Symposia




Advocacy vs science: The inherent tensions in community engagement in epidemiologic research

Session Chair(s):   Nichole Austin,   Patricia O'Campo
Date: 2018-06-20       Time: 10:00 am - 11:30 am
Location: President


     

Community based participatory research (CBPR) has gained popularity as a way to improve the questions we ask, identify appropriate populations, and interpret study findings. Done well, CBPR can enhance the research and give community members a stake in the findings. However, community engagement is not without its challenges. For example, the extent to which communities participate is context- and study-specific, which requires careful planning. Also, ethical challenges are particularly common: when community members are invested in a specific research outcome, it can create tension and blur the lines between advocacy and science, introducing opportunities for bias. Finally, as CBPR is rooted in the community, it is reasonable to question the generalizability of findings. When does CBPR lead to useful research outputs, and what should researchers look out for? In this session, we discuss the ethical and practical challenges of balancing community priorities and scientific integrity.

Session Chairs:
Nichole Austin, McGill University
Patricia O'Campo, University of Toronto


Presenters:
Matthew Fox, Boston University
"Introductory remarks"
Patricia O'Campo, University of Toronto
"Ethics and community engagement in research"
Michael Oakes, University of Minnesota
"Do community-based approaches improve or hinder the quality of research outputs?"
Sacoby Wilson, University of Maryland
"Use of community-engaged research to study and address environmental health disparities"
Nichole Austin, McGill University
"Does involving the community compromise objectivity?"


Is your soup toxic or nutritious? Methods for assessing exposure mixtures in environmental and nutritional epidemiology

Session Chair(s):   Anna Pollack,   Jessie Buckley
Date: 2018-06-20       Time: 10:00 am - 11:30 am
Location: Chesapeake B


     

Nutritional and environmental epidemiologists are faced with challenges, particularly regarding methods to address mixtures of related exposures. Novel statistical approaches to estimate effects of real world mixtures are developing in parallel. The application of these methods has lagged despite widespread recognition of the inadequacy of single exposure approaches. Yet, crosstalk between fields is limited. This symposium will describe novel methods to address exposure mixtures with applications in nutritional and environmental epidemiology. These methods include: lagged kernel machine regression, Bayesian g-formula, weighted quantile sum regression, dietary pattern analysis, and Bayesian hierarchical modeling. Speakers will describe the theory and implementation of their approach. This session will foster dialogue across disciplines and will be of broad interest to SER members seeking to estimate effects of multiple exposures.

Session Chairs:
Anna Pollack, George Mason University
Jessie Buckley, Johns Hopkins School of Public Health


Presenters:
Shelley Liu, Icahn School of Medicine at Mount Sinai
"Bayesian methods to assess the impact of chemical mixtures on neurodevelopmental trajectories"
Jill Reedy, National Cancer Institute
"Multidimensionality and dynamism: extending methods in dietary patterns research"
Alexander Keil, University of North Carolina
"Fixing the problem by changing the question: leveraging correlated exposures to estimate impacts of public health actions"
Patrick Bradshaw, University of California, Berkeley
"Bayesian hierarchical modeling of correlated exposures"
Chris Gennings, Icahn School of Medicine at Mount Sinai
"Using methods for environmental mixtures in nutrition research: Prenatal nutrition and neurodevelopment"


Suboptimal Sleep as an Understudied but Fundamental Contributor to Poor Health and Health Disparities: Challenges and Opportunities for All Epidemiological Research

Session Chair(s):   Chandra L Jackson,   Dayna A Johnson
Date: 2018-06-20       Time: 10:00 am - 11:30 am
Location: Chesapeake A


     

Sleep is an essential human need that is positively or negatively affected by modifiable social and environmental factors. Suboptimal sleep is associated with poor health outcomes, and well-observed yet underrecognized racial/ethnic disparities in sleep health may be fundamental contributors to recalcitrant health disparities. To advance the effectiveness of epidemiological research, it is critical to broaden and deepen the field’s understanding of the importance of incorporating sleep into all research along with its related challenges (e.g. measurement). This session will include an overview of (1) epidemiological findings regarding sleep health and its likely contribution to health and health disparities; (2) social and environmental determinants of sleep health; (3) challenges related to measuring sleep; (4) the validity and reliability of using wearable/biosensor devices to assess sleep in large-scale epidemiological studies; and (5) research opportunities for studying sleep.

