AMERICAN JOURNAL OF EPIDEMIOLOGY Vol. 124, No. 2Copyright © 1986 by The Johns Hopkins University School of Hygiene and Public Health Printed in U.Sj\.All rights reserved

SOCIAL AND CULTURAL FACTORS IN PREGNANCYCOMPLICATIONS AMONG NAVAJO WOMEN

W. THOMAS BOYCE,1 CATHERINE SCHAEFER,1 H. ROBERT HARRISON,'WILLIAM H. J. HAFFNER,3 MARGUERITE LEWIS,1 AND ANNE L. WRIGHT1

Boyce W. T. (Dept of Pediatrics, U. of Arizona, Tucson, AZ 85724), C. Schaefer,H. R. Harrison, W. H. J. Haffner, M. Lewis, and A. L. Wright Social and culturalfactors in pregnancy complications among Navajo women. Am J Epidemiol1986;124:242-53.

A population of 968 pregnant Navajo women was followed in a prospectivestudy conducted from 1980 to 1983 at the Indian Health Service Hospitals inGallup and Crownpoint New Mexico. The purpose of the study was to examinesocial and cultural influences on obstetric and neonatal complications. The extentof traditional cultural practices and the availability of social support were ascer-tained in structured interviews completed during each woman's first prenatalvisit In a subsample of women, the occurrence of stressful life events was alsomeasured during a final prenatal visit in the third trimester of pregnancy. Con-trolling for a variety of conventional risk factors and other potential confounders,traditional women sustained complications at a rate greater than twice that ofthe least traditional, most acculturated women (approximate relative risk = 2.1;p = 0.001). Social support and life events were modestly associated with maternalcomplications (approximate relative risk = 0.7, 0.8, respectively; p = 0.07), withpoorer outcomes found among those with low social support and low numbersof life events. It is proposed that the relationship of maternal complications to allthree sociocurtural variables—traditionally, social support, and life events—mayreflect the influences of social isolation on the course and outcomes of pregnancy.

acculturation; pregnancy complications; stress, psychological

Evidence from a variety of past work exposure to pathogenic agents, health be-suggests that sociocultural factors may be liefs and behaviors, and the experience ofimportant determinants of disease distri- acculturative stress. A number of studiesbutions in populations undergoing rapid have documented an increased risk ofcultural change. As reviewed by Marmot changes in both psychiatric and biologic(1) and Dubos (2), the culture in which health status under conditions of culturalpeople live may have significant effects on transition and social change. The work ofpatterns of disease through intervening in- Cassel and Tyroler (3), for example, dem-fluences on a range of factors—including onstrated significant elevations in reported

Received for publication January 4, 1985, and in vereity of Arizona, Tucson, AZ 85724.final form January 10, 1986. This work was supported by a grant from the

1 Department of Pediatrics, College of Medicine, Thrasher Research Fund and was presented in part atUniversity of Arizona, Tucson, AZ. the 1985 annual meeting of the Society for Epidemi-

1 Centers for Disease Control, Atlanta, GA. ologic Research in Chapel Hill, NC. Dr. H. Robert' Department of Obstetrics and Gynecology, Uni- Harrison is the recipient of a John A. and George L.

formed Services University of the Health Sciences, Hartford Fellowship.Bethesda, MD. Dr. William H. J. Haffner was for- The authors gratefully acknowledge the contribu-merly with the Obstetrics and Gynecology Service, tions of Dr. E. Russell Alexander, Dr. Dalice Sim,Navajo Area Indian Health Service, Gallup, NM. Julie Arthur, Cathy Arthur, Roxanne Gutman, and

Reprint requests to Dr. W. Thomas Boyce, Depart- the medical staffs at the Indian Health Service Hos-ment of Pediatrics, University Medical Center, Uni- pitals in Crownpoint and Gallup, New Mexico.

242

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from

PREGNANCY COMPLICATIONS AMONG NAVAJO WOMEN 243

symptoms among first generation factoryemployees who had recently undergone achange from a rural, folk culture to anindustrial social setting. Other work, suchas that by Leighton et al. (4) in rural NovaScotia, found an increased risk of psychi-atric impairment under conditions of socialdisorganization, and Scotch (5) and others(6, 7) reported higher rates of hypertensionin populations of traditional people exposedto modern, urban societies. Not only isacculturative change often accompanied byalterations in health, but continued affili-ation with the more traditional culture oforigin appears to be in some cases protec-tive. The studies of Marmot and Syme (8)and Yano et al. (9) suggest that amongJapanese-American men exposure to Jap-anese culture during childhood and reten-tion of traditional cultural elements intoadult life have sparing effects in the devel-opment of coronary heart disease, inde-pendently of conventional risk factors.

