H

Ba

b

a

ARRA

KSRHD

I

dvwsrwah22

map

v

1d

International Journal of Hygiene and Environmental Health 214 (2011) 162–166

Contents lists available at ScienceDirect

International Journal of Hygiene andEnvironmental Health

journa l homepage: www.e lsev ier .de / i jheh

ealth impact of small-community water supply reliability

atsirai Majurua, M. Michael Mokoenaa, Paul Jagalsa,∗, Paul R. Hunterb

Department of Environmental Health, Faculty of Science, Tshwane University of Technology, Pretoria, South AfricaSchool of Medicine, Health Policy and Practice, University of East Anglia, Norwich, UK

r t i c l e i n f o

rticle history:eceived 18 August 2010eceived in revised form 16 October 2010ccepted 28 October 2010

eywords:mall-community water supplyeliabilityousehold water treatmentiarrhoea

a b s t r a c t

There is still debate and uncertainty in the literature about the health benefits of community water supplyinterventions. This paper reports on a changing incidence of self-reported diarrhoea associated withthe implementation of two community water supplies. We conducted prospective weekly recording ofdiarrhoeal disease in three communities. Two of the communities were scheduled to receive an improvedwater supply and one was expected to continue to rely on an unimproved source during the study period.Data of self-reported diarrhoea was collected from each participating household on a weekly basis for upto 56 weeks, of which some 17 weeks were prior to implementation of the new water supply systems. Datawas modelled using Generalized Estimating Equations (GEE) to account for possible clustering withinhouseholds and within villages. For the two intervention communities in the study, the incidence rateratio (IRR) for all ages after the intervention was 0.43 (95% CI 0.24–0.79) when compared to the controlcommunity (who did not receive an intervention), implying a 57% reduction of diarrhoea. Both of thenew water systems were unreliable, one not operating on 4 weeks and the other on 16 weeks. The morereliable of the two intervention systems was also associated with less illness than in the least reliable

system (IRR = 0.41, 95% CI 0.21–0.80). We also noted anecdotal reports that during supply failures inthe new systems some people were starting to use household water treatment. The implementation ofimproved water systems does appear to have been associated with a reduction of diarrhoeal disease inthe communities. However the health impact was most obvious in the community with the more reliablesystem. Further research needs to be done to determine whether public health gains from community

ns ca

water supply interventioduring supply failures.

ntroduction

The World Health Organisation has estimated that 94% ofiarrhoea cases can be prevented through environmental inter-entions, amongst them increasing the availability of safe drinkingater (Bartram and Gordon, 2008). Yet despite general consen-

us on the importance of safe drinking water, some 272 millionural dwellers in sub-Saharan Africa still lack access to an improvedater supply (Rural Water Supply Network, 2009). It is in these

reas that one of the key millennium goals of reducing by onealf the proportion of people without access to safe water by015 is very unlikely to be met (Montgomery and Elimelech,007).

While problems with access to safe drinking water for rural com-unities are not restricted to developing country settings but are

lso an issue for most western countries (Hunter et al., 2009a),roblems related to access to improved safe drinking water are

∗ Corresponding author at: Department of Environmental Health, Tshwane Uni-ersity of Technology, Private Bag X680, Pretoria 0001, South Africa.

E-mail address: jagalsp@tut.ac.za (P. Jagals).

438-4639/$ – see front matter © 2010 Elsevier GmbH. All rights reserved.oi:10.1016/j.ijheh.2010.10.005

n be leveraged by occasional use of household water treatment (HWT)

© 2010 Elsevier GmbH. All rights reserved.

particularly acute for rural communities in developing countries.These communities traditionally have had to obtain their drinkingwater from untreated surface sources, often situated some distanceaway from their home (Hemson, 2007). In rural South Africa, wherewater supply infrastructure has been largely rudimentary or non-existent, small-community water supply systems have become thecommon mode of water supply (Momba et al., 2008). In these areas,water supply infrastructure is usually in the form of shared facili-ties such as communal taps (Peter-Varbanets et al., 2009). Thesesupplies have varying reliability, with an unacceptable propor-tion of apparently improved supplies being non-operational wheninspected (Rietveld et al., 2009).

