It should come as no surprise that one of the key factors in combating sepsis is recognizing it early on. To lay the ground for further improving sepsis recognition, the Surviving Sepsis Campaign (SSC) revised its guidelines, which primarily combine the 3-hour and 6-hour bundles into a single “hour-1 bundle.” This new bundle calls for the use of five vital elements in patients suspected of sepsis within the first hour of presentation, including measuring lactate levels and re-measuring if the initial result is >2mmol/L.
While having guidelines in place help facilitate rapid recognition, health systems are also exploring the use of decision support systems to streamline the identification process. Increasingly, research suggests that computer algorithms and artificial intelligence (AI) are useful in the medical world, with some studies finding that the application of AI technology helps to spot disease early and thus improve patient outcomes.
Interested in seeing how AI can improve sepsis identification for you? Check out a few projects initiated by hospitals that have employed machine learning below.
Sentara Healthcare’s AI Tool: This organization developed an AI tool to capture patient data like body temperature, heart rate, blood tests, and past medical history in the EHR. By running this data through an algorithm, the system assesses a patient’s risk for sepsis and will send an alert to notify physicians if a patient is deemed high risk. In particular, having this kind of system in place has helped the system to reduce human error, such as missing an important risk factor.
HCA Healthcare’s SPOT Technology: Healthcare professionals and IT experts at this system utilized data from approximately 31 million annual patient care episodes to inform the Sepsis Prediction and Optimization of Therapy (SPOT) algorithm, which also uses real-time data for the rapid detection of sepsis in the inpatient setting. The tool monitors patient background data, detects critical data points, and identifies subtle changes in a patient’s condition to help providers take appropriate action. This tool has been estimated to save around 8,000 lives in the last five years, which led to the system extending its use to the ED. Called SPOT-ER, the algorithm is also used to detect life-threatening conditions, such as shock, postop complications, and early signs of deterioration.
The Children’s Hospital of Philadelphia’s Machine Learning Model: As evidence of the increasingly popular machine learning model became more available, some organizations started to apply this technology to sepsis recognition in pediatric patients as well. Using readily available EHR data, the sepsis prediction model was designed to be used for infants hospitalized in the NICU. Through the system’s machine learning models, sepsis was often identified before clinical recognition.
Integrating data with clinical experience has allowed some facilities to better serve patients. Although AI can’t always replace human actions, it can certainly assist physicians and nurses to make more informed and faster clinical decisions.
If you want access to more information on the impact of AI on sepsis outcomes, inquire about joining our community today!
