Clinical pathway variance prediction using artificial neural network for acute decompensated heart failure clinical pathway

Abstract

Patients in modern healthcare demand superior healthcare quality. Clinical pathways are introduced as the main tools to manage this quality. A clinical pathway is a task-oriented care plan that specifies steps to be taken for patient care. It follows the clinical course according to the specific clinical problem. During clinical pathway execution, variance or deviation from the specified care plan could occur, and may endanger the patient’s life. In this paper, a proposed framework for artificial neural networks (ANNs) in clinical pathway variance predictions is presented. This proposed research method predicts the variance that may occur during Acute Decompensated Heart Failure Clinical Pathway. By using the Artificial Neural Network, 3 variances (Dialysis, PCI, and Cardiac Catherization) are predicted from 55 input. The results show that artificial neural networks with the Levenberg-Marquadt training algorithm with a 55-27-27-1 architecture achieve the best prediction rate, with an average prediction accuracy of 87.4425% for the training dataset and 85.255% for the test dataset.