Related Experiment Video
Updated: Jun 19, 2026
04:13
Using a Real-Time Locating System to Measure Walking Activity Associated with Wandering Behaviors Among Institutionalized Older Adults
Published on: February 8, 2019
6.8K
Improving postsurgical fall detection for older Americans using LLM-driven analysis of clinical narratives
Malvika Pillai1,2, Terri L Blumke3, Joachim Studnia4
1Veterans Affairs Palo Alto Health Care System, Palo Alto, California, USA.
Medrxiv : the Preprint Server for Health Sciences
|July 9, 2024
Summary
Large language models (LLMs) can automatically detect postsurgical falls from clinical notes, improving patient care and reducing costs. This technology shows promise for future fall prediction and prevention strategies.
Area of Science:
- Medical Informatics
- Artificial Intelligence in Healthcare
- Patient Safety
Background:
- Postsurgical falls present significant challenges in patient care and healthcare economics.
- Current methods for identifying and tracking postsurgical falls are often difficult and inefficient.
Purpose of the Study:
- To evaluate the effectiveness of large language models (LLMs) for automated detection of postsurgical falls.
- To compare different LLM prompting strategies across diverse healthcare systems.
Main Methods:
- Tested multiple LLM prompting approaches using three open-source LLMs.
- Evaluated performance in two distinct healthcare systems: Stanford Health Care and the Veterans Health Administration.
Main Results:
- The Mixtral-8×7B model, using a zero-shot approach, achieved the highest performance.
- Excellent results were observed at Stanford Health Care (PPV = 0.81, recall = 0.67) and the VA (PPV = 0.93, recall = 0.94).
Conclusions:
- LLMs can reliably detect postsurgical falls with minimal or no specific guidance.
- This research establishes a foundation for using LLMs in fall prediction and prevention across various healthcare settings.
