The 2024 mental health parity rule has stood unenforced for 15 months while 182 million people hold the private coverage it governs

What is happening?

The United States has a federal mental health parity rule that no federal agency is enforcing. On 2025-05-15 the Departments of Labor, Health and Human Services, and the Treasury jointly announced they would not enforce the 2024 rule implementing the Mental Health Parity and Addiction Equity Act. As of 2026-08-08 that announcement has stood for roughly 15 months, and the coverage the rule reaches is about 182.8 million people — 161,058,000 with employer-based insurance and 21,731,900 in the individual market, on KFF counts for 2024.

The rule was published at 89 FR 77,586 on 2024-09-23 and took effect 2024-11-22. It applied in two tranches. The general framework for nonquantitative treatment limitations, together with the specified contents of the comparative analyses plans must produce, applied from the first plan year beginning on or after 2025-01-01. The meaningful benefits standard, the prohibition on discriminatory factors and evidentiary standards, and the relevant data evaluation requirement applied from the first plan year beginning on or after 2026-01-01. That second tranche is the one that would have made plans collect outcomes data on actual access and take action where the data showed material differences. It never applied to anyone.

The rule was never repealed and has never been held unlawful by any court. The ERISA Industry Committee sued on 2025-01-17 in the District of Columbia. The court stayed the case on 2025-05-12 at the request of the government, and the nonenforcement statement followed three days later. On 2026-03-30 the Departments told the court they would not defend the rule at all and would instead issue a new proposed rule carrying anticipated significant revisions to the challenged provisions, with a notice of proposed rulemaking no later than 2026-12-31. As of 2026-08-08 no such notice has published, and the next joint status report is due on or before 2026-09-30.

Whose problem is this?

The body that announced it would not enforce is the same body that wrote the rule and published the evidence for it. Nothing in the parity statute obliges the Departments to enforce any particular regulation.

Where does this problem end?

The boundary matters here because the duty was never removed. What was removed is the check on whether anyone is meeting it.

What is the state now, and what should it be?

Now

Needs a new measurementthe target state: no source opened here names a target compliance rate, a baseline against which compliance would be measured, or a sampling frame for locating noncompliant plans. The Departments wrote in the rule itself that plans and issuers are not in full compliance although the extent across all plans and issuers is not known, and nothing opened here proposes how that extent would be learned. The nonenforcement statement names a litigation event as the endpoint, not any condition about parity.

How big is it?

Between 18,000,000 and 182,789,900 people, and the two ends answer different questions.

The high bound is an exposure ceiling. It is everyone holding private health coverage that carries a federal parity duty the enforcing agencies have announced they will not test. On KFF counts for 2024 that is 161,058,000 people with employer-based coverage plus 21,731,900 in the non-group market, giving 182,789,900 against a US total population of 331,160,300.

The low bound is not an exposure estimate at all. It is an outcome count. The 2025 MHPAEA Report to Congress records that EBSA enforcement work produced corrections affecting more than 18 million participants across more than 39,000 group health plans in the 24 months from 2023-08-01 to 2025-07-31. Cumulatively since 2021 the figure is over 23 million people across more than 77,000 plans and issuers. Read with its window, the low bound is what the enforcement machinery demonstrably delivered in one comparable two-year period, which makes it a floor on what a two-year pause costs. Neither number survives being stripped of that window. Eighteen million is not a count of people currently harmed. It is a count of people whose plan terms changed while someone was checking.

The high bound overcounts by an unknown amount, because the KFF employer figure does not separate self-funded employers with fewer than 50 employees, which are exempt from MHPAEA. For scale on the other side, the rule itself counted 46,080 self-funded plans and 2,129,516 ERISA-covered group health plans, and stated no total number of participants anywhere in the document.

Under what conditions does it arise?

1. The duty is comparative, and the comparison is invisible from outside. MHPAEA does not require any particular benefit. It requires that a nonquantitative treatment limitation — prior authorization, network admission criteria, medical necessity standards, reimbursement methodology — not be written or applied more restrictively for mental health and substance use care than for medical and surgical care. Whether that holds lives inside plan documents, vendor criteria and claims systems. A participant who is denied care sees the denial and never the comparison.

2. The only population-scale detector was the comparative analysis, and only the 2024 rule made it answerable. The 2021 amendments created the duty to perform and document the analysis without specifying what it must contain. The 2024 MHPAEA Report to Congress records the result: upon initial submission, every comparative analysis reviewed was in some way insufficient. The 2024 rule specified contents and added the relevant data evaluation requirement. That provision belonged to the tranche due 2026-01-01 and never applied.

3. Declining to defend a rule is a cheap administrative move, and defending it is not. An administrative challenge opens a standing option to seek abeyance, reconsider, and replace. The court granted abeyance three days after the government asked for it. Nonenforcement pending reconsideration sits inside enforcement discretion, which is largely unreviewable. The mechanism does not require anyone to want less parity. It only requires that defending cost more than replacing.

