The Care Stack

Denials are the revenue cycle problem data can actually move

Victor Wilson · · 4 min read

Revenue cycle leaders hear a lot of pitches about data. Better cash forecasting. Smarter patient collections. Some of these help at the edges. Many are hard to tie to cash.

Denials are different. If you’re going to put data effort anywhere in the revenue cycle, I’d put it here first.

Why denials are a good fit

Denials leave a trail. Every one comes back with a reason code, a payer, a date, and the claim that triggered it. That’s structured data you already have. You don’t need a new source to get started.

The causes repeat. Most denials aren’t random. They cluster around the same payers, the same services, and the same front-end steps. A missing authorization for one kind of imaging. An eligibility check skipped on a certain registration path. A coding pattern one payer keeps rejecting. When causes repeat, you can find them and fix them.

The money is easy to see. A denial is cash you earned and haven’t collected. When you prevent one or overturn one, the dollars land somewhere finance already tracks. You don’t have to argue about credit the way you do with softer projects.

This year you can also see how your payers compare. Under CMS’s prior authorization rule, many plans had to post their 2025 approval and denial rates by March 31. KFF analyzed that first data in August. Average denial rates for standard prior authorization requests were 12% in Medicare Advantage, 14% in Medicaid managed care and 18% in the ACA Marketplace. And 67% of appealed Medicare Advantage denials were overturned.

That last number is worth sitting with. Many denials don’t hold up when someone pushes back. Each one still cost staff time to fight.

Much of the fix is in your control. You can’t change a payer’s policy. You can change how your own teams register patients, get authorizations, document care, and code claims.

Where most organizations get stuck

Most health systems I’ve worked with already have a denials report. Usually several. The trouble is what those reports are built to do.

A typical denials dashboard shows how much was denied last month, by payer and by reason code. That’s a useful thermometer. It doesn’t tell anyone what to do on Monday morning.

Reason codes alone are too coarse. The same code can come from a dozen different root causes. To act, someone has to connect the denial back to where it started: which department, which registration workflow, which order, which step. That link is where the value is, and it usually lives across systems that don’t talk to each other.

The other common gap is follow-through. Appeals teams work the denials that come in. Fewer organizations close the loop by sending the pattern upstream to the team that caused it, with enough detail for that team to change something.

What I’d ask the data team to build

Start with a clean, joined view of each denied claim. The denial, the original claim, the encounter, the authorization record if there is one, and the registration details, all in one row you can trust. This is plumbing, and it’s most of the work.

Next, a short list of the top root causes, ranked by dollars rather than by count. A thousand small denials that get paid on resubmission matter less than a few dozen large ones that end up written off. Rank by money at risk and by how often each cause turns into a write-off.

Then, a worklist for each upstream owner. Patient access gets its eligibility and registration misses. The ordering departments get their authorization misses. Coding gets its own. Each list should be specific enough that someone can fix the process so the next claim goes through clean.

AI agents can take on much of the gathering. An agent can read a denial, pull the claim, encounter and authorization records, and suggest a likely root cause. A person confirms it. That keeps the joined view and the worklists current without an analyst stitching them together each week. What the agent can’t do is decide which process to change. That call belongs to the upstream owner.

Track two numbers over time for the causes you targeted: how often they still trigger a first-pass denial, and how many dollars from them get written off. Those are the numbers finance will believe.

Appeals still matter. They recover real money, and you should keep doing them well. But every appeal costs staff time, and not every one wins. A denial that never happens costs nothing to work. That’s why the upstream worklists are where the biggest gains sit. The goal is fewer denials arriving next quarter. A faster appeals queue is the smaller prize.

Be patient with this. Changing a registration workflow or an authorization step means training people and sometimes changing the EHR build. Expect the first results over a few months rather than a few weeks. Once a fix sticks, though, it keeps paying every month without anyone having to work it.

Here’s what to ask for. Go to your revenue cycle and data leaders and request the top five denial root causes by dollars written off over the last year, each with a named upstream owner.

If they can produce that list quickly, you’re in good shape. Start working it. If they can’t, building that list is the most valuable data project in your revenue cycle right now.

Discussion

    Join the discussion