PCOS Diagnostic Specificity: AI Documentation for REI — The Operations Playbook

Clinical Update — June 2026: This guide has been revised to incorporate the 2026 AMA E/M MDM documentation guidelines, updated international PCOM thresholds (≥20 follicles/ovary per the 2023 evidence-based PCOS guideline now fully adopted by ACOG/ASRM for 2026 coding cycles), and CMS Modifier 25 audit triggers effective Q2 2026. FHIR R5 resource mappings have been updated to reflect HL7's January 2026 ballot reconciliation for reproductive health Observation profiles.

TL;DR: REI clinics lose thousands per week on PCOS workup revenue because documentation fails to operationalize the Rotterdam Criteria inside E/M Medical Decision Making. The AMA's AI evaluation framework (and every competitor) describes diagnostic criteria generically but never maps them to MDM complexity elements, FHIR-structured data capture, or modifier 25 sufficiency logic. Scribing.io's REI-specific AI documentation system enforces a Rotterdam chain-of-evidence — from Ferriman–Gallwey scoring and OCP-confounded lab flagging to per-ovary PCOM quantification — then auto-binds these to MDM Data and Risk elements, justifying 99204/99205 and protecting same-day 76830 claims from denial. This playbook details how.

• The Revenue Gap Competitors Miss: Operationalizing Rotterdam Inside E/M MDM

• Scribing.io Clinical Logic: Handling the PCOS New-Patient Scenario

• Rotterdam Chain-of-Evidence: FHIR-Structured Documentation Architecture

• Technical Reference: ICD-10 Documentation Standards

• OCP Confounders and Lab Invalidation Logic

• Modifier 25 Sufficiency and Same-Day 76830 Compliance

• MDM Complexity Scoring: From Clinical Findings to 99204/99205 Justification

• Audit Defense: Exclusion Documentation and Payer-Proof Notes

The Revenue Gap Competitors Miss: Operationalizing Rotterdam Inside E/M MDM

The AMA's 2026 E/M MDM framework provides the reimbursement scaffolding. The 2023 International Evidence-Based PCOS Guideline provides the diagnostic scaffolding. No documentation tool on the market welds these two structures together at the point of clinical capture. That is the gap. That is where Scribing.io operates.

The Rotterdam Criteria for PCOS diagnosis require documentation of at least two of three features: oligo/anovulation, clinical or biochemical hyperandrogenism, and polycystic ovarian morphology (PCOM). Every EMR template and ambient AI scribe in this market describes these criteria. None operationalize them inside the E/M MDM framework that determines whether a payer reimburses at 99205 or downcodes to 99203. This distinction — between describing and operationalizing — is the difference between a note that reads well and a note that bills correctly.

For practices already observing how specialty-specific logic engines outperform generic documentation tools, our deployments in Cardiology and Family Medicine demonstrate this principle across service lines. The REI implementation extends the same architecture into reproductive endocrinology's most documentation-dense encounter type.

What the AMA Framework Misses for REI

The AMA's AI tool evaluation guidance asks whether a tool's "clinical function" aligns with workflows and whether "ground truth" matches specialty norms. Necessary questions. Insufficient questions for an REI Medical Director evaluating documentation AI for PCOS workups.

The anchor truth: Beyond IVF cycles, REI clinics lose substantial revenue on PCOS workups. Documentation must link Rotterdam Criteria findings (hirsutism, ultrasound, labs) to the MDM logic to justify high-complexity initial consults. Current payer audit data indicates E/M downcode rates for new-patient PCOS visits range from 18–32% across commercial payers when notes lack explicit Rotterdam criteria linkage in the assessment. This isn't a clinical knowledge deficit — REI physicians know the Rotterdam Criteria intimately — it's a documentation architecture failure that ambient AI tools must solve at the structural level.

Scribing.io Clinical Logic: Handling the PCOS New-Patient Scenario

The Scenario

A new patient presents to a high-volume REI clinic for infertility and irregular menses while on OCPs. The visit includes same-day transvaginal ultrasound and a broad PCOS lab panel. The claim for 99205 + 76830 is downcoded/denied because the note doesn't explicitly tie two Rotterdam criteria to the assessment, the ultrasound lacks per-ovary follicle counts and probe frequency, labs are confounded by OCP use without acknowledgment, and no "significant, separately identifiable" E/M is documented for modifier 25. Revenue lost: E/M downcode plus ultrasound denial across two payers.

Revenue Impact Quantification

For a high-volume REI clinic seeing 15–25 new PCOS evaluations per week, annualized losses from documentation failure range from $238,000 to $565,000 — revenue that was clinically earned but documentation-lost.

