Group | Control (n=374) | Lottery (n=374) |
Pooled non-payment (n=748) Cost per sampled woman (cost per response)‡ | Payment (n=125) | Lottery and payment (n=125) |
Pooled payment (n=250) Cost per sampled woman (cost per response)§ | Marginal cost per completed response of payment compared with non-payment¶ | |||||
Response rate | ||||||||||||
Initial mailing | 37.7% (141/374) | 41.7% (156/374) | 39.7% | 45.6% (57/125) | 49.6% (62/125)1-150 | 47.6% | 7.9% | |||||
£1.70 (£4.28) | £3.50†(£7.35) | £1.80 (£22.78) | ||||||||||
1st reminder | 20.4% (38/186)1-150 | 21.0% (39/186)1-150 | 20.7% | 22.0% (13/59)1-150 | 31.7% (19/60)1-150 | 26.9% | 6.2% | |||||
£1.70 (£8.21) | £3.05 (£11.34) | £1.35 (£21.77) | ||||||||||
2nd reminder | 17.6% (22/125)1-150 | 18.8% (24/128)1-150 | 18.2% | 29.6% (13/44)1-150 | 14.3% (5/35)1-150 | 22.8% | 4.6% | |||||
£1.70 (£9.34) | £2.84 (£12.46) | £1.14 (£24.78) | ||||||||||
Total | 53.7% (201/374) | 58.6% (219/374) | 56.1% | 66.4% (83/125) | 68.8% (86/125) | 67.6% | 11.5% | |||||
£3.12 (£5.56) | £5.85 (£8.65) | £2.73 (£23.74) | ||||||||||
Logistic regression analysis modelling of response by incentive method | ||||||||||||
Variable | Coefficient (β) | Odds ratios (95% CI) | χ2 1 | p Value | ||||||||
Payment | 0.53 | 1.70 (1.11 to 2.60) | 6.23 | 0.013 | ||||||||
Lottery | 0.20 | 1.22 (0.91 to 1.62) | 1.76 | 0.185 | ||||||||
Payment-lottery interaction | −0.09 | 0.92 (0.50 to 1.68) | 0.08 | 0.780 | ||||||||
Constant | 0.15 |
↵1-150 Subjects returning a blank form received no further reminders so that the denominator for reminder response rates are smaller than the numbers not responding to the previous mailing. †Assumes 29 women did not accept payment therefore costs are based on payment to 90 women not 119. ‡The cost per completed response in the pooled non-payment column was calculated by dividing the cost of the questionnaire (£1.70) by the proportion responding, for example, £1.70 / 0.397 = £4.28. The total cost per sampled woman was the total cost of questionnaires divided by women sampled (£2334.10 / 748) and the final cost per respondent by dividing the total cost by the number of responders (£2334.10 / 420). §The cost per sampled woman was calculated by adding the total cost of questionnaires sent to the cost of 90 respondents who accepted payment divided by the number of women sampled, for example, (£1.70 × 250) + (£5 × 90) / 250. The cost per response was calculated by dividing the cost by the number of respondents, for example, (£1.70 × 250) + (£5 × 90) / 119. The total cost per sampled woman was the total cost of questionnaires divided by women sampled (£1461.60 / 250) and the final cost per respondent was the total cost divided by the number of responders (£1461.60 / 169). ¶The marginal cost per completed response of payment compared with non-payment was calculated from the difference in response rates and costs from the pooled non-payment and pooled payment columns, for example, the total marginal cost was (£5.85−£3.12) / (0.676–0.561).