Prehospital Acute Stroke Severity Scale to Predict Large Artery Occlusion
Design and Comparison With Other Scales
Background and Purpose—We designed and validated a simple prehospital stroke scale to identify emergent large vessel occlusion (ELVO) in patients with acute ischemic stroke and compared the scale to other published scales for prediction of ELVO.
Methods—A national historical test cohort of 3127 patients with information on intracranial vessel status (angiography) before reperfusion therapy was identified. National Institutes of Health Stroke Scale (NIHSS) items with the highest predictive value of occlusion of a large intracranial artery were identified, and the most optimal combination meeting predefined criteria to ensure usefulness in the prehospital phase was determined. The predictive performance of Prehospital Acute Stroke Severity (PASS) scale was compared with other published scales for ELVO.
Results—The PASS scale was composed of 3 NIHSS scores: level of consciousness (month/age), gaze palsy/deviation, and arm weakness. In derivation of PASS 2/3 of the test cohort was used and showed accuracy (area under the curve) of 0.76 for detecting large arterial occlusion. Optimal cut point ≥2 abnormal scores showed: sensitivity=0.66 (95% CI, 0.62–0.69), specificity=0.83 (0.81–0.85), and area under the curve=0.74 (0.72–0.76). Validation on 1/3 of the test cohort showed similar performance. Patients with a large artery occlusion on angiography with PASS ≥2 had a median NIHSS score of 17 (interquartile range=6) as opposed to PASS <2 with a median NIHSS score of 6 (interquartile range=5). The PASS scale showed equal performance although more simple when compared with other scales predicting ELVO.
Conclusions—The PASS scale is simple and has promising accuracy for prediction of ELVO in the field.
Reperfusion with thrombolysis (intravenous tissue plasminogen activator [tPA]) and endovascular therapy (EVT) improves outcome in patients with acute ischemic stroke.1–4 The efficiency of both treatments is strongly dependent on early initiation after onset of symptoms.2,4–6 In case of an emergent large vessel occlusion (ELVO),7 EVT is superior to intravenous tPA only.4 Currently, Primary Stroke Centers and Comprehensive Stroke Centers form a 2-tier regionalized system of care for the efficient management of patients with acute stroke in Denmark similar to the organization in many other countries. Primary Stroke Centers provide acute stroke care with intravenous tPA, whereas advanced treatment with EVT is only available in Comprehensive Stroke Centers. Direct referral of patients with ELVO to a Comprehensive Stroke Centers even if a Primary Stroke Center is located nearer to the patient may be of importance with the recent findings that timely EVT improves outcome in acute ischemic stroke patients with confirmed occlusion of a large artery.8–10
Prehospital identification of ELVO by Emergency Medical Services (EMS) is a prerequisite for direct referral to Comprehensive Stroke Centers. Scoring systems identifying ELVO have been published previously, but have to date not been widely adopted in clinical practice.11–14
We report the development and validation of a new simple stroke scale, the Prehospital Acute Stroke Severity (PASS) scale, for prediction of ELVO and compare its predictive performance with other published stroke scales for ELVO.11–14
Materials and Methods
To derive the PASS scale, we used a historical cohort from the Danish Stroke Registry.15 The Danish Stroke Registry is a nationwide, population-based clinical registry, which holds data on patient characteristics and care, including intravenous tPA and EVT. In patients receiving intravenous tPA or EVT results of acute imaging are recorded. Large artery occlusion is defined as a visible clot in either the anterior or the posterior circulation on an intracranial computed tomographic angiography (CTA) or a magnetic resonance angiography (MRA). Registration of patients in the Danish Stroke Registry is mandatory for all hospitals treating patients with acute stroke in Denmark. The sensitivity and predictive value of the registration with acute stroke are >90% in the Danish Stroke Registry.15 We identified all patients receiving intravenous tPA in Denmark from January 2010 to April 2015, n=5864. We restricted the population to patients examined with CTA or MRA before intravenous tPA. Patients with inconclusive or missing information on CTA or MRA and National Institutes of Health Stroke Scale (NIHSS) and grip strength (included in Scandinavian Stroke Scale [SSS]) before intravenous tPA were excluded leaving a cohort of 3127 intravenous tPA patients. Flowchart of the study population is shown in Figure 1.
