1 août 2017 · Results: A MANOVA found significant differences (p < 0 001) between residency specialties and both USMLE Step 1 and Step 2 CK scores,
18 août 2022 · Understanding Your Score Reported scores for Step 1, Step 2 CK, and Step 3 range from 1 to 300 Small differences in
4 nov 2019 · The United States Medical Licensing Examination (USMLE) is a 3-step examination required for medical licensure in the United States The first
13 mai 2021 · Differences between Step 2 CK and Step 1 scores were stratified by Step 1 score Results A total of 1,688 applicants applied to our
20 sept 2021 · USMLE Step 1 score, NBME clinical subject exam scores, and USMLE Step 2 CK scores Pearson correlations were run between the performance
An applicant is registered to take Step 2 CK in the United States and requests to take the exam in Europe She is required to pay $300: the $90 region change
The United States Medical Licensing Examination® (USMLE®) is a three-step analogous, though not identical, to those of the USMLE Step 1 and Step 2
1 juil 2020 · women on Step 1, yet the trend is reversed for Step 2 and negligible for Step 3 Some racial differences have also been seen,
The USMLE Step 1 and Step 2 CK exams are the first two exams in a three-step Although everyone's schedule will look different, below is an example of
*jesse.rafel@nyulangone.orgAbstractIntroductionTheUnitedStatesMedicalLicensingExamination(USMLE)Step1andStep2Clinical
Knowledge(CK)areimportantfortraineemedicalknowledgeassessmentandlicensure, medicalschoolprogramassessment,andresidencyprogramapplicantscreening.Littleis knownabouthowUSMLEperformancevariesbetweeninstitutions.Thisobservational studyattemptstoidentifyinstitutionswithabove-predictedUSMLEperformance,whichmayindicateeducationalprogramssuccessfulatpromotingstudents'medicalknowledge.MethodsSelf-reportedinstitution-leveldatawastabulatedfrompubliclyavailableUSNewsand
WorldReportandAssociationofAmericanMedicalCollegespublicationsfor131USallo- pathicmedicalschoolsfrom2012-2014.Bivariateandmultiplelinearregressionwereper- formed.TheprimaryoutcomewasinstitutionalmeanUSMLEStep1andStep2CKscores outsidea95%predictioninterval(�2standarddeviationsaboveorbelowpredicted)basedonmultipleregressionaccountingforstudents'prioracademicperformance.ResultsEighty-nineUSmedicalschools(54public,35private)reportedcompleteUSMLEscores
overthethree-yearstudyperiod,representingover39,000examinees.Institutionalmean gradepointaverage(GPA)andMedicalCollegeAdmissionTestscore(MCAT)achievedan adjustedR2of72%forStep1(standardizedMCAT0.7,GPA0.2)and41%forStep2CK (standardizedMCAT0.5,GPA0.3)inmultipleregression.Usingthisregressionmodel,5 institutionswereidentifiedwithabove-predictedinstitutionalUSMLEperformance,while3 institutionshadbelow-predictedperformance. PLOSONE|https://doi.org/10.1371/journal.pone.0224675November4,20191/10a1111111111 a1111111111 a1111111111 a1111111111 a1111111111observedperformancedifferences.IntroductionTheUnitedStatesMedicalLicensingExamination(USMLE)isa3-stepexaminationrequired
formedicallicensureintheUnitedStates.Thefirsttwoexams,USMLEStep1andStep2Clin- icalKnowledge(CK),assessmedicalstudents'masteryofbasicbiomedicalprinciplesandtheir clinicalapplications[1,2].About40,000traineestakeeachexamannually,ofwhichover35% arenon-US/Canadianmedicalstudents[3].Bothexamsarehigh-stakesparametersofmedical studentperformancecriticalforadvancement[4],residencyapplicantscreeningandselection [ 5 , 6 ],andfutureboardcertification[7].Multiplestudieshavedemonstratedcorrelations betweenindividualfactors-includingMedicalCollegeAdmissionTest(MCAT)score[8], undergraduategradepointaverage(GPA)[9],andstudybehaviors[10]-andUSMLEperfor- mance.However,littleisknownaboutinstitutionalUSMLEperformancevariation.One groupanalyzingdatafromthe1990sdemonstratedthatinstitutionalvariables,includingcur- riculardifferences,didnotpredictUSMLEperformance[11,12].Arecentstudyusingone