Common problem
Why is our AR over 90 days climbing?
AR over 90 days is a lagging indicator. What you are seeing now reflects something that went wrong weeks or months earlier, at eligibility, coding, submission, or follow-up. The cause is almost always identifiable from the pattern in your own numbers rather than from the total. It is also worth knowing that part of the rise may not be your fault, because payer behaviour changed measurably between 2024 and 2026.
Start with three numbers
Over 15% of total AR sitting past 90 days
HFMA targets under 10 percent. Above 20 percent is systemic rather than seasonal.
Over 40 days in AR
MGMA's 2024 Cost and Revenue Survey put the median at 47 days, with better performing practices at 36.
The direction of both, over three months
A stable 18 percent is a different problem from an 18 percent that was 12 percent in April.
The trend matters more than the number. A practice whose AR over 90 has been flat for two years has a process that needs improving. A practice whose figure has climbed for two quarters has something specific that broke, and this page is about finding it.
What should AR over 90 days actually be?
Under 15 percent of total AR is the common working target, and HFMA sets a stricter bar at under 10 percent. Above 20 percent indicates something systemic. But benchmarks are less useful here than most articles suggest, because payer mix moves them more than performance does. A practice with heavy Medicaid or workers compensation carries structurally longer AR than a commercial practice doing everything wrong.
| Metric | Target | Concern | Source | Independent? |
|---|---|---|---|---|
| AR over 90 days | Under 15% | Over 20% | HFMA targets under 10% | Yes |
| Days in AR | Under 40 | Over 50 | MGMA 2024, median 47, better performers 36 | Yes |
| Net collection rate | 95% and above | Below 93% | MGMA, HFMA | Relayed |
| First-pass clean claim rate | 95% and above | Below 90% | HFMA MAP Keys | Yes |
| Denial rate | Under 5% | Over 10% | HFMA. Kodiak recorded 11.81% actual in 2024 | Yes |
A correction worth knowing
You will see a figure of roughly 13.5 percent quoted as the MGMA benchmark for AR over 90 days across a lot of billing company content. It is a real number in the wrong place. It comes from MGMA's 2021 DataDive figure for AR over 120 days in multispecialty practices, which is a different bucket, a different cohort, and five years old.
We mention it because if you have been comparing yourself against it, you have been comparing against the wrong thing.
What to actually compare against
Three comparisons, in order of usefulness:
1.Your own trend, month over month. The most informative number you have.
2.Your specialty cohort. Surgical specialties carry more bundling denials. Behavioural health carries more medical necessity denials. A blended benchmark tells you little.
3.Your own payer-by-payer breakdown. Two practices both averaging 35 days in AR can have completely different aging profiles. A single blended figure hides a concentrated problem.
Free tool
AR aging benchmark comparator
Enter your aging buckets and see where you sit against published benchmarks, plus a ranked list of what is most likely causing it. Runs in your browser, sends nothing anywhere, and needs no patient information.
Payer mix (%)
Educational only. Benchmarks vary by specialty and payer mix. Verify timely filing windows against your own payer contracts. Not financial or legal advice. Runs in your browser. Days in AR formula per HFMA MAP Keys: total AR ÷ (trailing 90-day gross charges ÷ 90).
Given your numbers, what is actually causing it?
The main causes of AR aging leave different fingerprints. The mistake most practices make is treating a rising over-90 bucket as a follow-up problem and putting more people on the phone. Sometimes that is right. Often the claim never reached the payer, or was denied for something the front desk could have caught, or is one payer behaving differently from the rest. Find the fingerprint first.
Eligibility not verified
Fingerprint
Denial codes CO-27 and CO-22 clustered shortly after service, spiking alongside new patient volume.
What distinguishes it
The denials spread across many payers and many procedure codes, and correlate with registration dates rather than with anything clinical.
Registration and demographic errors
Fingerprint
High clearinghouse rejection rate, CO-16 for missing or invalid information, payer returns for wrong member ID or date of birth.
What distinguishes it
The failures happen at the clearinghouse rather than at the payer, and often concentrate by front desk shift or by location.
Referral or authorization not obtained
Fingerprint
CO-197 and medical necessity denials concentrated in the procedures that require prior authorisation.
What distinguishes it
Concentrated by CPT and by payer rather than spread across everything.
Payer downcoding, which looks like a coding problem and is not
Fingerprint
A gap between the E/M level you submitted and the level that paid, flagged in remark codes on the remittance.
What distinguishes it
Run a submitted-versus-paid variance report by payer. If you sample the notes and 70 percent or more support the level you billed, this is the payer downcoding, not your documentation. That distinction changes the response from retraining your clinicians to running an appeal campaign.
