The cost effectiveness of including pencils and erasers with self-completion epidemiological questionnaires

It is cheap to process epidemiological data from optical mark read (OMR) questionnaires. Respondents should use a pencil to complete OMR questionnaires, but many will not unless these are supplied. Sending pencils and erasers is expensive. Does sending pencils and erasers increase the response rate as cost-effectively as sending reminders, or does this decrease the error rate and offset data checking costs? We mailed 300 smokers and half were randomised to receive pencils and erasers. The relative risk (95% confidence intervals) for the response rate for the pencil group relative to the non-pencil group was 0.77 (0.46-1.29) and for the error rate was 1.31 (0.78-2.21). Sending pencils and erasers was not cost-effective in sensitivity analysis with any response rate or using the confidence intervals. Including pencils with mailed epidemiological questionnaires probably has no benefit and any plausible benefit does not offset the costs of sending pencils and erasers.