- 27th November 2018
- Posted by: Manolis
- Category: Blockchain
At North Oaks Health System in Hammond, Louisiana, researchers have used big data from the Epic electronic health record (EHR) to develop a predictive analytics tool that has reduced sepsis mortality by 18 percent.
According to the CDC, sepsis leads to death for a quarter of a million Americans each year and causes two percent of all hospital admissions nationwide.
North Oaks implemented a Clinical Care Advisory (CCA) using tools developed by Epic, which helps notify providers of patients at high risk for sepsis based on the information in their medical records. CCA runs on a predictive model in the patient’s EHR to look for signs and symptoms of the condition.
“If left untreated, septic patients will die. So, treatment must begin within the first few hours of onset to offer the greatest chance of survival,” said Dr. Herbert Robinson, Chief Health Information Officer for North Oaks Health System.
“That’s why from the time patients arrive at our hospitals, we evaluate them carefully and then constantly monitor their clinical course to detect any developing signs or symptoms of sepsis. We do this so that appropriate and timely interventions can occur.”
As soon as a patient arrives in the emergency room, CCA scans the patient’s health information every 15 minutes and monitors over 80 data points to develop a sepsis risk score. If the patient’s score reaches a certain threshold, clinicians receive a warning that a patient is at high risk of becoming septic.
“By embedding machine learning into the existing workflow, minutes can be saved,” said Seth Hain, Epic’s director of analytics and machine learning. “As data flows into the chart, the algorithm looks for patterns indicating the early onset of sepsis. Then the results are put directly in front of clinicians, giving them the cue to intervene earlier and help patients.”
This strategy allows clinicians to recognize and evaluate high-risk patients more closely. CCA also recommends the best ways to treat sepsis, and helps providers make decisions to deliver life-saving medicine at the earliest possible time.
“Patients are now receiving antibiotics about 25 percent faster — or 30 minutes sooner — than they were before CCA. Combined with the clinical expertise of our doctors and nurses, our approach is saving lives,” Robinson said.
Researchers hope to assist in bringing the CCA tool to other organizations, helping them reduce sepsis mortality rates and improve patient outcomes.
“Our goal is to make this new tool more accessible to community health systems like North Oaks by ensuring the setup and statistical validation are efficient and easy to use,” said Hain. “North Oaks was able to set up the model quickly, allowing them to focus on the outcomes.”
Other organizations have also used predictive analytics tools and big data analytics to combat sepsis.
At Carnegie Mellon University (CMU) Heinz College, a team of researchers used a machine learning algorithm with EHR data to more accurately assign risk scores to patients and catch sepsis sooner.
“Anybody can get sepsis from a multitude of infections, at different sites, and with different comorbid profiles,” said Jeremy Weiss, MD, PhD, Assistant Professor of Health Informatics at CMU’s Heinz College.
“EHR data is very detailed. There’s a lot of time-stamped information. A lot of classical analyses don’t get to capture that kind of information. With EHRs, where this data is automatically entered, we can look at the temporal progression of disease and update our risk models more adeptly.”
In 2017, a team at the University of Pennsylvania Health System also developed a machine learning tool that predicted patients at high risk of developing sepsis a full 12 hours before the onset of the condition.
“We were hoping to identify severe sepsis or septic shock when it was early enough to intervene and before any deterioration started,” Craig Umscheid, MD, of the Hospital of the University of Pennsylvania, said at the time.
“The algorithm was able to do this. This is a breakthrough in showing that machine learning can accurately identify those at risk of severe sepsis and septic shock.”