Can fitness apps work long term? A 24-month quasiexperiment of 516 818 Canadian fitness app users

Objective To examine whether a multicomponent commercial fitness app with very small (‘micro’) financial incentives (FI) increased population-level device-assessed physical activity (PA) over 2 years. The secondary objective was to explore the influence of select covariates on longitudinal effects. Methods This 24 month pre–post quasiexperiment was conducted in Ontario, Canada’s largest province (December 2016–June 2019). Following a 1-to-2 week baseline period, users earned micro-FIs ($0.04 CAD/day) for achieving daily step goals. Multiple linear regression models estimated changes in weekly mean daily step count from baseline to key timepoints (eg, 24 months). To address the secondary objective, separate models were developed for each level of the selected covariates (eg, start season, baseline PA). Results The sample included 516 818 users (% female: 62.83; age (SD): 33.46 (12.65) years). Half were ‘low’ active at baseline (<5000 daily steps; 47.15%). Overall, daily step counts were greater than baseline at all key timepoints (eg, 242 steps/day at 24 months; p<0.001). Users from earlier start seasons and longer FI exposure exhibited larger differences from baseline (eg, 758 steps/day at 24 months; p<0.001). Differences were also more pronounced among ‘low’ active users (eg, 1986 steps/day at 24 months; p<0.001). Substantial daily step count reductions were observed among ‘very high’ active users (≥10 000 daily steps; eg, −3969 steps/day at 24 months; p<0.001). Conclusion Modest PA increases of about 250 steps per day were sustained over 2 years. For important subgroups (ie, earlier start seasons, ‘low’ active) increases approached or surpassed 1000 steps/day—a level indicative of clinical significance. Substantial daily step count reductions among higher active users were also observed. Data are available upon reasonable request.