Using an Artificial Neural Networks to Predict Utilization Rate of Hospital District among Children with Respiratory Disease

Abstract

This study aims to generate the equation for predicting the utilization rate of the local hospital in children aged 1-10 years old with respiratory disease by an artificial neural networks (ANNs). Using ANNs, the prediction equation was simulated by using retrospectively collected data including nine environmental and health indicators. The eight indicators were taken from the existing environmental and health data from 2010 to 2014 of the Air Quality and Sound Management Office in Thailand report and the Health Promoting Hospital District in Chaloem PhraKiat District, respectively. The factor related to the activity outside their home was obtained from the literature. The results showed that only five indicators were the predictors for estimating the number of children who visited the hospital district with respiratory diseases. These indicators included season, distance, number of children with respiratory disease, %weight for height and PM10 levels. The prediction equation had the correlation coefficient about 63.20%, accuracy with rate of 86.33% and relative error 13.88%. However, the equation was able to be performed for predicting the low respiratory disease rate, with sensitivity, precision and confidence rates of 92.38%, 89.57% and 77.20%, respectively. It is concluded that the ANNs can be applied to select the environmental and health indicators that were associated with the respiratory disease; and can generate the equation from the existing environmental and health datasets to estimate the number of children with respiratory disease who come to be treated in a hospital district.