Session Chairs:
Chandra L Jackson, National Institute of Environmental Health Sciences
Dayna A Johnson, Harvard Medical School


Presenters:
Chandra L Jackson, National Institute of Environmental Health Sciences
"Suboptimal Sleep as a Fundamental Contributor to Poor Health and Health Disparities"
Dayna A Johnson, Harvard Medical School
"Social and Environmental Determinants of Sleep Health"
Diane S Lauderdale, University of Chicago
"The Measure and Mismeasure of Sleep"
Peter James, Harvard Medical School
"Measuring Sleep with Consumer Wearable Devices"


Misclassification of underrepresented groups: Caring about smaller categories

Session Chair(s):   Nadia Abuelezam
Date: 2018-06-20       Time: 10:00 am - 11:30 am
Location: Frederick/Columbia


     

Sociodemographic information, collected through routine demographic data collection are summarized using categories and adjusted for in analyses. However, routine sociodemographic surveillance is often crudely categorical and individuals of underrepresented groups do not often have the choice of choosing an appropriate category (masking potential risks and exposures). Therefore, misclassification of sociodemographics is common for underrepresented groups and is often ignored because the magnitude of this misclassification is expected to be small (due to small group sizes) and “gold standard” information is unattainable. The extent to which this misclassification affects large study findings is unknown. In this symposium we hope to understand the potential impact of misclassification for a number of underrepresented groups without standardized sociodemographic categories on surveys and medical forms and explore the ethical and social consequences of this misclassification.

Session Chairs:
Nadia Abuelezam, Boston College


Presenters:
Brittany Charlton, Harvard University
"Assessment, misclassification, and politicization of sexual orientation data"
Sari Reisner, Harvard University
"Transgender population health: Methods, measures, and misclassification"
Nadia Abuelezam, Boston College
"Categorizing race: Discordance and misclassification challenges"
Discussants:
Sandro Galea, Boston University


Machine Learning in Epidemiologic Science

Session Chair(s):   Ashley Naimi
Date: 2018-06-20       Time: 10:00 am - 11:30 am
Location: Constellation B


     

Machine learning (ML) methods and/or data adaptive techniques are fast becoming trends in epidemiology. Though they have much to contribute, these approaches are surrounded by hype and misunderstanding. After a brief introduction (Ashley Naimi), we will delve into three issues related to ML. Edward Kennedy will discuss little recognized inferential challenges, and how these can be improved via doubly robust estimators. Machine learning algorithms are computationally intensive and difficult to implement with “big” (N or p) data. Eric Polley will discuss scalability and how to address big data challenges using state-of-the-art computing resources. Finally, though often assumed “objective,” recent work suggests that ML methods may be biased in minority populations. Alexandra Chouldechova will discuss the presence, evaluation, and handling of “algorithmic bias” in the applied sciences. All speakers have agreed to participate.

Session Chairs:
Ashley Naimi, University of Pittsburgh


Presenters:
Ashley Naimi, University of Pittsburgh
"Machine Learning in Epidemiology: A Magical Mystery Tour"
Edward Kennedy, Carnegie Mellon University
"Machine learning in causal inference: balancing bias and variance"
Alexander Chouldechova, Carnegie Mellon University
"Fairness and bias in predictive modelling"
Eric Polley, Mayo Clinic
"Scalable Machine Learning for Epidemiological Studies"
Discussants:
Maria Glymour, University of California, San Francisco


Changes in United States health policy: implications for substance use and injury

Session Chair(s):   Katherine Keyes
Date: 2018-06-20       Time: 10:00 am - 11:30 am
Location: Baltimore/Annapolis


     

Substantial changes have taken place in US health policies in the past ten years, including the enactment of the Affordable Care Act, legalization of medical and recreational marijuana use, restrictions of access to opioids, and changes in the regulations related to access to firearms for high-risk groups. We illustrate how a range of methods can be used to infer the causal effects of major U.S. policy changes on substance use and injury, including synthetic control group estimation, agent-based modeling, spatio-temporal models, and randomized trials. Current evidence as well as the advantages and disadvantages of each analytic approach for the specific health policy questions will be discussed.

