Determining the Length of Stay in the Intensive Care Unit of Patients With Sepsis Who Underwent Hemoadsorption Using the Artificial Neural Networks Model

Abstract

Objectives: The artificial neural network (ANN) analysis was used to predict the exact staying period of patients with sepsis have undergone hemoadsorption hospitalized in the General Intensive Care Unit (ICU) of X University Hospital. Methods: ANN model was successfully developed having 21 neurons in total (11 neurons at the first level, 9 neurons at the second level, and 1 neuron at the third level) using a computer program. The created network was trained using our real data and the fermi function of the system was determined by the program and then the program was able to predict possible staying time prior to real-time. Main variables of interest: Staying periods (days), ages, number of comorbidities, hemoperfusion periods (days), blood pH, C-Reactive Protein, procalcitonin, and blood lactate levels were inputs. Results: Predicted values were plotted to real data and the determination coefficient was found to be r2= 0.802. This relation was found to be good to predict the possible staying times of patients with sepsis treated using hemoadsorption in the ICU for a better organization and reducing the total cost. Conclusion: The artificial neural network modeling and this prediction were found to be useful for predicting patients staying time at ICU