One of the major problems facing the drive to provide access to,and maintain continuity of improved drinking water supplies hasbeen the lack of strong evidence of the effects of system failure onpublic health gains from these supplies (Lee and Schwab, 2005).There is a great need for effective studies of the public health gains

achievable from community safe drinking water supply interven-tions but more importantly there is also a need to investigate theeffect of the reliability of these systems and assess the impact thatpoor reliability has on the public health objectives (Hunter et al.,2009b).

ene an

twts

M

VPttws

S

pbpswatdtotfpwotd

npcacect

batwf

E

Tchhtm

S

m

B. Majuru et al. / International Journal of Hygi

The objective of the study presented here was to determinehe effect of the reliability of newly established small-communityater supply systems on community health gains expected from

hese interventions, using diarrhoea incidence as an indicator. Thisubmission presents evidence of this effect.

ethods

The study was conducted in a remote and rural area in thehembe District, situated in the north-eastern parts of Limpoporovince in South Africa. Diarrhoeal incidence for children underhe age of five in the area was 224.3 per 1000 in 2002, comparedo the national average of 133.4 per 1000. At the time, the districtas also below the national averages in terms access to safe water

upply (Gundry et al., 2009).

tudy communities

The study was done across three small communities in closeroximity, two of which initially were conveniently selectedecause of their being provided, by the local service provider, with aiped water supply during the study period. At the beginning of thetudy, their most consistent water supplies consisted of untreatedater from a small nearby river and its tributaries and when avail-

ble, water from drill-wells fitted with hand-pumps (often subjecto breakdown) and a water tanker service often prone to non-elivery. These rudimentary services were unreliable to the extenthat the two communities were collecting water for drinking andther domestic purposes from the open untreated sources for morehan half of the time. After the first 17 weeks for one and 18 weeksor the other village, these unreliable water supply services (hand-umps and tanker) were each replaced with a small-communityater supply system, distributing potable water through a network

f communal taps throughout each community. These new sys-ems were expected to be reliable by continuously supplying saferinking water.

The third community was studied as a reference group as itever had a formal water supply, nor did it receive any during theeriod of study, with households continuing their usual practice ofollecting and using untreated water from the nearby small rivernd its tributaries. This community then served as a reference foromparing the data of the other two intervention communities,specially in terms of any effects the data from the interventionommunities might show, as well as any possible effects shouldhe new systems fail during the remainder of the study period.

Shortly after commissioning the two small systems, the suppliesecame intermittent, mainly because of failures in maintenancend operation. During these non-operational periods, the interven-ion communities – because their initial services were dismantledhen their new services were established – reverted to using water

rom the same small river system as the reference community.

thics

Ethics approval was obtained from the Tshwane University ofechnology Research Ethics Committee. Consent to work with theommunities was granted by traditional leaders in the area. At theouseholds, informed and formal consent was obtained from theousehold head who then assigned a respondent suitably in a posi-ion to provide information related to the study, usually the senior

atriarch in the household.

ampling

In these remote communities, households did not have for-al addresses. To facilitate random sampling of the study groups

d Environmental Health 214 (2011) 162–166 163

within the selected communities, a GPS (global positioning sys-tem) address (a unique waypoint based on its location coordinates)was assigned to each study household using Garmin 60 Csx® GPSdevices. These addresses were used to randomly sample house-holds for participation as well as uniquely mark and identify allcollected data, its capturing and analysis for the particular house-hold throughout the study.

Data collection

The study was conducted over a period of 56 consecutive weeks(July 2007 to July 2008) and included an initial period of 17 weeksbefore the intervention (baseline data) and the rest of the weekscomprised the post-intervention period. All data collection sheets,i.e. structured questionnaires and symptom diaries were first com-piled in English then translated into the local language of TshiVendaand thereafter translated back to English to verify linguistic andcontextual accuracy.

