For Michael Draugelis, chief data scientist at Penn Medicine, analytics needs to be a strategy defined at the executive level. Leaders need to create a blueprint around key initiatives, such as readmission rates and hospital compliance, with focused goals that outline what they hope the organization can achieve.

And producing technology to support these goals requires the least effort, Draugelis explained. Rather, the biggest task for executives is “producing these predictions and delivering it in a way that we can understand the outcomes of patients – and how they’ve improved.”

“It takes this whole engineering process, or developing process, to figure out how all of these things are connected,” Draugelis said. “On the data science side, I often look for areas in the hospital with strained resources or the element they’re trying to detect is difficult to find - like a needle in the haystack.

“It’s an incredible burden for the clinical team to identify (high-risk) patients,” he added. “The medical team needs to have an initiative and targets…. And the executives need to be accountable.”

Penn Medicine integrates real-time data science into its clinical and operational practices. The system, named Penn Signals, processes real-time data streams from multiple sensors and sources across the organization. The data provides insights to the clinical team. It is delivered to the EMR, applications and secure text.

[Also: Best practices for healthcare data visualization.]

However, before Draugelis’ team can even begin to generate the results, a dialogue with the clinical team will determine the areas for improvement and where Penn Signals can make the biggest impact. He said, it’s the clinical team that “really defines the target we’re trying to modify.”

His team then translates these areas for improvement into a customized algorithms, which provide targets that “will give maturity to the data.” Penn Medicine’s data scientists are also responsible for determining the rate at which the  data is received and whether it's according to the needs of the clinical team.

According to Draugelis, data science should be used for two major functions: finding the unknown unknowns within the data and to produce real-time insights.

The final piece to successfully navigating an analytics platform? According to Draugelis, it’s about collaborating with other health systems, as health systems have a lot that can be learned together to improve patient care.

At Penn Medicine, this means utilizing an open-source technology platform, which combines data points from an individual patient with data from the Penn Data Store, a data repository of clinical patient information.

This article is part of our reporting on the Big Data and Healthcare Analytics Forum, taking place this week in San Francisco. Other stories include: Atul Butte says 'precision medicine makes doctors nervous, Mayo Clinic physician engineer says analytics need a system-baed approach, and Best practices for healthcare data visualization. 

Twitter: @JessieFDavis
Email the writer: jessica.davis@himssmedia.com
Like Healthcare IT News on Facebook and LinkedIn