The #SER2020 conference will be based on Mountain Time.

Caution and Creativity: Maximizing the Value of Electronic Health Record and Claims Data in Epidemiologic Research
Dec 16 @ 10:15 am – 11:45 am

Session Co-Chair: Nrupen Bhavsar, Duke University
Session Co-Chair: Lauren Cain, AbbVie

The volume & velocity of electronic health record (EHR) and claims data is immense; however, there are concerns about the veracity of these data. It is increasingly common for EHR and claims data to be used in epidemiologic research for cohort selection, disease surveillance, prediction and inferential research. These data are also used for regulatory purposes; the FDA recently released guidance for use of these data for drug, biologic, and device approval. A better understanding of the methodological challenges of these data is needed. The speakers in this session will provide a brief overview of how these data inform their work and the cautions/pitfalls/solutions they have encountered when using these data. Topics to be discussed include data linkage, missing data, validity of EHR/claims as compared to cohort data, and applications for environmental, aging, and social epidemiology. A significant portion of time will be allocated for discussion with panel and attendees.

Ann Marie Navar, Duke University
“Beyond Babel: Combining data across health systems and EHR vendors”

Bryan James, Rush University
“Only Built 4 Data Links: Medicare claims linked to cohort data in aging research”

Katherine Moon, Johns Hopkins University
“EHR^2: Electronic health records for environmental health research”

Annemarie Hirsch, Department of Population Health Sciences, Geisinger
“A look under the hood: Social determinants in the EHR data”

Colin Anderson-Smits, Takeda Pharmaceuticals
“Expediting innovation: Real world data for regulatory submissions”

Likely biased but possibly useful: the implications of conditioning on future events for interpretation of effects in perinatal epidemiology
Dec 16 @ 10:15 am – 11:45 am

Session Co-Chair: Mollie Wood, Harvard University
Session Co-Chair: Dominique Heinke, Massachusetts Department of Health

Perinatal epidemiologists often condition on the occurrence of pregnancy or live birth when studying the effects of exposures occurring before or during gestation. Consequently, selection occurs at multiple points from exposure to outcome: detecting pregnancy, identifying pregnancy losses, and correctly classifying outcomes (often only observable in live births). This creates challenges for (1) validly estimating causal effects of exposures on outcomes, and (2) using these estimates to generate practical information that corresponds to choices that people face before and during pregnancy. The target trial approach is instructive, but highlights problems in interpreting findings. For example, if we are interested in effects of pre-pregnancy weight loss on birth outcomes (e.g. C-section, childhood overweight), we can fit a regression model for birth outcomes comparing women who lost weight prior to pregnancy with those who did not, conditioning on getting pregnant. This analysis translates to a trial intervention of “weight loss plus pregnancy” compared to “no weight loss plus pregnancy”, which would be impossible to implement. This is both because pregnancy cannot be guaranteed but also because pre-pregnancy weight loss may affect the ability to become pregnant, complicating interpretation. In this session, we explore the implications for conditioning on future events in perinatal epidemiologic research, and the implications of conditioning on pregnancy or offspring survival.

Matt Fox, Boston University

Mollie Wood, Harvard University
“Identifying Pregnancy Intention in Administrative and Registry Data”

Jack Wilkinson, University of Manchester
“Outcome Truncation in Assisted Reproduction Trials: Does it Matter in Practice?”

Dominique Heinke, Massachusetts Department of Health
“In Perinatal Epi, You Can’t Always get the Unbiased Estimate that you Want. But You Might Find the Information you Need.”

Elizabeth Suarez, Harvard Medical School
“Competing risks in perinatal pharmacoepidemiology: are live-birth cohorts a problem?”

Olga Basso, McGill University

Jonathan Snowden, Oregon Health and Science University, Portland

Living and Dying in Rural Areas
Dec 16 @ 10:15 am – 11:45 am

Session Co-Chair: Katherine Ahrens, University of Southern Maine
Session Co-Chair: Lauren Rossen, National Center for Health Statistics, Centers for Disease Control and Prevention

Rural residents in the US are at higher risk for the five leading causes of death (CDC, 2017) and experience shorter life expectancy (Singh, 2014) as compared with urban residents. While rural residents are often older, lower-income, and have a higher prevalence of smoking and chronic medical conditions, it is unknown how much of the disparities in mortality are due to characteristics of living in a rural area or to inadequate access to healthcare. A recent publication questioned the utility of even examining the causal effects of rural residence, given it is a non-specific and often non-modifiable factor (Caniglia, AJE, 2019). This symposium will bring together researchers with a range of views on how to study and address health outcomes and disparities in rural settings, and discuss the methodological challenges of estimating rural health.

We will first present an overview of how rural areas have been defined over time in the US, and what we know about disparities in rural health. We will then describe the use of various modeling approaches to account for geographic factors, and to examine spatial and temporal trends in health outcomes at the county level using hierarchical Bayesian models. The symposium will close with a presentation that questions the validity of using rural residence as an exposure variable in health research.

