped by age Show forest plot	6		Risk Ratio (Random, 95% CI)	Subtotals only

1.1 Cluster‐RCTs	1	3266	Risk Ratio (Random, 95% CI)	1.24 [0.98, 1.57]
1.2 CBA studies	5	5895	Risk Ratio (Random, 95% CI)	0.68 [0.42, 1.09]
2 Diarrhoea: CBA studies subgrouped by age Show forest plot	5		Risk Ratio (Random, 95% CI)	Subtotals only

2.1 All ages	5	5895	Risk Ratio (Random, 95% CI)	0.68 [0.42, 1.09]
2.2 < 5 years	3	999	Risk Ratio (Random, 95% CI)	0.92 [0.79, 1.07]

Comparison 2. Source: water supply improvement versus control

Comparison 3. POU: water chlorination versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Diarrhoea: subgrouped by study design Show forest plot	19	34694	Risk Ratio (Random, 95% CI)	0.72 [0.61, 0.84]

1.1 Cluster‐RCTs	16	30746	Risk Ratio (Random, 95% CI)	0.77 [0.65, 0.91]
1.2 CBA studies	3	3948	Risk Ratio (Random, 95% CI)	0.51 [0.34, 0.75]
2 Diarrhoea: cluster‐RCTs: subgrouped by age Show forest plot	16		Risk Ratio (Random, 95% CI)	Subtotals only

2.1 All ages	16		Risk Ratio (Random, 95% CI)	0.77 [0.65, 0.91]
2.2 < 5 years	15		Risk Ratio (Random, 95% CI)	0.77 [0.64, 0.92]
3 Diarrhoea: cluster‐RCTs; subgrouped by adherence Show forest plot	16	30746	Risk Ratio (Random, 95% CI)	0.77 [0.65, 0.91]

3.1 Residual chlorine in 86 to 100% of samples	1	276	Risk Ratio (Random, 95% CI)	0.78 [0.73, 0.83]
3.2 Residual chlorine in 51 to 85% of samples	6	9994	Risk Ratio (Random, 95% CI)	0.60 [0.40, 0.91]
3.3 Residual chlorine in ≤ 50% of samples	4	12613	Risk Ratio (Random, 95% CI)	0.90 [0.76, 1.06]
3.4 Residual chlorine not reported	5	7863	Risk Ratio (Random, 95% CI)	0.85 [0.65, 1.12]
4 Diarrhoea: cluster‐RCTs by risk of bias by blinding of participants Show forest plot	16		Risk Ratio (Random, 95% CI)	Subtotals only

4.1 Low risk	5	15867	Risk Ratio (Random, 95% CI)	1.07 [0.97, 1.17]
4.2 High risk	11	14879	Risk Ratio (Random, 95% CI)	0.68 [0.56, 0.83]
5 Diarrhoea: cluster‐RCTs; subgrouped by additional water storage intervention Show forest plot	16		Risk Ratio (Random, 95% CI)	0.77 [0.65, 0.91]

5.1 Chlorination kit alone	8		Risk Ratio (Random, 95% CI)	0.75 [0.54, 1.05]
5.2 Chlorination kit plus water storage	8		Risk Ratio (Random, 95% CI)	0.80 [0.66, 0.97]
6 Diarrhoea: cluster‐RCTs: subgrouped by sufficiency of water quantity Show forest plot	16	30746	Risk Ratio (Random, 95% CI)	0.77 [0.65, 0.91]

6.1 Sufficient	3	5352	Risk Ratio (Random, 95% CI)	0.90 [0.69, 1.17]
6.2 Insufficient	2	3499	Risk Ratio (Random, 95% CI)	0.91 [0.66, 1.26]
6.3 Unclear	11	21895	Risk Ratio (Random, 95% CI)	0.67 [0.50, 0.88]
7 Diarrhoea: cluster‐RCTs: subgrouped by water source Show forest plot	16		Risk Ratio (Random, 95% CI)	Subtotals only

7.1 Improved water source	3	5880	Risk Ratio (Random, 95% CI)	0.82 [0.59, 1.14]
7.2 Unimproved water source	13	24866	Risk Ratio (Random, 95% CI)	0.75 [0.59, 0.93]
8 Diarrhoea: cluster‐RCTs: subgrouped by sanitation level Show forest plot	16		Risk Ratio (Random, 95% CI)	Subtotals only

