en to the following in designing health care surveys of. Spanish-heritage ... Administration Studies, Graduate School of Business, University of Chicago, 5720 S.
Methodological Issues in Health Care Surveys Of the Spanish Heritage Population LU ANN ADAY, PHD, GRACE Y. CHIU, PHD,
AND
RONALD ANDERSEN, PHD
Abstract: This paper examines national survey data on access to medical care to explore methodological issues associated with conducting health care surveys of Spanish-heritage persons. These include problems of identifying and sampling such groups, achieving respondent cooperation, designing valid interview protocols, and controlling biases that may result from the cultural specificity of the concepts being studied. The findings suggest that more attention should be given to the following in designing health care surveys of Spanish-heritage individuals: Cultural and economic heterogeneity of "Spanish-heritage" grouping, validi-
ty studies of health care utilization, and yea-saying tendencies related to health care attitude items. Given that there is a paucity of information available on methodological problems associated with health care surveys of Spanish-heritage persons, these analyses should serve to inform researchers of issues to be taken into account in designing such studies and to suggest useful hypotheses to explore in evaluating the validity of social survey data on minority (especially non-English speaking) populations' health care practices. (Am J Public Health 70:367-374, 1980.)
Introduction
tion, occupational stability, family size, and mobility. Less attention was given to the ethnic variables emphasized in earlier research. A number of studies have appeared in the health services research literature in recent years, examining the factors that influence the ability of the Spanish-heritage population to obtain medical care when they need it.9-'2 These studies are, for the most part, based on samples of Spanish-heritage individuals in very localized areas (cities or counties, for example). Estimates of Spanish-heritage access at the national level are rare.** There is even less information available on methodological problems that may affect the validity of data gathered from such respondents using the social survey approach. A national study of the health care of the United States' population conducted between September 1975 and February 1976, which included oversampling of Spanish-heritage persons in the Southwestern United States provides a unique opportunity to generate a comprehensive picture of medical
The Spanish-heritage population is one of the largest minority groups in the United States. During the past 25 years there has been increasing interest expressed in the health, education, and social service needs of this group. Salazar,* elaborating a theme proposed by Weaver,' describes three generations of health services research on persons of Spanish heritage. The first generation of researchers engaged in participant observation and emphasized the importance of ethnic factors and group personality traits to explain health attitudes and behaviors of this group.2 The second generation used small, often nonprobability samples and again concentrated on ethnic factors.3'5 Particular attention was devoted to the use of curanderas or curanderos as substitutes for "Anglo" or scientific medicine. The third generation tended to use probability sampling and large data bases to account for health attitudes and behavior.6-8 Important variables to third generation researchers were income, educaFrom the Center for Health Administration Studies, University of Chicago Graduate School of Business. Address reprint requests to Lu Ann Aday, PhD, Senior Research Associate, Center for Health
Administration Studies, Graduate School of Business, University of Chicago, 5720 S. Woodlawn Avenue, Chicago, IL 60637. This paper, submitted to the Journal July 16, 1979, was revised and accepted for publication November 16, 1979. The research was supported by a grant from the Robert Wood Johnson Foundation and Contract PHS HRA 230-76-0096 with the National Center for Health Services Research. *Jaime Salazar: unpublished manuscript, "Mexican-American Illness Attitudes," Los Angeles: Health Services Research Associates, Inc., 1978. Editor's Note: See also editorial, p. 353, this issue. AJPH April 1980, Vol. 70, No. 4
**Previous estimates came from a Lou Harris poll commissioned by the National Blue Cross Association in the late 1960s13 for the Hispanic population in the Southwest region of the U.S., and recent estimates come from the National Center for Health Statistics concerning use of physicians and hospitals by the Hispanic population of the entire country.'4 Currently there is a National Chicano Study (NCS) underway at the Institute for Social Research of the University of Michigan. The NCS involves a national representative sample of 2,500 Mexican-descent households and focuses on mental health needs and utilization, as well as a number of other sociocultural dimensions of Mexican-descent individuals' life styles. 367
ADAY, ET AL.
care access for the Spanish-heritage population in that region and to test some methodological hypotheses concerning the probable validity of the social survey data collected on that population. Descriptive information on access patterns per se for the Spanish heritage are reported elsewhere.'5 The analyses that follow deal with some methodological questions that have been of interest to health survey researchers concerned with studying Spanish-heritage health care problems.'6 They are based primarily on the 1975-76 national survey data.
