Congratulations to Martin Than and colleagues for their work being voted HDR UK’s very first Open Access Publication of the Month. They developed a way of using information from patient blood samples, to identify people at high risk of a heart attack, prior to it happening. This is valuable because as a patient living with a heart condition, if your doctor can tell from your blood samples when a heart attack may be looming, it gives them the opportunity to be able to intervene and prevent it from happening. The method they use to ‘predict the future’ in this way is based on artificial intelligence, which is completely dependent on using patient data to develop. Read on for a more technical description of the work, provided by one of the member’s of HDR UK’s Early Career Research Committee, Dr Xiaoxuan Liu:
In this study, the authors develop and validate a machine learning-driven prediction model for likelihood of myocardial infarction, using prospectively collected data from over 11,000 patients. Using pre-specified variables such as patient demographic data and dynamic changes in cardiac troponin, the authors trained a machine learning model (MI3) to predict the likelihood of myocardial infarction. This paper addresses an important clinical need: to improve our ability to risk-stratify patients with potentially life-threatening signs of a heart attack. The prediction model developed aims to move towards an individualised assessment of risk, incorporating age, gender and complex changes in serial cardiac enzyme changes in time. And with an eye to clinical implementation, the authors also provide a mock-up of the tool interface, illustrating how the healthcare provider might interact with the algorithm’s risk score output.
Out of the five papers reviewed this month, the Early Career Researcher Committee felt Than and Pickering et al best fulfilled the core criteria of the HDR UK ethos in terms of collaboration (geographical and sectoral), scale, openness and transparency, impact to patients, and diversity. The authors spanned multiple continents and disciplines, and also acknowledged the contributions of an extensive global team in the collaborative authorship list. The study’s scale and inclusiveness were equally impressive, with a large number of patients represented through prospective recruitment from 9 countries. The authors dedicated part of the methods section to promoting transparency and openness, by agreeing to make both the analysis code and the algorithm shareable to researchers upon request. In terms of impact to patients, clear consideration has been given to the use-case for such an algorithm in the real-world clinical context, as well as the benefits it may bring to patients. The authors took care to draw parallels with appropriate nomenclature in studies for derivation and validation of non-machine learning diagnostic biomarkers and have reported the study according to the well-established TRIPOD guidelines for studies for predictive modelling.
The Early Career Researcher Committee felt that, not only should Than and Pickering et al be congratulated and commended for this scientific contribution, but also for the way they have promoted collaboration, openness and inclusivity through their research.