Abstract

This study aims at assessing the determinants of microbiological contamination of household drinking water under multiple-use water systems in rural areas of Ethiopia. For this analysis, a random sample of 454 households was surveyed between February and March 2014, and water samples from community sources and household storage containers were collected and tested for fecal contamination. The number of Escherichia coli (E. coli) colony-forming units per 100 mL water was used as an indicator of fecal contamination. The microbiological tests demonstrated that 58% of household stored water samples and 38% of protected community water sources were contaminated with E. coli. Moreover, most improved water sources often considered to provide safe water showed the presence of E. coli. The result shows that households' stored water collected from unprotected wells/springs had higher levels of E. coli than stored water from alternative sources. Distance to water sources and water collection containers are also strongly associated with stored water quality. To ensure the quality of stored water, the study suggests that there is a need to promote water safety from the point-of-source to point-of-use, with due considerations for the linkages between water and agriculture to advance the Sustainable Development Goal 6 of ensuring access to clean water for everyone.

INTRODUCTION

Lack of access to safe and adequate water supply and the health risks associated with water-related diseases are major public health problems in many developing countries. Today, more than 663 million people, who mostly live in developing countries, are without access to improved water sources (WHO/UNICEF 2015; according to the WHO/UNICEF JMP definition, water sources such as piped water into dwelling/yard/plot, public tap or standpipe, tube-well or borehole, protected dug wells, protected spring and rainwater collection are considered to be improved while unprotected dug wells/spring, and surface water such as river, dam, lake, pond, stream, canal, irrigation channel are considered as unimproved sources). More than 0.8 million people die annually from diarrheal diseases due to unsafe drinking water, poor sanitation and inadequate hygiene (WHO/UNICEF 2015). Unsafe drinking water is considered to be one of the major causes of diarrhea (Zwane & Kremer 2007).

The WHO/UNICEF Joint Monitoring Program (JMP) for sustainable water supply and sanitation defines access to improved drinking water in terms of the types of technology and levels of service provided. In Ethiopia, it is estimated that 57% of households have access to an improved drinking water source, with a higher proportion of urban residents (93%) than among rural residents (49%) (WHO/UNICEF 2015). This definition of access to ‘improved’ water source does not consider the safety or quality of the water; consequently, it does not reliably predict either the microbiological or the physiological quality of the water being consumed. It is therefore argued that inclusion of water safety parameters will further reduce the actual coverage level of improved water sources reported by the WHO/UNICEF due to the high risk of microbiological re-contamination in many developing countries (Godfrey et al. 2011; Bain et al. 2014).

Ethiopia has made remarkable progress over the last decade to improve the water supply situation in the country. The Ethiopian government standard for the rural population is at least 15 liters of water for everyone per day within 1.5 km of their home. However, as a result of limited improved water availability, most rural populations rely on unimproved water sources such as unprotected springs, shallow wells, and rivers as a source of water for domestic uses which are easily polluted by human and animal feces. To make the matter worse, most of the existing protected community water sources are often contaminated with fecal matter (Butterworth et al. 2013; Amenu et al. 2014). It is estimated that unsafe drinking water and poor sanitary conditions account for 70% of the diarrheal disease burden in the country (Federal Ministry of Health 2005).

Given that the problem of point-of-use (POU) water quality is complex, subjective judgments about stored water quality based on the types of sources can be misleading. This paper, therefore, aims to identify key factors that influence the quality of drinking water stored in the households in two rural districts of Ethiopia. It investigates the quality of stored water used for human consumption at a household level and of community water systems for multiple uses, where drinking water supply and sanitation may be lacking. In doing so, this paper addresses two main research gaps. First, existing studies that examine the determinants of stored water quality and its relationship with rural water supply sources and household sanitary behaviors are rare: they primarily focus on the impact of water source types on stored water quality and ignore hygiene- and sanitation-related factors (Amenu et al. 2014). Second, determinants of stored water quality under multiple-use water systems is understudied (Scheelbeek 2005; Sutton et al. 2011; multiple-use water systems refer to where communities use a given water source for more than one economic activity such as drinking and/or washing, and for irrigation to grow crops and vegetables). Such a non-exclusive water supply system creates competition for water between domestic and productive uses. Water used for irrigation of crops may have complex interactions with drinking water in rural areas where access to improved drinking water supply is inadequate or lacking. Research in this area is therefore crucial to enhance our understanding of the determinants of the microbial quality of stored water in rural households of Ethiopia. Such studies will also help policy makers to design effective intervention to improve access to safe drinking water.

The remainder of the paper is structured as follows. In the following section we present the data with some descriptive analysis, with the next section presenting the estimation results. This is followed by the discussion and then the concluding section.

MATERIALS AND METHODS

Description of the study areas

This study was carried out in the Fogera and Mecha woreda (districts) of the Amhara National Regional State (ANRS) of Ethiopia. Wereta and Merawi are the respective administrative towns of Fogera and Mecha districts and are situated 615 and 523 km north of Addis Ababa, respectively. Merawi is located 34 km from Bahir Dar city – the capital city of ANRS, and Wereta is located 59 km away from Bahir Dar. As of July 2012, the population of Mecha and Fogera districts was estimated to be 334,789 (with an area of 1,481.64 km2) and 264,512 (with an area of 1,111.43 km2), respectively (Ethiopian Central Statistical Agency (CSA) 2013). Figure 1 depicts a map of Ethiopia and the study areas. The shaded area with light-green represents the ANRS map and the two selected districts (shaded in blue) are shown on the right panel of the figure.

Figure 1

Map of Ethiopia and of the two study districts. Source: Authors’ own illustration. Please refer to the online version of this paper to see this figure in color: http://dx.doi.org/10.2166/wh.2017.069.

Figure 1

Map of Ethiopia and of the two study districts. Source: Authors’ own illustration. Please refer to the online version of this paper to see this figure in color: http://dx.doi.org/10.2166/wh.2017.069.