Demographic and socio-economic data were collected dur-ing the baseline period by experienced fieldworkers throughstructured interviews with the respondents. Data on potentiallyconfounding variables were collected, including family size, genderand age as well as education, income, and type of housing.

Data on diarrhoea incidence were collected for each memberof each study household over the full period of 56 weeks, includ-ing the baseline and post-intervention periods. The case definitionfor diarrhoea was three or more loose stools within a 24-h period(WHO, year unknown). An episode of diarrhoea was considered tohave ended after 48 consecutive hours without symptoms (WHO,year unknown). Respondents were issued with symptom diarieson which they recorded the daily diarrhoea status for each mem-ber of the household. The fieldworkers trained respondents onhow to record, on a weekly basis, the occurrence of diarrhoea inthese diaries. The fieldworkers collected the diaries at the end ofeach week, double-checked and confirmed the recorded data withrespondents and other household members.

The fieldworkers also confirmed with the respondent, thehousehold’s water source for every week. This was done to iden-tify the periods of time the household would not have accessto an operable system (and safe drinking water) and thereforereverted to use water of poor quality from open sources. Duringthe post-intervention periods the fieldworkers, on a weekly basis,also interviewed the two water supply systems’ operators as wellas physically checked on the systems to record periods when thesystems were not operational.

Statistical analyses

Data analyses were done with SPSSTM 17.0. Comparison ofillness rates between intervention and non-intervention commu-nities were done using generalised estimating equations in SPSSto handle possible clustering within communities and householdsand repeat sampling from individual participants.

The water systems became operational in Intervention Com-munity 1 during week 18 (the first week of November 2007) andin Intervention Community 2 the following week. Analyses priorto the intervention were restricted to weeks 1–17 and then post-intervention analyses from weeks 19 to 56, making up a total of38 weeks of follow-up. Data from the first week after the interven-tion were excluded from the analyses as it was not clear when any

recorded diarrhoea episodes would have started around the timeof the intervention. There were a total of 20,148 person weeks offollow-up data post-intervention available of which 14,948 were inthe intervention communities and 5200 in the reference commu-nity.

164 B. Majuru et al. / International Journal of Hygiene and Environmental Health 214 (2011) 162–166

Table 1Household demographic and socio-economic comparison of the three communities.

InterventionCommunity 1

InterventionCommunity 2

ReferenceCommunity

p value (intervention versus referencecommunities)

Population samplesize (totalpopulation size)

45 (120) 38 (94) 31 (33)

Mean family size(standarddeviation)

5.7 (2.1) 4.8 (2.0) 4.9 (2.4) 0.356

Gender: % female 54 57 60 0.355Mean age(standarddeviation)

26 (21) 25 (24) 23 (21) 0.253

Mean income perhead (standarddeviation)

183 (175) 218 (227) 300 (383) 0.078

Median highestlevel of education

Grade 11 Grade 11 Grade 10 0.028

Mean number ofconstructeddwellings(standarddeviation)

1.3 (0.7) 1.3 (0.5) 1.5 (0.5) 0.308

Mean number of 1.7 (1.1) 2.1 (1.1) 1.9 (0.8) 0.917

twot

R

eclcwhrtrp(tuiiidc

ac5

ioawn1m

traditionaldwellings(standarddeviation)

To determine the effect that interruptions in supply had onhe incidence of diarrhoea, Generalised Estimation Equations (GEE)ere applied with community as a covariate ranked with number

f weeks in which the system was actually operational for each ofhe communities.

esults

Table 1 shows that for most of the demographic and socio-conomic data there was very little difference between the threeommunities. The exception was the level of education, which wasower in the Reference Community. Although statistically signifi-ant, the median level of education (one high-school grade lower)as not considered sufficient to regard this community as non-omogenous. Overall the three community groups were thereforeegarded as being sufficiently homogenous to limit considera-ion of confounding. Furthermore, the pre-intervention illnessates were very similar in the three communities in the baselineeriod in the intervention communities and reference communityTable 2). Indeed the baseline incidence rate (IRR) ratio betweenhe intervention and non-intervention communities was almostnity (IRR = 1.002, 95% CI 0.581–1.727). The incidence rate ratio

s the ratio of the incidence rates in the intervention and non-ntervention communities with IRR > 1 indicating higher incidencen the intervention communities and IRR < 1 indicating lower inci-ence in the intervention communities compared to the controlommunity.