The study was approved by the Danish Data Protection Agency reference no. 2007-58-0010 and Danish Clinical Registries. According to Danish law, registry-based studies do not require ethical approval.
Before analyzing data, an expert group of vascular neurologists identified the following criteria for an applicable prehospital stroke scale predicting ELVO:
Maximum of 4 elements in the scale derived from NIHSS
No grading of the clinical elements
No complex flow chart models
Operational over the phone by EMS dispatch personnel
Operational in prehospital settings by EMS paramedics
Expert consensus was that the NIHSS items visual fields, facial paresis, limb ataxia, sensory, and extinction/inattention were not fit for the use in a prehospital assessment because of demands of complex examination and the high grade of subjectivity, and they were excluded for further analysis. Dysarthria and aphasia were combined and an overall element termed “language problems” was used for further analysis.
Comparison With Other Stroke Scales for ELVO
A comparison of the PASS scale was made to other published stroke scales designed for identifying ELVO in the field. The 3-Item Stroke Scale (3ISS),11 the Los Angeles Motor Scale (LAMS),12 the Rapid Arterial Occlusion Evaluation Scale (RACE),13 and Cincinnati Prehospital Stroke Severity Scale (CPSSS)14 were all applied on the entire test cohort and compared with predictive performance of the PASS scale. 3ISS, RACE, and CPSSS only encompasses items from NIHSS, whereas LAMS include grip strength and this could be derived from SSS. The construct of 3ISS, LAMS, RACE, and CPSSS with correspondent NIHSS or SSS cut points of the elements of the scales are shown in Figure I in online-only Data Supplement.
Among the preselected NIHSS elements, we identified the elements with the highest positive predictive value for large artery occlusion verified by intracranial CTA or MRA in a random sample of 2 of 3 of the patient population. The associations of the NIHSS elements with large artery occlusion were first examined by computing odds ratios, and the predictive performance was subsequently determined using receiver operating characteristic curves and c statistics. Different combinations of these items were computed, and the associations with large artery occlusion were examined using multivariable logistic regression, receiver operating characteristic curves, and c statistics. The most optimal combination was identified and odds ratio, sensitivity, specificity, area under the curve (AUC), predictive values and likelihood ratios were calculated for the optimal cut off. The PASS scale was subsequently validated using the remaining 1/3 of the patient population.
When comparing the PASS scale with the previously published scales, a similar approach was used for the individual scales to identify the optimal cut points. Analyses were done using the Stata 13.0 package (StataCorp LP, College Station, TX).
Of the 3127 included intravenous tPA patients, 60% were males and the mean age was 69 years (SD ±14). Baseline characteristics are shown in Table 1. Median NIHSS score was 7 (interquartile range=9) and CTA or MRA showed arterial occlusion in 35% (n=1104). The PASS scale consisted of only 3 items derived from the NIHSS. The strongest predictor of large artery occlusion among the NIHSS items was abnormal gaze (NIHSS gaze >0) with a sensitivity=0.54 (95% CI, 0.51–0.58), specificity=0.87 (0.85–0.89), AUC=0.71 (0.69–0.73), and odds ratio=7.88 (6.36–9.76). The most optimal combination with other items from NIHSS complying with the expert criteria was a combination with motor function of the arm (NIHSS >0 in right/left or both arms) and to level of consciousness questions (month/age) (NIHSS >0), Figure 2. The PASS scale had an overall AUC of 0.76 for detecting occlusion on an acute intracranial CTA or MRA. Receiver operating characteristic curve for the PASS scale in derivation is shown in Figure 3. The best predictive value of PASS was at cut point ≥2 (yes to 2 or 3 elements); this cut point value showed a sensitivity=0.66 (0.62–0.69); specificity=0.83 (0.81–0.85); AUC=0.74 (0.72–0.76), and odds ratio=9.22 (7.50–11.40). When validating the scale in the remaining 1/3 of the patient population, the accuracy of the PASS scale was similar as shown in Table 2.