Village and household selection

Fogera and Mecha districts were purposely selected from the ANRS. Although there is no official statistic for the coverage of water and sanitation for these districts, the average rural water and sanitation coverage in ANRS region is about 43 and 6%, respectively (CSA & ICF International 2012). The inhabitants of the districts identified that waterborne and water-related diseases are primary health problems where irrigation farming (the application of water for the purpose of crop and vegetable production) and multiple-use water systems are widely used. Administratively, regions in Ethiopia are divided into zones, which are subdivided into administrative units called woreda (district). Each district is further subdivided into the lowest administrative unit, called kebele. To select the sample households, a total of 20 kebeles were identified, 11 kebeles from Fogera and nine kebeles from Mecha district.

A stratified two-stage cluster sample design was used to select the sample households. In the first stage, 61 villages were selected randomly from the 20 kebeles. As the villages had different sizes, the probability of selecting a village within each kebele is proportional to the village size. Among the 61 villages, 39 were in Fogera and 22 were in Mecha district. Subsequently, in the second stage, 454 households were selected based on a systematic random sampling method. Of all the 454 selected households, 277 were in Fogera and 177 were in Mecha district. The lowest administrative division of the region (i.e. kebele) was used to form the first level of stratification. We used structured questionnaires (in English and Amharic, the local language) to collect household and individual level information.

Microbial quality of drinking water and sample collection

Although water contamination can have various origins, this study primarily focuses on Escherichia coli (E. coli) bacteria – one of the most common microbial quality indicators of drinking water (the identification of E. coli bacteria from sampled water is not complicated, and the result is obtained quickly and efficiently; however, it is only an indicator of fecal contamination. On the other hand, testing for all known pathogens is a complicated and expensive process in the study areas). Human and animal excreta are the primary sources of fecal coliform that cause waterborne diseases such as diarrhea, typhoid, and cholera among others. As stated in the WHO drinking water quality guideline, the concentration of E. coli bacteria as a microbial water quality indicator should be zero per 100 mL for the water to be considered safe for drinking (WHO 2011).

Drinking water samples were collected from the storage containers of all participating households in order to analyze the microbiological quality of the water for a random sample of 454 households using a portable water test kit (a product of Wagtech WTD, UK) in the field (enumerators asked household members (usually an adult woman) the following question, ‘could you please give me some water for drinking’ in order not to change their behavior). Using a membrane filtration technique, the test kit detects the presence of the E. coli bacteria which indicates a recent fecal contamination of the water. Stored water samples were kept in coded glass bottles that were properly sterilized using autoclaves in the local health centers at a temperature of 121°C for 30 minutes. The quality of water samples was tested between February and March 2014 – which is considered as a dry period in Ethiopia. This choice helps emphasize the processes of water contamination between clean point-of-source (POS) water (relative to rainy season) and the POU.

Immediately after the water samples were collected, growth pads were dispensed into a sterilized petri-dish and a dissolved media solution was poured over the growth pad. Then the water sample (until then kept in an icebox) was filtered through the membrane. When all the 100 mL water had been filtered, we placed the membrane on top of the pad, which had been saturated with the Membrane Lauryl Sulphate Broth (MLSB) media. In the next stage, we replaced the petri-dish lid and labelled it with a sample identification number and time, and placed the petri-dish into the petri-dish rack. Finally, we placed the filled rack into the incubator, and incubated the samples for 18–24 hours at a temperature of 44°C. Upon completion of the incubation period, we enumerated the number of E. coli colonies. E. coli concentrations were reported as colony-forming units per 100 mL (CFU/100 mL) of water sample. In a membrane filtration method, accurate enumeration of bacteria colony is difficult when E. coli colony counts are greater than 200 CFU/100 mL water. The organisms also exhaust the nutrients in the media if there are too many and they cannot grow effectively. It is recommended that the time between water sample collection and analysis must not exceed 6 hours, and it is one of the strengths of this work that we performed the test on-site immediately after collecting the samples from household storage.

Twenty-nine improved community water sources (protected hand dug wells/springs) were also tested for the presence of E. coli. If a community had access to more than one improved water source, we considered only the primary source from which most of the village households obtained their drinking water. However, some communities do not have access to improved water sources. Due to accessibility and resource constraints, we could not collect and test water from all community sources and thus focused on the primary village sources.

DATA

Descriptive statistics

The descriptive statistics describing the respondents’ background characteristics and socio-demographic variables are presented in Table 1. We found that literacy level (for reading and writing in the local language) is 9% for primary caregivers (the mother or the adult woman in the household taking care of the children) and 45% for household heads. Few individuals had completed primary school, indicating that the majority of the respondents in this study are illiterate. Moreover, based on the JMP classification; we found that 50% of the households obtained their drinking water from improved water sources such as protected wells/springs (see Table 1). The proportion of households having an improved water source is similar to the WHO/UNICEF (2015) progress report. More than 57% of households in the study areas practice open defecation, which is much higher than the rural national average open defecation rate of 43% (WHO/UNICEF 2014). Although half of the households got their drinking water from unimproved sources, the proportion of households applying any form of water treatment was extremely low (8%). This suggests that a general lack of awareness of the need to treat drinking water may exist among rural households in the study region.

Table 1

Descriptive statistics – household and community characteristics (n = 454)

