Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

Pitfalls in Measuring the Health Status of Mexican Americans:

Comparative Validity of the English and Spanish Sickness Impact Profile


RICHARD A. DEYO, MD, MPH

Abstract: We tested a Spanish translation of the Sickness severity, however, important differences emerged. Group I respons-
Impact Profile (SIP) in a clinical study of low back pain, which es appeared to be highly valid, while group III responses did not;
included non-Hispanic patients (Group I), Mexican Americans who Group II responses appeared reasonably valid, but intermediate
used the English SIP (Group II), and Mexican Americans who used between the other groups. These differences appear unlikely to be
the Spanish SIP (group III). The reliability and clinical validity of due to clinical differences, interviewing, or translational problems
responses by these three groups were compared. Internal consisten- and seem to parallel the groups' levels of "acculturation." It may be
cy of responses by all three groups was excellent (Cronbach's alpha that certain aspects of acculturation, including familiarity with
for the overall SIP = .93 - .95). When construct validity was tested questionnaire research, critically affect the validity of responses to
by correlating SIP scores with several clinical measures of disease this questionnaire. (Am J Public Health 1984; 74:569-573.)

The large size and rapid growth of the Hispanic popula- spondents is often less than that of English-speaking respon-
tion in the United States have prompted a surge of research dents.2,9 Using a questionnaire on health beliefs which had
into this group's health needs and behavior. '-3 Mexican not been back-translated, Berkanovic found that the internal
Americans constitute the largest portion of this Hispanic consistency (Cronbach's alpha) of some scales was lower for
population, and are concentrated most heavily in the south- Spanish-speaking Hispanics than for English-speaking His-
western United States.4 panics.9 In a national survey of access to medical care,
For many types of health care research among Hispanic survey instruments were carefully back-translated, and in-
populations, it would be helpful to have a valid measure of cluded positively and negatively worded questions on the
health status. Several health-status measures have been same underlying attitudes. Contradictory responses to such
developed recently and extensively validated in general items were more common among Spanish-heritage respon-
populations and in clinical settings.5-7 All are questionnaires dents than among Whites.2
which inquire about a variety of daily functions, both Because of these problems, it was important to assess
physical and psychosocial. Such instruments are being used the internal consistency of responses to the English and
increasingly to study the impact of medical care on health, to Spanish versions of the SIP in our study. Because we
improve both treatment and resource allocation decisions. If performed extensive clinical evaluations, we were also able
such measurements are to be used for studies of Hispanic to assess the construct validity of both versions of the
populations, it is important to be sure that the Spanish questionnaire.
language version of a given questionnaire is as reliable and
valid as the English language version. Methods
To facilitate cross-cultural research among US Hispanic Patients and Research Activities
populations, one of these health status instruments, the Patients were recruited consecutively from a high vol-
Sickness Impact Profile (SIP), has been translated into ume walk-in clinic in San Antonio, Texas as part of a study
Spanish, using a Mexican American idiom. This instrument of low back pain. All gave informed consent to participate.
was designed in English, translated into Spanish, and then
independently back-translated into English by several bilin- Most patients seen in this clinic are indigent; those who are
guals. Discrepancies were resolved, and one item which employed generally hold blue collar jobs. Virtually all His-
could not be translated successfully was omitted. Weighting panics attending this facility are of Mexican American
of items in the scale by Mexican Americans was found to be heritage.
similar to weights used in the English version.8 This transla- Medical histories and examinations were performed by
tion procedure has been cited as a model for translating housestaff physicians. Clinical data were recorded on a
research instruments.9 standard coding form by the examining physicians. All
As part of a study on low back pain which was to be patients were judged to have uncomplicated mechanical low
conducted among both Engilsh and Spanish-speaking pa- back pain. Patients with disabling co-morbidity were exclud-
tients, we wanted to measure functional health status before ed, so that dysfunctions measured by the SIP would be
and after our intervention. Because of the painstaking trans- largely due to back pain alone.
lation procedure, and because the SIP had been used suc- A bilingual interviewer administered the SIP and other
cessfully in clinical studies of arthritis,'0 it seemed ideal for questions concerning demographic data and health beliefs.
this purpose. Of the 120 patients enrolled, 109 were interviewed by the
Other investigators have noted that the internal consist- same bilingual, Mexican American interviewer. Eleven pa-
ency of questionnaire responses by Spanish-speaking re- tients were interviewed by one of three "Anglo" interview-
ers, two of whom were fully bilingual and had previously
Address reprint requests to Dr. Richard A. Deyo, Assistant Professor of resided for several years in Central America. ("Anglo" is a
Medicine, Division of General Internal Medicine, Department of Medicine, common term used in the southwestern US to designate non-
University of Texas Health Science Center, 7703 Floyd Curl Drive, San Hispanic Whites.) The choice of English vs Spanish SIP was
Antonio, TX 78284. This paper, submitted to the Journal August 10, 1983, was
revised and accepted for publication November 2, 1983. made at the discretion of the interviewer.
Patients were re-examined and interviewed three weeks
© 1984 American Journal of Public Health 0090-0036/84 $1.50 after the initial walk-in visit. Both clinician and patient were

