Is It Clinically Possible to Distinguish Nonhemorrhagic Infarct From Hemorrhagic Stroke?
Background and Purpose Diagnosis of the nonhemorrhagic ischemic type of stroke by analysis of patients’ clinical features is considered unreliable because no clinical feature is specific. The diagnosis is so difficult to establish that we cannot hope to use the same method to make a reliable diagnosis in all stroke cases. In this study, we propose a simple scoring system with a positive predictive value of close to 100% to distinguish nonhemorrhagic infarct from hemorrhagic stroke. This scoring is available for all physicians in bedside diagnosis even if this score can be applied to a subgroup of patients.
Methods Twenty-six clinical variables that might potentially distinguish cerebral hemorrhage from infarction were recorded in patients consecutively admitted to our stroke unit for stroke lasting more than 24 hours with at least unilateral motor weakness affecting face and/or arm and/or leg (internal validity study). Patients previously receiving anticoagulant therapy were excluded. We used CT scan as the gold standard. We used multivariate logistic regression to establish a clinical score from which we derived the classification rule. This rule was validated with data from the next 200 consecutive patients hospitalized in the stroke unit (external validity study).
Results Three hundred sixty-eight patients were enrolled in the internal study. The obtained score was (2×alcohol consumption)+(1.5×plantar response)+(3×headache)+(3×history of hypertension)−(5×history of transient neurological deficit)−(2×peripheral arterial disease)−(1.5×history of hyperlipidemia)−(2.5×atrial fibrillation on admission). All patients with a score less than 1 (n=123) had a nonhemorrhagic infarct (ie, 40% of the 305 patients with a nonhemorrhagic infarct). No threshold was found to diagnose cerebral hemorrhage with a sufficiently high positive predictive value. Among the 200 patients enrolled in the external validity study, 72 patients with a score below 1 had a nonhemorrhagic infarct (ie, 43% of patients with a nonhemorrhagic infarct).
Conclusions Diagnosis of nonhemorrhagic infarct can be made in 36% (95% confidence interval [CI], 29 to 43) of patients with a high level of accuracy (100% in the external validity study, which gives a 95% CI of 93 to 100). Thus, 43% (95% CI, 36 to 50) of patients with a nonhemorrhagic infarct could receive a bedside diagnosis. The score is simple and can be calculated from information available to all physicians.
The management of a patient with acute stroke is based on the knowledge of stroke type: hemorrhagic or ischemic. In most developed countries, diagnosis is easily obtained by CT scanning, which allows the accurate distinction of hemorrhagic and ischemic types. However, quick access to CT scanning is not available in every country and hospital. It is well known that some clinical data may suggest a hemorrhagic or ischemic stroke even though no data are specific enough to allow a reliable diagnosis. In a postmortem study including 51 cases of cerebral hemorrhage and 35 cases of cerebral infarction, Schaafsma1 considered sudden onset, decerebrate spasms, hypothermia, and bloody cerebrospinal fluid as unequivocal criteria for cerebral hemorrhage, whereas only a history of previous stroke was considered as an unequivocal criterion for cerebral infarction. Clinical data recorded in stroke data banks have also shown that several risk factors may be more evocative of hemorrhagic or ischemic stroke type. Thus, hypertension was considered as an important risk factor for cerebral hemorrhage.2 3 In these studies, bivariate statistical analysis was used; currently the role of each risk factor remains controversial.
Scoring systems based on clinical data determining the relative likelihood of infarction or hemorrhage have been formulated and tested.4 5 They use multivariate discriminant analysis to generate a linear equation that predicts the occurrence of ischemic or hemorrhagic stroke. Although the clinical diagnoses made using these scores seem more accurate than those made by physicians,4 they present several problems.6 The Allen score requires data collected 24 hours after admission, such as level of consciousness and diastolic blood pressure, and must be calculated with a handheld calculator. Furthermore, the Siriraj score, which also includes the level of consciousness and the diastolic blood pressure, and the Allen score do not achieve a diagnosis with a positive predictive value of close to 100%.
