www.elsevier.com/locate/ajem

American Journal of Emergency Medicine (2009) 27, 558–562

Original Contribution

Use of the Triage Stroke Panel in a neurologicemergency serviceIgor Sibon MD, PhD⁎, François Rouanet MD,Wassilios Meissner MD, PhD, Jean Marc Orgogozo MD

Department of Clinical Neurosciences, Bordeaux University Hospital, France

Received 31 October 2007; revised 9 May 2008; accepted 13 May 2008

Abstract

Background: Acute stroke is associated with serum elevations of numerous markers. We evaluated theadditive accuracy of the Triage Stroke Panel (D-dimer, B-natriuretic peptide, matrix metalloproteinase 9,and S-100β) to the triaging nurse for acute stroke diagnosis.Methods: Consecutive patients with suspected stroke were included in this prospective, controlled,single-center study. A well-trained stroke center triage nurse assigned a probability that the patient hadexperienced a stroke (certain, very probable, probable, not likely, doubtful, or other); then, the TriageStroke Panel testing was performed. Patients' diagnosis was based on clinical and imaging data by aneurologist blinded to the test results.Results: Two hundred four patients were evaluated. Confirmed strokes and transient ischemic attacks(TIAs) were observed in 131 patients. When considering an experienced stroke nurse's assessment of“other,” “doubtful,” or “not likely” to be negative for stroke and categorizing TIAwith stroke, the strokepanel's Multimarker Index (MMX) value had identical accuracy (∼70%) and equivalent sensitivity(∼94%) and specificity (∼24%) for stroke diagnosis to that of the nurse. Combining nurse assessmentwith the MMX result significantly improved the specificity of diagnosing “mimic” vs stroke + TIA from25.4% (nurse assessment only) to 46.0% (nurse assessment + MMX; P b .001).Conclusions: The Triage Stroke Panel provides objective information that complements a triage nurse inthe assessment of a suspected stroke patient. Its performance compares favorably with that of a well-trained stroke center triage nurse, suggesting potential use in nonexpert centers for improving theaccuracy of stroke diagnosis.© 2009 Elsevier Inc. All rights reserved.

1. Introduction

In patients suspected of having an acute cerebrovascularinsult, time to diagnosis and accuracy of diagnosis are the 2

⁎ Corresponding author. Département de Neurologie, CHU Pellegrin,Université de Bordeaux II, Place Amélie Raba-Léon, 33076 BordeauxCedex – France. Tel.: +33 5 56 79 55 20; fax: +33 5 56 79 87 02.

E-mail address: igor.sibon@chu-bordeaux.fr (I. Sibon).

0735-6757/$ – see front matter © 2009 Elsevier Inc. All rights reserved.doi:10.1016/j.ajem.2008.05.001

main parameters for successful therapy. In the last 15 years,the introduction of thrombolytic agents for the treatment ofacute ischemic stroke has resulted in a transition in strokecare from supportive medical therapy to that of an acutemedical emergency [1]. However, the time-window forthrombolytic therapy remains narrow and requires rapid,accurate, and well-organized prehospital and in-hospitalsystems [2,3]. Several obstacles preclude early administra-tion of thrombolytic therapy in acute stroke patients such as

559Triage Stroke Panel in a neurologic emergency service

the following: (1) the lack of awareness of stroke symptomsby the patient or the patient's family, (2) the nonurgent triageof stroke patients in the emergency department (ED), or (3)the physician's uncertainty regarding administration ofthrombolytic medication [4]. To overcome these short-comings, different awareness campaigns have disseminatedstroke warning signs to the general population and havepromoted thrombolysis to health professionals [5-7].Furthermore, the competences of paramedics have beenimproved by specific stroke education programs [8-10].However, only 2% to 10% of acute stroke patients receiveintravenous thrombolysis [11-13], suggesting the need foradditional efforts to increase the number of treated patients.

