AI-guided clinical alignment can help avoid AI-driven denials

Providers have been grappling with rising costs, fewer resources and reimbursement pressure for years. Now on top of that, 84% of hospitals report that the cost of complying with payer policies is increasing, with zero percent indicating a decrease, according to a recent AHA survey. This escalating administrative burden intensifies the financial strain on providers; and it’s being enabled by AI-driven denials.^[1]

Over the past decade, major insurance companies have significantly invested in advanced AI solutions to review and increasingly deny large volumes of claims. Currently, payers reject 15% of claims upon receipt and issue 9% more request-for-information denials compared to 2022.

“With advanced AI capabilities, payers are moving care decisions further upstream, which has the potential to reduce care delays,” said Andrew Ray, senior vice president of operations excellence and innovation at Ensemble Health Partners. “Unfortunately, without efficient clinical document exchange, providers are becoming overwhelmed with requests for additional clinical information, cutting into their already razor-thin margins.”

Fighting AI-Driven Denials with AI-Generated Appeals

Providers are exhausting efforts to combat increasing information requests and denials, but their core competency is clinical care, not claim adjudication. As one health system CEO explained, “Payers are in the business of processing claims. We aren’t.” 

In addition to resource constraints and competing clinical priorities, the lack of standardization across payer policies typically makes it cost prohibitive for providers to develop solutions to combat AI-driven denials. Without widespread payer consensus on clinical guidelines, Ray said providers would need to build more complex AI models to accommodate each payer and clinical condition, which is simply not realistic at the scale of one hospital system.

“It’s a leap — especially for smaller hospitals and health systems — to get some of these solutions in place,” Ray said. “The development work, technical expertise and variety of data required is typically out of scope and out of reach for individual hospitals.”

Some hospitals are experimenting with standard appeal automation, leveraging templates and large language models like ChatGPT. While these base models help significantly reduce the time and effort required for manual processing, their effectiveness in overturning denials often falls short. Many are concerned that this process of combating AI-driven denials with AI-generated appeals might lead to an endless cycle of rapid-fire disputes, failing to reduce unnecessary administrative costs — a “battle of the bots,” with no real winners.^[2]

“Fighting fire with fire isn’t a long-term strategy,” Ray said. “The ultimate place we’re moving to is, ‘How can I prevent friction altogether and achieve clinical alignment between payers and providers, prior to or at the time of care, on the patient’s condition and the course of treatment?’”

Moving Beyond the Battle of the Bots

Achieving AI-guided clinical alignment and avoiding the battle of the bots requires a strategic approach. According to Ray, there are five critical components.

1. Strong revenue cycle operations. Optimizing revenue cycle processes is a crucial first step for any technology initiative, especially those involving advanced AI. This ensures that financial operations are accurate, efficient and reliable, providing a solid foundation for technological advancements. “The value of AI is significantly limited if it is not paired with strong (and often reimagined) operational processes,” Ray said.
2. Robust data and security infrastructure. A secure and comprehensive data ecosystem is the cornerstone of any robust AI framework. “AI can only go as far as your data infrastructure allows it to,” Ray said. Healthcare organizations and their partners must excel at collecting, structuring and storing diverse data while providing ample computational power to enable increasingly sophisticated AI models. Equally important is ensuring robust data security and privacy measures including encryption, access controls and regular audits to safeguard data integrity and confidentiality.
3. Specialized large language models. While general-purpose language models like ChatGPT are useful starting points, specialized tasks such as crafting persuasive clinical arguments specific to a unique patient episode require specific, clinically informed models to create effective results. “It’s important to go beyond payer policies and CMS guidelines when training an effective clinical alignment model,” said Ray. “Processing extensive patient records, diagnostics results, InterQual and Milliman criteria, peer-to-peer interactions and medical research improves the accuracy of clinical reasoning and ultimately produces more compelling arguments,” Ray said. He also emphasized the importance of rigorous testing under close clinical supervision to prevent errors or biases, saying: “It’s critical to include human validation and refinement of model outputs, especially in early model testing, to enhance model accuracy and completeness.”
4. Payer collaboration. AI-driven clinical alignment will require payers to standardize coverage criteria and align on data requirements to facilitate efficient information exchange. Many payers are starting to push for direct access to providers’ electronic health records (EHR), which has the potential to be a great step toward interoperability and automated clinical alignment if implemented correctly. “Since access to EHR data is so valuable to payers for risk scoring and quality measures, we encourage providers to take a thoughtful approach before giving it away without an equitable exchange,” Ray said.
5. Integration with real-time clinical documentation. Ray believes ambient listening technologies will significantly contribute to advancing AI-guided clinical alignment. He explained that ambient listening devices will be able to generate patient notes during a patient consultation and automatically update the EHR in real time. AI will then be able to package and automatically send all necessary clinical documentation to the payer, ensuring clinical alignment before claim submission. “Leveraging real-time clinical documentation will help significantly reduce administrative burden and ultimately help patients get timely access to care,” Ray said.

The financial strain of complying with payer policies and the surge in AI-driven denials are presenting significant challenges for healthcare providers. However, by strategically leveraging AI, providers can reduce administrative burden, achieve better clinical alignment with payers and ultimately reallocate resources towards delivering exceptional patient care.

“AI can serve as a unifying force, transforming the often-adversarial relationship between payers and providers into a collaborative partnership,” said Ray. This shift toward collaboration can lead to more efficient healthcare delivery and improved patient outcomes. 

About Ensemble Health Partners

Ranked Best in KLAS by providers, Ensemble Health Partners is the leading end-to-end revenue cycle managed services firm for mid-sized to large healthcare organizations. Through a combination of more than 11,000 certified revenue cycle operators, data-rich intelligence and AI-infused decisioning, Ensemble helps healthcare organizations sustain best-practice revenue cycle operations and maximize their current technology, so providers can focus on delivering exceptional care in their communities. For more information, visit EnsembleHP.com.

_{This published piece is provided solely for informational purposes. HFMA does not endorse the published material or warrant or guarantee its accuracy. The statements and opinions by participants are those of the participants and not those of HFMA. References to commercial manufacturers, vendors, products, or services that may appear do not constitute endorsements by HFMA.}

---

^[1]. “Survey: Commercial Health Insurance Practices that Delay Care, Increase Costs,” PDF, American Hospital Association.

^[2]. Rosenbluth, T., “In constant battle with insurers: Doctors reach for a cudgel: A.I.,” New York Times, July 11, 2024; Emerson, J., “Using ChatGPT to generate payer appeals? A Georgia system is ‘playing around’ with it,” Becker’s Health IT, July 15, 2024; and Williams, J., “Battle of the Bots: As payers use AI to drive denials higher, providers fight back,” hfm, April 2024.