yearofnationaldatafoundsomeevidenceofinter-institutionalUSMLEperformancediffer- ences,buttheshortstudydurationprecludesdefinitiveconclusions[13]. Inthisexploratory,institution-levelstudy,weanalyzeinstitutionalvariationinUSMLE Step1andStep2CKperformancerelativetomeanmatriculantGPAandMCAT.Ourprimary objectivewastoidentifyinstitutionswithabove-predictedUSMLEperformance,whichmayindicateeducationalprogramssuccessfulatpromotingstudents'medicalknowledge.MethodsThisobservationalstudywasconductedinaccordancewiththeSTROBEguidelinesforobser-
vationalstudiesinepidemiology[14].DatasourcesWemanuallytabulatedself-reportedinstitutionaldata-aggregatepercentagesandmeansrep-
resentingyearlymedicalstudentcohortsatsingleinstitutions-fromtheannualUSNewsand WorldReport"BestGraduateSchools"publication(2008-2016editions)[15]andtheAssocia- tionofAmericanMedicalColleges(AAMC)MedicalSchoolAdmissionRequirementspublica- tion(2008-2012editions)[16]forall131USallopathicmedicalschools.Osteopathic institutionswereexcludedfromthisstudy,asosteopathicstudentstypicallytaketheCOMLEX licensingexaminationratherthantheUSMLEandveryfewUSosteopathicinstitutions reportedUSMLEperformancedata.Asamplesizecalculationwasnotperformedbecausewe obtainedavailabledataforacensusofUSallopathicmedicalschoolsduringthestudyperiod. Nationalaveragesforallallopathicmatriculantsandexamineeswereobtainedfromofficial AAMC[17]andUSMLEsources[18,19].InstitutionalReviewBoardapprovalwasnot requiredasnohumansubjectsoridentifiabledatawereinvolved.InstitutionalUSMLEperformancedifferencesPLOSONE|https://doi.org/10.1371/journal.pone.0224675November4,20192/10Funding:Theauthorsreceivednospecificfunding
forthiswork.PrimaryoutcomemeasuresandpredictorvariablesTheprimaryoutcomemeasureswereinstitutionalmeanUSMLEStep1and2CKscores,aver-
agedoverthe3-yearstudyperiod2012-2014.Predictorvariablesincludedstudents"prioraca- demicperformance(institutionalmeanundergraduateGPAandMCAT,averagedover3 years)anddemographics(percentagenon-traditionalstudents,minoritystudents,undergradu- atebiologicalsciencesorhumanitiesmajors),andmedicalschoolfactors(acceptancerate,pub- lic/privatestatus,faculty-to-studentratio,NationalInstitutesofHealthresearchfunding, graduatesenteringprimarycare).MCATscoresrepresentedtotalscorescomputedasthesum oftheaverageinstitutionalscoresonall3sections(biologicalsciences,physicalsciences,verbal reasoning).InstitutionalUSMLEscoreswerematchedtoinstitutionalGPAandMCATaver- agesfromtwoorfouryearsprior(forStep1or2CK,respectively)toaccountforthetypicallagbetweenmatriculationandUSMLEtesting.StatisticalanalysisAllanalysiswasattheinstitutionlevel.Weperformedordinaryleastsquareslinearregression
analysis,withtestofPearson'srforbivariatecorrelations.Conditionsoflinearity,nearlynor- malresiduals,andhomoscedasticitywerechecked[20].Institutionswith3-yearaverage USMLEperformanceoutsidea95%predictioninterval(regressionresidual�2standarddevi- ations,SD,frompredicted)wereidentified[21].Hypothesistestswere2-sidedwith=.05; ANOVAwasusedtoconfirmoverallsignificanceofmultipleregressions.StatisticalanalysiswasdoneusingSPSSversion25.0(SPSSInc.,Chicago,Illinois).ResultsIntotal,89(54publicand35private)of131USallopathicmedicalschoolsreportedcomplete
USMLEscoresoverthe3-yearstudyperiod(68%reportingrate),representing39,615and.PredictorsofinstitutionalUSMLEperformanceThestrongestpredictorofinstitutionalUSMLEscoreswaspriorstudentacademicperfor-
