Diagnostic tests Flashcards
Screening vs diagnosis
Screening tests done on healthy animals (detect seroprevalence, subclinical disease)
Diagnostic tests used to confirm or classify disease, guide treatment and prognosis.
Analytic sensitivity and specificity
Analytic sensitivity is lowest concentration of chemical compound that test can detect
Analytic specificity relateds to ability of test to only identify one chemcial compound (vs cross-react with multiple compounds)
Accuracy vs precision
Accuracy is how well the test depicts the truth on repeated testing
Precision is how consistent the results of the test are on repeat testing
- Repeatability - same results at same lab
- Reproducibility - same results at different lab
Sensitivity
Proportion of diseased animals that test positive
- SNOUT: Sensitive tests are useful for ruling out a disease. WIth sensitive tests, there is a low false negative rate and therefore if a test comes back negative it is likely that the animal is a true negative.
Specificity
Proportion of disease negative animals that test negative
- SPIN: specific tests useful for ruling in a diagnosis (i.e. confirming diagnosis). WIth specific tests, there is a low false positive rate and therefore if a lest comes back positive the animal is likely to have the disease
False positive fraction
1 - specificity
False negative fraction
1 - sensitivity
Apparent prevalence
test positives ÷ # animals tested
True prevalence
(AP + Sp -1) /
(Se + Sp - 1)
Actual level of disease in population (= prior probability of disease)
Positive predictive value
Probability that an animal with a positive test actually has the disease
- PPV increases with increasing prevalence
Negative predictive value
Probability that an animal with a negative test actually does not have the disease
Effect of prior probability on predictive value - under high/low prevalence, changing prevalence
PPV is highest when prevalence is high (positives are likely to be true positives).
- At moderate-high prevalence or at the start of control campaign, most test positives will be true positives. Therefore we use a sensitive test to minimize the false negative rate.
- At low-zero prevalence or at the end of an eradication campaign, more animals that test positive will be false positives. Therefore we use specific test to minimize the false positives rate.
Increasing positive predictive value of a test (3)
- Target screening towards animals most likely to have disease (e.g culled animals, animals displaying symptoms)
- Use more specific test or change cut-point to make test more specific (reduces false positives to zero)
- Use more than one test
Using tests in parallel (individual animal)
Animals testing positive to any/both tests are declared positive. Testing stops on first positive result.
- Increases sensitivity (decreases false negative rate) but decreases specificity
- Animal effectively being asked to prove that they are healthy
- Recommended for rapid assessment because animal is considered positive on the first positive result
- Disease less likely to be missed (false positives more likely)
Using tests in series (individual animal)
Only animals that test positive to both tests are considered positive. Testing stopped as soon as soon as one negative result is received.
- Increases specificity (low false positive rate) but decreases sensitivity
- Animal effectively being asked to prove that they have the disease
- Use cheaper test first (reduces costs)
- First test should be highly sensitive, since negative test will be taken to be a negative result
- More likely diseased animals will be missed
- Used when consequences of positive finding are dire
- Often used in control and eradication programs
- Example: BSE (rapid test [ELISA] then confirmataory test)
Herd level testing
Inferences made about disease status of the herd on the basis of the test results of only a few animals. Analagous situation to parallel testing (herd declared positive following only one positive result).
Herd-level sensitivity - definition, determinants (4)
Probability of that a test is capable of detecting at least one of the positive animals, given that one or more animal in the herd has the disease.
Determinants:
- Individual level test sensitivity
- Number of animals tested (more animals tested the more likely we will detect positives - true and false)
- Frequency of disease within infected herds (apparent prevalence)
- Threshold number of test positives for declaring a herd positive; usually 1 animal, unless test specificity is less than 100% (in which case higher threshold may be acceptable to allow for false positives)
Herd level specificity - definition, determinants (2)
Probability that a test is capable of declaring a herd negative, given that no animals in that herd have the disease.
Deteriminants:
- Individual test specificity
- Number of animals tested
Receiver operating characteristic (ROC) curves - plot, interpretation
Plot
- Y axis: Sensitivity
- X axis: 1 - Specificity (false positive rate)
Each point represents a particular cut-off value
Interpretaion:
- Diagonal line: test cannot meaningfully distinguish diseased and non-doseases (no better than chance)
- Test closest top left is best (100% sensitivity and 100% specificity)