Cybersecurity donations priced risk by the connection
HHS’s October 2019 proposals to let hospitals donate cybersecurity to nearby practices treated risk as a property of each connection. That gave hospitals three levers over outside parties: shape the link, strengthen the far end, or restrict it.
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
HHS’s October 2019 proposals to let hospitals donate cybersecurity to nearby practices treated risk as a property of each connection. That gave hospitals three levers over outside parties: shape the link, strengthen the far end, or restrict it.
- The week On October 9, 2019, the Department of Health and Human Services announced proposed changes to the regulations that interpret two fraud and abuse laws.
- How a referral relationship turns into a network connection A community hospital and the independent practices around it exchange a steady flow of information.
- Shaping the connection OIG’s own example is the clearest statement of the first lever.
- Strengthening the far end The second lever is the one the proposal created.
- Restricting the connection The third lever is the bluntest.
- Health systems, ambulatory networks and the practices between them The three levers apply differently depending on who holds them.
- The case against managing risk one link at a time The strongest objection to the connection frame is that attackers do not respect tiers.
- Decisions to make before the next connection request The proposals of October 2019 became final in December 2020, and the information blocking rule took effect in mid-2020.
- With hindsight Both proposals were finalized and published in the Federal Register on December 2, 2020.
The week
On October 9, 2019, the Department of Health and Human Services announced proposed changes to the regulations that interpret two fraud and abuse laws. One was the physician self-referral law, known as the Stark Law. The other was the federal anti-kickback statute. HHS presented them as part of its Regulatory Sprint to Coordinated Care, which it described as examining federal regulations that impede efforts among providers to better coordinate care for patients. The two proposals were published in the Federal Register on Thursday, October 17. The Office of Inspector General (OIG) issued the anti-kickback proposal and the Centers for Medicare & Medicaid Services (CMS) issued the Stark proposal. Comments on both were due by 5 p.m. on December 31, 2019.
HHS’s announcement led with new protections for value-based arrangements. Inside both proposals sat a protection for one specific kind of gift. OIG proposed a new safe harbor at 42 CFR 1001.952(jj) for donations of cybersecurity technology and related services. CMS proposed a parallel exception to the Stark Law at 42 CFR 411.357(bb). A safe harbor or exception describes an arrangement that is protected if it meets every condition. Without one, a hospital that gave free security tools to a referring physician practice ran the risk that the gift would be treated as payment for referrals.
OIG explained why it wanted the change. It described health care as an interconnected ecosystem in which the weakest link can compromise the whole system, drawing on the June 2017 report of the Health Care Industry Cybersecurity Task Force. That task force had recommended that Congress evaluate an amendment to both laws so that health care organizations could help physicians acquire cybersecurity software, through donation or subsidy. OIG said that responses to its own earlier request for information showed overwhelming support for a safe harbor, and that some commenters described costs rising to the point where some providers could not invest adequately.
The proposed conditions were short. Donated technology and services had to be necessary and used predominantly to implement and maintain effective cybersecurity. Donors could not directly take into account the volume or value of referrals when choosing recipients or deciding what to give. Recipients could not demand a donation as a condition of doing business. The parties had to sign a written agreement describing the technology and services and any contribution by the recipient. The preamble added that the agreement should give a reasonable estimate of the donation’s value. Donors could not shift the cost to federal health care programs, for example by claiming it as an administrative expense on a hospital cost report.
OIG defined cybersecurity, borrowing from the NIST Cybersecurity Framework, as the process of protecting information by preventing, detecting and responding to cyberattacks. Covered technology meant any software or other information technology, other than hardware. OIG also set out an alternative that would protect hardware a donor judged reasonably necessary after a risk assessment. The existing safe harbor for donations of electronic health record (EHR) technology required recipients to pay 15 percent of the donor’s cost. The cybersecurity proposal had no such contribution requirement.
Both agencies proposed changes to the EHR protections as well. They would make clear that cybersecurity software and services could be donated alongside EHR technology, and they would remove the EHR protections’ sunset, which was then set for December 31, 2021.
