Written by Clay Smith
Why Not Test for Flu in Summer?
For my friends in the southern hemisphere, just pretend it's summer. There have been no documented cases of influenza in Tennessee in the past 3 weeks per the CDC. From what I can tell, there were only 12 cases in the entire U.S. last week. Flu prevalence is really low. If you want to keep track of flu trends, see the CDC flu tracker.
Here’s the deal. The absolute best sensitivity of our flu tests is 90% in kids, probably closer to 70-80% overall. Specificity is 98%. But you have to factor in the prevalence of disease. Per the CDC, for week 31 of 2018, there were 12 positive flu tests out of 4302 tested, for a prevalence of 0.27%.
Low Prevalence = Low PPV
Let’s do an example of 0.3% prevalence and assume 90% sensitivity and 98% specificity. If we have a sample of 1000 patients that we test for flu, 3 will actually have the disease.
When we calculate the positive predictive value, it is only 12%. PPV = TP/TP+FP = 2.7/2.7+20 = 12%. What that means is that during summer, when the prevalence of flu is extremely low, there is only a 12% chance that a positive flu test is a true positive. In other words, you will have 7 or 8 false positive flu tests for every 1 that is true. That’s not so great.
High Prevalence = High PPV
But let’s look at what happens when flu is prevalent. Sensitivity and specificity are inherent characteristics of the diagnostic test, so they remain 90 and 98, respectively. But note how the PPV changes with increasing prevalence.
In the winter, flu prevalence may be >20%. So if we have a sample of 1000 patients, 200 will have the flu.
Now when we calculate PPV, it is much better. PPV = TP/TP+FP = 180/180+16 = 92%. What that means is during the winter, when flu prevalence is high, there is a 92% chance that the positive flu test is a true positive.
What is the moral of this little tale?
In the summer, when flu prevalence is low, it is best not to test or treat for flu. A “positive” result is most likely a false positive.
In the winter, when flu prevalence is high, a positive test is almost certainly a true positive. In fact, at times, if the patient has classic symptoms during high prevalence and has a high risk condition, the CDC says to treat empirically and not test.
As for the title of this post, I actually don't have any other bad ideas to share. I just thought that made the title sound better. But I look forward to reading your bad ideas in the comments!
If you want to see a great video of this concept, take 5 minutes to indulge your inner nerd.