What counts as a headache day?

Any calendar day on which you had head pain, regardless of whether it was migrainous. A tension-type-like ache counts toward the 15. So does a day where you took a triptan at onset and the pain was gone in an hour. Only the days meeting migraine criteria (or responding to migraine-specific treatment) count toward the separate 8-day requirement.

Does a migraine that lasts three days count as one attack or three headache days?

Three headache days. The criteria are counted per calendar day, not per attack. This is the single biggest source of undercounting in event-based trackers, which record the attack once and move on. An attack that begins late at night and resolves the next morning takes two days off the calendar.

How long do I need to track before a headache diary is useful?

Three months of day-level records is the minimum that maps onto the ICHD-3 duration criterion, and it's roughly what a prior-auth reviewer expects to see. Start before you need it. Counts reconstructed from memory run low and read as reconstructed, which is the weakest position to be in during an appeal.

The quiet slide from episodic to chronic migraine, and why your tracker misses it

Article · 5 min read

Chronic migraine is fifteen days a month. Your app counts attacks.

Q: When does migraine become chronic?

Under ICHD-3, chronic migraine means headache on 15 or more days per month for more than three months, with at least 8 of those days meeting criteria for migraine with or without aura, or responding to migraine-specific treatment. All three parts have to hold together: the 15-day count, the 8-day migraine subset, and the duration beyond three months.

Q: Do I need a chronic migraine diagnosis to get a CGRP medication?

Not from the FDA labels, which approve CGRP monoclonal antibodies for migraine prevention without a headache-frequency floor. Your plan's prior-authorization policy is a separate document, and those commonly ask for a documented baseline of monthly headache days plus documented failures of prior preventive classes. OnabotulinumtoxinA is the different case: its label is specific to chronic migraine, 15 or more headache days a month lasting 4 hours or longer.

The ICHD-3 threshold for chronic migraine is measured in headache days, not attacks. Event-based trackers count the wrong unit, and patients arrive at the neurologist months behind the line.

Nine attacks, twenty-two days

The neurologist asks one question, and it's rarely the one you rehearsed on the drive over: how many days last month did you have any head pain at all.

Not attacks. Days.

You open the app you've kept faithfully for fourteen months and scroll back to August. Nine entries. You say nine, because nine is what's on the screen, and nine goes into the chart, and nine reads as episodic.

Except August had the Tuesday one that didn't let go until Thursday afternoon. The weekend one that took both days. The stretch after the flight where the pain never fully cleared, it just dropped to a level you'd stopped bothering to log. Count calendar squares instead of entries and August was twenty-two days with head pain in it.

The app never added them up. It was never counting that.

The headache-day gap

Call it the headache-day gap: the distance between how many attacks you log and how many calendar days you actually spend in pain.

For people whose attacks resolve inside an afternoon, that gap sits near zero. For everyone else it compounds fast. Untreated migraine runs 4 to 72 hours by definition, so one attack routinely covers two or three dates, and an attack that starts at 11pm on Sunday has already taken Monday.

Chronification is not a diagnosis moment, it's a threshold crossing, and the threshold is written in days. That's the argument. An app that opens a record at pain onset, closes it at pain resolution, and hands you a monthly attack count can't produce the number the criteria are defined in. It's measuring something else entirely, and it's measuring it well, which is why the gap goes unnoticed for so long.

Nobody feels the line being crossed

Ask people how they'd expect to know, and most describe an event: a worsening they'd feel arriving, or a doctor saying the word. Neither tends to happen.

A 14-day month and a 15-day month feel identical from the inside. The slide usually plays out over quarters, through small additions (one more low-grade day a week, a couple of extra triptan days, a run of broken sleep that never quite corrects) that nobody experiences as a transition. In the AMPP population study, roughly 2.5% of people with episodic migraine progressed to chronic migraine in a given year. That's a slow risk spread across a very large group, and it arrives with no announcement attached.

So the crossing is only ever visible in retrospect, in data, to someone who counted. And if nobody counted in days, it stays invisible until a person sits down and tries to rebuild a year out of memory.

What the criteria actually count

ICHD-3 defines chronic migraine (code 1.3) as headache on 15 or more days per month for more than three months, with at least 8 of those days meeting criteria for migraine with or without aura, or responding to migraine-specific treatment.

Four details in that definition do most of the damage to an event log.

A headache day counts even when the headache wasn't migrainous. Tension-type-like days count toward the 15. They just don't count toward the 8.

