AI Documentation for Chiropractors: The P.A.R.T. Rule — How Scribing.io Closes the 33.6% Improper Payment Gap

• Why P.A.R.T. Compliance Remains the Largest Revenue Leak in Chiropractic Medicare

• The Information Gain Competitors Miss: Why A/R Region-Mapping Is the Real Audit Firewall

• Scribing.io Clinical Logic: Handling a Medicare DC TPE Audit Scenario

• P.A.R.T. Grid Enforcement Architecture: Point-of-Dictation Constraints

• Technical Reference: ICD-10 Documentation Standards

• Cross-Specialty Documentation Parallels

• LCD-Aligned Audit Packet Generation

• Implementation Workflow for Medicare-Participating DCs

• Run Your Last 50 Medicare CMT Notes Through the PART Compliance Scanner

TL;DR: Medicare's 2024 data shows a 33.6% improper payment rate for chiropractic services—$178.3 million in projected losses—with 95.5% caused by insufficient documentation. The root cause is predictable: progress notes fail to capture exactly 2 of the 4 P.A.R.T. elements (with at least one being Asymmetry or Range of Motion), fail to tie findings to named spinal regions, or miscalculate the CMT code range (98940–98942). Scribing.io is the only AI documentation platform that enforces the P.A.R.T. grid at the point of dictation, region-maps A or R findings to auto-derive the correct CPT code and AT modifier, and exports LCD-aligned audit packets—transforming a DC's highest compliance liability into a structured, defensible workflow.

Why P.A.R.T. Compliance Remains the Largest Revenue Leak in Chiropractic Medicare

The CMS compliance page for chiropractic services states a plain fact that should alarm every Medicare-participating DC in the country: 33.6% of all chiropractic Medicare fee-for-service payments in 2024 were improper, totaling a projected $178.3 million. Of those improper payments, 95.5% were caused by insufficient documentation—not fraud, not upcoding, not missing claims. Documentation.

This means the overwhelming majority of recoupment actions, TPE (Targeted Probe and Educate) reviews, and ADR (Additional Documentation Request) failures trace back to a single structural problem: the clinical note did not contain what the MAC reviewer needed to see. Scribing.io was engineered specifically to solve this structural problem—not with reminders or checklists, but with hard constraints embedded in the AI's note generation logic that prevent a non-compliant note from ever reaching the EHR.

CMS's own guidance is explicit about what the physical exam must include. To demonstrate subluxation by physical exam (as opposed to X-ray), the provider must document exactly 2 of the 4 P.A.R.T. criteria, and at least one must be Asymmetry/misalignment (A) or Range of motion abnormality (R). This is codified in the Medicare Benefit Policy Manual, Chapter 15, §240.1.3. It is not a recommendation. It is a coverage requirement. Fail it, and the claim is denied on its face.

Yet CMS's compliance materials—while enumerating each P.A.R.T. element in detail—do not address the operational question every practicing DC confronts multiple times per day: How do I reliably produce a note that satisfies this requirement during a 7-minute CMT encounter, across a 30-patient day, when I'm also managing patient communication, technique execution, and billing?

That operational gap is precisely what Scribing.io was built to close. Before we detail the solution, the specific documentation failures that cause 95.5% of those improper payments need dissecting—because they are remarkably consistent across MAC jurisdictions.

The Five Failure Modes

Clinical benchmarks from published TPE and CERT (Comprehensive Error Rate Testing) audit data indicate that chiropractic documentation failures cluster into five recurring patterns:

Every one of these failures is structurally preventable. They persist not because DCs lack clinical knowledge, but because the documentation tools they use—general-purpose EHRs, generic templates, manual dictation workflows—do not enforce the P.A.R.T. grid as a hard constraint at the moment the note is generated. The parallel exists in other specialties: cardiology AI scribes must enforce specific documentation structures for E/M leveling and procedure documentation, and family medicine AI scribes must handle multi-problem visit complexity. Chiropractic's documentation challenge is narrower but higher-stakes per claim: a single missing element triggers 100% recoupment.

The Information Gain Competitors Miss: Why A/R Region-Mapping Is the Real Audit Firewall

CMS correctly states that at least one of the two documented P.A.R.T. elements must be Asymmetry/misalignment or Range of motion abnormality. But the compliance page stops there. It does not address three critical downstream dependencies that MAC LCD reviewers actually evaluate during TPE and ADR reviews:

1. The A or R finding must be tied to a named spinal region (cervical, thoracic, lumbar, sacral/pelvic), not to a vague anatomical reference. "Decreased ROM in the upper back" does not satisfy the requirement. "Cervical flexion limited to 30° (normal 50°) at C4-C6" does.

