Groundwater use and diarrhoea in urban Nepal: novel application of a geostatistical interpolation technique linking environmental and epidemiologic survey data.

Background: Groundwater is a common domestic water source in developing countries, but is persistently contaminated with enteropathogens. However, studies on determinants of diarrhoea have predominantly focused on piped water. This study examines the relationship between groundwater microbial quality and household diarrhoea occurrence (HDO).
Methods: Considering it as a proxy of enteropathogens, this study analysed Escherichia coli concentrations in groundwater wells. Ordinary kriging, a geostatistical technique in geographic information systems, was used to interpolate the E. coli concentration to survey points that had secondary survey data (n=942). The relationship between E. coli and HDO using simple and multivariate statistical analyses in SPSS was analysed.
Results: A total of 77% of households used groundwater. One-third of households were without piped-water access (PWA), and these households were significantly more likely to use groundwater than those with PWA. Of the 87 households that reported HDO, 77% were groundwater users. Of the groundwater users, the households with HDO consumed groundwater with significantly higher E. coli concentrations than the households without HDO. Of the households without PWA, the increase in the E. coli concentration increased the odds of HDO (adjusted odds ratio=3.15; 95% CI=1.07-9.22).
Conclusion: It is suggested that the groundwater microbial quality is a risk factor for HDO and illustrates this by an application of an interpolation technique relevant for developing countries.

Introduction

Groundwater constitutes half of the potable water used in countries such as Bangladesh, China, India, Indonesia, Nepal, the Philippines, Thailand and Vietnam,[1] particularly for the poorest urban households.[2] While an important source of consumable water, groundwater has widely been reported to have faecal contamination in countries at all levels of economic development.[3-9] In the Kathmandu Valley, which is a representative area of a developing country in Nepal, groundwater is used by 52% of the households and is a major water source.[10] A total of 3% of groundwater users use it for drinking and 88% use it for bathing purposes.[11] Shrestha et al.[12] recorded numerous studies conducted at different times and showed that the faecal contamination of shallow groundwater is a persistent problem in the valley. This discussion provided evidence that shallow groundwater is unprotected and cannot be considered to be an improved source as defined by the World Health Organization (WHO). In addition, the presence of Escherichia coli along with other pathogens and viruses[13] points towards a serious threat of waterborne diseases to public health in the valley.

According to the WHO, nearly 1.7 billion cases of diarrhoeal disease occur each year and it is the leading cause of malnutrition and the death of approximately 525 000 children under the age of 5 years each year worldwide. It is the second leading cause of years of life lost due to premature death in Nepal[14] and the under-5 mortality rate is 35 per 1000 people.[15] To establish preventive measures for diarrhoea, countless studies have been performed to identify its determinants over the last few decades. Numerous studies have emphasized piped water as an important determinant—access,[16] use and quantity.[17] In an effort to reach Millennium Development Goal 7 Target10c, the coverage of improved drinking water sources increased to 90% in South Asia. The developments of improved drinking water sources (90% in South Asia) basically refer to other improved water sources (65% coverage), rather than piped water (25%).[18] Based on the data collected by the Asian Development Bank (ADB) in the Kathmandu Valley, 21% of households did not have access to a piped-water supply in 2009,[17] which rose to 34% in 2015.[11] In addition, the households that have access receive this supply for six or fewer hours per week.[11] Despite the fact that alternative sources represent a huge share in household water consumption, these sources have been analysed in few studies[19] as determinants of diarrhoea.

In developing countries, conducting large-scale surveys is difficult due to institutional and technical reasons, which leads to insufficient data for scientific studies. In this situation, the application of a geographic information system (GIS) to interpolate data spatially could be an affordable solution. Most of the studies on risk factors associated with diarrhoeal diseases are based on a questionnaire survey. Combining survey data with microbial concentration data can provide rigorous findings. Interdisciplinary studies combining environment and health are difficult to conduct in developing country settings due to technical limitations. If this study can link separate datasets of different disciplines, it will be a cost-effective and easy approach to obtain significant findings.

