PUBLIC HEALTH BRIEFS

ACKNOWLEDGMENTS This investigation was supported by the Flinn Foundation, and by Contract N01-CN-55426 from the Biometry Branch, National Cancer Institute.

REFERENCES

1. US Bureau of the Census: Census of population: 1980. General population characteristics. Final report. New Mexico, PC80-1-B33, 1981. Washington, DC: Govt Printing Office, 1982. 2. Becker TM, Madger LS, Harrison HR, Stewart J, Humphrey DD, Hauler J, Nahmias AJ: The epidemiology of infections with the human herpes virus in Navajo children. Am J Epidemiol 1988; 127:1071-1078 3. Ortiz A (ed): Handbook of North American Indians. Vol 9, Southwest, Washington, DC: Govt Printing Office, 1979. 4. Ortiz A (ed): Handbook of North American Indians, Vol 10, Southwest. Washington, DC: Govt Printing Office, 1983. 5. Williams JL: New Mexico in Maps. Albuquerque, NM: University of New Mexico Press, 1988. 6. World Health Organization: Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death. Based on the recommendations of the Seventh Revision Conference, 1955. Geneva: WHO 1975. 7. World Health Organization: Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death. Based on the recommendations of the Eighth Revision Conference, 1967. Geneva: WHO 1967. 8. World Health Organization: Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death. Based on the recommendations of the Ninth Revision Conference, 1975. Geneva: WHO 1977. 9. Becker TM, Wiggins CL, Key CR, Samet JM: Ischemic heart disease mortality in Hispanics, American Indians, and other Whites in New Mexico, 1958-82. Circulation 1988; 78:302-309. 10. Samet JM, Wiggins CL, Key CR, Becker TM: Mortality for lung cancer and COPD in New Mexico, 1958-82. Am J Public Health 1988; 78:11821186.

11. US Bureau of the Census: Census of the population: 1960. General population characteristics. Final report. New Mexico. PC(1)-33B. Washington, DC: Govt Printing Office, 1961. 12. US Bureau of the Census: Census of the population: 1960. Subject reports: Persons of Spanish surname. Final report PC(2)-IB. Washington, DC: Govt Printing Office, 1963. 13. US Bureau of the Census: Census of the population: 1960. Subject reports: Nonwhite persons by race. Final report PC(2)-1C. Washington, DC: Govt Printing Office, 1963. 14. US Bureau of the Census: Census of the population: 1970. General population characteristics. Final report. New Mexico. PC(1)-C33. Washington, DC: Govt Printing Office, 1971. 15. US Bureau of the Census: Census of the population: 1970. Subject reports. Final Report. American Indians. PC(2)-1F. Washington, DC: Govt Printing Office, 1973. 16. National Center for Health Statistics: Provisional estimates of selected comparability ratios based on dual coding of the 1966 death certificates by the seventh and eighth revisions of the international classification of diseases. DHEW Publ. Vol. 17 No. 8. Washington, DC: Govt Printing Office, 1968. 17. National Center of Health Statistics: Estimates of selected comparability ratios based on dual coding of 1976 death certificates by the eighth and ninth revisions of the international classification of diseases. DHEW Pub. No (PHS) 80-1120. Washington, DC: Govt Printing Office, 1980. 18. SAS Institute Inc: SAS user's guide: Statistics, version 5 Ed. Cary, NC: SAS Institute Inc, 1985. 19. Omran AR: Epidemiologic transition in the United States. Population Bulletin 1977; 32(2):3-45. 20. Maduro R: Curanderismo and latino views of disease and curing. West J Med 1983; 139:863-874. 21. Scheper-Hughes N, Stewart D: Curanderismo in Taos County, New Mexico-possible case of anthropological romanticism. West J Med 1983; 139:875484. 22. Bennett M, Mantlo EJ: New Mexico Resources Registry, Statistical Summary 1986-87. Albuquerque, NM: University of New Mexico Medical Center Press, 1988. 23. Williams R. Meningitis and unpaved roads. Soc Sci Med 1987; 24(2): 109-115.

