Alcohol and Alcoholism Advance Access originally published online on March 21, 2006
Alcohol and Alcoholism 2006 41(3):349-352; doi:10.1093/alcalc/agl019
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
RAPID COMMUNICATION
RECRUITMENT OF HEALTHY PARTICIPANTS FOR STUDIES ON RISKS FOR ALCOHOLISM: EFFECTIVENESS OF RANDOM DIGIT DIALLING
1 Donald W. Reynolds Department of Geriatric Medicine and 2 Department of Psychiatry and Behavioral Sciences, University of Oklahoma Health Sciences Center, Oklahoma City, OK, USA, 3 Department of Psychology and 4 Bureau for Social Research Oklahoma State University, OK, USA and 5 Behavioral Sciences Labs, Veterans Affairs Medical Center, OK, USA
* Author to whom correspondence should be addressed at: University of Oklahoma Health Sciences Center, Donald W. Reynolds Department of Geriatric Medicine, VA Medical Center (151A), 921 NE 13th Street, Oklahoma City, OK 73104, USA. Tel.: +1 405 270 0501 (ext. 3131); Fax: +1 405 290 1839; E-mail: Kristen-sorocco{at}ouhsc.edu
(Received 1 September 2005; first review notified 12 November 2005; in revised form 14 February 2006; accepted 15 February 2006)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSCONCLUSIONSREFERENCES |
|---|
Aims: To compare the effectiveness of two strategies for recruiting healthy research volunteers. Methods: Demographic characteristics and recruitment costs of participants who completed a laboratory study examining risk factors for alcoholism recruited through random digit dialling (N = 11) and community advertisements (N = 102) were compared. Results: Advertisement yielded a more representative sample [76% Caucasian, less well educated (M = 15.2 years, SEM = 0.2; P < 0.05), more equally divided by family history of alcoholism (43% FH– and 57% FH+), and lower in SES (M = 42.8, SEM = 1.3; P < 0.05)] and was more cost effective ($72 vs $2272 per participant) than random digit dialling. Conclusions: Findings are relevant to alcohol researchers trying to determine the recruitment strategy that will yield the most representative sample at the lowest cost.
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSCONCLUSIONSREFERENCES |
|---|
A number of papers have been written on successful strategies for recruitment of persons with overt clinical disorders (Schechter et al., 1994
), such as alcoholism, but little attention has been paid to the recruitment of healthy persons varying in presence of risk factors for alcoholism. The recruitment and assessment of healthy individuals pose unique challenges to researchers examining risk factors for alcoholism. First, the recruitment of healthy individuals can be more difficult than the recruitment of clinical samples because healthy individuals might not have a personal interest in the study topic (Eufemia et al., 1985). Second, rigorous screening assessments on healthy participants is necessary given potential confounders and the high rates of mental illness found among healthy volunteers (Halbreich et al., 1989; Shtasel et al., 1991). Identification of mental illness is particularly important to psychophysiological studies, which have found that apparently healthy subjects who are poorly screened can account for variability or outliers in control data when biological data are analysed (Schechter et al., 1994). Inadequate screening also increases the difficulty of comparing results across studies (Schechter et al., 1994). In addition to these challenges, representativeness and generalizability are often desired in a community-based sample, and it is often assumed that the best means for achieving these goals is to conduct random sampling from the community population, using a method such as random-digit telephone dialling (RDD). Research on recruitment strategies has focused on the yield of various publicity strategies, with a heavy emphasis on newspaper advertisements and collaboration with community agencies. While RDD has been shown to be effective for random population sampling in marketing and public opinion polls, less is known about its effectiveness for identifying and recruiting research subjects with given risk factors for alcoholism.
This paper compares the effectiveness of RDD with community advertisement for identifying healthy persons for a laboratory-based, long-term study of demographic, psychological, and psychophysiological characteristics in young adults with and without a family history of alcoholism (FH+, FH–). We present here our experience with these methods of subject ascertainment owing to the fact that grant reviewers in the field of alcohol research encouraged the use of RDD, addressing: (i) subject yield from recruitment sources, (ii) sample representativeness, and (iii) cost effectiveness.
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSCONCLUSIONSREFERENCES |
|---|
Our population of interest is FH+ and FH– young adults, ages 18–30 years, who are representative of the Oklahoma population. Individuals are classified as FH+ if either biological parent has ever had problems with alcohol or other substance use, ascertained by the Family History Research Diagnostic Criteria (FH-RDC) (Andreasen et al., 1977
). FH– individuals are those reporting an absence of alcohol or substance abuse in their biological parents and grandparents. Confirmation of FH-RDC reports is obtained by parent interview whenever possible.
