About this video
- Video Title: POPH192 Lec 25-27
- Channel: Oban Hawkins
- Speakers: Oban Hawkins
- Duration: 00:44:08
Overview
This video lecture covers concepts of bias and confounding in epidemiological studies. It explains different types of bias, including selection bias and information bias, and how they can affect study results. The lecture also discusses confounding factors, their characteristics, and methods to control for them in study design and analysis.
Key takeaways
- Bias Defined: Bias is a systematic error in an epidemiological study that leads to an incorrect estimate of the association between exposure and disease risk. Unlike random error (chance), bias is inherent in the study's design or execution.
- Controlling Bias: Bias can only be controlled during the design and data collection phases of a study; it cannot be corrected after the study is completed.
- Selection Bias: Occurs due to systematic differences between study participants and non-participants, impacting who enters or remains in the study. It is particularly relevant in historical studies, especially case-control studies, where participation must be related to both exposure and outcome status.
- Information Bias: Arises from systematic differences in how data on exposure or outcome is obtained from different study groups. This can stem from inaccurate participant responses (like recall bias) or systematic measurement errors by researchers.
- Differential vs. Non-Differential Misclassification: Differential misclassification occurs unevenly between groups and can bias results, while non-differential misclassification occurs evenly and typically has less impact on the measure of association.
- Confounding Explained: Confounding occurs when the relationship between an exposure and an outcome is mixed with the effect of a third factor (confounder). A confounder must be independently associated with the exposure, independently associated with the outcome, and not on the causal pathway between the exposure and outcome.
- Controlling Confounding: Methods to control confounding include randomization, restriction, and matching, applied during the study design phase. Analysis phase adjustments can also be made.
