3A_Clinical Decision Analysis and ROC

Question 1

When you move the decision making threshold towards the curve with the individuals positive for the disease, which improves: sensitivity or specificity?

Answer 1

Specificity

Question 2

When you move the decision making threshold towards the curve with the individuals negative for the disease, which improves: sensitivity or specificity?

Answer 2

Sensitivity

Question 3

What is ROC?

Answer 3

Receiver Operating Curve

• ROC Plot shows True Positive Rate (%) as a function of False Positive Rate (%)
• TP rate = sensitivity (# positive tests / # with disease)
• FP rate = 1 - specificity (# neg tests / # disease absent)

Question 4

What is the clinical value of a ROC plot?

Answer 4

• it compares different test result criteria values for a single test (parameter)
• it compares across different tests or different test parameters

Question 5

What would an idealized ROC plot look like?

Answer 5

100% True positives, 0% False positive (so, bent towards upper left corner)

Question 6

What would the worst test ROC plot look like?

Answer 6

Straight diagonal line (pure chance - both condition curves overlap completely)