Indoor air quality of 5,000 households and its determinants. Part A: Particulate matter (PM2.5 and PM10–2.5) concentrations in the Japan Environment and Children's Study

Highlights: • Indoor and outdoor PM was measured twice each in 5000 Japanese households. • A random forest model was used to build prediction models of indoor PM. • The 10-fold cross-validation R² of the models ranged from 0.36 to 0.44. • Smoking, visible smoke, temperature and window opening increased indoor PM. • Formation of secondary organic aerosols (SOAs) may have contributed to indoor PM. Exposure to particulate matter (PM) is one of the important risk factors for morbidity and mortality. Although PM concentrations have been assessed using air quality monitoring stations or modelling, few studies have measured indoor PM in large-scale birth cohorts. The Japan Environment and Children's Study (JECS) measured indoor and outdoor air quality in approximately 5000 households when the participating children were aged 1.5 and 3 years. PM was collected using portable pumps for 7 days (total of 24 h), inside and outside each home. Prediction models for indoor PM concentrations were built using data collected at age 1.5 years and post-validated against data collected at age 3 years. Median indoor/outdoor PM_2.5 and PM_10–2.5 concentrations at age 1.5 years [3 years] were 12.9/12.7 [12.5/11.3] μg/m³ and 5.0/6.3 [5.1/6.1] μg/m³, respectively. Random forest regression analysis found that the major predictors of indoor PM_2.5 were indoor PM_10–2.5, outdoor PM_2.5, indoor smoking, observable smoke and indoor/outdoor temperature. Indoor PM_2.5, outdoor PM_10–2.5, indoor humidity and opening room windows were important predictors of indoor PM_10–2.5 concentrations. Indoor benzene, acetaldehyde, ozone and nitrogen dioxide concentrations were also found to predict indoor PM_2.5 and PM_10–2.5 concentrations, possibly due to the formation of secondary organic aerosols. These findings demonstrate the importance of reducing outdoor PM concentrations, avoiding indoor smoking, using air cleaner in applicable and diminishing sources of VOCs that could form secondary organic aerosols, and the resulting models can be used to predict indoor PM concentrations for the rest of the JECS cohort.