Lessons learned from a multimodal sensor-based eHealth approach for treating pediatric obsessive-compulsive disorder

Carolin S. Klein, Karsten Hollmann, Jan Kühnhausen, Annika K. Alt, Anja Pascher, Lennart Seizer, Jonas Primbs, Winfried Ilg, Annika Thierfelder, Björn Severitt, Helene Passon, Ursula Wörz, Heinrich Lautenbacher, Wolfgang A. Bethge, Johanna Löchner, Martin Holderried, Walter Swoboda, Enkelejda Kasneci, Martin A. Giese, Christian Ernst, Gottfried M. Barth, Annette Conzelmann, Michael Menth, Caterina Gawrilow, Tobias J. RennerFrontiers Digit. Health 2024 · DOI

TL;DR

This paper introduces a novel multimodal sensor-based eHealth approach for treating pediatric obsessive-compulsive disorder (OCD). By integrating real-time physiological data (heart rate, eye movements, motion) into online therapy sessions, the study demonstrates promising feasibility, high user satisfaction, and potential for more personalized and effective treatment for young OCD patients.

What Problem Does This Paper Address?

Treating pediatric obsessive-compulsive disorder (OCD) faces significant barriers, including limited access to specialized care, geographical limitations, and the inherent challenges of conducting exposure therapy effectively. Traditional eHealth interventions, while improving accessibility, often lack objective, real-time data to assess a patient's emotional and physical state during critical therapeutic moments like exposure with response prevention (E/RP). This gap hinders therapists' ability to accurately gauge stress responses, identify avoidance behaviors, and tailor interventions dynamically. The paper addresses this by exploring the feasibility and usability of a sensor-supported eHealth system designed to bridge this gap, offering a more ecologically valid assessment and potentially enhancing treatment efficacy.

Methods

The study developed and evaluated a multimodal sensor-based eHealth intervention delivered over 14 video sessions of cognitive-behavioral therapy (CBT) for 20 pediatric OCD patients (aged 12-18). The core of the methodology involved integrating several sensor technologies:

  • Eye Trackers: Recorded eye movements and gaze direction to identify focus on symptom-triggering objects and potential avoidance behaviors.
  • ECG Chest Strap: Captured heart rate (HR) to indicate stress responses.
  • Motion Sensors: Detected approach and avoidance behaviors through movement analysis.

This data was streamed in real-time to therapists via a web application (therapist UI), allowing them to monitor patients' physiological states and tailor sessions accordingly. Patients also used an ambulatory assessment web application with gamification elements to encourage daily and weekly self-reporting. The study focused on evaluating the feasibility, usability, acceptance, and satisfaction of this integrated system from the perspectives of patients, parents, and therapists.

Datasets

The study did not utilize pre-existing datasets for training. Instead, it collected data from a cohort of 20 pediatric OCD patients during a 14-week treatment period. This included:

  • Physiological Data: Real-time recordings from eye trackers, ECG chest straps, and motion sensors.
  • Questionnaire Data: Responses from the System Usability Scale (SUS), Client Satisfaction Questionnaire-8 (CSQ-8), and a Questionnaire for the Evaluation of the Treatment (FBB) administered at various points.
  • Ambulatory Assessment Data: Daily and weekly self-reported data collected via a web application.

No new benchmarks were introduced; rather, the study focused on evaluating the performance and user experience of the developed system within a clinical trial context.

Key Insights and Ideas

  • Real-time Physiological Data Enhances Therapy: The integration of real-time HR, gaze, and motion data provided therapists with objective insights into patients' stress and avoidance behaviors, enabling more informed session adjustments.
  • High User Acceptance and Satisfaction: Patients, parents, and therapists reported high levels of satisfaction with the sensor-supported eHealth approach, indicating a positive reception to this novel treatment modality.
  • Therapeutic Alliance Can Be Established Online: Despite the remote nature of the therapy, a strong and trusting therapeutic relationship was successfully formed between therapists and patients.
  • Gamification for Engagement: While the gamified ambulatory assessment app was generally well-received, adherence to daily/weekly questionnaires declined over time, suggesting that while motivating, it may not fully overcome the burden of consistent self-reporting.
  • Technical Challenges Remain: While the system showed promise, technical issues like internet connectivity disruptions and the need for robust IT support were noted as areas for improvement.

Key Results

  • Usability: Patients rated the technical components with an average SUS score between 81.52 and 85.32, indicating "excellent" usability. The video conferencing system and ECG strap received the highest scores.
  • Patient Satisfaction: Patients achieved a mean CSQ-8 score of 27.5 (out of 32), signifying high satisfaction with the treatment.
  • Treatment Effectiveness (Perceived): Patients and parents reported a reduction in OCD symptoms at the end of the therapy. Parents rated the therapy as successful, with patients showing fewer compulsions.
  • Therapeutic Relationship: Both patients and parents reported a trustworthy and positive therapeutic relationship.
  • Ambulatory Assessment Adherence: Compliance with daily questionnaires was low (76.7% missing data) and declined over time, while weekly questionnaire compliance was also moderate (55.4% missing data). However, all patients completed the 14 iCBT sessions with no dropouts.
  • Therapist Feedback: Therapists found the real-time physiological data beneficial for understanding patient tension and adapting sessions, though they noted some technical difficulties and challenges in simultaneously monitoring data and conducting therapy.

Limitations and Future Work

The authors acknowledge several limitations:

  • Small Sample Size and Single-Group Design: The study used a single group of 20 patients, limiting generalizability. Future research should involve larger, controlled trials.
  • Adherence Decline: The decrease in ambulatory assessment adherence suggests a need for further optimization of engagement strategies beyond gamification.
  • Technical Refinements: Issues with internet connectivity, the ergonomics of the eye-tracking glasses, and the need for more robust real-time data transmission require further development.
  • Limited Exposure Settings: Exposures were confined to the home setting due to data protection concerns regarding uninvolved third parties in eye-tracking videos. Future work could explore methods for anonymous recording in other environments.
  • Selective Sample: The study may have attracted a more technology-oriented patient sample, potentially limiting transferability to the general population.

Future work should focus on refining the technical aspects, exploring more effective adherence strategies, and expanding the application of this multimodal approach to other mental health disorders and diverse patient populations. The authors also provide a checklist of 10 key considerations for developing future sensor-based therapy approaches.

Summary updated

BibTeX
@article{klein2024lessons,
  author = {Carolin S. Klein and Karsten Hollmann and Jan Kühnhausen and Annika K. Alt and Anja Pascher and Lennart Seizer and Jonas Primbs and Winfried Ilg and Annika Thierfelder and Björn Severitt and Helene Passon and Ursula Wörz and Heinrich Lautenbacher and Wolfgang A. Bethge and Johanna Löchner and Martin Holderried and Walter Swoboda and Enkelejda Kasneci and Martin A. Giese and Christian Ernst and Gottfried M. Barth and Annette Conzelmann and Michael Menth and Caterina Gawrilow and Tobias J. Renner},
  title = {{Lessons learned from a multimodal sensor-based eHealth approach for treating pediatric obsessive-compulsive disorder}},
  journal = {Frontiers Digit. Health},
  year = {2024},
  doi = {10.3389/fdgth.2024.1384540},
}
Figure