Injury, Int. J. Care Injured (2008) 39, 598—603

www.elsevier.com/locate/injury

Early prediction of outcome in very severeclosed head injury

Sandeep Jain *, Satish B. Dharap, Madhuri A. Gore

L.T.M. Medical College, Sion, Mumbai 400022, India

Accepted 2 June 2007

KEYWORDSPrognosis;Outcome;Prediction;Severe head injury;GCS

Summary

Background: People with severe head injury and admission Glasgow Coma Scale (GCS)score �5 have a poor outcome and greatly strain limited resources.Aim: To identify patients with the best chances of survival, using routine clinicalmeasures.Methods: People attending the trauma intensive care unit, who had isolated blunthead injury and GCS � 5 and who had survived �4 h, were included in the study,resuscitated and clinically assessed. The GCS score was followed serially afteradmission. Bivariate analysis of various parameters with outcome was performedusing the chi-square test. Serial GCS scores were compared with admission GCS bypaired t-testing.Results: Of the 102 patients who were studied prospectively, 78 (76.5%) died and 24(23.5%) survived. Age, gender, pre-hospital delay and admission GCS scores werecomparable between the two groups. Adequate spontaneous respiration, brisk pupil-lary light reactivity on admission and increase in GCS by at least 2 at 24 h afteradmission significantly affected the outcome ( p < 0.05). In the presence of all thesefactors, the survival rate increased from 6.1% to 57.1% ( p < 0.001).Conclusions: People with GCS score �5 still have a reasonable chance of survival, soall patients should be aggressively managed initially. Better survival was observedamong those with adequate spontaneous respiration, good pupillary reaction andimprovement in GCS of at least 2 at 24 h. These clinical parameters can help to predictsurvival and thus make best use of limited resources.# 2007 Elsevier Ltd. All rights reserved.

* Corresponding author. Tel.: +91 22 24042095;fax: +91 22 24042095.

E-mail address: sjain7172@yahoo.com (S. Jain).

0020–1383/$ — see front matter # 2007 Elsevier Ltd. All rights resedoi:10.1016/j.injury.2007.06.003

Introduction

People with head injury scoring 5 or less on theGlasgow Coma Scale (GCS) on admission have apoor prognosis. The high incidences of death and

rved.

Early prediction of outcome in severe head injury 599

disability among such cases have been well docu-mented.3,7,20 These individuals need intensivetreatment in the Intensive Care Unit (ICU), and theirmanagement puts enormous strain on ICU resourceswhich are particularly scarce in developing coun-tries such as India. The authors’ institution, which isa major tertiary referral centre, deals with onaverage 2500 major trauma cases every year.Approximately 95% of these cases involve blunttrauma and 50% head trauma; 25% of the latterare severe, with 75% mortality, and occupy over75% of trauma ICU beds at any given time. Traumasurgeons are faced with the challenge of allocatinglimited resources to people with the best chance ofsurvival.

A number of studies have identified predictors ofoutcome after severe head injury. These predictorsinclude clinical parameters such as age, gender, GCSscore, mechanism of injury, pupillary reactivity,Abbreviated Injury Scale (AIS) score and pre-hospi-tal airway management.3,8,11,13,14,21,22 Variouslaboratory parameters such as hyperglycaemia, leu-kocytosis, serum S-100N, neuron-specific enolase,serum S-100b protein and auditory brain stemresponses have also been correlated with out-come.4,10,18,26 The GCS is the most widely usedpredictive variable because of its facility, repeat-ability and familiarity. However, the relationship ofpre-resuscitative GCS with mortality is non-linear,which casts doubt over its use as a continuousmeasure incorporated into outcome predictionmodels.23 Most severe head injury studies haveinvolved people with GCS score of 10 or less, andhave been retrospective in design.3,15,24,25

Some trauma centres do not admit cases of headtrauma with GCS less than 5 into the ICU, in order tomake best use of limited resources.1 It has been theauthors’ observation that, although the overall mor-tality among such individuals is high, some of themdo survive with an acceptable Glasgow OutcomeScale GOS score. The policy of not admittingpatients with GCS score of less than 5 means denyingthem a chance of survival that is real.

