A SHORT ACCULTURATION SCALE FOR MEXICAN-AMERICAN POPULATIONS.pdf

53Ethnicity & Disease, Volume 15, Winter 2005

A SHORT ACCULTURATION SCALE FOR MEXICAN-AMERICAN POPULATIONS

Acculturation is important to examine var-iables that differentiate members of ethnicgroups so that interventions can be appropri-ately targeted. By using a population-basedsample of Mexican-origin adults, we sought tovalidate an acculturation scale for Mexican-American populations. The acculturation in-strument included eight items adapted fromthe Acculturation Rating Scale for MexicanAmericans (ARSMA). By using principal com-ponent analysis, we calculated eigenvalues forthe eight items. The first principal componentaccounted for 66% of the variance. Languagespoken most of the time, by itself, explained62.4% of the variance of the full model,whereas birthplace, by itself, accounted for74%. Slight increases in correlation valueswere observed beyond a four-item model thatincluded language spoken most of the time,language thought, ethnic identity, and birth-place. Future studies should compare this scalewith other multidimensional scales. (Ethn Dis.2005;15:5362)

Key Words: Acculturation, Hispanic, Immi-gration, Latino, Mexican, Mexican-American

Fred Hutchinson Cancer Research Cen-ter (GC, BT, DM, SS); Department of Epi-demiology (GC, SS, TK), Department ofHealth Services (BT), University of Washing-ton; Seattle, Washington.

Address correspondence and reprintrequests to Gloria D. Coronado, PhD;1100 Fairview Ave. N.; MP-702; Seattle,WA 98109; 206-667-4216; 206-667-5977 (fax); [email protected]

Gloria D. Coronado, PhD; Beti Thompson, PhD; Dale McLerran, MS;Stephen M. Schwartz, PhD; Thomas D. Koepsell, MD, MPH

INTRODUCTION

The Hispanic population has sur-passed Blacks to become the largest eth-nic population in the United States.1,2

Data from the 2000 US Census showthat the Hispanic population grew57.9% from 1990 to 2000.1,2 This trendis partially explained by high immigra-tion rates. In fact, residents from LatinAmerican make up nearly half of theUnited States 25.8 million foreign-bornresidents.3 Immigrants from Mexico arereported to represent 40% of the totalMexican-origin population in the UnitedStates; more than one quarter of theseimmigrants arrived in the past five years.4

Previous research reported that im-migration affects the structure of rolesand values of Mexican immigrants andtheir families.5 National health surveil-lance data on disease incidence show dif-ferent patterns of occurrence among US-born Hispanics relative to foreign-bornHispanics for many health conditions,including cancer. Historic data from astudy conducted in Los Angeles county,for example, showed that Hispanics havea higher incidence than non-HispanicWhites of certain types of cancer (buccalcavity, colon, rectum, larynx, lung,breast, bladder, prostate, and testes) anda higher overall incidence for other can-cers such as stomach, gallbladder, liver,and cervix.6 Rates diverged from non-Hispanic Whites most in immigrantMexican Americans, and rates in US-born Mexican Americans fell betweenthe two groups.6 Some of the excess riskhas been attributed to lifestyle differences(eg, dietary behaviors, tobacco use, phys-ical activity, and cancer screening prac-tices).6,7 One key factor thought to pre-dict some of the difference is level of ac-culturationthe degree to which main-stream values are adopted by immigrants.

Acculturation is thought to involve

contact with and adaptation to divergentcultural customs, beliefs, and practices.The acculturation process varies by in-dividual and is influenced by degree ofinteraction with mainstream culture, ageat immigration, reasons for immigrating,etc. Acculturation is linked to attitudesand behaviors, such as language acquisi-tion and use,8,9 mental health status,9 cig-arette smoking,10 alcohol and drug use,10

and use of preventive health services.11,12

Knowing acculturation level allows re-searchers to identify groups within a cul-ture that may experience differential risksfor diseases or have distinct behavioralpatterns. Because acculturation is linkedto cultural beliefs and attitudes, under-standing level of acculturation can helpdesign health promotion programs andmaterials.

While the importance of accultura-tion is generally accepted among the sci-entific community, consensus is limitedabout factors that should be included ininstruments designed to measure it.Some studies have attempted to con-struct and validate acculturation scales(Table 1), but dimensions included insuch scales have varied markedly, in-cluding such items as language, socio-economic status, and ethnic interaction.

Previous research on acculturationhas relied on various premises to con-struct acculturation measures. A recentpremise is that acculturation is a mul-tidimensional construct and cannot beanalyzed by a single question or set ofquestions about one factor, such as ed-ucation, age, sex, or fluency in Spanishor English.13,14 A single question mayfail to distinguish important subgroupsin the immigrant population. Englishlanguage use alone may measure func-tional integration but not the adoptionof new values.15 Birthplace or generationstatus are unchanging factors and do not

54 Ethnicity & Disease, Volume 15, Winter 2005

ACCULTURATION SCALES - Coronado et al

Because acculturation is

linked to cultural beliefs and

attitudes, understanding level

of acculturation can help

design health promotion

programs and materials.

Table 1. Characteristics of previous acculturation scales for Mexican Americans

Author Year Number of Items Description of Items

Olmedo E 1978 20 Nationality and languageSocioeconomic statusPotency ascribed to the concepts of father and male

Cuellar I 1980 20 Language use and preferenceEthnic identification and classificationCultural heritage and ethnic behaviorsEthnic interaction

Ramirez M 1984 83 Demographic and linguistic characteristicsSocialization and educational historyCultural participation

Deyo RA 1985 4 Language use, preference, ability, and first languageBurnam MA 1987 26 Language use and preference

GenerationEthnic interaction

Hazuda H 1988 25 ChildhoodExperience with English vs Spanish language (2 questions)AdultProficiency in English (3 questions)Pattern of English vs Spanish language usage (10 questions)Value placed on preserving Mexican cultural origin (3 questions)Attitude toward traditional family structure and sex-role organization (7 questions)

