Explainability Elevated Obstructive Pulmonary Disease Care: Severity Classification, Quality of Life Prediction, and Treatment Impact Assessment

Chronic Obstructive Pulmonary Disease (COPD) is a progressive respiratory condition that significantly impacts individuals' quality of life (QoL). This study uses a comprehensive dataset to classify COPD severity and predict patients' quality of life (QoL). For severity classification, we employ four machine learning (ML) models: K-Nearest Neighbors (KNN), Logistic Regression (LR), Random Forest (RF), and Extreme Gradient Boosting (XGB). These models' Regression (REG) variants are utilized to predict QoL. All outcomes of classification and REG models are evaluated using 10-fold cross-validation. The results demonstrate that XGB achieves an outstanding accuracy of 99.55% in classifying COPD severity, while RF-REG demonstrates superior performance in predicting QoL with a Mean Squared Error (MSE) of 94.95. To understand the decision-making process of all ML models, we utilize Explainable Artificial Intelligence (XAI) techniques. Additionally, we explore survival rates based on classification and REG outcomes using the Cox Proportional Hazard (CoxPH) Model. This research also estimates the Average COPD Treatment Effect (ACTE) on QoL using the Causal Inference Model (CIM), finding that treatment improves QoL by an average of 4.26 units. This result holds true even when accounting for factors like disease severity. The study highlights the potential of ML and XAI in better understanding disease progression and improving patient care through more accurate QoL predictions and treatment effect assessments.