This research demonstrates how large language models can extract valuable insights from unstructured clinical notes to improve patient mortality predictions.

• Leverages GPT-4o-mini to answer clinical questions about patients based on discharge summaries
• Enhances traditional tabular data models with rich information from unstructured clinical notes
• Improves prediction accuracy for patient-level mortality in intensive care settings
• Demonstrates practical integration of AI into existing healthcare predictive frameworks

This approach matters because it bridges the gap between structured medical records and the wealth of information trapped in clinical documentation, potentially improving clinical decision support and resource allocation in healthcare settings.

Paging Dr. GPT: Extracting Information from Clinical Notes to Enhance Patient Predictions