Marin G 1987 12 Language useEthnic social relationsMedia

Marin G 1996 24 Language useLinguistic proficiencyElectronic media

Cuellar I 1995 18 each (for two scales) Language use and preferenceEthnic identity and classificationCultural heritage and ethnic behaviorsEthnic interaction

fully measure the continuous process ofacculturation.15

Approaches to measuring accultura-tion vary, but they typically rely on oneof three models that define acculturatedgroups. The first, defined as the single-continuum model, reflects replacing anative custom with a mainstream prac-tice, such as greeting an acquaintance

with a handshake rather than a kiss onthe cheek. The replacement is assumedto be consistent across all traits. A sec-ond model, known as the two-culturematrix model, defines acculturation ontwo independent axes, one representingthe traditional culture and a second rep-resenting mainstream culture.5 Themodel recognizes that an individual mayvary in adhering to and accepting thetwo cultures: he may exhibit behaviorscharacteristic of one culture, both cul-tures, or neither culture. A third model,the multidimensional model, recognizesthat accepting new cultural traits andlosing traditional traits vary.5 An indi-vidual can gain new cultural values, be-haviors, and customs, while retainingother traditional customs and values.For example, traits such as language maybe lost in one generation, while othertraits, such as religious affiliation or eth-

nic identification may persist throughmultiple generations.

In a study of Mexican-origin Hispan-ics, we assessed and validated a short ac-culturation scale. The purpose of thiswork was to construct a scale that wouldbe brief enough for large-scale studies ofpeople in population settings rather thanclinic settings. The scale was validated onthe basis of comparing single questionsand combinations of questions to the en-tire eight-item scale. The goal was to findan instrument that maintained the di-mensions of the larger scale while mini-mizing the number of questions.

METHODS

SettingThe Hispanic population in the

state of Washington is concentrated in



Yakima County, where Hispanics con-stitute 24% of the total population.16 Inthe Yakima Valley, a region that includesmany small agricultural communities,the percentage of Hispanics is estimatedat over 50%.17 This report is part of alarger community intervention studythat measured acculturation to assesscancer prevention attitudes, beliefs, andscreening practices.

Various characteristics of Mexicanimmigration to the Yakima Valley makeit appropriate for research on accultur-ation. First, unlike areas of the south-western United States, Mexican settle-ment in the Yakima Valley is a recentphenomenon. Large-scale Mexican im-migration to the valley began duringWorld War II, when the high demandfor agricultural labor led to the enact-ment of the Bracero Program (19421964), which brought more than35,000 Mexican laborers to Washing-ton.18 Since the Bracero Program endedin 1964, immigrants (predominantlyfrom Mexico) have continued to cometo the valley to find employment or tounite with family members and friendswho settled in the area.

Identifying the SampleSample selection and survey proce-

dures have been reported elsewhere.19

Briefly, the study population includedadults residing in any of the 20 com-munities of Yakima Valley. We identifiedCensus blocks by using each communi-tys geographic boundaries. Because wewished to survey enough Hispanics tomake inferences to both Hispanic andnon-Hispanic White populations, thedesign oversampled Hispanics. Censusdata (1990) were used to calculate thepercentage of Hispanic residents withineach Census block. The Census blockswere then arranged into three groups(tertiles) in order of the percentage ofHispanics. Approximately 160 house-holds were drawn from each community;50% of the sample came from the firsttertile (blocks with the highest percent-age of Hispanics), 33% of sample house-

holds came from the second tertile(blocks with the second-highest percent-age of Hispanics), and 17% of samplehouseholds came from the third tertile(blocks with the lowest percentage ofHispanics). The percentage of Hispanicsin each tertile varied by community:0%48% in the lowest percent-Hispanicgroup, 1%79.4% for the middle group,and 24.6%93.4% for the group withthe highest proportion of Hispanic resi-dents. Address lists were purchased frombulk mailing companies and overlaid onCensus block maps to identify house-holds. In areas where information was in-complete, project staff went to the com-munity to clarify addresses.

In the six communities with fewerthan 160 housing units, all householdsin the community were surveyed. With-in each randomly selected household,one adult was interviewed. Eligibilitycriteria required respondents to be 18years of age or older, have lived in thehousehold for at least the past week, andbe able to respond to questions. Wheretwo or more eligible adults lived in thehousehold, the first adult to have abirthday after December 31 was selectedfor the interview.

Survey ProceduresA number of steps were taken to

prepare residents for the face-to-face in-terviews. English and Spanish advertise-ments were played on the two mostpopular radio stations in the valley. En-glish and Spanish flyers were placed inhigh-profile areas in the 20 communi-ties. Letters introducing the study werewritten in English and Spanish and de-livered to each randomly selected house-hold. Letters described the study andgave potential respondents the oppor-tunity to telephone the project office ifthey had questions.

We used in-person interviewing be-cause many households did not havetelephones, education level of the pop-ulation was low, and many respondentscould not read well. Interviews wereconducted by 22 bilingual interviewers,

whose language proficiency was assessedby asking them to read aloud a sectionof the questionnaire in English andSpanish. Three training sessions of sixhours each were conducted by bilingualproject staff. Training addressed strate-gies for approaching households, meth-ods for asking questions in a standardmanner, methods of editing question-naires, and rules for documentinghousehold contacts and survey disposi-tions. All interviewers were tested, andonly those with adequate language pro-ficiency and demonstrated ability to se-lect respondents and approach house-holds were certified. Interviewing tookplace between October 1, 1998 and Jan-uary 31, 1999. Respondents who gaveverbal consent to participate were givena small incentive. Prior to survey imple-mentation, the survey and study proto-col were reviewed and approved by theinstitutional review board at the FredHutchinson Cancer Research Center(FHCRC).

