Development and validation of a new simpli fi ed anatomic trauma mortality score
Jih Huei Tan
a,b,*, Yuzaidi Mohamad
a, Rizal Imran Alwi
a, Chor Lip Henry Tan
a,b, Af fi rul Chairil Arif fi n
c, Razman Jarmin
baGeneralSurgeryDepartment,HospitalSultanahAminah,JohorBahru,Malaysia
bPusatPerubatanUniversitiKebangsaanMalaysia,Cheras,Malaysia
cUniversitiSainsIslamMalaysia,Ampang,Malaysia
ARTICLE INFO
Keywords:
Mortality
Traumaseverityindices Trauma
Prognosis
ABSTRACT
Background:Mosttraumamortalitypredictionscoresarecomplexinnature.GAP(GlasgowComaScale, Age,Systolicbloodpressure)andmGAP(mechanism,GlasgowComaScale,Age,Systolicbloodpressure) scoresarerelativelysimplescoringtools.However,thesescoreswerenotvalidatedinlowandmiddle incomecountriesincludingMalaysiaanditsaccuraciesareinfluencedbythefluctuatingphysiologic parameters.Thisstudyaimstodeveloparelevantsimplifiedanatomictraumascoringsystemforthe localtraumapatientsinMalaysia.
Method:Atotalof3825traumapatientsfrom2011to2016wereextractedfromtheHospitalSultanah Aminah Trauma SurgeryRegistry. Patients weresplitintoadevelopmentsample (n=2683)and a validationsample(n=1142).Univariateanalysisisappliedtoidentifysignificantanatomicpredictors.
Thesepredictorswerefurtheranalyzedusingmultivariablelogisticregressiontodevelopthenewscore and compared to existing score systems. The quality of prediction was determined regarding discriminationusingsensitivity,specificityandreceiveroperatingcharacteristic[ROC]curve.
Results:Existingsimplifiedscoresystems(GAP&mGAP)revealedareasundertheROCcurveof0.825and 0.806.ThenewlydevelopedHeCLLiP(Head,cervicalspine,lung,liver,pelvicfracture)scorecombines onlyfiveanatomiccomponents:injuryinvolvinghead,cervicalspine,lung,liverandpelvicbone.The probabilitiesofmortalitycanbeestimatedbychartingthetotalscorepointsontoagraphchartorusing thecut-offvalueof(>2)withasensitivityof79.2andspecificityof70.6%onthevalidationdataset.The HeCLLiPscoreachievedcomparablevaluesof0.802fortheareaundertheROCcurveinvalidation samples.
Conclusion:HeCLLiPScoreisasimplifiedanatomicscoresuitedtothelocalMalaysianpopulationwitha goodpredictiveabilityfortraumamortality.
©2019ElsevierLtd.Allrightsreserved.
Background
Traumaticinjuryisthefifthleadingcauseofdeathrecordedin health carecentresofMalaysia. [1] Itis asignificantburden in regardstomanpowerandlossineconomyduetothehighcost incurredtomanagethis cohortofpatients. Traumascoresisan important tool for outcome prediction (mortality or functional outcome), to triage patients in a pre-hospital setting with continuousauditingforbettermentofthetraumaservice.
CurrentexistingtraumascoresincludeMGAP[2],GAP[3],RTS [4],NISS[5],TRISS[6],TARN[7],ASCoT[8],PRESTO[9]orTrauma Mortalitypredictionmodel[10].Thesearecomplexscoringsystem which requirestraining andcomplex mathematicalformulasto calculateinexceptiontotheGAPandMGAPscores.GAPandMGAP scores developed in 2011 and 2010, is based on physiological Abbreviations:GCS,GlasgowComaScale;NISS,NewInjurySeverityScore;RTS,
revisedtraumascore;TRISS,traumaandinjuryseverityscore;HR,heartrate;RR, respiratoryrate;SBP,systolicbloodpressure;AIS,abbreviatedinjuryscale;MGAP, mechanismofinjury,GlasgowComaScale,Age,SystolicBloodPressure; GAP, GlasgowComaScale,Age,SystolicBloodPressure;TARN,traumaauditandresearch network;ASCoT,aseveritycharacterisationoftraumaASCoT(among)others;NTrD- TRISS,MalaysianNationalTraumaregistrydatabasederivedTRISS;HeCLLiP,head, cervicalspine,lung,liver,pelvicfracture;FAST,focussedassessmentofsonography intrauma.
* Correspondingauthorat:MRCS,Ireland.
E-mailaddresses:[email protected](J.H.Tan),[email protected] (Y.Mohamad),[email protected](R.ImranAlwi),[email protected] (C.L.HenryTan),affi[email protected](A.ChairilAriffin),[email protected] (R.Jarmin).
https://doi.org/10.1016/j.injury.2019.01.027 0020-1383/©2019ElsevierLtd.Allrightsreserved.
