Meet the Researcher January 2022
When scientists finish their research, they usually publish reports about it in specialised journals to inform clinicians about new health evidence and treatments and allow other researchers to evaluate their projects. Health research should, therefore, improve itself by constantly communicating what works and what doesn’t. However, these reports are not always perfect — far from it. Evidence shows that these reports are frequently incomplete and imprecise.
The UK EQUATOR Centre investigates this problem as well as alternatives to help eliminate the research waste created when science communication is suboptimal. We want to improve the way scientists communicate their studies both among themselves and to the general public.
I am a researcher in the EQUATOR team, but I am a different kind of researcher from those you usually meet at NDORMS. I am not a doctor, a health professional, or even a statistician. I have experience in journalism, editing, and research. I believe we should have more members of the public participating in our discussions about health research reporting. I’m curious about how you feel when you read about health science. Do you understand it? Do you think it could be communicated more clearly? I invite you to join my colleagues and me in this conversation!
I’m a senior researcher in medical statistics. I am working to improve the design, quality, and integrity of future medical research by advancing statistical methodology research and educating the next generation of researchers. My research interests lie in three areas, statistical methodology, applied statistics and meta-research, linked by a focus on prediction modelling for medical research.
I review the quality of reporting and methodology of published medical studies and highlight common problem areas in research. For example, I recently conducted a review of a cancer prediction model developed using machine learning or artificial intelligence (AI). Prediction models aim to predict things like someone’s chance of developing a particular disease or how well a particular treatment will work for a particular patient. The models are built using information from lots of people who have developed that disease or used that treatment. I found that many of the studies building these models poorly reported what they had done and did not use enough data when developing their models. This work has led me to develop a research study to investigate and develop guidance for researchers about how much data they should be using when developing a prediction model using machine learning.