OIG releases advisory opinion re-analyzing online healthcare directories

On July 11, 2023, the Office of the Inspector General (OIG) posted Advisory Opinion 23-04 (AO 23-04), which addresses an arrangement involving an online healthcare directory for patient-users to book medical appointments, as well as proposes functionality changes to the arrangement.

In it, OIG concluded that, although the technology company requestor’s payment model and scheme would generate prohibited remuneration under the federal Anti-Kickback Statute (AKS) and the Beneficiary Inducement Civil Monetary Penalty Law (Beneficiary Inducement CMP), OIG would not impose administrative sanctions on the requestor.

Notably, AO 23-04 is an update to an arrangement previously discussed in Advisory Opinion 19-04 (AO 19-04), analyzing how the requestor’s platform evolved over the preceding four years, with regard to both payments and functionality.

As healthcare continues to go digital, consumer-driven companies that seek to operate healthcare marketplaces and booking systems are encouraged to review AO 23-04 with a careful lens in order to structure arrangements with appropriate safeguards, as to not violate applicable laws.

AO 19-04 background

In AO 19-04, OIG analyzed the requestor’s website and mobile application-based healthcare directory platform (together, the Platform) where potential patient-users (Users) can search for different types of medical providers (Providers) using specified search criteria, before booking medical appointments with such Providers who were displayed based on search results generated by a proprietary algorithm.

The Platform displayed Providers to Users if the Provider contracted with the requestor via one of its payment methods, one of which was a proposed “per-click” fee, where the requestor charged Providers whenever Users clicked on a Provider’s profile from a search result.

Further, the Platform allowed Providers to purchase advertisements on the Platform and third-party websites, which displayed the Provider, but did not specify any services the Provider offered. The requestor clearly labeled the advertisements as such, and the advertisements were readily distinguishable from the search results. The requestor did not charge Users to use the Platform.

Based on the original proposed arrangement in AO 19-04, OIG determined that, although the requestor’s arrangement implicated AKS and the Beneficiary Inducement CMP, it would not impose administrative sanctions due to the low risk of fraud and abuse. OIG focused on the fact that the fees Providers paid to the requestor in exchange for the referral of federal healthcare beneficiaries did not exceed fair market value, nor would the payment of more fees lead to more frequent appearances for Providers in search results.

Additionally, the requestor’s advertisements did not target federal healthcare program beneficiaries and clearly displayed that they were paid advertisements.

AO 23-04 – Changes to the platform’s search result functionality and spending caps

In AO 23-04, the requestor’s Platform allows Users to search for Providers based on specific criteria, such as whether a Provider accepts a User’s insurance, what services are provided, and geographic proximity. The search results are generated by a proprietary machine learning algorithm, which identifies general preferences for Provider characteristics across Users and then orders such Providers based on engagement data for the Platform and User-centric criteria, to match a User’s preferences. Users still do not pay any fees to use the Platform, and Providers displayed on the Platform still contract with the requestor to display their profiles.

Providers are charged through a singular fee model, whereby the requestor charges Providers a per-click “Per-Booking Fee,” each time a new patient legitimately schedules their first appointment with the Provider (User attendance at the appointment is not considered).

The requestor obtained a fair market valuation of the Per-Booking Fees from a third-party valuation firm, and certified that the fees are not determined in a manner that considers a User’s insurance status or the volume or value of federal healthcare program referrals or business generated.

Additionally, the requestor allows Providers to set spending caps, limiting the number of new patient bookings they will accept per month; however, Providers may not set caps for specific visit reasons or to limit certain patient populations (eg, federal healthcare program beneficiaries). When a Provider hits their spending cap, the Provider will no longer show up in search results for new patient Users.

Further, similar to AO 19-04, the requestor continues to offer the opportunity to purchase paid advertisements. Payment for these advertisements can be through two possible models: the per-click model, which charges Providers based on how many times a user “clicks” their advertisement, or the per-impression model, which charges a Provider every time a User views their advertisement. The requestor stated that these fees do not exceed fair market value and do not consider a User’s insurance status, nor the volume or value of any business generated for Providers.

In addition to the current set-up of the Platform, AO 23-04 also analyzed two proposed changes the requestor would like to implement.

First, the requestor would like to continue to display Providers who hit their spending caps in search results, but only for Users who either indicate they are federal healthcare program beneficiaries or who decline to provide their insurance coverage. In the proposal, these Users would be able to view these Providers, but they would not be able to book appointments with them directly through the Platform until the Provider’s spending cap is reset, which the Platform would clearly indicate. Specifically, these Users would also be able to click on an icon that generally explains why appointments are temporarily unavailable and would be able to click a “notify me” button, allowing them to receive a notification when new appointments become available for the Provider through the Platform. Users could still book appointments directly with such Providers outside the Platform. 

Second, the requestor proposes to measure User engagement for Providers who have spending caps based on the clicks to Provider profiles and the “notify me” button to generate the order of search results. While the algorithm currently uses more than 180 different criteria for Users and Providers to order search results, the requestor would like the algorithm to also consider how frequently Users engage with Providers utilizing the requestor’s spend-cap model.