A limited number of studies have followedpatients with GCS scores of 5 or less.3,9,16 Thepresent study was planned to examine the outcome,and the factors affecting it, in this subset of patientsin order to enable more reliable prognosis early inthe course of treatment; thus potentially salvage-able patients may be identified.

Materials and methods

This was a prospective observational study carriedout in a tertiary level university hospital from

September 2004 to July 2005. It included all headinjury cases with admission GCS scores of 5 or less,but excluded those with multiple associated injuriesand Injury Severity Scale (ISS) scores of 30 or moreand also those who died within 4 h of admission.

All patients underwent intubation on admissionand further evaluation by arterial blood gas analysisfor adequacy of respiration and need for ventilatorysupport. Ventilation was defined as inadequatewhen respiratory rate was <10 or >29 breaths permin, or partial pressure of carbon dioxide ( pCO2)was >45 mmHg, or oxygen saturation was <90%.Patients with adequate spontaneous respiration asjudged by these parameters received supplementaloxygen after intubation; those judged to haveinadequate respiration received mechanical venti-latory support. Indwelling Foley’s catheters wereinserted in all cases, and so as nasogastric tubes(with appropriate precautions in all those with sus-pected skull base fracture). Cranial CTwas obtainedafter achieving adequate oxygenation. CT findingswere categorised either as diffuse lesions (includingmultiple intracerebral haemorrhagic contusions,massive cerebral oedema, subarachnoid haemor-rhage, intraventricular haemorrhage, diffuse axonalinjury and cerebral infarct) or mass lesions (includ-ing extradural haematoma, subdural haematomaand intracerebral haematoma with or without mid-line shift). All patients underwent evaluation byneurosurgeons in order to decide between conser-vative management and surgical intervention.

All patients were observed in the trauma ICU untilfit to be moved to the ward, i.e. haemodynamicallystable and breathing spontaneously after extuba-tion or tracheostomy. Enteral nutritional supportwas initiated in all cases by day 3. All patientsreceived supportive care in the form of chest andlimb physiotherapy, and specific care to eyes andback. Intracranial pressure monitoring was not car-ried out. All patients received prophylactic antibio-tics (in view of the risk of chest infection associatedwith intubation), histamine (H2) receptor blockersfor stress ulcer prevention and dilantin sodium for 1week unless injury precipitated a seizure. For 3 daysintermittent bolus mannitol therapy was adminis-tered to all individuals with evidence of cerebraloedema. If the GCS score was 8 or less by day 7,tracheostomy was performed to facilitate tracheo-bronchial toilet. Patients moved from the ICU onceoxygenation and respiration without supplementaryoxygen or ventilatory support were considered ade-quate, and were offered nursing care. They weredischarged after achieving tracheostomy extuba-tion.

In all cases the GCS score was measured at 4, 12,24 and 48 h, at 3, 4, 5, 7, 10, 15 and 20 days and then

600 S. Jain et al.

every 10 days until discharge or death. Data regard-ing age, gender, mode of injury, pre-hospital delay,pupillary light reaction, arterial blood gases, vitalparameters, CT findings and ICU and ward stayswere recorded. Every person who survived under-went evaluation by the GOS at discharge.

Bivariate analysis of these factors with outcomewas carried out using the chi-square test. The GCSscores at various time periods were paired withadmission GCS scores and analysis was performedby paired t-testing.

Results

A total of 102 individuals were included in the study;90 were male and 12 were female. Their mean agewas 31.74 (�12.7, range 6—75) years. Blunt headinjury was due to road traffic accident in 45 (44.1%)cases; to railway traffic accident in 37 (36.27%); tofall from height in 14 (13.7%); and to blunt assault in6 (5.8%). Table 1 depicts the impact of variousfactors on outcome.

In all, at admission 63 people were judged to haveadequate respiration and thus were managed byintubation and oxygenation, whereas 39were judgedto have inadequate respiration and received venti-latory support; there were 19 (30.2%) and 5 (12.8%)survivors, respectively. This difference was statisti-cally significant (p = 0.045). Adjusted odds ratio was2.936 with 95% confidence interval of 0.995—8.664.

