An ecological quantification of the relationships between water, sanitation and infant, child, and maternal mortality.

0.26 MB·English

by  Cheng, June J

Chengetal.EnvironmentalHealth2012,11:4
http://www.ehjournal.net/content/11/1/4
RESEARCH Open Access
An ecological quantification of the relationships
between water, sanitation and infant, child, and
maternal mortality
June J Cheng1,2*, Corinne J Schuster-Wallace2,3,4, Susan Watt2,5, Bruce K Newbold3,4 and Andrew Mente6,7
Abstract
Background: Water and sanitation access are known to be related to newborn, child, and maternal health. Our
study attempts to quantify these relationships globally using country-level data: How much does improving access
to water and sanitation influence infant, child, and maternal mortality?
Methods: Data for 193 countries were abstracted from global databases (World Bank, WHO, and UNICEF). Linear
regression was used for the outcomes of under-five mortality rate and infant mortality rate (IMR). These results are
presented as events per 1000 live births. Ordinal logistic regression was used to compute odds ratios for the
outcome of maternal mortality ratio (MMR).
Results: Under-fivemortalityratedecreasedby1.17(95%CI1.08-1.26)deathsper1000,p<0.001,foreveryquartile
increaseinpopulationwateraccessafteradjustmentsforconfounders.Therewasasimilarrelationshipbetweenquartile
increaseofsanitationaccessandunder-fivemortalityrate,withadecreaseof1.66(95%CI1.11-1.32)deathsper1000,p<
0.001.ImprovedwateraccesswasalsorelatedtoIMR,withtheIMRdecreasingby1.14(95%CI1.05-1.23)deathsper1000,
p<0.001,withincreasingquartileofaccesstoimprovedwatersource.Thesignificanceofthisrelationshipwasretained
withquartileimprovementinsanitationaccess,wherethedecreaseinIMRwas1.66(95%CI1.11-1.32)deathsper1000,p
<0.001.Theestimatedoddsratiothatincreasedquartileofwateraccesswassignificantlyassociatedwithincreased
quartileofMMRwas0.58(95%CI0.39-0.86),p=0.008.Thecorrespondingoddsratioforsanitationwas0.52(95%CI0.32-
0.85),p=0.009,bothsuggestingthatbetterwaterandsanitationwereassociatedwithdecreasedMMR.
Conclusions: Our analyses suggest that access to water and sanitation independently contribute to child and
maternal mortality outcomes. If the world is to seriously address the Millennium Development Goals of reducing
child and maternal mortality, then improved water and sanitation accesses are key strategies.
Keywords: Water, Sanitation, Maternal health, Infant health, Child health, Millennium development goals
Background safety related to environmental chemicals adds to the
AWorldHealthOrganization(WHO)reportfoundthat considerablediseaseburden[3,4].Althoughthesereports
almost one tenth of the global disease burden could be andothershavecalculateddiseaseburdenrelatedtopoor
preventedbyimprovingwatersupply,sanitation,hygiene watersupply,sanitation, andhygiene therehasnotbeen
and management of water resources [1]. Another esti- a quantification of the improvement in health related
matereportsthat4.0%ofalldeathsand5.7%oftotaldis- outcomesduetoimprovementsinwatersupplyandsani-
ability-adjusted life years can be attributed to water, tation. Quantification would further support the impor-
sanitation,andhygiene[2].Inaddition,waterqualityand tance of investing in water and sanitation as a
development strategy and would provide a mechanism
formonitoringprogress.
*Correspondence:[email protected] Worldwide, 1.4 million children die each year from
1PublicHealthandPreventiveMedicineResidencyProgram,Departmentof
preventable diarrheal diseases and some 88% of diarrhea
ClinicalEpidemiologyandBiostatistics,McMasterUniversity,Hamilton,ON,
Canada cases are related to unsafe water, inadequate sanitation,
Fulllistofauthorinformationisavailableattheendofthearticle
©2012Chengetal;licenseeBioMedCentralLtd.ThisarticleispublishedunderlicensetoBioMedCentralLtd.ThisisanOpenAccess
articledistributedunderthetermsoftheCreativeCommonsAttributionLicense(http://creativecommons.org/licenses/by/2.0),which
permitsunrestricteduse,distribution,andreproductioninanymedium,providedtheoriginalworkisproperlycited.
Chengetal.EnvironmentalHealth2012,11:4 Page2of8
http://www.ehjournal.net/content/11/1/4
or insufficient hygiene [2,5]. In addition, there are 860 Indicator 7·9: proportion of population using an
000 preventable child deaths per year due to malnutri- improved sanitation facility [12].
tion. Childhood underweight, defined as weight two
standard deviations below the mean, is implicated in Despite progress, the WHO/UNICEF Joint Monitoring
35% of all deaths of children under the age of five Programme reported that at the current rate, the world
worldwide, or about 70 000 deaths per year [5,6]. An will miss the MDG sanitation target by 13%, with 2.7
estimated 50% of this underweight or malnutrition is billion people lacking basic sanitation [11]. Even if the
associated with repeated diarrhea or with intestinal target is met, there will still be 1.7 billion people with-
nematode infections as a result of unsafe water, inade- out access to improved sanitation [11]. More positively,
quate sanitation or insufficient hygiene [5]. Children the world will meet the MDG water target at the cur-
also bear 54% of the burden of illness from environmen- rent rate [11]. However, even if the goal is met by 2015,
tal exposure to chemicals–estimated to be 4.9 million there will still be 672 million people without access to
deaths in total worldwide–some of which is caused by improved water sources [11].
