Our neurology ICD-10 specificity guide

Neurology ICD-10 specificity: stroke, TIA, MS, epilepsy, migraine.

Four diagnosis families carry most of neurology's medical-necessity risk, and every one of them fails the same way: a code that's technically valid but not specific enough for the LCD or payer policy checking against it. This guide is the full crosswalk — every code family verified live against the FY2026 ICD-10-CM code set, organized by the decision logic that actually separates one code from the next, not just a list.

Key takeaways

  • Imaging confirmation, not clinical impression, decides between I63 and G45. A completed infarct with imaging evidence is I63.x; a resolved event without imaging-confirmed infarction is G45.x — and a chart that never states which happened forces the coder to guess.
  • A follow-up visit for a residual deficit is never a repeat I63 code. It's I69.3xx, the sequelae family, regardless of how recently the stroke occurred — billing the acute code on a chronic-management visit is one of the more common specificity errors in stroke follow-up.
  • Plain G35 is a category header, not a billable code. Every MS claim needs a phenotype-and-activity code from the FY2026 replacement set — we verified all eleven current codes live.
  • Named epilepsy syndromes have their own dedicated codes. Juvenile myoclonic epilepsy, Lafora disease, Dravet syndrome, KCNQ2-related epilepsy, and Lennox-Gastaut each carry a specific G40 subcode — defaulting to unspecified epilepsy on a chart that names the syndrome under-codes what the record supports.

Stroke, TIA, and sequelae: the imaging-confirmation rule

Three code families cover what looks, at the bedside, like a single spectrum of cerebrovascular events — but they answer three different clinical questions, and the decision between them turns on one fact: did imaging confirm an infarction, and is today's visit about the acute event or a lasting deficit from a past one.

Decision logic separating I63, G45, and I69.3xx. All confirmed billable against the FY2026 ICD-10-CM code set.
Clinical pictureFamilyWhat has to be documented
Acute event, infarction confirmed on imagingI63.xCT or MRI evidence of infarct; artery and mechanism drive the specific subcode
Acute event, resolved without imaging-confirmed infarctionG45.xSymptom resolution documented; no infarct on imaging, or imaging not obtained but clinical picture is transient
Any later visit addressing a residual deficit from a prior infarctI69.3xxThe specific deficit type — never a repeat acute-event code, regardless of elapsed time

The trap runs in both directions. Coding a resolved, non-infarcted event as I63 overstates what happened and can trigger scrutiny on utilization review; coding a genuine completed infarct as G45 because the note never explicitly says "infarction confirmed on imaging" under-codes a real event and can undercut medical necessity for the treatment that followed. The fix is documentation, not memorization: the radiology report or the physician's note has to state explicitly whether imaging showed an infarct, because that single sentence is what the code family hinges on.

I63.x: cerebral infarction by mechanism and artery

I63 codes are built from two axes stacked together: the mechanism (thrombosis, embolism, or unspecified occlusion/stenosis) and the artery affected (precerebral — vertebral, basilar, carotid — or cerebral — middle, anterior, posterior, cerebellar). A final digit adds laterality (right, left, bilateral, or unspecified) wherever the artery is a paired structure.

I63 structure by mechanism and artery territory. Each cell is a range of codes; laterality (right/left/bilateral/unspecified) is the final digit within each artery-specific code.
MechanismPrecerebral arteries (vertebral, basilar, carotid)Cerebral arteries (MCA, ACA, PCA, cerebellar)
ThrombosisI63.00I63.09I63.30I63.39
EmbolismI63.10I63.19I63.40I63.49
Unspecified occlusion or stenosisI63.20I63.29I63.50I63.59

Example: I63.311 is cerebral infarction due to thrombosis of the right middle cerebral artery — mechanism (thrombosis, the "3" series), artery (middle cerebral), laterality (right). I63.9 (cerebral infarction, unspecified) is valid and billable, but it names neither mechanism nor artery; use it only when the workup genuinely didn't establish more, not as a default when the imaging report is sitting in the chart with the answer already in it. Two codes outside the mechanism-and-artery grid round out the family: I63.6, cerebral infarction due to nonpyogenic cerebral venous thrombosis, and I63.81, infarction due to occlusion or stenosis of a small artery — both frequently under-selected in favor of an unspecified code when the workup actually supports them.

G45.x: transient ischemic attack by syndrome

G45 codes describe a resolved event without imaging-confirmed infarction, and unlike I63, the subcodes are organized by clinical syndrome rather than a mechanism-and-artery grid.

G45 TIA and related syndromes, verified billable against the FY2026 ICD-10-CM code set.
CodeSyndrome
G45.0Vertebrobasilar artery syndrome
G45.1Carotid artery syndrome (hemispheric)
G45.2Multiple and bilateral precerebral artery syndromes
G45.3Amaurosis fugax
G45.4Transient global amnesia
G45.8Other transient cerebral ischemic attacks and related syndromes
G45.9Transient cerebral ischemic attack, unspecified

G45.9 is valid and reasonably common where the vascular territory genuinely wasn't localized, but a chart documenting hemispheric symptoms (contralateral weakness, aphasia) supports G45.1, not G45.9, and a chart documenting vertebrobasilar symptoms (vertigo, diplopia, ataxia) supports G45.0. The syndrome-specific codes exist because payers and specialists both use them to track recurrence risk and localization — coding everything as unspecified erases information the chart already has.

