Post-acute care is where MA review is hardest to get right

Post-acute care is where Medicare Advantage review is hardest to get right. The decisions come at a vulnerable moment, as a patient leaves the hospital and needs skilled nursing or rehabilitation. They depend on judgments about function and recovery that are harder to read from a chart than a diagnosis or a lab value, and they are made while a hospital bed and a discharge plan wait.

The evidence suggests that the hardest decisions are also among the most often denied. A 2024 investigation by the Senate Permanent Subcommittee on Investigations found that UnitedHealthcare’s prior authorization denial rate for post-acute care rose from 10.9% in 2020 to 22.7% in 2022, and that in 2022 Humana’s post-acute denial rate was 16 times its overall denial rate. CMS has since tightened the rules, and a federal court has let a lawsuit over an algorithm used in post-acute decisions proceed into broad discovery.

Key takeaways

• Post-acute denials stand out. The Senate investigation found post-acute denial rates in 2022 several times higher than overall denial rates at the 3 largest Medicare Advantage insurers.
• Traditional Medicare criteria apply. Since 2024, Medicare Advantage plans must follow traditional Medicare’s coverage criteria for post-acute settings, and may not substitute stricter internal criteria where those exist.
• Predictions cannot end coverage. CMS has said that software may predict how long a patient will need care, but that a prediction cannot be the basis for terminating it. The patient’s condition must be assessed.
• Concurrent decisions carry appeal rights. Since January 2026, decisions made during a stay are organization determinations with full appeal rights.
• The algorithm lawsuit is moving. In March 2026, a federal court ordered broad, class-wide discovery in the case over UnitedHealth’s use of a post-acute prediction tool.
• Good review is individual and fast. Decisions should rest on current function and the treating team’s plan, with a conversation offered before any denial.

How post-acute review works and why it is hard

Post-acute care covers several settings with different rules in traditional Medicare.

The table summarizes traditional Medicare coverage standards. Details vary by setting and are set in regulation and manuals.

Medicare Advantage plans typically review post-acute care at 2 points. Before admission, a plan decides whether to authorize the setting the hospital team recommends. During the stay, the plan reviews progress at intervals and decides whether continued care remains covered. When a plan decides that covered skilled nursing or home health services should end, the enrollee must receive a written notice before services stop and can request a fast review by an independent quality improvement organization.

Each of those decisions is harder than it looks. Eligibility turns on function rather than diagnosis, such as whether a patient can walk safely or tolerate 3 hours of therapy. Function changes day to day and is often documented in therapy notes that use their own scales and vocabulary. The hospital’s discharge team has a view of what the patient needs, and the facility’s team may see it differently after admission. The plan’s reviewer sees both only through documents.

Decisions are made against discharge timelines, so delay has its own cost. A patient who stays in the hospital waiting for an authorization is occupying a bed and losing time in recovery.

The stakes are also personal. A patient sent home too soon may fall or lose function that more rehabilitation would have restored. A patient kept in a facility longer than needed loses time and independence in a different way. Neither error is visible on the day of the decision.

What the evidence shows

The Senate subcommittee’s investigation, released in October 2024, examined prior authorization at the 3 largest Medicare Advantage insurers. Its findings on post-acute care drew the most attention.

Compiled from the Senate Permanent Subcommittee on Investigations majority staff report and release (October 2024) and Fierce Healthcare’s summary.

The report also described CVS’s use of AI beginning in 2021 to reduce spending in skilled nursing facilities, with a savings estimate that rose to $77 million over 3 years, and noted that Humana had contracted with naviHealth, a company that offered post-acute prediction tools, since 2017. Among its recommendations, it asked CMS to collect prior authorization data by service category and to audit plans when denial rates spike in particular areas.

The investigation’s findings build on earlier work. The HHS Office of Inspector General’s 2022 review of Medicare Advantage denials found that 13% of the prior authorization denials it sampled met Medicare coverage rules and would likely have been approved in traditional Medicare, and stays in post-acute facilities were among the examples it described.

