A Framework for Analyzing the Determinants of Maternal Mortality

Author(s): James McCarthy and Deborah Maine

Source: Studies in Family Planning, Vol. 23, No. 1 (Jan. - Feb., 1992), pp. 23-33
Published by: Population Council
Stable URL: http://www.jstor.org/stable/1966825 .
Accessed: 05/09/2014 10:12

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.

Population Council is collaborating with JSTOR to digitize, preserve and extend access to Studies in Family
Planning.

http://www.jstor.org

This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM

All use subject to JSTOR Terms and Conditions
A FrameworkforAnalyzing the
Determinantsof Maternal Mortality
JamesMcCarthyand Deborah Maine

Hundredsofthousands ofwomenindeveloping countriesdieeachyearfromcomplications ofpregnancy, attempted

abortion, and childbirth.Thisarticlepresents
a comprehensive and integrated framework foranalyzingthecultural,
social,economic, behavioral,and biological
factorsthatinfluence maternal mortality.Thedevelopment ofa compre-
hensiveframework was carriedoutbyreviewing thewidelyaccepted frameworks thathavebeendeveloped for
and childsurvival,and byreviewing
fertility theexistingliteratureon maternal mortality, includingtheresultsof
research studiesandaccountsofintervention programs. Theprincipalresultofthisexerciseis theframework itself.
One ofthe main conclusionsis thatall determinants ofmaternal mortality(and,hence,all efforts
toreducematernal
mortality) mustoperatethrough a sequenceofonlythreeintermediate outcomes. Theseeffortsmusteither(1) reduce
thelikelihood thata womanwillbecomepregnant; (2) reducethelikelihoodthata pregnant womanwillexperience a
seriouscomplication ofpregnancy orchildbirth;
or(3) improve theoutcomes forwomenwithcomplications. Several
typesofinterventions aremostlikelytohavesubstantial and immediate effectson maternal mortality,including
familyplanningprograms toprevent pregnancies, safeabortionservicestoreducetheincidence ofcomplications,
and improvements in laborand deliveryservicestoincreasethesurvivalofwomenwhodo experience complications.
(STUDIES IN FAMILY PLANNING 1992;23, 1: 23-33)

As recentlyas 1985, maternalmortalityin developing timesgreaterin theformercountries(WHO, 1985;United

countries was referredto as "a neglected tragedy" NationsPopulationDivision,1988). As dramaticas these
(Rosenfieldand Maine, 1985).Althoughaccuratedata on statisticsare, theytellonly part of the story,since many
maternalmortalityare not available, estimatesare that
otherwomen experiencesubstantialsufferingand per-
500,000 women in developing countriesdie each year manentinjuryas a resultofpregnancyand childbirth.
fromcomplicationsofpregnancy,abortionattempts,and In thelast fiveyears,a numberoforganizationshave
childbirth(WHO, 1985). In many countries,maternal participatedin the launch of the Safe Motherhood Ini-
mortalityis theleading cause ofdeath among women in tiativeby supportingboth researchon thedeterminants
thereproductiveages. Further,thediscrepancybetween of maternalmortalityand interventions to reduce levels
maternal mortalityrates in developing countries and ofmaternalmortality. These researchand programmatic
those in the developed world is greaterthanthatof any
effortshave considered a diverse set of factorsthought
otherdemographic indicator.The number of maternal to be associated with maternalmortality.For example,
deathsamong women ofreproductiveage in Bangladesh some recentpapers have included listsofthecauses and
and India is about 100 times the numberin the Unitedrisksofmaternalmortality, whichare usually organized
States. Infantmortality,by contrast,is about 10 to 15
into such categoriesas obstetric,health service,repro-
ductive, socioeconomic, and transportation factors
(Maine et al., 1987; Royston and Armstrong,1989).
Amongtheinterventions beingpromotedby varioussafe
JamesMcCarthy, Ph.D.is Director,
Center forPopulation motherhood/maternalmortalityprojectsare those that
andFamily HealthandProfessorofPublicHealth,Columbia encouragechangein thestatusofwomen; programsthat
60HavenAvenue,
University, NewYork, NY 10032;and offerfamilyplanning,prenatalcare,nutritionalsupple-
Deborah Maine,M.P.H.is Director,
PreventionofMaternal mentation,and tetanus immunization; programs that
MortalityProgram,CenterforPopulationandFamily providemoreeffective linkagesbetweentraditionalbirth
Health,ColumbiaUniversity. attendantsand themodernhealthsystem;and programs

