Leveraging causal modeling to predict future COPD

Abstract

Chronic obstructive pulmonary disease (COPD) is a complex, heterogeneous lung condition that claims the lives of 3 million people each year, making it the fourth leading cause of death worldwide¹. The disease is characterized by coughing exacerbations and dyspnea attributable to inflamed bronchial tubes (bronchitis) and damaged alveoli (emphysema). Previous efforts to predict the early stages of disease have seen limited success because COPD patients inconsistently exhibit these symptoms across diseases severities. In this study, we employ causal graph inference to identify clinical variables (from the longitudinal COPDGene® study) linked to COPD onset. Subsequently, these variables are used as predictors in a logistic regression model to identify healthy subjects that develop lung function abnormalities. Of the 115 input variables, 14 were identified as being causally linked to early-stage COPD progression, with spirometry measurements and airway wall thickness being the most prominent. Model fitting and cross-validation resulted in an area under the receiver operating characteristic curve (AUROC) of 0.78 demonstrating good predictability. Overall, this is an encouraging study, showing that unbiased approaches can determine factors that both differentiate and predict future COPD progression, which can potentially guide clinical treatments for individual patients. 1. Chronic obstructive pulmonary disease (COPD) fact-sheets. WHO (2017)

• Download figure
• Open in new tab
• Download powerpoint