As in the work cited above, most studiesof acculturation and health have found el-evated rates of disease among traditionalpeople experiencing the most intensive ex-posure to and involvement in modern, west-ern society. Nevertheless, research on cul-tural change has not been unequivocal inthis regard, and other work has found ab-sent, curvilinear, or positive associationsbetween acculturative status and health,depending in part upon the category ofoutcome studied. Beiser and colleagues (10,11), for example, found no greater preva-lence of psychologic distress or hyperten-sion among Senegalese migrants to urbanenvironments, and Rosen and Voorhees-Rosen (12) reported no differences in therates of psychiatric disturbance amongShetland Islanders who were differentiallyexposed to industrialization, following thediscovery of the North Sea oil fields. Onthe other hand, work such as that byMcGarvey and Baker (13) and Maoz et al.(14) suggests a curvilinear relationship be-tween modernity and health, with subjectshaving ambiguous cultural identities bear-ing the highest rates of disorder. Finally,

some investigators (15, 16) have reportedbeneficial effects that may accompanymodernization, including enhanced psycho-logic well-being and adjustment.

A population having both historical andcurrent experience with major socioculturalchange is the Navajo Indian people of theAmerican Southwest. The Navajo are Ath-abaskan people whose traditional way oflife is based economically on sheepherdingand subsistence agriculture, with a socialstructure dominated by a wide network ofextended family relationships. As docu-mented by Kunitz (17) and others (18),increasing economic interdependence be-tween Navajo and Anglo-American socie-ties has resulted in a growing heterogeneityin the cultural practices adopted by Navajopeople. Such practices range from a highlytraditional lifestyle—characterized by res-idence in an isolated area, unemployment,little formal education, and primary use ofthe Navajo language—to a more accultur-ated lifestyle—characterized by residencein a small town, work for wages, availabilityof modern conveniences, and the relativeneglect of traditional religious activities.Major changes in the health status of theNavajo people have occurred over the pastseveral decades (17), yet only limited atten-tion has been focused on the role of accul-turation and traditionality in the mediationof such changes.

We examined the effects of traditionalityand related social experiences on the preg-nancy outcomes of a cohort of Navajowomen, in a prospective study begun in1980. Outcomes of pregnancy were chosenas a variable of interest because of pastwork suggesting that both psychologic (19-25) and cultural (26, 27) factors may haverelevant effects on the course of pregnancyand the risk of obstetric or neonatal com-plications.

METHODSPopulation and design

A sample of convenience, consisting of968 pregnant Navajo women, was enrolledduring the first prenatal visit to the obstet-

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from

244 BOYCE ET AL.

rics clinics at two Indian Health ServiceHospitals, in Gallup and Crownpoint, NewMexico. Enrollments were completed dur-ing the two-year period between October1980 and October 1982, and the study sam-ple represented approximately 40 per centof the 2,421 women presenting for prenatalcare during that period. Recruitment wassequentially conducted during each prena-tal clinic, without prior knowledge of vari-ables intrinsic to the study design. At thetime of enrollment, basic demographic in-formation was collected, and a structuredinterview was completed to assess the avail-ability of social support and the woman'sdegree of cultural traditionality. Interviewswere conducted by a Navajo research as-sistant in either the Navajo or the Englishlanguage, according to the subject's prefer-ence. In a subsample of 566 enrollees (59per cent), a second structured interview wasconducted during a final visit in the thirdtrimester of pregnancy, in order to ascer-tain major life events occurring during thepreceding six months. Within two monthsfollowing delivery, a review of the medicalrecord was completed, and data on obstetricand neonatal complications of pregnancywere gathered, along with information onpast medical and obstetric histories. Medi-cal record reviews were conducted withoutknowledge of prenatal interview results.

Measurement of variables

Subjects reported their age, education,and the income of the household in whichthey lived. The presence of medical condi-tions with possible effects on the course ofpregnancy was ascertained, including ane-mia (hematocrit <28 per cent), diabetes,and pre-existing hypertension. Social sup-port was measured using a weighted indexof interview items modeled after the instru-ment of Berkman and Syme (28). The five-item index assessed marital status and theavailability of both emotional support andinstrumental support (or aid). For purposesof analysis, scores were collapsed into adichotomous variable representing high orlow availability of social support. Life

events were measured using a 40-itemchecklist derived from the instruments ofHolmes and Rahe (29) and Paykel et al.(30). Several items were added to addressmajor events specific to the Navajo culture,and a general question eliciting otherstressful occurrences was presented at theend of the interview.

