What might the true sensitivity be for lateral flow Covid-19 tests?

Disclaimer: I am in absolutely no way an expert on Covid-19 or tests for it. This post was motivated from undergraduate teaching for medical statistics, and probably makes assumptions which are implausible.

This morning I spoke to my undergraduate medical statistics students briefly about this piece in the BMJ published a few days ago raising concerns about the possible low test sensitivity of the rapid lateral flow tests which are being used in various contexts currently in the UK. This piece cites a UK government document which says concerning the lateral flow tests that:

Results of the PHE and Oxford University Innova evaluation show it has an overall analytical sensitivity of 76.8% for all PCR-positive individuals but detects over 90% of individuals with high viral loads


and that

In field evaluations, such as Liverpool, these tests still perform effectively and detect at least 50% of all PCR positive individuals and more than 70% of individuals with higher viral loads in both symptomatic and asymptomatic individuals.


Read more

What does correlation in a Bland-Altman plot mean?

The Bland-Altman plot is a very popular approach for analysing data from a method agreement study, which I was teaching students about today. We have measurements of a sample of subjects using one measurement technique or method, and a second measurement on each, taken using a new technique or method. The objective is to see how closely the measurements from the two methods agree. If they are very similar, we could use the new method, which may be cheaper, easier or less invasive to use, rather than the old method. The Bland-Altman plot plots the pairwise differences between the measurements against their average. Sometimes one sees a correlation between the pair-wise differences and averages. What is the interpretation of such a correlation?

Read more

Using Stata’s sem to adjust for covariate measurement error

Covariate measurement error is a common issue in epidemiology. Many statistical methods have been developed for allowing for covariate measurement error over the last three decades or so. I’ve been playing around with Stata’s structural equation modelling builder, which enables one to allow for covariate measurement error using maximum likelihood for estimation. I’m still very much a beginner with structural equation models and Stata’s implementation of them, but hopefully the following YouTube video is a useful illustration of just one of the things that’s possible with them: