This research explores how large language models can be leveraged to detect mental health symptoms in patients with chronic conditions through their secure messages.

• Evaluated multiple LLM optimization strategies including engineered prompts, systematic personas, and temperature adjustments
• Three out of five tested LLMs demonstrated excellent performance in detecting depression and anxiety symptoms
• Identified best practices for enhancing model accuracy in clinical communication analysis

This innovation could transform healthcare delivery by enabling earlier intervention for comorbid mental health conditions, reducing treatment gaps, and improving overall patient outcomes for those with chronic diseases like diabetes.

Optimizing Large Language Models for Detecting Symptoms of Comorbid Depression or Anxiety in Chronic Diseases: Insights from Patient Messages