ences 270 (2008) 60–66www.elsevier.com/locate/jns

Journal of the Neurological Sci

Predictors of survival and functional outcome in acute stroke patientsadmitted to the stroke intensive care unit

Jiann-Shing Jeng a,c,⁎, Sheng-Jean Huang b, Sung-Chun Tang a, Ping-Keung Yip d

a Stroke Center and Department of Neurology, National Taiwan University Hospital, Taipei , Taiwanb Department of Surgery, National Taiwan University Hospital, Taipei, Taiwan

c Section of Neurology, Department of Internal Medicine, Far Eastern Memorial Hospital, Taipei, Taiwand Department of Neurology, Cardinal Tien Hospital, Taipei, Taiwan

Received 15 October 2007; received in revised form 24 December 2007; accepted 22 January 2008Available online 4 March 2008

Abstract

Multivariate models have not been widely used to predict the outcome of acute stroke patients admitted to the intensive care unit (ICU). Thepurpose of this study was to determine potential measures observed in the first 12 h post-stroke that predict earlymortality and functional outcomesin ICU-admitted stroke patients. Eight hundred and fifty acute stroke patients (ischemic stroke, 508; intracerebral hemorrhage, 342) were includedin this analysis betweenNovember 2002 andDecember 2006.Measures of interest were obtained in the first 12 h after onset of strokewere analyzedfor three types of outcome: 3-month mortality, 3-month mortality or institutional care, and poor functional outcomes at discharge. Poor functionaloutcomes were defined as a Barthel index b80 or a Rankin scale N2.Multivariate regression models were used to determine the predictive value ofthe observed measures. After 3 months, 17% of patients had died; 21% were alive but being cared for in institutional settings; and 62% were aliveand living at home. Functional status at discharge indicated 16% of patients had died, poor function in 50%, and good function in 34% of patients.Initial stroke severity, measured by National Institute of Health Stroke Scale, and dependence on a ventilator predicts 3-month mortality and pooroutcome in all stroke patients. In addition, old age, previous stroke, and total anterior circulatory infarct were associated with poor outcome inischemic stroke patients; old age, low body mass index and the presence of intraventricular hemorrhage were associated with poor outcomes inintracerebral hemorrhage patients. In conclusion, early stroke mortality and outcome at discharge can be predicted in the first few hours followingan acute stroke for moderate to severe ICU-admitted stroke patients.© 2008 Elsevier B.V. All rights reserved.

Keywords: Intensive care unit; Intracerebral hemorrhage; Cerebral infarct; Critical care; Outcome; Survival

1. Introduction

Most stroke patients arrive at the emergency department(ED) and are admitted to the acute wards; however, in contrastto heart attack, only a minority of stroke patients are managedin the stroke intensive care unit (ICU). A systematic review ofobservational studies has shown that stroke unit care wassignificantly associated with a 21% decrease of death, and a

⁎ Corresponding author. Department of Neurology, National TaiwanUniversity Hospital, No. 7 Chung-Shan South Road, Taipei 100, Taiwan.Tel.: +886 2 23123456x5338; fax: +886 2 23418395.

E-mail address: jsjeng@ntu.edu.tw (J.-S. Jeng).

0022-510X/$ - see front matter © 2008 Elsevier B.V. All rights reserved.doi:10.1016/j.jns.2008.01.015

13% decrease of death or poor outcome within 1 year of strokethan conventional ward care [1]. A small study showedadmission of acute stroke patients to a stroke care monitoringunit may reduce mortality and poor outcome [2]. Some haveadvocated that intensive monitoring should be a routine pro-cedure in stroke unit design [3]. Patients with acute non-traumatic intracerebral hemorrhage (ICH) who were admittedto the neurologic/neurosurgical ICU had lower mortality ratethan those treated in the general ICU [4]. Furthermore, aspecialized neurocritical care team can significantly reduce in-hospital mortality and length of hospital stay [5].

Previous studies have examined predictors of functionaloutcome ormortality after stroke developing several predictive

61J.-S. Jeng et al. / Journal of the Neurological Sciences 270 (2008) 60–66

models [6–12]. Significant independent predictors of func-tional outcome and survival for stroke patients include age,initial stroke severity, hyperglycemia or diabetes mellitus, andstroke types. Other known predictors consist of gender,previous stroke, blood pressure, hyperthermia, and the volumeof infarction or hemorrhage. These studies examined con-secutive stroke or ischemic stroke patients, and some inhemorrhagic stroke patients. There have been few studies donein ICU-admitted stroke patients [13–15]. Therefore, the aim ofthis study was to investigate the influence of potential factorsobtained in the first 12 h after onset of stroke symptoms onearly strokemortality and functional outcome in ICU-admittedstroke patients.

