An assessment of availability of handwashing facilities in households from four East African countries

The World Health Organization provides guidelines on handwashing as part of the global campaign towards achieving proper hygiene. In East Africa, cholera and diarrhoea outbreaks and, most recently, COVID-19 remain a threat to public health – calling for the promotion of handwashing to prevent infection. Using data from demographic and health surveys in four East African countries (Kenya, Rwanda, Tanzania and Uganda), we estimate the prevalence and identify the predictors of the availability of handwashing facilities in households. Findings indicate that the presence of a handwashing facility is not universal in the four countries: Kenya (66.4%), Rwanda (76.4%), Tanzania (80.7%) and Uganda (59.2%). Results from the pooled binary logistic regression model indicate that age, sex and education of the household head, type of place of residence, number of children, and household wealth are strong predictors of having handwashing facilities in all countries. However, the likelihood of having a handwashing facility in Uganda is lower than other countries. This study provides a rich understanding of the factors that explain the availability of handwashing facilities. Findings indicate how prepared the four countries are in the face of the COVID-19 pandemic – and can guide the policy direction in the prevention of infection.

at all levels has been implemented (Ministry of Infrastructure ). In Tanzania, sanitation programmes have aimed at increasing water utility points in slums and rural settings (Thomas et al. ). Non-water hygiene practices (use of sanitizers) have been promoted in areas where there is water scarcity. The government of Tanzania has also promoted community-led programmes to increase sanitation practices but also disseminate sanitation messages through mass media (Madon et al. ). In Uganda, demonstration campaigns have been adopted using a multi-sectoral approach aimed at promoting sanitation programmes. For example, a participatory hygiene and sanitation transformation (PHAST) and engaging of CHEWs as key strategies to promoting basic public health have been promoted (Dumba et al.  How do the four countries compare with each other? Our main contribution to the field of hygiene is that this study makes use of data collected for the first time in some DHS datasets, to provide a regional-country comparison of handwashing facilities in households.

THEORETICAL FRAMEWORK
We adopt the 'Focus on Opportunity, Ability, and Motivation (FOAM)' framework ( Figure 1) that was designed to monitor and evaluate handwashing behaviour in low-resource settings, particularly in SSA (Coombes & Devine ).
According to the FOAM framework, for one to perform a certain behaviour, an opportunity and the ability to perform such a behaviour must prevail. The focus relates to the target population and the target behaviour one wishes to perform (Coombes & Devine ). Opportunity relates to the factors such as soap, water, handwashing facilities that would facilitate the behaviour to be performed. Ability measures one's capability to perform a behaviour (Coombes & Devine ).
Based on the literature, we hypothesize that: (1) households found in rural areas are less likely to have a handwashing place than households in urban areas; (2) households with no children are less likely to have a handwashing place than households with children; (3) households in the richest wealth quintile are more likely to have a handwashing place than their counterparts in the poorest wealth quintile; and (4) household heads with higher levels of education are more likely to have a handwashing place than their counterparts with no education.

Source of data and sample size
We used data from the DHS programme (https://dhsprogram.

Dependent variables
The variables we use in this study largely hinge on the FOAM framework although some aspects of the framework were not analysed due to data limitation. We adopted the opportunity aspect in the framework to be the dependent variable. The DHS groups households into five categories: observed fixed place, observed mobile place, not observed or not in a dwelling, not observed or not permitted to see, and not observed for any other reason. We created a binary variable coded yes (1) to indicate presence of a handwashing place for households whose handwashing facility was observed and no (0) for the absence of a handwashing place for those not observed.
The DHS also collects data about water and soapas opportunity factors. We ran checks to determine whether simply having a place for handwashing would yield a variable notably different from one that also takes into consideration the availability of water and soap. We found that having a place for washing hands was strongly correlated with also having soap and running water. Therefore, the findings we report in this study are generalizable to the scope of a handwashing place, water, and soap or detergent.

Independent variables
According to the FOAM framework, three variables formed the target population (focus): type of place of residence, sex of household head and number of children five years and below. The type of place of residence was either rural or urban. The number of children living in households who are younger than six years was categorized into three groups: households with no children, one to two children, and more than two children. Sex of the household head was a binary variable (male or female). Age, level of education and household wealth are ability factors while interruption of water supply, presence of soap/detergent, location of water source, and presence of water at handwashing place were adopted as opportunity factors. We created seven categories for the age of the household head from reported ages in single years, with the first six categories being in ten-year age groups (10-19, 20-29, 30-39, 40-49, 50-59, 60-69), and the last one being open ended (70 or more). We elected to start from age 10 because the lowest reported age of the household head was 12 years in Tanzania, 13 years in Uganda and 14 years in both Kenya and Rwanda, irrespective of marital status.
Data were collected on whether water was not available for at least a day in the last 2 weeks preceding the survey, from which we created a variable 'interruption in water supply'. Households which reported to have experienced an interruption in water supply for a full day or more were coded 'yes', otherwise 'no'. However, data on interruption in water supply were not collected for Kenya and Rwanda.
Presence of soap or detergent is a binary variable, 'yes' otherwise 'no'. Households whose source of water to use was in 'own dwelling, or yard or plot' were coded 'own dwelling' otherwise 'elsewhere'. The highest education of the household head was categorized into four groups: no education, primary, secondary, and higher. Household wealth was categorized into five groups: poorest, poorer, middle, richer, and richest. Households that never had water available at the handwashing place were coded 'not available' otherwise 'available'.

