Confounding

July 5 2022

The New Epidemiology of Fusion Designs

January 13 2022

Use of directed acyclic graphs (DAGs) to identify confounders in applied health research: review and recommendations

March 8 2021

Conceptual Papers on Race, Socioeconomic Status or Position, and Sexual & Gender Minority Health

Ellen Caniglia, NYU
Lorraine Dean, JHU
William Goedel, Brown University

In social epidemiology, we are concerned with understanding how the distribution of privilege and power translates to patterns of health and disease in a society. The following playlist focuses on the inclusion of race/ethnicity, sex, gender identity, sexual orientation, and socioeconomic position with a focus on (a) conceptual frameworks for distinguishing these social constructs from related systems of oppression (e.g., racism, sexism, heterosexism, classism) and their relations to health outcomes, (b) best practices for measuring these social constructs in epidemiological studies, and (c) best practices for use of these social constructs when interested in estimating casual effects. Each entry includes a summary and a note about which audiences might benefit most from reading a particular article. Within each broad area and sub-topic, we order the articles in the order we believe they should be read. This playlist is not intended to be an exhaustive resource on all literature relevant to these topics but rather as a starting point for deeper understanding.

RACE/RACISM

Invited commentary: “race,” racism, and the practice of epidemiology
Jones CP. Invited commentary: “race,” racism, and the practice of epidemiology. American journal of epidemiology. 2001 Aug 15;154(4):299-304.
Summary: Describes why race might be considered a “contextual” factor rather than an individual-level attribute and explains the different levels at which racism operates
Potential Audiences: All levels
https://www.researchgate.net/publication/11848610_Invited_Commentary_Race_Racism_and_the_Practice_of_Epidemiology

Structural racism and health inequities in the USA: evidence and interventions
Bailey ZD, Krieger N, Agénor M, Graves J, Linos N, Bassett MT. Structural racism and health inequities in the USA: evidence and interventions. The Lancet. 2017 Apr 8;389(10077):1453-63.
Summary: Reviews and explains what structural racism is and what implications it has for health in the United States. Differentiates institutional racism and structural racism from other types of racism.
Potential Audiences: All levels
https://pubmed.ncbi.nlm.nih.gov/28402827/

Racism and health: evidence and needed research
Williams DR, Lawrence JA, Davis BA. Racism and health: evidence and needed research. Annual review of public health. 2019 Apr 1;40:105-25.
Summary: Differentiates race, racism, and discrimination. Reviews evidence of how each have been linked to health outcomes and identifies gaps in research and interventions.
Potential audiences: All levels
https://www.aamc.org/news-insights/racism-and-health-reading-list

SOCIOECONOMIC STATUS (SES) & SOCIOECONOMIC POSITION (SEP)

Socioeconomic status in health research: one size does not fit all
Braveman PA, Cubbin C, Egerter S, Chideya S, Marchi KS, Metzler M, Posner S. Socioeconomic status in health research: one size does not fit all. Jama. 2005 Dec 14;294(22):2879-88.
Summary: Outlines the key problems with commonly used assumptions made when using SES measures, with specific guidance on measurement, interpretation of measures, and modeling.
Potential Audiences: All levels
https://pubmed.ncbi.nlm.nih.gov/16352796/

Using credit scores to understand predictors and consequences of disease

SEX/GENDER/SEXUAL IDENTITY

Genders, sexes, and health: what are the connections—and why does it matter?
Krieger N. Genders, sexes, and health: what are the connections—and why does it matter?. International journal of epidemiology. 2003 Aug 1;32(4):652-7.
Summary: Offers guidance on definitions and distinctions between sex, a biological factor, and gender, a social factor, and traces the historical roots of each term. Includes examples from public health about the interpretation of each as an exposure or outcome.
Potential audiences: All levels
https://pubmed.ncbi.nlm.nih.gov/12913047/

