Evidence-Based Approaches to Public Health: Epidemiology – Bias and Confounding: Types of Bias (Selection, Information, Publication)

In this tutorial, we will explore the different types of bias that can affect epidemiological studies: selection bias, information bias, and publication bias. Bias refers to systematic errors that can lead to incorrect conclusions about the relationship between exposures and outcomes in research. Understanding these biases is essential for the Certified in Public Health (CPH) exam and for interpreting public health research critically.

By the end of this tutorial, you will understand how these types of bias occur, how they affect study results, and how they can be minimized. Practice questions will also be included to reinforce your knowledge.

Table of Contents:

1. Introduction to Bias in Epidemiology
2. Selection Bias
   • Definition of Selection Bias
   • Examples of Selection Bias
   • How to Minimize Selection Bias
3. Information Bias
   • Definition of Information Bias
   • Types of Information Bias
   • How to Minimize Information Bias
4. Publication Bias
   • Definition of Publication Bias
   • Impact of Publication Bias
   • How to Minimize Publication Bias
5. Practice Questions
6. Conclusion

1. Introduction to Bias in Epidemiology

Bias is a systematic error that can distort the findings of an epidemiological study, leading to incorrect conclusions about the relationship between an exposure and an outcome. Bias can occur at any stage of research, from the selection of study participants to data collection and analysis. It is crucial for researchers to recognize and minimize bias to ensure the validity and reliability of study results.

There are several types of bias that can affect epidemiological studies, including selection bias, information bias, and publication bias. Each type of bias impacts research in different ways.

2. Selection Bias

Selection bias occurs when there is a systematic difference between those selected for the study and those who are not, leading to results that are not representative of the target population. This can occur if the selection process for participants is flawed or if certain groups are more likely to participate than others.

2.1 Definition of Selection Bias

Selection bias arises when the participants included in the study are not representative of the population that the study aims to analyze. This can lead to incorrect estimates of the association between an exposure and an outcome.

2.2 Examples of Selection Bias

• Non-response bias: When individuals who choose to participate in a study differ systematically from those who do not, such as when healthier individuals are more likely to participate in a health survey.
• Loss to follow-up bias: Occurs in longitudinal studies when participants drop out over time, and those who remain in the study are systematically different from those who drop out (e.g., healthier individuals may be more likely to stay in a study).
• Berkson’s bias: In case-control studies, when cases and controls are selected from a hospital population, which may not be representative of the general population, leading to a biased comparison.

2.3 How to Minimize Selection Bias

• Randomization: In experimental studies, randomly assigning participants to groups helps ensure that the study population is representative of the larger population.
• Careful participant recruitment: Ensuring that inclusion and exclusion criteria are well-defined and that the recruitment process targets a representative sample of the population.
• Tracking and follow-up: In longitudinal studies, maintaining contact with participants and minimizing loss to follow-up can help reduce selection bias.

3. Information Bias

Information bias occurs when there are systematic errors in how data are collected, measured, or classified. This bias can affect both exposure and outcome data and may lead to incorrect associations between variables.

3.1 Definition of Information Bias

Information bias refers to errors in the collection, recall, or recording of information from study participants. These errors can occur during data collection or when interpreting the data, leading to misclassification of exposure or outcome status.

3.2 Types of Information Bias

• Recall bias: Occurs when participants in a study do not remember past events accurately, which can affect retrospective studies. For example, cases may be more likely to recall a potential exposure than controls, leading to biased results.
• Interviewer bias: Occurs when the interviewer unintentionally influences the responses of participants, particularly in studies where the interviewer knows the participants’ case or control status.
• Misclassification bias: Can occur when participants are placed into the wrong exposure or outcome categories, leading to incorrect associations. Misclassification can be either differential (systematic) or non-differential (random).

3.3 How to Minimize Information Bias

• Blinding: Blinding interviewers and participants to the study’s hypothesis or to their group assignment (e.g., case or control) can reduce interviewer bias and recall bias.
• Standardized data collection: Using standardized, objective, and validated data collection tools can help ensure that information is collected consistently across participants.
• Training: Providing thorough training for interviewers and data collectors to ensure that they follow standardized protocols and do not inadvertently introduce bias.

4. Publication Bias

Publication bias occurs when studies with positive or significant results are more likely to be published than studies with negative or non-significant findings. This can lead to an overestimation of the effectiveness of interventions or associations between exposures and outcomes, as the published literature does not reflect the full range of evidence.

4.1 Definition of Publication Bias

Publication bias refers to the tendency for journals to preferentially publish studies with statistically significant or positive results. As a result, studies with null or negative findings may remain unpublished, leading to a skewed representation of the evidence in systematic reviews or meta-analyses.

4.2 Impact of Publication Bias

• Overestimation of effects: Since only positive findings are published, the effectiveness of an intervention may be overestimated in the literature.
• Skewed evidence base: Publication bias can lead to an incomplete understanding of the relationship between exposure and outcomes, as non-significant results are underreported.

4.3 How to Minimize Publication Bias

• Pre-registration of trials: Registering trials in a public database before data collection begins can help ensure that all results, positive or negative, are reported.
• Encouraging publication of all results: Journals and researchers should be encouraged to publish studies regardless of their findings, ensuring a more complete evidence base.
• Systematic reviews: Researchers conducting systematic reviews should make efforts to search for and include unpublished or gray literature to minimize the impact of publication bias.

5. Practice Questions

Test your understanding of bias in epidemiological studies with the following practice questions. Try answering these before checking the solutions.

Question 1:

A study on the relationship between diet and heart disease recruits participants through a social media campaign. Health-conscious individuals are more likely to respond, leading to an overestimation of the association between a healthy diet and heart disease. What type of bias is this?

Answer 1:

Question 2:

In a case-control study, participants with lung cancer are more likely to accurately recall their smoking history than participants without lung cancer. What type of bias is this?

Answer 2:

Question 3:

A systematic review of interventions to reduce obesity only includes studies that reported positive results, leading to an overestimation of the intervention’s effectiveness. What type of bias is this?

Answer 3:

6. Conclusion

Bias is an important consideration in epidemiological research, as it can distort the findings and lead to incorrect conclusions about the relationship between exposures and outcomes. Understanding the different types of bias, including selection bias, information bias, and publication bias, is critical for interpreting research results accurately.

Remember:

• Selection bias occurs when the participants in a study are not representative of the population being studied, leading to biased estimates of associations.
• Information bias refers to errors in the way data are collected or classified, leading to incorrect associations between exposures and outcomes.
• Publication bias occurs when studies with positive or significant results are more likely to be published, resulting in a skewed evidence base.

Final Tip for the CPH Exam:

Make sure you can identify different types of bias and understand how they can affect epidemiological research. Look into studies which have been identified to have each sort of bias and try to determine where the bias happens to exist. This knowledge will be essential for answering questions related to bias and confounding on the Certified in Public Health (CPH) exam.

Humanities Moment

The featured image for this article is Hydrangea (1920s, published post-humously) by Sakai Hōitsu (Japanese, 1761-1828). Sakai Hōitsu was a Japanese painter of the Rinpa school, renowned for reviving the style of Ogata Kōrin through reproductions and extensive studies of his work. Born into the Sakai daimyō clan, he trained in various artistic traditions before becoming a Buddhist priest in 1797, spending his final years in seclusion producing significant contributions to Rinpa art, including woodblock print books such as Kōrin Hyakuzu and Oson Gafu.