The objective of this study was to investigate the quality of on-plot piped water and rainwater at the point of consumption in an area with rapidly expanding coverage of ‘improved’ water sources.
The objective of this study was to investigate the quality of on-plot piped water and rainwater at the point of consumption in an area with rapidly expanding coverage of ‘improved’ water sources.
Cross-sectional study of 914 peri-urban households in Kandal Province, Cambodia, between July–August 2011. We collected data from all households on water management, drinking water quality and factors potentially related to post-collection water contamination. Drinking water samples were taken directly from a subsample of household taps (n = 143), stored tap water (n = 124), other stored water (n = 92) and treated stored water (n = 79) for basic water quality analysis for Escherichia coli and other parameters.
Household drinking water management was complex, with different sources used at any given time and across seasons. Rainwater was the most commonly used drinking water source. Households mixed different water sources in storage containers, including ‘improved’ with ‘unimproved’ sources. Piped water from taps deteriorated during storage (P < 0.0005), from 520 cfu/100 ml (coefficient of variation, CV: 5.7) E. coli to 1100 cfu/100 ml (CV: 3.4). Stored non-piped water (primarily rainwater) had a mean E. coli count of 1500 cfu/100 ml (CV: 4.1), not significantly different from stored piped water (P = 0.20). Microbial contamination of stored water was significantly associated with observed storage and handling practices, including dipping hands or receptacles in water (P < 0.005), and having an uncovered storage container (P = 0.052).
The microbial quality of ‘improved’ water sources in our study area was not maintained at the point of consumption, possibly due to a combination of mixing water sources at the household level, unsafe storage and handling practices, and inadequately treated piped-to-plot water. These results have implications for refining international targets for safe drinking water access as well as the assumptions underlying global burden of disease estimates, which posit that ‘improved’ sources pose minimal risks of diarrhoeal diseases.
Analyser la qualité de l'eau provenant de robinets dans les concessions et l'eau de pluie aux lieux de la consommation dans une zone en pleine expansion avec une couverture en sources d'eau «améliorées».
Une étude transversale portant sur 914 ménages périurbains dans la province de Kandal, au Cambodge, entre juillet et août 2011. Nous avons collecté des données dans tous les ménages sur la gestion de l'eau, la qualité de l'eau de boisson et les facteurs potentiellement liés à la contamination après la collecte de l'eau. Des échantillons d'eau potable ont été collectés directement à partir d'un sous-échantillon de robinets des ménages (n = 143), de l'eau du robinet stockée (n = 124), de l'eau stockée provenant d'autres sources (n = 92) et de l'eau stockée traitée (n = 79), pour l'analyse de base de la qualité de l'eau pour E. coli et d'autres paramètres.
La gestion de l'eau potable des ménages était complexe, avec différentes sources utilisées à chaque moment donné et selon les saisons. L'eau de pluie était la source d'eau potable la plus couramment utilisée. Les ménages mélangeaient de l'eau provenant de sources différentes dans des récipients de stockage, y compris celles de sources «améliorées» et «non améliorées». L'eau de robinets se détériorait au cours du stockage (p < 0,0005), avec des teneurs d’E. coli allant de 520 UFC/100 ml (coefficient de variation, CV: 5,7) à 1100 UFC/100 ml (CV: 3,4). L'eau stockée provenant de sources autres que le robinet (principalement l'eau de pluie) avait une teneur moyenne en E. coli de 1500 UFC/100 ml (CV: 4,1), ce qui n'est pas significativement différent de l'eau de robinet stockée (p = 0,20). La contamination microbienne de l'eau stockée était significativement associée aux pratiques de stockage et de manipulation observées, y compris le trempage des mains ou des récipients dans l'eau (p < 0,005) et le fait d'avoir un récipient de stockage non couvert (p = 0,052).
La qualité microbienne des sources d'eau «améliorées» dans notre zone d’étude n’était pas maintenue aux lieux de consommation, probablement en raison d'une combinaison du mélange d'eau de sources différentes à l’échelle des ménages, des pratiques de stockage et de manipulation à risque et d'eau de robinet des ménages inadéquatement traitée. Ces résultats ont des implications pour affiner les objectifs internationaux pour l'accès à l'eau potable ainsi que les hypothèses sous-jacentes sur les estimations de la charge mondiale des maladies, qui postulent que les sources «améliorées» posent des risques réduits de maladies diarrhéiques.