2. Method

2.1. Patients

This study was conducted at the stroke ICU of NationalTaiwanUniversityHospital (NTUH), fromNovember, 2002 toDecember, 2006. NTUH is a 2000-bed teaching hospitallocated in the southern part of Taipei municipality, servingboth primary and tertiary referral care for patients. NTUHserves a population of approximately 500,000. Patientsadmitted to the stroke ICU for care had severe neurologicaldeficits (e.g. National Institute Health Stroke Scale [NIHSS]score higher than 10), a stroke in evolution, or medicalconditions requiring intensive care (e.g. cardiac or respiratoryfailure). We excluded patients who arrived at ED more than12 h after the onset of stroke symptoms, that had rapid im-provement and were diagnosed as having a transient ischemicattack, anoxic–ischemic brain injury following cardiac arrest,or patients with traumatic intracranial hemorrhage. Patientswith subarachnoid hemorrhage were also excluded as theirprognostic factors may be different from those for ischemicstroke and ICH. The study protocol was approved by theInstitutional Ethics Committee.

2.2. Diagnostic evaluation

The NTUH stroke registry initiated since January, 1995and is an ongoing stroke registry [16,17]. The clinical,laboratory, imaging, and outcome in each admitted strokepatient were recorded prospectively. The medical records ofeach patient were including age, sex, body mass index (BMI,calculated as body weight in kilograms divided by the squareof body height in meters), clinical manifestations, stroke riskfactors (history of hypertension, diabetes mellitus, dyslipi-demia, cardiac diseases, left ventricular hypertrophy onelectrocardiogram, smoking status and alcohol use), com-plications and therapies (including life support techniques,and neurosurgical interventions). Stroke was defined usingthe World Health Organization (WHO) definition andincluded ischemic stroke and non-traumatic ICH [18]. Ac-cording to the Bamford criteria, ischemic stroke was cate-gorized as the total anterior circulation stroke, partial anterior

circulation stroke, lacunar stroke, and posterior circulationstroke [19]. ICH was categorized as primary hypertensive,other secondary etiology, or undetermined etiology. Thediagnosis of stroke and its subtypes was based on the clinicalfeatures and the data collected by laboratory examinations,such as brain imaging (computed tomography or magneticresonance image), echocardiography, ultrasonography ofextracranial and/or intracranial arteries, angiography (mag-netic resonance or conventional angiography), and labora-tory assessment of a prothrombotic state. All patients receivedat least one brain imaging study.

Initial systolic, diastolic, and brachial mean arterial bloodpressure (MABP) taken with the patient in the supine position,heart rate, respiratory rate, body temperature, and pulseoximeter saturation of arterial oxygen were recorded inpatients at triage when entering the ED. Additionally, serumlevels of glucose, hemoglobin, hematocrit, leukocyte andplatelet count, C-reactive protein, urea nitrogen, creatinine,aspartate and alanine aminotransferases, and electrolytes weremeasured in the ED. Neurological status at the time of EDadmissionwas determined byGlasgow coma scale (GCS), andNIHSS. All patients' status was followed for at least 3 monthsand categorized as death, survival with institutional care, andsurvival living at home. Functional outcomes at dischargewere assessed using the modified Rankin scale [20] andBarthel index [21]. A poor functional outcome was defined asa modified Rankin scale N2 or a Barthel index b80.

2.3. Statistical analysis

The categorical variables were presented as percentages, andthe continuous or discrete variables were presented as mean±SD or median (25th–75th percentile). The Kaplan–Meiermethod was used to estimate the probability of early strokemortality 3 months after the stroke event. Potential variableswere selected to fit into the predicting models, including age,sex, BMI, hypertension, diabetes mellitus, cardiac disease, atrialfibrillation, previous stroke, initial NIHSS, GCS, requiringmechanical ventilation, MABPb90 or ≥140 mmHg, heartrate≥100 bpm, body temperature≥37.5 °C, hematocritb30%,white blood cell count≥104/mm3, blood glucose≥10mmol/L,serum creatinine≥1.5 mg/dL, and stroke subtypes. For ICHpatients, an ICH volume 30 mL or more, the presence ofintraventricular hemorrhage, or supratentorial versus infraten-torial ICH were taken into analysis. Cox's proportional hazardsmodel was used to calculate the hazard ratio and 95%confidence interval (CI) of predictors of early stroke mortality.A logistic regression analysis was fit to evaluate the odds ratioand their 95% CI of predictors of patients' status 3 months afterstroke onset and functional outcome at discharge. All univariatepredictor variables with a p-value b0.1 were entered into themultivariate regression. Interaction terms of significant variableswere incorporated in the models. For all tested models, theReceiver Operator Characteristic (ROC), plotting sensitivityagainst one minus specificity was performed. The SPSS 12.0software package was used for statistical analyses.

