One state publishes Medicaid behavioral-health denial rates by service category, and it carries the first public ABA number: 9.31%, below the non-behavioral-health average in its own table
The question
Verbatim: "Does any state Medicaid agency, Medicaid MCO, or other state parity report publish behavioral-health denial rates BY SERVICE CATEGORY -- and does Virginia's pattern (SUD denials driven ~30% by preauthorization failure, not medical necessity) replicate in NY DFS, CA DMHC, IL or DE?"
Context: [[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]] found exactly one primary source, and it covers commercial fully-insured lives only (16 carriers, 2.48M lives), excluding Medicaid managed care and self-funded ERISA plans. The hard number we had was drawn from the payer line that matters least to an Applied Behavior Analysis (ABA) or substance use disorder (SUD) provider. Acronyms used throughout: ABA = Applied Behavior Analysis; SUD = substance use disorder; MH = mental health; BH = behavioral health; M/S = medical/surgical; MCO = managed care organization; PA = prior authorization; RCM = revenue cycle management; EOB = explanation of benefits; EQRO = External Quality Review Organization; MCPAR = Managed Care Program Annual Report; T-MSIS = Transformed Medicaid Statistical Information System; TAF = T-MSIS Analytic Files; CARC = claim adjustment reason code; ASAM = American Society of Addiction Medicine; DFS = New York Department of Financial Services; DMHC = California Department of Managed Health Care; HFS = Illinois Department of Healthcare and Family Services; HSAG = Health Services Advisory Group; MACPAC = Medicaid and CHIP Payment and Access Commission; OIG = HHS Office of Inspector General.
What we already know (from the vault)
- The one hard number is on the wrong payer line. [[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]]: Virginia's 2024 Bureau of Insurance report put SUD at 24.9% of claims denied, M/S at 22.5%, MH at 20.8%; excluding prescription transactions the recomputed ratios were SUD 2.73x M/S and MH 1.13x. Commercial fully-insured only.
- The durable finding there was a cause, not a rate. Excluding the prescription-drug subcategory, preauthorization or precertification accounted for 32.9% of Virginia SUD denials against 5.0% for M/S, while medical necessity was 1.4% of categorized SUD denials. Whether that cause mix replicates is what this brief was filed to test.
- The measurement was visibly unstable. Virginia's own overall denial rate moved 13.6% to 19.7% to 22.3% across three consecutive reports.
- No ABA-specific rate existed anywhere. The parent brief's cleanest null: no public dataset reported denial rates for ABA specifically, including the advocacy-side peer-reviewed literature with every incentive to publish a high one.
- Distribution, not denial rate, was already the binding constraint. [[2026-10-03-behavioral-health-billing-ai-bundled-vs-standalone]] found one of 24 ABA/SUD electronic health record vendors offering to run revenue cycle on a rival system of record.
What the web says
- PRIMARY, and the answer to the headline question is yes: Louisiana publishes Medicaid managed-care behavioral-health claim denial rates by behavioral-health service category, including ABA, quarterly. The Healthy Louisiana Claims Report, Response to R.S. 46:460.91 as amended by Act 233 of the 2022 Regular Legislative Session, Quarter 2 Calendar Year 2025 (Louisiana Department of Health, Bureau of Health Services Financing, February 2026) states at p.4 that "Behavioral Health providers are discreetly identified within these categories, as required by the Louisiana State Legislature." Table 1.8 gives, per MCO, adjudicated and denied counts plus a printed denial percentage for Applied Behavior Analysis, Behavioral Health - Residential, and All Other Specialized Behavioral Health; Table 1.2 gives Behavioral Hospital inpatient at 15.0% to 30.1% of days denied by MCO against Acute Hospital at 22.4% to 31.0%. Payer line: Medicaid managed care only, six MCOs, no commercial and no Medicare. Confidence: high -- I downloaded the PDF, extracted it, and read the tables directly.
