THE RATIONAL
CLINICAL EXAMINATION
Does This Patient Have a Hemorrhagic Stroke?
Clinical Findings Distinguishing Hemorrhagic Stroke
From Ischemic Stroke
Shauna Runchey, MD, MPH
Steven McGee, MD
CLINICAL SCENARIOS
Case 1
A 75-year-old woman with hyperten-
sion, hyperlipidemia, and diabetes
mellitus presents to the emergency
department with speech difficulty and
right arm weakness. Her symptoms
appeared abruptly 6 hours earlier,
when she awoke in the morning, but
several hours passed before she could
reach her daughter, who then trans-
ported her to the hospital. One year
ago she was diagnosed as having a
transient ischemic attack after present-
ing with a 10-minute episode of right-
hand incoordination and numbness;
subsequent neuroimaging and cardio-
vascular studies at that time were
unrevealing. In the emergency depart-
ment, her blood pressure is 140/75
mm Hg and her pulse is 70/min and
regular. She is alert and oriented but
struggles to generate words or name
objects. She reports a mild left hemi-
cranial headache. She exhibits right
central facial paralysis and right arm
weakness. Both great toes are down-
going. Do her findings suggest hemor-
rhagic or ischemic infarction? How
accurate are these findings?
See also Patient Page.
CME available online at
www.jamaarchivescme.com
and questions on p 2304.
2280 JAMA, June 9, 2010—Vol 303, No. 22 (Reprinted)
Downloaded From: http://jama.jamanetwork.com/ by a Fordham University User on 07/22/2013
CLINICIAN’S CORNER
Context The 2 fundamental subtypes of stroke are hemorrhagic stroke and ische-
mic stroke. Although neuroimaging is required to distinguish these subtypes, the di-
agnostic accuracy of bedside findings has not been systematically reviewed.
Objective To determine the accuracy of clinical examination in distinguishing hem-
orrhagic stroke from ischemic stroke.
Data Sources MEDLINE and EMBASE searches of English-language articles pub-
lished from January 1966 to April 2010.
Study Selection Prospective studies of adult patients with stroke that compared
initial clinical findings with accepted diagnostic standards of hemorrhagic stroke (com-
puted tomography or autopsy).
Data Extraction Both authors independently appraised study quality and ex-
tracted relevant data.
Data Synthesis Nineteen prospective studies meeting inclusion criteria were iden-
tified (N=6438 patients; n=1528 [24%] with hemorrhage stroke). Several findings
significantly increase the probability of hemorrhagic stroke: coma (likelihood ratio [LR],
6.2; 95% confidence interval [CI], 3.2-12), neck stiffness (LR, 5.0; 95% CI, 1.9-12.8),
seizures accompanying the neurologic deficit (LR, 4.7; 95% CI, 1.6-14), diastolic blood
pressure greater than 110 mm Hg (LR, 4.3; 95% CI, 1.4-14), vomiting (LR, 3.0; 95%
CI, 1.7-5.5), and headache (LR, 2.9; 95% CI, 1.7-4.8). Other findings decrease the
probability of hemorrhage: cervical bruit (LR, 0.12; 95% CI, 0.03-0.47) and prior tran-
sient ischemic attack (LR, 0.34; 95% CI, 0.18-0.65). A Siriraj score greater than 1 in-
creases the probability of hemorrhage (LR, 5.7; 95% CI, 4.4-7.4) while a score lower
than −1 decreases the probability (LR, 0.29; 95% CI, 0.23-0.37). Nonetheless, many
patients with stroke lack any diagnostic finding, and 20% have Siriraj scores between
1 and −1, which are diagnostically unhelpful (LR, 0.94; 95% CI, 0.77-1.1).
Conclusion In patients with acute stroke, certain findings accurately increase or de-
crease the probability of intracranial hemorrhage, but no finding or combination of
findings is definitively diagnostic in all patients, and diagnostic certainty requires neu-
roimaging.
JAMA. 2010;303(22):2280-2286 www.jama.com
Author Affiliations: Department of Medicine, Uni-
Case 2 versity of Washington, Seattle.
A 62-year-old man presents to the emer- Corresponding Author: Shauna Runchey, MD, MPH,
Department of Medicine, University of Washington,
gency department with new-onset left Box 356426, 1959 NE Pacific St, Health Sciences Bldg,
arm and leg weakness. While working Seattle, WA 98195-6426 (srunchey@u.washington
.edu).
in his workshop 2 hours earlier, he ex- The Rational Clinical Examination Section Editors:
perienced abrupt onset of a severe head- David L. Simel, MD, MHS, Durham Veterans Affairs
ache. Within 10 to 15 minutes, he had Medical Center and Duke University Medical Cen-
ter, Durham, NC; Drummond Rennie, MD, Deputy
difficulty holding tools in his left hand Editor.
©2010 American Medical Association. All rights reserved.

