Blue Button 2.0 let Medicare screen the apps; the plans and providers that followed could only warn patients

On March 6, 2018, CMS opened Medicare claims data to apps chosen by beneficiaries. As the data holder, CMS set terms for those apps. The plans and providers later told to open similar doors could refuse an app only on narrow grounds.

Looking Back revisits past weeks one at a time, up to the most recent, and asks what it meant for the banks, health systems and public agencies we serve. The series was written in 2026, with the benefit of hindsight, and each article says which week it looks back on.

Two-minute summaryThe dek and the first sentence of each section, quoted from the article

On March 6, 2018, CMS opened Medicare claims data to apps chosen by beneficiaries. As the data holder, CMS set terms for those apps. The plans and providers later told to open similar doors could refuse an app only on narrow grounds.

  1. The week On Tuesday, March 6, 2018, the Centers for Medicare & Medicaid Services announced a federal initiative called MyHealthEData at HIMSS18, the annual conference of the Healthcare Information and Management Systems Society.
  2. Admission: who decided which apps could connect The first moment is the decision to let an app connect.
  3. Consent: what the patient saw at the moment of authorization The second moment is the screen where the patient says yes.
  4. Custody: who answered for the data after it left The third moment begins once the app holds the data.
  5. Exit: how the patient stopped the flow The fourth moment is the end of the relationship.
  6. How the four moments differed by type of organization The four moments apply to every data holder, but the obligations differ.
  7. The strongest objections One objection is that the narrow refusal grounds are deliberate and good.
  8. What a health leader could check today The four moments translate into a short set of checks.
  9. With hindsight In our reading, Blue Button 2.0’s model held.

The week

On Tuesday, March 6, 2018, the Centers for Medicare & Medicaid Services announced a federal initiative called MyHealthEData at HIMSS18, the annual conference of the Healthcare Information and Management Systems Society. The agency published a summary the same day. It said the initiative aimed to let patients control their health care data and decide how it would be used, “all while keeping that information safe and secure.”

The summary described MyHealthEData as government-wide. It said the White House Office of American Innovation led it, with participation from the Department of Health and Human Services and the Department of Veterans Affairs. Within HHS, it named the Medicare and Medicaid agency itself, the Office of the National Coordinator for Health Information Technology and the National Institutes of Health. The stated aim was to break down barriers that kept patients from electronic access to their records “from the device or application of their choice.”

The concrete launch that day was Medicare’s Blue Button 2.0. The agency explained the history. Medicare first launched Blue Button in 2010, giving patients their claims data as a downloadable PDF file. Blue Button 2.0 would instead provide claims data “in a universal and secure digital format.” Beneficiaries could use it in applications designed to help them manage their health, or share it with their doctors.

The announcement gave the scale. It said Blue Button 2.0 contained four years of Medicare Part A, B and D data for 53 million Medicare beneficiaries. It listed prescriptions and primary care treatments among the types of information available. Part A is Medicare’s hospital insurance, Part B covers physician and outpatient services, and Part D covers prescription drugs.

The agency also described who would build on it. It said more than 100 organizations had joined a Blue Button 2.0 developer preview program, and that it expected more to sign on as the service launched to beneficiaries. The developer preview let application developers build and test apps against fake, or synthetic, claims data. The fact sheet expected more organizations to sign on as the service launched to beneficiaries, which kept building and testing apart from live use.

The announcement reached past traditional Medicare. One heading read “Calling on Private Plans to Provide Patients Their Data.” Under it, the announcement said the agency would re-examine its expectations for Medicare Advantage plans and for qualified health plans offered through the federally facilitated exchanges. It also called on all health insurers to release their data. It said it believed the private plans that contract through Medicare Advantage and the exchanges should provide the same benefit as Blue Button 2.0.

Providers were addressed as well. The agency said it intended to overhaul the Medicare and Medicaid EHR Incentive Programs, then commonly called Meaningful Use, and the Quality Payment Program for clinicians. The stated aim was more focus on interoperability and less time and cost to comply. It said it would prioritize quality measures and improvement activities that led to interoperability.

