Periodic Reporting for period 3 - GLASST (Global and Local Health Impact Assessment of Transport: methods for prioritising model development)
Berichtszeitraum: 2022-06-01 bis 2023-11-30
The Population Health Problem and Challenge
Transport is a major contributor to determinants of population health. Adverse health impacts are greatest in rapidly urbanising and motorising lower and middle income country cities. In 2016 there were an estimated 1.3 million road traffic deaths, 4.3 million deaths from outdoor air pollution, and 1.4 million from low physical activity. Road transport produces around 17% of energy related greenhouse gas emissions, and this is rising. Transport systems need to shift away from the car towards more active and sustainable modes. Many cities are investing in policies to support walking, cycling (including electric assist bikes), and transit, and wrong decisions have high costs. However, we lack academically rigorous and practically useful tools to estimate the health impact of these policies. The need to better integrate health evidence with transport research and decision-making is a major research gap – and opportunity.
The aim of GLASST is to create methods, models, and tools to support inclusion of health into transport decision making. It does this through three objectives and corresponding work streams:
Objectives:
1. To develop methods and associated computer programs that allow researchers to compare health impact models and data. By collating and comparing models across many settings and scenarios, we identify the circumstances in which variation in model structure and parameters make an important difference to model results. This information is being used to create and test models for new settings and problems.
2. To extend transport models to better represent health risk factors and include health impact modelling methods. By including the rich spatial and temporal behaviour data of transport models this integration will enable far more realistic health estimates,. It will also make health impacts visible to transport planners.
3. To use the methods from (1) and findings from (1) and (2) to work towards developing a global city-level model and tool that utilises the best data available in any setting to create comparable exposure and disease estimates. Initially I will select archetypical cities to model for various regions before working towards the global city level model. This will transform the opportunities for me and others to undertake modelling health impacts of transport policies and scenarios across the world.
Transport is a major contributor to determinants of population health. Adverse health impacts are greatest in rapidly urbanising and motorising lower and middle income country cities. In 2016 there were an estimated 1.3 million road traffic deaths, 4.3 million deaths from outdoor air pollution, and 1.4 million from low physical activity. Road transport produces around 17% of energy related greenhouse gas emissions, and this is rising. Transport systems need to shift away from the car towards more active and sustainable modes. Many cities are investing in policies to support walking, cycling (including electric assist bikes), and transit, and wrong decisions have high costs. However, we lack academically rigorous and practically useful tools to estimate the health impact of these policies. The need to better integrate health evidence with transport research and decision-making is a major research gap – and opportunity.
The aim of GLASST is to create methods, models, and tools to support inclusion of health into transport decision making. It does this through three objectives and corresponding work streams:
Objectives:
1. To develop methods and associated computer programs that allow researchers to compare health impact models and data. By collating and comparing models across many settings and scenarios, we identify the circumstances in which variation in model structure and parameters make an important difference to model results. This information is being used to create and test models for new settings and problems.
2. To extend transport models to better represent health risk factors and include health impact modelling methods. By including the rich spatial and temporal behaviour data of transport models this integration will enable far more realistic health estimates,. It will also make health impacts visible to transport planners.
3. To use the methods from (1) and findings from (1) and (2) to work towards developing a global city-level model and tool that utilises the best data available in any setting to create comparable exposure and disease estimates. Initially I will select archetypical cities to model for various regions before working towards the global city level model. This will transform the opportunities for me and others to undertake modelling health impacts of transport policies and scenarios across the world.
Please note all code is open source and on github.
Work Stream 1:
We conducted two systematic reviews to identify models used for transport health impact assessment, and to synthesise results. Paper are due for submission by March 2022.
We published a tutorial on the application of Value of Information methods to health impact models in Epidemiologic Methods (Jackson et al 2021), along with an R package and full code.
Work Stream 2:
SILO-MITO-MATSim are integrated transport models for simulating land use, travel demand, and road network assignment. We have extended these models to represent more accurately health behaviours and exposures, to sum exposures for each individual, and to simulate health outcomes. The model has been set up for Munich and is being set up for Manchester.
Work Stream 3:
ITHIM-Global is a new version of The Integrated Transport and Health Impact Model created for GLASST. Sufficient data for modelling the three health pathways (physical activity, road traffic injuries, air pollution) has been fully processed and we are analysing results for 20 cities in Latin America, 5 cities in Africa, and 3 cities in India.
Travel behaviours We have collated and harmonised anonymised travel survey datasets for 50 cities (across 17 different countries), and published two papers, one comparing cycling levels and one comparing gender differences in active travel.
Road traffic injury data We accessed and harmonised datasets of case-level road fatalities from multiple settings 76 cities from Latin America, North America, Africa, and Europe.
Air emissions and pollution data We are working with the European Commission’s Joint Research Centre (JRC) to use the EDGAR emissions database for impact modelling.
Health impact modelling methods We have implemented the proportional multi-state life table model in R. We have made advances in methods and software for modelling chronic disease burden from limited disease data (paper submitted).
Dose response functions: We undertook systematic reviews, including extensive exposure harmonisation and dose response meta-analyses for a wide range of diseases. This work is being published in three papers. Firstly, all-cause mortality, cardiovascular diseases, and 11 cancers. Secondly, depression. Thirdly, dementias and Parkinson’s disease.
