Fig. 2. Factors impacting participant retention in digital health studies.

a Proportion of active tasks (N = 3.3 million) completed by participants based on their local time of day. The centerline of the boxplot shows the median value across the studies and upper and lower whisker corresponding to outlier point (1.5 times the interquartile range). b Kaplan Meir survival curve showing significant differences (P < 1e-16) in user retention across the apps. Brighten App where monetary incentives were given to participants showed the longest retention time(Median = 26 days, 95% CI = 17–33) followed by Asthma(Median = 12 days, 95% CI = 11–13), MyHeartCounts(Median = 9 days, 95% CI = 9–9), ElevateMS(Median = 7 days, 95% CI = 5–10), mPower(Median = 5 days, 95% CI = 4–5), Phendo(Median = 4 days, 95% CI = 3–4), Start(Median = 2 days, 95% CI = 2–2) and SleepHealth(Median = 2 days, 95% CI = 2–2), c Lift curve showing the change in median survival time (with 95% CI indicated by error bars) based on the minimum number of days(1–32) a subset of participants continued to use the study apps, Kaplan-Meier survival curve showing significant differences in user retention across d Age group, with 60 years and older using the apps for longest duration(Median = 7days, 95% CI = 6–8, P < 1e-16) followed by 50–59 years (Median = 4 days, 95% CI = 4–5) and 17–49 years (Median = 2–3 days, 95% CI = 2–3). e Disease status; participants reporting having a disease stayed active longer(N₅₀ = 13days, 95% CI = 13–14) compared to people without disease(N₅₀ = 6 days, 95% CI = 5–6) and finally f Clinical referral; Two studies (mPower and ElevateMS), had a subpopulation, that were referred to the study by clinicians and showed significantly (P < 1e-16) longer app usage period(Median = 44 days, 95% CI = 27–58) compared to self-referred participants with disease (N₅₀ = 4 days, 95% CI = 4–4). For all survival curves the shaded region shows the 95% confidence limits based on the survival model fit.