Methodology The methodological problems to be explored include: 1) identifying and efficiently sampling groups that comprise only a small proportion of the total population in an area; 2) issues of cooperation and non-response due to cultural or language barriers; 3) technical difficulties involved in the preparation of field materials and the recruitment and training of staff for the conduct of interviews with nonEnglishspeaking respondents; and 4) biases in the data that might result from response sets, sensitivity of the subject matter relative to certain cultural groups, etc. The findings presented in the following sections represent an empirical test of the actual magnitude of these problems for a large and representative sample of Spanish-heritage people and will hopefully serve as a guide to assist other researchers and concerned with conducting valid and reliable surveys of similar populations. The Sample
Recognizing the dearth of information available on Spanish-heritage health care at the national level, the Robert Wood Johnson Foundation, Princeton, New Jersey, commissioned the Center for Health Administration Studies of the University of Chicago to include a special sample of the Spanish-heritage population of the Southwest U.S. as part of a national survey of access to medical care conducted in late 1975 and early 1976. The national sample involved interviews with 5,432 families representing the civilian noninstitutionalized population of the United States, carried out by the Center for Health Administration Studies (CHAS) and the National Opinion Research Center (NORC), University of Chicago. In each household, one adult and one child under 17 years of age (if there were a child in the family) were randomly chosen, yielding a sample of 7,787 people. The selected adults were personally interviewed and a responsible adult, usually the mother, was interviewed about the child. In addition to a general sample of the U.S. population, the sample design included oversampling of persons experiencing episodes of illness, rural Southern Blacks, and Spanish-heritage persons residing in the Southwest.*** The oversampling allowed detailed analysis of the access problems of these special groups. The data were then assigned weights to be used to correct for the oversampling and to allow estimates to be made for the total noninstitutionalized population. 368
The analyses reported here are concerned especially with the Spanish-heritage sample. The demographic composition of the Spanish-heritage sample compared to other racial groups and to the population as a whole is summarized in Table 1. The minority populations (including the Spanish heritage) have higher concentrations of children (under 18 years of age) than does the "other (majority) White" group. The smallest proportion of older adults (age 55 and over) are found among the Spanish heritage. There tends to be a somewhat higher proportion of females than males in the Southern Black sample, compared to the other groups. Spanishheritage persons tend to reside in urban areas; the vast majority of the "other nonwhite" population are Black residents of Standard Metropolitan Statistical Areas (SMSAs) particularly the inner city of these SMSAs. The majority White group is primarily middle or high-income people; the minority populations-particularly Spanish-heritage persons in the Southwest and non-SMSA Southern Blacks-tend to be low income. Of the eight per cent of the "other nonwhite" group who are not Black, most are Oriental-heritage persons or American Indians. Identifying and Sampling Spanish-Heritage Persons The U.S. Census considers a variety of approaches to identifying Spanigh-heritage individuals. These include determining: 1) the birthplace of the individual and his/her parents, e.g., Mexico, Cuba, Puerto itico, etc.; 2) whether the family has a Spanish surname; 3) whether or not Spanish was spoken in the person's home in early childhood; or 4) if the person claims to be of Mexican, Puerto Rican, Cuban, Central or South American, or other Spanish origin. In our study, we oversampled Spanish-heritage persons in the Southwestern states (Colorado, Texas, New Mexico, Arizona, California) and applied the Census definition of Spanish heritage developed for that region.: That classification was based on whether Spanish was spoken in the home ***The criteria for determining the Spanish-heritage households in the Southwest was based on the Census definitions: 1) did the family have a Spanish surname, or 2) was Spanish spoken in the home of the head or spouse of the family when he/she was d child? For the purpose of subsequent analyses reported here, people from the three other types of sample (general sample, plus episode of illness and rural Southern Black oversamples) who lived in the Southwestern states (Texas, New Mexico, Arizona, California, or Colorado) and who met these criteria were also included in the "Spanish heritage, Southwest" breakdowns, which served to increase the overall sample size for this group (from 865 to 1,092). The "Spanishheritage" category in all subsequent tables refers to these 1,092 persons. A more detailed description of the sample design is published elsewhere. Is tPlace of birth or parentage tends to be the criterion applied in the Mid-Atlantic states (New York, New Jersey, Pennsylvania). The Spanish language designation is applied in the District of Columbia and those states not in the Southwest or Mid-Atlantic regions. t*There is much confusion in the literature concerning the proper terminology to apply to this group-"Hispanic," "Chicano," "Latino," "Mexican." Many of thse terms reflect a great deal about the political or social consciousness of the persons who identify themselves with the respective categories. For the purposes of this paper, we use the Census term, "Spanish heritage," as it accurately reflects the criteria used for including selected individuals in the study.