Random digit dialling
RDD was conducted by the Bureau for Social Research (BSR) at Oklahoma State University during weekday evenings. Potential telephone numbers were randomly generated by Survey Sampling, Inc. of Fairfield, Connecticut using prefixes within the Oklahoma City Metropolitan Statistical Area. This yielded 9537 records that were provided to the BSR (Table 1).
|
Telephone numbers to be called were assigned a priority code by a Computer Assisted Telephone Interviewing (CATI) system. The protocol required that a minimum of five attempts be made to contact a sampled telephone number. If an interviewer reached an answering machine, the call outcome was recorded and the interviewer left one message per day stating they were calling from the Oklahoma State University and they would be calling back. Phone numbers were attempted a maximum of two times per day. For each call, the CATI system recorded the date, time, and disposition of the call as well as the interviewer's identification number. The final dispositions of the sampled telephone numbers are given in Fig. 1.
|
The telephone interview consisted of 23 questions. At each answered call, screen-out questions were asked to eliminate households without any eligible members. The screen-out questions asked the respondent whether anyone living in the household was 18–30 years of age. If no one in the household was 18–30 years old or that person was unavailable, the screening ended. If an eligible member was available to take the call, he or she was asked two additional screening questions regarding health and medication use. Persons who were potentially qualified then completed the 10 item alcohol use disorders identification test (AUDIT) (Babor et al., 1992
) and a 9 item demographic questionnaire that included the Hollingshead and Redlich questionnaire for socioeconomic status (Hollingshead, 1975). Persons meeting these preliminary inclusion criteria were given a brief description of the study and asked if they would agree to be contacted by the study staff. The BSR staff provided names of 133 qualified and willing individuals to the research staff, who then attempted to contact them to schedule an extensive screening at the laboratory. A minimum of 10 attempts was made to reach each qualified and willing individual provided by the BSR.
Community advertisement
Community advertisement included classified and display advertisements placed in local newspapers, including the main city newspaper, the evening entertainment paper, sports publications, and sales circulars. In addition, advertisements were placed on the exteriors of city buses, a television news spot was developed, and we also used personal referrals, and placed posters in public places (e.g. libraries). Additionally, participants were recruited from local technical schools and universities. The recruitment process by advertisement was as follows: (i) the potential participant telephoned in response to an advertisement, (ii) study personnel provided information on the study and requirements for participation, (iii) if interested, the potential participant completed a telephone screening similar to that used in the RDD, and (iv) if qualified, the participant was scheduled for an extensive screening.
Informed consent approved by the Institutional Review Board of the University of Oklahoma Health Sciences Center was obtained from all participants before completing the intensive screening, and volunteers were paid for their participation.
Statistical analysis
Response, cooperation, and contact rates were computed for the RDD effort using formulas specified in the American Association for Public Opinion Research document ‘Standard Definitions’ (AAPOR and Research, 2000). For example, the refusal rate was calculated as RR = [(I + P) + (R + NC + O) + UH + UP (UO)], where I = complete interviews, P = partial interviews, R = refusal and break-offs, NC = non-contact, O = other, UH = unknown household, and UO = unknown other. Cost data reported here represent only the costs of the advertisements or the RDD contractual costs. Research staff time is not included. Demographic comparisons were conducted using the Chi-square test and one-way analysis of variance. An alpha level of 0.05 was used for all statistical tests.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSCONCLUSIONSREFERENCES |
|---|
Yield
Figure 1 shows the results of RDD on the set of 9536 telephone numbers. It took an average of 2.8 (SEM = 0.10) calls to reach a final disposition of each sampled telephone number. A total of 2271 (Sum of did not qualify, refused, and completed interviews categories from Fig. 1) individuals completed the telephone prescreening, but only 380 individuals (17%) completed the full telephone interview (Fig. 1). Of these 380 individuals, 247 did not qualify or declined to be contacted about participation in the larger study. A final sample of 133 willing, potentially qualified individuals was passed along to the study staff. Figure 2 shows the disposition of each of these individuals. Of those we were able to contact, 11 individuals qualified and completed the study protocol as a result of phone or mail contact. The response rate for the BSR survey was 9.2%, and the cooperation rate (proportion of all cases interviewed of all eligible units ever contacted) was 73.1%. The contact rate (proportion of all cases in which some responsible member of the housing unit was reached) was 12.6% and the refusal rate was 0.05%.
|
By comparison 340 met the criteria for and participated in laboratory screening, as a result of other sources of recruitment besides RDD (N = 25; total N = 365; see Table 2). Of these, 105 (29%) persons qualified for inclusion in the study, and completed the protocol (see Table 2). Owing to missing data, demographic comparisons are presented for 102 of these participants.