Coder capacity
Fingerprint
Charge lag rising, meaning the days between the encounter and the charge being entered.
What distinguishes it
The delay sits between service and submission, before any payer has seen the claim. Visible in your own system, not on any remittance.
Clearinghouse rejections nobody retrieved
Fingerprint
A gap between charges entered and claims accepted by payers, with claims sitting in a rejected queue.
What distinguishes it
These claims do not appear in your practice management aging report at all, because they never reached a payer. They are invisible unless someone opens the clearinghouse dashboard. This is the most common silent cause, and the filing clock has been running the whole time.
Submission failures at the payer or batch level
Fingerprint
An entire payer missing from your remittances, or batches failing without anyone noticing.
What distinguishes it
Whole-payer gaps rather than scattered claims. Common after an EHR migration or a clearinghouse switch, and a particular risk for practices that moved clearinghouse after the Change Healthcare outage in 2024.
Denials not worked, or worked too late
Fingerprint
The over-90 bucket growing in denied rather than pending claims, low appeal volume relative to denial volume, appeals filed after the window closed.
What distinguishes it
The denials are sitting in your system with no follow-up activity logged against them.
No follow-up cadence
Fingerprint
Claims aging past timely filing with no touch history at all, and recovery skewed heavily toward large balances.
What distinguishes it
Audit the touch history. Claims with zero recorded follow-up actions are the tell.
Payments not posted or underpayments not caught
Fingerprint
Cash arriving but AR not reducing, and payments landing below your contracted rate without anyone flagging it.
What distinguishes it
Reconcile deposits against posted payments. If they diverge, your aging report is fiction. Check paid amounts against your fee schedule while you are there.
Credentialing or enrolment
Fingerprint
Denials concentrated on one provider, usually a recent hire, or a lapsed CAQH attestation.
What distinguishes it
The problem is provider-specific rather than payer-wide or procedure-specific. Worth noting: new provider claims should be held until enrolment is effective rather than submitted and denied.
One payer, not you
Fingerprint
A days-in-AR spike isolated to a single payer while everything else looks normal.
What distinguishes it
The payer-by-payer breakdown makes this obvious in about ten minutes, and it is the first report to run.
Patient balances, not insurance
Fingerprint
The patient responsibility bucket growing while insurance AR holds steady, with near-zero collection at time of service.
What distinguishes it
Separate patient aging from insurance aging. If you have never done this, do it before anything else on this list, because the two need completely different workflows and averaging them together hides both.
Something structural
Fingerprint
A step change in the trend beginning on an identifiable date.
What distinguishes it
The inflection point lines up with a calendar event. A go-live, a departure, an acquisition, a system change. If your AR chart has a corner in it, find out what happened that week.
What to do first
Re-sort your worklist before you do anything else. Most practices work aged AR by largest balance first, which feels right and costs money. Work by deadline proximity instead. A small claim a week from its filing deadline is worth more than a large claim with six months left, because one becomes zero and the other does not.
This week, and free
- 1. Re-sort the aged worklist by deadline, then by recoverability. Correctable-error denials first, then claims near their filing deadline, then high value recoverable, then formal appeals. Stop working genuinely dead claims entirely.
- 2. Pull the clearinghouse rejection queue. Every rejection nobody retrieved is a claim that never reached a payer and never appeared in your aging report.
- 3. Run a payer-by-payer days in AR breakdown. Ten minutes, and it either isolates the problem to one payer or rules that out.
- 4. Separate patient AR from insurance AR. Different problem, different workflow, and averaging them hides both.
- 5. Reconcile unposted remittances and confirm your ERA and EFT enrolments are intact. Particularly if you changed clearinghouse after 2024.
Over the next few weeks
- 6. Build a follow-up cadence with a worklist and logged touch history. Work denials within 14 days of receipt, before they migrate into the 90-plus bucket.
- 7. Run the submitted-versus-paid variance report to detect downcoding.
- 8. Verify credentialing and enrolment for every provider, and hold new provider claims until enrolment is effective rather than billing and being denied.
- 9. Fix eligibility verification at every visit, and capture prior authorisation before service for the procedures that need it.
- 10. Push the clean claim rate toward 95 percent. Every claim that fails on first pass adds roughly two to four weeks to its effective cycle.
Structural, over months
- 11. Close the loop from denials back to the front desk and coding, so a back-end denial becomes a front-end fix rather than a recurring cost.
- 12. Collect at time of service and offer payment plans, because the patient share of revenue is growing whether or not you adapt to it.
