FindZebra online search delving into rare disease case reports using natural language processing

Area of Science:

• Medical Informatics
• Rare Diseases
• Clinical Decision Support

Background:

• Early diagnosis significantly impacts rare disease patient outcomes and quality of life.
• Intelligent user interfaces can aid physicians in diagnosing complex rare diseases.
• Heterogeneous phenotypes in rare diseases often complicate accurate diagnosis.

Purpose of the Study:

• To enhance the rare disease search engine FindZebra.com by incorporating case report abstracts from PubMed.
• To improve diagnostic specificity by indexing age, sex, and clinical features.
• To evaluate the clinical relevance of the enhanced search engine for specific rare diseases.

Main Methods:

• Extended FindZebra.com to include PubMed case report abstracts.
• Built Apache Solr search indexes with extracted clinical features, age, and sex.
• Conducted retrospective validation using real-world outcome data for Gaucher and Fabry disease patients.

Main Results:

• Search results were deemed clinically relevant for Fabry disease patients.
• Results were less relevant for Gaucher disease patients, particularly from older case reports.
• A publication date filter was implemented to address relevance issues.

Conclusions:

• The enhanced search engine shows promise for rare disease diagnosis, particularly for Fabry disease.
• Discrepancies in reporting and understanding of diseases like Gaucher disease in literature impact search relevance.
• The addition of a publication date filter improves the tool's utility.