The first prior authorization equity reports from Medicare Advantage plans show near-identical rates for enrollees with and without social risk factors. That is the headline finding across the reports we read, and it is accurate as far as it goes. It does not go very far. The metrics CMS required pool every service a plan reviews into one plan-level number. Some reports give percentages without the counts behind them, and plans were left to choose their own denominators for at least one metric. A report built that way can show near-identical rates even when a particular service, or a particular group of patients, is treated very differently.
The reports exist because of the contract year 2025 Medicare Advantage final rule, which CMS issued in April 2024. It required each plan’s utilization management committee to include a member with expertise in health equity and to conduct an annual, plan-level health equity analysis of prior authorization, with the results posted on the plan’s website. The first analyses, covering 2024, were due in 2025. In June 2025, CMS said it would not enforce the requirement. In November 2025, it proposed to remove it.
This analysis follows our January piece on utilization management committee rules, which said the equity analyses deserved a closer look. We read the posted reports of 3 plan sponsors to see what the first round of public data can and cannot tell plans and regulators.
Key takeaways
- The reported gaps are small. In the reports we reviewed, standard approval rates for enrollees with and without social risk factors were within about 1 percentage point of each other.
- The design hides more than it shows. Every metric is aggregated across all items and services, so a plan that denies one service far more often for one group could still report near-identical overall rates.
- The same line item means different things. Blue Advantage reported “approved after appeal” as a share of all requests, while WellSense reported it as a share of appealed decisions. The 2 numbers cannot be compared.
- The largest gap we found is in the least-noticed row. In one report, the share of denials later approved on appeal appears to be about 6 times higher for enrollees without social risk factors. The report does not say why.
- Small groups produce noise, not findings. Mass General Brigham’s report included groups of 1 or 2 members, with no caution about what those numbers can support.
- The requirement may not survive 2026. CMS stopped enforcing it in June 2025 and proposed removing it in the contract year 2027 rule. Comments closed January 26, 2026.
What the rule required
The contract year 2025 final rule built on the utilization management committee requirements CMS had adopted a year earlier. Its fact sheet described the goal as a “plan-level annual health equity analysis of prior authorization policies and procedures” meant to identify disproportionate effects on enrollees with certain social risk factors.
The rule made 2 changes to the committee. Starting in January 2025, each committee had to include at least one member with expertise in health equity, under 42 CFR § 422.137(c)(5). And under § 422.137(d)(6), the committee had to conduct an annual health equity analysis of the use of prior authorization, with the final report approved by that member before it was posted.
The analysis compares enrollees with at least 1 of 2 social risk factors against enrollees with neither.
- Receipt of the Part D low-income subsidy, or dual eligibility for Medicare and Medicaid
- Disability, determined from the original reason for Medicare entitlement
For each group, the plan reports a fixed set of metrics from the prior year.
| Metric | Applies to |
|---|---|
| Percentage of requests approved | Standard and expedited requests, reported separately |
| Percentage of requests denied | Standard and expedited requests, reported separately |
| Percentage of requests approved after appeal | Standard requests |
| Percentage of requests approved after the review timeframe was extended | All requests |
| Average and median time from request to decision | Standard and expedited requests, reported separately |
From 42 CFR § 422.137(d)(6) as adopted in the contract year 2025 final rule, and the metrics as plans posted them.
These are the same metrics CMS finalized in January 2024 for public prior authorization reporting under its interoperability and prior authorization rule. Like those metrics, they are aggregated for all items and services. That choice made the reports easy to produce from data plans would have to assemble anyway. It also built the main limitation into every report.
Plans had to post results by July 1 of each year, starting in 2025, in a manner the rule’s preamble described as “easily accessible and without barriers.”
What the first reports show
We reviewed reports posted by 3 sponsors. Blue Advantage, the Medicare Advantage plan of Blue Cross and Blue Shield of Louisiana, posted a single plan-level table. WellSense Health Plan posted a report covering its New Hampshire Medicare Advantage product and its Senior Care Options product in Massachusetts. Mass General Brigham Health Plan posted tables for 3 Medicare Advantage products, splitting enrollees with social risk factors into 3 groups. All 3 cover calendar year 2024.
