Revenue cycle

Your denials queue is always full. Is anyone fixing what fills it?

A denials queue recovers money one claim at a time. A prevention program finds the intake step, authorization habit or documentation gap that keeps producing those claims, gives it an owner, and checks whether the fix held. This is how to build the second one without abandoning the first.

Samuel Jean, Co-Founder14 November 202616 min read
The short version
  • A denials queue works claims one at a time. A prevention program works the upstream workflows that keep producing them.
  • Tag every denial with a root cause in your own workflow terms, not just the payer's reason code, because one reason code can come from several different failures in different departments.
  • Owners belong to workflows such as intake, utilization review, credentialing, documentation and coding, and every fix needs a named owner, a written change and a date to check it.
  • Measure a fix by date of service, against a denominator, once that cohort of claims has had time to adjudicate. Measuring by denial date will show you improvement that has not happened yet.
  • Count denials you recovered on appeal. An overturned denial still cost staff time and delayed cash, and the workflow that caused it is still running.

The appeal went out on time. The payer reversed the denial. The biller closed the item and opened the next one, which was the same denial for a different patient, caused by the same gap at intake, for the third month in a row. Nobody did anything wrong in that queue. The queue did exactly what a queue does: it recovered the money. What nobody did was ask why the claim was deniable in the first place, hand that question to the person who runs the workflow where it started, and check a few months later whether the answer changed anything. That is the difference between denial management and denial prevention, and most behavioral health programs have the first while assuming it produces the second. This guide covers how to build the second: a root-cause taxonomy, tagging that survives a busy month, owners who can actually change the workflow, fixes written so they can be verified, and measurement that tells you whether the fix worked rather than whether the calendar moved.

What a denials queue does, and what it cannot do

A good denials queue is a recovery tool. Every denied or underpaid line lands in it with the payer's reason, the dollars at risk, the appeal or corrected-claim deadline and a person responsible for working it. That is necessary, and if you do not have it, build it first. Appeal windows close quietly when denials live in a shared spreadsheet tab.

But a queue is organized around the claim, and its definition of done is that the claim is resolved: paid on appeal, corrected and resubmitted, or written off. None of those outcomes changes what happens at the front desk next Tuesday, or how the utilization review coordinator tracks units on the PHP census, or whether the new clinician's payer enrollment was submitted before her first session.

A prevention program is organized around the workflow. Its definition of done is that a category of denial stops appearing for services delivered after a specific change was made. That is a different unit of work, with a different owner and a different clock.

  • The queue asks: can this claim be recovered, by when, and by whom?
  • Prevention asks: why did this claim leave the building in a state the payer could deny, and what has to change so the next one does not?
  • The queue is staffed by billing. Most of the causes it sees originate somewhere billing does not control: registration, authorization, credentialing, documentation, scheduling.

If you want a refresher on the common denial categories themselves, the five denial reasons behind most write-offs covers them. This guide is about the machinery that sits around them.

Start with a root-cause taxonomy in your own words

The 835 tells you what the payer decided, using claim adjustment reason codes and remittance remark codes. Those codes are useful and you should keep them. They do not tell you where in your operation the problem began.

Take a denial for missing or invalid authorization. Inside your program, that one reason could mean the authorization was never requested; that it was requested for the wrong level of care; that it was approved but the units ran out because nobody was counting; that the patient stepped down from residential to PHP and the new level was never requested; or that the authorization was on file and simply not attached to the claim. Those are five different failures owned by at least three different people. If your analysis stops at the reason code, all five land on the same row of the same report and nobody can act on it.

So write a short list of root causes in workflow terms, short enough that a biller can choose the right one in a few seconds without a manual. Group them by the part of the operation that owns them, and avoid vague catch-alls such as billing error that tell nobody what to change. A starting set might look like this, adjusted for your levels of care and payer mix.

