Skip to content
ParamergeParamerge
Stephen Lieberman
Through Paramerge
Contact

PAN Lab example

nH Predict Utilization Review

The order of operations inside a coverage determination

UnitedHealthcare and naviHealth, both owned by UnitedHealth Group, decide whether members leaving the hospital get skilled nursing or rehabilitation care covered, and for how long.

See more

nH Predict is software from naviHealth, the UnitedHealth Group subsidiary that reviews post-acute care requests for UnitedHealthcare. It compares a patient with similar past patients and estimates their needs, including a skilled nursing stay's length. UnitedHealthcare and naviHealth say coordinators use it to help set review dates, not to deny care, and patients suing them dispute that.

What is decided

Medicare Advantage plans are private insurance plans that provide the federal Medicare benefit under contract. UnitedHealthcare is the largest Medicare Advantage insurer. When a member is ready to leave an acute care hospital, the plan decides whether to cover the next stage of care, and for how long. That care can be a stay in a skilled nursing facility, an inpatient rehabilitation facility, or a long-term acute care hospital.

Who does the review

naviHealth, Inc. reviews requests for this care for UnitedHealthcare's Medicare Advantage plans. UnitedHealth Group, UnitedHealthcare, and naviHealth are defendants in the Lokken lawsuit described below. This page calls them "the companies". The companies say naviHealth began some of this work on or around 1 July 2019. A UnitedHealth Group subsidiary bought naviHealth in May 2020. The companies admit a price of $2.5 billion.

A report by a US Senate investigations subcommittee says that in March 2021, a UnitedHealthcare project announced the move of its Medicare Advantage post-acute services to naviHealth. The report calls 2022 the first full year naviHealth managed them.

In a statement quoted by federal inspectors, UnitedHealth Group calls naviHealth "an indirect wholly owned subsidiary". It also calls naviHealth "a first tier downstream contractor and delegate of various UnitedHealthcare entities". The company also reviews requests for other insurers, including Humana. The naviHealth brand was retired in early 2024. The company still appears under that name in June 2024 government data and in 2026 court filings.

Who does what inside naviHealth

The companies described the staff in a sworn court filing in February 2026. By that account, naviHealth has about 2,000 care coordinators, who are licensed professionals such as physical therapists or registered nurses. About 460 of them handle requests to admit a patient, and about 1,500 handle reviews of whether a stay should continue. Coordinators may approve care, but they may not deny it. About 122 physicians, called medical directors, review any case that may end in a denial.

What nH Predict is

The companies' February 2026 filing calls nH Predict "a data-driven care support tool". It estimates needs such as the patient's functional gains and the length of a skilled nursing stay, "based on past experience". The companies say coordinators use it "to assist in setting dates for medical necessity review".

The companies also say nH Predict is not used for rehabilitation facilities or long-term acute care hospitals. A judge accepted that point for the pretrial exchange of evidence. They say coordinators have not used it in reviews of requests to admit a patient since May 2023, and that it is not used to deny care. The people suing contest these statements.

A January 2022 slide obtained by the Senate subcommittee shows a coordinator completing nH Predict. The slide says it is done to "determine optimal [post-acute care] placement" while the patient is in the hospital. UnitedHealthcare told the subcommittee the slide is in error if it suggests the tool decided skilled nursing admissions. In March 2023, naviHealth told the news site STAT that the tool "is used as a guide". STAT later reported a 2023 naviHealth target of keeping rehab stays within 1 percent of the days its algorithm projected. That is a reported internal target, not a measured result.

The insurer's own review tools

UnitedHealthcare also runs its own review of requests for advance approval, separate from naviHealth's. Its Utilization Management Program Committee decides which services need approval in advance. In 2021 the committee voted on two review tools.

In April 2021 the committee voted to approve integrating Machine Assisted Prior Authorization into its reviews. The committee's minutes say the tool "[p]oints the clinician to significant sources of primary evidence". They also call it "never a valid source to justify approval or denial of a case". Asked whether the software "creates potential risk of bias", committee members were told the reviewing doctor or nurse had to verify that "the primary evidence is acceptable". The record does not say what that check compares. Internal testing found the tool cut review time by six to ten minutes. The record gives no starting review time.

