Hospital Sepsis Risks Could Be Lessened With AI Model: Study

Artificial Intelligence process that analyzes real-time patient data could be a powerful tool in preventing hospital sepsis deaths, researchers say.

A new study indicates an innovative model using artificial intelligence (AI) technology may reduce sepsis deaths more than other protocols.

According to findings published in the journal Precision Clinical Medicine last month, using a predictive AI framework to rapidly analyze patient data as it occurs could help to improve the accuracy of sepsis diagnosis and reduce patient deaths.

Sepsis is a life-threatening infection that damages the body’s immune system and shuts down the organs. It can progress quickly and carries a high risk of death if not detected early and treated. It is one of the deadliest conditions treated in the intensive care unit (ICU), with a fatality rate between 20% and 50%.

Because of the difficulty identifying sepsis early and the high mortality rate, a significant amount of research is focused on finding accurate early detection methods, some of which fail to be helpful. To resolve this problem, researchers designed a prediction framework using AI and tested it against existing detection protocols, like APACHE-II and SOFA.

Researchers from China and Spain led by Bairong Shen and Alejandro Pazos Sierra trained the AI model on data from the eICU Collaborative Research Database, which included more than 200,000 patients.

The framework processed patient health indicators on both an hourly and daily basis in two stages. Initially, it analyzed hourly data, including vital signs and lab results. Subsequently, it processed daily aggregated data to identify long-term trends. This approach allowed it to concentrate on critical predictors of mortality, such as lactate levels, respiratory rates and coagulation markers.

The study indicated the AI framework had stronger predictive power than other protocols, with an accuracy rate ranging from 77% to 82%, depending on the dataset.

The study indicated the framework was able to capture quickly changing health data during a patient’s stay, leading to real-time identification of high-risk patients with suggestions for personalized treatments.

Researchers determined that the predictive power of the AI model surpassed conventional assessment tools, predicting the likelihood of sepsis with high precision. They indicated that the AI framework has the potential to completely transform ICU triage and speed up medical decision-making for sepsis to help save lives.

Sepsis Screening Studies

Other studies have analyzed electronic screening programs to detect sepsis in patients. A program tested by researchers from Saudi Arabia used an electronic model to screen for sepsis “red flags.” Compared to facilities that didn’t use the screening method, the electronic model reported fewer deaths from sepsis or other serious infections.

However, not all screening models are created equal. Sepsis screening guidelines implemented by the Centers for Medicare & Medicaid Services (CMS) did not lead to fewer sepsis deaths, according to data published earlier this year.