Technical meeting, 15-16 July 2015, Ljubljana
The statistical strategy was discussed during the fourth action meeting “B4 Development of the statistical modelling strategy for the analysis of the long-term effects of air pollutants on health endpoints” that has been organised at mid July 2015. The meeting took place at National Institute of Public Health (NIJZ) in Ljubljana. The statistical protocol presented during the meeting reported all the analytical steps to be undertaken in the epidemiological evaluations, from the assumptions to be verified in the Cox regression model, to the evaluation of confounders and modifiers.
The exploratory analysis which have been completed in order to assess a statistical model to estimate the long-term health effects of air pollutants, is able to represent the association between the air pollutants exposure and the health endpoints analysed.
For each of the axes that will be taken into account in statistical analysis, the results from harmonization of variables have been presented together with a score, indicating the possibility to effectively use each of the variable in EU countries.
Different approaches have been considered:
- Cox proportional hazard model for total or cause-specific survival;
- Ecological analysis models.
Furthermore, a number of sensitivity analyses will be designed in order to make the results more robust to different modelling choices.
The differences between findings on urban versus non-urban need some further exploration of the data to understand its reasons.
Basic exposure assessment and sensitivity analyses on samples areas will be carried out to show the best way to represent population real exposure.
The time-dependent assignment of exposure is a useful strategy but sensitivity analyses will lead to a depth discussion on this relevant issue.
The meeting has been attended by all partners, someone joined the group in videoconference.