ICD-10 F31.9: Bipolar Disorder, Unspecified — The Complete Clinical Documentation & Prior Authorization Playbook for Outpatient Psychiatrists

TL;DR: F31.9 (Bipolar disorder, unspecified) is the most common initial bipolar code in outpatient psychiatry—and the single largest driver of preventable mood stabilizer PA denials. The root cause isn't clinical judgment; it's a structured-data gap. PBM ePA bots in 2026 parse FHIR Condition resources, not your narrative note. A lifetime hypomanic episode buried in prose without a discrete Problem List entry with onset date gets ignored. The bot sees no hypomania, classifies the patient as unipolar MDD, and denies cariprazine, lamotrigine, or whatever you prescribed. This playbook details the exact FHIR-level failure mode, provides the clinical decision logic for moving from F31.9 to F31.81 when criteria are met, and explains how Scribing.io eliminates this gap at the point of dictation.

• What Every Other Guide Misses: The Structured Data Gap That Triggers Medication Denials

• Scribing.io Clinical Logic: Resolving the F31.9 → F31.81 PA Denial Scenario

• The Anchor Truth: Understanding the 'Switch' Risk in Payer Logic

• Technical Reference: ICD-10 Documentation Standards for F31.9 and F31.81

• The F31.9 → F31.81 Decision Tree: Step-by-Step Clinical Logic

• FHIR Condition Resource Anatomy: What the ePA Bot Actually Reads

• DSM-5-TR Criteria Documentation: Language That Clears PA

• Five Documentation Failures That Convert Bipolar II to "Unipolar MDD" in Payer Systems

• EHR Workflow Changes: What You Must Do Differently Starting Today

• Switch-Risk Guardrails: How Scribing.io Solves This in Your EHR

• References & Authoritative Sources

What Every Other Guide Misses: The Structured Data Gap That Triggers Medication Denials

Stop reading documentation guides that treat ICD-10 coding as a charting exercise disconnected from medication access. That framing was adequate in 2018. It is clinically dangerous in 2026.

Here is what changed: the majority of electronic prior authorization (ePA) adjudication systems deployed by pharmacy benefit managers—Express Scripts, CVS Caremark, OptumRx, and their downstream processors—do not read your narrative note. They parse structured HL7 FHIR resources. Specifically, they query Condition resources with defined fields for category, clinicalStatus, verificationStatus, code, and onsetDateTime. The ONC Cures Act Final Rule mandated FHIR R4 API availability; payers built their automation on top of it. Your note's prose is invisible to these systems unless its clinical content is mirrored in structured EHR fields.

Scribing.io was built specifically to close this gap between clinical narrative and structured payer-facing data. The platform's NLP engine detects documentable criteria during dictation, maps them to discrete FHIR-compatible fields, and nudges the clinician to post them—turning a documentation exercise into a PA-clearing event. The Scribing.io ICD-10 Documentation Library provides the coding reference layer that supports this workflow for every F31.x subcode.

This creates a specific, high-stakes failure mode for bipolar disorder. A psychiatrist dictates a thorough history noting a five-day hypomanic episode in 2019 meeting DSM-5-TR criteria. The narrative is clinically impeccable. But the episode never reaches the EHR's Problem List as a discrete, structured entry. When cariprazine is prescribed, the PBM's ePA bot queries structured clinical data, finds no Condition resource documenting hypomania, defaults to classifying the presentation as unipolar major depressive disorder, and denies the medication. A step-therapy loop begins. Treatment is delayed by weeks. The patient decompensates.

This is not a hypothetical. A 2025 AMA prior authorization physician survey found that 94% of physicians reported care delays attributable to PA, with psychiatric medications among the most frequently delayed categories. The structured-data gap amplifies this problem specifically for bipolar spectrum diagnoses, where the justifying historical episode (hypomania) may have occurred years before the current prescription.