mance,includingundergraduateGPA(Step1,Pearson'sr=.64;Step2CK,r=.53;bothP< .001)andMCATscore(Step1,r=.84;Step2CK,r=.62;bothP<.001).Numerousstudent bodydemographicandinstitutionalfactorshadmoderatelystrongcorrelationswithinstitu- tionalUSMLEscoresinbivariateregression;however,whencontrollingforGPAandMCAT, thesecorrelationswereweakandnolongersignificant(Table2).Forexample,privateinstitu- tionswerecorrelatedwithhigherUSMLEStep1scores(r=.51,P<.001),butthiscorrelation vanishedaftercontrollingforGPAandMCAT(r=.12,P=.42),asprivateinstitutionsrecruit studentswithhigherMCATscorescomparedtopublicinstitutions(mean33.5vs.30.9,differ- ence2.7,95%CI1.9-3.5;P<.001). ThefinalregressionmodelutilizingGPAandMCATachievedanadjustedR2of72%for Step1(standardizedMCAT0.7,GPA0.2,modelP<.001)and41%forStep2CKInstitutionalUSMLEperformancedifferencesPLOSONE|https://doi.org/10.1371/journal.pone.0224675November4,20193/10
(standardizedMCAT0.5,GPA0.3,modelP<.001).GPAaddedsignificantbutincremental validityevidenceoverMCATalone(Step1,R22%,P=.009;Step2CK,R24%,P=.02); accordingly,forvisualization,institutionalUSMLEwasregressedonMCATscorealone ( Fig1).Table1.AverageGPA,MCAT,andUSMLEStep1and2CKscoreamong89USallopathicmedicalschoolsandnationally.GPA,UndergraduateGradePointAverage;MCAT,MedicalCollegeAdmissionTestscore;USMLE,USMedicalLicensingExamination;CK,ClinicalKnowledge;ns,
notsignificantatP<.05threshold�Two-tailedt-testcomparingpublictoprivatehttps://doi.org/10.1371/journal.pone.0224675.t001Table2.LinearregressionbetweenvariousinstitutionalcharacteristicsandinstitutionalUSMLEperformance,withoutandwithcontrolforaverageinstitutional
GPA,UndergraduateGradePointAverage;MCAT,MedicalCollegeAdmissionTestscore;USMLE,USMedicalLicensingExamination;CK,ClinicalKnowledge;NIH,
NationalInstitutesofHealth.�P<.05��P<.01†PartialcorrelationcontrollingforGPAandMCAT(2010-12)‡PartialcorrelationcontrollingforGPAandMCAT(2008-10)
https://doi.org/10.1371/journal.pone.0224675.t002InstitutionalUSMLEperformancedifferencesPLOSONE|https://doi.org/10.1371/journal.pone.0224675November4,20194/10
InstitutionalUSMLEperformancedifferencesPLOSONE|https://doi.org/10.1371/journal.pone.0224675November4,20195/10
Institutionswithabove-orbelow-predictedUSMLEperformanceUsingtheGPAandMCATregressionmodel,weidentifiedasubsetofinstitutionswith3-year
averageinstitutionalUSMLEscoresstatisticallyaboveorbelowpredicted(Table3).DiscussionInthisexploratorystudyof89USallopathicmedicalschools,weidentified5institutionswith
above-predictedinstitutionalUSMLEperformancebasedonthedescribedmodel.Theetiol- ogyoftheseinstitutions'relativesuccess(orthe3unnamedinstitutions'below-predictedper- formance)isunclear;wecanonlysaythatnumerousdemographicandinstitutionalfactorswe assesseddidnotaccountforthisvariation.Wehypothesizethatunmeasuredstudentfactors thatvarysystematicallybetweeninstitutions(e.g.,throughadmissionsprocesses)orinstitu- tion-specificfactors(e.g.,alignmentofcurriculawithUSMLEcontent)mayexplainthese institutionaldifferences.Forexample,medicalschoolsthatprovidecommerciallyavailable Step1questionbanks[22]orwherestudentstakeStep1afterthecoreclerkships[23]have reportedimprovedinstitutionalscores,demonstratingthatuniqueinstitutionalstrategiescan promotestudents'USMLEsuccess.Furtherstudyisneededtounderstandifthe5institutions identifiedherehaveuniquefactorsthatpromotedtheirstudents'successontheseexams. Wefoundthatinstitutions'averagestudentGPAandMCATaccountedforsubstantialvari- ationininstitutionalaverageUSMLEStep1andStep2CKscores,whichwasexpectedbased onpriorstudiesattheindividual[8,9]andinstitutionallevel[9,11,12].Importantly,institu- tionaldemographicfactors(suchaspercentminoritystudentsorbiologicalsciencesmajors) werecorrelatedwithinstitutionalUSMLEperformanceinbivariateregressionbutwerenot significantaftercontrollingforGPAandMCAT.NationalInstitutesofHealthresearchfund- ing,whichhadbeenpreviouslyshowntocorrelatewithinstitutionalUSMLEperformance [ 13 ],wassimilarlynotsignificantwhencontrollingforGPAandMCAT. Institutionswithapropensityformatchingstudentsintheprimarycarespecialtiesfamily medicine,pediatrics,andinternalmedicine-whichhavelowerUSMLEscreeningthresholds forresidencyinterviewsthanothermore"competitive"specialties[24]-tendedtorecruitstu- dentswithlowerGPAandMCATscores,andthuslowerinstitutionalUSMLEscores.Aswith