Patients with a large artery occlusion identified when applying the PASS scale (cut point ≥2) on the entire cohort had a median NIHSS score of 17 (interquartile range=6), whereas patients not identified by the PASS scale had a median NIHSS score of 6 (interquartile range=5).
The PASS score’s (≥2) predictive performance was subsequently tested in clinically moderately/severely affected patients (NIHSS ≥10) with a large artery occlusion in the entire cohort: sensitivity=0.89 (0.86–0.91), specificity=0.84 (0.82–0.85), AUC=0.86 (0.85–0.88), and odds ratio=41.3 (32.2–53.1).
When applying the PASS scale and the other prehospital scoring systems for ELVO on the entire patient population, none of the existing scales performed better in detecting a verified large artery occlusion than the PASS scale. Overall accuracy and characteristics of best cut-off values of 3ISS, LAMS, RACE, and CPSSS in comparison with the PASS scale are shown in Table 3. Receiver operating characteristic curves of 3ISS, LAMS, RACE, CPSSS, and PASS when applied on the entire test cohort are shown in online-only Supplementary Data, Figure II.
We developed the PASS scale after applying clinical consensus criteria and derived and validated the PASS scale in a large national cohort of patients with acute ischemic stroke and intracranial artery imaging performed before reperfusion therapy in Denmark from 2010 to 2015. The PASS scale identified ≈2 of 3 of all patients with CTA or MRA verified large artery occlusions while maintaining a high specificity and the performance improved in predicting patients with large artery and moderate to severe deficits (NIHSS ≥10). The patients with large artery occlusion identified by the PASS scale had severe deficits (median NIHSS 17) opposed to the patients not identified by the PASS scale (median NIHSS 6).
The PASS scale therefore identifies patients with ELVO who is the most likely to benefit from EVT.4 The most important clinical symptom of identifying ELVO was found to be abnormal gaze (comprising gaze palsy and gaze deviation), which alone accounted for identifying more than half of all the patients with verified occlusion on CTA or MRA in the test cohort with a high specificity. This was anticipated by the experts as gaze deviation indicates involvement of large areas of cortical or subcortical tissue primarily in the frontal eye field and is in patients with stroke associated with severe neurological deficits and is a clinical sign from both hemispheres despite being reported more frequent in right hemispheric involvement.16–18 Gaze deviation is also seen when the paramedian pontine reticular formation (pontine gaze center) is affected as seen in large artery occlusion in the posterior circulation affecting the brain stem. Affection of brain stem will also in some cases lead to a reduction of the eye movements because of nuclear cranial nerve palsy or internuclear ophthalmoplegia. Eye deviation often results in head deviation, which is an even more prominent clinical sign that is easily observed in the field by paramedics. A combination with the cortical sign (month and date also test for aphasic symptoms) with left hemispheric dominance and motor arm function further strengthens the prediction of large artery occlusion, although only modest because the symptoms are highly dependent.
Timely EVT has recently become evidence-based standard care4 and easily used triage systems for ELVO detection in the field are, therefore, of great importance. The already published scales11–14 may, however, be too complex or time-consuming and not optimal for the use in the prehospital phase.
We chose to set up predefined criteria of a scale to ensure usefulness in the prehospital phase before it was derived from the data set. A high specificity was considered crucial even though it may entail lower sensitivity because the key objective for the scale is to enable selection patients in the prehospital setting for direct referral to Comprehensive Stroke Centers for EVT. A high specificity ensures that only a small number of patients without ELVO and therefore no need for EVT will risk potential delays in receiving intravenous tPA by being transported to a more distant Comprehensive Stroke Center because ELVO was mistakenly suspected. It was also of importance to identify criteria that could easily be operated over the phone in case of long transport distances as to prepare for physician-assisted helicopter transport service if possible.
