Computerized Score Predicts Sepsis Risk Shortly After ED Admission

By Marilynn Larkin

NEW YORK (Reuters Health) - A risk score based on data that are routinely collected when patients are admitted to emergency departments can predict sepsis risk within about half an hour, researchers say.

The risk score is called cNEWS because it builds on the National Early Warning Score (NEWS; https://bit.ly/2Ur2tJ1) that was developed by the Royal College of Physicians in London, Dr. Mohammed A. Mohammed of the University of Bradford, UK, said in an email to Reuters Health.

"We have demonstrated that it is feasible to get the hospital computer to automatically provide a reasonably accurate risk of sepsis as soon as the patients' vital signs are recorded (usually within 30 mins of admission) without any additional data collection or calculation burden on busy clinicians," he said.

"This is important," he added, "because each hour of delay in sepsis treatment is associated with a 7% reduction in survival, and treatment delays are not uncommon in hospitals."

Dr. Mohammed and colleagues constructed three cNEWS models to predict sepsis risk: NEWS alone (M0); NEWS plus patient's age and sex (M1); and M1 plus subcomponents of NEWS plus diastolic blood pressure (M2).

They used data on more than 70,000 patients discharged from EDs over a two-year period (mean age, 67; close to half, male). One dataset from York Hospital was used to develop the models and a combined data set from two hospitals and the Goole National Health Service Foundation Trust (NHS) was used for external model validation.

As reported online April 8 in CMAJ, sepsis prevalence was lower in York Hospital (4.5%) than in NHS (8.5%). The concordance statistic increased across the three models. For a NEWS score of five or higher, sensitivity increased in York Hospital from 47.24% (M0) to 50.56% (M1) to 52.69% (M2), and in NHS from 37.91% (M0) to 43.35% (M1) to 48.07% (M2).

Similar increases were seen for the positive likelihood ratio and the positive predictive value.

The team concluded that M2 (M1 plus subcomponents of NEWS plus diastolic blood pressure) was the best model to predict sepsis. "This is unsurprising," they note, "because it incorporates additional information about the patient's age and other vital signs."

Dr. Donald Richardson, Deputy Medical Director York Hospital and clinical lead for the study, told Reuters Health, "We would stress that cNEWS is intended and designed to support, not undermine, clinical decision making and can be overridden by clinical concern."

"A crucial next phase of this work," he said by email, "is to field test cNEWS by carefully engineering it into routine clinical practice to see if it does support the earlier detection and treatment of sepsis without unintended adverse consequences."

Dr. Mohammed added, "We have used the same approach to provide an automatic computer-aided estimate of the patients' risk of death. The next stage of research is to use a suite of automated computerized risk scores in routine hospital care and evaluate their impact in enhancing the quality and safety of care for patients."

Dr. Waleed Javaid, Director of Infection Prevention and Control at Mount Sinai Downtown in New York City, told Reuters Health by email, "The system in its current form is a reasonable tool for data collection. Clinically, sepsis is a dynamic situation, and real-time information to clinicians is required."

"This system and others do have potential to be the early warning system that is needed to capture a person in early sepsis, or even before a person becomes septic, to treat them and minimize harm," he concluded.

SOURCE: https://bit.ly/2UoG273

CMAJ 2019.

(c) Copyright Thomson Reuters 2019. Click For Restrictions - https://agency.reuters.com/en/copyright.html
Source: