Conversational Artificial Intelligence for Achieving Activity Targets through Routine Physical Activity—Longitudinal Observational Study among People with Type 2 Diabetes

Objective: People with type 2 diabetes mellitus are prescribed to complete a minimum of 30 mins of activity every day. We evaluate the effectiveness of the Wellthy Diabetes (WD) digital therapeutic app in changing health behavior to achieve these activity goals through routine activity, by coaching and nudging them with the help of an artificial intelligence (AI) powered chatbot.

Methods: De-identified data from 527 users [Mean = 38 years (95% CI: 36.8- 39.3); 69.7% male] who had used the app for at least 21 days and had activity tracking data for at least 15 days was used. The AI-powered chatbot delivered guidance and education on the importance of exercise for persons with diabetes and encouraged users to engage in routine physical activity through personalized nudges and reminders delivered through an AI-Powered digital persuasion model. The WD app tracked the users’ physical activity via inbuilt sensors on the phone.

Results: A total of 46,320 person-days and 8,833 person-weeks of activity tracking was recorded, wherein users were tracked for an average of 87 days (95% CI: 81- 95). We recorded the daily active time (DAT) and weekly active time (WAT). Mean DAT was 35 min (95% CI: 33- 37) with a median of 29.75 min (25th-75th: 18-45 min); it was significantly higher for males (37.5 vs. 29.5 min; p=0.001). Mean WAT was 187 min (95% CI: 174-200) with a median of 153 min (25th-75th: 77-242 min). MDAT was noted to be higher on weekdays (38 min) vs. weekends (35 min, p=0.001). Daily activity targets were achieved on 46.5% person-days (n=21542) and weekly activity targets on 47.5% of person-weeks (n=4197).

Conclusion: The results confirm that an AI-powered Digital Therapeutic like the Wellthy Diabetes can be an effective prescriptive tool for physicians to help their patients achieve activity targets.