Spatial heterogeneity in the burden of respiratory syncytial virus infection attributable to low and high ambient temperature in Japan: a nationwide time-stratified case-crossover analysis

Respiratory syncytial virus (RSV) infection poses a considerable disease burden worldwide. Although multiple studies have estimated the short-term effects of ambient temperature on mortality and morbidity, the burden of RSV attributable to temperature is poorly characterized. This nationwide study aimed to quantify the total burden of RSV infection in Japan attributable to non-optimum ambient temperature conditions, and to partition this burden into contributions from low and high temperatures and from moderate and extreme components of non-optimum temperature. We collected weekly time-series data on the number of reported RSV cases and meteorological variables (mean temperature, relative humidity, wind speed, and sunshine duration) for all 47 Japanese prefectures from 2006 to 2019. An extended two-stage approach was used: First, a time-stratified case-crossover design fitted with conditional quasi-Poisson regression was used to estimate prefecture-specific associations. Second, multivariate random-effects meta-analysis was used to obtain pooled estimates. The number of attributable cases were estimated for temperatures above and below the optimum minimum morbidity temperature (MMT) and further categorized as moderate or extreme using the 2.5th and 97.5th percentiles as cutpoints. Data on 1,227,012 RSV cases reported during the study period were analyzed. The pooled exposure-response relationship between temperature and RSV incidence was U-shaped, with a higher RSV incidence at both low and high temperature levels. In most prefectures the MMT was between the 40th to 50th percentiles of the local temperature distribution. Overall, 21.32% (95% empirical confidence interval [eCI] 18.93% to 22.78%) of cases were attributable to non-optimum temperature, with marked geographic heterogeneity: from 8.24% (95% eCI: - 9.06-21.12%) in Akita, to 32.63% (95% eCI: 27.86-41.78%) in Kagoshima. A greater share of temperature-attributable cases was due to moderate high temperatures (12.27%, 95% eCI: 10.95-13.19%) than to low temperatures (9.05%, 95% eCI: 7.12-10.36%), whereas the contributions of extremely high and extremely low temperatures were relatively small. High weekly mean ambient temperatures over lag 0-3 weeks were associated with a substantial attributable risk of RSV infection. Public health efforts to reduce the RSV burden should take ambient temperature into account, paying particular attention to the impact of sustained, moderately warm conditions.