exposure through contaminated water [3]. Although links have been made between water and
Pregnant women face a similar, equally dire situation, sanitation and newborn, child, and maternal health,
particularlybecauseoftheirvulnerabilitytoanemia,vita- there have not been studies quantifying these relation-
min deficiency, trachoma and hepatitis, all of which can ships globally using country-level data [13]. Will, for
leadtoincreasedmorbidityandmortality[7].Theprovi- example, the failure to provide water and sanitation jeo-
sion of safe waterfor medical purposesto treat such ill- pardize the achievement of MDG four and five? In this
ness can improve newborn and child health in addition paper, we take the first steps towards quantifying the
to maternal health[8]. Currently,health centres provid- relationship between water and sanitation and infant,
ing maternal and delivery care can expose women to child, and maternal mortality: How much does safe
unsafe water, poor sanitation and poor management of water and sanitation contribute to these mortality out-
medical waste: 15% of all maternal deaths are caused by comes on a global scale? This quantification would
infectionsinthe6weeksafterchildbirthandhavemainly further enforce the importance of investing in water and
been found to be due to unhygienic practices and poor sanitation as a development strategy.
infectioncontrolduringlabouranddelivery[9].
Despite the importance of water and sanitation and Methods
the availability of interventions, 2.6 billion people in the Data source
world currently lack access to basic sanitation, while 884 Country level data were collected for 193 countries from
million people lack access to safe drinking water [10,11]. several organizations including the World Health Orga-
In order to promote access to improved water and sani- nization Statistical Information System [14], United
tation, the Millennium Development Goals (MDG), Nations Children’s Fund Childinfo [15], and World
including access to safe drinking water and basic sanita- Bank Open Data [16].
tion (target 7 C), infant and child health (MDG 4), and
maternal health (MDG 5) guide development and plan- Outcome variables
ning policies [12].The MDGs, along with suggested indi- Four key outcome variables were explored in this analy-
cators for measure progress, are as follows: sis: under-five mortality rate, IMR, MMR, and propor-
tion of under-five deaths due to diarrhea. Under-five
MDG 4: Reduce child mortality mortality rate is defined as the probability of a child
Target 4A: reduce by two thirds, between 1990 born in a specific year or period dying before reaching
and 2015, the under-five mortality rate the age of five per 1000 live births, if subject to age-spe-
Indicator 4·1: under-five mortality rate cific mortality rates of that period [17]. The proportion
Indicator 4·2: IMR of under-five deaths due to diarrhea is estimated by the
MDG 5: Improve maternal health WHO based on the causes of death that are entered on
Target 5A: reduce by three-quarters, between the medical certificate of cause of death in countries
1990 and 2015, the MMR and recorded by the civil (vital) registration systems. For
Indicator 5·1: MMR the analyses, the concept of the ‘underlying cause of
MDG 7: Ensure environmental sustainability death’ is defined by ICD. In countries with incomplete
Target 7C: halve, by 2015, the proportion of peo- or no civil registration, causes of death are those
ple without sustainable access to safe drinking reported as such in epidemiological studies that use ver-
water and basic sanitation bal autopsy algorithms to establish cause of death [18].
Indicator 7·8: proportion of population using an IMR is defined as the probability of dying between birth
improved drinking water source and exactly 1 year of age, expressed per 1 000 live births
Chengetal.EnvironmentalHealth2012,11:4 Page3of8
http://www.ehjournal.net/content/11/1/4
[19]. MMR is defined as number of maternal deaths per least once during pregnancy, as a percentage of live
100 000 live births during a specified time period, births in a given time period 2000-2009 [14]; and deaths
usually 1 year [20]. All outcome variables were obtained (1 000’s) attributable to diarrheal diseases 2002 [14].
from the World Health Survey 2010, and the time per-
iod used for our analyses were from the year 2008. Missing data
Due to the limited available sample size of 193 coun-
Independent variables tries, attempts were made to preserve the power of the
The proportion of population with access to improved study. Percentage of missing data ranged from 0 to 32%
water source and the proportion of population with (62/193). Missing data were imputed using a Multiple
access to improved sanitation, as defined by the WHO/ Imputation by Chained Equation (MICE) approach to
UNICEF Joint Monitoring Programme, are our key pre- obtain a full dataset [21]. See Table 1 for summary sta-
dictors of interest. Data were obtained for the year 2008. tistics of all raw variables in the study, including the
The following are detailed definitions of improved number of missing values.
access to water and sanitation:
Statistical analyses
Access to safe drinking water is measured by the Linear regression and ordinal logistic regression models
percentage of the population using improved drink- were used. The generic model for the linear regression
ing-water sources. is as follows:
(cid:129) Improved drinking water source is a source
y =β x +β x +...+β x +ε, i=1,...,n,
that, by nature of its construction, adequately i 1 i1 2 i2 p ip i
protects the water from outside contamination, where y is the outcome variable, b’s are the parameter
in particular from fecal matter. Common coefficients, and x’s are the predictor variable or covari-
examples: ates, and ε an error term.
