The Rational Clinical Examination Section Editors: David L. Simel, MD, MHS, Durham Veterans Affairs Medical Center and
Duke University Medical Center, Durham, NC; Drummond Rennie, MD, Deputy Editor, JAMA.
Author Affiliations: Department of Medicine
(Neurology), Duke Center for Cerebrovascular Disease (Dr Goldstein); Center
for Clinical Health Policy Research and Education, Duke University, and Department
of Veterans Affairs Medical Center (Drs Goldstein and Simel); Department of
Medicine, Duke University (Dr Simel), Durham, NC.
Context Patients suspected of having a stroke or transient ischemic attack require
accurate assessment for appropriate acute treatment and use of secondary preventive
Objective To update a 1994 systematic review of the accuracy and reliability of
symptoms and findings on neurological examination for the evaluation of patients
with suspected stroke or transient ischemic attack.
Data Sources We identified potential articles dated between 1994 and 2005 by multiple
search strategies of the MEDLINE database and review of article and textbook
bibliographies along with private collections.
Study Selection Selected articles provided primary data or appropriate summary statistics
of the accuracy and/or reliability of the history or physical examination
for diagnosis or short-term prognosis of patients with neurological signs
prompting a consideration of stroke. Articles addressing accuracy also needed
to provide a final diagnosis following neuroimaging and all relevant laboratory
Data Extraction The authors reviewed and abstracted data for estimating sensitivities,
specificities, positive and negative likelihood ratios (LRs). Reliability
assessment was based on reported kappa (κ) statistics or intraclass
correlation coefficients as appropriate.
Data Synthesis The prior probability of a stroke among patients with neurologically
relevant symptoms is 10%. Based on studies using modern neuroimaging, the
presence of acute facial paresis, arm drift, or abnormal speech increases
the likelihood of stroke (LR of ≥1 finding = 5.5; 95% CI, 3.3-9.1),
while the absence of all 3 decreases the odds (LR of 0 findings = 0.39;
95% CI, 0.25-0.61). The accurate determination of stroke subtype requires
neuroimaging to distinguish ischemic from hemorrhagic stroke. Early mortality
increases among those with any combination of impaired consciousness, hemiplegia,
and conjugate gaze palsy (LR of ≥1 finding = 1.8; 95% CI, 1.2-2.8
and LR of 0 findings = 0.36; 95% CI, 0.13-1.0). Symptoms associated
with high agreement for the diagnosis of stroke or transient ischemic attack
vs no vascular event are a sudden change in speech, visual loss, diplopia,
numbness or tingling, paralysis or weakness, and non-orthostatic dizziness
(κ = 0.60; 95% CI, 0.52-0.68). The reliabilities of individual
neurological findings vary from slight to almost perfect, but can be improved
with standardized scoring systems such as the National Institutes of Health
Stroke Scale. Based on examination findings, stroke vascular distribution
can be determined with moderate to good reliability (κ = 0.54;
95% CI, 0.39-0.68).
Conclusions The history and clinical findings provide the basis for evaluating patients
with possible stroke and choosing appropriate treatments. Focusing on 3 findings
(acute facial paresis, arm drift, or abnormal speech) might improve diagnostic
accuracy and reliability.
Goldstein LB, Simel DL. Is This Patient Having a Stroke?. JAMA. 2005;293(19):2391-2402. doi:10.1001/jama.293.19.2391