Health 2.0 Show Conversations: Data-Backed Decisions in the PICU

Optimized-PICUIn case you missed it, you can now watch the March Health 2.0 Show below and on Health 2.0 TV. Health 2.0 CEO Indu Subaiya on Wednesday hosted talks with two guests, David Kale and DJ Patil, about their respective data science projects. Patil’s appearance on the Big Data-themed show was particularly fitting since he, along with Jeff Hammerbacher, coined the term “data scientist.” Patil is now a data scientist in residence at Greylock Partners, a venture capital firm.

DJ Patil on delivering information to the user 

Patil used the term to describe what he was doing at the time while working at LinkedIn. “We were trying to say, what is the new way you can take data and make it into a user-facing product?” he told the Phenomlist. LinkedIn users today see the results of Patil’s then-vision. Upon signing into their homepage they are now served up content like “people you may know” and LinkedIn’s recommended daily news.

Data science is a combination of many fields including math, statistics, pattern recognition, high performance computing and uncertainty modeling. The key is that all of these data operations result in a product. A good product is one that the user has no idea how much brain and computing power went into its creation.

“One of the earliest models of thinking about data, and probably the most sophisticated data product that I think we have is weather forecasting,” Patil said on the Health 2.0 Show. Every day millions of people base decisions about what coat they will wear, whether or not to take an umbrella, and even upcoming weekend plans all on one little icon.

However, a whole lot of work went into creating that icon. Weather forecasts involve information from satellites, different kinds of sensors, computer models and human interpretation. The aim is to translate all of the data into a form in which it is useful. “And you’ve got to do it fast because if you predict tomorrow’s weather two days from now that’s not really helpful,” Patil said.

David Kale in the Virtual PICU

David Kale, lead data scientist in the VPICU Lab at Children’s Hospital Los Angeles, knows all too well the importance of quick decision making. In the pediatric intensive care unit, a child’s condition can go from stable to critical in an instant.

The virtual lab that Kale works in is a research group focused on finding the best approaches to creating and implementing clinical decision support. Kale, a PhD student in machine learning at the University of Southern California, has been working in the VPICU for three years, and he emphasized that the lab is still mostly working on ideas, not solutions yet.

Kale’s talk with Subaiya was mostly centered around two limitations that pediatric care faces. One, there are huge disparities in care. Two, there are inherent difficulties in treating a unique case for the first time. Doctors do best when they can recall a past experience with a similar patient, but without that experience, they have to start from scratch.

For those two reasons the VPICU team believes it’s important to connect databases across hospitals. That way the hospital systems can begin to dig into disparity issues. It will also allow the VPICU to create clinical decision support tools based on much more data. These tools will essentially allow doctors to peer into the memory banks of other hospitals in order to find possible relevant past cases that might help them treat the unique case in front of them.

“Our idea is to sort of speed up that process by helping them generate better hypotheses and by extending their memory,” Kale said. “The thing that computers are really, really good at is storage and memory.”

Both Kale and Patil strongly agree on one thing when it comes to designing data tools: decision support shout be passive. In other words the computer shouldn’t tell doctors what to do, rather it should equip them with useful context and aggregate statistics from which to make a decision. “The goal should be to augment people to make them more efficient, less likely to fail,” Patil said.

To hear the whole conversation, play the video above. This episode also includes a demo from Jog of War. To read more about the winning Berlin Code-a-thon app, click here. You can visit Health 2.0 TV for related Health 2.0 show content on Big Data and Quantified Self.

Leave a Reply

Your email address will not be published. Required fields are marked *