ORIGINAL CONTRIBUTION

Acute Headache Presentations to the

Emergency Department: A Statewide
Cross-sectional Study
Kevin H. Chu, MBBS, GCBiostat, MS, FACEM, Tegwen E. Howell, BEcon, MEcSt,
Gerben Keijzers, MSc (Biomed Health Sci, Epi), MBBS, FACEM, PhD, Jeremy S. Furyk,
MBBS, MSc, MPH&TM, FACEM, Robert M. Eley, BSc(Hons), MSc, PhD, CBiol, CSci,
FRSB, Frances B. Kinnear, MBChB, BSc(Hons), PhD, FACEM, Ogilvie Thom, MBBS,
GCClinResMethods, FACEM, Ibrahim Mahmoud, MBBS, GCEpid, MSc, MPH, PhD,
and Anthony F. T. Brown, MBChB, FRCP, FRCSEd, FRCEM, FACEM

ABSTRACT
Objectives: The objective of this study was to describe demographic and clinical characteristics including
features that were consistent with subarachnoid hemorrhage (SAH), use of diagnostic tests, emergency
department (ED) discharge diagnoses, and disposition of adult patients presenting with an acute headache to
EDs statewide across Queensland, Australia. In addition, potential variations in the presentation and diagnostic
workup between principal-referral and city-regional hospitals were examined.

Methods: A prospective cross-sectional study was conducted over 4 weeks in September 2014. All patients 18
years presenting to one of 29 public and ve private hospital EDs across the state with an acute headache were
included. The headache had to be the principal presenting complaint and nontraumatic. The 34 study sites attend to
about 90% of all ED presentations statewide. The treating doctor collected clinical information at the time of the ED
visit including the characteristics of the headache and investigations performed. A study coordinator retrieved
results of investigations, ED discharge diagnoses, and disposition from state databases. Variations in presentation,
investigations, and diagnosis between city-regional and principal-referral hospitals were examined.

Results: There were 847 headache presentations. Median (range) age was 39 (1892) years, 62% were female,
and 31% arrived by ambulance. Headache peaked instantly in 18% and 1 hour in 44%. It was worst ever in
37%, 10/10 in severity in 23%, and associated with physical activity in 7.4%. Glasgow Coma Scale score
was < 15 in 4.1%. Neck stiffness was noted on examination in 4.8%. Neurologic decit persisting in the ED was
found in 6.5%. A computed tomography (CT) head scan was performed in 38% (318/841, 95% CI = 35% to
41%) and an lumbar puncture in 4.7% (39/832, 95% CI = 3.4% to 6.3%). There were 18 SAH, six
intraparenchymal hemorrhages, one subdural hematoma, one newly diagnosed brain metastasis, and two
bacterial meningitis. Migraine was diagnosed in 23% and primary headache not further specied in 45%. CT
head scans were more likely to be performed in principal-referral hospitals (41%) compared to city-regional

From the School of Medicine, University of Queensland (KHC, RME, FBK, OT, IM, AFTB), Herston, Queensland; the Department of Emergency
Medicine, Royal Brisbane and Womens Hospital (KHC, THE, IM, AFTB), Herston, Queensland; the School of Medicine, Grifth University (GK),
Nathan, Queensland; the School of Medicine, Bond University (GK), Gold Coast, Queensland; the Department of Emergency Medicine, Gold Coast
University Hospital (GK), Gold Coast, Queensland; the College of Medicine and Dentistry, James Cook University (JSF), Townsville, Queensland;
the Department of Emergency Medicine, The Townsville Hospital (JSF), Townsville, Queensland; the Department of Emergency Medicine, Princess
Alexandra Hospital (RME), Brisbane, Queensland; the Department of Emergency Medicine, The Prince Charles Hospital (FBK), Brisbane, Queens-
land; and the Department of Emergency Medicine, Nambour General Hospital (OT), Nambour, Queensland, Australia.
Received February 25, 2016; revision received July 24, 2016; accepted July 25, 2016.
Presented at the 32nd Australasian College for Emergency Medicine Annual Scientic Meeting, Brisbane, November 2015.
Funded by the Emergency Medicine Research Foundation (http://emergencyfoundation.org.au)
The authors have no relevant nancial information or potential conicts to disclose.
Supervising Editor: James E. Olson, PhD.
Address for correspondence and reprints: A/Prof Kevin Chu; e-mail: k.chu@uq.edu.au.
ACADEMIC EMERGENCY MEDICINE 2017;24:5362.

2016 by the Society for Academic Emergency Medicine ISSN 1069-6563

doi: 10.1111/acem.13062 PII ISSN 1069-6563583 53
54 Chu et al. ED HEADACHE PRESENTATIONS

hospitals (33%). The headache in patients presenting to the latter was less likely to be instantly peaking or
associated with activity, but was no less severe in intensity and was more frequently accompanied by nausea
and vomiting. Their diagnosis was more likely to be a benign primary headache. Variations in CT scanning could
thus be due to differences in the case mix. The median (interquartile range) ED length of stay was 3.1 (2.2 to 4.5)
hours. Patients was discharged from the ED or admitted to the ED short-stay unit prior to discharge in 57 and
23% of cases, respectively.

