Epidemiology: Bias and Confounding

Epidemiology: Bias and Confounding

• Research errors include random errors (sampling, lack of precision, variability in measurement) and systematic errors (bias).
• Bias is any systematic error in a study that results in an incorrect estimate of the association between exposure and risk of an outcome.
• Complete elimination of research bias is often impossible. Researchers aim to avoid bias in the study design and minimize its effect during analysis.
• Bias is a result of an error in the design or conduct of a study.

Types of Bias: Selection

• Selection bias occurs when the way cases and controls are selected creates an apparent association that doesn't exist in reality.
• Selection bias in case-control studies can occur if cases exposed to a substance are more likely to be chosen than those not exposed or vice versa.
• Subjects in a study may not be representative of the overall population.
• Examples include coffee drinking and pancreatic cancer, where stratification by drinking levels might mitigate selection bias.

Types of Bias: Non-Response Bias

• Non-response bias occurs when the response rate of a population with a particular disease is higher among those exposed than those not exposed.
• A lower response rate among smokers in an asthma study via mail questionnaires could lead to bias.
• Researchers try to collect data from non-respondents to mitigate this bias.

Types of Bias: Information Bias

• Information bias occurs when methods to collect information about study participants are inadequate and lead to incorrect exposure or outcome data.
• Different levels of information given in the exposed and unexposed populations, or cases and controls, lead to incorrect conclusions.
• Misclassifying exposure status can also cause information bias, especially if based on interviews without objective verification.

Types of Bias: Information Bias- Recall Bias

• Recall bias occurs when past exposures are more accurately recalled by some individuals than others, leading to bias in results.

Types of Bias: Information Bias- Interviewer Bias

• Interviewer bias occurs when interviewers subconsciously influence responses in different study groups. Blinding interviewers to the outcome can help reduce this.

Types of Bias: Information Bias- Misclassification Bias

• Misclassification bias occurs when participants' exposure or disease status is incorrectly categorized during the study.
• This bias can be differential (misclassifying the exposure rates changes between groups) or non-differential (misclassifying rates don't change between groups).
• The rate of misclassification may change between cases and controls during the study.

Sources of Information Bias

• Bias in abstracting records
• Bias in interviewing
• Bias from surrogate interviews
• Surveillance bias
• Recall bias
• Reporting bias
• Wish bias

Impact of Bias on Estimation

• Bias can underestimate or overestimate the RR/OR.
• Researchers want to be more confident in their findings, especially if they underestimated the RR/OR.
• Understanding how different factors might bias results of studies is important to drawing accurate conclusions.

Evaluating Epidemiological Associations

• Consider if the association could be observed by chance.
• Evaluate the potential for bias.
• Assess confounding variables that may account for observed outcomes.

Confounding

• Confounding variables (C) interfere with the observed association between a causal factor (A) and outcome (B).
• They are associated with the putative causal factor and are causally associated with the outcome.
• Cigarette smoking can be a confounding variable in studies connecting coffee consumption to pancreatic cancer.

Controlling for Confounders

• Research design strategies such as randomization, restriction, and matching can help control for confounders.
• Analyze data by using stratification or multivariate analysis to adjust for confounding.
• Example: Stratifying results by sex, age and other characteristics and adjusting for them.

Types of Associations

• Associations can be real (causal) or spurious (due to confounds).
• An association between maternal smoking and low birth weight is causal.
• An association between coffee drinking and pancreatic cancer might be spurious if confounded by smoking.