In most hospitals, stroke patients are commonly triaged bynurses responsible for the management of all acute medicalconditions. Several studies have clearly demonstrated theeffectiveness of acute stroke triage protocols in reducingdoor-to–computed tomography (CT) and door-to-needletimes in patients presenting directly to the ED [14].Moreover,Lindsberg, et al [15] reported a favorable increase in thenumber of patients eligible for thrombolytic therapy after areorganization of the ED. Improved triage of stroke patientswas achieved using a nursing team that was trained in strokesymptom recognition and exclusively dedicated to the care ofneurologic patients in the ED [15,16]. However, in thepreadmission setting and in nonspecialized centers wherespecifically trained stroke nurses are unavailable, theaccuracy of clinical and paraclinical tools is critical to reducethe risk of misdiagnosing stroke and its medicolegalimplications [17]. In this case, a rapid and sensitive diagnosticassay for detecting acute stroke would be highly beneficial forthe triage of stroke patients during preadmission and in theED of nonspecialized centers. The Triage Stroke Panel(Biosite Incorporated, San Diego, Calif) is a rapid point-of-care test that measures plasma levels of B-type natriureticpeptide (BNP), D-dimer, matrix metalloproteinase 9 (MMP-9), and S-100β and provides the test results as a MultimarkerIndex (MMX) value. TheMMXvalue can be used as an aid inthe assessment and diagnosis of stroke. Here, we assessed theaccuracy of a trained stroke nurse and the Triage Stroke Panelfor identifying stroke patients in the ED.

ig. 1 Disease probability based upon single MMX cutoff.ercentages of mimic (gray), stroke (black), and TIA (white)iagnoses depending on the MMX cutoff. Mimic was significantlyss probable when MMX value was higher than 1.3. Stroke wasignificantly more probable when MMX value was higher than 1.3★ corresponds to P b .001).

2. Materials and methods

Two hundred four consecutive patients with suspectedstroke or transient ischemic attack (TIA, defined asspontaneous disappearance of neurologic symptoms withinthe 24 hours after stroke onset) presenting during autumn2005 to the ED of the Hospital Pellegrin, Bordeaux, wereincluded in this study. Patients with coma or traumatic injurywere excluded.

A well-trained stroke center triage nurse considered adiagnosis of stroke when a patient presented with suddenonset of limb weakness, difficulties in speech/language,

vision, mobility, or balance, or when a physician referred thepatient for suspected stroke or TIA. The nurse assigned astroke probability for each patient (certain, very probable,probable, not likely, doubtful, or other diagnosis).

During the assessment, the patient's blood was drawn forTriage Stroke Panel testing. The EDTA-anticoagulatedwhole blood was used for this purpose. The Triage StrokePanel is a point-of-care fluorescence immunoassay for therapid quantitative measurement of BNP and fibrin degrada-tion products containing D-dimer, MMP-9, and S-100β inEDTA-anticoagulated whole blood or plasma specimens. Inaddition to the values for the individual markers, aproprietary algorithm for the automatic calculation of asingle MMX result is obtained to simplify the assessmentand diagnosis of stroke (Triage Stroke Panel). The MMXranges from 0 to 10. If the MMX is less than or equal to 1.3,the likelihood of a stroke is low. An index greater than 5.9indicates a strong likelihood of stroke. A value between 1.3and 5.9 is abnormal and suggests the need for furtherevaluation [18]. Nurses were blinded to the MMX results.All patients underwent a clinical evaluation by a neurologist,and brain imaging was performed (brain CT scan and/ormagnetic resonance imaging) within 24 hours after admis-sion. The final diagnosis was based on clinical and imagingdata and was performed 1 day after admission by a seniorneurologist blinded to the test results. Strokes weresubclassified as infarctions, TIAs, and hemorrhages.

