The denial was wrong eight times out of ten. The platform that fixes it has to start there.
In Medicare Advantage last year, payers issued 4.1 million prior-authorization denials. When providers appealed, 80.7% were overturned. Most patients never appealed. The friction was the product.
We built Authora because three lines of that math do not square — and because, starting January 1, every payer in the country has to publish those numbers. The rule is no longer about whether you have an FHIR API. It is about whether the answer you publish at the end of the year is one anyone can defend.
U.S. administrative spend on prior auth, annually.
McKinsey, 2024
Medicare Advantage PA determinations issued in 2024.
KFF, 2024
Of appealed MA denials, overturned.
KFF, 2024
Per physician per week, on prior auth.
AMA, 2024
A prior auth platform is not a workflow. It is a record of judgment.
Every category of prior authorization software so far has tried to be one thing: a faster pipe to the payer, a cleaner inbox for the coordinator, a better robot for the portal. The category accepts the underlying logic — that PA exists, that denials are mostly correct, that the work is to move them along faster — and tries to optimize on top of it.
The KFF data refuses that frame. Eight in ten appealed denials are overturned. The initial determination is not a high-confidence judgment most of the time. It is a probabilistic decision rendered without the chart, made under load, and reversed on review when anyone bothers to look. A platform that simply moves those decisions faster is making a worse problem more efficient.
Authora was built on the premise that the only durable answer is to ground every determination in citable evidence — both the chart on the provider's side and the policy on the payer's side — and to make the trail visible to both audiences in the same surface. Not faster denials. Better-grounded ones. With a record that survives the audit.
This is also, not by coincidence, what CMS-0057-F starts requiring on January 1: specific reasons for every denial, public reporting of approval rates and overturn rates, and four FHIR APIs that make all of this auditable in real time. The platform that wins the next decade will be the one that publishes those numbers and stands behind them.
Three round-trips, all instrumented.
The HL7 Da Vinci stack defines the three handshakes that turn PA into a bidirectional FHIR conversation. Authora runs all three end-to-end and publishes the latency on each.
At order-sign in the EHR, a CDS Hooks card returns the payer's policy, the documentation it requires, and a SMART launch into Authora. The clinician knows in 200 milliseconds whether PA applies and what evidence the chart has to produce.
The payer's Questionnaire executes against the chart via CQL. Every criterion gets evidence pulled from the FHIR resources we already have — labs, encounters, imaging, prior conservative care — and the gaps are flagged before the request leaves the building.
A FHIR Bundle wraps the X12 278 transaction underneath. Every payer that accepts native FHIR gets the Bundle; every payer that still wants 278 gets the 278. The response — A1 Certified, A3 Not Certified, A4 Pended — comes back into the same case record, with the auth number and the period.
EHRs, payers, and the public-data plumbing that makes the chart speak.
We integrate at the layer that produces the evidence — not at the layer that produces the screenshot.
Every checkmark links to a specific line in the chart. Every denial cites the page in the policy.
Pre-flight check before submission. The platform tells you which criterion will fail and what document, page, or value would close it. The 7pm In Basket message asking for "more clinical info" stops happening because the gap was surfaced at the moment of order.
The case arrives pre-mapped. Each criterion sits next to the chart excerpt that supports it, with confidence scores, and the UM nurse keystroke is one of three: approve, request specific info, or escalate. Defensible determinations, not faster ones.
The deadlines on the wall.
Authora was built against the rule, not retrofitted to it. Public-reporting fields are first-class data; SLA telemetry is computed, not estimated; denial-reason specificity is enforced at the schema level.
7 calendar days standard · 72 hours expedited. Specific denial reasons mandatory regardless of submission method.
Volume, approval rate, denial rate, appeal volume, overturn rate, average decision time — by service category, by line of business.
Patient Access · Provider Access · Payer-to-Payer Data Exchange · Prior Authorization (PARDD).
Annual UM data collection: denial rate, denial reasons, approval rate, % of services subject to PA, appeal-overturn rate, timeliness.
The product is open. Inspect any case.
The worklist is seeded with ten real-shaped cases — auto-approved, in-review, peer-to-peer scheduled, denied, appeal-overturned. Click any row. Read the evidence. Pull the FHIR Bundle. Verify the X12 response. The work is here.