Two further items set expectations for hospitals and practices. The agency noted that, for some of its programs, it had finalized a requirement to use 2015 Edition certified EHR technology beginning in 2019. That edition, the summary said, included technical requirements for sharing data with patients and their care teams through application programming interfaces. It also said it was taking steps against information blocking, as required by law, by requiring hospitals and clinicians in some programs to show they had not engaged in it.

The summary closed with operational aims. The agency intended to specify what information hospitals must share with a patient’s receiving facility or post-acute care provider at discharge. It said it was considering changes to documentation guidelines for evaluation and management codes. It was also studying the extent of duplicate testing caused by providers not sharing data.

What the week did not contain matters for what follows. The announcement set out a direction and a working service for traditional Medicare claims. It did not yet impose an API requirement on private plans or providers, and it did not say how a plan or hospital should treat an app it had not chosen. Those questions were answered later, by rule and by guidance.

Interoperability announcements are easy to read as technology news. We read Blue Button 2.0 as a customer experience decision with a governance consequence. CMS did more than publish an API. It made a beneficiary’s choice of app the event that moved data out of a federal system. Every later rule that copied the model inherited that design.

The design created an asymmetry that the March 2018 materials did not dwell on. CMS ran Blue Button 2.0 as the holder of the data and the operator of the service. As operator, it could set terms for the apps that connected. The health plans and providers later required to offer similar access were in a different position. Their duties ran to the patient’s right to direct the data. Their room to judge the app was narrow.

Our argument is that the patient’s experience of sharing data was decided at four distinct moments, and that a plan or provider controlled less at each one than CMS did in 2018. The four moments do not overlap, and together they cover the trip a record takes from the organization’s system to the app and beyond:

  • Admission: which apps may connect at all, and on what grounds an organization may say no.
  • Consent: what the patient sees and agrees to at the moment of authorization.
  • Custody: who answers for the data once the app holds it.
  • Exit: how the patient stops the flow, and what happens to what the app already received.

After the four moments, we look at how the picture differs across types of health organization, at the strongest objections, and at what a health leader could check today.

Admission: who decided which apps could connect

The first moment is the decision to let an app connect. In 2018, CMS made that decision for its own service. The developer preview, as the fact sheet described it, let developers build against synthetic data. Production access, as CMS’s Blue Button page describes it today, is a separate step that CMS controls.

What the data holder could require

CMS’s Blue Button production access page, as it reads today, shows what an operator in that position can ask for. An organization applying for production access must submit its privacy policy and terms of service. The page says these must be publicly available and clearly accessible to the user. It also says the application should let the enrollee actively opt in to them, “instead of defaulting to agreement.”

The page goes further. Applicants attest that their privacy policy covers the requirements in the Blue Button API Terms of Service. They schedule a one-hour demonstration with the Blue Button team. The demonstration covers account creation, the authorization to share Medicare data, how the app displays the data, how it uses the data, and any sharing with others. Production credentials are handed over after concerns with the application, privacy policy and terms of service have been met.

We do not claim that every element of today’s process existed on March 6, 2018. The point is structural. An organization that offers an API as a service it operates can condition access on terms it sets. CMS did so for Blue Button.

What a plan could require once the rules arrived

The plans that CMS called on in 2018 later received a rule. The Interoperability and Patient Access final rule, published May 1, 2020, applied to Medicare Advantage organizations, Medicaid and CHIP programs and their managed care plans, and qualified health plan issuers on the federally facilitated exchanges. For Medicare Advantage, the requirement is codified at 42 CFR 422.119.

The text of that section frames admission differently from Blue Button. A Medicare Advantage organization must maintain a standards-based API that permits third-party applications to retrieve data “with the approval and at the direction of” a current enrollee or the enrollee’s personal representative. The enrollee’s direction is the trigger. The plan does not choose the app.

The section then limits refusal. A plan may deny or discontinue an app’s connection only on a security finding made consistent with its HIPAA risk analysis. The finding is that the connection would present an unacceptable level of risk to protected health information on the plan’s systems. The determination must use objective, verifiable criteria applied fairly and consistently across all apps and developers.

Security review as the only gate

The 422.119 text does say something about how a plan may judge risk. Its criteria may include, but are not limited to, criteria that rely on automated monitoring and risk mitigation tools. We read that as an invitation to judge behavior at the connection rather than the app’s intentions. Examples a plan might consider include how an app handles its credentials, whether it registers correctly with the authorization server and whether its request patterns look abusive.