One passage is easy to miss. Explaining how a donor could choose recipients without regard to referrals, OIG said selection criteria could rest on the type of connection between donor and recipient. It contrasted a read-only connection to a properly implemented, standards-based application programming interface (API) with a bi-directional read-write connection into the donor’s systems, which it said posed more risk.
One reading of the October 2019 proposals is permission to give. Once finalized, they would let hospitals give security tools to the independent practices around them under a defined protection, if every condition was met. That reading is accurate but incomplete. The more durable idea was in how OIG described the risk. It treated a practice’s security as the hospital’s problem only because, and only to the extent that, the practice was connected to the hospital’s systems. It even suggested that donors could choose recipients by the kind of connection they had.
Read that way, the proposal handed hospital technology and security leaders a frame worth using today. Every outside party plugged into a hospital is a connection with a direction, a scope and a far end the hospital does not control. A hospital has exactly three levers over the risk such a connection carries. It can shape the connection, deciding what can flow and in which direction. It can strengthen the far end, which is what the donation safe harbor allowed. Or it can restrict the connection, limiting or suspending it. Every response to an outside party’s weakness uses one of these levers or a mix of them, and each one changes a different thing.
After a short account of how a referral relationship turns into a network connection, the sections below take the three levers in turn. Three further sections follow. They cover how the levers differ for health systems, ambulatory networks and the practices between them, and the strongest case against managing risk one link at a time. The last sets out the decisions worth making before the next connection request arrives. Nothing here is legal advice. Any donation program or restriction policy should be reviewed by counsel familiar with the anti-kickback statute, the Stark Law and the information blocking rules.
How a referral relationship turns into a network connection
A community hospital and the independent practices around it exchange a steady flow of information. Referrals go out and come back. Orders for labs and imaging come in. Results, discharge summaries and appointment details go back out. Every one of those exchanges runs over some technical channel. It might be a login to the hospital’s EHR through a web portal, an interface engine feed carrying standard messages, a secure messaging service, a virtual private network, an API, or still a fax line.
Each channel gives the practice some ability to see into the hospital, and sometimes to write into it. A portal account lets a physician read the records of the patients she refers. An interface feed may let the practice’s system post orders straight into the hospital’s queue. A network tunnel may give a practice workstation a route to much more than the one application it needs. The far end of every one of these channels is a computer the hospital does not own, patch or monitor.
That is where the risk transfers. A stolen password at the practice becomes a valid login at the hospital. A workstation infected through a phishing email at the practice becomes a machine with a trusted path into the hospital’s network. The hospital’s security team cannot see the practice’s endpoints or email. It sees only traffic arriving over a channel it approved.
Before 2019, a hospital worried about this had awkward options. It could write security requirements into its access terms and hope a small practice could afford to meet them. It could pay for the practice’s security tools, and accept fraud and abuse exposure because free tools have value to the recipient. Or it could accept the risk. The anti-kickback statute and the Stark Law did not distinguish a free firewall service from any other gift to a referral source.
The two laws work differently, which is why both agencies had to act. In general terms, the anti-kickback statute is a criminal law that turns on intent: it prohibits knowingly and willfully offering or paying anything of value to induce referrals of federal health care program business. The Stark Law is narrower in scope but stricter in operation. It bars a physician from referring certain Medicare services to an entity with which the physician has a financial relationship, unless an exception applies, whatever anyone intended. A free security service is something of value under the first law and a financial relationship under the second.
So a hospital’s compliance team, asked about a security donation in 2018, had two separate reasons for caution. Even a well-meant gift could look like an inducement, and a Stark exception had to be found for any physician relationship. The existing EHR protections helped only where the security tools were part of an EHR donation. A stand-alone security program had no clear protection under either law.
OIG’s proposal changed the framing in two ways. It called a cybersecurity donation a self-protective measure for the donor, because a breach in the donor’s system can devastate the donor and everyone connected to it. And in a footnote it observed that if a system is only as strong as its weakest link, even a practice that sends very few referrals poses a cybersecurity risk. Both statements pull the security decision away from referral volume and toward the connection itself.
