Next generation phenotyping using narrative reports in a rare disease clinical data warehouse

Area of Science:

• Biomedical Informatics
• Rare Disease Research
• Clinical Data Mining

Background:

• Electronic Health Records (EHRs) offer valuable data for rare disease research.
• The Necker-Enfants Malades Hospital's Dr. Warehouse clinical data warehouse facilitates exploration of patient clinical narratives.
• Identifying phenotypes associated with rare diseases is crucial for advancing knowledge.

Purpose of the Study:

• To present a method for finding phenotypes linked to specific rare diseases using EHR data.
• To explore the association between clinical phenotypes and rare diseases through data mining.

Main Methods:

• Leveraged frequency and TF-IDF to analyze associations between clinical phenotypes and rare diseases.
• Applied the method to six rare disease use cases, including Rett syndrome and Activated PI3-kinase Delta Syndrome (APDS).
• Domain experts evaluated the relevance of identified phenotypes, calculating average and mean average precision.

Main Results:

• Experts identified 16-39 relevant phenotypes within the top 50 for frequency and 11-41 for TF-IDF.
• Average precision ranged from 0.55-0.91 (frequency) and 0.52-0.95 (TF-IDF), with a mean average precision of 0.79.
• The study validates the utility of EHR-derived phenotypes for rare disease specialists.

Conclusions:

• Clinical Data Warehouses enable Next Generation Phenotyping for rare diseases.
• A method was developed to detect patient phenotypes from free-text clinical narratives.
• EHR data mining provides valuable candidate phenotypes for rare disease research.