Investigating the effect of air quality on health Event

1. Fiona O'Connor Title: Modelling present-day and future air quality for the UK Abstract: Poor air quality is both a global and national problem, with implications for human health and natural and managed ecosystems. Here, we apply a nested suite of 3 composition models, based on the Met Office's Unified Model (MetUM), to dynamically downscale air quality and climate from the global scale to the UK national scale. The nested models have been run for the present day and the model performance over the UK is benchmarked against both observations and the UK's operational air quality forecast model, AQUM. It is also run for the 2050s using future climate projections of sea surface temperatures, sea ice and using aerosol and ozone precursor emissions for 3 different Representative Concentration Pathways (RCPs): RCP2.6, RCP6.0, and RCP8.5. A brief assessment of how air quality is projected to change over the UK in the 2050s across the different RCPs is also presented, with a particular focus on the spatial patterns of change as well as projected changes to exceedances in air quality standards.
2. Sujit Sahu Title: Spatio-temporal modelling of daily air pollution levels for five years for London and the rest of the UK. Abstract This article proposes and then compares several flexible hierarchical Bayesian models for studying air pollution levels in the UK during the four year period 2007--2011. The models make use of the observed data as well as output of an atmospheric air quality model. Model based out-of-sample spatial predictions are found to be more accurate than both the air quality model output and kriging of the observations using currently available methods.
3. Duncan Lee Title:An adaptive spatio-temporal smoothing model for estimating the health impact of air pollution. Abstract The long-term effects can be estimated using a spatio-temporal ecological study design, which makes use of spatio-temporal contrasts in air pollution and population level disease prevalence. The disease data are aggregated counts of the numbers of disease cases in geographical areal units for consecutive time periods, meaning that Poisson log-linear models accounting for spatio-temporal autocorrelation are used for the analysis. However, these data present a number of statistical challenges, which if ignored have the potential to bias the estimated health effects of air pollution. One such challenge when modelling these pollution and disease data are their spatial misalignment, resulting in the change of support problem. The result is that the pollution concentrations vary spatially within each areal unit at which disease data are available, which if ignored, for example by computing the average concentration in each areal unit, leads to the possibility of ecological bias. This spatial misalignment, coupled with the sparsity of the monitored pollution data, also means that point level pollution concentrations are mostly unknown and need to be estimated, and thus the uncertainty in the estimates should be propagated forward into the health model to ensure full uncertainty quantification. The other challenge we consider in this paper relates to modelling the spatio-temporal autocorrelation remaining in the disease data after the covariate effects have been accounted for. Traditional approaches utilise random effects that are globally smooth in space and time, but these are likely to be too simplistic to accurately capture the remaining spatio-temporal autocorrelation in the disease data, because the disease data are unlikely to exhibit the same level of spatial smoothness across the study region. Therefore this paper proposes the first rigorous statistical framework for estimating the long-term health effects of air pollution that addresses the statistical challenges outlined above. Our proposed modelling approach is then applied to a new study of air pollution and health in England between 2007 and 2011, which is one of the most comprehensive studies of its type ever to be conducted in terms of both the spatio-temporal scale of the study region, and the range of pollution and disease data considered.
4. Alessandro Fasso Title: Model based multiresolution European population exposure to airborne pollutants. Abstract We consider the distribution of population by exposure to multiple airborne pollutants at various spatial and temporal resolutions over Europe. To do this we use high resolution semiparametric estimates of daily average concentrations for seven pollutants in years 2009-2011. In order to exploit the spatial information content and allow the computation of daily multipollutant exposure distribution, uncertainty included, we use a multivariate spatio-temporal model capable to handle non Gaussian large datasets such as multivariate and multiyear daily air quality, land use and meteorological data over Europe.
5. Eliane Rodrigues Title: An application to ozone data from Mexico City of a non-homogeneous Poisson model with spatial anisotropy. Abstract: We consider a non-homogeneous Poisson process to count the number of times that a given threshold is surpassed in a time interval. An anisotropic spatial dependence on the parameters of the Poisson intensity function is assumed. The anisotropic dependence is to account for the possible correlation between measurements in different sites. Using spatial interpolation we estimate the number of threshold exceedances and the days where those exceedances occurred in sites where measurements are not available. This is a joint work with Dani Gamerman, Mario H. Tarumoto and Guadalupe Tzintzun.
6. Gary Fuller Title: Learning from receptor analysis – new measurements of particle composition for model validation and health studies. Abstract: PM10 and PM2.5 define airborne particle concentrations by size only and but comprise a range of particles with different chemical and physical properties. While dispersion and chemical models can predict pollution concentrations receptor analysis seeks to explain measured concentrations at a point. New advances in instrumentation are allowing more the measurement of more chemical properties of the PM mix and at higher time resolutions. This opens new opportunity for the validation of predictive models and health studies to determine if differential toxicity exists between PM components.
7. Gavin Shaddick Title:The effects of exposure mis-specification in spatio-temporal epidemiological studies. Abstract: In order to perform studies into the risks of environmental hazards on human health study there is a requirement for accurate estimates of exposures that might be experienced by the populations at risk. Often there will be missing data and often the locations and times of exposure measurements and health data do not match. To a large extent this will be due to the health and exposure data having arisen from completely different data sources and not as the result of a carefully designed study. In such cases, a direct comparison of the exposure and health outcome is often not possible without an underlying model to align the two in the spatial and temporal domains. In addition, there may be preferential sampling where monitoring locations in environmental networks may be located in areas where levels are expected to be high. Biased estimates of exposures may lead to biased estimates of risk. The Bayesian approach provides a natural framework for modelling, however the large amounts of data that can arise from environmental networks mean that inference using MCMC might not be computational feasible. Here we use Integrated Nested Laplace Approximation (INLA) to implement spatio-temporal exposure models.
8. Marta Blangiardo Title:Using ecological propensity score to account for residual confounding in the association between air pollution and health. Abstract: Small area ecological studies are commonly used in environmental epidemiology to assess the impact of area level factors like air pollution on health outcomes when data are only available in an aggregated form. However the estimates are often biased due to unmeasured confounders which cannot be taken into account. Extra confounding information can be provided through external datasets like surveys or cohorts, where the data are available at the individual level and typically they cover only a subset of the geographical units. We develop a framework of analysis which combines ecological and individual level data from different sources and that (i) summarises all the available individual level confounders into an area level scalar variable, which we call ecological propensity score (EPS) , (ii) it imputes the values of the scalar where missing and (iii) includes the scalar in the ecological regression linking the risk factors to the health outcome via a flexible function. In this talk I will show that integrating individual level data into small area analyses via EPS is a promising method to reduce the bias intrinsic in ecological studies due to unmeasured confounders and will present an application to evaluate the effect of PM10 on cardiovascular hospital admissions in Greater London.
9. Alison Gowers Title: To be announced.

Registration is free but mandatory through the Eventbrite site for the meeting