○ Piped household water connection The generic model for the ordinal logistic regression is
○ Public standpipe as follows:
○ Borehole
○ Protected dug well Logit[P(y≤j)]=αj+β1x1+β2x2+...+βpxp+εi, j=1,...J−1,
○ Protected spring
where b’s describe the effect of X on the log odds of
○ Rainwater collection
response (y) in category j or below. This model assumes
Access to sanitation is measured by the percentage
an identical effect of x for all j-1. In other words, it
of the population using improved sanitation facilities.
assumes proportional odds. This model was chosen for
(cid:129) Improved sanitation includes sanitation facil-
its potential for greater power and easier interpretations
ities that hygienically separate human excreta
compared to multi-category logit models [22].
from human contact.
In order to satisfy the underlying assumption of linear
(cid:129) Access to basic sanitation is measured against
regression, positively skewed variables including GNI,
the proxy indicator: the proportion of people
total fertility per woman, under-five mortality rate, and
using improved sanitation facilities (such as
IMR were log-transformed to follow a symmetrical dis-
those with sewer connections, septic system con-
tribution. Where log transformation was not possible,
nections, pour-flush latrines, ventilated improved
quartiles were created: MMR, proportion with access to
pit latrines and pit latrines with a slab or covered
improved water source, proportion with access to
pit)
improved sanitation, adult literacy rate, and percent
(cid:129) Shared sanitation facilities are otherwise-accep-
births attended by skilled health personnel.
table improved sanitation facilities that are
For continuous outcomes–log-transformed under-five
shared between two or more households. Shared
mortality rate and log-transformed IMR–linear regres-
facilities include public toilets and are not con-
sion model was used with estimates subsequently
sidered improved [11].
retransformed to their natural units. For categorical out-
comes–MMR quartiles and percent under-five mortality
Other covariates included in the analyses are: Gross
rate due to diarrhea quartiles–the ordinal logistical
National Income (GNI) 2008 [16]; total fertility rate per
regression model was used. To test the assumption of
woman 2008 [14]; WHO world region, adult literacy
proportional odds, the Stata Brant test for parallel
rate 2000-2008 [14]; percentage of births attended by
regression was performed [23]. Predictor variables and
skilled health personnel 2000-2008 [14]; percentage of
confounders included in each analysis were based on
women who used antenatal care provided by skilled
past literature and expert opinion. Highly correlated
health personnel for reasons related to pregnancy at
Chengetal.EnvironmentalHealth2012,11:4 Page4of8
http://www.ehjournal.net/content/11/1/4
Table 1Summarytable ofall variables used in regression analyses
Variable Mean Standard 25th 50th 75th Minimum Maximum Numberof
Deviation percentile percentile percentile value value missingvalues
%accesstoimprovedwater 85.99 17.00 80 94 100 30 100 24
source
%accesstoimprovedsanitation 71.14 30.18 48 84 98 9 100 23
Under-fivemortalityrate(per1 48.30 55.01 10 23 69 2 257 0
000livebirths)
%under-fivemortalitydueto 7.15 7.26 1 5 13 0 29 0
diarrhealdisease
IMR(per1000livebirths) 33.80 33.38 9 21 53 1 165 0
MMR(per100000livebirths) 206.60 307.66 8 44 317 0 1600 24
GNI 10790.76 16203.34 1045 3730 11940 140 87070 9
Fertilityperwoman 2.88 1.44 1.80 2.40 3.85 1.20 7.10 1
Adultliteracyrate 80.77 19.80 71 88.50 97 26 100 57
Deathsduetodiarrhea(1000’s) 8792.47 33797.29 0 300 5000 0 402200 7
%birthattendedbyskilled 79.80 25.62 61 94.5 100 6 100 15
healthpersonnel
%usingantenatalcare(atleast 85.37 16.56 80 91 97 16 100 62
1visit)
Regionsoftheworldisacategoricalvariable.Itsvaluesaredefinedasthefollowing:1-AfricanRegion;2-Europe;3-EasternMediterranean;4-Americas;5-
SoutheastAsia;6-WesternPacific
covariates as defined by r > 0.8 were not included in the under-five mortality rate, decreased odds of under-five
same equation. Akaike Information Criteria was used as mortality due to diarrhea, decreased IMR, and decreased
an additional aid to help determine the optimal regres- odds of MMR in our analyses (see Table 2 for univari-
sion equation. All statistical analyses were performed able models and Table 3 for multivariable models and
using Stata 10.0. results). Under-five mortality rate was seen to decrease
In all cases, adjusted and unadjusted models were esti- by 2.25 (95%CI 2.05-2.50) deaths per 1000 with
mated. Unadjusted models included only the dependent increased water access in the unadjusted regression.
