The next Royal Statistical Society Medical Section meeting on
CLUSTER RANDOMISED TRIALS will take place TOMORROW, Tuesday 27 June,
starting 2:30pm at the RSS (Tea 4pm). Directions to the RSS can be found at
http://www.rss.org.uk
Members and non-members are all welcome.
IZZY BUTCHER (Napier University)
The rise of the cluster randomised trial
The topic of cluster randomised trials is one of growing interest to
statisticians and a large number of papers around the subject have been
published in recent years. Over the past decade such trials have become
increasingly important in the evaluation of healthcare and educational
strategies, and more recently in the process of evidence-based policy
making. The speaker will give a brief introduction to some of the issues
arising from this type of experiment, illustrated with examples of their
application in different fields.
MIKE CAMPBELL (University of Sheffield)
Cluster randomised trials in primary care
Many of the initial applications of cluster randomised trials in health were
concerned with health promotion. However, there is increasing interest in
research in Primary Care, and the cluster randomised trial is being
increasingly used to evaluate health interventions. We will contrast primary
care trials with health promotion trials and illustrate some of the
practical problems in this area. Typical problems include guessing a
suitable intrapractice correlation coefficient for sample size estimation,
the use of baseline data and the meaning of informed consent. There has been
a notable lack of success in trials to educate Primary Care professionals
and reasons for this will be discussed.
SIMON THOMPSON (MRC Biostatistics Unit)
A Bayesian approach to the analysis of cluster randomised trials
We develop a Bayesian approach to the analysis of cluster randomised trials
using a hierarchical model, and compare this to a classical multilevel model
approach. The Bayesian approach has some notable advantages, permitting
informative priors based on information for intra-cluster correlation
coefficients, providing credible intervals for differences between variance
components, and allowing investigation of robustness of results to the usual
normality assumption of the cluster-level random effects. These issues are
illustrated by analysis of data from a trial in which general practices were
randomised.
Dr Keith R. Abrams
Senior Lecturer in Medical Statistics
Department of Epidemiology & Public Health
University of Leicester
22-28 Princess Road West
Leicester, U.K.
LE1 6TP
Tel +44 (0)116 252 3217
FAX +44 (0)116 252 3272
e-mail [log in to unmask]
WWW http://www.prw.le.ac.uk/epidemio/
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|