Headache differential diagnosis: red flags and causes

The differential diagnosis of headache is built on one decision: is this a primary headache disorder — migraine, tension-type headache or a trigeminal autonomic cephalalgia — or a secondary headache caused by something else, from subarachnoid haemorrhage to giant cell arteritis? Most headache presenting to primary care is primary and benign. The clinical task is not to diagnose every headache perfectly at first contact, but to systematically exclude the small proportion that will kill or disable the patient if missed, then classify the rest well enough to treat.

This is a working reference for clinicians: the red-flag framework that structures the search for secondary causes, the base rates that tell you how much to worry, the features that separate migraine from tension-type headache, when imaging is and is not indicated, and the features that — despite intuition — do not help at all.

Key takeaways

• Headache disorders affect roughly 40% of the world's population — about 3.1 billion people in 2021 (WHO).
• Around 18% of people presenting with headache have a secondary headache disorder; separately, age-stratified data in the same review put secondary headache at 11.2% in those aged 65 and over versus 8.0% in younger adults — a different study population, not a subset of the 18% (Practical Neurology).
• The SNNOOP10 list of 15 red and orange flags is the standard structured screen for secondary headache (Do et al., Neurology 2019).
• Thunderclap onset carries a pretest probability of serious intracranial pathology greater than 40% — it is a same-day imaging problem, not a clinic problem (AAFP 2022).
• Severity and response to treatment do not distinguish primary from secondary headache (NHS Scotland headache pathway).
• A stable primary-headache pattern with no red flags and a normal neurological examination does not need neuroimaging.

What is the differential diagnosis of headache?

The differential diagnosis of headache is the ranked list of conditions that could explain the presentation, organised as primary headache disorders versus secondary headaches attributable to another cause. The formal taxonomy is the International Classification of Headache Disorders (ICHD-3), published by the International Headache Society, but at the bedside the primary-versus-secondary split does the clinical work.

In practice the list you work through looks like this. Among primary disorders: migraine with and without aura, tension-type headache, and the trigeminal autonomic cephalalgias, of which cluster headache is the archetype. Among secondary causes, grouped by the system that produces them:

• Vascular — subarachnoid haemorrhage, intracerebral haemorrhage, reversible cerebral vasoconstriction syndrome, cervical artery dissection, cerebral venous sinus thrombosis, ischaemic stroke, hypertensive crisis, posterior reversible encephalopathy syndrome.
• Infective — meningitis, encephalitis, intracranial abscess, and systemic infection with headache as a symptom.
• Raised or lowered intracranial pressure — space-occupying lesion, idiopathic intracranial hypertension, hydrocephalus; and at the other extreme spontaneous intracranial hypotension.
• Inflammatory — giant cell arteritis above all, plus other vasculitides and inflammatory disease.
• Drug-related — medication-overuse headache, and headache beginning with a newly started drug.
• Traumatic — acute and persistent post-traumatic headache, subdural haematoma.
• Regional and referred — acute angle-closure glaucoma, pituitary apoplexy, temporomandibular and cervicogenic pain, sinus disease.

For the general method behind building and narrowing a list like this — hypothesis generation, pertinent negatives, and the reasoning errors that derail it — see our pillar guide, what is a differential diagnosis?

What are the red flags for secondary headache?

The red flags are the features that raise the probability of a secondary cause enough to change what you do next. The reference framework is SNNOOP10 — 15 red and orange flags published by Do and colleagues in Neurology in 2019, expanding the older SNOOP mnemonic. The authors recommend screening every new headache patient against the list.

• S — Systemic symptoms: fever, chills, sweats, myalgia, weight loss. Suggests infection, malignancy, or inflammatory disease including giant cell arteritis. Concurrent fever and headache have relatively high sensitivity for neurological infection — but an immunocompromised patient may not mount a fever at all.
• N — Neoplasm in the history: consider metastatic disease.
• N — Neurological deficit or dysfunction: focal signs, altered behaviour, seizures, reduced consciousness. Neurological findings carry positive likelihood ratios for a secondary headache disorder of 3.56 to 16.2. Diplopia, transient visual loss or pulsatile tinnitus point towards idiopathic intracranial hypertension.
• O — Onset that is sudden or abrupt: thunderclap headache, reaching maximum intensity in under a minute and lasting five minutes or more.
• O — Older age at onset: after 50 (the Do list uses 65 as the high-risk threshold) — consider neoplasm, giant cell arteritis, inflammation and infection. Onset before age 5 raises concern for brain tumour.
• P — Pattern change, or a genuinely new headache: a new aura, loss of headache-free days, or episodic headache becoming chronic daily headache.
• P — Positional headache: worse lying flat suggests raised intracranial pressure; onset within seconds of standing, relieved by lying down, suggests spontaneous intracranial hypotension (incidence around 5 per 100,000).
• P — Precipitated by Valsalva, cough, sneeze or exertion: cough headache needs imaging, the commonest cause being Chiari malformation type 1, with posterior fossa lesions and tumours in the differential.
• P — Papilloedema: always warrants further workup. Note that normal imaging does not exclude raised intracranial pressure.
• P — Progressive or atypical headache; Pregnancy or puerperium; Painful eye with autonomic features; Post-traumatic onset; Pathology of the immune system such as HIV; Painkiller overuse or a new drug at headache onset.

