Clustering at the GP and practice level in individually randomised studies – a review of 10 years of primary care trials

People who attend the same GP Practice or see the same GP are known to be more similar to one another than people who attend another Practice or see a different GP. This is known as a clustering effect. When we do a clinical trial, we need to take into account this clustering, or similarity, both when we decide how many people we need to recruit and when we analyse the data. If we don’t do this, we may end making incorrect judgements about whether an intervention or treatment worked.

This study aims to look back at 20 studies with data collected over 10 years from patients in GP practices, to measure how similar people are with respect to important outcome measures such as weight, depression, quality of life, etc. This will help to ensure that future studies recruit the right number of people.

The study also aims to compare the results we would get if we took these similarities into account with those we would get if we just ignored them. This will help us to understand whether this clustering tends to impact the judgements we make about whether treatments are effective or whether its impact is small enough that we can ignore it.