Spatial variability of indoor air pollutants in schools. A multilevel approach

The measurement of indoor air pollutants and their health effects are less often studied due to the costs of collection of such data. We have analysed the variability in the measurement of five indoor school air pollutants: fine particulate matter of size ≤2.5 μm (PM2.5), nitrogen dioxide (NO2), and three Volatile Organic Compounds (VOC), namely formaldehyde, acetaldehyde and acrolein, objectively measured over five days of a week at representative points in more than 400 classrooms of 109 schools and courtyards in six French cities spread out over the year. Separate 3-stage multilevel models were fitted to partition the different nested variance components (i.e., classroom, school and city levels), and intra-class correlation (ICC) coefficients were computed to bring out the similarities of pollutants' concentrations among these spatial units. The indoor PM2.5 and NO2 concentrations showed a high degree of similarity (ICC coefficients equal to 76% and 81%, respectively) between the classrooms of a school (and city), whereas the formaldehyde, acetaldehyde and acrolein concentrations showed low to moderate degree of similarity (ICC coefficients equal to 25%, 36% and 57%, respectively) between the classrooms. We conclude that to investigate the impact of indoor air pollutants, a multilevel approach taking into account the full design of the study would be the most appropriate.