Increase in COVID-19 inpatient survival following detection of Thromboembolic and Cytokine storm risk from the point of admission to hospital by a near real time Traffic-light System (TraCe-Tic)

w w w . e l s e v i e r . c o m / l o c a t e / b j i d

The

Brazilian

Journal

of

INFECTIOUS

DISEASES

Original

article

Increase

in

COVID-19

inpatient

survival

following

detection

of

Thromboembolic

and

Cytokine

storm

risk

from

the

point

of

admission

to

hospital

by

a

near

real

time

Traffic-light

System

(TraCe-Tic)

Marcela

P.

Vizcaychipi

_,

_Claire

_L.

_Shovlin

_,

_Alex

_McCarthy

_,

Andrew

Godfrey

_,

_Sheena

_Patel

_,

_Pallav

_L.

_Shah

_,

_Michelle

_Hayes

_,

_Richard

_T.

_Keays

_,

Iain

Beveridge

_,

_Gary

_Davies

_,

_Gary

_Davies

_on

_behalf

_of

_the

_ChelWest

_COVID19

Consortium

a_{Chelsea&WestminsterHospitalNHSFoundationTrust,AnaesthesiaandIntensiveCare,London,UnitedKingdom} b_{ImperialCollegeLondon,DepartmentofSurgeryandCancer,London,UnitedKingdom}

c_{ImperialCollegeLondon,NHLIVascularScience,London,UnitedKingdom}

d_{Chelsea&WestminsterHospitalNHSFoundationTrust,Information,DataQualityandClinicalCoding,London,UnitedKingdom} e_{Chelsea&WestminsterHospitalNHSFoundationTrust,Haematology,London,UnitedKingdom}

f_{Chelsea&WestminsterHospitalNHSFoundationTrust,Pharmacy,London,UnitedKingdom}

g_{Chelsea&WestminsterHospitalNHSFoundationTrust,RespiratoryMedicine,London,UnitedKingdom}

h_{WestMiddlesexUniversityHospitalNHSFoundationTrust,DepartmentofAnaesthesia&IntensiveCareMedicine,Isleworth,United}

Kingdom

a

r

t

i

c

l

e

i

n

f

o

Received17June2020 Accepted31July2020

Availableonline18August2020

Keywords: Anticoagulation C-reactiveprotein D-dimer Discharge Ferritin Mortality SARS-CoV2

a

b

s

t

r

a

c

t

Introduction:Ourgoalwastoevaluateiftraffic-lightdrivenpersonalizedcareforCOVID-19 wasassociatedwithimprovedsurvivalinacutehospitalsettings.

Methods:Dischargeoutcomeswereevaluatedbeforeandafterprospectiveimplementation ofareal-timedashboardwithfeedbacktoward-basedclinicians.Thromboembolism cate-gorieswere“medium-risk”(D-dimer>1000ng/mLorCRP>200mg/L);“high-risk”(D-dimer >3000ng/mLorCRP>250mg/L)or“suspected”(D-dimer>5000ng/mL).Cytokinestormrisk wascategorizedbyferritin.

Results:939/1039COVID-19positivepatients(medianage67years,563/939(60%)male) com-pletedhospitalencounterstodeathordischargeby21stMay2020.Thromboembolismflag criteriawerereachedby568/939(60.5%),including238/275(86.6%) ofthepatientswho died,and330/664(49.7%)ofthepatientswhosurvivedtodischarge,p<0.0001.Cytokine stormflag criteriawerereachedby212(22.6%) ofadmissions,including80/275 (29.1%) ofthepatients whodied,and 132/664(19.9%) ofthe patientswhosurvived, p<0.0001.

∗ _{Correspondingauthor.}

E-mailaddresses:Marcela.Vizcaychipi@chelwest.nhs.uk(M.P.Vizcaychipi),c.shovlin@imperial.ac.uk(C.L.Shovlin).

i _{MPV,CLSandAMcontributedequallytothiswork.}

https://doi.org/10.1016/j.bjid.2020.07.010

1413-8670/©2020SociedadeBrasileiradeInfectologia.PublishedbyElsevierEspa ˜na,S.L.U.ThisisanopenaccessarticleundertheCC BY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4.0/).

Themaximumthromboembolismflagdiscriminatedcompletedencountermortality(no flag:37/371[9.97%]died;medium-risk:68/239[28.5%];high-risk:105/205[51.2%];and sus-pected thromboembolism: 65/124 [52.4%], p<0.0001). Flag criteria werereached by535 consecutiveCOVID-19positivepatientswhosehospitalencountercompletedbefore traffic-lightintroduction:173/535(32.3%[95%confidenceintervals28.0,36.0])died.Forthe200 consecutiveadmissionsafterimplementationofreal-timetrafficlightflags,46/200(23.0% [95%confidenceintervals17.1,28.9])died,p=0.013.Adjustedforageandsex,theprobability ofdeathwas0.33(95%confidenceintervals0.30,0.37)beforetrafficlightimplementation, 0.22(0.17,0.27)afterimplementation,p<0.001.Insubgroupanalyses,olderpatients,males, andpatientswith hypertension(p≤0.01),and/ordiabetes(p=0.05) derivedthegreatest benefitfromadmissionunderthetrafficlightsystem.

Conclusion: Personalizedearlyinterventionswereassociatedwitha33%reductioninearly mortality.Wesuggestbenefitpredominantlyresultedfromearlytriggerstoreview/enhance anticoagulationmanagement,withoutexposinglower-riskpatientstopotentialrisksoffull anticoagulationtherapy.

©2020SociedadeBrasileiradeInfectologia.PublishedbyElsevierEspa ˜na,S.L.U.Thisis anopenaccessarticleundertheCCBY-NC-NDlicense(http://creativecommons.org/ licenses/by-nc-nd/4.0/).

