Methods: We used climate variables from the Korean Meteorological Administration, air pollution data provided by Korea National Institute of Environmental Research, and SAH admission data from the National Health Insurance Service. We estimated the temperature-SAH association by applying generalized additive models (GAMs) with nonparametric smoothing functions (splines) to describe nonlinear relationships. Associations were adjusted according to humidity, barometric pressure at sea level, and air pollutants including PM10 and NO2. We estimated the threshold temperature using the piecewise-defined function. The analysis was performed for the following subgroups: gender (male or female), age (<75 years or ≥75 years), insurance type (National Health Insurance for the general population (NHI) or medical care (Medicaid) for the poor), and area (rural or urban) with climate zones based on cooling degree-days (CDD) for summer and heating degree-days (HDD) for winter. Results: We found a delayed effect, between 22-28 days, on the incidence of SAH due to hot temperature. The maximum threshold temperature during heat exposure was 31.5°C. The maximum temperature increase of 1°C above the threshold temperature was associated with a significant increase in relative risk (RR) of 1.07 for the >75 years age group, and 1.03 for males, respectively. Apart from longer lasting heat effects, short-term cold effects were observed between 4-7 days. The mean threshold temperature during cold exposure was -3.5°C and the minimum threshold temperature was -13.5°C. A mean temperature decrease of 1℃ was associated with a significant increase in RR of 1.03 for the <75 years age group, and 1.02 for females. The increased risk associated with minimum temperature was especially strong for male patients (RR =1.09, CI=1.05-1.14). With regard to heat exposure, the Medicaid group showed RR of 1.11 (lag: 1 day), which was higher than the NHI group who showed RR of 1.02 (lag: 22-28 days). With regard to cold exposure, the RRs in the Medicaid group were 1.05 with the mean temperature lag of 4-7 days, and 1.11 at the minimum temperature with a lag of 15-21 days, respectively. Meanwhile, patients in the NHI group only showed a risk of 1.01 at the mean temperature. The RR due to heat exposure was higher in hot areas with higher cooling degree-days (CDD) than in warm areas with lower cooling degree-days (CDD). Meanwhile, the RR due to cold exposure was higher in cold areas with higher heating degree-days (HDD) than in warm areas with lower heating degree-days (HDD).
Conclusions: This is the first study to find an association between temperature and incidence of SAH related to heat and cold exposure (defined by threshold temperatures). An increase in temperature above the heat temperature threshold or decrease in temperature below the cold temperature threshold correlated with increased risk of SAH in susceptible populations with different lag effects and risks. Our findings provide useful information for identifying the risk of SAH in vulnerable groups that can be used to establish climate change adaptation strategies.