62 J.-S. Jeng et al. / Journal of the Neurological Sciences 270 (2008) 60–66

3. Results

During the study period, 1178 acute stroke patientsadmitted to stroke ICU. After excluding 70 patients withsubarachnoid hemorrhage and 258 patients who arrived atthe ED more than 12 h after onset of the stroke symptoms, atotal of 850 acute stroke patients were included in theanalysis. Of these patients, 491 (57.8%) were male and 359(42.2%) were female with a mean age of 65.3±14.4 years.The demographic and clinical information for all the patientsare presented in Table 1. Five hundred and eight (59.8%)patients were diagnosed as having had an ischemic stroke,and 341 (40.2%) as having had a non-traumatic ICH. Fivehundred and eleven (60.1%) patients arrived at ED within 3 hafter stroke onset. The average length of ICU stay was 9.4±8.8 days and the total length of hospital stay (includingrehabilitation) was 37.4±32.0 days. The mean and medianinitial NIHSS scores were 17.7±9.6 and 17 (10–24),respectively that indicated moderate to severe neurologicaldeficit. Of all patients, 66 (7.8%) had received neurosurgery,70 (8.2%) had tracheostomy, and 278 (32.7%) had mechan-ical ventilation. Seventy (13.8%) patients with ischemicstroke received thrombolytic therapy. Of the medicalcomplications occurred in the acute stage, respiratory tractinfection was noted in 426 (50.1%) patients; urinary tractinfection, 251 (29.5%); respiratory failure, 233 (27.4%);upper gastrointestinal bleeding, 110 (12.9%); sepsis, 57(6.7%); post-stroke seizure, 53 (6.2%); acute coronarysyndrome, 17 (2%); and deep vein thrombosis, 8 (0.9%).Of all patients, 3 months after the onset of stroke, 140(16.5%) were dead; 179 (21.1%), alive and cared for ininstitutional settings; and 531 (62.4%), were alive and living

Table 1Baseline characteristics of the study population

Ischemic stroke (n=508)

Age, years 67.8 (14.4)Male sex 282 (56)Body mass index, kg/m2 23.7 (3.8)Hypertension 363 (72)Diabetes mellitus 165 (33)Heart disease 301 (59)Atrial fibrillation 243 (48)Previous stroke 150 (30)Initial ED assessment

GCS score 11.2 (3.3)NIHSS score 17.3 (8.6)Requiring ventilator aid 151 (30)MABPb90 mmHg 68 (14)90–140 mmHg 374 (73)≥140 mmHg 66 (13)

Heart rate≥100 bpm 118 (23)Body temperature≥37.5 °C 55 (11)Hematocritb30% 21 (4)WBC ≥104/mm3 155 (31)Blood glucose≥10 mmol/L 112 (22)Serum creatinine≥1.5 mg/dL 69 (14)

Values are number (percentage) except age, body mass index, GCS and NIHSS (mGCS, Glasgow Coma Scale; NIHSS, National Institute Health Stroke Scale; MAB

at home. Examination of functional status at dischargerevealed 139 (16.3%) patients were dead, 423 (49.8%)patients had poor functional outcomes, and 288 (33.9%)patients were relatively well.

3.1. Predictors of ischemic stroke outcome

Of the 508 patients with ischemic stroke, the 7-day, 30-day,and 90-day survival were 91.3%, 87.2%, and 84.8%,respectively. The selected 18 potential predicting variablesobtained in the first 12 h after onset of ischemic stroke wereanalyzed for three types of outcome: 3-month mortality, 3-month mortality or institutional care, and poor functionaloutcome at discharge. To evaluate variables predicting deathwithin 3 months after the stroke event, Cox's regression–hazard analysis was used. The results of univariate andmultivariate analyses are shown in Tables 2 and 4. Increasedrisk of early stroke mortality was predicted by a more severeinitial neurological impairment (NIHSS score) on EDadmission, for patients who required mechanical ventilatoraid, had a low hematocrit (b30%), or an ischemic strokesubtype of total anterior circulatory infarct. Of 151 ischemicstroke patients requiring ventilator aid during the acute stage,60 (39.7%) were dead, 57 (37.7%) needed institutional care,and only 34 (22.5%) were home at 3 months. To evaluatevariables predicting death or institutional care within 3 monthsafter the stroke event, and poor functional outcome atdischarge, univariate and multivariate logistic regressionanalysis were used. The initial high NIHSS score, need forventilator aid, and low hematocrit level were independentvariables related to death within 3 months after stroke onset.This model explained R2=37.3% of the variability and the

Intracerebral hemorrhage (n=342) Overall (n=850)

61.6 (13.5) 65.3 (14.4)209 (61) 491 (58)23.9 (4.4) 23.7 (4.0)289 (85) 652 (77)63 (18) 228 (27)46 (14) 347 (41)21 (6) 264 (31)64 (19) 214 (25)

10.6 (4.2) 10.9 (3.7)18.0 (10.9) 17.6 (9.6)136 (40) 287 (34)22 (7) 90 (11)

185 (53) 559 (65)135 (40) 201 (24)66 (19) 184 (22)34 (10) 89 (11)24 (7) 45 (5)

130 (38) 285 (34)61 (18) 173 (20)54 (16) 123 (15)

ean [standard deviation]).P, mean arterial blood pressure; WBC, white blood cell count.

