NIHR Statistics Group Annual Meeting – Challenges and Opportunities for Applied Statisticians
The NIHR Statistics Group held their first Annual Event, entitled “Challenges and Opportunities for applied statisticians at the University of Sheffield on the 22-23rd June.
Invited speakers included Professor Deborah Ashby (Director of the Clinical Trials Unit, Imperial College London), Dr James Wason (MRC Biostatistics Unit) Mike Bradburn (University of Sheffield Clinical Trials Research Unit) and Professor Andy Vail (Director of the Centre for Biostatistics, University of Manchester).
As well as presentations, one of the main aims of the meeting was to involve new members in the formation of the Group’s future activities and priorities. There were discussion groups on various topics aiming to identify key practical and methodological challenges that NIHR statisticians face and discuss strategies to address them. Many of these issues had been raised by Senior NIHR Faculty who attended the NIHR Statistics Group Showcase event on HMS Belfast in November 2016.
Each of the four NIHR Statistics Group Research Sections gave presentations. These described how the research sections were established by individual members of the Group in response to issues that NIHR statisticians face in their day-to-day work. The Research Sections have each taken different approaches to their particular issues and their presentations focussed on the decision-making process and the results of their work so far. The intention is that these groups might provide exemplars of how issues raised in the conference discussion groups might be taken forward. A summary of the conference discussion groups can be found on the next page.
We summarise the discussions held by the working groups below:
Routinely collected data
- Obtaining permissions can be a difficult and lengthy process
- Challenges include missing data, duplicate cases and unmatched cases when linking datasets
- The data may not be collected for research purposes
- Incorporating public involvement
- Working with other key professionals e.g. computer scientists
- Registry of projects that have used routinely collected data
- Working group to share best practices and statistical code
Studies involving biomarkers
- ‘Biomarker’ is very broad: includes ‘omic’ discovery studies, predictive/diagnostic context, surrogate for a specific outcome etc
- Impact of measurement error: variation in measurement techniques, assessing agreement with respect to same biomarker
- Statisticians are too often called upon at the end of a study – need early involvement in study design
- MRC Methodology Hubs Stratified Medicine producing guidelines for designs involving biomarkers – may help
- Organise a half day meeting on design of studies involving biomarkers. Can be organised through the Laboratory Studies Section.
Development of diagnostic prognostic tests
- Conflicting views over the splitting of data for validation
- Correct interpretation of decision curve analysis
- Data completeness, rare events, proxy outcomes
- In small data sets shrinkage and need for cross validation are issues
- Organise a half-day workshop to consider various issues arising in diagnostic/prognostic modelling.
Opportunities for statisticians to be a Principal Investigator (PI)
- There are limited opportunities to take leadership roles
- Funding is limited for methodological research
- Difficult to show direct patient benefit with methodological work
- Start small— Apply for pump-priming funds, fellowships
- Lead workstreams within larger grants to demonstrate leadership
- Find out how funding panels work: volunteer to review, join a funding panel
- Develop a mentoring relationship with a PI
Improving statistical literacy through communication of best practice
- Mismatch between expectations of clinicians/researchers and expectations of statisticians
- A lack of trust from clinicians/researchers if complex methods suggested by statistician
- NIHR Statistics Group provide help when statisticians and clinicians/researchers disagree
- NIHR Statistics Group website house existing resources on best practice and communication of risk
Career development for applied statisticians
- Pressure to undertake a PhD to progress your career
- Pressure to publish, win funding: hard for junior staff with limited exposure
- Mentoring scheme for junior statisticians to be matched with friendly, experienced statistician to provide external perspective