Introduction: Online programs generate huge data sets that can be used to examine variables related to process and outcome measures, informing personalized medicine and advanced automation. Researchers need to find ways to both visualize and analyze this data. Methods: This study examined data from a large long-term study examining the effects of an Internet- and app-based, coached program (i.e., Student Bodies—Eating Disorders [SB-ED]) for reducing eating disorder (ED) symptoms. 382 college women with ED were offered the 40-session SB-ED program. With support from a trained, online coach (n=32 in the current study), users completed daily sessions at their own pace, having access to the intervention for 8 months (All participants will be followed for up to two years as part of a larger, randomized trial). For ease of analysis, a relational database was constructed, and each touch point (points of contact with the program) corresponded to unique keys labelled in the program. Analyses were conducted primarily with R, plus visualizing and text analysis software programs when appropriate. Results: Initially there were ~290,000 touch points which reduced to 229,000 after data cleaning; 16,859 message between coaches and users were generated. To illustrate visualizing techniques, a number of variables will be presented, including touch points and messages per user over time, daily ED symptoms plotted over “normalized time,” and user engagement by program technique plotted by time of recruitment. Data analyses illustrated will include machine learning to predict engagement and bingeing; text analysis to predict outcomes; and multifactorial longitudinal models to predict symptom improvement. Focusing on symptom improvement is reasonable given that daily average binge frequency dropped significantly over time (F = 9.312, P =.012). Discussion: Data analytic and visualization techniques can be very useful to study process and outcomes, ultimately leading to program improvement, personalization, and automation.

weitere Autoren: C. Barr Taylor, Ellen E. Fitzsimmons-Craft, Katherine Balantekin, Marie-Laure Firebaugh, Grace Monterubio, Corinna Jacobi, Denise Wilfley