The most expensive thing a large provider group does with a denial is work it twice. Not lose it. Work it twice, in the wrong order, by two people who did not know about each other. And the denial category where this happens most predictably is the one that should never occur at all: a service the payer authorized in advance, delivered as authorized, then denied on the back end.
I want to stay narrow here, because the general advice about appeals is worthless at scale. Everyone knows to appeal. The question inside a group running eight or eighty locations is what a single appeal costs to produce, which appeals justify that cost, and what sequence keeps you from paying for the same overturn three times. The post-authorization denial is the cleanest case to work through, because the evidence already exists somewhere in your own systems. You are not building an argument. You are retrieving one.
What one appeal actually costs to produce
Price it as labor, not as a percentage of the claim. A first-level appeal on a denial of this type consumes: someone to read the remittance and classify the reason code correctly, someone to pull the authorization number and the approval documentation, someone to compare what was authorized against what was billed line by line, someone to draft the letter, and someone to track the response clock. In a small practice those are one person and forty minutes. In a group, they are three departments and a handoff, and the handoff is where the money goes.
Every handoff adds queue time, and queue time adds re-reading. The second person re-reads what the first person already read because the note in the system says "appealed" and nothing else. That re-reading is the single largest hidden cost in the function. It does not show up on any report because nobody logs it as work. It shows up as a denial inventory that never falls below a certain floor no matter how many staff you add.
The other cost driver is the age of the claim when it reaches a human. A denial worked at eleven days costs a fraction of the same denial worked at sixty, because at eleven days the clinical documentation is still in front of the people who created it, the scheduler remembers the call, and the authorization portal still shows the transaction. At sixty days you are reconstructing. Reconstruction is where a fifteen-minute task becomes a two-hour task, and where a group starts outsourcing denial management not because it lacks staff but because its own oldest inventory has become more expensive to touch than to hand over.
The order, and why the first step is not an appeal
For an authorized-then-denied claim, the sequence that costs least runs like this.
- Classify the denial against the authorization, not against the code. A large share of these are not clinical disputes at all. They are mismatches: the authorization was issued to a different tax ID within your own group, or to a rendering provider who was swapped on the day, or for a unit count the delivered service exceeded by one. That is a correction, not an appeal, and correcting it costs a tenth of appealing it.
- Check whether the payer has a reopening or claim-correction path. Many do, and it sits outside the formal appeal ladder with a shorter turnaround and no consumption of your appeal levels. Groups routinely burn a level-one appeal on something a corrected claim would have fixed in a week.
- Request a peer-to-peer only if the denial is genuinely medical necessity. It is the most expensive step you have, because it spends physician time. Spending it on an administrative mismatch is the worst dollar in the whole process.
- File the formal first-level appeal, with the authorization approval attached as the first exhibit and the medical record second. The argument is short: you approved this, we delivered this, here is the transaction.
- Preserve the second level and external review as real options rather than afterthoughts, and calendar them from the date on the determination letter, not the date you opened the envelope.
Where the lens of scale changes the math
A solo practice appeals a claim. A group appeals a pattern, and if it does not notice the pattern it pays retail for every instance.
Say four hundred denials in a quarter carry the same reason code and the same payer. Worked one by one, that is four hundred letters. Worked as a pattern, it is one conversation with the payer's provider representative, one root cause, and usually a project fix on your side: an authorization field that maps to the wrong entity, a scheduling template that lets a swap happen without re-verification, a location added to a contract that was never loaded into the payer's system. The four hundred still need to be refiled, but refiling under a known fix is clerical. Appealing four hundred times individually is not.
This is the argument for a group structure that a single practice cannot use: separate the people who work volume from the people who work causes. If the same staff do both, causes never get worked, because the queue is always louder. Groups that split the function tend to find that a small cause team pays for itself against one or two payer-level fixes a year, and that the volume team's cost per touch falls once the recurring categories stop arriving.
The deadlines that set the ceiling on everything
Cost discipline is meaningless if the clock runs out, and in a multi-site group the clock is rarely one clock. Each payer contract sets its own appeal window, some as short as a few months from the remittance date, and those windows do not align across the contracts a group holds. Medicare Advantage plans run on their own rules. For employer self-funded plans, the claims and appeals procedures sit under federal benefits law that the Department of Labor is responsible for overseeing, which means the member-side appeal rights and timelines may differ from what your contract says about provider appeals. Both matter, and they are not the same ladder.
The practical move is a single deadline table, maintained centrally, keyed to payer and plan type, that every site reads from. Not a binder in the business office. A field on the denial record that populates automatically and turns red before it matters. The most costly denials I see in large groups are not the ones that were argued and lost. They are the ones that were correct, provable, and filed nine days late by someone reading the wrong contract.
What good looks like on a report
Three numbers tell you whether the function is healthy, and none of them is the overturn rate on its own. Average age at first touch, because it drives the cost of every step afterward. Percentage of denials resolved without a formal appeal, because that is the corrections path working. And the count of distinct recurring reason codes, because a falling count means causes are being closed rather than instances being processed.
A group that watches those three will find its appeal volume shrinking while its recovery holds steady. That is the outcome worth paying for: fewer appeals, filed earlier, on the merits, by people who touched the file once.
The post-authorization denial deserves this much attention because it is the category where the provider is unambiguously right and the money is most often left behind anyway. Get the order right, fix the cause behind the pattern, and the same staff you have now clear more of it in less time.