Your denial rate improved. Your payers took it back anyway.

• October 1 · FY 2027 ICD-10-CM took effect. 190 new reportable codes, 30 deleted, 4 revised by CMS's count. 15 of the deleted became non-billable headers requiring child codes.
• October 1 · NCCI PTP v32.3 took effect, the last edit set before January.
• September 17 · CMS issued CR 14611 / Transmittal 13946 for NCCI v33.0, effective January 1, 2027.
• September 14 · Comments closed on the CY 2027 Physician Fee Schedule proposed rule. Final rule expected around November 1.
• Through H1 2026 · Kodiak Solutions reports initial denials fell to 10.63% from 12.00%, takebacks rose to 1.57% of accounts receivable from 1.38%, appeal recovery fell to 2.21% from 2.48%.
• Coming January 1, 2027 · CMS-0057-F API requirements for impacted payers, with dates varying by payer type.

The denial rate is the wrong number

If your 2026 denial reporting looks better than 2025, you are reading it correctly. Kodiak Solutions, drawing on a platform covering more than 2,300 hospitals and 375,000 physicians, reports initial denial rates fell to 10.63% in the first half of 2026 from 12.00%.

Take the win. Then read the other two numbers Kodiak published beside it.

Takebacks, meaning money an insurer already paid and later recouped, rose to 1.57% of accounts receivable from 1.38%. That is a 13.8% increase. And the share of accounts receivable providers won back on appeal fell to 2.21% from 2.48%.

Put all three together and the picture inverts. Fewer claims get stopped at the front door. More money gets pulled back after you booked it. When you fight for it, you win less often than last year. Kodiak's own headline finding says the falling denial rate did not improve cash performance, which is the part most coverage dropped.

Why a takeback costs more than a denial

A front-end denial is cheap by comparison. You learn about it within days, the claim never got recognized as revenue, and the fix joins a rework queue your team already runs.

A takeback behaves differently. The money was recognized. For a billing company it may already have reached a client report. The documentation is months old, the coder who worked the chart may have left, and the payer wants you to defend a judgment made under a code set that has since changed. You are not reworking a claim. You are reconstructing a decision.

Medicare Advantage is where the exposure moved

MA plans showed higher initial and final denial rates than traditional Medicare, and more than double the rate of takebacks. If your payer mix has been shifting toward MA, your exposure grew even while your aggregate denial rate improved.

Run the numbers by payer class before you report an improvement to anybody.

What to do about it

Report takebacks next to denials, every month. Reporting that shows a denial rate with no recoupment line describes a world that stopped existing this year.

Audit what passed, not only what failed. Most coding QA samples denials, because that is where the known errors sit. That design finds front-end problems and stays blind to claims that clear edits and fail audits later. Pull a sample of paid, unchallenged claims from your highest-recoupment service lines and read them the way an auditor would.

Make the evidence retrievable at the code level. The cost of a takeback is mostly reconstruction cost: find the chart, find the passage, establish who decided what and when. At an hour per claim, a 200-claim audit eats five weeks of a senior coder.

• ~Nov 1 · What happens: CY 2027 PFS final rule expected · What to do before it: Change nothing for the same-day E/M reduction or the G2211 modifier yet. Read the final rule first.
• ~Nov 2 · What happens: NCCI v33.0 test files expected · What to do before it: Book scrubber maintenance time now.
• ~Nov 17 · What happens: NCCI v33.0 change reports expected · What to do before it: Diff against v32.3. Flag every modifier-indicator change among your top 20 code pairs.
• Dec 1–19 · What happens: Configuration and briefing window · What to do before it: Load v33.0. Brief coders on modifier-indicator changes only.
• Jan 1 · What happens: NCCI v33.0 effective. CMS-0057-F APIs. CY 2027 rates. Telehealth modifiers BB and BC. · What to do before it: January 1 is a Friday and a holiday. Real volume lands Monday, January 4.
• Jan 4 · What happens: NCCI v33.0 implementation · What to do before it: Watch week-one rejections. Whatever you missed appears here.

The FY 2027 deletions follow one pattern

Fifteen codes became non-billable headers on October 1. Four worth your attention:

• I42.0 dilated cardiomyopathy · Now requires: I42.00, I42.01 or I42.09
• I42.8 other cardiomyopathies · Now requires: I42.81 or I42.89
• Z68.1 BMI 19.9 or less, adult · Now requires: Z68.18 or Z68.19
• D69.1 qualitative platelet defects · Now requires: D69.11 or D69.19

In each case the clinical documentation usually supports the correct child code already. The etiology sits in the cardiologist's assessment. The BMI value sits in the vitals. The coder reached for the parent out of habit.

That distinction decides the fix. A check running against the current addenda catches habit. A query catches a documentation gap. Running coder retraining at what is actually a configuration gap costs you coder hours you cannot replace.

Z68.1 deserves a separate note. BMI Z codes sit in secondary positions and nobody watches them. They feed risk adjustment and support medical necessity for nutrition and metabolic services. A header code in a secondary position clears front-end edits and resurfaces in an audit, which is the deferred-denial pattern again.

What your next client actually costs you

AAPC's 2026 salary report, built on more than 20,000 member responses, puts certified medical records specialists at $67,260 against $55,721 for non-certified. Two certifications reach $74,557. Three or more reach $81,227. Geography swings almost $27,000, from $77,708 in Delaware to $50,393 in Mississippi.

Recruiting a credentialed coder with the right specialty experience takes months. New volume arrives on a contract date. The distance between those two facts caps what you can take on, and no salary band closes it.

So ask what proportion of your certified coders' hours went to charts that never required certified judgment. That ratio is measurable in your own data this week, and it tells you whether your next client is a hiring problem or a routing problem.

Per-chart routing is how CODIO AI answers it. Complete charts flow through AutoPilot for fully autonomous coding. Charts where the documentation is thin or contradictory stay in CoPilot, where certified coders review and finalize the output. Teams see 45% more output per coder, measured against a manual baseline: same team, same charts, before and after.

Every assigned code carries the chart passage it came from and the guideline that justified it, which turns the audit described above into a lookup rather than a reconstruction. Nothing runs autonomous on day one.

What our certified coders observed across chart volume this month. Qualitative practitioner observation, not measured data.

Deletion errors cluster in the first ten days, then fall off. Coders who know about a change still reach for the familiar code under time pressure, until rejections start arriving. The first two weeks after any October 1 change are when a front-end check earns its keep.

Laterality gaps concentrate in high-volume, low-complexity notes. The complex chart gets a careful note. The twelfth follow-up of the day gets a terse one. Laterality documentation fails where volume runs highest, which is exactly where FY 2027 added lateralized codes.

Secondary diagnoses hide the sequencing errors. First-listed selection is usually right. The errors sit further down, in companion codes an instructional note requires and the pointer linking a diagnosis to the service that justifies it.

• [Article 1] : why a coding manager should read it
• [Article 2] : one line
• [Article 3] : one line

Since switching to MediCodio, my office has had more time to focus on collections and patient billing. Our coding is so streamlined, I don't even have to think about it. — Chris McMenemy, Ortmann Healthcare Consulting Services