and he required assistance to get to his
car. He has chronic atrial fibrillation and
has been taking aspirin for the past 4
years. In the emergency department, his
blood pressure is 200/108 mm Hg and
his pulse is 104/min and chaotic. He has
a dense left hemiparesis, hemianesthe-
sia, and a Babinski response. During the
examination he becomes progres-
sively drowsy and vomits twice. Do his
findings suggest hemorrhagic or ische-
mic infarction? How accurate are these
findings?
WHY IS THE CLINICAL
EXAMINATION IMPORTANT?
In the United States, new or recurrent
stroke affects about 795 000 individu-
als each year and represents the third
leading cause of death.1 The 2 funda-
mental subtypes of stroke are hemor-
rhagic stroke (ie, either intracerebral or
subarachnoid hemorrhage) and ische-
mic stroke (ie, infarction from throm-
bosis or embolism). In the United
States, 87% of these vascular strokes are
ischemic and 13% are hemorrhagic
(10% are intracerebral hemorrhage; 3%
are subarachnoid hemorrhage),1 al-
though hospital-based surveys from
other countries suggest that the per-
centage due to hemorrhage may be as
high as 25% to 60%.2,3 Clinicians must
rapidly distinguish these 2 vascular sub-
types because they have distinct etiolo-
gies, prognoses, and treatments.
Herein, we assume that the clinical
diagnosis of stroke has already been
made, a subject extensively reviewed in
a prior Rational Clinical Examination
article.4 Classic descriptions of stroke
suggest that some clinical features may
differentiate these subtypes: head-
ache, neck stiffness, vomiting, and coma
are more common in hemorrhagic
stroke whereas previous transient is-
chemic attacks, atrial fibrillation, and
atherosclerosis risk factors are more
common in ischemic stroke. Clini-
cians frequently examine patients with
these findings in mind to form a clini-
cal impression of whether hemor-
rhage or ischemia is more likely. Even
so, all patients with new strokes must
undergo immediate computed tomog-
©2010 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a Fordham University User on 07/22/2013
DISTINGUISHING HEMORRHAGIC AND ISCHEMIC STROKES
raphy (CT) of the head, the best test that
rapidly distinguishes intracerebral hem-
orrhage from ischemia.
In patients presenting early enough to
be considered for thrombolytic therapy,
the clinical priorities are urgent CT
imaging, stabilization of the patient, iden-
tification of stroke mimics (eg, migraine,
seizures, hypoglycemia) and contrain-
dications to thrombolysis, and measure-
ment of the neurologic deficit.5 In such
patients, a clinician’s impression of hem-
orrhage vs ischemia is secondary to these
more pressing matters. Nonetheless, 3 of
4 patients presenting to emergency
departments with stroke are already
beyond the narrow time window for
thrombolysis,6 and in these patients the
clinical overall impression may be help-
ful while awaiting the CT results. The
clinical impression also is essential when
CT is not immediately available.2,3 Finally,
clinical impression is even important in
patients receiving thrombolytics. Accord-
ing to treatment guidelines, the infu-
sion should be immediately discontin-
ued (and imaging repeated) if severe
headache, acute hypertension, nausea, or
vomiting develop,5 a statement imply-
ing that these symptoms and signs sug-
gest hemorrhage.
The diagnostic accuracy of clinical
findings in distinguishing hemor-
rhagic and ischemic stroke has not been
systematically reviewed. In this ar-
ticle, we summarize the diagnostic ac-
curacy of all available clinical informa-
tion that is easily obtainable to clinicians
at the bedside—initial patient inter-
view, physical examination, and basic
laboratory tests. In addition, we re-
view the diagnostic accuracy of differ-
ent stroke scores that combine clini-
cal information to predict the risk of
hemorrhagic infarction.
PHYSIOLOGIC ORIGINS
OF SYMPTOMS AND SIGNS
Both cerebral hemorrhage and infarc-
tion cause abrupt dysfunction of neu-
rologic tissue, leading to neurologic defi-
cits such as hemiparesis, hemisensory
loss, aphasia, ophthalmoplegia, and vi-
sual field cuts. Cerebral hemorrhage, in
contrast, also causes leakage of blood
(Reprinted) JAMA, June 9, 2010—Vol 303, No. 22
into the brain, displacing and compress-
ing adjacent tissue, increasing intracra-
nial pressure, and eventually dissect-
ing into the ventricles and subarachnoid
space (FIGURE). Consequently, hemor-
rhage may cause additional symptoms
beyond neurologic deficits, such as
severe headache (from increased intra-
cranial pressure or meningeal irrita-
tion), progressive deterioration after
stroke onset (from continued bleed-
ing), vomiting (from increased intra-
cranial pressure), neck stiffness (from
meningeal irritation), bilateral Babin-
ski signs (from enlargement of the hem-
orrhage beyond the distribution of a
single vessel), and coma (from bilat-
eral cerebral dysfunction or uncal her-
niation).
METHODS
Literature Search Strategy
One author (S.R.) searched PubMed
(1970 to April 2010), using MEDLINE,
and EMBASE (1988 to April 2010), using
OVID, to identify English-language stud-
ies that evaluated the diagnostic accu-
racy of clinical history and physical signs
for detecting intracerebral hemorrhage.
The specific search strategy for PubMed
was (Stroke/diagnosis[Mesh] OR Intra-
cranial Embolism and Thrombosis/
diagnosis[Mesh]) AND Intracranial
Hemorrhages/diagnosis[Mesh]) and for
EMBASE was (exp *Stroke/et, di [Etiol-
ogy, Diagnosis] or exp Brain Hemor-
rhage/et, di [Etiology, Diagnosis] or exp
Brain Infarction/et, di [Etiology, Diag-
nosis] and diagnostic accuracy/ or diag-
nostic value/). Additional articles were
identified from liberal use of the search
tool “all related articles” in PubMed and
examination of bibliographies of se-
lected articles.
Study Selection
Both authors independently read all ar-
ticles relevant to the study question
and selected those representing unique
patient populations meeting the
following inclusion criteria: (1) The
study prospectively enrolled only
patients presenting to clinicians in a
hospital or emergency department with
a diagnosis of stroke, defined as focal
2281
 DISTINGUISHING HEMORRHAGIC AND ISCHEMIC STROKES