The pupillary reaction to light on admission wascategorised as either poor (fixed or sluggish pupils)or brisk. Of 83 individuals with poor reactivity, 14(16.9%) survived; whereas of 19 with good reactiv-ity, 10 (52.6%) survived. This was statistically sig-nificant (p = 0.001). On analysis, the risk estimate

Table 1 Study of factors known to affect outcome in seve

Factors Non-survivors

Age (years) < 50 66 (75.9%)Age (years) > 50 12 (80.0%)Male 68 (75.6%)Female 10 (83.3%)pCO2 (mmHg) < 45 54 (74.0%)pCO2 (mmHg) > 45 24 (82.8%)O2 saturation (%) < 90 17 (89.5%)O2 saturation (%) > 90 61 (73.5%)SBP (mmHg) < 90 04 (80.0%)SBP (mmHg) > 90 74 (76.3%)CT: diffuse injury 42 (77.8%)CT: mass lesion 36 (75.0%)Surgery 26 (68.4%)Conservative treatment 52 (81.3%)

SBP, Systolic Blood Pressure.

odds ratio was 5.476 with 95% confidence interval of1.882—15.936.

CT revealed mass lesions in 48 (47%) and diffuselesions in 54 (53%) of cases. There was no differencein survival. After neurosurgical consultation, of thetotal patient population 38 (37.3%) underwent sur-gery and 64 (62.7%) received non-operative treat-ment. Among those who underwent surgery, 26(68.4%) died and 12 (31.6%) survived. In the non-operative group, 52 (81.3%) died and 12 (18.8%)survived. Those requiring surgical evacuation hada better survival, but this did not reach statisticalsignificance ( p = 0.14).

The study population was divided into survivorsand non-survivors, and the groups were compared asin Table 2. Of the 24 survivors at discharge, 14(58.3%) had poor functional outcome (GOS 2 or 3)whereas 10 (41.7%) had good functional outcome(GOS 4 or 5). In the survivor group, pre-hospitaldelay was 3.55 � 5.09 h whereas in the non-survivorgroup it was 4.11 � 3.89 h; this was not statisticallysignificant ( p = 0.57). There was no significant dif-ference in mean pulse rate, systolic blood pressureor respiratory rate at admission between the survi-vors and non-survivors; hypotension (systolic bloodpressure �90 mmHg) was not found to affect theoutcome significantly ( p = 0.849) Mean arterialpCO2 was 40.8 � 10.1 mmHg in the non-survivorgroup and 39.6 � 8.04 mmHg in the survivor group(p = 0.6); 29 persons had hypercarbia (arterialpCO2 �45 mmHg) and 73 had normocarbia ( pCO2

�44 mmHg). Hypercarbia did not affect the out-come significantly (p = 0.34). Mean oxygen satura-tion was comparable in both groups. Hypoxia (O2

saturation � 89%) was found in 19 cases, whereas O2

saturation was �90% in 83 cases. This was statisti-cally insignificant (p = 0.139).

re closed head injury

Survivors Total p-Value

21 (24.1%) 87 0.7273 (20.0%) 15

22 (24.4%) 90 0.55102 (16.7%) 1219 (26.0%) 73 0.34505 (17.2%) 2902 (10.5%) 19 0.13922 (26.5%) 8301 (20.0%) 05 0.84923 (23.7%) 9712 (22.2%) 54 0.74112 (25.0%) 4812 (31.6%) 38 0.14012 (18.8%) 64

Early prediction of outcome in severe head injury 601

Table 2 Factors affecting survival after severe closed head injury

Factors Non-survivors Survivors p-Value

Mean years of age (S.D.) 32.13 (�13.12) 30.46 (�11.74) 0.578Hours pre-hospital delay (S.D.) 4.11 (�3.89) 3.55 (�5.09) 0.570mmHg SBP (S.D.) 112 (�18.75) 114.5 (�18.15) 0.566mmHg pCO2 (S.D.) 40.82 (�10.14) 39.66 (�8.04) 0.608% oxygen saturation (S.D.) 92.29 (�10.9) 96.03 (�9.14) 0.131Glasgow Coma Scale (S.D.) 4.10 (�0.71) 4.21 (�0.72) 0.528

SBP, Systolic Blood Pressure.