8.1 Improved sanitation	3	4876	Risk Ratio (Random, 95% CI)	0.64 [0.44, 0.92]
8.2 Unimproved sanitation	6	17352	Risk Ratio (Random, 95% CI)	0.81 [0.63, 1.05]
8.3 Unclear	7	8518	Risk Ratio (Random, 95% CI)	0.75 [0.54, 1.05]
9 Diarrhoea: cluster‐RCTs; subgrouped by length of follow‐up Show forest plot	16		Risk Ratio (Random, 95% CI)	0.77 [0.65, 0.91]

9.1 ≤ 3 months	2		Risk Ratio (Random, 95% CI)	0.42 [0.06, 3.03]
9.2 > 3 to 6 months	7		Risk Ratio (Random, 95% CI)	0.71 [0.51, 0.99]
9.3 > 6 to 12 months	5		Risk Ratio (Random, 95% CI)	0.82 [0.71, 0.96]
9.4 > 12 months	2		Risk Ratio (Random, 95% CI)	0.99 [0.66, 1.48]

Comparison 3. POU: water chlorination versus control

Comparison 4. POU: flocculation and disinfection versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Diarrhoea: cluster‐RCTs Show forest plot	7		Risk Ratio (Random, 95% CI)	0.48 [0.20, 1.16]

2 Diarrhoea: cluster‐RCTs: subgrouped by age; excluding Doocy 2006 LBR Show forest plot	6		Risk Ratio (Random, 95% CI)	Subtotals only

2.1 All ages	6	11788	Risk Ratio (Random, 95% CI)	0.69 [0.58, 0.82]
2.2 < 5	6	0	Risk Ratio (Random, 95% CI)	0.71 [0.61, 0.84]
3 Diarrhoea: cluster‐RCTs: subgrouped by adherence Show forest plot	7		Risk Ratio (Random, 95% CI)	Subtotals only

3.2 Residual chlorine 51 to 85%	1	2191	Risk Ratio (Random, 95% CI)	0.12 [0.11, 0.13]
3.3 Residual chlorine < 50%	4	6914	Risk Ratio (Random, 95% CI)	0.76 [0.67, 0.85]
3.4 Residual chlorine not measured	2	4874	Risk Ratio (Random, 95% CI)	0.41 [0.26, 0.64]
4 Diarrhoea: cluster‐RCTs: subgrouped by additional storage container Show forest plot	7		Risk Ratio (Random, 95% CI)	0.48 [0.20, 1.16]

4.1 No storage container	2		Risk Ratio (Random, 95% CI)	0.81 [0.69, 0.95]
4.2 Storage container	5		Risk Ratio (Random, 95% CI)	0.39 [0.14, 1.08]
5 Diarrhoea: cluster‐RCTs: subgrouped by sufficiency of water quantity Show forest plot	7		Risk Ratio (Random, 95% CI)	Subtotals only

5.1 Sufficient	1	3401	Risk Ratio (Random, 95% CI)	0.62 [0.47, 0.82]
5.2 Insufficient	2	5454	Risk Ratio (Random, 95% CI)	0.31 [0.05, 2.09]
5.3 Unclear	4	5124	Risk Ratio (Random, 95% CI)	0.64 [0.49, 0.85]
6 Diarrhoea: cluster‐RCTs: subgrouped by water source Show forest plot	7		Risk Ratio (Random, 95% CI)	Subtotals only

6.1 Improved water source	2	4874	Risk Ratio (Random, 95% CI)	0.41 [0.26, 0.64]
6.2 Unimproved water source	4	5704	Risk Ratio (Random, 95% CI)	0.49 [0.14, 1.68]
6.3 Unclear	1	3401	Risk Ratio (Random, 95% CI)	0.62 [0.47, 0.82]
7 Diarrhoea: cluster‐RCTs: subgrouped by sanitation level Show forest plot	7		Risk Ratio (Random, 95% CI)	Subtotals only

7.1 Improved sanitation	2	4874	Risk Ratio (Random, 95% CI)	0.41 [0.26, 0.64]
7.2 Unimproved sanitation	2	5592	Risk Ratio (Random, 95% CI)	0.27 [0.05, 1.36]
7.3 Unclear	3	3513	Risk Ratio (Random, 95% CI)	0.79 [0.69, 0.90]
8 Diarrhoea: cluster‐RCTs: subgrouped by length of follow‐up Show forest plot	7	13979	Risk Ratio (Random, 95% CI)	0.48 [0.20, 1.16]

8.1 ≤ 3 months	2	5592	Risk Ratio (Random, 95% CI)	0.27 [0.05, 1.36]
8.2 > 3 to 6 months	1	3263	Risk Ratio (Random, 95% CI)	0.83 [0.67, 1.03]
8.3 > 6 to 12 months	4	5124	Risk Ratio (Random, 95% CI)	0.64 [0.49, 0.85]

