A Quasi-experimental Study of Longitudinal Patterns in Tracked Physical Activity, Before and After a Change in Incentive Strategy. (Preprint)

Abstract

It is well established that sufficient physical activity is important for good physical and mental health [1]. Despite ongoing campaigns to encourage more physical activity (PA), the WHO still report 1 in 4 adults and 2 in 4 adolescents are inactive [1]. Interventions with wide population reach are required. One possible solution is to encourage more PA using smartphone apps, which are currently used by 39.9 million adults in the UK [3]. Recent systematic reviews suggest that financial incentives delivered via apps are effective in encouraging PA [2-6]. However, the question remains whether financial incentives offer a cost effective intervention that could be funded at population level [7]. Lottery-based rather than assured incentives, may offer a way to control cost by fixing the pay-out [8-16].

The aim of this study was to explore patterns of tracked PA by users of a commercial app before and after a change in incentive strategy. This quasi-experimental study took advantage of a business decision to alter the incentives in the BetterPoints app from Assured to Lottery focused, to control costs.

Weekly minutes of tracked PA was explored among a sample of 1666 people. Mean and median minutes of PA were plotted from one month before the start of the Assured condition to the end of the Lottery condition. Confidence intervals were calculated using bias-corrected bootstraps. Wilcoxon tests were used to assess the difference in weekly minutes of PA in the weeks before and after the change in incentive strategy.

There were significant differences in the weekly minutes of activity before and after the change in incentive strategy (Friedman χ2(2) = 42.0, P<.001). The results of the pair-wise tests showed a significant difference between the last week of the Assured condition (25 August) and the first week of the Lottery condition (1 September) (P<.001). Confidence intervals overlapped during the period before and after the change in incentives, suggesting that there was no meaningful change despite the finding of significant differences in our statistical tests.

Time-series analysis with confidence intervals show that important fluctuations happen over time in tracked PA data which we cannot explain. This suggests the significant before/after difference may not represent the effect of a change in incentive strategies between the Assured to Lottery conditions. The important point being that without the longitudinal analysis we may have been misled by the statistical tests to draw premature conclusions about the relative effectiveness of assured versus lottery-based incentive strategies. Longitudinal analyses that take moderating variables into account, such as prior PA, demographics and psychological variables are required to better understand what an effective incentive might be, for whom and at what cost.