diagnostic tests and evidence based medicine
TRANSCRIPT
حق بنام
Diagnostic tests &
EBMPrepared by :Dr Nooria Atta
Quantitative medical predicationsDiagnostic testing
• The purpose is to reduce uncertainty about the patient’s diagnosis or prognosis
• Aid clinician in making management decision• Variety of diagnostic testing • In the last years EBM stressed the attention
not only on the evaluation of therapeutic strategies but also on the efficacy of diagnostic phase
Diagnostic tests(prognostic tests)
• At least two results (+ , -)
Diseased Un disease
Positive T.P a b F.P
Negative F.N c d T.N(correct rejection)
Sensitivity, Specificity
• Sensitivity: the proportion of diseased people with (+ve) test result ,or
• Probability of (test +/ diseased)• Specificity: the proportion of non diseased
people with (–ve) test result ,or • Probability of (test -/ non diseased)• A perfect test is supposed to have 100%
sensitivity and 100% specificity
Test A:• False Positive Rate = b/b+d (350/1800 = 0.19) • True Positive Rate = a/a+c (630/880 = 0.68)
Test B:• FPR= b/b+d (600/1800 = 0.333)
• TPR = a/a+c (750/880 = 0.85)
Test C:• FPR= b/b+d (1050/1800 = 0.58)
• TPR = a/a+c (820/880 = 0.93)
False positive rate (1- specificity )
True
pos
itive
rate
(sen
sitiv
ity)
0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.650
0.10.20.30.40.50.60.70.80.9
1
A B C
A B C
Receiver Operating Characteristic (ROC)
Positivity criterion
• Defines the threshold value at or above which the test is considered “positive”
• If cut point is move to improve sensitivity , specificity typically falls and vice versa.
• Tradeoff between more accurate identification of subjects with disease versus those without disease often displayed graphically as a ROC curve.
Other set of data
Test A:• False Positive Rate = b/b+d (200/1800 = 0.11) • True Positive Rate = a/a+c (700/1800= 0.795)
Test B:• FPR= b/b+d (300/1800 = 0.17)
• TPR = a/a+c (800/880 = 0.91)
Test C:• FPR= b/b+d (500/1800 = 0.28)
• TPR = a/a+c (820/880 = 0.93)
Receiver Operating Characteristic (ROC)
False positive rate (1- specificity )
True
pos
itive
rate
(sen
sitiv
ity)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.70
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0.68
0.850000000000001
0.93
0.79
0.91 0.93
A2 B2 C2A B C
SNOUT & SPIN
• Approach to quantify the diagnostic ability of a test:
- Sensitivity = a/a+c - Specificity = d/b+d• SNOUT ( very sensitive test + result is not very
helpful, but – result is useful) Rule disease OUT.• SPIN (very specific test – result is not very
helpful, but + result is useful) Rule disease IN.
Predicative values• Positive predicative value (PPV)= a/a+bThe proportion of patients with positive results who are correctly diagnosed.
• Negative predicative value (NPV)= d/c+dThe proportion of patients with negative results who are correctly diagnosed.
• Depends critically on the prevalence of disease in the population tested.
Diseased Un diseased
Positive T.P a b F.P
Negative F.N c d T.N
Depression in severely ill kidney patients
• Sen =0.69 = 69%• Speci =0.83 = 83%• PPV =0.6 = 60%• NPV=0.88 =88%• Prevalence=• a+c/a+b+c+d = 26/98
= 26,5%
Dis Un dis+ 18 12 30
_ 8 60 68
26 72 98
Depression in pass students
• Sen = 69%• Speci = 83%• PPV =0.101 = 10.1%• NPV=0.98 =98%• Prevalence=a+c/
a+b+c+d = 265/10000 = 2.6%
Dis Un dis+ 183 1623 1806
_ 82 8112 8194
265 9735 10000
Prevalence
• High prevalence (26)----- High PPV =0.6 (60%)• Low prevalence (2.6)---- Low PPV=0.101 (10%)• Predictive values observed in one study do not
apply universally.• Role of prevalence in choosing a cut off point: - If non diseased(b+d) is high-----specific test - If diseased (a+c)is high------sensitive test (e.g. HIV in blood donors)
Likelihood Ratio
• LR describes how many times a person with disease is more likely to receive a particular test result than a person without disease.
• Binary tests have two LR:• LR+ve = sensitivity/ 1- specificity or TPR/FPR• LR-ve = 1- sensitivity /specificity or FNR/TNR
Bayes’ theorem
• Provides a simple mathematical way to calculate the post test probability of disease from 3 parameters: pretest probability , sensitivity & specificity
• Fagan’s nomogramPretest probability=p1 Pretest odds=p1/1-p1Posttest odds=pretest odds x LR Post test probability= o2/1+o2
Fagan’s nomogram
Parallel tests• Two or more tests each with possibility of + or – result• If test A or B…..is positive then overall result is positive• Test A & B:
• Both:• Combined sensitivity in parallels test is higher than
each. 1- {(1-sens of A)X(1-sensof B)} = 1-{(1-0.70)x(1-0.60)} = 0.88 =88%• Combined specificity: sp A x sp B = 0.8 x 0.7= 0.56 = 56%
Test B: sensitivity = 60%Specificity = 0.7 =70%PPV= 67%NPV= 63%
Test A: sensitivity = 70%Specificity = 0.8= 80%PPV= 78%NPV= 72%
Usage of parallel tests
• If cost of false positive is not high • In emergency cases( time is important) • When we need a high sensitivity
Serial testing
• Sequence is important : If A is (-) stop but if (+) then do B.........(+) then should take a decision.
• If A &B &C are all positive then ---- decision.• Combined sensitivity = sen A xsen B • Combined specificity = 1- {(1-sp A)X(1-sp B)}
Usage of serial testing
• When cost of F.P is very high• In rare cases• In screening of large population• When can't apply all test to all population
Thanks