The investigators’ data were from 2020 to 2022, before CMS’s new rules took effect, and the insurers disputed the report’s characterizations. The pattern it documented is nonetheless the backdrop for the rules that followed.

What the rules require

CMS has addressed post-acute review in several steps since 2023.

Compiled from CMS fact sheets and guidance described in the text.

The 2024 rule closed a gap that had allowed plans to apply their own criteria to post-acute admissions that traditional Medicare would cover. A plan may not, for example, deny an inpatient rehabilitation admission using a stricter internal standard when traditional Medicare’s criteria for that setting are established.

CMS’s guidance on algorithms is the most specific federal statement on post-acute prediction tools. According to Norton Rose Fulbright’s summary of CMS’s February 2024 questions and answers, CMS said that plans may use algorithms and AI, but coverage decisions must reflect the individual patient’s situation. Algorithms relying on larger data sets rather than the individual’s circumstances would not comply. On post-acute care specifically, CMS said software tools may predict how long a patient may stay, but the prediction cannot be the predicate for terminating coverage. The patient’s condition must be assessed before a termination notice is issued.

The contract year 2026 rule, which took effect in January, matters for concurrent review. As we described in April, a decision made while an enrollee is receiving services is now explicitly an organization determination, which carries notice and appeal rights. A plan that decides mid-stay that continued rehabilitation is not covered has made a decision the enrollee can appeal through the full process.

The algorithm question

The most prominent legal challenge to post-acute review involves a prediction tool. In Estate of Lokken v. UnitedHealth Group, plaintiffs allege that UnitedHealthcare used naviHealth’s nH Predict tool to supplant physicians’ clinical judgment and cut off post-acute coverage prematurely. Optum, which owns naviHealth, has said the tool is a care-support tool that is not used to make coverage determinations.

The case has moved forward. In February 2025, the court dismissed several claims as preempted by Medicare Advantage law but allowed claims for breach of contract and breach of the implied covenant of good faith and fair dealing to proceed. In March 2026, according to ArentFox Schiff, the court largely granted the plaintiffs’ motion to compel discovery. It ordered production of documents on post-acute care policies and nH Predict, including internal AI review board materials, with some categories reaching back to 2017. The court also rejected the company’s request to stage discovery and ordered class-wide discovery from the outset.

The legal outcome is open. The policy question is already settled in CMS’s guidance, which treats a model’s estimate of how long a typical patient needs care as something other than an assessment of the patient in the bed. Models can still help. A prediction can flag a stay for review or help a care manager plan a discharge. What it cannot do is substitute for a look at the patient’s current function and the treating team’s assessment.

State law is moving in the same direction. As we described in March, California requires that a physician or other qualified professional make medical necessity determinations when plans use AI in utilization review, and that tools base determinations on the enrollee’s own history rather than solely on group data.

A prediction of how long a typical patient needs care is not an assessment of the patient in the bed.

What good post-acute review looks like

The rules set a floor. Plans that want post-acute review to be both defensible and fair can go further with practices that are well within reach.

• Use traditional Medicare’s criteria, visibly. Reviews should cite the specific coverage standard for the setting and show which facts in the record meet or miss it. Internal criteria, where they apply, should be published and tied to evidence.
• Assess current function. Decisions should rest on documented, recent functional status and the treating team’s plan, not on how long patients with similar diagnoses usually stay.
• Talk before denying. When a reviewer expects to deny an admission or end a stay, the treating physician or facility team should be offered a conversation first, as we argued in March. Post-acute cases are where that conversation most often changes the outcome.
• Decide fast at discharge. Requests that hold up a hospital discharge should be treated as expedited in practice, with decisions in hours rather than days.
• Plan reviews, not cutoffs. Concurrent reviews should be scheduled around expected progress, with clear goals that would justify continuing or ending care, communicated to the facility and the family.
• Watch the independent reviews. Quality improvement organization decisions on termination appeals and independent review entity decisions on reconsiderations are an outside check. A plan whose post-acute decisions are frequently reversed should treat the pattern as a signal about its process.
• Measure by setting. Denial and overturn rates should be tracked separately for each post-acute setting, which is the service-level view the Senate report recommended.

