Advanced Epidemiology:

Systematic Error and

Quantitative Bias Analysis
3 DAY SHORT COURSE
23 - 25 OCTOBER 2013
BRISBANE

SYLLABUS AND TIMETABLE

Day 1

Models of Causation in Epidemiology

• Models: Causal criteria (Bradford Hill) and Sufficient Component Cause Model
(Rothman)
• Counterfactual / Potential Outcomes Model (Rubin)
Potential Outcome Model
• “Missing data” and how we substitute “exchangeable” populations for what we can’t
observe - why perfect RCTs work
• The inevitable problems with most RCTs and observational studies
• Confounding, Selection Bias, Information Bias
Directed Acyclic Graphs (DAGs)
• Formalizing assumptions for your causal model
• What’s your DAG workshop?
Quantitative Bias Analysis (QBA):
• Principles
• Why do it?
Selection Bias
• Conditioning on common effect(s) and dependent on participation by exposure and
outcome
• Study design options – complete case versus imputation
• QBA to explore implications of, and corrections for, selection bias

1
 Day 2

Confounding
• Conditioning on common cause(s) or variables on back-door path
• Regression model building strategies thought to achieve appropriate confounder
adjustment – the art and science
• Other methods for dealing with confounding – Propensity Scores, and Inverse
Probability Treatment Weighting (IPTW), Instrumental variables (IV)
• QBA to explore implications of unmeasured and residual confounding
Interaction and Effect Measure Modification

Day 3

Information Bias
• Misclassification (categorical variables) and measurement error (continuous variables)
• QBA on misclassification in exposures, outcomes and confounders
Dealing with “difficult issues”
• Workshop – group discussion of examples provided by course participants
Direct and Indirect effects
Probabilistic Bias Analysis
• Bringing it all together – simultaneous assessment of selection, confounding and
information bias.

Course Prerequisites
This course will assume knowledge of epidemiology study design and analytical methods,
the basic principles of systematic error (confounding, selection and information biases) and
biostatistics up to multivariable regression. For example, successful completion of a
Diploma or Masters of Public Health course in epidemiology and biostatistics (or similar) will
usually provide the necessary basis to undertake this course.

Participants
The course is primarily intended for PhD students, early career researchers, and perhaps
advanced MPH students who have a basic background in epidemiology and who wish to add
depth to their understanding of some of the fundamental issues in epidemiological
research.

Course materials and resources

• Lectures: Powerpoint slides used in teaching will be handed out as a course book at the
start of the course. (All readings, however, will be available via DROPBOX – see below.)
Often key points are missing from the handouts, as course attendees will be expected to
deduce these key points in class and write them into their course-book.

2
• Texts: The course draws strongly on two key text books.
1. Rothman, Greenland and Lash (2008). Modern Epidemiology. (3rd Ed.) Lippincott
Williams & Wilkins. It is strongly recommended that course attendees either have
their own, or have very ready access to, this text. Chapter 19, Bias Analysis, is
provided in the readings file on dropbox. This is a dense chapter, but extremely well
organised and comprehensive. For attendees wishing to enhance their learning,
reading pages 345-63 before the course and pages 363-80 will assist.
2. Lash, Fox, Fink (2009). Applying Quantitative Bias Analysis to Epidemiological Data.
Springer. This is an excellent text. A strong point is the range of easy to use Excel
spreadsheets for conducting quantitative bias analysis that accompany the text,
available at: http://sites.google.com/site/biasanalysis/. These Excel spreadsheets
will be provided to attendees, and used for class exercises during the course.

• Readings: We have selected a number of readings that correspond to the major topics
covered in the course. The readings are available on “DROPBOX”
(https://www.dropbox.com/)

Causation and Counterfactuals

o Kaufman, J, Poole C. Looking back on "Causal thinking in the health sciences". Annu Rev
Pub Health. 2000;21: 101-119.
o Little RJ, Rubin DB. Causal Effects in Clinical and Epidemiological Studies Via Potential
Outcomes: Concepts and Analytical Approaches. Annu Rev Public Health. 2000;21: 121-
145.
o Greenland S, Brumback B. An overview of relations among causal modeling methods. Int
J Epidemiol 2002;31:1030-1037.
DAGs
o Hernán MA, Hernández-Diaz S, Werler M, Mitchell AA. Causal knowledge as a
prerequisite for confounding evaluation: An application to birth defects epidemiology.
Am J Epidemiol 2003; 155:176-184
o Hernández-Díaz S, Wilcox AJ, Schisterman EF, Hernán MA. From causal diagrams to birth
weight-specific curves of infant mortality. Eur J Epidemiol. 2008; 23(3): 163–166.
o Hernan MA, Cole SR. Invited Commentary: Causal Diagrams and Measurement Bias. Am.
J. Epidemiol. 2009;170:959-62.
Confounding
o Greenland S, Robins J. Identifiability, exchangeability, and epidemiological confounding.
Int J Epidemiol 1986;15:412-18.
o Greenland S, Morgenstern H. Confounding in health research. Ann Rev Public Health
2001;189-212.
o Davey Smith G, Ebrahim S. Data dredging, bias, or confounding. BMJ 2002;325: 1437-
1438.
Multiple Imputation
o Sterne et al. Multiple imputation for missing data in epidemiological and clinical research:
potential and pitfalls. BMJ 2009. 38 doi: 10.1136/bmj.b2393
o Spratt et al. Strategies for multiple imputation in longitudinal studies. Am J Epidemiol
2010; 172: 478-87.