AJPH June 1984, Vol. 74, No. 6 569


DEYO

asked to assess whether the patient's pain was better, worse, questions: "Have you had any problem with 'bad nerves' or
or unchanged from the initial visit, and the SIP was repeated. psychiatric problems?"; "Are you worried that your back
pain is due to some serious illness?"; and "Do you feel sick
Questionnaire Measures all the time?" We expected higher (more dysfunctional)
The SIP is a standardized questionnaire consisting of Psychosocial Dimension scores for patients with positive
136 items grouped into 12 categories.* Each item is a responses than for those with negative responses.'2
statement describing a specific dysfunctional behavior, and Analysis
respondents indicate whether or not the item describes a
dysfunction they presently experience due to their illness. We divided the patients into three ethnic subgroups.
Scores are calculated for the overall instrument, each cate- Black and Anglo respondents were combined into one group
gory, and the two dimensions using predetermined weights since most spoke only English (Group I); Mexican Ameri-
based on the relative severity of each dysfunction."I cans were divided between those who took the English SIP
Interviewers estimated each patient's English fluency (Group II) and those who took the Spanish SIP (Group III).
on a five-point ordinal scale ranging from 1 (completely SIP reliability of the internal consistency type was
fluent) to 5 (no English at all). Ethnicity was simply judged measured with Cronbach's alpha. Test-retest reliability of
by the interviewer as Anglo, Black, or Hispanic. The inter- the SIP was assessed by Pearson correlation coefficients.
viewer also asked about any co-morbid medical conditions For validity testing, linear correlations were also performed.
(e.g., diabetes, hypertension), and recorded up to three For correlating age and Physical Dimension of the SIP, a
conditions per patient. simple bivariate Pearson correlation was performed. Be-
cause a relation of age with SIP score was expected, and
Validity Measures because age might also be related to other biologic validity
Aspects of the medical history and examination were measures, age was controlled for in assessing the other
used for comparison with SIP scores. Since none of these validity correlations by means of partial correlations. For
was a direct measure of function, very strong correlations each of the validity coefficients, 95% confidence intervals
were not expected. However, our biologic understanding of were calculated using standard methods for either bivariate
the disease suggested that each should show some correla- or partial correlations.'3'4 Since several of the validity
tion in a predictable direction with a valid measure of health measures were not continuous variables, or were not nor-
status, and that certain measures might correlate best with mally distributed, we also performed bivariate nonparamet-
specific subscales of the SIP. This type of "construct ric (Spearman) correlations between each validity measure
validation" is necessary since there is no "gold standard" of and the appropriate SIP subscale. Although not reported
patient health or function. here, these correlations were very similar to the partial
Physicians were asked to rate the severity of each correlations in nearly every instance. The major exception
patient's pain on a seven-point Likert scale, with one was the association between number of prior episodes and
extreme labeled "Almost no pain" (score = 1) and the other the SIP Physical dimension, which was much smaller after
extreme labeled "Extremely severe pain" (score = 7). controlling for age.
Patients made self-ratings of pain on a four-point ordinal
scale in response to the question, "how severe is your Results
pain?": 1 (almost none), 2 (minimal), 3 (somewhat), 4
(extremely severe). In addition to the scale rating of pain, The demographic and clinical characteristics of the
physicians were asked whether the patient appeared on three patient groups are summarized in Table 1. Demograph-
