Multiple and complex links between babyWASH and stunting: an evidence synthesis

Studies have shown linkages between water, sanitation and hygiene (WASH) and stunting in children under 2 years in sub-Saharan Africa. WASH interventions have been shown to reduce stunting rates; however, the biological mechanisms and socio-economic influences responsible for this trend remain poorly understood. This paper reviews the literature regarding these links, and the efficacy of both general WASH interventions and those targeted at children in their first 1,000 days, known as babyWASH, for stunting reduction. Fifty-nine papers published between 2008 and 2019 were reviewed, retrieved from Science Direct, Scopus and Web of Science databases, comprising field trials and data analysis, and literature and systematic reviews. Key findings showed that stunting is directly attributed to diarrhoea, environmental enteric dysfunction and undernutrition although a more comprehensive understanding of these biological mechanisms is necessary. Interventions to interrupt the faecal transmission cycle proved to effectively reduce stunting rates, particularly improved sanitation facilities to reduce open defaecation, increased proximity to water and widespread behavioural change. Methodologies should move away from randomised controlled trials towards selected contexts, mixed data collection methods and inclusion of broader social, cultural and environmental conditions. Improved cross-sectoral collaboration is encouraged, particularly to ensure the complexity of social and contextual factors is fully considered.

and adults of all ages for its potential application to children in their first 1,000 days. It initially questions the direct and indirect causes of stunting and how WASH can affect these, then goes on to discuss successful and unsuccessful interventions through a comprehensive literature review.
Finally, it provides an analysis of these findings and identifies gaps and directions for further study. In some cases, the criteria were extended to include research on children under 5 years where literature was of value. Articles with a broader geographical scope but the inclusion of sub-Saharan countries were also considered if deemed to add value to the overall study. Inclusion limits required papers to be published in English.
For the purpose of this study, WASH is defined as (a) water infrastructure, availability, accessibility and quality, (b) sanitation infrastructure, availability, accessibility and quality, and (c) hygiene practices including infant feeding, environmental cleanliness and associated educational and cultural behaviours.

Search methodology and data sources
This study comprises an extensive literature review of relevant papers published between 2008 and 2019. ScienceDirect, Scopus and Web of Science databases were used to extract relevant articles, using the search terms 'baby WASH stunting' and 'water sanitation hygiene stunting'.
ScienceDirect returned 459 and 852 articles for these search terms, respectively; Scopus 7 and 113, respectively; and Web of Science 8 and 95, respectively.
All 1,534 articles were reviewed by title and abstract against the criteria; 1,401 excluded as a result; the remaining 133 examined for relevant articles in reference lists; 35 articles subsequently added; 168 articles read for full review; 109 excluded; and 59 articles remained for inclusion in this study. Figure 1 visually illustrates the selection process.
The available applicable research consists of a combination of primary trials, secondary data analysis and literature or systematic reviews. A database format was used to process categorised and numerical data, to allow simple filtering of information for comparison. NVivo literature review software was used to process qualitative content.

Categorisation of literature
The literature included 59 papers from 2008, when the literature in this field began to increase, to 2019. Following selection of the papers, the literature was classified by primary trials (22), secondary data analysis (9) and literature and systematic reviews (28). Publication dates were recorded to explore trends of research in the field, which show that publications have rapidly increased since 2012 with a peak of 10 published in 2017, and 9 published in both 2018 and 2019 ( Figure 2). Papers were also categorised by location and age demographics of study groups in accordance with the eligibility criteria. Analysis of data collection methods of the 22 trials revealed that 9 studies measured quantitative data alone, 3 qualitative data alone, and 10 measured a combination of both. Trials using quantitative data recorded primarily anthropometric measurements alongside other biological biomarkers such as haemoglobin levels in some instances.
Qualitative methods were typically used to collect data on household income, hygiene behaviours, health, food-security, diarrhoea incidence and household cleanliness.
All 59 papers included the objective age criteria of the first 1,000 days; however, the 22 trials frequently also studied other age groups which were included if deemed to add value to the study. Only eight trials recorded data for the 0-24 months age group as a discrete group, indicating the urgent need for more studies for this cohort alone. Figure 4 shows the varied data collection groups across all papers.

