In the US, the dual role of artificial intelligence in the prior authorization process is being debated. The implementation of new algorithms into the Medicare system could both optimize costs and create risks for the accessibility of medical care.
What Happened
As part of the Medicare program, the WISeR (Wasteful and Inappropriate Service Reduction) model is being launched, which uses machine learning to reduce cases of fraud and wasteful healthcare spending through 2031. The program marks the beginning of testing automated decision-making algorithms to control medical payments.
Context
The prior authorization process is a critical stage that determines whether an insurance company will pay for a specific treatment. Experts' concerns stem from the fact that the use of AI could turn into an "arms race" between insurance companies striving to minimize payouts and doctors attempting to bypass algorithmic restrictions to ensure patient treatment.
Why It Matters for the Industry
For the insurance industry and AI developers, this creates new economic models where vendors may receive a percentage of the savings. This gives rise to a risk of conflict of interest: optimizing the objective function of an ML model to reduce costs could lead to a systemic bias toward automating unjustified denials instead of real process optimization.
Why It Matters for Users
For patients and doctors, this means that algorithmic decisions in critical areas can directly affect people's physical well-being. Automating denials of insurance coverage could reduce the accessibility of services, which may necessitate the emergence of new "defensive" AI tools for auditing and contesting algorithmic decisions.
What Is Not Yet Known / Limitations
Opinions regarding the extent of the market impact vary: from technical risks of objective function bias to the potential emergence of new market niches for tools protecting the rights of patients and doctors.
Sources
Author
Look at AI, Editorial Team