ChatGPT now pulls Epic EHR data and public health sources

OpenAI announced that ChatGPT for Healthcare now supports a direct connection to Epic electronic health records and a new Healthcare Public Data plugin. The Epic integration pulls authorized patient context, such as appointment notes, lab results, medication lists, and specialist reports, into the AI model, while the plugin provides structured access to nine official public datasets, including PubMed, DailyMed, CMS Coverage, ClinicalTrials.gov, and RxNorm.

The release builds on ChatGPT for Healthcare’s existing clinical search capabilities. Instead of searching multiple systems, clinicians can ask the model to retrieve and summarize relevant patient information, then receive citations that point back to the original chart entries. The public data plugin lets teams pull specific records, fields, identifiers, and versioned information from authoritative sources, enabling tasks like verifying trial eligibility, checking medication labels, or reviewing Medicare coverage policies without leaving the AI interface.

Healthcare organizations face a fragmented data landscape. Patient context, research evidence, and organizational knowledge often reside in separate silos, and any new tool must meet strict governance and compliance requirements. OpenAI addresses this by offering a governed workspace that includes role‑based access, single sign‑on, audit logs, and a Business Associate Agreement, allowing the same environment to support clinical, research, and administrative workflows while remaining HIPAA‑compliant.

Compared to earlier releases, this integration is the first to embed real‑time EHR data directly into ChatGPT. Prior versions relied on external search engines and manual data entry, which limited speed and accuracy. By contrast, the Epic connector streams live, authorized records, and the public data plugin consolidates multiple trusted sources into a single, structured API, reducing the need for clinicians to switch between applications.

OpenAI has conducted extensive physician evaluations to gauge safety and accuracy. In a study of 27 clinical use cases, including pre‑visit review, medication reconciliation, and handoff summaries, physicians rated 99.1% of responses as safe. A separate evaluation of nuanced clinical questions across five connected data sources found that more than 93% of responses achieved a “good” or higher accuracy rating. These results suggest that the system can reliably surface relevant information and cite supporting evidence, though ongoing oversight and continuous training remain essential.

Organizations interested in adopting the new capabilities can start with the features that best match their needs and expand over time. The same governed workspace can also host other enterprise plugins, such as Microsoft SharePoint, Google Drive, Salesforce, and Slack, while preserving existing permissions. Technical teams can use Codex to build custom applications that leverage the integrated data, and business teams can generate reports, analyses, and presentations directly from ChatGPT Work.

The announcement signals a broader trend toward tighter integration of AI with clinical information systems. By unifying patient records, research literature, and public health data, OpenAI’s solution promises to streamline clinical decision support, accelerate research workflows, and improve operational efficiency. Future updates may include additional EHR vendors, expanded public datasets, and deeper compliance features, but the current release already offers a powerful, secure, and compliant platform for healthcare teams.