Where Redlake stands

We build tools health plans use in utilization review, and post-acute care is where our rules matter most. A clinician makes every denial. A machine can say yes, but only a person can say no. When software sees that a request for skilled nursing or rehabilitation clearly meets the criteria, it can recommend approval and the plan can approve quickly, which is the fastest way to get a patient out of a hospital bed and into the right setting. When the answer may be no, a clinician decides, with the full picture.

Every recommendation shows its reasons, in the form of the coverage criteria for the setting and the facts in the record, including functional assessments and therapy notes, that met or missed them. That is what allows a reviewer to check the basis rather than accept a score, and what lets a family or an appeals reviewer understand a decision. And every decision can be reconstructed, which in post-acute care includes what the plan knew about the patient’s function on the day it decided.

We do not think prediction has no place. Estimates of expected recovery can help a plan schedule reviews and help care managers plan ahead. But they belong in the planning, not in the decision. The decision belongs to a person looking at the patient’s current condition.

What to watch in 2026

• The Lokken case. What discovery shows about how post-acute prediction tools were used, and whether the court certifies a class.
• Service-level data. Whether CMS acts on the Senate report’s recommendation to collect prior authorization data by service category, and what the first public prior authorization metrics, posted in March, reveal about post-acute care where plans broke them out.
• Concurrent review under the new rules. How plans adapt mid-stay reviews now that those decisions carry full appeal rights.
• Program audits. Whether CMS audits focus on post-acute determinations and terminations, and what findings they produce.
• State action. Whether more states adopt rules on AI in utilization review or on post-acute authorizations specifically.

About this analysis

Basis. The evidence section draws on the Senate Permanent Subcommittee on Investigations’ October 2024 release and the HHS Office of Inspector General’s 2022 review of Medicare Advantage denials. The rules section draws on CMS fact sheets and guidance, and the litigation section on reporting about Estate of Lokken v. UnitedHealth Group.

Data notes. The Senate figures cover 2020 to 2022 and come from documents the insurers provided to the subcommittee. They predate the rules that took effect in 2024 and 2026, and the insurers disputed aspects of the report. The coverage standards table is a summary, not a statement of every requirement. The description of the lawsuit reflects allegations and court orders, not findings of liability.

Sources. Senate Permanent Subcommittee on Investigations (October 2024). Fierce Healthcare (October 2024). HHS Office of Inspector General, OEI-09-18-00260 (April 2022). Norton Rose Fulbright on CMS’s February 2024 guidance. CMS fact sheets for the contract year 2024 and 2026 rules. ArentFox Schiff (April 2026).

Frequently asked questions

Do Medicare Advantage plans have to follow traditional Medicare rules for post-acute care?

Yes. Since 2024, Medicare Advantage plans must follow traditional Medicare’s coverage criteria for post-acute settings such as skilled nursing and home health care, and may use internal criteria only where traditional Medicare’s criteria are not fully established.

Can a plan use an algorithm to decide how long post-acute care lasts?

A plan can use software to estimate how long a patient may need care, but CMS has said that the estimate cannot be the basis for terminating coverage. The plan must assess the individual patient’s condition before issuing a termination notice.

What can a patient do if a plan ends skilled nursing coverage?

The enrollee must receive written notice before covered services end and can request a fast review by an independent quality improvement organization. Decisions made during a stay are also organization determinations with the full appeals process.

What did the Senate investigation find?

It found that post-acute denial rates at the 3 largest Medicare Advantage insurers rose or ran well above their overall denial rates between 2020 and 2022, including a rise in UnitedHealthcare’s post-acute denial rate from 10.9% to 22.7%.