Applying a two-stage generalized synthetic control approach to quantify the heterogeneous health effects of extreme weather events: A 2018 large wildfire in California event as a case study.

Extreme weather events, including wildfires, are becoming more intense, frequent, and expansive due to climate change, thus increasing negative health outcomes. However, such effects can vary across space, time, and population subgroups, requiring methods that can handle multiple exposed units, account for time-varying confounding, and capture heterogeneous treatment effects. In this article, we proposed an approach based on staggered generalized synthetic control methods to study heterogeneous health effects, using the 2018 California wildfire season as a case study. This study aimed to estimate the effects of the November 2018 California wildfires, one of the state's deadliest and most destructive wildfire seasons, on respiratory and circulatory health, document heterogeneity in health impacts, and investigate drivers of this heterogeneity. We applied a two-stage generalized synthetic control method to compare health outcomes in exposed (from 8 November to 5 December 2018) versus unexposed counties and used random-effects meta-regression to evaluate the effect modification of county-level socioeconomic variables on the observed health effects of the November 2018 wildfires. We observed an increase in respiratory hospitalizations for most exposed counties when compared with unexposed counties, with significant increases in Fresno, San Francisco, San Joaquin, San Mateo, and Santa Clara counties. No effect on circulatory hospitalizations was observed. County-level sociodemographic characteristics seem to not modulate the effects of wildfire smoke on respiratory hospitalizations. This novel two-stage framework can be applied in broader settings to understand spatially and temporally compounded health impacts of climate hazards. We provide codes in R for reproducibility and replication purposes.