With Scribing.io Running: The Seven-Step Prompted Documentation Chain

Step 1: Clinical Hyperandrogenism Capture

The clinician is prompted to verbalize the modified Ferriman–Gallwey (mFG) score. Scribing.io's voice prompts guide assessment across nine body areas (upper lip, chin, chest, upper back, lower back, upper abdomen, lower abdomen, arm, thigh), records the composite score, and flags whether it exceeds the ethnicity-adjusted threshold (≥4–6 depending on population norms per the Ferriman-Gallwey scoring validation literature) for clinical hyperandrogenism documentation. The mFG score is captured as a discrete FHIR Observation, not buried in free text.

Step 2: OCP Confounder Acknowledgment

The system detects OCP use from medication reconciliation (Epic/Cerner FHIR MedicationStatement resource) and triggers an explicit confounder alert:

• Documents that biochemical hyperandrogenism labs (total testosterone, free testosterone, DHEA-S) are potentially invalid under active OCP suppression due to SHBG elevation

• Prompts clinician to verbalize clinical plan: washout period (minimum 3 months) before re-testing vs. reliance on clinical signs (mFG score, acne grading per the Global Acne Grading System) as the hyperandrogenism criterion

• Records this reasoning as MDM complexity — independent interpretation of confounded data constitutes a high-complexity data element under CMS E/M guidelines

Step 3: Ultrasound PCOM Quantification

For same-day transvaginal ultrasound, the system enforces structured capture of:

• Per-ovary follicle count (right ovary: ___; left ovary: ___)

• Ovarian volume per ovary (length × width × height × 0.523, reported in mL)

• Probe frequency used (endovaginal transducer, typically 5–9 MHz)

• Cycle day at time of imaging

• Whether PCOM threshold is met (≥20 follicles per ovary measuring 2–9mm and/or ovarian volume ≥10 mL per the 2023 international guideline threshold)

Step 4: Exclusion of Mimickers

The system prompts documentation of ordered/reviewed exclusion labs:

• TSH (thyroid disease exclusion)

• Prolactin (hyperprolactinemia exclusion)

• 17-hydroxyprogesterone, drawn in early follicular phase AM (non-classic congenital adrenal hyperplasia exclusion)

• 24-hour urinary free cortisol or late-night salivary cortisol (Cushing syndrome exclusion if clinically indicated)

Each exclusion documented adds to MDM Data complexity and Risk assessment. This is where documentation generates billable complexity: ruling out dangerous differentials is high-risk MDM by definition.

Step 5: MDM Auto-Linking

Step 6: Modifier 25 Sufficiency Check

The system verifies that the E/M documentation contains a "significant, separately identifiable" evaluation beyond the ultrasound procedure:

• Confirms history elements (menstrual history, hirsutism timeline, fertility goals, OCP duration/type, weight trajectory) are documented as distinct from ultrasound indication

• Confirms physical examination (mFG scoring, BMI, acanthosis nigricans screening, thyroid palpation, galactorrhea check) is documented

• Confirms MDM elements exist independent of the ultrasound interpretation alone

• Generates a sufficiency flag: ✓ Modifier 25 supported or ⚠ Additional documentation required before note finalization

Step 7: Note Generation with Structural Separation

The final note auto-includes:

• Separate E/M language clearly delineating the cognitive evaluation from the procedural ultrasound

• Rotterdam criteria explicitly identified in the Assessment (e.g., "Rotterdam Criterion 1: oligo-ovulation confirmed by history of ≤8 cycles/year; Rotterdam Criterion 2: clinical hyperandrogenism confirmed by mFG score of 12/36")

• Independent ultrasound interpretation section with all required measurements, separated from E/M body

• MDM summary paragraph explicitly stating complexity level and supporting elements

Result: Clean first-pass payment for 99205 + 76830-25. No downcode. No denial. No appeal cycle consuming 45 minutes of staff time.

Rotterdam Chain-of-Evidence: FHIR-Structured Documentation Architecture

Scribing.io's REI template creates a structured, interoperable clinical data object — not merely a narrative note. The Rotterdam chain-of-evidence is encoded using HL7 FHIR R5 Observation resources with LOINC-coded elements, enabling bidirectional communication with Epic (via FHIR R4/R5 APIs) and Cerner (Oracle Health FHIR endpoints).

FHIR Resource Mapping

The DiagnosticReport Summary Object

The FHIR DiagnosticReport resource functions as the apex clinical object, referencing all constituent Observations and rendering a machine-readable and human-readable summary:

• Criterion 1 (Oligo/Anovulation): Status [Met/Not Met/Indeterminate] — Evidence: menstrual history reference, cycle frequency documentation

• Criterion 2 (Hyperandrogenism): Status [Met/Not Met/Indeterminate] — Sub-type: [Clinical/Biochemical/Both] — Evidence: mFG Observation reference / lab Observation references — Confounders: [OCP active — biochemical criteria deferred to washout labs or clinical criteria used]

• Criterion 3 (PCOM): Status [Met/Not Met/Indeterminate] — Evidence: bilateral follicle count Observations, bilateral volume Observations, probe frequency, cycle day

• Exclusion Status: [All mimickers excluded / Pending exclusion labs / Exclusion lab abnormal — alternate diagnosis under investigation]

This structured object persists in the EHR, travels with referrals via FHIR messaging, and — critically — provides the audit trail that billing compliance teams and payer medical directors require for high-complexity E/M justification.