We chose these reports because they were posted in full and together illustrate each problem discussed below. They are not a representative sample, and we did not contact the plans.
The table shows standard prior authorization requests in 2024.
| Plan | Without social risk factors | With social risk factors |
|---|---|---|
| Blue Advantage (Louisiana), approved | 98.78% | 98.68% |
| Blue Advantage (Louisiana), denied | 1.19% | 1.32% |
| WellSense New Hampshire, approved | 97.96% (96 of 98) | 98.07% (457 of 466) |
| WellSense New Hampshire, denied | 2.04% (2 of 98) | 1.93% (9 of 466) |
| Blue Advantage (Louisiana), median days to decision | 0 | 0 |
| WellSense New Hampshire, median days to decision | 8 | 8 |
From the plan postings linked above. Blue Advantage did not publish counts.
Mass General Brigham split its results differently. In its PPO product, for example, the standard approval rate was 97.6% both for enrollees with disabilities and for enrollees without social risk factors. Low-income enrollees came in at 98.1%. Its report concluded that the plan “has not identified any disparities among the member categories outlined above by CMS.”
WellSense reached a similar conclusion for its New Hampshire product, finding that “no significant disparities exist in the approval rates or processing times of standard prior authorization requests.” Blue Advantage posted its table without a narrative conclusion.
On their face, these are reassuring numbers. Approval rates near 98% for both groups, and differences in the tenths of a percentage point, are what a plan would hope to see. The rest of this analysis explains why the numbers cannot carry that much weight.
Why the rates look the same
The first reason is arithmetic. When about 98% of requests are approved, any difference between groups has to fit inside the remaining 2%. Blue Advantage’s expedited denial rate was 1.84% for enrollees without social risk factors and 2.39% for enrollees with them. That is a relative difference of about 30%, and it appears on the page as a gap of about half a percentage point. Whether it reflects a real difference depends on counts the posting does not include. By our calculation, each group would need roughly 5,300 expedited decisions for a gap that size to be statistically significant at the usual 95% level. For the standard denial gap of 1.19% against 1.32%, each group would need roughly 56,000 decisions.
The second reason is pooling. A plan’s prior authorization requests mix very different services, from advanced imaging to skilled nursing stays. Denial rates differ widely among them. When the metrics are aggregated for all items and services, a gap in one service can disappear into the total, especially if the 2 groups request different mixes of services.
A simple example shows how. Suppose a plan reviews 2 services. Service A is high volume and nearly always approved. Service B is rarer and denied more often, and the plan approves it less often for enrollees with social risk factors.
| Illustration | Without social risk factors | With social risk factors |
|---|---|---|
| Service A requests | 9,000 | 9,500 |
| Service A approval rate | 99% | 99% |
| Service B requests | 1,000 | 500 |
| Service B approval rate | 90% | 80% |
| Pooled approval rate | 98.10% | 98.05% |
A Redlake illustration, not plan data.
In this example, enrollees with social risk factors are denied service B twice as often, and the pooled report shows a difference of 0.05 percentage points. Nothing in the required metrics would reveal it.
The third reason is what the metrics leave out. They count decisions on requests that were submitted. They do not count requests that were never submitted because an office expected a denial, or care the patient went without. If prior authorization falls harder on some enrollees, much of that burden may sit outside the decisions the report measures.
A report that pools every service into one rate can only find a disparity large enough to survive the pooling.
Where the numbers break down
The pooled design is not the only problem. The first reports also show what happens when a metric is defined loosely and when groups are cut too fine.
Appeals measured 2 ways
The rule asks for “the percentage of standard prior authorization requests that were approved after appeal.” Plans read that line in at least 2 ways.
Blue Advantage reported 0.16% for enrollees without social risk factors and 0.03% for enrollees with them. Those figures appear to be shares of all standard requests. WellSense reported 100% for its New Hampshire enrollees without social risk factors and 88.24% for those with them, based on 2 and 17 appeal decisions. Those figures are shares of appealed decisions.
