Clinical Scales to Assess Patients With Stroke

The author acknowledges the contributions to the revision of this chapter by Liping Liu, MD, PhD, and Weiqi Chen, MD, PhD, of the Department of Neurology, China National Clinical Research Center for Neurological Diseases, Beijing, China.

Desired Qualities of Stroke Scales

Most of the currently used scales were developed to describe subjects enrolled in clinical trials, but their use has been expanded to a spectrum of clinical settings. The result or score of a stroke scale should be clinically relevant and should be obvious to the healthcare providers that will use the scale (face validity). The aim of the scale should be clear. For a stroke scale to be useful, it should be able to provide a numerical score of some other categorization that is clinically relevant. The results should provide a clear impression that provides a mental image of the patient’s status to a clinician. The derived information may affect decisions in diagnosis, treatment, or counseling. The purpose of the scale and the proper time for its use should be obvious. For example, using an outcome scale such as the modified Rankin Scale (mRS) at the time of admission for treatment of a stroke is not appropriate because the scale emphasizes functional outcomes that cannot be accurately assessed in the setting of an acute stroke. ^, Some scales provide quantitative items that may be calculated with scores of individual items added to form a total scale; the National Institutes of Health stroke scale (NIHSS) is a widely used example. Some scores involve the addition of points to achieve a total score. In other instances, the score entails subtraction from an initial baseline score; in such a case a low score reflects a serious brain injury. In either instance, the total score of a numerical system is important because it usually provides information about prognosis, which, in turn, may affect decisions about treatment. Thus, the clinician needs to understand the scoring system that is used for each individual rating instrument.

Scales should be examined in a manner that is similar to that used to validate a diagnostic test because, in fact, a stroke scale is an ancillary instrument even if it is based on clinical findings. The scale should be sensitive; it should be able to detect those findings that are of most interest. The scale should be specific; it should permit recognition and scoring of only those abnormalities that are important. As much as possible, the scale should not have high rates of false-positive or false-negative results that would affect patient care. In summary, if possible, both the positive and negative predictive values of the scale should be determined with comparisons to a standard. These features are especially important for those clinical scales that are used to differentiate stroke from other acute neurologic illnesses, hemorrhagic from ischemic stroke, or the subtype of ischemic stroke. The standards to which these diagnoses are compared include subsequent clinical diagnoses, outcomes, and the results of diagnostic studies such as brain imaging. In summary, a scale must be accurate. Unfortunately, some scales, such as those differentiating hemorrhagic from ischemic stroke, have not met these criteria.

There is no single scale that provides information about the gamut of the clinical aspects of stroke. In most circumstances, patients are assessed by combinations of rating instruments that are performed at different times. All the scales have individual strengths and weaknesses. For example, the widely used NIHSS has a bias toward higher scores being calculated with strokes affecting the left hemisphere. Practitioners need to know the limitations and idiosyncrasies of each scale.

A clinically used scale should be easy to perform and germane to the clinical situation. In addition, scales that are used in an emergency situation should be able to be performed rapidly. Some rating instruments use components based on history obtained from the patient or observers; information may reflect performance, demographic findings (i.e., age), or the presence of risk factors for stroke or accelerated atherosclerosis. Other scales are based solely on findings detected by direct physical examination and place an emphasis on the neurologic assessment. Many scales, especially those used to rate the severity of neurologic impairments, include gradations in scoring, for example, different scores for the severity of motor signs or levels of consciousness. Other scales use a numeric system of rating, but these numbers are not based on computation of points from items contained within the scale. For example, the Hunt-Hess Scale for subarachnoid hemorrhage (SAH) has five defined grades that do not include scoring of components to reach that grade.

Weighting of different items is included in some scales. For example, in the Canadian Stroke (Neurological) Scale (CNS) the item scoring consciousness is given more points than individual items rating language or arm motor function. ^, An even more elaborate system (the Japan Stroke Scale) was developed by Gotoh et al.; they selected 10 variables ranging from consciousness to pupillary changes. The weighted factors ranged from consciousness (49.8%) to plantar reflex (2.2%) to sensory impairment (2.1%). The utility of this very complex approach that results in a wide range of scores is not established. Regardless of the rating instrument, the bottom line for the success of the scale is its credibility. Healthcare providers need to recognize the utility of the scale in their assessment and care of patients with stroke. A few scales that have been used for several years have considerable cachet. For example, the Glasgow Coma Scale (GCS), which was originally developed to define the severity of brain injury among patients with craniocerebral trauma, has become a worldwide standard for assessing a wide variety of critically ill patients with impairments of consciousness including those with stroke.