Katherine Ahrens, University of Southern Maine
“Introduction to the speakers”

Erika Ziller, University of Southern Maine
“What Does it Mean to be Rural?”

Amy Branum, CDC
“Rural Residence and Health”

Lauren Rossen, CDC
“It Matters Where You Live”

Katherine Ahrens, University of Southern Maine
“What About All That Suppressed Data?”

Ellen Caniglia, NYU Langone
“Rural Residence as an Exposure: What Exactly Are We Studying?”

Plausible Counterfactuals in the Study of Structural Racism and Population Health
Dec 16 @ 10:15 am – 11:45 am

Session Co-Chair: Anusha Vable, University of California, San Francisco
Session Co-Chair: Zinzi Bailey, University of Miami
Session Co-Chair: Julia Raifman, Boston University

There is growing interest in racism as a structural determinant of racial health disparities. While much scholarship has noted there are not plausible counterfactuals for race, there do exist counterfactuals to racism. In this symposium, we will lead a rigorous discussion on plausible counterfactuals in the study of racism and population health. As public health researchers and students increasingly seek to address racial disparities, we hope for this discussion to foster research ideas that will impact how we all understand and seek to address structural racism.

Zinzi Bailey, University of Miami
“Frame topic of structural racism”

Atheen Venkataramani, University of Pennsylvania
“Affirmative action bans and health risk behaviors to college affirmative action bans and smoking and alcohol use among underrepresented minority adolescents in the United States: A difference-in-differences study?”

Julia Raifman, Boston University
“Discussant of Venkataramani study, and state and local policy counterfactuals in the study of racism”

Alyssa Mooney, University of California Berkeley
“Criminal justice reforms: reducing penalties for drug possession, and clearing criminal records”

Anusha Vable, University of California San Francisco
“Discussant of Mooney study, and causal inference in the study of racism”

Whitney Robinson, UNC Chapel Hill
“A (not-so)-fun house of mirrors: Logical fallacies as barriers to conceptualizing better counterfactuals”

The impact of the obesity and diabetes epidemics on cancer incidence and survival
Dec 16 @ 10:15 am – 11:45 am

Session Co-Chair: Deirdre Tobias, Brigham and Women’s Hospital
Session Co-Chair: Paulette Chandler, Brigham and Women’s Hospital

A report from the International Agency for Research on Cancer (IARC) identified excess body weight and obesity as significant risk factor for at least 12 cancer sites, including cancers of the breast, colorectum, other gastrointestinal sites, and reproductive organs. Evidence for the role of type 2 diabetes as a mechanism linking obesity with several of these sites is also mounting. As the obesity and diabetes epidemics continue and younger generations age, what will be the key preventive strategies to offset the expected rise in obesity-related cancers? Identifying factors underlying obesity’s and diabetes’ influence on cancer incidence, progression, and survival, may allow for the development of targeted approaches in the prevention of several site-specific cancers. The transdisciplinary panel of speakers we proposed each address unique perspectives of this important public health question, from observational epidemiology to randomized clinical trials.

JoAnn Manson, Harvard University
“Epidemiologic evidence linking obesity, type 2 diabetes, and cancer risk”

Elizabeth Cespedes Feliciano, Kaiser Permanente
“Importance of body composition in cancer outcomes”

Kimberly Bertrand, Boston University
“Racial and Ethnic Disparities in Obesity, Diabetes, and Breast Cancer”

Jennifer Ligibel, Dana Farber Cancer Institute
“Lifestyle intervention trials for cancer prevention and survival: intervening on obesity and diabetes”

The rapid decline in adolescent mental health in the 21st century: magnitude, causes, and treatment innovations from a life course perspective
Dec 16 @ 10:15 am – 11:45 am

Session Co-Chair: Katherine Keyes, Columbia University
Session Co-Chair: Erin Dunn, Massachusetts General Hospital

Throughout much of the 20th century, psychiatric epidemiological studies indicated that depression and anxiety among adolescents had a stable prevalence, even as the prevalence for help-seeking changed. But in the past 10 years, numerous, independent data sources indicate that depression, anxiety, suicidal behavior and completed suicide among adolescents have increased more than in at least the past 50 years. Such an increase requires urgent attention, both to continued efforts to identify causes of the increases as well as underlying causes of adolescent psychiatric disorders, neurobiological mechanisms through which depression and anxiety occur, and innovations in treatment to reach more adolescents and alleviate suffering. In this symposium we address the growing morbidity and mortality of adolescent mental health by leveraging expertise across a spectrum of research foci, from an epidemiological overview into the emergence of the adolescent mental health decline, a life course perspective on early life causes and neurobiological pathways underlying the emergence of adolescent psychiatric disorders, as well as novel treatments for adolescent mental health that leverage digital and mobile technology aimed at ameliorating psychiatric symptoms. We aim to frame the research agenda for a new generation of scholars to engage in innovative and transdisciplinary research bringing science to practice.

