Using the National Institutes of Health Stroke Scale

O R I G I N A L
A R T I C L E (NIHSS) to predict the mortality and outcome of
patients with intracerebral haemorrhage
CME
CM Cheung 張春明
TH Tsoi 蔡德康 Objectives To investigate whether the National Institutes of Health
Sonny FK Hon 韓方光 Stroke Scale (NIHSS) can be used to predict mortality and
M Au-Yeung 歐陽敏 functional outcome in patients presenting with intracerebral
KL Shiu 邵家樂 haemorrhage.
CN Lee 李至南 Design Retrospective study of a prospectively collected cohort.
CY Huang 黃震遐 Setting Regional hospital, Hong Kong.
Patients A cohort of 359 patients presented to our hospital from 1996 to
2001 with their first-ever stroke and intracerebral haemorrhage.
Main outcome measures The sensitivity and specificity of the NIHSS with a cut-off point of
20 in predicting mortality at 30 days and 5 years, and a favourable
functional outcome at 5 years.
Results A total of 359 patients were available for analysis and were
divided into three subgroups according to the site and the size
of the haematoma. The NIHSS can predict 30-day mortality with
a sensitivity of 81% and a specificity of 90%. The NIHSS can
predict 5-year mortality with a sensitivity of 57% and a specificity
of 92%. In predicting favourable functional outcomes at 5 years,
the NIHSS had a sensitivity of 98% and a specificity of 16%.
Conclusions The NIHSS performed on admission can be used to predict
mortality at 30 days and 5 years as well as favourable functional
outcome at 5 years, all with an acceptable sensitivity and
specificity.

Introduction
Intracerebral haemorrhage (ICH) is a major cause of stroke among Asians. It contributes to
about 10 to 15% of strokes in western countries.1 In Hong Kong, ICH contributes to about
30% of all strokes.2,3 The disease differs from ischaemic stroke, as it confers higher early
mortality and poorer long-term outcomes.4 A method of predicting mortality within 30
days and good long-term functional outcomes could facilitate interviews with patients and
their relatives in terms of decisions for invasive and/or supportive care. For this purpose,
complicated scoring systems had been created but were difficult to use in daily clinical
practice. In the recent 6 years, two less complicated scoring systems have been published.5,6
The ICH score involves a scoring system consisting of the Glasgow Coma Scale (GCS), age,
Key words infratentorial origin, ICH volume, and presence of intraventricular haemorrhage. The new
Cerebral hemorrhage; Outcome
assessment (health care); Predictive ICH score uses National Institutes of Health Stroke Scale (NIHSS), admission temperature,
value of tests; Sensitivity and specificity pulse pressure, presence of intraventricular haemorrhage, and subarachnoid extension of
haemorrhage. Whilst these scores are useful for clinical trials and sophisticated research,
Hong Kong Med J 2008;14:367-70 a system based on commonly assessed clinical parameters for stroke patients could be
much more useful. The NIHSS score is commonly obtained in patients presenting with
Department of Medicine, Pamela Youde acute stroke. It consists of 15 items and a total score of 42 points. A score of 0 indicates
Nethersole Eastern Hospital, Chai Wan, no clinically relevant neurological abnormality. If a patient scores more than 20, it usually
Hong Kong
CM Cheung, MRCP, FHKAM (Medicine)
indicates a dense paralysis with impaired consciousness. We studied whether the NIHSS
TH Tsoi, FRCP, FHKAM (Medicine) can provide adequate predictive information in the course of routine clinical practice.
SFK Hon, MRCP, FHKAM (Medicine)
M Au-Yeung, MRCP, FHKAM (Medicine)
KL Shiu, MRCP, FHKAM (Medicine)
CN Lee, MRCP
Methods
CY Huang, FRACP, FHKAM (Medicine)
In our hospital, all patients with acute stroke attending the Accident and Emergency
Correspondence to: Dr CM Cheung Department are admitted to the Medical Department. On admission, a stroke is defined
E-mail: chun-ming@graduate.hku.hk as acute if the onset of symptoms has ensued within 5 days. Patients are transferred to the