- PRIMARY, re-cut, and this is the finding that matters most. Aggregating Table 1.8 across all six MCOs, ABA is denied at 9.31% while every non-behavioral-health outpatient and professional category combined runs 16.79%. My own arithmetic from the published rows, shown so it can be rechecked: ABA 10,241 denied of 109,991 adjudicated = 9.31%; Behavioral Health - Residential 25,976 of 284,717 = 9.12%; All Other Specialized Behavioral Health 253,920 of 1,650,761 = 15.38%; all three behavioral-health rows combined 290,137 of 2,045,469 = 14.18%; all non-behavioral-health rows combined 2,490,805 of 14,836,610 = 16.79%. Comparators in the same table: Primary Care - Pediatrics 13.49%, Primary Care - Others 14.91%, Outpatient Hospital 12.65%, Therapies (physical, occupational, speech) 13.52%, Durable Medical Equipment 20.79%. Confidence: high on the arithmetic (the per-MCO printed percentages reconcile exactly to the counts); medium on interpretation, for three stated reasons. The report's own caveat at p.11: claims "are reported based on the status of the last adjudication decision made in the current reporting period. This may not be the final determination, as claims may be adjusted in future reporting periods." The rows are line-level, not header-level. And the statewide aggregate is my derivation, not a printed figure.
- PRIMARY. The per-payer spread inside that one state program is larger than any behavioral-health-versus-medical gap found anywhere in this sweep. ABA denial rates by MCO, read directly from Table 1.8: 11.0% / 5.0% / 11.4% / 34.7% / 8.6% / 35.9%. A 7.2x spread between the best and worst payer, same state, same quarter, same service category. Confidence: high on the figures; the report gives no provider-mix detail, so how much of the spread is payer behavior versus provider composition is not determinable from it.
- PRIMARY. Illinois publishes most of the Virginia table shape for Medicaid managed care, and finds behavioral health paid slightly better than medical/surgical in-network. 2023 2024 Mental Health Parity Analysis Summary Report (Illinois HFS, prepared by HSAG, December 2024), Table 1-2, six HealthChoice Illinois MCOs, calendar 2023, read at source: in-network MH/SUD inpatient 93.1% paid, outpatient 94.5%, emergency 95.9%, total 94.5%; in-network M/S total 93.4%. Out-of-network MH/SUD total 83.0% against M/S 85.1%. The report's own conclusion: "the analysis showed a minimal difference in the percentage of header-level paid claims between MH/SUD and M/S claims for INN (94.5 percent and 93.4 percent), with MH/SUD claims being paid at a higher rate than M/S claims." Shortfalls against Virginia: MH and SUD are merged into one line, prescription and non-emergency transport claims are excluded, there is no office-visit category, denial rate is the unprinted complement of the paid rate, and there is no cause mix -- the report states "Data were not available to determine specific procedure type codes that were paid versus denied." Confidence: high, verified at source.
- The federal streams are a clean and documented null, and the gap is deliberate. MACPAC's Denials and Appeals in Medicaid Managed Care (January 2023), read at source, states under Oversight: "No required monitoring of denial rates or reasons for denials / No required monitoring of outcomes of appeals / No required audit of denials or appeals to assess whether denials were clinically appropriate," and that states "must contract with External Quality Review Organization (EQRO) - focus is on compliance." Its own state scan found 14 states requiring MCOs to report denials by service type and 11 requiring the reason, with footnote 1 naming "Colorado collects data on denials for behavioral health only." On the regulatory side, 42 CFR 438.210(f) as amended by CMS-0057-F requires the denied-request percentage "aggregated for all items and services," a phrase appearing 48 times in the rule; commenters asked for disaggregation including by psychiatric provider type and the Centers for Medicare & Medicaid Services declined, responding that "too much data can also become overwhelming." MCPAR carries inpatient and outpatient behavioral health as named indicators, but for appeals, and the technical guidance makes reason and service type structurally un-crossable: an appeal "can only have ONE reason" while it "can apply to MORE THAN ONE service type." T-MSIS captures a mandatory CLAIM-DENIED-INDICATOR, but the TAF research files that researchers actually get are "active, final-action, non-voided, and non-denied claims (except for Illinois)." Confidence: high on the MACPAC quotes (verified by me at source); medium-high on the CMS regulatory and T-MSIS quotes (verified at source by a sub-agent, not re-opened by me).