Investigar la calidad del agua corriente y del agua de lluvia en el punto de consumo, en un área con una cobertura en rápida expansión de fuentes de agua “mejoradas.”
Estudio croseccional de 914 hogares periurbanos en la Provincia de Kandal, Camboya, entre Julio-Agosto 2011. Hemos recogido datos de todos los hogares sobre el manejo del agua, la calidad del agua potable, y los factores potencialmente relacionados con la contaminación del agua después de su recolección. Las muestras del agua potable se tomaron directamente de una submuestra de grifos de los hogares (n = 143), de agua corriente almacenada (n = 124), de otra agua almacenada (n = 92) y de agua almacenada tratada (n = 79) y se les realizaron análisis básicos para E. coli y otros parámetros.
El manejo del agua potable de los hogares era complejo, con varias fuentes utilizadas al mismo tiempo y a lo largo de las estaciones. El agua de lluvia era la más comúnmente utilizada como fuente de agua para beber. Los hogares mezclaban en contenedores, para su almacenaje, agua proveniente de diferentes fuentes, incluyendo agua de fuentes “mejoradas” con agua de fuentes “sin mejorar”. El agua corriente obtenida a través de los grifos se deterioraba durante su almacenaje (p < 0.0005), de 520 ufc/100 ml (coeficiente de variación, CV: 5.7) E. coli a 1100 UFC/100 ml (CV: 3.4). El agua almacenada que no provenía de la red de tuberías (principalmente agua de lluvia), tenía un conteo de E. coli de 1500 ufc/100 ml (CV: 4.1), nada significativamente diferente al agua corriente almacenada (p = 0.20). La contaminación microbiana del agua almacenada estaba significativamente asociada con las prácticas de almacenamiento y de manejo observadas, que incluían el introducir las manos o recipientes dentro del agua (p < 0.005), o mantener el contenedor de almacenaje sin cubrir (p = 0.052).
La calidad microbiana de las fuentes de agua “mejoradas” en nuestra área de estudio no se mantenía en el lugar de consumo, posiblemente debido a una combinación de mezclar las fuentes de agua dentro del hogar, un almacenamiento y prácticas de manejo poco seguras, y un agua corriente tratada inadecuadamente. Estos resultados tienen implicaciones a la hora de refinar los objetivos internacionales para el acceso a agua potable segura, al igual que los supuestos sobre los que se basan los cálculos de la carga global de la enfermedad diarreica, que postulan que las fuentes “mejoradas” de agua plantean un riesgo mínimo para la enfermedad diarreica.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The WHO/UNICEF Joint Monitoring Programme for Water Supply and Sanitation (JMP) tracks global coverage of water and sanitation, dividing water and sanitation sources into ‘improved’ and ‘unimproved’ categories, with ‘improved’ sources defined as those assumed to pose a lower risk to health (Table 1). In its 2012 Update, the JMP indicated that the global Millennium Development Goal (MDG) to ‘halve, by 2015, the proportion of people without sustainable access to safe drinking water’ (from 1990 levels) had been met in 2010, with global coverage of improved water sources at 89% (WHO/UNICEF 2012). The definition of ‘improved’ does not, however, include any measure of the consistency of access or the microbiological or chemical quality of water delivered. These definitions have wider implications: a 2012 Global Burden of Disease/Comparative Risk Assessment used the JMP definition, considering ‘improved sources’ as having negligible attributable health risks (Lim et al. 2012). Even so, there is a growing body of evidence suggesting that there are health benefits and other advantages offered by consistently treated on-plot water access over other improved sources that may be less safe (Cairncross & Valdmanis 2006; Bartram & Cairncross 2010; Brown et al. 2013). Conflicting viewpoints reflect the lack of global data on the relative safety and sustainability of different improved sources and underline the need, as noted by the JMP, for improvements to current indicators in post-2015 monitoring (WHO/UNICEF 2012).
