Pain-fatigue-sleep disturbance symptom (PFS) cluster is the most common symptom cluster in patients with lung cancer following chemotherapy, which significantly impacts their quality of life. This study aims to develop and validate a machine learning-based prediction model for the severe PFS cluster and identify the relevant factors in patients with lung cancer following chemotherapy.

A total of 612 patients were enrolled in the study, and logistic regression, along with four machine learning algorithms, was used. The area under the curve (AUC), accuracy, sensitivity, specificity, and Brier score were utilized for model evaluation. The Shapley additive interpretation and restricted cubic splines were employed to assess the significance of feature coefficients. A web-based application was developed to facilitate the practical implementation of the best model in clinical settings.

The random forest model was identified as optimal, exhibiting the best discrimination and calibration in the test set (AUC: 0.765 and Brier score: 0.159) and excellent performance in the validation set (AUC: 0.914 and Brier score: 0.124). The factors encompassed in the model construction comprised stress, C-reactive protein, depression, body mass index (BMI), anxiety, neutrophils, age, gender, pathological classification, and Eastern Cooperative Oncology Group performance status. A nonlinear relationship existed between stress, BMI, age, and the severe PFS cluster.

The developed web program would assist health care professionals in accurately identifying patients experiencing the severe PFS cluster in clinical practice and facilitating efficient symptom management.

Clinical nurses can use a web-based calculator developed in this study to effectively identify patients with the severe PFS cluster and provide targeted interventions.

Prediction model

Symptom cluster

Lung cancer

Machine learning

© 2025 Elsevier Inc. All rights are reserved, including those for text and data mining, AI training, and similar technologies.