(1044) A Machine Learning Model for Prediction of Angiotensin II Use in Septic Patients

Introduction: Prompt initiation of treatment is associated with reduced mortality in patients with septic shock. Angiotensin II is a vasoconstrictor that is used as an adjunctive agent to increase blood pressure in patients with septic shock who have inadequate response to vasopressor therapy. This study explores the use of machine learning techniques to predict angiotensin II use at the time of ICU admission.

Methods: This multi-site, single health system, retrospective study included adult patients diagnosed with sepsis and admitted to an intensive care unit (ICU) from September 1, 2021 to August 31, 2023. Patients with age, height, weight, and Sequential Organ Failure Assessment (SOFA) scores recorded within 24 hours of admission were included in the study. A logistic regression model was developed using R programming language with tidyverse, caret, caTools, and pROC libraries. The primary outcome was predictive capability of angiotensin II use.

Results: A total of 1715 patients were included in this study, 70 of which received angiotensin II. Patients were randomly split for model development into a training and testing cohort of 1039 and 352 respective patients. A separate validation cohort of 324 patients were used to gauge model performance. Model was able to predict angiotensin II use with an area under the receiver operator curve of 0.8626. When probability threshold was set to 6.89%, predictive sensitivity was 0.880 and specificity was 0.726. There was no difference in mortality observed between patients receiving angiotensin II and those who did not.

Conclusions: This study demonstrates the capability of a machine learning algorithm at predicting use of angiotensin II using SOFA score and patient demographics obtained at ICU admission. This model may be used for targeted sepsis treatment and earlier initiation of therapy. Future studies are needed to evaluate patient outcomes from early intervention.