Carl Heneghan

I knew you'd come in at that point.

I thought you'd come in.

One of the things to think about is when we talk about quality improvement, a lot of people say, is quality improvement in practice really difficult to understand what's going on?

And this design uses cluster-based design.

So it takes practices...

and randomizes the practice as opposed to randomizing an individual so you go to a whole practice and that's the cluster and what you're going to do is with because this is a service instead of saying we're going to give half the practices the intervention and half the nought you say at some point in the future all the practices are going to get the intervention

So you start at the beginning with everybody as a control, no intervention.

And say you've got 40 practices.

You randomise the practices in a stepped wedge design.

So, for instance, five might get it at the first month, five the next month, five the next month, five after.

Correct.

I like what you're saying.

So it's a bit like a staggered start.

And at some point, everybody's going to be doing the intervention.

And then you can evaluate the whole intervention.

And what it is, is it's really helpful because what it tries to do is say, well, all of these practices are clusters and you're trying to also account for there may be some features in the cluster that might explain the effect.

You know, it might be you're in a rural versus a city and in there the midwives practice differently and they're more informed and educated.

And you use the control arm and the intervention arm and evaluate both of the arms.

And what did they show?

They showed that actually there was an effect on picking up small for gestational age babies, but no effect on severe adverse perinatal outcomes, which is what you're really interested.