Harnessing Patient-Generated Data for Rare Disease Knowledge Enrichment: A Pilot Study.

Area of Science:

• Medical Informatics
• Patient-Reported Outcomes
• Rare Disease Research

Background:

• Rare diseases present significant diagnostic challenges, leading to a prolonged diagnostic odyssey for patients.
• Patient-generated data, particularly from social media, is an underutilized resource for rare disease research.
• Understanding patient perspectives is crucial for comprehensive rare disease characterization.

Purpose of the Study:

• To characterize rare disease symptoms described by patients on social media.
• To systematically compare patient-reported symptoms with those documented in clinical literature.
• To explore the potential of social media data in improving rare disease diagnosis and care.

Main Methods:

• A hybrid human-AI framework, utilizing GPT-4 with manual verification, was employed to extract symptoms from 229 social media posts representing 166 rare diseases.
• Patient-reported symptoms were standardized and compared against symptom data extracted from PubMed literature.
• Analysis focused on identifying discrepancies and overlaps between patient narratives and clinical documentation.

Main Results:

• Motor impairment, ataxia, pain, and fatigue were the most frequently reported symptoms by patients.
• A low average overlap was found between symptoms reported by patients and those documented in medical literature.
• Patient narratives highlighted quality-of-life impacts, while clinical literature focused on diagnostic markers.

Conclusions:

• Social media data provides unique insights into underreported symptoms and patient-centered concerns in rare diseases.
• Integrating patient-generated data can complement clinical findings and potentially shorten the diagnostic odyssey.
• Harnessing real-world patient insights can inform patient-centered care and guide future rare disease research.