ContentslistsavailableatScienceDirect
Injury
j o u r n a lh o m e p ag e : w w w . e l s e vi e r . c o m / l o c a t e/ i n j u r y
parameters which are simple and easy to use. However,these modelsweredevelopedandvalidatedinhighincomecountries.
Traumaepidemiology and mortality trendsis different in low- middleincomecountriesbasedonCRASH-2trialdatabase[7].Lack of trauma registries in these countries limits the validation of currentavailableprognosticmodels[11].
Therefore,thesuitabilityoftheabove-mentionedtraumascoring systemmaynotbeapplicabletopatientsinthelow-middleincome countries. Furthermore, these scores utilized the physiologic parameters which may be inaccurate if data is obtained after resuscitationorrecordedtooearlybeforephysiologiccompromise.
Thenewscoringsystemmayimprovetheefficacyoftriaging traumapatients,assistclinicianwithsurgicalresourceallocation, local clinical auditing for quality of care improvement, loco- regional centre comparison, improved communication among trauma care providers and assist in counselling to family on prognosistargetingtotheorganinjuriessustained.
Themainobjectiveofthisstudyistoformulateanewsimplified anatomic prediction score for trauma mortality suitable for a Malaysian trauma population. Secondary objectives are to comparetheformulatedscorewithotheravailabletraumascores andidentifythebetterscoringsystemtopredicttraumadeaths.
Methodology
HospitalSultanahAminahisatertiarygovernmentrun-hospital withspecializedtraumasurgeryservices.Thisspecializedtrauma surgeryserviceisledbya traumasurgeon(s), generalsurgeons, medicalofficerswithnursestrainedtohandletraumaticpatients.
Thiscurrenttraumadivisionundertheumbrellaofgeneralsurgical department may contribute to a better outcome in managing traumapatients. Nevertheless, theinjury pattern and patient’s characteristicsaresimilartoothergeneralsurgicalunitinMalaysia [12,13].Inadditiontoprovidingtraumasurgerycare,thetrauma coordinatorshasstartedtheelectronictraumaregistrysinceits establishmentin2011.Theelectronicprospectivetraumaregistry capturesdatafromeverytraumapatienttreatedbythetrauma surgeryteamsinceitsinceptionin2011.Thedatawasenteredbya trainedtrauma nursewhich actsas thetraumacoordinator. All discrepanciesindatawasreviewedandsupervisedbytwosenior traumasurgeon'spriordataentryintotheelectronicdatabase.
This is a retrospective analysis of the prospective Trauma SurgeryRegistrydatabase.Alltraumapatientstreatedfrom1st May2011to31stApril2016(5years)wasincluded.Thefollowing weretheinclusioncriteria:a)Allpatientsaged13yearsorolder,b) Allpatientsadmittedandtreatedbythetraumasurgeryunit,c) Admissiontootherdepartmentinthehospitalbutreferralmadeto traumasurgeryteamduetounstablehemodynamicstatus,d)Limb vascularinjuries,e)Injuriestoneckandtorso.
The exclusion criteria include: a) Injury resulting from pathologicalconditions(i.epathologicalfracturesresultingfrom malignancy)and injury resulting fromdegenerativechanges or medicalillnesses.b)Hanging,drowning,burnsandenvenomation.
c) Very late presentations or transfers or referrals from other hospitalsforconditionsnotasadirectresultoftheinitialtrauma insultwheredefinitivetreatmenthadbeenaccomplishedinthe hospitaloforigin(i.ebowelobstructionfollowinga laparotomy performed for trauma) or sequelae of complications occurring temporallydistantfromtheindexinjury.
The data of interest was retrieved from the registryanony- mously. Data extracted were demographics (age and gender), mechanismofinjury(blunt/penetrating),AISgradeofeachorgan injured,physiologicparameters(systolicbloodpressure,respira- toryrate,heartrate,temperature,Glasgowcomascore)andinjury severityscoring(revisedtraumascore,newinjuryseverityscore, andtraumaandinjuryseverityscore).
Chart1.OverallFlowChartofStatisticalAnalysisPlan.
TotaltraumacasesdocumentedinTraumaDatabasefrom2011to2016(n=3884)
TheorganinjurygradewasbasedonAIS2005whichclassifies theinjuryforeachorganfromgrade1to6.Theorgansrecorded werehead,face,neck,heart,cervicalspine,lumbarspine,chest wall,lung,thoracicvascular,diaphragm,esophagus,stomach,liver, spleen,biliary,pancreas,duodenum,smallbowel,colon,rectum, adrenal,kidney, ureter, bladder, urethra, uterus,pelvicfracture, fallopiantube,ovaries,vagina,vulva,testis,scrotum,penis,cervical vascular,abdominalvascular,peripheralvascularandextremities [14]. Blunt mechanism includes events involving car, heavy vehicles,motorcycle,bicycle,fall,pedestrianorassault.Penetrat- ing injuries included were mainlyinvolving weapons of knife, agriculturaltool,industrialtools,handgunorshotgun.