Introduction

Humaninfectionduetothenovelsevereacuterespiratory syn-dromecoronavirus2(SARS-CoV-2)1_{(commonlyreferredtoas} COVID-19disease)hashadunprecedentedglobalimpacton societyandhealthcareprovision.Inthefirstsevenmonthsof 2020,therehavebeenmorethan600,000fatalities,withthree countries(USA,BrazilandUK)eachreportingmorethan40,000 deaths.2

Current management protocols have generally emerged fromChina,whereriskfactorsforin-hospitaldeathincluded older age, higher Sequential Organ Failure Assessment (SOFA) score, and D-dimer greater than 750ng/mL on admission.3_{Cytokinepatterns}3–5_{wererecognizedas} compat-ible witha “cytokine storm” of dysregulated inflammation and hyperpyrexia.6–9 _{Adult respiratory distress syndrome} (ARDS),acutekidney,cardiacandliverfailure,andacytokine stormdrivinglaterdeteriorationshavebeenfociofresearch and clinical initiatives, alongside development of vaccines andantiviraltherapeutics.10_{Autopsystudies}11–14 _provideda slightlydifferentpictureastheriskoffatalthromboembolic diseaseemerged, inkeepingwiththe initialreportsof ele-vatedD-dimerassociations.3_{Thromboembolicdiseaseisnow} estimatedtoaffecthighproportionsofpatientswith COVID-19.15–18

COVID-19 displays heterogenous presentation patterns andoutcomes.Patientcomorbiditiesincludinghypertension, diabetes,obesity,andheartdiseaseareassociatedwithhigher riskofsevereCOVID-19disease.3,4,19–21 _{whilepatients with} additionalcommondiseases includingasthma andchronic obstructivelungdisease(COPD)areadvisedtotakeextra pre-cautionsintheUSA.22_{ReportedCOVID-19inpatientmortality} ratesvarybycity/region,23_{betweencountries,}1,3,4,16,17,19,23–25 andbydiseaseseverity.3,4,23–26_{Differentialsurvivalratescan} thereforereflecta multitudeofconfoundersfrom patients, geography,andhealthcaresystems,superimposedon patho-physiologicalvariabilitythatisnotyetunderstood.

Anearreal-timedecisiontoolwasintroducedearlyinthe pandemic in our institution, as we were concerned about thehighpublishedmortalityrates.Thistoolrefreshesinthe background every 10minand updates aclinical dashboard accordingly.Generatedontheelectronicpatientrecord,traffic lightswereusedtoalertclinicianswhenpatientsmet crite-riaforincreasedriskofthe threemostcommon causesof deathinCOVID-19–thromboembolism,cytokinestorm,and respiratorydeterioration/ARDS.

The goal of the current study was to evaluate whether prospectiveapplicationduringAprilandMay2020hadbeen associatedwithanychangeinpatientearlymortalityacross completedhospitalepisodes.

Materials

and

methods

Institutionanddetails

All patientswere consecutiveadmissionstotheemergency department. A COVID-19 battery of tests was developed in electronic medical records and introduced into clinical practice in early March 2020. The COVID pathology order was requested for all patients admittedto the emergency department with suspected and or confirmed COVID-19 infection,andsubsequentlyrequestedbyphysicians through-out patients’ staytomonitortrendsand adjusttreatments accordingly. This standardized COVID-19 careset spanned complete blood count (CBC); coagulation including pro-thrombintime,activatedpartialthromboplastintime(APTT), fibrinogen, and D-dimer; electrolytes (sodium, potassium); renalfunction(urea,creatinine);liverfunction(bilirubin, alka-linephosphatase,albumin,aspartatetransaminase,alanine transaminase, gamma-glutamyltransferase (GGT), lactate dehydrogenase (LDH); iron studies (serum iron, transfer-rin saturation index, ferritin), and additional acute phase responsemarkersincludingC-reactiveprotein(CRP).

Trafficlightsystem

Anearreal-timedecisionsupporttoolwasdevelopedandset uponeachelectronicpatientrecord.27_{Thiswascreatedforuse} withCOVID-19patientsinMarch2020(Supplementary mate-rial1).IninitialstagesoftheCOVID-19pandemic,acomposite ofvariableswhichincludedfirstvitalsignsandthe COVID-19 battery of test results on admission to the emergency departmentwasusedtomodelthedisease.Inbrief,our clin-icaldatasetwaspassedthroughaNeuralNetworkalgorithm containedinSQLServerAnalysisServices(SSAS)Data Min-ingPackage.Theinputvariableswereage;sex;firstrecorded systolicblood pressure(BP); firstrecorded diastolicBP;first recorded respiratory rate; first recorded temperature; first recorded oxygen saturation (SpO2);first recorded fractional

inspiredoxygenconcentration(FiO2);firstrecordedheartrate,

andfirstrecordedc-reactiveprotein(CRP).Werapidlylearnt thattheCRPshowedhighlypredictivevaluewhenpatients presentedwithD-dimer<3000ng/mL.Inviewofthelackof randomizedcontrolledtrialdata,andpost-mortem commu-nications predating references11–14 _{the authors felt it was} prudenttoincludetheminimalD-dimercutoffof750ng/mL publishedbyChinesecolleaguesattheendofJanuary2020.3 Forsimplicitythevaluewasroundedto1000ng/mL.

Between20thMarch2020and12thApril2020,adailyreport wasissuedonallCOVID-19positivepatientsinthe institu-tion, and sent to an experienced critical care clinician for review.Results were interpretedand fedback tothe clini-cal teams to review,and modifypathways as appropriate, withdatafrom thefirst phasedefining patternsofdisease and biomarkers.Therewere recommended anticoagulation protocolsandguidelinesformanagementofcytokinestorm (Table1)whichthetreatingphysicianadoptedandtailoredto patient’sresponsetotreatmentwhenindicated.

Thetrafficlightsystemwentliveacrossthefullinstitution on12th April2020,updatedevery 10min, for implementa-tion into prospectivepathways byall clinicians.Clinicians were alsoupdatedbywhatsApps,emails, and phone calls, followingelectronicmedicalrecord monitoringof prescrip-tioncomplianceonaregularbasis.Themosteffectivewayof communicationwasbyutilizationofwhatsAppsgroupswith updatesoftheinstitutionalhematologyguidelines.Thedate ofintroduction(12thApril2020)wasusedtocohortpatients inthestudy,asdescribedfurtherbelow.

Ethics

Dataprocessingwasauthorizedbythe ClinicalInformation Governancecommittees.Accesstothedataandauthorization forthepresentstudywasjointlygrantedbyboththe Institu-tion’sDataProtectionOfficeandbytheInstitution’sAnalytical Unit, under the General Data Processing Regulations (pro-cessing underpublic authority forpurposesin the area of publichealth).ToensurecompliancewithGeneralData Pro-tectionRegulations,datawereextractedfrompseudonymized datasets into aggregate reports only for the outcomes of interest.Thecorrespondingauthorhadfullaccesstoallthe datainthestudy and thefinalresponsibilitytosubmit for publication.

