This preprint introduces LLMD, a large language model (LLM) that uses medical records to characterize patient health over time, and isdeployed today in applications that improve health outcomes andpower clinical research. Along with general medical knowledge,LLMD is trained on labeled longitudinal medical records, givingit unique advantages over LLMs trained on knowledge alone, onunlabeled records, or on records from a single health system. Weshow that LLMD learns to make nuanced connections in information covering years of patient care documented across facilities,and that these are critical to real-world accuracy

An overview of structuring, abstraction and all the nuance that comes with building a personalized medical AI.

In our prior post, we introduced structuring and abstraction, two processes needed to get real-world medical records into a form that we can draw insights from. We also explained how those processes map to the latest AI techniques. In this post, we go deeper into what medical records look like in the wild and what happens when we try to apply AI techniques to them.