examination to be in severe pain (with a yes or no response). ically, Group III respondents were older and not as well
Spine flexion was assessed by measuring the distance educated; language differences were in the expected direc-
from fingertip to floor on full forward flexion. Patients with tion.
more severe back pain have limited flexion, resulting in a There were no substantial differences among the groups
greater finger-to-floor distance. Straight leg raising to pain in duration of the current episode of back pain, or in the
was recorded in degrees with the patient supine. Patients percentage who had previously sought care for back pain.
with more severe pain, and especially those with nerve By four other measures of disease severity, however, there
irritation, have limited straight leg raising ability, and there- was a consistent ordering of the three groups. Judged by
fore achieve fewer degrees. degree of spine flexion, straight leg raising, self-rated pain,
Our previous work suggested that SIP Physical Dimen- or physician's pain ratings, Group I consistently appeared to
sion scores showed moderate correlations with patient age, have the most severe illness, Group III an intermediate
pain severity, spine flexion, and straight leg raising. As severity, and Group II appeared to be least severely ill.
expected, the correlation with straight leg raising had a There were no substantial differences among the groups in
negative sign.'2 number of co-morbid conditions.
For validation of the psychosocial dimension of the SIP, Table 2 shows mean SIP scores for each group, includ-
we devised three questions unrelated to any SIP items, but ing the overall score, and scores for the Physical and
which we thought would distinguish patients with anxiety or Psychosocial subscales. It can be seen that there is a
psychiatric dysfunction from other patients. We recorded consistent monotonic relation between the three groups
dichotomous positive or negative responses to the following which parallels the clinical assessment of disease severity,
i.e., Group I has the highest scores on each scale (worst
*Three of the categories (Ambulation, Mobility, Body Care and Move- function); Group III has intermediate scores, and Group II
ment) are aggregated into a "Physical Dimension." Four other categories are has the lowest scores (best function).
aggregated into a "Psychosocial Dimension" (Social Interaction, Communi-
cation, Emotional Behavior, Alertness Behavior). The remaining categories To assess whether the three groups seemed to experi-
are not aggregated in any way (Eating, Work, Sleep and Rest, Household ence similar types of dysfunction, we examined the individ-
Management, Recreation and Pastimes). ual items on the SIP which received the greatest number of

570 AJPH June 1984, Vol. 74, No. 6


MEASURING MEXICAN AMERICAN HEALTH STATUS

TABLE 1-Demographic and Clinical Characteristics of Patients within Ethnic Subgroups

Patient Group

Group l: Group Il: Mexican- Group III: Mexican-


Blacks and Americans Using Americans Using
Clinical Characteristic Anglos (N = 23) English SIP* (N = 54) Spanish SIP* (N = 43)
Mean Age 42.8 34.8 50.9
Education (years) 11.4 9.6 3.7
English Fluency** 1.0 1.3 3.9
English 1st language as a child (%) 100% 16.7% 0%
English most often spoken in home (%) 100% 35.2% 16.3%
Duration of Pain (days)*** 25 20 28
Previously seen MD because of back
pain (%) 56.5 47.2 46.5
Spine flexion (inches)t 9.0 4.2 7.0
Straight leg raising (degrees) 51 76 66
Physician's pain rating4t 3.8 2.6 3.3
Patient's pain ratingftS 3.7 3.1 3.3
Number of Co-morbid Conditions 1.0 0.7 0.9
*SIP = Sickness Impact Profile.
"Fluency judged by bilingual interviewer on 5-point ordinal scale: 1 = completely fluent, 5 = no English at all.
**"For patients with chronic pain (>3 mos.) duration was coded as 99.
tFlexion measured as distance from fingertip to floor on maximum forward flexion.
#*7-point Likert Scale, 1 = almost no pain, 7 = extremely severe pain.
#*44-point Ordinal Scale, 1 = almost no pain, 4 = extremely severe pain.