Technical Reference: ICD-10 Documentation Standards

PCOS workup encounters require precise ICD-10-CM code assignment that reflects the full clinical picture. Undercoding or imprecise coding triggers automated payer edits that either deny claims outright or flag them for pre-payment review. Scribing.io's logic engine ensures maximum specificity by mapping documented Rotterdam findings directly to the highest-specificity applicable codes.

Primary Code Assignments

E28.2 — Polycystic ovarian syndrome; N97.0 — Female infertility associated with anovulation

Scribing.io ensures these codes reach maximum specificity through the following logic:

• E28.2 (Polycystic ovarian syndrome): Only assigned when the note explicitly documents ≥2 Rotterdam criteria met AND exclusion of mimickers documented. The system will not suggest E28.2 if the note contains only "suspected PCOS" without criteria linkage — it will flag the documentation gap for clinician resolution before code assignment.

• N97.0 (Female infertility associated with anovulation): Assigned as secondary diagnosis when oligo/anovulation is documented as the mechanism linking PCOS to the infertility chief complaint. The system verifies that anovulation is explicitly stated (not merely implied by irregular menses) and that the clinical note ties menstrual irregularity to failed conception timeline.

Supporting Code Logic

The system prevents the common coding error of assigning E28.2 without supporting documentation of diagnostic criteria — a practice that triggers CMS Recovery Audit Contractor (RAC) reviews when the diagnosis lacks clinical substantiation in the note. By enforcing criteria documentation before code suggestion, Scribing.io eliminates this audit vulnerability at the source.

OCP Confounders and Lab Invalidation Logic

Oral contraceptive pills suppress gonadotropin secretion, elevate sex hormone-binding globulin (SHBG), and reduce bioavailable androgen levels. A PCOS lab panel drawn under active OCP use produces systematically invalid results for the biochemical hyperandrogenism criterion. This creates a documentation paradox: the labs are ordered (contributing to MDM Data complexity), but their interpretation requires explicit confounder documentation to avoid audit vulnerability and to correctly characterize which Rotterdam criterion is being satisfied.

Scribing.io's OCP Confounder Logic Tree

1. Medication Detection: On encounter initiation, the system queries the active medication list (FHIR MedicationStatement) for combined oral contraceptives, progestin-only pills, and other hormonal preparations (spironolactone, finasteride) that confound androgen measurements.

2. Alert Generation: If hormonal medication detected, the system generates an in-encounter alert: "Active OCP detected. Biochemical hyperandrogenism labs may be suppressed. Document clinical plan for criterion satisfaction."

3. Clinician Decision Capture: The system prompts and captures the clinician's verbalized decision:

   • Option A — Washout: "Plan to discontinue OCP for 3 months, repeat total testosterone, free testosterone, and DHEA-S in early follicular phase. Will rely on clinical hyperandrogenism (mFG) and PCOM for current Rotterdam assessment."

   • Option B — Clinical reliance: "Given mFG score of [X] exceeding threshold, clinical hyperandrogenism criterion is satisfied independent of biochemical markers. OCP confounding acknowledged; biochemical testing deferred as unnecessary for Rotterdam diagnosis."

4. MDM Complexity Credit: This clinical reasoning — interpreting confounded data, weighing alternatives, documenting the rationale for using clinical over biochemical criteria — constitutes high-complexity MDM per the AMA's MDM table. Scribing.io auto-tags this reasoning as a Data element (independent interpretation of ambiguous results) and Risk element (management decision affected by complicating factors).

Without this documentation, auditors see labs drawn under OCP suppression without clinician acknowledgment of their limitations — a finding that triggers both clinical quality concerns and coding downgrades. With Scribing.io, the confounder is documented, the clinical reasoning is explicit, and the MDM credit is earned.

Modifier 25 Sufficiency and Same-Day 76830 Compliance

Modifier 25 appended to the E/M code indicates that the evaluation and management service was "significant, separately identifiable" from any procedure performed on the same day. For REI encounters combining a new-patient evaluation with same-day transvaginal ultrasound (76830), payers scrutinize whether the E/M note truly documents cognitive work beyond the procedural interpretation. CMS National Correct Coding Initiative (NCCI) edits and commercial payer policies are increasingly aggressive on modifier 25 denials.

