OpenAI's ChatGPT Health now links to Epic EHR for clinician data access
OpenAI has integrated ChatGPT Health with Epic's electronic health record system, giving clinicians read-only access to patient data for AI-assisted queries. The integration also includes a new Healthcare Public Data plugin for external sources like ClinicalTrials.gov and PubMed. OpenAI notes that while the AI can summarize and analyze, it does not write back to records and is not intended for diagnosis.
The integration grants clinicians read-only access to patient data within Epic's system, which serves over 325 million patients. A new Healthcare Public Data plugin pulls from sources like ClinicalTrials.gov, PubMed, and RxNorm, helping workers synthesize trial eligibility criteria and medication identifiers. OpenAI reports that 99.1% of physician responses across 27 clinical use cases were deemed safe, though the company maintains the AI is not intended for diagnosis.
The rollout follows recent legal challenges, including a lawsuit from a Florida pastor alleging ChatGPT provided a near-fatal recommendation, and a May suit from family members over dosage advice. OpenAI also noted that organizations with Business Associate Agreements can use ChatGPT Work, Codex, and connectors for compliant workflows, while consumers already submit roughly 300 million health-related queries weekly.
This integration could significantly streamline clinical workflows, allowing physicians to quickly summarize patient histories and prepare for appointments. However, even a small percentage of unsafe AI responses poses serious risks in healthcare, where errors can be life-threatening. While read-only access limits direct harm, clinicians may become over-reliant on AI summaries, potentially overlooking errors. Patients could benefit from more efficient care, but recent lawsuits highlight ongoing concerns about AI reliability in medical contexts.