|Improved sources||Unimproved sources|
|Piped water into dwelling, yard or plot||Unprotected dug well|
|Public tap or standpipe||Unprotected spring|
|Tube well or borehole||Cart with small tank or drum|
|Protected spring||Tanker drum|
|Protected dug well||Surface water (river, dam, lake, pond canal, irrigation channel)|
|Rainwater collection||Bottled water (considered improved when household uses drinking water from an improved source for cooking and personal hygiene)|
Cambodia has seen improved water access rise from 48% in 1990 to 63% in 2000 and 87% in 2010, with piped, on-plot access rising from 15% to 63% between 1990 and 2010 (WHO/UNICEF 2012). At the national level, Cambodia has handily met the MDG water access target. Rapid expansion there has followed national decentralisation policies encouraging construction of public and private water supply systems, with particularly rapid growth of privately operated supplies (ADB 2007; UNDP/UNICEF 2009). Many small water supplies remain within the informal sector, however, possibly due to complex registration and licensing procedures, limited monitoring and regulation, and associated formal and informal charges (ADB 2007). There may be considerable variability in terms of service reliability and management, treatment effectiveness, water quality and quantity available to households with access to piped-to-plot supplies and other improved sources (ADB 2007; Feldman et al. 2007; JICA 2010). However, these attributes have not been systematically assessed, and no formalised, nationally representative monitoring system is in place for water safety.
To gain perspective on the rapid growth of ‘improved’ water access in Cambodia, we examined water quality and household-level water management in two communities with high coverage (97%) of ‘improved’ water, including on-plot connections to a piped supply (25%) and rainwater harvesting (95%). The goal was to understand whether these improved drinking water sources could also be characterised as microbiologically low risk. The study setting provided an opportunity to characterise the safety of on-plot improved water sources. By contrast, the majority of research on this topic has focused on sources outside the household.
We conducted a cross-sectional survey in two peri-urban sites (‘A’ and ‘B’) between July and August 2011. We used a census-derived list of approximately 1000 community water sources in Kandal Province, which, in consultation with local government officials, was used to identify private piped water supplies. We identified 12 candidate sites meeting selection criteria of (i) high access to improved water, including piped-to-plot water, (ii) rapidly developing peri-urban or urbanising area, and (iii) proximity to the water quality laboratory in Kandal Province. As the province surrounding the capital city of Phnom Penh, Kandal is the most rapidly urbanising area of Cambodia (UNDP/UNICEF 2009), with rapid expansion of piped water networks. Sites A and B were selected from eligible areas based on consent of the village leadership and system operators and after consultation with local government authorities. Many households in both villages had private piped water connections, although the reliability and quality of water delivered was highly variable and a number of households relied on piped connections in the dry season only. Piped water at site A was subjected to basic but inconsistent treatment (occasional chlorine pre-treatment, followed by partial and inconsistent flocculation and sedimentation); piped water at site B was untreated. Both systems abstracted surface water and were operated intermittently. We selected households at random from within village boundaries and presented an adult member of the household with informed consent information. We then enrolled consenting households, who agreed to respond to a questionnaire covering a range of issues related to water use, water safety, health, and related knowledge, attitudes, practices and beliefs.
We also collected water samples from households in our sample with functional piped connections on the premises (either inside the living area or in the yard), which accounted for at least part of their drinking water. Approximately 80% of such sources were sampled, with sampling limited by daily laboratory processing capacity. Samples included piped water at tap, stored piped water, non-piped stored water (usually rainwater) and treated (usually boiled) water. We tested water samples for E. coli and basic physical–chemical measures (pH, temperature, turbidity and free and total chlorine at the point of sampling). For microbial samples, we used sterile Whirl-Pak™ sample collection bags (Nasco Corp., FA, USA) and transported samples on ice after collection for processing within 24 hours at our laboratory. Chlorine residuals were tested at the point of collection from the single site where piped water was treated with chlorine (site A) at the point of entry and at tap in a random selection of households.
We used membrane filtration for detection of E. coli in water samples. Following filtration through sterile 47-mm, 0.45-μm pore size cellulose membranes (Millipore, Bedford MA, USA), we placed membranes on 60-mm plates of selective agar medium (Bio-Rad Rapid E. coli 2™; Bio-Rad Laboratories, Inc., Hercules, CA, USA) and incubated at 35 °C for 18–24 h. We processed all water samples in duplicate, with three dilutions each. We measured pH and temperature using a Hach pH meter, turbidity using a Hach nephelometer (Hach, Loveland, CO, USA) and free and total chlorine using a Hach chlorine test kit (Hach).
