Clinical AI Practice

Automated medical codingin production

What works, and where most systems fall short. AI has significantly advanced automated CPT, modifier, and ICD-10 code mapping. Achieving production-scale performance now depends primarily on clinical ontology depth, documentation variability, and seamless revenue cycle integration.

chest painvague complaintR00.xHeart beatR04.xAirway bleedR07.1On breathingR07.0Pain in throatR07Pain in chestR07.9 — Chest pain, NOS

Production benchmarks

95%+
F1 on discharge summaries
60%
Reduction in coder review time
4 wks
Typical integration to go-live
Zero
EHR workflow changes required
  1. Ingest

    EHR · PDF · HL7

  2. De-identify

    PHI removed first

  3. Extract

    Contextual extraction

  4. Code

    ICD-10 · CPT · rules

  5. Review

    Human queue

01The Production Gap

Where automated coding systems consistently fall short

Systems that perform well in controlled evaluations degrade against the conditions that define real hospital environments: dictated notes, multi-procedure encounters, and payer-specific rules that exist outside any standard code set.

Dictation artefacts and non-standard abbreviations

Production notes contain dictation errors, unconventional abbreviations, and fragmented sentence structure that standard NLP pipelines are not trained to handle. "Pt c/o SOB x3d, r/o PE, started on LMWH" is representative.

1 of 5

Exhibit — abbreviation resolution: from dictation shorthand to coded entities

Production input — raw

"Pt c/o SOB x3d, r/o PE, started on LMWH"
  • SOBShortness of breathR06.00
  • r/o PEPulmonary embolismNEGATED — NOT CODED
  • x3dDuration: 3 daysQUALIFIER
  • LMWHLow-mol-weight heparinZ79.01

Dictation abbreviations are not standardised across hospitals or departments — "SOB x3d" and "r/o PE" require clinical-domain NLP to resolve. Systems trained on clean EHR text degrade substantially against production dictation.

“Accuracy on training data sets and accuracy on production live datasets are often separated by 15 to 25 percentage points. The gap is addressable through re-training and implementing SOPs.”

02How the System Works

Vendor-neutral agentic architecture for high accuracy and operational reliability

Each of the five stages corresponds to a specific failure mode above. Select a stage to explore it.

Stage 01 / 05Ingest

EHR notes · PDFs · dictations

Document ingestion — any format, any delivery method

The pipeline accepts clinical documentation in every format hospitals actually produce — dictated notes, scanned PDFs, structured EHR exports, and HL7 feeds. Format and delivery method are decoupled from downstream accuracy. An audit record is created at the point of ingestion, before any processing begins.

HL7 v2 and FHIR API connectors
Direct integration with Epic, Cerner, and Meditech. Triggers IT security review — plan for 4–8 additional weeks. Not the fastest initial path.
SFTP batch ingestion
The fastest deployment path. Notes are exported from the EHR and transferred via encrypted file exchange. Bypasses the IT security review process entirely. Most clients start here and migrate to direct API later.
OCR for scanned documents
Scanned PDFs processed through an optical recognition layer before NLP. Recognition quality is a direct input to downstream extraction accuracy — poor scan quality is surfaced at ingestion, not discovered at coding.
Ingestion audit log
Every document receives a unique ingestion record: timestamp, source system, format, and document type. Audit trail begins at entry — before de-identification, before any model sees the text.
Ingestion methods — trade-offs
  • SFTP / BatchEncrypted file transferFASTEST
  • HL7 v2 / FHIRDirect EHR connectorIT REVIEW REQ.
  • PDF / ScannedOCR preprocessing layerSCAN QUALITY
  • Direct APIReal-time single-noteIT REVIEW REQ.

SFTP is the recommended starting point for most deployments. IT security review for direct API access typically adds 4–8 weeks — plan for it early or start with SFTP.

Exhibit — ingestion sources
  • EHR DirectIT review
  • HL7 v2IT review
  • SFTP / BatchFastest
  • FHIR R4IT review
  • PDF / OCRScan qual.

Centigrade

Ingest

Audit log begins

De-identifyExtractCode

Every document receives a unique audit record at ingestion — before de-identification, before any model processing begins.

Human-in-the-loop design

Workflow re-imagined around fast human review and audit

Accurate code extraction is necessary, but not sufficient. The review interface is what converts accuracy into measurable time savings.

Coder review queue — interface model

Coding queue

4 pending

Discharge summary · [PHI redacted] · 03/28/2026

  • E11.9T2DM without complications"Hx of T2DM" — line 297%ACCEPT
  • R06.00Dyspnoea, unspecified"SOB x3d" — line 394%ACCEPT
  • Z79.4Long-term insulin useLow confidence — verify source before accepting74%REVIEW
  • I48.91Atrial fibrillation, unspec."parox. AF" — line 788%ACCEPT

3 accepted · 1 pending review · Submission authority: coding team

Submit to billing

Amber row = low confidence. Final submission authority stays with the coding team.