Front door: eligibility and registration

  • Coverage not verified before the first service
  • Coverage not re-checked after a change (new card, new plan year, reported change)
  • Subscriber or demographic data entered incorrectly
  • Other coverage or coordination of benefits not identified

Authorization and utilization review

  • Authorization not requested, or requested for the wrong level of care
  • Authorized units, days or visits exhausted before an extension was obtained
  • Level-of-care change without a new authorization
  • Authorization on file but not linked to the claim

Credentialing and payer enrollment

  • Rendering clinician not enrolled with the payer on the date of service
  • License, certification or supervision arrangement lapsed or not on file for the date of service
  • Billing entity, NPI or taxonomy mismatch for the location or program

Clinical documentation

  • Note missing, unsigned or not co-signed when billed
  • Documentation does not support the service or time billed
  • Treatment plan missing, expired or not covering the service

Coding, claim build and payer-side

  • Wrong service, revenue or procedure code, modifier or place of service
  • Units inconsistent with documentation or attendance
  • Timely filing missed, or duplicate or incorrectly coded corrected claim
  • Payer processing error, or a payer policy change not yet reflected in your workflow

Add one more tag: unknown, needs follow-up. It is a legitimate answer when the cause is not obvious, but every item carrying it should be looked at in the monthly review. If that pile grows, either the taxonomy is missing a category or tagging has become a formality. You can always split a category later when you see it is hiding two problems.

Tag at the point of work, and count the denials you won

The person working the denial should tag the root cause while the claim is in front of them. Tagging later from a report does not happen, or happens once and then stops. Each denied line should carry at least two fields: the payer's reason, taken from the remittance, and your root cause, from the taxonomy. A free-text line for the specific event helps later, for example authorization ended Friday, step-down to PHP Monday, PHP request sent Tuesday.

For high-dollar items and anything that repeats, add a second look by the owner of the suspected workflow. Billers usually identify the upstream cause well, but not always. Intake may know that eligibility was checked and the payer returned outdated information, which is a different problem from eligibility never being checked. The second look also keeps owners close to the evidence, which matters when you later ask them to change how their team works.

Recovered is not the same as prevented

The most common gap in denial analysis is that overturned denials disappear from it. The claim was paid on appeal, the item closed, and the root cause was never recorded. But a recovered denial still consumed staff time, delayed cash and sometimes required a clinician to pull records. If the workflow that produced it is still running, it is still producing more. Tag every denial, including the ones you win, and report recovered and unrecovered denials side by side.

Assign owners to workflows, not to denials

In the queue, every item has an owner, usually a biller, whose job is recovery. In a prevention program, every root-cause category has an owner who runs the upstream workflow, and whose job is recurrence. They do not work the appeal. They see the denials in their category every month, decide what to change, and report whether it worked.

  • Eligibility and registration: the intake or admissions lead
  • Authorization and utilization review: the UR lead or coordinator
  • Credentialing and payer enrollment: whoever maintains credential files and enrollment status
  • Clinical documentation: the clinical director, with supervisors for their teams
  • Coding, claim build and payer-side issues: the billing manager or a certified coder, with whoever manages payer contracts

The appeal is billing's job. The reason there was something to appeal is somebody else's.

The failure mode here is an owner who cannot change the workflow that causes the problem. Say authorization denials trace back to scheduling booking group sessions past the authorization end date. UR owns the category, but the change has to happen in scheduling. Name an accountable owner and the contributors whose processes have to move, and get the contributors into the review when their piece is on the agenda. An owner with responsibility and no authority will report the same problem every month, accurately, until they stop coming.

Keep ownership about workflows, not people. When the same error shows up across several staff members, it is a process problem, and coaching one person will not fix it.

Pick a small number of problems at a time

Once tagging has run for a month or two, you will have more root causes than capacity to fix them. Pick two or three per cycle, finish them, and measure them before adding more. A short list worked to completion is worth more than a long list that every owner nods at.