Variables Description Mean SD 
Demographic characteristics 
 Household head age Age in years 37.72 8.64 
 Household head literacy 1 = read and write; 0 = otherwise 0.45 0.50 
 Primary caregiver age Age in years 30.33 6.64 
 Primary caregiver literacy 1 = read and write; 0 = otherwise 0.09 0.29 
 Highest education The highest grade completed in a household 3.50 3.05 
 Number of adult females Female household members aged >14 years 1.22 0.49 
 Household size Number of household members 5.98 1.77 
 Household density Number of people living per room 3.30 1.27 
Water, sanitation and hygiene 
 Primary drinking water sources    
 Private-protected dug wells  0.05 0.22 
 Shared-protected dug wells/spring  0.44 0.50 
 Unprotected dug wells/spring  0.39 0.49 
 Surface water  0.12 0.32 
 Minutes to water sourcesa Time needed for water collection for a round trip 24.18 14.19 
 Stored drinking water quality 1 = Contaminated with 1 or more E. coli (CFU/100 mL) 0.58 0.49 
 Household water treatment Treating household drinking water (1 = yes) 0.08 0.27 
 Water collection container 1 = Jerry cans; 0 = Clay vessel 0.83 0.37 
 Pit latrineb Households with a pit latrine (1 = yes) 0.42 0.49 
 Handwashing with soap Handwashing with soap by primary caregiver during handwashing demonstration (1 = yes) 0.27 0.45 
Garbage disposal 
 Dugout/burning  0.11 0.31 
 Throw-away in the yard  0.54 0.50 
 Throw-away outside the yard  0.13 0.34 
 Used as a fertilizer  0.22 0.42 
Agriculture 
 Irrigation Practicing irrigation farming (1 = yes) 0.66 0.47 
 Livestock holding Total livestock holding in Tropical Livestock Units 3.97 1.87 
 Assets value Total asset value excluding livestock in 1,000 Birrc 5.88 6.05 
Community characteristics 
 WUA Presence of WUA in a village (1 = yes) 0.29 0.46 
 Distance to health center Distance to the nearest health center in km 4.97 4.09 
Variables Description Mean SD 
Demographic characteristics 
 Household head age Age in years 37.72 8.64 
 Household head literacy 1 = read and write; 0 = otherwise 0.45 0.50 
 Primary caregiver age Age in years 30.33 6.64 
 Primary caregiver literacy 1 = read and write; 0 = otherwise 0.09 0.29 
 Highest education The highest grade completed in a household 3.50 3.05 
 Number of adult females Female household members aged >14 years 1.22 0.49 
 Household size Number of household members 5.98 1.77 
 Household density Number of people living per room 3.30 1.27 
Water, sanitation and hygiene 
 Primary drinking water sources    
 Private-protected dug wells  0.05 0.22 
 Shared-protected dug wells/spring  0.44 0.50 
 Unprotected dug wells/spring  0.39 0.49 
 Surface water  0.12 0.32 
 Minutes to water sourcesa Time needed for water collection for a round trip 24.18 14.19 
 Stored drinking water quality 1 = Contaminated with 1 or more E. coli (CFU/100 mL) 0.58 0.49 
 Household water treatment Treating household drinking water (1 = yes) 0.08 0.27 
 Water collection container 1 = Jerry cans; 0 = Clay vessel 0.83 0.37 
 Pit latrineb Households with a pit latrine (1 = yes) 0.42 0.49 
 Handwashing with soap Handwashing with soap by primary caregiver during handwashing demonstration (1 = yes) 0.27 0.45 
Garbage disposal 
 Dugout/burning  0.11 0.31 
 Throw-away in the yard  0.54 0.50 
 Throw-away outside the yard  0.13 0.34 
 Used as a fertilizer  0.22 0.42 
Agriculture 
 Irrigation Practicing irrigation farming (1 = yes) 0.66 0.47 
 Livestock holding Total livestock holding in Tropical Livestock Units 3.97 1.87 
 Assets value Total asset value excluding livestock in 1,000 Birrc 5.88 6.05 
Community characteristics 
 WUA Presence of WUA in a village (1 = yes) 0.29 0.46 
 Distance to health center Distance to the nearest health center in km 4.97 4.09 

aThe mean is calculated for households whose water sources are not in their own yard/premise.

bHouseholds who reported that they do not have any toilet facility defecate in the open.

cThe exchange rate during the time of the survey was 1 Euro = 26.02 Ethiopian Birr.

Statistical analysis

To examine the determinants of microbial quality of household stored water, socio-demographics, water sources, water collection time, storage, sanitary conditions, and waste disposal behaviors were assessed using simple chi-square analysis followed by a multivariate regression analysis. Admittedly, due to the collinearity among the variables and the cross-sectional nature of the data, our analysis is constrained to make any causal interpretation of the results. We instead investigate the degree of relationship between the microbial quality of stored water and socio-demographic, water sources, and sanitary factors.

In the multivariate analysis, we examined two different measurement specifications for the dependent variable (water quality). First, the dependent variable indicates the number of E. coli CFU/100 mL water. We transformed the dependent variable (E. coli counts) into the inverse hyperbolic sine (IHS), which is defined as: ihs (y) = log(y + sqrt(y2 + 1)) where y is the number of E. coli and is estimated using ordinary least squares (OLS) (the reason for this transformation is that we cannot take the normal log of y as we have many observations with zero value, and the distribution of E. coli is positively skewed because coliforms naturally grow exponentially). This transformation is an alternative to log transformation when the dependent variable takes zero values (MacKinnon & Magee 1990), and we interpret the coefficients of the explanatory variables in a similar manner to the log transformation. Second, we measured the dependent variable as a binary outcome, which indicates the presence or absence of E. coli, that is, y is equal to 0 if E. coli is less than 1 and y is equal to 1 if E. coli is greater than or equal to 1, and is estimated using maximum likelihood estimator.

RESULTS

Bivariate analysis

The bivariate analysis helps to examine if there are statistically significant relationships between stored water quality and other specific variables of interest. In the bivariate analysis, the water quality indicator is measured as a dummy variable (the variable is equal to 1 if there is 1 or more E. coli CFU/100 mL, and 0 otherwise; the range of CFU/100 mL in stored drinking water of the surveyed households was 0–195). The relationships between water sources, collection and handling practices, and stored water quality are presented in Table 2. The results show that types of water sources, water collection containers, and garbage disposal patterns have a statistically significant influence on stored water quality. Households who had so called ‘improved’ water sources showed much better microbial water quality than households who had either unprotected dug wells/springs or surface water sources. The result in Table 2 also shows a significant association between the types of water collection containers and stored water quality (p < 0.001). Conversely, household water treatment practices did not appear to have a significant influence on stored water quality. Moreover, the proportion of households with water contaminated with E. coli was slightly lower among households who had simple pit latrines than those who did not (p < 0.05). Similarly, households in which the primary caregiver washes her hands with soap had better stored water quality than households whose primary caregiver did not. Safe disposals of household garbage has an influence on household water quality (p < 0.001). Although a higher percentage of non-irrigator households had better water quality than irrigator households, the relationship is not statistically significant.