That is the unit the rest of this article uses. A connection has five properties a hospital can describe. They are the party at the far end, the channel, the direction of flow, the scope of data it can reach, and how users and systems authenticate. The three levers act on different properties. Shaping changes the channel, direction, scope and authentication. Strengthening changes the far end. Restricting changes whether the connection exists at all, and for how long.
Shaping the connection
OIG’s own example is the clearest statement of the first lever. Compare two links. One is a read-only connection to a properly implemented, standards-based API that sends a copy of a patient’s record at the patient’s request. It poses less risk to the donor’s systems than a connection that lets information be written directly into them. A donor allowing a higher-risk connection, OIG said, could set selection criteria that reflect the difference.
The mechanism is easy to trace. A read-only, narrowly scoped API exposes some records if a credential is stolen, which is serious, but it cannot be used to alter a medication order or post a false result. A read-write interface can. A network-level tunnel can reach systems no one meant to expose. The same practice, with the same weak laptop, carries very different risk depending on which of these it holds.
What shaping looks like inside a hospital
Shaping starts with an inventory of external connections. The pieces may sit with different teams. The interface team knows its feeds. The identity team knows its portal accounts. The network team knows its tunnels. A useful inventory joins them and records, for each connection, the outside party, the channel, the direction, the data scope, the authentication method and a named internal owner.
With an inventory, a hospital can sort connections into tiers. One illustrative scheme, not a regulatory requirement, runs as follows:
- Tier one: read-only, scoped API access with short-lived tokens and logging of every call.
- Tier two: interactive portal access to the EHR, with multi-factor authentication and access limited to the practice’s own patients.
- Tier three: interface feeds that write orders, results or documents into hospital systems.
- Tier four: network-level access such as a site-to-site tunnel, which should be rare and justified case by case.
Each tier carries minimum controls, a review interval and an approver. Moving a practice down a tier, for example replacing a tunnel with portal access, can be the cheapest security improvement available. It costs the hospital some engineering and costs the practice some convenience. It costs nobody a donation.
Authentication across the connection
The fifth property of a connection, authentication, is where small practices and large hospitals can talk past each other. A hospital can require multi-factor authentication on its portal. It cannot see when a practice’s front-desk employee leaves, or whether a shared login sits on a sticky note beside the reception computer. The account lives at the hospital, but the person behind it is managed at the practice.
That gap has a practical fix. Access for each practice can be tied to a named sponsor at the practice who confirms the user list on a set schedule, and accounts that are not confirmed lapse. Interface connections use system accounts rather than people, so they need a different control: credentials stored in a vault, rotated on a schedule, and limited to the specific message types the interface carries. A write interface that authenticates with a password unchanged since installation is a tier-three connection with tier-four exposure.
Shaping decisions should also cover what happens when a practice changes vendors. If a practice moves to a new hosted EHR, its interface may be rebuilt by a vendor the hospital has never assessed. A change-notification clause in the access terms, and a short reassessment before the new connection goes live, keep the inventory honest.
Where shaping runs into clinical work
Narrowing a connection is never free. A practice that can place orders electronically avoids transcription errors and phone calls. A practice limited to reading records has to re-key or fax. Interoperability policy in this period pushed toward more electronic exchange, and OIG itself noted that rules designed to increase interoperability and data sharing create more targets. A security team that tightens every connection to read-only will find clinicians routing around it.
So shaping is a joint decision. The clinical owner of a referral program and the security owner of the hospital’s perimeter need to agree on the tier for each class of partner. They also need to agree on the evidence that would justify moving a partner up. That record is the evidence to have ready if an auditor or regulator asks how the hospital manages third-party access: the inventory, the tier assigned, who approved it, and when access was last reviewed.
Standards-based APIs deserve particular attention. OIG singled them out as a lower-risk connection. A well-built API can be scoped, logged and revoked token by token. The alternatives a practice might otherwise use, such as a shared portal login or a screen-scraping tool, are harder to scope and harder to audit. Making the safer channel the easier one is part of shaping.