Subjects' degree of traditionality wasevaluated using an abbreviated instrumentderived in part from the work of Milliganet al. (31). Classical factor analysis of in-terview responses revealed two factors, cor-responding to the traditionality of culturalpractices and the modernity of the homeenvironment. Five items in the first factorattained loadings of 0.45 or greater andwere retained as the measure of tradition-ality. These were 1) religious affiliation(Navajo Way or Native American Churchversus Christian or none), 2) use of cornpollen in ritual observances, 3) use of atraditional healer, 4) experiencing a Ki-naalda (Navajo puberty ceremony), and 5)planning a Blessingway ceremony for thebaby. In the subsequent analysis, tradition-ality scores were converted to a three-levelvariable, representing the most traditional(highest quartile), the transitional (middletwo quartiles), and the least traditional(lowest quartile). Three items from the sec-ond factor had loadings of 0.45 or greaterand were used as a measure of the modern-ity of the home. These items were the pres-ence of electricity, running water, and atelephone.

Outcome variables evaluated were thematernal and neonatal complications listedin table 1. Maternal complications includedprenatal events (such as vaginal bleedingand pre-eclampsia), intrapartum events(such as prolonged labor and Cesarean sec-tion), and post-partum problems (such asendometritis and post-partum fever). Neo-natal complications included prematurityand its associated problems (such as lowbirth weight and respiratory distress syn-drome), as well as conditions also affectingfull-term infants (such as birth asphyxiaand meconium aspiration). All maternal

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from

PREGNANCY COMPLICATIONS AMONG NAVAJO WOMEN 245

TABLE 1

Outcome variables among 968 Navajo women atGallup and Crownpoint, New Mexico, Service Units of

the Navajo Indian Reservation, 1980-1982

Maternal complications Neonatal complications

Vaginal bleedingPre-eclampsiaGestational diabetesSpontaneous abortionPremature rupture of

membranes (>1 hourprior to onset of labor)

Prolonged labor (a24hours)

Cesarean section,nonrepeat

AmnionitisEndometritisPost-partum urinary tract

infectionPost-partum fever (>100

F (37.8 C))

Prematurity (<36weeks)

Low birth weight(<2500g)

Apgar score <7SepsisRespiratory distress

syndromePneumoniaSeizureMeconium aspirationStillbirth or neonatal

death

and neonatal complications assessed wereevents requiring active medical interven-tion.

Statistical analysis

Maternal and neonatal complicationswere treated as dichotomous variables(present/absent) in the analysis of data.Bivariate relationships among pairs ofindependent variables and between inde-pendent and outcome variables were ex-amined using Pearson correlations and chi-square analyses, respectively. At the mul-tivariate level, multiple logistic regressionwas used to estimate the approximate rel-ative risk associated with an individual pre-dictor variable, while controlling for thepotentially confounding effects of other in-dependent variables. Finally, a confoundersummarization procedure (32) was used toexamine the association between tradition-ality and complications of pregnancy atvarious strata of confounder scores.

RESULTS

At least one complication occurred in 46per cent of women and 14 per cent of in-fants. These rates are comparable to thosereported by Nuckolls et al. (21), who found

maternal or neonatal complications in 47per cent of pregnancies, and by Norbeckand Tilden (25), who reported intrapartumcomplications in 39 per cent of cases andneonatal complications in 11 per cent. Therates for specific maternal complicationsranged from 1.5 per cent for spontaneousabortion to 11 per cent for pre-eclampsiaand post-partum fever. In the case of neo-natal outcomes, rates of specific complica-tions ranged from 0 per cent for seizures to10 per cent for an Apgar score less than 7.

Among independent variables, measuresof psychosocial experience and tradition-ality were normally distributed, with meansat or near the midpoint of each scale. Ma-ternal age ranged from 13 to 45 years, witha mean of 24 years, and parity ranged fromzero to 19, with a mean of 1.5. Income levelswere uniformly low, with only 18 per centof households reporting yearly incomesabove $10,000. Twenty-six per cent of sub-jects had experienced complications duringpast pregnancies, and 5 per cent had achronic medical condition. In keeping withprevious work on the health behaviors ofpregnant Navajo women (31), only 50 percent of the sample presented for prenatalobstetric care in the first trimester of preg-nancy, and 6 per cent were not seen untilthe third trimester.

Table 2 shows the interrelationshipsamong pairs of independent variables. Dueto the number of correlations examined,only those significant at the p < 0.01 levelor less are noted. Significant associations,in the expected directions, were foundamong the basic demographic variables ofage, education, and income. Medical andobstetric data also demonstrated the ex-pected relationships of age, education, andincome to parity and gestational age at thefirst prenatal visit. Among social and cul-tural variables, several potentially impor-tant interrelationships were found as well.Social support was significantly and posi-tively correlated with age, income, and par-ity. As expected, traditional women weresignificantly older, less well-educated, andmore likely to live in a less modem home

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from

246 BOYCE ET AL.

- d o

88$— odd

1 0.04

-0.0

70.

040.

00

8SS3S— dodo©

3 $$33©— d a S S d a

bs sssss4 © © ©•©'©©©

• * * •

° £ S2 SS'~tt'~tS3r^ O* C) C> ^^ ©* CO ^5

•-H©'©© ©'©'©©O

ls§S §88^3H © © © © © © © © ©

S#^SSS S2S32•"HO©©©*© ©'©*©'©'©

i 1S 1 o.

