Machine Learning–Based Diagnostic Model for Acute Exacerbation of Chronic Obstructive Pulmonary Disease Using Routine Clinical Laboratory Parameters

Acute exacerbation of chronic obstructive pulmonary disease (COPD), or AECOPD, significantly increases disease burden yet lacks objective diagnostic criteria. We aimed to develop a machine learning model for AECOPD diagnosis using routine laboratory parameters. We analyzed records from 25,965 COPD patients at the First Affiliated Hospital of Guangzhou Medical University, with patients randomized 7:3 into training and test cohorts. We evaluated 113 model combinations from 12 machine learning algorithms, assessing performance through receiver operating characteristic analysis, calibration curves, and decision curve analysis. The generalized linear model boosting + random forest (glmBoost + RF) model demonstrated excellent diagnostic performance (training area under the curve [AUC] = 0.993, test AUC = 0.834) utilizing only nine variables: age, lymphocyte percentage, calcium, hemoglobin, eosinophil percentage, potassium, platelet distribution width, monocytes count, and mean corpuscular hemoglobin concentration. This streamlined model showed performance comparable to the more complex Lasso + RF model (48 variables) with superior clinical applicability. Both models exhibited excellent calibration performance (mean absolute error = 0.012-0.013) and maintained consistent performance across gender-stratified populations. A machine learning model utilizing nine routine clinical laboratory parameters effectively distinguishes AECOPD from stable COPD, providing an objective diagnostic tool applicable across diverse healthcare settings, particularly in resource-limited facilities.