Quality ControlTo ensure the authenticity of survey

data, completed surveys were reviewedby project staff. An additional 10% ran-dom sample of completed surveys wasselected from each interviewer at regularintervals during the interview process.Participants from this group were re-contacted and asked to verify that theinterviewer had interviewed the respon-dent listed and that household rosteringinformation was authentic. If informa-tion was incorrect and the contactedperson reported living in the homewhen the survey was conducted, ques-tionnaire data were assumed to be er-roneous. If an interviewer was consis-tently found to have problematic surveydata, all questionnaires and trackingsheets linked to him were verified, andthe survey was readministered whennecessary. Three interviewers had prob-lematic data; 46 (2.6%) surveys werereadministered.

InstrumentThe interview instrument was a

100-item questionnaire that asked about



acculturation, healthcare access, smok-ing behavior, eating patterns, cancerscreening behavior, demographics, andpesticide exposure. The interview gen-erally took 45 minutes to complete andwas administered during a single house-hold visit. The acculturation questionsare relevant for this study.

We chose not to use an existing ac-culturation scale for various reasons.First, with the exception of the scalesdeveloped by Deyo et al and Marin etal, previous scales were too lengthy tobe included in our instrument.20,21 Wechose not to use the scale developed byDeyo et al because it included onlyquestions about language, and previousresearch has suggested that acculturationis multidimensional.22 While questionsin the scale developed by Marin et aladdressed multiple dimensions, somequestions were linked to socioeconomicstatus; eg, questions about language oftelevision programs may assess the pres-ence of cable television, rather thanviewing preferences.21 Second, existingscales would require adaptation to in-clude response categories appropriate forrespondents who belong to racial or eth-nic groups other than Hispanic. Somescales, such as the Acculturation RatingScale for Mexican Americans (ARSMA),were developed for self-administrationand would have to be adapted for in-terviewer administration.23 In a face-to-face interview, the interviewer can estab-lish trust with the respondent, whichenhances response rates.

In selecting acculturation questionsfor our survey, we reviewed previous ac-culturation scales. Of nine scales de-signed for Mexican-American adults, allincluded questions about language abil-ity or preferred language use.9,2027 Fivescales included questions about the eth-nicity of ones peers (ethnic interac-tion),9,2123,25 five scales included ques-tions about preferences for Mexicanfood, music, or television (cultural her-itage).2123,26,27 Some scales included so-ciodemographic factors, such as place ofbirth, place of growing up, and gener-ation level.9,24,26

Items in the acculturation section ofour instrument were adapted from theAcculturation Rating Scale for MexicanAmericans (ARSMA).23 The originalARSMA contains 20 questions. Thescale differentiates five types of MexicanAmericans based on level of accultura-tion: very Mexican, Mexican-orientedbicultural, true biculturals, Anglo-ori-ented biculturals, and very Anglicized.The dimensions of the scale were lan-guage familiarity and usage (What lan-guage do you prefer?; What languagedo you speak?), ethnic identity andclassification (What ethnic identifica-tion does [did] your mother use?;What ethnic identification does [did]your father use?; How do you identifyyourself?; Where were you born?),cultural heritage and ethnic behaviors(In which language, English or Spanish,do you write better?; In which lan-guage, English or Spanish, do you readbetter?; Where were you raised?),and ethnic interaction (What was theethnic origin of the friends and peersyou had as a child from age 618?;What was the ethnic origin of thefriends and peers you had as a child upto age 6?). The scale was designed toprovide reliable acculturation group as-signments for both non-clinical andclinic populations (such as psychoticsand schizophrenics).

We excluded questions that mightbe considered sensitive (eg, Have youever found it difficult to get a job or apromotion because you are of Mexicandescent?).5 Sensitive questions werethought to risk project acceptance andmight have resulted in low survey re-sponse rates. Further, we excluded ques-tions that we thought assessed generalcompetence (ie, Who is the currentpresident of Mexico?; Can you iden-tify this picture [Benito Juarez]?5) orthat referred to ability to read or write(Which language do you read better?;Which language do you write bet-ter?5) because previous studies in thevalley indicated that the population isgenerally of low literacy.

The eight questions we selectedmade up the first two dimensions in theARSMA (language familiarity and useand ethnic interaction). Specifically, thetwo questions asked about the respon-dents language used most often: Whatlanguage would you say you speak mostof the time? (Spanish, English, other)and What language do you mostlythink in? (mostly in Spanish, mostly inEnglish, about the same in English andSpanish, mostly in another language,about the same in English and the otherlanguage). Three questions asked aboutethnic identity of the respondent andhis or her parents: Of the following,how do you most identify yourself?;Of the following how does/did youmother identify?; Of the following,how does/did your father identify? Re-sponse categories were Mexican, Chica-no, Mexican American, Spanish Amer-ican, Anglo American, American, andother. A final set of questions askedabout birthplace: Where were youborn?; Where was your motherborn?; Where was your father born?Response categories for these questionswere Mexico, the United States, andother. The questions we chose are thesame as those used in the HispanicHealth and Nutrition Examination Sur-vey (HHANES) to measure accultura-tion level.28 Questions were adapted byproject staff to allow response categoriesappropriate for non-Hispanic respon-dents. Thus, our analysis can be consid-ered a validation for an abbreviated (andslightly modified) version of the originalARSMA scale.

AnalysisImportant response categories were

identified for the eight variables used todetermine acculturation. We excluded re-sponses labeled Dont know and Re-fused or those which had any variationof Other (Other, Mostly in anotherlanguage, About the same in Englishand the other language) as part of theresponse. Of the eight variables we usedto form an acculturation scale, the re-



sponse categories were ordered from lowto high acculturation; responses of speak-ing and thinking in Spanish, identifyingas Mexican, and being born in Mexicorepresented the lowest acculturation re-sponse. Principal component analysis(PCA) was used to assess the internal va-lidity of our instrument. Principal com-ponent analysis forms linear combina-tions of the variables to partition the to-tal variance in a set of variables into in-dependent components. The firstprincipal component is the combinationof variables that accounts for the greatestamount of overall variation. The secondprincipal component is the combinationof variables that is uncorrelated with thefirst principal component and accountsfor the second largest share of the totalvariance. Later principal componentscontribute less to the explanation of thetotal variance. The analysis has as manyprincipal components as variables. To en-sure equal weighting among principalcomponents, variables are standardizedso that those with a high number of re-sponse categories are given a weight pro-portionate to variables with few responsecategories.