Statisticalanalysisplan(refertoChart1forflowchartof analysis)
The data analysis was performed using SPSS version 16, Chicago,SPSSIncwithatotalof3825patientsincluded.Fortyone
patients were excluded due to missing data. The dataset was dividedintotwosubsets:70%(n=2683)wereusedasdevelop- mentdatasetandtheremaining30%(n=1142)wasanalyzedfor validation of new scoring tool (Chart 1). The independent variables included were age, gender, mechanism of injury, physiologic parametersand anatomicparameter (AIS 05 grade of organ injury). The outcome measure was any recorded in- hospitalmortality.
Fromdevelopmentdataset,univariateanalysiswasperformed with t-test and Chi-square test. Significant(p<0.05) anatomic variableswereincludedinmultivariableanalysis.Themultivariable analysis was done with binary logistic regression using Enter method.AvalueofP<0.05isconsideredstatisticallysignificant.
Fromthis,themodelanditscoefficientsweredetermined(Chart1).
Fromvalidationdataset,thenewlydevelopedanatomicscore and existing traumascores(GCS, GAP,mGAP,Shock Index(HR/
SBP)[15],AgeShockIndex(AgemultipliedwithShockIndex)[16]) wereanalysedwithAreaunderROCcurveforcomparison.From
Table1
Characteristicsofsamplepopulationsindevelopmentandvalidationdataset.
Characteristics Developmentdataset Validationdataset
N=2683 N=1142
Age,mean(SD) 36.0(16.1) 36.7(16.3)
GCS,mean(SD) 13.3(3.4) 13.4(3.4)
RR,mean(SD) 20.3(3.6) 20.2(3.6)
HR,mean(SD) 93.6(20.7) 93.2(21.2)
Temp,mean(SD) 37.0(0.2) 37.0(0.2)
SBP,mean(SD) 125.6(25.2) 126.5(26.1)
ShockIndex,mean(SD) 0.8(0.3) 0.8(0.3)
RTS,mean(SD) 7.4(1.0) 7.4(1.0)
TRISS,mean(SD) 0.92(0.2) 0.92(0.2)
NISS,median(IQR) 16(9to27) 16(9to27)
Gender,n(%)
Male 2389(89.0%) 999(87.5%)
Female 294(11.0%) 143(12.5%)
Ethnicity,n(%)
Malay 1216(45.3%) 521(45.6%)
Chinese 635(23.7%) 275(24.1%)
Indian 504(18.8%) 208(18.2%)
OtherLocalEthnicity 191(7.1%) 74(6.5%)
Foreigners 137(5.1%) 64(5.6%)
Mechanism,n(%)
Blunt 2462(91.8%) 1045(91.5%)
Penetrating 221(8.2%) 97(8.5%)
Outcome,n(%)
Alive 2403(89.6%) 1022(89.5%)
Death 280(10.4%) 120(10.5%)
TypeofBluntMechanism,(n%)
Car 413(16.8%) 156(14.9%)
HeavyVehicle 35(1.4%) 16(1.5%)
Motorcycle 1566(63.6%) 669(64.0%)
Bicycle 22(0.9%) 6(0.6%)
Fall 255(10.4%) 117(11.2%)
Pedestrian 95(3.9%) 44(4.2%)
Assaulted 29(1.2%) 16(1.5%)
OtherBluntInjury 47(1.9%) 21(2.0%)
Total 2462(100.0%) 1045(100.0%)
TypeofPenetratingMechanism,(n%)
Knife 153(69.2%) 60(61.9%)
ToolAgriculture 38(17.2%) 21(21.7%)
ToolIndustrial 17(7.7%) 11(11.3%)
Handgun 9(4.1%) 3(3.1%)
Shotgun 3(1.4%) 2(2.1%)
OtherPenetratingInjury 1(0.5%) 0(0.0%)
Total 221(100.0%) 97(100.0%)
GCS=GlasgowComaScale;RR=RespiratoryRate;HR=HeartRate;Temp=Temperature;SBP=Systolicbloodpressure;RTS=RevisedTraumaScore;NISS=NewInjurySeverity Score;TRISS=TraumaandInjurySeverityScore.
thecombineddataset, a graphchartis created toestimatethe probabilitiesofmortalityforeachtotalscorepoints.
Samplesizecalculationfordevelopingriskpredictionmodels
Itwasestimatedthatatleast10predictorswillbesignificantfor traumamortality,hencethetotalnumberofregressioncoefficients willbe10.BasedonpastregistryrecordsfromApril2011untilApril 2014,thereisanestimatedcountof239in-hospitalmortalities&
1969survivors.Assumed70%percentofthedatasetfordevelop- mentofriskmodelhasasimilarpatternofmortalities;eventsper
variablearecalculatedasfollows:Eventspervariable=(239in- hospitalmortalitiesdividedby10regressioncoefficients)=>23.9.
