Verified late August 2026 · every figure third-party and sourced
Prior-auth denial statistics, with receipts.
Denial numbers get quoted constantly and sourced rarely — and the most-quoted figure isn't even about prior auth. Here are the numbers that survive checking: what payers now publish under federal rule, what that rule still excludes, and what the oncology studies actually found.
The numbers that survive checking.
~12–14%
the median share of prior-auth requests DENIED by marketplace (QHP) payers in the first wave of federally required disclosures — with a huge spread across payers, from under 2% to over 27%
Payer PA metrics published under CMS-0057-F (first disclosures, March 2026)
~52.6%
the median share of appealed prior-auth denials that payers OVERTURNED in the same disclosures — most denials that get contested don't survive contest
Payer PA metrics published under CMS-0057-F (first disclosures, March 2026)
~18–20%
the often-quoted marketplace denial figure — which is CLAIM denials from HealthCare.gov transparency data, a different number that articles routinely conflate with prior-auth denials
KFF analysis of HealthCare.gov claims transparency data
71%
of denied cancer drugs at one academic hematology/oncology center were oral oncolytics — the pharmacy-benefit drugs patients pick up themselves
Published academic hem/onc center study (JHOP)
52%
of those oncology drug denials were for ON-LABEL use — care the evidence already supports, lost administratively rather than clinically
Published academic hem/onc center study (JHOP)
60% / 80%
of appealed Medicare Advantage cancer-therapy denials reversed when the denial reason was "not medically necessary" (60%) or "experimental/investigational" (80%), 2020–2025
JCO Oncology Practice analysis, 2026
How to read denial numbers without being fooled.
What payers must publish now
Since March 31, 2026, impacted payers must publicly report prior-auth metrics — approval, denial, and appeal-overturn rates and decision timelines — under the federal interoperability and prior-authorization rule (CMS-0057-F). The first wave is why payer-level denial numbers finally exist in public.
What the rule does NOT cover
Drugs. CMS-0057-F explicitly excludes drug prior authorization from both the API and the metrics. A separate proposed rule (CMS-0062-P) would fold medical-benefit drugs in starting October 2027 if finalized — until then, public PA metrics are about imaging, procedures, and services, not oncology drugs.
The two numbers people mix up
A claim denial happens after care (the bill gets rejected); a prior-auth denial happens before care (permission is refused). The ~18–20% marketplace figure is claims; the ~12–14% median is prior auth. Quoting one as the other is the most common error in coverage of this topic.
Our own outcome accounting is public: how we count appeal outcomes — and the questions to ask any vendor are at how to evaluate prior-auth software.
The denial-statistics questions.
What percentage of prior authorizations are denied?
In the first federally required disclosures (March 2026, under CMS-0057-F), marketplace payers reported a median prior-auth denial rate around 12–14% — with an enormous spread, from under 2% to over 27% depending on the payer. Payer-level variation is the real story, not the median.
How often are prior-auth denials overturned on appeal?
The same disclosures show a median appeal-overturn rate around 52.6% — and in oncology specifically, a 2026 JCO Oncology Practice analysis of Medicare Advantage cancer-therapy appeals found 60% of "not medically necessary" denials and 80% of "experimental/investigational" denials were reversed. Most contested denials don't survive contest; most denials are never contested.
Is the famous ~18–20% marketplace denial rate about prior auth?
No — that figure is claim denials (bills rejected after care) from HealthCare.gov transparency data, per KFF's analyses. Prior-auth denials are a different, separately reported number. The two are conflated constantly, including in press coverage.
Do the new public metrics cover cancer drugs?
Not yet. CMS-0057-F excludes drug prior authorization from its API and metrics requirements; the proposed CMS-0062-P rule would add medical-benefit drugs from October 2027 if finalized. Today's public PA metrics cover services like imaging, radiation, and procedures — oncology drug denial data still comes from studies and practice-level records, not federal disclosures.
How does Insurf count its own appeal outcomes?
By published rules, not marketing: denominators, outcome categories, physician attestation, source snapshots, and small-n suppression are all public at insurf.io/methodology. We publish no outcome percentages of our own on this page — every figure above is third-party, with its source attached.
The contested tail is where the money is lost.
Every figure on this page is third-party and carries its source; medians are labeled as medians and proposed rules as proposed. Inveto prepares source-cited prior authorizations and denial appeals; your team controls submission; no payer outcome is ever guaranteed. This page is an educational reference — not legal, medical, or coverage advice.