Many previous studies have used fac-tor analysis to identify items that explainportions of the variance in acculturationassignments.9,21,23 One scale relied onPCA. We chose to use principal com-ponent methods because the low numberof questions included in our instrumentwould render the assessment of factorsdifficult. Nevertheless, PCA is typicallythe first step in factor analysis. When fac-tor analysis is performed, the eigenvec-tors generated during the PCA step arerotated to achieve factor scores. If onlyone dimension exists, no rotation is re-quired. Hence, our PCA may be consid-ered as a factor analysis if there is a singleeigenvector that accounts for most of thevariance in the data. We keep only theprincipal components that have an eigen-value $1.00, which means that eachprincipal component retained must ac-count for at least as much variance as anyone single variable in the set of accultur-

ation questions. We did not consider us-ing structural equation modeling sinceno previous acculturation scale known tothese authors relied on structural equa-tion modeling, and our analyses had alow number of variables.

We conducted three tests of internalvalidity with principal components.First, we examined the proportion of to-tal variance explained by the retainedprincipal components. A dominant firstprincipal component that accounts formost of the variance in the set of ac-culturation questions indicates high in-ternal validity. Second, we examined theeigenvectors. Eigenvector values indicatethe relative importance of each item tothe total variance. Finally, we examinedthe correlation of the first principalcomponent for various subsets of the ac-culturation measures with the first prin-cipal component for the entire set ofmeasures. These correlations indicate re-dundancy among the set of accultura-tion questions, as well as internal valid-ity. In order to determine the contri-bution of specific acculturation items tothe overall model, we calculated Pearsoncorrelation coefficients comparing thePCA value of individual questions orsets of questions to the PCA value forthe full eight-item scale.

Principal component values are onlyindirectly useful for assigning accultur-ation levels. A range of principal com-ponent scores may represent Hispanicswho have the same broad acculturationlevel. That is, Hispanics of any specificacculturation level may have principalcomponent scores following an accul-turation level specific mean and vari-ance. We used normal distribution finitemixture models to determine the num-ber of levels of acculturation and to as-sign individuals into broad accultura-tion levels.29 The finite mixture modelfinds level-specific means and variancesthat maximize the likelihood for the ob-served distribution of first principalcomponent scores.

For 38 observations, one or more ac-culturation questions were marked on

the questionnaire as other or miss-ing. In order to assign an acculturationlevel for these observations, values formissing variables were assigned the mostcommon value among Mexican Ameri-cans with the matching responses tonon-missing and non-other accul-turation questions.

To assess the external validity of ourinstrument, the first principal compo-nents from the four- and eight-item anal-yses were assigned as acculturation scores.We then calculated Pearson correlationcoefficients for the acculturation scoreswith individual demographic character-istics. The characteristics included gender(male, female), number of years of edu-cation completed (no school completed,4th grade or less, 5th8th grade, 9thgrade, 10th grade, 11th grade, 12thgrade no diploma, high school gradua-tion or GED, some college but no de-gree, Associates degree in college, Bach-elors degree, Masters degree, Doctoraldegree [MD, PhD, JD]), annual house-hold income (#$5000, $5001$10,000,$10,001$15,000, $15,001$25,000,$25,001$35,000, $35,001$50,000,.$50,000), and years of residence in theYakima Valley. In addition to the rawcorrelations of each variable with the ac-culturation scores, correlations were cal-culated partialling other demographiccharacteristics.

In order to assess differences amongindividuals in the high and low accul-turation groups using the four-itemscale, frequencies of demographic char-acteristics were calculated. The charac-teristics included age (1824, 2534,3549, 501), education, household in-come, gender, years of residence in theYakima Valley, occupation (unem-ployed; agricultural, warehouse; service,technical, retail; administration, man-agement; professional; retired; student),marital status (married, living as mar-ried; widowed; divorced, separated; nev-er married), and whether the respondentlives year round in the Yakima Valley(yes, no).



Table 2. Eigenvector values for eight-item acculturation scale

Acculturation Items Eigenvector

Question 1: Language spokenQuestion 2: Language thoughtQuestion 3: Ethnic identity (self)Question 4: Ethnic identity

(mother)

0.350.360.33

0.35Question 5: Ethnic identity

(father)Question 6: Birthplace (self)Question 7: Birthplace (mother)Question 8: Birthplace (father)

0.340.380.370.35

Table 3. Correlation of individual items and sets of items with full 8-item scale

Acculturation Items R*

Q1: Language spokenQ6: Birthplace (self)Q1: Language spoken, Q2: Language thought, Q3: Ethnic identity (self)Q1: Language spoken, Q2: Language thought, Q3: Ethnic identity (self), Q6: Birthplace (self)Q1: Language spoken, Q2: Language thought, Q3: Ethnic identity (self), Q4: Ethnic identity (mother), Q5: Ethnic identity (father)

0.790.860.920.940.97

Q1: Language spoken, Q2: Language thought, Q6: Birthplace (self), Q7: Birthplace (mother), Q8: Birthplace (father)Q1: Language spoken, Q2: Language thought, Q3: Ethnic identity (self), Q6: Birthplace (self), Q4: Ethnic identity (mother), Q5: Ethnic

identity (father)

0.960.99

Q1: Language spoken, Q2: Language thought, Q3: Ethnic identity (self), Q6: Birthplace (self), Q7: Birthplace (mother), Q8: Birthplace(father)

0.98

Q1: Language spoken, Q2: Language thought, Q3: Ethnic identity (self), Q6: Birthplace (self), Q4: Ethnic identity (mother), Q5: Ethnicidentity (father), Q7: Birthplace (mother), Q8: Birthplace (father)

1.00

Q 5 question.* Pearson correlation coefficient.

