Bayesian Health, a machine learning startup has found in collaboration with Johns Hopkins University that a sepsis early detection tool has reduced sepsis associated deaths by approximately 18%. Sepsis is identified as a life-threatening reaction to an infection. It occurs when the immune system overreacts to an infection and creates inflammation. Blood clots and leaking blood vessels occur, and can cause damage to the body’s tissue and organs. Sepsis is extremely common, affecting approximately 1.7 million adults annually. However, despite its prevalence, over 250,000 instances resulted in death. Sepsis is often simple to overlook because its symptoms such as fever and disorientation are frequent in other illnesses. Therefore, the faster sepsis is identified, the better opportunity the patient has for survival.
The researchers at Bayesian Health and Johns Hopkins University sought to create a system capable of identifying the condition early. The result of which is the Targeted Real-Time Early Warning System. The system analyzes a patient’s medical records and clinical notes with their current symptoms and lab results to identify when they are at risk for sepsis. The system will in turn offer clinicians appropriate treatment protocols such as antibiotics
The study consisted of over 4,000 clinicians across 5 hospitals who used the AI to treat approximately 590,000 patients. Additionally, the AI system examined approximately 174,000 previous patient cases. The researchers discovered that in 82 percent of cases of sepsis, the AI system could detect sepsis 40 percent of the time. This is a remarkable improvement from previous systems whose results caught less than 20 percent of sepsis cases. In sepsis cases where the conventional techniques. Significantly, the AI system will enable clinicians to observe why the tool is choosing a specific recommendation.
The Bayesian Health team has worked in collaboration with Epic and Cerner, 2 of the largest electronic health record system providers in the U.S., to help provide the technology in other hospitals. The development team has also implemented the AI to identify patients who are at risk for pressure injuries and sudden deterioration.