Fig. 3: Cumulative relative risk for respiratory hospitalizations associated with a 1 µg m−3 increase in wildfire-specific PM2.5 during lag 0–1 days.
From: Respiratory risks from wildfire-specific PM2.5 across multiple countries and territories
a–d, Relative risk modified by sex (a), age (b), community non-wildfire PM2.5 level (c) and community income level (d). The centre of the error bars represents the mean effect estimates comparing days with lower versus higher wildfire-specific PM2.5 levels across 1,052 study communities, with error bars indicating uncertainty ranges derived from the 95% CIs of 1,000 Monte Carlo samples. All the models were controlled for temperature, relative humidity, non-wildfire PM2.5, day of week, seasonality and long-term trend. Income level was categorized according to tertiles of the community-level GDP per capita: GDP per capita ≤ tertile 1 as low income; tertile 1 < GDP per capita ≤ tertile 2 as middle income; and GDP per capita > tertile 2 as high income. Non-wildfire PM2.5 level was classified according to the 50th percentile of non-wildfire PM2.5 concentration distributions (17.25 µg m−3): average daily non-wildfire PM2.5 during the study period ≤17.25 µg m−3 as low non-wildfire PM2.5 level; and >17.25 µg m−3 as high non-wildfire PM2.5 level. A Wald-type test was used to test the significance of the effect modifications. *A two-sided P for the difference is statistically significant (<0.05).