Those are illustrations, not requirements, and each needs to be applied the same way to every app. The value of framing them in advance is evidential. A plan that discontinues an app after an incident can point to criteria it published and applied before the incident. A plan that improvises its reasons after the fact invites the charge that security was a pretext for blocking a competitor.

The section also requires plans to make technical documentation public. Paragraph (d), as codified, lists the API syntax, function names, required and optional parameters, return variables and the technical requirements for an app to register with the authorization server. A developer who reads that documentation can predict whether its app will be admitted. Predictability is part of the admission experience for the patient as well, because the patient is the one who sees a failed connection.

The difference in one sentence

Our reading of the two texts is this. CMS as Blue Button operator could ask whether an app would treat the beneficiary well. A plan under 422.119 may ask only whether the app endangers the plan’s own systems. The first question looks outward, at the patient’s interests after the transfer. The second looks inward, at the organization’s security posture.

Providers and the right of access

Providers sit under a different body of law for this purpose. The HHS Office for Civil Rights answered the question directly in its FAQs on the access right, health apps and APIs. It asked whether a covered entity could refuse to disclose ePHI to an app chosen by an individual because of concerns about how the app would use or disclose it. The answer given was no.

OCR explained that the HIPAA Privacy Rule generally prohibits a covered entity from refusing to send ePHI to an app the individual designates. The condition is that the ePHI is readily producible in the form and format the app uses. It cited 45 CFR 164.524. It added that the HIPAA Rules do not restrict how an individual or the individual’s designee, such as an app, may use information disclosed under the access right.

What admission control was left

For a health system, then, admission control narrowed to a few things. It could secure its own API endpoints. It could register apps for technical purposes. It could apply objective security criteria where a rule allowed them. It could not run a values test on the app’s business model. We treat that as the first moment where the 2018 model, copied outward, left organizations with less than CMS had.

Operational detail

The practical work at this moment belongs to the technology and security functions. Someone owns the app registration process. Someone writes the objective criteria a plan may use, and documents that they are applied consistently. Someone keeps the record of each denial or discontinuation, with the security reason tied to the risk analysis. An examiner or auditor reviewing a denial would look for that chain.

The trade-off is real. Criteria strict enough to stop a careless app may also stop a legitimate one, and the rule requires consistency across all apps. Criteria loose enough to admit everyone shift the whole burden of protection to the later moments. A plan cannot escape the choice; it can only make it visible and defensible.

The second moment is the screen where the patient says yes. In Blue Button’s case, the beneficiary authorizes an app to retrieve Medicare data. That moment is short. It is also where the patient’s understanding is either formed or skipped.

What CMS asked apps to show

CMS’s production access guidance names the problem directly. It says many users will click through and authorize access without reading privacy policies and terms completely. For that reason, it asks apps to describe in their own interface how user data will be collected, used and shared. It suggests short contextual messages before data is shared, giving the enrollee the ability to choose to proceed or not.

The guidance states a design principle: “A Medicare enrollee should never be surprised to learn how their data is being used.” It also asks apps to give users ways to opt into or revoke service and to request secure and complete deletion of their data. It recommends that apps consider a plain-language privacy notice, and points to ONC’s Model Privacy Notice template.

The privacy policy checklist on the same page lists what the policy should address. The items include whether and how data is shared, with whom, and whether once or continuously. They include any use of de-identified or pseudonymized data. They include what happens to data when a user revokes access, policies for dormant accounts, notice of policy changes, breach notification and the effect of a sale of the company.

What a plan had to say instead

A plan subject to 422.119 does not control the app’s consent screen. The rule gives it a different tool: education. Under paragraph (g), a Medicare Advantage organization must provide educational resources in non-technical, simple and easy-to-understand language. They must be in an easily accessible location on its public website and through its other usual channels to enrollees seeking their information.

The required content is specific. The resources must explain steps an individual may consider to protect the privacy and security of their information. That includes factors to consider in selecting an app, including secondary uses of data. They must give an overview of which organizations are and are not likely to be HIPAA covered entities. They must describe the oversight roles of OCR and the Federal Trade Commission, and how to complain to each.