variable and water (sanitation) on the right hand side. After adjustments for GNI, fertility per woman, MMR,
Adjusted models accounted for developmental effects. and region of the world as potential confounders, this
Water and sanitation were considered in separate mod- decrease continued to be significant at p < 0.001, 1.17
els in this study. Due to considerable overlap between (95%CI 1.08-1.26) deaths per 1000. The estimated odds
the two variables, water and sanitation access often can- ratio that increased access to water was significantly
celled each other in effect in test regressions. Their associated with increased odds of under-five child mor-
simultaneous effect, therefore, was not explored. tality due to diarrhea was 0.20 (95%CI 0.14-0.28). This
A note about variable interpretations: from this point odds ratio changed to 0.46 (95%CI 0.30-0.70), p < 0.001
on GNI and fertility rate per woman should be under- in the adjusted model. IMR, in the unadjusted model,
stood to be log transformed values. In addition, the fol- decreased by 2.12 (95%CI 1.93-2.29) deaths per 1000
lowing variables are presented as quartiles: percent with increased access to improved water source. This
under-five mortality due to diarrheal diseases, MMR, relationship retained its significance in the multivariable
proportion of population with access to improved water analysis, and the decrease in IMR was 1.14 (95%CI 1.05-
source, proportion of population with access to 1.23) deaths per 1000, p = 0.001. The estimated odds
improved sanitation, adult literacy rate, deaths (in thou- ratio that increased water access was significantly asso-
sands) due to diarrhea, and percent births attended by ciated with increased MMR is 0.20 (95%CI 0.14-0.28).
skilled health personnel. In ordinal logistic regression, This odds ratio was 0.58 (95%CI 0.39-0.86), p = 0.008 in
odds ratios (OR’s) represent the odds of being in a the adjusted analysis.
higher quartile of the outcome variable compared to a Increasing access to improved sanitation was signifi-
lower quartile. cantly associated with decreased under-five mortality
rate, decreased odds of under-five mortality due to diar-
Results rhea, decreased IMR, and decreased odds of MMR in
In our analyses, increased access to improved water our analysis. Under-five mortality rate was seen to
sources was significantly associated with decreased decrease by 2.45 (95%CI 2.27-2.66) deaths per 1000 with
Chengetal.EnvironmentalHealth2012,11:4 Page5of8
http://www.ehjournal.net/content/11/1/4
Table 2Unadjusted Regressions
Coefficient(95%Confidenceinterval(CI)) p-value
Predictor:proportionofpopulationwithaccesstoimprovedwatersourcea
Under-fivemortalityrate(per1000livebirths) -2.25(-2.50,-2.05)b <0·001
%under-fivemortalityduetodiarrheaa OR:0·20(0·14,0·28)c <0·001
IMR(per1000livebirths) -2.12(-2.29,-1.93) <0·001
MMRb(per100000livebirths) OR:0·20(0·14,0·28) <0·001
Predictor:proportionofpopulationwithaccesstoimprovedsanitationa
Under-fivemortalityrate(per1000livebirths) -2.45(-2.66,-2.27) <0.001
%under-fivemortalityduetodiarrheaa OR:0.20(0.14,0.28) <0.001
IMR(per1000livebirths) -2.29(-2.48,-2.12) <0.001
MMRb(per100000livebirths) OR:0.14(0.10,0.20) <0.001
avariabledividedintoquartiles
bSampleinterpretation:under-fivemortalityrateisseentodecreaseby2.25(95%CI2.05,2.50)withincreasingquartileofproportionofpopulationwithimproved
wateraccess
cSampleinterpretation:theestimatedoddsratiothatincreasedpopulationquartileaccesstowaterissignificantlyassociatedwithincreasedoddsofunder-five
childmortalityduetodiarrheais0.20(95%CI0.14,0.28)
increasing access to sanitation quartile in the unadjusted with increasing access to improved sanitation. This rela-
regression. After adjustments for GNI, fertility per tionship retained its significance in the multivariable
woman, MMR, and region of the world as potential con- analysis, and the decrease in IMR was 1.66 (95%CI 1.11-
founders, this decrease continued to be significant at p < 1.32) deaths per 1000, p < 0.001. The estimated odds
0.001, 1.66 (95%CI 1.11-1.32) deaths per 1000. See Table ratio that increased sanitation access was significantly
2 for univariable results and Table 3 for the multivari- associated with increased MMR is 0.14 (95%CI 0.10-
able models and results. The estimated odds ratio that 0.20). This odds ratio was 0.52 (95%CI 0.32-0.85), p =
increased access to sanitation was significantly asso- 0.009 in the adjusted analysis.
