Virginia's 16.2 million denied claims: the report never says initial or final, and 88% of the count is pharmacy transactions
The question
"Does the Virginia Bureau of Insurance's 16.2 million denied-claim count represent initial determinations or final post-resubmission status, per the report's own methodology text?"
Context: the 0.06% appeals ratio in [[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]] and the rework volumes in [[2026-10-17-behavioral-health-silent-rework-market-size]] both read 16.2M as initial determinations. This pass reads the report and its sibling editions to see whether the text supports that.
Verdict: AMBIGUOUS PER TEXT. No edition of the report defines "denied claim", and none says whether a claim that is denied, corrected, resubmitted and paid is counted as denied, as paid, or as both. Indirect evidence in the tables leans toward a per-submission count (closer to "initial"), but that is inference from structure, not a statement by the Bureau.
What we already know (from the vault)
- The 16.2M comes from the Bureau's 2024 report (calendar 2023 data, RD712): 16,233,560 denied of 72,730,407 received, 16 carriers, 2.48M lives. [[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]] already wrote that "the report does not define whether 'denied' counts initial determinations or final status after resubmission." That sentence was correct and this pass confirms it at source.
- The same brief computed 9,429 closed internal appeals / 16.2M denied = 0.06% and used it to say appeals are a rounding error.
- [[2026-10-17-behavioral-health-silent-rework-market-size]] applied a 40.8% resubmit-and-paid share (Medicare Advantage, Health Affairs) to Virginia's 133,009 SUD and 2,006,618 mental-health denials to get roughly 54,300 and 873,000 reworked claims, and named "these are initial denials" as its Assumption 2 and largest threat.
- [[2026-10-17-state-medicaid-behavioral-health-denial-rates-by-category]] flagged that Virginia's overall rate is unstable across editions (13.6% to 19.7% to 22.3%).
What the web says
All quotes below were read at source from PDFs downloaded from the Virginia legislative reports archive and extracted locally. Page numbers are PDF page / printed page.
- The 2024 report's entire methodology for the claims count is two sentences, and neither defines "denied". "Carriers surveyed for this 2024 report received a total of 72,730,407 claims, with 16,233,560 (22.3%) of those denied." and "Each carrier reported the total number of denied claims related to medical/surgical, mental health, and substance use disorder benefits. These claims were then separated into five service types: office visit claims, all other outpatient claims, inpatient claims, emergency care claims, and outpatient prescription (Rx) drug transactions." (RD712, PDF p.5 / printed p.2). The only other method statement is "the Bureau of Insurance conducted a data call of 16 health carriers" (PDF p.4). The words initial, final, resubmit, adjudicate, pended, corrected and reprocess do not appear anywhere in the claims section or appendices. Confidence: high (full-text search of the extracted PDF plus a read of the claims section and Appendix A).
- The definition is absent from every edition I could get, so it did not change across years; it was never there. I searched six editions covering calendar 2019 through 2024 (RD335 2020, RD408 2021, RD615 2022, RD582 2023, RD712 2024, RD721 2025) for the same terms. None defines a denied claim or addresses resubmission. The 2020 edition says only that "a separate survey was developed by the Bureau in conjunction with the Virginia Association of Health Plans" (RD335, PDF p.2) and "Carriers surveyed reported a total of 42,465,433 claims received with 5,418,407 (12.8%) of claims being denied" (PDF p.5). Confidence: high for those six; the 2018 and 2019 editions on scc.virginia.gov returned 404 and have no Wayback capture, so they were not read.
- One instruction did change, and it bears on what the count means. The 2020 through 2023 editions say: "Carriers were asked to report to the Bureau the total number of claims denied for which the denial would leave the member responsible for payment and to identify the top three denial reasons" (RD335, PDF p.19, Attachment A; same sentence in RD408 p.23, RD615 p.10, RD582 p.10). In the 2020 edition the total reported under that instruction is the same 5,418,407 as the Section I headline, which implies the headline denied count was at least nominally scoped to member-liability denials. The 2024 and 2025 editions drop the sentence without comment. Yet every edition's appendix lists reasons that are normally provider write-offs, not member liability: "Claim line denied by external bundling/fraud detection system", "Duplicate charges", NCCI edit denials (RD712, PDF p.18 / printed p.15). So carriers were not applying one definition. Confidence: medium that this reflects real carrier-to-carrier inconsistency; it is my reading of the text, the Bureau does not say so.
- Structural evidence leans per-submission, not final status. (a) In all 15 service-type rows of Tables 1 to 3, Received = Paid + Denied exactly (verified arithmetically), so every received item is forced into one of two bins with no pending, adjusted or reversed category. (b) "Duplicate charges" appears as a reported denial reason, which only makes sense if a second submission is counted as its own received-and-denied claim. (c) The largest single reason is "Prescription refill too soon" (1,870,315 medical/surgical denials, Table 4, PDF p.8), a point-of-sale transaction reject that is by nature an initial event later re-run and paid as a new transaction. (d) "Claim line denied" and "Claim Paid at 0" in Appendix A show at least some carriers reported at line level. None of this is a definition. Confidence: low-medium that the count is submission-level for most carriers.