Astheeventspervariableweremorethan10,thesamplesizefrom theregistryissufficepreventmodeloverfitting[17].
Results
Atotalof3825traumapatientswereincludedintheanalysis.
Theywererandomlydividedinto2separategroupsofdevelop- ment(70%,n=2683)andvalidation(30%,n=1142).Thecharacter- isticofthe2studygroupsisasillustratedin(Table1).Thepatients
Table2
Univariateanalysisofallvariablesversusmortality[developmentdataset].
Mortality Survivor OR(95%CI) p-value
Gender,n(%)
Male 254(10.6%) 2135(89.4%) 1.23(0.80To1.87) 0.344
Female 26(8.8%) 268(91.2%)
Mechanism,n(%)
Blunt 275(11.2%) 2187(88.8%) 5.43(2.22To13.30) <0.001
Penetrating 5(2.3% 216(97.7%)
TypeOfBluntInjury,n(%)
Car 45(10.9%) 368(89.1%) 0.217
HeavyVehicle 5(14.3%) 30(85.7%)
Motorcycle 168(10.7%) 1398(89.3%)
Bicycle 2(9.1%) 20(90.9%)
Fall 27(10.6%) 228(89.4%)
Pedestrian 19(20.0%) 76(80.0%)
Assaulted 2(6.9%) 27(93.1%)
Other 7(14.9%) 40(85.1%)
TypeofPenetratingInjury,n(%)
Knife 3(2.0%) 150(98.0%) 0.585
ToolAgriculture 1(2.6%) 37(97.4%)
ToolIndustrial 0(0.0%) 17(7.9%)
Handgun 1(11.1%) 8(88.9%)
Shotgun 0(0.0%) 3(100.0%)
Other 0(0.0%) 1(100.0%)
OrganInjuries (SeeTable3)
Age(SD) 39.3(18.8) 35.7(15.7) <0.001
SBP(SD) 119.3(37.2) 126.3(23.3) <0.001
RR(SD) 21.9(5.2) 20.1(3.4) <0.001
HR(SD) 102.6(28.1) 92.5(19.4) <0.001
Temp(SD) 36.9(0.3) 37.0(0.2) <0.001
GCS(SD) 9.1(4.9) 13.8(2.8) <0.001
Table3
Organsinjuredversusmortality[developmentdataset].
SurvivalN,(%) DeathN,(%) OR 95%CI p-value
NoHeadInjury 2009(94.30) 121(5.7)
HeadInjury 394(71.2) 159(28.8) 6.70 5.17to8.69 <0.001
NoHeartInjury 2397(89.6) 277(10.4)
HeartInjury 6(66.7) 3(33.3) 4.33 1.08to17.40 0.024
NoCervicalSpineInjury 2355(90.1) 260(9.9)
CervicalSpineInjury 48(70.6) 20(29.4) 3.77 2.21to6.46 <0.001
NoLungInjury 1493(93.7) 100(6.3) 2.95 2.28to3.82 <0.001
LungInjury 910(83.5) 180(16.5)
NoLiverInjury 2207(90.6) 2390(9.4) 2.45 1.74to3.44 <0.001
LiverInjury 196(79.7) 50(20.3)
NoPelvicFracture 2327(90.1) 257(9.9) 2.74 1.69to4.45 <0.001
PelvicFracture 76(76.8) 23(23.2)
*NonSignificantOrganInjuryAreListedInSupplementaryFile.
in 2 datasets sharedsimilar demographics characteristics(age, gender or ethnicity), physiologic parameters (GCS, RR, HR, Temperature,SBP),injuryscoring(RTS,NISS,TRISS),mechanism ofinjury(typeofbluntorpenetratinginjury),mortalityrate.
Bluntmechanismandphysiologicparametersisa significant predictorfordeathinunivariateanalysisasshowninTable2.
Univariateanalysiswithchisquaretestrevealedhead,heart, cervicalspine,lung,liver,pelvicinjuryweresignificantanatomic variablesleadingtomortality(Table3).
Multivariable analysis of the development dataset showed organinjury involvinghead,cervical spine,liver,lung orpelvic bonehavesignificanthigheroddsofdeathwithp-value<0.01 (Table4).Hence,regressionanalysesoftheseanatomicvariables wereperformed(Table5).
Thepredicted probabilityformortalityfor eachscorepoints werecalculatedbasedonthecoefficients(B)fromtheregression analyses(Table5,Fig.1).
Basedontheoddratiovaluefromregressionanalysis(Table5), score points for each component of the HeCLLiP score were estimated.Thescoreisanabbreviatedanatomicscoremadeupof5 differentorganswhichincludehead,cervicalspine,liver,lungand pelvicbone.Presenceofanyoftheinjuriesshallbeassignedscore pointsasbelow(Table6).