Mean ICU stay in the non-survivor group was4.14 � 5 days and in the survivor group was4.37 � 2.88 days (p = 0.71). Mean hospital staywas significantly (p < 0.001) longer among survivors(24.3 � 17.6 days) than among non-survivors(5.64 � 5.2 days).

There was no significant difference in admissionGCS score between the two groups. The GCS wasevaluated serially at different time periods afteradmission and these scores were compared with theGCS score at admission using paired t-testing(Tables 3a and 3b). In the non-survivor group, therewas no significant difference in GCS score until 5days after admission. In the survivor group, the GCSscore increased by 1.25 at 12 h after admission andby 2.13 at 24 h after admission. These differenceswere statistically significant (p < 0.001). Of thetotal 102 patients, 78 showed an increase in GCSscore by less than 2 points at 24 h after admission. Ofthese 78 patients, 67 (85.9%) died and 11 (14.1%)survived. Of the 24 patients who showed an increasein GCS score of 2 or more, 11 (45.8%) died and 13(54.2%) survived. Chi-square testing demonstratedthat the differences in the outcome were statisti-cally significant ( p < 0.001). The odds ratio was7.198 with a 95% confidence interval of 2.582—20.069.

We analysed the three significant factors affect-ing outcome, namely increase in GCS by 2, presenceof good pupillary reactivity and presence of spon-taneous ventilation (Tables 4 and 5). We found that

Table 3a Change in Glasgow Coma Scale score of non-surv

Hours since admission Mean GCS score (S.D.)

0 4.10 (�0.71)4 4.19 (�1.36)

12 4.36 (�2.11)24 4.38 (�2.13)48 4.49 (�2.28)72 4.52 (�2.54)96 4.68 (�2.61)

120 4.86 (�2.32)GCS, Glasgow Coma Scale.

when none of these factors was present, survivalrate was 6.1%; when only one of the factors waspresent, survival increased to 15.4%; in the presenceof two of these factors, survival rate was 52.2%; andwhen all three of the factors were present, 57.1% ofthe patients survived.

Discussion

Very severe head injuries (GCS � 5) have a highmortality and poor functional outcome. Predictionof outcome may assist resource allocation, particu-larly in resource-poor societies. There are very fewstudies of this group of cases. Demetriades et al. in2004 studied the outcome and prognostic factorsamong people with head injury and a GCS score of 3.They found an overall mortality of 76%.3 Our studyhad comparable results. Age has been reported toaffect the outcome independently,13,22 but we couldnot confirm this because there was only one patientolder than 60 years in our series. Gender5 and CTfindings6,14 have also been shown to affect outcome.However, in our study these seemed to have noeffect on the outcome of such severe head injuries.Hypoxia and hypotension2,19 have been found toaffect outcome adversely, but our results did notsupport this finding, perhaps because of the exclu-sion from our population of cases with significantassociated injuries (ISS 30 or more) and the rela-tively small sample size.

ivors (n = 78) over time

Improvement in GCSsince admission

p-Value

Baseline0.09 0.5350.26 0.2750.28 0.2410.39 0.1750.42 0.2090.58 0.1440.76 0.050

602 S. Jain et al.

Table 3b Change in Glasgow Coma Scale score of survivors (n = 24) over time

Hours since admission Mean GCS score (S.D.) Improvement in GCSsince admission

p-Value

0 4.21 (�0.72) Baseline4 4.79 (�1.21) 0.58 0.027

12 5.46 (�1.64) 1.25 0.00224 6.33 (�1.92) 2.13 0.00048 7.58 (�3.09) 3.38 0.00072 7.92 (�3.34) 3.71 0.00096 7.83 (�3.52) 3.63 0.000

120 8.17 (�3.79) 3.96 0.000

GCS, Glasgow Coma Scale.

