The outcome of severe closed head injury

LAWRENCE F. MARSHALL, M.D., TIIERESA GAUTILLE, B.S.N., AND

MELVILLE R. KLAUBER, M.D.
Division of Neurosurgery. University ~?fCa/!fornia Medical Center, San Diego, California

HOWARD M. EISENBERG, M.D.

Division of Neurosurgery, Univo:siO' ~[ Texas Medical Branch, Galveston, Texas

JOHN A. JANE, M.D.

Department t f Neurosurgery. University ~?[Virginia School of Medicine, Charlottesville, Virginia

THOMAS G. LUERSSEN, M.D.

Division of Neurosurgery, Indiana University Medical Center, Indianapolis, Indiana

ANTHONY MARMAROU, PH.D.

Division of Neurosurgery, 34edical College of Virginia, Richmond, Virginia

MARY A. FOULKES, PH.D.

Biometry and Field Studies Branch, National Institute of Neurological Disorders and Stroke, Bethesda,
Maryland

u- The outcome of severe head injury was prospectively studied in patients enrolled in the Traumatic Coma
Data Bank (TCDB) during the 45-month period from January 1, 1984, through September 30, 1987. Data
were collected on 1030 consecutive patients admitted with severe head injury (defined as a Glasgow Coma
Scale (GCS) score of 8 or less following nonsurgical resuscitation). Of these, 284 either were brain-dead on
admission or had a gunshot wound to the brain. Patients in these two groups were excluded, leaving 746
patients available for this analysis.
The overall mortality rate for the 746 patients was 36%, determined at 6 months postinjury. As expected,
the mortality rate progressively decreased from 76% in patients with a postresuscitation GCS score of 3 to
approximately 18% for patients with a GCS score of 6, 7, or 8. Among the patients with nonsurgical lesions
(overall mortality rate, 31%), the mortality rate was higher in those having an increased likelihood of elevated
intracranial pressure as assessed by a new classification of head injury based on the computerized tomography
findings. In the 276 patients undergoing craniotomy, the mortality rate was 39%. Half of the patients with
acute subdural hematomas died - - a substantial improvement over results in previous reports. Outcome
differences between the four TCDB centers were small and were, in part, explicable by differences in patient
age and the type and severity of injury.
This study describes head injury outcome in four selected head-injury centers. It indicates that a mortality
rate of approximately 35% is to be expected in such patients admitted to experienced neurosurgical units.

KEY WORDS 9 Traumatic Coma Data Bank 9 head injury 9 intracranial pressure .
outcome . coma

N 1979, a pilot study to determine the feasibility of nia, San Diego) in cooperation with the Biometry and

I a Traumatic Coma Data Bank (TCDB) was under-

taken. The methodology and preliminary results of
that study have been described previously. ~3 Beginning
Field Studies Branch of the National Institute of Neu-
rological Disorders and Stroke (NINDS) entered into
the main phase of the TCDB. The objective of this
on January l, 1984, four of the centers involved in the study was to gather prospective data on a large number
pilot phase (the University of Virginia, Charlottesville, of patients suffering severe head injury so that specific
the University of Texas, Galveston, the Medical College questions regarding particular subsets of these patients
of Virginia, Richmond, and the University of Califor- could be addressed.

