Predicting Depression, Anxiety, and Their Comorbidity among Patients with Breast Cancer in China Using Machine...

Area of Science:

• Psychology and Oncology
• Machine Learning in Healthcare

Background:

• High prevalence of depression and anxiety in breast cancer patients.
• Need for predictive models of psychological distress in this population.

Purpose of the Study:

• To evaluate the predictive capacity of personal resources for depression, anxiety, and comorbid depression and anxiety (CDA) in breast cancer patients.
• To compare the predictive performance of personal resource models with models incorporating demographics and COVID-19 impacts.

Main Methods:

• Cross-sectional survey of 707 breast cancer patients in China.
• Lasso logistic regression for initial personal resource models.
• Six machine learning methods with tenfold cross-validation for combined models (personal resources, demographics, COVID-19 impacts).

Main Results:

• Prevalence rates: 21.9% depression, 35.1% anxiety, 14.7% CDA.
• Key predictors identified: loneliness, vitality, mental health, bodily pain, self-control.
• Personal resource models significantly outperformed demographic and COVID-19 impact models (AUC 0.826-0.869 vs. 0.505-0.629).
• Support vector machine model showed best prediction (AUC 0.832-0.873) when combining all factors.

Conclusions:

• Personal resources are strong predictors of depression, anxiety, and CDA in breast cancer patients.
• Machine learning effectively utilizes personal resource features for prediction.
• Interventions targeting loneliness, bodily pain, vitality, mental health, and self-control are recommended.