Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
31 Cards in this Set
- Front
- Back
|
Analytical Sensitivity |
Refers to the lower concentration that a test can detect Also called detection limit |
|
|
Analytic specificity |
Refers to capacity of the test to react to only one chemical compound |
|
|
What is accuracy? |
Ability of a test to give a true measure of the substance of interest |
|
|
What is precision? |
Refers to how consistent the results for the tests are |
|
|
What is repeatability? |
Variability among tests results obtained from testing the same sample within the same laboratory |
|
|
what is reproducibility? |
Variability among test results obtained from testing the same sample in different laboratories |
|
|
How to measure accuracy? |
Run tests on samples with a known quantity of substance present |
|
|
How to measure precision |
Coefficient of variation (CV) CV = standard deviation (δ) / average of test results (µ) CV = δ / µ |
|
|
What is agreement? |
Refers to how well 2 tests agree |
|
|
What does agreement not tell you? |
No information about the accuracy of the test No info about the best tests of the two - useful when the true status/concentration is unknown |
|
|
What are 3 methods to measure agreement for 2 tests with quantitative outcomes? |
1) Pearson correlation coefficient (not recommended) 2) Concordance correlation coefficient - best one to use 3) Limits of agreement plots (also called Bland-Altman plot) |
|
|
How to measure agreement for 2 tests with a qualitative outcome? |
1) Cohen's kappa statistic (κ) - What is the level of agreement beyond what would have been expected by chance? Kappa = actual agreement beyond chance / potential agreement beyond chance *Use kappa if you're not 100% sure of sensitivity and specificity of the tests or if the animal is actually sick or not* |
|
|
Interpretation of kappa |
≤0 = poor agreement 0.01-0.2 = slight agreement 0.21-0.4 = fair agreement 0.41-0.6 = moderate agreement 0.61-0.8 = substantial agreement 0.81-1.0 = almost perfect agreement |
|
|
5 approaches to estimating Se and Sp of diagnostic tests |
1) use gold standard population 2) use gold standard reference test 3) use of a pseudo-gold standard reference test (or combo of tests) 4) use of a reference test with known Se and Sp 5) evaluation of test in absence of gold standard |
|
|
What is a gold standard population? |
A population assumed to be completely free of disease - use them to estimate Sp of test **Not often used** |
|
|
What is the issue with a gold standard population? |
Usually no population in which all animals assumed to be diseased for estimation of Se - experimentally infected population? Question: external validity? (potential bias) |
|
|
1 Stage approach to using a gold standard reference test |
Sample of animals from population are tested both by gold standard test and the test being evaluated |
|
|
Potential problem with 1 stage approach of gold standard reference testing? |
Need a very large sample size if the disease prevalence is low |
|
|
What is the 2-stage approach to using a gold standard reference test? |
Sample of animals is screened with test being evaluated, and then a subsample of T+ and T- is submitted to the gold standard |
|
|
Potential biases of 2-stage approach to using a gold standard reference test |
1) Must be random sampling size of T+ and T- - if not random = potential selection bias 2) Same size* of verified T+ and T- - if not same size = potential verification bias * Can adjust for the difference in size |
|
|
How do we use a pseudo-gold standard reference test/combo of tests? |
Use combo of imperfect tests as a substitute for a gold standard Example: - Cattle with BRD if pulled by pen rider with BRD sign, a rectal temp >104F and abnormal lung sounds = BRD - Cattle pulled by pen rider as healthy, rectal temp <104F and no abnormal lung sounds = healthy |
|
|
What is the issue with using a pseudo-gold standard reference test/combo of tests? |
Test to be evaluated cannot be part of the combination Potential bias = incorporation bias = overestimation of diagnostic tests Se and Sp |
|
|
How to evaluate a diagnostic test using a reference test with a known Se and Sp |
Recalculate the 2x2 table based on the Seref and Spref **not done very commonly** |
|
|
How to evaluate a test in the absence of gold standard test |
When there is no reasonable gold standard nor a test with known characteristics (Se and Sp) - use of latent class models (advanced computer modelling) |
|
|
What is external validity? |
How well the results can be generalized to the target population - easier to generalize results of analytic studies than descriptive studies |
|
|
What is internal validity? |
How accurately the results of the study are for members of the source population - do the characteristics and findings correctly represent the source population? |
|
|
What is review bias? |
Lack of blinding *Animals with unexpected results should not be re-tested (unless required for technical reason) Potential overoptimistic sensitivity and specificity estimates |
|
|
What is spectrum of disease bias? |
The distribution of disease stages in the sample is not representative of the source population (sampling bias) = overestimation of Sensitivity |
|
|
What is work-up bias? |
more extensive verification of positive results than negative results (or vice versa) = overestimation of Se and Sp |
|
|
What is selection bias? |
No randomization in the (sub)sample - may get rid of false positives and false negatives if makes tests look better than they actually are |
|
|
What is information bias? |
Low accuracy of reference test Underestimation of the Sp of the test you are analyzing if it is actually more sensitive than the reference test **If you do not have a perfect test = use latent class analysis** |