$28 J. Neurosurg. / Volume 75/November, 1991

Outcome of severe closed head injury

All patients admitted to the hospital with a Glasgow tality rates between TCDB centers were tested, taking
Coma Scale2~(GCS) score of 8 or less, or who deterio- into account patient age, pupil reactivity, postresusci-
rated to a GCS score of 8 or less within 48 hours of tation motor score, and injury (that is, chest or abdom-
admission, were included. The TCDB centers agreed inal vs. other or no multiple injury)Y For this, the
that in all cases where the patient was felt to be sal- BMDP statistical computer software was used. 6 This
vageable intracranial pressure (ICP) would be monitor- choice of predictors was based on a previous study of
ed and computerized tomography (CT) scans would be 7912 head-inj ured patients. ~2Goodness-of-fit (indicated
obtained. No absolute agreement regarding the level at by p values close to 1) of the logistic regression models
which ICP must be treated was established, but there was tested using the method of Hosmer and Leme-
was general agreement that sustained ICP levels of 25 show) In addition, a logistic model was obtained for
mm Hg or greater would be treated syslematically by the mortality rate associated with each of the CT diag-
all TCDB centers. Therefore, while no absolute man- nostic categories. The final model was obtained using
agement protocol was stipulated, a center-by-center backward elimination from the following limited list of
comparison demonstrated that treatment for closed predictors: patient age; highest motor score in the first
head injury was relatively homogeneous. 72 hours postinjury; lowest motor score (derived from
the GCS); highest systolic blood pressure; lowest systolic
Clinical Material and Methods blood pressure; highest ICP; lowest ICP; Abbreviated
Injury Scale ~ score; blood alcohol level; postresuscita-
Each of the four neurosurgical centers participating tion motor score; and postresuscitation pupil reactivity.
in the TCDB employed nurse clinicians to gather data A new classification of head injury was used to
on all patients admitted with severe head injury. The categorize the types of abnormalities visualized on CT
data-collection instruments and the methodology for scanning. Previously, head injury has been divided pri-
the data collection are described in a separate report. 7 marily into "mass lesion" and "no mass lesion" cate-
The details of the patient's evaluation at the scene, at gories. In addition, a category for brain-dead based
the first hospital, and/or at the TCDB receiving hospital solely on the clinical examination was employed to
were defined. In each case, the steps taken to resuscitate ensure that all of the severely injured patients were
the patient, the type of ICP monitoring device used, categorized even in the absence of a CT scan. This new
the intensive care unit and clinical course during the classification system is described in detail elsewhere, ~4
hospital stay, the results of CT scans, and the surgical but a brief description of each of the categories is as
procedure or procedures performed on the patient were follows: Diffuse Injury I = no visible intracranial pa-
also specified. At the patient's discharge or death, a thology on CT scan; Diffuse Injury lI = cisterns present
narrative was dictated describing the pertinent features with midline shift of 0 to 5 mm and/or there is no high-
of the clinical course as well as the initial outcome. or mixed-density lesion present greater than 25 cc (may
Outcome at discharge was assessed using the Glasgow include bone fragments and foreign bodies); Diffuse
Outcome Scale9 (GOS). All patients who were capable Injury III (swelling) = cisterns compressed or absent
underwent daily screening to assess their orientation with midline shift of 0 to 5 ram, no high- or mixed-
and short- and long-term memory using the Galveston density lesion greater than 25 cc; Diffuse Injury IV =
Orientation and Amnesia Test (GOAT). Baseline neu- midline shift greater than 5 mm, no high- or mixed-
ropsychological testing was carried out when the patient density lesion greater than 25 cc; evacuated mass le-
had achieved a GOAT score of 75 or greater on two sion = any lesion surgically evacuated: nonevacualed
consecutive examinations. Additional testing was per- mass lesion = high- or mixed-density lesion greater than
formed at 6 months, 1 year, and 2 years in order to 25 cc, not surgically evacuated; brain-dead = no brain-
assess the longitudinal behavioral and cognitive conse- stem reflexes (without improvement), flaccidity, fixed
quences of brain injury. At the neuropsychological nonreactive pupils, and no spontaneous respirations
follow-up examination, a neurological assessment was with a normal PaCO2 (spinal reflexes permitted).
carried out by the TCDB neurosurgeon.
Tests of significance in contingency tables were per-
formed using Pearson's chi-square (unordered varia- Results
bles) and the Kruskal-Wallis test (one variable ordered). As of September 30, 1987, 1030 patients had been
When data were sparse, either exact p values were entered into the TCDB. Of these, 284 were brain-dead
obtained or empirical randomization tests were per- following nonsurgical resuscitation or had a gunshot
formed using 10,000 randomizations. Many of the wound to the head and were excluded from this analy-
tables showed a high degree of association, thus requir- sis. The outcome of the remaining 746 patients is the
ing a great number of randomizations in order to make subject of this report. Outcome is described at last
statements regarding significance (such as p < 0.001) contact with the patient; however, the outcome at the
with confidence. All tables were analyzed using the time of acute hospital discharge is also shown in some
StatXact computer program. instances in spite of variations in time to discharge.
Logistic regression and the Cox proportional hazards This is to illustrate the relatively small changes in the
models were developed, and other differences in mor- number of deaths following hospitalization and the

J. Neurosurg. / Volume 75/November, 1991 $29

 L. F. Marshall, et al.

TABLE 1
Intracranial diagnosis correlated with mechanism of injury

Motor-VehicleAccident Pedestrian Motorcycle Fall or Other

Intracranial Driver Passenger Injury Accident Jump Injury Total
Diagnosis Cases
No. Percent No. Percent No. Percent No. Percent No. Percent No. Percent
unknown 5 2.5 2 2.2 1 1.0 3 2.7 3 2.5 3 2.0 17
Diffuse InjuryI (no 18 9.0 3 3.2 10 13.0 10 9.0 4 3.3 7 5.0 52
visiblepathology)
Diffuse InjuryII 67 34.0 26 28.0 17 22.0 28 25.0 15 12.6 2 16.6 177
Diffuse InjuryIII 46 23.0 29 31.2 17 22.0 24 21.0 11 9.0 26 18.0 153
(swelling)
Diffuse InjuryIV (shift) 5 2.5 5 5.4 8 10.0 5 4.4 2 1.6 7 5.0 32
evacuatedmass 45 23.0 26 23.0 22 29.0 37 33.0 77 65.0 69 47.9 276
nonevacuatedmass 12 6.0 2 2.0 2 2.6 6 5.0 7 6.0 8 4.8 36
brain-steminjury 2 1.0 0 0 0 0 0 0 0 0 1 0.7 3
totals 200 100 93 100 77 100 113 100 119 100 144 100 746