  1. Total denied dollars per root cause for the period, including denials later recovered.
  2. Count of denied lines per root cause. A cause with modest dollars and high volume can be eating more staff time than a single large denial.
  3. Estimated cost to work each denial in that category: a corrected claim is cheap, a medical necessity appeal with records is not.
  4. Whether the cause is fixable inside your organization. Payer processing errors are worth tracking and escalating, but they will not respond to a change in your intake script.

Be careful about ranking only by written-off dollars. Categories your team is good at appealing will look small on a write-off report and large on a staff-time report. Those are often the cheapest to prevent, because the cause is well understood.

Write the fix as a change someone can check

Remind intake to verify benefits is not a fix. Neither is be more careful with authorizations. A fix is a change to the workflow that names a trigger, a person, an action and a record that proves the action happened. If you cannot describe what you would look at to confirm the step was done, you have written an intention, not a fix.

  • Instead of re-verify coverage more often, write: coverage is re-checked for every active patient on the first business day of the month and whenever a new card is scanned; the result is recorded on the coverage record; the intake lead reviews the exceptions list weekly.
  • Instead of watch PHP authorizations, write: when remaining authorized days drop to a threshold the UR lead sets per payer, the case appears on the UR worklist with a due date for the continued-stay request; no PHP day is scheduled past the authorization end date without a recorded request status.
  • Instead of get new clinicians enrolled faster, write: no clinician is scheduled for a payer's patients until that payer's enrollment is recorded with an effective date; the credentialing coordinator owns the date and the scheduler checks it.

Fixes vary in how long they last. In rough order of durability: a system rule or hard stop; a required field or a checklist step that cannot be skipped; a scheduled routine with an owner; training; a reminder or memo. Training and memos decay as staff turn over. Use them, but pair them with something structural whenever you can.

Record every fix in one place: the root cause it addresses, the owner, what exactly changed, the date it took effect, what result you expect, and the date you will evaluate it. Without that last date, fixes are never revisited and the program turns into a list of things that were once decided.

When the fix belongs in a scrubbing rule

Some causes cannot be reliably prevented upstream, or need a backstop for the cases that slip through. Those are candidates for a pre-submission check: if a service is outside the authorization window, or the rendering clinician has no enrollment on file for that payer on that date, hold the claim and send it to the owner before it goes out. Seven claim problems you can catch before the payer sees them walks through common checks.

Two cautions. A claim caught by a scrubber is still an upstream failure; it is cheaper than a denial, but it is not prevention, so track scrubber holds by root cause with the same taxonomy. And a held claim is delayed cash and, depending on the payer, a timely filing clock that keeps running. Every hold needs an owner and a due date, just like a denial.

Measure whether the fix worked

This is where most prevention efforts go wrong, usually in the direction of declaring success too early. Four habits keep the measurement honest. For broader context on which revenue cycle numbers to track and how to count them, see revenue cycle KPIs for behavioral health.

Measure by date of service, not denial date

Denials arrive after adjudication, which can be weeks after the service. If you made a change on the first of the month and count denials received that month, you are mostly counting claims for services delivered before the change. Group claims by date of service and compare the cohort before the change with the cohort after.

Use a denominator

A count of denials goes down when volume goes down. Census drops over a holiday, a program pauses admissions, and the denial count improves without anything getting better. Express each root cause as a rate against the claims it could have affected: authorization denials per authorized-service claim submitted to that payer, enrollment denials per claim billed under newly hired clinicians. Pick the denominator when you write the fix, not after you see the numbers.

Wait for the cohort to mature, and check the step first

A date-of-service cohort is not complete until most of its claims have been adjudicated, and evaluating it early flatters the fix. Decide in advance how long you will wait, based on how your payers actually pay. Meanwhile, check the mechanism: of the admissions after the change, how many have the re-verification recorded? If the new step is happening inconsistently, the denial rate cannot tell you whether the fix would work, only that it was not done. This check is cheap and tells you something weeks before the denial data can.

