Comparative evaluation of Spatio-Temporal Exposure Assessment Methods for estimating the health effects of air pollution (STEAM)

Lead Research Organisation: King's College London
Department Name: Health and Social Care Research

Abstract

Background: Epidemiological studies have linked long- and short-term outdoor air pollution exposures to increased risks of death and morbidity. Cohort studies evaluate the health effects of long-term exposure by exploiting spatial heterogeneity in annual air pollution concentrations whereas time-series studies evaluate acute effects exploiting temporal variability in pollution concentrations. The extent to which the adverse health effects reported by these different study designs overlap has important implications for health impact assessment and policy development and to date this has not been investigated outside the US. In the past, both study designs have been limited by the availability of monitored pollution data but the development of dispersion and land use regression models have improved exposure assessment in cohort studies and enabled the investigation of a broader range of pollutants. The extension of these models to provide spatially resolved daily estimates of pollutant concentrations will enable an integrated assessment of health effects arising from both long- and short-term exposures to a wide range of pollutants and with applicability to locations with sparse or no monitoring.
Objectives: 1.To develop, validate and compare dispersion, land use regression and satellite data based pollution models for London at both fine spatial (postcode level) and temporal (daily) scales. 2. To assess the performance of these methods and their combinations in estimating the health effects of long and short-term exposures to air pollutants. 3. To use the optimum integrated exposure method to estimate the relative importance of the effects of long-and short-term exposure for selected health outcomes. To achieve these objectives the project will use the expertise developed within the MRC-PHE Centre for Environment and Health and will need to draw upon the expertise of pioneering international groups (in the U.S. & Greece) in modelling, simulation studies and time series analyses.
Methods: Models will be developed for particulate and gaseous pollutants and validated using measured concentrations from the London Air Quality Network. Their applicability to locations with sparse or no monitoring will be assessed using sequentially smaller parts of the available exposure information. An extensive simulation study, in which the "true" effect is known, will investigate the implications of using each exposure method on the health effect estimation (precision and bias).To enable proof of concept for the simultaneous estimation of long- and short-term effects on various health outcomes, the analysis of selected endpoints including GP consultations, hospital admissions and mortality data will be applied.
Innovation aspects:
1. Application of air pollution modelling approaches at a fine time and spatial resolution
2. Integration of the different exposure assessment methods for optimizing their performance.
3. Use of simulation methods to evaluate simultaneously the performance of models in assessing the effects of long- and short-term exposure to particulate and gaseous pollutants.
4. Simultaneous estimation of health effects of long- and short-term exposure to air pollution in existing large health data sets to enable a comparison of their relative importance.
Policy implications:
1. Information on the advantages and disadvantages of each exposure model will inform the optimum extent and density of monitoring systems and other inputs to local models.
2. The simultaneous estimation of the effects of short and long-term exposures will provide a rationale for the balance between emergency short-term action and long-term pollution management interventions.
3. Informing the development of modelling in areas of Europe and the World where monitoring is not so dense.
4. Informing health studies about the comparative importance of short vs long-term effects in areas of Europe and the World where health outcome data bases are limited

Technical Summary

The aim of this study is to develop, validate and compare three integrated air pollution exposure assessment methods for estimating simultaneously the health effects of short and long-term exposure to outdoor air pollutants. The dispersion model will use CMAQ-urban which couples the USEPA regional model, CMAQ and the Atmospheric Dispersion Modelling System roads model incorporating emissions data. This combined model will provide hourly predictions at 20x20m spatial resolution for Greater London for PM10, PM2.5, NO2 and O3. Spatio-temporal LUR models for PM10, PM2.5, NO2 and O3 will use a semiparametric approach with covariates that vary primarily in space (e.g. distance to major road, traffic counts) and covariates that vary primarily in time (e.g. meteorology) and a bivariate smooth thin plate function for the geographical co-ordinates. New methods, similar to Chudnovsky 2014, will use a hybrid model, with AOD, land use, and meteorological data to estimate daily PM2.5. Spatial predictors will include population and traffic density, land use (Landsat) and emissions as well as time varying predictors (e.g. vegetative index from MODIS, meteorology). Each method will be evaluated using simulation techniques developed from previous work. We propose to construct scenarios where 'true' temporally and spatially resolved air pollution data are associated with health outcomes by a priori defined concentration response functions (such as those proposed by W.H.O. HRAPIE project). Model performance will be assessed under a variety of scenarios using a number of parameters including bias, statistical power and coverage intervals. Finally the model predictions will be used in conjunction with selected health outcomes to demonstrate proof of concept in estimating simultaneous associations between health outcomes and long- and short-term exposure to air pollution.