Thescorepointofeachcomponentcanbeeasilyremembered as74,245(Table6).Themaximumallocatedscorepointsare22.A cutoff ofmore than2 pointswas obtainedwithYouden index analysis. The cut off value (>2) predicts the mortality with sensitivityof79.2andspecificityof70.6%onthevalidationdataset (Fig.2).
TheAUCvaluesofHeCLLiP,GAP,MGAP,GCS,ShockIndexand AgeShockIndexwere0.802,0.825,0.806,0.810,0.603and0.580 respectively (Fig. 3 and Table 7). The predictive ability of the HeCLLiP scoreis reflectedby theAUC.It revealeda fairlygood discriminationwhencomparedtoGAP,MGAP,GCSbutitwasnot statisticallysignificant,pvalue>0.05(Table7).
Discussion
The HeCLLiP score is a useful tool tosubstitute thecurrent available simplified scores that is able to accurately estimate mortalityrisk of individualtraumapatients. Having introduced only 5 components, the score performed equally well in
comparison tootherwellknownscoringsystems.Theaccuracy indiscriminatingdeathwasrepresentedwithacomparableAUC value of 0.802.It is also aneasy, objective and wellcalibrated anatomicscoringsystemtopredictmortality.
Itiseasilyscoredaftertheinitialevaluationofpatientsfrom thetraumabay.Simplechestandpelvicx-raywillidentifyanyrib orpelvicfractures.EasyavailabilityofHead,NeckandAbdominal CTscaninthemajorityoftertiaryhospitalswilleasethescoring forhead,cervicalspineorliverinjury.However,ifthereisnoCT imaging available, clinical judgement with supplementary investigations may assist in scoring. The clinical suspicion of head injurycan be diagnosedbased on poorGCS andobvious woundstothehead.Cervicalspineinjurycanbesuspectedbased Table4
MultivariableAnalysisofAllSignificantAnatomicOrganfromUnivariateAnalysis.
[DevelopmentDataset].
p-value OR 95%CI B
Head <0.001 6.97 5.26 9.24 1.942
CervicalSpine <0.001 4.08 2.25 7.40 1.405
Lung <0.001 2.36 1.78 3.13 0.858
Liver <0.001 3.79 2.60 5.52 1.331
PelvisFracture <0.001 4.78 2.78 8.19 1.563
Heart 0.029a 5.47 1.19 25.25 1.699
anotsignificant.
Table5
RegressionAnalysisofAllSignificantAnatomicalVariablestoCreateMortality PredictionModel.[DevelopmentDataset].
p-value OR 95%CI B
Head <0.001 6.88 5.20 9.12 1.929
CervicalSpine <0.001 4.02 2.22 7.29 1.391
Lung <0.001 2.38 1.80 3.16 0.868
Liver <0.001 3.77 2.59 5.50 1.327
PelvisFracture <0.001 4.85 2.84 8.31 1.580
Constant 3.591
0.000 0.100 0.200 0.300 0.400 0.500 0.600 0.700 0.800 0.900
2 3 4 5 6 7 8 9 10 11 12 13 14 15
Fig.1.PredictedProbabilitiesofInHospitalMortalitybasedonTotalScorePointsof HeCLLiPscore.(CombinedDataset).
Table6
ScoreComponentsofHeCLLiPscore.
ScoreComponents ScorePoints
Head 7
Cervicalspine 4
Lung 2
Liver 4
Pelvisfracture 5
Fig.2.ROCanalysisofHeCLLiPscorewithValidationdataset.
on cervical tenderness or swelling with neurological deficits.
Liver injurycan be suspectedbasedon focusedassessment of sonography in trauma (FAST) with presence of free fluid at Morrison’spouch.
The other advantages of HeCLLiP score is the avoidance of utilisingapotentialfalsephysiologicparameterwhichmaybedue tofluctuationofphysiologyeitherduetotooearlyorlaterecording duringtheresuscitativephase.
Theeasy andquickapplicationofHeCLLiP scoreinaclinical settingisexplainedasfollows:1.Presenceofheadinjury,cervical spinefracture,liverinjury,pelvicfracturerepresentsasignificant riskofmortalityasthetotalscorepointsareabovethecutoffvalue whichisgreaterthan2points.Presenceofanyofthese4injuries representsmorethan10%ofmortalityrisk.ThisisshowninFig.1 for which score points of 4 and above carries at least 10%
probabilitiesof death.Thisgroupofpatientsshallbetriagedto earlyresuscitationwithearlyinvolvementoftraumasurgicalteam members.2.Mortalityriskestimationbyusingtotalscorepoints fromHeCLLiPcanbereferedtoFig.1.Forexample,scorepointsof8 correspondstoaround30%riskofmortality(Fig.1).
Previousliteratureidentifiedthattheaccuracyoftheanatomic scoreincreasesifitisdonebyatrainedtraumacareprovideror traumasurgicalteam.[18]Themajorityoftraumavictimswould have received initial resuscitation at pre-hospital setting or emergencydepartmentresuscitationbaypriortotheinvolvement ofseniortraumacareproviderortraumasurgicalteammembers.