In March 2021 the committee heard about a second tool, the Healthcare Economics (HCE) Auto Authorization Model. It came from UnitedHealthcare's Healthcare Economics department. It sorted requests into ones approved automatically and ones set aside for the insurer's Clinical Coverage Review team. It could approve but could not deny. The Senate report says its first tests recorded "faster handle times" together with "an increase in adverse determination rate". An adverse determination is a decision to deny or limit care. The minutes explain the increase as "finding contraindicated evidence missed in the original review". In May 2021 the committee voted to "tentatively approve the model".

The record disagrees about what that increase means. The minutes call it validated, while the Senate report calls the missed evidence alleged. A footnote in the report says clinicians using the technology denied requests at a higher rate. The minutes themselves are not public. No public source says whether either tool was switched on after these votes. No source places either tool in front of naviHealth's coordinators or medical directors.

What the denial rates show

The US Senate Permanent Subcommittee on Investigations published its report in October 2024. It relied on more than 280,000 pages of documents it obtained from three insurers by subpoena. UnitedHealthcare denied 8.7 percent of requests for advance approval of post-acute care in 2019. It denied 10.9 percent in 2020, 16.3 percent in 2021, and 22.7 percent in 2022. Over 2019 to 2022, its denial rate for all requests needing advance approval moved from 7.3 to 7.6 percent. Its skilled nursing denial rate was 1.4 percent in 2019 and 12.6 percent in 2022, nine times higher. The report links the insurer's review tools to the post-acute rise by timing alone.

What the government counted

In June 2026, the Office of Inspector General of the Department of Health and Human Services published counts of decisions made in June 2024. The counts cover the 19 largest Medicare Advantage companies and the contractors that review for them. In that data, naviHealth denied 14.3 percent of the skilled nursing admission requests it reviewed for all its client plans (7,891 of 55,176). Plans that reviewed their own requests denied 11.0 percent. Other contractors denied 8.9 percent.

The plans decided appeals on 1,573 of naviHealth's denials and overturned 1,520 of them, or 96.6 percent. Other contractors' appealed denials were overturned at a similar rate, 97.4 percent. For people who already live in a nursing home, naviHealth denied 47 percent of these requests, against 13 percent for all other members. Three plans told the Inspector General that naviHealth had extra reviewer guidance for nursing home residents. The Inspector General reviewed no medical records. It says it "cannot determine from this data analysis alone whether or how many of these denials were inappropriate".

For comparison, KFF, a health policy research organization, analyzed the advance-approval decisions of all Medicare Advantage plans, for all services, in 2022. Of 46.2 million decisions, 7.4 percent were denied in whole or in part. It found that 9.9 percent of denials were appealed, and 83.2 percent of appeals overturned the first decision.

An earlier government audit, published in 2022, reviewed a sample of denials from one week in June 2019 across fifteen large insurers. It found 13 percent of the prior authorization denials met Medicare coverage rules. That week predates naviHealth's work for UnitedHealthcare, so the audit describes this kind of decision, not this deployment.

What is disputed, and what is settled

Two class-action lawsuits are under way. Estate of Gene B. Lokken v. UnitedHealth Group is in federal court in Minnesota. The Lokken lawsuit claims that artificial intelligence, instead of doctors, decided whether patients kept their skilled nursing care. The companies deny it. They say physicians apply Medicare rules to decide whether care is medically necessary and how long it lasts.

In February 2025 the Minnesota court let two contract claims go forward and dismissed five others for good. A judge wrote in March 2026 that whether nH Predict is used to make coverage decisions "is a merits-based argument". The case is in discovery, the stage where each side hands over evidence. Its trial-ready date is about February 2028. No court has ruled on the merits, and there is no settlement.

Barrows v. Humana is in federal court in Kentucky. In August 2025 that court let four claims go forward. They are breach of contract, breach of the implied covenant of good faith and fair dealing, unjust enrichment, and common-law fraud. Humana admits it hired naviHealth to review some post-acute care in some areas. Humana denies using nH Predict to deny care, and denies owning it.

