Heart failure is a worldwide epidemic and leading cause of morbidity and mortality. Though a multitude of risk factors for the development and prognosis of heart failure are well understood, their individual contribution levels remain uncharacterized, and robust algorithms to generate individual-level predictions still remain to be developed. The UK Biobank is the world’s largest dataset of donated medical information, and this study aims to use it to develop and validate risk prediction models for new-onset heart failure.
This study has two major components. First, we will use the intake assessment (clinical data, physical examination, biomarkers, genetics and lifestyle assessment) to predict the development of heart failure over the following decade. In order to classify patients and predict their outcomes, several methods will be explored, with their performance compared according to multiple clinical epidemiology metrics. Two broad modelling approaches – stacking and non-stacking machine learning models – will be considered in this study. A risk reclassification study will be performed to determine the effect of adding genetic information to the model generated from physical measurements and lifestyle alone.
The second aim of this study is to translate these models through data visualization tools deployed as part of a consumer-oriented website. Participants will be able to create a patient profile with their medical history, physical characteristics and lifestyle. This information will be used to generate an individualized risk profile using the algorithm previously described which will highlight the most important risk factors for each participant. As part of the intake questionnaire; participants will be asked questions about their personality. Once participants see their risk profile, an interactive display will ask them to make decisions about their lifestyle to reduce their future risk of heart failure. In a final step, intelligent recommender technology will be used to match risk profiles, personality and participant preferences. This algorithm will then be used to generate lifestyle recommendations for future participants.
By focusing on the predictive models that incorporate risk factors modifiable through lifestyle changes; we hope that this study will make contributions which reduce the burden of heart failure through improved awareness and prevention efforts.