OpenAI brings ChatGPT Health into Epic clinical record workflows

ChatGPT Health gains access to Epic records

OpenAI is integrating ChatGPT Health with Epic’s electronic health record system, enabling clinicians to import patient information and ask questions using AI. Epic’s EHR platform holds data for more than 325 million patients. In selected systems, OpenAI said ChatGPT will also be available directly within EHR workflows.

The integration gives clinicians access to appointment notes, laboratory results, medications and specialist documentation for summarisation. OpenAI said users can review patient history, identify changes and prepare for forthcoming appointments. In some deployments, they can generate pre-visit reviews and build clinical timelines without leaving a patient chart.

The records connection is limited to read-only access. ChatGPT does not write information back to the health record, OpenAI said. That constraint defines the announced role as a tool for reviewing and organising existing material rather than altering clinical documentation.

Public-data tools and compliant workspace access

OpenAI is also launching a Healthcare Public Data plugin. It can retrieve material from ClinicalTrials.gov, CMS Coverage, RxNorm, DailyMed and PubMed. The company said healthcare workers can use the plugin to synthesise information related to trial eligibility criteria, medication identifiers, coverage-policy versions and provider records.

Organizations with a Business Associate Agreement can use ChatGPT Work, Codex, apps and connectors in their workspace for compliant workflows. The announcement expands on ChatGPT Health access for U.S. adults by moving from consumer access toward tools designed around clinician record-review tasks.

Safety remains a stated limitation

OpenAI said ChatGPT Health was rolled out to all U.S. consumers last month and that people ask ChatGPT 300 million health-related questions each week. The company has nevertheless maintained that AI is not appropriate for diagnosis or treatment.

For its clinical-use evaluation, OpenAI collected more than 4,300 physician responses across 27 use cases, including pre-visit review, clinical timelines, medication review and handoff summaries. It reported that 99.1% of the responses were safe. The source also notes lawsuits alleging harmful recommendations, underlining that a small number of unsafe answers can still have serious consequences.

For healthcare organizations, the practical implication is to evaluate where read-only retrieval, clear workflow boundaries and clinician review can support record preparation without treating AI output as diagnostic or treatment guidance.