With the background of wide usage and the poor microbial quality of groundwater, as well as the lack of investigations regarding their relationship with diarrhoea, this study analyses the microbial quality of the groundwater as a determinant of diarrhoea in the Kathmandu Valley. The first part of this paper illustrates the spatial distribution of the groundwater microbial concentration using a geostatistical interpolation technique in GIS to integrate environmental data (microbial quality) with social data. An ordinary kriging interpolation is applied, which uses the statistical properties of the measured points to quantify the spatial autocorrelation between measured points and account for the spatial configuration of the sample points around the prediction location. The second part examines the potential impact of the groundwater microbial quality on household diarrhoea occurrence (HDO) with the aim to emphasize the necessity of pollution control interventions for water sources in addition to piped water.

Methods

Study area

The Kathmandu Valley lies in the central hilly region of Nepal and comprises 85% of the Kathmandu district, the entire Bhaktapur district and 50% of the Lalitpur district. The valley is situated at an elevation of 1300–1400 m above sea level, and is drained by the Bagmati River and its tributaries. The residents of the valley exploit two aquifers to meet their water needs—shallow and deep aquifers. The shallow aquifer is composed of up to 50 m of Quaternary arkosic sand, with some discontinuous interbedded silt and clay, and lies in the northernmost part of the valley.[20] The deep aquifer is beneath the interbedded clay and lignite aquitard of up to 200 m thickness and consisted of Pliocene sand and gravel, with interbedded lignite, peat and clay, and lies in the southern zone of the valley.[21] Based on the water table map and the piezometric surface elevation, the groundwater flow is towards the centre from the periphery.[22] The shallow aquifer is directly recharged via precipitation and small rivers, whereas the deep aquifer is recharged via precipitation from the northeast part of the valley.[23] Groundwater is tapped from the shallow aquifers via dug wells and tube wells, and from the deep aquifer via deep tube wells.

The Kathmandu Valley is the largest urban centre in the country; it has a population of 2.51 million people and shows an annual growth rate of 3.65%. The Kathmandu district alone has a population density of 4408 persons per km2, which is 30 times higher than the national population density. Most of the wards (administrative units) of the Kathmandu metropolitan city (KMC) and the Lalitpur sub-metropolitan city (LSMC) have population densities higher than 10 000/km2, especially the old settlements that lie in the centre, which have densities >20 000/km2.[24] Kathmandu Upatyaka Khanepani Limited (KUKL) supplies 106 million litres/day (MLD) and 76 MLD of water during the wet and dry seasons, respectively, countering a water demand of 320 MLD, and the water loss due to leakage is 40%.[25] Of the alternative water sources used, 52% of the households use groundwater.

Groundwater sample collection and microbial analysis

In this study, 36 groundwater wells were randomly and uniformly selected from the wards in KMC and LSMC (Figure 1). The samples were collected once during August–September 2009, which was the wet season. The daily rainfall data of the sampling period was compared with those of August–September in 2008 and in 2010. The rainfall amount and pattern in the sampling period were similar to those in the previous and following years, indicating an absence of extreme rainfall events within the sampling window. Before collecting water samples, tube wells and dug wells that were fitted with a hand pump were purged for 1–2 min. For open dug wells, the water samples were brought to the surface using the same method used by the owners. The autoclaved 250 mL polythene bottles were rinsed with the groundwater to be sampled, then the samples were collected and immediately stored in an icebox.

Figure 1.

Locations of groundwater wells and questionnaire survey points in the Kathmandu metropolitan city (KMC) and the Lalitpur sub-metropolitan city (LSMC) area. Open circles (Nwell=36) represent wells, and closed triangles (Nsurvey_points=35) represent survey points. The dark black boundary represents the boundary between KMC and LSMC. The river flowing from east to west separates KMC and LSMC.