AJPH April 1980, Vol. 70, No. 4
SPANISH HERITAGE HEALTH SURVEYS TABLE 1 -Soclodemographic Characteristics of Racial Groups Race Per Cent Distribution of Sociodemographic
Characterstics
Non-SMSA Blacks South
Nonwhite
%
Other White %
%
%
12 31 31 18 4 5
8 24 25 23 10 11
13 28 20 19 9 13
26 19 7 7
8 25 25 22 10 10
46 54
47 53
42 58
47 53
47 53
37 39 14 9 1
21 38 12 22 7
0 0 37 54 9
63 34 1 2 0
26 37 12 20 6
0 0 44 56
24 33 29 13
0 0 100 0
22 27 39 12
22 31 33 14
54 32 13
24 34 42
68 26 7
41 32 28
28 34 38
100 0 0 100
100 0 0 100
0 100 0 100
0 92 8 100
88 11 1 100
Spanish Heritage Southwest
Age (years) 1-5 6-17 18-34 35-54 55-64 65ormore
Sex Male Female Residence
SMSA Central City SMSA other Urban Non-SMSA Urban Rural Nonfarm Rural Farm
Negion
Northeast North Central South West Family Income Low Medium High Color White Black Other TOTAL
of the head or spouse when he/she was a child or whether the family had a Spanish surname. If either of these criteria were met the family was classified as being of Spanish heritage.#4 A problem with the application of this definition is that it did include a small number of people in the Spanish heritage category who did not consider their origin or descent to be Spanish, e.g., Portuguese-heritage people who had Spanishlike surnames, and Anglos whose family may have spoken Spanish, even though they did not think of themselves as Spanish heritage (see Table 2). Ninety-three per cent of persons who were classified as Spanish heritage by the surname or language definition did report a Spanish origin when asked their family's "origin or descent." The remaining seven per cent reported their primary ethnic identification as Portuguese, Anglo Saxon (English, Scottish, Welsh, French, Ger-
ttExamples: positively worded-"there
are
enough family
doctors around here." negatively worded "there is a big shortage of family doctors around here." Respondents were asked to indicate the extent to which they agreed with these items: strongly agreed, agreed, uncertain, disagreed, strongly disagreed. If the respondent tended to agree with both items, this was said to reflect a "yea-saying" tendency.
AJPH April 1980, Vol. 70, No. 4
Other
11
29
Total %
man, Irish, Italian), or American Indian.ttt These persons were included in the analysis of the Spanish heritage (based on Census definitions), however, to ensure conformity with Census estimates. The preponderance of those who were Spanish heritage, according to the Census definition for the Southwest, were Mexican heritage (83 per cent).* Others who indicated a Spanish-heritage ethnic background identified themselves as being from Puerto Rico, Central America, or other Spanish origins.**
tt1There is some evidence that the rate of not classifying one's origin or descent as Spanish heritage when other criteria are met is higher in the higher socioeconomic status groups. *Different operational definitions of Spanish heritage may, then to some degree define non-overlapping categories of people with very different socio-cultural experiences, e.g., American Indians, fifth or sixth generation Hispanics or recent Mexican immigrants. The number of observations and variables available for detailed identification and analysis of such groups is not sufficient in this particular sample. However, in general, the data suggest that the vast majority of the "Spanish heritage" persons in this sample are low or middle income and of Mexican descent. **There were a few people in the Southwestern states who were not classified as Spanish heritage, according to the Census definition, but who reported at least one (out of a maximum of three) of their ethnic origins as being Spanish (one per cent of Southwestern sample). "Cuban" was the Spanish heritage reported most often by this group. 369
ADAY, ET AL.
