Friday, November 4, 2011

Sensitivity and Specificity

On Wednesday at noon conference we briefly reviewed the concepts of sensitivity and specificity.

Every diagnostic test has a variety of characteristics which can be studied, (including sensitivity, specificity, and predictive values,) and which physicians need to understand the definitions of in order to read the medical literature fluently.  Most important test characteristics can be easily described and calculated using this two-by-two chart:


The sensitivity of a test is the percentage of patients with the disease being studied who also have a positive test result; so it's calculated as A / (A+C).  The specificity is the percentage of patients without the disease who have a negative test result; so it's calculated as D /( D+B).

The study we used to illustrate this was published in Chest this June, and was designed to determine the sensitivity and specificity of procalcitonin and CRP for differentiating between pneumonia and simple exacerbations of asthma and COPD.

I fudged the numbers a little for the demonstration, since they actually reported their results in terms of receiver operating characteristic curves (which are not as complicated as they sound), so I won't review the calculation.  However, it is intriguing that, for a CRP level of >48mg/L, they reported a sensitivity of 91% and a specificity of 93%.  This is one, small study and CRP as a test for pneumonia is certainly not ready for prime time.  However, it's worth considering that, if this does turn out to be true, a cheap blood test might be able to improve antibiotic stewardship in an era of increasing resistance.