Logo image
Back
Using wearable passive sensing to predict engagement in binge eating in response to negative affect: a multimethod investigation of predictive utility, feasibility, and acceptability
Dissertation   Open access

Using wearable passive sensing to predict engagement in binge eating in response to negative affect: a multimethod investigation of predictive utility, feasibility, and acceptability

Emily Kelley Presseller
Doctor of Philosophy (Ph.D.), Drexel University
Apr 2025
DOI:
https://doi.org/10.17918/00010967
pdf
Presseller_Emily_20251.88 MBDownloadView
PDF Open Access Open Access (License Unspecified)

Abstract

Affect Binge eating Psychophysiology
Objective: Binge eating, characterized by eating a large amount of food accompanied by a sense of loss of control over eating, is a public health crisis. The affect regulation model of binge eating posits that elevated negative affect increases momentary risk for binge eating, as engaging in binge eating alleviates negative affect and reinforces the behavior. However, the field's existing capacity to identify moments of elevated negative affect, and thus risk for binge eating, has exclusively relied on ecological momentary assessment (EMA), which involves the completion of surveys in real time on one's smartphone to report behavioral, cognitive, and emotional symptoms. EMA surveys are often delivered only 5-6 times per day, involve self-report of affect intensity only, and are unable to assess affect-related physiological arousal. Wearable, psychophysiological sensors that measure markers of affect arousal (e.g., heart rate, heart rate variability, and electrodermal activity) may augment EMA surveys to improve our capacity to accurately detect risk for binge eating in real time. However, it is unknown whether data from these sensors can adequately distinguish between positive and negative affect states, as physiological arousal may occur during both negative and positive affect states. The aims of this study were: 1) test the hypothesis that sensor features can distinguish positive and negative affect states in individuals with binge eating with > 60% accuracy, 2) test the hypothesis that a machine learning algorithm using sensor data and EMA-reported negative affect data to predict binge eating episodes can predict binge eating with greater accuracy than an algorithm using EMA-reported negative affect alone, 3) evaluate the feasibility and acceptability of using sensors among adults with binge eating. The study also involved an exploratory aim to evaluate user design preferences for a sensor-powered ecological momentary intervention system for adults with binge eating. Method: The study recruited 30 individuals with clinically significant binge eating who wore Fitbit Sense 2 smartwatches to passively measure heart rate and electrodermal activity and reported affect and binge eating on EMA surveys for four weeks. The participants and 6 community eating disorder clinicians completed self-report surveys and qualitative interviews on the feasibility, acceptability, and user design preferences for a future sensor-integrated momentary intervention. Results: The best performing model using sensor data to distinguish positive and negative affect demonstrated accuracy of 0.63, sensitivity of 0.20, and specificity of 0.93; the model's accuracy exceeded 0.60 indicating adequate accuracy. The best performing model using EMA data alone for predicting binge eating had accuracy of 0.65, sensitivity of 0.53, and specificity of 0.69. The optimal model combining EMA data and sensors data for the predicting binge eating demonstrated accuracy of 0.58, sensitivity of 0.90, and specificity of 0.48. Mixed methods data from both participants with binge eating and eating disorder clinicians substantiated the feasibility and acceptability of smartwatches and a future sensor-integrated digital intervention for binge eating. Participants with binge eating and community eating disorder clinicians preferred the just-in-time adaptive intervention format. Conclusions: Findings from the present study indicate that data from psychophysiological sensors included in commercial smartwatches can distinguish between instances of positive and negative affect among individuals with binge eating and that these data augment EMA data for accurate prediction of binge eating episodes. Individuals with binge eating and eating disorder clinicians are enthusiastic about the potential for a digital intervention to augment outpatient therapy and data collected in this study set the stage for the future user-centered design of such an intervention.

Metrics

2 File views/ downloads
137 Record Views

Details

Logo image