All other diagnoses were grouped as “mimics.”Descriptive statistics were summarized for patient demo-

graphic variables and final diagnosis. Patients with missingdata (MMX or nurse assessment) were excluded from thestatistical analyses. The clinical performance parameters ofsensitivity, specificity, positive predictive value (PPV), andnegative predictive value (NPV) were calculated based onthe final diagnosis of the patients for the MMX result and thestroke nurse's assessment. The accuracy for identifyingstroke patients by the MMX result (value, N1.3), the trainednurse's assessment, or their combination was defined by the

FPdles(★

Table 1 Sensitivity and specificity of nurse and MMX strokeassessment (TIA included)

n = 189 Mimic All strokes (+TIA)

Nurse+MMX+n 34 111% 17.99 59MMX−n 13 8

560 I. Sibon et al.

sensitivity, specificity, NPV and PPV, and the validity wassupported by the percent of true positives (ie, MMXvalidity = Number of positive MMX (N1.3) results / Totalnumber of strokes confirmed by the neurologist).

NcNemar tests were used to compare the clinicalperformance parameters between the MMX result, the strokenurse's assessment, and the combination of the MMX resultwith the stroke nurse's assessment. Results with P valuesless than .05 were considered to be statistically significant.

% 6.88 4.23Nurse−MMX+n 14 6% 7.41 3.17MMX−n 2 1% 1.06 0.53

Nurse: sensitivity, 94.4%; specificity, 25.4%; and accuracy, 71.4%.MMX: sensitivity, 92.9%; specificity, 23.8%; and accuracy, 69.8%.Nurse is negative if assigned stroke probability is “other,” “doubtful,” or“not likely.”MMX is negative if equivalent to 1.3 or lesser.

3. Results

Two hundred four patients with a suspected stroke wereevaluated (104men and 92 women), 8 patients were excludedbecause of technical failure. Mean age was 70.11 ± 15.6 years(67.69 ± 14.29 years for men and 72.85 ± 15.6 years forwomen). The diagnosis of stroke was confirmed in 131 caseswith the following subtypes: 85 infarction, 33 TIA, and 13hemorrhages. The remaining 65 patients received hetero-geneous diagnoses including epilepsy, encephalitis, ortumors. These patients received the diagnosis of stroke“mimic.” For 7 patients (5 patients with stroke and 2 patientswith mimic), the nurse-assigned stroke probability wasunavailable. Therefore, statistical analyses were performedfor 189 patients (126 patients with stroke and 63 patientswith mimic).

In patients whose MMX value was greater than 1.3, theprobability of stroke was greater than 50%, whereas inpatients whose MMX value was less than or equal to 1.3, theprobability of a stroke mimic or TIA was 80% (Fig. 1). Theprobability of stroke increased with the MMX value and was76.1% in patients whose MMX value was greater than 5.9.

When considering an experienced stroke nurse assess-ment of “other,” “doubtful,” or “not likely” to be negative forstroke, and categorizing TIA with stroke, the MMX resultand the nurse assessment had identical accuracy (agreementwith the final diagnosis) (71.4% nurse vs 69.8% MMX) andequivalent sensitivity (94.4% nurse vs 92.9% MMX) andspecificity (25.4% nurse vs 23.8% MMX) as well as NPVand PPV for the diagnosis of stroke compared with thediagnosis suggested by the nurse (Fig. 2, Table 1). Ninety-

Fig. 2 Nurse assessment and MMX result comparison in strokediagnosis. Comparison of MMX (gray) and nurse (black)performance for patients with a diagnosis of stroke mimic vspatients with a cerebrovascular event (TIA and stroke).

nine percent of strokes were identified if any positive nurseassessment or a positive Triage Stroke Panel MMX wasconsidered. Combining nurse assessment with the MMXsignificantly improved the specificity of diagnosing mimicvs stroke + TIA from 25.4% (nurse assessment only) to46.0% (nurse assessment + MMX; P b.001) (Table 2).

4. Discussion

The present study demonstrates that the measurement of acombination of biomarkers that are included in the TriageStrokePanel candetect cerebral infarctwith the sameaccuracyand specificity as a nurse trained in stroke management.