Asaresult,thetraumasurgicalteammaynothavethebestinitial physiologicparametersforaccuratescoringforscoresthatheavily reliesonphysiologicparameters.
Previouspapershowedafairlygoodpredictiveabilityforshock index(SI)inthecaseofpredictingearlymortalityduetotrauma [19,20].TheAUROCanalysisafterinclusionofallpatientsfromthe validation sethada reasonablygood valueofAUROCfor shock indexof0.6.(refertoTable7/Fig.3inthemanuscript).Onseparate sub-analysisofAUROCaccordingtoagegroups of15yearsand below,16to64yearsandgreaterthan65yearsrevealedthatthe second and third group had significant accuracy in predicting mortalitywithSI0.9(Tables2and3intheSupplementaryfile), similartopreviousreport[19].Fortheagegroupof15yearsand belowhadpoorassociationwithmortalityprobablyduetosmall samplesize.
TheauthorsidentifiedthatthelowerAUROCvalueofSIinour analysisincomparisontopreviousreportsisprobablyduetothe differentendpoint (whichis in hospital mortality) usedin this current study. We didnot restrict the definition of in-hospital mortalityincurrentstudywithaspecificduration.However,the maximumnumberofdaysfromtraumatodeathrecordedinour studywas140dayswithameantimetodeathof11days.Previous publishedarticlesreportontheshockindex(heartratedividedby blood pressure) revealed that it can accurately predict early mortality rather than in hospital mortality after adequate interventionwasperformed[16,19].Thescorecorrelatesnotonly toearlymortalitybutalsoheavilyreliesonhospitalresource.In contrary toSI, the HeCLLiP score is a better prediction for in- hospitalmortalityratherearlymortalitybecausetheanatomical componentsofthescore(foreg.head,cervicalandpelvis)which maybeassociatedwithhigherriskofsubsequentdebilitatingstate andriskofeventualdeath.
ThoughphysiologicscoressuchasmGAP,GAP,GCSmayappear easyintermsofscoring,butfromouranalysisthesetriagetoolsdid not showed a significanthigher AUC valueto suggest superior mortality predictive ability.The limitations of such physiologic scoreswereidentifiedinapreviousstudybyHueietal.[21].The discrepancyinphysiologicparameterscoringismoreapparentin bloodpressuremonitoring.Thisisduetodelayedbloodpressure recordingsafterrigorousresuscitationatthesceneofmishapor recordedtooearlypriortophysiologicaldeterioration.Thismay limittheaccuracyofphysiologicscores.Furthermore,accurateand continuousdocumentationofbloodpressurerecordingrequirea dedicatedandadequatestaffingformostaccuratedataprocure- ment. Addingthe inadequacyof staff in low tomiddle-income country, this may further reduce the accuracy of physiologic scoring.
InIndia,whichisalsoalowtomiddleincomecountry;Gerdin etaldevisedachartwhichrequirestwophysiologicalparameters (GCSandSystolicbloodpressure)topredictriskofmortality[22].
Thoughthisiseasytoscoreandhasshowntohavegoodpredictive ability,buttheusageofphysiologicparameterssharedthesame inconsistentissuesmentionedearlier[23].
Fig.3. ComparisonofAUROCofHeCLLiPscorewithGAP,mGAP,GCS,ShockIndex andAgeShockIndex.
Table7
AUCvaluesofeachscoreandtheirpairwisecomparison.
Scores AUC 95%CI Pairwisecomparisonbetweenthescores(pvalue)
GAP MGAP GCS ShockIndex AgeShockIndex
HeCLLiP 0.802 0.777to0.824 0.3353 0.8670 0.6998 <0.001 <0.001
GAP 0.825 0.801to0.846 – 0.0022 0.2709 <0.001 <0.001
mGAP 0.806 0.782to0.828 – – 0.7757 <0.001 <0.001
GCS 0.810 0.786to0.832 – – – <0.001 <0.001
ShockIndex 0.603 0.573to0.631 – – – – 0.4654
AgeShockIndex 0.580 0.551to0.609 –
Age, which is a component of the aforementioned scores involvedstratifyingpatientagedmorethan60yearswithhigher risk.However,itmaynotbeusefulinthelocalMalaysianhospital settingasmajorityofthemajortraumainvolvedmanlessthan60 yearsofage[18,21,24].
GCSisagoodobjectivescoretopredictprognosisormortalityin traumapatients. However,its predictive ability varieswith the presenceorabsenceofintracranialinjuries.It isshowntohave betterprediction inhead injured patient ratherthan non-head injuredpatient.Theaccuracyofscoreisalsoaffectedbyalcohol influence,priorsedationandendotrachealintubation[25].