2. The number of distinct spinal regions documented with A or R findings directly determines the correct CMT code. Per the AMA's CPT code definitions, 98940 covers 1–2 regions, 98941 covers 3–4 regions, and 98942 covers 5 regions. If a provider documents A or R findings in only two regions but bills 98941, the claim is vulnerable to downcoding or denial—even if the provider actually treated three or four regions.

3. Multiple MAC LCDs require that the A or R element demonstrate measurable change over time for subsequent visits. A static notation of "asymmetry present" on visit 14 that is identical to visit 1 raises a maintenance therapy flag per MAC local coverage determinations, potentially invalidating the AT modifier.

This is the original insight that Scribing.io operationalizes and that no competitor platform—including generic chiropractic EHRs and general-purpose AI scribes—has systematically addressed:

Multiple MAC LCDs require that one of the two documented P.A.R.T. elements be Asymmetry (A) or Range of Motion (R) tied to a specific spinal region. Scribing.io enforces an "exactly 2-of-4" P.A.R.T. grid and region-maps A/R findings to auto-derive the correct CMT code (98940–98942) and AT modifier—closing a common audit gap competitors miss.

The practical implication is significant. A DC using a standard template-based EHR might document "P: tenderness at T4-T6; T: hypertonicity paravertebral musculature thoracic spine." That note captures two P.A.R.T. elements—but they are P and T. Neither is A or R. The subluxation is not demonstrated by physical exam under CMS criteria. The claim will be denied on review, and the DC may not discover the error until a TPE letter arrives months later.

Scribing.io prevents this at the point of documentation by making the P.A.R.T. grid a structural requirement—not a checkbox, not a reminder, but a hard constraint in the AI's note generation logic. The system will not finalize a CMT encounter note unless exactly two P.A.R.T. elements are documented, at least one is A or R, and the A or R finding is mapped to a named spinal region with sufficient specificity to support the selected CPT code.

Scribing.io Clinical Logic: Handling a Medicare DC TPE Audit Scenario

Consider the following scenario—composited from patterns reported across MAC jurisdictions—that illustrates why structural enforcement matters:

A Medicare-participating DC in Ohio undergoes TPE review. 126 CMT claims are flagged, and $22,400 is recouped because progress notes only captured Pain and Tissue, referenced "upper back" instead of named spinal regions, and still billed 98941 with AT.

This is not an outlier. This is the median outcome of the 33.6% improper payment rate applied to a moderately busy chiropractic practice. Here is exactly how Scribing.io would have prevented each denial cause.

Denial Cause #1: Only Pain (P) and Tissue (T) Documented

The DC's notes consistently recorded tenderness to palpation (P) and hypertonicity/spasm (T). Both are valid P.A.R.T. elements. But the CMS requirement is unambiguous: at least one must be A or R. A note with only P and T does not demonstrate subluxation by physical exam.

Scribing.io Prevention: During ambient dictation capture, the Scribing.io P.A.R.T. engine monitors which elements the provider verbalizes in real time. If the provider mentions pain/tenderness and tissue findings but does not verbalize an asymmetry or ROM observation, the system flags the note as incomplete before finalization. The provider receives a structured prompt: "P.A.R.T. grid requires at least one Asymmetry or Range of Motion finding. Please dictate an A or R observation for [identified spinal region]." The note cannot be signed and locked without satisfying the 2-of-4 constraint with A or R inclusion.

Denial Cause #2: "Upper Back" Instead of Named Spinal Regions

The phrase "upper back" does not correspond to a specific spinal region in CMS's coding framework. It could mean upper thoracic, lower cervical, or a combination. MAC reviewers cannot validate the CPT code against a non-specific anatomical reference.

Scribing.io Prevention: The system's region-mapping module normalizes all anatomical references to CMS-recognized spinal regions: cervical, thoracic, lumbar, sacral, and pelvic. If the provider says "upper back," Scribing.io prompts for clarification: "Please specify spinal region: cervical (C1-C7), thoracic (T1-T12), or cervicothoracic junction." The A or R finding is then locked to the specified region, and the subluxation level is auto-populated with the corresponding vertebral segments.

Denial Cause #3: 98941 Billed but Region Count Unsupported

CPT 98941 requires chiropractic manipulative treatment of 3–4 spinal regions. If the documentation only references one region (even vaguely as "upper back"), the code is unsupported regardless of what treatment was actually rendered.