- New York and California are nulls on the Medicaid question, and both hold the data they do not publish. New York's DFS mental-health parity workbooks under Insurance Law section 343 are Virginia-shaped and better -- inpatient and outpatient by network status, emergency care, and prescription drugs split three ways, each with parallel M/S, MH and SUD approval and denial columns -- and the DFS reporting instructions say the data "should not include information related to the following types of coverage: Self-funded; Medicaid; Child Health Plus; Essential Plan; Managed Long-Term Care." Commercial only; does not answer the Medicaid question. New York's own 2022 parity compliance report for Medicaid managed care instructs plans to perform exactly this analysis ("Audit/review of denial and appeal rates (both medical and administrative) by service type or benefit category") and publishes zero denial figures, while confirming the data is held: "Information on service utilization and benefit coverage denials ... was available for inclusion as this data is currently reported to the State." The 2024 New York EQRO annual technical report contains zero occurrences of the string "deni". California is a harder null: the DMHC Behavioral Health Investigations exclude Medi-Cal by their own footnote, and their Phase 2 finding is that plans do not track this at all -- "One health plan does not identify, track and report behavioral health services with a high denial rate." Delaware is a null on both payer lines; its State Council for Persons with Disabilities asked DMMA to require MCO claims-denial data and DMMA declined. Confidence: high on the quoted document language (verified at source by sub-agents; I did not re-open these).
- SECONDARY, federal auditor rather than state publication. The only published New York Medicaid behavioral-health denial rates by service come from OIG, and they are the highest numbers in this brief. New York Did Not Ensure That Selected Medicaid Managed Care Organizations Complied With Mental Health and Substance Use Disorder Parity Requirements Related to Prior Authorization (A-02-24-01011, August 2026), read at source: one MCO's outpatient out-of-network denial rates for calendar 2023 were mental health 69.8% / 69.9% / 72.7% against medical/surgical 24.9% / 28.9% / 30.0% across quarters 2 through 4. Separately, "the annual denial rate for community psychiatric support and treatment (CPST) services was 42 percent," partial hospitalization 37% at a second MCO, continuing day treatment 20% at a third. Confidence: high that these are the reported figures, verified at source; low as a market fact -- three MCOs, anonymized, one quarter missing because "the MCO did not maintain data to support its analysis," and OIG's own footnote warns a higher rate "does not necessarily mean noncompliance."
Convergences and contradictions
- The question's premise is half right: by-category behavioral-health denial reporting exists on the Medicaid line, but only where a statute named behavioral health explicitly. Louisiana (Act 233), Illinois (305 ILCS 5/5-30.1 and the parity reporting mandate), Colorado (SB 21-137, SUD only) publish; New York, California and Delaware do not. The EQRO annual technical report -- the document class most likely to carry it on structural grounds -- is a near-universal null across 16 reports checked, and MACPAC independently confirms why: the EQRO's focus is compliance, and no federal rule requires denial-rate monitoring at all. The null is auditable rather than an absence of effort.
- The Virginia cause mix does not replicate, and the reason is a measurement-basis artifact that matters more than the divergence itself. Virginia measured claim denials, where a provider's failure to obtain authorization surfaces as a CARC; preauthorization was 32.9% of SUD denials and medical necessity 1.4%. The two Medicaid documents that publish a behavioral-health cause mix measured prior-authorization denials instead, and both invert the result: Colorado's SUD denial study splits 86% medical necessity / 14% administrative, and New Hampshire's behavioral-health-drug mix is 95.4% not medically necessary against 2.1% incomplete paperwork. That inversion is close to definitional -- in a dataset counting PA requests submitted, "never submitted a PA request" cannot appear. The one Medicaid claim-denial cause mix found, Illinois HFS's semiannual hospital claims report, lands back on Virginia's side: Authorization 9.90% of denials rising to 24.04% in a later quarter, Medical Necessity 0.25% falling to 0.10%, with HFS writing that "'Medical Necessity' of services continues to be a non-factor with respect to denials." But that report carries no behavioral-health breakout. Nobody, in any state, on any payer line, publishes a cause mix crossed with a behavioral-health service category. That cell is empty everywhere.
