Automated Chronic Obstructive Pulmonary Disease Phenotyping and Control Assessment in Primary Care: Retrospective...

Area of Science:

• Pulmonary Medicine
• Health Informatics
• Clinical Decision Support

Background:

• Chronic obstructive pulmonary disease (COPD) poses a significant global health challenge.
• Primary care settings often lack consistent spirometry and symptom scores for accurate COPD control assessment.
• There is a critical need for scalable, data-driven tools to improve COPD management.

Purpose of the Study:

• To validate the Seleida model for COPD control assessment and phenotyping using real-world primary care data.
• To evaluate the feasibility of integrating Seleida into electronic health record (EHR) systems.

Main Methods:

• Seleida estimates poor control probability (Pr) using annual short-acting bronchodilator dispensations and antibiotic courses for exacerbations.
• The model employs a deterministic, bijective approach for risk estimation and phenotype inference.
• A retrospective cohort of 106 patients was used to assess agreement between Seleida phenotyping systems and clinician classifications.

Main Results:

• Seleida demonstrated perfect agreement between its internal phenotyping systems (Cohen κ=1.00) and substantial concordance with clinician assessments (Cohen κ=0.70).
• The model operates transparently without machine learning and can be embedded in EHRs or used manually.
• It facilitates individualized risk estimation, phenotype-driven treatment, and population-level case identification.

Conclusions:

• Seleida offers a reproducible, interpretable framework for COPD control monitoring using prescribing data.
• Its transparent logic, low data requirements, and interoperability support integration into various digital infrastructures, including resource-limited settings.
• Seleida bridges predictive analytics with clinical decision-making for improved COPD care.