Computable Phenotype for Identifying Undiagnosed Hypermobile Ehlers-Danlos Syndrome: Protocol for a Development and...

Area of Science:

• Medical Informatics
• Genetics and Genomics
• Rheumatology

Background:

• Hypermobile Ehlers-Danlos syndrome (hEDS) is a complex hereditary connective tissue disorder with significant diagnostic challenges.
• Delayed diagnosis of hEDS leads to fragmented care and prolonged patient suffering, averaging over 22 years.
• Electronic health records (EHRs) offer a potential data source for identifying patients with undiagnosed hEDS.

Purpose of the Study:

• To develop and validate a computable phenotype for identifying patients who warrant clinical evaluation for hEDS using EHR data.
• To characterize the clinical signature of hEDS through data-driven analysis of EHRs.
• To assess the practical utility and potential of the algorithm to reduce diagnostic delays.

Main Methods:

• Retrospective analysis of a national EHR database (Cosmos) to define the hEDS clinical signature.
• Development of a machine learning model integrating structured EHR data and natural language processing of clinical notes.
• Validation of the model against an expert-adjudicated cohort and assessment of diagnostic delay reduction using discrete event simulation.

Main Results:

• Empirical results are pending as this is a study protocol.
• Expected outputs include a data-driven hEDS clinical signature by late 2026 and a validated computable phenotype by Q1 2027.
• A quantitative estimate of diagnostic delay reduction is anticipated by Q2 2027.

Conclusions:

• The study aims to produce a validated algorithm to identify patients needing evaluation for hEDS, potentially reducing diagnostic delays.
• The identified clinical signature will inform provider education and enhance clinical practice for hEDS.
• This framework advances computable phenotyping for diagnostically challenging conditions like hEDS.