From all these studies, it appeared that a single score established from clinical findings that could allow reliable diagnosis for all cases did not exist. Because approximately 80% of strokes are ischemic, the reliability of a diagnostic process involving a score has to be greater than 80%, which is reached by classifying all patients in the ischemic group.
In this study, hemorrhagic infarcts were pooled with hematoma because the aim was to predict the absence of blood at CT. The purpose of this study was to build a classification rule from a score as follows: if the value is inferior to a threshold, then we classify the case as a nonhemorrhagic infarct; if the value is superior to the same threshold, then the diagnosis remains unknown. The score has to be simple and has to be calculated from information available to all physicians at the patient’s bedside, in the emergency department, during ambulance service, or at home within a few minutes or hours of the onset of stroke. Thus, this score should display a subgroup whose elements can be classified in one of the diagnostic groups with a positive predictive value of close to 100%. We did not aim to diagnose the type of stroke in all cases and thus paid no attention to the the negative predictive value.
Subjects and Methods
In the first stage, we selected variables from data on a selected population to generate a score with an empirical positive predictive value of 100% in the diagnosis of nonhemorrhagic infarct in this population (internal validity study). In the second stage, we applied this score to a second population to validate it and to compute a confidence interval (CI) of the positive predictive value in the diagnosis of nonhemorrhagic infarct.
Internal Validity Study
Between November 1990 and July 1992, all patients admitted to the stroke unit for stroke lasting more than 24 hours with at least unilateral motor weakness affecting face and/or arm and/or leg were included. Patients with bilateral motor weakness and patients previously receiving anticoagulant therapy were excluded. The type of stroke was confirmed in all patients by CT scan performed within 24 hours of onset. All patients had a 12-lead electrocardiogram and standard blood tests on admission.
Twenty-six variables including age and sex were prospectively recorded during the time of hospitalization. Unknown and unassessable items were scored as absent because multivariate analysis would not function with missing values. Moreover, scoring unknown and unassessable variables as absent was more discriminative than creating a separate score category. Moreover, in some cases patients with unknown variables were aphasic or comatose, and these variables account for the information. The rating of the items is reported in the “Appendix.”
All data analyses were performed with the Statistical Package for the Social Sciences (spss).7 The CI for the positive predictive value was read with Abacus.8 We used Yates’ corrected χ2 test, t test for equality of means, and multivariate logistic regression. The level of statistical significance was defined as 5%. The dichotomous dependent variable was the type of stroke (hemorrhagic or nonhemorrhagic). Hemorrhagic infarcts were pooled with hematoma because the aim of the study was to predict the absence of blood at CT. A forward stepwise analysis with a likelihood ratio criterion was performed in the logistic regression to select variables. The score (Z) may be written as Z=B0+B1X1+. . .+BpXp, where B0 is a constant, B1 is the coefficient estimated from the data, and X1 is the independent variable. In the logistic regression model, the probability of a hemorrhagic stroke knowing the values of X1, . . . , Xp is expressed as e−Z/1+e−Z. A histogram of the score was drawn up to define a cutoff point.
External Validity Study
For the external study, the scoring was applied to 200 consecutive patients hospitalized from July 1992 through July 1993 in the stroke unit. The variables selected by logistic regression were prospectively recorded at entrance into the stroke unit, and the score was calculated. The data recorded at entrance into the stroke unit is comparable with that recorded at the bedside in the patient’s home.
Internal Validity Study
Among 408 patients admitted to our stroke unit for stroke lasting more than 24 hours, 368 patients (90%) satisfying the inclusion criteria were enrolled in the internal validity study. There were 209 men (56.8%) and 159 women (43.2%). They were aged from 21 to 97 years (mean±SD, 67.95±15.84 years; median, 72 years). The diagnosis was nonhemorrhagic infarct in 305 patients (82.9%; 95% CI, 79.1 to 86.7) and cerebral hemorrhage in 63 patients (17.1%; 95% CI, 13.3 to 20.9). Cerebral hemorrhage diagnoses consisted of 52 cerebral hematomas and 11 hemorrhagic infarcts.