- Contradiction the venture should care about most: on the Medicaid line, behavioral health is denied less, and ABA least of all. Louisiana's behavioral-health rows aggregate to 14.18% against 16.79% for everything else in the same table, with ABA at 9.31%. Illinois finds in-network MH/SUD paid at a higher rate than medical/surgical. Illinois's EQRO found behavioral-health prior-authorization denial rates below medical ones at five of six MCOs and concluded there were "no concerns with parity." Against that, OIG's New York audit found one MCO denying 70%+ of outpatient out-of-network mental-health requests, and Virginia's commercial data found SUD elevated. The reconciliation is network status and payer identity, not service line: the gaps appear out-of-network and at specific payers, and vanish in-network and in aggregate.
Synthesis for RDCO
The question had two halves and they resolve in opposite directions. The by-category reporting exists -- Louisiana publishes it quarterly, for Medicaid managed care, with Applied Behavior Analysis broken out as its own line, and has done so since roughly calendar 2019. That is a real answer, not a null, and it retires the parent brief's cleanest finding. There is now a public ABA denial rate. It is 9.31% across 109,991 adjudicated line-claims in a single quarter, and it sits below the 16.79% non-behavioral-health rate in the same table and below Primary Care - Pediatrics in the same table. The vendor claim of 15-30% ABA initial denials at "3-5x the medical average" is no longer merely unsourced, as the parent brief concluded. On the payer line that carries most ABA volume it is contradicted by the only public measurement of it. The venture premise should be treated as falsified on the ABA leg rather than open.
The cause half of the question resolves into something more useful than a yes or no. The Virginia preauthorization-driven mix does not look like a generalizable market fact, and the reason is that the cause mix is an artifact of which denial event you measure. Claim-denial datasets put administrative and authorization causes on top and medical necessity near zero; Virginia at 32.9% versus 1.4%, Illinois Medicaid at 24.04% versus 0.10%. Prior-authorization datasets put medical necessity on top and administrative near the bottom; Colorado at 86% versus 14%, New Hampshire at 95.4% versus 2.1%. Both are true of the same providers. They are counting different moments in the lifecycle. Any product pitch built on "most behavioral-health denials are authorization failures, not clinical refusals" is defensible only if the pitch also says which measurement it rests on, and the honest version is that no published dataset crosses a behavioral-health service category with a denial cause on the Medicaid line at all. Louisiana comes closest by accident: its behavioral-health prior-authorization denial rates run 0.0% to 6.1% at most ASAM levels of care while its behavioral-health claim denial rates run 9% to 15%, which is consistent with the authorization gate approving and the claim dying afterward. That is my inference from two tables in one report, flagged as such, not a published finding.
What this does to where the product has to sit is the most actionable result. The dominant term in the Medicaid data is not service line, it is payer. ABA denial rates inside one state program, one quarter, under one set of state rules, span 5.0% to 35.9% across six MCOs -- a 7.2x spread, larger than any behavioral-health-versus-medical gap anywhere in this sweep including Virginia's. The service-line framing that every competitor in [[2026-10-01-national-rcm-eob-ai-behavioral-health-scan]] markets on is the weaker signal; payer-specific adjudication variance is the stronger one. A product positioned on "behavioral health gets denied more" is positioned on a claim the Medicaid data contradicts. A product positioned on "your denial rate at this payer is multiples of your rate at that payer for the same service, and here is the rule you are missing" is positioned on something the data supports and that providers cannot see for themselves, because the comparison requires a book of business wider than any one practice. That is first-pass claim construction and payer-rule benchmarking, upstream of the EOB, not appeals or remittance parsing downstream of it. It also makes the claim lifecycle question concrete rather than philosophical: the leak is at authorization capture and payer-specific claim assembly, and the measurement that would price it is the one nobody publishes.