Written and analysis methodology
The literature categorisation graphically displays the trends found from the data extraction process, categorised by dates of publication, location of studies and age demographic. The following paper structure is split into three key sections: findings, discussion and gap analysis.

FINDINGS
Findings reveal two distinct pathways to childhood stunting, direct and indirect causes, and therefore two stages that present opportunity for intervention in early life. Direct causes comprise the factors that biologically enable growth faltering; namely protozoan and helminthic infections, diarrhoea and EED, a subclinical disease of the small intestine (Dodos et al. ). Malnutrition in the form of undernutrition also presents itself as stunting or wasting, often exacerbated by the causes above (Null et al. ).
Indirect causes comprise the environmental influences that enable these biological factors to manifest: chronic poverty, the broader political and socio-economic environment, cultural practices and climatic conditions (Dodos et al. Early literature typically attributes stunting to diarrhoea; however, over time studies have found a correlation between stunting and EED, distinguishable due to their symptomatic and asymptomatic natures, respectively (Prendergast et al. ). Stunting rates have typically been unaffected by WASH, behavioural and nutritional interventions to reduce diarrhoea, which has led to hypothesis of an alternative causal mechanism (Pickering et al. ).

).
Additionally, indirect pathways and contextual factors prove significantly influential (Budge et al. ). The poverty cycle through low income, restricted access to WASH resources and food, and ill health is a fundamental barrier to development (Esrey et al. ). It is itself influenced by the wider political and economic climate, worsened by political volatility, economic instability and conflict (Atinmo et al. ). Most removed from control yet significantly influential is the seasonal climate responsible for droughts and floods, straining food and water resources, increasing economic vulnerability and exposure to disease (Hyland & Russ ).  The following section will review these direct causes, indirect causes and interventions in further detail.

Environmental enteric dysfunction
EED is a subclinical disease of the small intestine; however, the exact aetiology is not fully understood and difficult to determine given its asymptomatic nature (Budge et al. ). Literature attributes EED to chronic pathogen contact and ingestion which alters the fundamental biological gut structure (Prendergast et al. ). However, statistical data   Wagner & Lanoix (1958)).
quantifying trends between EED and stunting are lacking, likely due to the absence of reliable biomarkers, lack of symptoms and use of diarrhoea prevalence as a proxy measurement (Prendergast et al. ).
Little research or evidence exists regarding the presence of EED at birth, which if present infers that causes are present during pregnancy, questioning the priority and efficacy of pre-natal versus post-natal interventions.

Diarrhoea
Diarrhoea is typically caused by viral or bacterial infection through faecal-oral pathogen transmission and is frequently used as an indicator of stunting due to its symptomatic nature (WHO ). A self-perpetuating cycle between diarrhoea and malnutrition is evident, which suggests a causal pathway linking diarrhoea to stunting via nutrition (Brown et al. ). There is also evidence that diarrhoea and EED can occur simultaneously, questioning whether studies could unknowingly indicate a false positive correlation between diarrhoea and stunting due to the (unknown) presence of EED (Schmidt ).

Indirect causes
The poverty cycle and economic influences Intergenerational and caregiver trends Evidence shows that stunting follows a generational cycle; stunted mothers, younger mothers and mothers with a

Water quality
Trials mostly examine the relationship between water quality and direct causal pathways, rather than stunting specifically. A relationship between water quality and diarrhoea is apparent; a study in Ethiopia found 3.7 times increased likelihood of diarrhoea in children who use an unimproved water source (Marugán et al. ).

Hygiene and behavioural interventions
Handwashing successfully interrupts the pathogen transmission pathway according to the 'F' diagram, specifically when encouraged before eating and after defaecation (Brown et al. ). In one study, maternal handwashing with soap at critical times after defaecation and before feed-