Scribing.io Prevention: As A/R findings are region-mapped, Scribing.io maintains a running count of distinct spinal regions with documented subluxation findings. The system auto-derives the appropriate CPT code:

If the provider's dictation supports only 2 regions but the practice management system has previously defaulted to 98941, Scribing.io alerts the discrepancy and suggests the correct code. The provider can then either add clinical findings for additional regions or accept the downcoded CPT. This prevents billing a code the note cannot defend.

Denial Cause #4: AT Modifier Applied Without Active Treatment Justification

The AT modifier indicates that the service is an active/corrective treatment, not maintenance therapy. For chronic subluxation patients, MAC reviewers specifically look for objective evidence of ongoing functional improvement—most commonly demonstrated through progressive changes in A or R findings across visits. The Medicare Benefit Policy Manual, Chapter 15, §240.1.1 explicitly distinguishes active treatment from maintenance care.

Scribing.io Prevention: The system tracks A and R values longitudinally across the patient's visit history. If a patient's cervical ROM has been documented as "flexion 35°" for six consecutive visits with no change, Scribing.io flags a maintenance therapy risk: "ROM values unchanged across 6 visits. AT modifier may not be supportable. Document clinical rationale for continued active treatment or consider maintenance care classification." This gives the provider the clinical decision point before the note is finalized—not after a TPE review triggers a recoupment demand.

P.A.R.T. Grid Enforcement Architecture: Point-of-Dictation Constraints

Scribing.io's P.A.R.T. enforcement is not a post-hoc audit tool bolted onto a generic scribe. It is embedded in the real-time transcription and structuring pipeline. The architecture operates in four sequential validation gates:

1. Gate 1 — Element Detection: As the provider dictates, natural language processing classifies clinical statements into P.A.R.T. categories. "Tenderness at C5-C6" → Pain. "Cervical flexion limited to 30 degrees" → Range of Motion. "Spinous process laterality at T4" → Asymmetry. "Paravertebral hypertonicity at L3-L5" → Tissue. Each classified statement is tagged with its element type and associated spinal region.

2. Gate 2 — Constraint Validation: Before note finalization, the system checks: (a) Are exactly 2 P.A.R.T. elements documented? (b) Is at least one of them A or R? If either condition fails, the note is held in draft status with a specific prompt identifying what is missing.

3. Gate 3 — Region-Code Mapping: The distinct spinal regions associated with A/R findings are counted. The CPT code is auto-derived from the region count. If the auto-derived code differs from any pre-selected code in the billing queue, the provider is alerted.

4. Gate 4 — AT Modifier Logic: The system evaluates whether the documented A/R findings show measurable change from the prior visit. If they do, AT is applied. If they do not, the provider receives the maintenance therapy flag described above and must either document additional clinical rationale or remove the AT modifier.

This four-gate architecture means that every CMT note exiting Scribing.io into the EHR has already passed the same validation a MAC reviewer would apply. The note is pre-audited.

Technical Reference: ICD-10 Documentation Standards

P.A.R.T. compliance is the exam documentation requirement. ICD-10 coding is the diagnosis documentation requirement. Both must align for a CMT claim to survive audit. The most commonly billed chiropractic ICD-10 codes—the M99.0x series for segmental and somatic dysfunction—require region-level specificity that directly mirrors the P.A.R.T. region-mapping discussed above.

The critical codes are: M99.01 Segmental and somatic dysfunction of cervical region; M99.03 Segmental and somatic dysfunction of lumbar region. These codes demand that the documentation specify which spinal region is affected—cervical, thoracic, lumbar, sacral, or pelvic. A note that references "upper back" without specifying a region cannot support M99.01 (cervical) or M99.02 (thoracic) with confidence, creating a mismatch between the billed diagnosis and the clinical narrative.

Scribing.io ensures these codes reach maximum specificity through three mechanisms:

• Auto-population from region-mapped findings: When the P.A.R.T. engine maps an A or R finding to "cervical region, C4-C6," the system automatically suggests M99.01 as the primary diagnosis. When lumbar findings are documented at L3-L5, M99.03 is suggested. The provider confirms or modifies, but the baseline suggestion is always region-specific.

• Multi-region code stacking: For encounters involving multiple spinal regions (supporting 98941 or 98942), Scribing.io generates a complete list of region-specific M99.0x codes corresponding to each documented subluxation level. This prevents the common error of billing 98941 (3–4 regions) with only one M99.0x code—a mismatch that triggers automated claim edits at many MACs.

