Problem: Over the last year, the average general surgery (ED only) time to complete a consult, for those discharged, was 215 minutes. The goal length-of-stay for the entire emergency department visit is 330 minutes, which would require a surgical consult time less than 60 minutes. The 330-minute goal is a priority for Duke Health, as emergency department flow has large impact on resource utilization and patient outcomes. This project aims to reduce the time to surgical consult completion.
Solution: We aim to create an artificial intelligence (AI) augmented workflow with large language models (LLMs) and natural language processing (NLP) to improve the flow of information between the emergency department and general/vascular surgery consultants. Our model generate suggested pages, acuity levels, and notes with evidence-based medicine.
Impact: This project decrease time to surgical consultation, reduce emergency department cognitive burden for routine diagnoses, and improve patient experience with shorter emergency department wait times and shorter time to surgical evaluation. This also has implications on hospital resource utilization, especially bed flow. Furthermore, this is scalable, both internally to other departments and externally.