What the regulator did

The Centers for Medicare & Medicaid Services, the federal regulator, wrote a rule and guidance instead of banning such tools. Its rule in force from 1 January 2024 requires medical necessity decisions to be based on the patient's medical history, their doctor's recommendations, and their clinical notes. Its guidance of 6 February 2024 says an algorithm may help a plan make a coverage decision. A predicted length of stay cannot, by itself, be the basis for ending post-acute care. The plan must re-assess the patient's condition before it issues a notice ending that care. No source shows UnitedHealthcare or naviHealth changing its process under the rule or the guidance.

The regulator proposed guardrails for artificial intelligence in December 2024 and did not finalize them in April 2025.

What the Senate recommended

The Senate subcommittee made three recommendations to the regulator. Collect prior authorization data broken down by type of service. Audit plans where adverse determination rates rise notably. Expand the rules for utilization management committees, so that predictive tools do not unduly influence human reviewers.

What the available tools can and cannot address

A failure pathway is a link between two parts of the network where a mistake made by one part can be passed on to the other.

Explore (No Targets) sets no targets. Under Service Targets Only, the targets can be met. The cheapest combination of tools that meets them costs 6 of this case's 11 budget units.

Under Service and Safety Targets and under All Governance Targets, this case is not fully addressable with the available tools. Those two levels ask you to keep the review useful to the work, to close every failure pathway, and to bring privacy protection back to its target. All Governance Targets adds further targets. Every combination of tools that fits the budget was tried, 22,518 in all. None meets the targets at either level.

The two demands pull against each other. The combinations that keep the review most helpful leave ten failure pathways open. The combinations that close the most pathways still leave four open. They also cut the review's benefit below the level the targets require, because the tools that close pathways also weaken the links the review's usefulness depends on.

The four pathways that stay open are the coverage process itself.

nH Predict is one tool, so an error in its comparison repeats the same way in every case it is used on.

Each decided stay can become part of the comparison data nH Predict uses for later patients. The public record does not show what that data contains.

Each decision is counted in the figures the insurer reports to the regulator.

Each recorded denial becomes a written notice to the patient and the care provider.

Every tool at once, ignoring the budget, still leaves open the last three: the comparison data, the reported counts, and the notices. Closing those four would mean switching the coverage process off. That is a finding about the deployment, not a flaw in your choices.

Stylized model of a documented deploymentClinical decision support & deterioration alerting

Open this example in PAN Lab v0.1 to apply pressures and levers and watch what the system does.

What this models

This example runs on the nH-Predict-class payer utilization review network: 12 components and 25 pathways between them. Every context in the Lab is a stylized model, never a reconstruction of any actual deployment, and each assumption behind it carries a provenance label.

Evidence base: 1 assumed · 9 published baseline. In the Lab, the shaded evidence band behind each headline readout draws its width from the least-established class below.

Show all 10 assumptions
  • baseline

    This network has twelve parts, each documented in the public record. Other cases in this domain include alerts to the clinician treating a patient. This one sits on the insurer's side. Its subject is the order of steps inside a coverage decision. It includes two software tools: nH Predict on naviHealth's path, and Machine Assisted Prior Authorization on UnitedHealthcare's own path. One group, the medical directors, stands for the physicians who deny care on both paths. The denial or termination notice is its own step, because the regulator's 2024 guidance sets what must happen before one issues. There is no list of cases waiting for review, because the record documents less time per review rather than a backlog. There is no automated screen on the tools' output, because none is documented. The documented control before a decision is a person verifying the primary evidence. Nothing outside the review is included. The record does show naviHealth sending structured authorization data to Humana, another insurer it reviews for, and that exchange is not part of this network. When the record shows one act working in both directions, the network draws it once and the text describes the other direction.

  • baseline

    The network assumes a heavy workload against limited staff capacity. On the workload side, Medicare Advantage plans made 46.2 million prior authorization decisions in 2022. UnitedHealthcare's post-acute denial rate rose from 10.9 percent in 2020 to 22.7 percent in 2022. Testing of one review tool cut six to ten minutes from the average review. Call handlers at naviHealth were told not to help providers answer the intake questions. On the capacity side, the review is done by real clinical staff with duties that federal rules name. By the companies' February 2026 account, naviHealth has about 2,000 coordinators and about 122 medical directors. Capacity is not assumed to be high. Review time was cut when the duty to verify the evidence was stated. The regulator restated the duty to re-assess each patient in February 2024. The one correction with a measured effect, the plan's review of appeals, reached about one in five of naviHealth's skilled nursing denials in June 2024.