Cohortadmissioncategorizations

ThestudycohortrepresentedallCOVID-19positivepatients whohadcompletedtheirhospitalencounterby21stMay2020. Patientswereeitherdischargedhomealive(includingto tem-poraryhomesand/orresidentialcarehomes),orhaddied.At the dateofreporting,patientswhoremained inthe admit-tinghospital,orhadbeentransferredtootherhospitals,were consideredtohavean“incomplete”encounterandwerenot includedintheanalysis.

Patientswithcompletedencounterswerecategorizedinto threecohorts,accordingtotiminginrelationto12thApril2020 whenthetrafficlightsystemwentliveacrossthefull institu-tion.The“Pre”cohortcompletedtheirhospitalencounter(to homedischargeordeath)before12thApril2020.The“Post” cohort were admitted on or after12th April2020, so their admissionwasfullymanagedbythereal-timesystem.Athird group wasadmittedbefore,but dischargedafter12thApril 2020(the“Bridge”cohort).

Thecomorbiditiesofasthma,COPD,diabetes,and hyper-tensionweredefinedwherelistedasaDiagnosisorProblem on the electronicpatient record or historiccoding.Aprior pharmacy record of salbutamol prescription also placed patients in the asthma category, and a requested glycated hemoglobin (HbA1c) level placed patients in the diabetic category. Patients could have one, several or none of the comorbidities.

Dataanalyses

Anonymized data were analyzed using Stata IC version 15 (Statacorp, Texas) across the full cohort and across all time periods.Comparisonswere performedusingthe non-parametricKruskal–Wallisequality-of-populationsranktest, withDunn’smultiplecomparisontestusedtoderivethe mul-tiplecomparison-adjusted,two-tailedp-values.Toadjustfor differencesinage,sexandothervariablesbetweenthetime periods,multivariatelogisticregressionwasperformedforthe binaryoutcomeofdeath(‘1)versushomedischarge(‘0).The probabilities ofdeathin the “Pre”and “Post” cohorts were calculated by post-test marginalcomparisons afterlogistic regression acrossthe full939casesusingmodelcovariates ofage,sexanda3-cohortvariable distinguishingthe“Pre”, “Bridge”and“Post”cohorts.Forillustrationofsummary statis-ticsandheatmaps,datawereexportedtoGraphPadPrism 8.1.1forWindows(GraphPadSoftware,SanDiego,California USA).

Calculation

Itisalready knownthat COVID-19causes severemortality, particularlyinpatientsrequiringinvasiveventilatorysupport, developingacytokinestorm,orexperiencingthromboembolic disease.Thereislittleevidenceonwhethertoinstitute pre-ventative measures, (andif so, inwhichsubpopulationsof COVID-19infectedpatients),orwhethertowaitforpatients todevelopend-stagedisease.

Table1–Trafficlightdiagnosticflagsandoveralloutcomes. Predictors Cases N(%) Recommendation Mortality N(%) C-reactive protein mg/L D-dimer ng/mL Ferritin ␮g/L Thromboembolism

Lowrisk Noneoftheflagsbelow 371(39.5%) Enoxaparin40mgod 37/371(9.97%).

Mediumrisk >200or >1000 239(25.5%) Enoxaparin40mgbd 68/239(28.5%)

Highrisk >250or >3000 205(21.8%) Therapeutica/ca _{105/205(51.2%)}

Suspected >5000 124(13.2%) Therapeutica/ca _{65/124(52.4%)} Cytokinestorm

Lowrisk Noneoftheflagsbelow 727(77.4%) 195/727(26.8%) Mediumrisk >1000 143(15.2%) ivFluids,a_{AICUreview 48/143(33.6%)}

Highrisk >2000 34(3.6%) ivfluids,VitC,a_{AICUreview 19/34(55.9%)}

Suspected >4000 35(3.7%) ivfluids,VitC,a_{AICU 13/35(37.1%)}

a _{Therapeutica/c,therapeuticanticoagulationunlesscontraindications;VitC,ascorbicacid1gbd;ivfluids,fluidmanagementadjustedto}

replaceinsensiblelosses(1mlkg−1h−1foreverydegreeabove38◦C).Iftherewasnoresponsetoinitialfluidmanagementanddiastolicblood pressureremained<35mmHg,oroxygenrequirementincreased,thenthepatientswerereferredtotheAICUforfurthermanagement.

(1) bloodtestresultsindynamicreassessmentswould dis-criminate patients at or becoming at higher risk of complications;

(2) ifrapidlycommunicatedtoclinicians,higherriskpatients couldbetargetedforearlyreviewandimplementationof acceleratedenhancedtreatments;

(3) early implementation of accelerated enhanced treat-ments, targeted at individual level at the point of entry, would translate to improved outcome demon-stratedbymorepatientssurvivingtobedischargedhome alive.

Results

Studycohort

Atotalof1039COVID-positivepatientshadbeenadmittedby 21stMay2020.Ofthose,889(85.6%)weremanagedongeneral wards,and150/1039(14.4%)wereadmittedtoadultintensive careforpartoftheirinpatientstay.Ofthe1039patients,664 had been dischargedhome,eitherto theirusual residence (n=614),atemporaryhome(n=16),oraresidentialcarehome (n=34),and275haddied.Afurther100patientswerestillin thehospital,orofunknownoutcomefollowingtransfertoa differenthospital.

The939/1039whohadcompletedtheirhospitalencounter by 21st May 2020 constituted the current study cohort. Patients’medianagewas67 years(interquartilerange, IQR 54–81years)i.e.morethan25%ofadmittedpatientswereat least81yearsold,and563/939 (60%)weremales.The pop-ulation comprised 62% white (370/597 recorded), and 38% blackandminorityethnic(227/597recorded).Comorbidities were common: 500/939 (53%) patients were hypertensive; 354/939 (38%) were in the diabetic category, 226/939 (24%) weredefinedashavingasthma,and95/939(10%)withCOPD (Table2).

