Abstract
Metabolic syndrome has become a significant public health problem worldwide, and Specific Health Checkup and Guidance (SHCG) on this syndrome began for the people aged 40 to 74 in Japan in 2008. To support this large undertaking with information technology, we have introduced ideas based on the Bayesian estimation in data mining technology and proposed a Bayesian network scheme connecting the information from physical examinations and daily lifestyle questionnaires. In this paper, we focused on data from physical examinations and questionnaires for the 2 years before SHCG was initiated to measure interannual transition rates between the support levels assigned to examinees without SHCG interventions. We introduced a novel 4-bit representation with 16 states, treating body shape, blood lipids, blood glucose, and blood pressure as equal binary factors, and analyzed relationships between the support level, physical examination, and daily lifestyle questionnaire. Our results demonstrate an improvement in our previous Bayesian network scheme to include stratified support levels and 16 complementary states to measure health status. Furthermore, we evaluated the time lag for daily lifestyle measurements to affect physical examination values through the interannual combination of the data. The time lag was found to be approximately 1 year, and assessing daily lifestyle with consideration for this point is important for SHCG. In addition, we applied this Bayesian network to a case and showed its utility in allowing an examinee to improve his or her lifestyle by demonstrating individual predictions. Through the efforts described above, we confirmed that the Bayesian network for SHCG has the potential to be an effective support tool for health guidance regarding metabolic syndrome.