E-S

nc>in-vi

e

o. c

environment. Of importance to subsequentanalyses of independent effects on preg-nancy outcomes was the general finding ofsignificant and widespread intercorrela-tions among predictor variables.

Table 3 presents bivariate associationsbetween independent variables and the twodichotomous outcomes, maternal and neo-natal complications. Again, only differ-ences significant at p < 0.01 are noted dueto the number of comparisons made. Forneonatal complications, no significant as-sociations were found. For maternal com-plications, on the other hand, several sig-nificant predictors were found, includingboth medical and sociocultural variables.Parity was related in a curvilinear mannerto complication rates, with the highestrates found among the nulliparous andthose with parity greater than three. Pres-ent medical problems were also stronglyrelated to maternal outcomes, with morecomplications among women with suchproblems. Finally, both social support andtraditionality were significantly related tocomplication rates as well, with higher ratesfound among the most traditional and theleast supported. Within categories of tra-ditionality, there was a stepwise increase inmaternal complication rates. Lowest rateswere found among the least traditionalwomen, with intermediate and highestrates found in the transitional and mosttraditional groups, respectively.

Because of the high level of intercorre-lation noted earlier among independentvariables, multiple logistic regression anal-yses were conducted to better understandthe joint and independent relationships ofthese variables to maternal outcomes. Fol-lowing the methods of Kleinbaum et al.(33), independent variables of interest wereselected on the basis of concerns for bothprecision (i.e., statistically significant rela-tionships with maternal complications) andvalidity (i.e., relationships with maternalcomplications in past studies or theoreti-cally important relationships with other in-dependent variables). Under this proce-dure, income was eliminated because of low

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from

PREGNANCY COMPLICATIONS AMONG NAVAJO WOMEN 247

TABLE 3

Bivariate relationships between independent variables and pregnancy complications in 968 Navajo women atGallup and Crownpoint, New Mexico, Service Units of the Navajo Indian Reservation, 1980-1982

Independentvariables

Age (years)13-1819-3536-45

Education (years)0-89-111213-18

Income$0-2^99$3,000-4,999$6,000-6,999$7,000-9,999$10,000+

Parity01-34-19

Past pregnancy complicationsPresentAbsent

Present medical problemsPresentAbsent

Gestational age at first prenatal visit(weeks)

3-1314-2627-41

Social supportLowHigh

Life eventst0-12-56-36

Modernity of home (no. ofconveniences)

0123

Traditionality1 (least traditional)2 (transitional)3 (most traditional)

by variablecategory

<%>

1779

4

14364011

3821111218

345511

2674

695

6044

6

2476

225622

24202927

235226

Rate ofmaternal

complications

494561

53425040

4451464748

573952

4746

7846

494443

6543

654640

44484549

384664

Outcome variables

4.2

8.4

2.6

26.3"

0

16.0"

1.7

8.9*

6.1

1.4

10.4*

Rate ofneonatal

complications(*)

151511

19121515

1317111217

181310

1016

1416

141515

1414

141013

14121417

161612

V 1

X

0.3

2.9

3.9

4.7

3.9

0

0.2

0

1.2

1.9

1.9

• p < 0.01; "p< 0.001.t n for this variable - 566 rather than 968.

variability and the absence of an associa-tion with maternal problems. The numberof life events was omitted because of itslimited measurement in a smaller subsam-ple of the total population; a second logisticanalysis, described below, tested for theindependent effects of life events withinthe subpopulation of 566 women for whominformation on life events was available.

Logistic analyses began with a forwardselection procedure evaluating all possibletwo-factor interactions involving eithertraditionality or social support (i.e., tradi-tionality X age, social support X education,etc.). None of the interaction terms at-tained significance levels of p < 0.05, thecriterion for entry into the model. The finallogistic model including only main effects

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from

248 BOYCE ET AL.

is summarized in table 4; coefficients wereestimated using the unconditional maxi-mum likelihood method. The fit of themodel with all variables entered was ac-ceptable (p = 0.26). Parity, present medicalproblems, and traditionally contributedsignificantly and independently to the riskof maternal complications. Controlling forthe effects of other independent variables,nulliparous women had a risk of complica-tions 2.4 times that of women with paritiesof one to three, and those with pre-existingmedical problems had a risk 4.7 times asgreat as those without such problems.Within the spectrum of cultural identity,the most traditional women bore a risk ofcomplications 2.1 times greater than theirleast traditional counterparts, indepen-dently of other variables' confounding in-fluences. Finally, women with high levelsof social support and more modern homeswere slightly less likely to sustain compli-

cations, relationships that reached onlymarginal levels of significance.