  • baseline

    The companies describe nH Predict as estimating from the experience of similarly situated members. So the network treats that comparison data as the tool's primary input. The patient's own record reaches the tool through the coordinator who uses it. This follows the regulator's stated objection to tools that decide coverage from a larger data set, instead of the patient's own history, doctor's recommendations, and clinical notes. The plan's review of appeals carries two separate facts. When a denial is appealed, it is usually overturned. Few denials are appealed. In June 2024, plans overturned 96.6 percent of naviHealth's appealed skilled nursing denials, and about one in five of those denials reached an appeal. Across all Medicare Advantage plans in 2022, 83.2 percent of appeals overturned the first decision, and 9.9 percent of denials were appealed. The network keeps the two facts apart and never averages them. It treats the overturn rate as an upper bound, because people choose which denials to appeal.

  • baseline

    The network assumes that decided stays can become part of nH Predict's comparison data. A stay that actually happened lasted as long as earlier coverage decisions allowed. So a decision can shape the reference that later estimates are measured against. The public record does not show the data's contents, and a March 2026 court order left them out of what the companies must produce. Because naviHealth reviews for several insurers, the network does not assume the data comes from UnitedHealthcare's decisions alone. No case in this domain built on alerts to a treating clinician has reference data that its own decisions help produce.

  • baseline

    UnitedHealthcare's Utilization Management Program Committee both sets the review's workflow and checks its tools. The record shows the first more than the second. It approved Machine Assisted Prior Authorization, whose testing cut review time. It tentatively approved the HCE Auto Authorization Model after seeing faster handle times together with an increase in adverse determination rate. It reserved a return to move some rules from automatic approval to review by a person, and no source shows it did. The minutes call the increase validated catches, and the Senate report calls the missed evidence alleged. This is not an accusation. The committee exists, meets, records its reasons, and can withhold approval. The Senate's third recommendation, to expand the rules for such committees so predictive tools do not unduly influence human reviewers, aims at this part of the review.

  • baseline

    The network treats the denial or termination notice as its own step, because a recorded denial becomes a notice as a matter of course. Across Medicare Advantage in 2022, 3.4 million requests were denied in whole or in part. The re-assessment before a notice is treated as a real check. The standing federal rule already required decisions based on the patient's own history, doctor's recommendations, and clinical notes, with the medical director involved where appropriate. The February 2024 guidance restated the order: a tool may assist, but a predicted length of stay alone cannot end post-acute care. The check is not assumed to work well. The guidance came after UnitedHealthcare's post-acute denial rate rose from 10.9 to 22.7 percent in two years, while its review tools were being made faster. No source shows either company changing its process under it.

  • baseline

    Some claims about this case are left out on purpose. The figure most often attached to it, that over 90 percent of denials are reversed on appeal, is an allegation in the Lokken complaint. The court recited it under the rule that takes pleaded facts as true when deciding a motion to dismiss, and the companies deny it. No figure here rests on it. The overturn figures used are the Inspector General's counts for naviHealth, with KFF's 2022 analysis as background. Humana admits hiring naviHealth for some post-acute reviews, but its use of nH Predict rests on the plaintiffs' pleadings alone, and Humana denies it. In December 2022, a UnitedHealthcare workgroup explored using appeal data to predict which cases would be appealed. The record shows two meetings, not a working system, so no third tool is included. The 2023 target of keeping rehab stays within 1 percent of projected days is a reported internal target, not a measured behavior of any coordinator.

  • baseline

    On nH Predict's side, a mistake means an estimate that misstates what a patient's circumstances require, judged against the coverage rules that govern the decision. It is never an accuracy figure about a person. One audit gives a sense of scale. The government's 2022 evaluation found that 13 percent of sampled 2019 prior authorization denials met Medicare coverage rules. It pooled fifteen large insurers over one week in June 2019, and named post-acute stays among its examples. That week predates naviHealth's work for UnitedHealthcare. So the audit describes this kind of decision in general and measures nothing about this deployment.