Following double entry and verification of all data by a local data services company (http://www.digitaldividedata.org), we performed data analysis in both Excel and in STATA 11.1 (Statacorp, TX, USA). For assessing trends and differences between covariate groups, we used non-parametric statistical tests of heterogeneity, including Kruskal–Wallis and Wilcoxon rank-sum tests. For microbial quality measures, we used arithmetic means as indicators of risk (Haas 1996), Williams' means as potentially more accurate indicators of central tendency (Alexander 2012) and geometric means for comparability with other published findings that report this measure. We used negative binomial analyses to investigate predictive factors for water quality at various household water points (Alexander 2012). We used raw mean counts of E. coli in negative binomial regression, with outcomes expressed as risk ratios (RRs), estimating the change in the relative mean number of events between categories (McElduff et al. 2010). As these data were drawn from a larger study powered for other indicators, water quality data were not powered for full multivariate analysis. We report possible correlation where our findings showed statistical significance in both our regression analyses and non-parametric tests of homogeneity.
We obtained ethical approval for this study from the London School of Hygiene and Tropical Medicine's Research Ethics Committee, Duke University's Institutional Review Board and the Cambodian National Ethics Committee for Health Research.
Table 2 summarises key descriptive data. We visited a total of 914 households, with an average of 5.3 members per household. We did not detect free or total chlorine in any sample from site A's distribution network (n = 50, detection limit: 0.10 mg/l). More than 97% of households in these two communities had access to an ‘improved’ water source at the household level. All households practised domestic water storage.
|Total number of HHs||914|
|Mean people per HH||5.3|
|HHs with children under 5||370 (41%)|
|Illiterate HHs||43 (4.7%)|
|Average years of adult education/HH||5.7|
|Mean self-reported HH income||125 USD|
|Mean self-reported HH expenditure||118 USD|
|HHs that change water source by season||77%|
|Major sources consumed untreated|
|Rainwater (n = 869)||54%|
|Water vendor (n = 275)||35%|
|Raw surface (n = 269)||38%|
|Bottled water (n = 98)||43%|
|Piped water (n = 225)||20%|
|Borehole private(n = 73)||15%|
|Neighbour (n = 57)||14%|
|Composition of household stored water|
|Stored piped water|
|Household pipe (n = 160)||65%|
|Rainwater (n = 34)||14%|
|Mixed (including unimproved sources) (n = 49)||20%|
|Other (n = 4)||1%|
|Stored non-piped water|
|Surface water (n = 13)||2%|
|Rainwater (n = 612)||72%|
|Rainwater mixed with other sources (n = 187)||22%|
|Water vendor (n = 15)||2%|
|Other (n = 20)||2%|
|Stored treated water|
|Household pipe (n = 29)||5%|
|Raw surface water (n = 22)||4%|
|Water vendor (n = 25)||4%|
|Rainwater (n = 314)||55%|
|Bottled (n = 12)||2%|
|Mixed (including unimproved sources) (n = 157)||28%|
|Other (n = 13)||2%|
|Sanitation and Hygiene|
|HHs with soap present during visit||732 (83%)|
|Access to sanitation|
Our questionnaire revealed heterogeneity in water-use behaviours among households in the sample, within and across seasons. Approximately 77% of study households used different quantities and/or sources between seasons, often using a primary source supplemented by one or more secondary sources. Rainwater was the primary source for 82% of households in the rainy season, and stored rainwater was the most widely used secondary source in the dry season. After rainwater, three other sources were commonly used, in roughly equal proportions: raw surface water, vended water and piped-to-plot water (Figure 1).
Households stored water from a number of different sources and consumed water from some source types directly without treatment. We divided household storage containers into three main categories: those primarily storing piped water (usually large concrete jars situated below the household's piped water tap); those primarily storing non-piped water (concrete jars either connected to rainwater drainage pipes or open to any source, often filled by vended water in the dry season); and those storing water treated at the household (usually a smaller container kept in the household). Approximately 94% of all households had stored, non-piped water, 30% had stored, on-plot piped water and 63% had stored treated water available during our visits. Approximately 56% of households reported drinking untreated water directly from household storage containers, most commonly using the major primary and secondary sources listed above (Table 2). Direct consumption without treatment differed significantly across different sources (P = 0.0001) with stored rainwater consumed the most frequently, followed by bottled water, surface water, vended water (largely untreated surface water) and piped water. Households commonly mixed water from various sources together, including from ‘improved’ and ‘unimproved’ sources. Three quarters of containers storing piped water also contained water from other sources, and all storage container types were at least 20% composed of mixed – including unimproved – sources (Table 2).
