Predicting code blue events, which include cardiac or respiratory arrest, is not an easy task for doctors. Medical professionals use a scorecard called the Modified Early Warning Score to estimate the risk for heart attack by checking vital signs like heart rate, blood pressure and temperature. Being able to predict if a patient is at high risk allows doctors and nurses to ultimately reduce arrest rates and even shorten hospital stays.
Sriram Somanchi of Carnegie Mellon University wanted to see if he could predict such events using technology.
"We had to understand what happens in Code Blue patients before they enter Code Blue," Somanchi said.
Somanchi and his team trained an algorithm with data from 133,000 patients. With these patients, doctors called Code Blue 815 times. The algorithm looked at 72 parameters in patients' medical history, including blood glucose counts, age, and vital signs. Using this information, it was able to tell, up to four hours prior to the event, that a patient was going to go into arrest. It was correct in two-thirds of incidents, while the scorecard only predicted 30 percent of Code Blue events.
The issue with using the algorithm in hospitals is that it is difficult to collect that much detail from patients. The scorecard, on the other hand, has just a few parameters to monitor.
Researchers are working toward making the algorithm more accurate. It reports a false positive in 20 percent of incidents. The team is now working to train the technology with data from other hospitals.