Watch for displacement

Some fixes move a problem rather than remove it. Hold every claim without a signed note and documentation denials fall, but the hold queue grows and days to bill lengthen. When you evaluate a fix, look at the adjacent numbers: scrubber holds, unbilled services, days from service to submission, and patient balances where relevant.

When the evaluation is done, record the result in plain language: worked, partly worked, did not work, or step not consistently happening. A fix that did not work is useful information. Revise it with a new evaluation date, or retire it and try a different approach.

Run a monthly denial prevention review

The program lives or dies on a recurring meeting with a fixed agenda, the category owners in the room and decisions written down. An hour a month is usually enough if the preparation is done beforehand. Quarterly is too slow: by the time a pattern is discussed, more months of claims have been built the same way.

  1. New denials by root cause, grouped by date-of-service month, recovered and unrecovered side by side.
  2. Status of open fixes: was the change made on the date planned, and is the new step actually happening.
  3. Matured cohorts: results for any fix whose evaluation date has arrived, including adjacent numbers for displacement.
  4. Unknown tags: what the follow-up turned up, and whether the taxonomy needs a new category.
  5. Payer changes: new policies, recurring payer errors and anything that needs escalation.
  6. Next problems: if a fix has closed, pick the next root cause, name the owner and set the date for a written fix.

Keep minutes: decisions, owners, dates. Over a year, those minutes are a record of problems identified, corrected and checked, which is also what an effective compliance program is supposed to produce. A compliance program is more than a binder covers how auditing, monitoring and corrective action fit together.

One boundary to respect: some denial patterns are not just revenue problems. If the analysis suggests services were billed that documentation does not support, or that claims already paid may share the same defect, take it to your compliance lead and counsel rather than handling it as a billing fix. Obligations around identified overpayments vary by payer and program; confirm against the applicable regulations and contracts with counsel.

A worked example

This is hypothetical, but the shape will be familiar. Say a 30-bed residential program runs a PHP step-down. Over two months, billing works a cluster of authorization denials on PHP days from two commercial payers. In the queue, each one is handled correctly: some are overturned after a retro request, some are written off.

Tagging with the workflow taxonomy shows the pattern. Most are level-of-care changes without a new authorization, and the free-text notes say nearly the same thing every time: the step-down was decided at the Friday treatment team meeting, the patient moved to PHP on Monday, and the PHP request went to the payer Monday afternoon or Tuesday.

The UR lead owns the category. The clinical director is a contributor, because the decision point is the Friday meeting, and so is scheduling, because PHP groups were booked before any request existed. The written fix:

  • Step-down decisions at treatment team are recorded on the chart with a planned transfer date.
  • The UR coordinator receives the decision the same day and submits the PHP request before the transfer date, recording the request date and status.
  • Scheduling does not book PHP group days without a recorded authorization status, and requested-only days are flagged to the UR lead.
  • A pre-submission check holds any PHP day billed outside an authorization window and routes it to UR.

The measurement plan is written at the same time: authorization denials per PHP admission from residential, by date of service, before and after the change, evaluated once the post-change cohort has adjudicated. The mechanism check is the share of step-downs with a request date earlier than the transfer date, reviewed monthly from the start. The displacement check is whether residential stays now run beyond authorized residential days while patients wait for PHP approval.

The fix does not delay a clinically indicated step-down to protect a claim. Whether and when a patient moves is a clinical decision; the fix only makes sure the paperwork starts when the decision is made. Payer rules on retroactive authorization and concurrent review vary by plan and contract, so the UR lead confirms them per payer. For more on the review side, see the utilization review guide.

Where prevention programs stall

  • Tagging stops in a busy month. The queue gets heavy, tags get skipped, and the next review has nothing to discuss. Make the root-cause field required to close an item.
  • The owner cannot change the workflow. Responsibility without authority produces accurate monthly reports about an unchanged problem.
  • Fixes are memos. Instructions without a trigger, a record and a check decay with the next round of hiring.
  • Measurement by denial date. The fix looks like a failure in month one and a triumph in month three, and neither is true.
  • No evaluation date. Fixes are never checked, and the same causes return under new names.
  • Blaming individuals. Treating a workflow problem as a performance issue teaches people to hide denials instead of reporting them.