Planned Impact

National and international governmental agencies responsible for air pollution policy development and monitoring and for health impact assessment will be the main beneficiaries of this research. In the UK, Defra, the Department of Health, Public Health England and Local Authorities will be able to utilise the methodologies derived from the project to plan the deployment of air pollution monitoring sites, assess health impacts and determine local and national air pollution mitigation strategies in both the long- and short-term. Similarly, the WHO, the European Environment Agency and other devolved government and government agencies or regulators can also utilise the methodology with particular application in locations where less comprehensive monitoring networks exist. Examples of health impact assessment exercises at national and international levels include exercises here in the UK [COMEAP report] and in the EC [HRAPIE]. The central topic of our project is better estimation of a population exposure to ambient air pollution and the subsequent impact on human health. We therefore anticipate that the general public, both here in the UK as well as abroad, will benefit from better pollution management policies and a greater awareness of the links between air pollutant exposure and health. The academic community (epidemiology/Public Health, and air pollution science) will also benefit especially from the methodological developments.

How will they benefit from this research?
Health effects of acute and chronic exposure to air pollution have usually been studied separately and their results used for separate health impact assessment exercises. In a number of cases the lack of evidence for health effects of long term exposure from cohort studies has been supplemented by the evidence from short-term exposure (time series) studies. The methods used to derive exposure measures in cohort studies often vary and make synthesis difficult. Transferability of results from the regions of the world where studies have been conducted to less well studied locations is also problematical. This project will attempt to address many of these difficulties by 1) assessing the advantages and disadvantages of each exposure model to inform the optimum extent and density of monitoring systems and other inputs to local models required in any region of the world; 2) inform the development of modelling in areas where monitoring networks are sparse to facilitate exposure assessment; 3) facilitate the simultaneous estimation of the effects of long- and short-term exposures and provide a rationale for the balance between emergency short-term action and long-term pollution management interventions applicable to local and national needs; and 4) inform health studies about the comparative importance of short vs long-term effects in areas of Europe and the World where health outcome data bases are limited. The novel methods developed in this project will provide the tools policy analysts require to undertake more informed policy formulation and assessment exercises providing relevant information to local, national and international agencies responsible for air pollution policy.

Publications


10 25 50
 
Title Air pollution and meteorological data 2004-2013 
Description We have recorded all available data on daily temperature, relative humidity, wind speed and direction from fixed sites within the London area, for the period 2004-2013. For the same period we have recorded daily data for all available pollutants including PM10, PM2.5, NO2, NOx and ozone. The data is still being checked. We are working on a model to fill in the PM2.5 data base predicting the concentrations by modeling, using meteorological variables and other pollutants as predictors of PM2.5 daily levels. 
Type Of Material Database/Collection of data 
Year Produced 2017 
Provided To Others? Yes  
Impact Until today the data base is still under the process of quality control and management. 
 
