Chronic migraine records break at the attack boundary: count days instead.

One attack, two, or three?

The pain lets go around four on Tuesday afternoon. The fog doesn't. By Wednesday morning your neck is stiff, you can't hold a sentence together on a work call, and the overhead light in the kitchen is still too much. Thursday you're yawning at 3 p.m. Thursday night the pain is back.

Your tracking app wants to know how many attacks that was.

One long one? Two, with a bad day wedged between them? Three, if Wednesday's fog was really the front edge of Thursday? You'll pick something because the form won't save without an end time. And next month, more exhausted, you'll probably pick differently.

The seam problem

Call it the seam problem. In high-frequency migraine, the after-phase of one attack and the warning phase of the next can occupy the same hours with the same symptoms: fatigue, neck stiffness, trouble concentrating, light sensitivity. There's no seam to find. Any tool that demands a start and an end makes you invent one.

Invented seams add noise, and worse, they make your count depend on how you felt the night you logged. That's the number a neurologist, and later an insurance reviewer, will read as a fact about your disease.

So stop counting the thing that has no reliable edges. Count days.

Why the attack model breaks above 15 days a month

Most migraine trackers were designed around episodic migraine, where an attack is a clean event: onset, peak, relief, a few clear days, repeat. The diagnostic criteria support that picture. ICHD-3, the International Headache Society's classification, describes an untreated or unsuccessfully treated migraine attack as lasting 4 to 72 hours, and a debilitating attack that runs past 72 hours gets its own label, status migrainosus.

Now apply that to chronic migraine, which ICHD-3 defines as headache on 15 or more days a month for more than three months, with migraine features on at least 8 of those days. At that density the attack model fails in two directions. Some people under-count, logging a nine-day run as one attack that no longer fits the 72-hour window. Others over-count, opening a new entry for every flare, and end up with a record that looks chaotic.

ICHD-3 more or less concedes the point. Its commentary on chronic migraine explains that one reason for separating it from episodic migraine is that individual episodes can't be distinguished when headaches are this frequent or continuous. The classification stopped counting attacks for this population. The apps kept going.

What the clinical record actually runs on

The research behind CGRP preventives was built on days as well. The pivotal trials of the monoclonal antibodies reported their main results as change in monthly migraine or headache days, and the IHS guidelines for controlled trials in chronic migraine (Tassorelli et al., Cephalalgia, 2018) define outcomes on calendar days, with specific criteria for what counts as a qualifying day. Prior-authorization criteria for these drugs tend to use the same unit. Medical Mutual, Cigna, and Mass General Brighton Health Plan CGRP policies require ≥4 migraine days per month as prior authorization criteria ([source](https://www.medmutual.com/-/media/MedMutual/Files/Providers/Prior-Auth-Rx/Calcitonin-GeneRelated-Peptide-CGRP-Antagonist.pdf)).

Postdrome is the phase that makes attack-counting hardest. In an electronic diary study published in Neurology, Giffin and colleagues found that 81% of participants reported at least one non-headache symptom after the pain resolved, with tiredness and difficulty concentrating among the most common. In Giffin et al. 2016, the postdrome lingered for at least one day in 49% of patients. ([source](https://www.researchgate.net/publication/304361542_The_migraine_postdrome_An_electronic_diary_study)).

But several of those same symptoms, fatigue, neck stiffness, poor concentration, also appear in published descriptions of the premonitory phase. Put the two phases back to back and no symptom tells you which side of the line you're on.

How do you count migraine attacks when they run together?

Take an illustrative week, not a real patient's, and log it both ways.

Counted as attacks, this week is one, two or three, depending on where you put the seams and whether Wednesday belongs to Tuesday's postdrome or Thursday's prodrome. Counted as days, it's four migraine days, two symptom-only days, one clear day and three acute-medication days.

Anyone who reads that log gets the same numbers. You, your neurologist, a reviewer who has never met you. That's what makes it defensible.

A logging rule that holds up in a prior-auth file

Give every calendar day one status: migraine day, other headache day, symptom-only day, or clear. Decide it from what actually happened that day.

Record the phases as notes on the day. 'Fog and stiff neck, no pain' is information. 'End of attack three' is a guess.

Log acute medication by day, separately. ICHD-3 flags medication overuse at 10 or more days a month for triptans and 15 or more for simple analgesics, so the med-day count matters as much as the pain-day count.

If you want attack counts too, agree on one rule with your neurologist, such as how long a pain-free gap has to last before a return counts as new, and apply it the same way every month. Consistency beats cleverness here.

A month logged like this answers a reviewer's questions without asking anyone to trust your memory of where the seams were.

Where a continuous timeline fits

Postdrome puts symptoms on one continuous timeline, before, during and after the pain, for exactly this reason. Wednesday's fog sits on the same line as Tuesday's headache and Thursday's yawning, and you never have to decide which attack owns it. The record stays on your device, your data is portable when your neurologist asks for it, and pricing is a one-time lifetime purchase instead of a subscription.

For chronic migraine, the attack was always a convenience of the form. The day is the unit your diagnosis, your trial evidence and your prior-auth file are written in. Log that one.