Conclusions: The majority of patients had a benign diagnosis, with intracranial hemorrhage and bacterial
meningitis accounting for only 3% of the diagnoses. There are variations in the proportion of patients receiving
CT head scans between city-regional and principal-referral hospitals. As 38% of headache presentations overall
underwent CT scanning, there is scope to rationalize diagnostic testing to rule out life-threatening conditions.

A cute headache as a presenting complaint is fre-

quently encountered by emergency physicians.1
Emergency departments (EDs) backed up by imaging
and laboratory resources are well suited to perform the
diagnostic workup of worrisome headaches,2 but diag-
nostic protocols and practice vary, and little is known
about which tests are done, or when. Hemorrhage,
infection, and tumor are the most important serious
differential diagnoses that need to be considered and
potentially ruled out. This exclusion of life-threatening
causes is complicated by the knowledge that the major-
ity of headaches are benign and do not require investi-
gations. Clinical decisions rules such as the Ottawa
subarachnoid hemorrhage (SAH) rule can assist in dis-
criminating between these conditions and guide the
need for investigations.3
Despite published rules and guidelines,4,5 there are
variations in the perceived need for specific investiga-
tions between individual emergency physicians and
between departments.6,7 Differences in practice could
be partially explained by the variety of different diag-
nostic approaches in the literature especially for sus-
pected SAH.8,9 Additionally, the benign nature of the
vast majority of headaches limits rigorous comparisons
of alternative diagnostic workups for infrequent life-
threatening causes due to the large sample sizes
required.
Perceptions about diagnostic approaches reported in
surveys may differ from actual practice. A snapshot or
cross-sectional study can provide evidence for actual
practice, give point prevalence estimations, and reveal
variations in practice. Knowledge of current practices
could inform efforts to establish a consistent evidence-
based diagnostic approach.
The objective of this study was to describe demo-
graphic and clinical characteristics including features
that were consistent with SAH, use of diagnostic
tests, and ED discharge diagnoses of adult patients
presenting with an acute headache to EDs statewide
across Queensland, Australia. In addition, potential
variations in the presentation and diagnostic workup
between principal-referral and city-regional hospitals
were examined.
METHODS
Study Design
A cross-sectional study was undertaken to collect
prospective clinical findings on consecutive adult
patients with a headache presenting to one of 29 pub-
lic and five private hospital EDs across the state. We
estimated that the study sites attend to about 90% of
all adult public and private ED presentations state-
wide. This estimation was based on data from the
national Nonadmitted Patient ED Care Database
(NPEDCD),10 which captures 95% of all state public
ED attendances. The noncaptured 5% of public ED
attendances are presentations to the small public EDs
scattered throughout the state. Furthermore, NPEDCD
does not capture attendances from private hospital
EDs. However, private hospitals attend to only about
6% of all ED presentations in Australia.
The study was conducted over a 4-week period in
September 2014. The study was designed by the
Queensland Emergency Research Collaborative
(QERC)11 with support from site investigators at each
participating ED. Individual hospital participation was
voluntary. The conduct was managed by a central
study coordinator. The states central human research
ethics committee approved the study.
Study Setting and Population
Queensland is one of eight states and territories in
Australia. It has a land area of approximately 1.7 mil-
lion square kilometers and a residential population of
4.7 million in 2014 with 2.3 million residing in the
southeast corner of the state.12 The 34 participating
hospitals are listed in Data Supplement S1 (available
as supporting information in the online version of this
ACADEMIC EMERGENCY MEDICINE January 2017, Vol. 24, No. 1 www.aemj.org
paper). For the purpose of this study, four hospitals
with onsite neurosurgical services were categorized as
principal-referral hospitals. For the remaining hospi-
tals, 25 were categorized as city-regional hospitals and
five as private hospitals.
Patients were eligible for inclusion if they presented
to an ED with a nontraumatic headache of any poten-
tial cause. The headache had to be the principal pre-
senting complaint. Some patients had multiple
presentations for a given headache episode. Moreover,
some patients had multiple headache episodes defined
as being separated by a headache-free period.
Patients < 18 years old were excluded. Interhospital
transfers were identified and the patients journey
through the state healthcare system tracked using per-
sonal identifiers.
Study Protocol and Measures
The treating doctor prospectively completed the first
one-page paper datasheet at the time of the ED visit.
He/she was specifically asked whether the headache
was the principle presenting complaint. This was deter-
mined at his/her discretion. The data variables included
the characteristics of the headache such as time from
onset to peak severity, pain severity at peak, whether
the headache was considered worst ever by the patient
or as thunderclap by the treating doctor, activity at
onset, past history of any headaches, and whether the
current episode differed from that in the past. Other
symptoms and signs included Glasgow Coma Scale
score, loss of consciousness, neck pain and stiffness,
and focal neurologic deficits.
The study coordinator completed the second one-
page datasheet upon the receipt of the first datasheet.
The second datasheet focused on the results of investi-
gations including computed tomography (CT) head
scans and cerebrospinal fluid (CSF) analysis when
sampled, ED length of stay and disposition, and ED
diagnosis (based on an ED subset of the International
Classification of Disease 10 codes) recorded at the
time of discharge. The CT and angiography results
abstracted were those reported by a consultant general
radiologist. Final diagnoses were checked at least 3
months post-ED presentation to ensure that serious
causes of headaches such as intracranial hemorrhage
and meningitis were not missed at the initial presenta-
tion.
The information for the second datasheet was
sourced using a statewide ED-based desktop software
application (The Viewer) designed for clinicians to
55
view patient information including emergency presen-
tations and laboratory and imaging results across all
state public hospitals. Data from private hospitals were
provided by site investigators in individual hospitals
based on their medical record, pathology, and radiol-
ogy systems. Primary outcomes were patient epidemiol-
ogy, clinical features, investigations (CT and CSF
analysis), and ED discharge diagnosis.
Data Analysis
The 4-week study period and sample size were primar-
ily determined by resource constraints and practicali-
ties of sustaining prospective data collection
simultaneously across 34 EDs. In 20122013, there
were a total of approximately 3,500 presentations per
day to Queensland public hospital EDs reporting to
NPEDCD.13 Conservatively estimating that 1% of ED
presentations were for headaches (half that reported by
Goldstein et al.1) a 28-day snapshot had the potential
to capture approximately 1,000 cases. A sample size
calculation was not formally made.
The headaches were characterized with descriptive
statistics calculated for patient demographic, clinical
findings, investigations, and ED diagnosis. In an
exploratory analysis, city-regional hospitals were com-
pared with principal-referral hospitals using the chi-
square test, Fishers exact test, Student t-test, or Wil-
coxon rank-sum test as appropriate. Statistical signifi-
cance was set at a < 0.05 (two-sided). As analyses
were considered exploratory, no formal adjustment for
multiple comparisons was performed. In another
exploratory analysis, clinical risk factors associated with
intracranial hemorrhage or bacterial meningitis were
quantified using unadjusted odd ratios (ORs) with
95% confidence intervals (CIs). The risk factors
selected were based on prior research by Perry
et al.3,14 The association was further explored using a
multivariate logistic regression analysis.
Statistical analysis was performed using Stata 14.1
(StataCorp) except the multivariate logistic regression,
which was performed using SPSS 22.0 (SPSS, Inc.).
Reporting followed the Strengthening the Reporting of
Observational Studies in Epidemiology (STROBE)
Statement for cross-sectional studies.
RESULTS
Over the 4-week period, 950 presentations were
recorded for 860 patients. The analysis excluded 13
presentations where headache was not the principal
56 Chu et al. ED HEADACHE PRESENTATIONS