Table 4 Factors significantly affecting outcome in severe closed head injury

Factor Number ofnon-survivors

Number ofsurvivors

Total p-Value Odds ratio

<2 points’ increase in GCS score at 24 h 67 (85.9%) 11 (14.1%) 78 0.000 7.198>2 points’ increase in GCS score at 24 h 11 (45.8%) 13 (54.2%) 24Poor pupillary reaction 69 (83.1%) 14 (16.9%) 83 0.001 5.476Brisk pupillary reaction 09 (47.4%) 10 (52.6%) 19Inadequate spontaneous ventilation 34 (87.2%) 05 (12.8%) 39 0.045 2.936Adequate spontaneous ventilation 44 (69.8%) 19 (30.2%) 63

GCS, Glasgow Coma Scale.

The presence of spontaneous ventilation has beenknown to be significantly associated with better out-come in severe head injury cases.24 This result wasrepeated in our study. Fixed dilated pupils associatedwith low GCS score have been found to be associatedwith absolutemortality11; we also observed very highmortality (83.1%) among individuals with poor pupil-lary reactivity. GCS score at 24 h has also beenreported to be correlated with outcome,12,17 aswas demonstrated in our present series.

The identification of factors which reliably pre-dict outcome can help in optimising the use oflimited resources. People with head trauma whopresent with briskly reacting pupils and adequaterespiration have higher chances of survival andhence should be given priority in intensive care

Table 5 Model to predict survival after severe closed head

Number of significanta

factorsb presentNumber of patients wdid not survive

None 31 (93.9%)One 33 (84.6%)Two 11 (47.8%)Three 03 (42.9%)

Total 78 (76.5%)a p-Value < 0.001.b Factors comprise: increase in Glasgow Coma Scale score by 2 at

briskly reacting pupils.

resource allocation. Those whose GCS scoreincreases by at least 2 by 24 h after admission alsohave significantly higher chances of survival andmaybe selected for continued resource allocation. In theabsence of any of these factors, survival rate was6.1%; it increased to 15.4%, 52.2% and 57.1% withpresence of one, two or all three factors, respec-tively. Thus combinations of these factors can indi-cate which patients would best benefit from limitedresources. Those who need ventilatory support onadmission should be reassessed at 24 h and, if theyshow brisk pupillary reaction and improvement inGCS by 2 points, care should continue; otherwiserelatives should be counselled regarding the prog-nosis. This model provides a scientific basis for thevery difficult ethical decision as to whether to wean

injury

ho Number of patientswho survived

Total

02 (06.1%) 3306 (15.4%) 3912 (52.2%) 2304 (57.1%) 07

24 (23.5%) 102

24 h after admission; presence of spontaneous ventilation; and

Early prediction of outcome in severe head injury 603

such a patient off a ventilator in favour of a moresalvageable person.

Conclusion

Thus it can be concluded that early prediction ofsurvival in very severe head injury cases is possible assoon as 12 h after admission. If patients show a riseover admission GCS score of at least 2 points by 24 h,they have a sevenfold greater chance of survival thanthose who show no such increase. The presence ofspontaneous ventilation multiplies the chances ofsurvival almost threefold. People with severe headtrauma who present with briskly reacting pupils havefive times higher chances of survival than those withsluggish pupils. These three predictive factors canindicate which patients should have priority in theprovision of limited resources; the model may helpthe physician resolve major ethical dilemmas.

Conflict of interest

None.

References

1. Boffard KB, Degiannis E, Goosen J, et al. JohannesburgHospital Trauma Unit Protocols. Available at http://www.wits.ac.za/trauma/Information/Unit%20Protocols/UnitProtocols.htm. p. 5. Last accessed on 1 July, 2007. Nowsame available at http://web.wits.ac.za/NR/rdonlyres/2BD954E1-3BA8-4E42-95CC-06585EBC9D9B/0/ResusProtocols2004.doc.

2. Combes P, Fauvage B, Colonna M, et al. Severe head injuries:an outcome prediction and survival analysis. Intens Care Med1996;22:1391—5.

3. Demetriades D, Kuncir E, Velmahos GC, et al. Outcome andprognostic factors in head injuries with an admission GlasgowComa Scale of 3. Arch Surg 2004;139:1066—8.

4. Dimopoulou I, Korfias S, Dafni U, et al. Protein S-100b serumlevels in trauma-induced brain death. Neurology 2003;60:947—51.