E. coli was selected as the faecal indicator bacterium, and the concentrations in the groundwater previously reported[12,13] were used for statistical analysis. The EPA-approved IDEXX Quanti-Tray method (USA) using the Colilert reagent was applied. After mixing the reagent, 100 mL of the sample water was poured into a tray, and the tray was sealed and incubated for 24 h at 35±0.5°C. After incubation, the wells in the tray producing a blue colour under UV light were counted. The most probable number (MPN) table is referred to here to determine the MPN of the E. coli in 100 mL of the sample. Only one sample was analysed per site, and field blanks, laboratory blanks and duplicates were not used.

In this study, E. coli was considered to be the faecal indicator bacterium to infer the presence of pathogens, and the Quanti-Tray method, which was used here, cannot discriminate between pathogenic and non-pathogenic strains of E. coli. E. coli is applicable for predicting total bacteria pathogen[26] and pathogenic viruses.[27]E. coli concentration has been widely applied in risk assessment studies in the form of faecal indicator ratio conversion.[28] Based on these lines of evidence, E. coli was treated as a proxy of pathogenic microorganisms to examine the impact of groundwater microbial quality on HDO.

Secondary data from household surveys

Secondary data were obtained from a household survey of the Kathmandu Valley Water Distribution, Sewerage and Urban Development Project conducted by the ADB from August to September in 2009.[10] The study used a multi-stage cluster survey design.[10] The ADB chose 84 geographical locations randomly and then drew a circle with each randomly generated geographical location at the centre. Furthermore, they selected 20 houses closest to the random location and all the households that were in those houses were interviewed. In the Kathmandu Valley, one house may contain more than one household. The total number of households interviewed was 2214. This study selected 35 geographical points among the 84 that lie within KMC and LMSC (Figure 1) for a total number of households of 942. Details of the survey strategy, and inclusion and exclusion criteria of households have been presented elsewhere.[10,17]

The secondary data included information on diarrhoea occurrences in at least one family member in the past month, which was defined as an HDO in this study. According to the WHO, diarrhoea is defined as the passage of three or more loose or liquid stools per day (or a more frequent passage than is normal for the individual); this definition was applied in the present study. In addition to health, information on water-use behaviour was also obtained from the secondary data for the statistical analysis. The data on the sources of water use and the purpose of use of each source, as well as socio-economic and demographic data, were also used.

Data analysis

Geostatistical spatial interpolation

Here, the ArcGIS Geostatistical Analyst Extension of the ArcMap software (Version 10.1, Environmental System Research Institute, USA) was used to interpolate the groundwater microbial quality data over KMC and LSMC, and to the 35 studied geographical points. Ordinary kriging was chosen as an appropriate tool based on previous groundwater studies.[29] Ordinary kriging weights surrounding the measured values were used to derive a prediction for each location, and the weights were based on the distance between the measured and predicted locations, as well as on the spatial arrangement of the measured points. Kriging requires spatial autocorrelation, where close samples are considered to be more similar than those farther apart. Based on this concept, an interpolated microbial concentration at each point was then assigned to all 20 houses associated with that point.

First, the spatial dependency was quantified using a semivariogram.[30] The empirical semivariogram is then fitted to a theoretical model, and the best-fit model of the spherical, exponential, linear and Gaussian models was used for the interpolation. The prediction accuracy of the fitted model was determined using the root mean square error (RMSE) and the coefficient of determination (R2) obtained via cross-validation. The RMSE should be as small as possible.[29]

Statistical analysis

A χ2 test of independence was used to compare the HDO between groundwater users and non-users. An independent t-test and its non-parametric alternative, the Mann–Whitney U-test, were performed to compare the E. coli concentrations between HDO and without HDO in different groups:

groundwater users;

households that use groundwater only for drinking;

households that use groundwater only for bathing;

socio-demographic and water use characteristics;

those with and without piped-water access (PWA).