Trafficlightcategories

Ofthe939COVID-positivepatientsinthecohort,371(39.5%) patients had no thromboembolism flag (D-dimer values were ≤1000ng/mL, and CRP values ≤200mg/L) For 239/939 (25.5%)thehighestthromboembolismcategoryreachedwas a“mediumrisk”flag(D-dimerbetween1001and3000ng/mL, oraCRPbetween201and250mg/L).329/939(35.0%)received aflagforhigherrisk–205/939(21.8%)the“highrisk” throm-boembolism flag(D-dimerbetween3001–5000ng/mLorCRP >250mg/L) and124/939 (13.2%)the“suspected thromboem-bolism”flag(D-dimer>5000ng/mL).

Across all patients’ completed encounters to discharge home or death, mortalitywas lowest in the patients who received no thromboembolism flag (37/371, 9.97%). Mortal-ity was intermediate inthose withthe “medium-risk” flag (68/239,28.5%),andhighestinthepatientswiththehigh-risk flag(105/205,51.2%),orsuspectedthromboembolism(65/124, 52.4%)(Fig.1).

Forcytokinestorm,727/939(77.4%)patientsdidnotreceive aflagduringtheiradmission. For143/939 (15.2%)the high-est cytokine storm category reached was a “medium-risk” flag, 34/939(3.6%) receiveda“high-risk” flag,and 35 (3.7%) wereflaggedwith“suspected”cytokinestorm.Mortality dif-fered overall between the four categories (Kruskal Wallis

p<0.0001),predominantlyattributabletohighermortalityin the69patientswhowereflaggedinthetwohigherrisk cate-gories(Fig.1).

Categorizationbytrafficlightreal-timeapplication

Ofthe939patientswithacompletedencounter,535(57%)were inthe“Pre”cohortadmittedanddischargedordiedbefore/on 12thApril2020.204(22%)patientswereinthe“Bridge”cohort, and 200 (21%)were inthe “Post”cohort admittedand dis-chargedordied after12thApril2020(SupplementaryTable A.1).

Table2–CohortCharacteristics.

Totaldatasets Median(IQR)/N(%) 1–99%

Riskfactors

Sex(male) 939 563–60.0% –

Hypertension 939 500–53.3% –

Diabetes 939 354–37.7% –

Asthma 939 226–24.1% –

Chronicobstructivepulmonarydisease(COPD) 939 95–10.1% –

Admissioncharacteristics

Age(years) 939 67(54,81) 1–96

Majora_{symptomdurationatpresentation(min)}a _{793 515(170,4858) 34–29,744}

Majora_{symptomdurationatpresentation(h)}a _{793 8.6(2.8,80.9) 0.6–496(20.6d)}a

Respiratoryrate(min−1) 893 23(19,28) 14–50 Oxygensaturation(SpO2,%) 893 96(93,98) 71–100

Urea(mmol/L) 848 5.9(4.0,9.9) 1.9–34.8

Creatinine(␮mol/L) 842 84(67,113) 30–590

Whitebloodcellcount(×109_{/L) 895 7.4(5.6,10.3) 2.6–28.5}

Lymphocytecount(×109_{/L) 895 0.9(0.6,1.4) 0.2–5.1}

Fibrinogen(g/L) 579 6.52(5.3,7.7) 1.99–12

Activatedpartialthromboplastintime(s) 637 32.1(29.4,35.6) 23.3–52.2

Hematocrit 895 0.40(0.36,0.44) 0.22–0.53

C-reactiveprotein(mg/L) 902 97.0(40,171) 1.8–409

D-dimer(ng/mL) 565 1520(818,2746) 270to>20,000(ULS)

Ferritin(␮g/L) 641 526(245,1100) 17–7280

Totallengthofstay(d)a _{939 7(3,13) <1–58}

Dischargedalive(d)a _{664 7(3,13) <1–58}

Deceased(d)a _{275 6(4,11) 1–74}

a _{Themostcommonsinglesymptomsweredyspnea(77.4%),chestpain(19.9%)andcollapse(6.1%).min:minutes.h:hours.s:seconds.d:days.}

ULS:upperlimitofscale.

Fig.1–Mortalityfor939completedencounterstodeathorhomedischargebasedontrafficlightcategory.(A)

Thromboembolism.(B)Cytokinestorm.Numberswithinthebarsindicatethetotalnumberofthe939patientswhoreached thecategoryastheirmaximumflagforthecategorytype.p-valueswerecalculatedforeachcategoryacrossall939patients usingtheKruskal–Wallisequality-of-populationsranktest,withDunn’stestmultiplecomparison-adjustedp-values presented.

Themedianageofthe“Pre”cohortwas67years(IQR52–80). Themedianageofthe“Post”cohortwas72years(IQR56,82.5),

p=0.037(means and 95% confidenceintervals displayed in Fig.2A).Incrudeanalyses,173/535patientsinthe“Pre”cohort died(32.3%[95%confidenceinterval28.0–36.0]),comparedto 46/200patientsinthe“Post”cohort(23.0%[95% confidence

interval 17.1–28.9]), Dunn’s test p=0.013 (Fig.2B).An inter-mediate mortalitywasobservedin the204patients whose admissionspannedreal-timetrafficlightintroductionon12th April2020–ofthese,56/204died(Fig.2B).Thiswasnot signif-icantlydifferenttothatofthe ¨Pre ¨or ¨Post ¨cohorts(Dunn’stest

Fig.2–Ageandsurvivaltodischargealiveinthe939COVID-19positivepatients.Meanand95%confidenceintervals indicatedfor(A)Age(years),(B)Survival(percentage),categorizedbyadmissionanddischargedatesinrelationto12thApril 2020whenthetrafficlightsystemwentliveacrossthefullinstitution.

Full comparisons of the cohorts are provided in Sup-plementaryTableA.2. Trends forthe“Post” cohortto have higherproportionsflaggedatmedium-risk(p=0.27)or high-risk(p=0.23) ofthromboembolismdid notreach statistical significance,thoughahigherproportionofthe“Post”cohort wereflaggedwith“suspected”thromboembolismcompared tothe“Pre”cohort(p=0.019,Fig.3A–C).Therewasno differ-encebetweenthe“Pre”and“Post”subgroupsinthesmaller proportionswithmedium-risk(p=0.67),high-risk(p=0.96)or suspected(p=0.86)cytokinestormflags(Fig.3D–F).