5. Farin A, Deutsch R, Biegon A, et al. Sex-related differences inpatients with severe head injury: greater susceptibility tobrain swelling in female patients 50 years of age and younger.J Neurosurg 2003;98:32—6.

6. Fearnside MR, Cook RJ, McDougall P. The Westmead HeadInjury Project outcome in severe head injury. A comparativeanalysis of pre-hospital, clinical and CT variables. Br J Neu-rosurg 1993;7:267—79.

7. Gan BK, Lim JH, Ng IH. Outcome of moderate and severetraumatic brain injury amongst the elderly in Singapore. AnnAcad Med Singapore 2004;33:63—7.

8. Gennrelli TA, Champion HR, Socco WJ. Mortality of patientswith head injury and extracranial injury treated in traumacenters. J Trauma 1989;29:1195—202.

9. Kotwica Z, Jakubowski JK. Head-injured adult patients withGCS of 3 on admission–—who have a chance to survive? ActaNeurochir (Wien) 1995;133:56—9.

10. Li N, Shen JK, Zhao WG, et al. S-100B and neuron specificenolase in outcome prediction of severe head injury. Chin JTraumatol 2004;7:156—8.

11. Lieberman JD, Pasquale MD, Garcia R, et al. Use of admissionGlasgow Coma Score, pupil size, and pupil reactivityto determine outcome for trauma patients. J Trauma2003;55:437—42.

12. Mamelak AN, Pitts LH, Damron S. Predicting survival fromhead trauma 24 hours after injury: a practical method withtherapeutic implications. J Trauma 1996;41:91—9.

13. Mosenthal C, Lavery RF, Addis M, et al. Isolated traumaticbrain injury: age is an independent predictor of mortality andearly outcome. J Trauma 2002;52:907—11.

14. Ono J, Yamaura A, Kubota M, et al. Outcome prediction insevere head injury: analyses of clinical prognostic factors. JClin Neurosci 2001;8:120—3.

15. Pillai SV, Kolluri VR, Praharaj SS. Outcome prediction modelfor severe diffuse brain injuries: development and evalua-tion. Neurol India 2003;51:345—9.

16. Quigley MR, Vidovich D, Cantella D, et al. Defining the limitsof survivorship after very severe head injury. J Trauma1997;42:7—10.

17. Ratanalert S, Chompikul J, Hirunpat S, et al. Prognosis ofsevere head injury: an experience in Thailand. Br J Neurosurg2002;16:487—93.

18. Rovlias A, Kotsou S. Classification and regression tree forprediction of outcome after severe head injury using simpleclinical and laboratory variables. J Neurotrauma 2004;21:886—93.

19. Selladurai BM, Jayakumar R, Tan YY, et al. Outcomeprediction in early management of severe head injury:an experience in Malaysia. Br J Neurosurg 1992;6:549—57.

20. Shigemori M, Tokutomi T, Yomamoto F, et al. The outcome ofsevere head injuries in children. Neurosurg Rev 1989;12:1436—40.

21. Signorini DF, Andrews PJ, Jones PA, et al. Predicting survivalusing simple clinical variables: a case study in traumatic braininjury. J Neurol Neurosurg Psychiatry 1999;66:20—5.

22. Susman M, Dirusso SM, Sullivan T, et al. Traumatic brain injuryin the elderly: increased mortality and worse functionaloutcome at discharge despite lower injury severity. J Trauma2002;53:219—24.

23. Udekwu P, Schiro SK, Vaslef S, et al. Glasgow Coma ScaleScore, mortality and functional outcome in head injuredpatients. J Trauma 2004;56:1084—9.

24. Waxman K, Sundine MJ, Young RF. Is early prediction ofoutcome in severe head injury possible? Arch Surg1991;126:1237—42.

25. White JR, Farukhi Z, Bull C, et al. Predictors of outcome inseverely head-injured children. Crit Care Med 2001;29:534—40.

26. Wilson WJ, Penn C, Saffer D, et al. Improving the predictionof outcome in severe acute closed head injury by usingdiscriminant function analysis of normal auditory brainstemresponse latencies and amplitudes. J Neurosurg 2002;97:1062—9.