I spent over a decade running clinical operations at infusion centers, building MPR processes from scratch, training reviewers, and cleaning up the messes that happen when the process breaks down. The worst of those cleanups took 18 months and nearly half a million dollars to resolve.
When our organization acquired another infusion center, I started reviewing their patient files. They hadn't been doing real medical policy review. Referrals came in, documentation came with them, and the team submitted for prior authorization with whatever they had. Payers were approving claims. No one flagged anything.
Then the recoupments started.
Payers had approved the claims. Then they audited them. The documentation didn't hold up. We spent the next 18 months tracking down records for claims that had been paid and closed, calling providers who had moved on, chasing files from offices that had changed their systems. We got the money back eventually. But I had never worked harder to fix something that should have been right in the first place.
That experience changed how I think about medical policy review. Done well, it is some of the most valuable work an infusion center does. Done poorly, the liability sits quietly in your billing until an audit finds it.
What medical policy review actually is
Medical policy review is the step between receiving a referral and submitting for prior authorization. Its job is to confirm that the patient's clinical documentation actually supports what you're about to ask the payer to approve, and that nothing in those records could later be used to deny or recoup the claim. It sounds like a clerical task. It isn't.
Not all drugs are the same
Most people outside of specialty pharmacy operations assume this is straightforward. They think of it as a checklist. And for some drugs, it more or less is. Ocrevus, for example, has a relatively clean set of criteria. The patient has to be 18 or older, have a confirmed diagnosis of multiple sclerosis, and not be on any of a list of contraindicated therapies. You can work through that review in 20 or 30 minutes if the chart is organized.
Then there are drugs like tepezza.
Tepezza is a treatment for thyroid eye disease, and doing a clinical review for it is one of the more demanding tasks in specialty infusion. The payer criteria require documentation from an ophthalmologist and an endocrinologist, specific lab values showing Free T3 and Free T4 levels (the active thyroid hormones) within defined limits, proptosis measurements, a full clinical activity score with the complete scoring breakdown, a baseline hearing test, and an attestation that the patient is not pregnant. The physician prescribing tepezza is rarely any of those specialists. You're coordinating documentation from at least two other providers, each of whom may have no idea what the other sent, and each of whom has their own documentation habits.
The part that trips up even experienced reviewers is the Clinical Activity Score, which quantifies the degree of eye inflammation a patient is experiencing. An endocrinologist will put "CAS score: 4" in a note. That is not enough. The payer wants to see the full scoring chart, each component listed individually, with the rationale for how the score was reached. A note that just states the number will get denied. I've seen this happen to clinics with otherwise solid processes. They knew the requirement existed. They just didn't catch that the document they received didn't fully satisfy it.
Approval does not mean you're done
Payers can approve a claim, pay it, and then come back months or years later and ask for the underlying clinical documentation to support medical necessity. If you don't have it, or if what you have doesn't fully satisfy the criteria, they can recoup the payment. That's what happened with the practice we acquired. Their claims had been approved. Nobody had caught the gaps. Then the audits came.
This is why the output of a medical policy review matters as much as the review itself. When my team does a clinical review, we log not just what the patient has, but where in the chart we found it. Specific page numbers. The billing team can pull up a clinical review from two years ago and go directly to the documentation that answers a payer's question. That dramatically reduces the time it takes to respond to a recoupment request, and in many cases it prevents the recoupment from succeeding.
Where most reviews actually break down
The other place where MPR breaks down is communication with referring providers. When you call or message a provider and ask for "clinical notes," you get clinical notes. You get whatever they think is relevant, in whatever format they keep their records. What you don't get is the thyroid stimulating hormone (TSH) measurement from within the last 90 days, or the ophthalmology note with the proptosis measurement taken before treatment initiation, or the specific audiogram your payer requires for tepezza coverage.
Specificity matters more than people realize. When I tell a referring office that I need the free T3 and T4 from within 90 days, the CAS score with the full component breakdown, and the most recent proptosis measurement, I get what I asked for. When I ask for clinical notes, I get the same packet four times. The back-and-forth adds days or weeks to a patient's wait time and creates friction with referral sources who get frustrated by repeated follow-up requests.
A lot of providers genuinely don't know what they need to send. They're busy. They assume someone downstream will sort it out. Some of them have sent records to other practices for years without ever being told exactly what was missing. Getting specific with them early, and being consistent about it, eventually trains the relationship. The referral office starts anticipating what you'll need. That doesn't happen overnight, but it does happen.
Always go to the source
On the question of policy versions: payer medical policies change, usually once or twice a year. The change is often small. A lab draw window that was valid for six months gets tightened to 90 days. A criterion that was implied gets written explicitly. If you're working from a PDF you pulled out of a search result six months ago, you may be reviewing against criteria that are no longer current.
The safest practice is to maintain a list of links to the policy pages on each payer's website, not links to the documents themselves. The page always reflects the current version. The document may not. It takes some upfront work to build that database, but once it exists, it's easy to maintain. Most policies don't change dramatically from update to update. When they do, it's usually visible at the top of the policy: the effective date will have changed.
What the front end is worth
I've worked in infusion long enough to do this the wrong way and the right way, and the difference downstream is significant. The extra time spent on the front end, confirming every criterion against the actual documentation, making specific requests to referring providers, logging page numbers in the clinical review, shortens turnaround on PA submissions, reduces denials, and protects the practice when payers audit months or years later. It also makes things better for patients, who spend less time waiting and fewer months caught in appeals.
There's real potential for automation to help here, particularly in the document work: finding the relevant criteria, scanning records to identify what's present and what's missing, flagging page numbers. That kind of tooling, built well, could take most of the time out of a routine clinical review and let experienced staff focus on the cases that actually require clinical judgment. For drugs like tepezza and lemtrada, where the documentation is complex and the stakes of getting it wrong are high, there will always be a point in the process where a trained clinician needs to review the labs and confirm the patient is safe to treat. I don't think we're at the point where that judgment can be fully automated. But everything leading up to it can be made faster, more consistent, and less dependent on individual reviewers knowing exactly where to look.
