Presentation (CTI)

OBJECTIVES: To describe the characteristics of body size data of weight and body mass index (BMI) in two major claims databases in Japan, the insurance-based database from JMDC Inc. (JMDC) and the hospital-based database from Medical Data Vision Co., Ltd. (MDV) and demonstrate their applications to medical research.
METHODS: We examined the characteristics of body size data from JMDC and MDV and applied these data to two practical cases. For JMDC, individual BMI data obtained from annual health checkups were tracked from age 30 in 2015 for 10 years. Then, the BMI trajectories were estimated using a Bayesian hierarchical model. For MDV, the distributions of body weight for inpatient data ranging from age 0 to 20 in 2015 were modeled using the Lambda-Mu-Sigma (LMS) method.
RESULTS: The characteristics of JMDC's body size data included BMI from annual health checkup in working-age adults, allowing for long-term tracking over a median of 2.8 years. However, JMDC does not provide weight and height measurements. In our study, 20,996 patients with body size data were identified in JMDC and the mean number of measurements over the 10-year period was 7.7 per person. The Bayesian model showed an increasing trend in BMI as getting old. MDV covered all ages including pediatric and elderly and contained weight and height measurements while the measurements were limited to inpatients. In our study, 163,091 inpatients of age 0 to 20 had weight measurements in 2015 in MDV and the mean number of measurements over the 10-year period was 1.8 per person. The LMS-based curves showed trends similar to those of national growth standards in Japan.
CONCLUSIONS: We clarified the characteristics of JMDC and MDV and demonstrated data application cases that matched each characteristic. The appropriate database should be selected based on characteristics of body size data for each study objective.