TABLE 2-Ethnic Origin of Spanish Heritage, Southwest United States Ethnic Origin
Mexican, Mexican American Puerto Rican Central America Cuban Other Spanish NonSpanish ethnic origin TOTAL
Per Cent of Spanish Heritage, Southwest %
83 1 3 6 7 100
Another issue concerns the variant concentration of Spanish-heritage persons (however defined) in the area comprising the sample universe. For example, some counties or census tracts may have a very high proportion of Spanishheritage persons and others quite low. The cost efficiency of a design intended to capture a reasonably large sample of these persons would be greatly increased if areas in which they are known to be clustered were sampled at higher rates. This was essentially the approach used in the 1975-76 CHAS study.*** The clustering of households resulting from this oversampling procedure resulted in a relatively high design effect for the Spanish-heritage oversample-2.7 compared to 1.8 for the sample as a whole. What this reflects is how the stratified cluster nature of the Spanish oversample departs from that of a simple random sample. The design effect should be multiplied by the standard error for an estimate computed assuming a simple random sample, to obtain a more accurate estimate of the standard error for this more complex design. The standard errors are therefore higher for samples involving oversampling of rare populations, as was the case in this particular design. The researcher must be aware of the trade-offs of cost and precision (size of standard error) that result when trying to save screening costs by clustering interviews in areas where large concentrations of the population group of interest reside. ***The Spanish oversample was composed of 75 ultimate segments distributed among 25 primary sampling units. Nineteen of these 25 primary sampling units were from the NORC master sample falling within the five Southwestern states (Arizona, California, Colorado, New Mexico, and Texas). The probabilities of selection of these sampling units were those that had been used in developing the NORC master frame. The six additional primary sampling units were developed for use in the Spanish oversample. These additional selections were made independently with probabilities proportional to 1970 estimates of the Spanish population. On the basis of available data, it was determined that an equal probability subsampling of listings from the 75 ultimate segments in the Spanish oversample would not produce the required number of Spanish households with an overall constraint of 2,250 household screens. As a result, an algorithm was developed (using 1970 Census estimates of the pro-
portion of Spanish households within each of the 75 segments) that would keep the required oversampling at a minimum, while producing a sample of 500 Spanish-heritage dwelling units. As noted earlier, Spanish-heritage individuals in the Southwestern states, who fall into the sample from one of the other sample types, were also included in the analyses reported here. 370
Non-Response Rates for Minority Samples A number of studies, both in marketing and academic research, have shown that characteristics such as potential respondents' age, marital status, social class, location, and type of residence are related to whether or not they agree to participate in social surveys.'6 Surveys conducted of ethnic (or racial) minorities are usually assumed to have higher non-response rates. Inherent problems such as the suspicion on the part of the respondents, language difficulties, or the transient and mobile nature of the ethnic groups are factors thought to contribute to their lower rates of cooperation. The majority of research on variations in response rates by racial-ethnic groups has focused on comparisons of Black and White respondents. Results from these studies reveal findings somewhat contradictory to what has generally been assumed about variability in response rate due to this factor. Some studies indicate no differences in response rate by race.t'6' 17 Whether the race of the interviewer and respondent was the same or different appears to have no discernible effect on the quality of the data collected, except in the case where the interview focused on racial topics.19 Even though there are little available data on the response rates for Mexican Americans (or Spanish-speaking Americans, or Americans of Spanish heritage), the evidence that is available for this minority group shows that ethnicity per se does not necessarily contribute directly to lower response rates.2022 The response rates for the subsamples in the 1976 CHAS-NORC survey were as follows: Self-weighting Main Sample, 83 per cent; Episode Oversample, 87 per cent; Non-SMSA Southern Black Oversample, 89 per cent; Southwestern Spanish-Heritage Oversample, 84 per cent; and Total Sample, 85 per cent. The response rate for the Spanish oversample was similar to that for the main sample, but somewhat lower than that for the other oversamplesfor people with illness episodes and for non-SMSA Southern Blacks. Design of Data Collection Instrument and Field Procedures for Minority Sample Another issue concerns the design of the data collection instrument and the field procedures for collecting information from nonEnglish-speaking minorities. People whose primary language is Spanish, for example, may not necessarily have had formal schooling in that language. Further, dialects may vary greatly among the general category of people who consider themselves "Spanish-speaking." In the CHAS-NORC study a survey firm from Southern California with considerable experience in conducting interviews in Spanish-speaking communities translated the questionnaire. A Chicano interviewer from Texas with formal language training in Spanish retranslated the Spanish version to English. We then assembled a committee-composed of
