Bookmark and Share

Saving Lives with Big Data

christian larsen  

Christian Larsen
Dean, Emory School of Medicine
President, Emory Healthcare
Physicians Group

 

What if we could look into a crystal ball and predict which ICU patients were going to experience a life-threatening event in the next few hours or days?

We could intervene, possibly preventing the event instead of treating it after the fact.

Emory University Hospital is, in fact, using a pioneering model of predictive care for its most vulnerable patients. The new system can identify patterns in physiological data and instantly alert clinicians to danger signs, using real-time streaming analytics.

Big Data, which has revolutionized the airline, food, manufacturing, retail, and countless other industries, is also transforming health care. Huge databases of coded, aggregate information from past patients (with individual identifiers removed, of course) make this leap possible. A patient's vital signs can be compared not only with his or her own physiological data from an hour or a day ago but also to data from hundreds, even thousands, of others previously treated for the same condition.

"Everybody is different, sure, but if you record enough patients and enough information about each patient, you start to see similarities within subgroups," says biomedical engineer Gari Clifford, an associate professor of biomedical informatics who helps interpret Emory's data. "Then you can find a model that will be predictive for that individual and will allow you to recognize early warning signs."

 Today's ICUs incorporate the best of design, functionality, and technology to create a nurturing, healing environment that is digitized, collaborative, and intuitive. (See the "Smarter ICU" infographic and a Q&A with Dr. Tim Buchman, director of Emory's Critical Care Center.)

Emory's prototype, created in partnership with IBM and Excel Medical Electronics, uses software to collect and analyze more than 2,000 data points per patient per second.

This metadata can illuminate patterns that precede complications like pneumonia or that indicate imminent, life-threatening conditions like septic shock. It can even let us know that Mr. Z, who is being released from the ICU today, is 20 percent less likely to return if he receives this medication or that follow-up care.

As we get better at parsing and analyzing this wealth of data, we will also improve the efficiency and effectiveness of patient care—now and in the future. That's an algorithm we can all celebrate.

Email the editor