Feasibility of a Telephone Survey to Study a Minority Community: Hispanics in San Francisco GERARDO MARIN, PHD, BARBARA VANOSS, PHD, Abstract: In two random digit dialing surveys conducted among Hispanics using a modified Mitofsky-Waksberg procedure, we found low refusal rates (4.7% and 3.1%), low assumed noncontact rate (14.0% and 18.3%), and high response rates (88.6% and 88.4%) with limited investment in time (1.58 hours and 1.66 hours per completed interview). These results suggest that Hispanics are willing to participate in telephone surveys and that this method may be feasible and useful for research and evaluation purposes. (Am J Public Health 1990; 80:323-326.)

Introduction

Surveys conducted over the telephone can be expected to be particularly useful to study minorities since they provide researchers with an efficient method to collect data Address reprint requests to Gerardo Marin, PhD, Associate Professor, Department of Psychology, University of San Francisco, 2130 Fulton Street, San Francisco, CA 94117-1080. Dr. B. Marin and Dr. Perez-Stable are with the Division of General Internal Medicine, Department of Medicine, University of California-San Francisco. This paper, submitted to the Journal April 17, 1989, was revised and accepted for publication September 21, 1989. © 1990 American Journal of Public Health 0090-0036/90$1.50

AJPH March 1990, Vol. 80, No. 3

AND

ELISEO J. PEREZ-STABLE, MD

with acceptable response rates and low interviewer bias.' -4 Telephone surveys also allow the use of moderately long interview schedules regardless of respondents' reading and writing abilities. In addition, telephone surveys generally cost less than personal interviews. On the other hand, several investigators have suggested there are problems in identifying and sampling Hispanics; obtaining respondent cooperation; designing valid interview schedules; and controlling biases that may be introduced due to the ethnicity of the interviewer and the language used in the interview.5-7 The absence of a telephone at home among a significant proportion of potential respondents can be expected to decrease the representativeness of a sample and limit the generalizability of the results of a survey. Most individuals in the United States have a telephone in their household.4 In San Francisco, for example, the 1980 Census found that 92 percent of self-identified Hispanic households in those census tracts with 400 or more Hispanics had at least one telephone compared with 94 percent for non-Hispanics. In other major cities telephone ownership among Hispanics averages 85 percent.8

We used the random digit dialing procedure for telephone surveys suggested by Mitofsky and Waksberg9 which eliminates the use of nonworking banks (prefixes) and avoids the use of directories (eliminating the problem of excluding 323

PUBLIC HEALTH BRIEFS

unlisted numbers) while guaranteeing equal probability of selection for all residential telephone numbers.' A recent study with Blacks'0 found this approach particularly efficient in obtaining samples from highly clustered subgroups as is the case of many ethnic/racial minorities living in urban areas of the United States. This paper describes two surveys in which the Mitofsky-Waksberg approach was used to sample Hispanics. Methods The surveys were designed to obtain estimates of the prevalence of cigarette smoking" and to serve as baselines for the evaluation of a community intervention designed to lower cigarette smoking prevalence among Hispanics in San Francisco. Each interview lasted approximately 15 minutes and included approximately 60 questions dealing with smoking behavior, standard socio-demographic variables, and an acculturation scale.12 The interviews were conducted in the language of choice of the respondents; 67 percent and 69 percent preferred to use Spanish in the first and second studies, respectively. Interviewers were bilingual (English and Spanish) men and women who received at least 12 hours of training, including role playing and supervised interviewing, before they participated in the study. Training of interviewers was based on procedures for telephone interviewing developed by the Minnesota Heart Health Program and on materials from the Survey Research Center of the University of Michigan.'3 A household was judged to be eligible when the respondent identified himself/herself as Hispanic, or identified the majority of the residents in the household as "Hispanics." Hispanic ethnicity was ascertained by the respondents' answer to the question "Do you or most of the people in your household consider yourselves to be Hispanics, Blacks, Asians, or Caucasians?" The response to this question was later checked with the respondents' and their parents' reported place of birth and with the respondents' identification with one of the Hispanic subgroups. Prefix Selection We selected prefixes that covered census tracts with the largest concentration of Hispanics, according to 1980 census data. This identified 27 tracts with 10 percent or more Hispanic residents. These tracts represented 66.6 percent (N = 55,541) of all Hispanics (N = 83,373) in the city of San

