Medical Expert Knowledge Meets AI to Enhance Symptom Checker Performance for Rare Disease Identification in Fabry...

Area of Science:

• Medical informatics
• Artificial intelligence in healthcare
• Rare disease diagnostics

Background:

• Rare diseases present diagnostic challenges, often leading to delays and increased healthcare costs.
• Artificial intelligence (AI)-powered symptom checkers (SCs) show promise for early rare disease detection.
• Current SCs rely on literature data, which may be incomplete for rare diseases, impacting accuracy.

Purpose of the Study:

• To enhance AI-powered SCs by incorporating expert interview vignettes.
• To evaluate if this novel approach improves diagnostic accuracy for rare diseases.
• To assess user satisfaction with the enhanced SC, focusing on Fabry disease.

Main Methods:

• A mixed-methods pilot study at Hannover Medical School, Germany.
• Expert interviews generated clinical vignettes to enrich the AI SC's Fabry disease model.
• Patients with Fabry disease evaluated both original and optimized SC versions for diagnostic accuracy and satisfaction.

Main Results:

• Expert vignettes improved AI SC performance in identifying Fabry disease.
• The optimized SC identified Fabry disease as the top suggestion in 33% of cases, versus 17% for the original.
• User satisfaction was higher with the optimized SC, citing better symptom coverage and completeness.

Conclusions:

• Integrating expert-derived vignettes enhances AI SCs for rare disease diagnosis.
• The optimized SC demonstrated improved diagnostic accuracy and user satisfaction for Fabry disease.
• Future research should include atypical presentations and larger validation studies.