One of the potential mediators of therelationship between traditionality and ma-ternal complications is the experience ofmajor life events. For this reason, a secondlogistic regression analysis was completedfor the subpopulation of women (n = 566)in which stressful life change was alsomeasured. As before, initial testing for two-factor interactions revealed no interactionterms that attained a 0.05 level of signifi-cance. Table 5 displays the approximaterelative risks of maternal complications forall independent variables, including stress-ful life events. The direction and magnitudeof logistic coefficients for independent vari-ables are congruent with those found forthe entire sample, indicating general rep-resentativeness of the life events subpopu-lation. Again, significant independent ef-fects were found for parity, present medical

TABLE 4

Associations between maternal complications and independentin 968 Navajo women at Gallup and Crownpoint, New Mexico,

1980-1982*

variables, by multiple logistic regression analysis,Service Units of the Navajo Indian Reservation,

Variable

Age (19-35 years vs.)13-1836-45

Education (0-8 years vs.)9-111213-18

Parity (1-3 vs.)04-19

Past pregnancy complications(yes, no)

Present medical problems (yes,no)

Gestational age at firstprenatal visit (trimester)

Social support (high, low)Modernity of home (no. of

conveniences)

Traditionality

Constant

• •LiOgistic

coefficient

-0.2390.242

-0.434-0.240-0.606

0.8930.298

0.236

1.556

-0.177

-0.326

-0.224

0.359

-0.078

Approximaterelative

risk

0.81.3

0.60.80.5

2.41.3

1.3

4.7

0.8

0.7

0.8

1.4<2.1)t

95%confidence

interval

0.5-1.20.6-2.6

0.3-1.20.5-1.30.3-0.9

1.7-3.50.8-2.2

0.9-1.8

2.2-10.4

0.7-1.1

0.5-1.0

0.6-1.0

1.2-1.8(1.7-2.5)t

x1

0.7

1.7

11.0

1.8

14.7

2.0

3.4

3.7

10.5

ApproximateX*to

removedegrees offreedom

2

3

2

1

1

1

1

1

1

p

0.48

0.17

0.001

0.19

0.001

0.16

0.07

0.05

0.001

* Hosmer goodness of fit test: x1 " 101, degrees of freedom " 8, p •=t Most vs. least traditional.

0.26.

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from

PREGNANCY COMPLICATIONS AMONG NAVAJO WOMEN 249

TABLE 5

Associations between maternal complications and independent variables in subpopulations with stressful lifeevents known, by multiple logistic regression analysis, in 968 Navajo women at Gallup and Crownpoint, New

Mexico, Service Units of the Navajo Indian Reservation, 1980-1982*

Variable

Age (19-35 years vs.)13-1836-45

Education (OS years vs.)9-111213-18

Parity (1-3 vs.)04-19

Past pregnancy complications(yes, no)

Present medical problems (yes,no)

Gestational age at first prena-tal visit (trimester)

Social support (high, low)Modernity of home (no. of con-

veniences)

Life events (low, medium, high)

Traditionality

Constant

Logistic

coBincient

-0.0320.116

-0.009-0.070-0.469

1.0400.629

0.462

1.270

-0.197

-0.320

-0.132

-0.269

0.366

-0.194

Approximaterelative

risk

1.01.1

1.00.90.6

2.81.9

1.6

3.6

0.8

0.7

0.9

0.8

1.4(2.1)t

95%confidence

interval

0.5-1.80.4-2.8

0.4-2.30.5-1.70.3-1.2

1.7-4.60.9-3.7

1.0-2.5

1.4-9.0

0.6-1.1

0.5-1.1

0.7-1.2

0.6-1.0

1.1-1.9(1.6-2.8)1

x'

0.1

3.7

19.4

3.8

8.2

1.4

1.9

1.0

3.4

7.0

ApproximateX*to

removedegrees offreedom

2

3

2

1

1

1

1

1

1

1

P

0.96

0.29

0.001

0.06

0.01

0.23

0.17

0.32

0.07

0.01

* Hosmer goodness of fit test:t Most v». least traditional.

11.1, degrees of freedom = 8,p-0.19.

problems, and traditionality. Life eventshad a marginally significant but inverserelationship with the risk of maternal com-plications. Women with low numbers of lifeevents had a complication risk 1.7 timesthe risk for women with high numbers ofevents (95 per cent confidence interval =1.3-2.3). Traditionality retained its signifi-cant association with maternal outcomes,with the most traditional women having a2.1 times greater risk of complications.

