Testing Ch 13 D&P Flashcards
What is accuracy?
How close the test result is to the True Value of the analyte, like comparing the test result to some gold standard or known value. Like your getting the shots near the bulls eye.
What is precision and what specific value is used to determine it?
Reproducibility. Like your getting all the shots in a tight cluster than may or may not be near the bullseye.
The coefficient of variation is used to determine the Precision.
What are the three general categories for sources of error in lab tests?
- Pre-analytical (made by the sampler prior to testing, like wrong technique, wrong tube etc, specimen ID, temp, storage etc.)
- Analytical (An error with the machine when the sample is being tested in the lab, requires proper maintenance and calibration to avoid this)
- Post Analytical (Incorrect data entry and generation of reports).
Important point regarding reference intervals.
They are specific to the test method and reference group and cannot be transferred between labs without validation even when equipment and test methods are the same.
How do you design a 2X2 contingency table?
Actual dz status goes across the top with Dz positive animals in the left column (includes TPs and FNs) and in the right column is dz negative animals (TNs and FPs). This puts the test result on the left side of the table with Positive test result in the top row (TPs and FPs) and Negative test result in the bottom row (FNs and TNs).
Sensitivity = TP/(TP+FN) (calcuated under pos dz column)
Specificity = TN/(TN+FP) (calculated under neg dz column)
PPV = TP/(TP+FP) (calculated in test pos row)
NPV = TN/(TN+FN) (calculated in test neg row)
When designing the table just remember true dz status is on top and then dz positive is on the left, and test pos is on the top.
Positives are in the two top blocks and negatives are in the two bottom blocks.
See table 13.4 page 377
How is sensitivity defined and how is it calculated?
The frequency of a positive test result in a dz positive animal.
Sen = TP/(TP+FN) X 100 (to get the percentage)
The measure of the ability of the test to identify positive individuals, the higher the Se, the more likely a dz’d patient will test pos.
What is the important characteristic of a screening test?
High sensitivity. Few patients will incorrectly test negative. A positive result DOES NOT confirm dz but a negative result effectively eliminates the dz.
What is specificity and how is it defined?
Defined as frequency of a negative test result when the dz is absent. (essentially the rate of TNs, so it also must consider FPs)
The ability of a lab test to ID patients that do not have the disease, OR to ONLY identify patients with the dz. The higher the Specificity the more likely a non dz’d patient will test negative.
- Sp = TN/(FP+TN) (ie you take the true/test negatives and divide it by all the negatives, which includes the false positives)
What is the important characteristic of a confirmatory or diagnostic test?
High specificity is important for a confirmatory test to keep the FPs as low as possible. A positive test result confirms dz, but a negative result doesn’t necessarily mean the animal is negative (ie FN).
What is the relationship between Se and Sp?
There is usually a trade off, you usually don’t have both high at the same time. So you first do test with high sensitivity to screen and this test has NO false negatives.
Then you follow up with a confirmatory test that has NO false positives or a highly specific test.
What is a ROC curve?
Its a graphical representation of the inverse relationship between Sensitivity and Specificity. It graphs the Sensitivity (or true positive rate) on the Y axis and specificity (or false positive rate) on the X axis. Ideally the line closely approaches the upper left corner. See Fig 13.3 page 379
What is the positive predicted value?
An indicator of the probability of having the disease if the patient tests positive. A high PPV means most test positive patients have the dz. Its defined as a percentage of dz’d patients among patients that test positive.
You calculate it using the top row of the table (TP / (TP+FP)), so its the probability of having the dz if the animal tests positive and is related to the prevalence of dz in the population.
What is negative predicted value?
Percentage of non dz’d patients among patients that test negative.
TN/(TN+FN)= NPV
An indicator of the probability of being disease free if you test negative. So if the NPV is high, most patients that test negative are dz free.
Predictive values are based on what factors?
Actual true dz prevalence. The more common the dz among the population your screening the more likely they going to be positive regardless of your test result. A positive test is more likely to be correct when the prevalence is high.
Conversely with NPV, the rarer the dz, the more likely the animal will be negative and a negative test is more likely to be right if the prevalence is already low.
How do you increase the PPV?
Perform lab tests in a series, ie the second test is done if the first is positive. Specificity is increased, but sensitivity is decreased with this approach. Good approach when txment could harm the patient.