The utility of resilience as a conceptual framework for understanding and measuring LGBTQ health
Colpitts E, Gahagan J. The utility of resilience as a conceptual framework for understanding and measuring LGBTQ health. International journal for equity in health. 2016 Dec;15(1):1-8.
Summary: Discusses how sexual minority health is often viewed from a heternormative perspective and deficit framework rather than a resilience framework. Includes a table of key terms for sexual minority populations.
Potential audiences: All levels
https://equityhealthj.biomedcentral.com/articles/10.1186/s12939-016-0349-1

Expanding Gender-Based Health Equity Framework for Transgender Populations
Restar AJ, Sherwood J, Edeza A, Collins C, Operario D. Expanding Gender-Based Health Equity Framework for Transgender Populations. Transgender Health. 2020 Jun 30.
Summary: Expands on the World Health Organization (WHO) Gender Responsive Assessment Scale categorizing policies and programs as gender unequal, gender blind, gender sensitive, gender specific, or gender transformative to include transgender people. Offers suggestions and language for how transgender people can be a focus of gender inclusion.
Potential Audiences: All levels
https://www.liebertpub.com/doi/10.1089/trgh.2020.0026

Best Practices for Measuring Race, Sexual Orientation, and SEP

MEASURING SYSTEMS OF OPPRESSION

Measures of racism, sexism, heterosexism, and gender binarism for health equity research: From structural injustice to embodied harm—An ecosocial analysis
Krieger N. Measures of racism, sexism, heterosexism, and gender binarism for health equity research: From structural injustice to embodied harm—An ecosocial analysis. Annual Review of Public Health. 2020 Apr 1;41:37-62.
Summary: The author describes the key specific and shared features of unjust “isms”: racism, sexism, heterosexism, and gender binarism. The author provides an overview of how these systems of oppression may manifest at different levels and provides recommendations for measurement at the individual and structural levels.
Potential Audiences: All levels
https://www.annualreviews.org/doi/full/10.1146/annurev-publhealth-040119-094017

MEASURING RACE AND ETHNICITY

Classification of race and ethnicity: Implications for public health
Mays VM, Ponce NA, Washington DL, Cochran SD. Classification of race and ethnicity: Implications for public health. Annual Review of Public Health. 2003 May;24(1):83-110.
Summary: The authors describe the historical development of measures of race and ethnicity in the United States, with a focus on the Office of Budget and Management (OMB) standards promulgated in 1977 and later revised in 1997. The authors provide a list of recommendations for measuring race and ethnicity in public health.
Potential Audiences: All levels
https://pubmed.ncbi.nlm.nih.gov/12668755/

Socially-assigned race and health: A scoping review with global implications for population health equity
White K, Lawrence JA, Tchangalova N, Huang SJ, Cummings JL. Socially-assigned race and health: A scoping review with global implications for population health equity. International Journal for Equity in Health. 2020 Dec 1;19(1):25.
Summary: Race is not a biological construct and, therefore, social processes inform an individual’s self-identified race as well as how race is assigned to them by others. The authors describe potential limitations associated with the use of self-identified race as a sole measure of race and introduce socially-assigned race as an additional construct.
Potential Audiences: All levels
https://www.researchgate.net/publication/339153170_Socially-assigned_race_and_health_a_scoping_review_with_global_implications_for_population_health_equity

MEASURING SEXUAL ORIENTATION AND GENDER IDENTITY

Measuring sexual orientation: A review and critique of US data collection efforts and implications for health policy
Wolff M, Wells B, Ventura-DiPersia C, Renson A, Grov C. Measuring sexual orientation: A review and critique of US data collection efforts and implications for health policy. Journal of Sex Research. 2017 Jun 13;54(4-5):507-31.
Summary: The authors describe progress made in efforts to improve the measurement of sexual orientation in population health studies supported by the United States Department of Health and Human Services. Using the National Survey of Family Growth (NSFG) and the National Survey on Drug Use and Health (NSDUH) as illustrative case studies, the authors describe a matrix approach for measuring the different dimensions of sexual identity, sexual attraction, and sexual behavior.
Potential Audiences: All levels
https://pubmed.ncbi.nlm.nih.gov/28010119/