ciated with increased odds of under-five child mortality
due to diarrhea was 0.20 (95%CI 0.14-0.28). This odds Discussion
ratio changes to 0.66 (95%CI 0.41-1.05), p = 0.08 in the Our analyses show interesting and statistically significant
adjusted model. IMR, in the unadjusted model, relationships between water and maternal, infant, and
decreased by 2.29 (95%CI 2.12-2.48) deaths per 1000 child mortality. Increasing access to water and sanitation
Table 3Adjusted Multivariable Regressions
Coefficient(95%Confidenceinterval(CI)) p-value
Predictor:proportionofpopulationwithaccesstoimprovedwatersourcea
Under-fivemortalityrateb(per1000livebirths) -1.17(-1.26,-1.08)f <0.001
%under-fivemortalityduetodiarrheaa,c OR:0.46(0.30,0.70)g <0.001
IMRb(per1000livebirths) -1.14(-1.23,-1.05) 0.001
MMRa,e(per100000livebirths) OR:0.58(0.39,0.86) 0.008
Predictor:proportionofpopulationwithaccesstoimprovedsanitationa
Under-fivemortalityrateb(per1000livebirths) -1.66(-1.32,-1.11) <0.001
%under-fivemortalityduetodiarrheaa,c OR:0.66(0.41,1.05) 0.08
IMRd(per1000livebirths) -1.66(-1.32,-1.11) <0.001
MMRa,e(per100000livebirths) OR:0.52(0.32,0.85) 0.009
avariabledividedintoquartiles
blinearregression,controlledforGNI,fertilityperwoman,maternalmortalityratio,regionoftheworld
cordinallogisticregression,controlledforGNI,deathsduetodiarrhea(1000s),MMR,regionoftheworld
dlinearregressionGNI,fertilityperwoman,MMR,regionoftheworld
eordinallogisticregressioncontrolledforGNI,fertilityperwoman,%birthsattendedbyskilledhealthpersonnel,regionoftheworld
fSampleinterpretation:under-fivemortalityrateisseentodecreaseby1.17(95%CI1.08,1.26)withincreasingquartileofpercentofpopulationwithimproved
wateraccess
gSampleinterpretation:theestimatedoddsratiothatincreasedpopulationquartileaccesstowaterissignificantlyassociatedwithincreasedoddsofunder-five
childmortalityduetodiarrheais0.46(95%CI0.30,0.70)
Chengetal.EnvironmentalHealth2012,11:4 Page6of8
http://www.ehjournal.net/content/11/1/4
are significantly associated with decreases in the nega- explained in several different ways. Access to clean
tive health outcomes of interest, namely under-five child water is essential for mothers who do not breastfeed;
mortality, under-five child mortality due to diarrhea, the use of dirty containers and unsafe water for formula
IMR, and MMR. These associations remain significant preparation puts infants’ health at risk. Infants who are
after adjusting for known and expected confounders, not breastfed are six times more likely to die from infec-
except in the case of under-five mortality due to diar- tious diseases, including diarrhea, in the first 2 months
rhea and poor access to sanitation. For the linear out- of life than those who are breastfed [25]. In addition,
comes, under-five mortality rate and IMR, the R2 value many basic birth practices, including hand washing, can
stands at 91% and 89%, respectively, for both access to affect infant mortality outcomes [26]. Unhygienic prac-
water and sanitation. This allows confidence that the tices and poor infection control have also been shown
models account for a large part of the variability in the to influence maternal mortality, which in turn will have
outcome measures. This is not to say that access to an effect on infant mortality, as evidence shows that
improved water and sanitation account for such high children who lose their mother are 10 times more likely
percentages of the outcome, but that in they do so in to die before their second birthday [8,27].
combination with all other variables. Indeed, we see that The estimatedoddsratiothatincreasedwateraccessis
the partial R2 of water access is 8.86% for under-five significantlyassociatedwithincreasedMMRis0.58inthe
mortality rate, and 5.82% for IMR. Sanitation access adjustedanalysis.Thecorrespondingoddsratioforsanita-
shows partial R2 of 9.23% for under-five mortality rate, tionis0.52.Bothoddsratiossuggestthatbetteraccessto
and 8.38% for IMR. A similar measure cannot be calcu- waterandsanitationareassociatedwithdecreasedmater-
lated for ordinal logistic regression, although the regres- nal mortality. Severalreasons articulated inthescientific
sions were highly significant as measured by the log- literaturemaycontributetothisrelationship.Betterwater
likelihood. qualityandsanitationhelpwomentoreducephysicalbur-
Under-five mortality rate is significantly related to densofcarryingwater,whichtranslatesintoimprovement
increasing water access, decreasing by 1.17 deaths per in the expectant mothers’ health. This improvement in
1000, after adjustments for GNI, fertility per woman, maternal healthinturnenablesparentsto providebetter
MMR, and region as potential confounders. A similar care to their children as they themselves are less vulner-
relationship is seen for sanitation access and under-five able to diseases. In additionto being vulnerable to water
mortality rate, with a decrease of 1.66 deaths per 1000. and sanitation related disease such as anemia, vitamin
We then performed an additional analysis of under-five deficiency,trachomaandhepatitis,theeffectsofsuchill-
mortality rate due to diarrhea since diarrhea is the sec- nesses can be more serious. For example, anemia and
ond leading cause of death in children under five [5]. associated nutritional deficiency can contribute to up to
We found that increased access to water is significantly 20%ofmaternaldeaths[28]. Maternal mortalitycanalso
associated with decreased odds of under-five child mor- be increased by exposure to unsafe water and sanitation
tality due to diarrhea, supporting the argument that andpoormanagementofmedicalwasteinhealthcenters.
water access is significantly associated with decreased Halfofallinfection-relateddeathscanbeavertedbyhygie-
odds of diarrhea mortality. Children are more suscepti- nicchildbirthtechniquesemployedbyskilledbirthatten-
ble to poor outcomes from diarrhea–such as life-threa- dants. Both clean water and skilled birth attendants are
tening dehydration–in part because water makes up a necessaryforlowermaternalmortality[29].