Data analyses
Comparative distribution of households was presented at the univariate level of analysis while associations between household characteristics and having a handwashing place were at the bivariate level of analysis. At the multivariate level of analysis, a binary logistic regression model was fitted to examine the household predictors of having a handwashing place because the dependent variable is a dummy variable. The largest category in each variable was used as a reference category. The analysis in the first model was performed for each country separately but the results generated were compared across the four countries.
For comparison between countries, we run a pooled model, first, without interactions and, later, with interactions between country and the selected household variables to estimate accurately the average estimate of the exposure effect (Stukel et al. ). This is possible because the variables we include in the analyses are uniformly coded and defined across all studies. However, the variable on the interruption in water supply was not included in the pooled model since it was not in some datasets. All the results presented are adjusted to take care of the appropriate weights for representativeness to the whole population and appropriate adjustments for non-response and missing values. All analyses were performed using the STATA version 14 software.

Characteristics of households
Results shown in Table 1 show a similar distribution for most variables in all countries. Most household heads in all countries were males, with primary level of education and in the age group 30-39 years. The results in Table 1 show that household wealth varied by country: the lowest proportion of households were in the poorest wealth quintile in Kenya (16.7%) and Tanzania (16.8%) while the richer (18.0%) and middle (18.5%) wealth quintiles constituted the lowest proportion of households in Rwanda and Uganda, respectively. On the other hand, the richest wealth quintile constituted the majority proportion of households in all countries except Rwanda.  Table 2 indicate that the proportion of households with no soap or detergent at a handwashing place is lowest in Tanzania (40.3%), but highest in Uganda (53.3%). Most households had their source of water located outside the dwelling, but the majority had water present at a handwashing place.

Prevalence of having a handwashing place/facility
The results in Figure 2 show that more than half of the households in the four countries had a handwashing place,  Association between selected household variables and the presence of a handwashing place The results in Table 3 show that most households had a handwashing place in their dwelling except for the poorest households in Uganda. This is an expected finding considering the challenges poor households face in terms of accessing clean water, gadgets for keeping clean water and detergents used when washing hands. The results presented in

Pooled model that predicts handwashing
Overall, the results from the pooled model (Figure 3) Figure 3 show that the likelihood to have a handwashing place for households in urban areas is higher than in rural areas in all countries.
Households with three or more children or those with no children emerged to be strong predictors of a handwashing place in all countries. However, the likelihood to have a handwashing place is lower for households with no children compared to households with one to two children.
In all countries, the likelihood to have a handwashing place is lower among households with heads who do not have any form of education compared to their counterparts.
Household wealth was significant at predicting a handwashing place, but the likelihood is lower for households in the poorest, poorer, middle and richer wealth quintiles com- The graphical representations in Figure 4 show a clear pattern of the type of place of residence by country on p-value 0.000**** 0.000**** 0.000**** 0.000**** Notes: Results for soap or detergent or presence of water are not shown because there was no variation. Similarly, all households had water for washing hands at a handwashing place. **p < 0.05; ***p < 0.01; ****p < 0.001.  Confidence intervals in parentheses. **p < 0.05; ***p < 0.01; ****p < 0.001. Of the four countries, Tanzania's approaches that increase awareness and promote better sanitation practices through CHEWs, increase of water utility points and promotion of non-water hygiene strategies are good approaches at promoting sanitation practices (Thomas et al. ). While all countries make use of CHEWs to promote handwashing with soap and water, only Tanzania has promoted nonwater hygiene practices but also increased water utility points. However, the results observed for Uganda may be influenced by the fact that the government is more concerned with policy formulation and not service delivery (Ekane et al. ). Although this paper did not explore actual handwashingbecause of data limitationswe assume that households with handwashing facilities are more likely to practise handwashing behaviours, even though this may not always be true.

Limitation
The main shortcoming relates to the dependent variable used in this studywhich is based on data collected through observation, yet direct observation of handwashing practice is difficult during household surveys.

Recommendation
It is important that the DHS programme consistently collects data on handwashing facilities to allow a trend analysis for definitive conclusions and to offer better-tailored solutions.

DATA AVAILABILITY STATEMENT
All relevant data are available from an online repository or repositories (https://dhsprogram.com/).