Session Chairs:
Katherine Keyes, Columbia University


Presenters:
Bryce Pardo, RAND
"Modeling the effect of health insurance on substance use: Lessons from the ACA's Dependent Care Expansion"
Katherine Keyes, Columbia University
"Consequences of medical and recreational cannabis legislation on marijuana and opioid-related harms"
Magdalena Cerda, University of California, Davis
"Prescription opioid and heroin regulations and opioid overdose in the United States: a spatio-temporal approach to policy evaluation"
Hannah Laqueur, University of California, Davis
"The Armed and Prohibited Persons System in California: a randomized controlled trial evaluation of its impact on firearm violence"
Jennifer Ahern, University of California, Berkeley
"Effects of the Mental Health Services Act on suicide in California: a synthetic control approach"


Virtuous circles: new ideas at the intersection of observational and randomized studies

Session Chair(s):   Daniel Westreich,   Michael Kosorok
Date: 2018-06-20       Time: 10:00 am - 10:30 am
Location: Constellation A


     

While epidemiologists are greatly concerned with methodology for observational studies, they sometimes regard randomized trials as existing more in the domain of biostatistics than epidemiology. Of course, the two approaches to understanding the world have a great deal in common, and much to teach each other. In this symposium, organized by an epidemiologist and biostatistician and featuring speakers from epidemiology, biostatistics, and statistics departments, we will discuss issues at the intersection of observational and randomized studies. Specifically we will introduce and give an overview of the topic (Westreich); discuss the generalizability of randomized trials (Lesko) and the implications of those lessons to trial designs (Stuart); address innovative trial designs and analysis informed by observational data analysis (Laber); and offer some high-level takeaways for epidemiologists conducting trials or working with trialists (Kosorok)."

Session Chairs:
Daniel Westreich, University of North Carolina at Chapel Hill
Michael Kosorok, University of North Carolina at Chapel Hill


Presenters:
Catherine Lesko, Johns Hopkins Bloomberg School of Public Health
"Generalizability of randomized trials"
Elizabeth Stuart, Johns Hopkins Bloomsburg School of Public Health
"Trial designs for achieving internal and external validity"
Eric Laber, North Carolina State University
"Innovative trial design and analysis"
Discussants:
Michael Kosorok, University of North Carolina at Chapel Hill


Trial Emulation Using Observational Studies: The Case of Aging Outcomes

Session Chair(s):   Michelle Odden,   Adina Zeki Al Hazzouri
Date: 2018-06-21       Time: 2:15 pm - 3:45 pm
Location: Baltimore/Annapolis


     

Observational epidemiology is often held inferior to randomized trials, largely due to inappropriate control for confounding. A growing body of literature considers statistical methods to estimate a counterfactual contrast for the treatment of interest without the benefit of randomization. Trial emulation explicitly aims at recreating the design principles from a randomized trial in analysis of observational data. Such principles include restriction of the study population to those eligible for the treatment and restriction to new users of a treatment to allow adjustment for pre-treatment levels of covariates. Trial emulation does not only improve the quality of observational data but is also useful when a trial is not feasible or not yet available. This is an important advantage in studies of older adults, where trial data are limited, especially among those with multiple chronic conditions and poor functional status. Data from observational studies may offer an unprecedented opportunity for emulating target trials of interest. Applied examples in studies of aging will help elucidate some of the real-world benefits and challenges of implementation of these methods.

Session Chairs:
Michelle Odden, Oregon State Unversity
Adina Zeki Al Hazzouri, University of Miami


Presenters:
Sonja Swanson, Erasmus MC, Rotterdam, the Netherlands
"Mendelian randomization and trial analogues in elderly populations"
Miguel Hernan, Harvard University
"How to choose time zero when emulating a target trial using observational data"
Adina Zeki Al Hazzaouri, University of Miami
"Statins and cognitive decline: Methods to address bias in observational studies"
Maria Glymour, University of California, San Francisco
"Closing the gap between observational and randomized research on prevention of Alzheimer's Disease"


The Utility and Feasibility of New Indicators of Health Inequality

Session Chair(s):   Zinzi Bailey,   Lorraine Dean
Date: 2018-06-21       Time: 2:15 pm - 3:45 pm
Location: President


     

In recent years, several scholars have proposed novel measures of socioeconomic position (SEP), race, and discrimination as indicators of health inequality; however, the usefulness of these new measures in academic research and beyond--to government, non-profit, policy makers, healthcare providers--is largely unknown. The symposium session will address several questions: What are the strengths and limitations of newer measures relative to traditional measures? What is the feasibility of using these novel measures in academia, government, nonprofit, or private sectors? What unintended consequences may result from their use in these different sectors? Focusing on real-world applications to how we study health and disease, this symposium introduces fresh perspectives from an inter-disciplinary panel of speakers from epidemiology, government, and the social sciences. Audience members will vote on new which measures might be useful, and meet in breakout groups by work sector to discuss.

