A team of researchers from China found that a predictive machine-learning model assisted healthcare professionals in identifying patients with lung cancer who are experiencing severe pain-fatigue-sleep disturbance (PFS) symptom clusters after undergoing chemotherapy. The study authors published their findings in Seminars in Oncology Nursing.
The investigators highlighted that PFS symptom clusters are “the most common symptom cluster in patients with lung cancer following chemotherapy, which significantly impacts their quality of life.” The goal of their research was to “develop and validate a machine learning-based prediction model for the severe PFS cluster” among patients in this population.
Investigators recruited 612 patients for the study, and the criteria for participation included individuals aged 18 years or older, those “pathologically diagnosed with primary lung cancer, either with or without solitary metastases,” and having underwent at least one cycle of chemotherapy.
The study team used four machine learning algorithms and logistic regression to develop the predictive model, which they evaluated through area under the curve (AUC), accuracy, sensitivity, specificity, and Brier score measurements. In addition, “a web-based application was developed to facilitate the practical implementation of the best model in clinical settings.”
According to the results, the predictive model was “identified as optimal, exhibiting the best discrimination and calibration in the test set (AUC, 0.765; Brier score, 0.159) and excellent performance in the validation set (AUC, 0.914; Brier score, 0.124).”
Furthermore, factors associated with the development of PFS symptom clusters identified by the predictive model included stress, C-reactive protein, depression, BMI, anxiety, neutrophils, age, gender, pathological classification, and Eastern Cooperative Oncology Group performance status. The findings also revealed that “a nonlinear relationship existed between stress, BMI, age, and the severe PFS cluster.”
In reflecting on the implications for nursing practice, particularly in the field of thoracic oncology, the investigators highlighted that “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.”