AHA Urges CMS to Delay Implementation of Prior Authorization Model That Would Employ Use of Artificial Intelligence

The American Hospital Association (AHA) is urging the Centers for Medicare & Medicaid Services (CMS) to make certain changes to its Wasteful and Inappropriate Services Reduction (WISeR) model, which would use artificial intelligence (AI) to review prior authorization requests.

In an Oct. 23 letter addressed to Abe Sutton, deputy administrator and director of the Center for Medicare and Medicaid Innovation at CMS, the AHA shares its recommendations for “guardrails” to be placed on the model “to ensure that it achieves its goal of eliminating waste, fraud and abuse without creating inappropriate barriers to patient care or administrative burden for providers.”

The model is slated to run for six performance years, from Jan. 1, 2026, to Dec. 31, 2031, in six states: Arizona, New Jersey, Ohio, Oklahoma, Texas, and Washington. The voluntary program will use “enhanced technologies,” such as artificial intelligence and machine learning, to expedite prior authorization processes for items and services found to be vulnerable to fraud or inappropriate use. Although technology will be used in the review process, licensed clinicians will make all final decisions.

The AHA recommends the following improvements to the model:

• A six-month delay in the implementation timeline, currently set for Jan. 1, 2026.
• A flat fee compensation structure for selected vendors, rather than compensation by receiving 10 to 20 percent of the savings associated with care denials, which AHA argues could incentivize egregious denials of care.
• Appeal rights for patients denied by participating vendors, rather than the option to resubmit a nonaffirmed prior authorization decision or submit it for reconsideration via peer-to-peer consultation with the vendor.
• Additional guardrails to specifically track physician involvement in prior authorization determinations, including tracking the time spent reviewing each AI-proposed denial, the number of AI recommendations that the vendor physician(s) overturned, and the service-specific qualifications of the clinician reviewer.
• Adequate vendor oversight, including strong control on the specific technologies used by plans.
• A caution against the prospect that the WISeR model could be expanded beyond the specific services identified.
• Clear identification of how CMS will evaluate program success, including a quarterly progress report with relevant vendor performance information.
• Utilization of Fast Healthcare Interoperability Resources-based Application Programming Interfaces standards.

Source:

https://www.aha.org/lettercomment/2025-10-23-aha-comments-cms-wiser-model