Background: Cognitive computing technology providing personalized insights based on correlations between behavior patterns and glycemic outcomes was leveraged to create an interactive mobile assistant for T1D patients.
Method: Medtronic MiniMed™ Connect users (N=256) were invited to take part in a 90-day Sugar.IQ app pilot program that began April 11, 2017. The percentage of time in target range (TIR, 70-180 mg/dL), <70mg/dL, >180mg/dL, and excursions (periods >20 minutes and <70mg/dL or >180mg/dL) were collected 30 days before Sugar.IQ onboarding, and compared to those 90 days later (August 2017). Insights delivered to users and user feedback were also analyzed.
Results: There were 11,356 sensor-wear days; 10,761 unique Sugar.IQ usage sessions collected; and 4,688 insights delivered (1 every 3 days). Insights included 655 and 699 identifying behaviors associated with more time <70mg/dL and >180mg/dL, respectively. The Sugar.IQ app was used 2.1 times/day. Compared to baseline, TIR was 33 minutes longer per day (P<0.15) and hypoglycemia events reduced by 1.0 per month (P<0.001). A week after receiving insights associated with hypoglycemia, 55% and 54% of users had fewer hypoglycemia and hyperglycemia events, respectively. Hyperglycemia events >2 hours reduced by 1.3 per month (P<0.001). After receiving insights about low glucose associated with boluses delivered at rapid rates of change, users tended to take smaller boluses and consume less carb in the following 7 days. Among insights that included user feedback, 86% of users rated them as “Helpful” vs. “Not helpful.”
Conclusion: Timely and personalized insights, such as those provided during the Sugar.IQ pilot, may advance patient understanding of glucose trends, aid in behavioral change that improves therapy adherence, and lead to better outcomes.
Y. Zhong: None. S. Arunachalam: None. P. Agrawal: None. H. Neemuchwala: None. T.L. Cordero: Employee; Self; Medtronic. F.R. Kaufman: Employee; Self; Medtronic.