Session Chairs:
Zinzi Bailey, University of Miami Miller School of Medicine
Lorraine Dean, Johns Hopkins School of Public Health


Presenters:
Asad L Asad, Cornell University
"Racialized Legal Status as a Social Determinant of Health"
Matthew Clair, Harvard University, University of Pennsylvania Law School
"Racialized Legal Status as a Social Determinant of Health"
Junia Howell, University of Pittsburgh
"Selecting Reimagined Variables: Using Empirical Evaluation Techniques to Choose Measures of Race, Poverty and Inequality"
Lorraine Dean, Johns Hopkins School of Public Health
"Consumer Credit and Health"


Rural health in the US: Do we have the right measures for the right questions?

Session Chair(s):   Christine Gray,   Annemarie Hirsch
Date: 2018-06-21       Time: 2:15 pm - 3:45 pm
Location: Chesapeake A


     

While the opioid crisis has brought rural health in the U.S. into sharp focus, the health gap between Americans living in rural vs. urban areas is stark across many leading health indicators (heart disease, cancer, injury), and has been growing. This widening gap demands that epidemiologists step back and clarify what we know, what we do not know, and our current limitations with respect to these populations and the communities in which they live. Our goal is to bring together speakers with diverse expertise to engage this important topic. Speakers will discuss their current research as it relates to rural health, emphasizing 1) what makes rural populations unique, 2) whether we have the right tools (e.g., social/environmental measures, data) to understand rural health, 3) critical methodological issues in rural health (e.g., scale selection, separation of contextual and compositional effects), and 4) questions we should be asking but are not.

Session Chairs:
Christine Gray, University of North Carolina at Chapel Hill
Annemarie Hirsch, Geisinger, Department of Epidemiology and Health Services Research


Presenters:
Kristen M Rappazzo, US Environmental Protection Agency
"Context matters: what metrics are used where for socioeconomic status and position."
Katherine Keyes, Columbia Unversity
"Are urban and rural health differences due to exposure prevalence variation or interaction: when and why does it matter?"
Brian S Schwartz, Johns Hopkins Bloomberg School of Public Health
"Environmental epidemiology in geographies with a range of place types: what does rural have to do with it?"
Lynne C Messer, OHSU-PSU School of Public Health
"Missing data in rural health: No literally. Its missing."
Marissa Seamans, Johns Hopkins School of Public Health
"Opioids in rural America"


Marching towards progress in causal inference for health disparities research

Session Chair(s):   John Jackson,   Chanelle Howe
Date: 2018-06-21       Time: 2:15 pm - 3:45 pm
Location: Frederick/Columbia


     

Recent work in the causal inference literature has laid the groundwork for asking questions about how disparities might change upon intervening to remove disparities in determinants of the outcome. Though much of the causal inference tools can be re-expressed for these problems, the adaptation may not always be straightforward for practical or conceptual reasons. For example, data may be lacking for the social groups considered; populations defined by disease illness, geography, or insurance plan enrollment may be of substantive interest but may open the door to selection-bias; and estimators may provide precise answers but to the wrong questions. This symposium will present ongoing work from four causal inference methods researchers who have considered these issues or similar challenges, providing up-to-date guidance on how to make progress in health disparities research despite these challenges. A discussant will consider the proposals for circumventing the identified challenges.