positive responses within each group. The items are ranked Finally, we wished to assess the construct validity of
in Table 3 for each group according to numbers of positive both versions of the SIP by correlating scores with indepen-
responses, with ties given equal rank. It is apparent that the dent clinical measures of disease severity. Table 4 shows the
14 items had similarly high rankings among all three groups, results of this analysis. Because we expected the Physical
with most having a rank of eighth or better in every group. Dimension to correlate best with measures of pain or physi-
This crude comparison was somewhat reassuring, since cal limitations (supported by previous work'2) we have
there were no obvious qualitative differences in the com- displayed only the Physical Dimension correlations with
monly experienced types of dysfunction. these physical measures. Likewise, we expected the Psycho-
Internal consistency of the overall SIP was uniformly social Dimension to correlate best with the questions about
high among all groups, ranging from 0.93 to 0.95. When the psychiatric problems and worry (also supported by previous
Physical and Psychosocial Dimensions were examined sepa- work12) so for these variables, only correlations with the
rately, alpha was again consistently very high, ranging from Psychosocial Dimensions are displayed.
.89 to .93 for all the ethnic subgroups. It can be seen that for Group I, most of the validity
Test-retest reliability could only be assessed for a small coefficients lie between 0.24 and 0.56, placing them in the
group of patients-those who were rated by both themselves range of moderate correlations. All these associations are in
and the examining clinician to be unchanged at the time of the expected directions. The mean of all 11 coefficients
the three-week follow-up visit. For the entire patient sample (using absolute values) is 0.35.
of 120, only 10 patients met this criterion. For these 10 For Group II, the validity coefficients fall in a roughly
patients, test-retest correlation for the overall SIP was 0.88. similar range, with most lying between .11 and .51. Again all
There were only two in Group I, making correlation mean- are in the expected direction. The mean absolute value of the
ingless. There were four each in Groups II and III. For validity coefficients is 0.23.
Group II (four patients) test-retest correlation for the overall For Group III, those using the Spanish SIP, the results
SIP was .99, and for Group III (four patients), the correlation are disappointing. The range of the coefficients is lower, with
was 0.85. Given the small numbers, we consider the values none greater than 0.32. Furthermore, three of the signs are in
for Groups II and III to be roughly similar, and the values for directions opposite to those expected. These are for physi-
all 10 patients suggest that test-retest reliability for the entire cian ratings of pain severity and for straight leg raising. Even
patient sample is reasonably high. between the SIP physical dimension score and the patient's

TABLE 2-Mean Sickness Impact Profile (SIP) Scores by Ethnic Subgroups

Patient Group

Group l: Group II: Mexican- Group IlIl: Mexican- Significance*


Blacks & Anglos Americans Using Americans Using of Differences
Scale (N = 23) English SIP (N = 54) Spanish SIP (N = 43) between Groups

Overall SIP 24.4 15.7 18.9 p = .03


Physical Dimension 25.3 14.3 18.6 p = .007
Psychosocial Dimension 21.1 16.1 20.5 p = .32

'Analysis of variance.