Who speaks for the patient

The codified text names two people who may direct an app: the current enrollee and the enrollee’s personal representative. That second category matters for consent design. A caregiver acting as personal representative sees the authorization screen in place of the patient. The educational material has to work for that reader too.

Personal representative status is a legal question that the plan or provider decides, not the app. The consent moment therefore depends on an earlier record: whether the organization has documented who may act for the member. Where that record is missing or out of date, the consent screen can be accurate and still reach the wrong person.

Why education is a weaker lever than design

The education duty and Blue Button’s guidance address the same moment from opposite sides. CMS could ask the app to explain itself on the screen where the decision happens. A plan can only publish material elsewhere and hope the enrollee reads it before choosing. Our view is that design at the point of decision does more for comprehension than a separate web page. We offer it as a judgment, not a measured finding.

For providers, the access right FAQs offer a similar, thinner tool. OCR said that where an individual asks for ePHI to be sent to an app in an unsecure manner, the covered entity is not responsible for unauthorized access in transmission. It said the covered entity may want to consider informing the individual of the potential risks the first time the request is made. That is a permission to warn, not a duty to vet.

Later rules extended consent design beyond patient apps, and they chose different defaults. Under 42 CFR 422.121, beginning January 1, 2027, a Medicare Advantage organization must offer a Provider Access API. Enrollees may opt out of that exchange, and the plan must explain the benefits and the opt-out in simple wording before data first flows.

The same section requires a Payer-to-Payer API beginning January 1, 2027. For that exchange, enrollees must opt in. The plan must explain the benefits of the exchange and the ability to opt in or withdraw permission. The contrast is useful. Within one section, the rule writers chose opt-out for sharing with an enrollee’s providers and opt-in for sharing with other payers.

Operational detail

For a plan, the consent moment produces a content and channel problem. Member services, the website team, compliance and the privacy office each touch the educational resources. The material must be current, accurate about OCR and FTC roles, and reachable through the channels members actually use to request data. A compliance reviewer would ask where the material is published, when it was last reviewed and how the member portal links to it.

For a health system, the problem is similar but voluntary. Patient portal teams decide whether to show a risk notice when a patient connects a third-party app. They decide the wording, how frequently it appears and whether it differs for apps the system itself offers. Those decisions belong to someone. If nobody owns them, the default is silence.

Custody: who answered for the data after it left

The third moment begins once the app holds the data. The question is simple to state. If the app misuses or loses the data, who is accountable, under which law, and to whom does the patient turn.

The HIPAA line

OCR’s FAQs set the line for covered entities. Suppose an app that is neither a covered entity nor a business associate receives health information from a covered entity at the individual’s direction. The FAQs say the information is then no longer subject to the HIPAA Rules. If the app was not provided by or on behalf of the covered entity, the covered entity would not be liable under HIPAA for the app’s later use or disclosure.

The FAQs draw the opposite conclusion for apps the covered entity provides. If the app was developed for, or provided by or on behalf of, the covered entity, the covered entity could be liable for a later impermissible disclosure. The reason is the business associate relationship. OCR gave the example of an app that a provider uses to deliver services to patients.

OCR also addressed the EHR developer. An EHR developer is a business associate of the covered entity. Whether it faces liability for an app depends on whether it owns the app or provides it to, through or on behalf of the covered entity. API terms of use agreed to by an app, on their own, do not create a business associate relationship, according to the FAQs.

What data travels

The scope of the data raises the stakes of custody. Under 422.119(b), the patient API must make available adjudicated claims, including claims that may be appealed or are under appeal, with provider remittances and enrollee cost-sharing. It must make available encounter data from capitated providers. It must also include the clinical data classes in a federal content standard that the plan maintains.

Timing is set as well. Claims data must be available no later than one business day after a claim is processed. From January 1, 2027, the API must also carry prior authorization information, including the status, the dates, the items and services approved and, for denials, a specific reason. The data reach back to dates of service on or after January 1, 2016.

A denial reason or a claim under appeal is the kind of detail a patient may not expect an app to keep. Once it reaches an outside app, its handling depends on that app’s policy and the FTC’s rules. For that reason, the custody moment carries more weight as the data set grows.

