Verified
ICD-10 K22.70: Barrett's Esophagus without Dysplasia — Clinical Documentation & Coding Guide
Master ICD-10 K22.70 coding for Barrett's esophagus without dysplasia. Expert guide on documentation, surveillance protocols & payer compliance for GI practices.
ICD-10 K22.70: Barrett's Esophagus without Dysplasia — The Definitive Clinical Documentation & Surveillance Coding Guide for GI Practices
TL;DR — What Every GI Medical Director Needs to Know
What Payers Actually Scrutinize: The Prague C & M Gap That Competitors Ignore
Scribing.io Clinical Logic: Handling the 3-Year Surveillance Error in Nondysplastic Barrett's Esophagus
Technical Reference: ICD-10 Documentation Standards for Barrett's Esophagus
Why Prague C & M Must Be Structured Data, Not Free Text
Anatomy of a Barrett's Post-Pay Audit: What Triggers Recoupment
Implementation Checklist: Rolling Out Structured Barrett's Documentation
See the Prague C & M Auto-Capture in Action
TL;DR — What Every GI Medical Director Needs to Know
K22.70 codes Barrett's esophagus (BE) without dysplasia, but the code alone does not protect your practice from audit recoupment. Payers now cross-reference surveillance EGD intervals against discrete Prague C & M measurements documented in the note. If M ≥ 3 cm (long-segment BE), current AGA guidelines support a 3-year surveillance interval for nondysplastic disease. If M < 3 cm, the guideline-supported interval is 5 years. The term "short-segment Barrett's" is not an ICD-10 concept, carries no reimbursement weight, and introduces ambiguity that triggers denials. Scribing.io auto-captures Prague C & M as FHIR Observation components, suppresses "short-segment" phrasing, pathology-gates code selection between K22.70 and K22.719, and prevents scheduling errors before the claim ever drops.
What Payers Actually Scrutinize: The Prague C & M Gap That Competitors Ignore
The Blind Spot in Every Existing ICD-10 Reference
Standard CMS ICD-10 Clinical Concepts resources provide code lists, laterality reminders, and broad documentation tips. They are indispensable for understanding the structure of ICD-10. What they do not address is the intersection of procedure-level clinical data, guideline-specific surveillance intervals, and post-pay audit logic that determines whether a Barrett's esophagus claim actually survives payer review. Scribing.io was built to close exactly this gap — not by replacing coding references, but by enforcing the clinical documentation chain that makes K22.70 audit-proof at the point of care.
Post-pay audits of surveillance endoscopy are increasing across commercial and Medicare Advantage plans, with Barrett's esophagus representing a disproportionate share of recouped procedures. The mechanism is straightforward: payers have operationalized the ACG Clinical Guideline and AGA guideline language into algorithmic audit rules. They are not reading your notes for clinical nuance. They are parsing data fields for Prague values, pathology dates, and interval justification. Refer to the Scribing.io ICD-10 Documentation Library for the full scope of Barrett's-related codes and their documentation requirements.
The Original Insight Competitors Miss
Payers scrutinize nondysplastic Barrett's surveillance intervals against explicit Prague C & M measurements captured as discrete data. This is not a theoretical risk — it is a reproducible audit pathway that GI practices encounter quarterly.
Here is the logic payers apply, distilled into the five checkpoints that determine whether your surveillance EGD survives review:
Audit Checkpoint | What the Payer Expects | What Triggers Recoupment |
|---|---|---|
Prague Classification | Discrete C (circumferential) and M (maximal extent) values in centimeters | Narrative-only descriptions like "short-segment" or "tongue of Barrett's" with no measurements |
Segment Length → Interval | M ≥ 3 cm → 3-year recall; M < 3 cm → 5-year recall (nondysplastic) | 3-year recall scheduled when M < 3 cm without documented dysplasia or other risk factor |
Pathology Linkage | Most recent biopsy result explicitly linked to the Barrett's problem, confirming "no dysplasia" | Code posted before pathology returns, or pathology result not referenced in the attestation |
ICD-10 Code | K22.70 (without dysplasia) only when pathology confirms absence; K22.719 when "indefinite for dysplasia" | K22.70 posted when pathology reads "indefinite for dysplasia" |
Terminology | Prague-based language; "long-segment" acceptable if M ≥ 3 cm | "Short-segment Barrett's" — not a recognized ICD-10 descriptor, creates ambiguity |
Why "Short-Segment" Is a Liability, Not a Descriptor
The term "short-segment Barrett's esophagus" (SSBE) persists in clinical parlance. It has legitimate historical usage in research and risk stratification. But it is not an ICD-10 concept, does not map to any code modifier, and — critically — creates confusion about which surveillance interval is guideline-appropriate. When a note reads "short-segment Barrett's" and the scheduler defaults to a 3-year recall, the practice has manufactured a documentable audit trail that contradicts published AGA recommendations.