|
Comparison of RDD and community advertisement
Demographic variables are reported in Table 3. The 11 participants recruited using RDD were all Caucasian, highly educated [M = 16.5 years, SEM = 0.5; 95% confidence interval (95% CI): 15.2–17.7], predominantly FH–, and had a mean Hollingshead and Redlich SES score of 52.8 (SEM = 3.1; 95% CI: 45.9–59.7). In comparison, community advertisements yielded a sample that was 76% Caucasian, less well educated (M = 15.2 years, SEM = 0.2; P < 0.05; 95% CI: 14.7–15.6), more equally divided by FH (43% FH– and 57% FH+), and lower in SES (M = 42.8, SEM = 1.3; P < 0.05; 95% CI: 40.2–45.5). The ethnic and racial make-up of the sample ascertained through community advertisement was more representative of the Oklahoma population, according to 2000 Census figures. When we compared the costs associated with each recruitment method, RDD proved to be more expensive than community advertisement: $2272 vs $72 per participant (see Table 2).
|
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSCONCLUSIONSREFERENCES |
|---|
Based on our experience, community advertisement was a more cost effective recruitment method than RDD. It yielded a greater proportion of volunteers who completed the study and who were more representative of the local population. The cost of RDD for recruitment was also higher in other studies (Ashery and McAuliffe, 1992
). The cost data presented here do not reflect the added staff time needed to recontact potential volunteers. Most importantly, RDD yielded a sample with less desirable demographic characteristics than the sample obtained by advertisement. These findings may be helpful to substance-use researchers considering alternative methods of subject ascertainment in studies similar to ours.
Our experience is in accord with a report of the effectiveness of newspaper advertisements in ascertaining healthy controls in a study of depression (Good et al., 1995). Advertisements placed in a wide range of publications, as we did, can help attract specific segments of the population, potentially yielding a more representative sample. The findings reported here further support the effectiveness of newspaper advertisements as a recruitment tool. By contrast, more costly methods, such as RDD, do not always yield representative samples.
One limitation to these findings is that the BSR stopped calling after five attempts and research staff stopped calling willing volunteers from the RDD phase after 10 attempts. More subjects may have been recruited if a longer series of calls had been attempted. Second, the small sample size obtained from RDD poses significant statistical limitations in comparing the methods of ascertainment. However, the fact that the RDD yielded such a small number of completed participants is noteworthy in itself.
RDD has proven to be a highly effective method for random sampling in population surveys and opinion polling. Our study had different goals. We sought persons meeting restrictive entry criteria who were also willing to visit the laboratory on three occasions to complete the study. In these circumstances, RDD yielded a less representative sample that was also lacking the risk factor of interest. RDD appears to be less effective when more screening criteria are employed and demands on participation are greater. In our study, newspaper advertisements were an effective recruitment tool that yielded a more representative sample having the desired risk factors for alcoholism. Our experience with these methods may prove useful to alcohol researchers in need of otherwise healthy volunteers having specific risk factors for alcoholism.
|
ACKNOWLEDGEMENTS |
|---|
This study was supported by the Medical Research Service of the Department of Veterans Affairs and by grants AA012207 and M01-RR14467 from the National Institutes of Health, Bethesda, MD, USA.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSCONCLUSIONSREFERENCES |
|---|
AAPOR and Research. (2000) Standard Definitions: Final Dispositions of Case Codes and Outcome Rates for Surveys. American Association for Public Opinion Research, Lenexa, KS.
Andreasen, N. C., Endicott, J., Spitzer, R. L. et al. (1977) The family history method using diagnostic criteria. Reliability and validity. Archives of General Psychiatry 34, 1229–1235.[Abstract]
Ashery, R. S. and McAuliffe, W. E. (1992) Implementation issues and techniques in randomized trials of outpatient psychosocial treatments for drug abusers: recruitment of subjects. American Journal of Drug and Alcohol Abuse 18, 305–329.[Medline]
Babor, T. F., de la Fuente, J. R., Saunders, J. et al. (1992) AUDIT: The Alcohol Use Disorders Identification Test: Guidelines for Use in Primary Health Care. World Health Organization, Geneva, Switzerland.
Eufemia, R., Weslowski, M. D. and Dowdy, C. A. (1985) Recruiting subjects for clinical research. Behavior Therapist 8, 214–215.
Good, B. G., O'Toole, S. and Sekula, K. (1995) Recruiting subjects into a research study on depression: Recruitment success and failure. Paper presented at the 55th Annual Conference of the American Medical Writers Association, Baltimore, MD, 1995.
Halbreich, U., Bakhai, Y., Bacon, K. B. et al. (1989) The normalcy of self-proclaimed "normal volunteers". [see comment]. American Journal of Psychiatry 146, 1052–1055.
Hollingshead, A. B. (1975) Four Factor Index of Social Status. pp. 1–22, New Haven, CT.
Schechter, D., Strasser, T. J., Santangelo, C. et al. (1994) "Normal" control subjects are hard to find: a model for centralized recruitment. Psychiatry Research 53, 301–311.[Medline]
Shtasel, D. L., Gur, R. E., Mozley, P. D. et al. (1991) Volunteers for biomedical research. Recruitment and screening of normal controls. Archives of General Psychiatry 48, 1022–1025.[Abstract]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||