  • baseline

    Most facts about this deployment come from documents the companies produced under compulsion. They include more than 280,000 pages subpoenaed from three insurers for the Senate report, and 2026 court filings. Others come from government audits, the regulator's guidance, court orders, and investigative reporting read only in its free preview. The Inspector General's 2026 reports measured naviHealth's decisions by name, but reviewed no medical records and did not examine the tool. No independent technical evaluation of nH Predict exists in the public record. The naviHealth brand was retired in early 2024. The company kept its role and still appears under the naviHealth name in 2024 government data and 2026 court filings.

  • assumed

    Patients are not part of what this network computes. Nothing here computes a coverage decision, a length of stay, a medical necessity judgment, or a health or money outcome for anyone. No score about a person appears anywhere in it. The denial, appeal, and overturn rates describe streams of requests, as the government and analyses of the plans' own reporting measured them. The Inspector General's 2026 reports add measured gaps between groups of patients, such as nursing home residents. The network records those as outside observations and computes nothing from them.

What this example does not show

Show all 5 limitations
  • Whether nH Predict was used to make coverage decisions is disputed, and no court has decided it. UnitedHealthcare and naviHealth say coordinators use it to help set review dates, not to deny care. In 2023 naviHealth told the news site STAT the tool is used as a guide. Both class actions, Lokken against UnitedHealth Group and Barrows against Humana, survived motions to dismiss in part, that is, the courts let some claims go forward. In Lokken, the contract and implied-covenant claims go forward. In Barrows, those two go forward along with unjust enrichment and common-law fraud. Both are in discovery, with no ruling on the merits and no settlement. So this example assumes neither side's answer. It draws on what documents record, including a January 2022 slide showing a coordinator completing the tool, and the committee votes with their recorded test results. UnitedHealthcare disputes what the slide implies, and the Senate report disputes the minutes' reading of the test results.
  • The figure most often attached to this case, that over 90 percent of denials are reversed on appeal, is an allegation. The court recited it under the rule that takes pleaded facts as true when deciding a motion to dismiss. The companies deny it, and this example does not use it. The appeal figures used are the Inspector General's June 2024 counts for naviHealth's skilled nursing admission denials. The companies say nH Predict has not been used in admission reviews since May 2023. So those counts may not cover the decisions the tool is said to inform. KFF's 2022 figures, 83.2 percent overturned with 9.9 percent appealed, cover all Medicare Advantage plans and all services. Every overturn rate counts only the denials someone chose to appeal.
  • A 2022 government audit of one June 2019 week of denials across fifteen large insurers found that 13 percent of prior authorization denials met Medicare coverage rules. It found the same for 18 percent of payment denials. That week predates naviHealth's work for UnitedHealthcare. The figures describe this kind of decision in general and measure nothing about this deployment. No error rate for nH Predict has been published, and no independent technical evaluation of it exists.
  • The record now includes a measured gap between groups of patients. In June 2024, naviHealth denied 47 percent of skilled nursing requests from people who live in nursing homes, against 13 percent for everyone else. Across the 19 largest insurers, people eligible for both Medicare and Medicaid were denied skilled nursing admission at 15.7 percent, against 10.5 percent for Medicare Advantage-only members. The Inspector General reviewed no medical records, so these figures show a gap, not wrong decisions. This example records them and computes nothing from them.
  • This example does not model patients. Coverage decisions, lengths of stay, and patients' health and money are recorded in the case file, not computed here. The Lab shows how mistakes move between the review's tools, staff, and records. The naviHealth brand was retired in early 2024. The company kept its role and still appears under the naviHealth name in 2024 government data and 2026 court filings.

Sources and evidence

What this example rests on, claim by claim. Every entry resolves to the same ledger the Evidence Registry publishes.

  • Internal records subpoenaed by the U.S. Senate Permanent Subcommittee on Investigations show that early-2021 testing of an auto-authorization model inside UnitedHealthcare produced faster handle times together with an increase in adverse determination rate - attributed to finding contraindicated evidence missed in original review - and the internal committee voted to tentatively approve the model at the following meeting; the April 2021 approval of 'Machine Assisted Prior Authorization' was paired with testing that removed six to ten minutes from the average review while the reviewing doctor or nurse still had to verify that the primary evidence is acceptable. Over the same period the insurer's post-acute prior authorization denial rate went from 10.9 percent (2020) to 16.3 percent (2021) to 22.7 percent (2022), and its 2019 skilled-nursing-facility denial rate was nine times lower than its 2022 rate. A January 2022 vendor presentation shows a naviHealth care coordinator completing nH Predict to determine optimal post-acute placement while the patient is still hospitalized, and an April 2022 vendor instruction told call handlers not to guide providers on the questions used to collect the information determinations are made from.