AJPH June 1984, Vol. 74, No. 6 571


DEYO

TABLE 3-Items on the Sickness Impact Profile Receiving the Greatest Number of Positive Response
Rank for Rank for Mexican- Rank for Mexican-
Anglos and Americans Using Americans Using
Category Item (Paraphrased) Blacks (N = 23) English SIP (N = 54) Spanish SIP (N = 43)

Emotional Behavior Rub areas of body that hurt 1 4 5


Body Care Change position frequently 1 2 1
Home Management Not doing heavy work around house 2 6 7
Emotional Behavior Often moan and groan in discomfort 3 10 7
Body Care Kneel, stoop, or bend only by holding on to something 3 7 5
Home Management Do work around house only for short periods 4 6 5
Home Management Do less of daily work around house 4 1 2
Social Interaction Often express concern about health 4 3 3
Ambulation Walk shorter distances, stop often 4 5 6
Body Care Get in and out of chairs, bed by using a support 5 4 11
Sleep & Rest Sleep less at night 5 7 4
Ambulation walk more slowly 5 8 6
Body Care Make difficult moves with help, e.g., getting into cars, bathtubs 6 6 10
Mobility stay home most of the time 6 7 6

self-rating of pain, there is virtually no correlation (r = 0.07). validity may not be assured for Spanish-speaking subjects.
Associations with the more objective measures of spine Such findings may be important in interpreting outcome
flexion and straight leg raising are similarly very small. Only studies of health care interventions which utilize health
one of the 11 validity coefficients has 95% confidence limits status assessments. The results indicate that our own back
which exclude zero, even though there are almost twice as pain study will require an analysis stratified according to
many subjects as in Group I. The mean absolute value of the vesion of the SIP used.
11 coefficients is just 0.15. We tested the SIP in a group of patients who had not
In summary, it appears that responses to the SIP are been examined in earlier studies of the Spanish SIP, namely
highly reliable, whether administered in English or Spanish. Spanish monolinguals. Although the SIP appeared to per-
The construct validity of responses by those who use the form well with Mexican American bilinguals,8 there is rea-
Spanish language version, however, apepars to be poor. The son to believe that Mexican Americans who speak only
validity of responses to the English SIP by Blacks and Spanish are different from bilinguals in several ways. As in
Anglos is very good. Valdity for Mexican Americans who other studies,'5 our respondents who used the Spanish
use the English version of the SIP is reasonably good but language questionnaire were older than other respondents.
probably is intermediate between the other groups. They were also very poorly educated, less fluent in English,
more likely to have learned Spanish as a first language, and
Discussion less likely to speak English at home. Thus, these patients
appear to be less "acculturated" by several criteria than
Our results add yet another caution to the growing Mexican Americans who responded in English and were
literature on methodologic difficulties in health care research bilingual.
among US Hispanics.29"'5 Even when high levels of reliabil- One could argue that clinical differences between the
ity are achieved in a health status questionnaire, clinical groups might account for the differences in validity. The

TABLE 4-Construct Validity of the Sickness Impact Profile (SIP): Correlations* between Scale Scores and Clinical Variables

Patient Group

Group I: Group II: Mexican- Group IlIl: Mexican-


Blacks and Americans Using Americans Using
SIP Subscale Used for Anglos (N = 23) English SIP (N = 54) Spanish SIP (N = 43)
Correlations Validating Clinical Variables r (95% C.l.) r (95% C.l.) r (95% C.l.)

Physical Dimension Age .48 (.08,.74)" .17 (-.10, .42) .18 (-.12, .46)
Pain duration (days) .28 (-.17, .64) .07 (-.21, .34) .01 (-.31, .33)
No. previous episodes .13 (-.33, .54) .02 (-.27, .31) .13 (-.22, .45)
Appears to be in severe paint .49 (.09, .76)" .11 (-.17, .37) -.16 (-.44, .15)
Physician's rating of pain .24 (-.21, .61) .20 (-.11, .48) -.06 (-.36, .26)
Patient's rating of pain .36 (-.07, .68) .40 (.15, .60)** .07 (-.24, .37)
Spine flexion .56 (.17, .80)** .51 (.27, .69)** .14 (-.18, .43)
Straight leg raising -.30 (-.64, .17) -.21 (-.45, .06) .03 (-.28, .33)
Psychosocial Dimension Always feel sickl# .43 (.01, .72)** .42 (.17, .62)** .23 (-.08, .50)
Worried about serious illnesstt .05 (-.22, .60) .16 (-.12, .42) .28 (-.03, .54)
Psychiatric problemst4 .54 (.15, .78)" .21 (-.07, .46) .32 (.01, .57)"
Mean absolute value of all validity coefficients .35 .23 .15

