Participant non-response is a source of bias in all research, especially in randomised controlled trials. Participants followed up remotely can have high non-response rates. Four such trials have been conducted of a cover letter with content informed by behaviour change theory to overcome hypothesised barriers to responding to a mailed questionnaire. Pooled results to date have suggested further research to be worthwhile. We conducted an embedded randomised study within a trial of such cover letters in the hope that we would improve response rates to our postal quality of life questionnaires. One hundred forty-eight participants in the CODIFI2 diabetic foot ulcer sampling trial were randomised 1:1 to receive one of two different cover letters at follow-up timepoints: either a standard cover letter accompanying their postal follow-up questionnaires or to an 'enhanced' (theory-informed) cover letter. Questionnaires were mailed at 39, 52 and (for some participants) 104weeks post randomisation. Outcome measures were response to mailing at each timepoint. Analysis was restricted to those for whom a questionnaire and letter was issued. Owing to limited recruitment, a reduced analysis plan, comprising solely observed response rates and 95% confidence intervals for difference in response rates was followed. Post hoc, we added our week 52 results to an already-published meta-analysis. Sixty-seven out of 74 enhanced cover letter group (Enhanced) and 67/74 standard cover letter group (Standard) participants who had not already died or withdrawn were sent their first mailing at 39weeks. The 39-week response rates were 47/67 (70.1%) and 39/67 (58.2%) for Enhanced and Standard participants, respectively. At week 52, the response rates were 45/64 (70.3%) and 35/63 (55.6%) for Enhanced and Standard participants, respectively. At week 104, the response rates were 24/33 (72.7%) and 19/33 (57.6%) for the Enhanced and Standard participants, respectively. Adding our week 52 results to a published meta-analysis increased the pooled estimate of differences in response rates to 0.04 (- 0.01 to 0.09) favouring enhanced letters. While this embedded randomised controlled trial observed greater response rates at all times among those randomised to the enhanced letter, the reduced sample size meant that these results are imprecise. ISRCTN registry ISRCTN74929588. Registered on 5 March 2019.
Read full abstract