역학적 연구의 편향

Bias is a systematic tendency for an estimate or expected value to deviate from the true value. For example, a thermometer consistently measuring body temperature 3 degrees lower produces biased measurements. In epidemiology, biases distort study results and can mislead conclusions about disease relationships, making recognition and control essential for research validity.

Selection bias occurs when the study population is not representative of the target population. This happens when participants are selected with higher or lower probability based on certain characteristics. For instance, surveying only urban dwellers to understand a national health issue introduces selection bias, compromising the generalizability of findings.

Attrition bias occurs when participants drop out during study follow-up, especially if dropouts differ systematically from those who remain. This differential loss skews results because the remaining sample no longer represents the original population. Non-respondents may differ regarding the variable of interest, further distorting study conclusions.

Spectrum bias occurs when a diagnostic test is evaluated using a non-representative patient population, inflating perceptions of test accuracy and reliability. Testing only severe cases or only mild cases produces biased estimates that do not reflect real-world performance. This misrepresentation can lead to inappropriate clinical decisions based on overstated test validity.

Observer bias arises when researchers subconsciously introduce skew during data collection or analysis through preconceived notions. This interviewer bias affects how responses are interpreted and recorded. Minimizing observer bias requires standardized protocols, blinding when possible, and awareness of personal assumptions that could distort findings.

Confounding bias occurs when an extraneous variable correlates with both the dependent and independent variables, distorting the true relationship. For example, in a smoking and lung cancer study, age could confound results because older individuals may have both higher smoking rates and higher cancer rates. Proper study design and statistical adjustment are needed to address confounding.

Memory bias is a type of information bias where participants fail to accurately recall past events, leading to misclassification of exposure or outcomes. This recall error is particularly problematic in retrospective studies relying on participant recollection. Accurate data collection requires careful questionnaire design and validation to minimize memory-related errors.