I69.3xx: sequelae of cerebral infarction, by deficit

This is the family that gets skipped most often, because it requires recognizing that a visit is about a lasting effect rather than the event itself. It's organized by deficit type, and several subfamilies add a side-and-dominance axis that requires the chart to document the patient's handedness.

I69.3 sequelae subfamilies, verified billable against the FY2026 ICD-10-CM code set.
SubfamilyCodesCovers
Cognitive deficitsI69.310I69.319Attention/concentration, memory, visuospatial/neglect, psychomotor, frontal-lobe/executive function, social/emotional
Speech and language deficitsI69.320I69.328Aphasia, dysphasia, dysarthria, fluency disorder
Monoplegia, upper limbI69.331I69.339By side and dominance: right dominant, left dominant, right non-dominant, left non-dominant, unspecified
Monoplegia, lower limbI69.341I69.349Same side-and-dominance breakdown as upper limb
Hemiplegia and hemiparesisI69.351I69.359Same side-and-dominance breakdown
Other paralytic syndromeI69.361I69.369Side-and-dominance breakdown, plus a bilateral option (I69.365)
Other sequelaeI69.390I69.398Apraxia, dysphagia, facial weakness, ataxia, other
Unspecified sequelaeI69.30Valid, but under-codes any chart with a named deficit

The dominance axis is the detail generic guides skip entirely: four of the seven subfamilies split into right-dominant, left-dominant, right-non-dominant, and left-non-dominant, plus unspecified. That distinction matters clinically (a dominant-side deficit carries different functional implications than the same deficit non-dominant) and it requires the chart to state handedness explicitly — a note that documents "left hemiparesis" without recording whether the patient is right- or left-handed can only support the unspecified-side code, not the dominance-specific one, even though the anatomical side is clear.

Chronic migraine: G43.7xx and G43.Exx

Migraine specificity runs on three independent axes stacked together: presence of aura, intractability, and status migrainosus. For chronic migraine specifically — 15 or more headache days per month, at least 8 meeting migraine criteria, sustained 3 months — the two governing categories are G43.7xx (without aura) and G43.Exx (with aura), each splitting the same way.

Chronic migraine codes, verified billable against the FY2026 ICD-10-CM code set. Full billing and PREEMPT documentation detail is in our Botox chronic migraine billing guide.
PresentationNot intractableIntractable
Without aura, without status migrainosusG43.709G43.719
Without aura, with status migrainosusG43.701G43.711
With aura, without status migrainosusG43.E09G43.E19
With aura, with status migrainosusG43.E01G43.E11

Beyond chronic migraine, the broader G43 category carries codes that get miscoded into the chronic family when they shouldn't be: G43.4 hemiplegic migraine, G43.5/G43.6 persistent migraine aura without/with cerebral infarction, and G43.A cyclical vomiting in migraine each have their own subcode families with the same intractable/status-migrainosus structure, and none of them are "chronic migraine" even when the patient also happens to meet the 15-day frequency threshold. Code the specific migraine subtype the chart documents, not the frequency pattern alone.

Epilepsy: G40.x by type, syndrome, and severity

G40 is neurology's largest and most granular diagnosis family, and its size is exactly why it gets under-coded — it's easier to reach for unspecified epilepsy than to navigate 90-plus subcodes. Every subcategory below splits further by intractability and status epilepticus; the table shows the category-defining codes, not every combination.

G40 major categories and named syndromes, verified billable against the FY2026 ICD-10-CM code set.
CategoryCode prefixNotes
Localization-related idiopathic epilepsyG40.0xFocal onset, idiopathic
Localization-related symptomatic, simple partial seizuresG40.1xFocal, symptomatic, no impairment of consciousness
Localization-related symptomatic, complex partial seizuresG40.2xFocal, symptomatic, with impairment of consciousness
Generalized idiopathic epilepsyG40.3x
Other generalized epilepsyG40.4xIncludes G40.42, CDKL5 deficiency disorder
Epileptic seizures related to external causesG40.5x
Other epilepsy, including named syndromesG40.8xLennox-Gastaut (G40.81x), epileptic spasms (G40.82x), Dravet syndrome (G40.83x), KCNQ2-related epilepsy (G40.84x)
Unspecified epilepsyG40.9xValid, but the least informative code in the family — see FAQ
Absence epileptic syndromeG40.Ax
Juvenile myoclonic epilepsyG40.BxImpulsive petit mal
Lafora progressive myoclonus epilepsyG40.Cx

The pattern to watch for on chart review: a note that names a specific syndrome — "consistent with juvenile myoclonic epilepsy" or "Dravet syndrome, genetically confirmed" — but the claim carries G40.909 anyway, usually because the EHR's problem-list mapping was never updated after the diagnosis was refined. That gap doesn't cause a denial by itself, since G40.909 is valid, but it under-documents medical necessity for the AED regimen, EEG frequency, or specialist referral the specific syndrome actually justifies.