A separate point worth internalising: red flags are sensitive but not specific. Do and colleagues are explicit that much of the evidence comes from clinical experience and large case series, that prospective epidemiological data on red flags are lacking, and that even with red-flag-driven imaging most scans return no findings. A flag tells you to take more history, examine further or investigate — it is not itself a diagnosis.

How common are serious causes of headache?

Base rates discipline the worry. Roughly 18% of people experiencing headache have a secondary disorder, and most of those secondary headaches are benign in themselves — medication-overuse headache is the commonest secondary headache worldwide and may affect up to 5% of some populations. The dangerous causes are rarer still, and the numbers below are the ones to carry:

• Significant intracranial pathology in about 1% of all headache patients imaged in an outpatient neurology clinic.
• About 0.2% in patients with migraine and a normal neurological examination (meta-analysis).
• Brain tumour risk under 0.1% in patients with no prior history of neoplasm.
• But greater than 40% pretest probability of serious intracranial pathology once the headache is thunderclap in onset.

Those four figures, all from the 2022 American Family Physician review, explain the whole shape of headache practice: investigate almost nobody with a stable pattern and a normal examination, and investigate immediately when onset was explosive.

How do you tell migraine from tension-type headache?

Tension-type headache is the most common primary headache and accounts for 46–78% of headache diagnoses; migraine has a one-year prevalence of about 12% (17% of women, 5.6% of men) and is the type more likely to bring someone to the clinic because it is more disabling. Trigeminal autonomic cephalalgias account for under 1%, with cluster headache at a lifetime prevalence of about 0.1%.

Two validated bedside aids help. The POUND mnemonic counts five features — Pulsating quality, duration of 4–72 h, Unilateral location, Nausea, and Disabling intensity; four or more make migraine the most likely diagnosis. The ID Migraine screen asks three questions covering light sensitivity, nausea or vomiting, and activity-limiting headache in the past three months, and is positive with two or more.

Migraine is therefore recognised positively, not by exclusion: episodic, unilateral, pulsatile, disabling attacks with nausea or photophobia, often with a family history, in a patient whose pattern is stable and whose examination is normal. Tension-type headache is the bilateral, pressing, non-disabling, non-pulsatile counterpart. Cluster headache is strictly unilateral, excruciating, short (minutes to a few hours), with ipsilateral autonomic features — lacrimation, conjunctival injection, ptosis, rhinorrhoea — and striking restlessness.

Which headaches need imaging, and which do not?

Imaging decisions follow the red flags, not the pain score. Where haemorrhage or mass effect is suspected, non-contrast head CT is the emergency investigation; for acute thunderclap headache the AAFP review advises CT within 12 hours of onset. Lumbar puncture to exclude subarachnoid haemorrhage comes after a reassuring CT, not instead of it. For non-urgent presentations that still carry concerning features, brain MRI with and without contrast is the preferred modality — specifically to exclude a secondary cause behind a suspected trigeminal autonomic cephalalgia, and to exclude a posterior fossa lesion in cough-precipitated headache.

Equally important is what not to do. Patients with a stable primary headache pattern, no red flags and a normal neurological examination do not need neuroimaging — and Choosing Wisely recommendations from the American College of Radiology, the American Headache Society and the American Academy of Neurology advise against imaging uncomplicated or stable migraine-pattern headache, and against EEG in headache evaluation. EEG is not useful in headache work-up other than for atypical migraine aura. Kernig and Brudzinski signs are of limited value, so their absence should not reassure you in suspected meningitis.

In a new headache over 50, check inflammatory markers — CRP, with ESR or plasma viscosity and a full blood count per local protocol — and if giant cell arteritis is a serious consideration, refer urgently and consider starting steroids under your local pathway rather than waiting for confirmation.

Which headache features are misleading?

Severity and response to treatment do not help you differentiate primary from secondary headache. A migraine can be the worst pain of someone's life; a subarachnoid haemorrhage can settle temporarily with simple analgesia. The NHS Scotland pathway names both explicitly as non-discriminating, and it also flags that "amber" features — the orange flags of SNNOOP10 — occur in primary headache too, so they prompt consideration of a secondary cause rather than confirming one.

Two more traps. First, a normal scan does not exclude raised intracranial pressure: examine the fundi, and refer same-day, usually to ophthalmology, if you suspect it. Second, most headaches presenting over the age of 50 are still benign — age raises the prior, it does not settle the question.