    empirical
    • Government U.S. Senate Permanent Subcommittee on Investigations (Committee on Homeland Security and Governmental Affairs), Majority Staff Report, Refusal of Recovery: How Medicare Advantage Insurers Have Denied Patients Access to Post-Acute Care, 17 October 2024 https://www.hsgac.senate.gov/wp-content/uploads/2024.10.17-PSI-Majority-Staff-Report-on-Medicare-Advantage.pdf
    • Investigative Casey Ross and Bob Herman, UnitedHealth pushed employees to follow an algorithm to cut off Medicare patients' rehab care, STAT, 14 November 2023 (part of the Denied by AI series; free preview read, full text paywalled) https://www.statnews.com/2023/11/14/unitedhealth-algorithm-medicare-advantage-investigation/

  • Across Medicare Advantage in 2022, of 46.2 million prior authorization determinations, 3.4 million (7.4 percent) were denied in whole or in part; 9.9 percent of denials were appealed; and 83.2 percent of appeals resulted in the initial decision being overturned (KFF analysis of the plans' own federal reporting). The figures are program-wide, not plan- or service-line-specific, and the overturn rate is conditioned on the self-selected minority of denials that were appealed, so it overstates what the same review would correct if applied to every denial.

    empirical
    • Reference KFF, Medicare Advantage Plans Denied a Larger Share of Prior Authorization Requests in 2022 Than in Prior Years (8 August 2024), analysis of federal Medicare Advantage prior authorization and appeals reporting https://www.kff.org/medicare/medicare-advantage-plans-denied-a-larger-share-of-prior-authorization-requests-in-2022-than-in-prior-years/

  • Whether nH Predict was used to make coverage determinations is contested and unadjudicated. The vendor's public statement in STAT's March 2023 series opener was that the tool 'is not used to make coverage determinations' and 'is used as a guide'; the Lokken order (D. Minn., Feb 13, 2025) recites - as pleaded allegations taken as true on a motion to dismiss - an allegation about the share of claim denials reversed on appeal, and records in the same paragraph that 'UHC denies any use of nH Predict'; the Barrows order (W.D. Ky., Aug 14, 2025) recites Humana's use of the tool as pleadings only. Both class actions survived motions to dismiss in narrowed form - contract and implied-covenant counts in both, plus unjust enrichment and common-law fraud in the Humana action and are in active discovery as of August 2026, with no dispositive merits ruling. CMS's contract-year-2024 rule and its February 6, 2024 FAQ constrain the order of operations: an algorithm may assist a coverage determination, the plan remains responsible, an algorithm determining coverage from a larger data set instead of the individual patient's medical history, physician recommendations, or clinical notes would not comply with 42 CFR 422.101(c), and a predicted length of stay alone cannot be the basis to terminate post-acute care services - only re-assessing the individual patient's condition can. The naviHealth brand was retired in Q1 2024 into Optum 'Home & Community Care'; the entity continues in the same delegated role and still operates under the naviHealth name in June 2024 regulator data and in 2026 court filings.