Two kinds of app, one portal

This line produces a practical split inside a single patient portal. An app the health system licenses and offers to its patients falls inside HIPAA, with a business associate agreement and the system’s accountability. An app the patient finds on their own falls outside it. To the patient, both may look like the same thing: an app that shows their records.

We think that split is where customer experience and compliance meet. A patient who has a bad experience with an outside app may still blame the hospital whose logo appeared during the connection. The hospital has no HIPAA duty for that app. It may still have a reputational stake, and it owns the explanation the patient received.

The FTC as the backstop

The custody gap was not left empty. The Federal Trade Commission’s Health Breach Notification Rule applies to vendors of personal health records and related entities not covered by HIPAA. The FTC announced changes to the rule on April 26, 2024. It said the changes clarified the rule’s application to health apps and similar technologies not covered by HIPAA.

The FTC listed what changed. A breach of security under the final rule includes an unauthorized acquisition resulting from a data security breach or an unauthorized disclosure. The rule covers entities offering products and services through the online services, including mobile applications, of personal health record vendors. Notices to consumers must include more content, including the identity of third parties that acquired the data.

The FTC also set timing. For breaches involving 500 or more individuals, entities must notify the FTC at the same time they notify affected individuals, without unreasonable delay and no later than 60 calendar days after discovery. The final rule took effect 60 days after publication in the Federal Register. The Commission approved publication by a 3–2 vote.

What custody means for an organization’s promises

The educational duty in 422.119(g) connects to this structure. A plan must tell enrollees which organizations are likely to be HIPAA covered entities and how to complain to OCR and the FTC. In effect, the rule asks plans to explain the custody handoff to their members before it happens. Our reading is that the rule writers saw the custody gap and chose disclosure as the remedy.

For any regulated organization, a further point follows. Statements on a website or in a portal about protecting patient data can be read as promises. If those statements do not distinguish between data held by the organization and data the patient has sent to an outside app, a patient could reasonably misunderstand them. Counsel should review whether privacy notices and portal language draw that line clearly.

Operational detail

Custody work is mostly contract and inventory work. The organization needs a list of the apps it offers or endorses, with the business associate agreement for each. It needs a separate view of the outside apps that patients connect, drawn from API logs. It needs a policy on whether it will feature, recommend or link to any outside app. Each of those choices can move an app across the HIPAA line.

The trade-off here is between helpfulness and liability. Recommending a well-built outside app may serve patients. Depending on the facts, it may also be read as providing the app on the organization’s behalf. Organizations should check with counsel before endorsing any app they do not contract with.

Exit: how the patient stopped the flow

The fourth moment is the end of the relationship. A patient who changes their mind must be able to stop new data from flowing. They also need to know what happens to data the app already holds. These are separate questions with separate owners.

Stopping new data

Stopping the flow rests partly with the data holder. The authorization the patient gave can be revoked at the source, through the organization’s portal or authorization server, or at the app. Blue Button’s guidance asks apps to give users opportunities to revoke service. A plan or provider decides whether its own portal shows the patient a list of connected apps and a way to disconnect each one.

The rules we reviewed do not prescribe that interface for patient apps. Section 422.121 does require a process for enrollees to change their permission at any time for the provider and payer exchanges. The opt-out for provider access must be available before data first flows, and payer-to-payer exchange waits for an opt-in. For patient-directed apps, the design of revocation is left largely to the data holder and the app.

Data already transferred

Data already in the app is beyond the data holder’s reach. Blue Button’s privacy checklist asks app developers to say what happens when a user revokes access: whether previously shared data is retained and used, or securely deleted. It asks about dormant or closed accounts, and about what happens if the company is sold. Those are the questions a patient would ask at exit.

A plan or provider that is not the operator of a federal service has no comparable gate. It cannot require an outside app to delete data on revocation. It can explain, in its educational material, that deletion depends on the app’s own policy. That is the custody gap again, seen from the end of the relationship.

Former members

The rule text recognizes that the relationship can end on the organization’s side as well. Paragraph (g) requires the educational resources to be available through the mechanisms the plan uses with current and former enrollees seeking their health information. A person who has left the plan may still need to know what an app holds.

The patient API itself is described in terms of current enrollees. A former member whose app keeps an old copy of claims data has no live connection to revoke at the plan. The plan’s remaining tool, again, is explanation: telling the person, before and after they leave, how deletion works at the app.