Table 2Univariate association of variables with death, institutional care at 3 months or dependency at discharge in ICU-admitted ischemic stroke patients (n=508)

HR (95% CI) for death at 3 months OR (95% CI) for death orinstitutional care at 3 months

OR (95% CI) for death ordependency at discharge

Age (per year) 1.00 (0.99–1.02) 1.01 (1.00–1.03)† 1.05 (1.04–1.07)*Male sex 0.78 (0.52–1.16) 0.67 (0.46–0.96)† 0.47 (0.32–0.70)*Body mass index (per kg/m2) 0.96 (0.91–1.02) 0.96 (0.91–1.00)* 0.92 (0.87–0.97)*Hypertension 1.12 (0.71–1.77) 1.13 (0.76–1.69) 1.96 (1.31–2.94)*Diabetes mellitus 1.09 (0.71–1.67) 0.95 (0.65–1.39) 1.39 (0.92–2.11)Heart disease 1.26 (0.83–1.92) 1.65 (1.13–2.39)†† 2.09 (1.43–3.07)*Atrial fibrillation 1.16 (0.78–1.74) 1.88 (1.31–2.71)†† 2.57 (1.73–3.83)*Previous stroke 1.29 (0.85–1.97) 1.56 (1.06–2.30)† 2.72 (1.69–4.37)*Initial ED assessment

GCS (per unit) 0.77 (0.72–0.82)†† 0.74 (0.69–0.79)†† 0.65 (0.60–0.71)*NIHSS (per unit) 1.11 (1.08–1.13)†† 1.14 (1.11–1.18)†† 1.23 (1.19–1.28)*NIHSS≥20 3.15 (2.06–4.82)†† 4.50 (3.07–6.61)†† 9.27 (5.37–16.0)††Requiring ventilator aid 5.61 (3.66–8.61)†† 13.6 (8.59–21.6)†† 11.1 (5.45–22.4)*MABPb90 mmHg 1.07 (0.60–1.89) 0.90 (0.53–1.55) 1.63 (0.88–3.01)90–140 mmHg 1.0 1.0 1.0≥140 mmHg 0.61 (0.29–1.26) 0.94 (0.55–1.63) 0.86 (0.50–1.50)

Heart rate≥100 bpm 0.71 (0.42–1.21) 1.01 (0.66–1.54) 0.83 (0.53–1.28)Body temperature≥37.5 °C 2.50 (1.51–4.13)†† 2.41 (1.37–4.25)†† 1.88 (0.95–3.75)Hematocritb30% 3.29 (1.71–6.34)†† 2.87 (1.17–7.06)† 4.37 (1.01–18.9)*WBC≥104/mm3 1.70 (1.13–2.57)† 1.82 (1.24–2.68)† 1.17 (0.77–1.78)Blood glucose≥10 mmol/L 1.18 (0.74–1.89) 1.18 (0.77–1.81) 1.28 (0.80–2.04)Serum creatinine≥1.5 mg/dL 1.28 (0.74–2.22) 1.36 (0.81–2.27) 1.70 (0.93–3.12)

Subtype of ischemic strokeTACI 4.92 (2.25–10.8)†† 8.92 (4.69–17.0)†† 9.96 (5.96–16.6)*POCI 3.83 (1.65–8.90)† 6.58 (3.29–13.2)†† 2.01 (1.21–3.33)*Others (PACI and LACI) 1.0 1.0 1.0

CI, confidence interval; GCS, Glasgow Coma Scale; HR, hazard ratio; LACI, lacunar infarction; MABP, mean arterial blood pressure; NIHSS, National InstituteHealth Stroke Scale; OR, odds ratio; PACI, partial anterior circulatory infarction; POCI, posterior circulatory infarction; TACI; total anterior circulatoryinfarction; WBC, white blood cell count.*, pb0.1; †, pb0.05; ‡, pb0.01.

63J.-S. Jeng et al. / Journal of the Neurological Sciences 270 (2008) 60–66

area under the ROC curve was 0.866 (95%CI, 0.829 to 0.903).Univariate and multivariate analysis indicated that initial highNIHSS score, need for ventilator aid, and total anterior orposterior circulatory infarct stroke type significantly increasedthe risk of dying or requiring institutional care in the first3 months post-stroke. This model explained R2=48.1% of thevariability and the area under the ROC curve was 0.866 (95%CI, 0.833 to 0.899). Multivariate analysis showed age,previous stroke, requiring ventilator aid, high NIHSS score,and total anterior circulatory infarct type increased risk of poorfunctional outcome of ischemic stroke at discharge, defined asBarthel index b80 or modified Rankin scale N2. The modelexplained R2=57.3% of the entire variability. The area underthe ROC curve was 0.904 (95%CI, 0.875 to 0.933). There wasno significant interaction term in the models. Of ischemicstroke patients with an initial NIHSS score b10, 76.2% ofpatients achieved a good functional outcome by discharge.But, only 8.3% of patients with an initial NIHSS score of 20 ormore had achieved a good functional outcome at discharge.

3.2. Predictors of ICH outcome

Of the 342 patients with ICH, the 7-day, 30-day, and 90-daysurvival were 87.4%, 82.2%, and 81.9%, respectively. Theresults of univariate and multivariate analysis of variables

