New Johns Hopkins Algorithm Improves Early Sepsis Detection Without Increasing False Positives

image

Researchers at Johns Hopkins have developed an improved algorithm for predicting sepsis in patients, boasting an 85 percent alert accuracy without increasing existing false positive rates. The numbers are promising, as sepsis continues to be a major contributor to hospital mortality rates in the US and abroad. According to the CDC, 10 percent of all hospitalized patients in the US are affected by sepsis, for a total of more than 750,000 cases annually. A 2014 study published in JAMA found that nearly 50 percent of all US hospital deaths are related to sepsis.

Sepsis is an inflammatory response to an existing infection in the body, and can develop into a fatal condition notoriously fast. Left untreated, sepsis quickly intensifies into severe sepsis and septic shock. Sepsis carries a mortality rate of 15 to 30 percent, a number that increases to 40 to 60 percent in patients with severe sepsis or septic shock. Every hour that a patient with severe sepsis is left untreated, the likelihood of death increases by 10 percent, meaning that early identification and rapid response times are critical to reducing the impact that sepsis is having on hospital mortality rates.

As EHRs move patient data from paper to servers, researchers have been busy looking for ways of monitoring changes to patient conditions in real time to identify sepsis as early as possible. This effort has resulted in a number of promising early algorithms, including one developed last year that accurately identifies severe sepsis and septic shock 93 percent of the time, and maintains a 98-percent specificity, meaning that it rarely triggers a false positive. In a pilot program, this algorithm identified severely septic patients an average of four hours before clinicians. Though this is a promising step forward, this earlier algorithm focused on identifying severe sepsis, a condition that already carries a 40 to 60 percent mortality rate.

Now, researchers at Johns Hopkins have designed an algorithm that targets early detection of sepsis, rather than severe sepsis or septic shock, a condition with a much lower 15-to-30 percent mortality. Researchers are calling the algorithm TREWscore, standing for targeted real-time early warning score. The new scoring system was unveiled in a study published as the cover story of Science Translational Medicine this week. To design the algorithm, researchers retrospectively analyzed EHR data from 16,234 patients admitted to intensive care units—including medical, surgical, and cardiac units, at Beth Israel Deaconess Medical Center between 2001 and 2007. TREWScore evaluates 27 biomarkers, and identifies sepsis 85 percent of the time, without increasing false positives over current algorithms. Equally impressive, the new algorithm identifies septic patients before any organ dysfunction has occurred two-thirds of the time, a 60-percent improvement over current algorithms. 

Now, with a viable new algorithm to pilot, researchers are moving forward to determine how these algorithms will be integrated into existing clinical workflows and EHRs in a way that limits alert fatigue but improves the persistently high sepsis mortality rates currently plaguing care delivery.


Enjoy HIStalk Connect? Sign up for update alerts, or follow us at @HIStalkConnect.

↑ Back to top

Founding Sponsors

Platinum Sponsors