The association of traditionality withmaternal complications was further exam-ined through the use of Miettinen's (32)confounder summarization procedure. Us-ing the coefficients from the multiple logis-tic analysis reported in table 4 to weightthe confounding variables, cases wereranked by their predicted probability ofcomplications. The ranked cases were di-

vided into quintiles, and noncases were as-signed to the defined strata. Stratum-spe-cific estimates of the relative risk associ-ated with traditionality were homogeneousand consistent with the approximate rela-tive risks derived from the multiple logisticmodel. A test of the homogeneity of theassociation (34) across the five strata wasnonsignificant (X2H« = 2.6, with 4 degreesof freedom, p — 0.62), indicating that themeasures of association did not differ sig-nificantly at the various levels of risk. Atest for linear trend (35) in the associationof levels of traditionality with complica-tions was significant (x2M-«t = 13.5, with 1degree of freedom, p < 0.001), confirmingthe results of the multiple logistic regres-sion.

A more detailed review of the associationbetween traditionality and maternal com-

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from

250 BOYCE ET AL.

plications revealed that the relative risk foreach individual category of complication(i.e., vaginal bleeding, pre-eclampsia, ges-tational diabetes, etc.) was increased in themore traditional women. The relationshipsof traditionally to complications cannottherefore be explained by a limited subsetof obstetric problems. Further, comparisonof prenatal versus post-partum complica-tions showed no differences in the magni-tude of the associated relative risks.

A final consideration was the possibilitythat the relationship of traditionally tomaternal complications was due to sys-tematic differences in either nutritionalstatus or illness reporting among the mosttraditional women. Since the infant'sweight for gestational age is known to co-vary with maternal nutritional status, theproportion of small, average, and large forgestational age infants was examinedwithin the three categories of traditionality.Similarly, illness reporting across levels oftraditionality was assessed by comparingthe proportions of women reporting eitherno intercurrent illnesses, an illness withoutfever, or an illness with fever in the courseof prenatal care. If traditional women hadsystematically underreported symptoms, asignificantly larger proportion would be ex-pected to have no illnesses recorded, and asmaller proportion would have had knownillnesses, with or without fever. No signifi-cant differences were found among womenat the three levels of traditionality foreither the proportion of infants small forgestational age (x2 = 1.1 with 4 degrees offreedom, p = 0.90) or the pattern of re-ported intercurrent illness experience (x! =5.5 with 4 degrees of freedom, p = 0.24).

DISCUSSION

This study found a significant associa-tion between obstetric outcomes and degreeof traditionality among pregnant Navajowomen. Controlling for a variety of conven-tional obstetric risk factors, traditionalwomen were found to experience compli-cations of pregnancy at a rate greater thantwice that of their least traditional coun-

terparts. This association is unlikely to beaccounted for by the effects of either clas-sification or selection bias. Misclassifica-tion is rendered less likely by the fact thatthe study was prospective in design andthat all sociocultural and medical assess-ments were made independently, withoutknowledge of complementary study data.Interviews were conducted by a single Na-vajo research assistant, and all medical rec-ord reviews were completed by the sameteam of three investigators, who wereblinded to the previously collected data.

A selection bias, in which the most tra-ditional Navajo women might have avoidedhospital deliveries or prenatal care exceptwhen complications arose, is also unlikely.The current rate of home delivery on theNavajo reservation is estimated between 0per cent (17) and less than 1 per cent (B.C. Milligan, Navajo Area Indian HealthService, personal communication, 1985). Itis therefore improbable that uncomplicatedbirths to traditional women took place athome. Further, if traditional women hadtended to present for prenatal care (and beenrolled in the study) only when compli-cations developed, traditionality wouldhave appeared disproportionately associ-ated with prenatal problems. As notedabove, however, the rates of prenatal andpost-partum complications were equally in-creased among the most traditional sub-jects. Finally, it is possible that our 40 percent nonrandom sample of women present-ing for prenatal care was not representativeof the entire population from which it wasdrawn. However, no aspect of the study'sdesign or enrollment methods would haveclearly predisposed toward oversamplingwomen within specific categories of expo-sure or outcome variables.

What then accounts for the relationshipof traditionality to obstetrical outcomes?Since it is unlikely that traditionality perse influences complications of pregnancy,we have regarded traditionality as a markeror proxy indicator for some other variableor condition with more direct and biologi-cally plausible influences on the course of

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from

PREGNANCY COMPUCATIONS AMONG NAVAJO WOMEN 251

pregnancy. A number of possible explana-tions can be raised. First, the associationmay reflect systematic differences in com-pliance with medical recommendations orother health behaviors among the most tra-ditional women. A relative lack of familiar-ity or comfort with western medical prac-tices, along with greater distances betweenhome and hospital, may have altered thetraditional women's use of available healthservices. Such alterations in the use of med-ical care could have influenced both thecourse of a given complication or the like-lihood that a problem would reach medicalattention.