As mentioned above, GAP and mGAP may be affected by inaccurate physiologic parameters which may be scored at inadequate times and the HeCLLiP score may serve as an alternative predictive score to be used at early trauma care setting.Nevertheless,thetraumacareproviderneedsbefamiliar- ized and trained in clinical assessment and interpreting basic investigation such as FAST Scan and Chest x-ray to determine presenceofthisinjury.
Withthesimplicityofthisnewscoringsystem,therearemultiple possibilitiesthattheusermayfalselyscoreahighermortalityrisk.
Forinstance,thepresenceofheadinjuryresultedinascoreofatleast 7 which is above the proposed cut off value. A mild cerebral concussionifscoredpositiveforthe‘headinjury’component,will resultedina falselyhighmortalityriskprediction.Therefore,we proposedthatthe‘headinjury’componentshouldbescoredpositive onlyifthereisanyintracranialhemorrhageonCTscanorGCSof8or less.Anotherlimitationofscoringiswithregardstopelvicfracture injuries.Aminorchipfractureofpelvisbonemayalsofalselyscore positiveforthe‘pelvicfracture’scorecomponent.As‘pelvicfracture’ contributes5points,whichisalsoabovethecut offvalue,weidentify thatit shouldonlybescoredpositiveonlywhen thefracture is involvingsacroiliacjointsorassociatedhemodynamicallyinstabili- ty.However,subsequentstudytorefinethese2scorescomponent intodifferent grades and more precise definition to each score componentmayimprovetheaccuracyforfutureuse.
Thisstudyis limitedby dataof a singletertiarycentre.The newlydevelopedHeCLLiPscoresmayrequireexternalvalidationin othertraumapatientsfromdifferentnations.Thoughthisstudyis based on a retrospective analysis but the data was collected prospectivelybythetraumanurse,whichminimizedtheriskof missingandinaccuratedata.
Conclusion
HeCLLiPScoreisanewlydevelopedabbreviatedanatomicscore suitedtothelocalMalaysiantraumasettingwithgoodpredictive abilityformortality.Thescoremayserveasanalternativetothe simplifiedphysiologicalscoresuchasGAPandmGAP,especially whenaccuratephysiologicparametersmaybealimitation.
Funding
Thiscasereportwasself-funded.
Authors’contributions
Tan Jih Huei was involved in data collection and analysis, draftingandreportingoftheworkdescribedinthearticle.Yuzaidi, HenryandAffirulwereinvolvedinthedraftingofthismanuscript.
RizalandRazmanwereinvolvedincriticallyreviewedthearticle, supervisedthestudy.
Competinginterests
Theauthorsdeclarethattheyhavenocompetinginterests.
Consentforpublication
Studywasdoneretrospectively.Subjectrecruitedweremade anonymous. Informed consent was waived by the Malaysia ResearchEthicCommittee.
Ethicsapprovalandconsenttoparticipate
Ethicalapprovalwas grantedbyMinistryofHealthMalaysia MedicalResearchEthicsCommittee.
Dataavailability
Thedatasetsusedandanalyzedduringthecurrentstudyare availablefromthecorrespondingauthoronreasonablerequest.
Acknowledgments
WewouldliketothankDirectorGeneralofHealthMalaysia,Dr HishamAbdullahforhispermissiontopublishthearticle.
AppendixA.Supplementarydata
Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.
injury.2019.01.027.
References
[1]Official Portal for Ministry of Health Malaysia. Principal causes of hospitalisationanddeath[Internet]. [cited2016Sep24].Availablefrom:.
2019.http://www.moh.gov.my/english.php/pages/view/409.
[2]SartoriusD,LeManachY,DavidJ-S,RancurelE,SmailN,ThicoïpéM,etal.
Mechanism,glasgowcomascale,age,andarterialpressure(MGAP):anew simpleprehospitaltriagescoretopredictmortalityintraumapatients.Crit CareMed2010;38(March(3)):831–7.
[3]KondoY,AbeT,KohshiK,TokudaY,CookEF,KukitaI.Revisedtraumascoring systemtopredictin-hospitalmortalityintheemergencydepartment:
GlasgowComaScale,Age,andSystolicBloodPressurescore.CritCareLond.
Engl.2011;15(4):R191.
[4]ChampionHR,SaccoWJ,CopesWS,GannDS,GennarelliTA,FlanaganME.A revisionofthetraumascore.JTrauma1989;29(May(5)):623–9.
[5]Injuryscoringscales-theAmericanassociationforthesurgeryoftrauma [Internet].2019..[cited2016Sep2].Availablefrom:http://www.aast.org/
library/traumatools/injuryscoringscales.aspx.
[6]BoydCR,TolsonMA,CopesWS.Evaluatingtraumacare:theTRISSmethod.
TraumaScoreandtheInjurySeverityScore.JTrauma1987;27(April(4)):370–8.
[7]PerelP,Prieto-MerinoD,ShakurH,ClaytonT,LeckyF,BouamraO, etal.
