Retaining young people in a longitudinal sexual health survey: a trial of strategies to maintain participation

A longitudinal survey was conducted with 8,430 eligible pupils from two sequential year groups from 25 Scottish schools. Wave 1 (14 years) and Wave 2 (16 years) were conducted largely within schools. For Wave 3 (18 years), when everyone had left school, the sample was split into 4 groups that were balanced across predictors of survey participation: 1) no incentive; 2) chance of winning one of twenty-five vouchers worth ￡20; 3) chance of winning one ￡500 voucher; 4) a definite reward of a ￡10 voucher sent on receipt of their completed questionnaire. Outcomes were participation at Wave 3 and two years later at Wave 4. Analysis used logistic regression and adjusted for clustering at school level.The only condition that had a significant and beneficial impact for pupils was to offer a definite reward for participation (Group 4). Forty-one percent of Group 4 participated in Wave 3 versus 27% or less for Groups 1 to 3. At Wave 4, 35% of Group 4 took part versus 25% or less for the other groups. Similarly, 22% of Group 4 participated in all four Waves of the longitudinal study, whereas for the other three groups it was 16% or less that participated in full.The best strategy for retaining all groups of pupils and one that improved retention at both age 18 and age 20 was to offer a definite reward for participation. This is expensive, however, given the many benefits of retaining a longitudinal sample, we recommend inclusion of this as a research cost for cohort and other repeat-contact studies.There is an increasing trend towards lower participation in questionnaire surveys [1]. This affects costs, as more people have to be approached in order to meet the target sample size, and generally reduces representativeness, since participation is biased to particular groups. The challenges are multiplied for longitudinal studies as they aim to retain the same individuals across a number of waves and, after the first wave, individuals who do not respond cannot simply be replaced by