Logisticregression

Inlogisticregressionanalyses,adjustedoddsratiosfor mor-talityweresmaller(i.e.morefavorable)thanthecrudeodds ratioforthe“Post”cohortfollowingadjustmentforage,sex, andallcombinationsofvariablesinTable1(datanotshown). Across all 939 patients, and adjusted for age and sex, theprobability ofdeathwas0.33 (95%confidenceintervals 0.30–0.37)inthe“Pre”cohort,and0.22(95%confidence inter-vals0.17–0.27)inthe“Post”cohort,p<0.001(Fig.4).

Cohortsubgroupanalyses

Death-ratesacrossthethreetimeperiodswereexaminedfor keypatientsubgroups.Comparedtothe“Pre”cohort,death rateswere lower inthe“Post”cohort formales(p=0.0014), forpatientsintheupperthreeagequartilesandabove(≥54 years, p=0.0014), particularly the third age quartile (68–79 years,p=0.0014);forpatientswithhypertension(p=0.0072), andforpatientswithdiabetes(p=0.05)(Fig.5).

Discussion

Wehaveshownthatearlyintroductionofnearrealtimetraffic light-driven,personalizedcarewasassociatedwithanoverall reductionincrudemortalityfrom32.3%(95%confidence

inter-val28.0,36.0),to23.0%(95%confidenceinterval17.1,28.9).The overallageandsex-adjustedprobabilityofmortalityreduced from 0.33(95%confidenceintervals 0.30,0.37), to0.22(95% confidenceinterval0.17,0.27),p<0.001.Thiswasdespitethe latercohortincludingmorepatientswiththromboembolism flags associatedwithpoorer survival. Thegreatestbenefits wereobservedinolderpatients,males,andinpatientswith hypertension(p-values≤0.01)and/ordiabetes(p=0.05).Taken together,thesedatasuggestthattheearlyclinical interven-tionsmodifiedprogressionofCOVID-19induceddisease,and reducedtheproportionofpatientsprogressingtocritical ill-nessanddeath.

There were several study limitations with many shared by other studies in emergency-based assessments of this newhumandiseasewithveryhighmortality.Interventions prioritized clinical care delivery, and the main purposeof this manuscript was to stratify the disease and introduce earlymedicalmanagement,ratherthandetailedexpositions ofpopulation characteristicsand symptomatology. Patients withveryprolongedhospitalizationdidnothavecompleted encountersandwerenotincluded:responsesinthis impor-tant subgroup will need evaluation in future prospective randomizedcontrolledtrials.Additionally,capturing mortal-itytoend-of-hospitaladmissionmayhavemissedsubsequent deathsinthecommunity,andthesemayhavedifferedin pro-portionsoverthestudyperiod.However,forcrudemortality inthe“Post”cohorttoreachthatofthe“Pre”cohort,an86% mortalityratewouldberequiredintheinpatientsadmitted on/after 12thApril 2020,when theircurrent lengthofstay IQRalreadyexceedstheIQRforlengthofstaybeforedeath (SupplementaryTableA.1).Therefore,thisreductioninearly mortalityisconsideredtoberobust,andimportantto com-municate.

Thestudywasconductedatatwo-siteinstitution, there-foreitisdifficulttocomparebaselinemortalitieswithother institutions.Lowerbaselinemortalitywasreportedfor2569 completedepisodesinadifferenthealthcaresettinginNew

Fig.3–Proportionsofthe939COVID-19positivepatientswiththromboembolismorcytokinestormflags.Meanand95% confidenceintervalsindicatedfor(A)Medium-riskthromboembolism;(B)High-riskthromboembolism;(C)Suspected thromboembolism;(D)Medium-riskcytokinestorm;(E)High-riskcytokinestorm;and(F)Suspectedcytokinestorm.

Fig.4–ProbabilityofDeathin“Pre”and“Post”TrafficLight cohorts.Illustrationofthepost-testmarginalcomparison of“Pre”and“Post”cohortsafterlogisticregressionofdeath (versusdischarge)acrossall939patients,usingmodel covariatesofage,sexanda3-cohortvariabledefining patientsinthe“Pre”,“Bridge”and“Post”cohorts.The overallmodelofthe939patientshadapseudor2_of12.2 andp-value<0.0001.

York,where665/2569(25.9%)patientsdied,though highlight-ingthepotentialproblemsofcomparingdifferentoutcomes in different institutions, the New York cohort24 _{were ∼6} years younger, and their interquartile range (IQR) for D-dimerdidnotoverlapwiththecurrentcohort(NewYorkIQR 237–714ng/mL,currentcohort818–2746ng/mL).

Withtheveryhighworldwidemortality2_{andstrongsteers} towardriskfactorsfromtheearliestreportsfromChina,1,3_it wasnotconsideredappropriatetorandomizepatients.Where lethality ishigh,as inCOVID-19,the paucityofsequential analysistechniqueshasbeennotedwithconcern.28_The cur-rentstudyemployedasequentialanalysistechnique,within adualcenter,singleinstitution,anddemonstratedthatthe introductionoffulltrafficlight-drivencareon12thApril2020 wasassociatedwitha33%reductioninageandsex-adjusted inpatientmortality.Werecognizethatalimitationofa sequen-tial analysis is that it is not possible to fully standardize variables,andacknowledgethatotherfactorscouldhave con-tributed.Itisfeasiblethatinthecurrentstudy,thebaseline “Pre” groupwere alreadybenefitingfrom the firstphaseof traffic light implementation,27 _{and different approaches to} care.29_{Thesurvivalsignalwasrobusttoageandsex} adjust-ment; adjustment for patient severity indices would have madetheassociationstronger,andothermajortreatment pro-tocolsdidnotchangeinthestudyperiod.Clinicalexpertise andpathwaysdidhaveextratimetoembed,thoughwenote unchanged mortality figures in major country-wideaudits overthesameperiod.30_{Itisthereforedifficulttoconcludethat} anyotherfactorcouldhavebeenresponsibleforgeneratinga completed encounterinpatientmortalityratethatat23.0% [95%confidenceintervals17.1,28.9]isnowlowerthaninNew York,24_{andlowerthantheratesreportedduringtheMiddle} East RespiratorySyndrome(MERS)epidemic.31 _Further,this completedencountermortalityratewasachievedwithmore than25%ofthepatientpopulationaged81yearsoroverat thetimeofadmission,andwithadvanceddiseaseat presen-tationasindicatedbytheinterquartilerangesforadmission

Fig.5–Heatmapforproportionofpatientsdischarged aliveordied,bysubcategories.Colorsrangefromgreen (100%survival/zeroriskofdeath)tored(50%chanceof survival/50%riskofdeath).Thetotalnumberofpatients perrow(n)isindicatedtotherightoftheheatmap.The heatmapstars(*)indicateadjustedp-valuecalculatedfor eachcategoryusingtheKruskal–Wallis

equality-of-populationsranktest,withDunn’smultiple comparisontestusedtoderivethemultiple

comparison-adjustedpvaluesforthe“Pre”versus“Post” comparison*p≤0.05;**p≤0.01;and***p≤0.005.

