33: Diagnostic Accuracy Flashcards
A test of coordination is scored from 0-100, with a cut-off score of 50. Scores below 50 indicate abnormal loss of coordination; scores 50 or above indicate a normal range of coordination. A patient is tested, achieving a score of 38 (a positive test), suggesting that she has decreased coordination. The test has a sensitivity of .73 and specificity of .95. Which of the following statements is a true interpretation of these data?
a. The false positive rate for this test is 27%.
b. The false negative rate for this test is 5%.
c. This test will be better at ruling in a coordination deficit.
d. This test will be better at ruling out a coordination deficit.
ANS: C
Rationale: Based on SpIN, the test has a higher probability of being able to rule in a deficit (having a higher specificity). The false positive rate is 5% and the false negative rate is 27%.
A test for the presence of HIV has a positive predictive value (PV+) of 83%. statements is an accurate interpretation of this value?
a. 83% of those who test positive for HIV are likely to have the virus.
b. 27% of those who test positive for HIV actually have the virus.
c. The negative predictive value (PV–) is 27%.
d. 27% of those who test negative for HIV will not have the virus.
ANS: A
Rationale: PV+ estimates the likelihood that a person who tests positive will have the disorder.
A positive likelihood ratio (LR+) is the ratio of:
Which of the following
a. Specificity divided by 1–sensitivity.
b. Specificity divided by sensitivity.
c. Sensitivity divided by specificity.
d. Sensitivity divided by 1–specificity.
ANS: D
Rationale: LR+ = sensitivity/1–specificity.
A negative likelihood ratio (LR–) tells us:
a. How many times more likely a negative test will be seen in those with the disorder than in
those without it.
b. How many times more likely a positive test will be seen in those with the disorder than in
those without it.
c. How many times more likely a negative test will be seen those without the disorder than in
those with it.
d. How many times more likely a positive test will be seen in those without the disorder than
in those with it.
ANS: A
Rationale: A test with a low LR– means that the disorder has a small probability of being present with negative test. A good test will have a low negative likelihood ratio.
Which of the following statements is true if the LR+ is large?
a. A positive test is good at ruling the disorder out.
b. A positive test is good at ruling the disorder in.
c. A negative test is good at ruling the disorder out.
d. A negative test is good at ruling the disorder in.
ANS: B
Rationale: To get a large LR+, specificity should be large (the denominator is 1–specificity). Using the same rationale as SpPIN, a high LR+ will indicate that a positive test is good at ruling the disorder in.
A study was done to consider the relative accuracy of a new test for carpal tunnel syndrome, testing it against nerve conduction velocity studies. The prevalence of the condition in the population is 25%. The sensitivity of the test was shown to be 90% and specificity was 94%. Which of the following statements is accurate?
a. The positive likelihood ratio is 1.06.
b. The pretest probability is 75%.
c. The posttest probability for a positive test is approximately 80%
d. The posttest probability for a negative test is approximately 80%.
ANS: C
Rationale: The pretest probability is .25. The LR+ = 15.0. The LR– = 0.106. Using either the nomogram or direct calculations (or an online calculator), the posttest probability for a positive test is approximately 80% and for a negative test is approximately 23%.
A study was done to determine the accuracy of a new test for scoliosis, testing it against radiographs. The test was scored from 0-10, with 10 representing the highest degree of scoliosis. A cutoff score of 3 was used to indicate a diagnosis of scoliosis. Which form of analysis would be most appropriate to determine if this test is useful?
a. Plotting coordinates for different cutoff scores on an ROC curve.
b. Determining the sensitivity and specificity of the test using the cutoff score of 3.
c. Calculating the likelihood ratios at the cutoff score of 3.
d. Calculating likelihood ratios at different cutoff scores.
ANS: A
Rationale: The ROC curve can show which cutoff score provides the best balance between true positives and true negatives.
All of the following indicate a purpose of a clinical prediction rule except:
a. To indicate the likelihood that a patient with certain characteristics will respond favorably to an intervention.
b. To indicate if a set of prognostic factors can predict an eventual outcome.
c. To indicate if a set of characteristics can predict the presence or absence of a disorder.
d. To indicate the likelihood that a particular cutoff score of a test can be used to determine the presence or absence of a disorder.
ANS: D
Rationale: A CPR is not intended to look at cutoff scores. An ROC curve is used for that purpose.
A clinical prediction rule (CPR) has been derived to indicate the presence of sleep apnea based on neck circumference, snoring, and hypertension. Using the CPR, the LR+ was 5.2 and the LR– was 0.25, with a pretest probability of 45%. All of the following statements are true about this test except:
a. The posttest probability for a negative test is approximately 17%.
b. The posttest probability for a positive test is approximately 80%.
c. This test will be good at identifying those with and without sleep apnea.
d. This test will be better at identifying those without sleep apnea.
D
Rationale: The posttest probability for LR– = ~17%. The posttest probability for LR+ is ~80%. These both indicate a strong test that will identify both positive and negative cases.
For the study of a diagnostic test to be valid, all of the following conditions should bet met except:
a. The index test and reference standard should be taken at approximately the same time.
b. Those taking measurements should know the actual diagnosis before they administer the test.
c. The validity of the reference standard should be established.
d. Participants should represent the population to which the test would apply.
b
Rationale: Those taking measurements should be blinded to the diagnosis to assure there is no bias.