- 88.1% of the 16.2M is outpatient prescription transactions. Summing the Rx rows: 12,370,680 + 1,856,913 + 69,449 = 14,297,042 of 16,233,560. Non-pharmacy denied claims total 1,936,518. Confidence: high (arithmetic on Tables 1 to 3).
- The newest edition shows the denominator is not stable enough to carry a ratio. The 2025 report (calendar 2024, 15 carriers) has 44,482,942 received and 7,953,579 denied (17.9%), with Rx transactions falling from 45.6M to 17,948,248 (RD721, PDF p.8 and p.12), and 8,894 closed internal appeals (PDF p.19). The same appeals-to-denials ratio is 0.11% for 2024, nearly double 2023's 0.06%, with almost no change in appeals. The Bureau offers no explanation for the Rx swing. Confidence: high on figures.
- The appeals numerator is not a subset of the denied-claims denominator. The report's definition: "An internal appeal is filed to obtain approval for services denied by a managed care health insurance plan as the result of utilization review or an administrative denial" (RD712, PDF p.11 / printed p.8). The 2020 edition is explicit that this includes "a denied request for pre-authorization, which is a pre-service appeal" (RD335, PDF p.13). Pre-service appeals concern services with no claim yet, and the count is appeals closed in the year, not appeals filed against that year's denials.
- The data-call instrument itself is not public as far as I could find. Three searches returned only the annual reports. No template, instruction letter or field definitions surfaced on scc.virginia.gov or the legislative archive. Not confirmed absent; only not found.
Convergences and contradictions
- Convergence: the 2026-10-04 brief's caveat was right. The report does not define the basis, in any year.
- Contradiction with how the caveat was then used: both parent briefs go on to treat 0.06% as "the share of denied claims appealed" and 16.2M as initial determinations. The text supports neither. The ratio divides closed utilization-review appeals (some pre-service) by a count that is 88% pharmacy transactions and of undefined basis.
- Contradiction on direction of risk: the 2026-10-17 brief says that if the counts are final, "the whole estimate collapses." It would actually move the other way. If Virginia's figure were claims still denied after resubmission, initial denials would be larger than reported and the reworked volume implied by a 40.8% share would be higher (about 0.69x the final count rather than 0.41x), not lower. The estimate becomes invalid as computed, but not smaller.
Synthesis for RDCO
The question as asked has a clean answer: the report's own text cannot settle it. Six editions, zero definitions. The tables behave like per-submission counts (two-bin partition, duplicates counted as denials, pharmacy rejects dominating), so "initial" is the more likely reading, but that is a lean, not a finding, and the dropped "member responsible for payment" instruction shows carriers were not reporting on one basis anyway. Treat Assumption 2 of the rework brief as still open and unresolvable from public documents.
What this does to the 0.06%: the arithmetic stands (9,429 / 16,233,560 = 0.058%), and the qualitative claim it supports, that formal appeals are tiny next to denial volume, survives under either reading. If the count is initial, the ratio is as stated. If it is final, initial denials are larger and the ratio is smaller still. So the initial-versus-final question was never the real threat to that number. The real problems are different and bigger: the denominator is 88% pharmacy transactions, the numerator includes pre-service appeals that are not claims, and the ratio jumps to 0.11% a year later on a reporting change. It should be described as "about 9,400 closed internal appeals alongside 16.2M denied claims and pharmacy transactions", not as an appeal rate. On non-pharmacy denials only, the same 9,429 appeals come to at most 0.49%, an upper bound because some appeals concern drugs.
What this does to the rework arithmetic: the larger correction is one this question did not ask about. The 2026-10-17 brief applied 40.8% to all 133,009 SUD and 2,006,618 mental-health denials, pharmacy included, even though the 2026-10-04 brief had already argued that pharmacy transactions should be excluded for an RCM product. On the non-pharmacy rows the bases are 63,560 SUD and 149,705 mental health. At 40.8% that is about 25,900 SUD and 87,000 behavioral-health claims reworked and paid, not 54,300 and 873,000. The behavioral-health total is overstated roughly tenfold and the SUD figure roughly twofold, independent of initial versus final. The "440 to 1" rework-to-appeals ratio becomes about 209 to 1 on non-pharmacy SUD, and even that divides by 124 appeals that include drug and pre-service appeals. All of this still borrows a Medicare Advantage percentage, and I did not re-read the Health Affairs study to confirm it excludes pharmacy claims.
For the family RCM venture the direction of the conclusion does not change: rework dwarfs appeals, and the upstream-authorization finding rests on Virginia's cause table, not on these volumes. What changes is that the Virginia-derived volume figures should not be quoted to Michael and Christian as sizing. They are an order-of-magnitude illustration on an undefined, unstable, pharmacy-dominated base.
Line-level corrections needed (parent briefs not edited):
[[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]]
- Line 30, final sentence: keep, and add that no edition from 2020 through 2025 defines it, and that editions through 2023 describe the count as denials that "would leave the member responsible for payment", language dropped in 2024.
- Line 31: "moved 13.6% to 19.7% to 22.3% across three consecutive reports" - append the 2025 edition's 17.9%, and that Rx transactions fell from 45.6M to 17.9M between the two latest editions.
