Background: Mobile application (app)-based self-tracking of weight, diet and exercise is growing in popularity. Both type 1 (T1D) and type 2 (T2D) diabetes require self-management of these areas, so there is potential for app-based tracking to improve diabetes control. Many studies have reviewed mobile apps for diabetes, but the purported health benefits of tracking behavior using apps have been hard to quantify. We investigated the relationship between wellness tracking and self reported blood glucose (BG) levels in individuals with T1D and T2D using the One Drop app.
Methods: Inclusion criteria was at least 20 BG readings between Jan. 2015 and Jan. 2016, with at least 90 days between the first and the last reading. Data included self-reported carbohydrate (carbs) intake, insulin doses, and physical activity reported through the One Drop app. Outcome of interest studied was effective BG control, approximated by %OOR (percentage of Out-Of-Range BG readings, below 70 or above 180 mg/dL). Explanatory variables studied were logarithm of the counts (log-count) of each kind of self-report. We created a panel dataset consisting of the %OOR and the recording log-counts measured each month for each user. We used fixed-effects panel regression to model the association between effective BG control and the amount of wellness tracking while controlling for heterogeneity between users. This controls for unobserved confounders such as socioeconomic status.
Results (BG tracking and %OOR): BG readings averaged 21±40 and 16±27 BG readings/month in the T1D and T2D populations, respectively. For T1D, 37% of BG readings were OOR compared to 20% for T2D. For T2D, there was a significant association between increased BG logging and decreased %OOR. Doubling BG logging was associated with a reduction in %OOR of 0.78 (95% CI [0.28 – 1.29], p = 0.002). That’s a relative improvement in BG control of 3.9%. Users with T1D had a similar, but non-significant pattern, decreasing the %OOR by 0.59 (95% CI [0.44 – 1.62], p = 0.262).
Results (Wellness tracking and %OOR): There were no significant changes in %OOR associated with increased activity tracking. There were significant and near-significant increases in %OOR in T2D and T1D respectively associated with increased tracking of insulin injections, suggesting OOR BG readings encourage users to inject insulin more frequently.
Conclusions: Increased BG logging with the One Drop app was associated with fewer BG readings outside the ideal 70-180 mg/dL range, after controlling for heterogeneity between users. Future analyses will examine any relationships by user age, insulin status, and duration of diabetes. This study is relatively low power. Larger studies are needed, particularly on the association between BG control and activity tracking and on the smaller T1D population. Further research should assess whether the increased stability in observed BG measurements reflects improved glucose control among One Drop users.