Two dated catalysts make this a live window rather than a settled null, and both cut in the venture's favor. MCPAR Topic XIII prior-authorization indicators became mandatory in June 2026, and 42 CFR 438.210(f) plan-website postings began March 2026. Both produce Medicaid prior-authorization denial rates per plan with no service-category and no behavioral-health dimension, because the Centers for Medicare & Medicaid Services declined that granularity on the record. So within a year there will be a universal public payer-level denial rate and still no public behavioral-health service-category breakdown -- precisely the shape of a defensible data asset. The proprietary-remittance test from [[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]] therefore survives and gets cheaper: a multi-site ABA operator's remittance history can now be benchmarked against a published Medicaid comparator in Louisiana and a published parity comparator in Illinois, which was not possible two weeks ago. The first question to Michael Holzum is unchanged and the framing is better: not "is behavioral health denied more," which the public data answers no, but "what is your denial rate by payer, and how far is your worst payer from your best."
Why this is in the vault
This closes two of the three follow-ups left open by [[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]] on the family co-founder venture thread, and it changes the go/no-go test rather than merely informing it: the ABA elevated-denial premise is now contradicted on the Medicaid line by Louisiana's Act 233 series, and the product thesis moves from service-line positioning to payer-variance benchmarking upstream of the explanation of benefits.
Open follow-ups
- Does the Louisiana ABA denial rate trend across the full Act 233 quarterly series back to calendar 2019, and does the 7.2x per-MCO spread persist or was Q2 2025 an outlier? A six-year single-state series is the only longitudinal behavioral-health denial data found anywhere, and one quarter of it is thin.
- Can the two known unpublished behavioral-health cause-mix datasets be obtained by records request? Illinois HFS collects quarterly from all six MCOs "the percent of prior authorization requests denied and top denial reasons, broken down by behavioral health and non-behavioral health benefits" and has never published it; New York's Office of Mental Health collects monthly inpatient and quarterly outpatient denial reports keyed to 21 behavioral-health outpatient service categories by named denial reason and review stage, also unpublished. Either would fill the empty cell.
- Do the MCPAR Topic XIII prior-authorization filings actually land after the June 2026 mandate, and at what granularity? As of the 2024 public use file only two plans had reported voluntarily, and both New York and Arizona answered "not reporting data."
- Is the claim-denial versus prior-authorization-denial cause-mix inversion a pure measurement artifact or a real difference? Testable by finding one dataset carrying both bases on the same population; none surfaced in this pass.
- What explains the Louisiana per-MCO ABA spread -- payer rules, provider mix, or data-quality variance in MCO self-reporting? The report is MCO-self-reported with a Louisiana Department of Health validation pass and prints "N/A" where gaps exist, so the spread may be partly reporting artifact.
- What is the ABA denial rate on self-funded ERISA plans? Still measured by nothing, on either the rate or the cause axis, and it is the second-largest payer line for ABA after Medicaid.
