Submitted by Administrator on Wed, 22/05/2019 - 08:23
An interdisciplinary team led by Prof Mihaela van der Schaar (DAMTP) has discovered that a person's self-reported walking speed predicts cardiovascular events over the next five years.
The study used a novel machine learning method called AutoPrognosis to analyse data from the UK Biobank cohort of over 420,000 volunteers.
Of the 4,801 cardiovascular events recorded over 5 years, AutoPrognosis was able to correctly predict 368 more events compared to the standard Framingham risk score, recommended in US guidelines.
Machine learning also identified new important predictors not usually considered in existing risk prediction models, such as an individuals' usual walking pace and their self-reported overall health rating.
Furthermore, the model improved risk prediction in potentially relevant sub-populations, such as in those with diabetes, where the prediction of future heart attacks and strokes often fails.
Work is ongoing to incorporate genetic, proteomic and blood biomarkers into the machine learning model.
The study was published in PLoS One.
