Sixteen randomised controlled trials were included in the review. One trial on vaccination and screening was not included in the meta-analyses due to high risk of bias and insufficient information. Risk of bias was low or unclear in most quality domains in most studies. The main potential sources of bias were allocation concealment and selection bias. Post-intervention follow-up ranged from four weeks to 24 months.
Overall (25 relevant comparisons), financial incentives successfully encouraged health behaviour change (RR 1.62, 95% CI 1.38 to 1.91; Ι²=84%). Meta-regression showed that effect size decreased as post-intervention follow-up and incentive value increased.
Smoking cessation (10 trials, 23,602 participants): Results up to six months follow-up showed incentives significantly improved smoking cessation (RR 2.48, 95% CI 1.77 to 3.46; eight trials; Ι²=21%). There was no evidence of association between effect size and follow-up time or total incentive value.
Statistically significant improvement was also seen at follow-up over six months (RR 1.50, 95% CI 1.05 to 2.14; six trials), though statistical heterogeneity was high (Ι²=76%). Subgroup analysis showed that the average effect of cash-only financial incentives was greater than for other formats (high heterogeneity remained). The average effect for other formats was not statistically significant. There was some evidence that effect size increased with incentive value.
Attendance for vaccination and screening (four trials, 9,942 participants): Incentives significantly improved attendance for vaccination and screening (RR 1.92, 95% CI 1.46 to 2.53; Ι²=89%). Subgroup analysis indicated that cash plus other motivational components was more effective than cash or vouchers alone (heterogeneity remained high). There was no evidence that effect size varied by incentive value.
Physical activity (one trial, 70 participants): A statistically significant difference was observed in favour of financial incentives in terms of increased daily physical activity behaviour over a four week intervention period (RR 5.71, 95% CI 1.35 to 24.26).
There was no evidence to suggest publication bias in any analyses.