Session Chairs:
John Jackson, Johns Hopkins School of Public Health
Chanelle Howe, Brown University


Presenters:
Ashley Naimi, University of Pittsburgh
"Construct Validity and Causal Inference: On the Measurement of Social Causes"
Whitney Robinson, University of North Carolina, Chapel Hill
"Back to basics: What was the question again?"
John Jackson, Johns Hopkins School of Public Health
"Explaining the disparity of interest when confounders are associated with race/ethnicity"
Chanelle Howe, Brown University
"Minimizing survival-related and other sources of selection bias in studies of racial health disparities: the importance of the target population and study design"
Discussants:
Linda Valeri, Harvard Medical School


Balancing internal and external validity – can one really have it all?

Session Chair(s):   Catherine Lesko,   Matthew Fox
Date: 2018-06-21       Time: 2:15 pm - 3:45 pm
Location: Constellation A


     

Transportability and generalizability have received considerable attention over the past decade as epidemiologists have begun to develop methods and articulate assumptions sufficient to go beyond estimating effects in their own study sample, to think about impacts on the larger populations from which their data arose or on external populations. This new emphasis takes the inward focus on epidemiologic research and training (which focuses on internal validity) and redirects it outward to external validity and population impact. While the new focus has important implications for how we think about interventions to improve the health of populations, epidemiologists may be less comfortable making the assumptions sufficient to generalize or transport effects. Namely, can we ever fully understand all the drivers of population level variation in effects? This session seeks to better understand the relationship between internal and external validity through point-counterpoint presentations.

Session Chairs:
Catherine Lesko, Johns Hopkins School of Public Health
Matthew Fox, Boston University


Presenters:
Catherine Lesko, Johns Hopkins School of Public Health
"Overview of the relationship between internal and external validity"
Laura Balzer, University of Massachusetts
"Introduction to and overview of the distinction between generalizability and transportability"
Til Sturmer, University of North Carolina, Chapel Hill
"Response: Representative sampling is not necessary"
Trang Nguyen, Johns Hopkins School of Public Health
"Sensitivity analyses for external validity"
Liz Stuart, Johns Hopkins School of Public Health
"Response: parallels with sensitivity analyses for internal validity"
Ellie Murray, Harvard University
"Implications of G-formula and agent based models: A Practical Example"
Jessie Edwards, University of Carolina, Chapel Hill
"Response to Implications of G-formula and agent based models"
Daniel Westreich, University of North Carolina, Chapel Hill
"Where are we headed with external validity?"


Challenges for Women in Epidemiology and Systemic and Equitable Solutions

Session Chair(s):   Stephanie Engel,   Andrew Olshan
Date: 2018-06-21       Time: 2:15 pm - 3:45 pm
Location: Constellation B


     

SER 2017 featured a provocative session on gender bias in epidemiology publications, editorial boards, and journal leadership, mirroring well-documented disparities in gender inclusion found across STEM fields. Continued focus on issues of equity/inclusion is essential in order to develop a comprehensive equity/inclusion strategy for professional epidemiology societies. This symposium will address systems-level challenges of equity/inclusion in epidemiology, including implicit biases, salary equity, advancement, work-life balance, and leadership opportunities. The focus will be eliciting solutions with the aim creating equitable work environments. The symposium will include an expert-led, facilitated discussion, and incorporate input from a distinguished panel. Audience participation will play a key role in the conversation. To facilitate discussion during the symposium, a plenary speaker will give a keynote on the topic of women in science prior to the symposium.

Session Chairs:
Stephanie Engel, UNC Chapel Hill
Andrew Olshan, UNC Chapel Hill


Discussants:
Donna Dean, Association for Women in Science
Diane Lauderdale, University of Chicago
Kay Dickersin, Johns Hopkins School of Public Health
Martha Werler, Boston University


Epidemiological approaches to identify policy solutions for the opioid epidemic

Session Chair(s):   Magdalena Cerda
Date: 2018-06-21       Time: 2:15 pm - 3:45 pm
Location: Chesapeake B


     

Opioid overdose deaths have quadrupled since 1999, and the epidemic worsens every year. The epidemic has shifted from a prescription opioid epidemic to an epidemic driven by synthetic opioids and heroin. Its shifting nature has required a rapid response to identify effective policy solutions. Our symposium will present some of the latest applications of epidemiologic methods to identify effective responses responses to the opioid epidemic. We will highlight the types of prescription drug monitoring program features that lead to the greatest reduction in opioid overdose, the joint effects that prescription opioid policies and marijuana laws have on opioid overdose, the impact of physician payments by the pharmaceutical industry on prescribing behaviors and opioid overdoses, and public health policies to prevent transitions to heroin use and mitigate related harms. By identifying effective policies targeting both prescription opioids and heroin, solutions to this epidemic may emerge.