Volume23 Number1 January/February

1992 23

This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM

All use subject to JSTOR Terms and Conditions
 thatundertakeoverallimprovementsin access to and the
qualityofemergencyobstetriccare (WHO, 1989;Boerma,
1987). Obviously,these causal factorsand potentialin-
terventionsoperateat quite differentlevels ofproximity
to the event of a maternaldeath; some are directlybio-
logical,whereas othersare relatedto aspects ofsocial or-
ganizationand healthservices.Relativelyfewreportsor
programs,however,have explicitlyor systematically con-
sidered the mechanisms or pathways through which
thesediversefactorsinfluencematernalmortality.There
are a few exceptions;several authorshave begun to ad-
dress more systematicallythe entireprocess thatculmi-
nates in maternaldeath or serious maternalmorbidity.
In some moving passages, Fathalla (1987) described the
"road to death" that women follow,a road that starts
withtheunderlyingsocioeconomicconditionsoflifeand
continuesto include thedemographicand healthservice
factorsthatcontributeto death.Othershave focusedtheir
efforts on thatsegmentoftheprocess thatstartswiththe
event of a pregnancycomplication,and have examined
various factorsthatinfluencedelays in deciding to seek
medical care,in reachinga place where care is available,
and in receivingappropriatecare (Thaddeus and Maine,
1990).
Although these papers have taken a more system-
atic approach to understandingthedeterminantsofma-
ternalmortality, theyhave notpresentedfullydeveloped,
comprehensiveframeworksor models forthe determi-
nantsofmaternalmortality.The understandingofother
demographicevents,notablyfertility and child survival,
has been advanced by the developmentof frameworks
that specifythe biological and behavioral mechanisms
throughwhich social, economic,and culturalfactorsop-
erateto produce a birthor the survival of a child to age
five(Davis and Blake,1956;Bongaarts,1978;Mosley and
Chen, 1984). Using the same approach as these earlier
efforts, we presenta frameworkforthe analysis of ma-
ternalmortality-a frameworkthatwe hope will be es-
peciallyusefulforresearchin developing countries.The
developmentof an analyticalmodel ofmaternalmortal-
ityis an indicationthattheSafe MotherhoodInitiativeis
comingofage.
Many of the issues discussed in this articleare fa-
miliarto people workingon theSafe MotherhoodInitia-
tive;sufficientresearchdata and programexperienceal-
ready exist to guide effortsto understandand change
patternsof maternalmortality.However, thereis now
enough evidence fromboth research studies and pro-
gramsto supportthedevelopmentofa "firstgeneration"
framework. Our overallgoals in developinga framework
foranalyzingthedeterminantsofmaternalmortalityare
to stimulatefurther discussion,furtherresearch,and new
programs.Our assumptionis thatconsiderationof such
24 Studies in FamilyPlanning
This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM
All use subject to JSTOR Terms and Conditions
a framework will result in programs and research
projectsthatare more focused and ultimatelymore ef-
fectivein savingwomen's lives. Furthermore, experience
fromthesefutureresearchefforts and programinterven-
tions will undoubtedlycontributeto refinementsof the
frameworkpresentedhere.
The primaryfocus of this articleis on the determi-
nants of maternalmortality, but the frameworkwe pro-
pose can also be applied to chronicmorbiditythat re-
sults frompregnancyor childbirth.Although the term
"maternalmorbidity"encompassesa wide varietyofcon-
ditions,some of which are relativelyminorand of very
shortdurationand othersof which are severe and long-
standing,our frameworkis designed to capturethe de-
terminantsoflong-term, serious morbidityonly.We use
the term"disability"to referto chronic,severe morbid-
itythatresultsfromeitherpregnancyor childbirth;and
we referto maternaldeath or disabilityas the ultimate
outcomeoftheframeworkpresentedbelow.
The Concepts behind the Framework
Figure1 presentsa relativelysimple frameworkforana-
lyzingthe determinantsof maternalmortalityand mor-
bidity.It includes thebasic stages in the process thatre-
sultin maternaldisabilityor deathand a briefdescription
of each of those stages. The frameworkis organized
around threegeneral stages or componentsof the pro-
cess of maternalmortality.Closest to the eventof a ma-
ternaldeathare a sequence ofsituationsor outcomesthat
culminatesin eitherdisabilityor death; these outcomes
are pregnancyand pregnancy-relatedcomplications.A
woman mustbe pregnantand experiencesome compli-
cation of pregnancyor childbirth,or have a preexisting
healthproblemthatis aggravatedby pregnancy,before
her death can be defined as a maternaldeath. This se-
quence of outcomes is most directlyinfluencedby five
setsofintermediatedeterminants: thehealthstatusofthe
woman; herreproductivestatus;heraccess to healthser-
vices; her health care behavior (including her use of
healthservices);and a set of unknownfactors.Finally,a
set of socioeconomicand culturalbackgroundfactorsis
at thegreatestdistancefroma maternaldeath.
Considering pregnancy and pregnancy complica-
tionsas part of the sequence of eventsor outcomes that
culminatesin maternaldisabilityor death leads to an ob-
vious but importantset of propositions.Any factorthat
is thoughtto influencematernalmortality, and therefore
any effortsto reduce maternalmortality,must operate
throughtheseevents.These efforts must
1 reduce thelikelihoodthata woman will become
pregnant;
 and morbidity
foranalyzingthe determinantsof maternalmortality
Figure 1 A framework

Distant Intermediate Outcomes

determinants determinants

I I
_ regnancy

AcstohealthservicsI
Socioeconomic/ .
and b 4 ( Complication
culturalfactorsII\
l ~~~Access
to healthservices f

l /
I D~~~~~~~~~~~~~~~~eath/disability
h carebehavior/usedofhealthoservices

Unknowno nrdce atr

2 reduce the likelihood that a pregnant woman
will experiencea serious complicationof preg-
nancyor childbirth;or
3 improvethe outcomes forwomen with compli-
cations.
Therefore, in all researchstudiesand forall proposed
interventions, the manner in which a given variable or
programactivityis expected ultimatelyto affectone of
the outcomes in the sequence should be made clear. A
review of existingresearchon this topic and programs
in place or proposed suggeststhattheseconnectionsare
rarelyexplicit.
The frameworkin Figure1 is completein thatitcov-
ersall ofthepossiblefactorsthatinfluencematernalmor-
tality,and it specifiesthe general mechanismsthrough
which themore distantfactorsmustoperate.The frame-
work is not, however, very precise; the boxes can be
thought of as general concepts that can encompass a
much largerset of variables. Figure 2 elaborates on the
basic frameworkby incorporatingmany of the specific
variables thatcould be used to measure the conceptsin
thebasic framework.Thereare manysuch variables,and
theframeworkpresentedin Figure2 is clearlyquite com-
plicated. However, in spite of thiscomplexity,a discus-
sion of these specific variables can illustratehow the
frameworkmightbe used in eithera programor research
setting.Our discussionofvariablesis orderedfromthose
closestto theeventof death and disabilityto thosemost
distantfromthatevent.
Sequence of Outcomes
Pregnancy
By definition,pregnancyis a necessarypreconditionfor
maternaldeath. It is the biological state throughwhich
all otherfactorsmustinfluencematernaldeath.Although
therecan be no maternaldeathuntilthereis a pregnancy,
itis importantto include pregnancyas thestartingpoint
ofthesequence ofoutcomesleading to maternalmortal-
ity,because the risk of pregnancyvaries considerably
fromwoman to woman, and because pregnancyrates
varyso greatlyamong different groups ofwomen.'
Complications
Maternal (or obstetric)deaths can be classifiedas either
director indirect.A directobstetricdeath is one due to
complicationsofpregnancy,delivery,or thepostpartum
period,includingabortioncomplications.Indirectobstet-
ric deaths are those due to existingmedical conditions
thatare made worse by the pregnancyor delivery.On
average,approximatelythree-quarters ofmaternaldeaths
in developingcountriesare due to directobstetriccauses,
and one-quarterare due to indirectcauses (WHO, 1985).
Furthermore, a verylimitednumberofcomplicationsare
responsibleforthevast majorityofdeaths.For directob-

