Pre-Conference Courses

Sunday August 26
Shaw Centre (meeting venue)

Pre-conference courses will be offered on Sunday August 26 at the Shaw Centre. Registration can be done via the online registration website. Payment of the course fee should be submitted along with the registration fee.

Lunch Box
The ISES-ISEE 2018 Joint Annual Meeting organizers offer you the option to purchase a lunch box on the Pre-Conference Courses day. During the registration you can indicate if you would like to order a lunch box.

Please note: We reserve the right to cancel a class if there is insufficient enrollment. If we must cancel, you will be offered the opportunity to switch to another class or given a full refund.

Please find an overview of the Pre-Conference Courses below. At the bottom of the page you can find the detailed description of the courses.

Room 1 Room 2 Room 3 Room 4 Room 5

Morning Courses

09:00 am – 12:00 pm

PC01A (Full Day Course): Big data, machine learning techniques to investigate health effects in environmental health studies PC02A (Full Day Course): Consumer Exposure Modeling for Human Health Risk Assessment – Advanced Tools PC03: Bayesian methods for environmental health researchers PC04: Application of New Approach Methodologies for Exposure Assessment and Prioritization – Tools for Researchers and Regulators Including use of Quantitative Structure Use Relationships (QSUR) PC05: Model-Based Geostatistics and Spatial Epidemiology: a practical introduction with R
Afternoon Courses

1:00 pm – 4:00 pm

PC01B (Full Day Course): Big data, machine learning techniques to investigate health effects in environmental health studies PC02B (Full Day Course): Consumer Exposure Modeling for Human Health Risk Assessment – Advanced Tools PC06: Predicting microscale urban features using street-level images – an introduction to machine learning PC07: Causal inference foundations and applications in environmental health sciences PC08: Advanced modelling techniques for time series analysis using R

Pre-Conference Course Fees:

Below you can find an overview of the Pre-Conference Course fees.

The full day course fee applies for one full day course or two half day courses.

Early bird registration fee
Before July 1
Standard registration fee
July 1 – August 16
On-site registration fee
As of August 16
Regular full day 150 CAD 180 CAD 200 CAD
Regular half day  100 CAD 130 CAD 150 CAD
Postdoc full day 115 CAD 135 CAD 150 CAD
Postdoc half day 75 CAD 100 CAD 115 CAD
Student full day 75 CAD 90 CAD 100 CAD
Student half day 50 CAD 65 CAD 75 CAD

PC01: Big data, machine learning techniques to investigate health effects in environmental health studies

Youssef Oulhote, Department of Environmental Health, Harvard T. H. Chan School of Public Health; Boston, USA

Laura Balzer, Department of Biostatistics & Epidemiology, School of Public Health & Health Sciences, UMass Amherst; Amherst, USA
Chirag J Patel, Department of Biomedical Informatics, Harvard Medical School; Boston, USA
Martin Tondel, Department of Medical Sciences, Occupational and Environmental Medicine, Uppsala University; Uppsala, Sweden

Description

Purpose of the course including specific learning objectives:

This course will present key methodological challenges that arise in environmental health, and provide recent methods that can be used to deal with these challenges. Our focus is on modern solutions to multiple testing, model misspecification, and causal inference as applied to environmental health data, with a focus on assessing health effects of chemical mixtures. Participants will gain both a theoretical understanding as well as practical experience with a hands-on session using R software. At the end of the course, participants will be aware of and be able to implement state-of-the-art epidemiologic methods, including Environmental Wide Association Studies, ensemble learning techniques, G-computation, and targeted maximum likelihood estimation. Participants will gain the skills to correctly interpret software outputs and conduct these analyses in their own research projects. R scripts and specific functions will be provided.

Outline

Morning session (9.00-12.00) will be devoted to theoretical and conceptual descriptions of the proposed methods with concrete examples. This session will include:

  1. Overview of methodological challenges in environmental epidemiology: multiple testing, model misspecification, and causal inference
  2. High-throughput associations of multiple environmental and non-genetic factors.
  3. Ensemble Learning Techniques
  4. G-computation
  5. Targeted Maximum Likelihood Estimation
  6. Ethics Guidelines for Environmental Epidemiologists in the era of big data.

Afternoon session (13.00-16.00) will be devoted to hands-on exercises using R software. Participants can choose to participate only in the morning session, but attending the morning session is a prerequisite for the afternoon hands-on session. In any case, all participants will also be given fully executable do-files, so no prior programming experience is necessary.