Comparison 4. POU: flocculation and disinfection versus control

Comparison 5. POU: filtration versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Diarrhoea: cluster‐RCTs: subgrouped by age Show forest plot	23		Risk Ratio (Random, 95% CI)	Subtotals only

1.1 All ages	23		Risk Ratio (Random, 95% CI)	0.48 [0.38, 0.59]
1.2 < 5 years	19		Risk Ratio (Random, 95% CI)	0.49 [0.38, 0.62]
2 Diarrhoea: cluster‐RCTs: subgrouped by type of filtration Show forest plot	23		Risk Ratio (Random, 95% CI)	Subtotals only

2.1 Ceramic filter	12	5763	Risk Ratio (Random, 95% CI)	0.39 [0.29, 0.53]
2.2 Sand filtration	5	5504	Risk Ratio (Random, 95% CI)	0.47 [0.39, 0.57]
2.3 LifeStraw®	3	3259	Risk Ratio (Random, 95% CI)	0.69 [0.51, 0.93]
2.4 Plumbed	3	1056	Risk Ratio (Random, 95% CI)	0.73 [0.52, 1.03]
3 Diarrhoea: cluster‐RCTs: subgrouped by blinding of participants Show forest plot	23		Risk Ratio (Random, 95% CI)	Subtotals only

3.1 Low risk	5		Risk Ratio (Random, 95% CI)	0.80 [0.68, 0.94]
3.2 High risk	18		Risk Ratio (Random, 95% CI)	0.41 [0.33, 0.52]
4 Diarrhoea: ceramic filter studies subgrouped by water source Show forest plot	12	5763	Risk Ratio (Random, 95% CI)	0.39 [0.29, 0.53]

4.1 Improved water source	8	3607	Risk Ratio (Random, 95% CI)	0.33 [0.23, 0.46]
4.2 Unimproved water source	4	2156	Risk Ratio (Random, 95% CI)	0.54 [0.48, 0.61]
5 Diarrhoea: ceramic filter studies subgrouped by sanitation level Show forest plot	12	5763	Risk Ratio (Random, 95% CI)	0.39 [0.29, 0.53]

5.1 Improved sanitation	7	4198	Risk Ratio (Random, 95% CI)	0.49 [0.38, 0.64]
5.2 Unimproved sanitation	4	1491	Risk Ratio (Random, 95% CI)	0.35 [0.22, 0.56]
5.3 Unclear	1	74	Risk Ratio (Random, 95% CI)	0.21 [0.18, 0.25]
6 Diarrhoea: sand filter studies: subgrouped by water source Show forest plot	5		Risk Ratio (Random, 95% CI)	0.47 [0.39, 0.57]

6.1 Improved water source	2		Risk Ratio (Random, 95% CI)	0.50 [0.33, 0.75]
6.2 Unimproved water source	2		Risk Ratio (Random, 95% CI)	0.44 [0.25, 0.76]
6.3 Unclear	1		Risk Ratio (Random, 95% CI)	0.47 [0.37, 0.60]
7 Diarrhoea: sand filter studies: subgrouped by sanitation level Show forest plot	5		Risk Ratio (Random, 95% CI)	0.47 [0.39, 0.57]

7.1 Improved sanitation	1		Risk Ratio (Random, 95% CI)	0.47 [0.37, 0.60]
7.2 Unimproved sanitation	3		Risk Ratio (Random, 95% CI)	0.48 [0.34, 0.68]
7.3 Unclear	1		Risk Ratio (Random, 95% CI)	0.46 [0.22, 0.96]
8 Diarrhoea: cluster‐RCTs: subgrouped by adherence Show forest plot	23		Risk Ratio (Random, 95% CI)	Subtotals only

8.1 86 to 100%	12	7300	Risk Ratio (Random, 95% CI)	0.43 [0.34, 0.55]
8.2 51 to 85%	4	2346	Risk Ratio (Random, 95% CI)	0.56 [0.33, 0.95]
8.3 ≤ 50%	1	1516	Risk Ratio (Random, 95% CI)	0.75 [0.60, 0.94]
8.4 Not reported	6	4420	Risk Ratio (Random, 95% CI)	0.46 [0.28, 0.75]
9 Diarrhoea: cluster‐RCTs: subgrouped by additional water storage intervention Show forest plot	19		Risk Ratio (Random, 95% CI)	Subtotals only

9.1 Filtration alone	8		Risk Ratio (Random, 95% CI)	0.60 [0.48, 0.76]
9.2 Filtration plus storage	11		Risk Ratio (Random, 95% CI)	0.38 [0.29, 0.49]
10 Diarrhoea: cluster‐RCTs; subgrouped by length of follow‐up Show forest plot	23		Risk Ratio (Random, 95% CI)	0.48 [0.38, 0.59]

