The U.S. Department of Health and Human Services Office of Inspector General (OIG) recently issued Advisory Opinion 23-04, approving the per-booking fees healthcare providers pay a technology company for each new patient who books an appointment through the company’s online platform and the per-click and per-impression fees the company charges for advertising on the platform.

The advisory opinion—which was requested by the same party that requested Advisory Opinion 19-04, itself approving a similar arrangement—addresses aspects of the platform that were not covered by the earlier opinion, along with proposed changes to its algorithm and search results. Although OIG concludes that the arrangement, both before and after the proposed changes, implicates the federal Anti-Kickback Statute and the Civil Monetary Penalty provision prohibiting inducements to beneficiaries (Beneficiary Inducement CMP), it would not impose sanctions under either law.

Advisory Opinion 23-04 serves as a reminder of the breadth of the Anti-Kickback Statute, particularly for technology companies that do not furnish items or services payable by federal healthcare programs but that may arrange for or recommend those items. It does the same for the Beneficiary Inducement CMP for those that provide free services that may influence beneficiaries’ choice of providers. The opinion also offers useful guidance on key risk factors and safeguards to mitigate these risks.

The Arrangement and Proposed Changes

The requestor operates an online platform through which users can search and book medical appointments with physicians, nurse practitioners, and other healthcare providers. The search algorithm uses over 180 criteria for users (e.g., location, time of search, visit reason) and providers in its directory (e.g., geography, star ratings) to order search results, aiming to match the user’s perceived preferences. It does this through machine learning based on user engagement data across the platform. If users engage more with providers who have a certain characteristic—say, being within five miles of the user—then providers who have this characteristic are prioritized in the search results. Each characteristic is weighted, also based on engagement data. The objective is to identify what matters most to users and sort results accordingly.

To appear in the platform’s searchable directory, providers must pay a fee for each new patient who books an appointment through the platform. The per-booking fee is based on fair market value assessments conducted by an independent valuation firm. The aggregate fees paid by a provider necessarily vary with the volume of business the platform generates for the provider (i.e., the new patient bookings), but the per-booking fee itself is fixed.

Although apparently not contemplated by the arrangement in Advisory Opinion 19-04, the platform allows providers to cap their monthly spend. Capping spend effectively limits the number of new patient bookings providers can receive through the platform. Currently, when providers hit their caps, they no longer appear in the search results (until spend is reset the following month).

The requestor proposed two changes to the existing arrangement. First, spend-capped providers would now appear in the search results for users who identify as federal healthcare program beneficiaries and those who decline to provide insurance information (who could also be federal healthcare program beneficiaries). A notation next to the provider’s listing would indicate that the provider has no current availability through the platform, and a hover-over icon would state that the provider may be able to accommodate the user if contacted directly. A “Notify Me” button would allow users to receive an automatic notification once the provider has availability through the platform.

Second, the requestor would modify the algorithm to include as a characteristic whether the provider has spend caps. That is, the algorithm would now measure user engagement with spend-capped providers and use the data when ordering search results. This, OIG points out, could result in those providers without spend caps being prioritized over those with them. The requestor, however, certified that it does not know if the algorithm will reach this result, the algorithm would continue to use a multitude of characteristics, and their relative weighting would not be fixed (but instead driven by the algorithm).

The requestor also displays paid advertisements in exchange for per-click or per-impression fees. This aspect of the arrangement in Advisory Opinion 23-04 is largely the same as that in Advisory Opinion 19-04. Providers can bid on sponsored placements that appear atop or beside the search results or on a third-party website. These sponsored results advertise the provider but not any particular item or service. Providers pay a fee either per impression (i.e., display on the platform) or per click (i.e., user click). These fees, the requestor certified, do not exceed fair market value and do not take into account the business generated by the requestor.

OIG’s Analysis

The arrangement implicates the Anti-Kickback Statute, OIG concluded, for three reasons. First, through the per-booking fees, the requestor is paid to arrange for and recommend the furnishing of items or services payable by federal healthcare programs. Second, through the per-click and per-impression advertising fees, the requestor is paid to recommend such items and services. Third, by offering free use of the platform, the requestor may induce beneficiaries to purchase such items and services. It would also implicate the Beneficiary Inducement CMP because free use of the platform may influence a beneficiary’s selection of provider. Although both laws are implicated, the opinion was favorable for the following reasons:

