Q: What are 2 uses of ROC curves?
show answer
A: ROC curves can be used to determine the optimal "cut point" for a diagnostic test with continuous outcomes. It can also be used to compare diagnostic tests: the test with the bigger "area under the curve" (AUC) is generally the better test.
Q: What would happen to diagnostic test specificity if you change the cut point of your test from a less sensitive to a more sensitive value?
show answer
A: When you change the cut point of your test from a less sensitive to a more sensitive value, you've moved from left to right along your curve. This means that you will mislabel more of the well as ill (have a look at the 3rd image). In other words, you'll end up with more true positives because you chose a more sensitive cut point (good news), but you also end up with more false positives (bad news). Remember, specificity refers to how well the test does at correctly identifying the well. More false positives will therefore result in lower specificity. If you move from left to right on the ROC curve, sensitivity increases and specificity decreases. If you move from right to left on the ROC curve, sensitivity decreases and specificity increases. Again, look at the third image. Phew!
Q: Why might you choose a more sensitive but less specific cut point?
show answer
A: If the consequences of missing something (e.g. cancer) are high and the consequences of a false positive are low (perhaps a different non-invasive test to see if the patient is really ill), then it makes sense to use the more sensitive cut point.