CRP(40–171mg/L),admissionD-dimer(818–2746ng/mL)and admissionferritin(245–1100␮g/L).

Despitemoreseverediseaseinthe“Bridge”cohort,survival ratewasnoworsethanforthe“Pre”cohort.Thissuggeststhat the“Bridge”cohortmayhavealsobenefittedinthefirstphase oftraffic-lightimplementation.27

Numerically,the keydifferences comparedtoourinitial practicewereearlier,enhancedanticoagulationforpatients identifiedasathigherrisk ofthromboembolism disease.A smallergroup ofpatientsflagged atriskofcytokine storm basedonserumferritinalsohadenhancedattentiontofluid balanceand intravenous fluiddelivery.For thrombosis,the flagshelpedtoidentifythestratifiedrisk,andtheintervention wasbasedontheriskregardlessoftheradiologicalfindings. InCOVID-19,thrombosisiscommonlymicrovascularand dif-ficulttodetect using currentimaging, withclinical reports tendingtofocusonthepresenceorabsenceofmajor throm-bus.Theobservedbenefitfromenhancedanticoagulationisin keepingwiththelimitedpublishedevidenceinCOVID-19.26,32 The prothrombotic COVID-19 state is currently considered less toreflect disseminated intravascularcoagulation (DIC) causedbythesystemicgenerationofthrombin,than hyper-coagulabilityinthesettingofasevereinflammatorystate,33 andparticularlyinsituthrombusformation.34–36_Thefindings wouldalsobesupportedbypreviousevidenceonmulti-organ failureinsepsiswhereinhibitingthrombingenerationby anti-coagulationmayhavebenefitinreducingmortality.37

Therewasearlyevidencethatthepresenceofanelevated D-dimer was associated with a poor prognosis in COVID-19.3,38 _{IthasbeensuggestedthathighD-dimervaluesmay} not reliably predict the presence of a thrombus that has

beendegradedbyfibrinolysis,butinsteadrepresentamarker ofpooroverall outcomeinCOVID-19.39_{However,datafrom} thecurrentstudytargetingD-dimer-identifiedsubgroups,and fromtwoothergroups,26,32_{indicatethatenhanced} anticoag-ulationcanimprovesurvival,inadvanceoffullmechanistic understanding.

Atthetimeofrevisedmanuscriptsubmission,thereremain nodataonwheretheriskofthromboembolismstartstoriseon thescaleofelevatedD-dimer.Interpretationofthestepwise accrualofsurvivalbenefitwithheparintreatmentasD-dimer exceeded the upper limit of normal26 _{varies, as} demon-strated bythesixdifferent classificationsforhigh D-dimer usedasinclusioncriteriaforCOVID-19anticoagulation stud-ies listed on clinicaltrials.gov. Combining elevated D-dimer with other biomarkershasbeen proposed for thromboem-bolicriskstratificationsinCOVID-19patients.40_Asproposed in,40_{wealsoexaminedfibrinogeninitially,andunexpectedly} found that CRP(whichwas not includedin many Chinese reports),3,4 _{was morediscriminatory. CRPand D-dimernot} onlyidentifiedhigherriskcategoriesforenhancedtreatments, but alsoalowrisksubgroup(allD-dimer≤1000ng/mL,and all CRP ≤200mg/L). The abilityto identify subgroups with better prognosis is generally helpful in establishingwhere risksofintensivetherapiesmaybemoredifficulttojustify. Sincetherapeuticanticoagulationcarrieshemorrhagic con-cerns,particularlyfordiseaseswithabnormalvasculatureas in COVID-19,11–14,18 _{we speculatethat restricting enhanced} thromboembolic prophylaxis topatients insubgroupswith higherthromboticrisksmayhavecontributedtotheoverall favorablesurvivalfiguresinthetrafficlights-managedcohort. Inconclusion,earlyrecognitionofthromboembolismrisk basedonCRPandorD-dimer,andimpendingcytokinestorm predominantlybasedonserumferritin,enabledhospital sur-vival for completed encounters of 77.0% (95% confidence intervals, 71.1–82.8%) when more than 25% of admitted patients were at least 81 years old. Many COVID-19 at-risk groupsbenefitted including olderpatients, males, and patients with concurrent hypertension and diabetes. The blood measurements are already part of most hospitals’ COVID-19 datasets, and interventions for thromboembolic disease,andfluidmanagementarewidelyavailableand inex-pensive. Thus broadapplication seems feasible,and wider implementationisencouraged.

Data

sharing

statement

The code and guidance isfreely availableon requestfrom Alex.McCarthy@chelwest.nhs.uk, and is to be published to GitHubinduecourse.Fullyanonymiseddatawillbeavailable postpeerreviewpublicationonreasonablerequest,in accor-dance withinstitutionalprotocols above. Adatadictionary definingeachfieldinthedatasetwillalsobemadeavailable toothers.

Funding

Thisresearchdidnotreceiveanyspecificgrantfromfunding agenciesinthepublic,commercial,ornot-for-profitsectors.

Supportforclinical care wasprovided byChelsea& West-minsterHospitalNHSFoundationTrust,London,UK,andthe NationalHealthService(NHS)inEngland.

Authors’

contribution

Thestudy was conceivedand overseenby MPV. Literature searcheswereperformedbyMPV,CLS,AM,AG,SP,andMH.The trafficlightbioinformaticssystemwasdesignedand imple-mentedbyMPV,AM,AH,and AB.Theclinical studydesign wasbyMPV.HaematologycarepathwaydesignwasbyMPV, AGandSP.Dataanalysisforreal-timefeedbackwasperformed byMPVandAM.Realtimeanalyticdataapplicationwas per-formedbyMPV,AG,PS,AB,SC,PD,PL,SP,TP,MH,RTK,IB,and GD.Data analysiswasdevisedand performedbyCLS.Data interpretationwasperformedbyMPVandCLS.Figureswere generatedbyCLS.ThemanuscriptwaswrittenbyCLS,and revisedbyMPV,AM,SP,PLS,SC,andMH.Allauthorsreviewed andapprovedthefinalmanuscript..

