Evaluation of water, sanitation and hygiene program outcomes shows knowledge-behavior gaps in Coast Province, Kenya

Introduction Water related diseases constitute a significant proportion of the burden of disease in Kenya. Water, sanitation and hygiene (WASH) programs are in operation nation-wide to address these challenges. This study evaluated the impact of the Sombeza Water and Sanitation Improvement Program (SWASIP) in Coast Province, Kenya. Methods This study is a cluster randomized, follow-up evaluation that compared baseline (2007) to follow-up (2013) indicators from 250 households. Twenty-five villages were selected with probability proportional to size sampling, and ten households were selected randomly from each village. Follow-up data were collected by in-person interviews using pre-tested questionnaires, and analyzed to compare indicators collected at baseline. Cross-sectional results from the follow-up data were also reported. Results Statistically significant improvements from baseline were observed in the proportions of respondents with latrine access at home, who washed their hands after defecation, who treated their household drinking water and the average time to collect water in the dry season. However, this study also observed significant decreases in the proportion of respondents who washed their hands before preparing their food, or feeding their children, and after attending to a child who has defecated. The analysis also revealed a knowledge-behavior gap in WASH behaviors. Conclusion SWASIP contributed to improvements from baseline, but further progress still needs to be seen. The findings challenge the assumption that providing infrastructure and knowledge will result in behavior change. Further understanding of specific, non-knowledge predictors of WASH related behavior is needed.


Introduction
It is estimated that 10% of the global burden of disease results from unsafe water, poor sanitation or inadequate hygiene [1]. Due to poor management of water resources and inadequate sanitation, the human consumption of unsafe water poses a major challenge to population health in many regions of the world [2]. The scope of these problems is broad and even though the etiologies of disease are varied, they are transmissible and thus, preventable [3].
Unfortunately, these diseases persist because 900 million people globally live without access to safe-water [1], and one billion people live without access to any type of sanitation facility whatsoever [3].
In Kenya, 17 million of the country's 40 million inhabitants do not have access to clean drinking water [4]. Water scarcity is becoming a more pressing concern as the population of Kenya is growing faster than infrastructure can be built for water and sanitation. The World Bank estimates that from 2011 to 2025, Kenya's per capita water consumption will drop from 630 to 235 cubic meters per person per year [4]. In the wake of the Millennium Development Goals (MDG), there are a number of programs operating that specifically target water and hygiene instability, yet many face sustainability challenges with infrastructure, continuity of funding and program policy support [5][6][7]. Sombeza Water and Sanitation SWASIP had three major program components. Firstly, the program constructed water and sanitation infrastructure in schools and communities such as roof water catchments, latrine blocks, hand hygiene stations, small farm reservoirs, public taps and community pipelines. Secondly, the program delivered health and hygiene promotion education to communities and schools, employing Community Led Total Sanitation (CLTS) methods, which have been adopted by over 60 countries worldwide as the primary means to improve sanitation in rural communities. CLTS aims to sensitize communities to the importance of sanitation and hygiene and eliminate open defecation [8]. Lastly, SWASIP constituted Water User Associations to manage and maintain the community WASH infrastructure. In this paper, we presented the results of a household survey, which was one component of an impact evaluation that was conducted in 2013 to assess the sustainability and impact at the household level of these WASH interventions.

Study Design
This study is a cluster randomized comparison study between baseline in 2007 and follow-up in 2013. We surveyed 250 households in the Kinango district of Coast Province, Kenya.

Sample Size and Participant Selection
This study was designed to detect a 15% change from baseline on key indicators including latrine coverage, distance to water source, and hygiene behaviors (α=0.05, two-sided, and power=80%).
Households were sampled by using probability proportional to size cluster sampling. The design effect of cluster sampling was calculated to be 1.27, based on the intra-cluster correlation coefficient of 0.03 from a WASH study in Nyanza Province, Kenya [9]. The required sample size was estimated to be 218 households.
Twenty-five of the 67 villages in Kinango that were intervened by SWASIP were selected by probability proportional to size cluster sampling. A total of 250 households were selected, 10 households were randomly selected from each of these 25 villages. One participant from each selected household was interviewed. This person had to match the following inclusion criteria: had been residing in that household for more than 3 years, was older than 18 years old, and was the primary caregiver of the household.