Francisco. In each census tract with high concentrations of Hispanics, three city streets were randomly selected to locate telephone prefixes assigned to residences. This process produced 49 three-digit prefixes out of the 121 prefixes in use in San Francisco. These procedures were conducted separately for each study and in both cases the same 49 three-digit prefixes were identified and used. Stage One Sampling A random list of 2,000 different four-digit numbers was generated by a computer and the numbers were randomly assigned to the 49 prefixes. These seven-digit telephone numbers were then assigned to interviewers. When a Hispanic household was reached, the first five digits of that telephone number became a Primary Sampling Unit (PSU). During Study I, a total of 4,326 working numbers were called as part of this first stage in the sampling procedure yielding 259 PSUs. Study II produced 355 PSUs out of the 4,680 working numbers called. Tables 1 and 2 present a summary of the outcome ofthe calls made in this stage during both studies. 324

TABLE 1-Summary of Telephone Calls for Study I

Outcome Total working numbers called Eligible Completed Refusals Ineligible Business Ethnicity Age Lost to survey Total screened Response rate Refusal rate Assumed noncontact rate

Stage N (%)

Stage II N (%)

Total N (%)

4,326 (100)

12,107 (100)

16,433 (100)

286 (6.6) 20 (.5)

1,383 (11.4) 63 (.5)

1,669 (10.2) 83 (5)

852 (19.7) 2,232 (51.6) 35 (1.0) 901 (20.1) 3,425 (79.9) (74.1) (6.5)

1,704 (14.1) 8,605 (71.1) 114 (1.0) 238 (2.0) 11,869 (98.0) (93.8) (4.3)

2,556 (15.6) 10,837 (65.9) 149 (1.0) 1,139 (6.9) 15,294 (93.1) 88.6 (4.7)

(42.4)

(4.0)

(14.0)

TABLE 2-Summary of Telephone Calls for Study II

Outcome Total working numbers called Eligible

Completed Refusals Ineligible Business Ethnicity Age Lost to survey Total screened Response rate Refusal rate Assumed noncontact rate

Stage N (%)

Stage II N (%)

Total N (%)

4,680 (100)

16,823 (100)

21,503 (100)

365 (7.8) 11 (.2)

1,694 (10.1) 55 (.3)

2,059 (9.6) 66(.3)

1,245 (26.6) 2,735 (58.4) 40 (9) 284 (6.1) 4,396 (93.9) (91.3) (2.9)

3,347 (19.9) 9,942 (59.1) 167 (1.0) 1,618 (9.6) 15,205 (90.4)

4,592 (21.4) 12,677 (59.0) 207 (1.0) 1,902 (8.8) 19,601 (91.2) (88.4) (3.1)

(15.9)

(18.8)

(87.5) (3.1)

(18.3)

Stage Two Sampling

A computer-generated list of 99 random two-digit numbers was used to produce the suffixes for each of the PSUs. Each interviewer continued using a given PSU until eight Hispanic households were reached including completed interviews or refusals, or until the list was exhausted including four follow-up calls made within a four-week period to every unanswered number. Tables 1 and 2 present a summary of the outcome of the numbers called in this stage during both studies. A total of 92 PSUs (36 percent) were exhausted before reaching eight Hispanic households during the first study. The corresponding figure for the second study was 95 (25 percent). Interviewee Eligibility

Within a Hispanic household, the respondent was selected by asking for the resident who had most recently celebrated a birthday and who was between 18 and 65 years of age. Approximately 8 percent of otherwise eligible households failed to meet this criterion (Tables 1 and 2). A total of 1,669 respondents were interviewed during Study I (Fall 1986). Of these, 44 percent were males and the majority of the respondents (69 percent) were foreign-born with 60 percent of them reporting being born in Central America, 29 percent in Mexico, and the remainder in South America, Cuba, or Puerto Rico. The respondents' mean age was 35.5 years and they reported a mean number of years of AJPH March 1990, Vol. 80, No. 3

