Predicting Depression Risk in Patients with Cancer Using Multimodal Data.

Area of Science:

• Oncology
• Psychiatry
• Data Science

Background:

• Depression is a common comorbidity in cancer patients.
• Untreated depression negatively impacts cancer care and treatment adherence.
• Early identification of depression risk is essential for timely intervention.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting depression risk in cancer patients within the first month of treatment.
• To compare the performance of models using structured data versus Natural Language Processing (NLP) on clinical notes.

Main Methods:

• Utilized machine learning, including LASSO logistic regression, on structured patient data.
• Developed a Natural Language Processing (NLP) model analyzing clinician notes.
• Compared predictive performance between the structured data model and the NLP model.

Main Results:

• The LASSO logistic regression model using structured data demonstrated good performance in predicting depression risk.
• The NLP model, relying solely on clinician notes, exhibited poor performance.
• Further validation of the structured data model is recommended.

Conclusions:

• Machine learning models utilizing structured data can effectively predict early depression risk in cancer patients.
• Early prediction models hold potential for earlier identification and treatment of depression.
• Improved depression management can enhance cancer care and treatment adherence.