300 Section - Decision Tree-Based Prediction of Pulmonary Function Recovery in Lung Cancer Patients Using Early Postoperative Performance Test

Objectives : Postoperative pulmonary function decline is common after lung cancer surgery, yet it is still difficult to predict which patients will have meaningful recovery. This study aimed to develop a decision tree model to identify patients at risk of poor pulmonary recovery, defined as ≥20% decline in forced vital capacity(FVC) or forced expiratory volume in one second(FEV1) at 1 month after surgery, using early postoperative physical performance tests.

Design: This study retrospectively analyzed patients who underwent lung cancer surgery and completed physical assessments within postoperative day 3, including handgrip strength, sit-to-stand repetitions, and 4-m walk test velocity. Preoperative pulmonary function test and body mass index (BMI) were also included as predictors. Poor recovery was defined as ≥20% reduction in FVC or FEV1 at 1 month compared with baseline. Decision tree classifiers were developed, and diagnostic performance was evaluated using 5-fold cross-validation.

Results: A total of 222 patients were included. For FVC, the decision tree identified handgrip strength and walk speed as the main discriminators. Patients with handgrip strength ≤16.3 kg or walk speed ≤1.1 m/s were at high risk of poor recovery. The FVC model showed accuracy 65.8%, sensitivity 79.5%, specificity 56.7% in the training set, with a cross-validated accuracy of 52.7%. For FEV1, walk speed, and BMI were key predictors. Patients with walk speed ≤0.5 m/s and BMI ≤29.7 had the highest risk of poor recovery. The FEV1 model achieved training accuracy 71.6%(sensitivity 26.7%, specificity 94.6%), with cross-validated accuracy 63.1%.

Conclusion: Decision tree analysis revealed clinically interpretable cut-off values of physical performance and pulmonary function that predict postoperative pulmonary recovery. The FVC model showed is more suitable for screening patients at risk, while the FEV1 model may help confirm preserved recovery. These findings suggests that simple performance tests may support early identification of patient who require intensive rehabilitation after lung cancer surgery.