After the intervention, the IRR in the intervention communitieslso showed a marked reduction in illness compared to the controlommunities (IRR = 0.43, 95% CI 0.24–0.79, p = 0.006), indicating a7% reduction in illness.

There were two post-intervention breakdowns of the systemn Intervention Community 1, amounting to non-operational peri-ds of four weeks. Intervention Community 2 had five breakdowns

fter intervention, amounting to non-operational periods of 16eeks with the longest non-operational period in this commu-ity being seven weeks. Consequently the system in Communitywas classified as a more reliable supply system and the Com-unity 2 system as a less reliable supply system. As can be seen

from Table 2 the crude incidence rate in the low reliability supplysystem was much lower than in the very low reliability supply sys-tem. This strongly supports the suggestion that system reliabilityindeed had an impact on the reduction of diarrhoeal incidence andconsequently the public health benefits in the study communities.

Discussion

To our knowledge this study is the first epidemiological inves-tigation of the impact of community water interventions that hasattempted to quantify the impact that unreliability may have onhuman health. The results of the study show reductions in diar-rhoeal morbidity of about 57%, which is higher than has beenreported in other studies and substantially greater than one wouldnormally expect from reporting bias seen in non-blinded inter-vention studies (Hunter, 2009). This overall reduction is higherthan the 25% reduction in diarrhoea due to improved water sup-ply reported in a systematic review by Fewtrell et al. (2005) forinstance. Consequently this study provides important evidence ofthe value of small-community water supply interventions. How-ever, the study was conducted over just three communities towhich interventions were not randomly allocated. However, thethree communities were very similar in all socio-economic datawith just once exception-level of education. Importantly also thepre-intervention diarrhoea incidence rates were very similar acrossthe three communities. Nevertheless, it is plausible that some otherevent may have effected diarrhoeal disease incidence in the com-munities, but nothing was obvious to these investigators.

Rates of reported illness in all three communities declined withtime, even in the reference community. This is not surprising asillness rates decline in all prospective studies of diarrhoeal dis-ease, almost certainly due to “reporting fatigue” (Feikin et al., 2010;Strickland et al., 2006). Another potential source of bias could beunder-reporting of diarrhoea, especially for the older age-groups,

as the study relied on the matriarch of the household to recorddiarrhoeal incidence. Zafar et al. (2009) suggested that matriarchstended to keep a closer focus on younger children, and thus notkeep a consistent record for the older children and other adults inthe household.

B. Majuru et al. / International Journal of Hygiene and Environmental Health 214 (2011) 162–166 165

Table 2Crude incidence rates by community.

Intervention Community 1 Intervention Community 2 Reference Community

Persons with data 234 173 146Pre-interventionPerson weeks of follow-up 3735 2592 1765Number of episodes of diarrhoea 55 42 27Crude incidence rate/episodes per person yr (95% confidence intervals) 0.77 (0.58–1.00) 0.85 (0.61–1.14) 0.80 (0.53–1.16)Post-intervention

.17 (0

lsinitAfctos

hbpvsp(tttthitwwotms

tfptcwmghgAmaic2ao

Person weeks of follow-up 8696Number of episodes of diarrhoea 28Crude incidence rate/episodes per person year (95% confidence intervals) 0

One of the most notable findings in this study has been the muchower illness reduction in the community with the system clas-ified as “very low reliability”. Indeed the observed reduction inllness in this community compared to the control community isot that much different than what one may expect due to report-

ng bias because participants were not blinded to which study armhey belonged to. Results of a recent Quantitative Microbial Riskssessment indicated that interruption in water supply for even a

ew days can negate many of the health benefits derived from alean water supply (Hunter et al., 2009b). Whilst the sample size inhis present study was too small to test this hypothesis, the trendf effectiveness with reliability as found in this current study doesuggest that this may be true.