PC02: Consumer Exposure Modeling for Human Health Risk Assessment – Advanced Tools

Eva Wong, U.S. Environmental Protection Agency, Office of Pollution Prevention and Toxics, Consumer Exposure Model (CEM); Washington, DC, USA
Heidi Hubbard, ICF, Consumer Exposure Model (CEM); Fairfax, USA
Gerlienke Schuur, Centre for Safety of Substances and Products, National Institute for Public Health and the Environment (RIVM); Utrecht, The Netherlands
Wouter ter Burg, Centre for Safety of Substances and Products, National Institute for Public Health and the Environment (RIVM); Utrecht, The Netherlands

Description

Purpose of the course including specific learning objectives:

This course will introduce and review two widely used and recently updated consumer exposure modelling tools, US EPA’s Consumer Exposure Model (CEM) and ConsExpo Web. The course will provide an introduction and overview of both models, for use in human health exposure and risk assessment. Course attendees will participate in a demonstration of the models and application of the models in case studies, to illustrate similarities and differences between the tools.

Morning session (9.00-12.00): Introduction to CEM and ConsExpo Web

The morning session will be devoted to providing an introduction and overview of the U.S. EPA’s Consumer Exposure Model followed by an introduction for ConsExpo Web

This session will include:

  1. Introduction and overview of the U.S. EPA’s Consumer Exposure Model which is used to estimate indoor air concentrations, indoor dust concentrations, dermal exposure, and mouthing exposure for a variety of consumer products and materials.
  2. Short introduction and overview of the different models for the inhalation, dermal and oral routes in ConsExpo Web, as well as an introduction on the ConsExpo fact sheets, with default choices for models and exposure parameters for different consumer product categories (such as cosmetics, paint, cleaning products).

CEM is a user-friendly computer program which estimates indoor air concentrations, indoor dust concentrations, dermal exposure, and mouthing exposure for a wide variety of consumer products and materials. The model was developed by the United States Environmental Protection Agency for use in implementing the requirements of the Toxic Substances Control Act which was recently modified through the Frank R. Lautenberg Chemical Safety for the 21st Century Act. The model estimates inhalation, ingestion, and dermal exposures, calculated as single day doses and chronic average daily doses. CEM (2.0) retains six existing models (CEM 1.2) within E-FAST V2.0 (.exe, 32MB) and adds nine additional models.

During 2016/2017, CEM (2.0) was tested by experienced users (i.e., via beta test) and peer reviewed by independent experts. CEM was developed using Microsoft Access and Visual Basic for Applications (VBA), and it is compatible with 2007 and 2010 versions of Microsoft Office. A freely-available Microsoft Access runtime environment is available on the Microsoft website for users that do not have Microsoft Access installed on their computer.

CEM facilitates the tailoring of the exposure scenario, based on the chemical, consumer product, receptor, and environment of interest. The model estimates acute and chronic exposures and provides a variety of exposure metrics.

ConsExpo is a computer program that enables and facilitates the estimation and assessment of exposure to substances from consumer products such as paint, cleaning agents and cosmetics. The model is developed by the National Institute for Public Health and the Environment (RIVM).

In October 2016 at ISES in Utrecht, the new freely available web version of ConsExpo Web was launched.  The update of ConsExpo is executed by RIVM in collaboration with the counterpart institutes ANSES (France), BfR (Germany), BAG (Switzerland) and Health Canada.

The program provides insight in exposure to substances in consumer products via multiple exposure routes. Users can choose the most appropriate consumer exposure scenario and use the default exposure models and exposure parameters set therein (from a database based on several product category Fact Sheets). Alternatively, users can perform an exposure assessment with user-specified exposure parameters. The program consists of both screening models and higher tier models for exposure estimation, and is referenced by REACH and other regulatory programs.

Afternoon session (13.00-16.00): Comparison of Consumer Exposure Models and Related Case Studies

Several models are developed to provide exposure estimates for chemicals in consumer products. These models are developed for different purposes, and have different capabilities (single chemical vs multiple chemicals), product representation (broad categories of products vs detailed product and use scenarios), and number of parameters required.

To get more insight in the similarities and differences with regard to model algorithms (under the hood), as well as choice of default values for exposure parameters, a comparison of ConsExpo Web (RIVM) with the Consumer Exposure Model CEM (U.S. EPA) is an interesting challenge.

Two case studies on a chemical in a consumer product will be put forward, one focusing primarily on a dermal exposure assessment, and one an inhalation exposure assessment. Participants will perform the exposure assessments, using CEM as well as ConsExpo Web.

Discussion will be facilitated on the similarities and/or differences in outcomes with regard to model algorithms as well as assumptions in the exposure scenarios. This should provide some insight (for a selected scenario) between the two models and can aid users in future exposure assessments when using exposure models (in general and for CEM and ConsExpo Web in particular).