Pathways to stunting
The lack of a finite understanding of causal pathways through diarrhoea, EED and malnutrition presents challenges when designing, conducting and interpreting trials, and implementing interventions elsewhere.
The lack of a useable biomarker to determine EED is a particular problem and has resulted in trials using diarrhoea incidence as a proxy for EED presence. However, as EED is often asymptomatic, diarrhoea presence can confirm either diarrhoea alone or in combination with EED (WHO ).
Also, given the mixed evidence of the relationship between diarrhoea and EED, it is possible that two very distinct issues are being combined complicating and possibly invalidating results (Mbuya & Humphrey ). Serology as an indicator of infection would provide additional information for possible EED presence; however, as EED does not as yet have known unique biomarkers, this would still not enable a conclusive diagnosis. Some studies have shown that improved sanitation reduces stunting prevalence but has no effect on diarrhoea and that studies showing a reduction in diarrhoea incidence typically involve water or combined interventions. This suggests another link between faecaloral transmission and stunting, and supports the argument that EED is a primary causal pathway warranting further research.

Sanitation
Most efficacious sanitation interventions were those that most effectively interrupted the faecal-oral transmission pathway.
Access to a latrine for adult use was shown to reduce childhood stunting rates in four trials reviewed in this Similarly, other trials examining sanitation interventions found solutions aimed at reducing open defaecation to adoption of any fixed-point facility to most effectively reduce stunting. Disposal of child faeces into an improved sanitation facility also showed reductions in stunting across a multiple country analysis, however found significant challenges to adherence and sustained behaviour change. It follows that to be effective, sanitation interventions must have widespread adoption else children continue to be exposed to contaminated land and water sources from a cycle of faeces contamination.
Trials commonly stressed the importance of behavioural education to achieve widespread behavioural change. Those that did not include strong educational elements, such as the SHINE trials, found a reduced effect on stunting, EED or diarrhoea prevalence.
It follows that sanitation interventions should focus on urgent provision in environments with no pre-existing facilities to reduce open defaecation and direct faecal-oral exposure. Significant investment in supportive social behavioural change programmes is justified to overcome issues of widespread use, sustained hygiene practices and cultural gender barriers.

Water availability
Piped water into the home proved to be strongly linked with a reduction in stunting rates through increased water use per capita, improved hygiene and household cleanliness, and in combination with improved sanitation. Typically, the more sophisticated the infrastructure (such as networked infrastructure with safe treatment), the greater the reduction in stunting and diarrhoea rates. However, much of this research failed to consider affordability for financially Little information is available regarding water storage and handling methods, which have high potential to contaminate water through stagnation and bacterial growth, likely negating any previous water quality treatments.
Efforts to investigate appropriate storage and handling methods would, therefore, be beneficial.

Hygiene and behavioural interventions
Trials examining handwashing typically assess it in relation to diarrhoea rather than stunting and have found that handwashing reduces diarrhoea rates when practised regularly at critical times before eating and after defaecation. The provision of soap is also critical due to its ability to kill bacteria and viruses. Promising Achieving the behavioural change required to increase handwashing practices requires continued hygiene education, necessitating interventions to target social factors.
Given the mixed results of WASH interventions alone, it provides hope that WASH efficacy may be bolstered with stronger, more thorough supportive social engagement programmes.
Greater improvements in stunting rates are observed when improved hygiene behaviours are adopted on broadscale community levels. For example, the benefits of the transition from open defaecation to fixed-point defaecation are only observed following widespread adoption; household and community environments will only remain free from animal faeces if all community members commit to separation of livestock; and change to inherent cultural beliefs will only be achieved as a collective. As mothers are not always the primary caregiver, others responsible for childcare must also maintain appropriate hygiene practices and interventions must target this demographic.
There may also be possible benefit to exploring community health centres to generate social inclusion and support which has potential to increase community-wide behaviour change.
There is evidence that higher maternal and caregiver educational levels correlate with decreased stunting rates in children under 2 years; attributed to a higher baseline knowledge of suitable care practices. However, in the absence of formal education, targeted nutritional and hygiene education may overcome the immediate barriers to childcare practices. In the longer-term, higher levels of formal education will assist with improved employment opportunities and the chance to interrupt the poverty cycle.