Perspectives from transgender and gender diverse people on how to ask about gender
Puckett JA, Brown NC, Dunn T, Mustanski B, Newcomb ME. Perspectives from transgender and gender diverse people on how to ask about gender. LGBT Health. 2020 Jun 26.
Summary: The authors introduce the current best practice for assessing sex and gender identity as two distinct constructs in population health studies and use a mixed methods approach to explore the perspectives of transgender and gender diverse people on how they would liked to be asked about sex and gender in population health studies.
Potential Audiences: All levels
https://pubmed.ncbi.nlm.nih.gov/32598211/

MEASURING SOCIOECONOMIC STATUS AND SOCIOECONOMIC POSITION

Measuring social class in US public health research: concepts, methodologies, and guidelines
Krieger N, Williams DR, Moss NE. Measuring social class in US public health research: concepts, methodologies, and guidelines. Annual Review of Public Health. 1997 May;18(1):341-78.
Summary: The authors first distinguish between common definitions of social class, socioeconomic position, socioeconomic status, and socioeconomic context. After doing so, the authors provide options for measuring each of these constructs at various levels (individual, household, and neighborhood) and over the life course and overview the definition and measurement of other constructs relevant to socioeconomic position (income, poverty, deprivation, wealth, and education). The appendix provides illustrative examples of the recommended measures in the extant literature.
Potential Audiences: All levels
https://pubmed.ncbi.nlm.nih.gov/9143723/

Measuring socio-economic position for epidemiological studies in low-and middle-income countries: a methods of measurement in epidemiology paper
Howe LD, Galobardes B, Matijasevich A, Gordon D, Johnston D, Onwujekwe O, Patel R, Webb EA, Lawlor DA, Hargreaves JR. Measuring socio-economic position for epidemiological studies in low-and middle-income countries: a methods of measurement in epidemiology paper. International Journal of Epidemiology. 2012 Jun 1;41(3):871-86.
Summary: The authors describe the challenges in applying measures of socioeconomic position used in high-income countries to population health studies in other settings in low- and middle-income countries and provide alternative, equivalent measures, such as asset-based measures, consumption expenditure, and participatory wealth ranking. Each measure includes a theoretical basis, suggested interpretation, and summaries of the strengths and limitations associated with a particular type of measure.
Potential Audiences: All levels
https://pubmed.ncbi.nlm.nih.gov/22438428/

Modeling Race/Ethnicity, Racism, Sex, Gender, SEP, etc., and Estimating Causal Effects

MODELING RACE

Beyond dummy variables and sample selection: what health services researchers ought to know about race as a variable
LaVeist TA. Beyond dummy variables and sample selection: what health services researchers ought to know about race as a variable. Health services research. 1994 Apr;29:1-16.
Summary: This paper reviews the socio-political roots of race data collection in the US census and reinforces when controlling for race is or is not appropriate. The author explains the fallacy in interpretations of adjusting for race, and why it is more important for health equity to use race as an effect measure modifier.
Potential Audiences: All levels
https://pubmed.ncbi.nlm.nih.gov/8163376/

Hidden in Plain Sight—Reconsidering the Use of Race Correction in Clinical Algorithms
Vyas DA, Eisenstein LG, and Jones DS. Hidden in Plain Sight—Reconsidering the Use of Race Correction in Clinical Algorithms. N Engl J Med. 2020 Aug 27;383:874-882.
Summary: Diagnostic algorithms and practice guidelines often adjust their outputs based on race and ethnicity. In doing so, these algorithms propagate race-based medicine to assess risk assessment and guide clinical decisions, and falsely lead people to interpretation of racial disparities as immutable facts rather than as injustices that require intervention.
Potential Audiences: All levels
https://www.nejm.org/doi/full/10.1056/NEJMms2004740?source=nejmtwitter&medium=organic-social

