Lack of access to health data said to limit potential of machine learning
As machine learning technology continues to advance at a rapid pace, providers are excited by the potential of this type of artificial intelligence to predict which patients are most at risk for clinical events that require early intervention.
However, these medical breakthroughs are being hampered by the lack of health data necessary to learn the complex patterns required to positively affect patient care.
That’s the consensus of healthcare stakeholders who gathered at Wednesday’s Machine Learning in Healthcare: Industry Applications conference in Boston to discuss the technology’s promise and challenges.
Research published earlier this month by MarketsandMarkets projected that the healthcare artificial intelligence market is expected to grow from $667.1 million in 2016 to more than $7.9 billion by 2022, a compound annual growth rate of 53 percent over the forecast period. Machine learning technology is accelerating at a rate beyond Moore’s Law, with algorithms and models doubling in capability every six months.
Among the potential applications: medical imaging, diagnostics, drug discovery, precision medicine, as well as patient data and risk analysis. In fact, a study presented this week at the American Thoracic Society International Conference in Washington showed that a machine-learning algorithm has the capability to identify hospitalized patients at risk for severe sepsis and septic shock using data from electronic health records.
According to Russ Wilcox, partner of venture capital firm Pillar, machine learning is currently benefitting from a “trifecta” of technology trends—big data (a flood of digital information that doubles every 3 years), better hardware (optimized processors and storage) and smarter algorithms.
“Ninety percent of the world’s digital information is less than two years old, and (that trend) is accelerating even faster,” Wilcox told the Machine Learning conference.
Yet, in healthcare, he lamented the fact that much of the data is trapped in silos, which is stifling machine learning’s promising applications in medicine.
“So many of the other industries are way ahead of us in terms of thinking about how to bring automation and digital tools to personalize our access” to data, said John Brownstein, chief innovation officer at Boston Children’s Hospital, who summed up the problem in healthcare as being a lack of data accessibility and quality.
On the flip side, Brownstein noted that the large consumer technology companies have access to good quality data that enables automation and machine learning, resulting in a high level of personalization.
In the public sector, artificial intelligence is not on the agenda of the Centers for Medicare and Medicare Services, even though the agency has increasingly been releasing updated healthcare data to improve transparency in the Medicare program and to provide more timely data for providers, researchers and beneficiaries, says Niall Brennan, former chief data officer at CMS.
“While we did a ton of data work and re-centered and reengineered CMS as a more data-driven organization, I’m afraid AI is so far off its radar screen that if you said AI to somebody at CMS they might think you were talking about Allen Iverson,” said Brennan.
CMS is the largest single U.S. health payer, generating enormous amounts of data. However, as Brennan pointed out, it is primarily claims data, with the agency “missing almost all of the clinical and genomic data.” Even so, he believes artificial intelligence and machine learning could give CMS the capabilities to use data in new and innovative ways and to generate actionable insights, he said.
Over the next five years, Brennan contends that one of the key challenges related to whether or not artificial intelligence and machine learning gain traction with large public payers is “translating it into something tangible that will resonate with payers and lead them to think them about realigning financial incentives” to improve patient outcomes and reduce healthcare costs.
On the provider side, he made the case that the single most important driver for spurring the adoption of AI and machine learning is the transition from fee-for-service to value-based care “because it creates incredible incentives for providers to innovate and try to provide better care at lower costs.” At the same time, Brennan warned against the hazards of “bolting innovative solutions on to a fee-for-service chassis.”