Development and Validation of an Electronic Health Record Algorithm to Predict the Presence of Chronic Obstructive Pulmonary Disease.

Purpose: Chronic obstructive pulmonary disease (COPD) is frequently diagnosed and treated based on clinical suspicion alone, without spirometric confirmation of expiratory airflow obstruction (AFO, defined by a forced expiratory volume in 1 second (FEV Patients and Methods: Our analysis included 15,065 patients who underwent pulmonary function testing between 2016-2022. Patients were categorized as having AFO (n=4632) or not (n=10,433) based on spirometry. Patients < 45 years, with cystic fibrosis, alpha-1 antitrypsin deficiency, or prior spirometric evidence of obstruction were excluded. Logistic regression assessed 65 variables, retaining those that optimized model discrimination. The data were randomly split into training (n=10,546) and validation (n=4519) sets. Results: Key predictors of AFO included older age, male sex, lower BMI, smoking history, prior COPD diagnosis, increased emergency department utilization, fewer outpatient visits, fewer chest X-rays, and higher cumulative beta-agonist prescriptions. The final model achieved an area under the receiver operating characteristic curve (AUC) of 0.82 (95% CI: 0.81-0.83) in the validation dataset and was well calibrated. Conclusion: We developed an EHR-based algorithm that accurately predicts AFO using routinely collected structured data. This tool provides a practical method for identifying patients for targeted COPD case finding. Future efforts will focus on external validation and integration into clinical workflows, enabling automated identification and provider or patient notification to facilitate appropriate pulmonary function testing.