Hong Kong Med J Vol 14 No 5 # October 2008 # www.hkmj.org 367

 # Cheung et al #

had been used prospectively by trained or certified

應用美國國家衛生研究院腦中風評估量表預 doctors to assess stroke patients within 2 days of
admission.
測腦出血病人的死亡率和康復效益
Patient data in the registry, in-patient hospital
目的 探討美國國家衛生研究院腦中風評估量表(NIHSS)能 records, out-patient follow-up notes, and subsequent
否用於預測腦出血病人的死亡率和康復效益。 hospital admission records were retrieved and
設計 對前瞻性收集數據的組群進行回顧研究。 retrospectively reviewed at 5 years or more after the
index stroke episode. For patients followed up in other
安排 香港一所地區醫院。 hospitals, their electronic hospital records, including
患者 1996至2001年，本醫院接收的359名首次中風兼腦出 discharge summary and out-patient progress notes,
血病人。 were traced. Patients were followed up in the
主要結果測量 以20為界點，應用NIHSS預測在第30天和第5年死亡 integrated clinics of our hospital and government
率，以及在第5年的康復效益情況的敏感度和特異性。 out-patient clinics. The patients could also have
been followed up by doctors in the rehabilitation
結果 本研究把359位病人按出血的位置和規模劃分為三
hospital, and sometimes in other hospitals (when
組作分析。NIHSS對預測第30天死亡率的敏感度為
they changed their residence). The modified Rankin
81%、特異性90%；NIHSS對預測第5年死亡率的敏感
score was estimated at 5 years, by using all of the
度為57%、特異性92%；NIHSS對預測第5年康復效益
written information collected in the medical record
情況的敏感度為98%、特異性16%。
and in the electronic record. We did not estimate the
結論 在病人進院時可應用NIHSS預測第30天和第5年的死 score before admission, so we cannot exclude other
亡率、以及第5年的康復效益情況，其敏感度和特異 factors affecting the score, eg chronic obstructive
性均在可接受的水平。 pulmonary disease. However, such factors were
not common in our cohort. Accurate classification
into five grades may be difficult but classification
into favourable outcome (a score of 0 to 2) or poor
neurosurgical team, only if neurosurgery is deemed outcome (a score of 3 to 5) appeared reasonable. We
necessary. From July 1996 onwards, all acute stroke usually described whether a patient could walk or
patients under the care of our department were was dependent for the activities of daily living during
assessed by the neurology team. We entered the data out-patient visits or admissions. The patients were
of all acute stroke patients into a stroke registry. This divided into three groups according to the size and
included: demographic data, risk factors for stroke, site of their haematoma. If the size of the haematoma
and stroke type (ischaemic, ICH, subarachnoid was estimated as more than 62.5 cm3, it was classified
haemorrhage). Non-contrast computed tomography as massive. Haematomas smaller than 62.5 cm3 were
of the brain was performed on all acute stroke classified into lobar (if within the brain parenchyma)
patients within 24 hours after admission, and the site or non-lobar (if in a deep part of the brain).
and the size of any haematoma recorded. All patients, The data were analysed by the Chi squared
who were enrolled in the first 5 years of our stroke test, if appropriate (using the Statistical Package for
registry with first-ever strokes and also diagnosed as the Social Sciences, Windows version 12.1; SPSS Inc,
having ICH, were identified for recruitment into the Chicago [IL], US). A P value of <0.05 (2-sided) was
present study. From 1997, in our institution the NIHSS taken to be statistically significant.

TABLE 1. Baseline demographic and clinical characteristics of the whole cohort and different types of intracerebral haemorrhage (ICH)
Characteristic Whole cohort Patients with small Patients with lobar Patients with
(n=359) non-lobar ICH (n=239) ICH (n=68) massive ICH (n=52)
Mean (range) age (years) 71.4 (27-98) 69.9 74.1 74.4
Male:female 192:167 141:98 28:40 23:29
No. (%) of patients with atrial fibrillation 15 (4) 13 (5) 1 (1) 1 (2)
No. (%) of patients with diabetes mellitus 66 (18) 49 (21) 8 (12) 9 (17)
No. (%) of patients with hypertension 201 (56) 144 (60) 31 (46) 26 (50)
No. (%) of patients with ischaemic heart disease 21 (6) 13 (5) 3 (4) 5 (10)
No. (%) of patients with hypercholesterolaemia 28 (8) 22 (9) 4 (6) 2 (4)
No. (%) of smokers 66 (18) 52 (22) 9 (13) 5 (10)
Mean NIHSS* score 16 13 17 32