The first thirty days

You do not need new software or a new hire to start. You need a list, a field and a meeting.

  1. Draft the root-cause taxonomy and review it with each prospective owner so the terms match how their team describes its own work.
  2. Add a required root-cause field wherever denials are worked today, alongside the payer reason code.
  3. Back-tag a recent sample of denials, for example last month's, including those already recovered, to get a starting picture and test the taxonomy.
  4. Name an accountable owner for each category and note the contributors whose workflows touch it.
  5. Pick two root causes, write the fixes, and set evaluation dates and denominators.
  6. Put the monthly review on the calendar with a fixed agenda and a person responsible for preparing the numbers.

After three or four cycles, you should be able to answer three questions for any denial category: who owns it, what was changed, and did it work. If you can only say how many denials were appealed, you still have a queue.

What software can and cannot do here

Whatever system you use, prevention depends on a denials work queue with reason, dollars, deadline and owner; a root-cause field you control; reporting by date of service, payer, program and cause; and a way to get from a denied claim to the upstream record that explains it. When those records live in separate systems, the second look becomes a project of its own and tagging quality drops. Software cannot decide the root cause, and it cannot make the Friday treatment team meeting send its step-down decisions to UR. That part is operational.

In Unisona Health, the denials queue carries the reason, dollars at risk, appeal deadline and a named owner on every item, grouped by reason, and because authorizations, credentials and eligibility sit on the same record, the upstream cause is a click away rather than a phone call. Recurring causes can then become pre-submission checks in claim scrubbing, which flags the failing field in plain language, including authorization windows, credentials, modifiers and level of care.

Frequently asked questions

What is the difference between denial management and denial prevention?

Denial management works each denied claim to recovery: appeal, correct and resubmit, or write off. Denial prevention works the upstream workflow that produced the denial, such as intake, authorization, credentialing or documentation, so the same category stops appearing for services delivered after a change. Most programs have the first and assume it produces the second.

How do you find the root cause of a claim denial?

Tag every denial with two things: the payer's reason code from the remittance, and a root cause from your own workflow taxonomy. One reason code, such as missing authorization, can come from several different failures owned by different people. The person working the denial tags it at the point of work, and high-dollar or repeating items get a second look from the owner of the suspected workflow.

Who should own denial prevention in a behavioral health program?

Each root-cause category should be owned by the person who runs that upstream workflow: intake for eligibility, the UR lead for authorization, credentialing for enrollment, the clinical director for documentation, the billing manager for coding and claim build. Owners decide the fix and report whether it worked; they do not work the appeal. Name contributors too when the change has to happen in another team's process.

How do you measure if a denial prevention fix worked?

Compare claims by date of service before and after the change, express denials as a rate against a denominator chosen in advance, and evaluate only after the post-change cohort has had time to adjudicate. Check first that the new step is actually happening, and watch adjacent numbers such as scrubber holds and days to bill for problems that moved rather than disappeared.

Should denials that were overturned on appeal be included in denial analysis?

Yes. A recovered denial still consumed staff time and delayed cash, and the workflow that caused it is still producing more. Tag the root cause on every denial, including the ones you win, and report recovered and unrecovered denials side by side so well-appealed categories do not disappear from view.

Is a claim scrubber enough to prevent denials?

No. Turning a recurring denial reason into a pre-submission check is a useful backstop, but a claim caught by a scrubber is still an upstream failure and a held claim is delayed cash. Track scrubber holds by root cause with the same taxonomy as denials, and keep fixing the upstream workflow.

Samuel Jean

Co-Founder at Unisona Health, the behavioral health platform built for the audit. More about us →

Get started

Start today, or take a look first.

Create your workspace in minutes and try the whole platform free for 14 days — no credit card. Or book a one-hour walkthrough first.

Start your free trial