Audio Computer-Assisted Self-Interviews Do Not Necessarily Yield Higher Reports of Sensitive Behaviors
Women who participated in audio computer-assisted self-interviews (audio-CASI) were less likely than those in face-to-face interviews to report ever having had sex (35% vs. 48%) in response to a question on age at first sex, according to a randomized study among unmarried young women in rural Malawi.1 However, women in the audio-CASI group were more likely than those in the conventional interview group to report having had multiple partners (27% vs. 17%) and having had sex with a friend or acquaintance (17% vs. 7%), as well as to give inconsistent responses regarding their sexual history. Respondents in the two interview groups had similar levels of STIs, although the association between sexual experience and STI status was stronger in the face-to-face group.
Most studies of audio-CASI have been conducted in developed countries, where computer literacy is high. To compare the effectiveness of the technique with that of face-to-face interviews in eliciting reports of sexual behavior in a developing country, researchers questioned 501 unmarried Malawian women aged 15–21 regarding their sexual history. The women lived in villages near centralized market areas in Balaka district, located in the region with the country's highest rates of HIV infection, teenage pregnancy and risky sexual behavior. Women were randomly assigned to a face-to-face interview or an audio-CASI interview on a laptop computer, during which they were asked about their age at first sex and, regardless of their answer, the types of partners they had had sex with (e.g., boyfriend, expected spouse, or friend or acquaintance). Following the interviews, researchers asked respondents whether they were willing to be tested for HIV and three other STIs (gonorrhea, chlamydia, trichomoniasis); 84% were tested for at least one infection using vaginal or oral swabs. Prior to testing, they were asked again—this time by trained nurses—about their age at first sex.
On average, respondents were 17 years old and had had seven years of schooling; about six in 10 were enrolled in school, and a little more than half were Muslim. Although the two interview groups were similar in most respects, a lower proportion of those in the face-to-face group than in the audio-CASI group were born-again Christians or had "made tauba" (a form of Muslim conversion—29% vs. 37%). To account for differences between the two interview groups, reports of sexual behavior were estimated as predicted percentages derived from logistic regression models that adjusted for respondents' background characteristics.
Lower proportions of respondents in the audio-CASI group than in the face-to-face group reported ever having had sex (35% vs. 48%) or having had sex with a boyfriend (21% vs. 31%)—findings that belie the common belief that relatively anonymous interviewing methods, such as audio-CASI, will elicit more reports of stigmatized or sensitive behavior than will face-to-face interviews. In contrast, higher proportions in the audio-CASI group reported having had sex with a friend or acquaintance (17% vs. 7%) or having had more than one sexual partner (27% vs. 17%).
To explore how different measures of sexual activity affected the reporting of premarital sex by interview mode, the researchers created composite variables based on the "ever had sex" and "sexual partner" variables. A higher percentage of women in the audio-CASI group than in the face-to-face group reported that they had had sex or had had at least one sex partner (58% vs. 48%). Furthermore, while only 25% of face-to-face respondents said they had had sex with someone other than a boyfriend or expected spouse, 42% of audio-CASI respondents reported having had such partners. These results are consistent with the expectation that computer-assisted interviews will elicit higher reporting of premarital sex than face-to-face interviews when a broader set of questions are considered.
The responses of women in the audio-CASI group were more inconsistent than those of women who were interviewed face-to-face: Thirty-five percent of women in the former group said they had never had sex, yet identified a sexual partner in later questions, while 22% of those who said they had had sex later failed to identify any partners. Only a few women in the face-to-face group gave inconsistent responses.
The study also gauged inconsistent reporting by comparing responses at the initial interviews with those elicited by the nurses who administered STI tests. Similar proportions of respondents in the audio-CASI and face-to-face groups originally said they had had sex but later denied it (8% and 7%, respectively), whereas a higher proportion of women in the audio-CASI group than in the face-to-face group originally said they had never had sex but later reported that they had (38% vs. 19%). When the composite measure was compared with women's later responses to the nurses, audio-CASI respondents had higher rates of inconsistent reporting than did face-to-face respondents for each of the preceding scenarios (18% vs. 7% for later denying having had sex, and 24% vs. 20% for later reporting having had sex).
Fourteen percent of the study sample tested positive for an STI; 6% were HIV-positive, 6% had gonorrhea, 2% had trichomoniasis and 1% had chlamydia. Among respondents who said at the initial interview that they had ever had sex, 26% of those in the face-to-face group and 15% in the audio-CASI group tested positive for an STI; among those who reported at the nurses' interview that they had ever had sex, the proportions testing positive were similar (23% and 14%, respectively). Chi-square tests found that the association between STI status and behavior was stronger among women in the face-to-face group, yet in both groups a number of women who denied ever having had sex tested positive for an STI.
Overall, the study suggests that audio-CASI does not necessarily elicit higher rates of sensitive behaviors than do face-to-face interviews, and the researchers suggest that data on such behaviors, particularly when collected from adolescents in developing countries, must be viewed with healthy skepticism. They assert that "in light of the importance of behavioral data for understanding both the etiology of the AIDS pandemic and the results of clinical trials…to reduce the transmission of sexually transmitted infections," further quantitative and qualitative research is needed on interview mode effects, "particularly among adolescents and young adults."—J. Thomas
1. Mensch BS et al., Sexual behavior and STI/HIV status among adolescents in rural Malawi: an evaluation of the effect of interview mode on reporting, Studies in Family Planning, 2008, 39(4):321–334.