Predictingearlydeathinpatientswithtraumaticbleeding:developmentand validationofprognosticmodel.BMJ2012;345(August):e5166.
[8]ChampionHR,CopesWS,SaccoWJ,FreyCF,HolcroftJW,HoytDB,etal.
Improvedpredictionsfromaseveritycharacterizationoftrauma(ASCOT)over TraumaandInjurySeverityScore(TRISS):resultsofanindependent evaluation.JTrauma1996;40(January(1))42–8discussion48–49.
[9]St-LouisE,BraccoD,HanleyJ,RazekT,BairdR.Developmentandvalidationof anewpediatricresuscitationandtraumaoutcome(PRESTO)modelusingthe U.S.NationalTraumaDataBank.JPediatrSurg2017(October).
[10]GlanceLG,OslerTM,MukamelDB,MeredithW,WagnerJ,DickAW.TMPM- ICD9:atraumamortalitypredictionmodelbasedonICD-9-CMcodes.Ann Surg2009;249(June(6)):1032–9.
[11]O’Reilly GM, Joshipura M, Cameron PA, Gruen R. Trauma registries in developingcountries:areviewofthepublishedexperience.Injury2013;44 (June(6)):713–21.
[12]SabariahFJ,RameshN,MahatharAW.NationalTraumaDatabase(NTrD)–
improvingtraumacare:firstyearreport.MedJMalaysia2008;63(September (SupplC)):45–9.
[13]PortalRasmiKementerianKesihatanMalaysia-PembedahanAm[Internet].
2019..[cited2016Jun21].Availablefrom:http://www.moh.gov.my/index.
php/pages/view/1329.
[14]GennarelliTA, Wodzin E. AIS2005: acontemporary injuryscale.Injury 2006;37(Dec(12)):1083–91.
[15]MontoyaKF,CharryJD,Calle-ToroJS,NúñezLR,PovedaG.Shockindexasa mortalitypredictorinpatientswithacutepolytrauma.JAcuteDis2015;4 (August(3)):202–4.
[16]ZarzaurBL,CroceMA,FischerPE,MagnottiLJ,FabianTC.Newvitalsafter injury:shockindexfortheyoungandagexshockindexfortheold.JSurgRes 2008;147(Jun(2)):229–36.
[17]PavlouM,AmblerG,SeamanSR,GuttmannO,ElliottP,KingM,etal.Howto developamoreaccurateriskpredictionmodelwhentherearefewevents.BMJ [Internet]2015(August)..[cited2017Feb22];351.Availablefrom:http://www.
ncbi.nlm.nih.gov/pmc/articles/PMC4531311/.
[18]TanJH,TanHCL,NohNAM,MohamadY,AlwiRI.Validationofthetrauma mortalitypredictionscoresfromaMalaysianpopulation.BurnsTrauma [Internet]2017;5(December)..Availablefrom:https://www.ncbi.nlm.nih.gov/
pmc/articles/PMC5740795/.
[19]McNabA,BurnsB,BhullarI,ChesireD,KerwinA.Aprehospitalshockindexfor traumacorrelateswithmeasuresofhospitalresourceuseandmortality.
Surgery2012;152(September(3)):473–6.
[20]ZarzaurBL,CroceMA,MagnottiLJ,FabianTC.Identifyinglife-threatening shockintheolderinjuredpatient:ananalysisoftheNationalTraumaData Bank.JTrauma2010;68(May(5)):1134–8.
[21]HueiTJ,MohamadY,LipHTC,NohNM,AlwiRI.Prognosticpredictorsofearly mortalityfromexsanguinationinadulttrauma:aMalaysiantraumacenter experience.TraumaSurgAcuteCareOpen.2017;2(May(1))e000070.
[22]GerdinM,RoyN,KhajanchiM,KumarV,DharapS,Felländer-TsaiL,etal.
Predictingearlymortalityinadulttraumapatientsadmittedtothreepublic universityhospitalsinurbanindia:aprospectivemulticentrecohortstudy.
PLoSONE[Internet]2014(September)..[cited2018Oct28];9(9).Available from:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4152220/.
[23]GerdinM,RoyN,KhajanchiM,KumarV,Felländer-TsaiL,PetzoldM,etal.
Validationofanovelpredictionmodelforearlymortalityinadulttrauma patientsinthreepublicuniversityhospitalsinurbanIndia.BMCEmergMed 2016;16(February):15.
[24]TanChorLipH,TanJH,MohamadY,AriffinAC,ImranR,AzmahTuanMatTN.
Clinicalcharacteristicsof1653injuredmotorcyclistsandfactorsthatpredict mortalityfrommotorcyclecrashesinMalaysia.ChinJTraumatol.Zhonghua ChuangShangZaZhi2018(November).
[25]ZuercherM,UmmenhoferW,BaltussenA,WalderB.TheuseofGlasgowComa Scaleininjuryassessment:acriticalreview.BrainInj2009;23(January (5)):371–84.