Every health system or hospital billing team has a pile of claims they don’t ever get to.
Ask any director of revenue cycle and they will describe it: small dollar denials, high volume, each one worth less than the effort of fighting for it. The pile ages, the timely filing windows close, and the balances get written off.
I want to be clear that this is not laziness, and it is not bad management. It is arithmetic, and until recently the arithmetic was correct.
The rule your team is actually following
Nobody writes it down, but every experienced biller runs this calculation:
Claim value, times the probability you win, versus the cost to rework it and how long till timely filing.
If the left side is smaller than the right, you walk away. That is a rational decision. A team with a finite number of hours should put those hours where they return the most money.
So how big is the right side of that equation? The published estimates vary, which tells you something in itself:
- $25 to $117 to rework a denied claim, depending on the study (MGMA, CAQH, and HFMA figures)
- Up to $181 at hospitals, per the Journal of AHIMA
- $57.23 as the average administrative cost of reworking a denial in 2023, up from $43.84 in 2022, per Premier
Take the middle of that range. Now price out a $90 denial with even odds of being overturned. You are spending something like $57 of labor to chase $45 of expected value. You should not do it. Your team is right.
So abandonment concentrates in high-volume, low-dollar denials. The per-claim economics decide it, not the total dollars, which is why the pile can be enormous in aggregate and still be correctly ignored one claim at a time.
Three forces push the same direction
Dollar size. Big claims get worked first.
The clock. You cannot touch all of them before they age out past timely filing, so you triage toward the ones still inside the window.
Capacity. Even inside the window, there is a fixed number of claims a team can research in a day, a week, even a month.
Put those together and you get the outcome everyone in this industry recognizes. In hospitals it can be explicit. I know plenty of organizations where outstanding accounts receivable under roughly $1,000 simply does not get touched, because the same person could be working the denial on a three-week inpatient stay instead. That is the right call with the tools they have.
What happens to the pile
The industry data here is uncomfortable. AHIMA figures, widely cited, put the share of denied claims that are never reworked at somewhere between 35 and 65 percent. MGMA and AHA numbers land near the top of that range, around 65 percent.
Now hold that against what happens when denials do get worked: roughly half get overturned and paid, according to Premier.
We have to be careful here, because those are two different studies and multiplying them together is not a real statistic. But the direction is not ambiguous. A large share of denied claims are never worked, and a large share of the ones that do get worked turn into cash. That is a lot of collectible revenue sitting in a pile that nobody can justify touching.
And notice the trap: the reason it is not worth working is the cost of the work. Not the merit of the claim. The claim might be perfectly good.
What actually changed
Two things have changed.
- The cost to work a denial
- The how fast you can do it
Cost
Almost all of the cost in that $25 to $117 range is diagnostic. It is a person figuring out what happened: checking the clearinghouse, then the payer portal, then the practice management system, then the lockbox, then waiting on hold with the payer. The fix at the end is usually fast. Getting to the fix is what takes time and costs money.
That research is exactly what AI agents are good at. An agent can check every channel on every claim, tirelessly, and do this across many claims at once. Agents can hand a biller a fully synthesized answer instead of a starting point. Across the workflows we have touched, we drive at least a 70 percent reduction in the cost to collect in those specific workflows.
Now go back to the equation. When the cost to rework falls by that much, every claim that was just below the line moves above it. The threshold drops. That is the entire thing. Not a clever appeal strategy, not a better letter template. The economics of touching a small dollar denial changed, and the pile stopped being rational.
For NextGen RCM Services, Claim Status took billers from 2,500 to 10,000 claims processed per month in a single quarter, at 75 percent lower cost per claim status. The unlock there was not speed for its own sake. It was that the long tail became reachable. Read the case study.
Speed
We have run 4,000 to 5,000 claim statuses in a couple of hours, complete with full denial root cause analyses across payer medical necessity policies, front end denials, bundling, appeals, etc. These denials can each take 20 to 30 minutes to diagnose. Substrate collapses this research to seconds.
What this is worth, honestly
Two numbers matter to a CFO, and they should stay separate. Do not add them together.
The first is recovery. Of the bad debt that is genuinely recoverable, you really can cut it by 75 to 80 percent. Good, in my experience, is bad debt at about 2 percent of net revenue. We have seen 6, 7, 8 percent and higher. If you are a reasonably sized organization, the gap between those two numbers is millions of dollars.
The second is cost to collect, which falls for the reason described above. It is a real number and it is not the same number as recovery.
Now the caveats, because I would rather you hear them from me.
I do not think you can recover 100 percent of claims, and I would not trust anyone who says otherwise. Some denials should exist. Practices make mistakes, and that is real.
Some bad debt is not recoverable at any cost. If a service needed a prior authorization and nobody got one, that claim is not going to be paid, and no amount of appealing changes it. That is an operational problem upstream of billing, not a denials problem.
What is recoverable is the honest mistake, the eligibility denial where a member ID was wrong, the policy denial where the payer does not believe the encounter was documented the way their policy requires, even though it was. Those are recoverable at any dollar amount now, down to the tens of dollars.
And agents will not do all of it. Substrate ARC covers roughly 60 to 70 percent of the denials workflow today. Medical necessity fights, peer-to-peer reviews, and contract disputes that require a call to the payer's contracting representative are human work, and will always be.
What agents buy you is the capacity to work the entire queue, and the time for your best people to go take on the fights that actually need them.
Watch the demo
Your team runs RCM. We give them time and leverage. See our agents work a real denial, on demand, no meeting required: watch the demo.