Scribing.io Clinical Logic: Resolving the F31.9 → F31.81 PA Denial Scenario

The following scenario occurs daily in outpatient psychiatry. It is the canonical example of the 'Switch' Risk—the documentation failure that causes payers to switch your bipolar patient's classification to unipolar depression.

The Problem

A community psychiatrist starts cariprazine for a patient coded F31.9 — Bipolar disorder, unspecified. The note describes a 5-day hypomanic episode in 2019 (elevated mood, decreased need for sleep, increased goal-directed activity), but the episode isn't added to the Problem List. The PBM's ePA bot can't find a structured "history of hypomania," flags the case as unipolar depression, and denies the medication—delaying treatment two weeks and triggering a step-therapy loop requiring trials of generic SSRIs the patient has already failed.

Data from CMS prior authorization reform initiatives confirms that unspecified diagnostic codes carry substantially higher denial rates than their specific counterparts across all payer types. For F31.9 specifically, the absence of structured hypomania documentation transforms a legitimate bipolar diagnosis into an apparent unipolar presentation from the payer bot's perspective.

The Scribing.io Resolution — Step by Step

The clinical consequences of a two-week treatment delay in bipolar disorder are well-documented. Research published in JAMA Psychiatry demonstrates that medication gaps during mood episodes are associated with increased hospitalization risk, suicidal ideation escalation, and functional deterioration. The documentation problem is, at root, a patient safety problem.

The Anchor Truth: Understanding the 'Switch' Risk in Payer Logic

The term "Switch Risk" in payer pharmacology typically refers to the clinical risk of antidepressant-induced mania. But there is a documentation Switch Risk that is equally dangerous and far more common: payers deny coverage for high-tier mood stabilizers unless the note documents the "History of Hypomania" that justifies Bipolar over Unipolar depression.

This is the Anchor Truth of bipolar PA management. Understand it and every downstream documentation decision becomes clear.

Why Payers Enforce This Logic

Payer formulary tiers are built around indication-specific coverage. Cariprazine (Vraylar), for example, has FDA indications for both schizophrenia and bipolar I depression—but payer coverage criteria for the bipolar indication require documentation that the patient has bipolar disorder, not just depression. When the only structured diagnosis is F31.9 with no supporting hypomania documentation, the payer's algorithm cannot distinguish the patient from someone with treatment-resistant unipolar MDD (F33.2), for whom cariprazine is off-label and therefore non-covered under standard formulary rules.

The National Institute for Health Care Management has documented how formulary enforcement increasingly relies on automated adjudication, with human peer-to-peer review reserved for appeals—a process that adds days to weeks of delay.

The Three Requirements the ePA Bot Checks

1. Structured bipolar diagnosis on the Problem List — not just an encounter diagnosis, but a persistent Condition resource with category=problem-list-item.

2. Evidence of mood polarity — a documented history of at least one hypomanic or manic episode, ideally as a separate inactive Condition entry with an onset date. Without this, the bot sees only depression.

3. Code specificity — F31.81 (Bipolar II) or F31.0–F31.7x (Bipolar I subtypes) pass at dramatically higher rates than F31.9 (unspecified). The unspecified code signals insufficient clinical evidence, which triggers additional review or outright denial.

When any of these three requirements is missing, the PA fails. Scribing.io's architecture ensures all three are satisfied at the point of care, before the prescription leaves the EHR.

Technical Reference: ICD-10 Documentation Standards for F31.9 and F31.81

The precise clinical and coding boundaries between F31.9 and F31.81 determine both diagnostic accuracy and payer acceptance. Below is a technical comparison built for the outpatient psychiatrist making this determination at the point of care.

F31.9 — Bipolar Disorder, Unspecified

F31.81 — Bipolar II Disorder

Per CMS ICD-10-CM Official Guidelines for Coding and Reporting, clinicians should code to the highest level of specificity supported by clinical documentation. F31.9 is appropriate only when specificity genuinely cannot be achieved—not as a convenience code.

The F31.9 → F31.81 Decision Tree: Step-by-Step Clinical Logic

This decision tree is designed for use during the clinical encounter. Each node represents a documentation checkpoint.

