Investigation of Machine Learning Models for Anxiety Levels Assessment in Patients With Breast Cancer Using Self-Questionnaires.

BACKGROUND: This study aimed to develop a machine-learning model using self-report questionnaires to screen for clinically significant anxiety symptoms in patients with breast cancer experiencing severe distress. METHODS: A cohort of 327 breast cancer clinic patients was included. Anxiety symptoms were evaluated using the State-Trait Anxiety Inventory-State (STAI-S), State-Trait Anxiety Inventory-Trait (STAI-T), Beck Anxiety Inventory (BAI), and Hospital Anxiety and Depression Scale (HADS) questionnaires. The high-risk anxiety group was determined based on the Mini International Neuropsychiatric Interview Patient Health Survey. Supervised machine learning models were analyzed and validated using MATLAB2022. RESULTS: The BAI showed an area under the curve (AUC) of 0.782 with the logistic regression classifier. The HADS showed an AUC of 0.784 with linear discriminant analysis. STAI-S and STAI-T exhibited AUCs of 0.770 and 0.791, respectively, using Support Vector Machine with a linear kernel. Models combining multiple questionnaires yielded higher AUCs: STAI-S and STAI-T combined resulted in 0.807, STAI-S and BAI in 0.794, STAI-T and BAI in 0.808, and the combination of STAI-S, STAI-T, and BAI achieved 0.810. CONCLUSION: The combination models demonstrated superior performance in detecting anxiety among patients with breast cancer compared to individual questionnaires. Combining multiple self-report scales enhances screening accuracy, indicating future research should optimize these models further through integration with other methodologies.