Patient-Centered Research Through Artificial Intelligence to Identify Priorities in Cancer Care

Area of Science:

• Oncology
• Medical Informatics
• Artificial Intelligence

Background:

• Patient-centered research is crucial for healthcare but often lacks patient perspectives.
• Bridging the gap between research and patient care requires understanding patient concerns.
• Existing health research inadequately represents patient voices.

Purpose of the Study:

• To utilize artificial intelligence (AI) and natural language processing (NLP) for analyzing patient messages.
• To identify patient concerns and generate relevant research topics from large datasets.
• To quantify the quality and relevance of AI-generated research topics.

Main Methods:

• A case series employed an automated framework with a two-stage unsupervised NLP topic model.
• Deidentified patient portal messages from breast and skin cancer patients were analyzed.
• A large language model (ChatGPT-4o) was used for topic interpretation, generation, and refinement.

Main Results:

• Over 614,000 patient messages from 25,000+ individuals were analyzed.
• AI-generated topics for breast cancer research scored 3.00 for meaningfulness and 3.29 for novelty.
• AI-generated topics for skin cancer research scored 2.67 for meaningfulness and 3.09 for novelty.
• A significant portion of AI-suggested topics demonstrated high meaningfulness and novelty.

Conclusions:

• AI/NLP analysis of patient messages can generate high-quality, patient-centered research topics.
• This methodology offers valuable guidance for future health research.
• The study highlights the potential of AI in cancer care research.