A valid scale must have attributes of strong inter-rater agreement and intra-rater reproducibility in scoring. These features are especially important for scales that describe subtypes of stroke, that measure the severity of neurologic impairments, or that rate outcomes. The goal is to achieve an accurate measure of the patient’s condition. A lack of agreement in diagnoses or assessments of the clinical findings is a potential disaster for multicenter clinical trials and also weakens the applicability of the data obtained from these trials. For example, during the development of the Trial of Org 10172 in Acute Stroke Treatment (TOAST) classification, the researchers found that physicians often disagreed about the subtype of ischemic stroke despite being presented by the same information. Agreement was very high when assessing a straightforward case, but was disappointingly low for those cases that had multiple potential explanations or when important supporting data were missing. Thus, the researchers found that the κ statistic (the degree of improvement about chance) was reasonably good for some subtypes of stroke but less acceptable for others. Measures to improve the agreement of the scales can be undertaken, but both researchers and clinicians should recognize that no scale will ever achieve unanimity or perfection when a complex, multifactorial disease such as stroke is being evaluated.

To improve agreement and reproducibility, developers of scales use arbitrary rules to define the various grades of scored items. For instruments that are based on historical information, specific questions to ask and specific answers that are sought are explained. For those scales based primarily on findings on an examination, the steps in the performance of the assessment and the methods to rate the findings are described. Some scales include scoring responses for contingencies such as an absent limb or the presence of a severe comorbid disease. The scales that perform the best included detailed definitions for all the potential scores for items that are being assessed. Still, even these definitions are open to interpretation by the rater, and they may not address the wide range of potential scenarios that may be found in a large group of patients with stroke. As an additional step to improve validity, researchers and other groups created programs to train clinicians on the use of several scales. The programs usually are supplemented by a certification process that tests the clinician’s ability to accurately use the rating instrument. Because the mRS and the NIHSS are the most widely used rating instruments in clinical trials, they have the most extensive educational and certification programs, which have become a critical quality component of clinical trials.

Because of marked differences between measurements of scales and among investigators in multicenter studies, some trials used central adjudication or scoring as an additional quality control measure. This approach is most commonly used to assess outcome data. The relevant information is sent to an individual or panel that reviews a videotape of the patient’s performance. This tactic improves consistency of scoring in the trial and may be accompanied by a reduction in the size of the trial that usually would be needed to compensate for differences in scoring among raters.

Scales Used by Emergency Medical Services

The GCS was developed to quantify the severity of neurologic impairments with head injuries ( Table 21.1 ). Potential scores range from 3 to 15 points and are based on the patient’s best verbal response, best motor response, and eye movement. For example, a patient with hemiparetic but volitional movement on the left and flexor posturing on the right would be rated as having 5 or 6 points on the motor item. In general, patients with scores less than 8 have a very serious brain injury and a poor prognosis. Although the GCS has not been tested for inter-rater agreement and intra-rater reproducibility, it has strong predictive power. It is used widely by EMS personnel, physicians, and other healthcare providers. The rating system is most useful in assessment of patients with alterations in consciousness. Thus, its value generally is greater among persons with hemorrhagic stroke than among those with cerebral infarction. Although the GCS usually is not helpful in evaluation of cases of suspected ischemic stroke because consciousness usually is not disturbed, it may be a strong predictor of outcomes among patients with severe strokes affecting the brainstem. A pared down (“slim”) version of the GCS has been developed to expedite its use by non-physicians; unfortunately, it has not been found to be useful. The World Federation of Neurological Surgeons (WFNS) Scale for measuring the severity of SAH is a derivative of the GCS. ^, The GCS score also is included in scales to rate the severity of intracerebral hemorrhage (ICH). Overall, the GCS continues to be an important component of assessment of patients in multiple care settings. The Japan Coma Scale also may be used to assess patients with impaired consciousness due to a variety of neurologic illnesses including SAH.