Shaping also covers what the hospital watches on its own side. Every connection has a normal pattern. A referring practice’s accounts open a predictable number of records at predictable hours, and an interface sends a predictable volume of messages. Logging each connection separately, and alerting when its pattern changes sharply, lets the hospital notice a compromised far end it cannot see directly. A practice account that usually opens a few dozen charts a day and suddenly opens thousands is a signal the hospital can act on within its own walls.
Strengthening the far end
The second lever is the one the proposal created. OIG listed the kinds of technology it expected donors to give. They included software for malware prevention, endpoint protections that support network access control, and business continuity software that limits the effect of attacks. Data protection, encryption and email traffic filtering were on the list too. It said cloud-based and locally installed software would be treated alike.
The list of services was broader. It included developing, installing and updating security software, and training on the tools and on how to prevent, detect and respond to threats. It covered help desk support specific to cybersecurity, and business continuity and data recovery services. It covered a cybersecurity-as-a-service model in which a third party manages or monitors the recipient’s security. It also covered risk assessments, vulnerability analysis, penetration testing and sharing information about known threats. Paying a ransom on a recipient’s behalf would not be protected.
Matching the donation to the connection
The connection frame tells a donor what to give. A practice with tier-one API access needs little from the hospital beyond good token handling. A practice with portal access is mainly a credential risk, so we would expect email filtering, multi-factor authentication support and phishing training to address much of what flows back to the hospital. A practice with a write interface or network access needs managed endpoint protection and monitoring, because its machines are effectively inside the hospital’s trust boundary.
This matters because OIG set no contribution requirement and no dollar cap. It explained that many providers lacked the resources to invest significantly in cybersecurity, and that omitting a contribution would leave recipients’ own money for other needs, such as replacing legacy hardware. Without a contribution to discipline demand, the donor needs another way to decide how much to give. Connection risk is a defensible one, and it is unrelated to referral volume.
The same frame gives the donor a way to judge whether a donation worked. If the purpose is to reduce the risk a connection carries, the measures are on the connection. They include fewer phishing-driven password resets for that practice’s accounts, managed endpoints reporting in, credentials rotated on schedule and no unexplained spikes in access. A donation that leaves those measures unchanged has not reduced the hospital’s risk, whatever it cost. That is a better test than counting licenses deployed.
When the far end is a vendor, not a practice
A small practice may not run its own systems. Its EHR may be hosted by a software company, and their computers may be looked after by a local managed service provider. In that case the far end of the connection is partly the practice and partly its vendors. A donation of endpoint software helps only if the practice’s service provider deploys and maintains it, and it does nothing for the hosted EHR.
This changes the shape of a sensible donation. Services such as a security assessment of the practice’s setup, or monitoring that covers the accounts the practice uses to reach the hospital, may do more than software the practice cannot run. The donor may also find that the most useful thing it can offer is a set of minimum requirements the practice can hand to its own vendors. The donor’s security staff can be available to explain them. That costs little. Counsel should confirm how any support, including advice, is characterized under both laws.
What running a donation program involves
A donation program is a managed service with a legal wrapper. Under the final rule, each donation needs a signed writing with a general description of what is provided and the amount of any contribution by the recipient. Someone has to onboard the practice, deploy the tools, take the alerts and decide what happens when a practice’s laptop starts beaconing to an unknown server at night. The donor’s finance team has to keep these costs out of the hospital cost report.
The proposal also allowed shared funding. OIG’s example was two hospitals and a large multi-specialty practice jointly subsidizing security for smaller practices in their area. Shared programs spread cost, but they also blur ownership. Before launch, the partners need to agree who runs the monitoring, who talks to the practice during an incident, and who decides that a practice’s connection should be cut.
Selection needs a paper trail. OIG said donors could use selective criteria, such as a risk assessment of the potential recipient or a limit to physicians on the hospital’s medical staff. Referrals could not be directly taken into account. A written policy that ties eligibility and scope to connection tier, and records each decision, gives counsel something concrete to defend.
The proposal also asked whether patients could be recipients. OIG did not propose to restrict who could receive donations and said patients would be included, while asking whether extra safeguards were needed. It expected any such donations to be limited, such as anti-malware tools. For a hospital, the patient-facing version of this lever is the security of the patient portal and the apps that connect to it, which is shaping rather than giving.
