Computing latent taxonomies from patients' spontaneous self-disclosure to form compatible support groups

Area of Science:

• Health Informatics
• Information Retrieval
• Computational Social Science

Background:

• Online platforms enable patient experience sharing, but finding compatible support groups is often left to chance.
• Existing methods lack systematic approaches for matching patients with similar health journeys.

Purpose of the Study:

• To apply Information Retrieval (IR) techniques for systematically identifying compatible patient groups for online mutual support.
• To develop and evaluate an algorithm for creating a latent taxonomy of patient self-disclosures to facilitate support group formation.

Main Methods:

• Utilized Information Retrieval (IR) algorithms to analyze patient self-disclosures from online platforms.
• Developed a novel algorithm to generate a latent taxonomy based on patient narratives.
• Compared taxonomy results derived from patient stories versus explicit questionnaires.

Main Results:

• Demonstrated the feasibility of using IR tools to systematically match patients for online support.
• The latent taxonomy effectively organizes patient self-disclosures for improved group formation.
• Preliminary results suggest differences between story-based and questionnaire-based taxonomies.

Conclusions:

• Information Retrieval offers a powerful framework for enhancing patient-to-patient support systems.
• The proposed algorithmic approach can systematically improve the formation of effective online patient support groups.
• Future work should further explore the nuances of taxonomy generation from diverse data sources.