Artificial neural networks in analysing health inequalities

This paper presents a method of applying artificial neural networks to health structure analysis for Devon, Southwest England. The network used in this study is called unsupervised probabilistic net, with which we can cluster the ward areas within Devon and then quantify the health inequalities in Devon. This paper also presents a new method to analyse the sensitivity of the indicators, by which, the role of each indicator in the formulation of a health inequality structure of Devon can be determined. The data set is provided by Devon health authority containing 252 ward areas within Devon and each ward area is composed of 12 indicators. Through this analysis, it has been found that 75% of the ward areas have similar and better health status whilst the remaining 25% of the ward areas need to be paid attention for reducing the health inequalities within Devon. The sensitivity analysis also shows that ´child poverty´, ´employment´, ´Breadline´ and ´Townsend´ play important roles in the health inequality structure of Devon.