AI ASC coding accuracy benchmark

CorePulse on real ortho ASC charts, measured by code type

Here is the short answer. On a production run of 75 real ambulatory surgery ortho charts, CorePulse hit 97% on surgical CPT, 78% on implant and device C-codes, and 75% on primary ICD-10-CM. Secondary-diagnosis capture is where it still trails a seasoned human coder, and we will show you exactly where.

97%
Surgical CPT
78%
Implant / device C-codes
75%
Primary ICD-10-CM
$0.055/chart
All-in compute cost

// The stat wall :: accuracy by code type

One number per code type, on charts that actually went out the door

No synthetic notes, no cherry-picked cases. These are the measured accuracies from a real ortho ASC production run, scored line by line against the coder of record. We break them out because a single blended "accuracy" number hides the thing that matters: CPT and device lines behave very differently from the diagnosis tail.

Surgical CPTSTRONG
97%
Primary procedure codes matched the human coder on nearly every chart. AMA CPT 2026 descriptors are where deterministic anatomy checks earn their keep.
Implant / deviceSOLID
78%
HCPCS Level II C-codes for hardware and pass-through devices. Misses cluster where the op note names the implant but not the billable descriptor.
Primary dxSOLID
75%
ICD-10-CM FY2026 principal diagnosis, including the laterality and encounter characters ortho lives and dies on.
Secondary dxTHE GAP
~14%
Honest number. Comorbidity and secondary-diagnosis capture is where the engine over- and under-reads against a human coder. This is the open work, not a rounding error.
75
Real ortho ASC charts
$0.055
Compute per chart
111s
Avg processing time
100%
De-identified before any model call

// Where the two disagree

Engine vs coder, read plainly

When CorePulse and the human coder diverge, the direction of the miss tells you the fix. On this run the engine's error was almost never a made-up code. It over-read the diagnosis tail, and in a companion 30-chart benchmark 98% of the extra diagnoses it added were actually documented in the chart. The coder simply did not bill them.

Code layer
Agreement with coder of record
Surgical CPT Primary procedure, AMA CPT 2026
Matches on 97% of charts
Implant / device C-codes HCPCS Level II, OPPS/ASC device lines
Matches on 78% of charts
Primary diagnosis ICD-10-CM FY2026, laterality + encounter
Matches on 75% of charts
Secondary diagnosis Comorbidities, status codes, tail dx
Trails the coder, over-reads the tail
The part vendors skip

Secondary-diagnosis capture still needs the coder

We will not sell you a 99% headline. On real ortho ASC charts, secondary and comorbidity diagnoses are the layer where CorePulse and an experienced coder disagree most. The engine tends to surface every documented condition, while the coder applies judgment about what belongs on an ASC claim under OPPS/ASC reporting rules.

That is exactly why we run CorePulse as a coder-in-the-loop tool, not an autopilot. It clears the CPT and device lines fast and cheap, then hands the human a short, ranked list of secondary-dx candidates to accept or drop. Speed on the certain layers, judgment reserved for the uncertain one.

// How a chart moves through CorePulse

Five steps, PHI never leaves the boundary

01
De-identify
Fail-closed PHI scrub before any model call.
02
Extract
Op note, implant log and dx pulled structured.
03
Match
Deterministic anatomy check against CPT descriptors.
04
Rule pass
NCCI edits, C-code and OPPS/ASC line logic.
05
Coder review
Ranked candidates, human accepts or drops.

// Scored against the real rulebooks

The 2026 guidelines this benchmark answers to

Accuracy only means something if it is measured against the code sets and payment rules a real ASC bills under. Every number above was scored against current, named authority, not a frozen internal reference.

CMS CY2026 OPPS and ASC Payment System Final RuleGoverns how ASC facility, procedure and device lines are packaged and reported. Sets the frame for what belongs on the claim.
AMA CPT 2026 code setThe surgical procedure descriptors CorePulse matches on, including the ortho-specific approach and laterality nuances.
ICD-10-CM FY2026 (effective Oct 1, 2025)Diagnosis coding, encounter characters and 7th-character rules that make ortho dx unforgiving to get slightly wrong.
HCPCS Level II C-codesDevice and implant pass-through codes reported on OPPS/ASC claims, the layer behind the 78% implant number.

See CorePulse scored on your own charts

Numbers on our charts are a starting point. The honest test is a run on your ASC's own ortho notes, scored line by line against your coder of record, with the secondary-dx gap laid out in the open. That is a benchmark you can actually trust to a staffing decision.

Request a benchmark run on your charts →
ASP-RCM Solutions  ·  Frisco, TX  ·  Senior Partner, Revenue Cycle

Related reading

More on this topic

Insight

The case is on the list. Does the facility fee actually cover the implant?

A scheduler-first read of the CY2026 OPPS/ASC final rule and CMS ASC Covered Procedures List. See which cases

Read →Whitepaper

Onboarding the CY2026 ASC Covered Procedures List, one dated step at a time

A dated onboarding timeline for adding CY2026 ASC Covered Procedures List additions to the chargemaster and ap

Read →Field note

The second line is where ASC money quietly leaks.

How the ASC 50% multiple-procedure reduction and modifiers 50, 51, 59, 73, and 74 change the actual paid amoun

Read →