Analyzing Patient Data Improves ICU Response Times

Being able to quickly analyze health data about a patient’s condition is obviously extremely important in an intensive care unit. At Atlanta’s Emory University Hospital, an ICU research team is taking that a step further by applying streaming analytics to big data.

“Situational awareness is what we seek and what we’re trying to use streaming analytics toward developing,” says Dr. Tim Buchman, MD, PhD, director of critical care at Emory University Hospital.

The research ICU, one of four ICUs within Emory Healthcare Hospitals, is making use of a streaming analytics platform developed by IBM that works together with an Excel Medical Electronics bedside monitor data aggregation application to provide quick analysis of big data. Clinicians are able to identify changes in a patient’s condition—including anomalies—through analysis of continual data that comes in through bedside monitors.

This process enables quicker comprehension of data and predictive analysis related to changes in a patient’s physiological patterns. It can lead to earlier identification of conditions such as heart failure, pneumonia and even sepsis, which often results from a bacterial infection through a wound.

In fact, sepsis, a life-threatening condition, is the leading cause of death in non-coronary ICUS, Of the approximately 750,000 Americans who contract sepsis annually, it’s estimated that between 28 and 50 percent die, according to the National Institute of General Medical Sciences. Earlier diagnosis and treatment could improve the survival rates for patients in an ICU.

“The revolution here is as powerful as the revolution of bringing the Internet to computing,” Buchman says.

The inability to quickly analyze patient health data is said to be a contributing factor in the 10 to 29 percent mortality rate that exists in ICUs, according to the Society of Critical Care Medicine. However, the research at Emory University Hospital is showing implications for these ICU mortality rates. “What we are seeing is prevention of failure to rescue,” Buchman says.

Given that a patient’s medical data points stream at approximately 1,000 to 2,000 points per second per patient, analyzing just 100,000 data points per second for an ICU with 100 patients can be a challenge. These data points are most meaningful when they present a patient’s health picture in real time, rather than an hour or even minutes ago.

“In health care, we need to stop depending on old information,” Buchman points outs. “The problem with patient data is that if you freeze it, it’s only good for retrospective analysis.”

Real-time analysis of an ICU patient’s health data sheds light on potential conditions that may otherwise remain hidden. Choosing the right types of data to process and display is important because it enables doctors and clinicians to deliver the “right care, right now, every time,” Buchman says.

Some 104 beds in the Emory University Hospital ICU, which usually has an 80 to 85 percent occupancy rate, are wired for real-time analysis as part of the research. The project continues in its research phase, but its implications for improved health care through predictive analytics may not yet be fully understood.

“This really is a step toward an environment where we have the same sort of intuitive displays around health care that we have with our car’s GPS,” Buchman says.