Rapid machine learning model for differentiating asthma and chronic obstructive pulmonary disease using age and blood...

Area of Science:

• Pulmonary Medicine
• Medical Diagnostics
• Machine Learning in Healthcare

Background:

• Differentiating Chronic Obstructive Pulmonary Disease (COPD) from asthma in emergency settings is challenging due to symptom overlap.
• Accurate and timely diagnosis is crucial for appropriate treatment initiation.
• Traditional spirometry is often impractical in urgent care settings.

Purpose of the Study:

• To develop a rapid, accessible diagnostic model for differentiating COPD from asthma.
• To utilize routine blood tests for timely clinical decision-making in emergency settings.

Main Methods:

• Analysis of clinical and laboratory data from 9,038 patients diagnosed with COPD or asthma.
• Development of a machine learning model using LASSO regression and AdaBoostClassifier.
• Validation of the model on training, internal, and external cohorts, assessing performance via sensitivity, specificity, accuracy, and AUC.

Main Results:

• The model identified seven key predictors: MCHC, age, LYMPH, HGB, PCT, MONO, and EO.
• Achieved high AUCs (0.890 training, 0.871 internal, 0.855 external) and accuracies (0.810 training, 0.804 internal, 0.792 external).
• Demonstrated strong predictive reliability and clinical utility, with an online tool developed for practical application.

Conclusions:

• A machine learning model using age and routine blood parameters effectively differentiates COPD from asthma.
• The model offers a convenient, accurate, and widely applicable solution for early diagnosis.
• Reliance on standard blood tests supports timely diagnosis in diverse healthcare settings, enhancing patient outcomes.