Measuring whether exit works

Measurement arrived late. Under 422.119(f), beginning in 2026, by March 31 after each year it offers a Medicare Advantage plan, an organization must report two metrics to CMS. One is the number of unique enrollees whose data were transferred through the Patient Access API to a health app they designated. The other is the number whose data were transferred more than once.

Those metrics count use, not exits. They do say something about continuing connections, since repeat transfers show apps that kept drawing data. An organization that wants to manage exit would need its own measures: connections revoked, connections dormant for a defined period, and complaints about apps that would not delete data.

Operational detail

The exit moment needs a named owner in member or patient services. Someone answers the call from a patient who wants an app disconnected. Someone confirms the token was revoked. Someone explains, accurately, that deletion of data already shared depends on the app. Scripts for those calls belong in the same review cycle as the educational resources.

How the four moments differed by type of organization

The four moments apply to every data holder, but the obligations differ. Precision about who is covered matters here.

Medicare Advantage organizations and other covered payers

Medicare Advantage organizations are subject to 422.119 and 422.121. The 2020 rule’s title shows that Medicaid and CHIP programs and their managed care plans, and qualified health plan issuers on the federally facilitated exchanges, received parallel requirements. For Medicare Advantage organizations, under the text we reviewed, admission is limited to security grounds, consent is supported by mandatory education, and from 2026 usage is reported.

Health systems

A health system is a covered entity under HIPAA. The access right governs its response to patient-directed apps, and OCR’s FAQs describe its limited room to refuse. A health system that also operates a Medicare Advantage or Medicaid managed care plan holds both roles. Its plan is subject to the payer rules while its hospitals answer to the access right.

That dual role can produce two different experiences for the same person. As a plan member, the person sees the plan’s mandatory educational resources. As a patient, the person may see whatever the hospital portal chooses to show. A health system that wants a consistent experience has to coordinate the two programs deliberately.

Ambulatory networks and smaller practices

Ambulatory networks and physician practices are also covered entities. In our view, they are likely to depend more heavily on their EHR developer’s API and portal design. Their admission and exit controls may be largely whatever the developer’s platform provides. Their main lever is the explanation they give patients, and the business associate terms they negotiate with the developer.

Public programs

State Medicaid agencies are public bodies as well as payers. Their educational resources and complaint channels are also public communications. For them, the consent and custody moments become citizen experience questions, subject to the same scrutiny as any other agency notice.

Agencies and plans that contract out the API

Some payers do not build the patient API themselves. A vendor may host it, run the authorization server and publish the documentation. The rule’s duties stay with the plan. The admission criteria, the denial records and the usage figures reported from 2026 are the plan’s obligations, whoever operates the system.

That arrangement shifts the operational questions into the contract. The plan needs the vendor to apply its criteria consistently, keep records of each denial and produce the usage counts in the form CMS requires. It needs rights to see connection logs for the exit moment. A contract that is silent on those points leaves the plan answerable for evidence it cannot obtain.

The strongest objections

That the asymmetry is the point

One objection is that the narrow refusal grounds are deliberate and good. If every plan and hospital could reject apps on privacy grounds, data holders could use that power to block competitors and protect their own apps. The access right exists to prevent that.

We accept the premise. The narrow grounds protect patients from gatekeeping by organizations with their own interests. Our argument does not call for wider refusal rights. It says that once refusal is narrow, the remaining moments carry more weight, and organizations should invest in them accordingly.

That patients choose and patients bear the risk

A second objection is that adults choose apps the way they choose banks or phones. On this view, a health system that explains too much becomes paternalistic. Our answer is that the rules themselves chose explanation. Section 422.119(g) requires plans to describe secondary uses of data and the OCR and FTC roles. Blue Button’s guidance asks apps to avoid surprise. Explanation is the policy, not a departure from it.

That the FTC rule closed the gap

A third objection is that the 2024 Health Breach Notification Rule changes made the custody gap less important. The rule does cover health apps outside HIPAA and requires breach notice, including for unauthorized disclosures. It is a notification rule. It tells consumers after the fact. The admission, consent and exit moments still decide what happens before that point.