larger percentage of their body weight [5]. Diarrhea can Our study is an observational and cross-sectional
also contribute to under-nutrition through mal-absorp- study based on major databases from international
tion. Diarrhea, along with nematode infections caused sources. We have found strong associations between
by poor sanitation, can lead to 3.5 million under-five access to water and sanitation and child, infant, and
child deaths per year [5]. A World Bank publication maternal mortality on a global scale. This is supported
found that in the absence of diarrhea, the nutritional by previous literature on smaller scales and reinforces
status of those who are undernourished improves the importance of water and sanitation access to child,
quickly [24]. These known relationships between water, infant, and maternal mortality. Due to the ecological,
sanitation and child health support the relationships observational nature of the study, this relationship can-
identified in our models. not be proven to be causal, though there is a strong sug-
Access to water and sanitation have also been found gestion of the interconnections. Such connections
to be related to IMR, with the IMR decreasing by 1.14 should be explored with longitudinal analyses to deter-
deaths per 1000 with increasing access to improved mine the causative relationships.
water source. This relationship retains its significance in Ecological studies such as this one have been viewed
the sanitation analysis, where the decrease in IMR was unfavorably within the epidemiological literature. They
1.66 deaths per 1000. This pattern of findings can be havebeencriticizedbecauseofecologicalfallacy andthe
Chengetal.EnvironmentalHealth2012,11:4 Page7of8
http://www.ehjournal.net/content/11/1/4
problem of geographic scale, the lackof statistical rigor, indicatorsareonlyestimatesofwhatwehopetomeasure.
and the inability to appropriately define risks or health Similarly, there are numerous impedimentsto having an
measures[30,31]. However,despite thedisadvantagesof accurateassessmentofMMR,asnotedinarecentarticle
ecologic studies, there are also a number of advantages, byHoganetal.[40],whileinfantandchildmortalitydata
including the ability to study a large population and to areoften lacking orhave very lowcoverageforcountries
addressquestionsofenvironmentalhealthbeforeturning withhighmortalityrates[41,42].Newtoolsandmethods
toexpensiveandintensivecase-controlorcohortstudies needto continuetobedevelopedforbettermeasureand
[30]. For example, ecological studies of mortality and monitor the global situation of water, sanitation, and
morbidity atthe regionalscale(includingcounties)have infant,childandmaternalmortalityoutcomes;ouranalysis
demonstrated important findings, while generating new alongwithpastliteraturestronglysupporttheexistenceof
avenuesofresearchatthesametime[30,32-38].Inparti- suchrelationships.
cular, ecologic studies provide a relatively inexpensive
way toexamine regionalvariationsinthe healthoflarge Conclusions
populations.Whiletheauthorsfullyrecognizetheweak- Our analyses show interesting and statistically significant
nesses of this approach, there is significant value to this relationships between access to water and sanitation and
type of analysisindetermining the relationshipsstudied maternal, infant, and child mortality. With this knowl-
on a global scale. The associations between water and edge, plans for development should include improving
sanitationand healthoutcomesneed tobefurthereluci- water and sanitation access. This will allow fuller
dated with longitudinal approaches and controlled stu- achievement in MDGs four and five–improving child
dies. Other types of analyses were considered, such as a and maternal health.
timeseriescomparison.However,withthelargeamount
of missing data on the global scale, the level of uncer-
Abbreviations
tainty addedby otherapproacheswasconsideredby our
MMR:Maternalmortalityratio;IMR:Infantmortalityrate;WHO:WorldHealth
group to be too large. With current available data, this Organization;UN:UnitedNations;UNICEF:UnitedNationsChildren’sFund;
approach was considered the most suitable and the best DG:MillenniumDevelopmentGoal;GNI:Grossnationalincome;OR:Odds
ratio;RADWQ:Rapidassessmentofdrinking-waterquality.
availablebytheauthorsfortheresearchquestionposed.
An issue we were not able to address in our models is Acknowledgements
the rural-urban divide. The known differences in water PrimaryauthorJJCwassupportedbytheUNU-INWEHforaportionofthis
project.CSWisemployedbytheUNU-INWEH.AMissupportedbya
and sanitation access, as well as mortality outcomes
ResearchEarlyCareerAwardfromHamiltonHealthSciences.JJC,SW,KBN,
between rural and urban areas cannot be examined andAMareaffiliatedwithMcMasterUniversity.
throughcountry-levelanalysis.Thesedifferencesmaylead
Authordetails
to bias in our calculations, especially in countries where
1PublicHealthandPreventiveMedicineResidencyProgram,Departmentof
rural and urbanpopulations experience large differences ClinicalEpidemiologyandBiostatistics,McMasterUniversity,Hamilton,ON,
in the indicators we used. Similar issues exist for other Canada.2UnitedNationsUniversityInstituteforWater,Environmentand
types of heterogeneity within a country–such as gender, Health(UNU-INWEH),Hamilton,ON,Canada.3SchoolofGeographyand
EarthSciences,McMasterUniversity,Hamilton,ON,Canada.4McMaster
race, and cultural differences. Lastly, we did not assess InstituteofEnvironmentandHealth,McMasterUniversity,Hamilton,ON,
water quality andsafety related toenvironmental chemi- Canada.5SchoolofSocialWork,McMasterUniversity,Hamilton,ON,Canada.
6DepartmentofClinicalEpidemiologyandBiostatistics,McMasterUniversity,
cals(eg, exposureto arsenicfrom drinkingwaterindeep
Hamilton,ON,Canada.7PopulationHealthResearchInstitute,DavidBraley
wells). While exposure to toxic chemicals in drinking Cardiac,Vascular,andStrokeResearchInstitute,HamiltonGeneralHospital,
water adds further to the considerable disease burden Hamilton,ON,Canada.