rather substantial variation in the n u m b e r of patients associated with a disproportionate number of evacuated
who were vegetative at the time of discharge but who mass lesions. This reflects not only the influence of the
subsequently improved or died. As in other reports on mechanism of injury, but also the importance of age
traumatic injury, there was a predominance of males, in determining the kind of intracranial pathology that
with a male:female ratio of approximately 3:1. The develops following injury. 23 Patients in these categories
mean age of the patients was slightly under 30 years tended to be older than the overall patient population.
(29.5 years); the median age was 25 years. The time interval between admission and the deter-
Most deaths occurred soon after admission: 28% mination of the postresuscitation GCS score was be-
during the first day of hospitalization, 78% within 1 tween 30 minutes and 1 hour. The number of patients
week, and 90% within 2 weeks; the median survival with a postresuscitation score of 3 or 4 is strikingly
time of those who died was 2 days. The distribution of similar among the TCDB centers (Table 2); however,
time to discharge was also highly skewed: the median the differences in preresuscitation scores (Table 3) in-
hospitalization time was 32.5 days; 46% were dis- dicate that the initial GCS score is not as good an
charged within 30 days and 79% within 60 days, with indicator of patient status as the postresuscitation
a m a x i m u m length of stay 266 days. The follow-up scores. Although there is no statistically significant het-
time to last contact for survivors ranged from 11 to erogeneity by TCDB center shown by either table, the
1199 days, with a median of 674 days. Two-thirds of Kruskal-Wallis chi-square test (0.94 vs. 4.18) confirms
the patients have been followed for over 1 year and the visual impression.
one-third for over 2 years. In contrast to the relative homogeneity of trauma
severity among the TCDB centers, there were significant
Intracranial Diagnosis and Patient Status differences in the intracranial diagnosis (Table 4). Re-
There was a strong association (p < 0.001) between gardless of whether the unknowns were included, the
intracranial diagnosis and mechanism of injury (Table randomization test was very highly significant (p <
1). The categories of "fall or j u m p " and "other" were 0.001). There were few patients with Diffuse Injury I

TABLE 2
Postresuscitation GlasgowComa Scale (GCS) scorecorrelated with center

Postresuscitation Center 1 Center 2 Center 3 Center 4 Total

GCS Score2~ No. Percent No. Percent No. Percent No. Percent Cases
3 23 17.0 62 24.0 40 21.5 35 21.1 160
4 24 17.8 34 13.1 26 14.0 27 16.3 111
5 31 23.0 25 9.7 13 7.0 13 7.8 82
6 9 6.7 36 13.9 39 21.0 29 17.5 113
7 19 14.l 54 20.8 43 23.0 32 19.3 148
8 11 8.1 21 8.1 13 7.0 17 10.2 62
_>9 10 7.4 27 10.4 12 6.5 13 7.8 62
unknown 8 5.9 0 0 0 0 0 0 8
totals 135 100 259 100 186 100 166 100 746

$30 J. Neurosurg. / Volume 75/November, 1991

 Outcome of severe closed head injury

TABLE 3
Preresuscitation Glasgow Coma Scale (GCS) score correlated with center

Preresuscitation Center 1 Center 2 Center 3 Center 4 Total

GCS Score:~ No. Percent No. Percent No. Percent No. Percent Cases
3 38 28.2 93 35.9 79 42.5 77 46.4 287
4 25 18.5 25 9.7 20 10.8 19 11.4 89
5 22 16.3 33 12.7 11 5.9 6 3.6 72
6 14 10.4 35 13.5 24 12.9 I1 6.6 84
7 17 12.5 16 6.2 24 12.9 20 12.1 77
8 5 3.7 12 4.6 13 6.9 11 6.6 41
-> 9 14 10.4 45 17.4 15 8.1 22 13.3 96
totals 135 100 259 100 186 100 166 100 746

TABLE 4
lntracranial diagnosis correlated with center
Center 1 Center 2 Center 3 Center 4 Total
Intracranial Diagnosis
No. Percent No. Percent No. Percent No. Percent Cases
Diffuse Injury 1 (no visible 2 1.5 20 7.7 18 9.7 12 7.2 52
pathology)
Diffuse Injury !I 26 19.3 58 22.4 53 28.5 40 24.1 177
Diffuse Injury !II (swelling) 42 3 l. 1 46 17.8 26 14.0 39 23.5 153
Diffuse Injury IV (shift) 6 4.4 18 7.0 4 2.1 4 2.4 32
evacuated mass 44 32.6 94 36.3 79 42.5 59 35.5 276
nonevacuated mass 7 5.2 18 7.0 2 1.1 9 5.5 36
brain-stem injury 0 0 3 1.0 0 0 0 0 3
unknown 8 5.9 2 0.8 4 2.1 3 1.8 17
totals 135 100 259 100 186 100 166 100 746

( n o visible pathology) in C e n t e r 1, whereas t h a t c e n t e r ing a good o u t c o m e . As w o u l d be expected, there was

h a d a h i g h p r o p o r t i o n o f p a t i e n t s w i t h Diffuse I n j u r y a n e x t r e m e l y strong relationship (p < 0.001) b e t w e e n
III (swelling). C e n t e r 2 h a d t h e highest p r o p o r t i o n o f i n t r a c r a n i a l diagnosis a n d o u t c o m e ( T a b l e 5). Diagnosis
p a t i e n t s w i t h Diffuse I n j u r y IV (shift). E v a c u a t e d m a s s o f Diffuse I n j u r y IV (shift), n o n e v a c u a t e d m a s s lesions,
lesions were fairly equally d i s t r i b u t e d a m o n g the cen- or b r a i n - s t e m i n j u r y was associated with a higher t h a n
ters, b u t C e n t e r 3 s h o w e d few n o n e v a c u a t e d mass le- average p r o p o r t i o n o f deaths, while Diffuse I n j u r y I (no
sions. T h e t h r e e p a t i e n t s w i t h b r a i n - s t e m i n j u r y were visible pathology) a n d Diffuse I n j u r y II caused relatively
all in C e n t e r 2. few deaths. T h e r e m a i n d e r o f the patients, those with
Diffuse I n j u r y III (swelling) a n d e v a c u a t e d mass lesions,
Outcome h a d o u t c o m e s close to t h e average. Relatively high
A t the t i m e o f hospital discharge, 33 % o f t h e p a t i e n t s p r o p o r t i o n s o f p a t i e n t s w i t h Diffuse I n j u r y III a n d I V
h a d died a n d 14% were vegetative, w i t h only 7% show- (swelling a n d shift) were vegetative. N o t a b l y only o n e