10.1 ≤ 3 months	3		Risk Ratio (Random, 95% CI)	0.26 [0.20, 0.33]
10.2 > 3 to 6 months	11		Risk Ratio (Random, 95% CI)	0.52 [0.44, 0.60]
10.3 > 6 to 12 months	8		Risk Ratio (Random, 95% CI)	0.51 [0.30, 0.87]
10.4 > 12 months	1		Risk Ratio (Random, 95% CI)	0.87 [0.74, 1.02]

Comparison 5. POU: filtration versus control

Comparison 6. POU: solar disinfection versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Diarrhoea: subgrouped by study design Show forest plot	6		Risk Ratio (Random, 95% CI)	Subtotals only

1.1 Cluster‐RCTs	4	3460	Risk Ratio (Random, 95% CI)	0.62 [0.42, 0.94]
1.2 Quasi‐RCTs	2	555	Risk Ratio (Random, 95% CI)	0.82 [0.69, 0.97]
2 Diarrhoea: cluster‐RCTs; subgrouped by age Show forest plot	4		Risk Ratio (Random, 95% CI)	Subtotals only

2.1 All ages	4		Risk Ratio (Random, 95% CI)	0.62 [0.42, 0.94]
2.2 < 5	3		Risk Ratio (Random, 95% CI)	0.55 [0.34, 0.91]
3 Diarrhoea: cluster‐RCTs; subgrouped by adherence Show forest plot	4		Risk Ratio (Random, 95% CI)	Subtotals only

3.1 86 to 100%	1	928	Risk Ratio (Random, 95% CI)	0.37 [0.29, 0.47]
3.2 51 to 85%	0	0	Risk Ratio (Random, 95% CI)	0.0 [0.0, 0.0]
3.3 ≤ 50%	2	1443	Risk Ratio (Random, 95% CI)	0.80 [0.57, 1.11]
3.4 Not reported	1	1089	Risk Ratio (Random, 95% CI)	0.73 [0.63, 0.85]
4 Diarrhoea: cluster‐RCTs; subgrouped by sufficiency of water supply level Show forest plot	4	3460	Risk Ratio (Random, 95% CI)	0.62 [0.42, 0.94]

4.1 Sufficient	2	1443	Risk Ratio (Random, 95% CI)	0.80 [0.57, 1.11]
4.3 Unclear	2	2017	Risk Ratio (Random, 95% CI)	0.52 [0.27, 1.02]
5 Diarrhoea: cluster‐RCTs; subgrouped by water source Show forest plot	4		Risk Ratio (Random, 95% CI)	Subtotals only

5.1 Improved water source	1	718	Risk Ratio (Random, 95% CI)	0.64 [0.39, 1.05]
5.2 Unimproved water source	3	2742	Risk Ratio (Random, 95% CI)	0.62 [0.38, 1.02]
6 Diarrhoea: cluster‐RCTs; subgrouped by sanitation level Show forest plot	4		Risk Ratio (Random, 95% CI)	Subtotals only

6.1 Improved sanitation	0	0	Risk Ratio (Random, 95% CI)	0.0 [0.0, 0.0]
6.2 Unimproved sanitation	2	1653	Risk Ratio (Random, 95% CI)	0.57 [0.24, 1.39]
6.3 Unclear	2	1807	Risk Ratio (Random, 95% CI)	0.72 [0.63, 0.83]
7 Diarrhoea: cluster‐RCTs; subgrouped by length of follow‐up Show forest plot	4	3460	Risk Ratio (Random, 95% CI)	0.62 [0.42, 0.94]

7.2 > 6 to 12 months	3	2371	Risk Ratio (Random, 95% CI)	0.59 [0.32, 1.09]
7.3 > 12 months	1	1089	Risk Ratio (Random, 95% CI)	0.73 [0.63, 0.85]

Comparison 6. POU: solar disinfection versus control

Comparison 7. POU: UV disinfection versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Diarrhoea: cluster‐RCT Show forest plot	1		Risk Ratio (Random, 95% CI)	Subtotals only

Comparison 7. POU: UV disinfection versus control

Comparison 8. POU: improved storage versus control

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Diarrhoea: cluster‐RCTs: subgrouped by age Show forest plot	2		Risk Ratio (Random, 95% CI)	Subtotals only

1.1 All ages	2		Risk Ratio (Random, 95% CI)	0.91 [0.74, 1.11]
1.2 < 5	1		Risk Ratio (Random, 95% CI)	0.69 [0.47, 1.01]

Comparison 8. POU: improved storage versus control