Sensitivity and Specificity: The Two Numbers Behind Every Test
Cutoffs, Prevalence, and What a Result Actually Tells You
A test that is 99 percent accurate can still mislead you
Picture a screening test for a disease that affects about 1 in 1,000 people in the population being tested. The test is genuinely good: it correctly identifies 99 percent of people who truly have the disease, and it correctly clears 99 percent of people who do not. Now imagine screening 100,000 people with it.
About 100 of those people actually have the disease. The test catches 99 of them and misses 1. The remaining 99,900 people do not have the disease, and the test correctly clears about 98,901 of them, but incorrectly flags about 999 healthy people as positive. Add it up, and out of roughly 1,098 total positive results, only 99 are real. That means a person who tests positive on this “99 percent accurate” test actually has the disease only about 9 percent of the time.
That is not a flaw in the math. It is what happens when a rare disease meets an imperfect test, and it is the reason two numbers, sensitivity and specificity, are only half the story behind any diagnostic test.
Sensitivity and specificity, defined
Sensitivity is the percentage of people who truly have a condition that the test correctly identifies as positive, the 99 percent in the example above. A highly sensitive test rarely misses real disease, which is why it matters most when missing a case is dangerous.
Specificity is the percentage of people who truly do not have the condition that the test correctly identifies as negative, also 99 percent in the example. A highly specific test rarely raises a false alarm, which matters most when a false positive leads to invasive follow up testing or unnecessary treatment.
Both numbers describe the test itself, how good it is at telling sick from healthy, when you already know the true answer for every person tested. Neither one, on its own, tells a patient what a single result actually means for them.
Where the cutoff comes from
Many tests are not naturally a clean positive or negative. A lab value, an antibody level, a screening questionnaire score, these produce a continuous number, and someone has to decide where to draw the line that separates a positive result from a negative one. That line is the cutoff, and moving it trades sensitivity for specificity in a direct, unavoidable seesaw.
Lower the cutoff, and the test flags more people as positive. It catches more true cases, raising sensitivity, but it also sweeps in more healthy people, lowering specificity. Raise the cutoff, and the opposite happens: fewer false alarms and higher specificity, at the cost of missing some real cases. There is no cutoff that maximizes both at once. Where a test's threshold gets set depends on what is more dangerous to get wrong, missing a case of an aggressive cancer argues for a lower cutoff and higher sensitivity, while confirming a diagnosis before starting a toxic treatment argues for a higher cutoff and higher specificity.
Why prevalence changes everything
Sensitivity and specificity are fixed properties of a test. They do not change depending on who walks through the door. But the number that a patient actually cares about, the odds that a positive result means real disease, depends just as much on how common that disease is in the population being tested. Statisticians call this predictive value, and it is the concept behind the opening example.
Run that same 99 percent sensitive, 99 percent specific test on a population where the disease is common instead of rare, a specialty clinic full of patients already referred for concerning symptoms, for example, and a positive result becomes far more likely to be real. The test has not changed. Only the starting odds have. This is why the same test can serve two very different roles: a blunt, first pass tool in a low prevalence screening setting, and a far more trustworthy confirmatory tool once a patient’s history and symptoms have already raised the odds of disease before the test is even run.
Sensitivity and specificity describe the test. Predictive value describes the patient in front of you, and it depends on both.
Key Takeaways
Sensitivity is the percentage of true cases a test correctly identifies as positive; specificity is the percentage of true negatives it correctly identifies as negative.
Many tests rely on a chosen cutoff, and moving that cutoff trades sensitivity for specificity in a direct tradeoff with no threshold that maximizes both.
Sensitivity and specificity are fixed properties of a test and do not depend on how common the disease is in the population being tested.
Predictive value, what a positive or negative result actually means for an individual patient, depends heavily on disease prevalence, not just on the test’s accuracy.
An accurate test can still produce mostly false positives when it is applied to a population where the disease is rare.
Up Next:
Next, we will put this tradeoff on a graph with the ROC curve, a way to visualize sensitivity and specificity across every possible cutoff at once, and see how it is used to compare one test against another.