Fecal contamination of drinking water in Kericho District, Western Kenya: role of source and household water handling and hygiene practices

The importance of water to human health is encapsulated in the Human Right to Water and Sanitation, which entitles everyone to ‘sufficient, safe, physically accessible and affordable water for personal and domestic uses’ (Committee on Economic, Social and Cultural Rights 2002; United Nations 2010). Unfortunately, over 275 million people in sub-Saharan Africa rely on unsafe drinking water sources from lakes, rivers, and open wells (WHO/UNICEF 2010). Consequently, in this region, many waterborne-related diarrheal diseases are responsible for significant morbidity and mortality among children, the elderly, and immunosuppressed individuals (Kariuki et al. 2006; WWAP 2006). The most common health problems associated with unsafe drinking water in Kenya include typhoid fever, cholera, diarrhea, dysentery, worms, and bilharzias (WWAP 2006).

The World Health Organization (WHO) guidelines for drinking water quality include criteria for assessing health risks and setting targets for improving water safety (WHO 2011). The guidelines recommend using either thermotolerant coliforms (TTCs) or Escherichiacoli in assessing fecal contamination of drinking water (WHO/UNICEF 2010; WHO 2011). A 100 mL water sample with <1 indicator organisms is considered ‘very low risk’; 1–10, ‘low risk’; 10–100, ‘medium risk’; >100, ‘high risk’ or ‘very high risk’ (WHO 1997).

Studies have given varied total and fecal (E. coli) coliform contamination of water samples in different settings. Over a quarter of samples from improved water sources in China, the United Kingdom, France, Portugal and in selected low and middle income countries were shown to contain fecal contamination (Bain et al. 2014). About 95% of water sources (dams, rivers, springs, and wells) in the informal settlements of Kisumu, Kenya had significant E. coli contamination (Opisa et al. 2012). In another informal settlement in Nairobi, Kenya, Chemuliti et al. (2002) identified 35% of standpipes and 95% of in-house storage containers as being contaminated with E. coli coliforms. Water source, storage practices, locality, poverty, hygiene, sanitary and environmental factors have been cited as sources for fecal water contamination (Gundry et al. 2006).

Currently, data are skewed on the quality of drinking water in Kericho District, Western Kenya. This district is faced with high population density and a growth rate of 3.6%, while the majority of urban and rural populations lack clean and safe piped water supply (Kenya Census 2010).

Kimani-Murage & Ngindu (2007) showed that 13% of the households in Kenya had their drinking water contaminated with thermotolerant E. coli. Applying the formula for estimating the population proportion with specified relative precision described by Lemeshow et al. (1990), setting the α at 0.05, and a detection rate of 20%, a total of 103 households were recruited to achieve 0.90 power. A two-stage sampling method was then used to select these households. First, a complete list of all the households in the 17 locations/villages in Kericho District was generated based on the Kenya Census of 2010. A total of 12 locations were then selected based on probability proportionate to size. Second, a systematic sampling method was then used to select every fifth household in each pre-selected location. The number of households per location varied based on their total household size.

Consent was gained from the heads of these households, who were interviewed using a structured questionnaire to gather information relating to the quality of drinking water. Further, two different water samples were collected from these households: one from source (tap, rivers, or wells) and the second from household drinking water storage vessels. This study was done between 2013 and 2014 and was approved by the Ethical Review Committee of Kenya Medical Research Institute (SSC No. 2579 on 18 July, 2013). This research has adhered to the STROBE guidelines for observational studies as outlined at: http://www.strobe-statement.org.

Structured questionnaires adopted from the WHO/UNICEF Joint Monitoring Programme (http://www.wssinfo.org/fileadmin/user_upload/resources/1268174016-JMP_Core_Questions.pdf) were used to gather information related to water-extraction patterns, type of water transport, water-treatment methods, hygiene and sanitation related issues.

Water samples were collected aseptically using sterile sampling containers. About 100 mL of the water sample was collected and immediately analyzed for bacteriological qualities and physical chemical properties (pH, temperature, turbidity, and free chlorine) on site using a DelAgua water testing kit. The cultured plates were then transported to the Micro Hub Kericho Walter Reed Project (WRP) of Kenya Medical Research Institute for further bacteriological analysis. Water source sampling (rivers, streams, or other surface waters in the district where enrolled households collect their waters) involved drawing water from 30 cm below the surface. Sampling from wells and boreholes involved drawing water using a bucket and taking 100 mL into a sterile container. This was considered to be more representative of what is actually being consumed by the household. Sampling water from taps was done directly into collecting containers (WHO 1998). Household water sampling was done by requesting the household head to draw water from their storage reservoir.

