The British Journal of Radiology, 76 (2003), 532–535 E 2003 The British Institute of Radiology

DOI: 10.1259/bjr/89012738

Red flags in patients presenting with headache: clinical

indications for neuroimaging
1
SOBRI M, MRad(USM), 2A C LAMONT, FRCR, 1N A ALIAS, MRad(UKM) and 3M N WIN, FRCP
1
Department of Radiology, University Putra Malaysia-Hospital Kuala Lumpur, 2Department of Radiology, Mater Childrens
Hospital, Brisbane, Australia and 3Department of Medicine, International Medical University, Malaysia

Abstract. Headache is a very common patient complaint but secondary causes for headache are unusual.
Neuroimaging is both expensive and has a low yield in this group. Most patients with intracranial pathology
have clinical features that would raise a ‘‘red flag’’. Appropriate selection of patients with headache for
neuroimaging to look for secondary causes is very important. Red flags act as screening tools to help in
identifying those patients presenting with headache who would benefit from prompt neuroimaging, and may
increase the yield. The aim of this study is to evaluate clinical features in patients with headache using
neuroimaging as a screening tool for intracranial pathology. 20 red flags were defined. A retrospective study of
111 patients was performed and the outcomes were divided into positive and negative. Abnormal neuroimaging
was present in 39 patients. Results were analysed using the Logistic Regression model. Sensitivity and specificity
of red flags were analysed to establish the cut-off point to predict abnormal neuroimaging and a receiver
operating characteristic (ROC) curve plotted to show the sensitivity of the diagnostic test. Three red flag
features proved to be statistically significant with the p-value of less than 0.05 on both univariate and
multivariate analysis. These were: paralysis; papilloedema; and ‘‘drowsiness, confusion, memory impairment
and loss of consciousness’’. In addition, if three or more red flags from the list were present, this showed strong
indication of abnormal neuroimaging, from cut-off point of ROC curve (area under the curve 50.76).

Headache is one of the most common medical
complaints. Perkins noted that about 10% of patients
with headaches seen in the emergency departments were
due to secondary causes, which may sometimes be over-
looked [1]. While primary headache is more common,
there is a high level of anxiety amongst both patients and
physicians that a particular headache may be due to a
secondary cause. The incidental discovery of a significant
lesion associated with symptoms thought to be a primary
headache is not uncommon; aneurysm occurs in approxi-
mately 2–5%, brain tumour 1% and arteriovenous malfor-
mation 0.8% [2].
Neuroimaging with CT and MRI are the most important
investigations for headache. However, these investigations
are expensive, are not without risk to the patient and
medicolegal concerns also influence decision-making
in some communities. We believe it to be important to
establish a set of appropriate selection criteria for con-
sideration as part of the investigation work-up before
neuroimaging is performed.
‘‘Red flag’’ features are those signs or symptoms that
may indicate headache with a serious cause [3]. In the
diagnosis of headache most patients with organic or
vascular disease that benefit from neuroimaging have been
shown to demonstrate clinical features that would raise a
red flag [4–6].
Although there have been some studies to identify
criteria that increase the yield of neuroimaging, no specific
studies have been carried out to discover which red flag
Received 5 August 2002 and in revised form 28 April 2003, accepted 9
May 2003.
Current address for Sobri M, Department of Radiology, UPM-HKL,
Jalan Masjid, 50586 Kuala Lumpur, Malaysia.
features would identify patients with headache who would
benefit from neuroimaging.
This study aims to establish a clinical screening tool for
patients who present with headache. We hope to show that
identification of red flags can help select patients for
neuroimaging, improve positive pick-up rates, avoid delay
in diagnosis and treatment, and in the long run may help
conserve scarce health funds.
Methods and materials
A retrospective study of all patients identified on the
Department of Radiology computerized database who
presented at Hospital Universiti Sains Malaysia, Kelantan
with headache and who underwent neuroimaging during
the 12-month period from January to December 1999 was
undertaken. We identified patients by choosing those with
‘‘headache’’ as the main indication on the neuroimaging
request. Patients with known underlying central nervous
system disorder were excluded. Medical records were
reviewed to identify patients with one or more of 20
baseline variables called ‘‘red flags’’. The list of red flags in
diagnosis of headache used in this study is adapted from
Cleveland Clinic Headache Centre (1998) (Table 1).
All patients received one or more investigation with either
CT, MRI or cerebral angiography. Axial 5 mm/10 mm
section CT Scan was performed using Siemens. Somatom
HiQ-s (Erlangen, Germany). MRI was performed using 1.0
Tesla Signa Horizon Lx General Electronic (Milwaukee,
WI). Axial and sagittal T1 weighted spin-echo, axial T2
weighted fast spin-echo, axial fluid attenuated inversion
recovery (FLAIR) and MR angiography (MRA) were
routine and gadolinium given when indicated. Catheter
angiography was performed using Sigma 16 inch General