That adoption was the real problem

A fourth objection is that the risk was theoretical because few patients connected apps. We do not know the numbers, and we do not assert any. The usage reporting that begins in 2026 exists because CMS wanted that information. A program whose use is not yet measured is a reason to design the moments carefully now, while volumes can still be managed.

What a health leader could check today

The four moments translate into a short set of checks. Each one maps to a function that can own it.

  • Admission: can the organization show the objective security criteria it applies to app connections, and the record of each denial tied to its risk analysis?
  • Consent: are the educational resources required of covered plans current, plainly written and linked from the screen where members request data?
  • Custody: does the organization keep separate lists of apps it provides under a business associate agreement and outside apps patients connect?
  • Exit: can a patient see and disconnect each connected app in the portal, and do service scripts explain what happens to data already shared?
  • Coordination: where one organization is both a plan and a provider, does the person see one consistent explanation in both roles?

The organization that answers these well has not made outside apps safe. It has made its own part of the experience clear, which is the part it controls.

With hindsight

In our reading, Blue Button 2.0’s model held. The idea announced on March 6, 2018, that a person’s choice of app should move data out of a regulated system, became the basis for binding rules.

The first came on May 1, 2020. CMS published the Interoperability and Patient Access final rule at 85 FR 25510, effective June 30, 2020. It applied to Medicare Advantage organizations, Medicaid and CHIP programs and their managed care plans, and qualified health plan issuers on the federally facilitated exchanges. The fact sheet’s call for private plans to match Blue Button had become a requirement.

The codified text for Medicare Advantage, at 42 CFR 422.119, shows the design choices. Plans must maintain a standards-based API that third-party apps can use at the enrollee’s direction. The applicability paragraph sets January 1, 2021 as the start for the core requirements, for data with a date of service on or after January 1, 2016. Plans may deny an app only on security grounds tied to their risk analysis, using objective criteria applied consistently. Plans must publish educational resources on privacy, the HIPAA boundary and how to complain to OCR and the FTC.

The second rule came on February 8, 2024, at 89 FR 8758, effective April 8, 2024. It extended the program to prior authorization and to exchanges among payers and providers. The codified text now requires prior authorization information in the patient API beginning January 1, 2027. It requires a Provider Access API, with an enrollee opt-out, and a Payer-to-Payer API, with an enrollee opt-in, both beginning January 1, 2027. It added annual reporting of patient API usage beginning in 2026.

HHS’s Office for Civil Rights set out the provider side in its FAQs on apps and APIs. Covered entities generally may not refuse to send ePHI to an app a patient designates because of concerns about the app’s use of it. Once an app that is neither a covered entity nor a business associate receives the data, HIPAA no longer protects it. The FAQ page, in an archived copy showing a last review on May 30, 2025, says the guidance remains in effect only to the extent it is consistent with the court’s order in Ciox Health, LLC v. Azar (D.D.C., January 23, 2020). The FAQs say a covered entity may want to consider telling the patient about risks the first time the patient asks for data to be sent insecurely.

The Federal Trade Commission addressed the custody gap from its side. On April 26, 2024, it announced final changes to the Health Breach Notification Rule. It said the changes clarified that the rule applies to health apps and similar technologies not covered by HIPAA, and that a breach includes an unauthorized disclosure.

CMS kept building on the 2018 platform. On August 3, 2026, it published a readout from the one-year celebration of its Health Technology Ecosystem. It listed initiatives that included the National Provider Directory, a Medicare App Library and expanded Blue Button Individual Access Services scopes. It also listed digital identity verification and QR code-based record sharing among the innovations shown.

What held up from 2018 was the direction of travel and the asymmetry. CMS, as the operator of its own service, still describes a production access process with privacy policy review and a demonstration. Plans and providers still operate under narrow refusal rights and explanation duties. What did not exist in 2018, and does now, is a required measure of use: under 422.119(f), the first annual usage reports were due by March 31, 2026. Whether the explanation duties worked will show in those numbers and in the complaints that reach OCR and the FTC. Neither had been published in a form we could cite when we checked these sources.

Spotted an error? Tell us and we will check it and correct the article where needed.

  • Blue Button 2.0
  • MyHealthEData
  • Patient Access API
  • HIPAA right of access
  • health apps