Feeding, nutritional and combined WASH interventions
Exclusive breastfeeding until 6 months has consistently shown correlation with lower stunting rates. High adoption rates are likely due to a preference for natural behaviours free from social barriers associated with other interventions.
Exclusive breastfeeding until age 6 months ensures that infants receive the required nutrients without need for complementary feeding. Where mothers or the primary caregiver are unable to breastfeed, clean water is of even greater importance to help protect against environmental contamination for sanitary food preparation.
This study does not focus on exclusive nutritional interventions; however, the SHINE and WASH Benefits Kenya trials have both shown promising results when implemented alone. The SHINE trials found nutritional interventions were as successful alone as in combination with WASH.
However, securing an adequate and varied diet is part of the larger pre-existing problem, and feasibility and sustainability of wide scale nutritional supplementation is questionable. Hence, further research into broader integrated WASH and nutritional interventions is justified through the necessity to find a more sustainable solution.

Study design
The wider environment and socio-economic context have been shown to significantly affect trial results, including climatic conditions, economic instability and political volatility. Longer-term studies with larger cohorts are suggested to better account for these variables, and to also enable intergenerational monitoring to reflect social development.
Secondly, infrastructural interventions are difficult to carry out under blind trial conditions, and many variables associated with this subject are inherently unamenable to randomised controlled trials (RCTs). WASH trials are particularly vulnerable due to the sensitivity of issues at hand.
Behavioural interventions are difficult to control and compare due to immeasurable consistency and reporting subjectivity. It may be beneficial to move away from blinded trials and RCTs in favour of locations and cohorts selected specifically for the situation, environment and WASH requirements. Lastly, there is often limited effort to consider or record pre-existing environmental conditions or existing facilities before implementing interventions. This is critical to enable accurate analysis, to determine the scale of change and to assess feasibility of replication elsewhere.

Cohort selection
The fundamental structure of these trials requires a continuous cohort throughout the trial period. Most studies only consider living children, despite that miscarriages and infant deaths may be attributable to the numerous primary and secondary factors discussed (Kaur et al. ).
No studies have been found including babies and children with pre-existing disabilities or illnesses. While adjusting for confounding illnesses would greatly complicate trials, it must be considered that pre-existing conditions and chronic poverty could pre-expose children to the primary biological causal factors, impact efficacy of WASH interventions or effect observed stunting rates.

Data collection methods
Most field trials comprise a combination of quantitative and qualitative data collection methods; each presenting merits and disadvantages and raising a case for studies using mixed methods. RCTs are typically considered preferable and benefit from more easily interpretable numerical data. However, numerous papers reviewed in this study fail to justify cohort or location selection, or contain uncontrolled variables preventing reliable interpretation of results.
Qualitative data collection using household surveys can achieve a deeper understanding of the issues at hand, however require a full understanding of the surrounding sociocultural behavioural factors. Qualitative surveys must be careful of subjective reporting for social reasons, and perception bias caused by narrow data collection time frames.
Mixed method studies to benefit from numerical data for analysis and qualitative data for interpretation of environmental and contextual factors would be of greater benefit.

Lag in anthropometric measurements
The lag between onset of causes and physical stunting evident through anthropometric measurements is unknown.
Similarly, there is a delay once interventions are put in place until changes in height are reflected in HAZ scores.
The papers studied in this report do not discuss or justify time frames chosen, and lack of evidence suggests that these periods are not fully understood.
The lack of a full understanding of the direct causal factors of stunting is an exacerbating factor. The incubation period of EED is unknown, so that the lag between onset, visibility of symptoms (if any) and observed stunting is unknown. An enhanced understanding may also help advise potential periods for catch-up growth.

Governance and the broader supporting context
Much of the available literature examines the efficacy of WASH interventions in field settings in isolation; however, little research considers the supporting mechanism required for implementation. Policy, finance and governance alongside public engagement must be in place to support and facilitate widespread application. The broader national political and economic climate has the potential to both aid and inhibit progress as the foundation of the poverty cycle.
Evidence supports that economic progression from lowerincome economies (LIC) to lower-middle-income economies (LMC) is reflected in population health and stunting prevalence.
Identification of this context would assist a full interpretation of results from studies, plan an optimal cross-sectoral approach and help achieve replication elsewhere. Finding effective solutions is fundamentally required, however ensuring their feasibility at scale, equity of access and a suitable mechanism to implement, manage and maintain them is as essential.

GAP ANALYSIS AND RECOMMENDATIONS
The following gap analysis presents concise recommendations for future work informed by the former findings and discussion.