The bivariate analysis of all qualitative variables is shown in Table 1⇓. The t test for age showed no difference for nonhemorrhagic infarct (68.44±0.92 years) and cerebral hemorrhage (65.59±1.90 years) (t value, 1.3).
The forward stepwise analysis selected nine variables to distinguish cerebral hemorrhage from nonhemorrhagic infarct. The equation was Z=(0.73×alcohol consumption)+(0.63×plantar response)+(1.90×headache)+(1.30×deviation of head and eyes)+(1.20×history of hypertension)−(2.00×history of transient neurological deficit)−(1.10×peripheral arterial disease)−(0.71×history of hyperlipidemia)−(1.58×atrial fibrillation on admission)−3.58.
The analysis of the Z values showed that no cerebral hemorrhage had a Z value below −3.41. Thus, in the population at hand, the empirical positive predictive value in the diagnosis of nonhemorrhagic infarct was 100% for Z less than −3.41. It concerned 95 patients with a nonhemorrhagic infarct (ie, 31% of all the patients with an infarct). No threshold was found to define a positive predictive value of close to 100% in the diagnosis of cerebral hemorrhage.
Since we wanted to distinguish “typical” nonhemorrhagic infarcts from cerebral hemorrhages, we first needed to define “typical” nonhemorrhagic infarcts, which was done in the first step. Thus, in the second step, considering the 95 cases as typical ones, we tried to distinguish them from the cerebral hemorrhages. We pooled the 63 patients with cerebral hemorrhage and the 95 patients with a Z value lower than −3.41. A forward stepwise logistic regression was performed for these 158 patients to try to find a more relevant score that would be able to distinguish these typical nonhemorrhagic infarcts from cerebral hemorrhages. Eight variables were selected to distinguish nonhemorrhagic infarct from cerebral hemorrhage. All these variables were previously selected in the first equation. The original equation was Z=(47.13×alcohol consumption)+(35.23×plantar response)+(71.01×headache)+(70.02× history of hypertension)−(116.36×history of transient neurological deficit)−(48.10×peripheral arterial disease)−(34.84×history of hyperlipidemia)−(59.55×atrial fibrillation on admission)−23.90.
The analysis of the Z values showed that none of the cerebral hemorrhages had a Z value lower than −1.58. On the other hand, no nonhemorrhagic infarction had a Z value above 10. A simplified equation was obtained by dividing each variable by the constant, followed by rounding off the coefficients. The equation was simplified to allow it to be applied at the bedside without the use of a calculator. Thus, we have defined a final score for the diagnosis of nonhemorrhagic infarct as (2×alcohol consumption)+(1.5×plantar response)+(3×headache)+(3×history of hypertension)−(5×history of transient neurological deficit)−(2×peripheral arterial disease)−(1.5×history of hyperlipidemia)−(2.5×atrial fibrillation on admission) <1.
We applied this score to the 368 patients. The empirical positive predictive value remained 100% in the diagnosis of nonhemorrhagic infarct (Fig 1⇓). This concerned 123 patients with a nonhemorrhagic infarct (40% of the patients with a nonhemorrhagic infarct and 33% of all the patients).
The bivariate analysis of the eight variables (Table 1⇑) shows that a history of hypertension and headache was significantly more frequent in patients with a hemorrhagic stroke, whereas history of transient neurological deficit, history of hyperlipidemia, and atrial fibrillation on admission were significantly more frequent in patients with a nonhemorrhagic ischemic stroke.
External Validity Study
There were 122 men (61%) and 78 women (39%) aged from 27 to 95 years (69.7±14.49 years; median, 72 years). The diagnosis on CT scan was cerebral hematoma in 31 patients (15.5%; 95% CI, 10.5 to 21.0), hemorrhagic infarct in 2 patients (1%; 95% CI, 0 to 2.4), and nonhemorrhagic infarct in 167 patients (83.5%; 95% CI, 78.4 to 88.6). We applied the scoring to the 200 patients. Seventy-two patients had a score strictly less than 1 (43% [95% CI, 36 to 50] of all the patients with nonhemorrhagic infarct and 36% [95% CI, 29 to 43] of all the patients) (Fig 2⇓). All these patients had a nonhemorrhagic infarct on CT. Thus, the empirical positive predictive value in the diagnosis of nonhemorrhagic infarct was 100%, yielding a 95% CI of 93 to 100.