Figure. Explanation of Additional Symptoms of Intracerebral Hemorrhage

Pathophysiological events Clinical findings

Initial hemorrhage
AXIAL VIEW CORONAL VIEW

Lateral Lateral
ventricle ventricle
Lateral
ventricle
Hemiparesis

Continued bleeding, compression of adjacent tissues, and increased intracranial pressure

AXIAL VIEW CORONAL VIEW S A G I T TA L V I E W

Lateral Lateral
ventricle Lateral ventricle Third ventricle
ventricle
Lateral Headache
ventricle
Tentorium Vomiting
cerebelli Hemiplegia
Third
ventricle Drowsiness

Fourth
Subarachnoid ventricle
space
Brainstem Median
aperture

Dissection of blood into the ventricles and extension of blood into the subarachnoid space; uncal herniation

AXIAL VIEW CORONAL VIEW S A G I T TA L V I E W

Subarachnoid
space

Lateral
ventricle
Neck stiffness
Coma

Subarachnoid
Extension of blood
space
Tentorium Uncal Brainstem via lateral aperture
cerebelli herniation

In the top panel, a small hemorrhage in the right basal ganglia causes left hemiparesis and a clinical presentation indistinguishable from ischemic stroke. As intracerebral
bleeding continues (middle panel), expansion of the hemorrhage exerts a mass effect on the brain, increasing intracranial pressure and causing a midline shift. Clinical
findings characteristic of hemorrhagic stroke manifest, such as progressive neurological deficits, headache, and vomiting. Eventually, blood may dissect into the ven-
tricles and extend into the subarachnoid space via the median and lateral apertures of the fourth ventricle (bottom panel), leading to neck stiffness. In severe hemor-
rhagic stroke, intracerebral expansion of the hemorrhage may result in coma from bilateral cerebral dysfunction or uncal herniation.

2282 JAMA, June 9, 2010—Vol 303, No. 22 (Reprinted) ©2010 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a Fordham University User on 07/22/2013