The PASS scale encompasses only 3 clinical elements and in case of the presence of minimum 2 of the elements, the PASS scale suggest ELVO with a high specificity and an NIHSS score comparable with what was found in published EVT trials.4
Large artery occlusions of both the anterior and the posterior circulation were included in this study as opposed to some of the other scales for ELVO focusing on the anterior circulation only when designed. All scales proved to be useful in this respect in the present test cohort. It is probably a strength to use a scale that can also identify patients with posterior circulation stroke for EVT because of the poor prognosis, for example, in basilar artery occlusion, without treatment although not proved from randomized trials.
This study has some limitations. First, the scale was derived and validated from a cohort all with a confirmed acute ischemic stroke receiving intravenous tPA or EVT. Applying the PASS scale in a general population with suspicion of stroke would presumably lead to selection of hemorrhagic strokes too, but this stroke type could also benefit from early care in a Comprehensive Stroke Center with advanced diagnostic imaging, continuous in-hospital neurosurgical availabilities, and neurocritical care.19 Second, the data did not include information on the specific location of the large artery occlusion because this was not registered in the Danish Stroke Registry. Some of the included large artery occlusions might not have been eligible to EVT because of a too distal location. This was probably true only in a few cases because the identified patients with large artery occlusions had a median NIHSS 17, which corresponds to the median NIHSS in resent EVT trials.4 Third, the elements were derived from the NIHSS done at the time of admission to hospital by experienced doctors. But since the NIHSS was done before intravenous tPA was administrated, the examination was done no later than 4.5 hours after onset of symptoms and probably correlates well to the clinical presentation on scene; and the elements from NIHSS included in the PASS scale have a good interobserver reproducibility by physicians and other health personals.20,21 Some patients recover during transportation to hospital, and this may also be seen when PASS is tested in the field and it needs to be confirmed that EMS scores correlate well with scores at admission. Finally, the proportion of stroke mimics needs to be determined in patients suspected of stroke and a PASS score of ≥2.
In summary, the PASS scale was developed to be a simple scale and it proved to identify ELVO in a thrombolysis cohort with acute intracranial angiographic information. The PASS scale with cut point ≥2 has promising characteristics in prediction of ELVO and translates to a high NIHSS score comparable with previous EVT trials, but we emphasize that the PASS scale should be validated in the field in a general population of patients with potential stroke by EMS personnel.
Sources of Funding
This study was supported by grants from the Lundbeck Foundation, Denmark; Laerdal Foundation, Norway and Aarhus University, Denmark.
Guest Editor for this article was Natan Bornstein, MD.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.115.012482/-/DC1.
- Received December 18, 2015.
- Revision received April 8, 2016.
- Accepted April 25, 2016.
- © 2016 American Heart Association, Inc.
- Campbell BC,
- Meretoja A,
- Donnan GA,
- Davis SM
- Khatri P,
- Yeatts SD,
- Mazighi M,
- Broderick JP,
- Liebeskind DS,
- Demchuk AM,
- et al
- McTaggart RA,
- Ansari SA,
- Goyal M,
- Abruzzo TA,
- Albani B,
- Arthur AJ,
- et al
- Antevy P,
- McTaggart RA,
- Jayaraman MV
- Singer OC,
- Dvorak F,
- du Mesnil de Rochemont R,
- Lanfermann H,
- Sitzer M,
- Neumann-Haefelin T
- Nazliel B,
- Starkman S,
- Liebeskind DS,
- Ovbiagele B,
- Kim D,
- Sanossian N,
- et al
- Perez de la Ossa N,
- Carrera D,
- Gorchs M,
- Querol M,
- Millan M,
- Gomis M,
- et al
- Katz BS,
- McMullan JT,
- Sucharew H,
- Adeoye O,
- Broderick JP
- Tijssen CC,
- Schulte BP,
- Leyten AC
- Singer OC,
- Humpich MC,
- Laufs H,
- Lanfermann H,
- Steinmetz H,
- Neumann-Haefelin T
- Fletcher JJ,
- Kotagal V,
- Mammoser A,
- Peterson M,
- Morgenstern LB,
- Burke JF
- Goldstein LB,
- Samsa GP