Read on Twitter
1/5 I have some difficulty understanding why there is so much discussion by @deeksj with @SMHopkins about the false -ve rate (1-sensitivity) of rapid Covid tests without mentioning prevalence or talking about PPV & NPV.
2/5 Assuming sensitivity of 60% and specificity of 99.5%, for 0.1% prevalence PPV & NPV are respectively 9.1% and 99.9% and for 1% prevalence 16.7% and 99.9%. It is not until 10% prevalence that NPV drops below 95%. Surely the very low PPV is also concerning. @deeksj @SMHopkins
3/5 A false negative test result of 0.1% may be worrying, since for every 100,000 people there will be 100 who may be infectious. But if so that should be clearly stated. Using the false -ve figure (1-sensitivity) surely creates the wrong impression? @deeksj @SMHopkins
4/4 A false +ve test result is much more likely. For population safety that could be desirable, but it means that testing in a population of 1% prevalence around 5 of 6 +ve results are false. Apologies if I have got this all wrong. If so please put me right. @deeksj @SMHopkins
5/5 More generally why the reluctance to emphasise the importance of pre-test prevalence when using these tests? Surely this is the most crucial thing when attempting to correctly interpret the test result? @deeksj @SMHopkins