Restricting the connection
The third lever is the bluntest. A hospital can limit what a connection allows, suspend it during an incident, or refuse it in the first place. In October 2019 that decision was largely a matter of contract and internal policy. Within months it acquired a regulatory frame of its own.
On May 1, 2020, the Office of the National Coordinator for Health Information Technology (ONC) published its final rule implementing the information blocking provisions of the 21st Century Cures Act. The rule applies to health care providers, developers of certified health IT, and health information exchanges and networks. It sets out exceptions for practices that interfere with access, exchange or use of electronic health information but are reasonable and necessary. One is a security exception.
What the security exception asks for
Under the security exception, a practice that interferes with access to electronic health information to protect its security is not information blocking if it meets three conditions. It must be directly related to safeguarding the confidentiality, integrity and availability of the information. It must be tailored to the specific security risk. And it must be implemented in a consistent and non-discriminatory manner.
The practice must also meet one of two further conditions. If it implements an organizational security policy, that policy must be in writing and must respond to assessed security risks. It must align with consensus-based standards or best practice guidance, and set objective timeframes and other parameters for handling incidents. If there is no such policy, the organization must have decided case by case that the restriction was necessary and that no reasonable, less disruptive alternative existed.
Why the three levers converge on one written policy
Set the security exception beside the donation safe harbor and a pattern appears. The safe harbor asks donors to show that who gets help, and how much, is not driven by referrals. The security exception asks providers to show that restrictions are tailored, consistent and based on assessed risk. Both reward the same artifact: a written policy that sorts connections by risk and applies the same treatment to the same tier.
A hospital that has that policy can answer both questions with one document. A hospital without it is making donation decisions and restriction decisions separately, case by case, and is exposed on both sides.
Restriction short of suspension
Restriction comes in degrees. Between full access and none there are several intermediate steps, and they can be easier to justify as tailored to a specific risk. A hospital can force every account at a practice to reauthenticate and reset its password. It can switch a write interface to read-only while an investigation runs. It can block access from unfamiliar locations or devices, or limit how many records an account can open in an hour.
Each of these keeps clinical information moving while reducing what an attacker could do with the connection. Writing them into the policy as a graded ladder, with the triggers for each step, gives the security team options that are faster to apply than a full cutoff. It also gives the hospital a clearer record if a partner later argues that a restriction went further than the risk required.
Suspending a connection during an incident
Restriction is most visible during an incident. If a connected practice is compromised, the hospital may need to disable its accounts, block its interface or drop its tunnel within the hour. The playbook should name who can make that call at night, what evidence is enough and how the practice is told. It should also set the clinical fallback while the connection is down, such as phone and secure fax for urgent results.
Restoration needs criteria too. A practice cut off for a week loses its view of its patients’ hospital records. Clear conditions for reconnection, such as a clean scan of affected machines and reset credentials, keep a temporary restriction from turning into an open-ended one that is hard to justify under the exception’s tailoring condition.
Health systems, ambulatory networks and the practices between them
The three levers apply differently depending on who holds them. A large health system with many independent practices in its referral area can run all three at scale. It can afford an inventory, a tiering scheme and a managed donation program, and it can fund engineering to move partners onto safer channels. Its risk is complexity: hundreds of connections, several interface teams and a donation program that drifts away from the written policy.
Practices the system owns are a different case. Their physicians are inside the system’s own security program, and the donation question does not arise in the same way. The safe harbor matters for the independent practices, labs, imaging centers and post-acute providers that are connected to the system but not part of it.
Large ambulatory networks and multi-specialty groups can sit on either side. They connect to hospitals as recipients, and they may connect smaller practices to their own systems as donors. OIG’s joint-subsidy example included a large multi-specialty practice alongside two hospitals. An ambulatory network that wants to play that role needs the same inventory and policy a hospital does.
Regional health information exchanges add another layer. A hospital that sends and receives data through an exchange holds one connection to the exchange rather than many to individual practices, and the exchange in turn connects to everyone else. That arrangement simplifies the hospital’s inventory and concentrates its risk. The hospital can shape its own connection to the exchange, but it depends on the exchange to shape, strengthen and restrict the connections on the other side. Its due diligence on the exchange should ask how the exchange tiers its participants and how quickly it can suspend one.
