Each year, to accompany our Healthcare Informatics 100 list of the largest companies in U.S. health information technology, we pick six fast-growing companies that we think could have a significant impact on the industry in the years ahead. Indeed, some of our picks have gone on to much bigger and better things. A 2013 pick, Health Catalyst, is having a major impact in the data warehouse and analytics space. Another from that year, Explorys, is now part of IBM Watson Health. One of the companies we chose in 2014, Evolent Health, is now publicly traded. This year, promising approaches include applying machine learning algorithms to clinical variation and the move to web services to address interoperability.
When people think of machine learning and healthcare, the name that likely comes to mind is IBM’s Watson, but the tech giant’s Jeopardy-winning technology is not the only game in town. Ayasdi, a Silicon Valley startup that grew out of years of research at Stanford University, has developed a series of “machine intelligence” applications for healthcare and has quietly partnered with sophisticated health systems such as Intermountain Healthcare, Mercy Health System and Mount Sinai School of Medicine.
Ayasdi (the name is a Native American word meaning “to seek”) has created clinical variation management tools that leverage both machine learning and topological data analysis (TDA) to extract insights from millions of data points. TDA brings together machine learning with statistical and geometric algorithms to create compressed representations and visual networks that allow organizations to more easily explore critical patterns in their data.
Ayasdi grew slowly out of government-funded research at Stanford. Gurjeet Singh, co-founder and CEO, developed key mathematical and machine-learning algorithms for TDA as a graduate student in Stanford’s Mathematics Department, where he was advised by Ayasdi co-founder Prof. Gunnar Carlsson.
Gurjeet Singh
Created in 2008, the company wasn’t marketed aggressively until 2013, Singh says. “It took a while for us to figure out the practical applications of our technology,” he adds. “We knew from a scientific perspective there were lots of applications, but it took us a while to figure out the commercial applications. And frankly the technology itself was hard enough to build that it took us a while before we had something to show customers.”
The path to working with healthcare providers was a little circuitous, Singh admits. Ayasdi started out doing work for government agencies, one of which was the Food & Drug Administration. “We started to work in bioinformatics, then in pharmaceuticals, and from there we got into working with providers.”
Singh used the company’s work on clinical variation management as an example of its value. As hospitals transition from volume to value, one of the biggest impediments is variation. When you are getting paid for value, standardization is a huge issue, he explains. Health systems want to wring out all the variation in their systems, so that they can determine which type of surgery is best for patients with specific co-morbidities. “A hospital system like Mercy believes that a system for discovering and operationalizing care paths could save them $50 million to $100 million over the course of three years,” he says. “Intermountain has been a leader in creating care paths. They have seen these benefits. Essentially what our software does is automate an ongoing continuous improvement process.”
The Menlo Park, Calif.-based company has grown to approximately 100 employees and has drawn $100 million in three rounds of venture investment. After flying under the radar for a few years, the company is now drawing attention. It was named one of the 10 most innovative companies in big data by Fast Company magazine in March 2015.
“You will see our name more and more in very sophisticated health systems on the provider side,” Singh says, “but also on the payer side. In general, machine learning is going to be exciting and large for healthcare. I think Watson is basically a very small embodiment of what machine learning is going to be.”