Katherine Keyes, Columbia University
“Adolescent mental health is declining: convergence across five independent data sources, and controversies regarding the cause”

Erin Dunn, Massachusetts General Hospital
“Does adolescent health begin in early life? Links between early life adversity, DNA methylation, and adolescent-onset depression”

David Weissman, Harvard University
“Neurobiological Pathways Underlying the Emergence of Depression and Anxiety”

Kathleen Merikangas, National Institute of Mental Health
“Application of mobile technology to identify targets for treatment and prevention in adolescent mood and anxiety disorders”

The Selected, The Generalizable and The Transportable: Dealing with Threats to Internal and External Validity
Dec 16 @ 10:15 am – 11:45 am

Session Co-Chair: Onyebuchi Arah, UCLA
Session Co-Chair: Caroline Thompson, San Diego State University

Selection bias, generalizability and transportability are hot topics in epidemiology and rightly so because they deal with the internal and external validity of causal effect estimates. Unlike internal validity threats like unmeasured confounders, selection bias, and measurement error, external validity concerns have been traditionally poorly studied and understood. Graphical and potential outcomes approaches to validity have been proposed to varying degrees of completeness and accessibility. Estimation methods are increasingly receiving attention in the literature, but actual applications remain scarce. Clinical trials reach only a fraction of eligible participants, and priority populations are often excluded despite a growing need to understand the effectiveness of available interventions at the population-level. The “big data” explosion of the 21st century has resulted in an unprecedented availability of observational data. However, many of these so-called “real world” data sources are mere convenience samples, leading to an increasing need for attention to issues of generalizability of the results derived from them. Further, as time passes, expensive epidemiologic cohorts are aging out and we are faced with the need to generalize results that have been obtained in the past to future generations of humans. In this symposium, we will explore both theoretical and applied issues surrounding selection bias, generalizability and transportability, giving them a general framework.

Caroline Thompson, San Diego State University
“Improving the generalizability of studies using electronic health records and other “real world data” sources”

Zeyan Liew, Yale University
“Live-Birth (Selection) Bias, Internal Validity, and Generalizability in Life-Course Epidemiology: Lessons from Denmark”

Roch Nianogo, UCLA
“Using simulations to generalize and transport treatment effects to a new environment”

Onyebuchi Arah, UCLA
“Adapting g-methods for generalizability and transportability studies”

Elias Bareinboim, Columbia University
“Getting the Causal Language of Selection Bias, Generalizability and Transportability Right”

Wrong Question Bias
Dec 16 @ 10:15 am – 11:45 am

Session Chair: Daniel Westreich, UNC Chapel Hill

At SER 2018, an overflowing room of epidemiologists watched as a host of panelists debated which bias was worst. The audience as jury decided overwhelmingly that bias due to missing data was worst. But we argue that a fundamental bias was missing from the debate: “Wrong question bias.” If our scientific questions are not good, then no answer – no matter how free from systematic error – will help us meet our goals, such as improving the healthy human lifespan.

But what does it mean for a scientific question to be good? What are some of the attributes of good scientific questions? What are some signs of poor scientific questions? We epidemiologists must begin the conversations that will help to sharpen our questions, and thereby help to improve our answers. Here, speakers will address these fundamental questions-about-questions, head-on.

Daniel Westreich, UNC Chapel Hill
“What we talk about when we talk about research questions”

Jessie K. Edwards, UNC Chapel Hill
“Zen and the Art of Asking Better Questions”

Maria Glymour, University of California San Francisco
“Advancing Population Health, Advancing Theory, and Advancing your Career: Working Backwards from your Goals to Define a Research Question”

Stephen Cole, UNC Chapel Hill
“Good Questions are Understandable, Unambiguous, at least partially computable (or identifiable), and important”

Elizabeth Stuart, JHSPH

Causal inference and molecular epidemiology: How can we best use -omics to answer epidemiologic questions of interest?
Dec 17 @ 1:45 pm – 3:15 pm

Session Co-Chair: Jonathan Y. Huang, Singapore Institute for Clinical Sciences
Session Co-Chair: Brian Whitcomb, University of Massachusetts

Driven by computational and methodological advances, developments in observational causal inference related to “-omics,” or high-dimensional molecular biomarkers, are proceeding apace in computer science and bioinformatics: from promises of machine learning of Directed Acyclic Graphs and Bayesian modelling of gene regulatory networks to agnostic “causal discovery” via genomic instrumental variables. Such work might suggest –omics can be leveraged to address a variety of goals in epidemiologic causal inference such as identifying biochemical targets for intervention; decomposing effects of modifiable disease pathways; or clarifying how “embodied” biological processes mediate the effects of social factors. And yet, very basic questions related to causal effect identification in molecular epidemiology and its relevance to population health remain unanswered.

