Machine learning is already outpacing humans when it comes to predicting certain illnesses like heart disease and diabetes—and those algorithms are likely to become even more accurate with the ability to factor in personal data captured on smartphones and wearables.
Engineers at Boston University are working with local hospitals like Brigham and Women’s Hospital and Boston Medical Center to manage heart disease and diabetes using algorithms that have the ability to predict hospitalizations up to a year in advance with 82% accuracy, Yannis Paschalidis, the director of the Center for Information Systems Engineering at Boston University, wrote in Harvard Business Review.
Comparatively, guidelines used by cardiologists to predict a patient’s risk of cardiovascular disease is about 56% accurate.
RELATED: Chronic conditions a major driver of healthcare spending
As Paschalidis pointed out, both of these illnesses are among the highest cost conditions. Research shows diabetes patients cost an additional $10,000 annually, and spending on diabetes, heart disease and back and neck pain cost consumers more than $275 billion in 2013, with diabetes alone accounting for more than $100 billion. Early identification could potentially save patients, providers and insurers billions.
Recognizing an opportunity for cost savings and better care, startups have been targeting diabetes—in some cases promising to reverse Type 2 diabetes in 100 million people over the next decade. Partners HealthCare recently announced a 10-year partnership with GE to develop and commercialize AI solutions specifically for the healthcare industry.
RELATED: Startup aims to reverse diabetes using tech tools
Doctors aren’t entirely convinced that machine learning and artificial intelligence are a cure-all to disease management. Last week, a Newsweek cover story proclaimed AI the savior to healthcare’s woes, but several physicians told CNBC new technology is not a cure-all. One U.K. nephrologist told the news outlet AI has “almost become an advertising tag line, which few people can define.”
But Paschalidis noted that machine learning will only become more accurate as data sets expand with access to personal data from wearables and smartphones. And as value-based payment models take hold, hospitals are likely to build analytics into their care processes.
“If we can now predict future hospitalizations with more than 80% accuracy using medical records alone, imagine what is possible if we can tap into this trove of personal data,” he said. “Recommender systems could be used to nudge us to adopt healthier eating habits and behaviors. The holy grail of heading off the emergence of conditions by keeping people well could be realized.”