1. Has the patient ever experienced a hypomanic episode meeting DSM-5-TR duration and symptom count thresholds?

   • Yes, documented with dates and ≥3 Criterion B symptoms → Proceed to Step 2.

   • Suspected but not yet confirmed → Retain F31.9. Document: "Bipolar spectrum suspected; awaiting [collateral/longitudinal observation/mood charting] to confirm hypomanic episode history. Plan: reassess at next visit on [date]."

   • No evidence of hypomania → F31.9 is inappropriate. Evaluate for unipolar MDD (F32.x/F33.x) or other diagnoses.

2. Has the patient ever experienced a major depressive episode?

   • Yes → Proceed to Step 3.

   • No → Consider cyclothymic disorder (F34.0) or hypomanic episode only (code per current episode type).

3. Has the patient EVER experienced a full manic episode? (Defined as ≥7 days, or any duration if hospitalization was required, with marked impairment in functioning.)

   • Yes → Diagnosis is Bipolar I. Code per current episode type (F31.0–F31.7x). Do not use F31.81.

   • No → Proceed to Step 4.

4. Are the episodes better accounted for by schizoaffective disorder, schizophrenia, or another psychotic spectrum diagnosis?

   • Yes → Code accordingly.

   • No → F31.81 (Bipolar II disorder) is the appropriate code. Post to Problem List with onset date. Document the ruling-out of mania explicitly.

Scribing.io automates this decision tree. When dictation content indicates a historical hypomanic episode with sufficient criteria, the system walks through these nodes silently and surfaces the code recommendation only when all conditions are met—preserving clinical autonomy while eliminating the cognitive overhead of payer-aware coding during a 16-minute visit.

FHIR Condition Resource Anatomy: What the ePA Bot Actually Reads

Most psychiatrists have never seen the data structure that determines whether their patient gets medication. Here it is, stripped to the fields that matter for bipolar PA adjudication:

This table explains why clinically excellent notes still produce PA denials. The narrative documents everything. The structured data documents nothing. The bot reads only the structured data. For a deeper understanding of the FHIR standard, see the HL7 FHIR Condition Resource specification.

DSM-5-TR Criteria Documentation: Language That Clears PA

Even when a human reviewer is involved—either at initial PA or on appeal—the speed and outcome of review depends on how quickly they can locate DSM-5-TR criteria language in your note. Vague language forces the reviewer to interpret; explicit criteria language allows them to check boxes.

Hypomanic Episode: Documentation Template

The following language, when present in the psychiatric history section and mirrored in structured Problem List data, satisfies both automated and human PA review:

• Episode identification: "Patient experienced a distinct period of abnormally elevated and expansive mood lasting approximately 5 days in [month/year or year]."

• Duration confirmation: "Episode duration met DSM-5-TR threshold of ≥4 consecutive days."

• Criterion B symptoms (name ≥3 explicitly):

  • "Decreased need for sleep (slept 3–4 hours/night without fatigue)"

  • "Increase in goal-directed activity (started two new business projects simultaneously)"

  • "More talkative than usual / pressured speech (per patient and spouse report)"

  • "Inflated self-esteem (described feeling 'unstoppable')"

• Functional context: "Episode represented an unequivocal change from usual behavior, observable by others. Episode did not result in hospitalization or marked functional impairment requiring emergent intervention, consistent with hypomania rather than mania."

• Ruling out mania: "No lifetime history of manic episode (≥7 days elevated mood with marked impairment or psychotic features). No prior psychiatric hospitalization for mood elevation."

This language pattern maps directly to DSM-5-TR diagnostic criteria published by the American Psychiatric Association. Scribing.io auto-seeds this language structure when its NLP engine detects hypomanic episode content in dictation, filling criterion placeholders with patient-specific details extracted from the ambient capture.