[3,4], this area was beyond the scope of our study. One
Authors’contributions
elementthatiscrucialtothistypeofanalysisisthequality JJCwastheprimarywriter,datacollector,analyst,andinterpreter.CSW
ofthedatacollectedbythevariouscountries.Doesaccess assistedwithdatacollection,interpretationandwriting.SWassistedwith
datainterpretationandwriting.KBNassistedwithdatainterpretationand
to improved water source mean safe water? What does
writing.AMassistedwithdataanalysis.Allauthorsreadandapprovedthe
safewaterreallymean?Arethecurrentmeasurementsof finalmanuscript.
water access the most useful? The rapid assessment of
Competinginterests
drinking-water quality (RADWQ) measurement is an
Theauthorsdeclarethattheyhavenocompetinginterests.
alternative to the current measurement of access. How-
ever,RADWQisnoteconomicallyviabletoperformglob- Received:29August2011 Accepted:27January2012
Published:27January2012
ally, and carries with it opportunity costs [39]. A cost-
effectiveandmeaningfulmeasurementforsafewaterisyet
References
tobedetermined.Sanitationindicatorsareevenmoredif- 1. Pruss-UstunA,BosR,GoreF,BartramJ:SaferWater,BetterHealth:Costs,
ficulttomeasure.Thecurrentstandardforimprovedsani- BenefitsandSustainabilityofInterventionstoProtectandPromoteHealth
Geneva:WHO;2008.
tationmeasuresaccess,notuse.Bothwaterandsanitation
Chengetal.EnvironmentalHealth2012,11:4 Page8of8
http://www.ehjournal.net/content/11/1/4
2. PrussA,KayD,FewtrellL,BartramJ:Estimatingtheburdenofdisease 29. WaterSupplyandSanitationCollaborativeCouncil:Forher,it’stheBigIssue:
fromwater,sanitation,andhygieneatagloballevel.EnvironHealth PuttingWomenattheCentreofWaterSupply,SanitationandHygiene
Perspect2002,110:537-542. EvidenceReportGeneva:WaterSupplyandSanitationCollaborativeCouncil;
3. Pruss-UstunA,VickersC,HaefligerP,BertolliniR:Knownsandunknowns 2006.
onburdenofdiseaseduetochemicals:asystematicreview.Environ 30. WalterSD:Theecologicmethodinthesurveyofenvironmentalhealth:
Health2011,10:9. 1.overviewofthemethod.EnvironHealthPerspect1991,94:61-65.
4. ChowdhuryUK,BiswasBK,ChowdhuryTR,SamantaG,MandalBK,BasuGC, 31. RichardsonS,MonfortC:Ecologicalcorrelationstudies.InSpatial
ChandaCR,LodhD,SahaKC,MukherjeeSK,RoyS,KabirS, Epidemiology:methodsandapplications.Editedby:ElliottP,WakefieldJC,
QuamruzzamanQ,ChakrabortiD:Groundwaterarseniccontaminationin BestNG,BriggsDL.London:OxfordUniversityPress;2000:205-220.
BangladeshandWestBengal,India.EnvironHealthPerspect2000, 32. GriffithJ,RigganWB,DuncanRC,PellomAC:CancermortalityinUS
108:393-397. countieswithhazardouswastesitesandgroundwaterpollution.Arch
5. UNICEFandWHO:Diarrhea:WhyChildrenareStillDyingandWhatCanbe EnvironHealth1989,44:69-74.
DoneNewYorkandGeneva:UNICEFandWHO;2009. 33. WalterSD:Assessingspatialvariationsindiseaserates.StatMed1993,
6. DeOnisM,BlossnerM,BorghiE,FrongilloEA,MorrisR:Estimatesofglobal 12:1885-1894.
prevalenceofchildhoodunderweightin1990and2015.JAMA2004, 34. WalterSD,BirnieSE,MarrettLD,TaylorSM,ReynoldsD,DaviesJ,DrakeJJ,
291:2600-2606. HayesM:ThegeographicvariationofcancerincidenceinOntario.AmJ
7. WHO/UNICEFJointMonitoringProgrammeforWaterSupplyandSanitation: PublicHealth1994,84:367-376.
WaterforLife:MakingitHappenNewYorkandGeneva:WHOandUNICEF; 35. JerrettM,EylesJ,ColeD,ReaderS:EnvironmentalequityinCanada:an
2005. empiricalinvestigationintotheincomedistributionofpollutionin
8. UnitedNationsWorldWaterAssessmentProgramme:Waterforthe Ontario.EnvironPlanA1997,29:1777-1800.
MillenniumDevelopmentGoals–WhyManagingWaterResourcesWiselyisKey 36. CurrieroF,HeinerKS,SametJ,ZegerS,StrugL,PatzJA:Temperatureand
toAchievingtheMDGsParis:UnitedNations;2010. mortalityin11citiesoftheEasternUnitedStates.AmJEpidemiol2002,
9. GoodburnE,CampbellO:Reducingmaternalmortalityinthedeveloping 155:80-87.
world:sector-wideapproachesmaybethekey.BMJ2001,322:917-920. 37. BurnettRT,GoldbergMS:Size-fractionatedparticulatemassanddaily
10. FewtrellL,KaufmannRB,KayD,EnanoriaW,HallerL,ColfordJMJr:Water, mortalityineightCanadiancitiesBoston:HealthEffectsInstitute;2003,85-89,
sanitation,andhygieneinterventionstoreducediarrheainless Inrevisedanalysesoftime-seriesstudiesofairpollutionandhealthspecial
developedcountries:asystematicreviewandmeta-analysis.LancetInfect report..