A worked example (illustrative)

A 54-year-old presents with bitemporal headache that has built up over four months and has been daily for the past three weeks, worse on waking, and newly different from the episodic headaches she has had for years. She has taken codeine-containing analgesia on most days across those four months. Examination is normal; fundi are difficult to visualise.

Flags present: pattern change, new headache over 50, painkiller overuse, and possible positional aggravation. Working differential, ranked: medication-overuse headache superimposed on her known primary headache; giant cell arteritis; raised intracranial pressure from a space-occupying lesion or idiopathic intracranial hypertension. Next steps follow directly — a proper fundoscopic view (dilated or with imaging of the optic disc), CRP and ESR with a full blood count, a careful temporal-artery and jaw-claudication history, and brain MRI rather than CT given the subacute course. The medication-overuse diagnosis is the most probable, but it is the one you confirm last, after the serious alternatives are excluded.

Note the duration test, because it is easy to get wrong. The ICHD-3 criteria for medication-overuse headache require headache on 15 or more days a month in a patient with a pre-existing headache disorder, plus regular overuse for more than three months — at 10 or more days a month for ergotamine, triptans, opioids or combination analgesics, and 15 or more for simple analgesics such as paracetamol or an NSAID (criteria as reproduced in Moavero et al., Frontiers in Neurology 2020). Four months of near-daily codeine-containing analgesia clears that bar; a fortnight of it would not, however plausible the mechanism looks. In the shorter case the right label is emerging analgesic overuse to intervene on — not a diagnosis that lets you stop looking for the serious alternatives.

Where does clinical AI fit into headache assessment?

Headache is a good test of decision support precisely because the reasoning is checklist-shaped: 15 flags, a handful of base rates, and a short list of time-critical diagnoses that are easy to skip when the clinic is running 40 minutes behind. Shifaa AI — an AI clinical assistant currently in private beta — is built for that gap: its differential-diagnosis support surfaces a ranked differential, raises the red-flag presentations ahead of the likely ones, and shows the guidance behind each suggestion so you can check it. It is decision support. It does not diagnose; the clinician reviews, edits and signs, as we argue in can AI diagnose patients?

The documentation side matters too, because a headache consultation is only as good as the record of it — onset, pattern change, pertinent negatives and the flags you actively excluded. Our guide to writing a SOAP note covers the structure, and the free printable SOAP note template gives you a layout that prompts for exactly those fields. For the same reasoning applied to the other high-stakes symptom, see chest pain differential diagnosis.

Frequently asked questions

What are the red flags for a headache?

The SNNOOP10 list: systemic symptoms, neoplasm history, neurological deficit, sudden (thunderclap) onset, older age at onset, pattern change, positional headache, precipitation by Valsalva or exertion, papilloedema, progressive or atypical course, pregnancy, painful eye with autonomic features, post-traumatic onset, immune pathology, and painkiller overuse or a new drug.

What percentage of headaches are secondary?

About 18% of people presenting with headache have a secondary headache disorder. Age-stratified data reported separately in the same review put secondary headache at 11.2% in those aged 65 and over versus 8.0% in younger adults — a different study population, so the two sets of figures are not directly comparable. Most secondary headaches are benign, medication-overuse headache being the commonest worldwide.

When does a headache need a brain scan?

When red flags or abnormal neurological findings are present. Non-contrast CT is the emergency choice for suspected haemorrhage or mass effect — within 12 hours of onset for thunderclap headache — followed by lumbar puncture if the CT is normal. MRI is preferred for non-urgent presentations with concerning features. Stable primary headache with a normal examination needs no imaging.

How do you distinguish migraine from a dangerous headache?

Not by severity or by response to analgesia — neither differentiates primary from secondary headache. Migraine is recognised positively by a stable pattern of episodic, unilateral, pulsatile, disabling attacks with nausea or photophobia and a normal examination. A dangerous headache is suggested by abrupt onset, new deficits, systemic features, or a change in an established pattern.

References

• World Health Organization. Headache disorders — fact sheet (prevalence, burden, under-diagnosis). who.int/news-room/fact-sheets/detail/headache-disorders
• Viera AJ, Antono B. Acute Headache in Adults: A Diagnostic Approach. American Family Physician, 2022. aafp.org/pubs/afp/issues/2022/0900/acute-headache-adults.html
• Do TP, Remmers A, Schytz HW, et al. Red and orange flags for secondary headaches in clinical practice: SNNOOP10 list. Neurology, 2019. research.birmingham.ac.uk
• Diagnosing Secondary Headaches (SNNOOP10 review, with secondary-headache frequency data). Practical Neurology, 2020. practicalneurology.com
• NHS Scotland / Right Decisions. National headache pathway — red, amber and green flags. rightdecisions.scot.nhs.uk
• Moavero R, Stornelli M, Papetti L, et al. Medication Overuse Withdrawal in Children and Adolescents Does Not Always Improve Headache. Frontiers in Neurology, 2020;11:823 — cited here for its reproduction of the ICHD-3 diagnostic criteria for medication-overuse headache. frontiersin.org