Table 2

Bivariate analysis showing the link between water sources, collection and stored water quality

Variables Water quality (%)
 
Chi-squared ([χ]2p-values 
Contaminated Uncontaminated 
Drinking water sources 
 Private protected wells 23 43.48 56.52 41.640 0.000 
 Shared protected well/spring 202 43.07 56.93   
 Unprotected wells/spring 176 72.16 27.84   
 Surface water 53 75.47 24.54   
Water collection container 
 Jerry cans 379 62.01 37.99 14.014 0.000 
 Clay vessel 75 38.01 61.33   
Water treatment practice 
 Yes 35 71.43 28.57 2.748 0.097 
 No 419 57.04 42.96   
Handwashing with soap 
 Yes 124 47.58 52.42 7.831 0.005 
 No 330 62.12 37.88   
Sanitation facility 
 Pit latrine 189 51.85 48.15 5.277 0.022 
 No facility (open field/bush) 265 62.54 37.36   
Garbage disposal 
 Dugout/burning 49 16.33 83.67 59.309 0.000 
 Throw away in the yard 245 71.43 28.57   
 Throw away outside the yard 59 42.37 57.63   
 Used as a fertilizer/compost 101 55.45 44.55   
Irrigated agriculture 
 Yes 302 58.94 41.06 0.232 0.630 
 No 152 56.58 43.42   
Variables Water quality (%)
 
Chi-squared ([χ]2p-values 
Contaminated Uncontaminated 
Drinking water sources 
 Private protected wells 23 43.48 56.52 41.640 0.000 
 Shared protected well/spring 202 43.07 56.93   
 Unprotected wells/spring 176 72.16 27.84   
 Surface water 53 75.47 24.54   
Water collection container 
 Jerry cans 379 62.01 37.99 14.014 0.000 
 Clay vessel 75 38.01 61.33   
Water treatment practice 
 Yes 35 71.43 28.57 2.748 0.097 
 No 419 57.04 42.96   
Handwashing with soap 
 Yes 124 47.58 52.42 7.831 0.005 
 No 330 62.12 37.88   
Sanitation facility 
 Pit latrine 189 51.85 48.15 5.277 0.022 
 No facility (open field/bush) 265 62.54 37.36   
Garbage disposal 
 Dugout/burning 49 16.33 83.67 59.309 0.000 
 Throw away in the yard 245 71.43 28.57   
 Throw away outside the yard 59 42.37 57.63   
 Used as a fertilizer/compost 101 55.45 44.55   
Irrigated agriculture 
 Yes 302 58.94 41.06 0.232 0.630 
 No 152 56.58 43.42   

Source: Authors’ estimate using survey data.

Regarding community water source quality, of the total 29 protected dug wells/springs water samples tested, 38% of the total samples were contaminated with E. coli. The mean level of E. coli was 6.83 (CFU/100 mL) water. It is worth mentioning that the water quality of community source samples does not necessarily reflect stored household water quality as some sampled households have their own private wells and some others get their drinking water from unprotected sources. Although this paper primarily focuses on stored household water quality, water source sample analysis is vital to understand the level of protected community water sources contamination. Our findings show that some protected community water sources are of unacceptable microbial quality for household consumption unless the water undergoes subsequent treatment and is made safer. The analysis of water quality of community water sources is, however, limited by inadequate inspection and risk assessment of the water source points, necessary to identify potential source of contamination.

Multivariate analysis

This section discusses the empirical results from the multivariate regression. The results of the OLS regressions are presented in Table 3 while the logistic estimated odds ratios are presented in Table 4. The OLS model was used to determine the factors associated with the natural logarithm of E. coli water quality measures, that is, the degree of fecal contamination. Given that drinking water is generally of poor quality among the sampled households, this approach allows us to investigate the incremental effects of the covariates on the level of E. coli concentration. On the other hand, the logistic regression was used to estimate the odds of unsafe water quality, that is, the odds of the binary outcome of potable or unpotable water (is equal to 1 if there is at least 1 or more E. coli CFU/100 mL water, and 0 otherwise). For both types of regression analysis, different model specifications were estimated and adjusted for covariant factors that affect the outcomes of interest.

Table 3

Estimates from OLS regression predicting the natural log of E. coli (N=454)

Variables Model 1 SE Model 2 SE Model 3 SE 
Primary drinking water sourcea       
 Unprotected well/spring 1.040*** 0.186 1.052*** 0.164 0.335** 0.152 
 Surface water 1.190*** 0.261 1.127*** 0.238 0.212 0.241 
Minutes to water source (1 = 30 min or less) −1.061*** 0.242 −0.981*** 0.225 −0.868*** 0.224 
Container (1 = jerry cans)a 1.197*** 0.198 1.146*** 0.191 1.048*** 0.177 
Highest education comp.   −0.081** 0.034 −0.034 0.026 
Household size   −0.002 0.068 −0.081 0.057 
Household density   0.398*** 0.082 0.310*** 0.067 
Number of adult females   −0.408** 0.178 −0.187 0.135 
Garbage disposal methodsa       
 Throw in the yard     1.480*** 0.230 
 Throw away outside the yard     0.559* 0.310 
 Used as fertilizer     0.860*** 0.245 
Handwashing with soap (dummy)     −0.547*** 0.161 
Log of assets value     −0.335*** 0.099 
Livestock holding     0.275*** 0.059 
Irrigating households (dummy)     0.532*** 0.134 
WUA (dummy)     −1.374*** 0.180 
Pit latrine (dummy)     −0.025 0.172 
Water source location (1 = on premises)     −0.496** 0.236 
Water source location X latrine     1.070*** 0.364 
Constant 1.520*** 0.281 0.970** 0.389 2.714*** 0.816 
R-squared 0.17  0.25  0.46  
Model F-stat 35.28  34.17  61.32  
Model P-value 0.000  0.000  0.000  
Variables Model 1 SE Model 2 SE Model 3 SE 
Primary drinking water sourcea       
 Unprotected well/spring 1.040*** 0.186 1.052*** 0.164 0.335** 0.152 
 Surface water 1.190*** 0.261 1.127*** 0.238 0.212 0.241 
Minutes to water source (1 = 30 min or less) −1.061*** 0.242 −0.981*** 0.225 −0.868*** 0.224 
Container (1 = jerry cans)a 1.197*** 0.198 1.146*** 0.191 1.048*** 0.177 
Highest education comp.   −0.081** 0.034 −0.034 0.026 
Household size   −0.002 0.068 −0.081 0.057 
Household density   0.398*** 0.082 0.310*** 0.067 
Number of adult females   −0.408** 0.178 −0.187 0.135 
Garbage disposal methodsa       
 Throw in the yard     1.480*** 0.230 
 Throw away outside the yard     0.559* 0.310 
 Used as fertilizer     0.860*** 0.245 
Handwashing with soap (dummy)     −0.547*** 0.161 
Log of assets value     −0.335*** 0.099 
Livestock holding     0.275*** 0.059 
Irrigating households (dummy)     0.532*** 0.134 
WUA (dummy)     −1.374*** 0.180 
Pit latrine (dummy)     −0.025 0.172 
Water source location (1 = on premises)     −0.496** 0.236 
Water source location X latrine     1.070*** 0.364 
Constant 1.520*** 0.281 0.970** 0.389 2.714*** 0.816 
R-squared 0.17  0.25  0.46  
Model F-stat 35.28  34.17  61.32  
Model P-value 0.000  0.000  0.000  