- Line 42: replace "9,429 internal appeals against 16.2 million denied claims (0.06%)" with wording that states 88% of the 16.2M is pharmacy transactions, that internal appeals include pre-service utilization-review appeals, and that the ratio is illustrative, not an appeal rate. Add the 2024-data equivalent (8,894 / 7,953,579 = 0.11%).
- Line 50: "Virginia providers appealed 0.06% of denied claims" - wrong on two counts. The appeals are filed by consumers or providers, and the ratio is not a share of denied claims. Reword to "closed internal appeals equal 0.06% of the denied-claim-and-transaction count."
- Line 63: add that Virginia's basis is undefined in the report text, so it cannot be used as the initial-denial base without that caveat.
[[2026-10-17-behavioral-health-silent-rework-market-size]]
- Line 27: same rewording of the 0.06% as line 42 above.
- Line 56: "133,009 ... roughly 54,300 ... near 440 to 1" - restate on non-pharmacy rows: 63,560 SUD denials, about 25,900 at 40.8%, about 209 to 1; note the 124 appeals are not restricted to non-pharmacy or post-service. "About 680x on Virginia's 0.06%" - flag as built on a non-rate.
- Line 62: "Virginia does not define ..." - keep; add "in any edition, 2020 to 2025."
- Line 64: replace "roughly 873,000 behavioral-health denials per year ... of which roughly 54,300 are substance use disorder" with about 87,000 and about 25,900 on non-pharmacy rows. Replace "applying 40.8% double counts and the whole estimate collapses" with: if the counts are final, the estimate is invalid as computed and the implied rework volume would be higher, not lower. State that Assumption 2 remains unresolved from public text, with structural evidence leaning initial.
- Line 70: "roughly 440 times the appeals market" - change to "roughly 200 times on non-pharmacy SUD rows, order of magnitude only."
- Line 83: mark the follow-up as answered "ambiguous per text" and link this brief.
Why this is in the vault
It closes the open Assumption 2 check behind the family RCM/EOB venture's Virginia-based rework sizing, and records that the sizing figures in the 2026-10-17 brief are overstated by the pharmacy rows before they get quoted to anyone.
Open follow-ups
- Does the Health Affairs resubmission study (Vabson, Hicks, Chernew) exclude Part D and pharmacy claims, so that its 40.8% share is only valid against non-pharmacy denial counts?
- Does any state parity data call (New York DFS section 343 workbooks, Illinois HFS, Texas HB 10 reporting) publish its carrier instructions with an explicit initial-versus-final definition of a denied claim, giving a defined-basis substitute for Virginia?
- What caused Virginia's outpatient Rx transaction count to swing from 19.7M (2022) to 45.6M (2023) to 17.9M (2024), carrier mix or a reporting-unit change, and does it affect the non-pharmacy rows too?
Related
- [[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]]
- [[2026-10-17-behavioral-health-silent-rework-market-size]]
- [[2026-10-17-state-medicaid-behavioral-health-denial-rates-by-category]]
- [[2026-10-01-national-rcm-eob-ai-behavioral-health-scan]]
Sources
- Vault: [[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]], [[2026-10-17-behavioral-health-silent-rework-market-size]], [[2026-10-17-state-medicaid-behavioral-health-denial-rates-by-category]] (read for the Virginia figures and the lines to correct). [[2026-10-01-national-rcm-eob-ai-behavioral-health-scan]] is listed as the venture's parent scan and was not re-read this pass.
- Virginia SCC Bureau of Insurance, 2024 Report (2023 data), RD712 - https://rga.lis.virginia.gov/Published/2024/RD712/PDF (source of 16,233,560; read in full for the claims, appeals and appendix sections)
- Same series, 2025 Report (2024 data), RD721 - https://rga.lis.virginia.gov/Published/2025/RD721/PDF
- Same series, 2023 Report (2022 data), RD582 - https://rga.lis.virginia.gov/Published/2023/RD582/PDF
- Same series, 2022 Report (2021 data), RD615 - https://rga.lis.virginia.gov/Published/2022/RD615/PDF
- Same series, 2021 Report (2020 data), RD408 - https://rga.lis.virginia.gov/Published/2021/RD408/PDF
- Same series, 2020 Report (2019 data), RD335 - https://rga.lis.virginia.gov/Published/2020/RD335/PDF
- JLARC staff memorandum, "Oversight of Mental Health Parity in Virginia" (December 1, 2020), RD318 - https://rga.lis.virginia.gov/Published/2021/RD318/PDF (searched; describes the data collection but gives no denied-claim definition)
Research budget used: QMD 0 of 5 (parent briefs named in the task and read directly); WebSearch 3 of 3 (all hunting the data-call instrument, none found it); WebFetch 0 of 3 (seven primary PDFs pulled with curl and searched locally with pdftotext).
Not reached: the Bureau's data-call template or carrier instructions (not found in public search); the 2018 and 2019 editions (scc.virginia.gov links return 404, no Wayback capture); the Health Affairs study at source. Editions other than 2024 were term-searched and read at the claims-methodology passages, not cover to cover.