The relationship between PWA and groundwater use was examined using the χ2 test. Finally, to examine the relationship between the groundwater microbial quality and HDO, a simple logistic regression model was applied separately for the households with and without PWA. Each model was adjusted for common influencing socio-demographic characteristics, e.g. ethnicity, income, education of the decision maker, the head of the household and family size. Statistical significance was set at a p-value of <0.05. The statistical program IBM SPSS Statistics Version 21.0 (IBM Corporation, Armonk, NY, USA) was used for all statistical analyses.

Results

Observed groundwater E. coli concentration

E. coli was detected in 61% (22/36) of the groundwater samples.[12,13] The minimum, maximum and mean (±SD) values of the observed E. coli concentrations were <0, 3.84 and 1.271 (±1.274) Log10 MPN/100 mL, respectively.

Interpolation of groundwater E. coli concentration and integration with survey data

Observed E. coli concentrations were used for the interpolation and, because the data best fit to a log-normal distribution, it was log transformed beforehand. The ordinary kriging method was used for interpolation. Different semivariogram models were tested for model prediction—circular, spherical, exponential, gaussian, k-Bessel and stable. Of the other semivariogram models, the gaussian model had the lowest RMSE (0.86) and the highest R2 (0.41). Therefore, it was used to prepare a prediction map of the E. coli concentration (Figure 2) over KMC and LSMC. The interpolated E. coli concentration ranged from 0.00 Log10 MPN/100 mL to 2.94 Log10 MPN/100 mL. The predicted spatial distribution suggests that the E. coli concentrations gradually increase moving from northeast to southwest.

Figure 2.

Observed and predicted surface of the E. coli concentration in the groundwater in the KMC and LSMC areas. The white bubbles represent the observed concentration, and the surface represents the interpolated E. coli concentration.

Using the prediction surface (Figure 2), the E. coli concentrations at 35 survey points (Figure 1) were determined. Of the 35 survey points, the mean (±SD), minimum and maximum E. coli concentrations were 1.707 (±0.796), 0.114 and 2.678 Log10 MPN/100 mL, respectively.

Because the interviewed houses were the closest buildings to the survey points and because all the households inside a house were interviewed, all households that belong to one survey point were assigned the same interpolated E. coli concentration for the respective survey point.

Descriptive information on socio-demographic and water using characteristics and HDO

Of the surveyed households, 47 (442/939), 52 (485/939) and 1 (12/939)% of households were from Brahmin/Chettri, Janajati and Dalit ethnicity, respectively, and 48% (366/762) of households had monthly incomes of >147 US$, 32% (242/762) had monthly incomes of 49–147 US$ and 20% (154/762) had monthly incomes of <49 US$ (US$ 1=NRs 101.94, as of 30 December 2017). A total of 73% (676/927) of the household heads were educated to above secondary level, 10% (97/927) to below secondary level and 17% (154/927) were illiterate. The average family size was 4.5 (±1.8).

A total of 75% (709/942) had PWA within their premises, 73% (684/942) households used groundwater and 29% (271/942) bought jar or bottled water. Of the groundwater users, 32% of households (184/575) bought jar water. Of the 942 households, 52% and 31% used piped water only for drinking and bathing, respectively. Likewise, 6% and 30% used groundwater only for drinking and bathing, respectively. Households without access to PWA had significantly higher (p-value <0.05) proportions of groundwater users compared with households that had PWA (Table 1). Likewise, the proportions of households that used groundwater only for drinking or for bathing were significantly greater in households without PWA than in those with PWA (Table 1). However, a significant difference between these two groups regarding other water sources such as vendor tankers or stone spouts was not observed.

Table 1.

Proportion of groundwater users among households with and without PWA

Water-use characteristics		Piped-water access**		p-value
		No	Yes
Groundwater use	No	23a	234c	<0.05#
	Yes	209b	475d
Drinking only groundwater	No	173a	707c	<0.05*
	Yes	59b	2d
Bathing only in groundwater	No	98a	556c	<0.05#
	Yes	134b	153d

Proportion of groundwater users in: without PWA=b/(a+b); with PWA=d/(c+d).

#, χ2 test; *, Fisher’s exact test; **, out of 942 households, one household has missing information on PWA.