WASH and babyWASH
More research is required to test WASH interventions targeted at children in their first 1,000 days specifically.
However, research studying WASH interventions across multiple age groups have shown to be effective at reducing stunting in children under 2 years, thereby warranting research into more targeted application to this age group. Social behaviour change and education interventions have proved a critical factor to achieve higher adherence.
The time burden of WASH activities and associated gender roles have shown strong correlation with the poverty cycle and stunting. However, no trials were found to explore this issue specifically despite the urgency warranted due to the potential scale of impact to childcare and stunting.

Nutrition and feeding practices
Trials such as SHINE found that WASH interventions did not improve efficacy of nutritional interventions alone.
However, more radical and robust WASH solutions should continue to be explored in combined trials to find a more sustainable solution to widespread nutritional supplements. Initiatives to better integrate WASH interventions into nutritional programmes, in addition to trialling groups concurrently, would be beneficial.
Breastfeeding until at least 6 months should be promoted as an effective, easily implementable and natural behaviour to ensure young babies are provided with necessary and adequate nutrients in their early months.
Breastfeeding interventions must be continued as an intrinsic part of stunting reduction, but supporting WASH interventions are necessary to provide an uncontaminated environment for sanitary food preparation and improved food safety.

Research methodologies
A move away from RCTs to include more varied methodologies would add value. Less reliance on RCTs and a conscious selection of locations and cohorts should be trialled to tailor interventions for each setting and achieve results which better reflect 'real world' application. Study locations should be aligned with areas of high stunting prevalence, with the aim of addressing the worst affected areas as a priority. Holistic contextual and environmental factors should also be considered in location selection and interpretation of results.
Cohort selection should consider children with existing disabilities and illnesses, to identify whether other illnesses pre-expose children to the primary biological causes of stunting. If so, there may be opportunity to address pre-existing conditions to reduce stunting.
Innovative research methodologies and radically revised data collection methods are required to find novel ways to address uncontrolled variables. Methods to reflect the sensitivity of issues discussed and discourage conformation to social pressures and desirability should be explored. Similarly, socio-cultural and behavioural studies to gain an understanding of the social context would help to more accurately interpret results. Communities need to be actively involved to ensure people are supportive and willing to adopt chosen interventions through mechanisms such as participatory research. The use of implementation science to optimise adoption of interventions would be beneficial.

Governance
WASH research should be considered as part of a broader, multi-sectoral programme. Papers fail to consider the wider supporting mechanism, which has potential to be both a limiting and enabling factor. If explored in context, outcomes of research would offer a fuller understanding of efficacy and a more realistic assessment of feasibility.
Cross-sectoral collaboration is required to involve multiple disciplines to ensure social and behavioural science is combined with health and engineering fields for valuable varied perspectives.

CONCLUSION
This report demonstrates the complex interlinkages between direct and indirect causal pathways and stunting. It exposes the heterogeneity of findings to date, and necessity for further research into both biological and contextual influences. However, despite mixed results, it identifies many promising findings showing the ways in which WASH and babyWASH interventions can reduce child stunting. WASH shows to be particularly effective in reducing faecal-oral pathogen transmission through interruption of the 'F' diagram, with particular success using sanitation interventions to reduce open defaecation, and water interventions to bring water nearer to the home to reduce contamination exposure points. Although not specifically targeted at children under 2 years, these interventions proved some of the most effective to reduce child stunting through the improvement of broader contamination and hygiene levels. The results show that these broader WASH interventions typically need to be in place first for more targeted babyWASH solutions to build on; sanitation must be in place for correct disposal of child faeces, water must be available to maintain hygienic feeding practices, and widespread social behavioural change is needed to improve child handwashing, play and feeding behaviours.
For future research, trial methodologies must be revisited to more fully consider broader social and contextual factors, and apply mixed quantitative and qualitative data collection methods. WASH must be considered as part of a broader interdisciplinary programme to address the impact of governance and the wider economic and political climate. Regional, national and international support is also critical to address the underlying socio-economic factors responsible for intergenerational stunting. Finally, and critically, future studies must explore WASH and baby-WASH as a broader network of research, drawing on expertise across multiple social science, medical and engineering fields.