Session Chairs:
Magdalena Cerda, University of California at Davis


Presenters:
Magdalena Cerda, University of California at Davis
"Prescription drug monitoring programs and their impact on prescription opioid and heroin overdose"
June Kim, New York University
"Joint effects of state medical marijuana laws and prescription drug monitoring programs on opioid overdose mortality among National Health Interview Survey participants, 1986-2011"
Ariadne Rivera-Aguirre, University of California at Davis
"Pharmaceutical industry payments to physicians and changes in prescribing behaviors and opioid overdoses in the United States"
Brandon Marshall, Brown University
"Epidemiologic approaches to evaluate public health policies that prevent transitions to heroin and mitigate related harms"
Discussants:
Pia Mauro, Columbia University


Doctoral education: What are we teaching and what are we missing?

Session Chair(s):   Matthew Fox,   Timothy Lash
Date: 2018-06-22       Time: 10:00 am - 11:30 am
Location: Constellation A


     

As epidemiologic methods further develop, topics like causal inference methods (directed acyclic graphs, instrumental variables, g-estimation, mediation analysis), Bayesian statistics, and transportability, vie for space in crowded doctoral curricula. Epidemiology is also looking to gain insights from fields like genomics, decision science, systems dynamics modelling, computer science and econometrics to expose students to more tools and approaches. At the same time, many doctoral students do not seek academic careers. NIH estimates 9 of 10 will work in industry, government public health agencies, consulting, contract research organizations, or not-for profits. Teaching these methods at the expense of more applied topics–such as sampling, survey design, community-engaged research and surveillance–may not be optimal preparation for the careers many students seek. We will present views on both sides of these trends and discuss what the future of doctoral training should look like.

Session Chairs:
Matthew Fox, Boston University
Timothy Lash, Emory University


Presenters:
Miguel Hernan, Harvard University
"How to learn what works (causal inference methods)"
Maria Glymour, University of California, San Francisco
"How to evaluate causality when RCTs are not feasible"
Laura Rosella, University of Toronto
"Alternative approaches to generating control groups"
Barbara Mahon, CDC
"How to do surveillance, how to recognize a consequential question, and how to choose the right methods to get a useful answer"
Brian Bradbury, Center for Obervational Research at Amgen
"How to make your epidemiologic research matter to regulatory and reimbursement authorities?"
Demian Christiansen, Cook County Department of Public Health
"All I Really Need to Know I Learned in Grad School: DIY"


Epidemiologic research with incomplete and imperfect data: making progress in the face of uncertainty

Session Chair(s):   Jessie Edwards
Date: 2018-06-22       Time: 10:00 am - 11:30 am
Location: Frederick/Columbia


     

Epidemiology informs public health action. But decisions are often required in settings where epidemiologic data are limited, incomplete, or nonexistent. Traditional epidemiologic methods may fail to produce results in such settings, and ad-hoc approaches may yield misleading results. How can epidemiology guide decision making when the paucity or inadequacy of data hampers use of our standard tools? This symposium will explore principled approaches for inference in real world settings in which use of standard methods is limited due to insufficient data. Examples include sparse data, incomplete networks, and unmeasured covariates, exposures, or outcomes. Speakers will address both philosophical questions (e.g., how to balance concerns about data quality with the need to answer urgent public health questions) and practical issues that arise in diverse substantive areas and under a variety of methodological paradigms (e.g., network analysis, systems science, and causal inference).

Session Chairs:
Jessie Edwards, University of North Carolina, Chapel Hill


Presenters:
Tiffany Breger, University of North Carolina, Chapel Hill
"Two-stage g-computation with missing exposure"
Eric Lofgren, Washington State University
"Crossing Data Gaps with Imaginary Bridges"
Kathryn Risher, London School of Hygiene and Tropical Medicine
"What can we infer about network structure from incomplete network data?"
Alexander Keil, University of North Carolina, Chapel Hill
"Little data inference: fitting marginal structural models with sparse strata"
Justin Lessler, Johns Hopkins School of Public Health
"Public health action in the face of uncertainty, where the rubber hits the road"


