Conditional versus non-conditional incentives to maximise return of participant completed questionnaires in clinical trials: a cluster randomised study within a trial
Cook J., Cook JA., Bongard E., Heneghan C., Butler CC.
Background: High participant retention enhances the validity of clinical trials. A monetary incentive can increase retention, but it is not known if when it is provided and if it is conditional matters. We aimed to determine whether there was a difference in the number of follow-up trial questionnaires returned when a monetary (gift voucher) incentive was given to participants at recruitment (non-conditional), compared to informing participants at recruitment that the incentive would be given only once their 14-day daily diary (questionnaire) had been returned (conditional). Method: A cluster randomised study within a trial embedded within the Antivirals for influenza-Like Illness, An rCt of Clinical and Cost effectiveness in primary CarE (ALIC4E) Trial. Matched site pairs (GP practices) were randomised using computer-generated random numbers, to either a non-conditional or conditional monetary voucher incentive (only once their 14-day daily diary (questionnaire) had been returned. Sites were matched on previous recruitment levels and practice list size. Analyses were conducted according to randomised groups irrespective of compliance with a two-sided 5% level statistical significance level. The main analysis of the primary outcome (site proportion of diaries returned) was linear regression accounting for site pair (using cluster-robust variance). Additional weighted, paired and non-parametric sensitivity analyses were conducted. Secondary outcomes were the site average number of completed pages, time to return diary, and cost related to the incentive (administration and postage). Results: Of the 42 randomised sites (21 for each intervention), only 28 recruited at least one participant with only 10 practice pairs recruiting participants at both constituent sites. Raw diaries return proportions were 0.58 (127/220) and 0.73 (91/125) for non-conditional and conditional incentive groups. Regression analysis adjusted for site pair showed no significant difference in returns, − 0.09, (95% CI, − 0.29, 0.10, p = 0.34); when weighted, there was still no clear difference: 0.15 (95% CI, − 0.02, 0.31, p = 0.07). There was no clear statistical evidence of a difference in time taken to return questionnaires, nor the proportion of pages completed, by the intervention group in the main analyses (all p > 0.05). The conditional incentive was approximately £23 cheaper per diary returned based upon observed data. Conclusion: There was no clear evidence of a statistically significant difference in the proportion of participant-completed diaries returned between conditional or non-conditional incentive groups. The time to questionnaire return and completeness of the returned questionnaires were similar in both groups. There was substantial statistical uncertainty in the findings. Some of the sensitivity analyses suggested that a meaningful benefit of a conditional incentive of a magnitude that would be meaningful was plausible. The conditional approach costs less in cash terms.