complaint as recorded by the treating doctor, four pre-
sentations where data on whether the headache was
the principal complaint were missing, and 86 presenta-
tions which were not the first presentation of the first
headache episode during the study period. This left
847 presentations from 847 patients available for anal-
ysis (see Data Supplement S2, available as supporting
information in the online version of this paper). Miss-
ing data are reflected in the denominators of the pro-
portions calculated for Tables 15. Data were missing,
for example, on headache onset to peak intensity in
11% (94/847), severity at peak in 13% (114/847),
loss of consciousness in 3% (25/847), and CT head
scans in 0.7% (6/847).
In these 847 headache presentations, 424 (50.1%)
presented to the 25 city-regional hospitals, 388
(45.8%) to the four principal-referral hospitals, and 35
(4.1%) to the five private hospitals. Patient demograph-
ics, arrival by ambulance, time of presentation, triage
category, and previous diagnoses are given in Table 1.
Principal-referral hospitals had a greater percentage of
males, patient arrivals by ambulance, and patients
triaged to higher urgency categories.
Characteristics of the headaches including those
consistent with the possibility of a SAH are given in
Table 2. In the 135 patients with instantly peaking
headaches, 94 (69.6%, 95% CI = 61.4% to 76.8%)
were considered as thunderclap by the treating doctor
when specifically asked. Conversely, in the 125
patients considered to have a thunderclap headache,
94 (75.2%, 95% CI = 67.0% to 81.9%) was instantly
peaking. Principal-referral hospitals had a greater per-
centage of patients with instantly peaking headache or
considered as thunderclap by the treating doctor.
Other symptoms and signs are given in Table 3. City-
regional hospitals also had a greater percentage of
patients complaining of nausea. Investigations per-
formed are listed in Table 4. A CT head scan was per-
formed in 38% (318/841, 95% CI = 3541%) of all
cases. The proportion was higher in principal-referral
compared to city-regional hospitals (42% vs. 33%;
respectively; difference = 9%, 95% CI = 216%). CT
scan by clinical variables is given in Data Supplement
S3 (available as supporting information in the online
version of this paper). Lumbar puncture was per-
formed in 4.7% (39/832, 95% CI = 3.4% to 6.3%).
ED discharge diagnosis, disposition, and ED length
of stay are given in Table 5. A headache diagnosis was
not specified in 44% (366/827, 95% CI = 41% to
48%) of cases. There were 18 patients with SAH (13
aneurysmal, five nonaneurysmal), six with intra-
parenchymal hemorrhage, three brain tumors (one
known primary brain tumor, one known brain metas-
tasis, one newly diagnosed metastasis), and two with
bacterial meningitis. Intracranial hemorrhage and bac-
terial meningitis accounted for 3.1% (95% CI = 2.2%
to 4.6%) of all the headache presentations. Patients
were discharged home from the ED or admitted to the

