The uploaded abstract presentation files are only visible for participants that are logged in.
Please log in to view the presentation files.
Detailed paper information
| Paper title | Development and validation of AI-model to predict refractory septic shock |
|---|---|
| Paper code | A08 |
| Authors | |
| Form of presentation | Oral abstract presentation |
| Topics |
|
| Abstract text |
Background: Sepsis is a severe, life-threatening condition with high morbidity and mortality, especially when it progresses to refractory septic shock1. This advanced stage is marked by persistent hypotension despite adequate fluid resuscitation and vasopressor therapy, significantly worsening patient outcomes2. Early identification of patients at risk for refractory septic shock is essential but remains a clinical challenge. Advances in artificial intelligence (AI) have opened new possibilities for predictive modelling in critical care, offering tools that can support timely and informed clinical decisions. Message: This study aimed to develop and validate an AI-based predictive model for refractory septic shock in patients with sepsis. The study was conducted in two phases. In the development phase, retrospective data from intensive care unit (ICU) patients with sepsis were analyzed to build an AI model using key predictive variables, including SOFA scores, vasopressor requirements, and timing of interventions. In the validation phase, the model was prospectively validated to assess its accuracy in predicting refractory septic shock. A total of 1,008 patients were included in the analysis, with 552 classified as non-survivors and 456 as survivors. Non-survivors had significantly higher SOFA scores, greater vasopressor support needs, and experienced delays in both antibiotic administration and ICU admission compared to survivors. There were no significant differences in age or weight between the groups. The model demonstrated reliable accuracy in identifying patients at high risk of developing refractory septic shock. Impact on patient care: The validated AI model enables early recognition of sepsis patients at risk for refractory septic shock, allowing clinicians to implement targeted interventions more rapidly. By improving risk stratification and promoting timely clinical responses, the model has the potential to enhance sepsis outcomes, reduce ICU resource strain, and support personalized care strategies in critically ill patients. |