Related
- [[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]]
- [[2026-10-01-national-rcm-eob-ai-behavioral-health-scan]]
- [[2026-10-03-behavioral-health-billing-ai-bundled-vs-standalone]]
- [[2026-10-05-behavioral-health-billing-firm-ehr-count]]
- [[2026-07-21-family-medical-careers-scenario]]
- [[03-contacts/michael-holzum]]
- [[hcls/problem-log|HCLS problem log]]
Sources
Per-state and per-stream verdicts
| Jurisdiction / stream | BH denial rates by service category | Cause mix published | Payer lines covered | Document |
|---|---|---|---|---|
| Louisiana | YES -- ABA, BH-Residential, All Other Specialized BH, Behavioral Hospital, per MCO, quarterly | NO -- zero denial-reason tables in the report | Medicaid managed care only (6 MCOs) | Healthy Louisiana Claims Report, Act 233, CY2025-Q2, Feb 2026 |
| Illinois | PARTIAL (strong) -- MH/SUD vs M/S x inpatient/outpatient/emergency x network status, statewide and per MCO | NO in the parity report ("Data were not available"); a 7-reason mix exists in a separate hospital-claims report with no BH breakout | Medicaid managed care (6 MCOs) | HFS/HSAG 2023 2024 Mental Health Parity Analysis Summary Report, Dec 2024 |
| New York | NO from New York itself. Commercial-only DFS section 343 workbooks are Virginia-shaped but explicitly exclude Medicaid. The only published NY Medicaid BH rates by service come from a federal auditor | NO -- the external-appeal cause axis is the statutory appeal grounds, and preauth/administrative denials are not externally appealable in NY | Commercial fully-insured (DFS parity); Medicaid (OIG audit only) | DFS MH/SUD parity reports + reporting instructions; OIG A-02-24-01011 |
| California | NO -- clean null on Medi-Cal. DMHC Behavioral Health Investigations exclude Medi-Cal by their own footnote and cite plans for not tracking BH denial rates | PARTIAL, commercial only -- Covered California QHP filings split BH-only claims with a reason mix | Commercial (Covered CA, DMHC BHI); Medi-Cal has no denial-rate publication | DMHC BHI Phases 1-3; Covered California denied-claims filings |
| Delaware | NO -- hard null on both lines; DMMA declined a request to require MCO claims-denial data | NO -- the only administrative-vs-medical-necessity split published statewide is dental, one vendor, 120 denials | Commercial market-conduct exam populations only (counts, not rates) | Delaware Quality Strategy 2023; 2024 EQR technical report |
| Virginia (carried forward) | YES by service type for MH/SUD/M/S + top-three-reason table | YES -- the only jurisdiction with both | Commercial fully-insured only -- excludes Medicaid and self-funded | Virginia Bureau of Insurance 2024 report |
| CMS federal (MCPAR, T-MSIS/TAF, CMS-0057-F) | NO. MCPAR has inpatient/outpatient BH indicators but for appeals, and reason and service type cannot be cross-tabulated. T-MSIS captures denials; TAF strips them. 438.210(f) is "aggregated for all items and services" | NO for Medicaid. The ACA transparency PUF has a BH medical-necessity denial field, Marketplace only -- does not answer the Medicaid question | Medicaid managed care (MCPAR); Marketplace (ACA PUF) | 42 CFR 438.210(f); MCPAR appeals/grievances technical guidance, Aug 2024; TAF Claims RIF Codebook v1.8 |
| EQRO reports | NO -- near-universal null across 16+ annual technical reports (HSAG, IPRO, Comagine, Myers and Stauffer, MetaStar), including behavioral-health carve-out reports in Arizona and New York. Exceptions are statute-driven focused studies, not the 438.364 report | PARTIAL in focused studies: New Hampshire 5-reason mix on BH drugs, Colorado 86/14 medical-necessity-vs-administrative on SUD | Medicaid | MACPAC Jan 2023 scan corroborates the null |
Vault