Work Stream 1:
We conducted two systematic reviews to identify models used for transport health impact assessment, and to synthesise results. Paper are due for submission by March 2022.
We published a tutorial on the application of Value of Information methods to health impact models in Epidemiologic Methods (Jackson et al 2021), along with an R package and full code.
Work Stream 2:
SILO-MITO-MATSim are integrated transport models for simulating land use, travel demand, and road network assignment. We have extended these models to represent more accurately health behaviours and exposures, to sum exposures for each individual, and to simulate health outcomes. The model has been set up for Munich and is being set up for Manchester.
Work Stream 3:
ITHIM-Global is a new version of The Integrated Transport and Health Impact Model created for GLASST. Sufficient data for modelling the three health pathways (physical activity, road traffic injuries, air pollution) has been fully processed and we are analysing results for 20 cities in Latin America, 5 cities in Africa, and 3 cities in India.
Travel behaviours We have collated and harmonised anonymised travel survey datasets for 50 cities (across 17 different countries), and published two papers, one comparing cycling levels and one comparing gender differences in active travel.
Road traffic injury data We accessed and harmonised datasets of case-level road fatalities from multiple settings 76 cities from Latin America, North America, Africa, and Europe.
Air emissions and pollution data We are working with the European Commission’s Joint Research Centre (JRC) to use the EDGAR emissions database for impact modelling.
Health impact modelling methods We have implemented the proportional multi-state life table model in R. We have made advances in methods and software for modelling chronic disease burden from limited disease data (paper submitted).
Dose response functions: We undertook systematic reviews, including extensive exposure harmonisation and dose response meta-analyses for a wide range of diseases. This work is being published in three papers. Firstly, all-cause mortality, cardiovascular diseases, and 11 cancers. Secondly, depression. Thirdly, dementias and Parkinson’s disease.
In Work Stream 1, building on the findings of the systematic review, we will undertake analysis of how the choice of modelling methods and input data affects results.
In Work Stream 2, the extension of transport models to include health outcomes and exposures is a first. We will continue to extend this, including by:
• Introducing a fully microsimulation version of the multistate life table model into the land use model SILO
• Comparing the results of the MATSim dispersion model using other methods that better account for chemistry, such as the JRC’s SHERPA source-receptor air quality tool.
• Completing a full implementation of the integrated modelling suite for Munich, Manchester and Melbourne.
• Improving the physical activity model to include hilliness and other micro environmental features
• Calculating individual noise exposure along the route
• Further improving the travel demand model to give more realistic distributions and behavioural sensitivity to policy intervention
• Conduct stakeholder workshops
In Work Stream 3, we have made progress beyond state of the art
• The processing and harmonisation of travel survey and traffic injury has gone far beyond previous studies, notably including many more cities in lower and middle income countries (LMIC).
• The dose response meta-analyses has included many new outcomes and far more studies than previous work due to our extensive harmonisation approach
• ITHIM Global has been parameterised for multiple LMIC cities
We will
• Assess the life cycle carbon benefits of active travel in LMIC cities; considering cyclists vs non-cyclists, and socio-economic variation
• Extend further our collation and harominsation of travel survey and traffic injury data
• Generate global and regional estimates of road traffic deaths attributable to different types of vehicles.
• Compare estimates of emission sources with those from EDGAR, with the European Commission’s Joint Research Centre.
• Develop statistical models to estimate travel patterns and injury data for cities without adequate survey data. We will start by doing this for Latin America.
• Continue to develop new methods for health impact modelling with limited data.
• Undertake health impact modelling of transport scenarios for a wider range of cities
• Conduct stakeholder workshops
In Work Stream 2, the extension of transport models to include health outcomes and exposures is a first. We will continue to extend this, including by:
• Introducing a fully microsimulation version of the multistate life table model into the land use model SILO
• Comparing the results of the MATSim dispersion model using other methods that better account for chemistry, such as the JRC’s SHERPA source-receptor air quality tool.
• Completing a full implementation of the integrated modelling suite for Munich, Manchester and Melbourne.
• Improving the physical activity model to include hilliness and other micro environmental features
• Calculating individual noise exposure along the route
• Further improving the travel demand model to give more realistic distributions and behavioural sensitivity to policy intervention
• Conduct stakeholder workshops
In Work Stream 3, we have made progress beyond state of the art
• The processing and harmonisation of travel survey and traffic injury has gone far beyond previous studies, notably including many more cities in lower and middle income countries (LMIC).
• The dose response meta-analyses has included many new outcomes and far more studies than previous work due to our extensive harmonisation approach
• ITHIM Global has been parameterised for multiple LMIC cities
We will
• Assess the life cycle carbon benefits of active travel in LMIC cities; considering cyclists vs non-cyclists, and socio-economic variation
• Extend further our collation and harominsation of travel survey and traffic injury data
• Generate global and regional estimates of road traffic deaths attributable to different types of vehicles.
• Compare estimates of emission sources with those from EDGAR, with the European Commission’s Joint Research Centre.
• Develop statistical models to estimate travel patterns and injury data for cities without adequate survey data. We will start by doing this for Latin America.
• Continue to develop new methods for health impact modelling with limited data.
• Undertake health impact modelling of transport scenarios for a wider range of cities
• Conduct stakeholder workshops