1. The fees do not exceed fair market value and are not based on historical spend or the business the requestor generates for the provider. The per-booking fees vary based on medical specialty, geographic location, and other relevant factors. But they do not exceed fair market value, and they are not based on providers’ historical spend, the business the requestor generates for them, or users’ insurance status, any of which would likely increase the risk. Similarly, the advertising fees vary based on users’ clicks (if per-click) or the number of times an ad is displayed (if per-impression), but neither varies with the business the platform generates for the party paying the advertising fees.
2. The algorithm does not favor providers based on the fees they pay or user engagement with specific providers. For any given provider, neither users’ past engagement with the provider, nor the fees the provider has paid the requestor, has any bearing on where it appears in the search results. All that matters is whether the provider has the characteristic the algorithm perceives as preferred and the relative weight the algorithm places on that characteristic. The characteristic and its relative weight—not the identity of the provider or how much it pays—dictate where the provider appears. More bookings means more per-booking fees, but neither leads to more favorable placements in the search results.
3. Although not present in the existing arrangement, transparency safeguards for spend-capped providers would be implemented. Spend-capped providers would now appear in the search results for users who are federal healthcare program beneficiaries and those who decline to provide insurance information. These transparency safeguards—notifying users the provider may have availability if contacted directly and allowing users to set automatic notifications of availability to book through the platform—would reduce the steering risk.
4. The requestor is not a healthcare provider. The requestor’s relationship differs from potentially problematic marketing by healthcare providers, such as “white coat” marketing by physicians. Because it does not furnish medical items or services and is not affiliated with anyone who does, the requestor is not in the same position of trust as healthcare professionals who can unduly influence patients.
5. The advertising is passive in nature and does not specifically target federal healthcare program beneficiaries. Users come to the requestor, not the other way around, and advertisements are only displayed when users are on the platform (as opposed to more targeted forms of advertising, such as emails, mailings, and text messages). Although the platform uses aggregated user engagement data, it does not use the specific user’s engagement to dictate how search results are ordered.
6. The marketing does not promote specific items or services. Although the platform sells sponsored placements, it does not market any particular items or services that users can purchase from providers as a result of appointments booked through the platform.
7. The platform does not target federal healthcare program beneficiaries. The platform is freely available at no cost. Although it does collect insurance information from users who supply it, the platform does not use this information to target federal healthcare program beneficiaries or influence their decision-making.
8. Other than the functionality of the platform, the requestor does not give federal healthcare program beneficiaries anything else of value to induce them to use the platform. The remuneration at issue is limited to the functionality and convenience of the platform. No additional remuneration is provided to users.

Takeaways

Advisory Opinion 23-04 illustrates the breadth of both the Anti-Kickback Statute and Beneficiary Inducement CMP. In particular, it shows the risks technology companies and others face—even if they do not themselves furnish items or services payable by federal healthcare programs—when they are paid to facilitate connections between patients and providers or advertise for providers, and when they offer free services to beneficiaries.

OIG’s analysis of the algorithm illustrates a delicate balance. Users control their searches, while the algorithm influences the order of search results and the sponsored ads that appear alongside them, but the platform still avoids improper steering. Aggregated engagement data may be fine, but user-specific engagement data, particularly with active marketing, may not be. Likewise, selling ads clearly labeled as such may be fine, but selling preferable placements indistinguishably intermingled with standard search results may not be.

The opinion, when compared against Advisory Opinion 19-04, also offers a rare opportunity to see OIG’s analysis of closely related arrangements. All three arrangements—the one in Advisory Opinion 19-04, the existing arrangement in 23-04, and the proposed modifications to that existing arrangement—involve a common set of facts, with slight variations that alter the analysis.

Take, for instance, the Beneficiary Inducement CMP analysis. In the earlier opinion, OIG concluded the Beneficiary Inducement CMP was not implicated. But in Advisory Opinion 23-04, it reached the opposite conclusion. The remuneration to users (free use of the platform) was the same, yet OIG’s conclusion differed on the extent to which free use of the platform would likely influence beneficiaries’ selection of providers. Whether this conclusion was driven by a change in the algorithm or a change of mind is not stated in the opinion.

Consider also the spend caps, which were not contemplated in the earlier opinion. In Advisory Opinion 23-04, OIG points out that “OIG became aware that some aspects of the [existing arrangement] may have differed from those described in [Advisory Opinion] 19-04.” It cautioned that if the requestor has been operating the arrangement “inconsistent[ly] with any material certifications made to OIG in relation to that opinion, [the opinion] would be inapplicable to such arrangement.”

Spend caps compound the pay-to-play nature of the platform. Providers must pay to appear on the platform, and they must also not cap spend to continue appearing in the search results. Although OIG declined to impose sanctions in connection with the existing arrangement, it stressed the importance of including the spend-capped providers in the search results and the transparency safeguards for federal healthcare program beneficiaries. OIG also emphasized that if the spend caps result in the algorithm categorically favoring providers without caps over those with caps, the arrangement would fall outside the protection of the advisory opinion. In other words, the extent to which the opinion protects the modified arrangement will turn on how the algorithm treats spend caps.

At bottom, OIG’s analysis was guided by familiar principles. Parties in a position to generate federal healthcare program business should not be paid more than fair market value by those who benefit from the business. Marketing is heavily scrutinized, particularly if the fee takes into account the business it generates. Even so, per-booking, per-click, and similar fees can pose low risk in limited circumstances. The risk is a function of several factors, including not just the fee amount and structure, but also the identity of the party marketing, its relationship with the target population, the items and services it markets, the target population to whom it markets, and the safeguards it imposes to prevent fraud and abuse. Finding the right balance requires a thoughtful analysis.