06-reference/research/2026-10-04-behavioral-health-claim-denial-rate-primary-sources.md(parent brief; Virginia figures and the ABA null carried forward)06-reference/research/2026-10-01-national-rcm-eob-ai-behavioral-health-scan.md06-reference/research/2026-10-03-behavioral-health-billing-ai-bundled-vs-standalone.md06-reference/research/2026-10-05-behavioral-health-billing-firm-ehr-count.md01-projects/life/2026-07-21-family-medical-careers-scenario.md03-contacts/michael-holzum.md01-projects/hcls/problem-log.md
Primary, verified at source by me (downloaded, extracted, tables read)
- Louisiana Department of Health, Bureau of Health Services Financing, "Healthy Louisiana Claims Report, Response to R.S. 46:460.91 as amended by Act 233 of the 2022 Regular Legislative Session, Quarter 2 Calendar Year 2025," February 2026 -- https://ldh.la.gov/assets/docs/LegisReports/Claims/Act233_HLA_ClaimsReport_CY_Q2_2025.pdf (Tables 1.2, 1.8, 1.9, 1.22; statewide aggregates are my derivation from the published per-MCO counts)
- Illinois Department of Healthcare and Family Services, prepared by Health Services Advisory Group, "2023 2024 Mental Health Parity Analysis Summary Report," December 2024 (doc ID IL2024_Mental Health Parity_Summ Report_F1_1224) -- https://hfs.illinois.gov/content/dam/soi/en/web/hfs/sitecollectiondocuments/il20232024mentalhealthparitysummaryreport.pdf (Tables 1-2, 1-3)
- HHS Office of Inspector General, "New York Did Not Ensure That Selected Medicaid Managed Care Organizations Complied With Mental Health and Substance Use Disorder Parity Requirements Related to Prior Authorization," A-02-24-01011, August 2026 -- https://oig.hhs.gov/documents/audit/11847/A-02-24-01011.pdf (p.7 table; p.8 service-specific rates)
- MACPAC, "Denials and Appeals in Medicaid Managed Care," January 26 2023 -- https://macpac.gov/wp-content/uploads/2023/01/10_Denials-and-appeals-in-Medicaid-managed-care.pdf (oversight slide; state scan and footnotes)
Primary, verified at source by sub-agents and not re-opened by me -- treat as one step removed
- New York DFS, "Mental Health and Substance Use Disorder Parity Reports" (30 insurers x 2023 and 2024) and the MH/SUD parity reporting instructions -- https://www.dfs.ny.gov/reports_and_publications/mhsud_reports ; https://www.dfs.ny.gov/system/files/documents/2026/03/mh_sud_parity_reporting_instructions.pdf
- New York OMH/OASAS/DOH, "Compliance with the Mental Health Parity and Addiction Equity Act Comprehensive Report: New York Medicaid Managed Care, Alternative Benefit Plan, and CHIP," March 14 2022 -- https://www.health.ny.gov/health_care/managed_care/reports/docs/compliance_report_3-22.pdf (instructs the analysis, publishes no figures)
- New York 2024 EQR Annual Technical Report -- https://www.health.ny.gov/statistics/health_care/managed_care/plans/reports/docs/2024_mmc_harp_snp_report.pdf (zero occurrences of "deni")
- California DMHC Behavioral Health Investigations Phases 1-3 -- https://www.dmhc.ca.gov/Portals/0/Docs/DO/BHISummaryReport_FINAL.pdf and the Phase 2 / Phase 3 summary reports in the same directory (Medi-Cal excluded by footnote 3)
- Delaware DMMA, "The Delaware Quality Strategy 2023" -- https://dhss.delaware.gov/wp-content/uploads/sites/11/2026/06/The-Delaware-Quality-Strategy-2023.pdf (SCPD request and DMMA response); Delaware EQR 2024 Technical Report -- https://dhss.delaware.gov/wp-content/uploads/sites/11/2025/04/2024_de_eqr_technical_report.pdf
- Illinois HFS, "Analysis of HFS-contracted MCO Claims Processing and Payment Performance," Q1-Q2 and Q3-Q4 CY2024 -- https://hfs.illinois.gov/content/dam/soi/en/web/hfs/sitecollectiondocuments/mcohospitcalclaimprq1q22024.pdf (Table 8A denial-reason mix, no BH breakout)
