To make machine learning ubiquitous in healthcare, data warehousing and analytics company Health Catalyst has launched healthcare.ai. The goal, company executives said, is to make machine learning not only pervasive, but also routine – and actionable.
To that end, Health Catalyst is embedding machine learning as a core capability across its entire product line, dubbing the move “catalyst.ai.” The company has machine learning models built into every one of its applications, making it easier to identify patients who are most likely to acquire deadly infections; finding those who may have trouble paying their medical bills; spotting possible canceled appointments before they happen; or intervening sooner with treatments for patients who are at risk for dangerous complications.
Health Catalyst executives see machine learning initiatives as driving “an orders-of-magnitude improvement in healthcare outcomes.”
“Predictive analytics powered by machine learning has truly vast potential in healthcare, but we lag other industries by several years largely because early efforts were extremely expensive one-off models requiring an army of data scientists to write and test the algorithms behind the technology,” said Dale Sanders, Health Catalyst executive vice president. “Catalyst.ai solves that problem by lowering the bar for entry and enabling data architects and analysts to become ‘citizen data scientists.’
Sanders noted Health Catalyst invested tens of millions of dollars over the last few years to create high-volume high-quality data content and embed those into clinician workflows.
As a result of employing machine learning, he said, health systems such as Indiana University Health have achieved significantly reduced central line-associated blood stream infections, often referred to as “CLABSI.”
In the past six months, the health system decreased CLABSI by 20 percent and also recorded a 30 percent drop in harm events overall, according to IUH.
The effectiveness of catalyst.ai is closely tied to Health Catalyst’s proven ability to integrate high-volume data from virtually every internal and external source available, Sanders said, adding because multiple sources of data are required for machine learning to develop the models that drive predictive analytics, the technology is more effective the more data is present.
What physicians want from machine learning, Sanders said, are suggestions for how to make a better decision. What they aren’t quite ready for are interventions the machine drives.”
“It’s not artificial intelligence to me until the machine starts acting on your behalf,” he said. “And it’s a big, big leap technically to do that.”
Sanders said the tools, the techniques and models for machine learning are becoming commoditized far sooner than he would ever have predicted.
“It’s absolutely blowing my mind how quickly non-data scientists can access incredibly complex machine-learning algorithms,” he said. “However, none of these algorithms are usable without the data underneath training them. So it’s the data content becoming the most proprietary part of machine learning.”