Confidence flags

Flags surface only the cases that need judgment, so review attention concentrates where it changes the outcome.

One-step review

Accept, modify, or reject per suggestion without leaving the queue.

Feedback loop

Every coder rejection is a labelled training signal. Accuracy compounds against your own documentation patterns.

10–15 min

Per chart, manual coding

60% less

Coder review time in production

03Integration & Deployment

Integration considerations and deployment timeline

The most common source of extended deployment timelines is not technical complexity — it is the security review process triggered by direct EHR API access. Understanding this early significantly affects project planning.

Exhibit — typical deployment timeline: integration to first production output
  1. Kickoff

    Discovery

    SOP alignment, payer mapping

  2. Week 2

    Pipeline setup

    PHI validation, integration

  3. Week 3

    Pilot — 500 charts

    Accuracy measurement on your data

  4. Week 4

    Production

    Full rollout + feedback loop

SFTP path — no IT security review triggered, typically 2 weeks faster

Direct EHR API — triggers full IT security review, add 4–8 weeks

Integration path selection is the single largest determinant of go-live timeline. Secure file exchange (SFTP) bypasses the IT security review process that direct API access triggers.

REST API

Structured JSON output, most common

Standard

Secure SFTP

Fastest to deploy, no security review triggered

Fastest

Direct EHR connector

Epic, Cerner, Meditech

IT review req.

RCM system feed

Optum, Waystar, Experian Health

Standard

04Accuracy

The benchmark standard for measuring operational reliability

F1 score — the harmonic mean of precision and recall — is the standard production measure. The OIG compliance threshold is ≥95%. Aggregate vendor figures routinely conceal the variance by document type and specialty that is most material to procurement decisions.

Precision

Of all codes assigned, what proportion were correct. High precision means a low false-positive rate — fewer incorrect codes submitted to payers.

Recall

Of all codes that should have been assigned, what proportion were captured. High recall means a low missed-code rate and lower revenue leakage.

F1 Score

Harmonic mean of precision and recall. The standard production metric — a single figure that penalises both incorrect codes and missed codes equally.

OIG Standard

The Office of Inspector General defines ≥95% accuracy as the compliance threshold for hospital coding programs subject to external audit.

Exhibit — ICD-10 ontology traversal: from vague category to valid code
  1. chest pain

    vague presenting complaint

  2. R00.xAbnormalities of heart beat
    R04.xHaemorrhage from airways
    R07Pain in chest
  3. R07.1On breathing
    R07.0Pain in throat
    R07.9Chest pain, NOS

A vague presenting complaint such as “chest pain” must be traversed down the ICD-10 hierarchy to a valid, billable code — R07.9. Category-level identification is not the same as code-level specificity.

Specialties by AI coding readiness

AI coding maturity varies by specialty — determined by documentation structure, code set depth, and payer rule density. The groups below run from production-proven to advanced-tuning required.

E&M Outpatient

Primary Care · Internal Medicine · Paediatrics

High-volume structured notes — ideal first deployment.

HCC Risk Adjustment

Medicare Advantage · CMS-HCC · RAPS

Surfaces chronic conditions documented but not coded, directly improving RAF scores.

Radiology

Diagnostic · Interventional · Nuclear

Structured reports yield the highest production F1 rates across all specialties.

Pathology & Labs

Clinical · Anatomic · Molecular

Well-defined code sets with minimal post-processing required.

05Standards & Compliance

Code systems and compliance architecture

The system outputs to the standard code sets and interoperability formats in use across hospital billing and RCM operations. PHI handling is architecturally integrated, not a configuration option.

ICD-10-CM
ICD-10-PCS
CPT
HCPCS Level II
SNOMED CT
RxNorm
LOINC
FHIR R4
HL7 CCD
HIPAA — PHI De-identification
SOC 2 Available

06Knowledge Base

Our extensive knowledge base

Technical depth for data and engineering teams, operational guidance for RCM leadership, and a coding practice track for coders adapting to AI-assisted workflows. Benchmark data and case studies are included within the relevant track.

  • Language models reliably identify the diagnostic category. Laterality, encounter type, and episode of care require live traversal of the ICD-10 hierarchy — a constraint no training corpus eliminates.

    ICD-10-CMUMLSOntology
    Read the article →

Contact

Talk to someone who has built this in production

Bring your documentation patterns, payer mix, and integration constraints. We will walk through accuracy expectations and a realistic go-live sequence.