Medical Decision Making, Second Edition

358 Pages·2013·2.613 MB·English

by  Harold C. Sox, Michael C. Higgins, Douglas K. Owens(auth.)

Medical Decision
Making
Medical Decision
Making
SecondEdition
Harold C. Sox,
M.D.
GeiselSchoolofMedicineatDartmouth,Hanover,NewHampshire
Michael C. Higgins,
Ph.D.
StanfordUniversity,Stanford,California
Douglas K. Owens,
M.D., M.S.
DepartmentofVeteransAffairsPaloAltoHealthCareSystem,PaloAlto,California
StanfordUniversity,Stanford,California
A John Wiley & Sons, Ltd., Publication
Thiseditionfirstpublished2013,©2013byJohnWiley&Sons,Ltd.
Firsteditionpublished1988©1988ButterworthPublishers,Stoneham,MA,USA
Wiley-BlackwellisanimprintofJohnWiley&Sons,formedbythemergerofWiley’sglobalScientific,
TechnicalandMedicalbusinesswithBlackwellPublishing.
Registeredoffice:JohnWiley&Sons,Ltd,TheAtrium,SouthernGate,Chichester,
WestSussex,PO198SQ,UK
Editorialoffices:9600GarsingtonRoad,Oxford,OX42DQ,UK
TheAtrium,SouthernGate,Chichester,WestSussex,PO198SQ,UK
111RiverStreet,Hoboken,NJ07030-5774,USA
Fordetailsofourglobaleditorialoffices,forcustomerservicesandforinformationabouthowtoapply
forpermissiontoreusethecopyrightmaterialinthisbookpleaseseeourwebsiteat
www.wiley.com/wiley-blackwell.
Therightoftheauthorstobeidentifiedastheauthorsofthisworkhasbeenassertedinaccordancewith
theUKCopyright,DesignsandPatentsAct1988.
Allrightsreserved.Nopartofthispublicationmaybereproduced,storedinaretrievalsystem,or
transmitted,inanyformorbyanymeans,electronic,mechanical,photocopying,recordingorotherwise,
exceptaspermittedbytheUKCopyright,DesignsandPatentsAct1988,withoutthepriorpermissionof
thepublisher.
Designationsusedbycompaniestodistinguishtheirproductsareoftenclaimedastrademarks.Allbrand
namesandproductnamesusedinthisbookaretradenames,servicemarks,trademarksorregistered
trademarksoftheirrespectiveowners.Thepublisherisnotassociatedwithanyproductorvendor
mentionedinthisbook.Thispublicationisdesignedtoprovideaccurateandauthoritativeinformation
inregardtothesubjectmattercovered.Itissoldontheunderstandingthatthepublisherisnotengaged
inrenderingprofessionalservices.Ifprofessionaladviceorotherexpertassistanceisrequired,the
servicesofacompetentprofessionalshouldbesought.
Thecontentsofthisworkareintendedtofurthergeneralscientificresearch,understanding,and
discussiononlyandarenotintendedandshouldnotberelieduponasrecommendingorpromotinga
specificmethod,diagnosis,ortreatmentbyhealthsciencepractitionersforanyparticularpatient.The
publisherandtheauthormakenorepresentationsorwarrantieswithrespecttotheaccuracyor
completenessofthecontentsofthisworkandspecificallydisclaimallwarranties,includingwithout
limitationanyimpliedwarrantiesoffitnessforaparticularpurpose.Inviewofongoingresearch,
equipmentmodifications,changesingovernmentalregulations,andtheconstantflowofinformation
relatingtotheuseofmedicines,equipment,anddevices,thereaderisurgedtoreviewandevaluatethe
informationprovidedinthepackageinsertorinstructionsforeachmedicine,equipment,ordevicefor,
amongotherthings,anychangesintheinstructionsorindicationofusageandforaddedwarningsand
precautions.Readersshouldconsultwithaspecialistwhereappropriate.Thefactthatanorganizationor
Websiteisreferredtointhisworkasacitationand/orapotentialsourceoffurtherinformationdoesnot
meanthattheauthororthepublisherendorsestheinformationtheorganizationorWebsitemayprovide
orrecommendationsitmaymake.Further,readersshouldbeawarethatInternetWebsiteslistedinthis
workmayhavechangedordisappearedbetweenwhenthisworkwaswrittenandwhenitisread.No
warrantymaybecreatedorextendedbyanypromotionalstatementsforthiswork.Neitherthepublisher
northeauthorshallbeliableforanydamagesarisingherefrom.
