A model of good research practice in clinical prediction

By Alexander Gibson, AusHSI PhD Scholar

Throughout my life, science has always been an intense interest of mine. In 2017, a BBC documentary “An Hour to Save Your Life” sparked my passion for clinical sciences, exploring the life-or-death decisions facing doctors in the first critical hour of emergency care and showing the direct impact science can have on an individual.

I then embarked on a Bachelor of Biomedical Science at QUT with the plan to continue into post-graduate medicine. Unsure if I was ready, I detoured to a Master of Philosophy at QUT, quickly discovering my passion for research. Working in sports science research alongside Dr David Borg, I was introduced to statistics and meta-research, or what can be considered research on research. This ultimately led me to AusHSI, where I am now completing my PhD with some incredible supervisors, Professor Adrian Barnett, Associate Professor Nicole White and Dr David Borg.

Drawing on these two degrees and my first year of experience as a PhD student, I have tried to uphold the values of conducting research with integrity and honesty. In the early years of my studies, it became clear that your actions through science and research — no matter how small — can impact patients’ outcomes. When it comes to health and medical research, I strongly believe that one should do the right thing, regardless of whether that decision or path is less straightforward. Clinical prediction models are one area of research that has this potential to directly influence patient treatments and outcomes.

Clinical prediction models use health information, such as blood pressure or age, together with a chosen statistical method, to predict a patient’s risk of an adverse health outcome like a stroke or heart attack. Unfortunately, when these models are not developed or used appropriately, patients who need treatments may not receive them, or patients who are healthy may be exposed to unnecessary risks. With hundreds of thousands of clinical prediction models currently circulating, it is critical that they are developed appropriately with correct statistical methods and interpretations of results so they are fit for use in practice. My PhD focuses on identifying statistical and research practice issues in clinical prediction models.

Not only can problematic models negatively affect patient outcomes, but with limited resources, there are added opportunity costs. These include the additional research needed to identify problems before they can be fixed, costing time and money. However, if good statistical and research practices are followed from the beginning — doing the right thing even if it is hard — additional research may not be necessary. Highlighting these issues to other researchers, journals and clinicians will help to guide policy and expert guidelines in clinical prediction research. This is another aim of my PhD, to promote good statistical and research practices.

During my time as an early career researcher, I have been presented with incredible opportunities to work with amazing teams of people. I have worked on a clinical trial with industry partners in QUT’s Exercise and Environmental Ergonomics lab, the Association for Interdisciplinary Meta-research and Open Science annual conference and, most recently, with AusHSI and the Australian Institute of Sport. Each of these opportunities have been extraordinary and provided me with unique experiences and insights that continue to shape both me as a person and my research. I’m delighted to have found a supportive and uplifting group of fellow students and colleagues at AusHSI, where I am continuing my PhD research.