lLeonard A. LoSciuto has indicated in an unpublished manuscript, "Measurers and Correlates of Teenage Fertility-Pretest and Methodological Study," that matching of the race of the interviewer and the respondent appears to make no difference in response rates. 18 AJPH April 1980, Vol. 70, No. 4
SPANISH HERITAGE HEALTH SURVEYS
TABLE 3-Per Cent of Cases with Missing Values Spanish
Spanish
Heritage, Mainly Spanish Used in
Interview
Heritage, Mainly English Used in Interview
Estimate
%
%
%
%
%
%
lation %
Per cent contacting a physician in the year Per cent having an exam in the year Per cent hospitalized in the year Per cent contacting a dentist in the year Mean physician visits for those with one or more in the year
2.6
0.3
1.0
1.4
0.2
0.5
0.7
0.1
0.0
0.2
0.3
0.7
0.2
0.2
0.2
0.0
0.2
0.7
0.4
0.2
0.3
0.0
0.1
0.0
0.3
0.2
0.2
0.2
5.9
0.0
1.9
2.1
2.6
0.6
0.9
the original translator, the retranslator, a Central American whose native tongue was Spanish, and an Anglo who had had formal training and considerable work experience in the Spanish language-to reconcile any existing differences and arrive at a consensus concerning the most appropriate wording. Differences of opinion regarding the appropriate phrasing of various concepts, primarily reflecting the regional differences in phrasing among the various Spanishspeaking participants, were discussed and resolved. The questionnaire was pretested in areas of the Southwest in which it was to be administered, and changes suggested by tht pretest were incorporated. Bilingual interviewers fluent in Spanish were recruited to interview in the Spanish oversample segments. Some concerns expressed during the hiring and training process that Spanish-heritage male respondents might be reluctant to provide information to female interviewers were not borne out in reports from the field. Data Biases due to Variability in Responses by Minority Samples The validity of the data collected from minority samples may, to some extent, be affected by variations in item nonresponse (due to refusals or misunderstandings of the question) or by other response sets (such as a tendency toward yea-saying by disadvantaged respondents). 1 8 Oftentimes the sample non-response rate does not reflect the degree of completion of particular items. Some items may, for example, have larger numbers of missing values although the sample nonresponse rate is relatively low. Examination of item non-response rates is, therefore, a further step in evaluating the overall quality of the data collectAJPH April 1980, Vol. 70, No. 4
Spanish Heritage
NonSMSA Black South
Other NonWhite
Other White
Total
Total
Popu-
ed. The number of missing values on key estimates were examined for the Spanish-heritage sample. Comparisons of the number of missing values were made with those of other racial groups and among the Spanish-heritage sample-those for whom the interview was mainly conducted in Spanish with persons who spoke English during the interview (Table 3). For only two of the five estimates examined does the per cent of missing values for the Spanish-heritage population exceed that for the population as a whole-per cent contacting a physician in the year, and mean physician visits for those with one or more. Among the Spanish-heritage sample, there was a tendency for the proportion of missing values on these particular estimates to be higher for those with whom the interview was mainly conducted in Spanish, compared to those with whom English was the main language spoken. The language in which the interview was conducted undoubtedly provides some indication of the integration of the respondent into the mainstream English speaking Anglo culture and, hence, his/her ability to understand the concepts and questions about health care practices, for example, that may differ from those in the country of origin. Interviewers were asked directly how cooperative respondents seemed during the interview and whether there were questions they felt were not answered correctly. The results for the Spanish-heritage sample compared to other racial groups showed that the per cent of Spanish-heritage cases for which the interviewer reported he/she thought some of the answers provided.were inaccurate were quite similar to the rates reported for other Whites and for nonSMSA Blacks in the South-around eight per cent. The pro371
ADAY, ET AL.
