Abstract
OBJECTIVE: To develop an accurate predictive model for identifying patients at high risk of pulmonary infection during radiochemotherapy. METHODS: We retrospectively analyzed data from 544 lung cancer patients treated at Hubei Cancer Hospital between May 2019 and October 2022. The patients were divided into training and validation groups (7:3 ratio). An external validation cohort of 100 patients treated from November 2022 to January 2024 was also included. Feature selection and model development were performed using machine learning algorithms, including Lasso regression, Random Forest, XGBoost, and Support Vector Machine (SVM). Model performance was evaluated using Receiver Operating Characteristic (ROC) curves, calibration curves, and decision curve analysis. RESULTS: Key predictive factors for pulmonary infection risk were identified, including diabetes, chronic obstructive pulmonary disease, chemotherapy intensity, chemotherapy cycles, antibiotic use, age, Karnofsky Performance Status score, systemic inflammation index, prognostic nutritional index, and C-reactive protein. A nomogram-based prediction model was constructed, achieving ROC curve Area Under the Curve values of 0.889 in the training set, 0.897 in the validation set, and 0.875 in the external validation set, demonstrating strong classification ability and stability. CONCLUSION: We developed a robust nomogram-based model incorporating eight key factors to predict the risk of pulmonary infection in lung cancer patients undergoing radiochemotherapy. This model can assist clinicians in early identification of high-risk patients, enabling timely interventions to improve patient outcomes and quality of life.