Covariates and mediator
Information on covariates was obtained from the questionnaire, including age (<5 years, 5-10 years, 10-15 years and ≥15 years), gender (boy/girl), exercise time (hours/week), family income per year (<10,000RMB, 10,000-29,999RMB, 30,000-99,999RMB, or ≥100,000RMB), parental education (≥high school or lower), low birth weight (birth weight <2500g), premature birth (gestational age <37), breastfeeding (child being mainly breastfed for at least three months), obesity (yes/no), passive smoke exposure (child living with someone of the household who smokes daily at residence), home coal use (yes/no) and pet kept (yes/no), particulate matter with aerodynamic diameters < 2.5 µm (PM2.5) exposure (<48.97µg/m3, 48.97-56.23µg/m3, 56.23-60.57µg/m3, or ≥60.57 µg/m3). The assessment of personal PM2.5 exposure has previously been described in detail.15,16 In brief, the average PM2.5 concentration during 2009-2012, estimated according to each participant’s residence using a machine learning method, was regarded as a surrogate of individual exposure. Participants were categorized into four groups based on quartiles of PM2.5 concentration they were exposed. A directed acyclic graph (DAG) was drawn by online DAGitty (http://dagitty.net/dags.html) (Fig. S1). Variables, containing passive smoke exposure, home coal use, pet kept and PM2.5 exposure, were identified as potential confounders that needed adjusting in the main model. In addition, parent with asthma was identified as a potential mediator.