Pilot Award

Methods to improve accuracy in self-reported sexual behavior are critical in reducing errors that impede accurate assessment of intervention outcomes and prevalence estimates of HIV-related risk. Research findings to date provide abundant empirical evidence that reliability and validity of self-reported sexual behavior is a major problem. Scientific models of the survey question-answer process specify question comprehension as a necessary precursor to enhancing data quality. The proposed study seeks funding to adapt findings from an RCT survey methods study to the computer survey technologies (e.g., ACASI, Web surveys) commonly used in HIV/AIDS and other health-related research. The recently completed experiment compared an innovative method of administering survey interviews, conversational interviewing, to the traditional standardized method. In standardized interviewing every respondent hears the same words for every question to minimize bias that could result from the same question read differently to some respondents. In contrast, the conversational method maintains the structure of a survey interview but allows interviewers to assist respondents to understand the same meaning for every question. The study (n=200) had two components (1) an experiment comparing the two survey methods to determine if respondents' had improved response accuracy in the conversational condition; and (2) a follow-up interview conducted immediately after the survey interview where respondents discussed problematic questions. The survey interview was video/audio taped; the follow up interview was audio taped. The proposed study will triangulate the video/audio recordings from the survey data with the rich qualitative data from the follow up interview to identify problematic questions and the reasons respondents encountered difficulties in providing responses about their own sexual risk behavior. We will focus on respondents' verbal explanations and nonverbal cues (survey videos) to explore how computerized surveys (ACASI; Web survey) can be structured to assist respondents in overcoming question difficulties and uncertainties to minimize self-reported biases.