Table 1
Patient Characteristics

All Hospitals,* City-regional Hospitals, Principal-referral Hospitals,

N = 847 (%) n = 424 (%) n = 388 (%) p-value
Age (y), median (range) 39 (1892) 39 (1888) 39 (1892) 0.360
Female 511/824 (62.0) 269/411 (65.5) 223/387 (57.7) 0.023
Arrival by ambulance 255/822 (31.0) 101/400 (25.3) 140/387 (36.2) 0.001
Time of presentation
00:0005:59 75/788 (9.5) 36/384 (9.4) 35/369 (9.5) 0.167
06:0011:59 221/788 (28.1) 120/384 (31.3) 91/369 (24.7)
12:0017:59 294/788 (37.3) 132/384 (34.4) 151/369 (40.9)
18:0023:59 198/788 (25.1) 96/384 (25.0) 92/369 (24.9)
Australasian Triage Scale score
1 9/821 (1.1) 2/408 (0.5) 7/387 1.8) 0.038
2 92/821 (11.2) 40/408 (9.8) 51/387 (13.2)
3 538/821 (63.5) 283/408 (69.4) 243/387 (62.8)
4 175/821 (21.3) 82/408 (20.1) 80/387 (20.7)
5 7/821 (0.9) 1/408 (0.3) 6/387 (1.6)
Previous diagnoses
Cerebral aneurysm 9/847 (1.1) 3/424 (0.7) 6/388 (1.6) 0.079
Subarachnoid hemorrhage 8/847 (0.9) 3/424 (0.7) 5/388 (1.3)
Brain tumor 13/847 (1.5) 3/424 (0.7) 10/388 (2.6)

*City-regional, principal-referral, and private hospitals.

Comparison between city-regional and principal-referral hospitals.
Ordinal scale ranging from 1 to 5 with 1 assigned to the most urgent cases.
ACADEMIC EMERGENCY MEDICINE January 2017, Vol. 24, No. 1 www.aemj.org 57

Table 2
Headache Clinical Characteristics on History

All Hospitals* City-regional Hospitals, Principal-referral

N = 847 (%) n = 424 (%) Hospitals, n = 388 (%) p-value
Time from onset to peak
Instantaneous 135/753 (17.9) 53/377 (14.1) 73/345 (21.2) 0.032
1 h (not instantaneous) 198/753 (26.3) 92/377 (24.4) 89/345 (25.8)
> 1 to 24 h 272/753 (36.1) 146/377 (38.7) 121/345 (35.1)
> 1 to 3 d 89/753 (11.8) 47/377 (12.5) 42/345 (12.2)
>3d 59/753 (7.8) 39/377 (10.3) 20/345 (5.8)
Severity at peak (110)
10 166/733 (22.7) 85/363 (23.4) 67/335 (20.0) 0.235
79 386/733 (52.7) 199/363 (54.8) 178/335 (53.1)
16 181/733 (24.7) 79/363 (21.8) 90/335 (26.9)
Worst headache ever 292/798 (36.6) 147/394 (37.3) 128/369 (34.7) 0.451
Thunderclap 125/832 (15.0) 51/418 (12.0) 65/379 (17.8) 0.048
Activity at onset (%)
Physical activity 53/838 (6.3) 25/420 (6.0) 26/383 (6.8) 0.615
Coitus 9/838 (1.1) 4//420 (1.0) 4/383 (1.0)
Daily activity 265/838 (31.6) 120/420 (28.6) 138/383 (36.0)
Resting 151/838 (18.0) 90/459 (21.4) 50/383 (13.1)
Sleeping 169/838 (20.2) 82/420 (19.5) 81/383 (21.2)
Others 191/838 (22.8) 99/420 (23.6) 84/383 (21.9)
Headache awoke patient from sleep 172/837 (20.6) 83/419 (19.8) 83/383 (21.7) 0.516
Able to walk after headache started 761/838 (90.8) 383/420 (91.2) 345/383 (90.1) 0.589
Past history of headaches 502/826 (60.8) 257/418 (61.5) 223/373 (59.8) 0.626
Current episode differs from past episodes 157/475 (33.1) 73/247 (29.6) 73/207 (35.3) 0.194

*City-regional, principal-referral, and private hospitals.

Comparison between city-regional and principal-referral hospitals.
Comparison after grouping activities as exertional (physical activity, coitus) and nonexertional (daily activity, resting, sleeping, others).