TABLE 5
Discharge Glasgow Outcome Scale score (GOS) correlated with intracranial diagnosis
Diffuse
Diffuse Diffuse Non-
Injury I Diffuse Injury III Injury IV Evacuated evacuated Brain-Stem Unknown Total
GOS9 At (no visible Injury II Mass Injury Cases
Discharge pathology) (swelling) (shift) Mass

No. Percent No. Percent No. Percent No. Percent No. Percent No. Percent No. Percent No. Percent No. Percent
good 14 27.0 15 8.5 5 3.3 1 3.1 14 5.1 1 2.8 0 0 0 0 50 7.0
moderate disability 18 34.6 46 26.0 20 13.1 1 3.1 49 17.7 3 8.3 0 0 1 5.9 138 18.5
severe disability 10 19.2 72 40.7 41 26.8 6 18.8 72 26.1 7 19.4 1 33.3 0 0 209 28.0
vegetative 5 9.6 20 11.3 35 22.9 6 18.8 34 12.3 6 16.7 0 0 0 0 106 14.0
dead 5 9.6 24 13.5 52 34.0 18 56.2 107 38.8 19 52.8 2 66.7 16 94.1 243 32.5
totals 52 100 177 1013 153 100 32 100 276 100 36 100 3 100 17 100 746 100

J. Neurosurg. / Volume 75/November, 1991 $31

 L. F. Marshall, et al.

TABLE 6 category, Diffuse Injury I (no visible pathology), had a

Outcome at last contact of Diffuse Injury H group correlated high proportion of good outcomes (27%).
with patient age The category of "nonevacuated mass lesion" deserves
Outcome Age _< 40 yrs Age > 40 yrs Total Cases
special corn ment. Of the 36 patients, 19 (53 %) suffered
At Last a fatal outcome during hospitalization (Table 5). Pa-
Contact* No. Percent No. Percent No. Percent tients were placed in this category under two circum-
good 15 10.0 0 0 15 8.5 stances: 1) although they met the mass lesion criteria,
moderate disability 44 28.7 2 8.3 46 26.0 the TCDB neurosurgeon considered that they were
severe disability 63 4l. 1 9 37.5 72 40.7 unsalvageable; or 2) the surgeon chose not to evacuate
vegetative 17 11.1 3 12.5 20 11.3 the clot because medical management was believed to
dead 14 9.1 10 41.7 24 13.6
totals 153 100 24 100 177 100
be adequate in treating the patient.
The patient cohort was further divided by age, group-
* Outcome assessed by the Glasgow Outcome Scale?
ing those aged 40 years or younger and those aged over

TABLE 7
Outcome at last contact correlated with postresuscitation Glasgow Coma Scale score *

Postresuscitation Glasgow Coma Scale Score

Outcome Total
Un- 3 3.3# 4 5 6 7 8 _>9 Cases
At Last known
Contact
No. % No. % No. % No. % No. % No. % No. % No. % No. % No. %
good 0 0 4 4.1 16 25.4 7 6.3 10 12.2 33 29.2 69 46.6 34 54.8 25 40.3 198 26.5
moderate disability 0 0 3 3.1 12 19.0 9 8.1 14 17.1 24 7"1.3 33 22.3 14 22.6 13 21.0 122 16.4
severe disability 0 0 10 10.3 11 17.5 21 18.9 19 23.2 26 23.0 17 11.5 4 6.5 8 13.0 116 15.6
vegetative 0 0 4 4.1 5 7.9 12 10.8 6 7.3 6 5.3 3 2.0 3 4.8 0 0 39 5.2
dead 8 100 76 78.4 19 30.2 62 55.9 33 40.2 24 21.2 26 17.6 7 11.3 16 25.8 271 36.3
totals 8 100 97 100 63 100 111 100 82 100 113 100 148 100 62 100 62 100 746 100
* Outcome assessed by the Glasgow Outcome Scale? Significance of difference (Kruskal-Wallis): p < 0.001.
t Patients who are untestable due to paralytic agents.

