The Reform Agenda · litigation question · not legal advice
Challenging an algorithmic denial
The question is not whether it feels unlawful. It is what a plaintiff would have to prove, what document would prove it, and whether that document currently exists.
The question
Whether an adverse medical-necessity determination produced or materially shaped by an automated system, and issued without an identified clinician of the relevant specialty, is unlawful — as a breach of the plan’s own promise of individualised review, as a violation of a utilisation-review statute, or as an unfair business practice.
Documentary baseline: the interoperability and prior authorization final rule, which requires a specific reason for every denial and public reporting of authorisation metrics; three state Medicaid managed-care parity audits located in utilisation management rather than in coverage terms; and the appellate record on administrative-law challenges, which sets the realistic posture for any claim against a federal programme.
What a case would require
A denial with a traceable automated component
The hardest element. It usually requires the plan’s own criteria, its configuration, and an admission or metric showing automated involvement. Absent disclosure, a plaintiff is inferring from pattern.
A promise to breach
Plan documents, state utilisation-review requirements, or the federal duty to give a specific reason. The theory is strongest where the plan itself promised clinician review.
A named comparator
Overturn rates on internal appeal and independent review, ideally split by whether automation participated. Public reporting makes this newly obtainable.
Damage that survives the appeal
A denial reversed on appeal after a delay may still have caused harm, but the delay has to be documented at the time. Contemporaneous records decide this element.
A plaintiff who is not the physician
Standing is cleanest for the enrollee. A treating physician’s injury is real and is usually derivative.
A forum that will reach the merits
Preemption, exhaustion and arbitration clauses dispose of most of these claims before any court considers the algorithm.
Where it fails today
On the current record, a case would most likely fail on the first and last elements: no document establishes that an automated system made the determination, and the forum question is resolved against the claim before the evidence is reached. That is a finding about the record, not about the merits.
Which is why the drafted payer standard matters more than a speculative complaint. Its disclosure and metrics provisions would produce, as ordinary compliance, exactly the documents this analysis says are missing. Legislation first is not timidity; it is how the evidence gets created.
The strongest arguments against
- Publishing a feasibility analysis that says a case would fail discourages people with real injuries.
- This is effectively legal advice.
- Naming the weak elements hands a roadmap to defendants.
Answers
- The alternative is encouraging claims that will be dismissed on grounds nobody warned about. A plaintiff is better served by knowing which document to preserve.
- It is not advice, does not concern any actual matter, and no reader is a client. It is a statement of what a responsible case would require.
- Defendants already know their own preemption and arbitration positions. The asymmetry runs the other way.
What would change the analysis. A payer metric showing automated participation; a regulator’s finding on utilisation-review practice; a plan document promising clinician review that was not provided; or a state statute like the one drafted here. Each is a specific, obtainable document. Related: the litigation docket, where actual cases are recorded with their real posture.