06-reference/research

behavioral health silent rework market size

2026-10-17·research-brief·source: deep-research·by Ray Data Co (deep-research synthesis)

40.8% of denied claims are resubmitted and paid with no appeal involved -- a peer-reviewed number, from Medicare Advantage, with no behavioral-health breakout anywhere

The question

"What share of initially denied behavioral-health claims are resubmitted and paid WITHOUT a formal appeal -- i.e. how big is the quiet rework market, as distinct from the near-zero appeals market?"

This is open follow-up #4 of [[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]], which established the appeals market as effectively nonexistent and then declined to estimate the rework market. This pass finds the quantity exists and is published, but only on an all-specialty Medicare Advantage base.

What we already know (from the vault)

What the web says

The volume question (share of denials resubmitted and paid). This is a percentage, not a dollar figure.

The cost question (dollars per reworked claim). This is a dollar figure, not a share, and none of it can be converted into the volume question.

Convergences and contradictions

Synthesis for RDCO

The answer to the question as asked is a split verdict: the quantity is published and large, and the behavioral-health version of it is a clean null. 40.8% of initially denied claims are resubmitted and paid with no appeal, measured on 270 million Medicare Advantage claims from 2019 by Vabson, Hicks and Chernew. Nobody publishes that share for behavioral health, applied behavior analysis or substance use disorder. Health Affairs has zero behavioral-health mentions; KFF cannot link denial reasons to services ("neither the share of total claims denied for a given reason nor the type of service most often denied can be calculated"); Virginia does not define whether its "denied" counts initial determinations or final status after resubmission, and does not report resubmission at all. The measurement is coming on a known date: KFF reports CMS will add a behavioral-health versus non-behavioral-health split of claims received and denied for plan year 2027. Until then, any behavioral-health rework share is a borrowed number.

Best defensible bound, with the arithmetic shown and every assumption labelled. Assumption 1: the behavioral-health resubmit-and-paid share approximates the all-specialty Medicare Advantage share of 40.8%. Direction of error is arguable in both directions and I will not pretend otherwise. Pushing it higher: Virginia attributes 20.3% of substance-use-disorder denials to provider and administrative billing and 32.9% of non-pharmacy substance-use-disorder denials to preauthorization or precertification, against only 1.4% to medical necessity, and correctable-cause denials are exactly the ones a resubmission fixes. Pushing it lower: a missing prior authorization is often unobtainable retroactively, and behavioral-health practices are smaller with thinner billing capacity than the hospital and Medicare Advantage mix behind the 40.8%. Treat 40.8% as a midpoint, not a floor. Assumption 2: Virginia's 133,009 substance-use-disorder and 2,006,618 mental-health denials are initial denials. If they are instead final post-resubmission denials, applying 40.8% double counts and the whole estimate collapses. This is the largest single threat to the arithmetic and the parent brief already flagged the ambiguity. On those two assumptions: roughly 873,000 behavioral-health denials per year are reworked and paid inside Virginia's state-regulated fully-insured commercial market alone, covering 2.48 million lives, of which roughly 54,300 are substance use disorder. Do not scale that to the nation off a population ratio. Virginia's segment excludes Medicaid managed care, which is where the larger substance-use-disorder book sits, so the figure is a floor measured on the wrong denominator rather than a unit that multiplies.

Separately, and never multiplied into the above as if it were the same quantity: the labor price of one rework event is roughly $25 and has no published sample behind it. The one arithmetic worth doing is an order-of-magnitude labor-spend sanity check, and it is arithmetic on an unsourced multiplicand: 54,300 reworked substance-use-disorder claims times $25.20 is about $1.4 million of annual rework labor, and the full behavioral-health figure is about $22 million, both inside one state's commercial segment. Read the shape, not the digits. Two things that shape tells us. First, a rework-automation product's addressable price per claim is bounded well below $25.20, because that figure is the provider's internal cost and the product has to beat it, not match it. Second, the economic character of the two products differs: automating rework on the 40.8% that already gets fixed sells cost takeout against a $25-ish internal line item, while preventing the non-recoverable denial sells recovered revenue. Recovered revenue is the easier sale, the larger per-unit value, and the one that survives a procurement conversation. That is a sharper product distinction than either brief had.

Verdict on the parent brief's relocation finding: supported, and from three directions it did not have. One, the Health Affairs resubmission structure itself is evidence the first submission was defective rather than the claim meritless, since payers reverse two-thirds of resubmissions. Two, "fewer than 1 percent of claims will get resubmitted conditional on a second denial" means the loop is one shot, so there is no downstream tail to build a durable product on. Three, and most directly, Change Healthcare's first-party data says 82% of denials are potentially avoidable, front-end causes are more than 41% of denials, and the single largest root cause, registration and eligibility at 22%, is frequently non-recoverable. Non-recoverable is the operative word: once the claim goes out with bad eligibility or a missing authorization, no amount of downstream parsing gets the money back. That is the healthcare analogue of the Cone of Accounting Bullshit in [[2026-09-17-cfo-secrets-double-entry-bookkeeping-ramp]], and it makes the convergence concrete rather than rhetorical. Shrink the gap between the clinical encounter plus its authorization and the first claim submission; do not build a better reconstructor. The honest qualification is that the Health Affairs authors explicitly could not separate clerical from clinical denials, so the strongest version of the upstream claim rests on Virginia's cause table and Change Healthcare's avoidability split, not on the 40.8% itself. The 40.8% sizes the labor. It does not prove where the leak is.

What this does to the family venture's shape. The rework market is real, roughly 440 times the appeals market in volume on Virginia's substance-use-disorder rows, and it is the right market to be in. But the pricing unit should be the prevented denial, not the processed denial, and the wedge should sit at authorization capture and first-pass claim construction for substance use disorder specifically, where Virginia's preauthorization gap runs about 6.6x medical/surgical. The honest risk to name to Michael and Christian is that this entire estimate borrows its central percentage from Medicare Advantage because the behavioral-health version does not exist until plan year 2027 data publishes.

Why this is in the vault

This closes open follow-up #4 of [[2026-10-04-behavioral-health-claim-denial-rate-primary-sources]] and is the sizing input for the family RCM/EOB AI venture with Michael and Christian: it decides whether the product's pricing unit is the processed denial or the prevented denial, and it supplies the one published percentage (40.8%) plus the two dates (plan year 2027 CMS behavioral-health split, and the 2019 vintage of the only measurement) that any pitch built on rework volume has to disclose.

Open follow-ups

Related

Sources

Vault

Primary, peer-reviewed, full text read at source

Primary, first-party industry publication, original PDF downloaded and searched

Secondary

Cost-figure provenance trail, vendor-published

Flagged paywalled, not retried

Research budget used: QMD 2 of 5; WebSearch 3 of 3; WebFetch 3 of 3 (2 returned HTTP 403, 1 usable); 2 Playwright retrievals and 1 direct PDF download as documented 403 workarounds; 1 Europe PMC application programming interface call; 1 sub-agent for parent-brief extraction.