- 13. If your team cannot clear the backlog and keep current work current at the same time, separate the two. Backlog cleanup that competes with daily follow-up usually loses, and then you have two problems.
What to expect
A practice starting around 55 days in AR can usually reach the low 40s within a quarter and the high 30s within two, with disciplined denial work and a better clean claim rate. Top quartile, around 30 days, generally takes six to nine months of sustained work. This is a reasonable expectation based on how the work compounds rather than a guarantee, and it depends heavily on payer mix.
Free tool
Timely filing triage
Enter a payer and a date of service to see how long you have, and how that claim should rank against the others on your worklist.
Timely filing countdown
Enter a payer and a date of service to see how long you have. Runs in your browser. Sends nothing. Needs no patient data.
General standards only. Your binding deadline is in your payer contract and can be shorter. Medicare: 365 days from date of service (42 CFR 424.44). Verify before relying on this.
Which old claims are dead, and which are just aged?
Age does not determine recoverability. The filing window does. A 135 day old Medicare claim is comfortably alive, because Medicare allows 365 days from date of service under 42 CFR 424.44. A 100 day old claim on a 90 day commercial contract is dead, and no amount of follow-up will change that. Practices routinely spend labour on the second category while the first quietly expires.
| Payer | Window from date of service |
|---|---|
| Medicare | 365 days (42 CFR 424.44) |
| Medicare Advantage | 365 day CMS floor, plan rules vary |
| Commercial | Typically 90 to 180 days. UnitedHealthcare and Aetna commonly enforce 90 |
| Medicaid | State specific. Texas 95 days, California Medi-Cal 180 |
Appeal windows are separate and shorter, and they run from the remittance date rather than the date of service. Corrected claims often must be filed within 60 days of the original denial, and the original filing clock does not reset.
A claim is dead when
- ✕ Its filing window has closed with no proof of timely submission and no applicable exception.
- ✕ A Medicare claim has been denied CO-29 with no basis for reopening.
A claim is merely aged when
- ✓ It is old but still inside its filing or appeal window.
- ✓ It has a correctable error, in which case fix and resubmit rather than appeal.
Sort your aged AR into these two piles before you assign anyone to work it. The labour you free up by abandoning genuinely dead claims is usually enough to properly work the live ones.
What changed between 2024 and 2026
If your AR over 90 has worsened and you cannot find an internal cause, the environment is a real part of the answer. Denial rates rose, the type of denial shifted toward the kinds that are harder to overturn, tens of millions of people lost Medicaid coverage, and patient deductibles kept climbing. None of this excuses a broken process, but it does mean a well-run practice can be losing ground while doing nothing differently.
Denials rose, and got harder
Kodiak Solutions, drawing on data from more than 2,100 hospitals and 300,000 physicians, reported that the initial denial rate rose to 11.81 percent of claims in 2024, up from 11.5 percent in 2023 and 10.2 percent in 2020. True days in AR rose 5.2 percent year on year.
The composition matters more than the total. Authorisation related denials fell, while medical necessity denials rose around 5 percent and requests for information rose around 5.4 percent. Those are the categories that take longest to resolve.
Kodiak's own reading, from its vice president of revenue cycle intelligence, is that payers appear to be using initial denials to slow payment even though roughly 90 percent of claims are ultimately paid.
Automated downcoding arrived
Cigna introduced reimbursement policy R49 on 1 October 2025, automatically downcoding certain office and consultation codes by one level where its records indicated a provider coded above peers. Physician groups including the American Medical Association opposed it, the California Department of Managed Health Care reviewed it, and Cigna paused the policy around 8 October 2025.
Several Blue Cross Blue Shield licensees apply downcoding to level 4 and 5 office visits. Anthem reversed a proposed anaesthesia time policy in December 2024 after opposition led by the American Society of Anesthesiologists.
The pattern is consistent: automated adjustment introduced, opposed, sometimes paused. What it means practically is that a submitted-versus-paid variance check now belongs in your monthly routine.
Medicaid coverage churned
The Medicaid continuous enrolment unwinding began on 1 April 2023. As of September 2024, KFF's tracker recorded at least 25.2 million people disenrolled. Around 69 percent of those disenrollments were for procedural reasons, meaning paperwork, rather than confirmed ineligibility. Total enrolment fell from roughly 94 million in March 2023 to about 80 million by September 2024.
The 2025 reconciliation law adds federal Medicaid work requirements, with states implementing from January 2027 and several moving earlier.
For a practice, this means patients presenting as insured who turn out to be retroactively ineligible, which converts expected insurance revenue into patient bad debt without warning.
