Assessing Diagnostic Assessing Diagnostic Articles: Likelihood Ratios Articles: Likelihood Ratios Evidence Based Medicine: Evidence Based Medicine: Pedi 101, Part 1 Pedi 101, Part 1 Section of Emergency Medicine Section of Emergency Medicine Baylor College of Medicine Baylor College of Medicine Houston, Texas Houston, Texas
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
ValidityValidity Will the study reflect the truth?Will the study reflect the truth?
Rapid review: Do they give a search strategy andRapid review: Do they give a search strategy andassess quality?assess quality?
ImportanceImportance Are the effects big enough to matter?Are the effects big enough to matter?
Rapid review: How precise is the answer?Rapid review: How precise is the answer?
ApplicabilityApplicability Is it useful here and now?Is it useful here and now?
Rapid review: Are the costs similar? Is the benefitRapid review: Are the costs similar? Is the benefitworth the cost?worth the cost?
Defining riskDefining risk
PrePre--test and posttest and post--test probabilitiestest probabilities
Likelihood ratiosLikelihood ratios
Sensitivity and specificity of a testSensitivity and specificity of a test
PrePre--test probabilitytest probability
Defined as the prevalenceDefined as the prevalence
The probability of a disease beforeThe probability of a disease beforeany testing resultsany testing results
The assumptions we function withThe assumptions we function with
CaseCase
7 year old with7 year old withsore throat,sore throat,tonsillartonsillarhypertrophy,hypertrophy,cervicalcervicallymphadenopathylymphadenopathy,,and absence ofand absence ofcoryzacoryza..
Google Images, 2005. “pharyngitis”
Given minimal but appreciable risksGiven minimal but appreciable risksfrom therapy with Penicillin,from therapy with Penicillin,
would you treat ifwould you treat if……
the probability ofthe probability ofdisease were 70%?disease were 70%?
the probability ofthe probability ofdisease were 98%?disease were 98%?
A diagnostic test (rapidA diagnostic test (rapidstrep), can change yourstrep), can change yourprobability of disease.probability of disease.
This is a postThis is a post--test probability.test probability.
PostPost--test probabilitytest probability
Probability of having a condition after theProbability of having a condition after thetest result is availabletest result is available
The reason a test is obtained is toThe reason a test is obtained is toincrease or decrease the probability ofincrease or decrease the probability ofdisease that existed prior to the testdisease that existed prior to the test
CaseCase
In the example, theIn the example, theprepre--test probabilitytest probabilitywas 70%, letwas 70%, let’’ssassume the postassume the post--test probability withtest probability witha positive rapida positive rapidstrep is now 98%.strep is now 98%.Now would youNow would youtreat?treat?
Google Images, 2005. “pharyngitis”
In essence, we are interestedIn essence, we are interestedin how a diagnostic test altersin how a diagnostic test alters
the likelihood of a disease.the likelihood of a disease.
what does it mean to the clinician?what does it mean to the clinician?
DiseasePresent
DiseaseAbsent
TestPositive
TruePositive
FalsePositive
TestNegative
FalseNegative
TrueNegative
Sensitivity:Sensitivity:
The proportion ofThe proportion ofpatients with thepatients with thedisorder under studydisorder under studywho have a positivewho have a positiveresult.result.
The probability thatThe probability thatgiven the disease isgiven the disease ispresent, an abnormalpresent, an abnormaltest result indicatestest result indicatespresence of disease.presence of disease.
SensitivitySensitivity
=TP/(TP+FN)=TP/(TP+FN)
DiseasePresent
DiseaseAbsent
TestPositive
TruePositive
FalsePositive
TestNegative
FalseNegative
TrueNegative
Specificity:Specificity:
The proportion ofThe proportion ofpatients who do notpatients who do nothave the target disorderhave the target disorderand whose test resultsand whose test resultsare negative.are negative.
The conditionalThe conditionalprobability that given theprobability that given thedisease is absent, adisease is absent, anormal test resultnormal test resultexcludes the disease.excludes the disease.
SpecificitySpecificity
=TN/(FP+TN)=TN/(FP+TN)
Determining postDetermining post--test probabilitytest probability
CalculateCalculate
Or do it theOr do it thesimple waysimple waywith awith anomogramnomogram!!