Title Land Use Regression (LUR) Database for the London area 
Description For the Work Package A.3.Spatio-temporal LUR modelling with daily predictions We have collected all spatial data required to develop the spatio-temporal LUR models for air pollutants in the greater area of London, in order to estimate daily ambient air pollution concentrations for any point in space for the study area and time period of interest (2004-2013). Subsequently, we will use the estimated concentrations for assessing either short- or long-term effects of air pollution, in health effects analyses. We have collected information on the: 1. Geographical coordinates (latitude, longitude) of each air pollution monitoring site & meteorological monitoring station 2. Digital cartographical data for the study area. Attributes including LSOA polygons of the complete study area. 3. Digital road network for the complete study area with linked traffic intensity data to the complete road network (Meridian dataset). 4. Land use/cover data for the complete study area (OS Mastermap dataset) 5. Building density data (Mastermap/LIDAR dataset) GIS analyses is being conducted to derive the values for the potential predictor variables for the coordinates of the air pollution monitoring sites, in order to be taking into account in the spatio-temporal LUR modeling development procedure. Up to now, we have derived land use/cover variables for the monitoring sites located in the "small" study area, thus all sites located in LSOAs within or intersecting with the M25 motorway. These variables are part of the final spatio-temporal LUR database, which once complete, will incorporate the following information, over the same geographical study area and timescale (daily values for the years 2004-2013): 1. the daily air pollution measurement data 2. spatial data (all GIS potential predictor variables for the coordinates of each air pollution monitoring site) 3. temporal daily data 4. descriptive data This dataset will be used to develop the spatio-temporal LUR models for NOx, NO2, PM10, PM2.5 and O3 concentrations in the greater area of London. 
Type Of Material Computer model/algorithm 
Provided To Others? No  
Impact This data base will provide input for modelling the spatio temporal distribution of pollutants in the London area for the period 2009-2013. The data for 2004 to 2008 will be used as comparison data. 
 
Description King's College and Imperal College for STEAM 
Organisation Imperial College London (ICL)
Department Imperial Clinical Trials Unit (ICTU)
Country United Kingdom of Great Britain & Northern Ireland (UK) 
Sector Academic/University 
PI Contribution King's has provided data for pollutants and meteorological variables and is implementing one exposure model.
Collaborator Contribution Imperial provided access to Land Use variables.
Impact The land use variables were managed with a Geographical information system to produce variables used as input for various pollution exposure models
Start Year 2016
 
Description King's College and St George's for STEAM 
Organisation St George's University of London
Department St. George's Hospital Medical School
Country United Kingdom of Great Britain & Northern Ireland (UK) 
Sector Academic/University 
PI Contribution We have prepared the data sets and contribute to the exposure modelling.
Collaborator Contribution St George's are preparing the simulation study and will later compile a health outcome data set based on HES and ONS data
Impact Data sets have been prepared for use. The collaboration is multi-disciplinary between epidemiologists, statisticians and exposure scientists.
Start Year 2016
 
Description King's College and University of Athens for STEAM 
Organisation National and Kapodistrian University of Athens
Country Greece, Hellenic Republic 
Sector Academic/University 
PI Contribution The collaboration has been described in the contract of STEAM. We, from King's, provided data bases and are implementing one exposure model during the reporting period.
Collaborator Contribution The University of Athens are working on the calculation of variables from the Geographical Information System and implementing the Land Use Exposure modelling.
Impact So far data bases have been prepared and shared, containing measurements for air pollutants and meteorological variables for the London area.
Start Year 2016
 
Description First meeting of researchers and stakeholders 
Form Of Engagement Activity A formal working group, expert panel or dialogue
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Policymakers/politicians
Results and Impact We had a joint meeting of study participants from the U.K. (King's College, Imperial College and St George's), a U.S. Harvard University professor, and colleagues from the University of Athens Greece and members of Public health England and the Department for Environment, Food and Rural Affairs (DEFRA). The participants commented and made suggestions on the project plan for the next 2 and a half years which were very important for our future work.
Year(s) Of Engagement Activity 2016
 
Description Presentation of the STEAM project 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Other audiences
Results and Impact I presented the STEAM project, its objectives, completed activities and project plans to the Annual International Scientific Advisory Board Meeting of the MRC-PHE Centre for Environment and Health on June 2, 2016.
Year(s) Of Engagement Activity 2016