Rating systems are used by EMS personnel to determine if it is likely that a patient has had a stroke. A variation of the Cincinnati Pre-Hospital Stroke Scale (CPHSS) called the FAST (Face, Arm, and Speech Test) scale is used in the UK. Another derivative of the CPHSS is the Medic Prehospital Assessment for Code Stroke; it contains all the components of the CPHSS but adds assessments of gaze and leg function. Besides screening the patient at the scene, these scales are used to transmit information to a hospital emergency department. The Miami Emergency Neurologic Deficit (MEND) scale has two versions; one involves a brief examination that is performed on the scene, and the second includes a more extensive examination that is done in the ambulance as the patient is being transported to the hospital. The scales include a limited number of findings obtained by history or examination. The MEND scale includes information about severe headache or a stiff neck, information that points toward an intracranial hemorrhage. The MEND scale also includes questions that are directly related to possible exclusions for treatment with recombinant tissue plasminogen activator (rtPA) including the use of medications such as warfarin. Screening for marked disturbances in blood glucose levels (hyperglycemia or hypoglycemia) is used to detect these possible causes of acute neurologic signs that could mimic a stroke. The most commonly assessed features of the emergency scales are language, facial weakness, and arm weakness, which is based on the assumption that unilateral weakness or speech/language trouble points to a stroke ( Tables 21.2 and 21.3 ; Box 21.2 ). These scales can be rated quickly with a reasonable degree of accuracy. Kothari et al. compared the scores obtained by paramedics using the CPHSS with those obtained by physicians using the NIHSS. They found that an abnormality in any of the three scale items was associated with 66% sensitivity and 87% specificity for identifying a stroke; the sensitivity was much higher for vascular events in the anterior circulation than for strokes in the posterior circulation. The diagnoses of paramedics using the FAST scale were compared with diagnoses by vascular neurologists; agreement was best for detection of arm weakness, but acceptable levels of concurrence were noted for both facial weakness and speech disturbance. These scales have been validated by in-field tests, and educational programs are available to train EMS personnel. In another study, Studnek et al. compared the CPHSS with the Medica Prehospital for Code Stroke system; they compared assessments by EMS personnel with hospital discharge diagnoses of stroke. While the CPHSS was the more sensitive instrument, they concluded that the specificity was marginal. The CPHSS probably is the most widely used scale in the US. Its advantages are its ease of use and limited information that is required. The MEND scale does require a reasonably high level of sophistication, which may limit its use to only highly trained paramedics and EMS specialists. This scale may not be as applicable for use by volunteer community services that often exist in smaller communities or rural areas. Much of the information in MEND is similar to that collected as part of the assessment for the NIHSS, and an alternative to MEND is to train EMS in the use of the NIHSS itself; for example, a potential scenario would be its use during air evacuation.

The Recognition of Stroke in the Emergency Room (ROSIER) scale was developed to assist emergency medicine physicians in their rapid diagnosis of stroke. The goal is to differentiate stroke from stroke mimics including seizures, syncope, or other acute neurologic illnesses and to facilitate the selection of patients who may be treated with emergency therapies such as intravenous rtPA. The scale includes seven items with a total score ranging from −2 to +5 ( Table 21.4 ). In general, a stroke is unlikely if the total score is less than 0; scores of 4 or 5 are strongly correlated with the diagnosis of stroke. The ROSIER scale was tested in a prospective study and showed a sensitivity of 93% (95% confidence interval [CI], 89%–97%), specificity of 83% (95% CI, 77%–89%), positive predictive value of 90% (95% CI, 85%–95%), and negative predictive value of 88% (95% CI, 83%–93%). Information about the utility and implementation of the ROSIER Scale is not available. Whiteley et al. compared the ROSIER scale with the FAST scale and found that the sensitivity and specificity of the two scales were similar. They concluded that the FAST system, which is quicker and easier to perform, probably is sufficient for screening in an emergency department. In another study that tested the utility of the FAST and ROSIER systems for diagnosis in stroke in children, the investigators concluded that these scales had reasonable sensitivity in pediatric patients. Singer et al. developed a simple system involving measurement of consciousness, gaze, and motor function. Each of the items is scored from 0 (normal) to 2 (severe). They found that scoring would predict an occlusion of the proximal middle cerebral artery with reasonable accuracy and that the tool could be used for triage of patients with suspected stroke.