- Illinois HFS / HSAG, "2025 UM Service Authorization Data Analysis Summary Report," April 2026 -- https://hfs.illinois.gov/content/dam/soi/en/web/hfs/info/reports/miscellaneous/2025umsadasummaryreport.pdf (BH vs medical PA denial rates per MCO)
- HSAG for New Hampshire, "State Fiscal Year 2024 Service Authorization Quality Study," January 2025 -- https://medicaidquality.nh.gov/sites/default/files/2025-10/NH_SFY%202024%20MCO%20Quality%20Study%20Report_F2.pdf (BH-drug denial rates with a 5-category cause mix)
- HSAG for Colorado, "FY 2025-2026 Inpatient and Residential Substance Use Disorder Service Denial Determination Analysis," January 2026 -- https://hcpf.colorado.gov/sites/hcpf/files/CO2025-26_SUD%20Denial_Aggregate%20Report_F1-B.pdf (5.99% SUD denial rate; 86% medical necessity / 14% administrative)
- 42 CFR 438.210(f), eCFR current as of 2026-10-01 -- https://www.ecfr.gov/current/title-42/section-438.210 ; CMS-0057-F preamble response declining disaggregation, 89 FR 8893
- CMS, "MCPAR Technical Guidance -- Topic: Appeals and Grievances," August 2024 edition -- https://www.medicaid.gov/medicaid/managed-care/downloads/mcpar-appeals-grievances-tech-guidance.pdf (one reason per appeal, many service types per appeal)
- CCW, "T-MSIS Analytic Files (TAF) Claims RIF Codebook," December 2025, v1.8 -- https://www2.ccwdata.org/documents/10280/19022436/codebook-taf-claims.pdf ("non-denied claims")
- HHS OIG, "High Rates of Prior Authorization Denials by Some Plans and Limited State Oversight Raise Concerns About Access to Care in Medicaid Managed Care," OEI-09-19-00350, July 2023 -- https://oig.hhs.gov/oei/reports/OEI-09-19-00350.pdf (12.5% overall; zero hits for "mental health" or "substance use")
- HHS OIG, "Community Behavioral Health Did Not Comply With Requirements When Denying Prior Authorization Requests," A-03-24-00204, June 2026 -- https://oig.hhs.gov/documents/audit/11700/A-03-24-00204.pdf (within-BH denial breakout, one Pennsylvania BH-MCO)
- North Carolina Medicaid Managed Care Claims Monitoring Dashboard, Tailored Plan filter, July 2026 data -- https://dashboards.ncdhhs.gov/t/Medicaid/views/ManagedCareClaimsMonitoring-Website/HomePage (BH/IDD Tailored Plan paid/denied by claim type; "SERVICE IS NOT AUTHORIZED" and "PRIOR AUTHORIZATION REQUIRED" in top-three reasons)
Research budget used: QMD 1 of 5 (lex + vec on vault priors). Zero WebSearch and zero WebFetch calls from the parent; the web sweep was fanned out to five parallel sub-agents (New York, California, Illinois+Delaware, CMS federal streams, EQRO cross-state), each running its own searches and retrievals, per the AGENTS.md rule on routing long artifacts away from parent context. The parent then independently re-retrieved and re-read the four load-bearing documents by curl plus pdftotext and recomputed every Louisiana figure used here, satisfying the primary-source gate. One sub-agent claim was checked and rejected: Louisiana's Table 1.9 is average days to adjudicate, not a denial-cause mix.
Not reached: Massachusetts MBHP behavioral-health-only EQR technical reports 2020-2025 and Pennsylvania HealthChoices behavioral-health annual technical reports (both HTTP 403); Virginia DMAS RD267's companion "Top Claim Denial Reasons by Provider Type" spreadsheet, which would convert Virginia's Medicaid side to a full behavioral-health rate plus reason mix; Illinois HFS quarterly QBR behavioral-health denial-reason data (collected, never published); New York OMH monthly and quarterly behavioral-health denial reports (collected, never published); roughly 20 New York plan-level behavioral-health focused-survey PDFs that are scanned images without a text layer; Texas UMCM 5.24.1 and 5.24.6 behavioral-health denial collections (collected, never published).