• Laterality and segment-level notation: Where the provider documents laterality (e.g., "left rotation restriction at C3-C4"), Scribing.io preserves this specificity in the note narrative even though M99.0x codes do not have laterality modifiers. This additional detail strengthens the note's defensibility under audit, as it demonstrates clinical precision beyond the minimum coding requirement. Per WHO ICD-10 classification standards, maximum available specificity is always the documentation target.

The relationship between P.A.R.T. findings and ICD-10 codes is not incidental—it is the same data mapped to two different compliance frameworks. A note with properly region-mapped A/R findings automatically supports the correct M99.0x code. A note without region-mapping fails both.

Cross-Specialty Documentation Parallels

The structural enforcement approach Scribing.io applies to chiropractic P.A.R.T. documentation is an instance of a broader principle: specialty-specific AI documentation must enforce the specific compliance grid that governs that specialty's highest-risk claim type. In cardiology, that grid involves procedural documentation for catheterization and electrophysiology studies, where missing elements trigger identical denial patterns. In family medicine, the grid involves medical decision-making (MDM) complexity leveling for E/M visits under the 2021 AMA E/M guidelines. The principle is the same: identify the specific compliance structure that the payer evaluates, embed it as a hard constraint in the AI's note generation, and prevent non-compliant notes from reaching the billing queue.

For chiropractic, the P.A.R.T. grid is that structure. No other specialty has a comparably specific, binary pass/fail documentation requirement attached to its primary service code. And no other specialty has a 33.6% improper payment rate as direct evidence that current tools are failing to enforce it.

LCD-Aligned Audit Packet Generation

When a TPE or ADR request arrives, the DC's response time and packet quality directly influence the outcome. The CMS TPE process gives providers 45 days to respond. Industry data shows most chiropractic practices spend 2–3 weeks compiling records, reformatting notes, and writing cover letters—time that could be clinical hours.

Scribing.io exports a pre-formatted audit response packet containing:

• Each visit note with the P.A.R.T. grid highlighted: A/R region-mapping is visually flagged with the subluxation level, making it immediately apparent to the reviewer that the 2-of-4 requirement with A or R inclusion is satisfied.

• A longitudinal A/R summary: A tabular view of A/R findings by spinal region across all visits under review, demonstrating measurable change (ROM degree improvements, asymmetry reduction) that supports the AT modifier.

• CPT code derivation logic: For each visit, the packet shows: region count → code selection → documented regions. This preemptively answers the "how did you arrive at 98941?" question that MAC reviewers ask.

• AT modifier justification: Each visit's AT modifier is linked to the specific A/R findings that demonstrate active subluxation requiring corrective treatment, with comparison to prior visit findings.

• ICD-10 code-to-finding mapping: Each billed M99.0x code is linked to the specific P.A.R.T. finding and spinal region in the note that supports it.

The packet is auto-generated in the format required by the provider's specific MAC jurisdiction (CGS, Novitas, NGS, Palmetto, WPS, or First Coast). This reduces response time from weeks to hours—typically same-day or next-day export.

Implementation Workflow for Medicare-Participating DCs

Deploying Scribing.io's P.A.R.T. enforcement in a chiropractic practice follows a defined implementation sequence:

The typical practice reaches full autonomous operation—where the provider's natural dictation consistently triggers zero P.A.R.T. prompts—within 15–20 clinical days. At that point, the system is functioning as a silent validator: confirming compliance rather than correcting deficiencies.

Run Your Last 50 Medicare CMT Notes Through the PART Compliance Scanner

The 33.6% improper payment rate is not an abstraction. For a DC billing 40 Medicare CMT visits per week at an average of $45 per visit, that rate translates to approximately $25,000 in annual revenue at risk of recoupment—before accounting for the administrative cost of responding to TPE reviews, the opportunity cost of restricted billing during audit periods, or the reputational impact of a failed Round 1 TPE escalating to Round 2 and Round 3.

See our MAC-aware PART validator that enforces exactly 2-of-4 with A-or-R, region-links findings to auto-code 98940–98942, applies AT, and generates a one-click LCD audit packet—live in your EHR. Book a demo to run your last 50 Medicare CMT notes through the PART compliance scanner.

The scanner identifies which of the five failure modes are present in your current documentation, quantifies your recoupment exposure by failure mode, and shows you exactly what the corrected note would look like with Scribing.io's four-gate enforcement active. Every DC who has run this scan has found at least one systemic failure mode they did not know existed in their notes. Most find two or three.

The $178.3 million in projected improper payments is not a single catastrophic failure. It is 33.6% of routine claims, denied one note at a time, because the documentation tool did not enforce what the payer requires. Scribing.io enforces it. Every note. Every visit. Every region.