Dis2005,5:42-52. 38. KaiserR,LeTertreA,SchwartzJ,GotwayCA,DaleyWR,RubinCH:The
11. UNICEFandWHOJointMonitoringProgrammeforWaterSupplyand effectofthe1995heatwaveinChicagoonall-causeandcause-specific
Sanitation:ProgressonDrinkingWaterandSanitationNewYorkandGeneva: mortality.AmJPublicHealth2007,97:S158-S162.
UNICEFandWHO;2010. 39. WHO:ReportoftheWHO/UNICEFjointmonitoringprogramme:progress
12. UnitedNations:OfficiallistofMDGIndicators.[http://mdgs.un.org/unsd/ onsanitationanddrinkingwater2010update.[http://www.who.int/
mdg/Host.aspx?Content=Indicators/OfficialList.htm]. water_sanitation_health/monitoring/key_terms/en/].
13. YardleyS:JoiningtheDots:WhyBetterWater,SanitationandHygieneare 40. HoganMC,ForemanKJ,NaghaviM,AhnSY,WangM,MakelaSM,
NecessaryforProgressonMaternal,NewbornandChildHealthTeddington: LopezAD,LozanoR,MurrayCJ:Maternalmortalityfor181countries,
Tearfund;2010. 1980-2008:asystematicanalysisofprogresstowardsMillennium
14. WHOStatisticalInformationSystem(WHOSIS):Worldhealthstatistics2010. DevelopmentGoal5.Lancet2010,375:1609-1623.
[http://www.who.int/whosis/whostat/2010/en/index.html]. 41. WHO:Measuringchildmortality.[http://www.who.int/
15. UNICEFChildinfo:Thestateoftheworld’schildren2010.[http://www. child_adolescent_health/data/children/en/index.html].
childinfo.org/statistical_tables.html]. 42. Boschi-PintoC,VelebitL,ShibuyaK:Estimatingchildmortalitydueto
16. WorldBankOpenData.[http://data.worldbank.org/]. diarrhoeaindevelopingcountries.BullWorldHealthOrgan2008,
17. WHOSIS:Probabilityofdying(per1000)underage5years(under-5 86:710-717.
mortalityrate).[http://www.who.int/whosis/indicators/mortalityunder5/en/
index.html]. doi:10.1186/1476-069X-11-4
18. WHOSIS:Causesofdeathamongchildrenagedlessthan5years Citethisarticleas:Chengetal.:Anecologicalquantificationofthe
(percentageoftotal).[http://www.who.int/whosis/indicators/ relationshipsbetweenwater,sanitationandinfant,child,andmaternal
mortcauseslessthan5years/en/index.html]. mortality.EnvironmentalHealth201211:4.
19. UNICEF:Basicindicators.[http://www.unicef.org/infobycountry/
stats_popup1.html].
20. WHOSIS:Maternalmortalityratio(per100000livebirths).[http://www.
who.int/whosis/indicators/maternalmortratio/en/index.html].
21. RoystonP:Multipleimputationofmissingvalues:Furtherupdateofice,
withanemphasisoncategoricalvariables.StataJ2009,9:466-477.
22. AgrestiA:AnIntroductiontoCategoricalDataAnalysisNewYork:JohnWiley
andSons,Inc;1996.
23. UCLAAcademicTechnologyServices,StatisticalConsultingGroup:Stata
dataanalysisexamplesordinallogisticregression.[http://www.ats.ucla.
edu/stat/stata/dae/ologit.htm].
24. WorldBank:EnvironmentalHealthandChildSurvival:Epidemiology, Submit your next manuscript to BioMed Central
Economics,ExperiencesWashington:WorldBank;2008. and take full advantage of:
25. WHOCollaborativeStudyTeamontheRoleofBreastfeedingonthe
PreventionofInfantMortality:Effectofbreastfeedingoninfantandchild
• Convenient online submission
mortalityduetoinfectiousdiseasesinlessdevelopedcountries:a
pooledanalysis.Lancet2002,355:451-455. • Thorough peer review
26. RheeV,MullanyLC,KhatrySK,KatzJ,LeClerqSC,DarmstadtGL,TielschTM:
• No space constraints or color figure charges
Maternalandbirthattendanthandwashingandneonatalmortalityin
southernNepal.ArchPediatrAdolescMed2008,162:603-608. • Immediate publication on acceptance
27. UniteNationsPopulationFund:GivingBirthShouldnotbeaMatteroflife • Inclusion in PubMed, CAS, Scopus and Google Scholar
andDeathNewYork:UnitedNations;2007.
• Research which is freely available for redistribution
28. WHO:Water-relateddiseases.[http://www.who.int/water_sanitation_health/
diseases/anemia/en/].
Submit your manuscript at
www.biomedcentral.com/submit