aOmitted reference categories are protected well/spring, clay vessel, and dugout/burning.

Robust standard errors adjusted for clustering at the village level.

Significance ***p < 0.01, **p < 0.05, *p < 0.1.

Table 4

Estimates from logistic regression predicting E. coli concentration (1 if E. coli >= 1, N = 454)

Variables Model 1 SE Model 2 SE Model 3 SE 
Primary drinking water sourcea       
 Unprotected well/spring 3.246*** 0.717 3.582*** 0.752 1.958** 0.550 
 Surface water 3.693*** 1.277 3.619*** 1.213 1.161 0.433 
Minutes to water source (1 = 30 min or less) 0.387*** 0.116 0.396*** 0.123 0.373** 0.160 
Container (1 = jerry cans)a 2.635*** 0.759 2.702*** 0.769 3.470*** 1.236 
Highest education completed   0.901** 0.037 0.909** 0.038 
Household size   1.031 0.080 0.904 0.085 
Household density   1.428*** 0.165 1.466*** 0.181 
Number of adult females   0.655* 0.160 0.758 0.213 
Garbage disposal methodsa       
 Throw in the yard     14.755*** 7.629 
 Throw away outside the yard     2.869* 1.765 
 Used as fertilizer     5.770*** 3.245 
Handwashing with soap (dummy)     0.394*** 0.120 
Log of assets value     0.734** 0.101 
Livestock holding     1.571*** 0.176 
Irrigating households (dummy)     1.640* 0.435 
WUA (dummy)     0.157*** 0.055 
Pit latrine (dummy)     0.981 0.250 
Constant 0.795 0.277 0.471 0.243 0.672 0.910 
Pseudo R-squared 0.10  0.15  0.35  
Model Chi2 55.79  105.46  170.34  
Model p-value 0.000  0.000  0.000  
Variables Model 1 SE Model 2 SE Model 3 SE 
Primary drinking water sourcea       
 Unprotected well/spring 3.246*** 0.717 3.582*** 0.752 1.958** 0.550 
 Surface water 3.693*** 1.277 3.619*** 1.213 1.161 0.433 
Minutes to water source (1 = 30 min or less) 0.387*** 0.116 0.396*** 0.123 0.373** 0.160 
Container (1 = jerry cans)a 2.635*** 0.759 2.702*** 0.769 3.470*** 1.236 
Highest education completed   0.901** 0.037 0.909** 0.038 
Household size   1.031 0.080 0.904 0.085 
Household density   1.428*** 0.165 1.466*** 0.181 
Number of adult females   0.655* 0.160 0.758 0.213 
Garbage disposal methodsa       
 Throw in the yard     14.755*** 7.629 
 Throw away outside the yard     2.869* 1.765 
 Used as fertilizer     5.770*** 3.245 
Handwashing with soap (dummy)     0.394*** 0.120 
Log of assets value     0.734** 0.101 
Livestock holding     1.571*** 0.176 
Irrigating households (dummy)     1.640* 0.435 
WUA (dummy)     0.157*** 0.055 
Pit latrine (dummy)     0.981 0.250 
Constant 0.795 0.277 0.471 0.243 0.672 0.910 
Pseudo R-squared 0.10  0.15  0.35  
Model Chi2 55.79  105.46  170.34  
Model p-value 0.000  0.000  0.000  

aOmitted reference categories are protected well/spring, clay vessel, and dugout/burning.

Robust standard errors adjusted for clustering at the village level.

Significance ***p < 0.01, **p < 0.05, *p < 0.1; and coefficients are odds ratio (OR).

The OLS regression results presented in Table 3 show that types of primary water sources influence stored water quality. Stored household water from protected wells/spring had lower E. coli compared to unprotected wells/springs and surface water sources – implying that water from unprotected wells/spring and surface water sources had significantly higher levels of E. coli than protected sources (model 1, Table 3). It is shown that simple spring protection significantly improves the microbial quality of both POS as well as POU water (Kremer et al. 2011). This association remains significant after further adjustment for household demographic characteristics. However, the pattern of relationship between water sources and E. coli level of stored water does not remain the same after controlling for sanitary characteristics. The result suggests that stored water from unprotected wells/spring had higher levels of E. coli than other alternative water sources (model 3, Table 3). Similarly, the results from the logistic regression estimates presented in Table 4 suggest that stored water from surface water is 3.7 times more likely to be contaminated with fecal materials compared to protected wells/springs; however, this difference disappears after controlling for sanitary factors (model 3, Table 4). On the other hand, water from unprotected sources is 2–3.6 times more likely to be contaminated than from protected sources.

The time to walk to a water source (a proxy for distance to water source including waiting time) is highly positively associated with the level of E. coli. Traveling long distances to collect water increases the risk of water being contaminated, and limited water availability hinders proper hygiene practices (for example, washing hands after defecation). This relationship remains strong after controlling for household demographic and sanitary characteristics. The association between household water collection container vessels and the level of E. coli is strong even after controlling for variations in households' socio-demographic and sanitary characteristics. Households who use jerry cans for water collection activities had higher E. coli levels than households using clay vessels. This could be due to inadequate cleaning of the jerry cans.