How do we assess a racial disparity in health? Distribution, interaction, and interpretation in epidemiological studies
Ward JB, Gartner DR, Keyes KM, Fliss MD, McClure ES, Robinson WR. How do we assess a racial disparity in health? Distribution, interaction, and interpretation in epidemiological studies. Annals of epidemiology. 2019 Jan;29:1-7.
Summary: In this paper, authors argue that looking only at an interaction term for race overlooks potential information that might still point to disparities, and suggest thinking beyond race as an effect measure modifier. Using case studies and examples, it offers a broader framework for examining disparities that goes beyond the significance of a race interaction term. Specifically, authors propose considering the distribution of the outcome and exposure across racial groups in addition to assessing interaction terms and stratum-specific effects.
Potential Audiences: Career epidemiologists
https://pubmed.ncbi.nlm.nih.gov/30342887/

MODELING SEP

Relative roles of race versus socioeconomic position in studies of health inequalities: a matter of interpretation
Nuru-Jeter AM, Michaels EK, Thomas MD, Reeves AN, Thorpe RJ, LaVeist, TA. Relative roles of race versus socioeconomic position in studies of health inequalities: a matter of interpretation. Annual review of public health. 2018 April;39:169-188.
Summary: In this paper, conceptualizations of race and SEP in the United States are reviewed. The paper details challenges in conceptualization and modeling, as well as the problems with interpretation when race and SEP are perfunctorily adjusted for in analysis. The paper offers guidance on how to model race and SEP in a meaningful way.
Potential audiences: Career epidemiologists
https://www.annualreviews.org/doi/full/10.1146/annurev-publhealth-040617-014230#_i23

MODELING INTERSECTIONALITY

Future Directions for Incorporating Intersectionality Into Quantitative Population Health Research
Agénor M. Future Directions for Incorporating Intersectionality Into Quantitative Population Health Research. American Journal of Public Health. 2020 Jun;110:803-806.
Summary: This paper introduces the six core ideas of intersectionality and discusses how quantitative health research can incorporate intersectionality to assess the relationships between multiple forms of inequity and health outcomes. The author also describes how population health researchers can engage with communities to improve how they incorporate intersectionality into their research. The paper offers a framework for “putting it all together” and examining how sexism, racism, heterosexism, etc. together contribute to health inequity.
Potential Audiences: All levels
https://ajph.aphapublications.org/doi/abs/10.2105/AJPH.2020.305610

ESTIMATING CAUSAL EFFECTS

On the causal interpretation of race in regressions adjusting for confounding and mediating variables
VanderWeele TJ and Robinson WR. On the causal interpretation of race in regressions adjusting for confounding and mediating variables. Epidemiology. 2014 Jul;25(4):473-84.
Summary: This paper discusses potential interpretations of the “effect of race” when regressions are run with race as an exposure variable, and how interpretation changes given other variables included in the model. The authors propose scenarios under which regression coefficients for race can be interpreted as the extent to which a racial inequality would remain if socioeconomic distributions in early and/or late life across racial groups could be equalized. The paper offers guidance on when/how to interpret the “effect of race” in a regression model and when/how we might estimate direct and mediated racial inequality measures.
Potential Audiences: Career epidemiologists
https://pubmed.ncbi.nlm.nih.gov/24887159/

Evaluating the population impact on racial/ethnic disparities in HIV in adulthood of intervening on specific targets: a conceptual and methodological framework
Howe CJ, Dulin-Keita A, Cole SR, Hogan JW, Lau B, Moore RD, Mathews C, Crane HM, Drozd DR, Geng E, Boswell SL, Napravnik S, Eron JJ, Mugavero MJ, for the CFAR NEtwork of Integrated Clinical Systems. Evaluating the population impact on racial/ethnic disparities in HIV in adulthood of intervening on specific targets: a conceptual and methodological framework. American journal of epidemiology. 2018;187(2):316-325. Link.
Summary: This paper describes and applies causal mediation analysis methods to answer the question: Would racial disparities in HIV outcomes remain between African Americans and Caucasians had both groups received the same level of HIV treatment and adherence? The authors reframe relevant scientific questions in terms of a controlled direct effect and corresponding proportion eliminated, and argue for estimating the proportion of a disparity we could eliminate to advance equity and to go beyond just documenting disparities.
Potential Audiences: Career epidemiologists
https://academic.oup.com/aje/article/187/2/316/3964397