RESULTS

From a sample of 2,862 addresses,2,345 households were approached forthe study. The remaining addresses werevacant buildings (190), organizations(eg, schools, churches, fire stations)(162), nonexistent dwellings (109), andbusinesses (56). Interviews were com-pleted by 1,795 individuals. There wasno answer after five or more visits in186 households, which yielded a con-servative response rate of 76.5. The re-sponse rate of the known eligible house-holds (N52,159) was 83.1%. Of the re-spondents, 735 (41%) were Hispanic.Other respondents identified themselvesas non-Hispanic White (N5954), Af-rican American (N56), Asian American/ Pacific Islander (N59), American In-dian (N579), other (N55), or values

were missing (N57). Data from 735Mexican Americans, constituting 40.4%of the total sample, were used in theseanalyses.

Of Mexican-American respondents,66.5% completed the interview inSpanish. A large percentage of Mexican-American respondents reported Mexicoas their birthplace (63.8%). An addi-tional 18.5% of Mexican Americans re-ported being born in the United Statesand having at least one parent born inMexico. The mean age of the Mexican-American respondents was 39 years,whereas the mean age of the non-Mex-ican-American White respondents was52 years. Fifty-four percent of Mexican-American respondents had completed 8or fewer years of education; 74% ofnon-Mexican-American Whites hadcompleted 12 or more years of educa-tion. Eighty percent of Mexican Amer-icans earned a household income,$25,000; 54% of non-Mexican-American Whites earned $$25,000.

The eigenvalues for complete eight-item scale show that a high proportionof the variance is accounted for by thefirst principal component (data notshown). Component 1 accounts for66% of the variance in the total modeland has an eigenvalue of 5.24. Com-ponent 2 accounts for 10% of the var-iance, but has an eigenvalue ,1.00.Only the first component is thereforeconsidered meaningful.

Each question in the scale explainsan approximately equal proportion ofthe variance in the first principal com-ponent (Table 2). The value of theCronbachs alpha for the eight items is0.92 and does not increase with the re-moval of any item, which indicates thata single scale is being measured. Squaredeigenvector values are interpretable asthe percent of the variation in the prin-cipal component that is attributable toeach item. Overall, items range from ex-plaining 11% of the variance (ethnicidentity [self]) to 14% of the variance(birthplace [self]).

Correlations of individual and com-bined acculturation items to the fullmodel showed a trend of higher corre-lation for combinations of greater num-bers of items (Table 3).Language spokenmost of the time, by itself, explained62.4% of the variance of the full model,whereas birthplace, by itself, accountedfor 74% of the variance. Only slight in-creases in correlation values were ob-served beyond a four-item model thatincluded language spoken most of thetime, language thought, respondentsethnic identity, and respondents birth-place. From this analysis, a scale con-taining four items (Q1: language spo-ken, Q2: language thought, Q3: ethnicidentity [self], and Q6: birthplace[self]), accounted for only slightly loweramount of variability than the full mod-el (0.94 vs 1.00).



Table 4. Correlation of acculturation level group assignments using the four-itemand eight-time scales*

Four-Item Scale

Eight-Item Scale

Level 1 (low)N (%)

Level 2N (%)

Level 3N (%)

Level 4 (high)N (%)

Level 1 (low)Level 2Level 3Level 4 (high)

391 (53.20)000

4 (0.54)55 (7.48)6 (0.82)

0

2 (0.27)2 (0.27)

50 (6.80)20 (2.72)

01 (0.14)

17 (2.31)187 (25.44)

* Kappa 5 .89; weighted kappa 5 .95.

Table 5. Pearson correlation coefficients of acculturation with demographic characteristics

Characteristic

Four-Item Scale

R Adjusted R

Eight-Item Scale

R Adjusted R

Gender*Years of school completedAnnual household incomeLength of residence in Yakima Valley (years)

0.110.620.280.39

0.150.560.050.38

0.120.590.280.35

0.170.530.070.33

* Adjusted for years of school completed (no school completed, 4th grade or less, 5th8th grade, 9th grade, 10th grade, 11th grade, 12th grade no diploma, high schoolgraduation or GED, some college but no degree, Associates degree in college, Bachelors degree, Masters degree, Doctoral degree [MD, PhD, JD]), annual household income(#$5000, $5001$10,000, $10,001$15,000, $15,001$25,000, $25,001$35,000, $35,001$50,000, .$50,000), and length of residence in the Yakima Valley (numberof years).

Adjusted for gender, annual household income, and length of residence in the Yakima Valley. Adjusted for gender, years of school completed, and length of residence in the Yakima Valley. Adjusted for gender, years of school completed, and annual household income.

When we examined individual ac-culturation level group assignments byusing the four- and eight-item scales, wefound a high correlation between thefull and reduced scale (Kappa5.89) (Ta-ble 4). Using a mixture distributionanalysis of the first principal compo-nent, we identified four levels of accul-turation.29 A total of 391 study partici-pants were assigned the lowest level ofacculturation when the eight- and four-item scales were used. Similarly, boththe full and reduced scale classified 187participants in the highest level of ac-culturation. Fifty-five participants wereclassified as the second level of accultur-ation on both the four- and eight-itemscales; 50 were assigned the third levelof acculturation using both scales. Theremaining 52 participants were assigneddiffering acculturation levels by usingthe four- and eight-item scales.

Pearson correlation coefficients forthe relationship between individual ac-culturation scores and demographic

characteristics varied by characteristic(Table 5). Among the demographiccharacteristics tested, the highest coeffi-cient was reported for years of schoolcompleted, and the coefficient remainedhigh after adjustment for gender, annualhousehold income, and length of resi-dence in the Yakima Valley. A substan-tial reduction in the correlation coeffi-cient was noted for annual householdincome after adjustment for gender,years of school completed, and length ofresidence in the Yakima Valley. Coeffi-cients were similar for the four-itemscale as for the eight-item scale.

When we grouped individuals whowere assigned the two lowest levels ofacculturation and those that were as-signed the two highest levels of accul-turation, we observed notable differ-ences in demographic characteristics be-tween the two combined groups (datanot shown). Specifically, we found thatMexican Americans assigned a low levelof acculturation were younger, had com-

pleted fewer years of education, had alower annual household income, andwere more likely to report working inagriculture or in a warehouse than Mex-ican Americans assigned to a high levelof acculturation.