Regarding household demographic characteristics, the household level of education, as expected, negatively affects the level of E. coli in stored water. Furthermore, household density is strongly and positively associated with stored water levels of E. coli across all model specifications. On the other hand, the number of adult female household members positively influence stored water quality but the association vanishes after controlling for household sanitary conditions. Our study also indicates that households' methods of garbage disposal patterns are highly associated with stored water levels of E. coli.

In the third model specification, although the relationship between latrines and levels of E. coli is not statistically significant, it is positively associated with the level of E. coli when we introduce the interaction terms between latrine and water source location (Table 3) (the interaction term captures the simultaneous influence of both latrine and drinking water source location on stored water quality). This implies that availability of pit latrines may increase the risk of fecal contamination of stored water if a water source is located close to it. Our results also suggest that handwashing with soap and household assets (a proxy for wealth) are negatively associated with stored water levels of E. coli, and similar results are shown in all model specifications.

Most of the households practice mixed farming, and mostly livestock is living in close proximity to human beings, that is, keeping livestock generally creates more crowded living conditions. The negative relationship is expected and the effect size is relatively large. Households engaged in irrigated agriculture also had poor stored water quality. As irrigated agriculture has complex interactions with drinking water, household water can easily become degraded through irrigated agriculture practices or through multiple water use. The existence of a water user association (WUA) in the village is also robustly associated with better stored water quality. In most cases, the two regression tables produce similar results with expected signs across all model specifications. Finally, the r-squared for the OLS regression is modest for a cross-sectional study, and it ranges from 0.17 to 0.46 when we adjusted for socio-demographic and sanitary characteristics.

However, contrary to the OLS regression results presented in Table 3, some of the variables such as irrigation practice, water source location, and its interaction effect with pit latrines, which greatly influence stored water level of E. coli at different levels, do not have a statistically significant influence in the logistic regression results presented in Table 4. This indicates that these variables could be proximate causes for poor stored water quality as their effects depend on the level of E. coli (CFU/100 mL) concentration.

DISCUSSION

Our results generally indicate a common problem of poor stored water quality in the study areas with more than 58% of the households having at least one E. coli CFU/100 mL water. This result is not surprising when compared to earlier findings elsewhere in Ethiopia. For instance, a study in Kersa district of Eastern Ethiopia found that more than 78% of sampled households' stored water was contaminated with E. coli (Mengistie et al. 2013). Our results further suggest that stored water quality is strongly associated with water source, water collection time and types of containers, presence of WUA in the village, household demographic structures, and households’ overall sanitary characteristics. Our findings are also consistent with other studies that demonstrate substantial levels of fecal contamination of stored water after collection from improved sources that are less prone to high level of fecal contamination (Clasen & Bastable 2003; Wright et al. 2004).

In the bivariate analysis, the influence of household water treatment practice (such as boiling, addition of bleach or filtration through layers of material) on stored water quality is not strong (Table 2). The weak relationship between household water treatment practice and stored water quality is likely due to the lack of regular use of any form of water treatment in our sampled households. For instance, among the households who use some form of water treatment, more than 80% of these households are applying chlorine-based methods, of which 72% households used this method during the month before the survey. The empirical evidence that household water treatment and safe storage practice in improving the microbiological quality of drinking water is well-documented (for example, see Mintz et al. 1995; Mengistie et al. 2013; Clasen 2015). We also observed that there is a lack of awareness about domestic water quality and its health consequences: people often perceive that clean water is ‘clean’ as long as it is not turbid. About 87% of urban households and 91% of rural households do not practice any form of household water treatment in the country (CSA & ICF International 2012).

The types of household storage container can also influence household water quality (Levy et al. 2008; Günther & Schipper 2013). In this study, types of water collection containers are significantly associated with the quality of water consumed by the household. More than 83% of the households identified jerry cans as their preferred container for hauling and storing their drinking water, and only 24% of households had separate water storage containers. Households opt to store water for future use when the water supply is unreliable and intermittent; however, drinking water contamination will also be higher if water is stored for a longer period (Brick et al. 2004). Our result shows that jerry cans increases the risk of stored water contamination and this could be due to inadequate cleaning. Although jerry can containers have the advantage of being narrow-mouthed, rural households do not properly clean it. It is difficult to clean inside with simple washing. Previous studies elsewhere showing that storage container characteristics, such as narrow versus wide mouth and covered versus uncovered, are key factors in determining stored water quality (Mintz et al. 1995). It is argued that water pouring is safer than dipping, but this research also questioned whether narrow-necked containers such as jerry cans are the safest method of water storage (Mintz et al. 1995; Ogutu et al. 2001).

Our results also highlighted that increased water collection time increases the risk of stored water contamination. This is in line with studies showing that the microbiological quality of household water obtained from sources with initially acceptable quality significantly deteriorates during collection and transportation due to unhygienic storage and handling practices such as dipping hands and/or receptacles into the water and uncovered storage containers (Clasen & Bastable 2003; Wright et al. 2004). Moreover, water collection time determines the quantity of water a given household can collect and consume (Cairncross 1987), which is a critical determinant of key hygiene practices (Cairncross 1997; Curtis et al. 2000). On the other hand, more time allocation for household water collection may allow households to collect sufficient water and to maintain key hygiene practices such as washing hands at critical times, which can influence stored water quality (Curtis et al. 2000).

Household demographic variables, particularly household density, are strong predictors of stored water quality. It can be argued that crowded living conditions may influence the overall hygiene and sanitation environment that probably increases the risk of stored water contamination. It is also a common understanding that the level of E. coli in stored water is expected to positively correlate to household size due to increased chance of contact with householders’ hands, but the effect of this variable turns out to be statistically insignificant. Higher household education is also expected to correlate with better understanding of water quality and sanitary behaviors, which in turn could influence household water quality through improved water handling and hygiene practice. However, our results show that the effect of education is small, and even become statistically insignificant (model 3, Table 3). This could be explained by the low level of school attainment, and possibly that the primary caregiver's level of education may be more important in determining stored water quality than any other household members.

