The effect of monetary incentives on physical activity goal achievement. A systematic review and meta analysis

Physical inactivity is a huge issue in today’s world, leading to high costs for society. Interventions aiming at the improvement of physical activity are widely spread. One approach uses monetary incentives for improving activity. General positive effects on the increase of steps or gym visits are known, but not whether these increases are driven by some top performers or all participants. Furthermore, little is known about the effect on absolute goal achievement because it is seldom reported. The combination of activity which can be measured as degree of target achievement (i.e. 56 percent for 5600 steps in a 10,000 step goal condition) and absolute target achievement (dichotomous) offers a more wholesome picture of the functionality of physical activity goals. Furthermore, these two variables can be calculated for all sorts of physical activity, wherefore i.e. step and attendance goals can be compared. Also, this approach offers a contribution to the goal setting literature, which is highly theoretical in this context. In a first step a systematic literature review is conducted to showcase the different studies and whether the degree of target achievement and absolute goal achievement is reported. Therefore, only randomized controlled trials (RCTs) were used. Furthermore, physical activity had to be measured objectively in some sort, e.g. with pedometers or attendance logs. Studies encouraging physical activity with the goal to reduce weight were excluded, except if a part of the incentive was assigned to the improvement of physical activity. Third, studies using all forms of financial incentives received directly by the performing subject (i.e. studies where parents could earn money for the performance of their children were excluded) were valid. This includes the usage of direct monetary means, coupons, gift cards or gifts, but also interventions where the money was spent to charity, since this may lead to tax reductions, therefore having a direct monetary effect. Also, studies with all types of monetary incentives like lotteries and direct gifts were included. Fifth, it is only looked at the intervention part of the respective study, whereas long-term effects concerning habit formation are not considered because the lack of data. Sixth, all intervention durations and only studies in which a clear relationship between an action and the earned payoff was clear were eligible. This includes studies were no clear goal was given, but a theoretical maximum goal achievement could be calculated. Also, studies using an incentive configuration with the combination of financial with other incentives were included. To date 30 studies with 10,635 participants could be included. In a next step, a pooled random-effects analysis (DerSimonian and Laird) of mean differences between treatment arms and control groups is conducted in R with degree of target achievement and absolute goal achievement as dependent variables. For treatment arms with identical control groups, the number of participants in control groups was divided up, while leaving means and standard deviations unchanged. It is not yet clear if this approach will be changed since the resulting comparisons remain correlated. Pooling the intervention groups is no option because the variation of monetary means is too large in most cases. Concerning the comparison in the degree of target achievement variable, 59 treatment arms are included, while 31 treatment arms are included in the absolute target achievement arm analysis. Authors will be contacted for missing data in the next weeks when a 2nd coder has finished her work. Forest plots are created indicating a mean difference of the degree of target achievement favoring treatment groups with 21.60 percentage points. The mean difference concerning absolute target achievement is 20.00 percentage points in favor of treatment conditions. In a next step a subgroup analysis is conducted, comparing i.e. the effects of different activity goals, intervention lengths, incentive types and incentive values. The most notable findings are that interventions with young participants (< 25 years), high incentives (> median incentive value of $1.53 daily), direct gifts and above median step goals (> 49,000 steps) lead to the largest mean differences. No difference is found between treatments with above median attendance goals (> 2 visits per week) and those under. Concerning absolute goal achievement, most striking mean differences become exposed for treatment arms with above median incentive values, direct gifts, step goals (opposed to attendance goals) and above median goals (> 49,000 steps and > 1.5 visits). In a last step, meta-regressions are done. The model with the best fit (R2 = 0.3) for the explanation of the degree of target achievement shows significant effects for the incentive value, attendance goals and intervention length. The model with the best fit (R2 = 0.71) for absolute goal achievement shows significant positive effects regarding incentive value, step goals and direct gift interventions while age has a significant negative effect. Also, bias is tested, by using the Egger’s test which indicates a significant asymmetry in the absolute goal achievement context. First conclusions on the basis of these findings are that higher incentive values and longer interventions lead to higher degrees of and absolute rates of target achievement. Additionally, it seems logical that attendance goals have higher DTA rates, since one additional visit already has a high impact, while it is harder for individuals with step goals to increase their DTA. However, individuals with step goals reach these goals significantly more often than those with attendance goals. A contradicting finding is that individuals that are active before the intervention show higher DTA rates but fail to achieve the respective goals more often. However, one can argue, that inactive individuals are more likely to be motivated to achieve the goal, while individuals who take the intervention seriously and are active overshoot goals more easily. Furthermore, already active individuals tend to have already achieved a way they exercise, that is different from the incentivized and measured one. Nevertheless, these preliminary results have to be handled with care and a further analysis in the next months after the respective data is acquired is awaited with excitement.