Construction and evaluation of bayesian networks related to the specific health checkup and guidance on metabolic syndrome

Abstract

Metabolic syndrome has become a significant problem worldwide, and health checkups and guidance aimed at preventing this condition were initiated in 2008 in Japan. Through this guidance, people considered at high risk of developing metabolic syndrome are expected to be made aware of their own problems in terms of their daily lifestyle choices and to improve their daily life behaviors by themselves. To this end, the instructors should be able to supply satisfactory and evidence-based information for these subjects. In order to support this large undertaking from the point of information technology, we here introduce our novel ideas based on data mining technology using Bayesian networks. The Bayesian network has emerged in recent years as a powerful technique for handling uncertainty in complex domains, and it is expected to represent an appropriate method for the health checkup domain, where medical knowledge is required for the analysis of the results. In this study, we constructed Bayesian networks connecting the findings from a physical examination and questionnaire on daily lifestyle choices, and evaluated the relationship between them. We applied these network models to the field data of 5423 subjects. The proposed method was found to provide good performance, and its usefulness was revealed by evaluating the level of change of the responses to the questionnaire.