Intersectional decomposition analysis with differential exposure, effects, and construct
Jackson JW and VanderWeele TJ. Intersectional decomposition analysis with differential exposure, effects, and construct. Social Science & Medicine. 2019;226:254-259.
Summary: This paper reviews various decomposition methods used for conducting intersectionality analyses and how they can be applied in the context of causal analysis. The paper provides a nuanced discussion of intersectionality measurement from a causal inference perspective.
Potential Audiences: Career epidemiologists
https://www.sciencedirect.com/science/article/abs/pii/S0277953619300358?casa_token=I_9V09NpCXgAAAAA:pfQag8Wq-Vr8A1KpyIgezlOzaxT4hm1lP5w9mi-rIsVGZx4t1o4UpRjuUGvtGjGNAfocTOHcPA

December 17 2020

Likely biased but possibly useful: the implications of conditioning on future events for interpretation of effects in perinatal epidemiology

Session Chairs: Mollie Wood, Dominique Heinke

January 29 2020

New Versus Prevalent User Designs In Pharmacoepidemiology: Time To Get Principled About When We Can Be Pragmatic

SER-ISPE Collaboration
Jess Edwards and Til Sturmer

January 15 2020

Association of Changes in Air Quality With Incident Asthma in Children in California, 1993-2014

Erika Garcia, University of Southern California

Link to Article

January 15 2020

Causal Directed Acyclic Graphs (DAGs)

Jessica L. Rohmann
@JLRohmann

Backdoor paths, d-separation, collider-stratification bias, and mediation, oh my! Increasingly, epidemiologists are using structural causal models in the form of directed acyclic graphs (DAGs) to illustrate their assumptions about the causal structure underlying a set of variables based on subject-matter knowledge. Causal DAGs are convenient visual representations backed up by consistent nonparametric mathematics, which link the graphs to statistical assumptions that the user does not have to understand in detail to use in practice. Though there is much more to causal inference than DAGs, this intuitive tool has afforded the field of epidemiology valuable insights into confounding, selection bias, and measurement bias. DAGs can also be used to inform variable selection for causal questions, and offer intuitive explanations for several so-called paradoxes that kept epidemiologists scratching their heads for decades. It is no wonder causal DAGs have recently found their way into curricula of introductory epidemiology courses and textbooks.

Though large in size, this list is most certainly not comprehensive. I am grateful to the #epitwitter community for sharing additional suggestions for resources following the DAGs session at the SER 2019 Annual Meeting.

