Research Question: Can we predict incident heart failure using routinely collected data ?
Objective – To develop and internally validate a prognostic machine learning model that predicts the risk of first heart-failure hospitalization at 1-, 3-, and 5-year horizons among UK Biobank participants who are free of heart failure at baseline.
Scientific rationale:
In the UK, heart failure leads to significant healthcare utilization, with roughly one million hospital bed-days each year, equivalent to about 2% of the NHS total and is the cause of around 5% of all emergency admissions. In the United States, heart failure spending in 2020 was estimated at $32 billion in direct costs and $14 billion in indirect costs. Globally, the economic burden was estimated at $284 billion across 179 countries. These figures motivated us to develop heart failure risk prediction model.