Mental Health Parity Rule Paused: The Actuarial Liability Gap

Federal regulators suspended enforcement of the 2024 mental health parity rule's toughest provisions in May 2025, and on March 30, 2026 told a federal court they intend to replace the rule entirely by December 31, 2026. That leaves actuaries pricing against a baseline where 100% of the 210 comparative analyses the Department of Labor reviewed in its last report to Congress failed on first submission (EBSA, 2023 Report to Congress).

What the Pause Actually Suspends, and What Keeps Applying

The 2024 final rule amended the Mental Health Parity and Addiction Equity Act's implementing regulations to add four things that did not exist before: a formal fiduciary certification requirement for the comparative analysis, a fixed response timeline once a regulator or participant requests one, a "meaningful benefit" standard requiring parity in outcomes rather than just design, and a prohibition on "discriminatory factors and evidentiary standards" used to build the analysis itself. On May 15, 2025, the Departments of Labor, Health and Human Services, and Treasury said they would not enforce those four additions "prior to a final decision in the litigation," plus an 18-month tail after that decision (DOL, Statement Regarding Enforcement of MHPAEA, May 2025). On March 30, 2026, the same three departments told the D.C. federal court hearing the ERISA Industry Committee's challenge that they intend to go further: rather than defend the 2024 rule at all, they will issue an entirely new proposed rule by December 31, 2026, with "significant revisions" to the provisions ERIC challenged (court filing, reported by ERISA Litigation & Compliance, April 2026). The case has been stayed since May 2025 with 90-day status reports, and the next one is due September 30, 2026.

What did not pause is the statutory core: the 2013 MHPAEA regulations and the Consolidated Appropriations Act, 2021 requirement that group health plans perform a comparative analysis of non-quantitative treatment limitations and produce it on request. A plan still has to demonstrate, in writing, that any process, strategy, or evidentiary standard used to limit mental health and substance use disorder benefits is comparably applied to medical and surgical benefits. The pause removes the newest layer of proof and the deadline pressure to produce it fast; it does not remove the underlying parity obligation. That distinction is easy to lose in trade coverage that treats the pause as a blanket compliance holiday. It is not one, and the deficiency data below shows why the gap between "not currently enforced" and "not owed" is where the reserving exposure sits.

The Deficiency-Rate Baseline as an Exposure Proxy

Health actuaries do not have a public loss-triangle equivalent for parity violations, but EBSA's own audit history is a workable proxy. Between February 2021 and July 2023, the department issued 199 initial letters requesting comparative analyses, covering more than 480 distinct NQTLs, and followed with 183 insufficiency letters addressing more than 330 of those NQTLs (DOL data, cited in Oliver Wyman's March 2026 compliance brief). Its 2023 Report to Congress found that all 210 comparative analyses it reviewed that cycle failed to demonstrate compliance on first submission, EBSA's fourth consecutive report finding a materially deficient starting rate (EBSA, MHPAEA Comparative Analysis Report to Congress, 2023). Most of those deficiencies were eventually cured through follow-up documentation rather than penalty, but curing a deficiency after the fact is exactly the remediation cost a reserving actuary should be modeling now, not discovering when a request letter arrives.

The department's more recent enforcement has moved from letters to formal findings. As of its March 2026 update, DOL had issued final noncompliance determinations to 5 plans covering 7 NQTLs, and CMS had issued its own determinations to 6 plans covering 10 NQTLs (Oliver Wyman, "Why Insurers Must Act on Mental Health Parity Now," March 2026). EBSA holds primary enforcement jurisdiction over roughly 2.5 million private, employer-sponsored group health plans covering about 133 million Americans, and it now devotes close to 25% of its total enforcement program to MHPAEA NQTL work (DOL, EBSA MHPAEA enforcement fact sheet). None of that enforcement infrastructure went away with the May 2025 pause; it simply stopped being pointed at the 2024 rule's newest requirements. The examiners, the request letters, and the deficiency rate are all still live against the 2013 baseline, and they will be pointed at whatever the December 2026 rule adds back on top of it.

Reserving for a Rule That Isn't Being Enforced Yet

This is the part of the story trade press covering the pause as an HR and compliance question has not framed as a reserving problem, and it should be. A health actuary setting incurred-but-not-reported reserves treats claims that occurred but have not yet been reported as a known, quantifiable liability, because loss development patterns are stable and well understood. A comparative-analysis deficiency identified today, but not yet the subject of a request letter, a state market conduct exam, or a plan-participant complaint, behaves the same way: it is a liability that already exists in fact, is quantifiable in principle from the deficiency-rate data above, and simply has not yet been reported through an enforcement channel. Call it a regulatory contingent liability rather than IBNR in the technical sense, since ASC 450 and its insurance-reserving equivalents govern the accounting treatment, but the actuarial logic of estimating an unreported liability from a known reporting-lag pattern is the same discipline health actuaries already apply to claims.