Multiple sclerosis: the full FY2026 G35 replacement set

Plain G35 is a category header only as of the FY2026 restructuring — it is not valid for HIPAA transactions and will reject a claim before medical necessity is ever evaluated. The replacement set is built from phenotype (relapsing-remitting, primary progressive, secondary progressive, or unspecified) crossed with disease activity where applicable.

The complete billable G35 code set, verified live against the FY2026 ICD-10-CM code set. Plain G35 (header) is not on this list because it is not billable.
CodePhenotype and activity
G35.ARelapsing-remitting multiple sclerosis
G35.B0Primary progressive multiple sclerosis, unspecified activity
G35.B1Active primary progressive multiple sclerosis
G35.B2Non-active primary progressive multiple sclerosis
G35.C0Secondary progressive multiple sclerosis, unspecified activity
G35.C1Active secondary progressive multiple sclerosis
G35.C2Non-active secondary progressive multiple sclerosis
G35.DMultiple sclerosis, unspecified (phenotype not established — still billable, unlike plain G35)

G35.D is the code to reach for when the phenotype genuinely isn't determined yet, not plain G35 — it's the closest equivalent to the old unspecified code and it passes HIPAA validation. The activity designation (active/non-active) on the progressive phenotypes tracks whether the patient has had a relapse or new MRI activity within a defined recent period; a neurologist's note stating "no new relapses or MRI activity in the past 12 months" supports the non-active code, while silence on activity status defaults the coder to the unspecified-activity variant, not an assumption in either direction.

Do
  • Run get_hierarchy-style code lookups (or your encoder's equivalent) on any G35, G40, G43, I63, or I69.3 chart before defaulting to the unspecified variant.
  • Audit EHR problem-list mappings for plain G35 specifically — it's the one code in this guide that rejects rather than just under-codes.
  • Document handedness explicitly on any stroke sequelae chart where a dominance-specific I69.3 code could apply.
Don't
  • Don't bill a repeat I63 code on a chronic stroke follow-up visit — move to I69.3xx once the acute event is over.
  • Don't let a named epilepsy syndrome sit undocumented on a claim as G40.909 just because the problem list was never updated.
  • Don't guess at TIA localization (G45.0 vs. G45.1) when the note already describes the symptom pattern that determines it.

Worried your neurology diagnosis coding is under-specified?

We'll run a sample chart audit against the current FY2026 code set, flag every under-coded diagnosis, and show what it's costing in denied or unsupported claims.

Book a free claims review

Frequently asked questions

What ICD-10 code do we use for a follow-up visit addressing residual weakness two years after a stroke?

An I69.3xx sequelae code, specifically the hemiplegia/hemiparesis subfamily (I69.351–I69.359), not a repeat I63 acute-infarction code. I63 codes describe the acute cerebral infarction itself, confirmed by imaging at the time of the event; once that event is over and the visit is about a residual deficit, the encounter has moved into sequelae coding regardless of how much time has passed. The I69.35x codes further specify which side is affected and whether it's the patient's dominant or non-dominant side — a distinction that requires documented handedness in the chart, not an assumption.

Is G40.909 ever the right code, or should we always code more specifically?

G40.909 is valid for HIPAA transactions and is the right code when the chart genuinely doesn't support more — a new patient with a single unprovoked seizure and no established syndrome, for instance. It's the wrong code when the chart documents a named epilepsy syndrome (juvenile myoclonic epilepsy, Lafora disease, Dravet syndrome, KCNQ2-related epilepsy, Lennox-Gastaut) or an established intractability and status-epilepticus picture, because G40.909 doesn't capture any of that and under-documents a patient whose record supports far more specificity. Repeat EEG, video-EEG monitoring, and AED-management services frequently lean on that specificity for medical necessity, so a lazy default to G40.909 on a well-documented chart is coding below what the record supports.

Our EHR problem list still shows plain G35 — what happens if we bill it?

The claim rejects, not denies — plain G35 is a category header only as of the FY2026 restructuring and is not valid for HIPAA transactions, so it fails validation before medical necessity is ever evaluated. Replace it with a phenotype-and-activity code: G35.A for relapsing-remitting, G35.B0/B1/B2 for primary progressive by activity status, G35.C0/C1/C2 for secondary progressive by activity status, or G35.D when the phenotype genuinely isn't established yet. Audit every problem-list mapping and superbill template that references G35 alone before the next MS visit goes out, because every claim carrying it will fail the same way.

Verify before billing. CPT is a registered trademark of the American Medical Association; codes here are paraphrased, not reproduced from the CPT Professional edition. CPT, HCPCS and ICD-10 codes, coverage policy, and bundling edits change, including annual code-set updates. This page reflects standard industry practice and is provided for general education — it is not a substitute for your own compliance review, your current payer contracts, or the current-year code sets. Confirm requirements against your specific payer mix before submitting claims.

Related resources