Here we ask a panel of presenters to briefly address a few challenging questions relevant to causal inference in molecular epidemiology.

These will be considered by a discussant and the panel itself:
1. How do we address standard causal inference assumptions in molecular epidemiology? For example, can we ensure consistency is reasonably fulfilled?
2. How do we best leverage biological knowledge to investigate epigenetic mediation?
3. What are the implications of genetic instruments for agnostic causal discovery?
4. How can –omics network discovery or other data reduction strategies help understand causal mechanisms?

Laura Balzer, University of Massachusetts-Amherst
“Mediation Madness: Analyses with Multiple Mediators Occuring at Multiple Levels”

Ema Perkovic, University of Washington
“Graphical Criteria for Efficient Total Effect Estimation in Causal Linear Models”

Jon Huang, Singapore Institute for Clinical Sciences
“Leveraging Molecular Biomarkers in Multiple Tissues to Strengthen Inference on Developmental Mechanisms”

Jeremy Labrecque, Erasmus MC
“Instruments can’t keep time: Evidence for Genes with Time-Varying Effects and How to Use Them in Mendelian Randomization”

Causal mediation analysis meets consequential epidemiology: Recent insights and advances on the cross-world assumption from across the world
Dec 17 @ 1:45 pm – 3:15 pm

Session Co-Chair: Ryan M. Andrews, Leibniz Institute for Prevention Research and Epidemiology
Session Co-Chair: Wen Wei Loh, Ghent University

Understanding the causal mechanisms between an exposure and outcome is a crucial step towards designing interventions to improve population health. Epidemiologists have used causal mediation methods to estimate natural direct and indirect effects to answer such mechanistic questions; however, these methods have been criticized for relying on a “cross-world” independence assumption that is experimentally unverifiable and sometimes unrealistic. Also, natural and path-specific effects involve a non-manipulable parameter, which raises the question of whether causal mediation is more philosophical than scientific in nature. In this symposium, we will provide counterpoints to these criticisms via alternative conceptualizations of causal mediation under which mediation effects can be identified and estimated under realistic assumptions that are empirically testable, including when there are multiple mediators.

We believe this symposium is an excellent fit for SER 2020. There is ongoing debate at SER, on social media, and in the scientific literature on whether causal questions must be answered in terms of well-defined manipulable parameters. This symposium adds to the debate by showing new perspectives under which causal mediation is scientifically grounded. It also brings together an impressive group of mediation analysis researchers across three continents, and we believe the diversity of examples and opinions within this group will interesting to the SER membership.

Ryan M. Andrews, Leibniz Institute for Prevention Research and Epidemiology
“Insights into the “cross-world” assumption of causal mediation analysis: Theoretical and practical considerations”

Isabel Fulcher, Harvard Data Science Initiative
“Identification and estimation of indirect effects robust to unmeasured confounding”

Wen Wei Loh, Ghent University
“Interventional effect models for multiple mediators”

Margarita Moreno-Betancur, University of Melbourne
“Defining mediation effects for multiple mediators using the concept of the target randomized trial”

Vanessa Didelez, Leibniz Institute for Prevention Research and Epidemiology – BIPS

Climate change and health: where are we?
Dec 17 @ 1:45 pm – 3:15 pm

Session Co-Chair: Sandie Ha, University of California, Merced
Session Co-Chair: Carrie Nobles, NICHD

Climate change is an unprecedented threat to human health and the greatest public health challenge of the 21st century. With environmental impacts already observable and more severe impacts anticipated even when meeting global emissions goals, there is an immediate need to not only reduce carbon emissions but to develop strategies to mitigate anticipated health impacts of climate change. Beyond extreme temperatures, climate change is anticipated to impact health through many routes including severe weather events, droughts, wildfires, changes in the distribution of infectious diseases and population displacement due to sea-level rise, with disproportionate impacts in communities with the fewest resources. Epidemiology will play a vital role in estimating and predicting the health impacts of climate change, with efforts requiring cross-disciplinary collaborations and close partnership with policymakers. Despite this, climate change has not received the emphasis within the field of epidemiology needed to match its unprecedented threat to health. We propose a symposium inviting five key leaders in climate change epidemiology to a) present the latest cutting-edge research in climate change epidemiology, b) discuss climate change equity issues facing our communities, and c) discuss challenges and opportunities for research as well as the policy making process related to climate change.

Kristie Ebi, University of Washington
“Climate change and health: how did we get here and where are we going?”