The analysis of the eight variables (Table 2⇓) shows no difference between the two groups except for headache, which remained significantly more frequent in patients with hemorrhagic stroke. The comparison of the eight variables between the internal and the external validity studies shows no difference except for peripheral arterial disease, which was significantly more frequent in the internal validity study (P=.02).
Clinical diagnosis of stroke type at the bedside is inaccurate. However, using our scoring system, we can affirm diagnosis of nonhemorrhagic infarct in 43% of patients with a nonhemorrhagic infarct and in 36% of all the patients in our population. We know that a positive predictive value of 100% cannot be proved. However, with an external validity study including 200 patients, we can conclude that the positive predictive value was between 93% and 100% with a 95% CI. In our scoring system, the presence of a history of transient neurological deficit, peripheral arterial disease, history of hyperlipidemia, and atrial fibrillation on admission leads toward the diagnosis of nonhemorrhagic infarct, which is in accordance with registry data.2 3 Conversely, the presence of alcohol consumption, plantar response type, a history of hypertension, and headache leads toward an unknown diagnosis. When these variables are known to be present, it is more difficult to differentiate nonhemorrhagic infarct from cerebral hemorrhage. This is also in accordance with previous studies that have shown that headache, hypertension, and alcohol consumption are frequently found in patients with cerebral hemorrhage.9 However, the overall importance of these variables studied separately is slight for the diagnosis of cerebral hemorrhage.
In bivariate analysis, the population in the internal validity study does not differ from the population of the external validity study, except for peripheral arterial disease. In the external validity study, although there is no major difference between the eight selected variables for hemorrhagic strokes and nonhemorrhagic infarcts using bivariate analysis (Table 2⇑), the score shows that a subgroup of nonhemorrhagic infarcts may be isolated from all the other strokes. Moreover, although there were more unknown values in the external validity study, we might still affirm the diagnosis of nonhemorrhagic infarct in 43% of patients with a nonhemorrhagic infarct. The Allen score has been evaluated in 228 patients among 323 consecutive patients from the Oxfordshire Community Stroke Project and in 130 patients from the National Hospital for Nervous Diseases.10 Using a validating study of the Siriraj score, the empirical positive predictive value could be 100% for only 1 of 64 patients with cerebral infarct for a score above 6 and for 29 of 142 patients with cerebral hemorrhage for a score below −7.5 Thus, the “typical” cases (those for which a reliable diagnosis can be made) concerned 30 patients of 206 (14.6%), which is fewer than found with our score. A recent study including 231 patients has shown that the positive predictive values of the Allen and the Siriraj scores were 91% and 93%, respectively.11 However, these positive predictive values were calculated in a subgroup of patients because the Allen score was uncertain in 19% of cases and the Siriraj score in 16%. Moreover, another recent study including 1059 patients concluded that none of these scores are useful in excluding hemorrhagic stroke.6 Nevertheless, it is difficult to compare our score with the Allen and the Siriraj scores because all three scores have not been applied to the same group of patients.
When using the Allen score, one of the main problems in the misdiagnosis of intracranial hemorrhages is the association between intracranial hemorrhages and atrial fibrillation.10 Four patients with atrial fibrillation had an intracranial hemorrhage, and two of them would have been misdiagnosed as having cerebral infarct. In the internal validity study, five patients showed the association between intracranial hemorrhages and atrial fibrillation, and no one was misdiagnosed. In the external validity study, no cerebral hemorrhage had this association.
Our score is simple and can be calculated from information available to all physicians at the patient’s bedside at home. It seems to give a reliable diagnosis in about 36% of selected stroke patients. While it cannot replace CT in the management of all strokes, this score may be very useful for patients who do not undergo CT, as in nondeveloped countries.5 In addition, this score may be useful in the very early management of acute ischemic strokes in future stroke trials.