The Anchor Truth: To justify a 3-year surveillance EGD, the note must explicitly document the Prague C & M scores and the absence of dysplasia confirmed by pathology. If M < 3 cm, guidelines favor a 5-year interval. If you are scheduling 3-year surveillance on a C0M2 patient without dysplasia and without a documented clinical rationale for deviation, you are building the audit record that payers will use to recoup.
For the full code specifications, see K22.70 - Barrett's esophagus without dysplasia; K22.719 - Barrett's esophagus with unspecified (indefinite) dysplasia.
Scribing.io Clinical Logic: Handling the 3-Year Surveillance Error in Nondysplastic Barrett's Esophagus
The Scenario That Costs GI Practices Thousands
A 56-year-old man undergoes surveillance EGD. The endoscopy note reads "short-segment Barrett's," with Prague measurements of C0M2. Biopsies are pending. The coder posts K22.70 and the scheduler sets a 3-year recall.
Eighteen months later, in a post-pay audit, the payer applies the following logic:
M = 2 cm → This is not long-segment Barrett's by any published definition.
Nondysplastic BE with M < 3 cm → Per ACG guidelines, the supported interval is 5 years, not 3.
No discrete Prague field in the structured note → No machine-readable data to verify the endoscopist's clinical rationale for deviating from the 5-year interval.
No explicit "no dysplasia" attestation tied to a pathology date → The K22.70 code was unsupported at the time it was posted because biopsies were still pending.
Result: The payer recoups the surveillance EGD. If the practice performed multiple surveillance procedures under the same flawed workflow, the recoupment cascades. We have seen practices lose five figures in a single audit cycle on Barrett's surveillance alone.
How Scribing.io Prevents This — Step by Step
Workflow Step | Scribing.io Action | Clinical & Financial Impact |
|---|---|---|
1. Structured Prague Capture | The scribe workflow captures Prague C and M as FHIR Observation components (LOINC-coded, cm units) — not as free-text narrative. The endoscopist enters C = 0, M = 2. These values populate discrete, queryable fields in the encounter record. | Creates the exact discrete data element that payer audit algorithms require. Eliminates narrative ambiguity. Enables downstream analytics on Barrett's segment trends over time. |
2. Terminology Gating | If M < 3 cm, the system auto-prompts the clinician to remove "short-segment" phrasing and replaces it with the Prague classification (e.g., "Barrett's esophagus, Prague C0M2"). If M ≥ 3 cm, "short-segment" is suppressed because it contradicts the measurement. The clinician sees a real-time notification: "'Short-segment' is not an ICD-10-recognized descriptor. Prague C0M2 documented. Recommended phrasing: 'Barrett's esophagus, C0M2, nondysplastic.'" | "Short-segment" never appears in the final note. The terminology matches ICD-10 concepts, guideline language, and payer expectations. No ambiguity survives to the claim. |
3. Interval Logic Engine | With M = 2 cm (< 3 cm) and no dysplasia, the system sets the recommended recall to 5 years. The surveillance interval recommendation is displayed to the clinician in the encounter summary. If the clinician overrides to 3 years, the system requires a documented clinical rationale (e.g., family history of esophageal adenocarcinoma, prior indefinite pathology, confirmed high-risk histological features) entered as a structured override field. | The 3-year scheduling error is prevented at the point of documentation, not discovered 18 months later at audit. Override rationales are captured as discrete data, creating a defensible record if the deviation is ever questioned. |
4. Pathology-Gated Attestation | The system holds the claim in a pre-submission queue until pathology results are received. When pathology returns "negative for dysplasia," the clinician receives a prompt to attest: "Pathology [date] confirms no dysplasia. Confirm K22.70." The attestation, pathology date, and pathology accession number are linked in the encounter record. If pathology returns "indefinite for dysplasia," the system auto-resolves the code to K22.719, flags the encounter for clinical review, and adjusts the surveillance interval recommendation to 6–12 months per published guideline standards. | K22.70 is never posted prematurely. The claim is audit-proof because the code is pathology-confirmed and time-stamped. If pathology upgrades the finding, the code escalates without manual coder intervention, eliminating the most common Barrett's coding error in GI practice. |