On the other hand, pit latrine availability increases the level of E. coli on stored water for households who use well water sources in their own premises. As a standard practice, the WHO recommended a distance of 30 meters between water points and latrines. Megha et al. (2015) showed that the microbiological quality of ground water deteriorates where pit latrines are placed close to the source. In support of this, our result shows that households having a pit latrine and using own wells located in premises have high levels of E. coli in stored water. In addition to the type of well, the risk of water quality problems with groundwater supplies is directly related to how close it is to potential sources of contamination. The site of private wells, therefore, should be chosen carefully to minimize the risk exposure from external contamination as its location determines the water quality. Furthermore, as many of the private-well water sources in the study areas are bucket wells, and they are often shallow and inadequately protected, this might increase risks of contamination from animal feces, flood-washed wastes, dirty well surroundings, and water-drawing buckets. Although the relationship is not statistically significant in the logistic regression, source water contamination from households’ own latrines could be one possible source of the high E. coli levels present in stored household water. Our findings also suggest that households' methods of garbage disposal patterns are strongly correlated with stored water quality. Piped water from private or public systems generally has fewer pathogens than surface or well water, which are affected by the drainage of human, animal and other wastes, particularly when sanitary waste disposal systems are lacking or poorly maintained. If waste is not disposed of properly, it can contaminate surface and well water sources.

In rural areas, agriculture and livestock rearing, which are the primary sources of livelihood, have complex interactions with household water quality. Most households keep livestock, and often livestock live together with human beings, increasing the risk of household water contamination. Households engaged in irrigated agriculture have lower water quality. Irrigation often provides multiple-use water such as drinking, cooking, and bathing in addition to its prime use for crop production. Where access to improved drinking water is limited, households opt to use irrigation water for domestic purposes, which is often of poor quality. A significant portion of households reported that they directly withdraw water from irrigation water sources for household consumption. Although irrigation water increases water availability for domestic purposes, it could potentially increase the risk of contamination of stored waters. Similarly, household members working in close contact with irrigation waters may come into contact with domestic water and contaminate it if proper personal hygiene and handwashing are lacking.

We also expect households with more accumulated assets (wealth) to live in a more sanitary environment, which can reduce drinking water contamination. Livestock ownership may offset the net positive gains of household assets on stored water quality, as livestock can be a source of pathogens, which directly affect the water quality. Moreover, household assets are moderately correlated with livestock ownership, irrigation practices, and household education level. This suggests that a combination of variables can influence water quality indirectly through various pathways.

Another interesting finding is that the existence of WUA in a village appears to possibly influence the quality of stored water. Households that belong to a village where there is a WUA reported better stored water quality. This association is consistently statistically significant across all model specifications. The WUA is primarily responsible for monitoring and supervising community sources and handling conflicts among household users of community water sources. The influence of WUA on stored water quality could be via improving the protection of water sources from external contamination. However, some community water sources considered to be ‘improved’ and believed to provide safe water showed the presence of E. coli, which is not in compliance with the WHO guideline standards.

In summary, this study provides evidence that beyond POS water quality, POU water contamination is a critical issue and could pose significant health problems. It is important to take proactive measures to ensure that the water supply chain as a whole can deliver acceptable quality and quantity of water to meet domestic needs. As part of supplying safe drinking water, the government has implemented the water safety plan (WSP) approach, which is recommended by the WHO. The WSP is a comprehensive risk assessment and management approach to reduce drinking water contamination effectively in all stages of the water supply system (WHO 2012). Furthermore, the government took a new approach by launching the ‘ONE WASH program’ in 2013, which brings together Ministries of Water Resources, Education, Health, and Finance and Economic Development, as well as development partners to improve the provision of water and sanitation services. The ONE WASH national program is paving the way for greater cooperation between sectors to address the problem of access to improved water and sanitation in the country.

CONCLUSIONS

Currently, many rural householders still have to travel long distances in search of improved drinking water sources. Lack of access to clean and adequate drinking water and poor sanitary environment is a critical public health problem in Ethiopia, contributing to about 70% of the diarrheal diseases burden in the country (Federal Ministry of Health 2005). Using primary household survey data and microbiological water testing for E. coli, this paper aims at assessing key drivers of stored drinking water quality in rural Ethiopia.

The study suggests that stored water quality is strongly associated with water source, water collection time and types of containers, household demographic structures, and households’ overall sanitary characteristics. The results show that households' stored water collected from unprotected wells/springs had higher levels of E. coli than other drinking water sources. Distance to water sources and water collection with jerry cans are also associated with poorer stored water quality. Moreover, the quality of drinking water is affected by agricultural practices involving irrigation and/or livestock rearing.

The study suggests that there is a need to promote water quality at both the POS and POU to advance the Sustainable Development Goals (SDG6) of ensuring access to clean water for everyone. In addition to expanding the water supply infrastructures, available water source points should be adequately protected, and ad hoc water quality testing and quality control mechanisms need to be in place to ensure safety of rural water supply. Promoting household water treatment practices to make water safer would also be a worthy intervention to improve drinking water quality, given that most households draw their drinking water from unprotected sources. Moreover, providing safer and convenient storage containers or promoting how to clean jerry cans properly would avoid substantial risk of water contamination. The provision of drinking water through community water schemes is the only conventional way to increase access to clean water supply in many rural areas, therefore, building the capacity of WUA (such as training in water source protection and environmental sanitation) is critical in the provision of safe water supply. As the relationship between improved rural water supply and safe stored water is generally complex, a mix of instruments is needed to address the problem of drinking water safety and to make progressive improvements in the SDG6 in the next decade.

One possible drawback of this study is that we use only a one-time water sample test results. This does not allow us to capture the seasonal impacts on water quality. As our sampled households entirely rely on non-piped water sources, seasonal changes could likely affect water quality in the household, which may have also influence the level of water quality measured. Conducting subsequent water sample testing over time could provide a more representative water-quality indicator.

ACKNOWLEDGEMENTS

We are grateful to Joachim von Braun for his valuable comments and suggestions. We would also like to acknowledge the support from the Center for Development Research (ZEF), the Ethiopian Economics Association (EEA), the Welthungerhilfe and the Organization for Rehabilitation and Development in Amhara (ORDA) during our field research in Ethiopia. Financial support by the Bill & Melinda Gates Foundation is gratefully acknowledged. Lastly, we thank the anonymous reviewers who helped improve the text substantially. Any remaining mistakes and inconsistencies are the responsibility of the authors.