portion of other non-whites (the vast majority of whom were urban Blacks) for whom inaccuracies were thought to exist was somewhat higher (12 per cent). When interviewers were asked to indicate how cooperative they thought respondents were during the interview, Spanish heritage as well as other Whites were most often reported to have provided "excellent" cooperation. Spanish-heritage persons and other Whites had the lowest proportion of fair or poor ratings-five and six per cent, respectively. Eight per cent of rural Southern Blacks were reported to have provided fair or poor cooperation, and 12 per cent of other nonwhites (urban Blacks) were rated only fair or poor cooperators. In summary, it appears that, according to the accuracy and cooperation evaluations of respondents by the interviewers, the Spanish-heritage sample was often similar to the majority White population. Nonwhites received somewhat lower ranking on these dimensions. In examining the actual number of missing values on key estimates for the respective ethnic groups, the Spanish-heritage respondents who spoke Spanish during the interview fared worse than the population as a whole on two of these estimates (physician contact and number of visits). English speaking Spanish-heritage respondents had less missing information than the population as a whole. The per cent of cases with missing values on these estimates also tended to be high for the non-SMSA Southern Black group. There have been suggestions in the literature that minorities tend to agree with items in an effort to "please" the dominant group. In the 1976 CHAS-NORC survey, pairs of positively and negatively worded phrasings of the same underlying attitude item were included in a Health Opinions Questionnaire.tt To the extent respondents tended to agree with both items (although they said opposite things) a tendency toward yea-saying was said to exist. Analyses of a number of these item pairs showed that whereas 14 per cent of the total U.S. population demonstrated some "yeasaying" tendency, 24 per cent of the Spanish-heritage group were found to demonstrate this effect; this compared to around 17-18 per cent of the nonwhites, and 13 per cent of
other Whites.
The field bias in social survey estimates reflects the difference between the "true" estimate and what respondents themselves report that may be a consequence of some of the response effects just described. Data from a 1970 national survey of health care utilization and expenditures provide an opportunity to develop approximations of field biases for the Spanish-heritage population compared to the general population average.23 In that study, data were obtained from providers (physicians, hospitals, and third-party payers) reported by social survey respondents in an effort to verify the information provided by the survey respondents. Differences between the respondent and provider estimates yielded an approximation of the field bias for these estimates. The definition of Spanish heritage available in that study is not entirely comparable to that used in the 1975-76 CHASNORC survey. In the 1970 study, the ethnicity designation is based on the interviewer's perception of whether or not the respondent appeared to be "Mexican" or "Puerto Rican," whereas in the 1976 study, Spanish surname and self-reports of Spanish language spoken in the family of origin were the criteria used. Selected estimates and the field biases associated with them are reported in Table 4. These data show that the magnitude of the field biases (as a proportion of the actual estimate) vary considerably by the type of estimate, the lowest being for the hospital admission rate estimate and highest for the average physician visit measure. Although the relative magnitude of the hospital admission field bias for the Spanish-heritage sample does not differ from that of the population as a whole, the sign of the bias does differ. Apparently the Spanish-heritage persons tended to underestimate the rates of admission (the survey estimate should be higher), while the opposite tendency was the case for the population in general. With respect to length of hospital stay and number of physician visits, the magnitude of the field bias for the Spanish-heritage sample appears to be considerably greater than that for the population as a whole. This is particularly the case for the mean physician visit estimte. Both Spanish-heritage respondents and the population as a whole tended to overestimate these types of medical contact, i.e., the actual
TABLE 4-Selected Field Biases for Spanish-Heritage Subgroup and the Total U.S. Population, 1970 Field Bias
Field Bias
of Estimate
as Per Cent
Population Group
Actual Estimate
Total Spanish heritage
14.3 11.4
- .4
- 3%
.3
3%
Total Spanish heritage
10.4 9.7
- .6
- 6% -10%
5.7 7.3
- .8
Type of
Estimate
Hospital admissions per 100 person-years Hospital days per admission Mean physician visits for those seeing physician
Total
Spanish heritage 372
-1.0
-1.8
of Estimate
-14% -25%
AJPH April 1980, Vol. 70, No. 4
SPANISH HERITAGE HEALTH SURVEYS
estimate should be less than that obtained in the social survey.