Table 1. Definitions and Diagnostic Accuracy of Stroke Scores Detecting Hemorrhage: Pooled Results a
No. of Hemorrhages,
Score Patients No. (%)
Siriraj stroke 3439 1051 (31)
score13,15-17,26-34
Besson 261 46 (18)
score10,14
Abbreviations: CI, confidence interval; LR, likelihood ratio.
a See eTable 2 for results from the individual studies.
b Within each set of parentheses is the number of points awarded if the criterion following the times sign is satisfied. For example, in the Siriraj score, (2.5⫻semicoma or 5⫻coma) means
2.5 points are awarded if the patient has semicoma or 5 points if the patient has coma.
(or at times global) neurologic impair-
ment of sudden onset and of pre-
sumed vascular origin (World Health
Organization definition).7 (2) The study
used either neuroimaging (CT or mag-
netic resonance imaging) or autopsy to
distinguish hemorrhage from ische-
mia (all but 1 study8 exclusively used
CT). (3) Intracranial hemorrhage re-
ferred only to intracerebral hemor-
rhage (studies enrolling patients with
subarachnoid hemorrhage were ex-
cluded). (4) The study presented suf-
ficient information to create 2⫻2 tables
and calculate sensitivity, specificity, and
likelihood ratios (LRs). Each study was
assigned a quality grade similar to other
articles in the Rational Clinical Exami-
nation series9 according to whether
sample sizes were adequate, cases were
consecutive, and comparison with the
diagnostic standard was independent
(in assigning quality grades, we con-
sidered studies with ⱖ100 total pa-
tients and ⱖ30 patients with hemor-
rhage to be large; see eTable 1 [available
at http://www.jama.com]). In all stud-
ies, patients were enrolled at the time
of presentation and clinical data were
collected before the criterion standard
for the subtype of stroke (hemorrhage
vs ischemia). No patient received
thrombolytic medications. Several in-
vestigators8,10-18 supplied additional un-
published information from their origi-
nal studies.
Statistical Analyses
Both authors independently extracted
data from each study to calculate sen-
©2010 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a Fordham University User on 07/22/2013
DISTINGUISHING HEMORRHAGIC AND ISCHEMIC STROKES
Definition b
(2.5 ⫻ semicoma or 5 ⫻ coma) ⫹ (2 ⫻ vomiting) ⫹
(2 ⫻ headache within 2 h) ⫹ (0.1 ⫻ diastolic
blood pressure) − (3 ⫻ ⱖ1 of diabetes, angina,
intermittent claudication) − 12
(2 ⫻ alcohol consumption) ⫹ (1.5 ⫻ plantar response
both extensor) ⫹ (3 ⫻ headache) ⫹ (3 ⫻ history
of hypertension) − (5 ⫻ history of transient
ischemic attack) − (2 ⫻ peripheral arterial disease)
− (1.5 ⫻ history of hyperlipidemia) − (2.5 ⫻ atrial
fibrillation on admission)
sitivity, specificity, and positive and
negative LRs using standard defini-
tions.19 Any differences in data entry
were settled by consensus. If any cell
in the 2⫻2 table had the value of zero,
0.5 was added to all cells before calcu-
lating LRs or pooled estimates. Pooled
estimates were calculated using the Der-
Simonian and Laird random-effects
model.20 Data are presented using LRs
because they function as “diagnostic
weights” that are easily translated to
posttest probability of disease.21 Excel
2007 (Microsoft, Redmond, Washing-
ton) was used for statistical analyses.
RESULTS
Study Characteristics
We retrieved 109 citations for full-
text review, 19 of which met our in-
clusion criteria and presented data from
6438 patients worldwide with stroke
(eFigure and eTable 1). The methods
and enrollment of these studies varied
widely. The maximal allowed interval
between onset of neurologic symp-
toms and CT scanning was less than 3
days in 68% of studies, less than 15 days
in 16%, and 15 days or more or un-
known in 16%.
We identified 6 stroke scores that
combined clinical findings to calcu-
late probability of hemorrhagic
stroke.10,17,22-25 Three were excluded
because they lacked extensive exter-
nal validation beyond the original
cohort.22,23,25 A fourth score, the once
popular Allen (Guy’s Hospital) score,
was excluded because it required
clinical information first available 24
(Reprinted) JAMA, June 9, 2010—Vol 303, No. 22
Threshold LR for Hemorrhage
Values (95% CI)
⬍−1: infarction 0.29 (0.23-0.37)
−1 to 1: uncertain 0.94 (0.77-1.1)
⬎1: hemorrhage 5.7 (4.4-7.4)
⬍1: infarction 0.23 (0.01-5)
ⱖ1: hemorrhage 1.4 (0.92-2.2)
hours after presentation, which is
clinically unrealistic.24 The definitions
of the included stroke scores appear
in TABLE 1.
Prevalence of Hemorrhage Stroke
in Patients Presenting With Stroke
The prevalence of hemorrhagic stroke
in these studies was 24% (n = 1528),
with individual study prevalences
ranging from 12% to 57% (random-
effects summary prevalence, 29%;
95% confidence interval [CI], 24%-
35%). In general, the prevalence of
hemorrhage is lower in studies from
the United States and Europe (sum-
mary prevalence, 15%; 95% CI, 13%-
18%) than in those from Asia and
Africa (summary prevalence, 34%;
95% CI, 27%-41%). Some of this dif-
ference probably reflects real differ-
ences around the world but it also
may indicate referral bias (ie, in coun-
tries with more limited access to CT
and thrombolytic therapy, patients
thought to have hemorrhage may be
more likely referred for CT). A small
number of patients (range, 2.0%-
2.9%) experienced a stroke mimic that
was neither hemorrhagic nor ischemic
(eg, attributable to tumor, subdural
hematoma, or tuberculosis).13,26,27,35
Individual Findings
Several findings significantly increase
the probability of hemorrhagic stroke
(LR 95% CI does not cross 1.0). In or-
der of their LRs (the higher the value
of the LR, the greater the increase in
probability of hemorrhagic stroke),
2283
 DISTINGUISHING HEMORRHAGIC AND ISCHEMIC STROKES

these findings include coma (LR, 6.2;
95% CI, 3.2-12), neck stiffness (LR, 5.0;
95% CI, 1.9-12.8), seizures accompa-
nying neurologic deficit (LR, 4.7; 95%
CI, 1.6-14), diastolic blood pressure
greater than 110 mm Hg (LR, 4.3; 95%
CI, 1.4-14), vomiting (LR, 3.0; 95% CI,
1.7-5.5), headache (LR, 2.9; 95% CI,
1.7-4.8), and loss of consciousness (LR,
2.6; 95% CI, 1.6-4.2) (TABLE 2 and
eTable 3 ). The finding of xanthochro-
mia during lumbar puncture also
greatly increases the probability of hem-
orrhage (LR, 15; 95% CI, 7.7-29), al-
though this test is no longer routinely
performed to identify intracranial hem-
orrhage. The findings of diastolic blood
pressure greater than 110 mm Hg, loss
of consciousness, and xanthochromia
were each investigated in only a single
study.
Several findings significantly de-
crease the probability of hemorrhagic
stroke, thus increasing the probability
of ischemic stroke. These findings in-
clude the presence of a cervical bruit
(LR, 0.12; 95% CI, 0.03-0.47), ab-

