AI Agents Design Smarter Cognitive Health Queries

A novel approach using AI agents to design optimal queries for detecting cognitive issues in clinical notes has been developed by researchers at Mass General Brigham. These AI agents, through iterative debate, created prompts that matched or improved upon those designed by clinical specialists. This fully automated system aims to offer a more affordable and scalable method for identifying patients who might benefit from additional screening or interventions.

The technology functions by scanning clinical notes for references to cognitive concerns, thereby flagging individuals for potential follow-up. While challenges exist, such as distinguishing a noted concern from an actual biological problem, the rapid advancements in AI offer a promising solution to early detection of degenerative brain diseases. This is especially critical with an aging population and the advent of new medications that can slow disease progression.

Another initiative at the Cleveland Clinic uses predictive machine learning on electronic health records to identify patients at elevated risk for neurological conditions. This system aims to assist primary care physicians by providing transparent reasoning for patient flags, such as weight decline or reported falls. The goal is to facilitate early conversations and interventions to slow disease progression.

AI's potential to reduce screening costs is significant. The Mass General Brigham team focused on using less expensive models, enabling their system to run on a laptop and reducing the cost of screening to 2 cents per patient, a stark contrast to a human-driven workflow costing approximately $50. Despite these advances, further refinement and integration into electronic health records are necessary before widespread clinical implementation.