LibraryofCongressCataloging-in-PublicationData
Sox,HaroldC.
Medicaldecisionmaking/HaroldC.Sox,MichaelC.Higgins,DouglasK.Owens.--2nded.
p.;cm.
Rev.ed.of:Medicaldecisionmaking/HaroldC.Sox,Jr....[etal.].c1998.
Includesbibliographicalreferencesandindex.
Summary:‘‘Thistitleprovidesathoroughunderstandingofthekeydecisionmakinginfrastructureof
clinicalpracticeandretainsitsemphasisonpracticeandprinciplesofindividualizedmedicaldecision
making’’–Providedbypublisher.
ISBN978-0-470-65866-6(paper)
I.Higgins,MichaelC.(MichaelClark),1950-II.Owens,DouglasK.III.Medicaldecisionmaking.IV.
Title.
[DNLM:1. Diagnosis,Differential.2. Cost-BenefitAnalysis.3. DecisionMaking.4. Probability.
WB141.5]
(cid:2)
616.075–dc23
2012047951
AcataloguerecordforthisbookisavailablefromtheBritishLibrary.
Wileyalsopublishesitsbooksinavarietyofelectronicformats.Somecontentthatappearsinprintmay
notbeavailableinelectronicbooks.
CoverdesignbyRobSawkinsforOptaDesignLtd.
Setin9.5/12ptPalatinobyLaserwordsPrivateLimited,Chennai,India
12013
ToJeanandKathleen
ToSara,Rachel,andChristopher
ToCarol,Colin,andLara
Theirsupporthasmeanteverything.
Contents
Foreword,xi
Preface,xv
1 Introduction,1
1.1 HowmayIbethoroughyetefficientwhenconsideringthepossible
causesofmypatient’sproblems?1
1.2 How do I characterize the information I have gathered during the
medicalinterviewandphysicalexamination?2
1.3 HowdoIinterpretnewdiagnosticinformation?5
1.4 HowdoIselecttheappropriatediagnostictest?5
1.5 HowdoIchooseamongseveralriskytreatmentalternatives?6
1.6 Summary,6
2 Differentialdiagnosis,7
2.1 Introduction,7
2.2 Howcliniciansmakeadiagnosis,8
2.3 Theprinciplesofhypothesis-drivendifferentialdiagnosis,11
2.4 Anextendedexample,21
Bibliography,26
3 Probability:quantifyinguncertainty,27
3.1 Uncertaintyandprobabilityinmedicine,27
3.2 Usingpersonalexperiencetoestimateprobability,34
3.3 Usingpublishedexperiencetoestimateprobability,46
3.4 Taking the special characteristics of the patient into account when
estimatingprobability,57
Problems,58
Bibliography,59
4 Understandingnewinformation:Bayes’theorem,61
4.1 Introduction,61
4.2 Conditionalprobabilitydefined,64
4.3 Bayes’theorem,65
4.4 TheoddsratioformofBayes’theorem,69
4.5 LessonstobelearnedfromBayes’theorem,76
4.6 TheassumptionsofBayes’theorem,82
4.7 UsingBayes’theoremtointerpretasequenceoftests,84
vii
viii Contents
4.8 UsingBayes’theoremwhenmanydiseasesareunder
consideration,88
Problems,90
Bibliography,91
5 Measuringtheaccuracyofdiagnosticinformation,93
5.1 How to describe test results: abnormal and normal, positive and
negative,93
5.2 Measuringatest’scapabilitytorevealthepatient’struestate,98
5.3 Howtomeasurethecharacteristicsofadiagnostictest:ahypothetical
case,106
5.4 Pitfallsofpredictivevalue,109
5.5 Sources of biased estimates of test performance and how to avoid
them,110
5.6 Spectrumbias,116
5.7 Expressingtestresultsascontinuousvariables,125
5.8 Combiningdatafromseveralstudiesoftestperformance,134
Problems,137
Bibliography,140
6 Expectedvaluedecisionmaking,143
6.1 Anexample,145
6.2 Selectingthedecisionmaker,148
6.3 Decisiontrees:structuredrepresentationsfordecisionproblems,149
6.4 Quantifyinguncertainty,152
6.5 Probabilisticanalysisofdecisiontrees,156
6.6 Expectedvaluecalculations,158
6.7 Sensitivityanalysis,161
6.8 Foldingbackdecisiontrees,163
Problems,168
Bibliography,168
7 Markovmodelsandtime-varyingoutcomes,170
7.1 Markovmodelbasics,170
7.2 Exponentialsurvivalmodelandlifeexpectancy,189
Problems,198
Appendix:Mathematicaldetails,200
Bibliography,203
8 Measuringtheoutcomeofcare–expectedutilityanalysis,204
8.1 Basicconcept–directutilityassessment,205
