Analysing healthcare coverage with data mining techniques
Mohammad Hossein Tekieh, Bijan Raahemi, Seyed Abdolmotalleb Izad Shenas · International Journal of Society Systems Science · 2015
In this study, we analyse healthcare coverage and propose an approach to identify significant factors in healthcare coverage by building several data mining models including decision trees, neural networks, and k-means. The results of this study can help policy makers, governments, and insurance companies to enhance the quality of health services in various directions. In our proposed approach, the models are built and tested on a real dataset with 98,175 records retrieved from the medical expenditure panel survey databases including nationally representative samples of the USA. We prepare a dataset with 26,932 records and 23 variables first by preprocessing the data. We then build 44 predictive models with accuracies between 76% and 81% using IBM SPSS Modeller. We observe that the decision tree models provide higher accuracy than neural networks. Moreover, we discover four significant factors in healthcare coverage, namely 'access to care', 'age', 'poverty level of family', and 'race/ethnicity'. These results are verified and compared with the outcomes of our clustering models, and also with the related literature and data statistics.