1992 25
Volume23 Number1 January/February

This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM

All use subject to JSTOR Terms and Conditions
 thedeterminants
foranalyzing
Figure2 A detailedframework mortality
ofmaternal and morbidity

Distant Intermediate Outcomes

determinants determinants

r?I Healthstatus
Women's
status status
Nutritional
in familyand community (anemia,height,weight)
Education I I Infectionsand parasiticdiseases
(malaria,hepatitis,
tuberculosis)
I
Occupation
Income
Otherchronic
(diabetes,
conditions
hypertension) I Pregnancy
Social and legal autonomy, Priorhistoryof pregnancycomplications

I I I I__ _ _ _/_ _

I Reproductive
status I
Age/
Parity| /
Maritalstatus
status
Family's Complication
in community
Familyincome
I I |/_Hemorrhage
Access to healthservices PrInfection
Land Location of servicesfor Pregnancy-induced
Educationof others -family planning O thypertension
Occupation of others -prenatal care ut ObstRucted
e r
-otherprimarycare Ruptureduterus
-emergencyobstetriccare
Range of servicesavailable
I I Quality of care
to information
~~Access
A about
_ services
,_ \
I ~ ~ ~ ID ~ I,
1
__ __ _

I Community'sstatus
I I Health care behavior/useof healthservices Death/disability
Aggregatewealth I I Use of familyplanning
Community resources Use ofprenatal
care
of moderncare forlabor and delivery
| |(e.g. doctors,clinics, ] ? I ?Use
Use of harmfultraditionalpractices
| gambulances) t | | | I
ambulanc)Use ofillicitinducedabortion

I~~~~~~~~~~~~nnw
orupeice facor

stetricdeaths, hemorrhage,infection,sequelae of illicit ityor death. Althoughtherehas not been extensivere-

induced abortion,2pregnancy-induced hypertension,and
obstructedlabor/ruptureduterus are the major causes
(Maine et al., 1987). There can be considerable overlap
among thesecauses; forexample,a hemorrhagemay re-
sult froma ruptureduterus,or a serious infectioncould
be a sequela ofprolongedand obstructedlabor.
The category "complications" in the framework
refersto theseand possibly othercomplicationsthatdi-
rectlycontributeto maternaldeaths.Conditionsthatcon-
tributeindirectlyto maternaldeath are more appropri-
ately included as a component of a woman's health
status,one of the intermediatedeterminantsdiscussed
below.
DisabilityorDeath
The finaloutcomes in the frameworkare eitherdisabil-
search on pregnancyand childbirth-related disabilityin
developing countries,itis likelythatthereare onlya few
nonfatalconditionsthatare responsibleformostserious
disability.These include chronicurinarytractinfection,
uterineprolapse,and vaginal fistulae,each of whichis a
serious,chronicconditionthatcan have a considerable
impacton thephysicaland social well-beingofwomen.
Intermediate
Determinants
HealthStatus
A woman's personal healthstatuspriorto and duringa
pregnancy can have an important influence on her
chancesofdevelopingand survivinga complication.The
leadingpreexistinghealthconditionsthatare exacerbated

