Prediction Length of Stay with Neural Network Trained by Particle Swarm Optimization

Prediction Length of Stay with Neural Network Trained by Particle Swarm Optimization

Azadeh Oliyaei (Iran University of Science and Technology, Tehran, Iran) and Zahra Aghababaee (Sharif University of Technology, Tehran, Iran)
Copyright: © 2017 |Pages: 18
DOI: 10.4018/IJBDAH.2017070102

Abstract

This article describes how the prediction of the length of stay demonstrates the severity of the disease as well as the practice patterns of hospitals. Also, it helps the hospital resources management provide better services for inpatients and increase inpatients' satisfaction. In this article, an efficient model based on neural network algorithms is trained by a stochastic optimization technique called particle swarm optimization is proposed to predict the length of stay for coronary artery diseases. Real world data is used to generate the model. According to the number of missing values, some policies are considered. Since the outlier data has negative impact on the prediction model, it would be eliminated. The parameters of the proposed model are adjusted by Taguchi method. The applied algorithm evaluation result on the test data indicates that the model has the capability to predict the length of stay with 90 percent accuracy.
Article Preview

1. Introduction

Coronary Artery Disease (CAD) is the most usual type of heart disease that is seriously fatal especially in the developing countries. Patients with CAD require a cure time in hospital to get better. So, the estimation of each patient’s length of stay (LOS) is highly crucial. LOS is the number of days that each patient must spend at a hospital to become better. In this duration, it is required to use the medical facilities for each patient (Gomez & Abasolo, 2009). The cost of hospital equipment and medical facilities are really high, so the hospitals try to estimate and decrease LOS as much as possible.

The inpatient’s LOS is a significant criterion to measure the use of hospital resources. If it would be possible to predict LOS by a model accurately, this prediction can remarkably assist the hospital managers in the hospital resources planning. Furthermore, the prediction of LOS can assist in planning of hospital beds, and the demands of various hospital resources for each inpatient based on his demographic and clinical features. Also, LOS prediction helps the patients estimate the duration of their treatment, also helps the insurance institutes provide better treatment plans and services, and helps the hospital schedule the discharges and give proper consultation to the patients.(Suresh, Harish, & Radhika, 2015)In other words, a prediction model, which could estimate LOS with high efficiency, is helpful; it is a proper device for health system preventing the overlaps and helps provide better services and management, and plans properly to use optimally the resources of hospital in which, for example, it is possible to estimate when each hospital bed is available.

The LOS of various diseases, because of their different demographic and clinical features, is different. Providing that a model could estimate the LOS by considering the demographic and clinical features for each patient, it is possible to estimate the use of hospital resources for that patient, so that it is helpful in optimal use of hospital resources (Chang et al., 2002; Jiang, Qu, & Davis; Lim & Tongkumchum, 2009). On the other hand, precise detection and estimation of the amount of hospital resources based on each patient’s demographic and clinical feature is highly crucial.

Various researches to predict the LOS have done to find a model to estimate the LOS of diseases with high precision. Wren et al. consider using hospital resources for inpatients suffer from lung cancer. They estimate the required resources by considering the patients’ demographic and clinical features and the medical trace of their disease. The result shows that use of decision tree and neural network combination results the more accurate prediction in compared to using regression (Nouaouri, Samet, & Allaui, 2015). Nouaouri et al. propose to apply data mining techniques to predict the LOS and introduce the Evidential Length of Stay Prediction Algorithm (ELOSA) that allows the prediction of the length of stay of a new patient (Wren, Sharkey, & Dy, 2015). Chung et al. mention the influential factors on LOS. They try to detect the features which increase LOS of psychiatric inpatients among all data (Chunga et al., 2010).

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 5: 2 Issues (2020): Forthcoming, Available for Pre-Order
Volume 4: 2 Issues (2019): Forthcoming, Available for Pre-Order
Volume 3: 2 Issues (2018)
Volume 2: 2 Issues (2017)
Volume 1: 1 Issue (2016)
View Complete Journal Contents Listing