New HIPAA rule strengthens patients’ access through personal health apps
Kristen Valdes and Lisa Bari welcomed the finalized privacy changes while pressing for simpler identity verification and usable records for AI.
A newly finalized HIPAA rule explicitly recognizes personal health apps as a way people can obtain their electronic health information, a change two health-data executives said could help patients get records into the tools they choose. Speaking Wednesday, b.well Connected Health founder and CEO Kristen Valdes called that kind of request a “first party request” and welcomed the change as a step toward reducing the work patients do to assemble their records.
HHS announced the rule November 16. The final text provides for an effective date 60 days after Federal Register publication and a general compliance period extending another 180 days. The changes therefore have an implementation period, with specific treatment of requests through personal health apps becoming part of the work covered organizations must prepare to support.
Valdes described navigating more than 40 patient portals for her daughter, who has a rare disease and sees specialists across several states. “We have to stop competing on the data itself,” she said. Her proposed starting point was “one universal modern identity,” allowing a patient’s verified request to work across the organizations holding their information.
Valdes spoke with Lisa Bari, Innovaccer’s market president for public policy and public sector, during Can Patient Data Unleash AI? at HLTH USA 2026 in Las Vegas. STAT senior writer and editorial director of events Matthew Herper moderated the conversation. The discussion followed an EHR panel and shifted the emphasis from developers’ access to what patients and caregivers can actually accomplish.
A complete record is still work for the patient
Valdes said a patient seeking information across multiple organizations can still encounter separate usernames, passwords and portal accounts, even when the organizations use the same underlying software. She described the danger of incomplete information through an incident in her daughter’s care, attributing it to systems that had not shared a relevant contraindication. Her account illustrated the consequences she sees in fragmented records, without supplying a clinical case record for independent review.
Herper asked how many of the portals involved the same software operating at different institutions. Valdes said many did, arguing that a familiar interface can still leave someone moving among separate accounts. She said the person also needs a usable way to contact clinicians across those organizations, reducing the work of maintaining multiple message threads.
Valdes argued that incentives and decisions about access are now major barriers. She pointed to identity verification and record-location services as capabilities that some organizations already use. The existence of those capabilities at one organization, however, leaves a separate question about whether a patient can use them to retrieve information from every place they receive care.
The final rule establishes that a person’s own access right includes receiving electronic information through a personal health app when a covered organization can produce it in that form and format. It also distinguishes apps used primarily by or for the individual from products acting for a healthcare organization. That distinction matters because the same technology company can offer different kinds of services, and the nature of the request affects how it should be handled.
Who the software was built to serve
Bari challenged the use of patient-access language by companies whose products were built primarily around transactions between healthcare organizations. She said “the tools are for different things,” pointing to the difference between an institutional system of record and an application designed around a patient’s task. She described her own employer as a third-party intermediary that aggregates information from multiple sources for applications, including AI.
“We’re trying to retrofit a system that was built for billing,” Bari said. Her argument was that the record should be available as a foundation on which different tools can operate. A clinician completing a prior authorization and a patient trying to understand their history may need different interfaces and workflows, even when both depend on some of the same information.
Both speakers work for businesses with a commercial interest in making data available across healthcare organizations. In its announcement of an OpenAI relationship, b.well said its infrastructure would let users authorize connections to their medical records for ChatGPT Health. Bari described Innovaccer’s aggregation role onstage, making clear that the argument involves companies building services around access as well as people seeking records for their own care.
The discussion did not establish that a single new identity layer would make every record immediately available. Organizations still have to match a request to the correct person, identify the records they hold and determine the authority under which they may release them. The speakers’ proposal is to make those processes usable across applications and organizations, with less repeated work for the person requesting the information.
Caregivers expose weaknesses in the account model
Valdes described using her daughter’s portal credentials while acting as a caregiver. Bari offered a similar example involving a family member’s accounts. They used those examples to argue that a login can obscure the identity of the person actually doing the work, even when the patient has chosen to involve them.
That creates a design problem for an AI service as well as for a portal. An application receiving instructions from a caregiver needs to know who is requesting the action and what they are authorized to do. A verified identity can help establish who someone is, while consent and representative authority govern what information and actions should be available to that person.
Valdes argued for patients choosing applications based on the value those products offer in their lives. She said that choice should extend to tools that act for the person, helping with tasks such as finding care, consolidating information and identifying clinical trials. She presented those uses as a direction for the market, giving no new outcome study showing that a single application had delivered all those benefits.
Scheduling and imaging show the next practical tests
The conversation also returned to scheduling, a point disputed during the preceding EHR discussion. Valdes said an open approach already exists through HL7 and urged organizations to examine it. HL7’s SMART Scheduling Links product brief describes a framework for sharing appointment availability across EHRs or booking portals and connecting people to booking.
Publishing appointment availability is one part of the work required to complete a booking. A patient may also need to find a clinician who accepts their coverage, meet referral requirements and select an appropriate visit. The test of an integration is how much of that sequence it can support, along with how clearly it tells the patient what remains to be done.
Valdes also welcomed efforts to move diagnostic images without physical disks, while pressing for access through outside applications. Her argument was that a health system’s ability to retrieve an image should also translate into useful access for the patient’s chosen tools. The panel supplied no technical evaluation of a specific imaging interface or finding that a particular company had violated an access rule.
For health AI, the implication is that better reasoning cannot compensate for a record it never receives. The ability to obtain relevant information, understand where it came from and act with the patient’s authorization is part of the service itself. Valdes and Bari’s challenge was to judge that service by the work it removes for the patient, including the work required before an AI conversation can begin.