However, when considering the User-engagement data of the spend-capped Providers, it is possible the algorithm would deprioritize these Providers who inherently limit the amount they are willing to pay the requestor. While the requestor is unsure whether this de-prioritization would actually occur, OIG noted that, if the algorithm actually deprioritizes spend-capped Providers based on User-engagement data, this would be inconsistent with the certification that search results are not prioritized based on the amount Providers pay or are willing to pay or whether they have a spending cap. 

Considering both the arrangement as it exists and the proposed changes, OIG found that the arrangement did not qualify for any AKS safe harbor and implicated both AKS and the Beneficiary Inducements CMP because of (i) payments made by Providers to the requestor in exchange for the Platform to recommend Providers to potential patients; (ii) payments made by Providers to the requestor in exchange for arranging for the provision of federally reimbursable services; and (iii) remuneration (in the form of free usage) provided by the requestor to Users of the Platform, which include federal healthcare program beneficiaries, and which may induce them to seek out services from Providers on the Platform.

However, OIG also noted that there was a low risk of fraud and abuse, and thus, OIG would not impose sanctions on the proposed arrangement due to the following safeguards, which included the following:

1. The requestor certified that, while Per-Booking Fees could vary based on certain criteria, the fees were set in advance and did not exceed fair market value for the services provided by the requestor to Providers. Additionally, the Per-Booking Fees did not take into account the value of federal healthcare program referrals or business generated by the requestor for Providers. Further, the per-impression and per-click advertising fees did not exceed fair market value and did not consider Users’ insurance status, nor the volume or value of any business generated for Providers.
   
   
2. The technology company requestor is not itself a provider or supplier of medical items or services, does not have a relationship (beyond the offering) with the Providers listed on the Platform, and does not expressly recommend specific Providers to Users of the Platform.
   
   
3. The advertisements for the Platform do not target federal healthcare program beneficiaries and can only be viewed if a User initiates contact first by visiting the Platform or a third-party website that displays the advertisements. The advertisements are also clearly marked as paid advertising, thus reducing potential User confusion that the Providers shown in the advertisements are recommended in good faith by the requestor.
   
   
4. Advertisements only promote Providers, not services, and do not influence User search results. The requestor also confirmed that search results are ordered by which Providers best meet the criteria input by the User and are not ordered based on any non-User-centric criteria. Additionally, the Platform clearly states that Providers must pay Per-Booking Fees to be on the Platform, reducing the chance that Users would believe the Platform contains a full scope of all available healthcare professionals.
   
   
5. The Platform collects and stores insurance information from Users solely for User convenience – Users who provide their insurance information will generate search results of Providers who accept their insurance.
   
   
6. The requestor proposed changes that would reduce the risk of inappropriate steering to certain Providers, such as allowing federal healthcare program beneficiaries to view Providers who have hit their spending caps, clearly displaying on the profiles of the Providers that Users would be temporarily unable to make appointments with them through the Platform, but providing a “notify me” button where Users may be notified when appointments become available again on the Platform, and stating that Users may be able to make appointments with Providers currently unavailable of the Platform by contacting them directly.
   
   
7. The algorithm does not filter or order Providers in generated search results based on User engagement with specific Providers, the amount Providers pay or are willing to Pay the requestor, whether Providers have a spending cap, Providers’ historical use or amount of spending caps, or the volume or value of federal healthcare program business generated by the Platform. Rather, the order of Providers is based on User-centric criteria matching a User’s preferences. This allows Providers who may have had historically less popular search generations to have a fair opportunity to be displayed higher in the search results order if any of their characteristics change (eg, if they open a new location in a more populated area). Additionally, even if a Provider pays more or higher fees to the requestor, this does not result in more frequent, or more favorable, exposure of their profiles to Users of the Platform.
   
   
8. Other than allowing federal healthcare program beneficiaries to see all Providers – including those who hit their spending caps – these Users are not differentiated from other Users of the Platform and are not offered any other benefits that would influence their decision to use the Platform or choose any particular Provider.

Takeaways

Any company that may consider implementing a marketplace solution arranging of the provision of healthcare items and services covered by federal healthcare programs, or that may consider recommending Providers (and by extension the purchase of their items and services) to federal healthcare program beneficiaries, should consider the above safeguards when determining pricing mechanisms and the use of algorithmic results.

Further, we note that AO 23-04 does not address the impact of the OIG’s determination that it would not impose sanctions in the event that the algorithm does not deprioritize the spend-capped Providers, or if any of the other assumptions in the opinion are or become untrue. Advisory Opinions are narrow in their application, and typically, when an assumption is no longer true, the OIG’s determination not to impose sanctions against the parties to the arrangement no longer holds.

Accordingly, we expect that OIG puts the onus on the requestor to continue to test and confirm that its algorithm functions appropriately and does not prioritize Providers in an impermissible fashion, even where the requestor acknowledged that it was possible that the algorithm would deprioritize Providers in this manner.

AO 23-04 may therefore highlight that the OIG (and likely other governmental agencies) view the use of artificial intelligence and machine learning in healthcare as requiring post-launch audit and monitoring to ensure that the software functions as intended and in accordance with applicable law.

Lastly, we note that OIG would review all future proposed arrangements on a fact-specific case-by-case basis, and thus AO 23-04 is helpful guidance, but it does not provide any true protection to other arrangements.

If you have any further questions regarding AO 23-04 or the contents of this client alert, please contact your DLA Piper relationship attorney, any member of the DLA Piper Healthcare group, or the authors of this alert.