otherinstitutionalfactors,however,institutions'primarycarespecialtyratewasnotassociated withdifferentialUSMLEperformancebeyonditsassociationwithGPAandMCAT. Suchfindingshighlightthecriticalimportanceofcontrollingforprioracademicperfor- mancewhenattemptingtoexplainUSMLEperformancedifferences.However,wedoubtthat pre-medicalstudents-akeyconsumeroftheannualUSNewsandWorldReportdata-consider thesecovariateswheninterpretinginstitutionalUSMLEscoresandidentifyingmedicalschools ofinterest.Indeed,undergraduatesmightconclude(erroneously)thatprivatemedicalschools outperformpublicschoolsontheUSMLE,wheninfactstudentsattendingprivateschools havehighertestscoresatmatriculation.Theremaybearoleforbettercontextualizingthisdatasothatpre-medicalstudentscanbeinformedconsumers.TheNationalBoardofMedicalFig1.RegressionanalysisofinstitutionalMCATversusUSMLEperformance.(A)Regressionanalysisofinstitutional
averagematriculantMedicalCollegeAdmissionTest(MCAT)score(2010-2012)versusinstitutionalaverageUSMedical
LicensingExamination(USMLE)Step1score(2012-2014)acrossn=89USallopathicmedicalschools,representing39,615
examinees.(B)RegressionanalysisofinstitutionalaveragematriculantMCATscore(2008-2010)versusinstitutionalaverage
USMLEStep2ClinicalKnowledge(CK)score(2012-2014)acrossn=89USallopathicmedicalschools,representing39,252
examinees.Forbothplots,eachbubblerepresents3-yearaverageatoneinstitution,withbubblesizereflectingnumber
examinedateachinstitution.Ordinaryleastsquaresbestfitline(solid)and95%predictioninterval(dashedlines)areshown,
withcoloreddatapointshighlightinginstitutionsoutsidethepredictioninterval. https://doi.org/10.1371/journal.pone.0224675.g001InstitutionalUSMLEperformancedifferencesPLOSONE|https://doi.org/10.1371/journal.pone.0224675November4,20196/10
Examiners(NBME),whoproducetheUSMLE,arepositionedtomorerigorouslyexploretherelationshipbetweeninstitutionsandexamperformance.LimitationsThisstudyreliedonself-reportedinstitutionaldataviaathird-partypublication,astheNBME
doesnotpublishinstitutionalscoreperformance.Misreportingispossible,althoughwevali- datedthereportedscoresfromseveralinstitutions.USNewsandWorldReportprovidestheir methodologyfordatacollectionwitheachannualrelease[25],butdonotstatespecificsrelated tohowdataisvalidatedorstandardizedwithin-orbetween-schools.Forexample,itisunclear ifinstitutionshavediscretioninhowtheyformulatetheirinstitutionalMCATaverage,includ- inghowindividualswithmultipletestresultsarehandled,whichcanintroducebiasintothe relationshipbetweenMCATandUSMLEperformance[26]. Althoughweassessednumerousstudentandmedicalschoolfactors,somepotentially importantcovariates-suchaspercentofstudentswithadvanceddegrees,curricularstructure, timingofUSMLEexaminations,andschoolage-werenotincorporatedintothisstudybutare importantareasforfutureinvestigation.Forexample,someinstitutionshavemovedthe USMLEStep1testwindowtoaftercoreclinicalclerkships[23],withsmallbenefitsinscores andreducedfailurerates[27]. Moreover,only89of131USallopathicmedicalschools(68%)reportedcompletedata; non-reportersmaydifferinimportantways.Wefoundthatreportinginstitutions,ascom- paredtoanaverageofallstudentsnationally,hadslightlyhigheraverageGPAandMCAT scores,withanassociated1.5-pointhigheraverageUSMLEStep1score.Statisticalcompari- sonsofthesedifferencesarenotadvisablegiventhedifferentunitsofreporting(institutions vs.individuals);yettheverysmalldifferencessuggestthatthereportinginstitutionswere nationallyrepresentative.Therelativelyshort3-yearstudyperioddoesnotprecludethatthe observedinstitutionaloutliersmayrepresentrandomvariation;replicationwithlongerobser-vationisneeded.Finally,ourstudywasecological;noinferencecanbemadethatinstitution-Table3.USallopathicmedicalschoolswithabove-orbelow-predictedainstitutionalUSMLEStep1orStep2CKperformance,2012-2014.