26 Studies in FamilyPlanning

This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM

All use subject to JSTOR Terms and Conditions
by pregnancy and delivery and account for approxi-
mately one-quarterof maternal deaths in developing
countriesare malaria, hepatitis,anemia, and malnutri-
tion (Maine et al., 1987; Roystonand Armstrong,1987).
Furthermore,the presence of some of these conditions
may put women at higherriskof dyingfromone of the
directcomplicationsofpregnancy.Malaria, forexample,
may notonly be moresevere in pregnantwomen,but it
may also contributeto anemia, which in turnmay de-
crease a woman's chance ofsurvivinga hemorrhage.
Reproductive Status
The relationshipsbetween maternalmortalityand cer-
tainreproductivecharacteristics are among thebestdocu-
mentedin the literature(Maine, 1981). These character-
isticsincludeage and pregnancyorder,whichare known
to have a classic "J-shaped"relationwith the maternal
mortalityratio,with risksthatare high forveryyoung
women, older women, women with no children,and
those with many children,but are lower forwomen in
between. Age, especially veryyoung age, is also associ-
ated with disability that results frompregnancy and
childbirth. For example,vesico-vaginalfistulaeare much
more common among very young mothers,who are
more likely than othersto experienceprolonged labor
as a resultof immaturepelvises (Tahzib, 1989). Parity
can influenceone of the othermajor disabilitiesthatre-
sults frompregnancy,uterineprolapse, which is much
more common among high-paritywomen (Omran and
Standley,1976).
The wantedness of a pregnancyis also an impor-
tantvariable, especially since women who have an un-
wanted pregnancyare more likelythan othersto seek
an abortion,even iftheonlyproceduresavailable are un-
safe,illicitabortionsthatgreatlyincreasetheriskofdeath
and disability(Kwast and Liff,1988).
Duringthelast decade, evidenceaccumulatedshow-
ing a strong relationship between the spacing of a
woman's birthsand the survival of her children(Maine
and McNamara, 1987). Birthspacing, which is not in-
cluded on thelist,can be influencedby contraceptive use,
breastfeeding,or postpartumabstinence.Since many of
thereproductivevariablesthataffectinfantsurvivalalso
affectmaternalsurvival(forexample,age and pregnancy
order),one mighthypothesizethatbirthspacing influ-
ences the outcome forthe motheras well. To date, how-
ever,thereare no studies thatsupportthishypothesis.
AccesstoHealthServices
The medicaltechnologyto preventalmostall deathsfrom
common obstetriccomplicationshas been available for
decades. This technologyfallsintotwo majorcategories:
treatmentsforwomen who want to have a safe and suc-
cessfulbirth,and optionsforwomen who want to avoid
pregnancyand childbirth.The formercategoryincludes
This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM
All use subject to JSTOR Terms and Conditions
blood transfusions,antibioticsand otherdrugs,and ce-
sarean sections;thelatterincludes use of contraceptives
and safe abortionprocedures. But access to preventive
and curativehealth services thatcan provide this tech-
nology is limitedin developing countries.In many set-
tings, the physical distances between services and
women in need of reproductivehealth care are consid-
erable. Physicaldistancefromfacilitieshas been shown
to be associated withmaternalmortalityin several stud-
ies (Fortneyet al., 1985;Walkeret al., 1985).
Access, however, is a much broader concept than
physicaldistance.It includes financialaccess and access
to adequate care. (Each of thesevariables could, in turn,
be subdivided intomuch more preciseoperationaldefi-
nitions. For example, the World Health Organization
[1986] has identifieda numberof specificservices that
are essential to the care of pregnantwomen.) There is
ample evidence that financial barriers, shortages of
trainedpersonnel,especiallyin ruralareas,and poor per-
formanceon thepartof trainedpersonnelall contribute
to highlevels of maternalmortalityin developing coun-
tries(Ekwempu et al., 1990;Omu, 1981;WHO, 1985).
HealthCareBehavior/Use ofHealthServices
For servicestobe effective,women have to use them.The
use ofprenatalcare (to diagnose eitherpreexistinghealth
problemsor to detectcertaincomplications)and theuse
ofcare duringand afterlabor and delivery(to treatcom-
plicationsthatmay arise then)are particularlyimportant
in the case of maternalmortality.Other health care be-
haviors are also likelyto have importantinfluenceson
theoutcomeofpregnancyforwomen.Obvious examples
include theuse ofillicitabortionistsand harmfulbut tra-
ditional practicesduring pregnancyand childbirth.In
some areas,traditionalpracticesincludetheimproperuse
of drugs, pushing on the abdomen to hasten delivery,
and even the use of certainsurgicalprocedures.For ex-
ample, in northernNigeria, traditional healers make
"Gishiricuts" (incisionsin the vagina) on women who
are notmakingprogressin labor.More removedin time
fromlabor and delivery,but stillquite harmful,are the
more radical formsof female circumcisionpracticedin
manysocieties.
Unknownor Unpredicted Factors
Anotheressentialfactto consideris thatpregnancycom-
plicationscan arise fromfactorsotherthan a lack of ac-
cess to or use of health services,or poor health condi-
tions prior to or during a pregnancy, or a woman's
reproductivestatus.Women who are fromadvantaged
backgrounds, who have ample access to high-quality
healthservices(includingprenatalcare),and who are in
good health prior to pregnancydo experience serious
obstetriccomplications for reasons that cannot be ex-
Volume23 Number1 January/February
1992 27
plained or predicted.A recentstudyofnonhospitalbirth
centersin the United States is particularlyilluminating.
Of almost 12,000women who went to the birthcenters
forprenatal care, 15 percentwere referredto hospitals
because theywere considered to have an unacceptably
high risk of unfavorableoutcomes. Those who contin-
ued at the birthcenterswere deemed to be "low risk"
and had an average of 11 prenatalvisits.Despite thisin-
tensivescreeningand prenatalcare,almost 8 percentof
these women had a serious maternalor fetalcomplica-
tion (Rooks et al., 1989). In developing countrysettings,
predictingobstetriccomplicationsis equally difficult.For
example,althougha bad obstetrichistoryis a knownpre-
dictorof obstructedlabor, a study in Zaire found that
only 29 percentof the cases of obstructedlabor could
have been predictedbased on obstetrichistory.The other
71 percentof cases occurredin women with no known
riskfactors(Kasongo ProjectTeam, 1987).
Distant Determinants
Socioeconomic Status
It is well known thatthe riskof dying is stronglyinflu-
enced by one's positionin society.In mostcircumstances
and formostdiseases, includingmaternalmortality, the
poor and disadvantaged are more likelyto die than are
more affluentpeople. Differentialsin maternalmortal-
ityby socioeconomic status exist among countriesand
within countries. However, socioeconomic status is a
complex concept, one that operates at the individual,
family,and communitylevel.
Figure 2 illustratessome of the variables that can
serve as indicatorsof socioeconomicstatus.For women,
theirstatus in the familyand in the communitycan be
relatedto theirlevel ofeducation,theiroccupation,their
level of personal income or wealth,and theirautonomy
(forexample, theirabilityto travel on theirown or to
make independentdecisions to use healthfacilities).At
thefamilylevel, statuscan be associated withaggregate
familyincome as well as with the occupation and edu-
cationoffamilymembers.Finally,thecollectiveresources
and wealth of a local communityare also importantdi-
mensionsof socioeconomicstatusthatare likelyto have
an influenceon thehealthofcommunitymembers.
As consistentas findingsin the literatureare about
the association between socioeconomic status and ma-
ternalmortality, it is importantto recognizethatthereis
nothingautomatic or directabout the association. The
influenceof socioeconomicstatusmustoperate through
some set of intermediatedeterminantsthat affectsone
oftheoutcomesin theframework(pregnancy,pregnancy
complications,and death or disability).The best illustra-