DISCUSSION

The findings of our study suggestthat the abbreviated four-item scale hasnearly as high an internal validity as theoriginal eight items. Internal validity ofour four-item scale was demonstrated bya dominant first principal componentthat was found to account for 66% ofthe model variance, the relative equalimportance of each scale item to the to-tal model variance, and by the high cor-relation of the principal components forsubsets of acculturation measures to theprincipal component of the entire set ofmeasures. External validity was demon-strated by high Pearson correlation co-efficients for demographic characteris-tics.

The distribution of acculturationlevel group assignments using the four-and eight-item scales suggests that ac-culturation measures that assess lan-guage used for speaking, language usedfor thinking, ethnic identification (ofself ), and birthplace (of self ) producesimilar group assignments as accultura-tion scales that also assess self-identifi-cation (of parents) and birthplace (of



parents). The strong correlation be-tween acculturation level assignmentsbased on the four-item scale and thosebased on the original eight items is fur-ther supported by the high kappa value(.89).

Factor analysis performed on theoriginal ARSMA scale showed that fourfactors accounted for total model vari-ance.23 The first factor, language fa-miliarity and usage, accounted for64.6% of the total variance in accultur-ation scores, ethnic identity and gen-eration accounted for 18.9%, culturalheritage accounted for 11.4%, andethnic interaction accounted for5.2%. We found a similar distributionof variance in our study sample. Thus,factors that accounted for the largestshare of the variance in the originalARSMA were included in our shortscale.

The portion of the variance that isexplained by each of our scale questionswas consistent with that reported byCuellar et al.23 While the language fac-tor accounted for nearly 65% of theoverall difference in acculturation scores(the model variance) in the ARSMAscale, it accounted for 45% of the modelvariance in analysis performed by Mont-gomery and Orozco on the originalARSMA scale. Deyo et al reported that90% of second-generation MexicanAmericans had a working knowledge ofEnglish, whereas only 40% of first-gen-eration Mexican Americans had thisknowledge.20 Language preference anduse was found to account for 74% ofthe larger factor labeled Ethnic Loyaltyin a study conducted by Padilla et al.5

Marin et al reported that the language-based factor accounted for about 55%of the total variance in their scale.21 Ol-medo et al found that language account-ed for 40% of the total model vari-ance.24 Similarly, we found that lan-guage spoken most of the time was thefirst principal component and account-ed for .62% of the overall variance.

The importance of language in ac-culturation instruments is likely related

to two factors. The first is that changesin language appear to occur more quick-ly than changes in other acculturationcomponents. Hazuda et al reported thata greater percentage of Mexican Amer-icans were highly acculturated on adultlanguage dimensions than on dimen-sions involving attitudes or values. Var-ious other studies have reported21,25 thatlanguage change can occur indepen-dently of changes in values, attitudes,and beliefs. The second factor is thatlanguage proficiency is likely to be pos-itively correlated with demographic fac-tors that may allow greater contact withmainstream society, such as education,income, and occupation. Thus, the abil-ity to speak English may result in greateropportunities for employment and ac-cess to health and social services.

Because of the importance of lan-guage in acculturation, Deyo et al sug-gested that a language-based accultura-tion scale may be useful in studyingclinical phenomena, such as doctor-pa-tient interactions and patient compli-ance.20 Other authors, however, havediscouraged the use of language prefer-ence as a proxy for acculturation statusin assessing cancer screening behaviorbecause of the strong association of lan-guage use and ability with healthcare ac-cess or because it measures access tocare, rather than changes in cultural val-ues.30 Hispanics who speak English arereported to have greater access to health-care services and will score high on lan-guage-only acculturation scales. Thus,the strong association between health-care access and acculturation may ob-scure the independent effects of thesefactors on a given health outcome.

Acculturation assignments based onthe four-item scale were correlated withcertain sociodemographic characteris-tics, which is consistent with previousstudies. Various studies have assessed therelationship between acculturation andlevel of education. All of the studieshave reported higher acculturationscores for greater number of years of ed-ucation completed.20,24 Deyo et al, for

example, reported a trend of increasingmean years of education with increasinglanguage score category, beginning withan average of 3.7 years associated withthe lowest language score up to 13.3years for the highest language score (offour ordered scores).20 Olmedo and Pa-dilla also reported an increasing level ofeducation associated with increasingquartile of acculturation.24

Previous studies have also reported ahigh correlation of acculturation withannual household income. Deyo et alreported an increasing trend for higherlevels of income associated with higherlanguage scores, beginning with a meanof US$7,800 for those in the groupwith the lowest language score andUS$12,500 for the group with the high-est language score.20 We found that cor-relation diminished when other demo-graphic factors were adjusted. This re-sult is likely because of the high corre-lation between years of schoolcompleted and annual household in-come, and the greater precision withwhich education was recorded (greaternumber of available response catego-ries). Further, a greater percentage of re-spondents reported their level of edu-cation than household income. Despitedifficulties inherent in documenting ed-ucation among immigrant populations(whose members may have completedyears of education in other countries),education is nevertheless likely to bemore accurately reported than income.

Because of the strong correlation ofacculturation level with years of educa-tion completed and with annual house-hold income, some authors have sug-gested that for certain health behaviors,it is important to distinguish socioeco-nomic status and education level fromacculturation level.3032 The rationale isthat many previous scales have reliedheavily on language use alone as a sur-rogate measure for acculturation, whichmay simply measure the effects of edu-cation and socioeconomic status on as-sociated behaviors.30 Nevertheless, edu-cation by itself, may not distinguish cer-



Understanding and

characterizing the process of

cultural change is essential to

the conduct of culturally

relevant health intervention

in the Hispanic community.

tain subgroups, since contact withmainstream culture can occur in occu-pational or other settings.