The E. coli pathotypes were determined using multiplex polymerase chain reaction (PCR) as described by Gomez-Duarte (2009). The first PCR of this multiplex contained M1 primers for amplification of eae, bfpA, VT, and aggR genes for identification of shiga toxin-producing E. coli (STEC), enteropathogenic E. coli (EPEC), and enteroaggregative E. coli (EAEC) pathotypes. The second PCR contained M2 primers for amplification of LT, ST, daaE, ipaH, and virF gene targets for identification of enterotoxigenic E. coli (ETEC), diarrheagenic E. coli (DAEC), and enteroinvasive E. coli (EIEC) pathotypes. About 1 μL of genomic DNA was mixed with 24 μL of a premade mix containing primers at a 0.2 μM final concentration and Platinum Blue PCR SuperMix polymerase (Invitrogen, Carlsbad, CA, USA). The PCR amplification consisted of 2 min at 94 °C denaturing temperature, followed by 40 cycles of 30 sec at 92 °C denaturing temperature, 30 sec at 59 °C annealing temperature, and 30 sec at 72 °C extension temperature. The PCR products were visualized and recorded under ultraviolet light using a 2% agarose ethidium bromide-stained gel. The verotoxin of STEC isolates were analyzed using PCR specific for shiga-like toxin 1 (VT1) (Vidal et al. 2005) and shiga-like toxin 2 (VT2) (Nguyen et al. 2005).

Antimicrobial susceptibility patterns of E. coli pathotypes were undertaken using the disk diffusion technique of Kirby-Bauer. The susceptibility was checked against the following antibiotics: ampicillin, nalidixic acid, chloramphenicol, tetracycline, cefotaxime, cotrimoxazole, ceftazidime, and ciprofloxacin. The zones' inhibition was interpreted using standards for antimicrobial susceptibility testing (CLSI 2010).

Frequency (%), mean, standard deviation, and medium (interquartile ranges at 25% and 27%) were used to describe the qualitative and laboratory parameters. Chi-square or Fisher's exact test were used to test for significance where applicable. In bivariate analyses, odds ratios (OR) and 95% confidence intervals (CI) for the association between pathogenic thermotolerant E. coli contaminating water and socio-demographic, hygienic and environmental characteristics were calculated using Poisson regression. In multivariate analyses, a manual backward elimination approach was used to reach the most parsimonious model, including factors that were associated with contamination with pathogenic thermotolerant E. coli at the significance level of P ≤ 0.05. All statistical analyses were performed using STATA v 13 (StataCorp LP, College Station, TX, USA).

Table 1 shows the summary of characteristics of households recruited in this study. The mean age of the 103 household heads was 21.59 years (range 18–29 years) with slightly under a quarter (42.7%) aged 21–30 years. About 68.9% of households were located in rural areas, with 95.1% of them being female headed. About 57.3% had primary level education, 27.2% were farmers, while 54.4% had children below 5 years of age. Close to 36.9% of the households sourced their drinking water from rivers. Nearly all of them (91.3%) did not treat their drinking water, with 81.6% stating that their drinking water was clean regardless of the source. Most of the households (64.1%) stored the drinking water for more than 2 days, with the majority (83.5%) of them drawing drinking water by dipping the cup into the water storage container. Nearly half of the households reported having contact with drinking water during drawing. Some 60.2% used clay pots for water storage, with 64.1% of them never washing these storage containers.

Nearly half of the households (49.5%) used pit latrines or open pits for their toilet facility. About 32% of these toilets were moderately clean, with flies but no visible fecal matter. About 31.1% of households disposed of their children's fecal waste in the pit latrine. Slightly over half of them (59.2%) did not wash their hands before drawing water. The majority (96.1%) often rinsed water drawing utensils. About 88.3% were aware of waterborne diseases while only 30% had suffered from diarrhea, vomiting, and fever in the past 3 months.

Table 2 shows a summary of total and TTC levels in source and household drinking water. The mean total coliform count in water sources was 1,029.7 (SD 1,574.1 cfu/100 mL) and a range of 5,595 (80–5,675 cfu/100 mL). Close to half (50; 48.5%) of the water sources were contaminated (had total coliforms above >10 cfu/100 mL) (P = 0.768).

The mean household drinking water total coliform count was 529.1 (SD 857.5 cfu/100 mL) with a range of 5,462 (98–5,560 cfu/100 mL). The majority (85; 83.5%) of household drinking waters were contaminated, with total coliforms above >10 cfu/100 mL (P < 0.001).

