AI in Prior Authorization: Balancing Innovation and Oversight

The Healthcare Efficiency Dream

It’s no secret that prior authorization (PA) processes have long been a pain point in healthcare—burdening providers, delaying patient care, and clogging up already strained administrative systems. So when artificial intelligence (AI) entered the scene, promising faster decisions and reduced costs, many of us leaned in. Could this be the solution we’ve been waiting for?

Maybe. But the recent class-action lawsuit against UnitedHealth and its AI tool nH Predict reminds us that technology without transparency and oversight can do real harm—especially when it’s making decisions about vulnerable patients’ access to care.

The Upside: Faster, Cheaper, Smoother

The benefits of AI in prior authorization, on paper, are impressive:

• Lightning-fast approvals: Optum Rx's “PreCheck MyScript” pilot cut PA decision times for diabetes drugs from 8.5 hours to just 29 seconds. Nearly two-thirds of requests were approved automatically, appeals dropped 88%, and denials from missing info declined 68%¹.

  
  

• Cutting red tape: In 2024, UnitedHealthcare announced plans to eliminate 20% of prior authorizations, especially for home health and durable medical equipment in Medicare Advantage plans².

  
  

• Lower provider burnout: By removing repetitive tasks, AI can reduce administrative fatigue and help doctors focus on care—not paperwork.

The Downside: Denied by Algorithm

But speed isn’t everything. Behind the efficiency gains lurk troubling risks, especially when AI tools prioritize cost control over clinical judgment:

• Increased denials: According to a June 2025 report, UnitedHealth's denial rate for post-acute care more than doubled between 2020 and 2022 after implementing its AI system³. The American Medical Association (AMA) found 61% of physicians believe AI is increasing denials, with some noting rates up to 16 times higher than expected⁴.

  
  

• Patient harm: A staggering 93% of physicians say PA delays care. Nearly 30% have seen serious adverse events—including hospitalizations and deaths—as a result⁴.

  
  

• Elderly at risk: The UnitedHealth lawsuit centers on elderly Medicare Advantage enrollees being denied continued skilled nursing facility (SNF) care. One patient, Gene Lokken, was abruptly denied coverage by an algorithm—forcing his family to pay $12,000/month out of pocket⁵.

  
  

• Opaque AI logic: The nH Predict model reportedly applied a 20-day maximum stay rule, ignoring physician recommendations and individual patient needs⁶. CMS has since clarified that MA plans must make decisions based on individual circumstances, not “algorithmic averages”⁷.

  
  

The Balancing Act

Here’s how it breaks down:

Pros	Cons
Near-instant decision-making	Denials may override medical necessity
Reduced appeals and delays	Patients drop treatment due to denied access⁴
Administrative cost savings	AI logic often unexplainable and unchecked
Scalable across drug and service lines	Lawsuits, reputational risk, and patient harm are on the rise

The real challenge? Making AI serve clinical intent—not just insurer bottom lines.

What Compliance Officers Should Do Now

If your organization uses or plans to use AI in utilization management, now is the time to ask hard questions:

• Is a clinician involved in the final decision? CMS requires it⁷.

  
  

• Can patients and providers appeal easily? Make the process human-friendly.

  
  

• Is the AI explainable? Black-box tools invite scrutiny and lawsuits.

  
  

• Do you monitor denial trends by population or service line? Pay special attention to SNF, rehab, and other high-volume PA areas.

  
  

Final Thoughts

AI has the potential to transform prior authorization for the better—but only if it’s used with intention, transparency, and accountability. UnitedHealth’s nH Predict case should serve as a flashing warning sign: If we hand over too much control to opaque algorithms, we may solve one problem (delays) by creating another (denials that kill).

If your organization is evaluating AI tools for utilization management or prior authorization, we can help ensure compliance keeps pace with innovation. From policy reviews to program oversight and staff training, Simply Compliance helps healthcare organizations navigate the balance between efficiency and accountability.

Endnotes

1. UnitedHealth Group. (2024). Optum Rx streamlines prior authorizations with PreCheck MyScript. Retrieved from unitedhealthgroup.com

2. UnitedHealthcare. (2024). Streamlining the prior authorization process. Retrieved from uhc.com

3. Investopedia. (2025, June). Denied by a bot? Doctors warn AI is blocking your Medicare Advantage coverage. Retrieved from investopedia.com

4. American Medical Association. (2024). Physician Survey on Prior Authorization and AI. Retrieved from ama-assn.org

5. Financial Times. (2024, Dec). US health insurers face pressure over AI role in claim decisions. Retrieved from ft.com

6. CBS News. (2023). UnitedHealth lawsuit alleges AI tool denied necessary post-acute care. Retrieved from cbsnews.com

7. Centers for Medicare & Medicaid Services (CMS). (2023). MA Final Rule Guidance on Prior Authorization. Referenced via FT reporting ft.com