Several aspects of the data are inconsist-ent with this account, however. The asso-ciation of traditionality with maternal com-plications was independent of the timing ofthe first prenatal visit, a variable that is atleast partially reflective of health behaviorand the women's comfort with the hospitalsetting. Moreover, no differences in thereporting of intercurrent illnesses werefound among women in the three categoriesof traditionality, suggesting that the mosttraditional were no less likely to conveycomplaints or symptoms to their physi-cians. Finally, it is unlikely, for most of thecomplications assessed, that a delay in thereporting of early symptoms would haveincreased the likelihood of the complica-tion's occurrence. Premature rupture ofmembranes or Cesarean section for fetaldistress, for example, are generally not pre-ceded by a "subclinical" phase in whichsome form of intervention could have beenpreventive.

Second, the association of traditionalitywith maternal outcomes could representdifferences in nutritional or general healthstatus between women with different cul-tural identities. If the most traditionalwomen were in generally poorer medicalcondition, this might account for a higherrate of obstetric problems. However, ourresults are again not supportive of this ex-planation. Traditionality was significantlyrelated to complications even when con-trolling for the presence of the three most

prevalent, coexisting medical problems inthis population: anemia, diabetes, and hy-pertension. While no direct assessment ofnutritional status was made (except by theascertainment of anemia), examination ofinfants' weights for gestational age—aknown correlate of nutritional status—showed no differences along the continuumof traditionality.

Neither of the two previous explanationsfor the relationship of traditionality to ma-ternal outcomes can be excluded by thedata available from this study, yet both arecalled into question by the analyses re-viewed above. A final and, we believe,equally credible account is the possibilitythat traditionality is reflective of someother social condition or set of social con-ditions that characterizes the lives of highlytraditional Navajo women. This account issupported, in part, by the other social vari-ables identified in this study as importantto the maternal outcomes of pregnancy.Both social support and life events boreindependent and marginally significant re-lationships to maternal complications.Women with low levels of social supportwere somewhat more likely to experiencecomplications, and, paradoxically, womenreporting fewer life events sustained higherrates of complications as well. In assessingthe possible coherence of these two find-ings, it is important to note that nearly allof the life events evaluated in this studyinvolve perturbations of social relation-ships and thus reflect some form of positiveor negative interaction with other people.In the context of this population, extremelylow levels of reported life events may there-fore reflect a paucity of available socialexchanges and a life setting characterizedby marginality and isolation. With the pro-gressive acculturation of Navajo society, itis plausible to regard the very traditionalNavajo woman as a person at risk for in-creasing social isolation, as her peers andchildren move, both culturally and geo-graphically, toward a different life-styleand system of belief. It is thus reasonableto propose that the relationship of maternal

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from

252 BOYCE ET AL.

complications to all three socioculturalvariables—traditionally, social support,and life events—may reflect the influencesof social isolation on the course and out-comes of pregnancy. This formulation iscompatible both with prior work demon-strating significant relationships betweensocial support and pregnancy outcomes inhuman populations (21, 25) and with pastevidence in animal models indicating ad-verse effects of social isolation on a rangeof reproductive outcomes (36-40).

Whatever the actual link between tradi-tionality and obstetric outcomes, this studyprovides additional evidence that social andcultural processes are salient factors in thedevelopment of disease, particularly in pop-ulations experiencing major culturalchange. Increasing attention to such proc-esses in both the provision of medical careand the development of public health pro-grams may bear previously unanticipatedrewards.

REFERENCES

1. Marmot M. Culture and illness: epidemiologic ev-idence. In: Christie MJ, Mellett PG, eds. Foun-dations of psychosomatics. John Wiley and Sons,1981:323-40.

2. Dubos R. Determinants of health and disease. In:Landy D, ed. Culture, disease, and healing: studiesin medical anthropology. New York: MacMillan,1977:31-41.

3. Cassel J, Tyroler HA. Epidemiological studies ofculture change: health status and recency of in-dustrialization. Arch Environ Health 1961;3:31-9.

4. Leighton DC, Harding JS, Macklin DB, et al. Thecharacter of danger: the Stirling County study ofpsychiatric disorder and sociocultural environ-ment. New York: Basic Books, 1963.

5. Scotch NA. Sociocultural factors in the epide-miology of Zulu hypertension. Am J Public Health1963;53:1205-13.

6. Beaglehole R, Salmond CE, Hooper A, et al. Bloodpressure and social interaction in Tokelauan mi-grants in New Zealand. J Chronic Dis 1977;30:803-12.

7. Patrick RC, Prior IAM, Smith JC, et al. Relation-ship between blood pressure and modernityamong Ponapeans. Int J Epidemiol 1983; 12:36-44.

8. Marmot MG, Syme SL. Acculturation and coro-nary heart disease in Japanese-Americans. Am JEpidemiol 1976;104:225-47.

9. Yano K, Blackwelder WC, Kagan A, et al. Child-hood cultural experience and the incidence ofcoronary heart disease in Hawaii Japanese men.Am J Epidemiol 1979; 109:440-50.

10. Beiser M, Collomb H, Ravel JL, et al. Systemicblood pressure studies among the Serer of Senegal.J Chronic Dis 1976;29:371-80.