Table 2. Accuracy of Findings for Diagnosing Hemorrhagic Stroke a

No. of Hemorrhage, Sensitivity, % Specificity, % Positive LR Negative LR
Finding Patients No. (%) (95% CI) (95% CI) (95% CI) (95% CI)
Risk factors
Age ⱕ60 y35 1510 237 (16) 50 (43-56) 70 (68-73) 1.7 (1.4-1.9) 0.71 (0.63-0.82)
Alcohol consumption10 178 27 (15) 48 (29-67) 70 (62-77) 1.6 (1-2.5) 0.75 (0.51-1.1)
Male16-18,28,35 3107 635 (20) 57 (53-61) 51 (47-54) 1.2 (1.1-1.3) 0.85 (0.77-0.94)
Hypertension10,11,16-18,28,35 4193 776 (19) 68 (60-75) 40 (33-47) 1.1 (1.0-1.2) 0.88 (0.77-1.01)
Cigarette smoking11,28 1187 216 (18) 38 (22-55) 52 (45-79) 0.79 (0.45-1.4) 1.2 (0.79-1.8)
Diabetes mellitus11,13,16,28,29,35 3866 681 (18) 17 (9-25) 74 (66-81) 0.64 (0.43-0.95) 1.1 (1.0-1.2)
Prior stroke17,35 1622 295 (18) 11 (4-18) 79 (67-91) 0.59 (0.17-2.0) 1.1 (0.88-1.4)
Hyperlipidemia10,11,16 2028 300 (15) 7 (0-15) 78 (68-89) 0.48 (0.2-1.1) 1.1 (1.1-1.1)
Coronary artery disease11,16,35 3420 523 (15) 6 (0-13) 83 (67-100) 0.44 (0.31-0.61) 1.1 (1.0-1.3)
Atrial fibrillation11,16,35 3420 523 (15) 4 (0-7) 90 (89-91) 0.44 (0.25-0.78) 1.1 (1.05-1.1)
Peripheral artery disease10,35 1710 270 (16) 3 (1-5) 91 (86-96) 0.41 (0.2-0.83) 1.1 (1.0-1.1)
Prior transient ischemic 3478 523 (15) 7 (3-11) 79 (75-84) 0.34 (0.18-0.65) 1.2 (1.1-1.3)
attack10,11,16,35
Symptoms
Seizures accompanying neurologic 2497 385 (15) 9 (6-12) 98 (97-100) 4.7 (1.6-14) 0.93 (0.9-0.96)
deficit11,18,35
Vomiting13,16-18,29,35 2947 577 (20) 34 (17-52) 93 (90-96) 3.0 (1.7-5.5) 0.73 (0.59-0.91)
Headache10,11,13,16-18,29,35 3974 708 (18) 46 (41-52) 82 (75-89) 2.9 (1.7-4.8) 0.66 (0.56-0.77)
Loss of consciousness17 174 75 (43) 47 (35-58) 82 (74-89) 2.6 (1.6-4.2) 0.65 (0.52-0.82)
Acute onset of deficit11 887 109 (12) 44 (35-53) 32 (29-35) 0.65 (0.52-0.81) 1.7 (1.4-2.1)
Physical signs
Kernig sign, Brudzinski sign, 50 23 (46) 15 (0-29) 98 (93-100) 8.2 (0.44-150) 0.87 (0.73-1.0)
or both29
Level of consciousness: 1161 223 (19) 35 (19-50) 94 (89-99) 6.2 (3.2-12)
coma11,17,18
Neck stiffness17,29 223 97 (43) 20 (12-28) 97 (93-100) 5.0 (1.9-12.8) 0.83 (0.75-0.92)
Diastolic blood pressure 50 23 (46) 48 (27-68) 89 (77-100) 4.3 (1.4-14) 0.59 (0.39-0.89)
⬎110 mm Hg29
Level of consciousness: 1161 223 (19) 32 (20-44) 82 (69-96) 2.0 (1.0-3.9)
drowsy11,17,18
Plantar response: both extensor10,17 370 106 (29) 16 (9-23) 92 (89-96) 1.8 (0.99-3.4)
Plantar response: 370 106 (29) 62 (44-81) 39 (27-51) 1 (0.87-1.2)
single extensor10,17
Hemiparesis11,16,35 3420 523 (15) 63 (25-100) 33 (0-66) 0.96 (0.9-1.0) 1.1 (1.0-1.2)
Plantar response: both flexor10,17 370 106 (29) 11 (5-17) 74 (69-80) 0.45 (0.25-0.81)
Level of consciousness: alert17,18 274 114 (42) 23 (15-30) 31 (12-51) 0.35 (0.24-0.5)
Cervical bruit35 1510 237 (16) 1 (0-2) 93 (91-94) 0.12 (0.03-0.47) 1.1 (1.0-1.1)
Laboratory
Xanthochromia in cerebrospinal 231 41 (18) 71 (57-85) 95 (92-98) 15 (7.7-29) 0.31 (0.19-0.49)
fluid8
Atrial fibrillation on 1087 142 (13) 2 (0-4) 82 (64-100) 0.19 (0.06-0.59) 1.2 (1.0-1.5)
electrocardiogram10,11
Clinician’s overall impression
Hemorrhage most likely diagnosis28 300 111 (37) 76 (68-84) 88 (83-92) 6.2 (4.2-9.3) 0.28 (0.20-0.39)
Abbreviations: CI, confidence interval; LR, likelihood ratio.
a See eTable 3 for results from the individual studies.