Kate Weinberger, University of British Columbia
“Opportunities for Epidemiologic Research to Inform Adaptation to Climate Change”

Jesse Berman, University of Minnesota
“Existing Challenges, Opportunities and Recent Advances in Drought and Health Research”

Carina Gronlund, University of Michigan
“Translating Climate and Health Knowledge into Testable Interventions in the Industrial Midwest”

Wael Al-Delaimy, University California, San Diego
“Climate change is a reason to revisit our approach as epidemiologists: the role of policy”

Epidemiologists and data scientists: Can you tell the difference?
Dec 17 @ 1:45 pm – 3:15 pm

Session Co-Chair: Barbra Dickerman, Harvard University
Session Co-Chair: Miguel Hernan, Harvard University

Recent applications of machine learning to healthcare data are fueling an intense debate: will the value match the hype? Remarkable advances in areas such as visual pattern recognition have already enabled machine learning algorithms to outperform physicians on prediction tasks involving subtle patterns. This wave of success has heightened interest and expectations around the use of machine learning to uncover better treatment strategies that get the right interventions to the right people at the right time.

The challenge of drawing causal inference from observational data, which epidemiologists have long grappled with, is now in the spotlight of data scientists’ work on these complex tasks of sequential decision-making. This presents a significant opportunity to bring epidemiologists and other data scientists together, to borrow from each other’s theoretical and practical advances and learn from each other’s mistakes in the shared pursuit of optimal treatment strategies.

The aim of this symposium is to bring together investigators from different fields (epidemiology, biomedical informatics, computer science) and different institutions (Harvard, Stanford, MIT) who are approaching similar causal questions with different terminologies and analytic techniques. This symposium will bridge these perspectives and advance our understanding of how machine learning can be used to answer causal questions and how causal inference can support the development of machine learning methods to improve healthcare decision-making.

Barbra Dickerman, Harvard University
“Counterfactual Prediction: How to Use Data to Support Healthcare Decision-Making?”

Nigam Shah, Stanford University
“Learning from Patients like Mine at the Point of Care”

David Sontag, Massachusetts Institute of Technology
“Machine Learning to Improve a Precision Approach to Healthcare”

Moving forward: Theories, philosophies and histories for education in Epidemiology
Dec 17 @ 1:45 pm – 3:15 pm

Session Co-Chair: Michael Windle, JHSPH
Session Co-Chair: Daniel Antiporta, JHSPH

Paradigms in epidemiology have changed during the past 50 years, moving through what some have termed “traditional” and “modern” epidemiology to the epidemiology of the contemporary era. Efforts have been made to reconcile and consolidate specialized branches of the field, along with its diverse philosophical approaches. Some call for a unified epidemiology and others argue the historical record makes clear that a shared interest in investigating population health does not presume a single theoretical or philosophical orientation. Changing times represent a challenge for graduate training that should comprehend the full scope of an inherently synthetic field: competence in methodological innovations and preservation of its diverse philosophical underpinnings. As doctoral students, we noticed that our training is appropriately heavy on methods and statistical training. Less well represented are the theory, philosophy and history of the field, which tend to be incorporated tangentially when not omitted entirely. We believe that graduate students should be equipped to query and understand the raison d’être of epidemiology and the history of its methodological tools, which will enable us to push the boundaries of the discipline. This symposium was proposed accordingly, and we believe epidemiologists globally, both students and teachers, share this interest. The symposium aims to highlight contemporary aspects in epidemiology that deserve consideration in the context of graduate education. We hope to start a discussion on the need of emphasizing the diverse theory, philosophy, and history of the discipline during graduate training. We expect this discussion highlights current and potential concrete efforts towards that goal.

Alison Abraham, Johns Hopkins University
“Core competencies for epidemiologists: a multi-national effort”

Nancy Krieger, Harvard University
“Epidemiologic Theories and the People’s Health”

Maria Glymour, University of California, San Francisco
“Methods that Matter for Epidemiology in the Current Era”

Alfredo Morabia, Columbia University
“Why history matters for epidemiologists?”

Daniel Antiporta, JHSPH
“Student Voices: Programs, Expectations and Transitions in Epidemiology”

Michael Kramer, Columbia University

Polygenic risk scores: challenges and opportunities for population health equity
Dec 17 @ 1:45 pm – 3:15 pm

Session Co-Chair: Lindsay Fernandez-Rhodes, Penn State University
Session Co-Chair: Ann Von Holle, NIEHS

Polygenic and genetic risk scores have become popular in the medical literature. Nonetheless best practices for their use in population health studies at times are unclear and may prove problematic as future genetic studies include more racially and ethnically diverse populations.
Our symposium focuses on the construction and interpretation of polygenic and genetic risk scores as well as their implications for health disparities research. As such, this symposium lies at the intersection of genetic epidemiology, methodologic issues, and the social and ethical implications of genomics by using examples from a diverse panel of speakers on a number of complex diseases and population health relevant contexts.

The symposium is divided into three parts:
1) Dr. Chatterjee will provide an introduction to polygenic and genetic risk scores in the context of public health;
2) Dr. Wojcik will explain how polygenic risk estimation currently underperforms in ancestrally diverse populations, and will provide steps to address this inequity; and
3) Dr. McCarthy will translate recent innovations in polygenic risk scores to clinical setting and will discuss potential barriers to equal access.