Conflicts

of

interest

Theauthorsdeclarenoconflictsofinterest.

Acknowledgments

TheauthorswishtothanktheChelsea&WestminsterNHS FoundationTrustpersonnelandespeciallytheAdult Inten-siveCareUnitnurses,forthedeliveryofpersonalisedcareto allpatientsadmittedwithCOVID-19infection.Theauthors wouldalsowishtothanktheCWPluscharityforinvaluable supportthroughouttheCOVID-19outbreak.

Appendix

A.

Supplementary

data

Supplementary data associated with this article can be found, in the online version, at https://doi.org/10. 1016/j.bjid.2020.07.010.

r

e

f

e

r

e

n

c

e

s

1. ZhuN,ZhangD,WangW,etal.Anovelcoronavirusfrom patientswithpneumoniainChina,2019.NEnglJMed. 2020;382:727–33.Epub2020Jan24.

2. COVID-19DashboardbytheCenterforSystemsScienceand Engineering(CSSE)atJohnsHopkinsUniversity.

https://coronavirus.jhu.edu/map.html[Accessed19July2020]. 3. ZhouF,YuT,Du,etal.Clinicalcourseandriskfactorsfor

mortalityofadultinpatientswithCOVID-19inWuhan,China: aretrospectivecohortstudy.Lancet.2020;395:1054–110.Epub 2020Mar11.

4. HuangC,WangY,LiX,etal.Clinicalfeaturesofpatients infectedwith2019novelcoronavirusinWuhan,China. Lancet.2020;395:497–506.Epub2020Jan24.

5. XuZ,ShiL,WangY,etal.PathologicalfindingsofCOVID-19 associatedwithacuterespiratorydistresssyndrome.Lancet RespirMed.2020;8(4):420–2,

http://dx.doi.org/10.1016/S2213-2600(20)30076-X.Apr.

6.WuD,YangXO.TH17responsesincytokinestormof COVID-19:anemergingtargetofJAK2inhibitorFedratinib.J MicrobiolImmunolInfect.2020;53:368–70.Epub2020Mar11. PMID:32205092;PMCID:PMC7156211.

7.ZhangY,YuL,TangL,etal.Apromisinganti-cytokine-storm targetedtherapyforCOVID-19:theartificial-liver

blood-purificationsystem.Engineering(Beijing).2020, http://dx.doi.org/10.1016/j.eng.2020.03.006[Epubaheadof print].PMID:32292628;PMCID:PMC7118608.

8.MehtaP,McAuleyDF,BrownM,SanchezE,TattersallRS, MansonJJ.COVID-19:considercytokinestormsyndromes andimmunosuppression.Lancet.2020;395:1033–4.Epub2020 Mar16.PMID:32192578.

9.ChenC,ZhangXR,JuZY,HeWF.Advancesintheresearchof cytokinestormmechanisminducedbyCoronaVirusDisease 2019andthecorrespondingimmunotherapies.Zhonghua ShaoShangZaZhi.2020;36:E005[inChinese].Epub2020Mar 2.PMID:32114747.

10.WorldHealthOrganisation.R&DBlueprintandCOVID-19. https://www.who.int/teams/blueprint/covid-19[Accessed19J 2020].

11.FoxSE,AkmatbekovA,HarbertJL,LiG,BrownJQ,Vander HeideRS.PulmonaryandcardiacpathologyinCovid-19:the firstautopsyseriesfromNewOrleans.LancetRespirMed. 2020;8(7):681–6,

http://dx.doi.org/10.1016/S2213-2600(20)30243-5.EpubPMID: 32473124.JulMay272020.

12.MagroC,MulveyJJ,BerlinD,etal.Complementassociated microvascularinjuryandthrombosisinthepathogenesisof severeCOVID-19infection:areportoffivecases.TranslRes. 2020;220:1–13,http://dx.doi.org/10.1016/j.trsl.2020.04.007. Epub2020Apr15.Jun.

13.WichmannD,SperhakeJP,LütgehetmannM,etal.Autopsy findingsandvenousthromboembolisminpatientswith COVID-19:aprospectivecohortstudy.AnnInternMed. 2020;173(4):268–77,http://dx.doi.org/10.7326/M20-2003.Epub 2020May6.PMID:32374815.Aug.

14.AckermannM,VerledenSE,KuehnelM,etal.Pulmonary vascularendothelialitis,thrombosis,andangiogenesisin Covid-19.NEnglJMed.2020;383(2):120–8,

http://dx.doi.org/10.1056/NEJMoa2015432.Epub2020May21. Jul9.

15.KlokFA,KruipMJHA,vanderMeerNJM,etal.Incidenceof thromboticcomplicationsincriticallyillICUpatientswith COVID-19.ThrombRes.2020;191(2):145–7,

http://dx.doi.org/10.1016/j.thromres.2020.04.013.Epub2020 Apr10.Jul,pii:S0049-3848(20)30120-1.

16.M.P.Vizcaychipi,C.L.Shovlin,M.Hayes,etal.Earlydetection ofsevereCOVID-19diseasepatternsdefinenearreal-time personalisedcare,bioseverityinmales,anddecelerating mortalityrates.Publishedonline2020May11;

https://medrxiv.org/cgi/content/short/2020.05.08.20088393v1. 17.PatelBV,ArachchillageDJ,RidgeCA,etal.Pulmonary

angiopathyinsevereCOVID-19:physiologic,imagingand hematologicobservations.AmJRespirCritCareMed.2020, http://dx.doi.org/10.1164/rccm.202004-1412OC[Epubaheadof print].PMID:32667207.

18.PetrilliCM,JonesSA,YangJ,etal.Factorsassociatedwith hospitaladmissionandcriticalillnessamong5279people withcoronavirusdisease2019inNewYorkCity:prospective cohortstudy.BMJ.2020;369:1966,

http://dx.doi.org/10.1136/bmj.m1966,369:m1966.Published 2020May22.