The mean TTC count in the water sources was 798.6 (SD 1,302.6 cfu/100 mL) and a range of 3,539 (1–3,540 cfu/100 mL). Only 23 of the 103 (22.3%) water sources were contaminated with TTC (>10 cfu/100 mL) (P < 0.001).

The mean TTC count in households' drinking water was 321.2 (SD 761.5 cfu/100 mL) with a range of 3,799 (1–3,800 cfu/100 mL). Slightly below half (48) of the 103 (46.6%) households' drinking water was contaminated with TTC levels of >10 cfu/100 mL (P < 0.001).

In all 48 (46.6%) households with TTC, each had thermotolerant E. coli as well. Among the 48 thermotolerant E. coli, 5 (10.4%) were pathogenic E. coli: 2/5 (40%) EAEC, 2/5 (40%) ETEC, and 1/5 (20%) EPEC. There were 35 other types of TTCs isolated from household drinking water. These included 8/35 (22.8%) Serratia, 7/35 (20%) Enterobacter, 5/35 (14.3%) Klebsiella, 5/35 (14.3%) Moraxella, 4/35 (11.4%) Pseudomonas, 2/35 (5.7%) Shigella, 2/35 (5.7%) Acinetobacter, 1/35 (2.9%) Aeronomonas, and 1/35 (2.9%) Yersinia.

All the five pathogenic E. coli were multidrug resistant to four commonly prescribed antibiotics in Kenya, including sulfamethoxazole/trimethoprim, ampicillin, tetracycline and ampicillin/sulbactam.

Table 3 summarizes the factors associated with TTC contamination (>10 cfu/100 mL) of household drinking water. In the bivariate analyses, drinking water of households located in the rural set-up were more likely to have TTC contamination than households located in the urban areas (OR 2.01, 95% CI 1.09–4.12). Households with a piped water supply or water from a municipal source were less likely to have TTC contamination than those households whose source of water was from a river/spring (OR 0.38, 95% CI 0.16–0.91). Households which reported hand contact with drinking water during withdrawal were more likely to be contaminated with TTC compared to households which had no hand contact during water withdrawal (OR 1.11, 95% CI 1.11–3.39). Households that washed their water storage containers regularly were less likely to be contaminated with TTC than those households who did not wash the water storage container (OR 0.58, 95% CI 0.31–0.99). Households who practiced hand washing before drawing drinking water were less likely to have TTC contamination than those households that did not wash hands (OR 0.33, 95% CI 0.15–0.67). Households whose total coliform count was less than 10 cfu/100 mL were less likely to have TTC drinking water contamination than those households with more than 10 cfu/100 mL (OR 0.45, 95% CI 0.26–0.81). Further, households whose main water source temperatures were between 15 and 20 °C were less likely to have TTC drinking water contamination than those households with water source temperatures greater than 25.1 °C (OR 0.39, 95% CI 0.16–0.96). Lastly, households whose main water source had a free chlorine concentration of less than 1 mg/L were less likely to have their drinking water contaminated with TTC compared to households with a water source free chlorine concentration of above 1 mg/L (OR 0.41, 95% CI 0.24–0.71).

The following variables, which have been linked with drinking water contamination, were included in the multivariate model: socio-demographic characteristics (e.g., household locality, household headship, gender, age, educational level, and occupation), attitude and knowledge, water handling (e.g., water collection, treatment, storage, and handling practices) and sanitation (e.g., waste disposal and pollution alongside water source). None was found to be independently associated with TTC contamination of household drinking water.

Kenya has experienced recurrent cases of waterborne diseases like cholera (TDN 2007; Wambua 2008). In the last 10 years, the morbidity patterns ranked diarrhea as the fourth priority disease (HSSR 2005). Western Kenya and part of Rift Valley Province, notably the districts of Bungoma, Busia, Kakamega, and Kericho are among the most affected regions (Onyango & Angienda 2010). This study was a build up of the limited data in Kericho District, evaluating the quality of household drinking water and highlighting the role of source and household water handling and hygiene practices in the contamination of drinking water with TTC. Invariably, the extent of safe water handling was determined by the local people's knowledge and attitudes towards water safety and sanitation (Özdemir et al. 2011).

The current study detected significantly high (46.6%) numbers of household drinking water contaminated with TTC. The fecal contamination of water sources and a further deterioration between the collection points and homes has been observed in other studies. In Kibera, Nairobi, Chemuliti et al. (2002) observed higher contamination levels of fecal coliforms (95%) in in-house sources and (35%) of outdoor sources. In Western Kenya, Muruka et al. (2012) observed a lower (2%) source water fecal coliform contamination. In Masaba, Kisii, Nyagwencha et al. (2012) observed that 16% of the households accessed unsuitable water for human consumption. In Chinese, English, French, Portuguese, and Spanish populations over a quarter of samples from improved sources were shown to contain fecal contamination (Bain et al. 2014). In Peru, Gil et al. (2014) found TTCs in 48% of all water samples. In India, Boisson et al. (2013) found that 20% of the household drinking water had fecal contamination. Our study shows that a significant population in Kericho District, Kenya lacks safe clean drinking water according to the WHO guidelines.

