When an Algorithm Denied Your Claim: The Lawsuits Against AI Prior Authorization Systems
When an Algorithm Denied Your Claim: The Lawsuits Against AI Prior Authorization Systems
Abstract
Multiple major health insurers have deployed algorithmic and artificial intelligence systems to assist in or automate prior authorization denial decisions. Two systems have received particular scrutiny: UnitedHealth Group's nH Predict tool, used to deny post-acute care claims for Medicare Advantage beneficiaries, and Cigna's PXDX system, which allowed Cigna medical reviewers to reject claims without reading individual patient files. Both systems are the subject of litigation and investigative reporting. This brief summarizes the journalism, the litigation, what CMS has done in regulatory response, and what recourse exists for patients who believe their claims were denied by algorithmic review.
Background
Prior authorization has historically involved a physician or nurse reviewer employed by the insurer examining an individual patient's case and making a determination about medical necessity. The process was slow, frustrating, and inconsistent, but it nominally involved human review of individual cases.
The adoption of algorithmic review tools has changed this dynamic in ways that were not fully disclosed to patients or regulators. Insurers began integrating predictive model outputs into the prior authorization workflow, in some cases using algorithm outputs as the basis for denials with minimal or no individual case review.
Two specific systems received significant public attention beginning in 2023.
The Cigna PXDX System: ProPublica Investigation
In March 2023, ProPublica published an investigation into Cigna's PXDX system, based on internal documents and interviews with former Cigna employees.¹ The investigation found that PXDX, an acronym for "procedure-to-diagnosis," was a software tool that automatically flagged claims for denial based on predetermined matching rules pairing diagnosis codes with procedure codes.
According to ProPublica's reporting, Cigna medical reviewers used PXDX to reject claims at a rate of approximately 1.2 seconds per claim.¹ The system allowed a single physician reviewer to process thousands of claims per day, with the algorithm flagging which claims to deny and the physician approving the denial without reviewing the individual patient's medical file.¹
The legal issue is not that Cigna used technology in its review process. The legal issue is the allegation that Cigna was representing to patients and regulators that their claims received individualized medical review when, according to ProPublica's reporting and subsequent litigation, many claims did not.
What the PXDX Litigation Alleges
Following the ProPublica investigation, multiple class action lawsuits were filed against Cigna in federal courts. The complaints allege that Cigna's use of PXDX constituted:
The litigation is ongoing. Cigna has disputed characterizations of the system and its review process.
UnitedHealth's nH Predict System: Medicare Advantage and the Class Actions
The nH Predict system at issue in litigation against UnitedHealth Group is a tool developed by NaviHealth, a care management company acquired by UnitedHealth.³ nH Predict is a predictive algorithm that generates estimates of the expected duration of post-acute care, including skilled nursing facility stays, for Medicare Advantage beneficiaries.
The allegation in multiple class action lawsuits is that UnitedHealth and its affiliated Medicare Advantage plans were using nH Predict's output as a basis for denying coverage for continued skilled nursing facility stays, overriding the assessments of the treating physicians and the patient's documented clinical condition.³
The lawsuits allege that nH Predict had a documented error rate, that it systematically underestimated the care needs of elderly patients, and that patients were being denied coverage for medically necessary care based on what the algorithm predicted a "typical" patient at their stage would need, rather than on the individual patient's actual clinical status.³
A federal class action, Sargent v. UnitedHealth Group, filed in the Northern District of California, alleged that UnitedHealth violated the Medicare Advantage plan rules requiring that coverage decisions be based on individual medical necessity, not algorithmic predictions.³ Multiple other similar suits have been filed in federal courts.
The litigation gained additional traction following a November 2023 STAT News investigation that obtained internal data suggesting nH Predict's denial recommendations were followed by UnitedHealth reviewers at a high rate, and that appeals were frequently successful when patients submitted additional clinical documentation, suggesting the underlying denials were not clinically sound.⁴
What CMS Has Done
The Centers for Medicare and Medicaid Services issued a final rule in April 2024 that directly addressed algorithmic decision-making in Medicare Advantage prior authorization.⁵
The 2024 CMS final rule requires that:
CMS's rulemaking was explicit about the nH Predict controversy, noting that the agency had received significant public comment about the use of predictive tools in post-acute care coverage decisions and had determined that additional guardrails were required.⁵
For commercial insurance, the regulatory response has been more limited. The No Surprises Act and its implementing regulations addressed some aspects of prior authorization transparency, and the 2024 CMS rule on prior authorization (covering MA, Medicaid, and CHIP) requires more transparency about denial criteria.⁵ Legislation has been introduced in Congress to extend similar requirements to commercial insurance, but as of 2025, no comprehensive federal commercial insurance AI prior authorization rule had been finalized.