tionof thisidea comes fromexamples of the association
28 StudiesinFamilyPlanning
This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM
All use subject to JSTOR Terms and Conditions
betweenelementsofstatusand mortalitythatare theop-
posite of what one would expect.In one group of well-
nourished,well-educated,and relativelyaffluentpeople
in the United States,the maternalmortalityratio (num-
ber of maternaldeaths per 100,000live births)in 1983
was 872, at a timewhen the ratio forthe entirecountry
was eight.This level ofmaternalmortalitywas 100 times
thenationalaverage,higherthanthatofurbanIndia and
comparable to levels in ruralIndia (Kaunitz et al., 1984;
Bhatia,1985;U.S. DepartmentofHealth and Human Ser-
vices,1987).This economicallyadvantaged group,a fun-
damentalistreligious sect called the Faith Assembly of
God, has such highlevels of maternalmortalitybecause
its membersdo not believe in using any modern medi-
cal care,includingobstetriccare. Conversely,in Bahrain
and Kuwait, levels of maternalmortalityare quite low,
despite the factthatwomen in these societieshave low
status.Theydo, however,have readyaccess to relatively
high-quality obstetriccare (Royston and Armstrong,
1989). Further,althoughwomen's education and socio-
economic statusare inverselyassociated with maternal
mortalityin developed countriestoday,in the late 19th
and early20thcenturiesin Britain,women in themiddle
and upper classes actuallyexperiencedhighermortality
than did poorer women (Loudon, 1986). This situation
was a resultoftheuse of "interventionist obstetrics,"in-
cludingexcessiveuse ofchloroformand forcepsby phy-
sicians,withtheresultthatmidwivesusing less invasive
approachesprovidedsaferdeliveriesto theirclients,who
were usually poorerthantheclientsofphysicians.
The numberofpossible combinationsofall thevari-
ables listedin Figure2 is obviouslyvast,and clearlynot
all variablescould be includedin all studiesor addressed
by all programs.The listofvariablesin thefigureis also
notexhaustive;otherfactorscould be includedunderthe
categoriesof distantand intermediatedeterminants.In
makingchoices forvariablesto be included in studiesor
addressed by programs,considerationshould be given
to thosevariablesthatclearlymeasuretheprecisemecha-
nisms throughwhich the more general concepts oper-
ate. More detailed frameworkscould be developed to re-
flecthow different variables operate withineach of the
concepts in our basic model; and frameworkscould be
developed to convey,in much more detail, the process
of movement from one stage to another in the basic
model. For example, among the intermediatedetermi-
nants,a woman's healthstatusis certainlyinfluencedby
bothheraccess to healthservicesand herhealthcare be-
havior; access to health services also influenceshealth
care behavior,especiallytheuse ofhealthservices.
The frameworkcould also be furtherelaboratedby
describingin more detail the sequence of ev7entsfrom
pregnancyto death or disabilityand by consideringhow
each oftheintermediatevariablesoperatesat each ofthe eral differentpatterns,singlyor in combinationwithone
morespecificstagesin thesequence.Forexample,Mosley another.More educated women mightbe betterinformed
has suggested thatthreestages of pregnancyand deliv- about the symptomsof complicationsand could there-
erybe considered:theearlystages of pregnancy;thepe- forebe morelikelyto make a timelydecisionto seek care
riod immediatelysurroundinglabor and delivery;and when a complicationarises. Such women mightalso be
the postpartumperiod. Complicationscan arise at each concentratedin urban areas and thus would live closer
of these stages-complications that in many cases are to healthcare facilities,or theymighthave betteraccess
likely to be confinedto that specificstage. Abortionis to the transportation needed to reach those facilities.Fi-
most likelyto be associated with complicationsearlyin nally,educated women mightbe more likelyto receive
pregnancy,whereas access to emergencyobstetricalcare appropriateand timelycare when theydo reacha health
will be associated with complicationsaround the time facility,eitherbecause theyare betterable to pay forthat
of labor and delivery and in the postpartum period care or because, by virtueof theirstatus,theyare more
(Mosley, 1990).Depending on therelativeimportanceof likelyto be well-treated.
differentcomplicationsin a given setting,programin- Each ofthesemechanismscould well operatein some
terventionsand researchefforts mightbe focusedon one or all societies.An importantquestion,however,is the
of these more specific stages between pregnancy and magnitude of the effectoperatingthrougheach of the
death or disability. possiblemechanismsand therelativeimportanceofeach.
However, even in its general form,the framework These magnitudesare verylikelyto differconsiderably
does provide suggestionsforresearchon the determi- both within and among societies. Moreover, some of
nants of maternalmortalityand forprogramsdesigned theseassociationsare causal in natureand othersare not.
to preventmaternaldeaths. Knowledge of the relativemagnitudeof the effectsand
of the causal or noncausal natureof associations should
be ofinterestbothto researchersstudyingmaternalmor-
Implications forResearch talityand to officialsimplementingprogramsto reduce
maternaldeaths.
A frameworkforanalyzingthedeterminantsofa demo- Similarsetsofresearchquestionscould be developed
graphiceventimprovesresearchifthatframeworkclari- to examinetherole of otherbackgroundvariablesin ma-
fies the mechanismsthroughwhich social, behavioral, ternalmortality.However, researchbased on the frame-
and biological factorsinteractto produce an outcome.3 workneed notstartwitha focuson a backgroundvariable.
Suggestions of this kind of clarificationcan easily be Forexample,researchquestionscould be developed to ex-
drawn fromthe maternalmortalityframeworkjust de- aminetheinfluenceofpoor maternalnutrition on mortal-
scribed.For example, a common assertionin the litera- ityby studyingtheassociationbetweennutritional status
ture is thatthe status of women or women's education and thedevelopmentofcomplicationsofpregnancyor the
is relatedto maternalmortality(Roystonand Armstrong, treatmentof complications(if certaincomplicationsare
1989). Less common is a precise statementof the inter- moreseriousin thepresenceofpoor nutrition).
mediate mechanismsunderlyingthisobserved associa- In addition to the need forstudies of maternalmor-
tion.Severalthemesforresearchthatmightproducesuch talityto specifythe intermediatemechanisms through
statementsare impliedby theframework.To answer the which more distantfactorsaffectthe outcomes thatcul-
question of how women's education,forexample, may minatein death or disability,researchmustalso address
influencematernalmortality, we firstneed to determine the relativeimportanceof different factors.The frame-
theprecisemechanismsand sequences ofeventsthrough work itselfcan be used to help suggest the pathways
which education affectsmaternal death or disability. fromdistantdeterminantsthroughintermediatedeter-
Threemechanismsare possible.First,education(through minantsand throughthesequences ofoutcomesthatlead
itsassociationwithlaterage at marriageor increaseduse to maternaldeathand disability.Resultsofresearchstud-
of contraceptiveswithinmarriage)is likelyto be associ- ies are requiredto evaluate thestrengthofdifferent path-
ated withlower fertility and hence withfewerpregnan- ways, and hence to evaluate therelativepower ofdiffer-
cies. Education could also be associated with the devel- ent interventionsto reduce maternalmortality,an issue
opmentoffewercomplicationsamong pregnantwomen thatwill be discussed at greaterlengthin the following
ifbetter-educatedwomen are in generalin betterhealth section.
than othersbeforeand duringpregnancy.Finally,edu- A singleresearchstudythatspecifiedall thevarious
cation could be associated with a greaterlikelihood of pathways that culminate in maternal death and that
receivingappropriatecareforcomplicationsthatdo arise. evaluated therelativeimportanceofthosepathways,al-
This last association,in turn,could resultfromsev- though potentiallyof great interest,would be an enor-

