TL;DR: You want the most sensitive test first to rule out the disease in negatives (least likelihood of false negative) and the most specific test next to "rule in" only true positives (least likelihood of false positives) [FA2020 p257]
The way I always think of sensitivity and specificity is in relation to false negatives and false positives.
High Se N setivity means you are less likely to have false Negatives (someone who has the disease but tests negative)
High s P eceificity means you are less likely to have false Positives (someone who tests positive but does not have the disease)
In this question, you want to ensure you are only treating those who actually have the disease, or in other words that you want to minimize your false positives. The best way to do this is maximize sPecificity. Test 1 has 100% specificity meaning it will have essentially NO false positives which is great so we definitely want to use it (eliminating all but two options)! The better way to ensure you are catching as many cases as possible while still only getting true positives is to test with test 3 first, which has a much higher sensitivity. That way we are ruling out disease in the negatives of test 3. This also adds to our ability to only truly treat those with disease as we have ruled out disease in our first negative population, then we test the positives to "rule in" disease with a highly specific test.
The issue with using Test 1 first and only testing the positives is we know the positives from test 1 are likely true positives, as the highly specific test would have a basically zero false positive rate. At that point there would be no point in testing with test 3, as we have basically a true positive population. The issue is we have missed a lot of diseased due to the low sensitivity and high false negative rate of test 1.
Another way to think of this is to remember the formulas (as seNsitivity contains false Negatives, and sPecificity contains false Positives in the formulas; see FirstAid):
sensitivity = true positive / (true positive + false negative) = 1 - false negative rate --> higher sensitivity means less rate of false negatives
specificity = true negative / (true negative + false positive) = 1 - false positive rate --> higher specificity means less rate of false positives
SPIN/SNOUT also helps to remember that specific tests rule in disease, and sensitive tests rule out disease
the_crown_7Great Explanation. This question tripped me up, with thinking that the best test, would maximize both parameters. So can someone please explain why "Test 2 only" which has the highest sen & spec, is incorrect.+
shiggins8Why I didn't choose Test 2 only is because the stem stressed the 'toxicity of the drug,' and thus, treating those who actually have the disease. You would still be treating individuals who may not have the illness, thus exposing them to unnecessary toxicity. Or at least that's how I viewed it.+
submitted by โcassdawg(1781)
TL;DR: You want the most sensitive test first to rule out the disease in negatives (least likelihood of false negative) and the most specific test next to "rule in" only true positives (least likelihood of false positives) [FA2020 p257]
The way I always think of sensitivity and specificity is in relation to false negatives and false positives.
In this question, you want to ensure you are only treating those who actually have the disease, or in other words that you want to minimize your false positives. The best way to do this is maximize sPecificity. Test 1 has 100% specificity meaning it will have essentially NO false positives which is great so we definitely want to use it (eliminating all but two options)! The better way to ensure you are catching as many cases as possible while still only getting true positives is to test with test 3 first, which has a much higher sensitivity. That way we are ruling out disease in the negatives of test 3. This also adds to our ability to only truly treat those with disease as we have ruled out disease in our first negative population, then we test the positives to "rule in" disease with a highly specific test.
The issue with using Test 1 first and only testing the positives is we know the positives from test 1 are likely true positives, as the highly specific test would have a basically zero false positive rate. At that point there would be no point in testing with test 3, as we have basically a true positive population. The issue is we have missed a lot of diseased due to the low sensitivity and high false negative rate of test 1.
Another way to think of this is to remember the formulas (as seNsitivity contains false Negatives, and sPecificity contains false Positives in the formulas; see FirstAid):
SPIN/SNOUT also helps to remember that specific tests rule in disease, and sensitive tests rule out disease