Effect of computer-assisted interviewing on self-reported sexual behavior data in a microbicide clinical trial

In a microbicide safety and effectiveness trial (HPTN 035) in Malawi, 585 women completed the same questionnaire through a face-to-face interview (FTFI) and an audio computer-assisted self-interview (ACASI). Concordance between FTFI and ACASI responses ranged from 72.0 % for frequency of sex in the past week to 95.2 % for anal intercourse (AI) in the past 3 months. Reported gel and condom use at last sex act were marginally lower with ACASI than FTFI (73.5 % vs. 77.2 %, p = 0.11 and 60.9 % vs. 65.5 %, p = 0.05, respectively). More women reported AI with ACASI than FTFI (5.0 % vs. 0.2 %, p < 0.001). Analyses of consistency of responses within ACASI revealed that 15.0 % of participants in the condom-only arm and 28.7 % in the gel arm provided at least one discrepant answer regarding total sex acts and sex acts where condom and gel were used (19.2 % reported one inconsistent answer, 8.1 % reported two inconsistent answers, and 1.4 % reported three inconsistent answers). While ACASI may provide more accurate assessments of sensitive behaviors in HIV prevention trials, it also results in a high level of internally inconsistent responses.