Scenario: you have two diagnostic “blackboxes”. One higher sensitivity like 99% but low specificity like 75%; the other higher specificity like 95% but low sensitivity like 70%

To get an intuive feel of the difference, we need to imagine an extreme scenario of high-low. If you have a diagnostic that’s superior in both sensitivity and specificity, then clearly superior… not useful for this discussion.

Both specfiCity and sensitivity are all about error rates:

• high specifiCity means Low false_positive rate. Therefore, positive results are reliable.
• high sensitivity means Low false_negative rate. Therefore, negative results are reliable.

Sensitivity and Specificity Explained for Medical Professionals

— 99% sensitivity means out of every 100 true positives, one would be missed out i.e. false negative (99 would show positive) in this diagnostic. A highly sensitive perhaps over-sensitive test (eg: CT, Hba1c)

• 🙂 is good for rule-out, as it is unlikely to miss a true positive case.
• 🙁 it can misfire, give false_positive like spurious wake-up.
• .. if specificity = 75%, then 25% of the true negatives will show false positive.

— specificity .. is all about false_positives or misfire

A positive result of high specificity is “a highly specific indicator”.

95% specificity means out of 100 true negatives, 5 would show false_positive in this diagnostic. Some definitions of specificity focus on the 95% .. often convolunted and non-intuitive.

A highly specific diagnostic

• 🙂 has low false_positive rate, therefore good for rule-in
• 🙂 is an accurate indicator of a specific condition. A positive is unlikely to be some look-alike indistinguishable conditions.
• 🙁 but can miss some true positives… 70% sensitivity means 30% of true positives missed.