Data Collection
The survey tool combined relevant items from the USAID Hygiene Improvement Project [10], and the SWASIP tools used in 2007 for a baseline study. Behaviour change questions were modelled on the "Focus on Opportunity, Ability and Motivation" (FOAM) framework for hygiene and sanitation behaviour change [11,12]

Data Management and analysis
Data were entered into a data entry screen using EpiInfo 7 [13]. To minimize data entry errors, 50% of the data were re-checked for accuracy and were found to be accurate. Statistical analyses were done using STATA-12 [14]. Descriptive statistics were conducted on survey outcomes to report summary statistics. Eleven indicators were identified with sufficient baseline (2007) data to allow for direct comparisons with 2013 data (follow-up analysis). Two sided one-sample t-tests were conducted on these indicators to compare with baseline estimates. Results were expressed as mean ± 95% confidence interval (95% CI) for continuous variables and proportions (or percentages) ± 95% CI for dichotomous variables.
Logistic regression analysis was used to find predictors of hand washing behavior, latrine ownership, and household drinking water treatment. Results were reported as odds ratio (OR) ± 95% CI. A p < 0.05 was considered for statistical significance.

Ethics
Ethics approval was obtained for this study from the University of Alberta Research Ethics Office in Edmonton, Canada, and from the Aga Khan University Research Ethics Committee in Nairobi, Kenya.
Informed verbal consent was obtained from study participants.

Confidentiality
was strictly maintained throughout data management, analysis and report writing.

Participant Demographics
Survey respondents were predominantly 18-30 year-old primary caregivers, with low education levels. Forty-two percent (42%) of the total respondents reported having no education whatsoever and 62% of the total had less than Class 6. Two-thirds (68%) of respondents were employed as farmers or unemployed, while the remainder were either in small business or a working professional.
When respondents were asked if they felt responsible for their own health, and the health of their family, 90% answered yes.
Demographics for baseline respondents were unavailable.

Follow-up analysis
Eleven indicators were compared with baseline data from 2007 ( There was a decrease in the percentages of self-reported hand washing at the remaining four critical moments for hand hygiene behavior which are: before preparing food (-9%), before feeding children (-28%), before eating (-4%), and after attending to a child who has defecated (-37%).

Knowledge behavior gap
Six indicators covering hand hygiene, water treatment and toilet use were selected for a knowledge-behavior gap investigation.
Differences between knowledge and behavior were observed, at varying degrees, in each of the six indicators and are displayed in Figure 1

Discussion
The improvements in water access and sanitation facility coverage were significant and are a testament to successful programming.
There is a known, complementary health benefit to communities when latrine coverage and water consumption are improved concurrently and these benefits will likely be appreciated [16]. impractical from them to adopt [21]. Another theory to explain suboptimal latrine use posits that cultural taboo influences latrine use.
In Kilifi, a neighbouring district to where this study was conducted, some residents believe that a man's feces should never mix with his daughter-in-law's or that a person's feces can be used in witchcraft to bewitch him [22]. Our study did not find evidence that could support or refute this theory. None of our respondents mentioned taboo as a barrier to latrine use, however there may a social desirability bias to answering questions on latrine use in a socially acceptable way. This collection bias may have also artificially inflated the proportion of respondents reporting that they practice good hand hygiene and use a latrine, which has been described by other researchers working in South Asia [23]. In addition to the collection bias described above, this study was limited by baseline data that were incomplete and variance statistics of mean point estimates could not be utilized in the analysis but were assumed to

Conclusion
Significant improvements from baseline were observed, yet overall levels of latrine coverage are still low. This is likely a symptom of a successful project that was terminated before larger gains could be realized as self-sustaining behavior change may take longer commitments than a three-year program. Healthy WASH practices are still hindered, predominantly, by non-knowledge barriers such as convenience and financial insecurity. There are two recommendations for further practice. The first is to reinstate the successful health and hygiene promotion interventions to continue progress with increasing latrine coverage and healthy WASH practices. Along with this, it is recommended that funders consider this needed longevity when describing funding terms. Secondly, future programming must not rely on an unverified assumption that providing knowledge and infrastructure, even together, will result in changes in hygiene or sanitation behaviors. The socio-cultural context in which WASH decisions and behaviors are operating is complex and intermingled. It would be prudent to first understand and describe the non-knowledge predictors of WASH practices in a community when conceptualizing future WASH programs for implementation.
What is known about this topic  Unsafe water and poor sanitation are significant contributors to global morbidity and mortality.
 Kenya is experiencing water scarcity and low latrine coverage with population needs outgrowing infrastructure support.
 Challenges of sustainability with WASH infrastructure exist due to inconsistent funding and policy support.

What this study adds
 Short term or intermittent funding for WASH infrastructure precludes its safe and reliable functioning.
 There is a knowledge behavior gap with WASH practices, likely due severe financial constraints, inconvenience and to a lack of felt responsibility for health.
 It is unfounded to assume that providing WASH infrastructure and education, even together, will affect practices. The socio-cultural context needs to be considered when designing health behavior change programming.

Competing interests
The authors declare no competing interests.

Acknowledgments
The lead author would like to extend his sincerest thanks to the study participants and his gratitude and admiration to the staff at the Department of Community Health in Mombasa. In particular, Chris Mwaringa was instrumental to the success of the fieldwork required for this evaluation and deserves high commendation.

Personal Hygiene
Most common reasons for not washing hands (n=187