Development and validation of a claims-based prediction model for COPD severity

Area of Science:

• Medical Informatics
• Pulmonology
• Health Services Research

Background:

• Administrative claims are vital for Chronic Obstructive Pulmonary Disease (COPD) research.
• A validated measure for patient COPD severity is lacking in claims data.
• COPD severity is a key determinant of treatment and patient outcomes.

Purpose of the Study:

• To develop and validate a predictive model for COPD severity using administrative claims data.
• To enable COPD patient classification when spirometry or clinical data is unavailable.
• To support COPD research by providing a method for severity assessment from claims.

Main Methods:

• Identified COPD patients with spirometry from an electronic health records database linked to claims.
• Classified patients into GOLD-Unclassified, Mild/Moderate, and Severe/Very Severe COPD groups.
• Constructed a multinomial logistic regression model using claims data for severity prediction, validated on 10% of patients.

Main Results:

• The model, incorporating age, sex, comorbidities, and healthcare utilization, achieved 62.7% overall accuracy in the validation sample.
• Specific accuracy for predicting severity groups: GOLD-Unclassified (73.5%), Mild/Moderate (70.6%), and Severe/Very Severe (81.4%).
• Model performance indicated moderate agreement (kappa = 0.41) for COPD severity classification.

Conclusions:

• A predictive model using administrative claims data was successfully developed to classify COPD severity.
• This model offers a valuable approach for researchers to categorize patients when clinical measures are absent.
• The findings facilitate more robust COPD research by leveraging readily available claims data for severity assessment.