Listen instead
An individual reviewing a laboratory report on the athenahealth portal can now bypass the typical search engine queries for unfamiliar clinical terms. By configuring a connector within their personal Anthropic Claude account, the patient can grant the artificial intelligence assistant permission to read and interpret their electronic health records directly.
This capability became functional following the release of the athenaPatient MCP server. Developed on the Model Context Protocol, which is an open-source standard designed by Anthropic, the system links the athenaOne platform to consumer-facing artificial intelligence tools. Currently, US users with paid Claude subscriptions can activate this connector within their application settings. Once authenticated using their standard patient portal credentials, users can query their records using conversational prompts. This development marks a shift where patient data integration moves directly into consumer-managed applications rather than remaining confined to enterprise clinical systems.
Technical and Security Architecture
The implementation relies on granular permission controls managed by the patient. Users retain the ability to select which categories of clinical information are shared with the external assistant. The interface allows patients to toggle access for ten specific categories of data: laboratory results, current medications, vital signs, diagnosed health conditions, diagnostic reports, immunization records, clinical visit history, documented allergies, past medical procedures, and active care plans.
To address security requirements, the architecture prevents patient portal credentials from being shared with Anthropic. The connection uses encrypted transit protocols to protect data during transmission. Furthermore, patients can revoke the permission at any time through their portal settings, immediately severing the link between the assistant and the medical record.
The selection of partners for this framework is subject to specific corporate guidelines. Paul Brient, the chief product and operations officer at athenahealth, stated in an interview with Fierce Healthcare that the company restricts access to artificial intelligence assistants that do not use patient information to train their underlying models. While the Anthropic Claude integration is the first deployment utilizing the Model Context Protocol server, the developer plans to support additional external assistants. According to Brient, the company does not intend to mandate which specific assistant a patient must use, but aims to provide secure pathways to the tools patients choose to adopt.
Industry Context and Alternative Approaches
This release occurs alongside several distinct strategies for integrating artificial intelligence into medical workflows and patient communication. Major technology companies and electronic health record developers are pursuing different models of data access.
For instance, Epic Systems has integrated artificial intelligence capabilities via cloud partnerships to assist clinicians with message drafting and clinical documentation, focusing primarily on provider-facing tools. In contrast, Oracle has introduced generative features within its patient portal to let users query their records, keeping the artificial intelligence interface native to the healthcare vendor's ecosystem.
The adoption of these technologies matches changes in patient behavior. A survey published by the Pew Research Center reported that 34 percent of adults in the United States have used artificial intelligence tools for health-related tasks. As this consumer interest increases, healthcare software developers are responding by establishing connections to the consumer tools that patients already use.
Clinical and Operational Stakes
The strategy of building a connector to an external assistant rather than developing a proprietary chatbot represents a distinct operational choice. By positioning the electronic health record as a secure data utility for third-party software, athenahealth avoids the development and maintenance costs of consumer-facing artificial intelligence models.
However, this model introduces new dynamics into the patient-provider relationship. When a patient asks an external assistant to interpret medical data, the practicing physician has no visibility into the explanation provided. This lack of oversight could lead to discrepancies between the assistant's interpretation and the clinical intent of the care team. Furthermore, clinics may face an increase in portal messages generated by these automated interpretations, potentially increasing the administrative workload for primary care teams. Because the conversation occurs entirely within a third-party application, the medical practice does not capture a record of the interaction, creating potential gaps in the documentation of the patient's care journey.
By adopting the Model Context Protocol, athenahealth is helping to establish an open standard for how software applications connect with patient portals. The broader industry must now determine how to manage the clinical boundaries, liability, and documentation standards for these patient-led data transfers.
Source: The HealthTech Signal
