Content Recommendation Systems in Web-Based Mental Health Care: Real-world Application and Formative Evaluation

Area of Science:

• Digital mental health
• Recommender systems
• Personalized medicine

Background:

• Recommender systems offer scalable personalization for self-guided mental health content.
• These systems can supplement traditional mental health treatments.

Purpose of the Study:

• To describe and evaluate two knowledge-based recommender systems within the Ginger mental health platform.
• The goal was to increase user engagement with self-guided mental health content.

Main Methods:

• Developed two algorithms: one using onboarding responses, another using conversation transcripts for recommendations.
• Evaluated recommendation relevance and content completion rates across 14,018 users.

Main Results:

• Content from the recommendations section had higher completion rates (42.6%) than other app sections (37.35%).
• Conversation-based recommendations outperformed onboarding-based (11.4% higher completion) and random recommendations (26.1% higher completion).
• Older users (>35) and male users showed higher responsiveness to personalized recommendations.

Conclusions:

• Recommender systems effectively scale and personalize digital mental health care.
• Onboarding-based systems are useful for new users, while conversation-based systems adapt to evolving needs.
• Future work includes randomized controlled trials for outcome-driven personalization.