5. Structured EHR Export | Prague C & M fields, the pathology-linked attestation, the finalized ICD-10 code (K22.70 or K22.719), and the surveillance interval are exported as structured FHIR resources to the EHR via one-click integration with Epic, athenahealth, and other certified platforms. Scheduling systems receive the guideline-appropriate interval directly — no manual transcription. | Downstream systems (scheduling, cancer registry, quality reporting, MIPS measures) inherit clean, guideline-compliant data. The scheduler cannot override the interval without the same structured rationale requirement. The entire chain — from endoscope to recall letter — is internally consistent. |
The Outcome
The 3-year scheduling error is prevented. The audit recoupment is avoided. The patient receives surveillance at the guideline-appropriate 5-year interval. And if pathology later reveals indefinite dysplasia, the code automatically escalates to K22.719 without manual coder intervention. Every data point — Prague values, pathology date, attestation, code, interval — is linked, structured, and defensible.
Technical Reference: ICD-10 Documentation Standards for Barrett's Esophagus
K22.70 — Barrett's Esophagus without Dysplasia
Data Element | Specification |
|---|---|
ICD-10-CM Code | K22.70 |
Full Descriptor | Barrett's esophagus without dysplasia |
Category | K22 — Other diseases of esophagus |
Required Documentation | Endoscopic identification of columnar-lined esophagus; biopsy confirming intestinal metaplasia (per institutional criteria); pathology report explicitly stating "negative for dysplasia" or "no dysplasia" |
Prague Classification | C (circumferential extent in cm) and M (maximal extent in cm) must be documented as discrete values — per the original Prague C & M validation study |
Surveillance Interval (Nondysplastic) | M ≥ 3 cm → 3 years; M < 3 cm → 5 years (per AGA/ACG guidelines) |
Common Documentation Errors | Posting K22.70 before pathology results return; using "short-segment" without discrete Prague values; defaulting to 3-year recall regardless of segment length; failing to link pathology accession number to the attestation |
K22.719 — Barrett's Esophagus with Dysplasia, Unspecified
Data Element | Specification |
|---|---|
ICD-10-CM Code | K22.719 |
Full Descriptor | Barrett's esophagus with dysplasia, unspecified |
Clinical Application | Used when pathology reads "indefinite for dysplasia" — the pathologist cannot confirm or exclude dysplastic changes |
Required Documentation | Endoscopic findings; biopsy with pathology report stating "indefinite for dysplasia"; Prague C & M values; documentation of PPI optimization plan |
Surveillance Interval | Typically 6–12 months with repeat biopsy after PPI optimization — per ACG Clinical Guideline (2022) |
Critical Distinction from K22.70 | K22.719 should not be used interchangeably with K22.70. "Indefinite for dysplasia" is a distinct pathological finding that demands shorter follow-up, PPI dose escalation, and repeat biopsy — a fundamentally different management pathway |
The K22.70 vs. K22.719 Decision Tree
This decision tree is the core logic that Scribing.io enforces at the point of documentation. No code finalizes without the pathology result driving the selection:
Pathology Finding | Correct ICD-10 Code | Surveillance Implication |
|---|---|---|
Negative for dysplasia | K22.70 | M ≥ 3 cm → 3 yr; M < 3 cm → 5 yr |
Indefinite for dysplasia | K22.719 | Optimize PPI → repeat EGD in 6–12 months |
Low-grade dysplasia | K22.710 | Confirm with expert GI pathologist; consider endoscopic eradication therapy per ASGE guidelines |
High-grade dysplasia | K22.711 | Endoscopic eradication therapy recommended |
Biopsy pending / not yet resulted | Do not finalize code | Hold claim until pathology returns |
Scribing.io enforces this decision tree automatically. When pathology results are "indefinite," the system resolves K22.70 to K22.719 and alerts the clinician. When results are pending, the claim is held in the pre-submission queue. The coder never has to make this judgment call manually — and the practice never posts an unsupported code.