Pre-TestProbability
Multiply by LR
Post-TestProbability
Convert to Post- Test Odds
Convert to Pretest Odds
Adapted from Comprehensive Case Studies: ACEP, EBM Listserve 2003
Relationship between odds andRelationship between odds andprobabilityprobability
To use LR: convert between odds and probabilityTo use LR: convert between odds and probability
yProbabilit1
yProbabilitOdds
1Odds
OddsyProbabilit
What if we donWhat if we don’’t know thet know thelikelihood ratio?likelihood ratio?
Likelihood ratio positiveLikelihood ratio positive
The LR for a positive test result is theratio of true positive results to falsepositive results.
A helpful test will have a large LR positive.
yspecificit1ysensitivite)LR(positiv
Likelihood ratio negativeLikelihood ratio negative
The LR for a negative test result is theratio of false negative results to truenegative results.
Adapted from Comprehensive Case Studies: ACEP,EBM Listserve 2003
TheThe nomogramnomogram!!
For our child withFor our child withpharyngitispharyngitis: assume: assumethe rapid strep used inthe rapid strep used inthe unit carries a LR+the unit carries a LR+of 25.of 25.
For our rapid strep (QFor our rapid strep (Q--test)test)
Given a sensitivity of 94.2% and a specificityGiven a sensitivity of 94.2% and a specificityof 96.3%:of 96.3%:
yspecificit1ysensitivite)LR(positiv
25.9631
.942e)LR(positiv
TheThe nomogramnomogram!!
For our child withFor our child withpharyngitispharyngitis: the rapid: the rapidstrep used in the unitstrep used in the unitreally does carry areally does carry aLR+ of 25.LR+ of 25.
SummarySummary
Related EBM to practiceRelated EBM to practice
PICO questionsPICO questions
QueriesQueries
Diagnosis studiesDiagnosis studies
Pre and post test probabilitiesPre and post test probabilities
Likelihood ratiosLikelihood ratios
Optional case exerciseOptional case exercise
2 1/2 year old girl is brought in by mother2 1/2 year old girl is brought in by mother
Fever to 103Fever to 103°°FF
You remember a SBI lecture and wonderYou remember a SBI lecture and wonderabout the etiology of fever as it relates toabout the etiology of fever as it relates toUTIUTI
PICO: In children with fever without aPICO: In children with fever without afocus, what is the prevalence of UTI?focus, what is the prevalence of UTI?
BayesBayes' theorem: 0.1 x 0.61 = 0.06,' theorem: 0.1 x 0.61 = 0.06,
Post test odds = 0.06Post test odds = 0.06
If test is negative, still 6% chance of UTI.If test is negative, still 6% chance of UTI.
Adapted from Comprehensive Case Studies: ACEP, EBM Listserve 2003
SummarySummary
Related EBM to practiceRelated EBM to practice
PICO questionsPICO questions
QueriesQueries
Diagnosis studiesDiagnosis studies Pre and post test probabilitiesPre and post test probabilities
Likelihood ratio: how many times more likelyLikelihood ratio: how many times more likelyis the test to be positive when the disease isis the test to be positive when the disease isthere than when the test is positive but thethere than when the test is positive but thedisease is absentdisease is absent
Questions?Questions?
CreditsCredits
Society for Academic Emergency Medicine:Society for Academic Emergency Medicine:Evidence Based Medicine Course 2005Evidence Based Medicine Course 2005
JAMA. UserJAMA. User’’s Guides to the Medical Literatures Guides to the Medical Literature20022002
BarrattBarratt A, et al. Risk Reduction. 2005. EBMA, et al. Risk Reduction. 2005. EBMTeaching Tips Working GroupTeaching Tips Working Group
Moyer V et al. Evidence Based Pediatrics andMoyer V et al. Evidence Based Pediatrics andChild Health 2000 BMJ BooksChild Health 2000 BMJ Books
Katz D. Clinical epidemiology and evidenceKatz D. Clinical epidemiology and evidencebased medicine 2001 Sage Publicationsbased medicine 2001 Sage Publications