Collider Project allows students to collaborate with an industry partner on data blending.  

By Paul Nicolaus

March 18, 2016 | “I love it when creative people think about problems in new and unexpected ways,” says Nicholas Marko, chief data officer with Geisinger Health System. So it comes as no surprise that he’s currently engrossed in a collaborative project that harnesses the innovative minds of college students to tackle real-world data dilemmas.

During Phase 1 of the UC Berkeley and Geisinger Health Collider Project, which took place during the Fall 2015 semester, student teams were asked to address one of three problems related to multidisciplinary data analysis of obesity, heart/lung failure and mood disorder.

“The first phase was largely about finding data sources and being creative,” Marko said. “One of the big things about data science that I think a lot of people under-appreciate is how important it is to be creative and think outside the box, to bring novel hypotheses in, and to look at data from different perspectives.”

Upon culmination of Phase 1, all reports were evaluated, and the competition was whittled down to five teams who progressed to Phase 2. The remaining teams are busy pulling data together, writing code, performing computations, and testing their hypotheses.

Between Research and Real-World

The project originated with a conversation between Marko and Roberto Zicari, a professor of database and information systems at Goethe University in Germany who has also worked at UC Berkeley.

Already familiar with UC Berkeley’s Sutardja Center for Entrepreneurship & Technology (SCET), Zicari told Marko of the Center’s Innovation Colliders that combine education, research, and diverse individuals—such as students, researchers, investors, entrepreneurs, and industry leaders—to come up with fresh ideas.

After some initial brainstorming, the two reached out to Ikhlaq Sidhu, chief scientist and founding director of SCET, who said the involvement of an industry stakeholder like Geisinger places this particular Innovation Collider project at a unique juncture.

“If it was any closer to an actual industry situation, I think it would not be in the university; it would literally be at the company or at the outside institution,” Sidhu said. “We are right at the boundary between the research that’s done at a university and what is applicable in the industry.”

Designed for Diversity

While healthcare data has traditionally been collected by asking patients to complete forms during visits or tracking procedures and treatments, what hasn’t been thought about nearly as much is what occurs when patients walk outside the walls of a hospital or clinic, Marko explained.

It’s a crucial consideration because a whole family of diseases have a behavioral component. “There’s a whole world of data out there that healthcare hasn’t been tapping into,” he said, “so we’re very interested in accessing that space and trying to figure out how to use it to help us do the best possible job that we can for the people that we take care of.”

While a wealth of data exists, the real challenge comes along when attempting to blend data generated in different places about differing sets of patients in order to extract something meaningful. “The idea is we’re looking for novel ways to put disparate data sources together and still extract more knowledge from the two different data sources than you could extract from any one independently,” Marko said.

In order to boost innovation, the goal was to structure a challenge that—while data and information centric—wouldn’t be overly limiting in terms of the type of student who could get involved and contribute meaningfully.

“We ended up getting this very diverse set of people,” noted Sidhu. Included in that mix were pairings from Frankfurt, Germany; UC Berkeley; and Stanford representing disciplines ranging from math and computer science to psychology and the biological sciences. Students also ranged from undergraduates to Ph. D. candidates.

Everyone Wins

As the project comes to an end in April, student teams will submit a 5 to 15 page document, and each of the teams will present a brief summary of their findings. Based on the details of those final reports and related presentations, one team will be awarded a paid summer internship at Geisinger Health.

From Marko’s vantage point, however, there’s a whole lot more than just one victorious team with this type of endeavor. “I think this project is just an all-around win,” he said. “It’s good for the students, it’s good for the patients, it’s good for us here at Geisinger, and it challenges all of us in novel ways.”

Editor's Note: Nicholas Marko will be part of the plenary panel at the 2016 Medical Informatics World Conference in Boston, April 5-6.