A day counts even when you treated it successfully. Triptan at onset, pain gone in forty minutes, the day still counts.

The pattern has to hold for more than three months. One brutal month isn't chronification, and one good month doesn't undo it.

Acute medication has its own separate thresholds: 10 or more days a month of triptans, ergots, opioids or combination analgesics, or 15 or more days of simple analgesics, meets the medication-overuse definition. That sits on top of the day count rather than replacing it.

≥15

headache days per month, of any headache typeICHD-3, code 1.3

≥8

of those days meeting migraine criteria or responding to migraine-specific treatmentICHD-3, code 1.3

consecutive months the pattern must holdICHD-3, code 1.3

~2.5%

of episodic migraine cases progressing to chronic per yearAMPP study, Bigal et al. 2008

Event-based logging against ICHD-3 day counting. Same month, two different numbers.
What happened	What an event log records	What the criteria count
One attack, Tuesday evening through Thursday afternoon	1 attack	3 headache days
Attack starting 11pm Sunday, resolved 2am Monday	1 attack	2 headache days
Triptan at onset, pain gone within the hour	Often not logged at all	1 headache day
Low-grade ache, no photophobia, no nausea	Usually not logged (doesn't feel like a migraine)	1 headache day toward the 15, not toward the 8
Three days of fog and light sensitivity, no head pain	Not logged	0 headache days, though it belongs in the impairment record

The 15-day line is also a key

That line does more than name the condition. It gates what's available to you.

OnabotulinumtoxinA has been FDA-approved since 2010 for headache prophylaxis in adults with chronic migraine, and the label defines that population explicitly: 15 or more headache days a month, each lasting 4 hours or longer. Below the line it isn't an indicated option. CGRP monoclonal antibodies are approved for migraine prevention with no frequency floor, but an FDA label and a plan's utilization-management policy are two separate documents. Commercial prior-auth criteria commonly ask for a documented baseline of monthly headache days alongside documented failures of prior preventive classes, and payers vary in exactly how much diary history they want before they'll sign off. The specifics move often enough that the only safe move is to pull your own plan's current medical policy before you assume anything about what counts as enough.

The appeal is where the gap gets expensive. A denial citing inadequate documented impairment is asking for a number you either have or you don't.

A reconstructed count is not a diary

Recall of headache days drifts low, and a reviewer can usually tell the difference between a contemporaneous record and a calendar rebuilt from memory the week the appeal is due. Start the day count before you need it, not after the denial letter arrives.

What it takes to see the slide

A tracker that can surface chronification has to do five unglamorous things.

Count days rather than events, and put a rolling three-month headache-day figure somewhere you'll actually see it.

Let a single attack span midnight and multiple dates without collapsing into one dated entry.

Record aborted attacks. The triptan day is the one most likely to go unlogged, and it counts in full.

Accept the low-grade, non-migraine headache day, because 15 and 8 are two separate counts and most logs only ever capture the second.

Export the record in a form a neurologist or an appeals reviewer reads as a diary, with dates, not as a screenshot of a summary screen.

We built Postdrome around a continuous symptom timeline instead of an event that terminates at pain resolution, originally because the 24 to 72 hours after the headache were the part every tracker we tried threw away. The day count fell out of the same decision. When the record doesn't end where the pain does, the calendar arithmetic is already sitting there when somebody finally asks for it.

What we're watching next

Two open questions.

Reverse transition is the first. People do move from chronic back to episodic, and that crossing is documented even worse than the first one, because once chronic sits in the chart nobody goes back and recounts.

The second is whether payers start accepting patient-held, app-exported headache diaries as primary documentation, instead of asking a clinician to attest to a number the patient recited from memory in a fifteen-minute appointment.

Neither is settled. This part is: the threshold is 15 days, the unit is the day, and nobody is going to do the counting for you. If you track one thing this month, track days.

Start counting in days. Postdrome keeps the full timeline, exports it date by date, and costs one payment, not a subscription.

Postdrome records a continuous symptom timeline rather than a discrete pain event, so an attack running Tuesday night to Thursday afternoon appears in the record as three headache days, not one entry. Aborted attacks, low-grade non-migraine days, and acute medication days each get their own count, and the export is date-level, which is the form a neurologist or an appeals reviewer can read. Everything stays on device, the data is yours to take elsewhere, and the price is paid once. We built it because every tracker we tried for the people in our lives stopped recording the moment the pain stopped.

Get Postdrome