Theoretical Basis for an Edge-based, mHealth App to Guide Self-Management of Recurrent Medical Conditions

Background N-of-1 trials have been proposed as an approach to identify the optimal individual treatment for patients with a number of recurrent medical conditions, including chronic pain and mental health. When inserted into mHealth applications, this approach holds great promise to provide an automated, efficient method to individualize patient care; however, prior to implementation, an understanding of the properties of the recurrent condition needed to draw conclusions with sufficient power is needed. Methods We applied simulation studies and power calculations to determine statistical properties of the N-of-1 approach employed by an mHealth application for self-management of chronic recurrent medical conditions called the iMTracker. Results In 1000 simulated patients with a single recurrent medical condition and 5 possible associated conditions, we found that ~90 days of data collection was sufficient to identify associated risk factors with odds ratio (OR > 5.0) at power ≥ 80%, with an absolute event rate of 50% being optimal. Power calculations based on Fisher’s Exact test showed that 90 days was also sufficient to detect a decrease of 20% in the rate of the primary outcome after an intervention, but that shorter data periods could be used to identify stronger effect sizes, down to 15 days with a 90% reduction in rate. Repeat analysis with Bayesian models did not significantly change power calculations, but did allow for a flexible approach that we leveraged to create a web-based tool to allow users to perform power calculations prior to using the iMTracker for self-management. Conclusions We found that the N-of-1 approach employed in the iMTracker app for self-management of recurrent medical conditions is statistically feasible, given the right conditions. More work is needed to examine the impact of autocorrelation, seasonality, and trends in data, on statistical validity and power calculations.