4. The stated endpoint presupposes an event the parties are no longer heading toward. Relief runs until a final decision in the litigation plus a further 18 months. The case has been stayed since 2025-05-12, and the filing of 2026-03-30 replaces merits litigation with a promised proposal. If a replacement issues and the challenger withdraws, there may never be a final decision, and the stated endpoint never arrives. No source opened here names an alternative trigger.

5. The beneficiaries are not parties and hold no equivalent instrument. The plaintiff is a trade association of large plan sponsors with a direct financial stake and clear standing. Participants have claim-by-claim remedies that are individual and retrospective. They cannot compel production of a comparative analysis at population scale, and they cannot generate the outcomes data the suspended tranche would have required. The state backstop is partial by construction, because federal law preempts state regulation of self-funded plans, and HHS encouraged states that are the primary enforcers for issuers to adopt a similar approach to enforcement.

What has been tried?

Two directions ran at once for four years, and then the federal half was withdrawn by the same agencies that had built it.

What was found?

Why is it still unsolved?

Enforcement absent — the obligation survives, the instrument survives in statute, and the three agencies holding the authority have announced they will not use it.

The distinction that carries this document is written into the rule itself. The Departments stated that the provisions added to the parity statute in 2021 are self-implementing and took effect on 2021-02-10, and that a delayed applicability date for the comparative analysis requirements would not alter what a plan or issuer owes under the statute. Plans must still perform and document comparative analyses. The 2013 rules remain enforceable. Nothing about what is owed changed on 2025-05-15. What changed is that the three agencies with the authority to demand those analyses said they would not act on failures to comply with the standards that made the analyses answerable in the first place.

The second movement is that the suspended part is precisely the measuring part. The 2021 amendments created a duty to produce a document without saying what the document had to contain, and the enforcement record shows what followed: on first submission, every comparative analysis reviewed was in some way insufficient. The 2024 rule answered that by specifying contents, and it went further with a relevant data evaluation requirement that would have made plans collect outcomes data on real access and take action where the data showed material differences. That requirement belonged to the tranche due on the first plan year beginning on or after 2026-01-01. It never applied to a single plan. So the pause does not merely suspend enforcement. It removes the one mechanism that would have produced evidence about how much enforcement was needed.

The third movement is the one that holds the state in place. The relief was announced as running until a final decision in the litigation plus a further 18 months, which reads like a bounded interval. But the litigation stopped moving because the government asked it to stop, and the filing of 2026-03-30 substitutes a promise to propose something else for a decision on the merits. If a replacement issues and the challenger withdraws, no final decision ever arrives and the clock never starts. Meanwhile the record produces no visible event. Reports to Congress continue, the rule stays in the code, comparative analyses remain legally required, and no enforcement fact sheet has yet covered a fiscal year in which the policy was in force. A pause that never announces its own ending is hard to tell apart from a repeal that was never voted on.

What observation would mean it is solved?

Candidates — (a) the Departments withdraw the nonenforcement statement and resume enforcement, or a replacement rule takes effect carrying equivalent testing requirements (b) the out-of-network utilization gap between behavioral health and medical care narrows toward parity (c) findings of noncompliance and participants covered by corrections return to the levels recorded before 2025-05-15.

(a) alone is weaker than it looks. The promised replacement carries anticipated significant revisions to exactly the provisions under challenge. A rule can arrive that keeps the name and drops the relevant data evaluation requirement, in which case enforcement resumes against a standard that no longer detects anything. The observation would read as solved while the detector stayed switched off.

(b) alone is worse. The out-of-network gap moves with network economics, workforce supply and reimbursement levels, none of which parity enforcement controls directly. RTI reports the disparities remained large over nine years and that provider shortages do not explain them, which means the ratio is stubborn against good enforcement and bad enforcement alike. A narrowing could also come from behavioral health clinicians leaving practice rather than from any of them joining networks.

(c) alone counts the enforcer rather than the harm. Determination letters and correction counts measure how hard an agency worked, not how many people obtained care. They also rest on a denominator nobody publishes, since the Departments state the extent of noncompliance is not known. A rise in findings is therefore compatible with better detection, worse compliance, or both at once. The three have to be read together, and (b) has to be read against the span of years it covers.

What is it connected to?

Fills with researchMedicaid managed care parity under its separate regime, ERISA preemption of state insurance regulation, the mental health workforce shortage that places 73.3 million people in designated shortage areas including 24.4 million in rural areas, network adequacy regulation, and the general litigation environment for federal agency rulemaking. Relation type and evidence grade were not confirmed in this round.