Following the previous analysis (Hunter et al., 2009b), one mayave expected an even bigger decline in the estimated public healthenefit given the time that the systems were non-functioning. Oneossible explanation for the finding that illness rates in the inter-ention communities were not higher even with the failures inupply could be due to residual high levels of immunity in theopulations from the time before the water supply improvementsSwift and Hunter, 2004; Frost et al., 2005). Another possibility ishat once people have access to an improved water supply theyake more care about treating their drinking water on the occasionshat their supply fails. Indeed we are aware of anecdotal evidencehat within the intervention communities some people introducedome treatment of raw water during non-functioning periods. This

s an important issue. In the developed world, it is often the casehat during periods of inadequate water quality, a boil-water noticeill be issued. It is plausible that in developing countries householdater treatment will not be practiced effectively when people have

nly an unimproved supply such as untreated open water. But ashey become used to an improved supply (potable tap water), they

ight introduce household water treatment during periods of theirystem not functioning.

This suggestion that people may treat raw water more effec-ively when they have to revert to this during temporary supplyailures compared to people living with an unimproved supply is aotentially very important observation. If this observation is true,his could act as a means of leveraging the public health gains asso-iated with community drinking water interventions. For example,e are increasingly realising that many household water treat-ent (HWT) interventions are not delivering their promised health

ains over time. A recent analysis of the literature suggested thatousehold water disinfection was not associated with any healthain after the first few months post-intervention (Hunter, 2009).lthough there may be several reasons, it is probable that theain reason is that people simply stop using the technology reli-

bly with time. Indeed recent research has shown that many HWT

nterventions such as solar disinfection may have pitifully smallontinued use one or two years post-intervention (Mäusezahl et al.,009; Sobsey et al., 2008). We would argue that when installingn improved water supply in a community people are educatedn the importance of not simply reverting back to consumption of

6259 520749 62

.11–0.24) 0.41 (0.30–0.54) 0.62 (0.48–0.80)

raw water if/when that supply fails. In this context targeted use ofHWT or boiling for the duration of the supply failure may then addsubstantially to the health gains that can accrue from access to animproved drinking water supply.

Conclusion

We have shown that community water interventions can beassociated with a significant reduction in diarrhoeal disease muchgreater than what is likely to be explained by reporting bias seen inunblinded studies. We have further shown that most of the healthgains may be lost in those systems that have poor reliability. How-ever, there are suggestions that following implementation of a newcommunity water system some people at least use HWT when thenew system fails temporarily.

Conflicts of interest

PRH is chair of the executive board of the Institute of PublicHealth and Water Research, chair of the Science Advisory Com-mittee for Suez Environment and has done consultancy work forDanone beverages.

Acknowledgements

This work was funded by the South African Water ResearchCommission as well as the National Research Foundation. Theauthors also thank the people in the communities in Vhembe forparticipating in the study.

References

Bartram, J., Gordon, B., 2008. The global challenge of water quality and health. WaterPract. Technol. 3 (4), doi:10.266/WPT.200890.

Feikin, D.R., Audi, A., Olack, B., Bigogo, G.M., Polyak, C., Burke, H., Williamson, J.,Breiman, R.F., 2010. Evaluation of the optimal recall period for disease symptomsin home-based morbidity surveillance in rural and urban Kenya. Int. J. Epidemiol.39, 450–458.

Fewtrell, L., Kaufmann, R.B., Kay, D., Enanoria, W., Haller, L., Colford, J.M., 2005.Water, sanitation, and hygiene interventions to reduce diarrhoea in less devel-oped countries: a systematic review and meta-analysis. Lancet Infect. Dis. 5,42–52.

