Dynamic Chiropractic

Dynamic Chiropractic Facebook Twitter
Dynamic Chiropractic
Find
Advanced Search
Wellness Blog
Dynamic Chiropractic
Dynamic Chiropractic PracticeINSIGHTS
Current Graphic
Facebook

Even more about more on diagnostic testing

It’s like a sore spot I can’t stop poking! If you’ve been following this blog you will have read a number of submissions by my colleague Dana Lawrence and myself on the shortcomings of Sensitivity and Specificity in clinical decision making. From a clinician’s point of view you really need to use likelihood ratios. Likelihood ratios move you from a pretest to a posttest probability. You will find more information about moving from pretest to posttest probability in my previous blog called “How Clinicians Think.”

As Dr. Lawrence pointed out in his blog “And Even More on Diagnostic Testing,” sensitivity and specificity will vary in their interpretation in relation to the prevalence of the disorder. For example, in a population where there are 100 people with the disorder and 1000 people without, a test with a sensitivity of 90 and a specificity of 91 will be positive in 90 individual with the disorder and 90 without. Not helpful. A negative test, though, is likely to be correct. (SNOUT works). When the prevalence goes up, false positives will go down, but the opposite is true when prevalence goes down.

The sensitivity and specificity of a test does not change in populations with different prevalence of a disorder even though the interpretation of a positive test does change. You may think that using something like positive and negative predictive values may be more helpful because they are very sensitive to changes in prevalence. However, they are so sensitive to prevalence that unless the prevalence in your practice is exactly the same as in the population where the values were obtained the numbers are essentially worthless.

So now that we’ve listed out all the problems with interpretation what is a clinician to do?
Use likelihood ratios.

Likelihood ratios take into account both those with the disorder and without the disorder. The likelihood ratio numbers don’t change with differing prevalence but the application does. Likelihood ratios move you from a pretest probably to a posttest probably. The pretest probability is strongly influenced by the prevalence. Using likelihood ratios is an easier and more systematic way of looking at diagnostic testing. I’ve heard it said multiple times “Likelihood ratios are the clinician’s friend.”

Leave a Reply