2284 JAMA, June 9, 2010—Vol 303, No. 22 (Reprinted) ©2010 American Medical Association. All rights reserved.

Downloaded From: http://jama.jamanetwork.com/ by a Fordham University User on 07/22/2013


sence of xanthochromia on lumbar
puncture (LR, 0.31; 95% CI, 0.19-
0.49), history of transient ischemic
attack (LR, 0.34; 95% CI, 0.18-0.65),
peripheral artery disease (LR, 0.41; 95%
CI, 0.2-0.83), and history of atrial fi-
brillation (LR, 0.44; 95% CI, 0.25-
0.78). The finding of cervical bruit was
investigated in only 1 study.
Combinations of Findings
The diagnostic accuracies of the Bes-
son score (2 studies) and the Siriraj
score (13 studies) are shown in Table 1.
The Siriraj score is most accurate, in-
creasing the probability of hemor-
rhage when it indicates hemorrhage
(LR, 5.7; 95% CI, 4.4-7.4) and decreas-
ing it when it indicates ischemic in-
farction (LR, 0.29; 95% CI, 0.23-
0.37). Nonetheless, in all patients
presenting with stroke, 20% were clas-
sified as uncertain by the Siriraj score
(LR, 0.94; 95% CI, 0.77-1.1), a diag-
nostically unhelpful result because it
changes probability of hemorrhage little
or not at all (LRs near the value of 1.0
do not change probability).
Overall Clinical Impression
In a single study, clinician overall im-
pression of hemorrhage (LR, 6.2; 95%
CI, 4.2-9.3) or ischemia (LR, 0.28; 95%
CI, 0.20-0.39), without using explicit
rules or allowing classification into an
“uncertain” category, was as accurate
as the Siriraj score. This suggests that
these experienced clinicians are using
findings not addressed in our study,
combining them in nonlinear ways, or
deriving greater accuracy from more
thorough familiarity with the actual pa-
tients seen in these studies.
Limitations
One potential limitation is the meth-
odological diversity of these studies,
making it unwise to pool results, but
when we reexamined the Siriraj score
in various subgroups—patients receiv-
ing CT within 72 hours, studies meet-
ing grade 1 criteria, or studies meet-
ing grade 2 or 3 criteria—the LRs were
largely unchanged (eTable 4). Also, we
found no correlation between preva-
©2010 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ by a Fordham University User on 07/22/2013
DISTINGUISHING HEMORRHAGIC AND ISCHEMIC STROKES
lence of hemorrhage in a particular
study and that study’s hemorrhage LR
(P = .38 by Pearson correlation) or ra-
tio of hemorrhage LR to infarction LR,
an indicator of overall discriminatory
power (P=.47).
Another concern might be the defi-
nition of stroke used by most investi-
gators, a post hoc definition by the
World Health Organization requiring
that neurologic deficits persist for longer
than 24 hours. This raises the possibil-
ity that our conclusions may not ap-
ply to patients presenting to emer-
gency departments with only 1 to 2
hours of symptoms. Nonetheless, we
believe that the patients in these stud-
ies are similar to those with shorter du-
ration of symptoms for the following
reasons. First, the 24-hour time limit
was based on an outdated and arbi-
trary distinction between transient is-
chemic attack and stroke that no longer
is applicable and probably was mis-
leading because most patients with tran-
sient ischemic attack have less than 30
minutes of symptoms and their symp-
toms are unwitnessed by physicians.36
Second, few patients with only 1 to 2
hours of symptoms experience com-
plete resolution by 24 hours (unless
they receive treatment). In the pla-
cebo group of the National Institute of
Neurological Disorders and Stroke trial
of recombinant human tissue plasmino-
gen activator, only 2% of patients pre-
senting with less than 3 hours of neu-
rologic deficits had resolution of
findings at 24 hours.37
SCENARIO RESOLUTION
Case 1
The patient’s findings suggest ische-
mic stroke. Her pretest probability of
hemorrhagic stroke is 13% (in the
United States), and the findings of pre-
vious transient ischemic attack, alert
mental status, and absence of severe hy-
pertension or Babinski sign decrease the
probability of hemorrhagic stroke even
further. Even with the presence of head-
ache, her calculated Siriraj score is −5.5,
which is test-positive for ischemic in-
farction (scores ⬍−1: LR, 0.29), de-
creasing her probability of hemor-
(Reprinted) JAMA, June 9, 2010—Vol 303, No. 22
rhagic infarction from 13% to 4%. Her
CT scan reveals an infarction in the left
posterior frontal lobe.
Case 2
The patient’s findings of headache, se-
vere hypertension, emesis, drowsi-
ness, and rapid neurologic progres-
sion all strongly suggest hemorrhagic
stroke, and the calculated Siriraj score
is 5.3, which is test-positive for hem-
orrhage (scores ⬎1: LR, 5.7), a result
that increases the probability of hem-
orrhage from 13% to 46%. In this pa-
tient, urgent CT scanning reveals an in-
tracerebral hemorrhage centered in the
right basal ganglia, with surrounding
