AI, Payment Integrity, and the Need for a Human-in-the-Loop

For a discipline that once required scrolling through pages and pages of medical records to find the information you were looking for, artificial intelligence heralds a sea change improvement for payment integrity and fraud detection. These models can sift through millions of documents in mere seconds to find improperly coded claims for further review.

---

But you don’t have to look any further than the class-action lawsuits against big health insurers over prior authorization denials to see the risks of relying on AI in healthcare.
Change is a constant in healthcare. There are regular coding changes, advances in medical technologies and clinical protocols, updated compliance requirements, and even novel new fraud schemes. That poses major challenges to AI models, which rely on historical data to generate answers.

Moreover, healthcare documentation is not natural language. Algorithms can miss critical context and misinterpret clinical nuance.

Why is human oversight more important than ever in payment integrity? How do you incorporate AI while ensuring results are accurate, defensible, compliant, and engender trust from providers?

MedReview explored these questions during a presentation with the Illinois Association of Medicaid Health Plans titled “The Payment Integrity Shake-up: Taking AI in a Human Direction. During their discussion, Spencer Young, CEO, and David Servello, Vice President of Business Development, explored findings and perspectives on the pervasive use of AI in payment integrity and the importance of implementing a Human-in-the-Loop process:

• Pervasive Use of AI and Technology – AI, machine learning, natural language processing, and generative AI are now commonplace terms in payment integrity, enabling computers to learn from data and process information much more quickly than humans. MedReview uses AI strategically to identify and flag claims for further review by clinical experts.

• Human-in-the-Loop Acknowledges AI as a Tool, Not a Decision-maker – MedReview’s core philosophy is that AI should be applied as a tool in the decision-making process. Decisions concerning healthcare dollars directly impact patient care, risking delays and regulatory scrutiny. Our Human-in-the-Loop model ensures that the final determination is always reviewed, approved, and documented by a human with clinician or coding experience.

• Risk of Unchecked AI – Using unchecked AI can lead to “de-skilling” or “cognitive offloading,” reducing critical thinking. Further, AI models are rewarded for providing an answer, even if they sometimes guess. Their authoritative tone can lead to “lazy thinking” and reinforce systematic errors not verified by human clinical acumen.

• Governance and Accountability – Comprehensive cross-functional governance is necessary to ensure AI augments and assists human reviewers, rather than replacing staff. Overriding governance principles mandate validating automation, auditing logic, prioritizing transparency and ethics, and mitigating bias.

• Quality, Desirable Outcomes – A Human-in-the-Loop approach produces superior quality and defensibility. Our appeal overturn rate averages less than 5%. It also helps build provider trust, as providers respect that review decisions are made by a qualified team of medical professionals, rather than just an algorithm.

We’d love to speak with you about our Human-in-the-Loop model, approach to AI in payment integrity, and thoughts on the future of payment integrity. If interested, you can contact us here.