For the full code reference and related esophageal condition codes, visit the Scribing.io ICD-10 Documentation Library. The authoritative source for all ICD-10-CM code definitions remains the CMS ICD-10-CM Official Guidelines.
Verified code links: K22.70 - Barrett's esophagus without dysplasia; K22.719 - Barrett's esophagus with unspecified (indefinite) dysplasia.
Why Prague C & M Must Be Structured Data, Not Free Text
The Problem with Narrative-Only Documentation
Most EHR endoscopy modules capture Prague classification as a free-text field within the procedure note body. A typical note reads:
"There is a tongue of salmon-colored mucosa extending 2 cm above the GEJ. Short-segment Barrett's is suspected. Biopsies taken."
This note contains clinically useful information. It is also, from an audit defense perspective, nearly worthless. Here is why:
No discrete C value. The circumferential extent is not stated. An auditor cannot determine whether the Barrett's is circumferential at any level.
No explicit M value. "2 cm above the GEJ" describes maximal extent, but it is buried in prose. Automated audit tools parsing structured fields will not find it.
"Short-segment" is the documented descriptor. This term has no ICD-10 mapping. It does not appear in the CMS tabular list. It tells the payer nothing about whether a 3-year or 5-year interval is appropriate.
"Biopsies taken" without pathology linkage. The note confirms biopsies were obtained but does not reference a pathology result. If the code was posted before pathology returned, the claim is unsupported on the date of service.
What Structured Prague Data Looks Like in Scribing.io
Scribing.io captures Prague C & M as FHIR Observation resources with the following structure:
FHIR Element | Value (Example: C0M2) | Purpose |
|---|---|---|
| LOINC for Prague Classification | Standard clinical terminology binding |
| 0 cm | Circumferential extent — discrete, queryable |
| 2 cm | Maximal extent — drives interval logic |
| Date of EGD | Links measurement to the specific procedure |
| Reference to pathology DiagnosticReport | Creates the pathology → Prague → code → interval chain |
This structure means that every downstream system — the billing engine, the scheduler, the quality registry, the MIPS reporting module — receives the same discrete values. There is no transcription. There is no interpretation. The M value of 2 cm drives the 5-year interval recommendation automatically, and that recommendation flows to the scheduler as a structured order, not as a note the scheduler has to read and interpret.
The Segment Length Threshold: 3 cm
The 3 cm threshold separating shorter-segment from longer-segment nondysplastic Barrett's is derived from risk stratification data published in the AGA Clinical Practice Update (2022) and the ACG Clinical Guideline. The evidence demonstrates that the risk of neoplastic progression is meaningfully lower when M < 3 cm and no dysplasia is present, supporting a longer surveillance interval. Scribing.io encodes this threshold as a configurable parameter — practices that adopt different risk stratification criteria (e.g., based on institutional cancer registry data) can adjust the cutoff while maintaining the same structured workflow.
The key clinical point: the threshold must be applied against a discrete M value, not against a subjective "short" or "long" designation. "Short-segment" is not a measurement. Prague M = 2 cm is.
Anatomy of a Barrett's Post-Pay Audit: What Triggers Recoupment
The Audit Sequence
Understanding the payer's audit logic is not optional — it is the foundation of defensive documentation. Based on published OIG work plans and commercial payer audit protocols, here is the sequence a Barrett's surveillance EGD undergoes during post-pay review:
Claim pull: The payer identifies all EGDs billed with K22.70 or K22.71x within a rolling 36-month window per patient.
Interval check: If two EGDs are less than 5 years apart and the code is K22.70 (nondysplastic), the claim is flagged.
Documentation request: The practice must produce the endoscopy note, pathology report, and any relevant prior records.