The relationship between genderand level of acculturation also has beenexamined in a limited number of pre-vious studies. Burnam et al and Szapo-cznick et al reported a higher mean ac-culturation score for men than forwomen and suggested that the differ-ence may be a result of greater employ-ment and educational opportunitiesavailable for males.9 Marin et al foundno significant differences in accultura-tion scores between males and females.21

This finding was consistent with thework of Montgomery and Orozco.33

Our observation of slightly higher ac-culturation score for females is inconsis-tent with previous findings and may berelated to the high number of unmar-ried foreign-born males who are drawnto the valley to find work in agriculture,which suggests more generally that thefindings may depend on specific char-acteristics of the Hispanic population.

Our scale minimizes important lim-itations of previously developed scales.Most previous instruments are lengthyand produce high respondent burden.Five of the acculturation scales reviewedcontain 20 or more questions. 9,2326 Theextra burden may be particularly prob-lematic in studies that require longquestionnaires to assess other complexconstructs or many variables. Shorteradaptations of these scales have beenused widely in health research amongHispanics.12,28 Typically, however, thepsychometric qualities of these versionsare unknown. Some previous scales re-quire psychological judgments or in-clude questions that are potentiallythreatening.

Other studies, with the exception ofscales developed by Padilla,5 Deyo20

(language-only scale), and Burnam,9

have relied on study participants whoare college or university students or areconvenience sampled. College and uni-versity students may display limitedranges of acculturation dimensions;

most speak English as a requirement ofadmissions, and may represent the up-per range of socioeconomic status. Con-venience sampling may exclude partici-pants who are less acculturated (as theymay be less familiar with research stud-ies and may more likely be illiterate).Testing the reliability of a scale is im-portant in populations that vary in age,socioeconomic status, education, andcultural experiences. This study samplelargely comprised Hispanics who origi-nated from Mexico and were from thesame Hispanic population subgroup.Further, data were collected by usingrandom sampling techniques at thepopulation level.

Some limitations of this study alsoshould be considered. The relatively lownumber of survey questions related toacculturation limited our ability to com-pare our questions with previous scalesor to evaluate certain acculturation con-structs. Constructs that assess adherenceto traditional Mexican family values, forexample, were omitted from our scale,even though they have been document-ed in previous research as important inpredicting use of cancer screening ser-vices.30 Questions chosen for our study,by and large, arguably fail to assess fac-tors related to ones interest or willing-ness to adopt mainstream customs. Pa-dilla argues that the freedom of choicein acculturation is often overlooked inacculturation research.5 Factors such aspersonal preference for one cultural ori-entation over the other (ethnic loyalty),as demonstrated by the choice of ethnicidentification and the ethnic identity offriends and cultural awareness, may bet-ter tap the individual motivation toadopt customs and values.

Another limitation is the inability ofour instrument to assess acculturation insubgroups of Hispanics other than Mex-ican Americans. In particular, our ques-tion about ethnic identity is tailoredspecifically to the Mexican-Americanpopulation. Further, the use of differentterminology in different geographic re-gions to define Hispanics may limitassessments of Hispanics in other areas.

Another potential limitation of ourstudy is that the two-culture matrixmodel cannot be evaluated with ourdata. Such a model has been developedand tested by Cuellar et al and was pub-lished in 1995.22 It measures Mexicanand Anglo orientation by using two in-dependent axes and assesses four modesof acculturation: integration, assimila-tion, separation, and marginalization.The model defines biculturals who scorehigh on orientation toward Hispanicculture and toward mainstream culture.Similarly, the model defines personswho score low on orientation towardeach culture. Thus, the model is basedon the assumption that gaining and los-ing traits from one culture is unrelatedto gaining or losing traits from anotherculture. Some authors, however, arguethat this assessment method is con-founded by general competency and so-cial involvement. Others argue that thescale is lengthy and produces a high re-spondent burden and that there are sta-tistical limitations to using four catego-ries to define acculturation level. Thus,in order to describe the general processof acculturation in a population, thematrix model offers various advantages.However, when only a crude assessmentof acculturation is needed, complex sta-tistical analysis of other variables thatmay be related to acculturation is re-quired, or where instrument length mayreduce response rates, a shorter scale,such as ours, may be advantageous.



CONCLUSION

Various studies have shown that cul-tural factors play an important role inhealth decisions. Understanding andcharacterizing the process of culturalchange is essential to the conduct of cul-turally relevant health intervention inthe Hispanic community. Knowledge ofthe differences between groups isthought to be important for the designand implementation of health interven-tions in this population.31,34 Futurestudies should compare this scale withother multidimensional scales andshould evaluate this scale on popula-tions in other geographic areas.

ACKNOWLEDGMENTSThis work was supported by grant numberCA-74968 from the National Cancer Insti-tute.

AUTHOR CONTRIBUTIONSDesign and concept of study: Coronado, Thomp-son, McLerran, KoepsellAcquisition of data: Coronado, ThompsonData analysis and interpretation: Coronado,Thompson, McLerran, Schwartz, KoepsellManuscript draft: Coronado, Thompson, Mc-Lerran, SchwartzStatistical expertise: McLerran, SchwartzAcquisition of funding: ThompsonAdministrative, technical, or material assistance:Coronado, Thompson, KoepsellSupervision: Coronado, Thompson

REFERENCES1. US Census Bureau. Differences in Population

by Race and Hispanic or Latino Origin, for theUnited States: 1990 to 2000. (Table 4). Wash-ington, DC: Bureau of the Census; 2000.

2. US Census Bureau. Population by Race andHispanic or Latino Origin, for All Ages and for18 Years and Over, for the United States: 2000.Table 1. Washington, DC: Bureau of theCensus; 2000.

3. US Census Bureau. Coming from the Ameri-cas: A Profile of the Nations Latino AmericanForeign Born (2000 update). Washington, DC:Bureau of the Census; 2000.

4. Suarez-Orozco MM. Introduction: crossings:Mexican immigration in interdisciplinary per-spective. In: Suarez-Orozco MM, ed. Cross-ings: Mexican Immigration in InterdisciplinaryPerspective. Cambridge, Mass: Harvard Uni-versity Press; 1998:350.