Scribing.io's Modifier 25 Sufficiency Engine

The system performs a real-time sufficiency analysis before note finalization. If the E/M documentation fails any of the four pillars, the clinician receives a specific prompt (e.g., "Physical examination documentation does not contain elements beyond the ultrasound. Please document additional exam findings to support modifier 25"). The note is not finalized with modifier 25 flagged as supported until all four pillars are satisfied.

Independent Interpretation Credit

When the REI physician performs and interprets the transvaginal ultrasound themselves (as is standard in most REI practices), this qualifies as an "independent interpretation of a test performed by another physician/QHP" MDM data credit — but only if the interpretation is documented with sufficient specificity. Scribing.io enforces the required elements: measurements, findings, clinical correlation statement, and a separately signed interpretation report. This interpretation credits the MDM Data category while the procedural work justifies 76830 — a dual benefit that generic scribes miss entirely.

MDM Complexity Scoring: From Clinical Findings to 99204/99205 Justification

The 2026 AMA MDM framework scores complexity across three categories: Number and Complexity of Problems, Amount and/or Complexity of Data, and Risk of Complications and/or Morbidity. Two of three must meet threshold for the billed level. Scribing.io maps Rotterdam documentation directly to these categories.

99205 Justification Map for New PCOS Consult

Scribing.io's MDM auto-scoring engine tallies these elements in real time. If documented findings support 99205, the system confirms. If documentation falls short (e.g., external records not explicitly summarized, or Risk language doesn't reach "high"), the system identifies the specific gap and prompts clinician verbalization to close it — before the note is signed.

Audit Defense: Exclusion Documentation and Payer-Proof Notes

Payer audits of high-complexity new-patient E/M codes (99205) target three vulnerabilities: (1) MDM complexity not supported by documentation, (2) modifier 25 applied without separately identifiable service, and (3) diagnosis codes assigned without clinical criteria substantiation. Scribing.io's architecture addresses all three simultaneously.

The Exclusion Documentation Advantage

Documenting exclusion of PCOS mimickers serves dual purposes: clinical rigor and billing defense. Per NIH's Endotext, the differential diagnosis of hyperandrogenism includes non-classic congenital adrenal hyperplasia (prevalence 1–10% depending on ethnicity), thyroid dysfunction, hyperprolactinemia, Cushing syndrome, and androgen-secreting tumors. Each exclusion documented:

• Adds a differential diagnosis to the Problem complexity (undiagnosed problem requiring exclusion of dangerous alternate = high complexity)

• Adds ordered tests to the Data complexity (17-OHP, TSH, prolactin, cortisol = multiple test categories)

• Adds risk awareness to the Risk category (excluding life-threatening diagnoses = high-risk clinical scenario)

Audit-Defense Log Architecture

Scribing.io generates a parallel audit-defense metadata log for each encounter that includes:

• Timestamp of each Rotterdam element captured during encounter

• Clinician voice confirmation of mFG score, follicle counts, and clinical reasoning

• Modifier 25 sufficiency checklist completion status

• MDM element-by-element justification with reference to specific note language

• FHIR resource references for all structured data points

• Rotterdam criteria summary confirming ≥2 of 3 documented or documenting which are pending (with plan)

This log is accessible to billing compliance officers and can be produced within 24 hours of an audit request — compared to the typical 2–3 week scramble to reconstruct documentation rationale from a narrative note alone.

Payer-Specific Compliance Flags

The system maintains updated payer-specific policies for major commercial insurers and Medicare Administrative Contractors. Known triggers include:

• UnitedHealthcare: Requires explicit documentation that E/M service would have prompted a separate encounter "but for" the same-day procedure

• Aetna: Requires separate chief complaint documentation for E/M vs. procedure

• Medicare (Novitas, CGS): Requires medical necessity documentation for same-day US separate from E/M indication

Scribing.io's note template includes payer-appropriate language automatically based on the patient's insurance profile loaded from the EHR eligibility check.

The Bottom Line for REI Medical Directors

Documentation is not clerical work. In REI PCOS workups, documentation is the mechanism by which clinical complexity becomes reimbursable complexity. Every Rotterdam criterion you assess but fail to explicitly link to MDM logic is revenue foregone. Every same-day ultrasound you perform without structural separation in the note is a modifier 25 denial waiting to happen. Every OCP-confounded lab you order without documenting the confounder and your clinical reasoning is an audit vulnerability.

Scribing.io doesn't ask you to change your clinical workflow. It captures the reasoning you're already performing and structures it into documentation that survives first-pass adjudication, appeals, and RAC audits.

Book a 15-minute demo to see our Rotterdam-to-MDM auto-justification engine with modifier-25 guardrails, Epic/Cerner FHIR bindings for Ferriman–Gallwey, follicle count, and ovarian volume, and an audit-defense log that protects 99205 + 76830 on PCOS workups.