Part 1: Foundational Readings and Reviews

Causal Diagrams for Epidemiologic Research
Greenland S, Pearl J, Robins JM. Causal diagrams for epidemiologic research. Epidemiology. 1999 Jan;10(1):37–48.
https://www.ncbi.nlm.nih.gov/pubmed/9888278
- Although there were earlier papers, this one arguably introduced DAGs to a wide epidemiology audience
Causal Knowledge as a Prerequisite for Confounding Evaluation: An Application to Birth Defects Epidemiology
Hernán MA, Hernández-Díaz S, Werler MM, Mitchell AA. Causal knowledge as a prerequisite for confounding evaluation: an application to birth defects epidemiology. Am J Epidemiol. 2002 Jan 15;155(2):176–84.
https://www.ncbi.nlm.nih.gov/pubmed/11790682
- An example of how the use of subject-matter knowledge and DAGs to inform analysis changed the results from Slone Epidemiology Unit Birth Defects Study, a case-control study on folic acid supplementation and risk of neural tube defects.
Quantifying Biases in Causal Models: Classical Confounding vs Collider-Stratification Bias
Greenland S. Quantifying biases in causal models: classical confounding vs collider-stratification bias. Epidemiology. 2003 May;14(3):300–6.
https://www.ncbi.nlm.nih.gov/pubmed/12859030
- This paper compares the magnitude of biases introduced by confounding and collider-stratification.
A Structural Approach to Selection Bias
Hernán MA, Hernández-Díaz S, Robins JM. A structural approach to selection bias. Epidemiology. 2004 Sep;15(5):615–25.
https://www.ncbi.nlm.nih.gov/pubmed/15308962
- What’s in a name? The many faces of selection bias boiled down to biases resulting from conditioning on common effects.
When is Baseline Adjustment Useful in Analyses of Change? An Example with Education and Cognitive Change
Glymour MM, Weuve J, Berkman LF, Kawachi I, Robins JM. When is baseline adjustment useful in analyses of change? An example with education and cognitive change. Am J Epidemiol. 2005 Aug 1;162(3):267–78.
https://www.ncbi.nlm.nih.gov/pubmed/15987729
- To adjust or not to adjust? That is the question… In a world of change scores, graphical models are used to illustrate why analyses not adjusted for baseline values are sometimes less biased
The Birth Weight "Paradox" Uncovered?
Hernández-Díaz S, Schisterman EF, Hernán MA. The birth weight “paradox” uncovered? Am J Epidemiol. 2006 Dec 1;164(11):1115–20.
https://www.ncbi.nlm.nih.gov/pubmed/16931543
- Can DAGs help explain why small babies in high-risk populations usually have lower risk than small babies in low-risk populations?
Causal Directed Acyclic Graphs and the Direction of Unmeasured Confounding Bias
VanderWeele TJ, Hernán MA, Robins JM. Causal directed acyclic graphs and the direction of unmeasured confounding bias. Epidemiology. 2008 Sep;19(5):720–8.
https://www.ncbi.nlm.nih.gov/pubmed/18633331
- This paper uses DAGs to visualize the direction of bias when confounding control is inadequate
Chapter 12 "Causal Diagrams" in Modern Epidemiology (Textbook)
Invited Commentary: Causal Diagrams and Measurement Bias
Hernán MA, Cole SR. Invited Commentary: Causal diagrams and measurement bias. Am J Epidemiol. 2009 Oct 15;170(8):959–62; discussion 963–4.
https://www.ncbi.nlm.nih.gov/pubmed/19755635
- Using DAGs to represent 4 types of measurement error of exposure and outcome (independent nondifferential, dependent nondifferential, independent differential, and dependent differential)
"Toward a Clearer Definition of Confounding" Revisited with Directed Acyclic Graphs
Howards PP, Schisterman EF, Poole C, Kaufman JS, Weinberg CR. “Toward a clearer definition of confounding” revisited with directed acyclic graphs. Am J Epidemiol. 2012 Sep 15;176(6):506–11.
https://www.ncbi.nlm.nih.gov/pubmed/22904203