Both readings are defensible, and they produce numbers that cannot be compared across plans. A reader looking at both reports side by side would see 0.16% in one and 100% in the other for the same line item.
The Blue Advantage figures also contain the largest relative gap we found in any of the reports. If both of its appeal figures share the same base as its denial figures, then about 1 in 7 denied requests for enrollees without social risk factors was later approved on appeal, compared with about 1 in 44 for enrollees with them. That is roughly a 6-fold difference. It could mean that enrollees with social risk factors appeal less often or win less often when they do. It could also mean that the 2 groups face denials for different services. The posting offers no explanation and no counts, so none of these can be ruled out.
The question matters because appeals are where many denials are corrected. KFF’s analysis of 2024 data found that only 11.5% of denied requests in Medicare Advantage were appealed, and that 80.7% of appeals were partly or fully overturned. When most appeals succeed but few denials are appealed, the share of each group that appeals is one of the most important equity measures a plan could report. The required metrics do not ask for it.
Groups too small to read
Mass General Brigham’s report shows the opposite problem from pooling. It splits enrollees into groups so small that the numbers describe individuals rather than patterns. Its dual-eligible groups contain no more than 2 members in any of its 3 products. In one product, the dual-eligible group shows 0% approved and 0% denied, which suggests no requests at all. In another, a single dual-eligible member accounts for an average decision time of 13 days.
The report draws its conclusion of no disparities across all of these groups without noting that some are too small to support any conclusion. Decision times are also reported in whole days, which makes the timing comparison coarse. In one product, the reported average for one group is 0 days while the median is 1 day, a combination that can only come from rounding or truncation.
Small cells also raise a privacy question. CMS’s own public data releases generally suppress cells covering fewer than 11 people. A posted figure that describes the prior authorization experience of one member of a named plan is not much of an aggregate.
WellSense’s Senior Care Options report shows a third structural gap. Every enrollee in that product meets the social risk criteria, so there is no comparison group. WellSense compared the product with its New Hampshire plan instead, which is a different program in a different state. The rule’s design assumes each plan has both groups. Integrated plans for dually eligible enrollees often do not.
The enforcement pause and the proposed repeal
The requirement had a short life as an enforced rule. According to Georgetown University’s Center on Health Insurance Reforms, CMS indicated in June 2025 that it would not enforce the previously finalized requirements for health equity expertise on utilization management committees and for the public, plan-level equity reports. Some plans, including the 3 we reviewed, posted analyses anyway. We do not know how many plans posted, and we could not find a public inventory of the reports.
In November 2025, CMS released its proposed rule for contract year 2027. As summarized by the American Society for Radiation Oncology, the proposal “rescinds the annual health equity analysis of utilization management (UM) policies and procedures.” The law firm Hooper Lundy described the change as eliminating the health equity requirements for the committees, including the yearly posted analyses and the health equity expert. Comments were due January 26, 2026. CMS has issued its annual Medicare Advantage final rule in early April in each of the last 2 years.
The first reports give each side of this debate something to point to. Those who see the analysis as reporting for its own sake can point to pages of near-identical percentages that led no plan to change anything we could see. Those who want it kept can point to the design choices that made near-identical percentages likely, and argue that the fix is better metrics rather than none.
What a useful equity analysis would measure
Whatever CMS decides, plans can run a better analysis internally, and the same design would serve a public requirement if one returns. Based on what the first reports missed, it would include 6 elements.
- Service-level results. Approval and denial rates by group for the services with the most requests and the highest denial rates, not only a single pooled rate.
- Counts and uncertainty. The number of decisions behind every percentage, with confidence intervals. Cells too small to report should be suppressed.
- Appeal rates. The share of denials appealed by each group, and the share of appeals overturned, each with a stated denominator.
- Denial reasons. Whether denials rested on medical necessity or on missing documentation. Administrative denials can fall harder on enrollees whose providers have fewer staff to answer requests for information.