What the Legal Theory Is
The core legal theory in both sets of litigation involves a relatively simple principle: if an insurer represents to patients and regulators that their claims receive individualized medical review, and the insurer's actual practice substitutes algorithmic output for that review, the insurer may be:
The cases are legally complex because the insurers argue that the algorithms are tools that support, rather than replace, human reviewer judgment, and that the final decision always involves a human reviewer. Plaintiffs argue that when review takes 1.2 seconds per claim, the human reviewer is functionally a rubber stamp for the algorithm, and the representation of individualized review is false.
This distinction is not yet resolved in the courts, and different judges in different jurisdictions have reached different conclusions about which claims can proceed.
What It Means for Patients Whose Claims Were Denied
If you received a denial from UnitedHealth (including UnitedHealthcare, Optum, or a Medicare Advantage plan affiliated with UnitedHealth) or from Cigna and you believe the denial may have been automated rather than individually reviewed, you have several potential options.
Request the Specific Basis for Denial
Under ERISA claims procedure regulations, you have the right to the specific reason for your denial and the specific plan provisions on which the denial was based.² For MA plans, CMS regulations require similar specificity. A denial letter that provides only a general reference to an algorithm or a predicted length of stay without specific clinical criteria applied to your individual case may be procedurally deficient under these regulations.
Request the Credentials and Review Documentation
You can request information about who reviewed your claim, their credentials, and the criteria they applied. If the denial was generated by an automated system, ask for the specific clinical criteria applied to your individual case and the documentation supporting the determination.
File an Internal Appeal With Clinical Documentation
The most effective immediate response to an algorithmic denial is often to submit an internal appeal with robust clinical documentation from your treating physician. Litigation and investigative reporting have found that these appeals succeed at high rates, suggesting the underlying algorithmic denials were not clinically supportable in many cases.
Consider a Class Action Inquiry
If you are a Medicare Advantage beneficiary who was denied coverage for a skilled nursing facility stay, home health services, or post-acute care by a UnitedHealth-affiliated plan, class action litigation is actively ongoing. You may be a member of a potential class. An ERISA or Medicare benefits attorney can assess whether your situation fits the parameters of current litigation.
For Cigna PXDX-related denials, similar class actions have been filed. A benefits attorney can advise on whether your specific claim type falls within the scope of current litigation.
File a CMS Complaint for Medicare Advantage Denials
If you are in a Medicare Advantage plan, you can file a complaint with the CMS Beneficiary and Family Centered Care Quality Improvement Organization (BFCC-QIO) if you believe your care was inappropriately discontinued or denied. For MA enrollees, independent appeals to Medicare's standard appeal process exist separately from the plan's internal process.⁶
The Transparency Gap
One persistent challenge for patients is that the internal workings of AI and algorithmic systems used by insurers are generally treated as proprietary trade secrets. Patients and their physicians do not know what predictive inputs the algorithm uses, what the model's documented error rates are, or whether the model has been validated against the patient population it is applied to.
The CMS 2024 rule creates some additional disclosure requirements for MA plans, but comprehensive algorithmic transparency requirements for commercial insurance do not yet exist at the federal level.⁵ This gap is central to why litigation has become one of the primary mechanisms for accountability.
What This Means for You
If your claim was denied after a very short review period, or if the denial reasoning does not reflect your individual clinical situation, you may be looking at the output of an automated system. That does not automatically mean the denial was wrong, but it does mean you have the right to know the specific clinical basis for the decision and to have that decision reviewed by a qualified individual.
The appeals process exists for exactly this situation. The data consistently shows that algorithmic denials, when properly appealed with clinical documentation, are reversed at significant rates.
Ellen can help you decode your denial letter, identify the specific denial basis, and prepare an appeal that directly addresses the clinical criteria at issue.
Frequently Asked Questions
How do I know if my claim was denied by an AI system?
Most denial letters do not disclose whether automated systems were used. Indicators include very brief denial rationale, predicted length of stay language, or denials that seem inconsistent with your physician's documented assessment. You can request the specific clinical criteria and reviewer qualifications used.
Can I sue my insurer over an algorithmic denial?
This depends on whether you are in an ERISA plan, a Medicare Advantage plan, or a state-regulated individual plan, and on the specific facts of your situation. An ERISA or Medicare benefits attorney can assess your case. Filing an appeal first strengthens any later legal action by creating a record.
What did CMS's 2024 rule actually change for Medicare Advantage?
The rule explicitly prohibited MA plans from using algorithms as the basis for denial decisions without applying individual beneficiary medical information. It also increased prior authorization transparency requirements and added audit requirements for AI tools used in coverage decisions.
Are Cigna and UnitedHealth the only insurers using these systems?
No. Multiple insurers have adopted algorithmic tools for prior authorization review. Cigna and UnitedHealth have received the most public scrutiny due to investigative reporting and litigation, but algorithmic review is an industry-wide practice.
References
If your insurer denied your claim and you are not sure the decision reflects your actual medical situation, Ellen can help you decode the denial and build an appeal with the clinical documentation that challenges it. Start at EllenRx.com.