2.4 Rapid Influenza Diagnostic Testing

a)      What definition of False Positive Rate did the CDC use in this table?

 

You can see that in each row the "False Positive Rate" is 1 – PPV = P(D–|Test+) = FP/(FP+TP))  This is different from the more commonly used definition, which is (1-specificity) = P(Test+|D-) = FP/(FP+TN).

 

b)      In the first row of the table, the pre-test probability is 2.5% and the PPV ranges from 39% to 56%.  What sensitivity for the RIDT did they use for the 39% PPV estimate?

 

You need to start with the formula for post-test odds given pretest odds and work backwards from there: 

Pre-test Odds × LR(+) = Post-test Odds.  So LR(+) = Post-test Odds / Pre-test Odds.  So let’s start with finding the LR(+):

Pre-test prob = 2.5% è Pre-test Odds = 2.5/(100 – 2.5) = .0256

Post-test prob = 39% è Post-test Odds = 39/(100-39) = .639

LR(+) = Post-test Odds / Pre-test Odds = 0.639/.0256 = 25

Now LR(+) = Sensitivity/(1 – Specificity), so Sensitivity = LR(+) × (1 – Specificity)

Sensitivity = 25 × 2% = 50%

 

c)      The “GOOD” specificity of 98% may be too low.  The Quidel (QuickVue) rapid antigen test has a specificity of at least 99%.{Faix, 2009 #1065}  How would using 99% instead of 98% specificity, change the LR(+)?

 

LR(+) = Sensitivity/(1 – Specificity), so changing specificity from 98% to 99% would change (1-specificity) from 2% to 1% and double the LR(+) from 25 to 50.

 

d)      Repeat the calculation of the PPV in the first row of the table using 99% instead of 98% specificity.

 

Doubling the LR would double the post-test odds calculated in part b: 2 ×  0.639 = 1.28.  So the post test probability would be 1.28/(1+1.28) = 1.28/2.28 = 56%

 

The CDC website says that when the pre-test probability of influenza is relatively low and the RIDT is positive,

 

"If an important clinical decision is affected by the test result, the RIDT result should be confirmed by a molecular assay, such as reverse transcription polymerase chain reaction (RT-PCR)."

 

e)      Assume that the “important clinical decision” is whether or not to treat with oseltamivir (Tamiflu®) and the patient is a pregnant woman at high-risk for complications.  Further assume that the RT-PCR will not further identify the strain or sensitivities of the flu virus,[1] and it will take 3 days to get the results back.  Do you agree with the CDC about confirming a positive result?  Why or why not?

 

The cost of failing to treat if she has the flu is higher than the cost of treating unnecessarily if she doesn’t, so a post test flu probability of 56% would prompt us to prescribe oseltamivir without further testing.

 

However, if the clinical decision were something (e.g., quarantining a village and setting off widespread panic) where doing it unnecessarily is worse than failing to do it when indicated, we would want to confirm a positive result.