Accurate self-report data on individual sexual risk practices are critical for judging the efficacy of HIV risk- reduction interventions. These reports are most often collected by retrospective assessment. More attention is now being devoted to survey design and administration, including widespread use of computer-assisted self-interviews (CASI). However, relatively little research has systematically evaluated whether these methods improve the accuracy of retrospective self-report data. In our recent study, daily sexual activity diaries were used to study the effects of commonly-used partner-specific versus aggregate question formats and modes of data collection SAQ, CASI, and audio-CASI on the accuracy of retrospective sexual behavior reports. Findings indicated that CASI modes and partner-specific format resulted in greater accuracy than more traditional survey methods. However, the rate of error in the data remained relatively high (between 30% and 38%), when compared to the diary-based measures. Despite their potential for yielding highly-detailed and recall-error free data, the use of traditional (paper) diaries has been limited in HIV research, mainly due to problems of maintaining participant compliance, missing or inconsistent data, and the costs of implementation. The proposed study will examine whether electronic diaries both telephone and internet-based can mitigate the obstacles associated with paper diaries. 302 men and 302 women will be randomized to electronic and paper diary conditions for reporting sexual activity over a 3-month period. Rates of compliance, accuracy and costs relating to electronic diaries will be compared to those for paper diaries. Determining the feasibility of using diary-based measures to evaluate outcomes of HIV prevention interventions is a critical aspect of our study;however, we acknowledge that retrospective surveys will remain the standard mode for collecting outcome data in HIV research. Therefore, we also propose to evaluate the accuracy of sexual self-reports obtained from 2 audio-CASI assessments using question formats consistent with cognitive theory partner-specific and timeline follow-back when the data are compared to the diary-based measures for the same period. By developing methods to ascertain the accuracy of risk outcome measures, the proposed study seeks to improve our ability to evaluate the efficacy of HIV prevention models, thus supporting public health strategies for identifying effective HIV interventions.