8.2 Sensitivityanalysis–testingtherobustnessofutilityanalysis,210
8.3 Shortcut–usingalinearscaletoexpressstrengthofpreference,212
8.4 Exponentialutility–aparametricmodel,213
8.5 Exponentialutilitywithexponentialsurvival,218
Contents ix
8.6 Multidimensionaloutcomes–directassessment,220
8.7 Multidimensionaloutcomes–simplifications,223
8.8 Multidimensionaloutcomes–quality-adjustedlifeyears
(QALY),228
8.9 Comparison of the two models for outcomes with different length
andquality,232
Problems,235
Appendix:Mathematicaldetails,237
Bibliography,242
9 Selectionandinterpretationofdiagnostictests,243
9.1 Takingactionwhentheconsequencesareuncertain:principlesand
definitions,244
9.2 Thetreatment-thresholdprobability,247
9.3 Thedecisiontoobtainadiagnostictest,252
9.4 Choosingbetweendiagnostictests,259
9.5 Choosingthebestcombinationofdiagnostictests,261
9.6 Settingthetreatment-thresholdprobability,263
9.7 Takingaccountoftheutilityofexperiencingatest,275
9.8 Aclinicalcase:testselectionforsuspectedbraintumor,279
9.9 Sensitivityanalysis,281
Bibliography,287
10 Cost-effectivenessanalysisandcost–benefitanalysis,288
10.1 Theclinician’sconflictingroles:patientadvocate,memberofsociety,
andentrepreneur,288
10.2 Cost-effectiveness analysis: a method for comparing management
strategies,291
10.3 Cost–benefit analysis: a method for measuring the net benefit of
medicalservices,298
10.4 Measuringthecostsofmedicalcare,301
Problems,304
Bibliography,305
11 Medicaldecisionanalysisinpractice:advancedmethods,307
11.1 Anoverviewofadvancedmodelingtechniques,307
11.2 Useofmedicaldecision-makingconceptstoanalyzeapolicy
problem:thecost-effectivenessofscreeningforHIV,311
11.3 Useofmedicaldecision-makingconceptstoanalyzeaclinical
diagnosticproblem:strategiestodiagnosetumorsinthelung,323
11.4 Useofcomplexmodelsforindividual-patientdecisionmaking,330
Bibliography,333
Index,337
Foreword
The maturation of medical science during the last half of the twentieth
century was most impressive. Clinical trials displaced observational studies
thattypicallyconsistedofadozenorfewerpatients;thepathophysiologyand
genetics of many diseases were discovered; and diagnostic and therapeutic
methods advanced. Crude diagnostic tests such as cholecystography and
barium enemas and risky tests such as air encephalograms,needle biopsies,
and exploratory laparotomies were made obsolete by technology. Flexible
tubes, some outfitted with lights and cameras, CT, MRI, and PET scanners,
and sophisticated immuno-analyses of blood and other body fluids gained
immediateacceptance.Manytherapiesthatwereformulatedbygrindingup
organs, desiccatingthem, and hoping that they would correct a deficitwere
replacedbynew,potentchemicals.
Clinicalreasoning,theprocessesbehindbothdiagnosisandmedicaldeci-
sionmaking,includingthecomplextradeoffsbetweentherisksandbenefits
of tests and treatments, lagged behind advances in medical science. In the
run-uptothelastquarterofthecentury,studentslearnedhowtoreasonabout
patientproblemsbyobservingexpertcliniciansatwork,and(iftheydared)
byaskingthemwhytheyorderedthistestorthat,whytheygaveonedrugor
another.Becausethisapprenticeshipapproachwasnotcodified,objectified,or
quantified,medicaltextsstruggledtoexplainclinicalreasoning,andstudents
struggledtolearnit.Andwhentheevidenceofconfusionabouttheuseoftests
andtreatmentsfirstemerged,alarmbellsclanged.Researchershaddiscovered
extremevariationsintheuseoftestsandtreatmentsfromonecommunityto
anotherandinregionsacrossthecountrywithoutacorrespondingbenefitfor
patients.Irrationaltestingandtreatinghadbeguntocontributesubstantially
toanimpossibleescalationinthecostofcare.