TABLE 8
History of pupil reaclivity in patients dead or vegetative at discharge and at last contact

No. of % Cases Dead or Vegetative

Pupil Reactivity
Cases At Discharge At Last Contact
pupils normal postresuscitation, remaining normal throughout hospital course 235 10.0 8.5
pupils normal postresuscitation, then one pupil abnormal for at least one observation 18 61.0 5O.0
pupils normal postresuscitation, then both abnormal for at least one observation 108 79.0 80.0
both pupils abnormal postresuscitation 209 82.0 74.0
one pupil abnormal postresuscitation 70 47.0 34.3
pupil status unknown 106 unknown unknown
total cases 746

TABLE 9
Outcome at hospital discharge correlated with mechanism of injury*

Motor-Vehicle Accident Pedestrian Motorcycle Fall or Other

Outcome Driver Passenger Injury Accident Jump Injury Total
At Discharge Cases
No. Percent No. Percent No. Percent No. Percent No. Percent No. Percent
good 10 5.0 9 9.7 6 7.8 7 6.2 7 5.9 11 7.6 50
moderate disability 32 16.0 19 20.4 7 9.1 30 26.6 23 19.3 27 18.8 138
severe disability 63 31.5 27 29.0 20 26.0 33 29.2 27 22.7 39 27.1 209
vegetative 34 17.0 9 9.7 11 14.3 17 15.0 14 11.8 21 14.6 106
dead 61 30.5 29 31.2 33 42.8 26 23.0 48 40.3 46 31.9 243
totals 200 100 93 100 77 100 113 100 119 100 144 100 746
* Outcome assessed by the Glasgow Outcome Scale.9 Significance of difference (Kruskal-Wallis): p = 0.040.

$32 J. Neurosurg. / Volume 75/November, 1991

Outcome of severe closed head injury

40 years (Table 6). This split was selected to demon- reactive pupils throughout their hospital course, only
strate a maximum divergence of mortality rates; there 8.5% were dead or vegetative at last contact. In contrast,
was little difference in overall mortality rates in patients among patients who had reactive pupils following re-
between the ages of 1 and 40 years. Only 14 (9%) of suscitation and then developed one pupillary abnor-
153 patients under 40 years of age died versus a mor- mality, nine (50%) of 18 were dead or vegetative. When
tality rate of 41% in patients over 40 years of age. both pupils were fixed and unreactive immediately
Furthermore, no patients over 40 years of age with a following resuscitation, 151 (74%) of 209 died or were
CT diagnosis of Diffuse Injury II obtained a good left vegetative.
outcome. There was little difference in outcome between pas-
The well-established high degree of association be- sengers and drivers of motor vehicles (Table 9). Partic-
tween GCS and GOS scores is confirmed by our data ularly noteworthy was the relatively lower mortality
(Table 7). However, in our study the mortality rate for rate of patients who were injured as a result of operating
patients with a GCS score of 3.3 and 9 fell between the a motorcycle when compared to the group as a whole.
rate for those with a GCS score of 5 or more. A GCS This is a product of two factors: patients who suffer
score of 3.3 is allotted to patients with induced physi- motorcycle or moped injuries are younger compared to
ological paralysis since their true motor score is not the TCDB cohort as a whole (median age 24 years),
known due to the use of muscle relaxants. These pa- and they have slightly higher motor scores following
tients were placed in a separate category and were not resuscitation. An older group, the pedestrians, fared
analyzed with patients who had a GCS score of 3. It is particularly poorly. This is a product of not only their
of interest that patients with GCS scores of 9 or greater age but also a greater frequency of systemic compro-
who had to have deteriorated in order to be in the study mise, particularly hypotension and hypoxia.
did substantially worse than patients with initial GCS There were 276 patients with surgically treated intra-
scores of 6, 7, or 8. Therefore, the GCS scores, as com- cranial hemorrhages. In contrast to the entire cohort of
puted in the present study, did not lead to a strictly the TCDB (in which the median age was 25.2 years),
increasing measure of severity; hence, the Kruskal- patients with surgical intracranial hemorrhages had a
Wallis test was used to test GCS as an unordered var- median age of 30 years. The outcome at discharge by
iable versus GOS as an ordered variable (p < 0.001). type of hematoma is shown in Table 10, which is
The importance of pupillary reactivity in predicting divided into two major age epochs in order to empha-
outcome is shown in Table 8. Among patients who had size the influence of age on outcome in this group. Fifty
percent of patients with evacuated subdural hematomas
died, while only 26% of those with intracerebral hem-
orrhages and 18% of those with epidural hematomas
TABLE 10 suffered mortal injuries. While the sample size of pa-
Outcome at last contact by type of evacuated mass lesion and tients with extradural hematomas is small, the mortality
patienl age* rate of 18% is almost a 50% improvement over that
reported during the pilot phase of the TCDB by Seelig,
Age Age Total Cases et al., '7 in spite of the fact that there were no detectable
Outcome _<40 yrs > 40 yrs differences between the two patient cohorts. This sug-
No. % No. % No. % gests that the improvement in the rapidity with which
evacuated epidural hematomal" these patients were treated, either initially or when
good 6 15.4 0 0.0 6 13.3 deterioration first occurred, may have been a significant
moderate disability 13 33.3 2 33.3 15 33.3 factor in altering outcome. Although the influence of
severe disability l0 25,6 0 0.0 10 22.2
vegetative 5 12.8 1 16.7 6 13.3
age in acute severe head injury is described in detail
dead 5 12.8 3 50.0 8 17.8 elsewhere, 23 Table 10 dramatically illustrates the im-
totals 39 100 6 100 45 100 portance of age and its influence on mortality in pa-
evacuated subdural hematoma~ tients with evacuated mass lesions. The lack of statistical
good 4 4.0 0 0.0 4 2.5 significance of the association between outcome and
moderate disability 14 14.0 4 6.8 18 11.3 age for evacuated epidural hematomas is clearly due to
severe disability 31 31.0 10 6.9 41 25.8
vegetative 12 12.0 4 6.8 16 10.1
the small number (six) of such patients over age 40
dead 39 39.0 41 69.5 80 50.3 years. Clearly, the association with age is similar to that
totals 100 100 59 100 159 100 for the other two lesion types.
evacuated intracerebral hematoma, Patients were followed for a median postinjury period
good 4 7.5 0 0.0 4 5.6 of 674 days to determine outcome status after hospital
moderate disability 14 26.4 1 5.6 15 21.1 discharge. Note that there are significant differences in
severe disability 18 34.0 3 16.7 21 29.6
vegetative 7 13.2 5 27.8 12 16.9
outcome by TCDB center when no confounding factors
dead 10 18.9 9 50.0 19 26.8 are taken into account (Table 11); however, regardless
totals 53 100 18 100 71 100 of the analytical approach used, statistical differences
*Outcomeaccordingto the GlasgowOutcomeScaler Significance: in mortality rates are similar when additional covariates
? = p = 0.10; :~ = p < 0 . 0 0 1 . are considered. An excellent fit of the logistic model

