Effects and Impact of Fluid Creep in Intensive Care Unit Patients Receiving Intravenous Fluid Therapy After One Week

The management of fluids plays a crucial role in the Intensive Care Unit (ICU) requiring specific customization to meet the needs of the individual patient with the primary goal of restoring cardiovascular stability. We know that excessive fluid administration is associated with increased mortality among hospitalized patients. In this context, we can define the term”fluid creep” as the amount of intravenous fluids used to dilute medications and to maintain catheter patency. Our study focused on fluid creep by conducting a retrospective analysis at a single center using data from the MargheritaTre(M3) database. The aim was to identify variables associated with changes in fluid creep distribution through a linear model (LM), and to implement a classification model to identify patients at risk of receiving a high quantity of creep after one week of ICU admission. The study included 4786 patients receiving an average of 1.6 liters of fluids within the first 24 hours and an average of 2.4 liters during the first week of hospitalization. For classification, we tested 8 machine learning models using grid search for hyperparameter tuning with 5-fold cross-validation. The hyperparameters were selected to attain the highest average accuracy across the 5 folds. The best machine learning prediction results were obtained with the multilayer perceptron, which showed an accuracy of 0.80 on the hidden test set in predicting patients at risk of receiving high fluid creep.