Notes on Bias in Epidemiological Studies

Systematic Error: A systematic error occurs repeatedly throughout a study, impacting the outcome in a specific direction, unlike random errors (occasional mistakes).
It can lead to overestimating or underestimating causal relationships between exposure and disease.

Occurs when the method of selecting cases and controls creates a misleading association.
Examples:
- Nonresponse Bias: Individuals who choose to participate in studies may differ significantly from those who do not. For instance, in an asthma study, non-respondents were found to have higher rates of asthma and were more likely to smoke.
- Volunteer Bias: Over time, particularly in cohort studies, individuals may drop out due to relocation, death, or loss of interest, possibly skewing the data.
Impact: Selection bias compromises the validity of study results by influencing the detected association between exposure and disease.

Defined by different eligibility criteria applied between cases and controls in a case-control study.
Example: In studies about breast cancer and a medication, excluding users of the medication for other indications from the control group distorted the comparison.
- Proper study design requires consistent inclusion/exclusion criteria established before the study starts to avoid this bias.

Ensuring that the same criteria are applied to both cases and controls. This helps prevent skewing the results due to arbitrary exclusions or inclusions.
- Recruitment of participants should occur before diagnosis to maintain a balanced identification of cases and controls.

When the technique used to gather data about subjects is flawed or yields inaccurate results.

Wrongly categorizing individuals as cases or controls can dilute the strength of the associations observed between exposure and disease.
- Often arises from the limited sensitivity/specificity of diagnostic tests.

Medical records, not originally meant for research, may have incomplete data, impacting the accuracy of study findings.

A systematic difference in the recollection of past exposures between groups.
Example: In studies involving women with children with birth defects, those mothers may remember over-the-counter medication use better than mothers with healthy children due to emotional factors surrounding their experience.
This bias can skew results as one group systematically reports better than the other, which can be misinterpreted as a true association.

Distinct from recall bias; it occurs when individuals intentionally misreport information, which can occur due to social desirability or stigma related to subjects like smoking.

Internal Validity: Refers to the degree to which the study accurately reflects the population studied. Threats to internal validity stem from biases like selection bias, misclassification, or other methodological flaws.
External Validity: Concerns the extent to which findings can be generalized to a broader population. Selection bias particularly affects external validity as it may result in an unrepresentative sample.

Recognizing and addressing the various types of bias is critical for the integrity of research findings in epidemiology and understanding causal relationships in health research.
Proper study design, including clearly defined criteria and recruitment strategies, along with robust data collection methods, can help mitigate these biases, leading to more accurate conclusions.