J. Neurosurg. / Volume 75/November, 1991 $33

 L. F. Marshall, e t al.

was obtained using the mortality risk factors: patient 24 hours after trauma; for 97% of them, pronounce-
age, pupil reactivity, and postresuscitation motor score ment occurred less than 6 hours postinjury. In the
for the Hosmer-Lemeshow test (p = 0.82). With these Multinational Data Bank results reported by Jennett
four predictors in the model, T C D B center differences and Teasdale, ~~patients deemed unsalvageable at first
could be substantially explained by chance (p = 0.84). hospital assessment (a broader category of exclusion
Likewise, the Cox proportional hazards model, which than the one used here) were excluded from analysis.
takes into account the follow-up period and time to At a median postinjury follow-up time of 674 days,
death, showed no significant TCDB center differences approximately 41.6% of the patients were left vegetative
(p = 0.06, unadjusted for the three predictors, and p = or dead following severe head injury. Some improve-
0.30, adjusted). The time frame for the comparison of ment in the mortality rate may be achievable by further
the TCDB center results described above was up to the improvements in prehospital care and by a reduction
time of death or last contact. in management errors in the TCDB centers, although
An additional analysis was performed using logistic the number of management errors occurring in the
regression to test for TCDB center differences in the present cohort was quite small. Furthermore, whereas
proportion of patients with good or moderate outcome differences in mortality rates among centers could be
among the nonvegetative survivors. While many vari- explained by confounding variables, differences in good
ables were tested in the model, the covariates found to or moderate outcome could not. This suggests that
be of value in this model were: TCDB center, patient future studies of severe injury must focus much more
age, duration of coma, and follow-up period. Center closely on management alterations that affect morbid-
differences were marginally nonsignificant at the 1% ity, as a large number of deaths in this group appear
level (p = 0.011), whereas the other three factors were not to be preventable. Further improvements in the
each highly statistically significant (p < 0.006). The outcome of head injury will require new therapies that
Hosmer-Lemeshow statistic showed a good model fit can reverse or mitigate ischemic or axonal injury.
(p = 0.60). How do these results compare to those previously
reported, and has outcome improved over the last dec-
ade? The results are in keeping with and support fa-
Discussion vorable reports on outcome previously described by
The present report has summarized the outcome for Becker, et aL, 2 Bowers and Marshall, 3 and Marshall, et
746 consecutively studied patients with closed head al. ~5 The present results represent the most carefully
injuries entered in the TCDB of the NINDS who sur- collected data set in neurosurgical units in the United
vived their initial resuscitation. It is important to em- States because of the high level of internal as well as
phasize that these outcome data must be viewed in light external scrutiny. Comparison of data sets is frequently
of the fact that the TCDB hospitals have had a long difficult because of differences in definitions of the
history of dedication to head-injury care and research. types of injuries and documentation of the prehospital
Patients within each of these institutions were treated phase. Thus, in this study severity of injury is defined
using comparable methods, although there was no ab- for comparison purposes by the postresuscitation GCS
solute management protocol. Also, regional changes, score, which we believe to be a truer representation of
such as in prehospital services, occurred over time. The the actual state of the patients.
results reported here should be interpreted in that light. Eighty-nine percent of the patients in the present
The charts of all patients who were coded as brain- series were in coma immediately following injury, in-
dead following nonsurgical resuscitation were scruti- dicating that in the overwhelming majority of patients
nized to be certain that this category was not used to in this cohort the initial impact injury was quite severe.
exclude potentially salvageable patients. All patients in Perhaps the most vivid illustration that care in our
this category were dead or pronounced dead less than TCDB centers has improved over time is the fact that

TABLE 11
Outcome at last contact correlated with center*

Outcome Center 1 Center 2 Center 3 Center 4 Total Cases

At Last Contact No. Percent No. Percent No. Percent No. Percent No. Percent
good 22 16.0 89 34.0 36 19.0 51 31.0 198 27
moderatedisability 25 18.5 48 19.0 32 17.0 17 10.0 122 16
severedisability 25 18.5 30 12.0 41 22.0 20 12.0 ! 16 16
vegetative 5 4.0 14 5.0 7 4.0 13 8.0 39 5
dead 58 43.0 78 30.0 70 38.0 65 39.0 271 36
totals 135 100 259 100 186 100 166 100 746 100
* Outcomeassessedaccordingto the GlasgowOutcomeScale?Significanceof difference(Kruskal-Wallis):p = 0.0011.