A total of 9% (87/942) of the total households, 10% (68/684) of the groundwater users and 8% (21/258) of the groundwater non-users reported HDO. Even though there was no statistical significance, the tendency to HDO seemed to be higher in groundwater users compared with non-users. Of the households reporting HDO, 77% (67/87) used groundwater.

Relationship between HDO and groundwater microbial quality

Groundwater users

Groundwater users were selected for further analysis and the E. coli concentration was compared between users with HDO and without HDO across different household characteristics (Table 2). Table 2 shows an apparently higher mean E. coli concentration in the HDO group compared with the without HDO group across all household characteristics except for Dalit ethnicity. However, a significant difference was achieved for groundwater user households, where the household head’s education level was below secondary level or was illiterate, households with no PWA and households in which members bathed only in groundwater. However, the E. coli concentrations were not significantly different between the HDO and without HDO groups among the households that drank only groundwater and among those that bathed in all types of water.

Table 2.

Comparison of the E. coli concentration between households with and without diarrhoea occurrence among different strata of groundwater users

Groundwater users (N=684)	Diarrhoea frequency		E. coli concentration in groundwater Log10 MPN/100 mL		p-value
	No	Yes	Diarrhoea no	Diarrhoea yes
Groundwater users	617	67	1.58	1.79	0.05
Ethnicity
Brahmi/Chettri	299	39	1.60	1.83	0.09
Janajati	309	26	1.56	1.71	0.35
Dalit	7	1	2.05	1.64	1.00*
Household income#
>US$ 147	245	22	1.61	1.87	0.09
US$ 49–147	162	20	1.55	1.87	0.11
<US$ 49	94	10	1.53	1.81	0.33
Education of household head
Above secondary level	451	38	1.60	1.66	0.65
Below secondary level	66	9	1.64	2.10	0.047*
Illiterate	91	18	1.48	1.86	0.042*
Water treatment
No	132	10	1.48	1.81	0.20
Yes	479	56	1.61	1.77	0.16
Piped-water access
No	168	10	1.58	2.07	0.008*
Yes	449	57	1.58	1.72	0.23
Drinking only groundwater
No	558	65	1.61	1.78	0.10
Yes	59	2	1.37	1.90	0.45
Bathing only in groundwater
No	350	47	1.60	1.72	0.35
Yes	267	20	1.57	1.95	0.035*

*Mann–Whitney U-test; #, US$ 1=NRs 101.94, as of 30 December 2017.

Households without PWA

Among households without PWA, the E. coli concentration was significantly (p-value <0.05) higher in the HDO group than in the without HDO group. However, this relationship was not evident in households that had PWA. In a simple logistic regression analysis, the E. coli concentration in the groundwater was significantly and positively associated with HDO (Table 3). With a one unit (Log10 1 MPN/100 mL) increase in groundwater E. coli concentration, the odds of HDO increased three-fold (adjusted odds ratio [AOR]=3.15; 95% CI=1.07–9.22; p-value=0.03). Another factor that was associated with HDO was education. Compared with households with well-educated household heads, those with less-educated or illiterate household heads were at 4.6-fold (AOR=4.59; p-value=0.05) and 5.3-fold (AOR=5.3; p-value=0.03) more risk of HDO, respectively. Conversely, neither the E. coli concentration nor the groundwater use was significantly associated with HDO in households with PWA. The factors that were associated with HDO were ethnicity and the education level of the household head.

Table 3.