Summary In summary, the following findings have been documented in examining various methodological issues associated with data collection in the Spanish-heritage sample: (1) Ninety-three per cent of the persons classified as Spanish heritage according to the criteria of having a Spanish surname or the Spanish language spoken in the home as a child did report Spanish origin when asked their families' origin or descent. Seven per cent reported Anglo-Saxon (English, Irish, etc.) or American Indian identification. The preponderance of Spanish-heritage persons in the Southwest reported Mexican origins, however. (2) The clustering of households resulting from oversampling in areas of higher Spanish heritage concentration (based on 1970 Census data) resulted in a design effect about 50 per center higher than that for the population as a whole (2.7 compared to 1.8 for the sample as a whole). As a result, standard errors of the estimates for observations obtained from this oversampling process are higher. (3) The overall response rate for the Spanish-heritage oversample was similar to that for the general U.S. sample. (4) Item non-response for selected physician visit estimates for the Spanish-heritage population was somewhat higher than that of the total U.S. population for those who spoke Spanish during the interview but somewhat lower for those who spoke English. (5) Interviewer evaluations of Spanish-heritage respondents' cooperation and accuracy in reporting was comparable to that of the total U.S. population, while ratings of nonwhites' reporting behavior was somewhat less favorably characterized. (6) The tendency toward "yea-saying" on selected attitude items was higher for the Spanish-heritage population than for the U.S. population as a whole. (7) Based on 1970 national survey data, there appears to be a tendency for Spanish-heritage persons to overestimate physician contact rates to a greater extent than does the U.S. population in general.
Implications for Future Research Some implications of these findings for future methodological work on social surveys of Spanish-heritage health care are as follows: (1) Attention should be given to the heterogeneity within categories of "Spanish heritage", based on particular definitions. What is the self-reported ethnic origin of respondents, for example, and what does this suggest about within group economic or cultural AJPH April 1980, Vol. 70, No. 4
variability and its potential impact on outcome variables of interest? (2) More validity studies of Spanish heritage reporting of health services use would be helpful. Our analysis suggests that for some services measurement error may be a particular problem for Spanish-heritage respondents. Confirmation of this finding is required, along with consideration of methods to improve the accuracy of reporting and/or to introduce techniques to adjust for the bias. (3) Detailed investigation of "yea-saying" among Spanish-heritage respondents would be useful. We found some indication that contradictory answers to attitude questions about medical care were given more often by the Spanish heritage persons than by the population as a whole. Replication of this work, analyzing more questions, is necessary. Further, the effect of interviewers, language problems, and the cultural specificity of particular items should be considered.
REFERENCES 1. Weaver JL: National Health Policy and the Underserved: Ethnic Minorities, Women, and the Elderly. St. Louis: CV Mosby Co., 1976. 2. Saunders L: Cultural Difference and Medical Care: The Case of the Spanish-Speaking People of the Southwest. New York: Russell Sage Foundation, 1954. 3. Madsen W: Mexican Americans of South Texas. Chicago: Holt, Rinehart and Winston, 1964. 4. Rubel AJ: Across the Tracks: Mexican Americans in a Texas City. Austin: University of Texas Press, 1966. 5. Nall FC and Speilberg J: Social and cultural factors in the responses of Mexican Americans to medical treatment. Journal of Health and Social Behavior. 8:299-308, 1967. 6. Edgerton M and Karno RB: Perception of mental illness in a Mexican-American community. Arch Gen Psychiatry 30:233-
238, 1969. 7. Moustafa AT and Weiss G: Health Status and Practices of Mexican Americans. Mexican American Study Project, Advance Report 11. Los Angeles: University of California, Division of Medical and Hospital Administration, 1968. 8. McLemore SD: Ethnic attitudes toward hospitalization: An illustrative comparison of Anglos and Mexican Americans. Social Science Quarterly. 43:341-346, 1963. 9. Galvin ME and Fan M: The utilization of physicians' services in Los Angeles County, 1973. Journal of Health and Social Behavior 16:74-94, 1975. 10. Hoppe SK and Heller PL: Alienation, familism, and the utilization of health services by Mexican Americans. Journal of Health and Social Behavior 16:304-314, 1975. 11. McHugh JP, et al: Relationships between mental health treatment and medical utilization among low-income MexicanAmerican patients: some preliminary findings. Medical Care. 15:439-444, 1977. 12. Quesada GM and Heller PL: Sociocultural barriers to medical care among Mexican Americans in Texas: A summary report of research conducted by the Southwest Medical Sociology Ad Hoc Committee. Medical Care 15(Supplement):93-101, 1977. 13. Orshansky M, et al: Sources: A Blue Cross Report on the Health Problems of the Poor. Chicago: Blue Cross Association, 1968. 14. National Center for Health Statistics: Health characteristics of minority groups, United States, 1976. Advance Data 27 (April 14), Washington, DC: U.S. Govt Printing Office, 1978. 373
ADAY, ET AL. 15. Aday LA, et al: Health Care in the United States: Equitable for Whom? Beverly Hills, CA: Sage Publications (in press, 1980). 16. National Center for Health Services Research: Second Biennial Conference on Health Survey Research Methods. Research Proceedings Series, DHEW Pub. No. (PHS) 79-3207, Washington, DC: U.S. Govt Printing Office, 1979. 17. Scott JC: Response rate trends 1955-1970, and response by season, size of place of residence and length of interview. Technical Report 52, Field Officer, Survey Research Center, Institute for Social Research, The University of Michigan, Ann Arbor, 1971. 18. National Center for Health Services Research: Advances in Health in Health Survey Methods: Proceedings of a National Invitational Conference. Research Proceedings Series, DHEW
19.
