ArticlesAudio-computer interviewing to measure risk behaviour for HIV among injecting drug users: a quasi-randomised trial
Introduction
HIV is transmitted primarily through unprotected sexual intercourse and the sharing of equipment for injecting illicit drugs. The AIDS epidemic has greatly increased the need for good measurement of such socially sensitive behaviour.1, 2 Accurate measurement of private and socially stigmatised behaviour is difficult in social science and health research. Some research participants will falsely claim to have engaged in the most extreme behaviours, but overall socially sensitive behaviour is under-reported.3, 4
Good measurement of socially sensitive behaviours is of particular importance in evaluation of HIV-prevention programmes.5 Calculation of HIV incidence is the most desirable outcome measure, but very few research programmes can afford the expense of that method. The counselling and testing usually needed to collect data on HIV incidence may in itself influence social behaviours associated with HIV risk, and reduce the degree to which results can be generalised. Other biological outcome measures, such as other sexually transmitted diseases, hepatitis B, and hepatitis C, can be used, but these are also expensive to measure, and the relation between incidence of these other infectious disorders and HIV incidence may vary in space and time.
Most research evaluation of HIV-prevention programmes has to rely upon self-reported risk behaviours as its primary outcome measure.5 Self-reported HIV-risk behaviour among participants in HIV-prevention programmes may be biased. In addition to the usual pressures to under-report sensitive behaviours, participants in HIV-prevention programmes may feel pressure to please programme staff or to defend what they see as valuable services, which may lead to additional under-reporting of HIV-risk behaviour.
Audio-computer-assisted self-interviewing (audio-CASI) has been developed recently for interview-based research. In audio-CASI, computers are programmed to display a question on a computer screen, while simultaneously the respondent hears the question and response categories through headphones. The respondent answers each question by pressing the appropriate computer key. Audio-CASI thus allows greater respondent privacy than traditional face-to-face interviews. When audio-CASI is used properly, no-one other than the individual respondent sees their responses to any questions. Audio-CASI does not require the respondent to be literate, since questions and answers are heard through headphones, nor must the respondent be questionnaire-literate, since appropriate skip patterns can be programmed into the computer without the respondent having to follow complicated instructions. Several studies of audio-CASI with populations at low risk for AIDS show that this method reduces under reporting of sensitive behaviours compared with traditional face-to-face interviews or self-administered questionnaires.6, 7, 8.
We studied the use of audio-CASI to measure HIV risk behaviours among injecting drug users who participated in syringe-exchange programmes. Our hypothesis was that the use of audio-CASI would lead to higher rates of reporting of behaviours associated with HIV infection. We also assessed the implications of the audio-CASI data for the control of blood-borne viruses among injecting drug users.
Section snippets
Data collection
We collected data at four different syringe-exchange programmes in the USA: New York City (NY), Chicago (IL), Tacoma (WA), and Los Angeles (CA). All of these programmes provide single unit syringes with attached needles, and additional on-site and referral services to participants. Data were collected in 1997 and 1998. The four programmes exchanged syringes at 12 locations that served different subpopulations of injecting drug users in each city.
Two fieldworkers worked at each syringe-exchange
Results
1594 individuals were asked to participate. We did 724 audio-CASI and 757 face-to-face interviews (figure), and all interviews were fully completed. 113 people declined to participate, 58 in weeks of audio-CASI and 55 in weeks of face-to-face interviews. Although fieldworkers may not have reported all instances of potential participants declining to participate, declining to participate was rare. When people were asked why they declined to participate, they usually replied that they “did not
Discussion
Readiness to participate, the high degree of interview completion, the small amount of missing data for sensitive questions, and the lack of frivolous responses all suggest that our study group found both audio-CASI and face-to-face interviews acceptable methods for data collection. We believe that the use of random-number tables for participant selection and weekly alternation of interview methods created unbiased assignment of participants to interview method. The nature of the study
References (22)
- et al.
Survey measurement of drug use: methodological studies
(1992) - et al.
Underreporting of abortion in surveys of US women: 1976–1988
Demography
(1992) - et al.
Adolescent sexual behavior, drug use, and violence: increased reporting with computer survey technology
Science
(1998) - et al.
Automated self-interviewing and the survey measurement of sensitive behaviors. Computer-assisted survey information collection
(1998) - et al.
Abortion and breast cancer risk: fact or artifact?
Women drug injectors who have sex with women: heightened risk and unknown reasons suggest the need for further research
AIDS Soc
(1998)- et al.
AIDS risk reduction and reduced HIV seroconversion among injection drug users in Bangkok
Am J Public Health
(1994)