“The Power of Prevalence..”

A serious disease is not much fun for the patient, their doctor, or even the government funding the treatment. However having a lot of the one disease in a population can actually favour laboratarians.

Let me explain….

Let’s say you have a test for Disease X which has 99% sensitivity (the proportion of patients with the disease who have a positive result) and 99% specificity (the proportion of people without the disease who have a negative result). You then apply that test to a population which has a prevalence of Disease X of 1%. A hypothetical 2×2 table using these figures on a 10,000 population is demonstrated below.

  Disease X No Disease X  
+ve test 99 99  
-ve test 1 9801  
  100 9900 10000

 Thus the Positive Predictive Value (PPV) of the test (the proportion with a positive result who truly have the disease) is 50% (99/198). Half of the positive tests are false positives.

 

Now let’s keep the test sensitivity and specificity the same but increase the prevalence of Disease X to 10%. Here is a 2×2 table modelled on these figures for a 10,000 population.

  Disease X No Disease X  
+ve test 990 90  
-ve test 10 8910  
  1000 9000 10000

With the prevalence increasing to 10% the positive predictive value (PPV) of the test (the proportion with a positive result who truly have the disease) has increased to 91.6%, and the number of false positive results has decreased to 8.4%.

Now if disease X was something like HIV or Gonorrhoea, you would be keen that your laboratory produced as few false positive results as possible (in the interest of your career!). Thus if diagnosing an important disease with low prevalence it is essential you have a test with as high a specificity as possible.

Note that sensitivity and specificity are functions of the test and doesn’t change wherever you use the test. Positive and negative predictive values can only be calculated when you know what population you are applying the test to. Test manufacturers should not be giving you positive and negative predictive values. It is nothing to do with the manufacturers. That is the job of the laboratarians depending on who they are using the test on.

Whichever field of laboratory medicine you are involved with, understanding the concepts of sensitivity, specificity, positive and negative predictive value are critical. If I was an examiner (thankfully not), I wouldn’t care if you knew all the steps of the Krebs cycle, but I would care that you had an understanding of the above….

Michael

 p.s. Here is a link to a powerpoint on this topic produced by the WHO

 

Leave a Reply

Your email address will not be published. Required fields are marked *