S34 J. Neurosurg. / Volume 75/November, 1991

Outcome of severe closed head injury

the number of deaths in patients who talked prior to nonsurgical resuscitation. Thus, the initial GCS score
deteriorating fell from over 50% to 26%. This indicates may give an erroneous view of the likelihood of salvage
more rapid detection of intracranial hematomas and in some patients. The utilization of a uniform time for
better treatment of brain swelling.16 assessment following nonsurgical resuscitation is rec-
In spite of the aforementioned limitations in making ommended if comparisons between institutions are to
comparisons with data collected by others, some such be made.
comparisons seem appropriate. If one compares the The GCS score following resuscitation was closely
present report with the outcome of the treatment of associated with the outcome at last patient contact.
intracranial hematomas in New South Wales, Australia, Hence, patients who had a GCS score of 3 following
in a series of patients with remarkably similar ages and postsurgical resuscitation fared very badly (78 % of them
apparently less severe degrees of injury following ad- died); whereas only 18% of patients with GCS scores
mission, a greater than 50% reduction in mortality is of 6, 7, or 8 following resuscitation succumbed.
noted in this group. ~8If one compares the outcome of The fate of patients with abnormal pupils illustrates
patients with GCS scores of 6, 7, and 8 with that the role of deterioration in adversely affecting patient
reported by the Multinational Data Bank, 9 the mortality outcomes. Patients with one abnormal pupil prior to
rate is almost halved in this series. Patients with less and following resuscitation (< 10% had a third nerve
severe injuries, who with optimum treatment poten- palsy from direct trauma to the nerve) fared better
tially have a better outcome, fared better in the present (34.3% dead or vegetative) than if one pupil became
cohort. abnormal for at least one observation following resus-
A sizeable reduction in mortality rate is always sus- citation (50% dead or vegetative). Early intervention
pect, as the fate of many patients with severe closed for those preventable factors that cause deterioration
head injury is determined at the time of impact. Fur- resulting in brain-stem compression should be taken
thermore, a publication from the Glasgow group sooner rather than later.
indicating a rather remarkable reduction in patients As expected, the mechanism of injury played some
suffering intracranial hematoma associated with the role in determining outcome. Pedestrians fared partic-
introduction of the CT scan 19suggests that further im- ularly poorly, which is not surprising as the frequency
provements in outcome, if a new multinational study of shock and hypoxia as a consequence of multiple
were forthcoming, would be likely. injuries is more frequent in this population. The rela-
The utilization of a new series of diagnostic catego- tively good outcome for those injured in motorcycle
ries to describe types of brain injury appears to be a accidents (mortality rate 23%), although somewhat
substantial advance. The traditional description of dif- surprising, is a reflection of the relative youth of this
fuse versus focal injury is seen to be flawed in terms of cohort and somewhat higher postresuscitation motor
prognostic discrimination. The use of more accurate and GCS scores.
diagnostic criteria not only allows for more homoge- As the influence of patient age on outcome is the
neous comparisons but also permits the targeting of subject of a separate report, 23it will not be described in
treatment for groups known to be at risk based on the detail here. Suffice it to say that age, in a series of well-
clinical examination and the initial CT scan. For ex- treated patients, becomes an extremely important var-
ample, the overall mortality rate was approximately iable in describing differences in outcome. This is well
13% for patients with Diffuse Injury II while it was 34% illustrated in Table 10 where the deleterious influence
for patients with Diffuse Injury III (swelling), indicating of age, particularly in those patients with intracranial
that the latter group would be a more appropriate one hematomas, is illustrated.
for alternative therapies. This is in keeping with the Intercenter differences in the present study were rel-
emphasis placed on the status of the basal cisterns de- atively small, supporting the concept that data from
scribed by Toutant, el al., 2~and van Dongen, et al? 2 these institutions can be combined for analysis. The
The distribution of outcomes in patients with diffuse mortality differences by center almost disappear when
swelling was remarkably similar to that in patients with known risk factors are taken into account. On the other
evacuated mass lesions. Given the fact that much of hand, the predictors of good or moderate outcome are
the brain damage attributable to mass lesions is a prod- not well established. Other factors that were found by
uct of intracranial hypertension, it is logical to begin to inspection to be associated with good or moderate
think of these groups as similar, not only because the outcomes were considered in the model; however, miss-
outcomes were so similar here but also because intra- ing data caused a drastic reduction in percent usable
cranial hypertension is likely to be the major cause of sample size (more than 250 cases eliminated), making
morbidity and death in both. the results based on cases without missing data unreli-
Comparison between the GCS score at first evalua- able. Furthermore, the designation of death is certain,
tion and the postresuscitation GCS score illustrates the whereas good or moderate outcome is subjective.
dangers of relying on the initial GCS score when pre- Clearly, a more detailed measurement of outcome,
dicting outcome. In the present report, 42% of the which is under study by the TCDB investigators and
patients at Center 3 were noted to have a GCS score of will contain more objective and normative elements, is
3 at first evaluation, which fell to 21.5% following greatly needed.