* NIHSS denotes National Institutes of Health Stroke Scale

368 Hong Kong Med J Vol 14 No 5 # October 2008 # www.hkmj.org

 # National Institutes of Health Stroke Scale #

Results TABLE 2. The 30-day and 5-year mortality and favourable functional outcome rates at 5
years related to NIHSS assessments on admission*
A total of 431 patients suffering from first-ever stroke
NIHSS No. (%)
and diagnosed as having had cerebral haemorrhage
were enrolled in our stroke registry during the 5-year 30-Day mortality 5-Year mortality Favourable 5-year functional
outcome for survivors†
period from 28 July 1996 to 27 July 2001. The outcome
of 24 first-ever ICH patients could not be traced at >20 96/120 (80) 106/120 (88) 2/24 (8)
the time of review. The remaining 407 patients were 11-20 15/61 (25) 35/61 (57) 9/46 (20)
included in this analysis. Among these 407 patients, 6-10 3/56 (5) 14/56 (25) 17/53 (32)
only 359 patients had NIHSS assessments on the day 0-5 5/122 (4) 31/122 (25) 73/117 (62)
of admission.
* NIHSS denotes National Institutes of Health Stroke Scale
The baseline characteristics of these 359 † P<0.0001 when using 0-20 vs >20 for 30-day, 5-year mortality and good outcome
patients are shown in Table 1. The relationships
between NIHSS assessments on admission and 30-
day mortality, 5-year mortality, and the favourable TABLE 3. The 30-day mortality and favourable functional outcome rates at 5 years for
subgroups with massive, lobar, and non-lobar (small) intracerebral haemorrhage (ICH)
functional outcome at 5 years for the whole cohort are
shown in Table 2. The corresponding relationships for NIHSS* in various No. (%)
subgroups
the three different ICH subgroups are shown in Table 30-Day mortality Favourable 5-year functional
3. Except for the massive ICH subgroup (in which outcome for survivors
there were too few 30-day survivors), the relationship Massive ICH †

between NIHSS assessments on admission and 30- >20 38/40 (95) 0/2 (0)
day mortality or 5-year functional outcome holds true 0-20 5/12 (42) 2/7 (29)
for all subgroups.
Lobar ICH
‡

Age did not affect the poor outcome of those

>20 18/25 (72) 0/17 (0)
with NIHSS scores of higher than 20. However, among
0-20 5/44 (11) 14/39 (36)
those with scores of less than 20, younger patients
survived better (Table 4). Compared to older patients, Small non-lobar ICH §

those who were younger also had better 5-year >20 40/55 (73) 2/15 (13)
functional outcomes regardless of NIHSS category 0-20 13/183 (7) 83/170 (49)
(Table 5). In all, 41 patients had a second stroke after
* NIHSS denotes National Institutes of Health Stroke Scale
surviving the first 30 days. Adjustment of the results † P<0.0001 when using 0-20 vs >20 for 30-day mortality, P=1 for functional outcome
to their modified Rankin scores just before their ‡ P<0.0001 when using 0-20 vs >20 for mortality and P<0.01 for good outcome
second stroke shows that there would have been § P<0.02 for both mortality and good outcome
80 favourable outcomes for those with initial NIHSS
scores of 0-5, and 19 such outcomes for those with TABLE 4. The 30-day mortality for different age-groups
scores of 6-10. This would have further improved
NIHSS* No. (%)
the overall outcome of the whole cohort with NIHSS
scores of <20 (108 favourable outcomes instead of 99), <60 Years ≥60 Years
and accentuated the disparity of outcomes between >20 †
13/16 (81) 83/104 (80)
those with NIHSS scores of >20 and <20. By itself, age 0-20 1/56 (2) 22/183 (12)
did not increase the risk of recurrent stroke over 5
* NIHSS denotes National Institutes of Health Stroke Scale
years (in the group aged <60 years, the rate was 15%; † P>0.05 for NIHSS>20; P<0.05 for NIHSS≤20
and in those aged ≥60 years, it was 10%). Recurrent
stroke therefore was not the cause of less favourable
functional outcomes in older patients. TABLE 5. Favourable functional outcomes at 5 years for different
age-groups
In total, 186 patients died during the study
period, 119 within 30 days, and 67 between 30 days and NIHSS* No. (%)
5 years. The number of patients who died of vascular <60 Years ≥60 Years
causes (ie not counting aspiration pneumonia) was >20† 2/16 (13) 0/104 (0)
35. 0-20 43/56 (77) 56/183 (31)

* NIHSS denotes National Institutes of Health Stroke Scale

Discussion †
P<0.05 when comparing two age-groups for favourable functional
outcome in different NIHSS
This study was hospital-based. Therefore patients
with very minor deficits (not hospitalised whatever
the reason), those who refused admission, those with our cohort.
severe deficits who died before admission, and those When the NIHSS is used to predict 30-day
admitted to private hospitals were not included in mortality, it has good sensitivity (81%) and specificity

