Assessing participation bias in a sexual behaviour survey: implications for measuring HIV risk

To assess the likelihood of participation bias in a large population-based sex survey, and its possible effect on estimates of HIV risk behaviours.

Construction of general hypotheses about non-participants through comparisons of willing and unwilling participants.

British adults aged 16-59 years were surveyed in 1990-1991. Interviews consisted of a face-to-face section combined with a self-completion booklet (n = 18876). Interviewers recorded interviewee embarrassment. Homosexual experience and number of lifetime heterosexual partners (grouped 0, 1, 2 or more) were recorded prior to booklet offer. Logistic regression was performed, with embarrassment and booklet refusal as outcome variables, assessing their association with sexual behaviour after controlling for demographic variables. Assuming that, in sexual behaviour, non-participants are closer to the embarrased and the booklet refusers ('unwilling' participants) than to others, these analyses provide an indication of the nature of participation bias.

Higher refusal an embarrassment rates were associated with both reporting no homosexual experience and fewer heterosexual partners.

Under our untestable assumption, these results are consistent with non-participants being generally at lower risk of HIV infection. Methods need to be developed both to reduce participation bias in sex surveys, and to incorporate it in analysis.