Supervised machine learning to allocate emergency department resources in disaster situations

Abstract

Despite implementing the Hospital Emergency Plan (HEP) in Morocco that manages the massive influx of victims in disaster situations, it is still difficult for healthcare decision-makers to deal effectively with these situations, which has negative impacts on the performance of the emergency department. Thus, managers need decision systems that help save lives and reduce disabilities. This paper aims to improve the HEP by developing a model based on supervised machine learning as a support tool, to allocate the needed human and materials resources according to injuries number. We propose applying a feedforward neural network (FFNN) and backpropagation. We evaluate the proposed model performance indicators. Our framework was conducted on previous experiences between 2009 and 2019 of 4 public hospitals in Casablanca city. The application of the developed model on our data sample showed that the FFNN provided satisfactory precision for the direct implementation and gave feasible solutions according to the available resources. Allocating resources can be performed using FFNN and capitalizing from lessons learned through previous experiences. In addition, this solution can be used as an international reference to provide a new solution that is more performant taking account of the available resources.