-Correlations are partial correlations controlling for age in every case except the correlation between Physical Dimension and age (bivariate correlation).
-The 95% confidence intervals exclude zero; significant at p < .05.
tDichotomous variable coded 1 = no, 2 = yes; as assessed by examining physician.
**Dichotomous variables coded 1 = no, 2 = yes; as reported by the patient.

572 AJPH June 1984, Vol. 74, No. 6


MEASURING MEXICAN AMERICAN HEALTH STATUS

Spanish respondents were older than English respondents Efforts to measure "acculturation" have only recently
and perhaps their disease is different. One might expect begun,'8"19 and it appears that further efforts in this regard
older patients to have greater co-morbidity, which could will be essential. Studies of well educated Spanish monolin-
affect questionnaire responses. Perhaps SIP validity is not guals, and studies of other Hispanic groups such as Cubans
uniformly good across the entire spectrum of disease sever- and Puerto Ricans, may help in determining the relative roles
ity. These explanations seem unlikely, however, since pa- of education, cultural biases, and acculturation in the appar-
tients with disabling co-morbidity were excluded, and Group ent validity of questionnaire responses. Perhaps with studies
III respondents did not report more co-morbid conditions of this sort, we will be able to better determine the variables
than the other groups. Group III had by several measures a which affect questionnaire reliability and validity, and to
degree of disease severity that was intermediate between design better cross-cultural studies.
Groups I and II. If SIP validity suffered at one or both Our data emphasize that Mexican Americans are a very
extremes of disease severity, Groups I or II should have heterogeneous group, and studies of health behavior which
shown worse validity coefficients. ignore this fact may be misleading.
Other explanations for the observed differences lie in
the translation itself or the interviewing process. Perhaps, REFERENCES
despite elaborate efforts, the Spanish translation is simply 1. Hayes-Bautista DE: Identifying "Hispanic" populations: the influence of
research methodology upon public policy. Am J Public Health 1980;
not equivalent in all connotations to the English version. 70:353-356.
Again this seems unlikely because of the very careful 2. Aday LA, Chiu GY, Anderson R: Methodologic issues in health care
translation methods and because items which could not be surveys of the Spanish heritage population. Am J Public Health 1980;
successfully translated were excluded. The Spanish version, 70:367-374.
3. Roberts RE, Lee ES: The health of Mexican-Americans: evidence from
examined by two bilingual, bicultural research assistants the human population laboratory studies. Am J Public Health 1980;
who were both long-term residents of San Antonio, was 70:375-384.
judged to be appropriate for use in this locale without 4. US Department of Health, Education, and Welfare: Health, United
alteration. The overwhelming majority of interviews were States, 1979. DHEW Pub. No. (PHS) 80-1232. Washington, DC: GPO,
1980.
conducted by the same bilingual, bicultural interviewer, so 5. Gilson BS, Gilson JS, Bergner M, et al: The Sickness Impact Profile:
that differences in interviewing style should have been development of an outcome measure of health care. Am J Public Health
minimal, and there should have been no perception of a 1975; 65:1304.
cultural distance from the interviewer by Spanish respon- 6. Brook RH, Ware JE, Davies-Avery A, et al: Overview of adult health
status measures fielded in Rand's Health Insurance Study. Med Care
dents. 1979; 17 (suppl 7):1-131.
Finally, since the Spanish respondents were very poorly 7. Sackett DL, Chambers LW, MacPherson AS, et al: The development and
educated compared to the English respondents, it may be application of indices of health: general methods and a summary of
that their understanding of the questions, even in Spanish, results. Am J Public Health 1977; 67:423-428.