Duringthelastthreedecadesofthetwentiethcentury,clinician–scientists
begantoexaminetheprocessesofdiagnosisanddecisionmakingwithtools
from other disciplines, including cognitive science, decision science, proba-
bility, and utility theory. From these diverse sources the clinical science of
medical decision making was hatched. Elements of the diagnostic process
were identified and a language for explaining and teaching diagnosis was
formulated. Cognitive errors in diagnosis were sought and methods devel-
opedtoavoidthem.Thecriticalimportanceofaprobabilisticrepresentation
of diagnosis, in terms of prior probabilities, conditional probabilities, and
likelihoodratios,wasrecognizedandputtouseintheformofacenturies-old
formulation of Bayes’ Rule. Decision analysis, a discipline formerly used by
themilitary,wasappliedfirsttoindividualclinicalproblems,latertoclassesof
xi
xii Foreword
problems,andeventuallytoissuesofcostandefficacyoftestsandtreatments.
Beforetheendofthetwentiethcentury,ascienceofmedicaldecisionmaking
andalanguageforteachingithadbeenborn.
Implementingthenewscience,however,provedmoredifficultthandevel-
opingit.Skepticsaverredthatphysicians’estimatesofprobabilitieswereoften
flawed,thatapplyingBayes’Rulewasnoteasy,andthatdecisiontreeswere
eithertoosimple(andthusdidnotrepresentaclinicalproblemsufficiently)
ortoocomplex(andthuscouldnotbeunderstood).Manywonderedwhether
medicinecouldbeconvincedtoadoptthesenewapproachesandwhetherthe
averagephysiciancouldbeexpectedtousethemintheirday-to-daypractices.
As the field has evolved, some of these questions have been answered.
Now, in the second decade of the twenty-first century, Bayes’ Rule is used
to design clinical trials, to develop decision rules that help physicians judge
whethertoadmitpatientssuspectedofhavinganacutemyocardialinfarction,
and to develop compiled strategies for diagnosing and treating pulmonary
emboli, to name a few applications. Decision analysis has been used to
formulate answers for individual patients’ dilemmas, but this use is time
consuming, expensive, and requires special expertise. Nonetheless, decision
analysishasfoundextensiveapplicationinclinicalpracticeguidelinedevelop-
ment,cost-effectiveanalyses,andcomparativeeffectivenessstudies.Acadre
of physicians has become sufficiently skilled in the methods to apply them
in active clinical teaching environments and to integrate them into medical
studentandresidencycurricula.
It is legitimate to ask why a student or resident should spend the intel-
lectual capital to learn these methods. The answer is compelling. First, they
help in learning and teaching the process of diagnosis. Second, the princi-
ples of screening and diagnostic and management decision making become
transparentfromanunderstandingofBayes’Rule,diagnosticandtherapeutic
thresholds,decisionaltoss-ups,anddecisionanalysis.Subjectingsuchissues
torationalexaminationimprovesdecisionmakingand,consequently,patient
care.Moreover,becausethesemethodsarethebasisforsomanyanalysesof
health practices, appreciation of their limitations provides a healthy skepti-
cism of their applications. Lastly, the approaches are powerful tools to pass
on the concepts to others, as well as critical templates to understand honest
differencesofopiniononcontroversialmedicalpractices.
Forthepast25years,MedicalDecisionMakinghasbeenanidealvenuefor
developing a rich comprehension of these methods and for understanding
howtoapproachdiagnosis;thenewsecondeditionisevenbetter.Itschapter
onBayes’Rule,forexample,isexemplary,explainingthemethodinmultiple
different formats. The chapters on selection of diagnostic tests and decision
analysis are meticulously crafted so as to leave little uncertainty about the
methods.Anewchapteronmodelingmethodsisrichlyillustratedbyactual
analyses;thechaptersonexpectedvaluedecisionmaking,utilityassessment,
andMarkovmodelinghavebeenextensivelyrevised.
Foreword xiii
In short, this book has been a standard of the field, and the new edition
will continue its dominance. There is little doubt that in the future many
clinical analyses will be based on the methods described in Medical Decision
Making,andthebookprovidesabasisforacriticalappraisalofsuchpolicies.
Teachersofmedicaldecisionmakingwillrequireit;medicalstudentswilldig
intoitrepeatedlyastheylearnclinicalmedicine;residentswillgobacktoit
again and again to refresh their diagnostic and therapeutic skills. And from
itslucidpages,practicingphysicianswillattainaricherunderstandingofthe
principlesunderlyingtheirwork.
JeromeP.Kassirer,M.D.
DistinguishedProfessor,TuftsUniversitySchoolofMedicine,US
VisitingProfessor,StanfordMedicalSchool,US