Leveraging Machine Learning and Real-World Data to Predict Chronic Obstructive Pulmonary Disease Exacerbations.

Area of Science:

• Pulmonary Medicine
• Medical Informatics
• Machine Learning

Background:

• Previous research demonstrated AI and NLP's utility in identifying patients at risk of chronic obstructive pulmonary disease (COPD) exacerbations using EHR data.
• This study builds upon prior work by developing a predictive model for COPD exacerbations.

Purpose of the Study:

• To establish a predictive model for identifying patients at risk of COPD exacerbations within 24 months of diagnosis.
• To leverage real-world electronic health record (EHR) data for improved patient risk stratification.

Main Methods:

• Utilized structured and unstructured data from Epic EHR.
• Employed Bayesian Additive Regression Trees (BART), a flexible machine-learning approach, for multivariable prediction.
• Assessed model performance using receiver operating characteristic (ROC) curves and area under the ROC curve (AUC).

Main Results:

• Analysis included 3007 patients; 886 experienced exacerbations within 24 months.
• Bivariate analyses showed associations between exacerbations and cor pulmonale, dyspnea, and comorbidities.
• The BART model identified eosinophil count, pack years, and dyspnea as key predictors, achieving an AUC of 0.69.

Conclusions:

• Eosinophil count and dyspnea are significant predictors of COPD exacerbations.
• Active monitoring of these factors can help identify high-risk patients.
• Tailored therapies based on identified risk factors may improve health outcomes for COPD patients.