Volume 23 Number 1 January/February

1992 29

This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM

All use subject to JSTOR Terms and Conditions
mous and probablyimpossibleundertaking.Answersto cision mightpreventsome directobstetricdeaths.
all the questions raised about the association between However, an importantimplication for programs
women's educationand maternalmortality, forexample, thatis clearfromtheframeworkis thatpreventionis not
would requirea large,complicated,and veryexpensive enough. Complete preventionof all high-riskcases and
researchproject.In theabsence ofsuch a comprehensive substantialimprovementsin women's healthstatuswill
study,theframeworkcan stillbe used to guide morefo- not eliminate complications. In a comprehensivepro-
cused researchefforts,providingthosestudieswithclear gramto preventmaternaldeaths,therewill be a need to
boundaries and a sense of theircontributionsand limi- detectand treatcomplicationsof pregnancy.Programs
tations. mustbe able to provide adequate servicesto treatcom-
plications thatcan be detected early in pregnancyas a
resultof prenatalcare, as well as to provide emergency
Implications forPrograms care for complications that are not predicted early in
pregnancy.Not only are some complicationsdifficultto
The major implicationsof the frameworkforprograms predict,but manydo notdevelop untillaborbegins.The
aimed at reducing maternal deaths are related to the level of treatmentrequireddoes not necessarilyimplya
identificationof three outcomes that culminate in death: focus on high-technology care in largehospitals.
tertiary
a conception,a complication,and the treatmentof com- Referralsystemsthattakemaximumadvantage ofexist-
plications.Societiesthathave achieved thelowest levels ing facilitiesand improvementsin the abilityof firstre-
of maternalmortalityhave done so by preventingpreg- ferralcentersto provideessentialobstetricserviceswould
nancies,by reducingthe incidence of certaincomplica- have substantialimpacts on maternalmortality(WHO,
tions,and by having adequate facilitiesand well-trained 1986).
staffto treatcomplications. Obviously, not all programs can address each of
In countriesthatare stillexperiencinghigh rates of these outcomes and intermediatedeterminantsat once.
maternalmortality, programsmustbe developed to im- However, even ifprogramsneed to limittheirinterven-
provebothpreventionand treatment. The mosteffective tions because of a lack of resources,officialsdesigning
preventivemeasure will be the widespread acceptance the programsshould be aware of the total picture,rec-
offamilyplanningpractices.By reducingthenumberof ognizing the contributionthat specific initiativescan
pregnancies, family planning will reduce the risk that a make as well as the situationsthatthose initiativeswill
woman will die frompregnancy-related causes. In soci- leave unchanged. Whereas the above discussion on the
eties characterizedby both high fertility and high ma- implicationsforprogramsis verygeneral,it is possible
ternalmortality,the contributionof familyplanning to to be much more specificby takingeach proposed inter-
increased maternalsurvival could be considerable. In- vention,tracingits potentialimpact throughthe entire
creases in the age at marriage,especially in those popu- framework,and using the resultsof researchto identify
lationsin which marriageage is verylow, could also in- therelativecontributionofdifferent interventions to the
fluencematernalmortality, notonlyby reducingfertility reductionofmaternaldeaths.
but also by reducingthe chance thata woman's pelvis We can demonstratetheusefulnessoftheframework
will be immatureat the time of her firstbirth,a condi- by examiningseven interventionsthatare oftenconsid-
tionthatis associated withobstructedlabor. ered as partoftheSafe MotherhoodInitiative.4 These in-
In additionto preventingpregnancies,programsalso terventionsare: (1) provisionoffamilyplanningservices;
need to focuson preventingcomplicationsamong preg- (2) improvementsin thesocioeconomicstatusofwomen;
nant women. The most effectivestrategiesforpreven- (3) provision of safe, legal abortionservices; (4) provi-
tion of complicationsinvolve effortsto ameliorate the sion of prenatal care; (5) improvementsin emergency
prepregnancyhealth status of women and to improve obstetriccare; (6) trainingoftraditionalbirthattendants;
certainhealthbehaviors.Nutritionprogramsforyoung and (7) education and mobilizationofthecommunity.
girls, for example, might increase the chances that a The firstconclusionof such an exerciseis therecog-
woman will be physicallymaturebeforeher firstpreg- nitionthatonly threeof these interventionswill have a
nancy and thus less likelyto experienceobstructedla- directeffecton maternalmortality.Familyplanningwill
bor.Therefore, programsthatimprove thenutritional sta- reduce maternalmortalityby reducing the number of
tus of young girlscould, eventually,have an impact on pregnancies; safe, legal abortion will reduce maternal
maternal mortalityby reducing the incidence of ob- mortalityby greatlyreducingthe complicationsthatre-
structedlabor. The effectof improvementsin treating sultfrombotchedabortions;and improvementsin emer-
other health indicators,such as malaria, anemia, and gency obstetriccare will reduce maternalmortalityby
hepatitis,could be more immediate. Also, eliminating improvingthesurvivalrateamong women who develop
harmfulpracticessuch as Gishiricutsand femalecircum- complications.Each of the otherinterventionscan only