The diagnosis of acute ischemic stroke in most hospitalsis only made on clinical grounds after exclusion of anintracranial hemorrhage or a mass lesion by CT. However,other potential causes of an acute focal neurologic deficit,including complex migraine, postictal paresis, tumors,demyelinating disorders, or even metabolic disturbancessuch as hypoglycemia, may be difficult to differentiate fromstroke. Here, we demonstrate that the MMX might assist indifferentiating these mimics from real stroke patients, mainlyin centers without an experienced triage nurse and in theprehospital setting.

Acute stroke is associated with serum elevations ofnumerous inflammatory and anti-inflammatory mediatorssuch as interleukin 6 and MMP-9 [19-23], markers ofimpaired hemostasis and thrombosis such as D-dimer [24,25]and markers of glial activation such as S-100β [26,27].Several of these markers are elevated within hours afterischemia and correlate with infarct volume [28,29]. How-

Table 2 Statistical comparison between MMX, nurse and Nurse + MMX in stroke diagnosis

MMX Nurse Nurse + MMX Fisher exact or χ2 test a

Sensitivity 0.929 (0.870-0.962) 0.944 (0.890-0.973) 0.992 (0.956-0.999) 0.06598 NS b

TP 117 119 125FN 9 7 1Specificity 0.238 (0.149-0.356) 0.254 (0.163-0.373) 0.460 (0.343-0.582) 0.025148 b

TP 15 16 29FN 48 47 34Likelihood ratio+ 1.2 1.3 1.8Likelihood ratio− 3.3 4.6 58.0PPV 0.709 (0.636-0.773) 0.717 (0.644-0.779) 0.786 (0.715-0.843) 0.5682 NS c

NPV 0.625 (0.427-0.788) 0.696 (0.491-0.844) 0.967 (0.833-0.994) 0.1605 NS c

Accuracy 0.698 (0.630-0.759) 0.714 (0.646-0.774) 0.815 (0.753-0.864) 0.5454 NS c

Parenthetical values are 95% confidence interval.NS indicates nonstatistically significant; TP, true positive; FN, false negative.

a Comparison between nurse vs nurse + MMX (n = 189).b Fisher exact.c χ2

561Triage Stroke Panel in a neurologic emergency service

ever, although highly correlated with cerebral ischemia, noindividual biochemical marker has the necessary sensitivityand specificity to be used as a clinically independentdiagnostic marker [30-32]. Given the absence of a singleperipheral diagnostic marker of stroke, an alternativeapproach is the creation of a panel of several serumbiomarkers [30,33,34].

Among the biochemical markers that are involved in thepathogenesis of stroke and neuronal injury, Lynch et al [30]identified 4 markers as highly correlated with stroke: 1marker of glial activation (S-100β), 2 markers of inflamma-tion (matrix metalloproteinase 9 and vascular cell adhesionmolecule), and 1 marker of thrombosis (von Willebrandfactor). S-100β, MMP-9, and von Willebrand factor werealso identified by Reynolds et al [34] as highly correlatedwith the diagnosis of stroke as well as B-type neurotrophicgrowth factor and monocyte chemoattractant protein-1.Moreover, they demonstrated the ability of these differentbiomarkers as a “marker panel” to diagnose hemorrhagic andischemic stroke and, in particular, acute ischemic strokewithin the first 6 hours of onset [34]. The Triage Stroke Panelwas designed according to the results of these differentbiomarker studies in stroke patients. In a recent study,Grifoni et al [35] found the MMX result to have an accuracyfor stroke diagnosis of 74%, a result close to the 69.8%observed in the present study. Grifoni et al [35] also reportedthe ability of the Triage stroke panel to contribute to thediagnosis of stroke when clinical data are doubtful and whenthe values of a prehospital stroke scale are normal. Togetherwith the present study, these results suggest that the TriageStroke Panel is an attractive tool to improve the accuracy ofrapid stroke diagnosis. Two cutoffs associated with a high andlow probability of stroke have been defined. In the presentstudy, we only used the lower cutoff (1.3). This provided theopportunity to identify the benefit of combining the MMXwith the nurse assessment, suggesting that a single cutoffshould be used in the clinical practice.