- Howards et. al. revisit Weinberg’s 1993 paper about confounding, using DAGs to visualize the original assumptions made, and show that they are consistent with the original findings.
The Table 2 Fallacy: Presenting and Interpreting Confounder and Modifier Coefficients
Westreich D, Greenland S. The table 2 fallacy: presenting and interpreting confounder and modifier coefficients. Am J Epidemiol. 2013 Feb 15;177(4):292–8.
https://www.ncbi.nlm.nih.gov/pubmed/23371353
- This paper uses DAGs to visualize how presenting multiple coefficient estimates from multivariable regression models in a single table can lead to misinterpretation.
Causal Inference in Statistics: A Primer (Textbook)
Pearl J, Glymour M, Jewell NP. Causal Inference in Statistics – A Primer. 1 edition. Wiley; 2016.
http://bayes.cs.ucla.edu/PRIMER/
- Classical statistics meets causality in this user-friendly yet rigorous primer. Pearl and coauthors use examples to show why causal methods are needed to answer some of the “paradoxes” observed in data. The structural causal models introduced are intuitively linked with counterfactual thinking in the final chapter. Be sure to check the latest errata online; there are some important typos in the 2016 print version.
The Book of Why: The New Science of Cause and Effect (Book)
Pearl J, Mackenzie D. The Book of Why: The New Science of Cause and Effect. Basic Books; 2018.
http://bayes.cs.ucla.edu/WHY/
- This book was written for a general audience but doesn’t shy away from more complex formulations that may lose the casual reader. A nice historical overview that is critical (sometimes harshly so) of many big names in statistics. Lots of examples and interesting anecdotes to help explain and understand “why?” (and of course, causal diagrams!) This book was chosen for the #epitwitter #epibookclub in 2018. You can find some useful chapter-by-chapter tweetorials of the content by @EpiEllie here: https://twitter.com/EpiEllie/status/1005487535706066946
Epidemiology by Design: A Causal Approach to the Health Sciences (Textbook)
Westreich D. Epidemiology by Design: A Causal Approach to the Health Sciences. Oxford University Press; 2019.
https://global.oup.com/academic/product/epidemiology-by-design-9780190665760
- An introductory epidemiology textbook that takes a causal approach. Chapter 3 introduces causal DAGs and they are used throughout the remaining text to illustrate key concepts and examples. Unlike other texts, DAGs are introduced first and not last. Classic observational and intervention designs are explained through a “causal lens” with DAG visuals.
Causal Inference: What If (Textbook)
Hernán MA, Robins JM (2020). Causal Inference: What If. Boca Raton: Chapman & Hall/CRC.
Currently freely available online: https://www.hsph.harvard.edu/miguel-hernan/causal-inference-book/
- This is comprehensive causal inference text (currently ~300 pages) has grown through a series of online revisions and is a must-read for those interested in conducting research on causal questions. Part I, Chapter 6 “Graphical representation of causal effects” is most directly relevant for this playlist and Chapters 7-9 use causal DAGs to explain confounding, selection bias, and measurement error. For the more advanced readers, Part II explores causal inference with models, and Part III dives into making causal inferences using complex longitudinal data. Corresponding computer code for major statistical software programs is also available.
Online edX Course: 'Causal Diagrams: Draw Your Assumptions Before Your Conclusions'
Miguel Hernán
https://www.edx.org/course/causal-diagrams-draw-your-assumptions-before-your-conclusions
- This MOOC covers the basic of causal diagrams, how to identify common biases, and how to use DAGs to inform data analysis. Both free auditing and paid (with completion certificate) options are available.