The presence of TTCs in water indicates actual contamination with feces (human and non-human) and potential contamination by disease-causing pathogens of all kinds. In this study, pathogenic bacteria including E. coli, Serratia, Enterobacter, Klebsiella, Moraxella, Pseudomonas, Shigella, Acinetobacter, Aeromonas, and Yersinia were isolated. This is worrying given that these bacteria are associated with waterborne human diseases. Similar coliforms have been isolated by others from water sources (Kämpfer et al.2008). All of the thermotolerant pathogenic E. coli were resistant to various important antibiotics including ampicillin, sulfamethoxazole/trimethoprim, and tetracycline. In Nigeria, resistance patterns of 80.9% ampicillin, 95.4% tetracycline, and 46.5% chloramphenicol were observed; while in Tanzania, resistance rates of 83.1% to ampicillin, 57% chloramphenicol, 87.7% tetracycline, and 90.8% co-trimoxazole were found (Vila et al. 2000). Of major concern with our findings is that antibiotics such as ciprofloxacin and cefotaxime, which are prescription antibiotics in Kenya, showed considerable resistance. This indicates misuse of these and other classes of antibiotics which would have major implications in the treatment of E. coli causing diarrhea in Kenya.

The rural locality of the household and the type of water source contributed significantly to TTC household drinking water contamination. These findings mirror those of Bain et al. (2014), who showed water sources in low-income countries and rural areas as being more likely to be contaminated. As expected, households located in rural areas are prone to poor waste disposal, poor household hygiene, lack of resources, and knowledge of water treatment. These factors were likely to contribute to the higher TTC drinking water contamination. This argument is supported by the study of Bain et al. (2014), who observed that a defective water delivery system and inadequate environmental sanitation are shown to be a potential source of contamination for household drinking water. The poor levels of environmental hygiene coupled with a dilapidated water delivery system are major contributors to TTC water contamination (Chemuliti et al. 2002).

Household hygiene, and practices such as hand washing, hand contact with drinking water during withdrawal, household water storage type, cleanliness, water treatment, and human waste disposal were important for TTC household drinking water contamination. Eshcol et al. (2009), in India, showed that fecal contamination occurs principally during storage due to poor water handling. In Zimbabwe, homes where traditional drinking water containers have been replaced with covered, narrow-mouthed pots with a tap outlet have significantly less contamination (Mazengia et al. 2002). A combination of special storage vessels with point-of-use treatment has been shown to be very effective (Rose et al. 2006). Hand washing initiatives and the introduction of point-of-use disinfection can reduce diarrheal incidence (Ejemot et al. 2008).

Other water quality surveys have shown natural ecosystems (geological, topographical, hydrological in the drainage basin), seasonal differences, weather conditions, water levels as well as other human hygiene and practices affect the quality of water. The cross-sectional nature of this study, relatively small sample size, inadequate assessment using the full WHO/JMP standardized questionnaires for household water handling and sanitation, could partly explain the observed lack of association between household drinking water TTC contaminations in our multivariate analyses.

Given these limitations, a reasonable conclusion that can be drawn from these data is that a significant number of household drinking waters in Kericho District are unsafe in line with WHO guidelines, and some are contaminated with pathogenic multidrug-resistant E. coli. Water source and household hygiene and practices contribute significantly to the quality of drinking water.

This work was carried out at the Micro Hub Kericho Walter Reed Project (WRP), Kenya Medical Research Institute, Kericho District, Western Kenya. The work was funded by Field Epidemiology Laboratory Training Programme, Ministry of Health, Nairobi, Kenya. We thank the study participants for their invaluable support by consenting to the use of their samples in the study. This work was part of Master of public health degree for J.K.T. The work is published with the permission of Director KEMRI. The authors declare that they have no competing interest. J.K.T., W.S., Z.N., and M.O.N. conceived the study. J.K.T. supervised sample collections and testing while W.S. and M.O.N. supervised laboratory analysis. M.O.N. and J.K.T. analyzed the data and prepared the draft manuscript. W.S., Z.N., and M.O.N. provided guidance and mentorship during the implementation of the study. All authors reviewed and approved the final manuscript.