8. Gilson BS, Erickson D, Chavez CT, et al: A Chicano version of the
was poorer than that of English respondents. Perhaps the Sickness Impact Profile (SIP). Cult Med Psychiatry 1980; 4:137-150.
interpretation of questions differs subtly despite the adequa- 9. Berkanovic E: The effect of inadequate language translation on Hispan-
cy of translation, or there are cultural response biases or ics' responses to health surveys. Am J Public Health 1980; 70:1273-1276.
culturally different responses to the same disease condition. 10. Deyo RA, Inui TS, Leininger JD, et al: measuring functional outcomes in
chronic disease: a comparison of traditional scales and a self-administered
For example, it may be that certain symptoms are not seen health status questionnaire in patients with rheumatoid arthritis. Med
as a legitimate reason to limit certain activities among Care 1983; 21:180-192.
traditional Mexican Americans. The hard necessity of per- 11. Carter WB, Bobbitt RA, Bergner M, et al: Validation of an internal
forming certain tasks despite illness may vary among ethnic, scaling: the Sickness Impact Profile. Health Serv Res 1976; 11:516-528.
12. Deyo RA, Diehl AK: Measuring physical and psychosocial function in
economic, or age-related subsets of the clinical population, patients with low back pain. Spine 1983; 8:635-642.
resulting in functional patterns that are unexpected within 13. Dixon WJ, Massey FJ: Introduction to Statistical Analysis. New York:
the dominant culture. Furthermore, discrepancies are often McGraw-Hill, 1969.
observed between patient and provider ratings of functional 14. Afifi AA, Azen SP: Statistical Analysis, A Computer Oriented Approach,
2d Ed. New York: Academic Press, 1979.
ability or symptom severity.'6'7 These discrepancies may 15. Howard CA, Samet JM, Buechley RW, et al: Survey research in New
vary with patient age, sex,'6 and perhaps ethnicity. Such Mexico Hispanics: some methodological issues. Am J Epidemiol 1983;
differences might also apply to the interpretation of symp- 117:27-34.
toms and physical findings used here as validity measures, 16. Nelson E, Congor B, Douglass R, et al: Functional health status levels of
primary care patients. JAMA 1983; 249:3331-3338.
and alter the expected correlations with questionnaire 17. Kaufert JM, Green S, Dunt DR, et al: Assessing functional status among
scores. elderly patients: a comparison of questionnaire and service provider
As previously suggested, there may be differences in the ratings. Med Care 1979; 17:807-817.
seriousness with which a questionnaire is perceived,9 the 18. Cuellar I, Harris LC, Jasso R: An acculturation scale for Mexican-
American normal and clinical populations. Hispanic J Behav Sci 1980;
literalness with which questions are taken, or the attentive- 2:199-217.
ness of respondents. Spanish respondents are probably less 19. Olmedo EL, Padilla AM: Empirical and construct validation of a measure
familiar with questionnaires and research in general, and of acculturation for Mexican-Americans. J Soc. Psychol 1978; 105:179-
therefore less familiar with the spirit, intent, and assump- 187.
tions with which they are administered. One might group all ACKNOWLEDGMENTS
these social and cultural explanations together by suggesting Thanks to Andrew K. Diehl, MD and to Helen Hazuda, PhD, for critical
that "acculturation" in a broad sense is a critical factor in review of the manuscript; and to Margaret Carrillo and Helen Provot for
the validity of responses to this questionnaire. This hypothe- technical assistance.
sis would explain the gradient in validity coefficients that This research was supported by a grant from the Robert Wood Johnson
Foundation Medical Practices Research and Development Program. The
was observed among the groups, which paralleled the gradi- opinions expressed are those of the author and not necessarily those of the
ent in their apparent degree of acculturation. Robert Wood Johnson Foundation.

AJPH June 1984, Vol. 74, No. 6 573

You might also like