GW4 BioMed MRC DTP Studentship: Avoiding bias in medical research due to data that are Missing Not At Random (MNAR) in the Clinical Practice Research Datalink (CPRD)

This project is one of a number that are in competition for funding from the ‘GW4 BioMed MRC Doctoral Training Partnership’ which brings together the Universities of Bath, Bristol, Cardiff and Exeter to develop the next generation of biomedical researchers. Students will have access to the combined research strengths, training expertise and resources of the four research-intensive universities. The training programme has three strands: research skills; professional and career development skills; and opportunities to broaden horizons, which might include placements, research visits, public engagement internships and a mini-MD programme of bespoke clinical exposure.

Start date: 2 October 2017.

Supervisory team for this project:
Lead supervisor: Dr Alison Nightingale, [Email Address Removed] (Department of Pharmacy & Pharmacology, University of Bath)
Other team members: Dr Gavin Shaddick (Reader in Statistics, University of Bath); Prof Kate Tilling (Professor of Medical Statistics, University of Bristol); Dr Hughes (University of Bristol)

Project description:

A collaborative project with the Universities of Bath and Bristol using the Clinical Practice Research Datalink primary care database to develop methods of identifying and handling data that are Missing Not At Random (MNAR). The project offers an interdisciplinary environment and training in advanced epidemiology and statistical modelling.

The UK Clinical Practice Research Datalink (CPRD) is used for studies (>1500 to date) of disease, service delivery and benefits/harms of medicines. We are currently using the CPRD to investigate factors associated with diagnostic delay in rheumatoid arthritis (RA), psoriatic arthritis (PsA) and systemic lupus erythematosus (SLE) and factors associated with the development of PsA. In some inflammatory rheumatic diseases, tests for inflammation may be ‘normal’, potentially leading to delay in diagnosis and treatment. However, during our research into the impact of negative rheumatoid factor (RF) tests on diagnostic delay in RA we found that positive RF tests and RF test results were being preferentially recorded on the CRPD [1]. We hypothesised that negative RF tests were not recorded because they may be seen as of less clinical significance than positive tests. In our current study on risk factors for the development of PsA, there are missing data on lifestyle factors such as smoking and obesity. These data may be missing because the patient doesn’t smoke or because they are of ‘normal’ weight. Where data are missing based on certain characteristics, for example ‘normality’, they are said to be ‘Missing Not At Random’ (MNAR) and MNAR data can introduce selection bias. In the past, researchers combined missing data into a ‘missing’ category in their analyses however, there are now multiple imputation methods to fill in the missing data. Multiple imputation methods are not valid If data are MNAR because these methods are based on the assumption that the data are missing at random (i.e. missingness is not related to the true value of the variable, e.g. positive and negative diagnostic tests equally likely to be missing). However, it is difficult to determine (a) whether data really are missing (i.e. did an individual not have the test, or did they have the test but the result was not recorded) and (b) if data are MNAR. Moreover, there are no established methods to minimise the impact of the bias that they introduce. The aim of this project is to ensure that key clinical questions can be answered in the most efficient and least biased way by using CPRD data and potentially data from a second primary care research database for patients with musculoskeletal disease to (i) develop statistical methods to investigate whether data are MNAR and (ii) develop methods and statistical software programmes for the analysis of data that may be MNAR.

IMPORTANT: In order to apply for this project, you will need to complete BOTH an application to the GW4 BioMed MRC DTP for an ‘offer of funding’ AND to the University of Bath for an ‘offer to study’ (please select "PhD programme in Pharmacy & Pharmacology (full-time)"). Please quote the full project title in your application to the University of Bath.

The Research Theme Panels of the DTP will complete the selection for interviews and shortlisted applicants will be informed on 26 June 2017. Interviews will take place on 30 June 2017.

For more information on how to apply and eligibility criteria, please see http://www.gw4biomed.ac.uk/available-projects/national-productivity-investment-fund-studentships/

Project code: MRC17PHBaDTP Nightingale