Related Experiment Video
Updated: Feb 28, 2026
06:58
An Application for Pairing with Wearable Devices to Monitor Personal Health Status
Published on: February 3, 2022
3.4K
Low-Burden Detection of Clinical Worsening in Body Dysmorphic Disorder Using Smartphone Sensor and Demographic Data
Hilary Weingarden1, Vincent Holstein1, Geneva K Jonathan1
1Massachusetts General Hospital and Harvard Medical School.
Behavior Therapy
|February 25, 2026
Summary
This study shows smartphone sensor data can predict worsening body dysmorphic disorder (BDD) symptoms, including suicidal ideation (SI), enabling early intervention. Machine learning models accurately forecast daily clinical acuity without patient burden.
Area of Science:
- Psychiatry
- Digital Health
- Machine Learning
Background:
- Body dysmorphic disorder (BDD) involves distressing appearance preoccupations, functional impairment, and suicidal ideation (SI).
- Current monitoring methods (self-reports, clinician assessments) have limitations like infrequent administration and recall bias.
- Real-time monitoring via smartphones offers a low-burden approach for early detection of clinical deterioration.
Purpose of the Study:
- To test the feasibility of using smartphone sensor and demographic data to predict daily clinical acuity in individuals with BDD.
- To develop and evaluate machine learning models for real-time monitoring of BDD symptom severity.
Main Methods:
- Eighty-two participants with BDD completed 28-day ecological momentary assessments (EMA) for SI, avoidance, and BDD-related time.
- Smartphone sensor data (GPS, accelerometer) and demographic data were collected over 3 months.
- Random forest (RF) machine learning models were trained to predict same-day clinical outcomes.
Main Results:
- RF models significantly outperformed baseline models in predicting SI, avoidance, and time spent on BDD-related behaviors.
- Strong predictive performance was observed for BDD-related time (r=.74-.75) and mean/max SI (r=.70-.73).
- Step count and demographic factors were key predictive features.
Conclusions:
- Smartphone sensor and demographic data can effectively monitor real-time clinical worsening in BDD without increasing patient burden.
- This approach holds potential for developing just-in-time interventions to prevent symptom escalation.
- Further research is needed to validate these models in real-world settings and integrate them into clinical practice.
