Contributor: MA Plans Can Leverage Advanced Technology to Shore Up Risk Adjustment Practices

Medicare Advantage organizations (MAOs) are facing new government regulations that are motivating them to take a close look at their risk adjustment programs and policies. In January 2023, CMS released its Risk Adjustment Data Validation (RADV) final rule, which includes changes that will result in the government recouping an estimated $4.7 billion through audits. Also this year, the Medicare Advantage final rule was released, and while it was more lenient for MAOs in terms of the compliance timeline, the urgency among plans to shore up risk adjustment practices and ensure accurate coding is palpable.

While over-coding is a regulatory and financial problem for payers that can lead to penalty repayments, both over- and under-coding have significant impacts on members. The industry standard for coding accuracy is 95%, but findings from the Office of Inspector General^1-3 show many MAOs consistently do not hit that mark. As accurate coding becomes increasingly complex, it’s clear that payers can no longer rely on manual processes for risk adjustment, which can be inefficient and error prone. Fortunately, technology has advanced to the point where it can transform traditional risk adjustment processes, providing MAOs with the ability to increase the accuracy of their coding, mitigate their risk in today’s stringent regulatory environment, and ensure appropriate care for their members.

Ensuring Accuracy and Streamlining Coder Workflow With Clinically Intelligent NLP

It’s easy to miss risk-adjusted diagnosis codes and efficiently search for the clinical support needed to validate each diagnosis using traditional, manual methods. As technology has evolved, purpose-built solutions that leverage sophisticated business logic and intuitive workflows have emerged to help plans increase coding accuracy while reducing the time and resources required.

Advanced artificial intelligence technology, such as natural language processing (NLP) tools, are well suited for identifying specific search terms within unstructured text, which can make up as much as 85% of clinical information stored within electronic health records. However, without an understanding of medical terminology and proper refinement for medical use cases, such as risk adjustment, NLP can pick up erroneous information (commonly referred to as “noise”) that may lead to incorrect coding or frustrated coders. To prevent this, NLP must incorporate clinical intelligence that allows it to be used to validate diagnoses, find potential missed diagnoses, or identify clinical indicators in the context of the data in a medical record.

Clinically intelligent tools can also help rule out irrelevant information, saving coders valuable time by excluding up to 90% of the medical record containing information that isn’t needed for risk-adjustment purposes. It can also be used to automatically connect diagnoses to supporting clinical evidence in the chart, increasing coder efficiency by more than 40% and allowing a coder to review more than 60 records per day compared with the industry average of 20 to 40 records.

Supporting Physician and Coder Education on Documentation

With the latest regulatory changes, physicians and coders alike will be expected to adjust to new standards and intimately understand what is needed to support accurate, compliant risk adjustment policies. With all the changes to risk adjustment coding specifically, such as constrained coefficients for some hierarchical condition categories, individual coders will need additional education that payers would be wise to invest in alongside new technology tools to support accuracy through the learning curve.

Education for providers will be necessary, as well. Despite the large role coding plays in today’s health care marketplace, physicians don’t learn to code as part of their medical training. Additional education will help ensure they aren’t introducing coding errors through their own clinical documentation. Another way technology can help is by identifying patterns of poor documentation to identify clinicians who would benefit from additional coding education and which codes have the highest rates of error or not enough documentation for highly targeted training.

More Opportunity Ahead

Through retrospective review, clinically intelligent NLP tools can capture a diagnosis that the physician has documented anywhere in an individual’s medical record as well as the supporting clinical evidence, helping to defend diagnoses that MAOs have submitted to CMS in the event of an audit. Also possible in the future are concurrent reviews, which can be conducted when the individual is in the hospital and may involve a customer data integration expert querying a clinician directly.

Leveraging advanced technology, a prospective review offers the most complete approach to coding, looking back through years’ worth of records for supported historical conditions not diagnosed in recent visits or that have been lost to follow-up. Unlike a retrospective review, which is limited to a narrow set of data, prospective reviews also look at sources that are ineligible for retrospective review, such as prior lab data, device data, medications, and claims data. This breadth offers MAOs the opportunity to capture more risk-adjustable diagnoses, and also enables clinicians to provide better care for their patients.

Accurate Coding Benefits All

Payers have been managing risk adjustment processes without purpose-built technology for some time, but 2023 may be the tipping point as coding becomes even more complex and regulatory enforcement heats up. With the new rules, MAOs face more pressure and potentially greater financial risk than ever before, meanwhile enrollment in these plans continues to grow, providing great incentive to ensure accuracy and the resources to provide appropriate care.

References

1. Office of the Inspector General. Medicare Advantage compliance audit of specific diagnosis codes that UPMC Health Plan, Inc. (contract H3907) submitted to CMS. Published November 2021. Accessed September 19, 2023. https://oig.hhs.gov/oas/reports/region7/71901188.pdf

2. Office of the Inspector General. Medicare Advantage compliance audit of specific diagnosis codes that BlueCross BlueShield of Tennessee, Inc. (contract H7917) submitted to CMS. Published May 2022. Accessed September 19, 2023. https://oig.hhs.gov/oas/reports/region6/61805002.pdf

3. Office of the Inspector General. Medicare Advantage compliance audit of specific diagnosis codes that BlueCross BlueShield of Tennessee, Inc. (contract H7917) submitted to CMS. Published September 2022. Accessed September 19, 2023.https://oig.hhs.gov/oas/reports/region7/71901195.pdf