PUBLIC HEALTH BRIEFS

education of 11.6. In Study II (Summer 1987), 2,059 Hispanics were interviewed (42 percent men). The respondents' mean age was 35.5 years while the mean number of years of education was 11.6. The majority (70 percent) were born outside the United States, 60 percent of them in Central America, 28 percent in Mexico, and the remainder in either Cuba, Puerto Rico, or South America. Results

Response/Refusal Rates Tables 1 and 2 show the response and refusal rates for both studies. We have adapted Kviz's14 definition of response rate which is the proportion of completed interviews in terms of the number of eligible respondents (defined here as Hispanics between 18 and 65 years of age). This calculation assumes that those not contacted after four follow-up calls ("lost to survey") were eligibles.15 This proportion was assumed to be the same as the proportion of eligibles found among telephone numbers that were fully screened. The Stage I response rate, assumed that the proportion of eligibles found among the screened calls 306/3425 (8.9 percent) would be similar to that among the 901 numbers that were "lost to survey." Adding the 80 assumed eligible numbers yielded a response rate of 286/386 (74.1 percent). Response rates for the other stages of both studies were calculated in the same manner noted here and show an overall response rate of 88.6 percent for Study I and 88.4 percent for Study II. The refusal rates were defined' as the proportion of eligible respondents (Hispanics between 18 and 65 years of age) who were contacted but declined to be interviewed. Refusal rates were only 4.7 percent for Study I and 3.1 percent for Study II. The Assumed Noncontact Rate' is an index of the accessibility of respondents or the probability of researchers locating respondents. This index is made up of the proportion of respondents not contacted after four follow-up calls ("lost to survey" in both tables) in relation to the total number of possible eligible respondents. Given that a significant proportion of telephone numbers in San Francisco are unlisted (38 percent according to Pacific Bell), it is not appropriate to estimate the true number of eligible households from an alphabetic telephone directory. Thus, in computing the assumed noncontact rate we are assuming that 25 percent of those respondents "lost to survey" were eligible Hispanics. This assumption reflects twice the number of Hispanics reported to be living in San Francisco in 1985 and therefore produces a very conservative computation of nonresponse rates. In Study I for example, 25 percent of the 1,139 "lost to survey" (N = 285) represent 14 percent of the total number of assumed Hispanic numbers: completed interviews (N = 1,669) plus refusals (N = 83) plus lost to survey (N = 285). The comparable noncontact rate for Study II is 18.3 percent. Figures taken from Wiseman and McDonald's16 review of the literature show a median noncontact rate of 39.1 percent. Cost of Survey Based upon the time spent by all interviewers, we estimated costs of obtaining a completed interview. On the average, interviewers needed to spend 1.58 hours in order to obtain a completed interview during Study I and 1.66 hours during Study II. Each of the completed interviews cost approximately $15 based on paid interviewer and supervisor salaries but excluding overhead, training time, telephone

AMJPH March 1990, Vol. 80, No. 3

lines, investigators' time, and data management. Interviewers were paid $7.90 per hour in this study. Discussion These results suggest that it is feasible and efficient to use random digit dialing techniques to obtain a sample of Hispanics. The similarity in the outcomes of the two studies imply that the findings are fairly stable across time. The high completion rates and the correspondingly low refusal rates obtained in these surveys compare favorably with those found in studies of other Whites, which range between 66 percent in a study in Chicago4 to 73.5 percent in a regional survey.'7 The low refusal rates (4.7 percent in Study I and 3.1 percent in Study II) are lower than those of other studies. Wiseman and McDonald'6 in their review of 182 studies using telephone surveys found a median refusal rate of 28 percent while Frey' reports refusal rates ranging between 20-25 percent. Low refusal rates among Hispanics have also been reported in a study conducted in Texas18 where a telephone survey of a sample drawn from Spanishsurnamed individuals produced a refusal rate of 8.8 percent. The high rate of participation in these surveys by Hispanics was not due to advance publicity about the study. Respondents were informed of the purpose of the study only at the time of the interview. One possible explanation for these results is the presence of a cultural characteristic that promotes cooperation and that dictates positive interpersonal relations-the social script of "simpatia."'9 This cultural norm would move Hispanic respondents to agree to be interviewed in order to maintain a fluid and positive social relationship with the interviewer when the relationship is perceived as nonthreatening and as sanctioned by a respected local institution (the study was introduced as being conducted by a state university in San Francisco). The generalizability of our results may be limited by the following specific concerns: * The sampling frame used excluded approximately one-third of the San Francisco Hispanic population according to the 1980 census. Hispanics who are living in census tracts excluded in this study (less than 10 percent Hispanic) may behave differently when asked for an interview over the telephone. * At least 8 percent of San Francisco Hispanic households may not have a telephone at home and these persons are also excluded from the sample. * The approach utilized in choosing the prefixes may not produce all of the relevant ones although a third study survey recently concluded by the authors has again identified the same 49 prefixes as the first two studies. * Finally, San Francisco Hispanics (Central American and Mexican in this survey) may differ from Hispanics residing elsewhere. Future studies should compare samples gathered with this method with census data in order to address the representativeness of a sample of minority-group respondents obtained through a random digit dialing technique such as the