Factors associated with household diarrhoea occurrence (HDO)

Factors	Frequency (N=232)	B	p-value	Exp(B)	95% CI for Exp(B)
					Lower	Upper
E. coli concentration in groundwater (Log10 MPN/100 mL)	1.68 (±0.80)	1.22	0.03	3.15	1.07	9.22
Groundwater use
No	23	Ref.
Yes	209	−0.51	0.56	0.60	0.11	3.33
Ethnicity
Brahmin/Chettri	119	Ref.
Janajati	109	−1.13	0.09	0.32	0.09	1.18
Dalit	4	−20.01	1.00	0.000	0.000	0.000
Household income*
>US$ 147	82	Ref.
US$ 49–147	68	−0.31	0.68	0.74	0.17	3.19
<US$ 49	44	−0.07	0.94	0.94	0.19	4.64
Education of household head
Above secondary level	156	Ref.
Below secondary level	34	1.52	0.05	4.59	1.01	20.90
Illiterate	39	1.67	0.03	5.29	1.16	24.08
Family size (people)	4.5 (±1.8)	−0.06	0.73	0.94	0.66	1.33

B, beta coefficient; Exp(B), adjusted odds ratio; CI, confidence interval; *, US$ 1=NRs 101.94, as of 30 December 2017; Ref., reference group.

Discussion

The aim of this study was to identify the association of groundwater microbial quality with HDO by applying interpolation tools to link the groundwater microbial data (environmental data) with epidemiologic survey data. A total of 61% of the observed groundwater samples in this study detected E. coli and exceeded the National Drinking Water Quality Guideline, which states that E. coli should not be present in collected samples. These results are in line with previous groundwater studies in the Kathmandu Valley.[13,31] These comparative findings indicated persisting faecal contamination in the aquifer and confirmed that the water sources in most of the parts of study area fell under the ‘unimproved source’ category as defined by the WHO. As discussed in Shrestha et al.,[12] the possible sources of the microbial contamination of the groundwater wells in this urban area could be anthropogenic, e.g. leaky sewers, improperly constructed septic tanks, and improperly designed and managed storm drains that carried a mixture of sewage and storm water, which frequently overflowed, polluting the land surface.

These experimental E. coli concentration data were used for a geostatistical interpolation applying GIS for integration with the survey data. Even though the observed high E. coli concentration values were underestimated and the low values were overestimated, which is an inherent feature of interpolation algorithms,[32] the E. coli interpolation revealed that the quality of water worsened from northeast to southwest within the study area. JICA[33] (1990) divided the Kathmandu Valley watershed into three groundwater districts—northern (NGD), central (CGD) and southern (SGD). The northeastern part of the study area lies in the NGD, and the rest lies in the CGD. Previous studies have suggested higher chemical contamination in CGD compared with in NGD.[33,34] Most of the groundwater wells in CGD had a high ammonia, high nitrogen content[33] and high electrical conductivity compared with those in NGD.[34] The sediments in NGD consist of highly permeable sand and gravel, and exhibit the greatest groundwater potential, whereas those in CGD are comprised of very thick impermeable black clay, and a shallow layer, up to a depth of 20 m, overlies the thick clay. Such differences in the permeability of the sediments could produce a better dilution effect of the precipitation on the chemical, as well as the microbial contaminants in NGD, given the similar subsurface pollution.

Groundwater is the most commonly used alternative water source:

75% (709/941) households had PWA;

73% (684/942) households used groundwater;

29% (271/941) bought jar water.

Possessing a groundwater well within a compound makes households use it more often (ADB 2010).[10] It was also observed in this survey data that, compared with households that had PWA, those without access were more likely to use groundwater, and more likely to rely solely on groundwater for drinking and bathing purposes (Table 1), with 10% of groundwater using and 8% of non-using households reporting HDO, respectively. Similarly, 77% (70/87) of the records of HDO were groundwater users. These higher tendencies of HDO among groundwater users give a weak indication of some connection between groundwater use and HDO.

Across all household characteristics among groundwater users, the E. coli concentration was apparently higher in the HDO group than in the without HDO group. The E. coli concentration was significantly higher in the HDO groups than in the without HDO groups among the groundwater users and among the households that use groundwater only for bathing. However, such a significant difference was not observed among those households that drank only groundwater, but bathed in all types of water. In this study, 25% of the households did not have PWA. Among these households, the E. coli concentration was significantly higher in households with HDO than those without HDO. Among the households without PWA, the likelihood of HDO increased by three-fold with a log unit increase in E. coli concentration (Table 3). However, such an association was not significant among households with PWA. The association of the E. coli concentration with HDO under different conditions indicates that the microbial contamination of groundwater could be an important factor aggravating diarrhoea in the study area; this association was prominent when PWA was lacking.