- Timing that means something. Decision times measured in hours, from first receipt and from receipt of a complete request, so a plan can see where requests wait.
- A common format. The same definitions across plans, so that a reader can compare one plan with another, which the first reports did not allow.
Most of this needs no new data collection. A plan that makes decisions in software already records each decision with its reason and time stamps, along with any appeal that followed. The analysis is a query, not a project.
Where Redlake stands
We build tools health plans use in utilization review, and 3 rules guide what we build. A clinician makes every denial. A machine can say yes, but only a person can say no. Every recommendation shows its reasons, in the form of the criteria applied and the facts from the record that met or missed them. And every decision can be reconstructed from its log.
The third rule is what makes a real equity analysis cheap. If each decision records its criteria and record facts along with the reviewer and the timing, a plan can break its results down by service or enrollee group whenever it wants. It does not have to wait for a regulation to define the table.
We think plans should keep running this analysis whatever happens to the public requirement. Medicare Advantage plans remain subject to federal nondiscrimination law, and as we wrote earlier this month, Section 1557 now requires covered entities to identify and mitigate discrimination risk in the decision support tools they use. A plan that cannot say whether its own prior authorization process treats groups differently is in a weak position to make that showing.
We also think the first round of reports deserves a fair reading. The plans that posted them did what the rule asked, and in some cases more. The limitations came mostly from the metrics they were given.
What to watch in 2026
- The contract year 2027 final rule. Whether CMS finalizes the removal of the equity analysis and the health equity expert, likely in the spring.
- A second round of postings. Whether any plans post 2025 analyses by July 1 regardless of enforcement, and whether they change the format.
- The first public prior authorization metrics. Under CMS’s interoperability and prior authorization rule, impacted payers must post their first annual prior authorization metrics by March 31, 2026. These use the same aggregated measures, for all enrollees, without any comparison between groups.
- Section 1557 enforcement. How the Office for Civil Rights applies the decision support tool provision to utilization management, which would bring the equity question back through a different door.
About this analysis
Basis. We read the posted 2024 health equity analyses of 3 Medicare Advantage sponsors and the regulation as codified in the 2024 edition of the Code of Federal Regulations. We also used CMS fact sheets for the 2 rules discussed here, along with published summaries of the enforcement pause and the contract year 2027 proposal.
Data notes. Percentages are as the plans posted them. Everything else, including the relative differences and the sample sizes, is our calculation. The sample sizes assume 2 groups of equal size and a 2-proportion test at 95% confidence. The service A and service B table is an illustration, not plan data. The 3 reports are not a representative sample.
Sources. CMS, contract year 2025 Medicare Advantage and Part D final rule fact sheet (April 2024). 42 CFR § 422.137, 2024 edition. CMS, interoperability and prior authorization final rule fact sheet (January 2024). Blue Cross and Blue Shield of Louisiana, Blue Advantage annual health equity analysis. WellSense Health Plan, Health Equity Analysis of Prior Authorizations 2025. Mass General Brigham Health Plan, utilization management health equity report (June 2025). Georgetown CHIR (November 2025). ASTRO (December 2025). Hooper Lundy (December 2025). KFF (January 2026).
Frequently asked questions
What is a Medicare Advantage health equity analysis of prior authorization?
It is an annual report, required by the contract year 2025 final rule, in which a plan’s utilization management committee compares prior authorization outcomes for enrollees with social risk factors against outcomes for other enrollees. The plan posts the results on its website.
Which enrollees count as having social risk factors?
Enrollees who receive the Part D low-income subsidy or are dually eligible for Medicare and Medicaid count as having a social risk factor. So do enrollees with a disability, as determined from their original reason for Medicare entitlement.
Is the requirement still in effect?
It remains in the regulation, but CMS said in June 2025 that it would not enforce it. The contract year 2027 proposed rule would remove it, and a final decision is expected in the 2027 final rule.
Why do the first reports show such similar rates?
Mostly because of design. Approval rates near 98% leave little room for visible gaps, and every metric pools all services together. The metrics also leave out appeal rates and requests that were never submitted.