“No one went to medical school to do prior auth,” Geeta Nayyar, MD, MBA, a health tech expert, former Chief Medical Officer and author of Dead Wrong: Diagnosing and Treating Healthcare’s Misinformation Illness, shared during HFMA’s Annual Conference.
It’s an area where Nayyar believes artificial intelligence (AI) can help eliminate friction in patient care delivery by automating processes associated with prior authorization.
But we’re not at a point in medicine where AI could be used to issue instantaneous prior authorization decisions. Instead, AI’s true value for prior authorization lies in facilitating prior auth in ways that aren’t considered medically dangerous, including by extracting the information needed to make a clinical decision.
Going Beyond the Hype
Today, 75% of health system executives view digital and AI transformation as a top priority for their organization, according to a McKinsey & Company survey. Prior authorization often enters this conversation as an area of opportunity. “AI has the potential to create the next generation of [prior authorization] workflow design,” a McKinsey analysis states.
But what recently sparked conversation around AI for prior authorization was a Politico interview with Epic’s director of research and development, who said AI isn’t likely to transform prior auth soon.
In his view, the challenge is connectivity—namely, the lack of a national directory for hospitals and health systems to communicate. That’s true—and that’s why providers are relying on technologies they know, like fax, and applying AI to fax as a “FHIR wrapper” to bridge provider-to-provider and provider-to-payer communication gaps. This approach could provide a modern entry point for achieving the connectivity needed for prior auth regardless of where such requests originate. It also levels the prior auth playing field for organizations with less advanced tech resources.
For more than 10 years, there has been a push for electronic prior authorization. The goal has been to replace a process where forms are completed by hand and submitted via fax with highly automated workflows. Such a move increases efficiency, lowers costs, and helps reduce administrative burden.
But providers have been slow to adopt advanced technologies that would digitize prior authorization. Moreover, the ability to truly transform prior authorizations requires more than an electronic fax. What is needed is the ability to integrate structured data into the prior authorization process to ensure health plans have the right information in the right format to accelerate decisions.
It’s an area where AI could make a difference by enabling providers to make prior auth connections using the technology they want to use, the way they want to use it. And, with AI as part of the equation, healthcare providers can eliminate one of the biggest obstacles to streamlining prior auth processing: translating unstructured data into clinical intelligence.
Uncovering AI’s True Value for Prior Auth
Just 31% of prior authorizations were submitted fully electronically in 2023, according to the CAQH Index, published by the Center for Affordable Quality Healthcare. That means that nearly seven out of 10 prior authorization requests are submitted with unstructured data.
This is an area where AI holds significant value. When a prior authorization request is received by a health plan, AI can accelerate prior authorization decisions by extracting unstructured data from digital faxes and other unstructured data sources and providing a structured data set for faster processing. It can even access supportive resources, such as a drug database, to provide critical information for the health plan to make a decision.
And, with the ability to extract data from unstructured prior authorization requests, AI makes it easy for health systems to process these requests. It can automatically provide the data points that are used by prior auth workflow tools and engines. That’s why health plans want healthcare providers to make the move to electronic prior authorizations: because they are structured and feed their tools.
AI can also speed communication to providers once a decision has been made, getting the decision back to where it matters: the patient’s clinical team. This reduces the administrative burden for providers as well as patients. With AI, status information and the prior authorization decision can be applied to the structured data field in the EHR to close the loop for clinical teams and those they serve.
The industry also is seeing examples of physician practices that are leaning into AI to appeal prior authorization decisions. A New York Times article published in July found physicians are increasingly leaning into tools like ChatGPT to write appeals letters in seconds. This is a clever approach, but it’s also risky, given the potential for AI hallucination: instances where incorrect or misleading information is presented as fact. There are also growing concerns around reliance on AI in healthcare decision-making.
However, it’s also important to understand AI’s limits. For example, one area where AI shouldn’t be applied is in making an automatic decision around prior authorization based on previous prior auth decisions. This is controversial and risky due to the potential for bias in large language models related to race, gender, age, and more.
Applying AI in the Right Way for Prior Auth
The recent Epic User Group Conference showcased a number of AI innovations in play for providers and health plans and those yet to come—including AI for fax. While we may not have reached a point where AI can approve or deny a prior authorization or challenge decisions, we can leverage AI to help patients get faster care and reduce the provider’s frustration with delayed responses. Applying AI to extract data from unstructured documents is the right step for digital transformation in an AI-hungry healthcare world.
Johnny Hecker
Johnny Hecker is CRO and EVP of Operations for Consensus Cloud Solutions.