Small and rural hospitals face the hardest arithmetic. They may be the weakest link in a regional system rather than the strongest. The proposal did not restrict who could receive donations, which leaves room for a larger system to support a smaller hospital it exchanges data with, subject to the same conditions. Public and government-owned hospitals have an extra question. State law on gifts of public funds and procurement may limit what they can give or accept, so counsel should look at that before a program is designed.
Labs, imaging centers and post-acute providers such as skilled nursing facilities and home health agencies deserve separate attention. They may hold write connections, because their work produces results and documents that flow back into the hospital record. Where they do, they land in tier three of the scheme above, which means the hospital should know exactly which messages each interface can write and who at the partner can change the interface’s configuration.
The proposal also raised a question about donors. In its 2006 EHR rulemaking, OIG had excluded pharmaceutical, device and durable medical equipment manufacturers, among others, from the donor pool. For cybersecurity donations it did not propose to restrict donors, but it asked for comment on whether some types of individuals and entities should be excluded, and why. Any organization that is not a direct care provider and is considering a donation program should check how the final rule settled that.
The case against managing risk one link at a time
The strongest objection to the connection frame is that attackers do not respect tiers. A read-only API token that is stolen still exposes patient records, and a confidentiality breach is reportable whether or not anything was altered. Shaping a connection reduces some harms and leaves others in place.
A second objection is upkeep. An inventory of external connections decays from the day it is finished. Interfaces are added for new service lines, partners change vendors, and portal accounts outlive the staff who used them. A tiering scheme that is not reviewed becomes a record of past intentions.
A third objection is about where the largest risks sit. The connection frame directs attention to the many small parties at the edge. The most damaging failures may come through a few large shared intermediaries, such as clearinghouses, hosted EHR platforms or managed service providers, that connect to many hospitals at once. Managing each link does nothing about that concentration.
A fourth concerns competition. Donations from a dominant hospital can tie practices more closely to it. OIG acknowledged in its final rule that larger, wealthier organizations may be better placed to donate, and said the safe harbor’s conditions balanced that risk against the benefits.
A fifth objection is about responsibility. Once a hospital monitors a practice’s machines, the practice may assume the hospital is watching for everything. The hospital may then find itself expected to respond to problems that have nothing to do with the connection. A donation can blur the line between helping a partner and running its security. If the line is not written down, the first serious incident may draw it instead, at an inconvenient moment.
A sixth is about exit. Donated tools and services create dependence. If the hospital ends the program, merges, or changes its own security provider, practices that relied on the donation may be left with nothing in place. The obvious response, a long commitment, raises the program’s cost and its fraud and abuse profile at the same time.
The frame survives these objections, with limits. The connection is the one place where a hospital has authority. It cannot patch a practice’s laptop, but it can decide what that laptop can reach. Large intermediaries are connections too, and the same inventory and tiering apply to them, even if the hospital has less power to shape them. Where the objections land is concentration. Link-by-link management needs a companion analysis of shared dependencies, asking which single provider’s failure would cut many connections at once. That analysis has to be done at the level of the whole network, not one link at a time.
Decisions to make before the next connection request
The proposals of October 2019 became final in December 2020, and the information blocking rule took effect in mid-2020. A hospital that has not yet joined these pieces can do so with a short list of decisions, each with a named owner.
- Assign one owner for the external connection inventory, and set how often it is reconciled against interface, identity and network records.
- Adopt a tiering scheme for connections, with minimum controls for each tier and a clinical and a security approver for any move to a higher tier.
- Write donation eligibility and scope criteria that rest on connection tier and assessed risk, and have counsel review them against the safe harbor and the Stark exception.
- Write the restriction and suspension policy so that it meets the conditions of the information blocking security exception, including objective timeframes for incident handling.
- Build an incident playbook for a compromised partner: who can cut the connection, what clinical fallback applies, and what evidence restores it.