Mitofsky-Waksberg approach. Unfortunately, this analysis cannot be done at the present time due to the high undercount of Hispanics in the 1980 census and the later significant migrations from Cuba and Central America.

ACKNOWLEDGMENTS

This investigation was supported by PHS grant number R18 CA39260 awarded by the National Cancer Institute, US Department of Health and

325

PUBLIC HEALTH BRIEFS Human Services, Eliseo J. Perez-Stable, Barbara VanOss Marin and Gerardo Marin, principal investigators. Eliseo J. Perez-Stable is a Henry J. Kaiser Family Foundation Faculty Scholar in General Internal Medicine. The authors wish to express their appreciation for the extraordinary contribution of Rosa Marcano to this study as the supervisor of the surveys and to Raymond J. Gamba for his help in data management.

REFERENCES 1. Frey JH: Survey Research by Telephone. Beverly Hills, CA: Sage, 1983; 38-40. 2. Groves RM, Kahn RL: Surveys by Telephone: A National Comparison with Personal Interviews. New York: Academic Press, 1979. 3. Kidder LH, Judd CM: Research Methods in Social Relations. New York: Holt, Rinehart & Winston, 1986. 4. Lavrakas PJ: Telephone Survey Methods. Newbury Park, CA: Sage, 1987. 5. Aday LA, Chiu GY, Anderson R: Methodological issues in health care surveys of the Spanish heritage population. Am J Public Health 1980; 70:367-374. 6. Howard CA, Samet JN, Buechley RW, Schrag SD, Key CR: Survey research in New Mexico Hispanics: Some methodological issues. Am J Epidemiol 1983; 117:27-34. 7. Welch S, Conner J, Steinman M: Interviewing in a Mexican-American community: An investigation of some potential sources of response bias. Public Opinion Q 1973; 37:115-126. 8. Adams-Esquivel H, Lang DA: The reliability of telephone penetration estimates in specialized target groups: The Hispanic case. J Data Collect 1987; 27:35-39.

9. Waksberg J: Sampling methods for random digit dialing. J Am Stat Assoc 1978; 73:40-46. 10. Blair J, Czaja R: Locating a special population using random digit dialing. Public Opinion Q 1982; 46:585-590. 11. Marin G, Perez-Stable EJ, Marin B: Cigarette smoking among San Francisco Hispanics: The role of acculturation and gender. Am J Public Health 1989; 79:196-198. 12. Marin G, Sabogal F, Marin B, Otero-Sabogal R, Perez-Stable EJ: Development of a short acculturation scale for Hispanics. Hispanic J Behav Sci 1987; 9:183-205. 13. Guenzel PJ, Berkmans TR, Cannell CF: General Interviewing Techniques. Ann Arbor: Institute for Social Research, 1983. 14. Kviz FJ: Toward a standard definition of response rate. Public Opinion Q 1977; 41:265-267. 15. Fowler FJ: Survey Research Methods. Beverly Hills, CA: Sage, 1984. 16. Wiseman F, McDonald P: Noncontact and refusal rates in consumer telephone surveys. J Market Res 1979; 16:478-484. 17. Siemiatycki J: A comparison of mail, telephone and home interview strategies for household health surveys. Am J Public Health 1979; 69:238-245. 18. Shoemaker PJ, Reese SD, Danielson WA: Media in ethnic context. Austin, TX: College of Communication, University of Texas at Austin, 1985. 19. Triandis HC, Marin G, Lisansky J, Betancourt H: Simpatia as a cultural script of Hispanics. J Pers Soc Psychol 1984; 47:1363-1375. 20. Vernon CL, D'Augelli AR: Community involvement in prevention programs: The use of a telephone survey for program development. J Community Psychol 1987; 15:23-28.

