2.5. Ozone
We used measurements of ozone from 30 stations (background and urban background), and all individuals were assessed with measurements using the nearest station approach. If the nearest station had missing data the individual was assessed at the sec- ond-nearest station or the third-nearest station. We excluded data if there was more than 10% missing data or if an individual lived further away from a station than 32 km according to the EPA re- commendations (Abbey et al., 1991). The mean distance to stations used in the model was 8.5 km, with a range from 0.1 to 32 km. We used 24 h average ozone data and aggregated into trimester data. Ozone in the study area has been extensively studied and in transport of ozone (and precursors) local production dominates, leading to small spatial variation (Malmqvist et al., 2014).
The ozone levels were categorised into quartiles based upon the first trimester distribution among the controls in the study (in ppb using the WHO conversion rate for Ozone):