Interviewer Effects in the Elicitation of Sexual and Drug Injection Partners

Interviewer effects, or variations in interviewee responses associated with interviewers, are not uncommon in survey research. Such effects tend to be more likely when interview questions are open-ended. Indeed, recent research showed moderate interviewer effects in the number of personal network members elicited, with intra-class correlations ranging between .13 and .15 after adjustment for numerous respondent and interviewer characteristics (Marsden, 2003; van Tilburg, 1998). It is crucial that network elicitation be as complete as possible, because most network measures are very sensitive to missing data.Sexual and drug injection networks shape the transmission of sexually transmitted diseases, HIV, and other infections. The elicitation of sexual and drug injection partners involves open-ended questions that are of a sensitive nature, which may increase the likelihood of interviewer effects. Thus, in this project I am assessing the extent to which interviewer effects exist in the elicitation of such partners.The analyses are ongoing and involve multiple data sets from research studies and disease control activities. The results from the first two data sets suggest no interviewer effect on the number of partners elicited (intra-class correlations <= .01). However, there does seem to be a consistent interaction between interviewer sex and interviewee sex, such that more sexual partners are elicited when the interviewer and interviewee are of the same sex. Comprehensive results will be presented for at least three data sets. The discussion will focus on the implications for the collection and analysis of sexual and injection network data.