edema and shift of the midline struc-
tures to the left.
CLINICAL BOTTOM LINE
Other than previous transient ische-
mic attack, classic risk factors for ce-
rebral atherosclerosis—such as diabe-
tes mellitus, hyperlipidemia, cigarette
smoking, and hypertension—fail to dis-
tinguish the 2 subtypes of stroke, either
because the risk factor is only weakly
linked to ischemic stroke (eg, in the
Framingham Stroke Profile,38 a tool for
assessing risk of stroke, diabetes or ciga-
rette smoking increases the risk of
stroke only half as much as atrial fibril-
lation) or because the risk factor is
linked to both hemorrhage and ische-
mia, thus nullifying its diagnostic value
(eg, hypertension is linked to both ce-
rebral atherosclerosis—and, thus, is-
chemia—and intracerebral hemor-
rhage).
We conclude that in patients with
stroke, additional clinical findings char-
acteristic of hemorrhage—headache,
vomiting, severe hypertension, neck
stiffness, and coma—increase the prob-
ability of hemorrhagic stroke. Even so,
many patients with hemorrhage lack
any distinctive findings or are unclas-
sifiable by stroke scores. Neither the
clinical impression of experienced cli-
nicians nor the most accurate stroke
score can improve the posttest prob-
ability of hemorrhage to greater than
50%. While combinations of findings
are more predictive than individual
2285
 DISTINGUISHING HEMORRHAGIC AND ISCHEMIC STROKES
findings, diagnostic certainty requires
neuroimaging.
Author Contributions: Drs Runchey and McGee had
full access to all of the data in the study and take re-
sponsibility for the integrity of the data and the ac-
curacy of the data analysis.
Study concept and design: Runchey, McGee.
Acquisition of data: Runchey, McGee.
Analysis and interpretation of data: Runchey, McGee.
Drafting of the manuscript: Runchey, McGee.
Critical revision of the manuscript for important in-
tellectual content: Runchey, McGee.
Statistical analysis: Runchey, McGee.
Study supervision: McGee.
Financial Disclosures: None reported.
Online-Only Material: The eFigure and eTables 1
through 4 are available online at http://www.jama
.com.
Additional Contributions: We appreciate the exper-
tise and assistance of Sherry Dodson, University of Wash-
ington, who helped conduct the literature search. Michael
Fitch, MD, PhD, Wake Forest University School of Medi-
cine, and Amanda Green, MD, Paris Regional Medical
Center, provided helpful advice on earlier versions of
the manuscript. No compensation was received.
REFERENCES
1. Lloyd-Jones D, Adams R, Carnethon M, et al; Ameri-
can Heart Association Statistics Committee and Stroke
Statistics Subcommittee. Heart disease and stroke sta-
tistics—2009 update: a report from the American Heart
Association Statistics Committee and Stroke Statis-
tics Subcommittee. Circulation. 2009;119(3):480-
486.
2. Brainin M, Teuschl Y, Kalra L. Acute treatment and
long-term management of stroke in developing
countries. Lancet Neurol. 2007;6(6):553-561.
3. Durai Pandian J, Padma V, Vijaya P, Sylaja PN,
Murthy JM. Stroke and thrombolysis in developing
countries. Int J Stroke. 2007;2(1):17-26.
4. Goldstein LB, Simel DL. Is this patient having a
stroke? JAMA. 2005;293(19):2391-2402.
5. Adams HP Jr, del Zoppo G, Alberts MJ, et al; Ameri-
can Heart Association; American Stroke Association
Stroke Council; Clinical Cardiology Council; Cardio-
vascular Radiology and Intervention Council; Athero-
sclerotic Peripheral Vascular Disease and Quality of
Care Outcomes in Research Interdisciplinary Work-
ing Groups. Guidelines for the early management of
adults with ischemic stroke: a guideline from the Ameri-
can Heart Association/American Stroke Association
Stroke Council, Clinical Cardiology Council, Cardio-
vascular Radiology and Intervention Council, and the
Atherosclerotic Peripheral Vascular Disease and Qual-
ity of Care Outcomes in Research Interdisciplinary
Working Groups: the American Academy of Neurol-
ogy affirms the value of this guideline as an educa-
tional tool for neurologists. Stroke. 2007;38(5):
1655-1711.
2286 JAMA, June 9, 2010—Vol 303, No. 22 (Reprinted)
Downloaded From: http://jama.jamanetwork.com/ by a Fordham University User on 07/22/2013
6. Kleindorfer D, Kissela B, Schneider A, et al; Neu-
roscience Institute. Eligibility for recombinant tissue
plasminogen activator in acute ischemic stroke: a
population-based study. Stroke. 2004;35(2):e27-
e29.
7. WHO MONICA Project Principal Investigators. The
World Health Organization MONICA Project (Moni-
toring Trends and Determinants in Cardiovascular
Disease): a major international collaboration. J Clin
Epidemiol. 1988;41(2):105-114.
8. Britton M, Hultman E, Murray V, Sjoholm H. The
diagnostic accuracy of CSF analyses in stroke. Acta Med
Scand. 1983;214(1):3-13.
9. Simel DL, Rennie D, Bossuyt PM. The STARD state-
ment for reporting diagnostic accuracy studies: ap-
