Cause-specific mortality and its rural–urban patterning across 1,890 Japanese municipalities before (2017–2019) and during (2020–2022) the COVID-19 pandemic: a nationwide small-area ecological study of rurality and deprivation

Abstract Background National and regional mortality summaries can conceal local inequalities. We compared cause-specific mortality and its associations with rurality and area deprivation across Japanese municipalities before and during the COVID-19 pandemic. Methods This nationwide ecological study compared 2017–2019 with 2020–2022 across 1,890 municipalities. Expected deaths applied common pre-pandemic national age-, sex-, and cause-specific mortality rates to municipal populations estimated by interpolation to 2018.5 and extrapolation to 2021.5 from the 2015 and 2020 censuses. We calculated Empirical Bayes-smoothed standardised mortality ratios (EB-SMRs) and fitted period-specific Besag–York–Mollié 2 spatial models for 34 causes and all-cause mortality, with and without area deprivation adjustment. Pooled models with period-specific spatial fields tested period × rurality interactions. Multiplicity was assessed using Benjamini–Hochberg adjustment. Results Observed deaths increased by 7.3%, while the median municipal all-cause EB-SMR changed little (100.1 to 99.5). The largest decreases in median EB-SMR occurred for asthma (− 38.2), pneumonia (− 28.5), traffic accidents (− 24.4), and tuberculosis (− 21.8). Increases occurred for senility (+ 21.2), pancreatic cancer (+ 4.7), and suicide (+ 4.6; +13.5 in women), with smaller increases for liver disease, leukaemia, and renal failure. Pandemic-period deprivation-adjusted rurality rate ratios were 1.103 (95% credible interval 1.062–1.146) for traffic accidents, 1.029 (1.012–1.046) for unintentional injuries, and 1.026 (1.008–1.045) for suicide; these associations survived multiplicity adjustment. Senility showed a positive adjusted association (1.027, 1.003–1.051) that did not survive multiplicity adjustment. No period × rurality interaction identified a detectable between-period change. Conclusions Under common pre-pandemic rates and period-appropriate populations, median all-cause relative mortality remained stable, while cause-specific changes varied and rural excesses persisted for external causes and, less robustly, for senility. The two-period comparison does not isolate pandemic effects from secular trends. Municipal monitoring can inform geographically targeted transport, emergency care, long-term care, and suicide-prevention planning.