    empirical
    • Government Estate of Gene B. Lokken v. UnitedHealth Group, Inc., UnitedHealthcare, Inc., and naviHealth, Inc., Civil No. 23-3514 (JRT/DJF), U.S. District Court for the District of Minnesota, Memorandum Opinion and Order Granting in Part and Denying in Part Defendants' Motion to Dismiss, Doc. 91, 13 February 2025 (Tunheim, J.) https://storage.courtlistener.com/recap/gov.uscourts.mnd.211721/gov.uscourts.mnd.211721.91.0.pdf
    • Government Centers for Medicare & Medicaid Services, Health Plan Management System memo to all Medicare Advantage Organizations and Medicare-Medicaid Plans, Frequently Asked Questions related to Coverage Criteria and Utilization Management Requirements in CMS Final Rule (CMS-4201-F), 6 February 2024 https://www.aha.org/system/files/media/file/2024/02/faqs-related-to-coverage-criteria-and-utilization-management-requirements-in-cms-final-rule-cms-4201-f.pdf
    • Government 42 CFR 422.101(c), Medical necessity determinations and special coverage provisions (Centers for Medicare & Medicaid Services, as amended by the contract-year-2024 Medicare Advantage final rule CMS-4201-F, applicable to coverage beginning 1 January 2024) https://www.ecfr.gov/current/title-42/section-422.101
    • Investigative CBS News, UnitedHealth uses faulty AI to deny elderly patients medically necessary coverage, lawsuit claims https://www.cbsnews.com/news/unitedhealth-lawsuit-ai-deny-claims-medicare-advantage-health-insurance-denials/
    • Investigative Casey Ross and Bob Herman, Denied by AI: How Medicare Advantage plans use algorithms to cut off care for seniors in need, STAT, 13 March 2023 (series opener; free preview read, full text paywalled) https://www.statnews.com/2023/03/13/medicare-advantage-plans-denial-artificial-intelligence/
    • Government Barrows, et al. v. Humana, Inc., No. 3:23-cv-654-RGJ, U.S. District Court for the Western District of Kentucky, Memorandum Opinion and Order Granting in Part and Denying in Part Defendant's Motion to Dismiss, Doc. 82, 14 August 2025 (Jennings, J.) https://litigationtracker.law.georgetown.edu/wp-content/uploads/2023/12/Barrows-et-al_2025.08.15_MEMORANDUM-OPINION-ORDER.pdf
    • Investigative Bob Herman and Casey Ross, UnitedHealth discontinues a controversial brand amid scrutiny of algorithmic care denials, STAT, 23 October 2023 https://www.statnews.com/2023/10/23/unitedhealth-optum-navihealth-rebranding-algorithm/
    • Reference Georgetown Law, O'Neill Institute Health Care Litigation Tracker, Estate of Gene B. Lokken et al. v. UnitedHealth Group Inc. et al. (tracker entry, procedural posture read August 2026) https://litigationtracker.law.georgetown.edu/litigation/estate-of-gene-b-lokken-the-et-al-v-unitedhealth-group-inc-et-al/
    • Reference Georgetown Law, O'Neill Institute Health Care Litigation Tracker, Barrows et al. v. Humana Inc. (tracker entry, procedural posture read August 2026) https://litigationtracker.law.georgetown.edu/litigation/barrows-et-al-v-humana-inc/

  • The closest ground-truth audit of this decision class is HHS OIG evaluation OEI-09-18-00260 (April 2022): reviewing a stratified random sample of 250 prior authorization denials and 250 payment denials issued by 15 of the largest Medicare Advantage organizations during one week of June 2019, health-care coding experts and physician reviewers found 13 percent of the prior authorization denials and 18 percent of the payment denials met Medicare coverage rules, with identified causes including internal clinical criteria applied beyond Medicare rules, insufficient-documentation findings the reviewers judged unfounded, manual processing errors, and system programming failures, and with stays in post-acute facilities among the report's own examples. The evaluation predates naviHealth's management of this benefit and pools fifteen organizations, so it orders - and does not measure - an error term for this deployment.

    empirical
    • Government evaluation U.S. Department of Health and Human Services, Office of Inspector General (2022, April 27). Some Medicare Advantage Organization Denials of Prior Authorization Requests Raise Concerns About Beneficiary Access to Medically Necessary Care (OEI-09-18-00260) https://oig.hhs.gov/oei/reports/OEI-09-18-00260.asp

Where this connects

Institutional pressures in this domain

  • Workload surge — Demand outruns staffing; per-case attention shrinks and review becomes triage.
  • Reviewer bottleneck — One fixed-capacity checking stage sits between AI output and consequence; everything queues behind it.
  • Vendor opacity — The deploying institution cannot inspect the model, data, or update pipeline it is accountable for.
  • Data & policy drift — The world, the intake process, and the rules change under a system trained on how things used to be — two mechanisms with different remedies: the statistical properties of what the system processes move (concept drift), or the mixture of inputs arriving in deployment differs from the mixture it was trained on (covariate shift).
  • Deadline pressure — Statutory or managerial timeliness rules reward fast approval of machine output over slow disagreement.

All of them in context on the Clinical decision support & deterioration alerting domain page.

Levers available here and the patterns behind them

Documented case histories