Graph I presents an imaginary picture of how cases and non-cases distribute
themselves on a screening test.
Graph II presents a realistic picture of how cases and non-cases distribute
themselves on a screening test.
What’s the difference between Graph I and Graph II?
In Graph I cases and non-cases never receive the same screening test scores. All cases score 7 or
above and all non-cases receive scores below 7. Thus, selecting a cutpoint of 7 as the criterion of
positivity means every case will be scored + and every non-case will be scored -. Since all cases will
be scored +, the sensitivity of the test will be 100%. And since all non-cases will be scored -, the
specificity of the test will also be 100%.
In Graph II, however, there is overlap between the cases and non-cases regarding the screening test
scores. Overlap means that some cases and some non-cases receive the same scores on the
screening test. Specifically, on this screening test some cases score between 5 - 7 and some
non-cases also score between 5 - 7.
This means if the cutpoint of 5 was selected, all cases would be correctly screened +, resulting in a
sensitivity of 100%. But some non-cases would also be screened +, resulting in a non-zero rate of false
positives, and thus a specificity of less than 100%.
On the other hand, if the cutpoint was increased to 7, all non-cases would be correctly screened -,
resulting in a specificity of 100%. But some cases would also be screened -, resulting in a non-zero rate
of false negatives, and thus a sensitivity of less than 100%.
themselves on a screening test.
Graph II presents a realistic picture of how cases and non-cases distribute
themselves on a screening test.
What’s the difference between Graph I and Graph II?
In Graph I cases and non-cases never receive the same screening test scores. All cases score 7 or
above and all non-cases receive scores below 7. Thus, selecting a cutpoint of 7 as the criterion of
positivity means every case will be scored + and every non-case will be scored -. Since all cases will
be scored +, the sensitivity of the test will be 100%. And since all non-cases will be scored -, the
specificity of the test will also be 100%.
In Graph II, however, there is overlap between the cases and non-cases regarding the screening test
scores. Overlap means that some cases and some non-cases receive the same scores on the
screening test. Specifically, on this screening test some cases score between 5 - 7 and some
non-cases also score between 5 - 7.
This means if the cutpoint of 5 was selected, all cases would be correctly screened +, resulting in a
sensitivity of 100%. But some non-cases would also be screened +, resulting in a non-zero rate of false
positives, and thus a specificity of less than 100%.
On the other hand, if the cutpoint was increased to 7, all non-cases would be correctly screened -,
resulting in a specificity of 100%. But some cases would also be screened -, resulting in a non-zero rate
of false negatives, and thus a sensitivity of less than 100%.
Comments
Post a Comment