19.GuanWJ,LiangWH,ZhaoY,etal.Comorbidityanditsimpact on1590patientswithCOVID-19inChina:anationwide analysis.EurRespirJ.2020;55(5):2000547,

http://dx.doi.org/10.1183/13993003.00547-2020,3May14, 2000547,Print2020May.PMID:32217650.

20.AlbercaRW,OliveiraLM,BrancoACCC,PereiraNZ,SatoMN. ObesityasariskfactorforCOVID-19:anoverview.CritRev FoodSciNutr.2020:1–15[Onlineaheadofprint].PMID: 32539446.

21.CostaFF,RosárioWR,RibeiroFariasAC,deSouzaRG,Duarte GondimRS,BarrosoWA.MetabolicsyndromeandCOVID-19: anupdateontheassociatedcomorbiditiesandproposed therapies.DiabetesMetabSyndr.2020;14:809–14. 22.CentersforDiseaseControlandPrevention.Coronavirus

Disease2019(COVID-19).PeopleWhoNeedtoTakeExtra Precautions.

https://www.cdc.gov/coronavirus/2019-ncov/need-extra-pre cautions/asthma.html[Accessed28May2020].

23.LiangWH,GuanWJ,LiCC,etal.Clinicalcharacteristicsand outcomesofhospitalisedpatientswithCOVID-19treatedin Hubei(epicenter)andoutsideHubei(non-epicenter):a NationwideAnalysisofChina.EurRespirJ.2020;55(6):2000562, http://dx.doi.org/10.1183/13993003.00562-2020,2000562. Published2020Jun4.

24.RichardsonS,HirschJS,NarasimhanM,etal.Presenting characteristics,comorbidities,andoutcomesamong5700 patientshospitalizedwithCOVID-19intheNewYorkCity area.JAMA.2020;323(20):2052–9.Epub2020Apr22.

25.DohertyAB,HarrisonEM,GreenCA,etal.Featuresof20133 UKpatientsinhospitalwithcovid-19usingtheISARICWHO ClinicalCharacterisationProtocol:prospectiveobservational cohortstudy.BMJ.2020;369:1985.

26.TangN,BaiH,ChenX,GongJ,LiD,SunZ.Anticoagulant treatmentisassociatedwithdecreasedmortalityinsevere coronavirusdisease2019patientswithcoagulopathy.J ThrombHaemost.2020;18:1094–9.

27.VizcaychipiMP,ShovlinCL,McCarthyA,etal.Development andimplementationofaCOVID-19nearrealtimetrafficlight systeminanacutehospitalsetting.EmergMedJ.2020.In Press.

28.ElTaguriA,NasefA.Theworldiswaiting,usesequential analysisandgetustheevidence-basedtreatmentweneed forCOVID-19.LibyanJMed.2020;15:1770518.

29.ShovlinCL,VizcaychipiMP.ImplicationsforCOVID-19triage fromtheICNARCreportof2,204COVID-19casesmanagedin UKadultintensivecareunits.EmergMedJ.2020,

http://dx.doi.org/10.1136/emermed-2020-209791.Online aheadofprint.May4:emermed-2020-209791.

30.IntensiveCareNationalAuditandResearchCentre(ICNARC). ICNARCreportonCOVID-19incriticalcare.17July2020 availableat

https://www.icnarc.org/Our-Audit/Audits/Cmp/Reports. 31.HabibAMG,AliMAE,ZouaouiBR,TahaMAH,MohammedBS,

SaquibN.Clinicaloutcomesamonghospitalpatientswith MiddleEastrespiratorysyndromecoronavirus(MERS-CoV) infection.BMCInfectDis.2019;19:870.

32.ParanjpeI,FusterV,LalaA,etal.Associationoftreatment doseanticoagulationwithin-hospitalsurvivalamong hospitalizedpatientswithCOVID-19.JAmCollCardiol.2020, http://dx.doi.org/10.1016/j.jacc.2020.05.001.

S0735-1097(20)35218-9.[Epubaheadofprint][Accessed07 June2020].

33.PanigadaM,BottinoN,TagliabueP,etal.Hypercoagulability ofCOVID-19patientsinintensivecareunit.Areportof thromboelastographyfindingsandotherparametersof hemostasis.JThrombHaemost.2020,

http://dx.doi.org/10.1111/jth.14850.

34.MariettaM,AgenoW,ArtoniA,etal.COVID-19and haemostasis:apositionpaperfromItalianSocietyon ThrombosisandHaemostasis(SISET).BloodTransfus.2020, http://dx.doi.org/10.2450/2020.0083-20[Epubaheadofprint]. 35.CattaneoM,MoriciN.Isthromboprophylaxiswithhigh-dose

enoparinreallynecessaryforCOVID-19patients?Anew “prudent”randomizedclinicaltrial.BloodTransfus.2020, http://dx.doi.org/10.2450/2020.0109-20[Epubaheadofprint]. 36.ShovlinCL,VizcaychipiMP.Vascularinflammationand

endothelialinjuryinSARS-CoV-2infection:Theoverlooked regulatorycascadesimplicatedbytheACE2genecluster.QJM. 2020,http://dx.doi.org/10.1093/qjmed/hcaa241.Onlineahead ofprint.Aug10:hcaa241.

37.IbaT,LevyJH,WarkentinTE,etal.Diagnosisand managementofsepsis-inducedcoagulopathyand

disseminatedintravascularcoagulation.JThrombHaemost. 2019;17:1989–94.

38.TangN,LiD,WangX,SunZ.Abnormalcoagulation parametersareassociatedwithpoorprognosisinpatients withnovelcoronaviruspneumonia.JThrombHaemost. 2020;18:844–7[Epub13March2020].PMID:32073213;PMCID: PMC7166509.

39.BikdeliB,MadhavanMV,JimenezD,etal.COVID-19and thromboticorthromboembolicdisease:implicationsfor prevention,antithrombotictherapy,andfollow-up.JAmColl Cardiol.2020;75(23):2950–73,

http://dx.doi.org/10.1016/j.jacc.2020.04.031.Jun16,Epub2020 Apr17.

40.ThachilJ,TangN,GandoS,etal.ISTHinterimguidanceon recognitionandmanagementofcoagulopathyinCOVID-19.J ThrombHaem.2020;18(5):1023–6,