Big data epidemiology: sampling frames for 21st century data sources

Session Chair(s):   Steven Mooney
Date: 2018-06-22       Time: 10:00 am - 11:30 am
Location: Baltimore/Annapolis


     

Recent growth in diversity and availability of secondary data has broadened epidemiology’s scope and efficiency. Large databases of health records, environmental exposures, biological and neurological samples, and ‘effluent data’ drawn from internet use (Tweets, search terms, etc.) allows study of human experiences at unparalleled detail and scale. Yet the sample frame underlying many of these sources remains unclear. Epidemiologic theory suggests that inference requires defining the source population underlying data sources and the probability of selecting particular participants for observation. When using secondary ‘big data’, the underlying source population can be difficult to quantify. In this symposium, we will discuss issues of sampling frames in big data epidemiology, including how issues of sampling frame matter (and when they do not), and how we can develop a set of best practices for robust and consequential inference from rich secondary data sources.

Session Chairs:
Steven Mooney, University of Washington


Presenters:
Daniel Westreich, University of North Carolina, Chapel Hill
"Big Data or Big Anecdotes?"
Elizabeth Rose Mayeda, University of California, Los Angeles
"Does selection matter in research on racial/ethnic inequalities?"
Ali Rowhani-Rahbar, University of Washington
"Epidemiology of Crime and Elusive Sampling Frames"
Katherine Keyes, Columbia University
"Who thinks like that? Survey methods for non-survey data"
Discussants:
Jay Kaufman, McGill University


My breadth is better than your depth, or, can we get the best of both? Lessons and opportunities for causal inference integration across study designs from behavioral and mental health

Session Chair(s):   Daniele Fallin,   Elizabeth Stuart
Date: 2018-06-22       Time: 10:00 am - 11:30 am
Location: Chesapeake A


     

Epidemiologic studies of etiologic factors for mental health use a broad range of study designs. Each design, from small, deeply characterized, cohorts to large population-level registries has particular strengths and limitations. The field has not integrated causal inference across designs to fully capitalize on the strengths of each. We propose a panel discussion with three “Acts”. An argument for depth, often at the sacrifice of breadth, sample size, or generalizability. An argument for breadth, often sacrificing deep measures of exposure, outcomes, or context factors. deas for causal inference integration across designs, harnessing the strengths offered by each while minimizing the limitations. We hope to call upon the field to integrate across designs via existing and novel statistical approaches to enable better causal inference on effect sizes and confounding, as well as better accommodation of mediation and effect modification.

Session Chairs:
Daniele Fallin, Johns Hopkins School of Public Health
Elizabeth Stuart, Johns Hopkins School of Public Health


Presenters:
Daniele Fallin, Johns Hopkins School of Public Health
"The value of traditional prospective cohorts"
Peter Zandi, Johns Hopkins School of Public Health
"Learning health systems for clinical psychiatric epidemiology"
Lisa Croen, Kaiser Permanente
"Opportunities for developmental disabilities research in an integrated healthcare delivery system"
Colleen Barry, Johns Hopkins School of Public Health
"Using Administrative Claims Data to Inform Policy and Improve Population Health"
Diana Schendel, Aarhus University
"Epidemiologic insights drawn from total population registry-based research"
Nilanian Chatterlee, Johns Hopkins School of Public Health
"Generalized Meta-Analysis for Multivariate Regression Models Across Studies with Disparate Covariate Information"


Novel Methodological Approaches in the Study of Work and Health

Session Chair(s):   Ellen Eisen,   Sadie Costello
Date: 2018-06-22       Time: 10:00 am - 11:30 am
Location: President


     

Occupational class is a fundamental axis of social stratification that together with workplace chemical and physical exposures, presents a critical set of health determinants. Importantly, many occupational risk factors are plausibly modifiable through changes in permissible exposure limits or organizational practice. Moreover, longitudinal studies of occupational cohorts pose unique methodological challenges to investigators due to worker self-selection leading to the healthy worker survivor effect. This session will showcase research that addresses trenchant questions in the study of the organizational, social, and chemical exposures at work, with a particular focus on the public health impact of interventions, both real and hypothetical through application of novel epidemiological research methods.

