Laure Wynants

I'm Laure Veenans.

I work in the Netherlands at Maastricht University.

We are not medical doctors.

We are statisticians by training.

And we mostly work in the field of prediction modeling, so diagnostic and prognostic models.

I was on holiday.

I was enjoying a very nice holiday.

And before I left, there was no talk of COVID in Europe.

It was something that happened far away and right when I got back home.

the first cases started appearing in Belgium and in our neighboring countries.

And my anxieties sort of grew with the news that was coming in and what I was seeing on Twitter.

And then at one point one morning, I remember very clearly, I saw a tweet of someone trying to build a prediction model or proposing to build a prediction model for COVID.

And all alarm bells in my head started going off because I saw so many problems.

And that is how this project actually got started then.

I contacted Maarten and that's where it took off.

In general, when we build a prediction model, we usually like consecutive series of patients.

So people

who are for a diagnostic model, people who are suspected of having the disease.

And it should reflect really the situation in which you are going to apply your diagnostic model.

So as soon as you get into a case control sampling or just these case series for building a diagnostic or prognostic model, that won't work.