The potential impact of recanting on the assessment of prevention outcomes

Recanting is where study respondents deny previous positive reports of sensitive behaviours (e.g. saying you used an illicit drug use at time 1 but then claiming that you have never used at time 2). Recanting is a well recognised form of measurement error in longitudinal studies, with an emerging international literature examining reporting inconsistencies across a range of existing cohort and panel surveys (see for example, Percy et al., 2005; Shillington et al., 2010; Stanton et al., 2007). One key finding to emerge is that recanting is not evenly distrusted across study populations, but varies with particular individual characteristics and experiences. In particular, recanting has been show to be higher amongst males, young people with positive attitudes towards school, and those who report fewer other illicit behaviours (Percy et al., 2005).

However, one area that has received relatively limited attention is the potential impact that recanting may play in the assessment of behavioural change in prevention trials. This paper will examine recanting behaviour within an ongoing longitudinal cohort study. It will identify the various pattern of reporting within the cohort and examine the individual factors associated with various recanting patterns. In particular, it will test the extent to which the receipt of prevention materials may increase the level of recanting behaviour. Such an effect, if undetected, may lead to a significant overestimation of behaviour change within a prevention trial. Consideration will also be given to possible methods for correcting identified recanting within the repeated measurement of sensitive behaviours.