Background
Fine particulate matter (PM2.5) produced by landscape fires is thought to be more toxic than that from non-fire sources. However, the effects of “fire-sourced” PM2.5 on acute respiratory infection (ARI) are unknown.

Methods
We combined Demographic and Health Survey (DHS) data from 48 countries with gridded global estimates of PM2.5 concentrations from 2003 to 2014. The proportions of fire-sourced PM2.5 were assessed by a chemical transport model using a variety of PM2.5 source data. We tested for associations between ARI and short-term exposure to fire- and “non-fire-sourced” PM2.5 using a bidirectional case-crossover analysis. The robustness and homogeneity of the associations were examined by sensitivity analyses. We also established a nonlinear exposure–response relationship between fire- and non-fire-sourced PM2.5 and ARI using a two-dimensional spline function.

Results
The study included 36,432 children under 5 years who reported ARI symptoms. Each 1 µg/m3 increment of fire-sourced PM2.5 was associated with a 3.2 % (95 % confidence interval [CI] 0.2, 6.2) increment in the risk of ARI. This effect was comparable to that of each ∼5 µg/m3 increment in PM2.5 from non-fire sources (3.1 %; 95 % CI 2.4, 3.7). The association between ARI and total PM2.5 concentration was significantly mediated by the proportion of fire-sourced particles. Nonlinear analysis showed that the risk of ARI was increased by both fire- and non-fire-sourced PM2.5, but especially by the former.

Conclusions
PM2.5 produced by landscape fire was more strongly associated to ARI among children under 5 years than that from non-fire sources.