predicting outcome are shown in Tables 3 and 4. Cox'sregression–hazard analysis revealed significant independentpredictors of early stroke mortality included a higher initialNIHSS score on ED, requiring mechanical ventilator aid, alower BMI value, and a ICH volume ≥30 mL. The modelexplained R2=68.2% of the complete variation. The areaunder the ROC curve was 0.961 (95% CI, 0.936 to 0.985). Of136 ICH patients requiring ventilator aid during the acutestage, 50 (36.7%) were dead, 57 (41.9%) needed institutionalcare, and only 29 (21.3%) were home at 3 months. TheBMIb21 kg/m2 had higher death rate (19 of 77, 25%) thanBMI≥21 kg/m2 (38 of 265, 14%). A higher initial NIHSSscore, and requiring ventilator aid significantly increased therisk of dying or requiring institutional care in the first 3 monthspost-stroke. Multivariate analysis showed increased risk ofpoor functional outcome of ICH at discharge included olderpatients, requiring ventilator aid, a higher NIHSS score, alower BMI value, and the presence of intraventricularhemorrhage. The model explained R2=64.1% of the completevariation. The area under the ROC curve was 0.903 (95% CI,0.866 to 0.940). There was no significant interaction term inthe models. Eighty-four percent of ICH patients with an initialNIHSS score b10 had achieved a good functional outcome atdischarge. But, only 8.5% of patients had a score of 20 ormorecan achieve a good functional outcome at discharge.

Table 4Multivariate analysis of predictors of death or institutional care at 3 months or dependency at discharge in ICU-admitted acute stroke patients

HR (95% CI) for death at 3 months OR (95% CI) for death orinstitutional care at 3 months

OR (95% CI) for death ordependency at discharge

Ischemic strokeAge (per year) 1.00 (0.98–1.01) 1.01 (0.99–1.03) 1.05 (1.03–1.07)‡Previous stroke 0.92 (0.59–1.44) 1.19 (0.71–1.98) 2.39 (1.26–4.51)†NIHSS (per unit) 1.07 (1.04–1.10)‡ 1.10 (1.06–1.14)‡ 1.17 (1.11–1.23)‡Requiring ventilator aid 3.50 (2.15–5.57)‡ 9.24 (5.56–15.4)‡ 6.97 (3.02–16.1)‡Hematocritb30% 2.32 (1.17–4.59)* 2.49 (0.71–8.70) 5.38 (0.83–35.1)Subtypes of ischemic stroke

TACI 1.80 (0.78–4.13) 2.94 (1.39–6.21)† 2.33 (1.18–4.63)*POCI 1.40 (0.55–3.46) 2.73 (1.18–6.29)* 1.17 (0.57–2.38)Others (PACI and LACI) 1.0 1.0 1.0

ICHAge (per year) 1.00 (0.98–1.02) 1.01 (0.98–1.04) 1.05 (1.02–1.08)‡Body mass index (per kg/m2) 0.94 (0.89–1.00)* 0.92 (0.84–1.01) 0.91 (0.84–0.99)*NIHSS (per unit) 1.08 (1.04–1.13)‡ 1.19 (1.12–1.26)‡ 1.17 (1.11–1.24)‡Requiring ventilator aid 3.97 (1.49–10.6)‡ 9.65 (4.04–23.1)‡ 7.49 (2.54–22.1)‡ICH volume≥30 mL 2.51 (1.27–4.93)* 1.12 (0.46–2.75) 1.03 (0.40–2.65)ICH with ventricular extension 1.57 (0.89–2.77) 2.07 (0.92–4.68) 3.26 (1.52–6.96)‡

CI, confidence interval; HR, hazard ratio; ICH, intracerebral hemorrhage; LACI, lacunar infarction; NIHSS, National Institute Health Stroke Scale; OR, oddsratio; PACI, partial anterior circulatory infarction; POCI, posterior circulatory infarction; TACI; total anterior circulatory infarction.*, pb0.05; †, pb0.01; ‡, pb0.005.

Table 3Univariate association of initial measures with death or institutional care at 3 months or dependency at discharge in ICU-admitted ICH patients (n=342)

HR (95% CI) for death at 3 months OR (95% CI) for death orinstitutional care at 3 months

OR (95% CI) for death ordependency at discharge

Age (per year) 1.00 (0.99–1.02) 1.01 (1.00–1.03) 1.04 (1.02–1.05)††Male sex 1.35 (0.83–2.19) 1.21 (0.77–1.89) 0.98 (0.63–1.54)Body mass index (per kg/m2) 0.93 (0.88–0.99)† 0.93 (0.88–0.99)† 0.92 (0.87–0.97)††Hypertension 0.75 (0.42–1.33) 0.53 (0.30–0.96)† 0.46 (0.24–0.91)†Diabetes mellitus 1.25 (0.72–2.17) 1.28 (0.74–2.23) 1.90 (1.04–3.48)†Heart disease 1.58 (0.88–2.84) 1.30 (0.69–2.44) 1.71 (0.86–3.37)Atrial fibrillation 1.34 (0.58–3.08) 1.87 (0.77–4.53) 2.82 (0.93–8.57)*Previous stroke 0.82 (0.44–1.53) 1.84 (1.06–3.18)† 2.37 (1.27–4.44)††Initial ED assessment

GCS (per unit) 0.68 (0.63–0.74)†† 0.61 (0.55–0.67)†† 0.68 (0.62–0.74)††NIHSS (per unit) 1.16 (1.13–1.20)†† 1.26 (1.20–1.32)†† 1.23 (1.17–1.28)††NIHSS≥20 10.9 (5.74–20.8)†† 20.9 (11.9–36.8)†† 13.5 (7.13–25.4)††Requiring ventilator aid 13.7 (6.83–27.6)†† 29.4 (16.2–53.3)†† 18.3 (9.08–36.9)††