PC03: Bayesian methods for environmental health researchers

Ghassan B Hamra, Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health; Baltimore, USA

Description

Purpose of the course including specific learning objectives:

The goal of this course will be to provide a primer to the rationale and use of Bayesian statistical tools for environmental health research. The course will consist of a modest didactic component and will then allow participants to practice Bayesian analyses using R statistical software. Participants should have a basic working knowledge of R statistical software and will have installed RStudio and the Just Another Gibbs Sampler softwares as well as the rjags package on their laptops (which will be necessary for the practical exercises that will be included); instructions for installation of software can be found here.

Outline (3 hours)

Part 1:

  • Didactic
  • Bayes philosophy and why everyone is basically Bayesian
  • The case for Bayes in environmental health research

Part 2:

  • Practical
  • Coding a simple linear regression model
  • Applying a prior
  • Diagnosing model convergence

PC04: Application of New Approach Methodologies for Exposure Assessment and Prioritization – Tools for Researchers and Regulators Including use of Quantitative Structure Use Relationships (QSUR)

John Wambaugh, National Center for Computational Toxicology, U.S. Environmental Protection Agency; Research Triangle Park, USA
Katherine Phillips, Computational Exposure Division, National Exposure Research Laboratory, U.S. Environmental Protection Agency; Research Triangle Park, USA
Kristin Isaacs, Human Exposure and Atmospheric Sciences Division, National Exposure Research Laboratory, U.S. Environmental Protection Agency; Research Triangle Park, USA

Description

Purpose of the course including specific learning objectives:

This course will cover new approach methodologies for exposure assessment as available from the EPA’s CompTox Chemistry Dashboard (http://comptox.epa.gov). The objective of the course is to provide the attendees with specific examples of how Dashboard tools can be used to obtain innovative and up-to-date exposure information for chemicals in support of exposure assessment or non-targeted analyses.

We will briefly introduce high-throughput exposure modeling, including consensus exposure modeling results from the Systematic Empirical Evaluation of Models (SEEM) analysis of exposure tools such as SHEDS-HT and other near-field models developed under the ExpoDat initiative. We will then present a brief overview of non-targeted analysis (NTA) and demonstrate Dashboard tools and workflows for “exposure forensics” of chemical unknowns in NTA. We will discuss the chemical and product use data in CPDat, the US EPA’s Chemical-Product database, as well as the Dashboard mass search, data source rankings, and and retention-time prediction tools. Finally, we will describe machine-learning based Quantitative Structure Use Relationships (QSUR) models for chemical function, which were developed to fill gaps in existing databases.  We will present an NTA case study demonstrating how these QSUR models can inform NTA.

Outline (3 hours)

This session will include:

  1. Public Exposure Information on the CompTox Dashboard (John Wambaugh)
  2. A Chemical Forensics Workflow for Non-Targeted Analysis (Kristin Isaacs)
  3. Quantitative Structure Use Relationships (QSUR) for Exposure Assessment and Application in NTA (Katherine Phillips)

PC05: Model-Based Geostatistics and Spatial Epidemiology: a practical introduction with R

Patrick Brown, Centre for Global Health Research, St Michael’s Hospital, Department of Statistical Science, University of Toronto, Toronto, Canada

Description

Purpose of the course including specific learning objectives:

Participants will become familiar with methods and tools for use with spatially referenced data of the type frequently encountered in environmental epidemiology research and described in Brown (2016). A typical problem involves a study population where health outcome (i.e. lung cancer) is presumed to depend on an environmental exposure (air pollution) and one or more spatially varying counfounding variables (neighbourhood-level income and ethnic distribution). A standard linear model (logistic regression or survival model) would assume individual’s health outcomes are independent of one another given the explanatory variables. The Generalized Linear Geostatistical Model (GLGM) allows for the possibility that some form of spatial dependence (or autocorrelation) may be present, possibly due to an unknown or unmeasured risk factor. Fitting a geostatistical model allows for this residual spatial variation in risk to be estimated and mapped, and takes spatial dependence into account when inferring the effect of the exposure.

The emphasis of the course will be on understanding the GLGM and the results produced when using it. The first two hours of the course will be lecture-style, and a number of examples (air pollution and mortality in India, cancer survival in north-west England) will be worked through. The final hour will be a practical session where code will be provided to fit the GLGM to one or more datasets using the geostatsp package (see Brown 2015). The emphasis of this session will be on visualizing and understanding the results, and exploring how changing the modelling assumptions affects the conclusions.

Outline (3 hours)

Models and methods

  • Spatial correlation and Gaussian Random Fields
    – Simulating random fields
    – Understanding spatial correlation functions
  • Generalized Linear Geostatistical Models for spatially referenced data
  • Linear regression models with spatial random effects
  • A model for relating mortality to air pollution in India
  • Bayesian inference, Fitting models to data
    – 
    Bayesian inference and prior distributions
    – Making and mapping spatial predictions
    – Posterior distributions of model parameters

Practical session

  • Data and R scripts will be provided
  • Coffee shops and elections in Toronto
  • Does the downtown intellectual elite drink lattes?
  • Soil mercury in Europe
  • Where is it the highest?