Scales to Differentiate Hemorrhagic Stroke From Ischemic Stroke

Clinical rating instruments to distinguish hemorrhagic from ischemic stroke have been constructed to provide a diagnosis that would obviate the need for brain imaging before treatment, in particular emergency administration of intravenous rtPA. These scales usually include historical information, blood pressure measurements, and findings on the neurologic examination. These instruments have not reached adequate levels of either sensitivity or specificity for use in either clinical trials or patient care. As a result, the use of these scales should not be considered as a substitute for modern brain imaging, particularly if decisions about emergency treatment of stroke, such as the administration of intravenous rtPA, are under consideration.

Differentiation of Ischemic Stroke Syndromes

In ischemic stroke, the findings on neurologic examination generally reflect the location and size of the brain injury. Still, the clinical manifestations of stroke are diverse and often idio-syncratic to individual patients. Each patient has her or his own individual pattern of impairments. Still, there are features that are common to all and that may be aggregated and correlated with specific patterns or stroke syndromes. These general features may be classified into groupings that could be useful in addressing questions about prognosis, etiology of stroke, acute treatment, and long-term management. For example, most isolated infarctions affecting the cerebral cortex are secondary to branch occlusions from emboli that arise either in the heart or from proximal segments of major extracranial or intracranial arteries. Generally, patients with cortical infarctions have a good prognosis for survival. Those patients with small deep infarctions (lacunes) restricted to deep hemispheric structures (basal ganglia, thalamus, or internal capsule) usually have diseases of small penetrating arteries, and their acute prognosis for survival also is good. On the other hand, those patients with major hemispheric infarctions with involvement of both cortical and deep structures often have occlusions of major intracranial or extracranial arteries, in particular the middle cerebral artery or internal carotid artery. These patients are at greatest risk for malignant brain edema and death. The clinical features of ischemic strokes affecting the brainstem and cerebellum may differ from those seen with events affecting the cerebral hemispheres. The prognosis of patients with cerebellar and/or brainstem lesions generally reflects the extent of the ischemic lesion, which in turn is manifested by the pattern of neurologic impairments.

The Oxford Community Stroke Project (OCSP) Classification is the most widely used system to categorize patterns of ischemic strokes based on the types of neurologic impairments found on examination. It divides events into four groups: (1) total anterior circulation infarction (TACI)—a large cerebral infarction that usually is the result of an occlusion of a major intracranial or extracranial artery, (2) partial anterior circulation infarction (PACI)—smaller cerebral infarction that affects primarily the cortex and/or adjacent (lobar) white matter and that is usually due to embolic occlusion of a pial or cortical branch artery, (3) lacunar anterior circulation infarction (LACI)—a small infarction restricted to the structures deep in the cerebral hemisphere and secondary to occlusion of a small penetrating artery, and (4) posterior circulation infarction (POCI)—an infarction involving primarily the brainstem and/or cerebellum ( Box 21.3 ). There is some uncertainty about some of the definitions for the categories. For example, an isolated homonymous hemianopia, presumably due to embolic occlusion of the calcarine or posterior cerebral artery, is categorized as POCI because these arteries are usually derived from the basilar artery. It also could be considered to be a PACI event because it is restricted to the cerebral cortex and patients with this infarction are more likely to behave similarly to patients with anterior circulation events than those patients with brainstem strokes. The definition of pure motor hemiparesis for the LACI category does not include dysarthria while many patients with these strokes do have mild-to-moderate impairments in circulation. Although the OCSP was developed for differentiating patterns of cerebral infarction, Barber et al. used the system to define patterns of clinical findings among patients with ICH. This approach has not been implemented widely.

