What Causes Imaging Prior Authorization Denials

What the payer actually approved

A prior authorization is not general permission to image a patient. It is a record with fields in it: a CPT code, a contrast protocol, a rendering facility, a date range, a patient and plan ID. When the claim arrives, adjudication compares those fields against the ones on the claim. Anything that doesn't line up produces a denial, and the remittance code that comes back — CO-197, authorization absent — reads the same whether nobody ever pulled an auth or somebody pulled a perfectly good one for CPT 70551 and the tech ran 70553.

This is why the reflex to defend the exam clinically usually misses. The payer has not questioned whether the patient needed an MRI. It has noticed that the MRI on the claim is not the MRI in its approval record.

That number covers all claims, not just imaging, and advanced imaging sits above the average. CT, MRI, and PET have been the standing target of utilization management for over a decade, and the review lists keep growing: payers continue to publish new prior-auth requirements for imaging codes going into 2026, including BCBSTX's 2026 program changes. Every code added to a review list is one more field that has to match at claim time.

The order does not hold still

Picture the ordinary path of one study. A referring office enters the order on a Tuesday. Benefits are verified Thursday. The RBM approval comes back Friday for a named code at a named site. The appointment is three weeks out. In week two the patient reschedules. In week three the scanner in that suite goes down and the study moves across town. At the table, the radiologist protocols contrast because of what the indication suggests.

Nobody in that sequence did anything wrong. But three of the fields the payer will compare have moved since Friday, and the approval record still says what it said on Friday. That is the mechanism behind essentially every preventable imaging denial: the study evolved, the authorization didn't, and the two were never re-reconciled.

The fields that move most often are a short list:

• Modality, body region, or protocol — with contrast instead of without, abdomen instead of abdomen and pelvis, a different sequence than the one described in the request.
• The date — RBM windows typically run 30 to 90 days, and a no-show plus a rebook can eat that quietly.
• The billed code — the code the ordering clinician chose and the code the coder lands on after reading the report are not always the same code.
• The site of service — hospital outpatient versus freestanding center is a different approval, even for the same patient and the same exam.
• The linkage itself — an auth number attached to the wrong order, the wrong encounter, or a stale insurance ID.

Each of these has its own failure mode and its own catch point; we walk through all five in detail in 5 front-end failures behind imaging denials.

A valid authorization is not a paid claim

The version of this that frustrates RCM teams most is the study that had a clean, on-file, correctly obtained authorization and got denied anyway. Premier's member survey put initial denials on private-payer claims at roughly 15%, and some share of those had already cleared prior authorization before the claim went out. The auth was real. Something downstream of it moved.

Which makes "did we get the auth?" the wrong question to close out at scheduling. The useful question is whether the auth still describes the study you are about to perform, asked on the morning of, not two weeks prior.

Why this is hard to catch by hand

The obstacle is arithmetic, not competence. An order changes hands at entry, scheduling, verification, protocoling, and acquisition, and a reconciliation pass has to happen after the last of those handoffs to be worth anything. Do that across a full day's schedule and you are asking a small team to hold six fields in mind, per study, at the busiest point in the workflow, for studies that will mostly turn out to be fine.

Audits of radiology request quality land in the same place. Missing and incomplete information on imaging requests keeps recurring even at sites that trained staff specifically to eliminate it, which is the signature of a manual check under volume rather than a knowledge gap. (Al Qassabi et al., 2025)

How to prevent it

The changes that move denial rates are scheduling-desk changes, not clinical ones. In rough order of payoff:

1. Re-verify the authorization against the final order — code, contrast, site — at the last point before the study runs, not only when the auth was first requested.
2. Make every reschedule a trigger. A new appointment date is a new question about the authorization window, and it is the cheapest denial to avoid.
3. Close the loop on protocol changes made at the table, so a contrast decision reaches whoever owns the auth the same day rather than surfacing in the coding queue weeks later.
4. Re-check the site whenever a patient is moved between locations for capacity or equipment reasons.
5. Give every mismatch a named owner and a resolution before the study is performed. A flag nobody owns is a denial with extra steps.

These are completeness checks, all of them, which is what makes the problem tractable without more radiologist hours or a bigger appeals desk. They belong to the same pre-read pass as the rest of a read-ready check.

Where Corena fits

Corena is a radiology operations agent that runs read-ready QC before a study reaches the radiologist. Part of that pass compares the order against the authorization record and flags where the two have diverged. Corena recommends and never interprets, diagnoses, or makes a clinical call; a named person on your team reviews and signs off on anything it raises. The scope is deliberately small: catch the divergence while the study is still in the queue and the fix is still cheap.

If prior-auth mismatches are showing up in your denial data, request a demo to see how Corena's read-ready checks work against your own order and authorization data.