Why providers can't get paid for using clinical AI tools

The problems with today's healthcare payment models

Using the most common healthcare payment models would either inflate costs or insufficiently incentivize adoption of clinical AI, according to the report.

Costs would be a problem under fee-for-service, which is the most popular healthcare payment methodology per data from America's Health Insurance Plans and Blue Cross Blue Shield Association. PHTI researchers explained that clinical AI enables clinicians to be more productive, thus increasing the number of billable services under the payment model.

This is already happening in some capacity with revenue cycle AI. Solutions that leverage AI for clinical documentation and charge capture are producing more billable codes, which researchers say is driving a rising 9% trend in medical costs.

AI's inflationary impact on healthcare could worsen as providers move toward more autonomous clinical AI, which relies on inputs that differ from those used in fee-for-service models, they explained. For example, clinician time is not the primary driver of value in an AI model despite being the top driver in fee-for-service. Rather, the AI delivers care continuously and at scale at lower marginal costs, they said.

However, common value-based payment models also lack the design features to fully capture the value of clinical AI, which could impact adoption, the report continued.

Health system leaders agreed that pay-for-performance and even capitation models lack the right incentives and coverage rules to support the workflow overhauls required for clinical AI use, especially as the industry moves toward more autonomous technologies.

The business case for adopting clinical AI is also still weak among value-based providers, such as accountable care organizations, since attributing cost savings or improvements specifically to an AI solution is difficult. ACOs are still on the line for total cost of care savings.

How healthcare should pay for clinical AI

The use of clinical AI necessitates a whole new healthcare payment model, health system leaders agreed. As such, they identified three core principles to guide the design of a future payment model that supports AI-enabled care.

First, providers should only get paid for clinical AI use if the technology improves upon current care, they said. This means the AI must improve clinical outcomes, reduce total cost of care or expand access, measured against what providers would have achieved without the technology.

Second, payment needs to account for how clinical AI changes the relationship between utilization and value, they suggested. This means using inputs such as clinical or economic value rather than clinicians' time, effort and resources.

This future payment model must also include safeguards, such as price ratcheting, to prevent overutilization and excessive spending as clinical AI scales up its interventions.

Finally, a future payment model must reward innovation and adoption, then adjust as adoption increases, health system leaders said. Rates should start high enough to incentivize adoption, though, and evolve to account for declining marginal costs and new real-world evidence.

Additionally, they wanted the payment structure to account for the clinicians' evolving role in AI-enabled care. How clinicians deliver care will change as they provide greater oversight and orchestration of clinical AI, thereby requiring more quality assurance and population health management (versus individual encounters).

These core principles point to a complete overhaul of healthcare payment models, not merely modifications to existing ones, health system leaders stressed. These principles should also be used to design different models for different use cases, not a single model to support AI adoption, they said.

Enabling ACCESS to health AI reimbursement

While the healthcare industry works toward comprehensive payment model reform, operational groundwork is already underway. The American Medical Association is developing Current Procedural Terminology codes specifically for autonomous clinical AI, creating the billing infrastructure necessary for providers to seek reimbursement for AI-enabled services.

These specialized codes would allow providers to document and bill for AI-driven clinical services separately from traditional physician services, addressing the fundamental mismatch between clinician time-based billing and the continuous, scaled delivery model of autonomous AI that health system leaders identified as problematic.

Still, the healthcare industry isn't quite there in properly reimbursing providers for clinical AI use, especially for applications that reduce costs and improve population health management. But leaders did say the industry is getting closer with a new model from the Centers for Medicare & Medicaid Services.

The CMS recently launched the Advancing Chronic Care with Effective, Scalable Solutions, or ACCESS, Model to incentive clinician use of technology-enabled care, including AI solutions. The 10-year voluntary model through the agency's Innovation Center replaces Medicare fee-for-service billing with outcome-aligned payments. Specifically, participants receive half of the monthly payment, with the remainder withheld until they meet specific clinical targets.

PHTI researchers said this model is "the most concrete step toward a payment model that accommodates AI-enabled delivery," as it provides long-term payment stability with rates that support technology substitution.

However, researchers said ACCESS isn't perfect for AI-enabled care. The model pays providers only for managing qualifying conditions for attributed Medicare beneficiaries, making it difficult to apply more broadly and limiting the payment structure to chronic disease management. The latter makes it harder to apply to primary care, they explained.

Analysts have also said the ACCESS model's rates are financially unviable at $180 to $420 per beneficiary per year. These rates support more automated workflows but fail to adequately fund clinician-driven chronic care management services, including remote patient monitoring.

As such, providers have been deeply skeptical about the model, unlike digital health companies, which have welcomed it. ACOs also fear that ACCESS could compete with their existing value-based payment structures, creating more care coordination issues rather than solving them.

Health system leaders are still looking for the right path forward for health AI reimbursement, but PHTI researchers stressed that how the industry pays for AI-enabled care now will affect not only the pace of adoption but also AI's impact on costs and quality.

Jacqueline LaPointe is an Executive Editor at Xtelligent Healthcare Media, covering revenue cycle management, healthcare payers, health policy and health IT since 2016.