Medicare Advantage Plans and OIG Denial Concerns

Notably, the OIG found three common and easily preventable causes for these inappropriate denials:

• First, MA organizations sometimes relied on additional clinical criteria not contained in Medicare coverage rules, such as requiring an X-ray before more advanced imaging, which led to denied requests.
• Secondly, some PA requests were denied due to missing or inadequate documentation, despite sufficient documentation being present in the member’s medical records.
• Thirdly, the majority of inappropriate payment denials were caused by human error during the manual clinical review process or inaccurate system programming.

1. Automate clinical intelligence

For years, the push to fix prior authorization has centered on the digitization of existing processes. Many health plans have implemented electronic prior authorization (ePA) systems, and there has been some movement to stratify providers and selectively apply authorization requirements based on past performance.

However, simply digitizing PA does not get to the root of the matter, namely, a fundamental distrust of how health plans have designed their utilization management (UM) programs. Since MA plans are committed to providing members with access to the most appropriate healthcare services, they should leverage a system which incorporates clinical intelligence to guide high-value care choices. Providers should have confidence that the health plan is evaluating these requests not for the purposes of saving money, but to ensure safe, necessary care for each patient.

For UM to function smoothly, the rules should be fully transparent to providers, as specified in the 2018 consensus issued by six national advocacy associations. A UM program must use evidence-based clinical criteria which are clearly defined and referenceable. And, most importantly, it must offer providers meaningful support to help achieve the fastest and best outcomes for patients, which is possible with the integration of AI paired with clinical intelligence. An intelligent authorization platform can easily preempt the three main causes of inappropriate denials found by the OIG team.

2. Clarify the clinical criteria hierarchy

One of the main advantages of MA plans over traditional Medicare is their flexibility. MA plans are allowed to offer supplemental benefits, define their own high-value services, and provide varying coverage to designated populations; generally speaking, they enable a more person-centered approach to benefit design. For example, a MA plan might cover two brand-name drugs as well as a generic alternative, or it might offer a benefit to a population not typically covered under National Coverage Determinations or Local Coverage Determinations (NCDs/LCDs).

However, a MA plan’s UMprogram must have firm controls in place to ensure that the clinical criteria it uses to assess medical necessity are not more restrictive than Medicare’s criteria. This can be very difficult to do without an intelligent authorization platform, which provides automated rules to preserve the prioritization hierarchy.

For example, a MA plan might choose to follow evidence-based guidelines established by a national medical society, as they provide more nuanced, up-to-date guidance than the federal standards. An intelligent authorization platform can evaluate a PA request using the health plan’s policies, while ensuring that the rules default to NCDs/LCDs in instances when these standards diverge. In addition, an automated system can immediately reflect—and enforce—changes in policies as best practices evolve.

3. Solve the documentation problem

Concerns about the performance of Medicare Advantage UM programs are not new. In 2015, CMS cited 56% of 140 audited MA contracts for two types of violations related to the inappropriate denial of service requests and/or payments. Health plans were cited both for making the wrong clinical decision based on the submitted information, and for not gathering the appropriate information from the provider before making a clinical decision.

Let’s take a look at the documentation process. Typically, providers exchange information with between five to ten health plans, each with its own authorization portal, fax process, 278 EDI process, and coverage policies, which are generally not transparent to the provider. Physicians submit individual authorization requests for each separate care service or medication. The health plan can then take up to two weeks to review the case and ask the provider for any missing documentation. Once the provider has faxed the requested information to the health plan, the documentation must then be manually attached to the correct authorization request before the case moves to clinical review.

Given the inherent complexity of this process, it’s no wonder OIG investigators found that some authorization requests were inappropriately denied due to purportedly missing documentation.

In addition to providing the necessary clinical context for service requests, AI-driven authorization platforms can also ensure the proper completion of all PA requests—before they are submitted. Machine learning (ML) models can parse the request as it’s entered, triggering automated prompts when expected information is omitted. For example, a ML model might detect that a physician requesting an injection for a patient has not supplied evidence of advanced imaging within the clinical notes, prompting the platform to request imaging documentation before the request can be submitted.

4. Leverage AI to reduce human error

Understanding the member’s healthcare journey is more important to preventing denials and improving outcomes than most health plans realize. By taking a more holistic approach to care management, health plans can better anticipate, manage, and approve a patient’s needs across an episode of care. An intelligent authorization platform can extract patient-specific data from the EHR, mapping unstructured clinical notes to the correct authorization request to provide the health plan with a more complete patient record.

Using physician input, patient data, and historical data sets, an intelligent platform can also determine longitudinal patient journeys, or care paths, for a wide range of conditions—which enables the automatic suggestion of additional services that might be appropriate for a bundled authorization. Instead of submitting several disconnected requests for one patient, physicians can get multiple services approved in advance for an entire episode of care, reducing the time and cost of PA for both providers and health plans. Such a platform can also prompt physicians to make a higher-value care choice when warranted, such as skipping straight to a gold-standard imaging modality instead of requesting a series of lower-value tests.

Perhaps most importantly, applying AI to UM drives better auto-determination rates for PA requests, as this technology can identify which requests actually require review. An intelligent authorization platform can decrease denial rates by 60% or more, while increasing provider confidence that most requests will be immediately approved. For requests that need manual review, AI algorithms can quickly pinpoint the correct area of focus within the case, making the reviewer’s job faster and easier. An intelligent authorization platform can detect evidence that the health plan’s specific criteria have or have not been met, linking relevant text within the clinical notes to the plan’s policies.

As of last year, 42% of all Medicare beneficiaries were enrolled in a Medicare Advantage plan, a figure that is expected to reach 51% by 2030. With this rapid rise in enrollment, MA plans will face even greater scrutiny regarding the efficacy of their processes. Regardless of what happens at the legislative level, MA plans must ensure that their UM program is built upon sound principles, and is easily scalable to handle a greater volume of requests. While MA plans will never completely eradicate human error, incorporating AI and ML into their UM practices will go a long way toward preventing inappropriate denials.