MRC BIOSTATISTICS UNIT
CAMBRIDGE
2 MRC funded PhD studentships
The MRC Biostatistics Unit has 2 PhD studentships available from October 2000,
for a period of three years. The Unit is situated in Cambridge and is a leading
centre for research in statistics with applications to the biological sciences.
Main interests of the Unit are Markov chain Monte Carlo methodology and software
development, genetic epidemiology, cancer screening, cancer epidemiology, AIDS
projections, BSE, mental health and ageing, performance indicators,
transplantation, non-standard aspects of clinical trials and meta-analysis.
Applicants should have a good first degree in mathematics or a closely related
subject, and preferably an MSc or Diploma.
Applications by post or e-mail, including a copy of the full CV and contact
details of two referees should be sent to Dr. Linda Sharples, MRC Biostatistics
Unit, Robinson Way, Cambridge, CB2 2SR.
Proposed PhD projects.
Performance indicators
In recent years there has been considerable interest in clinical effectiveness
and the use of indicators to identify centres or individuals whose performance
is, by some definition, discordant. Examination results have been used for some
time to tabulate the performance of schools although there has been some
resistance to publishing 'value-added' comparisons. There are many parallels to
the medical context: case-mix adjustment is considered of fundamental
importance, although outcomes are generally events rather than continuous
measures, and there is a similar danger of naive creation of 'league tables'.
Among the issues to be addressed in this project are:
1. Strategies for case-mix adjustment
2. Relative merits of survival methods compared with logistic regression
3. Impact of assumptions regarding missing data
4. Detecting outliers in multi-level data
5. Assessment of trends across time
6. Accommodation of ceiling effects
The student will be directly supervised by Dr. Linda Sharples, with Dr. David
Spiegelhalter and Dr. Sheila Gore acting as close advisors. In addition, close
collaborative links with the Royal College of Surgeons and UK Transplant Support
Services Authority are in place. Access to some national audits in
transplantation and surgery will be available for use as case studies.
***
Methods for missing outcome data: application to large scale trials in
psychiatry
Outcome data for all the subjects participating in a randomised trial are not
always available. For
example, outcomes in psychiatry trials often depend on assessments or
questionnaires of the patients
involved, and will be missing for those patients who do not attend or respond at
follow-up. Missing
outcome data prevent intention-to-treat analysis, and cause bias because those
subjects with missing
data are not representative of all those in the trial. Missing data can
therefore create substantial
interpretational problems for trials in this area.
Issues to be addressed in this project include:
1. Strategies for describing the pattern, or determining the mechanism, of
missing data
2. The assumptions made by methods used in practice
3. How available baseline, interim, or other concurrent outcome data can assist
analysis
4. Use of ancillary data (such as reasons for missingness)
5. Bayesian modelling approaches
6. Strategies for the use of sensitivity analyses in practice
7. Designs which can help mitigate the problems caused by missing data
The project will be based initially on two large scale pragmatic randomised
trials. A recently
completed UK trial of 700 psychotic patients investigated the effectiveness of
intensive case
management, and has substantial data at baseline, 1 and 2 years. Questionnaire
assessment scores were
missing in around 20% of subjects at 2 years. The second trial, which will be
completed in February
2001, is evaluating cognitive behavioural therapy for the prevention of
parasuicide (deliberate
self-harm) in at-risk patients. Extensive psychometric and health care cost
data are available at
baseline, 6 and 12 months.
The project will be supervised by Professor Simon Thompson, with Dr Wally Gilks
and Dr Tony Johnson
acting as advisors.
***
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|