PC06: Predicting microscale urban features using street-level images – an introduction to machine learning

Mahdi Shooshtari, Canadian Urban Environmental Health Research Consortium (CANUE) Department of Geography, University of Victoria; Victoria, Canada
Joseph Paul Cohen, Montreal Institute for Learning Algorithms, University of Montreal, Montreal, Canada
Evan Seed, Canadian Urban Environmental Health Research Consortium (CANUE), Dalla Lana School of Public Health, University of Toronto; Toronto, Canada

Description

The Canadian Urban Environmental Health Research Consortium is developing a wide range of metrics related to urban form for approximately 800,000 postal codes in Canada, annually from the early 1980s onward. One area of interest is identifying the local climate zones, which are defined according to building and vegetation type, height and density, for every postal code. In this course, CANUE specialists will guide participants through a hands-on exercise using street-level images to create a training dataset for local climate zones, and then categorize a set of postal code locations into local climate zones.  The process, software and scripts provided/created in the course can later be used by participants on their own for identifying many other kinds of objects in images.

Outline (3 hours)

  • Overview of course objectives
  • Introduction to machine learning for image processing
  • Building a tensor flow image classifier
  • Creating training data – categorizing urban form by postal code
  • Training the neural network and applying to new images
  • Q/A, group discussion of application ideas, strengths and limitations

PC07: Causal inference foundations and applications in environmental health sciences

Jay Kaufman, Department of Epidemiology, Biostatistics and Occupational Health, McGill University; Montreal, Canada
Alexander Keil, Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina; Chapel Hill, USA

Description

Purpose of the course including specific learning objectives:

This course will present emerging methodologic challenges in environmental epidemiology and introduce foundational concepts and practical approaches for addressing these challenges. Our focus is on introducing causal concepts that sharpen the common approaches to data analysis, and motivating the use of modern statistical methods of inverse probability weighting, g-estimation, and g-computation. These methods will be applied to environmental problems, with particular focus on assessing health effects of long-term exposure and estimating policy impacts from observational studies of non-representative populations. Participants will gain a theoretical background for deciding between methods, and will be provided with SAS, Stata, and R code for applying each approach. At the end of the course, participants will be aware of ways in which their current research questions may benefit from modern approaches to data analysis, and in which ways these modern approaches may allow them to ask and answer new questions about their data. Participants will gain skills in interpreting software output from existing packages, as well as implementing causal inference analyses using standard software packages. SAS, Stata, and R scripts will be provided to all participants.

Outline

The first part of the course will be devoted to theoretical and conceptual descriptions of the proposed methods with concrete examples. This session will include:

  1. Definitions and identification of causal effects
  2. Inverse probability weighting and marginal structural models for point-exposures and time-varying exposures
  3. G-estimation and structural nested models
  4. G-computation and policy impact estimation

The second section of the course will be devoted to hands-on course with instructor-led discussions and guided exercises. R, SAS and Stata scripts and specific functions fromguided exercises will be provided to all participants.

PC08: Advanced modelling techniques for time series analysis using R

Antonio Gasparrini, Ana Maria Vicedo-Cabrera and Francesco Sera
Department of Social and Environmental Health Research, London School of Hygiene and Tropical Medicine; London, UK

Description

Purpose of the course including specific learning objectives:

Time series analysis has become a key tool for investigating short-term effects of environmental risk factors. In the last two decades, there has been an intense activity to develop more sophisticated study designs and statistical models for using time series data in this context. This course will offer an overview of recent methodological advancements, focusing on their application through the statistical software R. Participants will be provided with a theoretical introduction, as well as practical experience with a hands-on session using real-data examples, using a mix of mini-lectures and mini-practicals. At the end of the course, participants will be able to apply state-of-the-art methodologies for time series analysis using R, and will gain skills to correctly interpret software outputs and conduct these analyses within their own research projects. R scripts and specific functions will be provided to all participants.

Outline (3 hours)

The session will involve a mix of mini-lectures and mini-practicals on the various topics covered in the course, including illustrative examples and real-data analyses. The session will cover:

  1. Introduction to time series analysis with R
  2. Study designs and statistical models for time series analysis: an overview of packages and functions in R
  3. Modelling non-linear and delayed effects: an introduction to distributed lag linear and non-linear models and the R package dlnm
  4. Pooling results in two-stage multi-location analyses: multivariate meta-analysis and meta-regression and the R package mvmeta
  5. From aggregated to individual-level time series analysis: an introduction to the novel case time series design