Prague verification: The auditor looks for discrete Prague C & M values. If M ≥ 3 cm, the 3-year interval may be justified. If M < 3 cm, the auditor expects either a 5-year interval or a documented clinical rationale for deviation.
Pathology linkage: The auditor verifies that the pathology report — with an accession number and result date — explicitly states "no dysplasia" and is linked to the encounter. If K22.70 was posted before pathology resulted, the code is unsupported.
Recoupment decision: If the interval is not guideline-justified, if Prague values are absent, or if pathology linkage is missing, the claim is recouped. Appeals succeed only when the practice can produce the structured data that should have been in the note from the start.
Where Practices Fail Without Structured Workflows
Failure Point | Frequency | Scribing.io Prevention Mechanism |
|---|---|---|
Prague values documented in free text only | High — estimated in the majority of GI practices using standard EHR templates | FHIR Observation components with mandatory C and M fields; note cannot be signed without values entered |
"Short-segment" used as primary descriptor | Moderate — legacy terminology persists in auto-populated templates | Terminology gating suppresses the phrase and substitutes Prague-based language |
K22.70 posted before pathology returns | High — coders under pressure to close encounters same-day | Pathology-gated claim hold; K22.70 cannot finalize until "no dysplasia" attestation is completed |
3-year recall set for M < 3 cm without dysplasia | Moderate — schedulers follow physician verbal orders without interval verification | Interval logic engine sets 5-year default; override requires structured rationale |
K22.70 used when pathology reads "indefinite for dysplasia" | Low-moderate — requires coder to read pathology carefully | Auto-resolution to K22.719 when "indefinite" language is detected in the pathology interface |
Implementation Checklist: Rolling Out Structured Barrett's Documentation
For GI Medical Directors ready to operationalize this playbook, here is the implementation sequence. Each step can be deployed incrementally — you do not need to overhaul your entire EHR to start capturing structured Prague data.
Audit your current Barrett's notes (Week 1). Pull 20 recent surveillance EGD notes coded K22.70. Count how many contain discrete Prague C & M values (not embedded in prose). Count how many reference a specific pathology date and result. Count how many use "short-segment." This is your baseline.
Identify your EHR's structured data gaps (Week 2). Determine whether your endoscopy module supports discrete Prague fields. If it does, verify they export to the billing and scheduling systems. If it does not, this is where Scribing.io's FHIR-based capture layer integrates — it sits between the clinician and the EHR, capturing structured data that the EHR's native template cannot.
Deploy terminology gating (Week 3). Configure your documentation workflow — whether Scribing.io or a custom template — to flag "short-segment" and prompt Prague-based language. Train scribes and physicians on the rationale: "short-segment" costs money; Prague C0M2 saves it.
Implement pathology-gated claim release (Week 4). Establish a pre-submission queue for all Barrett's-coded encounters. Claims with K22.70 do not release until pathology is linked. Claims with "indefinite for dysplasia" auto-resolve to K22.719. Review the CMS guidelines to confirm your code selection protocols are current.
Activate interval logic (Week 5). Connect the Prague M value to your scheduling system. M < 3 cm + no dysplasia = 5-year recall. M ≥ 3 cm + no dysplasia = 3-year recall. Override requires documentation. This single change eliminates the most common Barrett's audit vulnerability.
Run a mock audit (Week 6). Apply the five-checkpoint audit table from the first section of this playbook to your last 50 Barrett's surveillance claims. Score each claim pass/fail. Any claim that fails a single checkpoint is an active recoupment risk. This is your ROI justification for the entire workflow.
See the Prague C & M Auto-Capture in Action
The workflow described in this playbook is not theoretical. It is live in Scribing.io today, running in GI practices that have eliminated Barrett's surveillance audit recoupments entirely.
See our Prague C & M auto-capture with pathology-gated ICD-10 engine and length-based (3y vs 5y) surveillance guardrails, with one-click FHIR export to Epic/athena. Book a 15-minute demo to watch it prevent a real denial in seconds.
Every day you document Barrett's without structured Prague data, you are building the audit trail that payers will use against you. The fix is five discrete data fields and one attestation. The cost of not fixing it is every surveillance EGD you have billed in the last three years.