532 The British Journal of Radiology, August 2003

Red flags in patients with headache

Table 1. List of red flags and their frequencies

Red flag Frequency Percentage

Onset of new or different headache 64 57.7
Nausea or vomiting 33 29.7
Worst headache ever experience 32 28.8
Progressive visual or neurological changes 20 18.0
Paralysis 15 13.5
Weakness, ataxia or loss of co-ordination 14 12.6
Drowsiness, confusion, memory impairment or loss of consciousness 13 11.7
Onset of headache after age of 50 years 12 10.8
Papilloedema 10 9.0
Stiff neck 6 5.4
Onset of headache with exertion, sexual activity or coughing 6 5.4
Systemic illness 5 4.5
Numbness 4 3.6
Asymmetry of pupillary response 2 1.8
Sensory loss 1 0.9
Signs of meningeal irritation 1 0.9

Electric Digital Subtraction Angiography using either 5
French Head-Hunter or MANI catheter.
Results were analysed using SPSS version 9.0 (Chicago,
IL). Confidence intervals of 95% and p-value of ,0.05
were considered statistically significant. Logistic regression
analysis was used to determine the association between
each red flag feature and outcome.
Sensitivity and specificity 262 table was used to identify
the cut-off number of red flags. Receiver operating
characteristic (ROC) curve (a plot of sensitivity versus
one minus specificity, where the area under the curve
determines the sensitivity of the diagnostic test) was then
used to show the sensitivity of the diagnostic test. The
point where both sensitivity and specificity are optimum
was taken as the cut-off point.
after multivariate regression analysis was seen in only three
out of the five patients with red flag features (Table 2).
(1) Papilloedema
(2) Drowsiness, confusion, memory impairment or LOC
and
(3) Paralysis.
Of the 20 red flag features used, 4 features were not seen
in any patient in this study population: weight loss; tender
poorly pulsatile temporal arteries; recurrent lymphadeno-
pathy; and persistent tinnitus.
From the ROC curve, the cut-off point that is optimally
derived from the sensitivity and specificity 262 table is red
flags of 3 (Figure 1).

Results
111 patients who complained of headache and complied
with the selection criteria were identified from the
computerized radiology database between January 1999
and December 1999.
The age ranged from 7 years to 73 years and the mean
age was 31.42¡17.77 years. Males represented 44.14% of
the group (n549) with mean age of 29.41¡2.36 years, and
females represented 55.86% (n562) with mean age of
33.02¡2.37 years. Patients presented either with single or
multiple red flags with a maximum of five.
There were 39 patients with abnormal radiological
findings, while the rest were interpreted as normal. The
onset of new or different headache was the most common
red flag feature, presenting in 57.7% (n564). Nausea or
vomiting occurred in 29.7% (n533) and ‘‘worst headache
ever experienced’’ in 28.8% (n532) (Table 1).
Significant linear correlation was seen with five red flag
features, these were:
(1) Papilloedema
(2) Drowsiness, confusion, memory impairment or loss of
consciousness (LOC)
(3) Paralysis
(4) Asymmetrical pupillary response
(5) Progressive visual or neurological changes.
A positive correlation, with p-value of less than 0.05,
Discussion
We selected a broad set of 20 different red flags for this
study and three were found to have a strong positive
predictive value. They were onset of paralysis, presence of
papilloedema and ‘‘onset of a combination of confusion,
drowsiness, memory loss and LOC’’. Any one of these is
sufficient indication for performing advanced imaging
investigations.
Paralysis is an important sign of intracranial disease
[1, 7, 8]. A wide range of causes for paralysis was seen,
including primary brain tumours, tuberculosis meningitis,
infarction, carotico-cavernous fistula and bleeding aneurysm.
Most had paralysis of the upper or lower limbs while
some presented with diplopia. 10 out of 15 patients presented
with paralysis and a headache had been shown to have
abnormal neuroimaging.
Papilloedema indicates increased intracranial pressure and
its importance in pointing to the presence of an intracranial
Table 2. Red flags with p-value less than 0.05 on multivariate
regression analysis
Variable p-value
Papilloedema 0.0124
Drowsiness, confusion, memory 0.0069
impairment or loss of consciousness
Paralysis 0.0190

