Measuring change in sexual behavior: Concordance between survey measures

Abstract

Measuring change in sexual behavior as reliably as possible is a crucial methodological issue in evaluation research. We compared two methods of assessing behavior change: the retrospective change method, which asks respondents about change explicitly by asking whether they have made any changes in their sexual behavior in a past time period, and the longitudinal or panel method, which asks respondents at two different times about their sexual behavior and compares the two sets of answers. Among heterosexual men and women (N = 962) in the AIDS in Multi‐Ethnic Neighborhoods (AMEN) Cohort Survey, the two methods of measuring change in condom use agreed, with a kappa of 0.38. The same was true among gay men (N = 596) where kappa for agreement was 0.35. However, among both heterosexuals and gay men, the two methods of measuring change in number of partners had a small but significant level of disagreement. Among heterosexuals, kappa was ‐0.14; among gay men, kappa was ‐0.16. Where the two types of change measures disagree, the data do not permit us to say which measure is more valid. Further research on concordance between measures of change in sexual behavior and on validity of change measures is recommended.