Part 2: Applied Papers, Examples, and Endorsements of the Use of Casual DAGs in Research

Illustrating Bias Due to Conditioning on a Collider
Cole SR, Platt RW, Schisterman EF, Chu H, Westreich D, Richardson D, et al. Illustrating bias due to conditioning on a collider. Int J Epidemiol. 2010 Apr;39(2):417–20.
https://www.ncbi.nlm.nih.gov/pubmed/19926667
- Useful hypothetical, worked examples to teach concepts underlying bias due to conditioning on a collider.
Endogenous Selection Bias: The Problem of Conditioning on a Collider Variable
Elwert F, Winship C. Endogenous Selection Bias: The Problem of Conditioning on a Collider Variable. Annu Rev Sociol. 2014 Jul;40:31–53.
https://www.ncbi.nlm.nih.gov/pubmed/30111904
- Examples from sociology, social network analysis, and social demography
Causal Models and Learning from Data: Integrating Causal Modeling and Statistical Estimation
Petersen ML, van der Laan MJ. Causal models and learning from data: integrating causal modeling and statistical estimation. Epidemiology. 2014 May;25(3):418–26.
https://www.ncbi.nlm.nih.gov/pubmed/24713881
- Suggested by Sherri Rose on #epitwitter, this is a fantastic paper that walks us through how to integrate our causal model information into common statistical analytic techniques, like regression modeling. Includes a “roadmap” for approaching causal questions and a walk-through on how to translate a scientific question into a target causal quantity.
Graphical Presentation of Confounding in Directed Acyclic Graphs (In Nephrology)
Suttorp MM, Siegerink B, Jager KJ, Zoccali C, Dekker FW. Graphical presentation of confounding in directed acyclic graphs. Nephrol Dial Transplant. 2015 Sep;30(9):1418–23.
https://www.ncbi.nlm.nih.gov/pubmed/25324358
- Gentle introduction to DAGs with clinical examples from the field of nephrology
Constructing Causal Diagrams for Common Perinatal Outcomes: Benefits, Limitations and Motivating Examples with Maternal Antidepressant Use in Pregnancy
Bandoli G, Palmsten K, Flores KF, Chambers CD. Constructing Causal Diagrams for Common Perinatal Outcomes: Benefits, Limitations and Motivating Examples with Maternal Antidepressant Use in Pregnancy. Paediatr Perinat Epidemiol. 2016 Sep;30(5):521–8.
https://www.ncbi.nlm.nih.gov/pubmed/18633331
- An overview of causal graphs and clinical examples from perinatal epidemiology
The C-Word: Scientific Euphemisms Do Not Improve Causal Inference from Observational Data
Hernán MA. The C-Word: Scientific Euphemisms Do Not Improve Causal Inference From Observational Data. Am J Public Health. 2018 May;108(5):616–9.
https://www.ncbi.nlm.nih.gov/pubmed/?term=29565659
- A good article to have on hand when writing a rebuttal to a peer reviewer asking you to replace “we aimed to estimate the causal effect of X on Y” with “we aimed to estimate the association between X and Y”.
Thinking Clearly About Correlations and Causation: Graphical Causal Models for Observational Data. Advances in Methods and Practices in Psychological Science
Rohrer JM. Thinking Clearly About Correlations and Causation: Graphical Causal Models for Observational Data. Advances in Methods and Practices in Psychological Science. 2018 Mar 1;1(1):27–42.
https://doi.org/10.1177/2515245917745629
- Gentle, applied intro to DAGs with examples from psychology
Control of Confounding and Reporting of Results in Causal Inference Studies. Guidance for Authors from Editors of Respiratory, Sleep, and Critical Care Journals
Lederer DJ, Bell SC, Branson RD, Chalmers JD, Marshall R, Maslove DM, et al. Control of Confounding and Reporting of Results in Causal Inference Studies. Guidance for Authors from Editors of Respiratory, Sleep, and Critical Care Journals. Ann Am Thorac Soc. 2019 Jan;16(1):22–8.
https://www.ncbi.nlm.nih.gov/pubmed/30230362
- 47 editors of respiratory, sleep, and critical care journals came together to endorse the use of DAGs for design and reporting of observational studies with causal questions.

Bonus Tracks: Web- and R-based Tools to Create and Manipulate your own DAGs

Robust Causal Inference Using Directed Acyclic Graphs: The R Package "Dagitty"
Textor J, van der Zander B, Gilthorpe MS, Liskiewicz M, Ellison GT. Robust causal inference using directed acyclic graphs: the R package “dagitty.” Int J Epidemiol. 2016 Dec 1;45(6):1887–94.
https://www.ncbi.nlm.nih.gov/pubmed/28089956
- See also the browser-based environment for creating, editing, and analyzing causal models: http://dagitty.net/. Want to understand how dagitty works? Check out the complete criteria and algorithmic framework: https://arxiv.org/abs/1803.00116
Paradoxical Collider Effect in the Analysis of Non-Communicable Disease Epidemiological Data: A Reproducible Illustration and Web Application
Luque-Fernandez MA, Schomaker M, Redondo-Sanchez D, Jose Sanchez Perez M, Vaidya A, Schnitzer ME. Educational Note: Paradoxical collider effect in the analysis of non-communicable disease epidemiological data: a reproducible illustration and web application. Int J Epidemiol [Internet]. 2018 Dec 14.
http://dx.doi.org/10.1093/ije/dyy275
- Fun, interactive R shiny app to explore colliders: http://watzilei.com/shiny/collider/
GGDAG: An R Package for Visualizing and Analyzing Causal Directed Acyclic Graphs

November 20 2019

Emulating target randomized trials to reduce bias in observational studies of comparative effectiveness

Goodarz Danaei, Bernard Lown Associate Professor of Cardiovascular Health
SER-AHA Collaboration

November 1 2019

An Introduction to Quantitative Bias Analysis

Richard MacLehose, University of Minnesota

October 24 2019

Epidemiology Counts – Episode Eight: “Gun Violence”

Click HERE to view “Gun Violence”