A review of studies on global outbreaks and epidemiology, and studies on pathogens and risk assessment demonstrated that consumers served by an untreated groundwater supply remain at risk for enteric diseases.[35] Previous risk assessment studies on enteric infections from a variety of pathogens in groundwater estimated high risk and advised proper risk management strategies.[36] The only risk assessment study in the Kathmandu Valley[31] estimated a high risk of diarrhoea from groundwater and suggested the implementation of risk reduction programmes. Comparable with this finding of a high E. coli concentration among the ‘bathing only in groundwater’ group, Shrestha et al.[31] estimated a high risk of diarrhoea via the bathing pathway. During swimming and bathing, people usually accidentally ingest water and children ingest higher amounts than adults.[37] Bathers in polluted seawater are exposed to an increased risk of gastrointestinal infections.[38] These outcomes have opened an avenue for researching the role of bathing in disease transmission, especially among young children. Overall, with a huge proportion of residents relying heavily on shallow groundwater, the findings of this study strongly indicate a potential risk from contaminated groundwater to public health, as well as suggesting the immediate application of groundwater pollution control approaches.

The biggest limitation of the present study is the lack of microbial data for other water sources. However, the observational data were compared with published data of piped, stored[39] and stone spout water.[40] The mean (558) and maximum (6867) values of E. coli (MPN/100 mL) in groundwater are higher than those in piped (mean=44; maximum=488), stone spout (mean=107; maximum=500) and stored (mean=301; maximum=2420) water. This comparison indicates that groundwater is a highly polluted major water source. However, it is strongly recommended that the microbial qualities of other water sources, together with groundwater, should be analysed to obtain a more realistic picture in future studies. The second limitation is the use of the faecal indicator bacterium, E. coli; the use of enteropathogens is recommended for quantitative microbial analyses, even though it was not feasible in this study due to limited resources and time. Data on pathogens such as protozoa are being accumulated in continuing research,[13] and further analyses will be reported in the future. The third limitation is the moderate regression accuracy (R2=0.41) of the interpolation method, which could be due to the small sample size. Small sample size is thought to be a factor that leads to reduced accuracy. Therefore, using a larger sample size in future studies is recommended. Moreover, the recall period of HDO was 1 month in this study; this may have led to a recall bias. A previous study showed a 42–44% drop in diarrhoea reporting when the recall period was only 1 week.[40] In this study, it is highly likely that the respondents might have forgotten the frequency of diarrhoea, which might have resulted in an underestimation of HDO. In addition, single water sampling instead of daily or weekly sampling to characterize water quality in the wet season is another limitation of this study. A previous study[12] showed consistency in the E. coli detections and concentrations in wet seasons of different years and showed significant variations between dry and wet seasons in a single year. Despite these limitations, this study is an example of an integration of environmental with social data, which is especially important for developing countries that face financial restraints to conducting wider scale and interdisciplinary studies.

Conclusions

This study successfully integrated environmental data with survey data to perform an interdisciplinary analysis. With the application of the geostatistical interpolation technique, the observed E. coli concentration of groundwater samples in the wet season of 2009 was estimated for 35 survey points that had survey data of the same season and year. The E. coli concentration was significantly higher in the HDO group compared with the without HDO group among groundwater users and among households that use only groundwater for bathing. A quarter of the households still did not have PWA on their premises in the urban area, and these households were more likely to use groundwater, as well as relying only on groundwater for drinking and bathing purposes. The increased risk of HDO with the increasing E. coli concentration of the groundwater among households without PWA suggests the need for further studies on interventions to improve groundwater quality and their effect on reducing HDO. The interpolation method and interdisciplinary approach used in this study are relevant to other developing countries with similar settings.