MABPb90 mmHg 2.09 (0.98–4.51)* 3.30 (1.32–8.30)† 4.41 (1.26–15.4)†90–140 mmHg 1.0 1.0 1.0≥140 mmHg 1.14 (0.70–1.86) 1.18 (0.75–1.88) 1.08 (0.69–1.70)

Heart rate≥100 bpm 1.88 (1.13–3.12)† 2.15 (1.25–3.70)†† 1.88 (1.04–3.39)†Body temperature≥37.5 °C 2.84 (1.61–5.01)†† 3.93 (1.84–8.36)†† 2.19 (0.96–5.00)*Hematocritb30% 4.14 (2.31–7.43)†† 3.58 (1.49–8.62)†† 3.37 (1.13–10.1)†WBC≥104/mm3 2.92 (1.82–4.68)†† 1.93 (1.23–3.03)†† 2.43 (1.51–3.90)††Blood glucose≥10 mmol/L 1.72 (1.02–2.89)† 1.62 (0.92–2.82)* 1.63 (0.90–2.97)Serum creatinine≥1.5 mg/dL 2.99 (1.84–4.88)†† 2.14 (1.19–3.84)† 1.97 (1.02–3.78)†

Subtypes of ICHHypertensive 0.44 (0.28–0.71)†† 0.56 (0.35–0.90)† 0.54 (0.33–0.89)†Non-hypertensive 1.0 1.0 1.0

ICH volume≥30 mL 10.7 (5.97–19.2)†† 9.25 (5.54–15.4)†† 7.37 (4.04–13.4)††ICH with ventricular extension 4.02 (2.45–6.60)†† 5.09 (3.17–8.16)†† 6.52 (3.77–11.3)††Infratentorial ICH 1.56 (0.91–2.69) 1.67 (0.94–2.94)* 1.03 (0.58–1.85)

CI, confidence interval; GCS, Glasgow Coma Scale; HR, hazard ratio; ICH, intracerebral hemorrhage; MABP, mean arterial blood pressure; NIHSS, NationalInstitute Health Stroke Scale; OR, odds ratio; WBC, white blood cell count.*, pb0.1; †, pb0.05; ‡, pb0.01.

64 J.-S. Jeng et al. / Journal of the Neurological Sciences 270 (2008) 60–66

65J.-S. Jeng et al. / Journal of the Neurological Sciences 270 (2008) 60–66

4. Discussion

Initial severity of stroke, often measured by the NIHSSscore, has been shown to be a powerful predictor of mortalityand functional outcome after stroke in many studies [6–12].In the TOAST trial, one additional point on the baselineNIHSS score had a 17% decrease in likelihood of achievingan excellent outcome at 3 months after ischemic stroke [7].Besides, symptoms or signs of initial stroke severity, such asdepressed consciousness, impaired orientation, limb paresis,truncal ataxia, and dysphagia have been reported aspredictors of early poor outcome [12,22–24]. Our studydemonstrated that the initial NIHSS score at ED admissionwas an independent, strong, prognostic factor for functionaldependence or death for moderate to severe stroke patients.An initial NIHSS score of 20 or higher predicted a highprobability of death or severe disability. In addition to theNIHSS score, the GCS or APACHE score were able topredict hospital mortality in stroke patients admitted to anICU [4,13–15]. Our study results also confirmed that theGCS score is an important predictor of prognosis for patientswith ischemic stroke or ICH admitted to the stroke ICU.

This study showed the prognosis for acute stroke patientsrequiring mechanical ventilator was poor. Only about one-fifthof these patients can be living at home at 3 months. In asystematic review of prognosis amongmechanically ventilatedstroke patients, a 30-day death rate was approximately 58%[25]. In a study of 124 stroke patients requiring mechanicalventilation in an ICU, only one-third survived at 1 year afteradmission, and age N65 years and GCS b10 on admissioncarried the poorest prognosis [14]. In another study of 230stroke patients with mechanical ventilation admitted to neuro/neurosurgical ICU, more than one-half of patients died duringhospitalization, particularly with impaired brainstem signs[26]. Factors influencing survival in mechanically ventilatedstroke patients may include accompanied neurological deficits(particularly brainstem function), ventilator-associated pneu-monia, life-sustaining treatment considerations, etc. [25–27].

Early hypotension in patients with acute ischemic strokewas associated with early mortality and poorer functionaloutcome [28–30]. One study reported an initial lower MABP(b100mmHg) was associated with an approximately two-foldincrease in mortality risk at 90 days after ischemic stroke [28].Some studies also showed a U-shaped relationship betweenmortality and admission blood pressure in patients with acuteischemic or hemorrhagic stroke [29,30]. But our study did notshow significant association of initial MABP with earlymortality or functional outcome.Our study population targetedICU-admitted patients with moderate to severe stroke andincluded few patients with minor stroke and normotension.Therefore, the association of blood pressure and outcome islikely to be insignificant.

