Course P282/03 Introduction to Quantitative Bias Analysis for Epidemiologic Research

ECTS: 3.3

Course leader: Oleguer Plana-Ripoll

Language: English

Graduate school: Faculty of Health

Graduate program: PH

Course fee: 3,960.00 DKK

Max seats: 25

Applicants: 17

Status: Course is open for application

Semester: Spring 2021

Application deadline: 10/05/2021

Start date: 07/06/2021

Administrator: Johanne Gregor Nielsen

Postponement

Please notice this course has been postponed until 7 -11 June 2021.

Title: Introduction to Quantitative Bias Analysis for Epidemiologic Research

Reg.no: P282/03

is being offered by the Graduate School of Health, Aarhus University, spring 2021.

Criteria for participation: University degree in medicine, dentistry, nursing, or Master’s degree in other fields and/or postgraduate research fellows (PhD students and research-year medical students).

Aim: This course aims to provide participants with an introduction to Quantitative Bias Analysis. A compilation of bias analysis methods for use with epidemiologic data will be introduced, including methods for corrections to address selection bias, uncontrolled confounding, and classification errors, as well as methods to quantify different sources of biases simultaneously. The course will consist of a combination of lectures, short presentations, group work, discussions, and individual exercises.

Learning outcomes: At the end of the course, students should be able to:

  • Identify different sources of biases in epidemiological studies.
  • Conduct simple, multidimensional and probabilistic bias analysis using summary data in Microsoft Excel and interpret the output.
  • Conduct probabilistic bias analysis using individual level data (record level correction) in STATA and interpret the output.
  • Discuss the strengths and limitations of each approach.

Content: The conventional epidemiologic approach in observational studies is to reduce systematic error (bias) in the design of their studies, and to describe random error by estimating 95% confidence intervals (or p-values) for point estimates. In addition, associations are usually adjusted for confounding variables using regression models. However, there is rarely a quantification (or correction) for selection bias, measurement error, confounding by unmeasured or unknown confounders, or residual confounding by measured confounders that are poorly specified or poorly measured. Quantitative bias analysis (QBA) provides a methodology for assessing the impact of bias on study results (both in direction and magnitude), by making assumptions about the bias parameters. Such analyses allow investigators to go beyond speculation about the bias in the discussion section of manuscripts.

This course will provide an introduction to QBA methods. First, the course will cover in detail simple and multidimensional bias analysis methods, that can be used to quantify potential bias due to selection bias, unmeasured confounding, and misclassification. These basic methods can be applied to nearly any dataset, including summary data presented in the literature (e.g. 2x2 tables). We will then continue with probabilistic bias analysis, that builds on the more basic methods to create intervals accounting for the uncertainty in the systematic error, and methods to account for the total error (systematic and random) in the study results. Finally, the course will cover bias analysis using individual level data (record level correction).

The course will consist of a combination of lectures, short presentations with examples of quantitative bias analysis, group work and individual exercises with real data, and discussions – both general but also specific about student’s concerns with their own research related to bias analysis.

Recommended knowledge for participation (if any): The main necessary concepts to understand bias analysis will be covered during the course, but it is recommended that the student has basic previous knowledge of epidemiology and biostatistics, such as measures of association, study design, selection bias, confounding, information bias, basic regression models, etc.). The course material is based on Microsoft Excel and the statistical package STATA. Users of other similar packages can also attend the course, but they need to be familiar with the software.

Instructors: Bodil Hammer Bech, Chunsen Wu, and Oleguer Plana-Ripoll.

Place: Det Blå Auditorie (1266-222), The Victor Albeck Building, Vennelyst Boulevard 4, 8000 Aarhus C

Participation in the course is without cost for:

  • PhD students, Research Year students and Research Honours Programme students from Aarhus University
  • PhD students enrolled at partner universities of the Nordoc collaboration
  • PhD students from other institutions in the open market agreement for PhD courses

Course dates:

  • 07 June 2021 09:00 - 16:00
  • 08 June 2021 09:00 - 16:00
  • 09 June 2021 09:00 - 16:00
  • 10 June 2021 09:00 - 16:00
  • 11 June 2021 09:00 - 16:00