USMLE,USMedicalLicensingExamination;CK,ClinicalKnowledge;SD,standarddeviation.aBasedonregressionmodelsincorporatinginstitutionalaverageMedicalCollegeAdmissionTest(MCAT)scoreandundergraduategradepointaverage(GPA)of
enteringstudents,asfollows:InstitutionalUSMLEStep1score=122+1.7�MCAT+14.1�GPA;InstitutionalUSMLEStep2CKscore=149+1.0�MCAT+15.6�GPA.bThenamesofinstitutionswithbelow-predictedinstitutionalUSMLEperformancewerewithheldduetothesensitiveandexploratorynatureofthisdata.
https://doi.org/10.1371/journal.pone.0224675.t003InstitutionalUSMLEperformancedifferencesPLOSONE|https://doi.org/10.1371/journal.pone.0224675November4,20197/10
levelfindingstranslatetoindividualstudents(i.e.,theecologicalfallacy),andindeedonlyinsti- tutionalaveragesandcountswerereported(withoutanymeasureofstudent-to-studentvari-ability).Nevertheless,thepurposeofthisstudywasonlytocompareinstitutions.ConclusionsWefoundthatinstitutionalaverageGPAandMCATscorescorrelatestronglywithinstitu-
tionalUSMLEperformance.Numerousstudentdemographicandinstitutionalfactorswere insignificantwhencontrollingforGPAandMCAT.Weidentifiedseveralinstitutionswithsig- nificantabove-orbelow-predictedUSMLEperformance,raisingquestionsaboutinter-institu- tionalUSMLEperformanceparity.Methodstoassessinstitutions'overallperformanceon knowledge-basedexamsmayofferaparametertoevaluatemedicalschoolsandtheircurricula, whileprovidingprospectivestudentswithvaluabledataregardingthesehigh-stakesexams. Additionalstudyisneededtoexploretheetiologyanddurabilityoftheobservedperformance differences,andtoincorporateotherstudentandinstitutionalfactorsthatmaybeimportantpredictorsofperformance.SupportinginformationS1Table.AverageGPA,MCAT,andUSMLEStep1and2CKfor89USallopathicmedical
schoolsreportingcompletedatafrom2012-2014,withassociatednationalaverages. (XLSX)AcknowledgmentsThisworkwasinspiredbyananonymouspostonMarch13,2013,intheAnastomosedblog titled:"Valueaddedintopmedicalschools?MCAT/GPAaspredictorsofUSMLEscores"(see https://anastomosed.wordpress.com/).JoelPurkiss,PhD(BaylorCollegeofMedicine,Hous- ton,Texas,USA)providedearlyguidanceonthedesignofthiswork.MartinV.Pusic,MD, MA(NewYorkUniversityLangoneHealth,NewYork,USA)andKentHecker,PhD(Univer- sityofCalgary,Alberta,Canada)providedhelpfulmethodologicadviceandmanuscript critiques.AuthorContributionsConceptualization:JesseBurk-Rafel,JosephC.Kolars.InstitutionalUSMLEperformancedifferencesPLOSONE|https://doi.org/10.1371/journal.pone.0224675November4,20198/10
References1.HaistSA,KatsufrakisPJ,DillonGF.TheevolutionoftheUnitedStatesMedicalLicensingExamination (USMLE):Enhancingassessmentofpractice-relatedcompetencies.JAMA.2013;310(21):2245-InstitutionalUSMLEperformancedifferencesPLOSONE|https://doi.org/10.1371/journal.pone.0224675November4,20199/10
InstitutionalUSMLEperformancedifferencesPLOSONE|https://doi.org/10.1371/journal.pone.0224675November4,201910/10