HealthcareFree
Why Prior Authorizations Get Denied
Every submission joined to its outcome, so the one missing element that actually predicts denial at each payer stops hiding behind a stated reason.
River's denial to authorization feedback loop reads your submission history alongside its outcomes and works out which missing documentation element actually predicts a denial, payer by payer and service by service. It counts approvals as well as denials, because a rate needs a denominator. And it mines the cases you resubmitted and got approved, since a case that flipped after one element was added is the closest thing your history holds to a controlled experiment. Those pairs carry more information than every reason code combined.
Every denial report that ranks for this search makes the same two mistakes. It tabulates the reason the payer stated, which describes their process rather than your submission, and it looks only at denials, which leaves it no control group. So an element missing from half your denials looks like the cause when half your approvals are missing it too. That is a base rate, not a finding, and acting on it changes nothing.
This is about authorizations decided before the service, not claims denied after one. Reading remittance advice for paid claims is a denial root cause analysis, a different pipeline with different data and a different remedy. This one is for prior authorization teams, practice managers and anyone who has watched the same service get denied at the same payer for the fourth time and suspected there was a rule nobody ever wrote down anywhere.
The payer now publishes the number you need
Two rules changed what is learnable here. From January 2026 a Medicare Advantage denial sent to a provider must state a specific reason, so the reason field stops being a shrug. The same section requires plans to post their own annual approval rate, denial rate, appeal overturn rate and median decision time on a public website by March 31. That is an external benchmark that no practice has ever had. The list of services requiring authorization has to be posted with it.
Sedgemoor Orthopaedics ran 1,842 submissions in a year, 231 of them denied. At one payer, 384 submissions produced 43 denials. A functional score was absent from 21 of those 43, nearly half, which looks like the answer. But 174 of the 384 submissions omitted it, and the denial rate moved from 10.5 percent to 12.1 percent. A difference of 1.6 points. The share of denials was almost entirely base rate. Acting on it would have added a field to every submission and changed nothing.
The element that mattered was a dated conservative care trial. Present on 288 submissions it drew a 4.5 percent denial rate. Absent on 96 it drew 31.3 percent, a gap of nearly 27 points. The resubmission pairs confirmed it: of 47 cases resubmitted with that one element added, 44 were approved. Of 18 where only a functional score was added, 8 were. Same payer, same services, one variable at a time. Rewriting the narrative without adding anything worked on 14 of 31.
How it works
Send the history
Every authorization submitted over a period, with what went in it and how it came back.
Include the approvals
Not just the denials. Without the approved cases there is no denominator and no finding.
Split by element
Denial rate with each documentation element present and absent, per payer and service family.
Change the submission
One change per payer, the one with the largest measured gap and enough cases behind it.
What you get
- Denial rate with each element present against absent, at each payer separately
- Approvals counted alongside denials, because a predictive rate needs a real denominator
- Resubmission pairs mined as natural experiments, since one element changed and the outcome flipped
- Your rates set against whatever the plan published on its own website
- Cell counts shown on every finding, and no predictor named below the minimum
- The specific submission changes to make next, ranked by how much they move
Common questions
Why do you need the approvals?
Because without them there is no rate, only a share. An element absent from half your denials means nothing until you know how often it is absent from your approvals. Sedgemoor's functional score was missing from 49 percent of denials and 45 percent of everything, which made it noise dressed as a cause.
Is this the same as analysing claim denials?
No, and conflating them wastes both. An authorization decision happens before the service, on a submission you controlled. A claim denial happens after, on a remittance with its own codes and its own remedies. Reading those is a denial root cause analysis. The data, the deadlines and the fix are all different.
How many cases before a finding is real?
More than most reports admit to having. Every cell gets its count printed next to it, and nothing is named as a predictor below the minimum you set. A twenty point gap on six cases is a coincidence with a confident label on it, and a report that hides the count is how it gets acted on.
What makes a resubmission pair so useful?
It holds everything constant except the thing you changed. Same patient, same service, same payer, one element added, and the outcome moved or it did not. That is as close to an experiment as an authorization log ever gets, and most denial reports discard the resubmission history entirely before they start counting.
Can I compare my rates to the payer's published ones?
On the aggregate, yes, and that is worth doing. The posted figures cover all items and services together, so they cannot be read service by service. Medicaid plans separately owe consistent application of review criteria. Sedgemoor's overturn rate on appeal was 34 percent against a published 62.
We submit by fax and portal, not electronically. Does this still work?
Yes, because the element list does not come from the transport. The adopted standard for a referral certification and authorization transaction already enumerates what a submission is made of, so the same fields describe a faxed packet. What you lose is automatic capture, not the analysis.
What do we actually do with the finding?
One change per payer, starting with the biggest measured gap that has enough cases behind it. Where the requirement itself is unclear, read it out of the policy with a payer policy criteria summary first. Where a denial is already in hand, the argument goes in a medical necessity letter.
Why Prior Authorizations Get Denied
Fill in the form and your workspace opens with the work already underway.