The British Journal of Radiology, August 2003 533


Figure 1. Receiver operating characteristic (ROC) curve. This
graph plots sensitivity vs 12specificity giving rise to ROC
curve. The graph shows area under the curve50.76. Red
flag53 (point50.62) represents the cut-off point (both sensis-
tivity and specificity are optimum derived from specificity and
sensitivity 262 table).
lesion has previously been proven [9]. Up to 38% of patients
with a primary brain tumour may show this sign [7].
However, we were unable to find a study that demonstrates
the relationship between headache and papilloedema as an
indicator of intracranial neuropathology.
The group of symptoms; drowsiness, confusion, memory
impairment or loss of consciousness was the third signi-
ficant red flag in this study and were considered together
as they signify reduction in conscious level [10, 11]. Of 13
patients with this feature in our study, 9 were noted to
have abnormal neuroimaging findings.
Other red flags when individually analysed did not reach
statistical significance. We found that neither age nor sex
of the patients was statistically significant, and this was
taken into account during analysis.
The percentage of positive neuroimaging outcomes was
higher among males (41%), than females (31%), and this
finding was similar to a study by Ramirez-Lassepas et al
[12]. However, it was not statistically significant. Primary
headache has previously been shown to be more common
in females and some studies show up to 3:1 female to male
ratio for migraine [13, 14]. In our study, only a small
difference (44%:56%) in the incidence of headache between
males and females was shown. This difference is likely to
be due to our sample population, which was not random-
ized, and therefore not representative of headache in the
general population.
Headache at the extremes of age are usually significant
[14] and patients older than 55 have been shown to have a
higher incidence of neuropathology [12]. Our sample
however, contained a high proportion of young patients
with only 10% greater than 50 years.
The importance of progression of symptoms was
examined, but this can be difficult to elicit in some cases,
as they can sometimes be insidious. Progressive refractive
changes can cause vision deterioration and it can be asso-
ciated with headache. Therefore this sign can be misleading.
However, 55% of our patients with history of progressive
deterioration of vision had positive neuroimaging.
534
Sobri M, A C Lamont, N A Alias and M N Win
Both patients with pupil asymmetry showed intracranial
abnormalities on neuroimaging, the number of patients
(n52) was too low for statistical evaluation.
Four red flag features that we looked for did not occur
in this series; weight loss, tinnitus, tender temporal arteries
and recurrent lymphadenopathy. Weight loss associated
with headache has previously been noted as a feature in
brain tumour patients [7]. Although not a single patient
in this study had persistent tinnitus, it is nevertheless
considered an important sign as it can be due to vesti-
bulocochlear disease and associated with headache. Tender
poorly pulsatile temporal arteries are seen in temporal
arteritis, which is more common among females after
the 5th decade [15]. The headache is characteristic with
unilateral tenderness over the temporal artery region.
Neuroimaging is negative in this condition and we feel that
because the signs are characteristic, and as the pathology is
extracranial, this feature could possibly be excluded from
the list of red flags.
It was interesting to note that a change in the type or
pattern of headaches was considered important by the
requesting clinicians, and was the most common reason
for requesting neuroimaging [16, 17]. Similar findings are
noted in this study. There were three red flag features
within this group. They were onset of new or different
headache, a worst headache ever experienced and onset of
headache with exertion, sexual activity or coughing. We
found however, that this group of red flags individually
did not show any statistical significance.
Nausea and/or vomiting is a recognized feature of raised
intracranial pressure and has been reported in up to 68%
of patients [7, 8]. In our study, 45% of patients that