Patient balances grew
KFF's 2025 Employer Health Benefits Survey put the average single coverage deductible at $1,886, up 17 percent over five years. At firms with 10 to 199 workers the average is $2,631, and 36 percent of covered workers at those firms face a deductible of at least $3,000.
Kodiak reported that providers collected 34.46 percent of what insured patients owed in 2024, down from 37.58 percent the year before.
More of your revenue is in the slowest, hardest bucket, and less of it is being collected.
The Change Healthcare outage
The ransomware attack on Change Healthcare on 21 February 2024 halted claims submission, eligibility verification and remittance across a clearinghouse handling roughly a third of US patient records. Nearly 94 percent of hospitals reported financial repercussions.
The lasting effect for independent practices is quieter. Practices that migrated to another clearinghouse in 2024 often carry residual mapping and enrolment gaps: wrong payer IDs, incomplete ERA re-enrolment, whole payers silently failing to submit. If your AR started worsening in 2024 and never recovered, check this before anything else.
Is it our billing company?
Possibly, and it is worth ruling in or out properly rather than guessing. The test is not your numbers, it is what your billing company can produce when you ask. A company managing your revenue cycle can give you an AR aging split by payer and provider, a denial breakdown by reason code, and an appeal overturn rate, on request and the same day.
We have written that diagnosis separately, including the six reports to ask for and what evasion looks like on each.
Is your billing company actually working your denials?Free 48-hour AR audit
Get a free 48-hour AR audit
We look at your claim data and tell you which of the causes above you actually have, what is at risk before each filing deadline, and what it is costing you. No obligation and no requirement to change anything. If your aging is normal for your payer mix, we will say so.
What is your specialty?
Roughly what percentage of your AR is over 90 days?
When did it start climbing?
You told us
— · — over 90 · —
Common questions
What percentage of AR over 90 days is normal for a small practice?+
Under 15 percent is the common target and HFMA sets a stricter bar at under 10. Small practices are noisier, because a single large claim moves the percentage in a way it would not at higher volume. Look at the dollar figure alongside the percentage, and weight the three month trend more heavily than any single month.
Is it worth chasing claims over 120 days?+
It depends entirely on whether the filing window is still open, not on the age. A 130 day Medicare claim is well within its 365 day window and fully recoverable. A 100 day claim on a 90 day commercial contract is already gone. Sort by window remaining, not by age, and stop working the dead ones.
How do I clean up a large backlog of old AR?+
Sort into dead and live first, and abandon the dead. Then work the live pile by deadline proximity rather than balance. The critical structural point: backlog cleanup that competes with daily follow-up usually loses, and then you have an aged backlog and a growing current one. Separate the two, whether internally or externally.
Our days in AR went from 35 to 55. What happened?+
Find the inflection point first. If the rise is gradual, it is usually follow-up cadence decaying or denials going unworked. If it is a step change on an identifiable date, look for what happened that week: a departure, a system go-live, a clearinghouse change. Then run the payer breakdown to see whether it is one payer or all of them.
Could Medicaid unwinding be causing our eligibility denials?+
Quite possibly. At least 25.2 million people were disenrolled from Medicaid between April 2023 and September 2024 per KFF, and around 69 percent of those were procedural rather than for confirmed ineligibility. That produces patients who present as covered and turn out not to be. The fix is front-end: verify eligibility at every visit rather than at registration only.
How often should we run the aging report?+
Monthly at minimum, weekly if it is climbing. But the report that changes decisions is not the aging total, it is the payer-by-payer breakdown and the insurance versus patient split. Running the same blended number every month tells you the temperature without telling you where the fire is.
Does the denial rate benchmark apply to my specialty?+
Loosely. Surgical specialties carry more bundling and modifier denials, behavioural health more medical necessity denials, and practices with heavy Medicaid or workers compensation carry structurally longer AR regardless of how well they are run. Compare against your own trend first, your specialty cohort second, and general benchmarks last.
What is the single most common cause you see?+
Clearinghouse rejections nobody retrieved. They are invisible, because a rejected claim never reached the payer and therefore never appears in the practice management aging report. The filing clock runs the entire time. It is worth checking before anything else because it costs ten minutes.
Sources
- ●42 CFR 424.44, Medicare timely filing
- ●CMS, Interoperability and Prior Authorization Final Rule CMS-0057-F
- ●KFF, Medicaid enrolment and unwinding tracker
- ●KFF, Employer Health Benefits Survey 2025
- ●MGMA, Cost and Revenue Survey 2024
- ●HFMA MAP Keys
- ●Kodiak Solutions revenue cycle data, 2024 and 2025
- ●Cigna, reimbursement policy R49