Monday, May 14, 2007

Sensitivity and Specificity

The April case study mentions both sensitivity and specificity, statistical measures you may encounter as you read medical research papers. As the case states:

"Sensitivity represents the probability of a positive result for the novel diagnostic test in people who definitely have the disease in question, as defined by the gold standard test."

"Specificity is the probability of a negative test result for the novel diagnostic test in people who definitely do not have the disease, as defined by the gold standard."
In more detail:
Sensitivity is the likelihood that a test will be positive in those who really do have disease. For example, if you took a pregnancy test, the sensitivity of that test would give you an idea how often the test will turn up positive when you really are pregnant (and so the test would be accurate).

Specificity is the likelihood of getting a negative result in those who really do not have the disease. In our pregnancy test example, knowing the specificity would let you know how often you could expect a negative pregnancy test result when you really are not pregnant (and again, the test will be accurate).

Note that both definitions make reference to a "gold standard" test - that would be the reference test for determining whether someone does or does not have a disease. Measures of sensitivity and specificity for an alternative diagnostic test (e.g. a newly available type of diagnostic test) are measured against that gold standard.

Calculating these values requires a bit of math:

Sensitivity = (number of people who have disease and test positive for it)/(number of people who have disease and test positive for it PLUS the number of people who have the disease and test negative for it)
OR
The number of people who have a disease and test positive for it divided by the total number of people with the disease who are tested
OR
How good is the test at identifying people who really do have the condition?

Specificity = (number of people who don't have the disease and test negative)/(number of people who don't have the disease and test negative PLUS number of people of who don't have disease and test positive)
OR
The number of people who don't have the disease and test negative divided by the total number of people without disease who are tested
OR
How good is the test at identifying people who really don't have the condition?

Remember, when we talk about the number of people who actually have the disease, we are using our gold standard test (such as an older, established test) and comparing it to our new test of interest (such as clinic-based pregnancy testing versus a home pregnancy test).

Sensitivity and specificity are generally expressed as percentages or decimals. For example, if a pregnancy test has a sensitivity of 0.91 (or 91%), then 91% of those who are pregnant will test positive. If our specificity were only 0.50, only 50% of those who test negative would actually not be pregnant.

For more information and tables that may make the formulas a bit more clear, visit any of the following sites:
-Glossary of EBM Terms > More on Sensitivity & Specificity: Centre for Evidence-Based Medicine
-How to read a paper: Papers that report diagnostic or screening tests
-Sensitivity and Specificity: Medical University of South Carolina, using an HIV testing example

Labels: , ,

0 Comments:

Post a Comment

Links to this post:

Create a Link

<< Home