Connecting Electronic Health Records to AI Systems
Healthcare providers now have a bridge between fragmented patient data and generative AI tools. OpenAI has introduced an integration for electronic health record (EHR) systems that allows clinicians to pull authorized patient information directly into ChatGPT. This update aims to reduce the time staff spend searching across disparate platforms. By connecting systems like Epic, the software provides summaries of laboratory results, medication lists, and specialist documentation within a single interface.
The system operates through two distinct deployment methods. Clinicians can pull EHR context into the main ChatGPT interface to prepare for upcoming appointments by identifying recent changes in a patient's status. Alternatively, the technology can reside directly inside the EHR workflow. This placement allows staff to receive AI assistance without leaving the patient chart. Such integration helps address the time-intensive nature of manual chart reviews while maintaining the security protocols required for protected health information.
Accessing Official Medical Datasets through Plugins
Beyond internal patient records, medical teams require access to external clinical and public health data. A new Healthcare Public Data plugin centralizes information from nine government and professional sources. These include PubMed, DailyMed, RxNorm, and ClinicalTrials.gov. The plugin enables users to cross-reference medication identifiers, verify coverage policies, and assess trial eligibility criteria without navigating multiple websites.
The utility of this tool spans various departments within a hospital. Pharmacy teams can confirm current label warnings, while research teams track actively recruiting trials. Population health planners can aggregate research, coverage guidelines, and provider data to inform program designs. By focusing on authoritative sources, the tool provides a verifiable record that supports decision-making in high-stakes environments.
Rigorous Validation and Clinical Safety Standards
Building tools for the medical industry requires a high threshold for accuracy. OpenAI collaborated with hundreds of physicians across 26 specialties to refine the output quality. These partners have reviewed over 700,000 model responses to ensure the information aligns with real-world medical practice. This iterative feedback process specifically targets the reliability of clinical answers.
The results of recent performance testing indicate high levels of reliability. Physicians evaluated ChatGPT responses across 27 common clinical tasks, such as creating handoff summaries and reviewing medication lists. Out of 4,363 independent ratings, 99.1% of responses were marked as safe. In a separate assessment focusing on clinical accuracy using U.S.-based datasets, more than 93% of answers received a rating of good or better. These benchmarks confirm that the model can interpret complex healthcare information when given access to specific, context-rich data sources.
Governance and Future Clinical Adoption
Security remains a primary concern for any organization integrating AI into patient care. The platform offers a governed workspace that includes role-based access controls, audit logs, and single sign-on features. When combined with a Business Associate Agreement, these tools allow hospitals to maintain HIPAA-compliant workflows while using AI for operational and clinical tasks. Organizations can also connect business systems like Salesforce or Slack to streamline administrative reporting alongside their medical data.
Initial rollout includes major health systems such as Cedars-Sinai, UCSF, and Memorial Sloan Kettering Cancer Center. These partners represent a broad spectrum of care delivery models. The focus for administrators is now on determining which clinical use cases benefit most from this connectivity. As healthcare organizations grow their comfort with these tools, they can expand the information available within the workspace. The shift marks a move toward integrating generative models into the core digital infrastructure that already drives hospital operations.