Non-identifiability of autochthonous chikungunya risk from surveillance data, Japan, 2011–2025

Japan notifies imported chikungunya but has never detected autochthonous chikungunya, although Aedes albopictus transmits both chikungunya and dengue and Japan experienced an autochthonous dengue outbreak in 2014. This absence is often read as reassuring. We reconstructed weekly, prefecture-level dengue, chikungunya, and Zika notifications from 1,357 national surveillance bulletins (1999–2026) and combined a temperature-suitability index with notification-based introduction proxies in a branching-process model with an explicit cluster-detection process. A descriptive temperature-suitability–notification overlap index was highest in metropolitan prefectures and in late summer; this geographic ordering was stable across prespecified proxy variants, population and per-capita proxies, and prefecture-specific seasonality, but was driven predominantly by the notification proxy, not temperature. Zero detected chikungunya clusters over approximately 14.9 surveillance-years placed a one-sided 95% upper bound of about 0.20/year on the detected-cluster rate; the data constrained only the product of introduction pressure and per-introduction detection probability, identifying neither. Chikungunya risk relative to dengue was therefore not identifiable from surveillance counts alone. Absolute probabilities and importation multipliers were calibrated to two events and are scenarios, not forecasts. The absence of detected autochthonous chikungunya does not, by itself, establish low underlying transmission risk when ascertainment is unknown; resolving this non-identification requires disease-specific diagnostic and case-level data.