30 Studies in FamilyPlanning

This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM

All use subject to JSTOR Terms and Conditions
work throughsome otherintermediatedeterminantsor
outcomes.Improvementsin thesocioeconomicstatusof
women, forexample, would have to have an effecton
theirhealthstatus,theirreproductivestatus,or theirac-
cess to or use of healthservices.Prenatalcare,the train-
ing of traditionalbirthattendants,and communityedu-
cation would each have to be linked in some way to a
referralnetworkthatincluded a range of services,prob-
ably including emergencyobstetricalcare. Therefore,
programsthatpropose interventions such as these,which
are not directlylinked to the sequence of outcomes that
culminatein death or disability,need to specifyin some
detail the mechanismsthroughwhich the intervention
is expected to operate and to considerwhetherfacilities
or programsare in place to make sure thatthosemecha-
nismswill function.
Conductinga similarexercisebeforemakinga final
decision on any particularinterventionwould provide
program officialswith a useful perspectiveon the po-
tentialimplicationsof theirdecisions. The exerciseneed
not be difficultor particularlytime-consuming.It re-
quires a basic understandingof the fullrange of factors
known to influencematernalmortalitycombinedwitha
thoroughknowledgeofthesettingintowhicha program
will be introduced.Familiaritywith the settingshould
includeknowledgeoftheservicescurrently available and
ofthecurrentbeliefsand practicesofwomen in thecom-
munityas theyrelateto pregnancy,labor,and delivery.
Previous researchhas shown thatthisinformationdoes
notrequireextensiveand expensive sample surveys.Se-
lected focus-groupdiscussions and institutionalinven-
tories can provide the needed informationrelatively
quicklyand at low cost (Preventionof MaternalMortal-
ityNetwork,1990). Programofficialscan also, without
too much difficulty, consult the resultsof researchthat
have recentlybeen summarized in several publications
(Thaddeus and Maine, 1990; Royston and Armstrong,
1989; Maine, 1991). Informationon local conditionsand
awareness of the resultsof priorresearchon the deter-
minantsof maternalmortality,combined withthe com-
prehensiveperspectiveoftheframework, should provide
programofficialswith the materialtheyneed to design
realisticand effectiveprograms.
Conclusions
The frameworkpresented above provides a context
withinwhichdiscussionsofbothresearchand programs
on maternalmortalitycan take place. Assumingthatthe
frameworkis complete(thatis, thatall possible mecha-
nisms and pathways can be accommodated),the major
task facingboth researchersand programofficialsis the
estimationoftherelativeimportanceofthevariouspaths
This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM
All use subject to JSTOR Terms and Conditions
to reducingmaternalmortality.To develop estimatesof
the relativecosts and benefitsof different interventions,
programofficialsneed to considerresearchfindingson
themagnitudeof given effectsas well as information on
the costs of differentprograms.Furtherresearchis re-
quired beforesuch cost-benefitexercises,based on ac-
tual data, can be carriedout. Exercisesof thistypecon-
ducted to date have been based largelyon assumptions,
albeit very reasonable assumptions (Maine et al., 1987;
Herz and Meascham, 1987).
In addition to cost-benefit considerations,program
officialsmust also consider the timeperiod over which
a given intervention can be expectedto influencemater-
nal mortality.When one thinksin termsof a complete
framework, itis clearthatchangesin certainfactors(usu-
ally those closest in the frameworkto the outcome) will
have a more immediate impact on maternalmortality.
Improvementsin thetreatmentofcomplications,forex-
ample, will effectthe mortalityof the currentcohortof
pregnantwomen. Improvementsin the nutritionalsta-
tus of women mighthave some immediateimpact,but
the fullimpactwill not be feltuntilan entiregeneration
of women has experiencedbetternutritionsince birth.
The timingof the effectof eliminatingfemalecircumci-
sion will depend on the age at which the procedure is
traditionallyperformed:The fartherthatage is removed
fromthechildbearingyears,thelongeritwill takeforits
eliminationto have an effecton maternalmortality.The
implicationof these differencesin the timingof the ef-
fectsof interventionsis not thatprogramsshould con-
centrateexclusivelyon interventionsthathave only im-
mediate effects;the implications are that,when they
design programs,officialsshould be aware ofthefullset
offactorsthataffectmaternalmortality, themechanisms
throughwhich theywork,the potentialpower ofdiffer-
entinterventions to reducematernaldeaths,and thetime
frameover which improvementsin mortalitycan be ex-
pected to occur.
By statingtheneed forfurther research,however,we
do not mean to underestimatewhat we already know
about both the determinantsof maternalmortalityand
the relativepower of differentinterventionsto reduce
maternaldeaths. A recentanalysis of existinginforma-
tionconcludedthatfamilyplanningprograms,programs
to providesafeand legal abortionservices,and programs
to improve the treatmentof obstetric complications
would have substantialand immediateeffectson mater-
nal mortality.Programsto increase women's status,to
extend access to prenatal care, to traintraditionalbirth
attendants,and to educate and mobilize communities
would be less likelyto have an effecton maternalmor-
talityunless theywere complementedby improvedfam-
ily planning, abortion,or labor and delivery services
(Maine, 1991).Furtherresearchon theseissues,especially
Volume 23 Number 1 January/February
1992 31
operations research linked to program interventions, Fathalla, Mahmoud. 1987. "The long road to maternaldeath." People
would providea valuable contributionto theSafe Moth- 14,3:8-9.

erhood Initiative,eitherby confirmingor by question-
ing theseconclusions.
Acknowledgments
Theauthorswishtothankseveralcolleagueswhoreadearlier
draftsofthisarticleand providedhelpfulcomments,includ-
ing Magda Ghanma,StephenIsaacs, HenryMosley,Allan
and JoeWray.JackKilcullen,Rafaeldel Rosario,
Rosenfield,
and YolandaRomanhelpedin theproduction ofthearticle,
and MaryO'Connoreditedthemanuscript.
Fortney,JudithA. et al. 1985. "Maternal mortalityin Indonesia and
Egypt."Paper presentedat theWHO InterregionalMeetingon the
PreventionofMaternalMortality,Geneva, 11-15 November.
Herz, Barbaraand AnthonyR. Meascham. 1987."The Safe Motherhood
Initiative:Proposal foraction." Washington,DC; The World Bank.
Kasongo Project Team. 1987. "Antenatal screening for fetopelvic
approach to the choice of simple in-
distocias: A cost-effectiveness
dicatorsforuse by auxiliarypersonnel."Journal ofTropicalMedicine
andHygiene87,4: 173-183.
Kaunitz,Andrew M. et al. 1984. "Perinataland maternalmortalityin a
religious group avoiding obstetricalcare." AmericanJournalofOb-
and Gynecology
stetrics 150,7: 826-831.
Kwast, Barbara E. and J.E.Liff.1988. "Factorsassociated with mater-
nal mortalityin Addis Ababa, Ethiopia."InternationalJournalofEpi-
demiology 17,1:115-121.

