Health Systems Deploy LLMs to Query Patient Records — SkimNews
Get the Health newsletter
Daily health & science — research, biotech, public health, the studies worth knowing. Free.
- Stanford deployed ChatEHR, an LLM-powered tool that lets clinicians query and summarize patient electronic health records, and a physician on a tough case used it to crack a cancer diagnosis that six pathologists couldn't solve after staining the lymph node biopsy cells 70 times.
- ChatEHR uncovered a prior diagnosis of sarcomatoid squamous cell carcinoma from a different health system that "completely explained the findings in the lymph node," according to the doctor's feedback to the tool.
- Health systems are moving toward broad implementation of chatbots for EHRs, building both homegrown and vendor-built tools, because modern electronic health records have become so bloated that clinicians struggle to find the information they need to deliver care.
- Despite the dramatic diagnostic-mystery anecdotes, solving those cases is "the least of their selling points" — the real value, per the article, is routine searching and synthesis of patient records.
- The underlying pain point is that EHRs have grown so unwieldy that finding the right note often takes longer than the clinical reasoning it supports, which is what these chatbot tools are designed to compress.
Why it matters: Clinicians currently lose time to bloated EHRs that bury the right note under pages of boilerplate, and a chatbot that can surface the relevant history in seconds changes the economics of every patient encounter. If ChatEHR-style tools reliably pull buried diagnoses across health systems, the gain extends from routine throughput to the rare cases — like the lymph node biopsy — that currently take a half-dozen specialists to crack.
Ask SkimNews