In each of these presentations, the speakers will highlight the challenges and opportunities of polygenic and genetic risk scores, and will specifically address how their use may impact equity in population health. As such this symposium will provide a continuum of different perspectives that a broad audience of epidemiologists may consider when incorporating risk predictions in their own research.

Nilanjee Chatterjee, Johns Hopkins University
“Polygenic Risk Scores for Cancer Prevention”

Genevieve Wojcik, Johns Hopkins University
“Challenges in the Transferability of Polygenic Risk Scores to Diverse Populations”

Anne Marie McCarthy, University of Pennsylvania
“Implementing Polygenic Risk Scores Clinically – Challenges and Opportunities”

Translational social epidemiology: Identifying and evaluating policy interventions that address the health implications of social stratification
Dec 17 @ 1:45 pm – 3:15 pm

Session Co-Chair: Ellicott Matthay, University of California, San Francisco
Session Co-Chair: Holly Elser, Stanford Medical School

The importance of labor, education, and the social safety net for population health inequalities is well-established. This motivates explicit consideration of cause-effect relationships between social systems and health outcomes. It also motivates identification of social policies that have the potential to reduce disparities in population health that arise from the uneven allocation of social services. For this symposium, we emphasize the work of junior researchers (students, postdoctoral scholars, and junior faculty) whose work examines the implications of education, labor, healthcare, criminal justice, and the social safety net for health. Rather than demonstrating the existence of health inequalities, these speakers will present research that evaluates or identifies specific social policies or interventions. Speakers will review the latest evidence, discuss key methodological challenges, and present studies aimed at translating epidemiologic evidence into effective social policies.

Holly Elser, Stanford Medical School
“Paid parental leave and maternal health: An impact evaluation of the San Francisco Paid Parental Leave Ordinance”

Danielle Gartner, Michigan State University
“Assessing the ACA’s impact on racial disparities in health: What is trust responsibility and why does it matter for understanding the health of Native nations?”

Alyssa Mooney, UC Berkeley
“Linking administrative data from criminal justice and health agencies to evaluate policy impacts: Challenges and opportunities”

Laura Rosella, University of Toronto
“Methodological considerations when using secondary data to evaluate the impact of health and social safety net policies in a universal health system”

Catherine Duarte, UC Berkeley
“School Discipline, Education Trajectories, and Health: A Lifecourse Analysis in the National Longitudinal Survey of Youth 1979 Cohort”

Ellicott Matthay, University of California, San Francisco

Documentation status, human rights, and the role of epidemiology in the immigration debate
Dec 18 @ 10:15 am – 11:45 am

Session Co-Chair: Christine Gray, Duke University
Session Co-Chair: Marissa Seamans, University of California, Los Angeles

Near-daily headlines about immigration suggest that it is one of the greatest challenges facing countries across the globe. In the United States, undocumented persons have become a particular target of immigration policies. As leaders and communities grapple with issues ranging from protection to equity to humanitarian crises, the role of epidemiology in addressing this challenge must be clarified. Our goal is to bring together speakers with diverse expertise to engage this important topic.

Speakers will discuss emerging research and policy implications as it pertains to immigration, focusing on the undocumented population in the U.S. Discussions will highlight health consequences related to documentation status, including the role of a human rights framework in addressing health consequences of documentation status. Speakers will pose critical questions and cast a vision for how epidemiology can inform immigration policies that are among the most consequential social issues of the 21st century.

Lynne Messer, OHSU-PSU
“The Hispanic Paradox revisited – quantifying the effect on pregnancy outcomes of being an undocumented Mexican woman in the United States”

Sarah Andrea, University of Washington
“Structural racism isn’t so black and white: are inequities in pregnancy outcomes for undocumented Mexican women driven by stratification into disadvantaged neighborhoods?”

Jasmine Aqua, Freedom University
“Education and Health are Human Rights: Freedom University’s Fight for Undocumented Youth”

Arijit Nandi, McGill University

Epidemiologic methods for applied research: Answering important public health questions that are not causal
Dec 18 @ 10:15 am – 11:45 am

Session Co-Chair: Catherine Lesko, Johns Hopkins University
Session Co-Chair: Matthew Fox, Boston University

Epidemiology has been called “the basic science of public health” and has been defined as the study of the “distribution and determinants of diseases and health states in human populations.” Academic epidemiology has tended to emphasize methods for studying the “determinants of disease”: answering clinical questions and identifying effects of interventions. The implication is that answering questions about the “distribution of disease” is intuitive and straightforward, yet the methods challenges inherent in this research can be quite complex. In this session, we will highlight sophisticated epidemiologic analyses that answer important applied public health questions. Nested in each speaker’s talk will be some discussion of the similarities and differences between “academic” and “applied” research and in particular the relative emphasis placed on different sources of bias as a threat to inference (for example, adjustment for confounder control versus dealing with missing data or measurement error as a source of information bias) and relative importance of adhering to a strict “causal” framework in designing studies.