We also asked households about their perceptions of different sources. Approximately 72% of households reported no aesthetic or health concerns related to water in the rainy season, compared with 57% in the dry season. Rainwater was perceived to be of the highest quality, with over 90% of users reporting ‘no concerns’ in either season. By contrast, only 55% of piped water users had no concerns across both seasons and reported the major issues as being that it ‘looked bad’ and was ‘bad for health’. Responses were similar for raw water and vended water where 50–60% reported no issues, while the main issues mentioned included physical appearance and health concerns. Water quality perceptions among households in this study are further explored in another publication from the same data set (Orgill et al. 2013).
We investigated E. coli counts in four water sample types: (i) piped at tap (n = 143), (ii) piped and stored (n = 124), (iii) non-piped and stored (n = 92), and (iv) treated and stored (n = 79). We only tested microbial quality across sources in households who also had running piped water, to observe the difference between the tap, storage and point of use. Microbial levels were significantly different across all four sources (P = 0.0001). E. coli counts in piped water at tap were far lower than those of stored piped water (P < 0.0005), suggesting household-level recontamination or growth of E. coli in storage. The arithmetic mean of tap water was 520 cfu/100 ml [CV (coefficient of variation): 5.7], rising to 1100 cfu/100 ml (CV: 3.4) during storage (Table 3). Williams' means increased from 11 to 68 cfu/100 ml.
|E. coli cfu/100 ml||Piped: tap||Piped: stored||Non-piped: stored||Treated: stored|
|<1||67 (47%)||22 (18%)||28 (30%)||57 (72%)|
|1–10||7 (5.0%)||12 (10%)||6 (7.0%)||6 (8.0%)|
|11–100||35 (24%)||33 (26%)||19 (21%)||5 (6.0%)|
|101–1000||27 (19%)||41 (33%)||24 (26%)||7 (9.0%)|
|1000+||7 (5.0%)||16 (13%)||15 (16%)||4 (5.0%)|
|Arithmetic mean (95% CI)||520 (34–1000)||1100 (430–1700)||1500 (220–2700)||350 (73–770)|
|Coefficient of variation (arithmetic mean)||5.7||3.4||4.1||5.4|
|Geometric mean (95% CI)||120 (81–180)||170 (120–250)||200 (120–340)||120 (45–330)|
Water quality at tap was the only source found to differ significantly between sites A and B (P = 0.001). Williams' means of E. coli were lower at site A (6.9 cfu/100 ml, n = 95), where there was rudimentary water treatment in place, than in site B (34 cfu/100 ml; n = 48). We detected no difference in stored piped water between sites, however (P = 0.77).
We found no significant difference between E. coli counts in piped and non-piped stored sources (P = 0.203), although mean stored non-piped water was more contaminated than stored piped water, with an arithmetic mean of 1500 cfu/100 ml (CV: 4.1) and a Williams' mean of 40 cfu/100 ml (Table 3). Many households had stored, treated (usually, boiled) water on hand, which was of significantly lower risk compared with all other household sources of drinking water (P < 0.0005). Stored treated water had a Williams' mean of 3 cfu E. coli/100 ml and an arithmetic mean of 350 cfu/100 ml (CV: 5.4).
Unhygienic water storage and handling practices were strongly correlated with microbial contamination in water samples. Households accessing water by dipping hands or using a receptacle (as observed during sample collection) significantly increased E. coli counts compared with households who used pouring (E. coli: RR 10 95% CI 3.2–34, P < 0.005; Table 4). E. coli counts in samples from households having a covered storage container were approximately half of those in samples from households with uncovered containers (RR 0.49 95% CI 0.24–1.0, P = 0.052). Variables in our hierarchical model that did not yield significant differences across categories included sanitation access (according to JMP definitions), self-reported hand-washing practices, self-reported frequency of water treatment, original water source, education levels, income or household size. Our analysis of risk factors for household water contamination was constrained by our number of water samples, however, allowing only univariate analysis. The above-mentioned non-significant factors all had very low or highly uneven sample sizes across categories of hypothesised explanatory variables.
