A machine learning platform accurately predicted mortality in patients with heart disease, outperforming models created by medical experts, according to an Aug. 31 study in PLOS One.
In collaboration with colleagues at the Farr Institute of Health Informatics Research and University College London Hospitals NHS Foundation Trust, a team from the Francis Crick Institute in London used electronic health data from more than 80,000 patients to create the new algorithm.
More than 600 variables were used by the machine learning algorithms to self-teach. That platform was compared to models designed by experts—based on coronary artery disease—which made predictions based on 27 variables including age, gender and chest pains.
The novel machine learning algorithms not only predicted cardiac prognoses more accurately than prior models, it brought to light variables cardiologists had never considered.
“Along with factors like age and whether or not a patient smoked, our models pulled out a home visit from their general practitioner as a good predictor of patient mortality,” said Andrew Steel, first author of the study, in a statement. “Home visits are not something a cardiologist might say is important in the biology of heart disease, but perhaps a good indication that the patient is too unwell to make it to the doctor themselves, and a useful variable to help the model make accurate predictions.”
Steele believes the proof-of-principle study results could be used to create a similar model that may arrive in clinics sooner rather than later.
“It won’t be long before doctors are routinely using these sorts of tools in the clinic to make better diagnoses and prognoses, which can help them decide the best ways to care for their patients,” Steele added.