Table 3
Headache Clinical Features on Physical Examination

All Hospitals,* City-regional Hospitals, Principal-referral Hospitals,

N = 847 (%) n = 424 (%) n = 388 (%) p-value
Glasgow Coma Scale score
15 788/822 (95.9) 397/411 (96.6) 356/376 (94.7) 0.213
1314 16/822 (2.0) 9/411 (2.2) 7/376 (1.9)
1012 7/822 (0.9) 2/411 (0.5) 5/376 (1.3)
39 11/822 (1.3) 3/411 (0.7) 8/376 (2.1)
Temperature (C)
38 22/677 (3.3) 10/313 (3.2) 12/330 (3.7) 0.066
37 to <38 199/677 (29.4) 101/313 (32.3) 88/330 (26.7)
36 to <37 404/677 (59.7) 187/313 (59.7) 195/330 (59.1)
35 to <36 49/677 (7.2) 14/313 (4.5) 33/330 (10.0)
<35 3/677 (0.4) 1/313 (0.3) 2/330 (0.6)
Heart rate (beats/min), mean (SD) 79 (15) 78 (15) 79 (15) 0.325
Blood pressure (mm Hg), mean (SD)
Systolic 136 (24) 136 (23) 136 (24) 0.994
Diastolic 79 (14) 80 (14) 78 (15) 0.040
Loss of consciousness
Yes, witnessed 30/822 (3.7) 12/408 (2.9) 18/379 (4.8) 0.369
Yes, unwitnessed 19/822 (2.3) 11/408 (2.7) 8/379 (2.1)
Nausea 462/817 (56.6) 254/407 (62.4) 183/375 (48.8) <0.001
Vomiting 260/820 (31.7) 137/406 (33.7) 106/379 (28.0) 0.080
Neck pain or stiffness on history 185/815 (22.7) 88/405 (21.7) 87/375 (23.2) 0.623
Neck stiffness on examination 37/776 (4.8) 20/372 (5.4) 12/369 (3.3) 0.155
Neurologic decit
Yes, transient 109/818 (13.3) 63/404 (15.6) 46/379 (12.1) 0.094
Yes, persistent 53/818 (6.5) 21/404 (5.2) 32/379 (8.4)

*City-regional, principal-referral, and private hospitals.

Comparison between city-regional and principal-referral hospitals.

ED short-stay (observation) unit in 80% (654/813, Intracranial hemorrhage (SAH or intraparenchymal

95% CI = 78% to 83%) of the time. The median ED hemorrhage) and bacterial meningitis were associated
length of stay was 3.1 hours. with being older, being male, and arriving by
58 Chu et al. ED HEADACHE PRESENTATIONS

Table 4
Investigations Performed

All Hospitals,* City-regional Hospitals, Principal-referral Hospitals,

N = 847 (%) n = 424 (%) n = 388 (%) p-value
CT 318/841 (37.8) 139/420 (33.1) 162/386 (42.0) 0.009
Cerebral angiography 35/823 (4.3) 12/409 (2.9) 23/387 (5.9) 0.038
Lumbar puncture 39/832 (4.7) 24/420 (5.7) 13/377 (3.5) 0.129
CSF bilirubin analysis 13/832 (1.6) 12/420 (2.9) 1/377 (0.3) 0.004

*City-regional, principal-referral, and private hospitals.

Comparison between city-regional and principal-referral hospitals.
CSF = cerebrospinal uid.

Table 5
Diagnosis and Disposition

All Hospitals,* City-regional Hospitals, Principal-referral Hospitals,

N = 847 (%) n = 424 (%) n = 388 (%) p-value
Primary headache
Unknown cause 365/827 (44.1) 204/412 (49.5) 154/387 (39.8) <0.001
Migraine 194/827 (23.5) 106/412 (25.7) 80/387 (20.7)
Tension 57/827 (6.9) 30/412 (7.3) 25/387 (6.5)
Cluster 3/827 (0.4) 1/412 (0.2) 2/387 (0.5)
Secondary headache
Hemorrhage
Subarachnoid 18/827 (2.2) 9/412 (2.2) 9/387 (2.3)
Intra parenchymal 6/827 (0.7) 3/412 (0.7) 3/387 (0.8)
Subdural 1/827 (0.1) 0 1/387 (0.3)
Infection
Meningitisbacterial 2/827 (0.2) 0 1/387 (0.3)
Meningitiscryptococcal 1/827 (0.1) 1/412 (0.2) 0
Meningitisnegative CSF culture|| 4/827 (0.5) 3/412 (0.7) 1/387 (0.3)
Mastoiditis 1/827 (0.1) 0 0
Meningoencephalitis/encephalopathy 2/827 (0.2) 0 2/387 (0.5)
Neoplasm of brain 3/827 (0.4) 0 3/387 (0.8)
Infarction 17/827 (2.1) 5/412 (1.2) 12/387 (3.1)
Transient ischemic attack 15/827 (1.8) 3/412 (0.7) 12/387 (3.1)
Other headache diagnosis
Postlumbar puncture/epidural 4/827 (0.48) 2/412 (0.5) 2/387 (0.5)
Postcoital 1/827 (0.12) 0 0
Viral infection 25/827 (3.0) 6/412 (1.5) 18 /387 (4.7)
Respiratory causes 15/827 (1.8) 8 /412 (2.0) 4/387 (1.0)
Other nonhead-related causes 93/827 (11.3) 31/412 (7.5) 58/387 (15.0)
ED disposition
Discharged home 465/813 (57.2) 234/395 (59.2) 214/383 (55.9)
Did not wait (self-discharged) 14/813 (1.7) 11/395 (2.8) 3/383 (0.8)
Admitted
Short-stay unit (observation unit) 189/813 (23.3) 91/395 (23.0) 92/383 (24.0)
Internal medicine 99/813 (12.2) 39/395 (9.9) 52/383 (13.6)
Neurosurgery 13/813 (1.6) 0 11/383 (2.9)
Intensive care unit 13/813 (1.6) 1/395 (0.3) 10/383 (2.6)
Interhospital transfer 19/813 (2.4) 19/395 (4.8) 0
Died in ED 1/813 (0.12) 0 1/383 (0.3)
ED length-of-stay (h), median (IQR) 3.1 (2.24.5) 3.1 (2.14.4) 3.4 (2.34.7) 0.054

*City-regional, principal-referral, and private hospitals.