Frost, F.J., Roberts, M., Kunde, T.R., Craun, G., Tollestrup, K., Harter, L., Muller, T., 2005.How clean must our drinking water be: the importance of protective immunity?J. Infect. Dis. 191, 809–814.

Gundry, S.W., Wright, J.A., Conroy, R.M., du Preez, M., Genthe, B., Moyo, S., Mutisi,C., Potgieter, N., 2009. Child dysentery in the Limpopo Valley: a cohort study ofwater, sanitation and hygiene risk factors. J. Water Health 7, 259–266.

Hemson, D., 2007. ‘The toughest of chores’: policy and practice in children collectingwater in South Africa. Policy Futures Educ. 5, 315–326.

Hunter, P.R., 2009. Household water treatment in developing countries: compar-ing different intervention using meta-regression. Environ. Sci. Technol. 43,8991–8997.

Hunter, P.R., Pond, K., Jagals, P., Cameron, J., 2009a. An assessment of the costs andbenefits of interventions aimed at improving rural community water suppliesin developed countries. Sci. Total Environ. 407, 3681–3685.

Hunter, P.R., Zmirou-Navier, D., Hartemann, P., 2009b. Estimating the impact onhealth of poor reliability of drinking water interventions in developing countries.Sci. Total Environ. 407, 2621–2624.

1 ene an

L

M

M

M

P

R

R

World Health Organization. Diarrhoea morbidity rate in children aged 0–4

66 B. Majuru et al. / International Journal of Hygi

ee, E.J., Schwab, K.J., 2005. Deficiencies in drinking water distribution systems indeveloping countries. J. Water Health 3, 109–127.

äusezahl, D., Christen, A., Pacheco, G.D., Tellez, F.A., Iriarte, M., Zapata, M.E., Ceval-los, M., Hattendorf, J., Cattaneo, M.D., Arnold, B., Smith, T.A., Colford, J.M., 2009.Solar drinking water disinfection (SODIS) to reduce childhood diarrhoea in ruralBolivia: a cluster-randomized, controlled trial. PLoS Med. 6 (8), e1000125.

omba, M.N.B., Obi, C.L., Thompson, P., 2008. Improving disinfection efficiency onsmall drinking water treatment plants. Water Research Commission (WRC),Report number 1531/1/08. ISBN: 978-1-77005-683-1.

ontgomery, M.A., Elimelech, M., 2007. Water and sanitation in developing coun-tries: including health in the equation. Environ. Sci. Technol. 41, 17–24.

eter-Varbanets, M., Zurbrugg, C., Swartz, C., Pronk, W., 2009. Decentralised systemsfor potable water and the potential technology. Water Res. 43, 245–265.

ietveld, L.C., Haarhoff, J., Jagals, P., 2009. A tool for technical assessment of ruralwater supply systems in South Africa. Phys. Chem. Earth 34, 43–49.

ural Water Supply Network, 2009. Myths of the Rural Water Supply Sector, RSWNPerspectives No. 4.

d Environmental Health 214 (2011) 162–166

Sobsey, M.D., Stauber, C.E., Casanova, L.M., Brown, J.M., Elliott, M.A., 2008. Point ofuse household drinking water filtration: a practical, effective solution for pro-viding sustained access to safe drinking water in the developing world. Environ.Sci. Technol. 42, 4261–4267.

Strickland, M.J., Crawford, J.M., Shen, L., Wilkins, J.R., 2006. Time-dependentrecordkeeping fatigue among youth completing health diaries of unintentionalinjuries. J. Saf. Res. 37, 487–492.

Swift, L., Hunter, P.R., 2004. What do negative associations between potential riskfactors and illness in analytical epidemiological studies really mean? Eur. J.Epidemiol. 19, 219–223.

years, http://www.who.int/ceh/indicators/0 4diarrhoeamorb.pdf (accessed July2010).

Zafar, S.N., Luby, S.P., Mendoza, C., 2009. Recall errors in a weekly survey of diarrhoeain Guatemala: determining the optimal length of recall. Epidemiol. Infect. 138,264–269.