Each speaker will present for 10-12 minutes, leaving 18-30 minutes for questions and audience discussion.

Catherine Lesko, Johns Hopkins University
“Single-sample biases: Describing the natural course in the presence of measurement error, missing data, loss to follow-up, and competing risks”

Matthew Fox, Boston University
“Smart supplemental data collection: Designing validation/tracing studies to account for loss to follow-up and competing risks when describing the HIV care continuum in South Africa”

Magdalena Cerda, NYU Langone Health
“Working in quicksand: evaluating the impact of prescription opioid policies on opioid overdoses when exposure and outcome definitions are evolving and context matters”

Elizabeth Cromwell, University of Washington
“Dealing with sparse data: Using detailed spatial estimates for decision making in the context of Neglected Tropical Disease programs”

Lauren Tanz, North Carolina Department of Public Health
“To know where to go you need to know where you are: Piecing together surveillance data for situational awareness in the opioid overdose epidemic”

Fausto Bustos Carillo, University of California at Berkeley
“Casual causal inference: the role of risk factor analyses in infectious disease outbreak response”

Got multiple exposure troubles? How causal inference and machine learning can help
Dec 18 @ 10:15 am – 11:45 am

Session Co-Chair: Jessie Buckley, Johns Hopkins Bloomberg School of Public Health
Session Co-Chair: Alexandra White, National Institute of Environmental Health Sciences

Epidemiologists across substantive subject areas are increasingly adopting methods to characterize the health effects of multiple correlated exposures, or mixtures, due to interest in potential joint effects and concerns about co-exposure confounding. Methods development in this area has led to extensions of older methods as well as new approaches designed to answer questions about how exposure mixtures affect health. This symposium will highlight recent advancements in the area of multiple exposure research with a focus on how causal inference and machine learning methods can be utilized to better address these research questions.

We will begin with a discussion of the target parameters for research on multiple exposures, followed by a comparison of results across a variety of established methods, and ending with presentations of several novel applications of machine learning and causal inference approaches to multiple exposure studies.

Throughout the symposium, we will use Poll Everywhere to facilitate interactive audience participation. Specifically, we will address the following key questions via live polling and a panel Q&A discussion:
1) What multiple exposure questions can causal inference and machine learning approaches address?
2) Are causal inference and machine learning techniques an improvement over previous methods?
3) Do benefits of these approaches outweigh difficulties of implementation and interpretation?

Jennifer Ahern, University of California, Berkeley
“Target parameters for effects of mixtures”

Marianthi-Anna Kioumourtzoglou, Columbia University
“A comparison of different mixtures methods to study the link between persistent organic pollutants and leukocyte telomere length”

Youssef Oulhote, University of Massachusetts, Amherst
“Beyond variable importance measures: Interpreting ensemble learning methods for multiple exposures”

Alexander Keil, UNC Chapel Hill & NIEHS
“Using Bayesian causal thinking to generate actionable results from exposure mixture data”

Jeanette Stingone, Columbia University
“Can data-driven causal discovery methods inform our studies of multiple exposures?”

Policy-Relevant Epidemiology and State Violence: Policing, Mass Incarceration, and Immigration Detention
Dec 18 @ 10:15 am – 11:45 am

Session Co-Chair: Abby Cartus, University of Pittsburgh
Session Co-Chair: Seth Prins, Columbia University

Over the past decade, epidemiologists and public health practitioners have begun to document the public health consequences of mass incarceration and policing in the United States, clearly linking state violence to population health. In addition to this research, schools of public health have integrated coursework, symposia, and conferences on state violence into their regular offerings. The results are clear: state violence creates and perpetuates persistent health disparities along lines of class, race, national origin, gender, and sexuality. But what is epidemiology’s role in confronting the health effects of state violence? What if the current policy space only permits collaboration with agencies engaging in state violence? Should we work with them “from the inside,” support social movements for decarceration and abolition fighting for non-reformist reforms “from the outside,” or chart a middle path? The proposed symposium will grapple with these questions by featuring trans-disciplinary empirical research at the intersection of social science and public health. The goal of the symposium is to critically engage with calls for consequential epidemiology by exploring different perspectives on maximizing population health in the context of state violence.

Zinzy Bailey, University of Miami
“Structural Racism of State-Sanctioned Violence”

Justin Feldman, NYU Langone
“When Epidemiologic Surveillance Becomes Police Surveillance: The Case of Weapons-Related Injuries”

Seth Prins, Columbia University
“The Criminalization of Youth: Public Health Ramifications of the School-To-Prison Pipeline”

Carl Williams, Water Protector Legal Collective

The development of the ‘test-negative’ study design for monitoring vaccine effectiveness
Dec 18 @ 10:15 am – 11:45 am

Session Co-Chair: Kylie Ainslie, Imperial College London
Session Co-Chair: Ben Cowling, Hong Kong University

The test-negative design (TND) is a variation of the case-control study design, in which the same clinical case definition is used for enrollment of both cases and controls. Laboratory testing is subsequently used to distinguish which patients were cases and which were controls. An important advantage of this approach is the efficiency of enrolling cases and controls in the same location with the same case definition, thereby assuring that they have arisen from the same source population and reducing potential selection biases due to differential healthcare-seeking behavior.

