Chapter 5- Assessing the Validity and Reliability of Diagnostic Screening Tests Flashcards
Bimodal curve
A graph with 2 peaks. However, most human characteristics are not bimodal
Validity
The ability of a test to determine who has a disease and who does not. It has 2 components- sensitivity and specificity
Sensitivity
The ability of the test to correctly identify who has a disease
Specificity
The ability of the test to correctly identify who does NOT have a disease
Risks of false positives
People who test positive may have to undergo more expensive and maybe more invasive testing. It may place unnecessary burdens on the healthcare system. Also, it causes stress for the patient who thinks that they have that specific disease
Risks of false negatives
If the disease being screened for is serious and has effective intervention, false negatives can delay diagnosis and negatively impact the patient’s outcome.
Cutoff level
This is necessary if a test doesn’t have a positive or negative result- there is no such thing as a “positive” blood pressure because it is a continuous variable. In this case, a unimodal curve is most common. It must be determined which blood pressure levels qualify as hypertension or hypotension. If the level is too high, people may be falsely considered negative when they actually do have hypertension. The cutoff level essentially determines sensitivity or specificity
Trade-off between sensitivity and specificity
If we increase the sensitivity by lowering the cutoff level, we decrease the specificity. However, if the cutoff level is raised to increase the specificity, the sensitivity will decrease
Sequential testing
Two stage testing, where tests are done one after the other. A less expensive, less invasive, or less uncomfortable test is conducted first. Those who test positive are recalled for another testing that might be more invasive but may have greater sensitivity or specificity. The goal in bringing people back for additional testing is to reduce the issue of false positives. These tests are frequently used for screening programs.
Simultaneous testing
When both tests are conducted at the same time. This may be done with clinical workups, like when people are being admitted to the hospital. Results are interpreted as positive if the patient is positive on either test, and negative if negative on all tests
Screening tests
Tests performed on asymptomatic
patients. Goal is to evaluate likelihood of
disease with the goal of early
detection
Diagnostic tests
Tests performed on symptomatic
individuals –or– persons with
abnormal screening tests. Goal is to establish presence or absence of disease
Dichotomous tests
Tests that have either a positive or negative result
Impact of sequential testing on sensitivity and specificity
With sequential testing, there is a loss in net sensitivity. There is a gain in net specificity
Impact of simultaneous testing on sensitivity and specificity
There is a net gain in sensitivity and a net loss in specificity
Positive predictive value (PPV)
If the test results are positive for the patient, what is the probability that the patient has the disease?
Negative predictive value (NPV)
If the test result is negative, what is the probability that the patient does not have the disease?
2 factors that affect PPV
- The prevalence of the disease in the population tested
- The specificity of the test being used- in situations when the disease is infrequent
Intrasubject variation
How an individual patient’s test results can vary over time. For example, a person’s blood pressure can be different depending on the time of day
Intraobserver variation
Variation that occurs between 2 or more readings of the same test results made by the same observer. If a radiologist reads the same group of X-Rays at 2 different times, how will the results change?
Interobserver variation
The variation between observers- two examiners may not give the same result
Reliability
Ability to get the same results if the test is repeated. There are 2 measures- percent agreement and the kappa statistic
Percent agreement
Does not account for agreement between observations due to “chance”
Kappa statistic
Accounts for agreement between observations due to “chance”. Less than 40% is poor agreement, greater than 40% is intermediate to good
When is disease detection worthwhile?
“Bear” diseases are dangerous, but move slowly enough to catch. These diseases benefit the most from screening programs. “Birds” move too quickly to be caught, and it will be too late to treat. “Turtles” move too slowly to ever be dangerous, and don’t need treatment.
Lead-time bias
Occurs when a disease is detected by a screening or surveillance test at an earlier time point than it would have been if it had been diagnosed by its clinical appearance. This time lag or “lead time” during which the disease is asymptomatic is not taken into account during the survival analysis