Epidemiologic evaluation of diagnostic tests Mr.sci. Sabina Šerić-Haračić TCDC/TCCT consultant – Aquatic epidemiology [email protected].

Epidemiologic evaluation of diagnostic tests

Mr.sci. Sabina Šerić-Haračić

TCDC/TCCT consultant – Aquatic [email protected]

Where to sample What to sample How to properly package/ship samples How many samples Where to ship Who is paying

Diagnostic consideration in field

Is sampled material appropriate for test used Have samples been shipped properly Condition of samples upon arrival Having and knowing to use recommended

test/s Where to send results Who is paying

Diagnostic consideration in lab

Having relevant and detailed sampling, shipping, testing procedures

Level of compliance to prescribed procedures Do we have all necessary information Is number of samples tested enough to safely

conclude presence/absence of disease Are test/s used precise enough to confirm/rule

out disease How much this costs

Diagnostic consideration in decision making

Diseased Healthy

Test positive

Test negative

Test sensitivity and specificityAre test/s used precise enough to confirm/rule out disease

100 100

105

95

95 10

5 90

Diseased Healthy

Test positive95 10

105

Test negative5 90

95

100 100 200

Test sensitivity and specificity

Sensitivity =

Specificity= =

Prevalence == - Prevalence based on test result (apparent

prevalence) Apparent prevalence ==

True prevalence==

Se/Sp and prevalence

UNKNOWN!!!

KNOWN!!!

Positive tests results came for 92% of diseased animals 3% healthy animals

What is test Sensitivity and Specificity for Prevalence = 7,3% Prevalence =15% 

Assignment

Diseased Healthy

Test positive

Test negative

Prevalence = 0 SP=90% Out of 100 healthy animals we will have 5

positive results (false positive) Repeat testing Use another test Use better test Demonstrating disease freedom requires

Sp=100%

SE/SP and disease freedom

Fair coin -50% cahnce of heads/tails Assume unfair coin – 90% of heads 10%of

tails Number of trows – 20 Chance to have all heads from 20 trows if

chance head in each trow is 90%

Test results on herd level(farm/pond/pooled samples...)

Result of a test from at least 2 animals or from pooling of 2 samples

Interpretation of the herd status is more important than interpretation of each individual's test results

Herd Sensitivity is the probability that an infected herd has a positive herd test result

Herd Specificity is the probability that an non-infected herd has a negative herd test result

Test results on herd level(farm/pond/pooled samples...)

Factors affecting Herd Sensitivity (HSe) and Herd Specificity (HSp)

Individual Se and Sp Within herd prevalence of disease Number of animals tested in the group Number of reactor animals per group that

will designate a positive or negative herd

Test results on herd level(farm/pond/pooled samples...)

For and infected herd (P>0):

For a non-infected herd (TP=0):

Test results on herd level(farm/pond/pooled samples...)

Probability of obtaining a positive test AP=p(T+)=Se * P + (1-P)*(1-Sp)

Probability of obtaining a negative test 1-AP

Probability of finding zero positive individuals (false negative herd proportion)

(1-AP)n

Herd Sensitivity (HSe) 1-(1-AP)n

Probability of obtaining a positive test (replace P=0 in eq 1)

AP= (1-Sp)

Probability of finding at least one positive animal (false positive herd proportion)

1-Spn

Probability of obtaining a negative test 1-AP = p(T-) = 1-(1-Sp)=Sp

Herd Specificity (HSp) Spn

Number of animals on farm = 200 Within farm prevalence = 10% Sample size = 20 Se = 95% Sp=90% HSe=1- Within farm prevalence =0 HSp==0,122

Example/Assignment

1

100% 10

0 0,864

100

Number of animals on farm = 200 Within farm prevalence = 1% Sample size = 20 Se = 95% Sp=100% HSe=1- Within farm prevalence =0 HSp==1

Example/Assignment

100

100 0,61

5

100