Notes Loudon, Irvine.1986. "Obstetriccare, social class, and maternalmor-

293: 606-608.
tality."BritishMedicalJournal
1 A similarsituationexistsinmodelsofthedeterminants
offertility Maine, Deborah. 1981.FamilyPlanning:ItsImpacton theHealthofWomen
thatincludecoitalfrequency as a proximate determinant. In-vitro and Children.New York: Columbia University,CenterforPopula-
andothertechniques
fertilization ofassistedreproduction notwith- tionand FamilyHealth.
standing, forthevastmajority ofwomen,intercourse is a neces- Programs:Optionsand Issues.New York:
. 1991. SafeMotherhood
sarycondition Butsincepatterns
forfertility. ofintercourse vary Columbia University,CenterforPopulation and FamilyHealth.
amongindividuals andsocieties,itis logicaltoincludeintercourse Maine, Deborah et al. 1987. "Preventionof maternaldeaths in devel-
as a variableina modeloftheproximate determinantsoffertility.
oping countries:Program options and practical considerations."
2 Although useofillicitinducedabortion isoftendescribed as a "com- Paper prepared fortheInternationalSafe MotherhoodConference,
plication,"it is includedin ourframework underthecategory of Nairobi, 10-13 February.
"Healthcarebehavior/use of healthservices."The resultof this
Maine, Deborah and Regina McNamara. 1987. BirthSpacingand Child
particularbehavior canbehemorrhage twoofthemore
orinfection,
Survival.New York: Columbia University,Center forPopulation
prevalent complications.
and FamilyHealth.
3 Although we presentseparatesectionsonimplicationsforresearch
Mosley,W. Henry.1990.Personal communication,December.
and implicationsforprograms, in factthetwoissuesare closely
related.Forexample,resultsfromresearch thatilluminate
thepro- Mosley, W. Henry and Lincoln C. Chen. 1984. "An analyticalframe-
cessculminatinginmaternal mortalitywillinmanycaseshaveob- workforthestudyofchild survivalin developingcountries."Popu-
viousimplicationsforprograms. Hence,thedivisionofthesesec- ReviewlOS: 25-45.
lationand Development
tionsshouldbe viewedas a convenience. National CenterforHealth Statistics.1987. VitalStatisticsoftheUnited
canbe foundin
accountoftheseinterventions
4 A moreexhaustive States1983: Vol. II-Mortality,PartA. Hyattsville,MD: US Depart-
Maine(1991). mentofHealth and Human Services.
Omran, A.R. and C.C. Standley (eds.) 1976. FamilyFormation Patterns
Collaborative
and Health:An International Studyin India,Iran,Leba-
and Turkey.
non,Philippines, Geneva: WHO.
References Omu, A.E. 1981. "Traditionalmidwives in Nigeria." Lancet1:620-621.
PreventionofMaternalMortalityNetwork.1990."Barriersto treatment
Bhatia,J.C. 1985. "Maternal mortalityin AnanthapurDistrict,India: of obstetricalemergenciesin West Africa:A reporton focus group
Preliminaryfindingsof a study." Paper presented at the World researchin Nigeria, Ghana and Sierra Leone." Backgroundpaper
Health Organization(WHO) InterregionalMeetingon the Preven- forthe Safe Motherhood Conference,Abuja, Nigeria, 11-13 Sep-
tionofMaternalMortality,Geneva, 11-15 November. tember.
Boerma,J.Ties. 1987. "Levels ofmaternalmortalityin developingcoun- Rooks,J.P.et al. 1989. "Outcomes ofcare in birthcenters:The National
tries."StudiesinFamilyPlanning18,4: 213-221. BirthCenterStudy."New EnglandJournal ofMedicine321,26: 1,804-
Bongaarts,John.1978. "A frameworkforanalyzing theproximatede- 1,811.
terminantsoffertility." Review4, 1: 105-
Populationand Development Rosenfield,Allan and Deborah Maine. 1985. "Maternal mortality-a
132. neglectedtragedy:Where is the M in MCH?" Lancet2, 8, 446: 83-
Davis, Kingsleyand JudithBlake. 1956. "Social structureand fertility: 85.
An analytical framework." EconomicDevelopmentand Cultural Royston,Erica and Sue Armstrong(eds.) 1989. PreventingMaternal
Change4: 211-235. Deaths.Geneva: WHO.
Ekwempu, C.C. et al. 1990. "Structuraladjustmentand health in Af- fistula."
onvesicovaginal
Tahzib,Farhang.1989."Aninitiative Lancet,
rica." Lancet366: 56-57. 1,8,650:1,316-1,317.

32 Studies in FamilyPlanning

This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM

All use subject to JSTOR Terms and Conditions
Thaddeus, Sereen and Deborah Maine. 1990. Too Far to Walk:Maternal World Health Organization (WHO). 1985. Prevention
ofMaternalMor-
Mortalityin Context.New York: Columbia University,Center for tality:Reportofa WHO Interregional
Meeting,Geneva,11-15 Novem-
Population and FamilyHealth. ber1985.Geneva: WHO.
United Nations Population Division. 1988. UnitedNationsWorldPopu- . 1986. EssentialObstetricFunctionsat FirstReferralLevelto Re-
lationChart1988.New York: United Nations. duceMaternalMortality. Reportofa TechnicalWorking Group,June23-
Walker,G.J.A.et al. 1985. "Maternal mortalityin Jamaica:A confiden- 27,1986. Geneva: WHO.
tial inquiryinto all maternaldeaths in Jamaica 1981-1983." Paper . 1989. SafeMotherhoodInitiatives:ReportoftheSecondMeetingof
presentedat WHO InterregionalMeetingon thePreventionofMa- Interested
Parties.Geneva: WHO.
ternalMortality,Geneva, 11-15 November.

1992 33
Volume 23 Number 1 January/February

This content downloaded from 151.228.24.8 on Fri, 5 Sep 2014 10:12:07 AM

All use subject to JSTOR Terms and Conditions