20. 21.
22. 23.
Pub. No. (HRA) 77-3154. Washington, DC: U.S. Govt Printing Office, 1975. Schuman H and Converse JM: Effects on black and white interviews: black respondents in 1968. Public Opinion Quarterly 35:44-68, 1971. Freeman D: A note on interviewing Mexican Americans. Social Science Quarterly 19:909-918, 1969. Grebler L, et al: The Mexican-American People. New York: The Free Press, 1970. Welch S, et al: Interviewing in a Mexican-American community: An investigation of some potential sources of response bias. Public Opinion Quarterly 37:115-126, 1973. Andersen R, et al: Total Survey Error: Applications to Improve Health Surveys. San Francisco: Jossey-Bass, Inc., 1979.
Ideas for Safe Workplace Sought in National Contest
374
A $75,000 awards competition to solicit new ways of reducing accidents in America's workplaces has been undertaken jointly by The Hartford Insurance Group and the National Safety Council. Senator Abraham Ribicoff (D.-Conn.), whose public service career includes a direct role in several major safety-related programs and laws, will serve as chairperson of the blue-ribbon panel of judges which will evaluate papers submitted in the year-long national competition. Called "The Hartford Loss Prevention Awards Competition," the endeavor will be administered by the National Safety Council under a grant from The Hartford. The Hartford will present a total of 77 awards in the competition, with $15,000 for first place, $12,000 for second, and $10,000 for third. There will be additional winners in eight designated regions of the country. The National Safety Council outlined the following criteria for entries in the competition: 1. The problem under consideration must be broad in scope and of such significance that its solution would be agreed by the judges to have high priority. 2. The proposed solution must be cost-effective, have wide application and be capable of immediate implementation. 3. The problem and its proposed solution cannot consist of the plans for or the development of a patentable device or use of an already patented device. 4. The paper must concern itself with commercial and occupational problems exclusive of traffic safety. It may describe programs, techniques, procedures, etc., that promise to alleviate the problems. The competition is designed to stimulate knowledgeable and concerned individuals to propose ideas that will help solve the manifold problems of occupational safety and commercial loss prevention, bring new insight into the problems of safety and loss in the workplace, and apply that to a reasonable cost-effective solution. Contributors ought to be conversant with loss prevention and occupational safety and health matters, although they need not be gainfully employed in these professions exclusively. To qualify, individuals must follow the Hartford Competition Rules and Criteria, as published, and prepare submissions on the authorized entry form. Both are available by request from: Administrator The Hartford Loss Prevention Awards Competition The National Safety Council 444 N. Michigan Avenue Chicago, IL 60611 Papers will be judged principally for clarity of presentation, ingenuity, method, and for the significance, uniqueness and freshness of the solution presented. Judges for the competition panel will be chosen by an independent committee established by the National Safety Council. Papers must be received in National Safety Council offices in Chicago postmarked no later than 5:00 pm, May 1, 1980. First place regional award winners and the national award winners will be announced at the National Safety Congress and Exposition in Chicago, October 20-23, 1980. AJPH April 1980, Vol. 70, No. 4