J. Neurosurg. / Volume 75/November, 1991 S3fi

 L. F. Marshall, et al.

Other reports from the TCDB focus more deeply on 12. Klauber MR, Marshall LF, Luerssen TG, et al: Determi-
specific factors that affect outcome in the head-injured nants of head injury mortality: importance of the low
patient. While some factors such as age cannot be risk patient. Nenrosurgery 24:31-36, 1989
influenced by the kind of care delivered, others such as 13. Marshall LF, Becker DP, Bowers SA, et al: The National
Traumatic Coma Data Bank. Part 1: Design, purpose,
hypotension and to a lesser extent hypoxia play a major goals, and results. J Neurosurg 59:276-284, 1983
role in determining the outcome of patients with severe 14. Marshall LF, Marshall SB, Klauber MR, et al: A new
head injury. Alterations in preventive measures such as classification of head injury based on computerized to-
seat-belt requirements and airbags in automobiles are mogmphy. J Neurosurg 75 (Suppl):S 14-$20, 1991
likely to have a favorable impact in this area. Certainly, 15. Marshall LF, Smith RW, Shapiro HM: The outcome with
an improvement in prehospital care, which is now oc- aggressive treatment in severe head injuries. Part I: The
significance of intracranial pressure monitoring. J Neu-
curring in many regions in the United States, is also rosurg 50:20-25, 1979
likely to have a favorable effect. However, the high 16. Marshall LF, Toole BM, Bowers SA: The National Trau-
mortality and morbidity rates associated with severe matic Coma Data Bank. Part 2: Patients who talk and
head injury, shown here in a series of patients cared for deteriorate: implications for treatment. J Neurosurg 59:
by experienced clinicians working in institutions with 285-288, 1983
substantial resources, indicate that there is room for 17. Seelig JM, Marshall LF, Toutant SM, et al: Traumatic
acute epidural hematoma: unrecognized high lethality in
considerable improvement in the care of such patients.
comatose patients. Neurosurgery 15:617-620, 1984
18. Sclecki BR, Ring IT, Simpson DA, et al: Injuries to the
References head, spine and peripheral nerves. Report on a study.
Initiated and sponsored by The Neurosurgical Society of
1. Abbreviated Injury Scale. Arlington Heights, 111:Ameri- Australasia, conducted by its Trauma Subcommittee, in
can Society of Automotive Medicine, 1985 collaboration with the Division of Health Services Re-
2. Beeker DP, Miller JD, Ward JD, et al: The outcome from search, Health Commission of New South Wales, in:
severe head injury with early diagnosis and intensive Epidemiology of Neurotrauma in New South Wales,
treatment, d Neurosurg 47:491-502, 1977 South Australia and the Australian Capital Territory.
3. Bowers SA, Marshall LF: Outcome in 200 consecutive New South Wales: Australian Neurological Foundation,
cases of severe head injury treated in San Diego County: 1977
a prospective analysis. Neurosurgery 6:237-242, 1980 19. Teasdale G, Galbraith S, Murray L, et al: Management
4. Cox DR: The Analysis of Binary Data. London: Methuen of traumatic intracranial haematoma. Br Med J 285:
& Co, 1970 1695-1697, 1982
5. Cox DR: Regression models and life-tables. J R Stat Soc 20. Teasdale G, Jennett B: Assessment of coma and impaired
B 34:187-220, 1972 consciousness. A practical scale. Lancet 2:81-84, 1974
6. Dixon WJ (ed): BMDP Statistical Software. Berkeley: 21. Toutant SM, Klauber MR, Marshall LF, et al: Absent or
University of California Press, 1985 compressed basal cisterns on first CT scan: ominous
7. Foulkes MA, Eisenberg HM, Jane JA, et al: The Trau- predictors of outcome in severe head injury. J Neurosurg
matic Coma Data Bank: design, methods, and baseline 61:691-694, 1984
characteristics. J Neurosurg 75 (Suppl):S8-Sl3, 1991 22. van Dongen KJ, Braakman R, Gelpke GJ: The prognostic
8. Hosmer DW, Lemeshow S: Goodness of fit tests for the value of computerized tomography in comatose head-
multiple logistic regression model. Cnmmun Star 10(A): injured patients. J Neurosurg 59:951-957, 1983
1043-1069, 1980 23. Vollmer DG, Torner JC, Jane JA, et al: Age and outcome
9. Jennett B, Bond M: Assessment of outcome after severe following traumatic coma: why do older patients fare
brain damage. A practical scale. Lancet 1:480-484, 1975 worse? J Neurosurg 75 (Suppl):S37-S49, 1991
10. Jennett B, Teasdale G: Management of Head Injuries.
Philadelphia: FA Davis, 1981 Address reprint requests to." Lawrence F. Marshall, M.D.,
11. Jennett B, Teasdale G, Galbraith S, el al: Severe head Division of Neurosurgery, University of California Medical
injuries in three countries. J Neurol Neurosurg Psychiatry Center, 225 Dickinson Street H-501, San Diego, California
40:291-298, 1977 92103-1990

$36 J. Neurosurg. / Volume 75/November, 1991