presented with nausea and vomiting associated with
headache were positive on neuroimaging. However, this
did not reach statistical significance as an isolated sign.
From the ROC curve and 262 sensitivity and specificity
table, the cut-off point at red flags of three represents the
highest value for both sensitivity and specificity of the
diagnostic test. The area below the curve is equal to 0.76
calculated using trapezoidal rules [18]. Therefore, the curve
is sensitive and the red flag number of three should be
considered as strong indicator for predicting positive
neuroimaging. We concluded from this study that the
presence of at least any of three red flag features is the
optimum to predict positive neuroimaging in headache
patients. Those who have three or more red flag features
should undergo prompt neuroimaging to look for secondary
causes of headache.
A review of the literature failed to provide well-studied
list of red flags in patients with headache, in which
neuroradiological testing would have high diagnostic yield.
To our knowledge, this study is a pioneer in trying to use
red flag features as a screening tool for neuroimaging in
identifying patients with headache due to serious under-
lying pathology.
Conclusion
This study reveals three statistically significant red flag
features in predicting abnormal neuroimaging in patients
with headache. They were paralysis, reduced conscious
level and papilloedema. Patients with headache who have
three or more red flag features should have a prompt
neuroimaging study.
The British Journal of Radiology, August 2003
Red flags in patients with headache
A large multicentre prospective study would be helpful
to identify the significant red flag features. Scoring
according to the weighting of its significance would pave
the way to a more detailed and reliable scoring system.
References
1. Perkins AT, Ondo W. When to worry about headache; Head
pain as a clue to intracranial disease. Postgrad Med
1995;98:197–208.
2. Mohr JP, Caplan LR, Melski JW, et al. The Harvard
Cooperative Stroke Registry: A prospective registry.
Neurology 1978;28:754–62.
3. Dodick D. Headache as a symptom of ominous disease.
Postgrad Med 1997;101:46–66.
4. Benjamin M, Frishberg MD. Neuroimaging in presumed
primary headache disorders. Semin Neurol 1997;58:373–82.
5. Tan AKY, Yeow YK. Warning symptoms of sinister
headache. Singapore Med J 1994;35:294–7.
6. Harris JE, Draper HL, Rhodes AI, Stevens JM. High yield
criteria for emergency cranial computed tomography in adult
patients with no history of head injury. J Accid Emerg Med
1999;17:15–7.
7. Edgeworth J, Bullock P, Bailey A, Gallagher A, Crouchman M.
Why are brain tumours still being missed? Arch Dis Child
1996;74:148–51.
8. Forsyth PA, Posner JB, Jerome BP. Headaches in patients
with brain tumours: A study of 111 patients. Neurology
1993;43:1678–83.
The British Journal of Radiology, August 2003
9. Synder H, et al. Signs and symptoms of patients with brain
tumours presenting to emergency department. J Emerg Med
1993;11:253–8.
10. Fontanarosa PB. Recognition of subarachnoid haemorrhage.
Ann Emerg Med 1989;18:1199–205.
11. Kassiner JP. Our stubborn quest for diagnostic certainty.
A cause of excessive testing. N Engl J Med 1989;320:
1489–91.
12. Ramirez-Lassepas M, Espinosa CE, Cicero JJ, Johnston KL,
Cipolle RJ, Barber DL. Predictors of intracranial pathologic
findings in patients who seek emergency care because of
headache. Arch Neurol 1997;54:1506–9.
13. Rozen TD, Swanson JW, Stang PE, McDonnell SK,
Rocca WA. Incidence of medically recognized migraine: a
1989–1990 study in Olmsted County, Minnesota. Headache
2000;40:216–23.
14. Fettes I. Menstrual migraine. Postgrad Med 1997;101:67–75.
15. Hunder GG, et al. The American Collage of Rheumatology
1990 criteria for the classification of giant cell arteritis.
Arthritis Rheum 1990;33:1122–8.
16. David RM, et al. Practical evaluation and diagnosis of
headache. Semin Neurol 1997;17:307–12.
17. Dumas MD, Pexman JH, Kreeft JH. Computed tomography
evaluation of patients with chronic headache. CMAJ
1994;151:1447–52.
18. Hanley JA, McNeil BJ. The meaning and use of area under a
receiver operating characteristic (ROC) Curve. Radiology
1982;143:29–36.
19. Phillips SJ, Whisnant JP. Hypertension and the brain. Arch
Intern Med 1992;152:938–45.
535