Lindley et al. tested the inter-rater agreement of the OCSP and found it was moderate-to-good (κ: 0.54; 95% CI, 0.39%–0.68%). Another group that tested the OCSP classification among nurses and physicians found only moderate agreement (95% CI, κ: 0.31%–0.45%). Because the features of the classification are straightforward and the number of categories is limited, the relatively low rate is surprising. Pittock et al. found that the primary utility of the OCSP was to differentiate the patients with TACI from the other three groups because those with TACI had poor outcomes. This finding is not unexpected given the extensive nature of the brain injury that occurs in this group of patients. The pattern diagnosis achieved by the OCSP was compared to the results of brain imaging in another study; the classification predicted the site of stroke in 80 of 91 patients (88%; 95% CI, 77%–92%). The system was most successful for cases of large hemispheric infarctions and did least well among small subcortical lesions. In a study using magnetic resonance imaging (MRI), Mead et al. found that the OCSP classification successfully predicted the site of the lesion. In another study that focused on classification when done in an emergency setting, Smith et al. found that the subtype diagnosis was correct in approximately 65% of cases. The sensitivity and specificity were highest for the POCI category (1.00 and 0.97, respectively) and lowest for the LACI category (0.33 and 0.88, respectively). While the early use of vascular and brain imaging may improve the accuracy of the classification system, it seems that the reliance on MRI or other imaging to make the OCSP subtype diagnosis partially defeats the original intent of the scale that was based solely on clinical findings. Although Zhang et al. found that narrowing of the extracranial segment of the internal carotid artery was correlated with the TACI category, a finding that is not surprising, the OCSP classification categories may not be associated with either the site or presence of a demonstrated occlusion of an intracranial artery.

Despite some limitations, the OCSP is widely used in both clinical trials and epidemiologic studies. It also is an excellent way to teach the basic principles of neurologic localization in stroke to medical students and other trainees. The OCSP system has advantages. It mimics the process that physicians use when they are assessing their patients. It is relatively simple, and it gives general information about localizing the stroke. The TACI category forecasts a poor prognosis. When compared to scales that quantify the severity of neurologic impairments following stroke, the OCSP classification is not as robust in forecasting outcomes. The OCSP does not predict the location of arterial pathology with sufficiently high frequency to be useful. It does not distinguish the etiology of stroke; for example, patients with the PACI pattern may have an embolus arising from either the heart or an extracranial artery and subsequent management of these underlying causes differs considerably.

Scales to Quantify the Severity of Intracerebral Hemorrhage

Scales to quantify the severity of ICH differ from most systems that rate the severity of SAH or ischemic stroke, in that they include both clinical information and the results of ancillary studies, usually brain imaging. Several scales are available and all have limitations. The Unified Neurological Stroke Scale was developed to be used in patients with either hemorrhagic or ischemic stroke; it incorporates components from the Scandinavian Stroke Scale (SSS) and the Middle Cerebral Artery Neurological Scale. It includes assessments of consciousness, language, eye movements, tone, and power of the face, arm, hand, leg, and foot. The ICH Scale is the most widely implemented instrument; it contains the following features: GCS score, age, infratentorial location of the hematoma, volume of the hemorrhage, and the presence of intraventricular bleeding ( Table 21.5 ). ^, The range of scores is 0–6. In one study, all patients with a score of 5 or 6 died and the mortality was 0 among those with a score of 0. This observation was confirmed in a study by Appelboom et al. who reported that the ICH score was strongly correlated with findings using the standard measures of stroke outcomes including the mRS. A variation of the ICH score involves the substitution of the GCS with the NIHSS score; in a comparison study, the modified scale was equal to the traditional ICH score in predicting mortality and was slightly better in forecasting favorable outcomes. Another variation of the instrument, called the ICH-GS, includes an expanded range of points and results in an improved ability to predict both mortality and functional outcomes. The severity of neurologic impairments after ICH also has been examined through the use of the CNS, SSS, or the NIHSS. ^{, ,} Smith et al. concluded that the NIHSS could be used effectively to predict in-hospital mortality among patients with ICH. Weimar et al. developed the Essen ICH score based on the patient’s age, the score on the consciousness item of the NIHSS, and an aggregate score based on the range of total NIHSS scores ( Table 21.6 ). The scores range from 0 to 10 with favorable outcomes predicted among patients with scores less than 3 and mortality being high among those with more than 7 points. In one study that compared the prognostic utility of the several scales, the Essen ICH score was the best to discriminate for different outcome measures. An advantage of the Essen ICH scale is that it is based solely on clinical variables and the score may be calculated without the results of brain imaging studies being available.