Comparison between city-regional and principal-referral hospitals. Diagnosis is categorized into ve groups: primary headache, sec-
ondary headache, other headache, viral infection, respiratory causes, and other nonhead related causes.
One Haemophilus inuenza on CSF culture, one Streptococcus pneumoniae on CSF polymerase chain reaction.
Immunosuppressed patient with known cryptococcal meningitis.
||Three viral meningitis, one Mollaret meningitis.
CSF = cerebrospinal uid; IQR = interquartile range.

ambulance. The headache was more likely to be associated with impaired consciousness, neck stiffness
instantly peaking, more severe, worst ever, different on examination, and persistent neurologic deficits. A
from past episodes, and related to exertion. It was also past history of any headache was more likely with
ACADEMIC EMERGENCY MEDICINE January 2017, Vol. 24, No. 1 www.aemj.org 59

benign primary headaches. The unadjusted ORs are specificity was 99.8% (95% CI = 98.8% to 99.9%),
given in Data Supplement S4 (available as supporting positive predictive value was 91.7% (95% CI = 61.5%
information in the online version of this paper). The to 99.8%), and negative predictive value was 99.4%
adjusted ORs from the multivariate logistic regression (95% CI = 98.1% to 99.9%).
analysis are given in the Table 6.
The Omnibus Tests of Model Coefficients showed
DISCUSSION
that the multivariate logistic regression model was sta-
tistically significant (v2(16) = 102.208, p < 0.0001). Emergency physicians are familiar with red flag
The casewise plot was not produced because no out- symptoms and signs in the evaluation of patients pre-
liers were found and no multicollinearity detected senting with a headache, consistent with a potentially
(variance inflation factor < 3). The model explained serious underlying cause such as hemorrhage or infec-
82.9% (Nagelkerke R2) of the variance in life-threaten- tion. Critical clinical features include, but are not lim-
ing headache and correctly classified 99.2% of cases. ited to, sudden onset, maximal at onset, and different
Sensitivity was 78.6% (95% CI = 49.2% to 95.3%), than previous headaches for SAH; fever and menin-
geal irritation for bacterial meningitis; and progres-
sively worse over time or worse in morning or head
Table 6
Multivariate Logistic Regression Model for Clinical Features Associ- down position for space-occupying lesion.15 The
ated With a Risk of Intracranial Hemorrhage or Bacterial Meningitis majority of headaches are, however, benign but painful
exacerbations of chronic primary headaches such as
Adjusted OR (95% CI) p-value
migraine, tension-type, and cluster.16
Age (y)
>40 5476 (3.9769891) 0.020 In this study we prospectively characterized over
40 1.0 800 nontraumatic headache presentations in adult
Sex
Male 1234 (2.367264) 0.027 patients across 34 EDs. Acute life-threatening causes
Female 1.0 such as intracranial hemorrhage and bacterial meningi-
Arrival by ambulance
Yes 12 (0.3528) 0.197 tis accounted for about 3%. The majority of headaches
No 1.0
Time from onset to peak
were benign primary headaches. The assignment of a
Instantaneous 50 (1.51700) 0.029 specific diagnosis to these headaches was challenging.
1 h (not instantaneous) 1.2 (0.113.8) 0.895
>1 h 1.0 About 44% of the headache episodes in this study
Severity at peak were labeled as a headache, without a more specific
10 224 (0.95614) 0.055
79 17 (0.12132) 0.255 International Classification of Headache Disorder diag-
16 1.0 nosis.17 This is higher than the 36% reported by
Worst headache ever
Yes 0.1 (0.004.9) 0.239 Friedman et al.,18 who conducted detailed structured
No 1.0 patient interviews with the assistance of trained
Associated with physical activity or coitus
Yes 3614 (8.7150235) 0.008 research associates. Migraine, tension, and cluster
No 1.0
Past history of headache
headaches with their specific treatments represented
Yes 0.001 (0.000.18) 0.009 about 31% of cases in our study. The combined viral
No 1.0
Current episode differs from past episodes infections and respiratory causes comprised about 5%
Yes 1.6 (0.0644.3) 0.770 of cases.
No 1.0
Glasgow Coma Scale score In our setting, CT head scans were more likely to
15 1.0 0.256 be performed in principal-referral hospitals (42%)
1314 43 (0.0728067) 0.142
312 2472 (0.0784204036) compared to city-regional hospitals (33%). Patients pre-
Loss of consciousness
Yeswitness 81 (0.415004) 0.099
senting to the latter were more likely to be female, not
Yesunwitnessed 358 (0.04180) 0.830 arrive by ambulance, and be assigned a lower triage
No 1.0
Neck stiffness on examination category. Their headache was less likely to be instantly
Yes 79811 (24264269) 0.006 peaking or associated with activity, but was no less sev-
No 1.0
Neck pain or stiffness on history ere in intensity and was more frequently accompanied
Yes 0.08 (0.003.4) 0.188 by nausea and vomiting. Their diagnosis was more
No 1.0
Neurologic decit in ED likely to be a benign primary headache. The difference
Yes 2.7 (0.240) 0.481 likely reflects the case mix and clientele presenting to
