Trends and Disparities in Adult Body Mass Index Across the 47 Prefectures of Japan, 1975-2018: A Bayesian Spatiotemporal Analysis of National Household Surveys.

BACKGROUND

Among high-income countries, Japan has a low prevalence of obesity, but little is understood about subnational trends and variations in body mass index (BMI), largely owing to the lack of data from representative samples of prefectures. We aimed to examine long-term trends and distributions of adult BMI at the prefecture level in Japan from the late 1970s using a spatiotemporal model.

METHODS

We obtained cross-sectional data for 233,988 men and 261,086 women aged 20-79 years from the 44 annual National Health and Nutrition Surveys (NHNS) conducted during 1975-2018. We applied a Bayesian spatiotemporal model to estimate the annual time series of age-standardized and age-specific mean BMI by 20-year age group and sex for each of the 47 prefectures. We assessed socioeconomic inequalities in BMI across prefectures using the concentration index, according to population density.

RESULTS

In men, the age-standardized prefectural mean BMI ranged from 21.7 kg/m (95% credible interval, 21.6-21.9) to 23.1 kg/m (22.9-23.4) in 1975 and from 23.5 kg/m (23.3-23.7) to 24.8 kg/m (24.6-25.1) in 2018. In women, the age-standardized prefectural mean BMI ranged from 22.0 kg/m (21.9-22.2) to 23.4 kg/m (23.2-23.6) in 1975 and from 21.7 kg/m (21.6-22.0) to 23.5 kg/m (23.2-23.8) in 2018. Mean BMI was highest in the southernmost prefecture for most of the study period, followed by northeast prefectures. The increase in mean BMI was largest in southwest prefectures, which caught up with northeast prefectures over time. The concentration index was negative, indicating higher BMI in less-populated prefectures. Absolute values of the concentration index were greater in women than in men and increased over time.

CONCLUSIONS

There were variations in adult mean BMI across prefectures, and geographic distributions changed over time. Further national and local efforts are needed to address the rising trend in mean BMI, particularly among men in rural prefectures, and socioeconomic inequalities among women. Bayesian hierarchical modeling is useful for reconstructing long-term spatiotemporal trends of mean BMI by integrating small-sized survey samples at the prefecture level in the NHNS.