ChatGPT Health connects to Epic records so clinicians can read charts in chat
OpenAI said healthcare organisations can now connect an electronic health record and extra industry data to ChatGPT, so clinicians can reach patient context and medical research securely. TechCrunch reports ChatGPT Health is wired to Epic's system, which covers more than 325 million patients. Clinicians at organisations with a business associate agreement can pull appointment notes, laboratory results, medications and specialist documentation into the chat; access is read-only and the model writes nothing back. OpenAI also released a public healthcare-data plug-in.
Artificial Intelligence··Midday
OpenAI says health records can now be connected to ChatGPT
OpenAI said on 1 September that healthcare organisations can connect an electronic health record and additional industry data to ChatGPT. The company says ChatGPT can now connect to trusted healthcare data, so clinicians can securely access patient context and medical research. TechCrunch reports that OpenAI connected ChatGPT Health to Epic's electronic health record system, which covers more than 325 million patients. Clinicians at organisations with a business associate agreement can pull appointment notes, laboratory results, medications and specialist documentation into ChatGPT to summarise a history or prepare for a visit.[1], [2]
The Epic link is read-only; some deployments put the chat in the workflow
TechCrunch says access is read-only and the model writes nothing back to the patient chart. In some deployments ChatGPT is reachable inside the Epic workflow for pre-visit review and for building a clinical timeline. OpenAI's announcement also says clinicians can securely access patient context; it does not name a vendor, a patient count or a write-back setting.[1], [2]
A public healthcare-data plug-in searches trial and drug questions
OpenAI also released a Healthcare Public Data plug-in that queries ClinicalTrials.gov, CMS coverage data, RxNorm, DailyMed and PubMed for trial eligibility and medication questions. OpenAI's announcement says clinicians can access medical research as well. TechCrunch reports that OpenAI's own evaluation collected 4,300 physician ratings across 27 clinical use cases and rated 99.1 per cent of responses safe. That figure leaves a small share of unsafe answers in a setting where a wrong answer can do harm.[1], [2]