|Selected covariates||Rate ratio (RR)|
|Piped vs. tap||2.1||1.1–4.0||0.03|
|Boiled vs. piped||0.57||0.37–0.87||0.008|
|Boiled vs. non-piped||0.24||0.080–0.70||0.009|
|Stored piped vs. non-piped||1.4||0.70–2.6||0.37|
|Dipping vs. pouring (stored treated)||10||3.2–34||<0.005|
|Covered vs. uncovered (stored piped)||0.49||0.24–1.0||0.052|
Our results suggest that ‘improved’ drinking water sources, considered safe by the global monitoring framework and burden of disease analyses (Lim et al. 2012; WHO/UNICEF 2012), may be unsafe in some settings, and improved water on site can be subject to further contamination once stored at the household level. Household drinking water storage and handling practices that vary across seasons, by source, and involve mixing may be associated with poor drinking water quality and therefore potentially increased risk of waterborne disease. In our study, mean E. coli counts across all tested household sources exceeded the Government of Cambodia's limit of <1 cfu/100 ml for E. coli (MIME 2004). These levels of E. coli probably indicate a higher health risk than that recommended by the WHO's risk-based target for drinking water quality (10−6 DALY per person per year) (WHO 2011c). International standards for drinking water safety vary, but E. coli counts exceeding 100 cfu/100 ml are widely considered to be ‘high risk’ (ibid., Brown et al. 2008).
There are at least three reasons why near-universal access to improved water sources in these study sites has not resulted in consistent access to low-risk water. First, intermittent access to rainwater and piped water has made water storage necessary, and water from all sources is subject to contamination from post-collection handling and storage. Our results are consistent with a number of studies reporting post-collection recontamination (Jensen et al. 2002; Clasen & Bastable 2003; Brick et al. 2004; Wright et al. 2004; Crampton 2005; Trevett et al. 2005; Oswald et al. 2007; Levy et al. 2008, 2009; Copeland et al. 2009; Sodha et al. 2011; Brown et al. 2013), although most previous studies have focused on recontamination after collection outside the household. Faecal contamination during water handling was highly plausible in our study setting given the high percentage of households lacking ‘improved’ sanitation (45%), the low prevalence of self-reported hand-washing behaviours (21% after defaecation, 10% after cleaning babies) and the reported and observed unsafe water handling and storage practices (Table 4).
Second, households commonly mixed improved and unimproved sources (Table 2) with the composition of mixed water depending on the available sources and preferences. Furthermore, respondents’ reported perceptions of the relative safety of sources were not correlated with microbial contamination, a finding that is explored further in a separate paper from this study (Orgill et al. 2013). Although most households in this study believed rainwater to be the safest source, our data suggested it to be the most faecally contaminated source of drinking water in these settings. Previous studies of rainwater harvesting have identified risk factors affecting rainwater quality, including contamination of rooftops or other collection surfaces, as well as unsafe storage and handling (Meera & Mansoor 2006; Fewtrell & Kay 2007; Kahinda et al. 2007; Ahmed et al. 2010).
Third, piped water supplies delivered abstracted river water that was treated only partially and inconsistently (site A) or was untreated (site B). No chlorine residue was detected in any sample from household taps. Both systems operated intermittently, a risk factor for microbial contamination through back siphonage and infiltration when systems lose positive pressure (Basualdo et al. 2000; Cotruvo & Trevant 2000; Agard et al. 2002; Lee & Schwab 2005), although it is unclear whether this is a meaningful contributor to risk when systems are delivering untreated or minimally treated river water. Arithmetic mean E. coli count was 520 cfu/100 at the tap, with further microbial contamination during storage. This high level of microbial contamination is likely to pose health risks to consumers, which should be considered alongside the many health and non-health advantages conferred by access to an on-plot water supply connection (Tomkins et al. 1978; Cairncross & Cliff 1987; Churchill et al. 1987; Aiga & Umenai 2002; Tumwine et al. 2002; White et al. 2002; Bartram & Cairncross 2010; Devoto et al. 2011; Sorenson et al. 2011; Brown et al. 2013). While microbiologically safe, piped water delivered directly and reliably to households is the end goal for water services provision (Cairncross & Valdmanis 2006), achieving this in practice presents significant challenges. Our data suggest that, in these settings, the relative advantages of having a household tap may not be so clear. Our observations may apply to other regions as well given the rapid global increase in access to improved sources in areas where the infrastructure and human resources required to maintain and regulate supplies may need further development. Indeed, similar conclusions on the need for appropriate supporting infrastructure to make piped water safer and more sustainable were recently drawn in studies in India (Jalan & Ravallion 2003), Brazil (Gamper-Rabindran et al. 2010) and Yemen (Lechtenfeld 2012).