Although a high BMI value or obesity has been estab-lished as a risk factor for stroke [31], the relationship be-tween BMI and mortality in critically ill patients remains afocus of controversy [32–35]. A prospective, multicenter

study of 1698 ICU patients revealed that a low BMI wasindependently associated with a higher mortality [33]. A lowBMI also predicted mortality in older patients admitted to amedical ICU [34]. The increased mortality of low BMI waspossibly explained by illness-related weight loss. Weightloss is also a potential preclinical marker for some chronic ordegenerative diseases [36]. However, another study showedthat BMI had minimal effect on outcome in 2148 ICUpatients [35]. Here we found that low BMI values wouldindependently increase the risks of early mortality and poorfunctional outcomes in ICH patients.

Stroke type and ischemic stroke subtype may influencestrokemortality. The NorthernManhattan Stroke Study showedischemic stroke patients with major hemispheric or basilarsyndrome had double the risk of mortality [37]. In the GermanStroke Study, the infarction of the lenticulostriate arteries was apredictor of poor functional outcome [38]. Our study demon-strated that total anterior and posterior circulatory infarctionincreased risk of early mortality and poor outcomes ascompared to lacunar and partial anterior circulatory infarction.

There are some potential limitations of our study. First,the study was conducted in a university hospital that mayinclude more severe and complicated stroke patients, and itprobably may introduce potentially biased information andlimiting the generalisability of the results. Second, this wasan observational study. Therefore, serial measurements ofvital signs and blood tests were not uniformly available. Thisimposed limitations on the availability of the data. Third, thelack of APACHE II or III scores or other validated measureof severity of systemic illness in this study.

5. Conclusions

Both initial severe neurological impairment and therequirement of ventilator assistance were independent pre-dictors of mortality and poor outcome in ICU-admittedpatients with ischemic stroke or ICH. This analysis indicatesthat several measures obtained at within 12 h of the onset ofstroke symptoms are valuable in predicting mortality andfunctional outcomes. Further validation of these modelsawaits assessment in a larger, multicentered study.

Acknowledgments

We thank Miss Yu-Chin Chan and Miss I-Chan Deng forexcellent assistance in this study. This study was supportedin part by the research grants DOH92~94-HP-1114 from theDepartment of Health, Taiwan.

References

[1] Seenan P, Long M, Langhorne P. Stroke units in their habitat:systematic review of observational studies. Stroke 2007;38:1886–92.

[2] Sulter G, Elting JW, Langedijk M, Maurits NM, De Keyser J.Admitting acute ischemic stroke patients to a stroke care monitoringunit versus a conventional stroke unit: a randomized pilot study. Stroke2003;34:101–4.

66 J.-S. Jeng et al. / Journal of the Neurological Sciences 270 (2008) 60–66

[3] Steiner T. Stroke unit design: intensive monitoring should be a routineprocedure. Stroke 2004;35:1018–9.

[4] Diringer MN, Edwards DF. Admission to a neurologic/neurosurgicalintensive care unit is associated with reduced mortality rate afterintracerebral hemorrhage. Crit Care Med 2001;29:635–40.

[5] Suarez JI, Zaidat OO, Suri MF, Feen ES, Lynch G, Hickman J, et al.Length of stay and mortality in neurocritically ill patients: impact of aspecialized neurocritical care team. Crit Care Med 2004;32:2311–7.

[6] Henon H, Godefry O, Leys D, Mounier-Vehier F, Lucas C, RondepierreP, et al. Early predictors of death after cerebral ischemic event. Stroke1995;26:392–8.

[7] Adams Jr HP, Davis PH, Leira EC, Chang KC, Bendixen BH, ClarkeWR, et al. Baseline NIH Stroke scale score strongly predicts outcomeafter stroke: a report of the Trial of Org 10172 in Acute StrokeTreatment (TOAST). Neurology 1999;53:126–31.

[8] Johnston KC, Connors AF, Wagner DP, Knaus WA, Wang X, HaleyEC. A predictive risk model for outcomes of ischemic stroke. Stroke2000;31:448–55.

[9] Petty GW, Brown Jr RD, Whisnant JP, Sicks JD, O'Fallon WM,Wiebers DO. Ischemic stroke subtypes: a population-based study offunctional outcome, survival, and recurrence. Stroke 2000;31:1062–8.

[10] Counsell C, Dennis M. Systemic review of prognostic models inpatients with acute stroke. Cerebrovasc Dis 2001;12:159–70.

[11] Appelros P, Nydevik I, Viitanen M. Poor outcome after first-everstroke. Predictors for death, dependency, and recurrence stroke withinthe first year. Stroke 2003;34:122–6.

[12] German Stroke Study Collaboration. Predicting outcome after acuteischemic stroke. An external validation of prognostic model.Neurology 2004;62:581–5.

[13] Navarrete-Navarro P, Rivera-Fernández R, López-Mutuberría MT,Galindo I, Murillo F, Dominguez JM, et al. Outcome prediction interms of functional disability and mortality among ICU-admittedsevere stroke patients: a prospective epidemiological study in the southof the European Union (Evascan Project, Andalusia, Spain). IntensiveCare Med 2003;29:1237–44.