Sex-specific and Race-specific Hip Fracture Rates SHIRLEY E. KELLIE, MD, MSC, Abstract: Sex-, race- and age-specific hip fracture rates were determined using Health Care Financing Administration data for Medicare-reimbursed hip fracture hospitalizations from 1980 to 1982. Rates were highest in White women, lowest in Black men, and intermediate in White men and Black women. Proportions of hip fracture patients dying during hospitalization and those discharged to nursing homes, respectively, were: White men (10.5%; 49%); Black men (9.3%; 32%); White women (5.0%1; 54%); and Black women (8.2%; 30%). (Am J Public Health 1990; 80:326-328.)

Introduction

In 1986 in the United States, there were in excess of 250,000 hospital admissions for hip fracture.' Elderly individuals who sustain hip fractures experience increased morbidity2-7 and mortality.8-'2 In an earlier paper, Farmer, et al,'3 reported that age-specific hip fracture rates were higher in White women than in White men, Black women, and Black men, but the differences among age-specific rates for White men, Black men, and Black women could not be estimated precisely. In this paper we report age-, sex-, and race-specific hip fracture rates extended through age 85 years. In addition, we present findings for proportions of patients dying during Address reprint requests to Shirley E. Kellie, MD, MSc, American Medical Association, 535 N. Dearborn Street, Chicago, IL 60610. Dr. Kellie is also Clinical Assistant Professor, Department of Medicine, University of Illinois College of Medicine. Dr. Brody is Dean, University of Illinois, School of Public Health. This paper, submitted to the Journal March 14, 1988, was revised and accepted for publication September 21, 1989. C 1990 American Journal of Public Health 0090-0036/90$1.50

326

AND

JACOB A. BRODY, MD

hospitalization, and discharged to nursing homes, not previously reported for Blacks. Methods Numbers of Medicare-reimbursed hospitalizations for hip fracture in Illinois residents over 65 years of age were obtained from data collected by the Professional Standards Review Organization in Illinois during 1980, 1981, and 1982. The Uniform Hospital Discharge Format was used in the data collection, and the data tape was made available through the Health Care Financing Administration (HCFA). Hip fracture was defined according to the ninth revised International Classification of Diseasesl4 categories 820.0-820.9: transcervical fractures, closed and open; pertrochanteric fractures, closed and open; and fractures of unspecified parts of the neck of the femur, closed and open. Populations at risk for hip fracture were obtained from the 1980 United States Census figures for the Illinois population by age, race and sex. Mean numbers of hip fractures for 1980, 1981, and 1982 were used to calculate the sex-, race- and age-specific annual incidence rates for hip fracture. Miettinen test-based confidence intervals were calculated for the hip fracture relative risks. 15 Age-adjusted odds ratios for death and nursing home placement were determined using the Statistical Analysis System multiple logistic regression program.16 Race- and sex-specific means for days of hospital stay were adjusted for age with the Statistical Analysis System linear regression program. 17

Results Age-group specific hip fracture rates were highest for White women, and lowest for Black men in all age groups after age 70 years (Figure 1). The rates in Black women are AJPH March 1990, Vol. 80, No. 3

Feasibility of a telephone survey to study a minority community: Hispanics in San Francisco.

In two random digit dialing surveys conducted among Hispanics using a modified Mitofsky-Waksberg procedure, we found low refusal rates (4.7% and 3.1%)...
881KB Sizes 0 Downloads 0 Views