REFERENCES

REFERENCES
Amenu
,
K.
,
Spengler
,
M.
,
André
,
M.
&
Valle Zárate
,
A.
2014
Microbial quality of water in rural households of Ethiopia: implications for milk safety and public health
.
J. Health Popul. Nutr.
32
(
2
),
190
197
.
Bain
,
R.
,
Cronk
,
R.
,
Wright
,
J.
,
Yang
,
H.
,
Slaymaker
,
T.
&
Bartram
,
J.
2014
Fecal contamination of drinking-water in low- and middle-income countries: a systematic review and meta-analysis
.
PLoS Med.
11
(
5
),
e1001644
.
Brick
,
T.
,
Primrose
,
B.
,
Chandrasekhar
,
R.
,
Roy
,
S.
,
Muliyil
,
J.
&
Kang
,
G.
2004
Water contamination in urban south India: household storage practices and their implications for water safety and enteric infections
.
Int. J. Hyg. Environ. Health
207
(
5
),
473
480
.
Butterworth
,
J.
,
Sutton
,
S.
&
Mekonta
,
L.
2013
Self-supply as a complementary water services delivery model in Ethiopia
.
Water Alternat.
6
(
3
),
405
.
Cairncross
,
S.
1987
Water, women and children
. In:
The Benefit of Water Supply: Developing World Water II
(
Pickford
,
J.
&
Leedham
,
B.
, eds).
Grosvener
,
London
.
Cairncross
,
S.
1997
More water: better health
.
People Planet
6
(
3
),
10
11
.
Central Statistical Agency (CSA)
2013
Population Projection of Ethiopia for all Regions at Wereda Level From 2014–2017
.
CSA
,
Addis Ababa
,
Ethiopia
.
Central Statistical Agency & ICF International
2012
Ethiopia Demographic and Health Survey 2011
.
Central Statistical Agency (CSA) and ICF International
,
Addis Ababa
,
Ethiopia
and Calverton, Maryland, USA
.
Clasen
,
T.
&
Bastable
,
A.
2003
Faecal contamination of drinking water during collection and household storage: the need to extend protection to the point of use
.
J. Water Health
1
,
109
115
.
Curtis
,
V.
,
Cairncross
,
S.
&
Yonli
,
R.
2000
Review: domestic hygiene and diarrhoea – pinpointing the problem
.
Trop. Med. Int. Health
5
(
1
),
22
32
.
Federal Ministry of Health
2005
National Hygiene and Sanitation Strategy to Enable 100% Adoption of Improved Hygiene and Sanitation
.
Federal Democratic Republic of Ethiopia Ministry of Health (FMoH)
,
Addis Ababa
,
Ethiopia
.
Godfrey
,
S.
,
Labhasetwar
,
P.
,
Wate
,
S.
&
Pimpalkar
,
S.
2011
How safe are the global water coverage figures? Case study from Madhya Pradesh, India
.
Environ. Monit. Assess.
176
(
1–4
),
561
574
.
Kremer
,
M.
,
Leino
,
J.
,
Miguel
,
E.
&
Zwane
,
A. P.
2011
Spring cleaning: rural water impacts, valuation and property rights institutions
.
Q. J. Econ.
126
(
1
),
145
205
.
Levy
,
K.
,
Nelson
,
K. L.
,
Hubbard
,
A.
&
Eisenberg
,
J. N.
2008
Following the water: a controlled study of drinking water storage in northern coastal Ecuador
.
Environ. Health Perspect.
116
(
11
),
1533
1540
.
MacKinnon
,
J. G.
&
Magee
,
L.
1990
Transforming the dependent variable in regression models
.
Int. Econ. Rev.
31
(
2
),
315
339
.
Megha
,
P. U.
,
Kavya
,
P.
,
Murugan
,
S.
&
Harikumar
,
P. S.
2015
Sanitation mapping of groundwater contamination in a rural village of India
.
J. Environ. Protect.
6
(
1
),
11
.
Mintz
,
E. D.
,
Reiff
,
F. M.
&
Tauxe
,
R. V.
1995
Safe water treatment and storage in the home: a practical new strategy to prevent waterborne disease
.
J. Am. Med. Assoc.
273
(
12
),
948
953
.
Ogutu
,
P.
,
Garrett
,
V.
,
Barasa
,
P.
,
Ombeki
,
S.
,
Mwaki
,
A.
&
Quick
,
R. E.
2001
Seeking safe storage: a comparison of drinking water quality in clay and plastic vessels
.
Am. J. Public Health
91
(
10
),
1610
1611
.
Scheelbeek
,
P.
2005
Two Containers a Day: The Search for Proper Water Sources in Eastern Ethiopia. Water Quality and Sanitation in the Lege Dini Watershed Area, Ethiopia
.
MSc Thesis in Irrigation and Water Engineering
,
Wageningen University
,
The Netherlands
.
Sutton
,
S.
,
Mamo
,
E.
,
Butterworth
,
J.
&
Dimtse
,
D.
2011
Towards the Ethiopian Goal of Universal Access to Rural Water: Understanding the Potential Contribution of Self Supply: RiPPLE Working Paper No. 23
.
WaterAid Ethiopia
,
Addis Ababa
,
Ethiopia
.
WHO
2011
Guideline for Drinking-Water Quality
.
4th edn
.
World Health Organization
,
Geneva
,
Switzerland
.
WHO
2012
Water Safety Planning for Small Community Water Supplies: Step-by-Step Risk Management Guidance for Drinking-Water Supplies in Small Communities
.
World Health Organization (WHO)
,
Geneva
,
Switzerland
.
WHO/UNICEF
2014
Progress on Drinking Water and Sanitation: 2014 Update
.
United Nations Children's Fund (UNICEF)
,
New York, NY
,
USA
and
World Health Organization (WHO), Geneva, Switzerland
.
WHO/UNICEF
2015
Progress on Sanitation and Drinking Water: 2015 Update and MDG Assessment
.
United Nations Children's Fund (UNICEF)
,
New York, NY
,
USA
and
World Health Organization (WHO), Geneva, Switzerland
.