Causal Inference in Epidemiology: Concepts and Methods
Many observational studies aim to make causal inferences about effects of interventions or exposures on health outcomes. This course defines causation, describes how emulating a ‘target trial’ can clarify the research question and guide analysis choices, introduces methods to make causal inferences from observational data and explains the assumptions underpinning them, which can be encoded using directed acyclic graphs (DAGs). Learning is consolidated by interactive discussion-based and computer practical sessions. The course is taught by academics and researchers from the University of Bristol’s Department of Population Health Sciences, MRC Integrative Epidemiology Unit and NIHR Bristol Biomedical Research Centre who are experts in the field with extensive experience of developing and applying relevant methods. This is an advanced course. Familiarity with regression models including Cox models for time-to-event data and their implementation in statistical software (R or Stata) is essential.
| Dates | 29 June - 3 July 2026 |
|---|---|
| Fee | £1250 |
| Format | Online |
| Audience | Open to all applicants (prerequisites apply) |
Course profile
This course aims to define causation in biomedical research, describe methods to make causal inferences in epidemiology and health services research, and demonstrate the practical application of these methods.
Please click on the sections below for more information.
This 5-day course will run online and will consist of a mixture of lectures, small group work and computing practicals. Computer practicals take place in virtual breakout rooms with the help of tutors, using participants’ own computers or our Posit Cloud environment. All sessions will be live.
By the end of the course participants should be able to:
- describe the potential (counterfactual) outcomes approach to defining causal effects;
- use Directed Acyclic Graphs (DAGs) to document assumptions and inform analysis plans;
- recognise the key sources of bias in causal analyses of observational data, and how to investigate them using DAGs;
- define a research question comparing health interventions using a hypothetical ‘target trial’
- apply key methods to estimate causal effects by emulating a target trial, and recognise the situations in which they are appropriate and the assumptions underlying them.
- recognise the challenges of making causal inferences about effects of exposures
This advanced course is aimed at epidemiologists, statisticians and other quantitative researchers. Applicants must have knowledge and experience of a variety of regression models, including Cox models for time-to-event data, and their implementation in Stata or R. Such knowledge must be to beyond the level achieved in the Introduction to Linear and Logistic Regression Models course. We recommend that you do not attend this course in the same year that you have attended Introduction to Linear and Logistic Regression Models.
The course will cover:
- potential (counterfactual) outcomes;
- causal diagrams (DAGs);
- confounding and methods to control for confounding (stratification, regression, propensity scores and inverse probability weighting);
- selection and information biases;
- inverse probability weighting to deal with informative censoring
- target trials to define a causal question about health interventions
- instrumental variable estimation;
- intention-to-treat and per-protocol effects in randomized trials and observational studies;
- time-varying confounding, marginal structural models and other g-methods;
- sequential approaches to emulating a target trial using observational data;
- avoiding bias caused by immortal time: the clone-censor-IP weight approach;
- model selection for causal inference studies
- study designs for causal inference; and
- reporting and triangulating causal inference studies
To make sure the course is suitable for you and you will benefit from attending, please ensure you meet the following prerequisites before booking:
| Knowledge |
Applicants must have knowledge and experience of a variety of regression models, including Cox models for time-to-event data, and their implementation in Stata or R. Such knowledge must be to beyond the level achieved in the Introduction to Linear and Logistic Regression Models course. We recommend that you do not attend this course in the same year that you have attended Introduction to Linear and Logistic Regression Models |
|---|---|
| Software |
You must have a recent version of Stata* or R installed in advance of the course. We recommend running this through RStudio Desktop or Posit Cloud**. *Internal University of Bristol participants are given access to Stata. Go to Stata Installation Instructions (internal only) for help setting it up before the start of the course. External participants are responsible for providing their own access to Stata, however if you are a student, Stata offer a short term free Student licence (one week). **A link to create an account and access Posit Cloud will be provided. |
Before booking this course, please make sure you read the information provided above about the target audience and prerequisites. It is important that you have access to the relevant IT resources needed for the course and meet the knowledge prerequisites to ensure you can get the most from the course.
Bookings are taken via our online booking system, for which you must register an account. To check if you are eligible for free or discounted courses please see our fees and voucher packs page. All bookings are subject to our terms & conditions, which can be read in full here.
For help and support with booking a course refer to our booking information page, FAQs or feel free to contact us directly. For available payment options please see: How to pay your short course fees.
Participants are granted access to our virtual learning platform (Blackboard Ultra) 1 to 2 weeks in advance of the course. This allows time for any pre-course work to be completed and to familiarise with the platform.
To gain the most from the course, we recommend that you attend in full and participate in all interactive components. We endeavour to record all live lecture sessions and upload these to the online learning environment within 24 hours. This allows course participants to review these sessions at leisure and revisit them multiple times. Please note that we do not record breakout sessions.
All course participants retain access to the online learning materials and recordings for 5 months after the course.
University of Bristol staff and postgraduate students who do not wish to attend the full course may instead register for access to the 'Materials & Recordings' version of this course: Further information and bookings.
94% of attendees recommend this course*.
*Attendee feedback from 2025.
Here is a sample of feedback from the last run of the course:
“World experts delivering course material clearly. Coverage of complex topics with practicals to support learning and actual use of methods." - Course feedback, July 2025.
“A great overview of a broad range of causal inference concepts and methods, with opportunities to implement them in practical sessions. I now feel much more comfortable with the properties of DAGs and how to interpret them, as well as thinking about biases in studies, and broadly how to deal with different types." - Course feedback, July 2025.
“I found it really brilliant - to navigate key papers in field, key resources, to get familiar with the language of DAGS and Causality which I have not found the most accessible in the past, and really the chance to practice some of the solutions/ methods." - Course feedback, July 2025.
“I really appreciated the depth of content in the course, the approach-ability of the instructors, and the readily available resources. I thought the practicals were engaging." - Course feedback, July 2025.
“I thought the course content and design was really excellent. Probably the best short course I have done (joint with the genetic epidemiology course). The facilitators were all great! They explained stuff super clearly (especially Kate Tilling) and also had a good sense of humour. It was nice to have R scripts for the practicals (rather than just stata)." - Course feedback, July 2025.
“Really enthusiastic knowledgeable course tutors, very keen to make sure we understood the material. Really appreciated that the sessions were all live rather than pre-recorded as being able to ask questions throughout was very helpful." - Course feedback, July 2025.
“Really great mixture of teaching styles, detail and attention in sessions. Paul in particular was really good at explaining complicated concepts in the practical's. The expertise was clear, it moved with good pace, huge amount of content. Well worth the cost." - Course feedback, July 2025.
“Teachers explained key concepts really well. Good mix of lectures, pen and paper and computer practicals. Excellent introduction to DAGs." - Course feedback, July 2025.
“The lectures and lab sessions went well, questions were answered which made it engaging and understandable." - Course feedback, July 2025.
“The quality of expertise of the tutors was excellent and the teaching of DAGS initially worked really well. The build up and 'denouement' (as it were) of the course structure was good." - Course feedback, July 2025.