Hong Kong Med J Vol 14 No 5 # October 2008 # www.hkmj.org 369

 # Cheung et al #

(90%) using a cut-off point of 20 (0-20 vs >20). Using Furthermore, the ICH scores have not
the same cut-off point to predict 5-year mortality, previously been studied in terms of predicting long-
the NIHSS has a lower sensitivity (57%) but good term prognosis; only outcome at 30 or 100 days has
specificity (92%). When using an NIHSS cut-off point been reported.5-7 For predicting favourable outcome
of ≤20 to predict a good outcome among survivors from stroke onset at 5 years, the NIHSS alone already
at 5 years, its sensitivity was 98% but specificity was achieved an acceptable negative predictive value.
16%. If the cut-off point is changed from 20 to 5 (0-5 This information is important as the busy clinician can
vs >5), sensitivity was reduced to 72% but specificity interview relatives in terms of life-support decisions;
increased to 68%. When the NIHSS is used to predict a score of >20 means a high chance of death in 30
30-day mortality, it has good sensitivity (81%) and days and virtually no chance of favourable long-term
specificity (90%) using a cut-off point of 20 (0-20 vs recovery, even in patients surviving 30 days. In which
>20). Using the same cut-off point to predict 5-year case, statistically the chance the patient would die
mortality, the NIHSS has a lower sensitivity (57%) but within 30 days would be 80%, and within 5 years it
good specificity (92%). When using an NIHSS cut- would be 88%. The chance of a poor outcome at 5
off point of ≤20 to predict a good outcome among years from the stroke onset would be 98%.
survivors at 5 years, its sensitivity was 98% but Taking age into consideration did not affect the
specificity was 16%. If the cut-off point is changed predictive value of a high NIHSS on 30-day mortality.
from 20 to 5 (0-5 vs >5), sensitivity was reduced to 72% However, chronological age may reflect concomitant
but specificity increased to 68%. disease burden, and less favourable response to the
In two earlier studies which have examined the neurological insult. Compared with persons aged 60
impact of NIHSS on outcome in cerebral haemorrhage, to 80 years, younger patients with an NIHSS score of
Cheung and Zou6 found that the NIHSS assessment ≤20 had a lower 30-day mortality rate (2% vs 12%),
but not the GCS was an independent predictor of which was statistically significant (P<0.05, Table 4). In
mortality and outcome at 30 days. A study published predicting favourable functional outcomes among
in 2006 also used the NIHSS to predict outcome those with an NIHSS score of ≤20, younger patients
at 100 days, when the patients were assessed at (<60 years) also faired better, although this did not
admission.7 The investigators assigned scores for: detract from the adverse prognosis of a high score on
NIHSS assessments (0-5=0; 6-10=1; 11-15=2; 16-20=3; functional outcome (Table 5).
>20=4), the level of consciousness (alert=0; drowsy=1; In conclusion, NIHSS assessments performed
stuporous=2; comatose=3), and age (<60=0; 60-69=1; at admission can be used to predict the 30-day
70-79=2; ≥80=3). Using a total score cut-off point of <3 and 5-year mortality as well as long-term outcome
to predict complete recovery and >7 to predict death, among survivors. Irrespective of age or type of
yielded a sensitivity of 74% and specificity of 84% for the cerebral haemorrhage, an NIHSS score of >20
the former and corresponding figures for the latter is a strong predictor of death or poor functional
were 44% and 98%. outcome. Whilst additional information, such as
Compared to previous scoring systems, the age, precise location and size of the haemorrhage,
NIHSS alone is much simpler to use. A cut-off point at intraventricular and subarachnoid extension,
0-20 versus >20 already achieved sufficient sensitivity temperature and blood pressure may improve
and specificity for predicting 30-day mortality, close prognostic precision, the busy clinician may find
to what was reported for the original, modified and the NIHSS sufficient for most clinical management
Essen ICH scores. decisions and counselling.

References

1. Caplan LR. Intracerebral haemorrhage. Lancet 1992;339: Johnston SC. The ICH score: a simple, reliable grading scale
656-8. for intracerebral hemorrhage. Stroke 2001;32:891-7.
2. Huang CY, Chan FL, Yu YL, Woo E, Chin D. Cerebrovascular 6. Cheung RT, Zou LY. Use of the original, modified, or
disease in Hong Kong Chinese. Stroke 1990;21:230-5. new intracerebral hemorrhage score to predict mortality
3. Kay R, Woo J, Kreel L, Wong HY, Teah R, Nicholls MG. and morbidity after intracerebral hemorrhage. Stroke
Stroke subtypes among Chinese living in Hong Kong: the 2003;34:1717-22.
Shatin Stroke Registry. Neurology 1992;42:985-7. 7. Weimar C, Benemann J, Diener HC; German Stroke Study
4. Cheung CM, Tsoi TH, Hon SF, et al. Outcomes after first- Collaboration. Development and validation of the Essen
ever stroke. Hong Kong Med J 2007;13:95-9. Intracerebral Haemorrhage Score. J Neurol Neurosurg
5. Hemphill JC 3rd, Bonovich DC, Besmertis L, Manley GT, Psychiatry 2006;77:601-5.

370 Hong Kong Med J Vol 14 No 5 # October 2008 # www.hkmj.org