11. Beiser M, Collomb H. Mastering change: epide-miological and case studies in Senegal, West Af-rica. Am J Psychiatry 1981;138:455-9.

12. Rosen DH, Voorhees-Rosen D. The Shetland Is-lands: the effects of social and ecological changeon mental health. Cult Med Psychiatry 1978;2:41-67.

13. McGarvey ST, Baker PT. The effects of modern-ization and migration on Samoan blood pressures.Hum Biol 1979;4:461-79.

14. Maoz B, Antonovsky A, Apter A, et al. The effectof outside work on the menopausal woman. Ma-turitas 1978;l:43-53.

15. Meleis Al. Effect of modernization on Kuwaitiwomen. Soc Sci Med 1982; 16:965-70.

16. Inkeles A, Smith DH. Becoming modern: individ-ual change in six developing countries. Cambridge,MA: Harvard University Press, 1974.

17. Kunitz SJ. Disease change and the role of medi-cine: the Navajo experience. Berkeley: Universityof California Press, 1983.

18. Shepardson M, Hammond B. Change and persist-ence in an isolated Navajo community. Am An-thropol 1964;66:1029-50.

19. Grimm ER. Psychological and social factors inpregnancy, delivery, and outcome. In: RichardsonSA, Guttmacher AF. Childbearing: its social andpsychological aspects. Williams and Wilkins,1967.

20. McDonald RL. The role of emotional factors inobstetrical complications: a review. PsychosomMed 1968;30:222-37.

21. Nuckolls KB, Cassel J, Kaplan BH. Psychosocialassets, life crisis and the prognosis of pregnancy.Am J Epidemiol 1972^5:431-41.

22. Williams CC, Williams RA, Griswold MJ, et al.Pregnancy and life change. J Psychosom Res1975;19:123-9.

23. Newton RW, Hunt LP. Psychosocial stress inpregnancy and its relation to low birth weight. BrMed J 1984;288:1191-4.

24. Chalmers B. Psychological aspects of pregnancy:some thoughts for the eighties. Soc Sci Med1982;16:323-31.

25. Norbeck JS, Tilden VP. Life stress, social support,and emotional disequilibrium in complications ofpregnancy: a prospective, multivariate study. JHealth Soc Behav 1983;24:30-46.

26. Mead M. Male and female. London: Victor Gol-lanz, 1950.

27. Newton N, Newton M. Childbirth in crossculturalperspective. In: Howells JG, ed. Modem perspec-tives in psycho-obstetrics. New York: Brunner-Mazel, 1972.

28. Berkman LF, Syme SL. Social networks, hostresistance, and mortality: a nine-year follow-upstudy of Alameda County residents. Am J Epide-miol 1979; 109:186-204.

29. Holmes TH, Rahe RH. The social readjustmentrating scale. J Psychosom Res 1967;11:213-18.

30. Paykel ES, Prusoff BA, Uhlenhuth EH. Scalingof life events. Arch Gen Psychiatry 1971;25:340-7.

31. Milligan BC, Dalton M, Swoboda VC, et al. Nurs-

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from

PREGNANCY COMPLICATIONS AMONG NAVAJO WOMEN 253

ing care and beliefs of expectant Navajo women.Am Indian Q 1984;8:83-101.

32. Miettinen OS. Stratification by a multivariateconfounder score. Am J Epidemiol 1976;104:609-20.

33. Kleinbaum DG, Kupper LL, Morgenstern H. Ep-idemiologic research: principles and quantitativemethods. Belmont, California: Lifetime LearningPublications, 1982.

34. Breslow NE, Day NE. Statistical methods in can-cer research. Vol. I: The analysis of case-controlstudies. Lyon: International Agency for Researchon Cancer, 1980.

35. Mantel N. Chi-square tests with one degree offreedom: extensions of Mantel-Haenszel proce-dure. J Am Stat Assoc 1963;58:690-700.

36. Terman CR Pregnancy failure in female prairie

deermice related to parity and social environmentAnim Behav 1969;17:104-8.

37. Anderson CO, Denenberg VH, Zarrow MX. Ef-fects of handling and social isolation upon therabbit's behavior. Behavior 1972;43:165-75.

38. Sackler AM, Weltman AS, Schwartz R, et al. Pre-matemal isolation effect* on behavior and endo-crine function of offspring. Acta Endocrinol1969;62:367-84.

39. Barlow S, McElhatton P, Morrison P, et aL Ef-fects of stress during pregnancy on plasma corti-costerone levels and foetal development in themouse. J Physiol 1974;239:55P-o6P.

40. Ely DL, Henry JP. Effects of prolonged socialdeprivation on murine behavior patterns, bloodpressure, and adrenal weight J Comp PhysiolPsychol 1974;87:733-40.

at University of A

rizona on October 5, 2011

aje.oxfordjournals.orgD

ownloaded from