The mechanism that makes this urgent rather than theoretical is retroactivity. Georgia's $25 million in January 2026 fines against 11 insurers, led by a $10.2 million penalty against Oscar Health Plan of Georgia and a $4.6 million penalty against Blue Cross Blue Shield Healthcare Plan of Georgia, stemmed from market conduct examinations of how those insurers administered mental health benefits during calendar year 2022 (Georgia Office of Insurance and Safety Fire Commissioner, January 2026). A four-year gap separates the conduct year from the fine. If the December 2026 federal rule reinstates a version of the meaningful-benefit standard or the fiduciary certification requirement, and DOL or CMS then examines plan years 2025 through 2027 against it, the exposure a plan is carrying today for those years does not disappear because enforcement was paused while the years were being lived. It sits on the books, unpriced, until an examiner's letter turns it into a number. Georgia's own enforcement history adds a second wrinkle worth flagging for reserve adequacy purposes: an 11Alive investigation found that, months after the fines were announced, none of the $25 million had actually been collected, a reminder that headline penalty figures and realized cash liability are not the same thing when building a remediation reserve (11Alive, 2026).

Network Adequacy and Reimbursement: The Two NQTLs Carrying the Most Risk

Regulators have not treated all NQTLs as equally risky, and neither should a pricing actuary allocating scarce compliance-review time. DOL and CMS enforcement activity concentrates on network adequacy and composition standards, and on the reimbursement methodology used to set out-of-network payment rates for behavioral health providers, because those two NQTLs have the most direct effect on whether a plan participant can actually access care (DOL/CMS enforcement patterns, cited in Oliver Wyman, March 2026). If a plan's own claims data shows materially higher out-of-network utilization for mental health and substance use disorder services than for medical and surgical services, that gap is itself evidence of a potential violation, and the plan must show the disparity is not the product of a network that is simply thinner for behavioral health providers or a reimbursement schedule that pays them less relative to billed charges than it pays medical providers.

For pricing actuaries, that finding has a direct line to two assumptions already sitting in every behavioral health rate filing: the credibility-weighted trend applied to out-of-network behavioral health claims, and the provider reimbursement schedule feeding network-adequacy modeling. A plan that has been pricing to a reimbursement gap between behavioral health and medical providers, even one justified on ordinary market or specialty-mix grounds, is pricing to a methodology regulators have specifically flagged as the likeliest NQTL to fail a comparative analysis. That is a pricing assumption with a compliance shelf life, and the December 2026 rule is the event that could shorten it abruptly.

The State Patchwork Layered on Top

The federal pause did not create a single national baseline, because state insurance commissioners retain independent enforcement authority over fully insured business and several have moved to lock in the 2024 rule's protections regardless of what Washington does with the federal version. Washington state enacted legislation requiring insurers to comply with the 2024 federal rule as published, so that a future federal rescission or weakening does not automatically loosen the state-level standard (state legislative action, cited in Commonwealth Fund, 2026). Colorado took a related path through HB 25-1002, which requires nationally recognized clinical criteria for behavioral health coverage decisions independent of the federal rulemaking timeline. Georgia's enforcement, discussed above, runs on its own state parity statute layered on top of MHPAEA and shows no sign of pausing while the federal rule is rewritten.

The result for a multistate carrier is a compliance and reserving patchwork rather than a single national exposure estimate. A plan operating in Washington or Colorado should treat the 2024 rule's meaningful-benefit and fiduciary-certification standards as live obligations today, federal enforcement pause notwithstanding, while the same carrier's business in a state with no independent parity statute is genuinely operating under the relaxed 2013-era standard until December 2026. Building one reserve estimate for "MHPAEA exposure" across a multistate book will misstate the liability in both directions: understating it in states that have already locked in the tougher standard, and overstating it in states waiting on the federal rule to reset expectations.

What Health Actuaries Should Build Before December

Watching enforcement pauses resolve into retroactive liability is a familiar pattern, most recently in the ACA risk-adjustment cleanups that followed years of relaxed data-validation enforcement; relaxed enforcement rarely makes exposure disappear, it typically lets it compound until the next rule change forces a settling of accounts. The defensible response between now and the December 2026 proposed rule is not a full remediation program across every NQTL a plan uses. It is a single, genuinely defensible comparative analysis built now for whichever NQTL carries the most claims-data risk, almost certainly out-of-network behavioral health reimbursement or network adequacy given where DOL and CMS enforcement has concentrated. That analysis becomes the template the rest of the NQTL inventory follows once the new rule's actual requirements are known, and it gives a reserving actuary one real, tested data point, rather than the EBSA-wide 100% deficiency rate, to anchor a plan-specific liability estimate. Carriers that wait for the December 2026 rule to start that work will be building both the analysis and the reserve estimate against a deadline instead of a runway.

Why This Matters

The gap between today's paused enforcement and a December 2026 rule that regulators have already signaled will restore many of the same requirements is not a compliance footnote; it is an unpriced liability sitting on carrier books right now, with EBSA's own audit history showing almost every plan examined has something to find. Health actuaries who treat the interim period as a genuine holiday, rather than as a reporting lag on a liability that statute still requires, will be reserving for the 2027 rule reset the same way plans are reserving for 2022 conduct in Georgia today: after an examiner's letter has already turned an estimate into a fine.