[14] Steiner T, Mendoza G, De Georgia M, Schellinger P, Holle R HackeW.Prognosis of stroke patients requiring mechanical ventilation in aneurological critical care unit. Stroke 1997;28:711–5.

[15] Rordorf G, Koroshetz W, Efird JT, Cramer SC. Predictors of mortalityin stroke patients admitted to an intensive care unit. Crit Care Med2000;18:1301–5.

[16] Jeng JS, Lee TK, Chang YC, Huang ZS, Ng SK, Chen RC, et al.Subtypes and case-fatality rates of stroke: a hospital-based strokeregistry in Taiwan (SCAN-IV). J Neurol Sci 1998;156:220–6.

[17] Yip PK, Jeng JS, Lee TK, Chang YC, Huang ZS, Ng SK, et al.Subtypes of ischemic stroke. A hospital-based stroke registry inTaiwan (SCAN-IV). Stroke 1997;28:2507–12.

[18] Aho K, Harmsen P, Hatano S, Marquardsen J, Smimov VE, Strasser T.Cerebrovascular disease in the community: result of a WHOcollaborative study. Bull World Health Organ, Suppl 1980;58:113–30.

[19] Bamford J, Sandercock P, Dennis M, Burn J, Warlow C. Classificationand natural history of clinically identifiable subtypes of cerebralinfarction. Lancet 1991;337:1521–6.

[20] Van Swieten JC, Koudstaal PJ, Visser MC, Schouten HJA, van Gijn J.Interobserver agreement for the assessment of handicap in strokepatients. Stroke 1988;19:604–7.

[21] Mahoney FI, Barthel DW. Functional evaluation: the Barthel index.Md Med 1965;14:61–5.

[22] Johnston KC, Connors Jr AF, Wagner DP, Haley Jr EC. Predictingoutcome in ischemic stroke: external validation of predictive riskmodels. Stroke 2003;34:200–2.

[23] Paolucci S, Antonucci G, Gialloreti LE, Traballesi M, Lubich S, PratesiL, et al. Predicting stroke initial rehabilitation outcome: the prominentrole of neuropsychological disorders. Eur Neurol 1996;36:385–90.

[24] SmithardDG,O'Neill PA, Parks C,Morris J. Complications and outcomeafter acute stroke: does dysphagia matter? Stroke 1996;27:1200–4.

[25] Holloway RG, Benesch CG, Burgin WS, Zentner JB. Prognosis anddecision making in severe stroke. JAMA 2005;294:725–33.

[26] Gujjar AR, Deibert E, Manno EM, Duff S, Diringer MN. Mechanicalventilation for ischemic stroke and intracerebral hemorrhage: indica-tions, timing, and outcome. Neurology 1998;51:447–51.

[27] Rabinstein AA, Wijdicks EFM. Outcome of survivors of acute strokewho require prolonged ventilatory assistance and tracheostomy.Cerebrovasc Dis 2004;18:325–31.

[28] Stead LG, Gilmore RM, Decker WW, Weaver AL, Brown Jr RD.Initial emergency department blood pressure as predictor of survivalafter acute ischemic stroke. Neurology 2005;65:1179–83.

[29] Leonardi-Bee J, Bath PMW, Philips SJ, Sandercock PAG, for the ISTCollaborative Group. Blood pressure and clinical outcomes in theInternational Stroke Trial. Stroke 2002;33:1315–20.

[30] Vemmos KN, Tsivgoulis G, Spengos K, Zakopoulos N, Synetos A,Manios E, et al. U-shaped relationship between mortality andadmission blood pressure in patients with acute stroke. J Intern Med2004;255:257–65.

[31] Kurth T, Gaziano JM, Berger K, Manson JE, Buring JE, Levey AS, et al.Body mass index and the risk of stroke in men. Arch Intern Med2002;162:2557–62.

[32] O'Brien JM, Phillips GS, Ali NA, Lucarelli M, Marsh CB, LemeshowS. Body mass index is independently associated with hospital mortalityin mechanically ventilated adults with acute lung injury. Crit Care Med2006;34:738–44.

[33] Garrouste-Orgeas M, Troché G, Azoulay E, Caubel A, de Lassence A,Cheval C, et al. Body mass index: an additional prognostic factor inICU patients. Intensive Care Med 2004;30:437–43.

[34] Bo M, Massaia M, Raspo S, Bosco F, Cena P, Molaschi M, et al.Predictive factors of in-hospital mortality in older patients admitted to amedical intensive care unit. Am J Geriatr Soc 2003;51:529–33.

[35] Ray DE, Matchett SC, Baker K, Wasser T, Young MJ. The effect ofbody mass index on patient outcomes in a medical ICU. Chest2005;127:2125–31.

[36] Gustafson D. Adiposity indices and dementia. Lancet Neurol2006;5:713–20.

[37] Sacco RL, Shi T, Zamanillo MC, Kargman DE. Predictors of mortalityand recurrence after hospitalized cerebral infarction in an urbancommunity: the Northern Manhattan Stroke Study. Neurology1994;44:626–34.

[38] Weimar C, Ziegler A, König IR, Diener HC, on behalf of the GermanStroke Study Collaborators. Predicting functional outcome andsurvival after acute ischemic stroke. J Neurol 2002;249:888–95.