No 1.0
community versus university hospitals. Variations in
60
CT scanning could thus be due to differences symp-
tomatology and diagnosis.
Overall, the proportion of patients receiving CT
head scans in this snapshot was higher than antici-
pated. CT scans are ordered at a rate of 10.8 per 100
headache problems in Australian general practice,19
while neuroimaging was obtained in 12.4% of all
headache visits to outpatient office-based care in the
United States.20 Comparative figures for ED patients
are elusive but are expected to be higher than non ED
patients. Lumbar punctures were slightly more likely
to be performed at city-regional hospitals (5.9%) com-
pared to principal-referral hospitals (3.7%), possibility
because CT scanning is less available after hours in
the former, although the difference was not statistically
significant.
LIMITATIONS
The strength of this study is its generalizability as it
included almost all ED presentations in the state for a
period of 4 weeks. In addition, the final diagnosis was
reviewed after at least 3 months to ensure that serious
causes of headache were not missed. There are, how-
ever, limitations that might threaten its internal validity
before external validity can be considered. Patient
recruitment was based on the doctors assessment that
the headache was a primary symptom. Eligibility was
not verified nor were missing cases specifically sought.
Systematic selection bias is possible but unlikely given
cases were enrolled in the ED 24 hours per day by
many clinicians across many sites.
The clinical data were collected by the treating doc-
tor and not by dedicated trained research personnel.
Data quality was not assessed, for example, by examin-
ing agreement with a second data collector. The
prospectively collected clinical data in the preprinted
datasheet were, however, usual information required
for the assessment of an ED headache presentation
and as such represent real-world data.
Missing data in this prospective study were consid-
ered as missing and not assumed to be the absence of
a criterion. For the data collection, questions on clini-
cal history stipulated a yes or no response. The only
exception was for past history of SAH, cerebral aneur-
ysm, or brain tumor, where there was a list of tick
boxes for these conditions. Whether missing data may
bias the results will depend on whether the data were
missing at random or not. The latter could be prob-
lematic. Data were missing from the first datasheet if
Chu et al. ED HEADACHE PRESENTATIONS
the patients impaired consciousness precluded a
description of the headache and circumstances sur-
rounding the presentation or because data were not
recorded by the treating doctor. Data were missing
from the second datasheet when patients left the ED
before the assessment was completed, there was no
identifier on the first datasheet to link patients to elec-
tronic medical records, or patients placed a privacy
request on their electronic record so that it could be
accessed by the investigators.
Patients without neuroimaging and without repre-
sentations for at least 3 months were assumed not to
have had a change of their diagnosis. Representations
and revisions in diagnosis, however, will not be cap-
tured in our local state databases if the patient repre-
sented interstate.
Finally, the multivariate logistic regression analysis
remains exploratory. Peduzzi et al.21 suggested that
the minimum number (N) of cases required for a
multivariate logistic regression analysis is 10k/p,
where k is the number of covariates and p is small-
est of the proportions of negative or positive cases.
In this study, there were 847 cases with 3.1% caused
by intracranial hemorrhage or bacterial meningitis.
The number of covariate (k) that would be supported
is 2.6 (847 = 10k/ 0.031). This study is thus under-
powered to provide robustness and stability of the
regression model. Furthermore, the model needs to
be cross-validated before any definitive conclusions
can be made.
CONCLUSIONS
In Queensland, acute headache is a common presenta-
tion to the ED, with the majority of patients having a
benign diagnosis, while intracranial hemorrhage and
bacterial meningitis account for only 3% of the diag-
noses. Computed tomography head scans are more
commonly performed at principal-referral hospitals
compared with city-regional hospitals likely as a result
of differences in the case mix. As 38% of headache
presentations overall underwent CT scanning, there is
scope to rationalize diagnostic testing to rule out life-
threatening conditions.
The following individuals in addition to the writing
group contributed to the conduct of this study:
Anthony Bell, Queen Elizabeth II Hospital; Pauline
Calleja, Barcaldine Hospital; Jaimi Greenslade, Royal
Brisbane & Womens Hospital; Casey Kean Khoo,
Hervey Bay Hospital; Jason Lindeman, Gympie
ACADEMIC EMERGENCY MEDICINE January 2017, Vol. 24, No. 1 www.aemj.org
Hospital; Stephen Margolis, Logan Hospital; Shane
Martin, Ipswich Base Hospital; Chris May, Redlands