- Run a separate review of shared intermediaries whose failure would sever many connections at once, and decide what fallback is worth paying for.
Useful measures follow from these decisions. Two are shares: external connections in the inventory with a named owner, and interactive connections protected by multi-factor authentication. Two are times: how long it takes to suspend a partner’s access in an exercise, and how long to restore it. Each is a number a board or an examiner can follow from year to year.
Sequencing matters. The inventory comes first, because tiering, donation criteria and restriction rules all depend on it. Tiering comes second, and the first tier decisions should target the riskiest class of connection, which in the scheme above means network-level access held by outside parties. Donation and restriction policies can then be written against tiers that already exist, rather than inventing their own categories.
The evidence pack is the last piece. For each connection, the record should show when it was approved and by whom, its tier, its last access review, and any restriction applied with the reason. For each donation, it should show the signed agreement, the estimated value, the eligibility decision and how cost was kept out of the cost report. These are the items to have ready for an auditor reviewing third-party access, or for counsel reviewing a donation program. A hospital that can produce them in an afternoon has made the three levers part of its operations rather than a set of intentions.
With hindsight
Both proposals were finalized and published in the Federal Register on December 2, 2020. The anti-kickback rule took effect on January 19, 2021. The Stark rule took effect the same day, except for one amendment to the group practice rules, which took effect on January 1, 2022.
OIG kept the cybersecurity safe harbor and changed it in ways that mattered. It extended protection to cybersecurity hardware that meets the safe harbor’s conditions, and dropped the alternative that would have required a risk assessment before hardware could be donated. It did not adopt the deeming provision that would have treated donations supporting a written program aligned with a recognized framework as meeting the purpose test. OIG said such a provision could end up protecting multifunctional hardware. It encouraged parties to follow widely recognized frameworks anyway. The final text protects technology and services used to implement, maintain or reestablish effective cybersecurity, adding reestablishment to the purpose. The written agreement and the ban on shifting costs to federal programs stayed.
CMS finalized its Stark exception at 42 CFR 411.357(bb), with a requirement that the arrangement be documented in writing. It also made the EHR exception permanent by removing its sunset. The regulatory text of neither rule says anything about connection type. Selection remains the donor’s choice, subject to the condition that it does not directly take referrals into account.
The third lever arrived first. ONC’s information blocking rule was published on May 1, 2020, and took effect on June 30, 2020. As published, it required compliance with part 171, which holds the information blocking provisions and the security exception described above, by November 2, 2020; an interim final rule published on November 4, 2020 moved that date to April 5, 2021. Once compliance was required, a restriction that interfered with access had to meet that exception to avoid being information blocking: tailored, consistent, and backed by a written policy or a case-by-case determination.
The pressure on the weak links grew. In January 2025, HHS proposed a major revision of the HIPAA Security Rule. The proposal said that more than 160 million individuals were affected by large breaches in 2023, a new record. It discussed the February 2024 Change Healthcare incident, in which the clearinghouse disconnected its systems after detecting a threat. Among its proposed changes was a new definition of multi-factor authentication, the control that matters most for the portal connections described above. It devoted a section to small and rural providers. It acknowledged their difficulty attracting security staff and still said strong security was just as important for them as for larger organizations. When we searched the Federal Register on November 14, 2026, we found no final rule from that proposal. Hospitals planning donation programs therefore still work from the 2020 safe harbor and the existing Security Rule, with the 2025 proposal as a signal of direction rather than a requirement.
Two parts of the 2019 framing held up. The safe harbor survived the move from proposal to final rule and remains in the Code of Federal Regulations as of November 2026. The connection lens also proved useful in a way OIG did not emphasize. When the Change Healthcare clearinghouse was attacked in 2024, it disconnected its systems, and the 2025 proposal cited surveys indicating that an overwhelming majority of US health care providers were affected. We read that incident as support for the concentration objection above: the most consequential connections can be the few large ones at the center rather than the many small ones at the edge. A hospital that managed only its small partners, and not its dependence on shared intermediaries, had covered the smaller part of the problem.
Spotted an error? Tell us and we will check it and correct the article where needed.