Session Chairs:
Ellen Eisen, University of California, Berkeley
Sadie Costello, University of California, Berkeley


Presenters:
Ashley Naimi, University of Pittsburgh
"The logic of causal inference3"
Andreas Neophytou, University of California, Berkeley
"Assessing hypothetical interventions using observational data: envisioning regulatory steps to limit occupational exposures"
J Michael Oakes, University of Minnesota
"Promise and pitfalls in organizational field experiments"
Sally Piccioto, University of California, Berkeley
"Short-term disability leave, return to work, and recurrence: an analysis using g-methods"
David Rehkopf, Stanford University
"Using a natural experiment to examine the long-term impact of household unemployment on mortality"


Bayes is not a pain in the posterior

Session Chair(s):   Richard MacLehose
Date: 2018-06-22       Time: 10:00 am - 11:30 am
Location: Constellation B


     

Bayesian statistics offer the promise of including prior information into analyses while avoiding some of the drawbacks of frequentist statistics. We have found that when introduced to Bayesian methods, many researchers are curious, yet rarely apply these methods in their own work. This may be because most epidemiology and biostatistics curricula, as well as the epidemiologic literature at large, are primarily based on frequentist statistics. But it is also likely due to the fact that many presentations are highly technical, providing statistical detail at the expense of practical implementation. We will discuss why Bayesian methods are useful, why they are under-utilized and will provide several detailed examples of Bayesian methods that can aid in epidemiologic research. To aid in the adoption of Bayesian methods, the focus of this symposium will be on applied, worked examples and we will post all data, computer code and slides, prior to the meeting at:
http://ghassanbhamra-phd.org/presentations-and-such/


Session Chairs:
Richard MacLehose, University of Minnesota


Presenters:
Matthew Fox, Boston University
"The need for simple Bayesian methods and why they aren’t used enough"
Richard MacLehose, University of Minnesota
"The why, what, when and how of Bayesian methods in epidemiology"
Jessie Buckley, Johns Hopkins School of Public Health
"Bayesian methods for limit of detection problems"
Ghassan Hamra, Johns Hopkins School of Public Health
"Bayesian hierarchical models in environmental epidemiology"
Tom Ahren, University of Vermont
"Bayesian meta-analysis and visual examination of assumptions via an online app"
Lawrence MacCandless, Simon Fraser University
"Mediation analysis with Bayes using Stan"


Thinking beyond diagnosis, procedure, and drug codes: Shaping the future of epidemiologic research using electronic healthcare databases

Session Chair(s):   Jessica Chubak,   Sophia Newcomer
Date: 2018-06-22       Time: 10:00 am - 11:30 am
Location: Chesapeake B


     

Epidemiologists frequently use electronic healthcare data, such as insurance claims and electronic health records (EHR), to address pressing research questions. These data offer many opportunities but also present unique challenges around lack of standardization, misclassification, and unmeasured confounding. Presenters will highlight major advantages of electronic healthcare data as well as commonly encountered methodological problems and innovative solutions, drawing upon examples from vaccine safety and efficacy, cancer screening, perinatal outcomes, and injury research. The symposium will look ahead to cutting-edge advances in data collection (e.g., e-health, mobile apps, patient-reported outcomes) and methodology (e.g., data linkage, machine learning, natural language processing). We will close with a facilitated audience discussion on how epidemiologists can advance research in a “big data” world.

Session Chairs:
Jessica Chubak, Kaiser Permanente Washington Health Research Institute
Sophia Newcomer, Kaiser Permanente Washington Health Research Institute


Presenters:
Jennifer Lund, University of North Carolina, Chapel Hill
"An overview of the current electronic healthcare database landscape"
Micheal L Jackson, Kaiser Permanente Washington Health Research Institute
"Vaccine safety surveillance using big healthcare data"
Jessica Chubak, Kaiser Permanente Washington Health Research Institute
"Using claims and electronic health record data to study cancer screening: challenges and opportunities"
Kristin Palmsten, HealthPartners Institute
"Harnessing electronic healthcare data for perinatal health research"
Mackenzie Herzog, University of North Carolina, Chapel Hill
"Getting back in the game: Understanding sports injuries and musculoskeletal conditions using claims and EHR data"
Sophia Newcomer, Kaiser Permanente Washington Health Research Institute
"How will epidemiologists shape the future of electronic health data-based research?"