AI Algorithm Monitors Vital Signs and Lab Results to Detect Sepsis before Symptom Onset

Sepsis, a serious blood infection, can initiate a life-threatening chain reaction throughout the body and poses a significant global health challenge. As a dysregulated host response to infection, sepsis affects over 48.9 million people annually worldwide, resulting in approximately 11 million deaths. Early detection of sepsis is crucial for effective treatment, including fluid resuscitation, antibiotic administration, and source control. However, identifying sepsis can be difficult due to its heterogeneous nature. Algorithms designed to aid early sepsis recognition could potentially enhance patient outcomes, yet there is limited research on their real-world impact.

At UC San Diego Health (San Diego, CA, USA), researchers have developed an AI model named COMPOSER to rapidly identify patients at risk of sepsis. This model leverages real-time data to predict sepsis before clear clinical signs emerge. Operating discreetly, COMPOSER continually monitors each patient from the moment they enter the emergency department, analyzing over 150 variables linked to sepsis, including lab results, vital signs, medications, demographics, and medical history. The advanced AI algorithms in COMPOSER can detect subtle patterns not immediately apparent to clinicians. By evaluating these risk factors, the system generates highly accurate sepsis predictions.

Should a patient exhibit a combination of high-risk factors for sepsis, COMPOSER alerts the nursing staff through the hospital’s electronic health record. The nurses then collaborate with physicians to decide the best course of action. If the algorithm determines that the risk patterns are more likely attributed to other conditions, it does not send an alert. Since its activation in December 2022, COMPOSER has been implemented in various in-patient units at UC San Diego Health. A study involving over 6,000 patient admissions before and after deploying COMPOSER in UC San Diego Medical Center and Jacobs Medical Center emergency departments revealed a 17% reduction in mortality, marking the first reported instance of improved patient outcomes through an AI deep-learning model.

“It is because of this AI model that our teams can provide life-saving therapy for patients quicker,” said study co-author Gabriel Wardi, MD, chief of the Division of Critical Care in the Department of Emergency Medicine at UC San Diego School of Medicine.

Related Links:
UC San Diego Health