The ideal biomarker (or panel) should be able to identifyacute stroke, differentiate ischemic from hemorrhagic stroke,correlate with infarct volume, relate to stroke mechanism,and predict the long-term outcome. To date, no singlebiomarker meets these criteria, but studies have demon-strated that a combination of biomarkers, including D-dimerand BNP, might be more specific of cardioembolic stroke andthat D-dimer or MMP-9 could be linked to infarct volume orstroke outcome [21,34]. Further studies are necessary toconfirm these data.

In conclusion, the use of the Triage Stroke Panel providesobjective information that complements a triage nurse in theassessment of a suspected stroke patient. Its performancecompares favorably with that of a well-trained stroke centertriage nurse, suggesting potential use in nonexpert centers forimproving the accuracy of stroke diagnosis. Moreover,experienced EDs might also benefit from use of the TriageStroke Panel because the specificity for the diagnosis of amimic was increased, and 99% of stroke patients could beidentified when combining the nurse assessment with theMMX result. However, the interaction of the nursingassessment and the Triage Stroke Panel result was notinvestigated and is a limitation to this study. Because thenurses were blinded to the results of the Triage Stroke Panel,a 99% rate of stroke identification (by a positive nurseassessment or a positive Triage Stroke Panel) may over-estimate the actual rate of stroke identification. Furtherresearch will be needed to investigate the interaction of thenurse assessment and the Triage Stroke Panel result.

References

[1] Citerio G, Galli D, Cadore B, et al. How to improve ischemic stroketreatment in the fibrinolysis era. Minerva Anestesiol 2006;72:407-12.

[2] Hacke W, Brott T, Caplan L, et al. Thrombolysis in acute ischemicstroke: controlled trials and clinical experience. Neurology 1999;53:S3-S14.

562 I. Sibon et al.

[3] Wardlaw JM, Zoppo G, Yamaguchi T, et al. Thrombolysis foracute ischaemic stroke. Cochrane Database Syst Rev 2003:CD000213.

[4] Kwan J, Hand P, Sandercock P. A systematic review of barriers todelivery of thrombolysis for acute stroke. Age Ageing 2004;33:116-21.

[5] Kwan J, Hand P, Sandercock P. Improving the efficiency of delivery ofthrombolysis for acute stroke: a systematic review. QJM 2004;97:273-9.

[6] Alberts MJ, Perry A, Dawson DV, et al. Effects of public andprofessional education on reducing the delay in presentation andreferral of stroke patients. Stroke 1992;23:352-6.

[7] Morgenstern LB, Staub L, Chan W, et al. Improving delivery of acutestroke therapy: the TLL Temple Foundation Stroke Project. Stroke2002;33:160-6.

[8] Sandercock P, Berge E, Dennis M, et al. A systematic review of theeffectiveness, cost-effectiveness and barriers to implementation ofthrombolytic and neuroprotective therapy for acute ischaemic stroke inthe NHS. Health Technol Assess 2002;6:1-112.

[9] Bray JE, Martin J, Cooper G, et al. An interventional study toimprove paramedic diagnosis of stroke. Prehosp Emerg Care 2005;9:297-302.

[10] Bray JE, Martin J, Cooper G, et al. Paramedic identification of stroke:community validation of the Melbourne ambulance stroke screen.Cerebrovasc Dis 2005;20:28-33.

[11] Rosamond W, Flegal K, Friday G. Heart disease and stroke statistics—2007 update. http://circahajournalsorg/cgi/content/full/115/5/e69#TBL5179743; 2007.

[12] Katzan IL, Furlan AJ, Lloyd LE, et al. Use of tissue-type plasminogenactivator for acute ischemic stroke: the Cleveland area experience.JAMA 2000;283:1151-8.

[13] Wang DZ, Rose JA, Honings DS, et al. Treating acute stroke patientswith intravenous tPA. The OSF stroke network experience. Stroke2000;31:77-81.

[14] Mehdiratta M, Woolfenden AR, Chapman KM, et al. Reduction in IVt-PA door to needle times using an Acute Stroke Triage Pathway. Can JNeurol Sci 2006;33:214-6.

[15] Lindsberg PJ, Happola O, Kallela M, et al. Door to thrombolysis: ERreorganization and reduced delays to acute stroke treatment.Neurology 2006;67:334-6.

[16] Ornstein SM. Translating research into practice using electronicmedical records the PPRNet-TRIP project: primary and secondaryprevention of coronary heart disease and stroke. Top Health InfManage 2001;22:52-8.

[17] Brous E. A patient with an undetected evolving stroke: legal lessonslearned. J Emerg Nurs 2005;31:580-2.

[18] Laskowitz D. A point of care diagnostic for acute stroke. 14th AnnualEuropean Stroke Conference, Bologna, Italy; 2005.

[19] Kim JS, Yoon SS, Kim YH, et al. Serial measurement of interleukin-6,transforming growth factor–beta, and S-100 protein in patients withacute stroke. Stroke 1996;27:1553-7.

[20] Dziedzic T, Bartus S, Klimkowicz A, et al. Intracerebral hemorrhagetriggers interleukin-6 and interleukin-10 release in blood. Stroke 2002;33:2334-5.

[21] Montaner J, Alvarez-Sabin J, Molina C, et al. Matrix metalloproteinaseexpression after human cardioembolic stroke: temporal profile andrelation to neurological impairment. Stroke 2001;32:1759-66.

[22] Perini F, Morra M, Alecci M, Galloni E, et al. Temporal profile ofserum anti-inflammatory and pro-inflammatory interleukins in acuteischemic stroke patients. Neurol Sci 2001;22:289-96.

[23] Vila N, Castillo J, Davalos A, Chamorro A. Proinflammatorycytokines and early neurological worsening in ischemic stroke. Stroke2000;31:2325-9.

[24] Fon EA, Mackey A, Cote R, et al. Hemostatic markers in acutetransient ischemic attacks. Stroke 1994;25:282-6.

[25] Takano K, Yamaguchi T, Uchida K. Markers of a hypercoagulablestate following acute ischemic stroke. Stroke 1992;23:194-8.

[26] Buttner T, Weyers S, Postert T, et al. S-100 protein: serum marker offocal brain damage after ischemic territorial MCA infarction. Stroke1997;28:1961-5.

[27] Martens P, Raabe A, Johnsson P. Serum S-100 and neuron-specificenolase for prediction of regaining consciousness after global cerebralischemia. Stroke 1998;29:2363-6.

[28] Tarkowski E, Rosengren L, Blomstrand C, et al. Early intrathecalproduction of interleukin-6 predicts the size of brain lesion in stroke.Stroke 1995;26:1393-8.

[29] Fassbender K, Rossol S, Kammer T, et al. Proinflammatory cytokinesin serum of patients with acute cerebral ischemia: kinetics of secretionand relation to the extent of brain damage and outcome of disease.J Neurol Sci 1994;122:135-9.

[30] Lynch JR, Blessing R, White WD, et al. Novel diagnostic test for acutestroke. Stroke 2004;35:57-63.

[31] Jauch EC, Lindsell C, Broderick J, et al. Association of serialbiochemical markers with acute ischemic stroke: the National Instituteof Neurological Disorders and Stroke recombinant tissue plasminogenactivator stroke study. Stroke 2006;37:2508-13.

[32] Rainer TH, Lam NY, Man CY, et al. Plasma beta-globin DNA as aprognostic marker in chest pain patients. Clin Chim Acta 2006;368:110-3.

[33] Dambinova SA, Khounteev GA, Skoromets AA. Multiple panel ofbiomarkers for TIA/stroke evaluation. Stroke 2002;33:1181-2.

[34] Reynolds MA, Kirchick HJ, Dahlen JR, et al. Early biomarkers ofstroke. Clin Chem 2003;49:1733-9.

[35] Grifoni S, Vanni S, Caini R, et al. Biochemical markers for thediagnosis of stroke: do they work in the real world. Conferenzanationale dell ictus cerebrale, Firenze, Italy; 2006.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.