In terms of protecting drinking water quality, our study communities would probably benefit from safer and more consistently available piped water as well as support for managing household water safety. Adequate treatment, improvements in design (e.g. pressure regulation) and basic support for operation and maintenance are needed in informal systems, many of which are constructed by local entrepreneurs with limited access to resources or knowledge about engineering or safety aspects of water supply infrastructure (Lee & Schwab 2005). Locally viable mechanisms for providing this support are unclear, but optimisation of existing systems that deliver water to the household represents ‘low-hanging fruit’ for rapidly expanding safe water access. Extension of local regulation and monitoring may help drive improvements where there is sufficient willingness to pay. Where local capacity constrains sustainable improvements in infrastructure services, communities may also benefit from access to household-based water treatment and safe storage (HWTS) options and information about risks associated with unsafe water. Rigorous studies of the effect of household-specific water quality information show that households, when informed about contamination, do adjust their water sourcing and/or treatment practices to utilise safer alternatives (Jalan & Somanathan 2008; Hamoudi et al. 2012). However, more research is needed on the content and delivery of such information to effect sustained behaviour change (Lucas et al. 2011; Huber & Mosler 2013) and on evidence-based behaviour change interventions to more generally improve and sustain water, sanitation and hygiene practices (Fiebelkorn et al. 2012; Mosler 2012). While we acknowledge challenges to achieving high adherence to HWTS (Schmidt & Cairncross 2009; Brown & Clasen 2012), the near-universal reliance on water storage and the quality disparity between available sources suggests that there may be a role for targeted, context-specific and effective HWTS methods as interim solutions. Many households in our study reported point-of-use treatment of stored water (boiling, in 80% of instances), with 72% of boiled water samples being low risk (≤10 cfu/100 ml E. coli) after treatment. Another study from peri-urban Cambodia found that boiling may not be consistently practised by most households who report it (Brown & Sobsey 2012), however, and recontamination in storage is possible. Other options such as filtration and point-of-use or point-of-collection chlorination have shown promise in terms of improving water quality, but many challenges related to stimulating correct, consistent and exclusive use of such technologies remain (Ahuja et al. 2010; Ahmed et al. 2011; Kremer et al. 2011; Whittington et al. 2012). More studies may be needed to identify appropriate and effective solutions with scalable promise, while concurrent efforts are underway to improve the safety and reliability of piped, treated water supplies.
Ensuring sustained access to safe drinking water requires that we move beyond the simplified typology of sources as ‘unimproved’ and ‘improved’. It is notable that the words ‘safe’ and ‘sustainable’ are explicit in the wording of the MDG target (WHO/UNICEF 2012), although neither of these qualities is systematically monitored at the global level, partly because there are no international consensus definitions of ‘safety’ or ‘sustainability’ of water supplies. The JMP acknowledge the limitations of current global water metrics and have been leading efforts to overcome the significant challenges to using measures of safety and sustainability in the post-2015 targets for expanding access (WHO/UNICEF 2012). While Cambodia and the globe are on track for achieving the UN MDG water target, diarrhoeal morbidity still ranks as one of the top five causes of global DALYs (Lim et al. 2012). Cholera cases have increased dramatically in Cambodia, South-East Asia and globally (WHO 2013). In addition to increasing access to basic water, sanitation and hygiene infrastructure and services, a greater focus is needed on assuring the safety of such services, especially to those most at risk. Such efforts would contribute to MDG targets for poverty reduction, nutrition, childhood survival, school attendance and gender equity and would contribute to fulfilling WHO and Member State obligations (WHO 2011a,b). While real progress has been made in global access to water, the estimated 780 million without ‘improved’ water (WHO/UNICEF 2012) and the potentially much higher number lacking access to microbiologically or chemically safe drinking water deserve sustained international attention and continued meaningful investment.
We would like to thank WaterSHED-Cambodia for their invaluable support in hosting this study, study participants for their patience in responding to our many questions and two anonymous reviewers for their the insightful feedback on a previous draft of this paper. We gratefully acknowledge funding support from WaterSHED-Asia and the Duke Global Health Institute.