plication to the history and physical examination. J Gen
Intern Med. 2008;23(6):768-774.
10. Besson G, Robert C, Hommel M, Perret J. Is it clini-
cally possible to distinguish nonhemorrhagic infarct
from hemorrhagic stroke? Stroke. 1995;26(7):
1205-1209.
11. Bogousslavsky J, Van Melle G, Regli F. The Lau-
sanne Stroke Registry: analysis of 1000 consecutive
patients with first stroke. Stroke. 1988;19(9):1083-
1092.
12. Hui AC, Man CY, Tang AS, Au-Yeung KM. Com-
puted tomography evaluation in acute stroke: retro-
spective study. Hong Kong Med J. 2002;8(3):177-
180.
13. Kolapo KO, Ogun SA, Danesi MA, Osalusi BS,
Odusote KA. Validation study of the Siriraj stroke score
in African Nigerians and evaluation of the discrimi-
nant values of its parameters: a preliminary prospec-
tive CT scan study. Stroke. 2006;37(8):1997-2000.
14. Mader TJ, Mandel A. A new clinical scoring sys-
tem fails to differentiate hemorrhagic from ischemic
stroke when used in the acute care setting. J Emerg
Med. 1998;16(1):9-13.
15. Ng PW, Sha KY. Clinical stroke scores in Chinese
population. Stroke. 1996;27(4):773-774.
16. Nouira S, Boukef R, Bouida W, et al. Accuracy of
two scores in the diagnosis of stroke subtype in a mul-
ticenter cohort study. Ann Emerg Med. 2009;53
(3):373-378.
17. Poungvarin N, Viriyavejakul A, Komontri C. Siri-
raj stroke score and validation study to distinguish su-
pratentorial intracerebral haemorrhage from infarction.
BMJ. 1991;302(6792):1565-1567.
18. Shah FU, Salih M, Saeed MA, Tariq M. Validity
of Siriraj stroke scoring. J Coll Physicians Surg Pak.
2003;13(7):391-393.
19. Sackett DL, Haynes RB, Tugwell P. Clinical Epi-
demiology: A Basic Science for Clinical Medicine.
Boston, MA: Little Brown; 2006.
20. DerSimonian R, Laird N. Meta-analysis in clinical
trials. Control Clin Trials. 1986;7(3):177-188.
21. McGee S. Simplifying likelihood ratios. J Gen In-
tern Med. 2002;17(8):646-649.
22. Woisetschläger C, Kittler H, Oschatz E, et al. Out-
of-hospital diagnosis of cerebral infarction versus in-
tracranial hemorrhage. Intensive Care Med. 2000;
26(10):1561-1565.
©2010 American Medical Association. All rights reserved.
23. Panzer RJ, Feibel JH, Barker WH, Griner PF. Pre-
dicting the likelihood of hemorrhage in patients
with stroke. Arch Intern Med. 1985;145(10):1800-
1803.
24. Allen CM. Clinical diagnosis of the acute stroke
syndrome. Q J Med. 1983;52(208):515-523.
25. Efstathiou SP, Tsioulos DI, Zacharos ID, et al. A
new classification tool for clinical differentiation be-
tween haemorrhagic and ischaemic stroke. J Intern
Med. 2002;252(2):121-129.
26. Badam P, Solao V, Pai M, Kalantri SP. Poor ac-
curacy of the Siriraj and Guy’s Hospital stroke scores
in distinguishing haemorrhagic from ischaemic stroke
in a rural, tertiary care hospital. Natl Med J India. 2003;
16(1):8-12.
27. Hui AC, Wu B, Tang AS, Kay R. Lack of clinical
utility of the Siriraj stroke score. Intern Med J. 2002;
32(7):311-314.
28. Ozeren A, Bicakci S, Burgut R, Sarica Y, Bozdemir
H. Accuracy of bedside diagnosis vs Allen and Siriraj
stroke scores in Turkish patients. Eur J Neurol. 2006;
13(6):611-615.
29. Nyandaiti YW, Bwala SA. Validation study of the
Siriraj stroke score in north-east Nigeria. Niger J Clin
Pract. 2008;11(3):176-180.
30. Wadhwani J, Hussain R, Raman PG. Nature
of lesion in cerebrovascular stroke patients: clinical
stroke score and computed tomography scan brain
correlation. J Assoc Physicians India. 2002;50:
777-781.
31. Kochar DK, Joshi A, Agarwal N, Aseri S, Sharma
BV, Agarwal TD. Poor diagnostic accuracy and appli-
cability of Siriraj stroke score, Allen score and their com-
bination in differentiating acute haemorrhagic and
thrombotic stroke. J Assoc Physicians India. 2000;
48(6):584-588.
32. Kan CH, Lee SK, Low CS, Velusamy SS, Cheong
I. A validation study of the Siriraj stroke score. Int J
Clin Pract. 2000;54(10):645-646.
33. Ilić T, Jovicić A, Lepić T, Magdić B, Cirković S. Ce-
rebral infarction vs intracranial hemorrhage—validity
of clinical diagnosis. Vojnosanit Pregl. 1997;54
(4):299-309.
34. Daga MK, Sarin K, Negi VS. Comparison of Siri-
raj and Guy’s Hospital score to differentiate supra-
tentorial ischaemic and haemorrhagic strokes in the
Indian population. J Assoc Physicians India. 1994;
42(4):302-303.
35. Massaro AR, Sacco RL, Scaff M, Mohr JP. Clini-
cal discriminators between acute brain hemorrhage and
infarction: a practical score for early patient
identification. Arq Neuropsiquiatr. 2002;60(2-A):
185-191.
36. Fisher CM. Transient ischemic attacks. N Engl J
Med. 2002;347(21):1642-1643.
37. National Institute of Neurological Disorders and
Stroke rt-PA Stroke Study Group. Tissue plasmino-
gen activator for acute ischemic stroke. N Engl J Med.
1995;333(24):1581-1587.
38. Wolf PA, D’Agostino RB, Belanger AJ, Kannel WB.
Probability of stroke: a risk profile from the Framing-
ham Study. Stroke. 1991;22(3):312-318.