Since first being used to assess influenza vaccine effectiveness (VE) in 2005, TND studies have become the most popular design for assessing annual influenza VE. In addition to influenza, the TND has been used for many years to measure pneumococcal VE, and has more recently been applied to evaluate vaccines for rotavirus, cholera, and meningococcus. Because of the benefits of the TND design, it has been adopted into public health practice quickly and before careful consideration of the underlying assumptions and potential sources of bias. Recently, a number of studies have been reported to validate and characterize the TND. This symposium will focus on the development of the TND, and the speakers will share current research on this study design including the potential for biases to affect VE estimation, applications to different diseases, and important considerations in design and analysis of data.

Kylie Ainslie, Imperial College London
“Overview of the test-negative design and how to control for bias and confounding”

Ben Cowling, Hong Kong University
“How the test-negative design is used for different vaccines, and the potential for a test-negative design ‘platform'”

Ben Lopman, Emory University
“Test-negative design for assessing rotavirus vaccines”

Marc Lipsitch, Harvard University
“Waning vaccine effectiveness in the test-negative design: real or apparent?”

Unpacking Decreases in US Life Expectancy
Dec 18 @ 10:15 am – 11:45 am

Session Co-Chair: Meredith Shiels, National Cancer Institute
Session Co-Chair: Renee Gindi, CDC

Despite decades of success in the prevention and treatment of chronic diseases in the U.S., sustained decreases in life expectancy have been observed for the first time since 1993. The increases in life expectancy seen over the past 30 years had primarily been driven by reductions in death rates among older men and women, but detailed analyses reveal sustained increases in death rates in 25-49-year-old non-Hispanic white men and women over the past two decades. These increases can be attributed in part to increasing death rates due to certain causes of death, notably drug overdose deaths, suicide and alcohol-related liver disease. A leveling off in the rate of cardiovascular disease deaths has also contributed to the life expectancy decrease in some groups. The proposed symposium will focus on disaggregating the overall mortality increase in the US to describe the demographic groups most impacted and the causes of death driving these trends. In addition to describing the trends, speakers will be invited to give insight into the underlying causes of these increases, and opportunities for prevention.

Meredith Shiels, National Cancer Institute
“Diverging Trends in Mid-Life Mortality Rates in the US”

Brandon Marshall, Brown University
“A Tale of Two Crises: Untangling the Social and Corporate Determinants of Overdose in the United States”

Katherine Keyes, Columbia University
“Increases in Suicide in the US: How Changes in Media and News Reporting Contribute to Increases in Suicide Contagion, and How to Disrupt the Cycle for Prevention”

Tiffany Powell-Wiley, National Heart, Lung, and Blood Institute
“Trends in Premature Cardiovascular Mortality in the United States: The Role of Interventions to Address Social Determinants of Health”

Renee Gindi, National Center for Health Statistics
“Health, United States as an exploratory tool to examine nationally representative trends”

Diversity, inclusion, and artificial intelligence: Challenges for population health
Dec 18 @ 10:15 pm – 11:45 pm

Session Chair: Geetanjali Datta, Cedar Sinai Medical Center Department of Medicine and the Research Center for Health Equity

“Artificial Intelligence (AI) is defined by the Merriam-Webster dictionary as “the capability of a machine to imitate intelligent human behavior.” Among its many perceived virtues, AI is often seen as a boon to streamlined prediction in our fast-paced world. As Artificial Intelligence is increasingly being used by health systems and social institutions to assist in decision making, it is becoming evident that AI algorithms can be flawed and mirror biased human behavior. As such, there have been calls to improve oversight to ensure the use of AI does not propagate inequalities. This symposium, submitted on behalf of the SER Diversity and Inclusion Committee, aims to bring together experts in artificial intelligence and health researchers utilizing AI to discuss how historic and existing social biases are reflected in AI algorithms, and what diversity and inclusion considerations we should make as we integrate AI-based analyses in our work. The participants will address the following questions: 1) How might artificial intelligence sustain and augment issues involving diversity and inclusion? 2) How can we improve the quality of the data we collect to limit these biases? and 3) How can we better train existing data to avoid perpetuating inequalities through our analyses?”

Irene Chen, Massachusetts Institute of Technology
“Diagnosing Sources of Unfairness in Supervised Algorithms”

Rediet Abebe, Harvard Society of Fellows
“Data Inequalities and Access to Information”

Adam Kalai, Microsoft Research New England
“What are the Biases in My Word Embedding”

Ashley Naimi, University of Pittsburgh
“Training Fair Algorithms: Considerations for Prediction and Causal Inference”

David Rehkopf, Stanford University