What these sources do not say

• Why the Departments decided not to defend the rule. The filing of 2026-03-30 states that they further evaluated the rule and the challenge and then reached that conclusion. No source opened here carries a reason, an internal analysis, or a named provision they concluded was indefensible.
• What the replacement will contain. Anticipated significant revisions is the whole of the description. Nothing opened here says whether the meaningful benefits standard, the relevant data evaluation requirement, or the specified contents of the comparative analyses survive in any form.
• How much noncompliance there actually is. The Departments state the extent across all plans and issuers is not known. No source opened here offers a compliance rate, a baseline, or a sampling frame, so the denominator for every enforcement count in this document is missing.
• What the comparative analyses say. CMS posts final determination letters, but none of the three agencies publishes the analyses themselves. The count of violating plans is therefore structurally unpublishable by anyone outside the Departments, including by the researchers whose work the rule cites.
• What happened to open investigations on 2025-05-15. The 2025 report gives totals for its 24-month period. No source opened here states how many in-flight investigations were closed, narrowed, or continued under the 2013 standard once the policy issued.
• Whether any plan changed course between 2024-11-22 and 2025-05-15. The rule was effective for nearly six months and applicable to calendar-year plans from 2025-01-01. Nothing opened here reports whether plans that had begun implementing it reversed that work.
• What the states did in aggregate. HHS asked states to match the federal approach. One trade account describes eight states moving in opposite directions, and nothing opened here counts how many of the fifty followed the request or how many insured lives sit under each posture.
• Where the after picture is. The enforcement fact sheet series stops at FY2023, so nothing published covers the fiscal years in which nonenforcement took effect. The statutorily annual report to Congress covered 24 months rather than 12, no source explains the merged period, and no 2026 report appears on the DOL index as of 2026-08-08.

See the evidence

Two primary documents were read end to end. The full text of the 2024 rule was obtained as the Federal Register PDF from govinfo and parsed locally, and the joint status report of 2026-03-30 was obtained as a PDF whose file header confirms the case number, document number and filing date, so it is the filing itself rather than a description of it. The complaint and the RTI study were likewise read as PDFs. Five rows carry r.jina.ai proxy URLs rather than canonical dol.gov and beckersbehavioralhealth.com addresses, because those hosts returned HTTP 403 to every direct fetch and the proxy URL is what was actually opened. Three rows carry no URL at all for the same class of reason, and they are kept rather than dropped because the absence is part of the record. Where sources overlap they agree: the three operative clauses of the nonenforcement statement appear identically in the DOL text and in an independent law-firm account, and the date of 2025-05-15 is fixed by that account and by the 2025 report to Congress against a proxy extraction that mis-dated the statement to the filing date of the suit. Where they disagree the disagreement is left visible rather than resolved. One secondary source dates the rule to 2024-09-09 while the Federal Register publication is 2024-09-23, which a pre-publication release would explain but which is recorded rather than reconciled. At least one secondary account restates the 2026-12-31 commitment as replacing the rule entirely by that date, whereas the filing promises only a notice of proposed rulemaking, which is the start of a replacement and not its completion. Several low-credibility sites assert specific state fines in 2026; the only trade publication opened here reports state actions and no dollar penalties, so the fine claims are not repeated. Two dates for the internal adoption of the policy circulate in a source that was opened but not carried into this table, and they are omitted because no primary source opened here corroborates them. The absence claim about a replacement rulemaking is bounded to one RIN, since a new rulemaking could carry a different one. This is a Path A output (research-based definition), so observation_refs is empty and provenance_mode: press-derived.

People affected

Estimated range 18,000,000–182,789,900 As of coverage counts 2024; enforcement window 2023-08-01 to 2025-07-31

Sensitivity The width is not uncertainty about one quantity. The two ends answer different questions and the true figure is not a point between them. The high bound, 161058000 plus 21731900 equal to 182789900, is the population whose federal parity right is now nominal rather than checked. The low bound, 18000000, is the population for whom a correction was actually obtained in the 24 months ending 2025-07-31; the cumulative figure since 2021 is over 23 million people across more than 77000 plans and issuers, and neither number may be quoted without its window, because stripped of it either reads as a snapshot of people currently harmed, which is not what it measures. What this range fails to count: Medicaid managed care enrollees, who fall under a separate parity regime untouched by this action, and Medicare beneficiaries, 49117800 on 2024 KFF counts, who were never subject to MHPAEA at all; people who never filed a claim because they expected a denial or never sought care, the largest silent category and one with no record to count; and the financial burden already borne by people who did obtain care out of network, since the underlying study reports rates rather than headcounts. The limit in the opposite direction is that the high bound overcounts. Self-funded employers with fewer than 50 employees are exempt and are not separable in the KFF figure, and some unknown share of plans would meet the 2024 standard without being prompted, since the Departments state that the extent of noncompliance across all plans and issuers is not known and no source opened here offers a compliance baseline.

Regional breakdown No source opened here gives state-level or regional counts of people covered by plans subject to the 2024 rule, nor state-level counts of participants reached by corrections. The enforcement reports give national totals only, and the coverage denominator was taken at the national level. Splitting either figure by state population share would be apportionment, which is not permitted and would in any case be wrong here, because ERISA preemption means the federal and state enforcement postures differ by plan funding type rather than by geography.