5. Padilla AM, ed. Acculturation, Theory, Models,

and Some New Findings. Boulder, Colo: West-view Press for the American Association forthe Advancement of Science; 1980.

6. Menck H, Henderson B, Pike M, et al. Can-cer incidence in the Mexican American. JNatl Cancer Inst. 1975;55(3):531536.

7. Crespo CJ, Smit E, Carter-Pokras O, et al.Acculturation and leisure-time physical inac-tivity in Mexican-American adults: resultsfrom NHANES III, 19881994. Am J PublicHealth. 2001;91(8):12541257.

8. Arcia E, Skinner M, Bailey D, et al. Modelsof acculturation and health behaviors amongLatino immigrants to the US. Soc Sci Med.2001;53(1):4153.

9. Burnam MA, Telles CA, Hough RL, et al.Measurement of acculturation in a commu-nity population of Mexican Americans. HispJ Behav Sci. 1987;9(2):105130.

10. Otero-Sabogal R, Sabogal F, Perez-Stable EJ.Psychosocial correlates of smoking among im-migrant Latina adolescents. J Natl Cancer InstMonogr. 1995;18:6571.

11. Solis JM, Marks G, Garcia M, et al. Accul-turation, access to care, and use of preventiveservices by Hispanics: findings from HHA-NES 198284. Am J Public Health. 1990;80(suppl):1119.

12. Harmon MP, Castro FG, Coe K. Accultura-tion and cervical cancer: knowledge, beliefs,and behaviors of Hispanic women. WomenHealth. 1996;24(3):3757.

13. Mendoza R. Acculturation and socio-culturalvariability. In: Mendoza R, Martinez JL, eds.Chicano Psychology. New York, NY: AcademicPress; 1984:6175.

14. Garza RT, Gallegos PI. Environmental influ-ences and personal choice: a humanistic per-spective on acculturation. In: Padilla A, ed.Hispanic Psychology: Critical Issues in Theoryand Research. Thousand Oaks, Calif. SagePublications, Inc; 1995:314.

15. Suarez L, Lloyd L, Weiss N, et al. Effect ofsocial networks on cancer-screening behaviorof older Mexican-American women. J NatlCancer Inst. 1994;86(10):775779.

16. US Department of Commerce. 1990 Censusof Population and Housing, Yakima, Washing-ton, USA. Washington, DC: Bureau of theCensus; 1995.

17. US Bureau of the Census. Detailed Tables:Yakima County, Washington. Washington,DC: US Census Bureau, American Fact Find-er; 2002.

18. Gamboa E. Mexican Labor and World War II:Braceros in the Pacific Northwest, 19421947.1st ed. Austin, Tex: University of Texas Press;1990.

19. Thompson B, Coronado GD, Solomon CC,et al. Cancer prevention behaviors and socio-economic status among Hispanics and non-Hispanic Whites in a rural population in theUnited States. Cancer Causes Control. 2002;13(8):719728.

20. Deyo RA, Diehl AK, Hazuda H, et al. A sim-

ple language-based acculturation scale forMexican Americans: validation and applica-tion to healthcare research. Am J PublicHealth. 1985;75(1):5155.

21. Marin G, Sabogal F, Marin BV, et al. Devel-opment of a short acculturation scale for His-panics. Hispanic J Behav Sci. 1987;9(2):183205.

22. Cuellar I, Arnold B, Maldonado R. Accultur-ation Rating Scale for Mexican Americans-II:a revision of the original ARSMA scale. His-panic J Behav Sci. 1995;17(3):275304.

23. Cuellar I, Harris LC, Jazzo R. An accultura-tion scale for Mexican-American normal andclinical populations. Hispanic J Behav Sci.1980;2(3):199217.

24. Olmedo EL, Padilla AM. Empirical and con-struct validation of a measure of acculturationfor Mexican Americans. J Soc Psychol. 1978;105:179187.

25. Hazuda HP, Stern MP, Haffner SM. Accul-turation and assimilation among MexicanAmericans: scales and population-based data.Soc Sci Q. 1988;69:687706.

26. Ramirez MI. Assessing and understanding bi-culturalism-multiculturalism in Mexican-American adults. In: Mendoza R, MartinezJL, eds. Chicano Psychology. New York, NY:Academic Press; 1984:7794.

27. Marin G, Gamba RJ. A new measurement ofacculturation for Hispanics: The Bi-dimen-sional Acculturation Scale for Hispanics(BAS). Hispanic J Behav Sci. 1996;18(3):297316.

28. Delgado JL, Johnson CL, Roy I, et al. His-panic Health and Nutrition Examination Sur-vey: methodological considerations. Am JPublic Health. 1990;80(suppl):610.

29. Qu P, Qu Y. A Bayesian approach to finitemixture models in bioassay via data augmen-tation and Gibbs sampling and its applicationto insecticide resistance. Biometrics. 2000;56(4):12491255.

30. Suarez L, Pulley L. Comparing acculturationscales and their relationship to cancer screen-ing among older Mexican-American women.J Natl Cancer Inst Monogr. 1995;18:4147.

31. Negy C, Woods DJ. The importance of ac-culturation in understanding research withHispanic Americans. Hispanic J Behav Sci.1992;14(2):224247.

32. Negy C, Woods DJ. A note of the relation-ship between acculturation and socioeconom-ic status. Hispanic J Behav Sci. 1992;14(2):248251.

33. Montgomery G, Orozco S. Validation of ameasure of acculturation for Mexican Amer-icans. Hispanic J Behav Sci. 1984;6:5363.

34. Flores ET, Castro FG, Fernandez-Esquer ME.Social theory, social action, and interventionresearch: implications for cancer preventionamong Latinos. J Natl Cancer Inst Monogr.1995;18:101108.

A SHORT ACCULTURATION SCALE FOR MEXICAN-AMERICAN POPULATIONS.pdf

Documents

Transcript of A SHORT ACCULTURATION SCALE FOR MEXICAN-AMERICAN POPULATIONS.pdf