Five Documentation Failures That Convert Bipolar II to "Unipolar MDD" in Payer Systems

Every one of these failures is preventable at the point of care. The challenge is cognitive load: a psychiatrist managing a complex bipolar patient in a 16-minute visit while navigating EHR clicks, safety screenings, and therapeutic alliance cannot also be expected to think about FHIR resource field population. This is precisely why ambient AI scribe technology with structured-data awareness—not just note generation—is a clinical necessity in 2026.

EHR Workflow Changes: What You Must Do Differently Starting Today

If you are not using an ambient scribe with structured-data capabilities, the following manual workflow changes reduce PA denial risk. They are not as efficient as automated solutions, but they are essential.

1. Post historical hypomanic episodes to the Problem List during the encounter in which they are first discussed. Do not defer. Set clinicalStatus to inactive. Set verificationStatus to confirmed. Add at minimum the year of onset.

2. Use F31.81 instead of F31.9 the moment you have confirmed Bipolar II criteria. F31.9 should not persist beyond the initial 2–3 evaluation visits unless you have documented, specific reasons for ongoing diagnostic uncertainty.

3. Name DSM-5-TR Criterion B symptoms explicitly in every note that references the hypomanic episode. Count them. State the duration in days. This is the language PA reviewers—human and automated—are trained to find.

4. Document the absence of mania. "No lifetime history of manic episode" is a single sentence that distinguishes Bipolar II from Bipolar I and prevents inappropriate step-therapy redirection.

5. Before prescribing any Tier 3/4 mood stabilizer, verify that your Problem List contains a confirmed, dated bipolar spectrum entry. This 30-second check prevents the most common denial pathway.

Switch-Risk Guardrails: How Scribing.io Solves This in Your EHR

Everything described in this playbook—hypomania detection, Problem List posting, DSM-5-TR criteria seeding, ICD-10 code optimization, and ePA packet embedding—is automated within Scribing.io's platform. Here is what the system delivers at the point of care:

• Real-time hypomania capture: NLP engine detects hypomanic episode descriptions during ambient listening, extracts duration, symptom count, and onset timing.

• Auto Problem List posting via FHIR: One-click confirmation pushes a fully structured Condition resource to your EHR's Problem List—category=problem-list-item, clinicalStatus=inactive, verificationStatus=confirmed, onsetDateTime populated.

• One-click ePA packet embedding: Structured "History of hypomania" is embedded directly into the ePA submission, so mood stabilizer PAs pass on first submission—not after weeks of appeals and peer-to-peer calls.

• Code specificity nudging: When dictation content supports F31.81 over F31.9, the clinician receives a non-intrusive recommendation with the clinical logic displayed. Acceptance is one click. Override is always available—Scribing.io recommends, the clinician decides.

Book a 15-minute demo to watch these Switch-Risk Guardrails operate in your EHR environment. See real-time hypomania capture, auto Problem List posting via FHIR, and one-click ePA packet embedding with structured "History of hypomania"—so mood stabilizer PAs pass on first submission. Visit Scribing.io Pricing to select the plan that fits your practice.

References & Authoritative Sources

• CMS ICD-10-CM Official Coding Guidelines — Authoritative source for code selection rules, specificity requirements, and unspecified code usage policies.

• AMA Prior Authorization Physician Survey (2025) — Data on care delays, denial rates, and physician burden attributable to PA processes.

• American Psychiatric Association: DSM-5-TR — Diagnostic criteria for Bipolar II Disorder, hypomanic episode, and manic episode.

• HL7 FHIR R4 Condition Resource Specification — Technical standard for structured clinical data exchange, including field definitions for category, clinicalStatus, verificationStatus, and onsetDateTime.

• ONC Cures Act Final Rule — Regulatory mandate for FHIR R4 API availability that enabled payer ePA automation.

• CMS Prior Authorization Reform Initiatives — Federal policy context for PA automation, transparency, and interoperability requirements affecting psychiatric prescribing.

• Scribing.io ICD-10 Documentation Library — Complete reference for psychiatric ICD-10 codes with PA-readiness documentation standards.