Heart disease is the leading cause of death in the United States and claims the lives of nearly 20 million people around the world every year [1]. Doctors rely on guidelines set out by the American Heart Association (AHA) to aid in identifying risk factors essential to heart attack prevention efforts. These guidelines are based on factors that include blood pressure, cholesterol level, smoking, and age [1]. A recent study carried out by epidemiologists and computer engineers at the University of Nottingham found that artificial intelligence algorithms correctly predicted up to 7.6% more heart attacks than doctors [2].
The study compared AHA guidelines to four types of machine-learning algorithms developed by the researchers [3]. The key to the success of the algorithms was the fact that they tested a large number of patient variables over a long period of time. The algorithms used data from over 10 years from 295,000 patients to search for possible patterns and to build guidelines of their own [1]. Due to the breadth of available data, the machine-learning mechanisms were able to take into account variables such as medication interaction, kidney disease, ethnicity, and arthritis, all of which are not mentioned in AHA guidelines. [3] All four artificial intelligence methods predicted between 7.4 and 7.6% more heart attacks than doctors using the standard method [1]. The algorithms also predicted up to 1.6% fewer false positives, or wrongly assigned “high-risk” labels, which can lead to the administration of strong medications with severe side effects [4]. Of the 83,000 patients in the sample tested by the best algorithm, the accuracy of the predictions amount to an additional 355 individuals whose lives could have been saved by prevention efforts [3]. Many of the top risk factors identified by the algorithms are not outlined by the AHA guidelines, such as severe mental illness, ethnicity, and the use of oral corticosteroids [1]. Additionally, none of the algorithms identified diabetes, a risk factor on the AHA list, among the top 10 predictors of heart attacks [3].
The early identification of risk factors is essential to preventing heart disease. The success of such algorithms hints at a future of medicine that sees further collaboration between health professionals and experts in the technology sector in order to find the most precise methods of disease prevention. When asked if technology would ever replace medical professionals, Dr. Stephen Weng, the leader of the research team at Nottingham responded, “You’ll always need doctors and nurses” [2].
References:
Weng, Stephen F., Jenna Reps, Joe Kai, Jonathan M. Garibaldi, and Nadeem Qureshi. "Can machine-learning improve cardiovascular risk prediction using routine clinical data?" PLOS One, April 4, 2017. http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0174944
Belani, Abhilasha. “AI predicts heart attacks better than doctors,” NBC News, April 27 2017, http://www.nbcnews.com/mach/science/ai-predicts-heart-attacks-better-doctors-n752011
Hutson, Matthew. “Self-taught artificial intelligence beats doctors at predicting heart attacks,” Science Magazine, April 14 2017, http://www.sciencemag.org/news/2017/04/self-taught-artificial-intelligence-beats-doctors-predicting-heart-attacks
Avery, Thomas. “AI beats doctors at predicting heart attacks,” Popular Mechanics, April 17 2017, http://www.popularmechanics.com/science/health/a26117/ai-beats-doctors-at-predicting-heart-attacks/