3
Propensity Scores
o Austin P. An introduction to propensity-score methods for reducing confounding in
observational studies, Multivariate Behavioral Research (2011) 46: 3, 399 - 424.
o Oakes M, Johnson PJ. Propensity score matching for social epidemiology. In: Oakes M and
Kaufman J. (eds) Methods in Social Epidemiology. Jossey Bass, San Francisco: 2006, pgs
370-392.
o Glynn RJ, Schneeweis S, Sturmer T “Indications for propensity scores and review of their
use in pharmacoepidemiology” Basic Clin Pharmacol Toxicol 2006; 98(3):253-9
Instrumental Variables
o Greenland, S. An introduction to instrumental variables for epidemiologists. Int J
Epidemiol. 2000;29(4): 722-729.
o Newhouse JP, McClellan M. Econometrics in outcomes research: The use of instrumental
variables. Annu Rev Public Health 1998;19:17-34.
o Glymour M. Natural experiments and instrumental variable analysis in social
epidemiology. In: Oakes M and Kaufman J. (eds) Methods in Social Epidemiology. Jossey
Bass, San Francisco: 2006, pgs 429-460.
Marginal Structural Models and Inverse Probability Treatment Weighting
o Hernan MA, Robins JM. Estimating causal effects from epidemiological data. J Epidemiol
Community Health 2006; 60:578-586.
o Arijit N, Glymour M, Kawachi I, VanderWeele TJ. Using Marginal Structural Models to
Estimate the Direct Effect of Adverse Childhood Social Conditions on Onset of Heart
Disease, Diabetes, and Stroke Epidemiology. 2012; 23:223-232
o De Stavola BL, Daniel RM. Commentary: Marginal Structural Models: The Way Forward
for Life-course Epidemiology? Epidemiology. 2012; 23:233-237.
Quantitative Bias Analysis
o Modern Epidemiology, Chapter 19: Bias Analysis.
o Greenland S. Quantifying Biases in Causal Models. Epidemiology 2003;14(3):300-06.
o Greenland S. Bayesian perspectives for epidemiologic research: III. Bias analysis via
missing-data methods. Int J Epidemiol 2009:dyp278.
o Fewell et al. The Impact of Residual and Unmeasured Confounding in Epidemiologic
Studies. Am J Epidemiol 2007;166(6):646-55.

4
TIMETABLE
Session Day 1: Wed 23 Oct Day 2: Thur 24 Oct Day 3: Fri 25 Oct
0900 to • Course Overview [TB] • Confounding [TB]: • Information bias [TB]:
1030 • Quantitative bias analysis (QBA) overview o Properties, counterfactual, DAGs o Definitions, DAGs, etc
[TB] o Approaches to regression model building • Information bias analysis 1:
• Causation [JL] o QBA Class exercises – Excel

1100 to • Potential Outcomes Model [JL] • Other methods for confounding [JL]: • Probabilistic bias analysis 2:
1230 • DAGs [JL] o Propensity scores o Distributions, Monte Carlo, etc
• What’s your DAG? (class workshop) [JL/TB] o MSM / IPTW o Class exercise – Excel

1330 to • Selection bias 3 [TB]: • Other methods for confounding (cont) [JL]: • Class workshop: [TB, LJ]
1500 o Definitions and DAGs o Instrumental Variables o Case studies of ‘difficult’ issues. Structure: brief
o Study design options – complete case vs • Interaction (if time): [TB] presentation, class discussion.
imputation
• What does OR measure [TB]

1530 to • Selection QBA 4 [TB]: • Confounding bias analysis 5 [TB]: • Direct and Indirect Effects [TB]
1700 o Formulas o Formulas • Multiple bias analysis [TB]:
o QBA Class exercises – Excel o QBA Class exercises – Excel o Modern Epi Ch 19 example
• Quiz • Quiz • Final Quiz

1
Spreadsheet for class: 2c. Misclassification for TB Adv Epi Class Exercises.xls
2
Recommended reading: Modern Epidemiology, Chapter 19: Bias Analysis (p.363-80). Recommended reading: Lash et al, Applying Quantitative Bias Analysis to Epidemiological
Data, Chapters 8 and 9. Spreadsheets for class: 2d. Multiple Probabilistic Bias Analysis for TB Adv Epi Class Exercises.xls; 2d. Multiple Probabilistic Bias Analysis for TB Adv Epi Class
Exercises.xls
3
Spreadsheets for class: ‘Selection Bias, TB, I.xls’ and ‘Selection Bias, TB, II.xls’
4
Spreadsheets for class: 2a. Selection Bias for TB Adv Epi Class Exercises.xls
5
Spreadsheet for class: 2b. Unmeasured Confounding for TB Adv Epi Class Exercises.xls

5
6