History of stroke, history of transient neurological deficit (including amaurosis fugax),12 history of ischemic heart disease, history of atrial fibrillation, history of hypertension, history of diabetes mellitus, history of hyperlipidemia, and history of current cigarette smoking were scored as 0 if absent or unknown and 1 if present.
Alcohol consumption was scored as 0 if absent or unknown and 1 if the patient drank alcohol every day, whatever the amount. The use of contraceptive pills was scored as 0 if absent or unknown and 1 if present.
Clinical Signs and Symptoms
Peripheral arterial disease was scored as 0 if absent and 1 when the patient had a documented history of lower-limb claudication or if physical examination showed the loss of at least one arterial ankle pulse. Vomiting, headache within 2 hours before onset, and/or headache after onset were scored as 0 if absent or unknown and as 1 if present. We used the eye-opening and motor-response items of the Glasgow Outcome Scale (GOS) score.13 The eye opening was scored as 0 when the GOS score was below 3 and 1 when the GOS score was 3 or 4. Motor response was scored as 0 when the GOS score was below 5 and 1 when the GOS score was 5 or 6. The onset of deficit was scored as 0 if it was unknown or if it was discovered on waking, 1 if it was acute, and 2 if it appeared progressively. Partial motor weakness was scored as 0, and hemiplegia or hemiparesis was scored as 1. Partial sensory deficit was scored as 0, and hemianesthesia or hemihypoesthesia was scored as 1. Deviation of head and eyes was scored as 0 if absent and 1 if present. Plantar responses were scored as 0 if absent, 1 if extensor ipsilaterally to the deficit, 2 if extensor contralaterally to the deficit, and 3 if they were both extensor.
Electrocardiogram on Admission
Atrial fibrillation was scored as 0 if absent and 1 if present, myocardial infarction was scored as 0 if absent and 1 if present, atrioventricular block was scored as 0 if absent and 1 if present, and left ventricular hypertrophy was scored as 0 if absent and 1 if present.
- Received October 21, 1994.
- Revision received April 6, 1995.
- Accepted April 7, 1995.
- Copyright © 1995 by American Heart Association
Mohr JP, Caplan LR, Melski JW, Goldstein RJ, Duncan GW, Kistler JP, Pessin MS, Bleich HL. The Harvard Cooperative Stroke Registry: a prospective registry. Neurology. 1978;28:754-762.
Bogousslavsky J, Van Melle G, Regli F. The Lausanne stroke registry: analysis of 1000 consecutive patients with first stroke. Stroke. 1988;19:1083-1092.
Allen CMC. Clinical diagnosis of the acute stroke syndrome. Q J Med. 1983;52:515-523.
Poungvarin N, Viriyavejakul A, Komontri C. Siriraj stroke score and validation study to distinguish supratentorial intracerebral haemorrhage from infarction. BMJ. 1991;302:1565-1567.
SPSS. Statistical Package for the Social Sciences. Windows version, release 5.0. Chicago, Ill: SPSS Inc; 1992.
Anonymous. Abaque 2.1. Rev Stat Appl. 1973;1:28.
Kase CS, Mohr JP, Caplan LR. Intracerebral hemorrhage. In: Barnett HJM, Mohr JP, Stein BM, Yatsu FM, eds. Stroke: Pathophysiology, Diagnosis, and Management. New York, NY: Churchill Livingstone; 1992:561-616.
Sandercock PAG, Allen CMC, Corston RN, Harrison MJG, Warlow CP. Clinical diagnosis of intracranial haemorrhage using Guy’s hospital score. BMJ. 1985;291:1675-1677.
Celani MG, Righetti E, Migliacci R, Zamppolini M, Antoniutti L, Grandi FC, Ricci S. Comparability and validity of two clinical scores in the early differential diagnosis of acute stroke. BMJ. 1994;308:1674-1676.
Ad Hoc Committee for Stroke. A classification and outline of cerebrovascular disease. Stroke. 1975;6:564-616.