Hospital; Doug Morel, Redcliffe Hospital; Vicki Mul-
ler, Maryborough Hospital; Ulrich Orda, Mt Isa
Hospital; Peter Pereira, Cairns Base Hospital; Liz Platz,
Kingaroy Hospital; Donna Powell, Caloundra Hospi-
tal, Louise Prowse, Capricorn Coast Hospital; David
Rosengren, Greenslopes Private Hospital; Sean Roth-
well, St. Andrews War Memorial Hospital; Henk
Sigle, Rockhampton Base Hospital; Neale Thornton,
Mackay Base Hospital; Joseph Ting, Mater Adults
Hospital; Greg Treston, Bundaberg Base Hospital;
Yusuke Ueno-Dewhirst, Nambour General Hospital;
Eric Van Puymbroeck, Noosa Private Hospital; and
Ben Walters, John Flynn Private Hospital and Pin-
darra Private Hospital.
References
1. Goldstein JN, Camargo CA Jr, Pelletier AJ, Edlow JA.
Headache in United States emergency departments: demo-
graphics, work-up and frequency of pathological diagnoses.
Cephalalgia 2006;26:68490.
2. Friedman BW, Lipton RB. Headache in the emergency
department. Curr Pain Headache Rep 2011;15:3027.
3. Perry JJ, Stiell IG, Sivilotti MA, et al. Clinical decision
rules to rule out subarachnoid hemorrhage for acute head-
ache. JAMA 2013;310:124855.
4. Edlow JA, Panagos PD, Godwin SA, Thomas TL, Decker
WW. Clinical policy: critical issues in the evaluation and
management of adult patients presenting to the emergency
department with acute headache. Ann Emerg Med
2008;52:40736.
5. Australasian College for Emergency Medicine. Guidelines
on Diagnostic Imaging. Available at: https://acem.org.au/
getattachment/17ee11f1-bdaf-4ca4-a524-396690123ba7/Dia
gnostic-Imaging.aspx. Accessed Aug 24, 2016.
6. Rogers A, Furyk J, Banks C, Chu K. Diagnosis of sub-
arachnoid haemorrhage: a survey of Australasian emer-
gency physicians and trainees. Emerg Med Australas
2014;26:46873.
7. Perry JJ, Eagles D, Clement CM, et al. An international
study of emergency physicians practice for acute headache
management and the need for a clinical decision rule.
CJEM 2009;11:51622.
8. Chu KH, Bishop RO, Brown AF. Spectrophotometry, not
visual inspection for the detection of xanthochromia in
suspected subarachnoid haemorrhage: a debate. Emerg
Med Australas 2015;27:26772.
9. Edlow JA. What are the unintended consequences of
changing the diagnostic paradigm for subarachnoid hemor-
rhage after brain computed tomography to computed
61
tomographic angiography in place of lumbar puncture?
Acad Emerg Med 2010;17:9915.
10. Australian Institute of Health and Welfare. National Non-
admitted Patient Emergency Department Care Database.
Available at: http://www.aihw.gov.au/hospitals-data/na
tional-non-admitted-patient-emergency-department-care.
Accessed Aug 24, 2016.
11. Queensland Emergency Medicine Research Foundation.
QERC at QEMRF. Available at: http://emergencyfoun
dation.org.au/research/research-support-network/.
Accessed Aug 24, 2016.
12. Queensland Government Statisticians Office. Queensland
Regional Profiles, Available at: http://statistics.qgso.
qld.gov.au/qld-regional-profiles. Accessed Aug 24, 2016.
13. Australian Institute of Health and Welfare 2013. Aus-
tralian Hospital Statistics 201213: Emergency Depart-
ment Care. Health Services Series No. 52. Cat. No. HSE
142. Canberra: AIHW, 2013.
14. Perry JJ, Stiell IG, Sivilotti ML, et al. High risk clinical
characteristics for subarachnoid haemorrhage in patients
with acute headache: prospective cohort study. BMJ
2010;341:c5204.
15. Swadron SP. Pitfalls in the management of headache in
the emergency department. Emerg Med Clin North Am
2010;28:12747.
16. Friedman BW, Grosberg BM. Diagnosis and management
of the primary headache disorders in the emergency
department setting. Emerg Med Clin North Am
2009;27:7187.
17. International Headache Society. HIS Classification ICHD-
II. Available at: http://ihs-classification.org/en. Accessed
Aug 24, 2016.
18. Friedman BW, Hochberg ML, Esses D, et al. Applying
the International Classification of Headache Disorders to
the emergency department: an assessment of reproducibil-
ity and the frequency with which a unique diagnosis can
be assigned to every acute headache presentation. Ann
Emerg Med 2007;49:40919.
19. Charles J, Ng A, Britt H. Presentations of headache in
Australian general practice. Austr Fam Physician
2005;34:6189.
20. Callaghan BC, Kerber KA, Pace RJ, Skolarus LE, Burke
JF. Headaches and neuroimaging: high utilization and
costs despite guidelines. JAMA 2014;174:81921.
21. Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein
AR. A simulation study of the number of events per vari-
able in logistic regression analysis. J Clin Epidemiol
1996;49:13739.
Supporting Information
The following supporting information is available in
the online version of this paper:
62
Data Supplement S1. Participating hospitals.
Data Supplement S2. Number of patient presenta-
tions.
Data Supplement S3. CT head scan rate by hospi-
tal category for selected clinical characteristics.
Chu et al. ED HEADACHE PRESENTATIONS
Data Supplement S4. Clinical features associated
with a risk of intracranial hemorrhage or bacterial
meningitis.