Blindness and vision loss in Japan: A spatio-temporal analysis

Objectives: To analyze the spatio-temporal patterns in the prevalence of blindness and vision loss (BVL) across Japanese prefectures from 1990 to 2019. Methods: The spatial Bayesian hierarchical regression model is used to estimate the risks and trends in the prevalence of BVL in Japanese prefectures. Three models are tested: (1) purely spatial model, (2) spatio-temporal model, and (3) spatio-temporal interaction model. Results: Overall, the purely spatial results suggest that relative risks do not have a consistent spatial pattern over time. There are a few prefectures with exceptionally high (1.5 to 4.9) and low (0.3 to 0.5) risks that appear to emerge and disappear over time. In addition, an increased risk, characterized by posterior probabilities above 0.8 (excess risk), can be seen in some prefectures in Japan in a very heterogeneous way. On the other hand, the spatio-temporal results show that the period between 2005 and 2007 seem to work as a spatial transition of BVL prevalence in Japanese prefectures. Conclusions: Overall, no evidence of a spatial pattern is found, such as a north-south divide, either in the purely spatial model or in the spatio-temporal interaction model.