From the Guidelines
Selection bias occurs when the sample chosen for a study is not representative of the population, leading to an incorrect estimate of the true effect of an exposure on the outcome of interest.
Types of Selection Bias
- Differential attrition: when participants drop out of a study at different rates, potentially due to the exposure or outcome of interest 1
- Differential enrollment: when participants are selected into a study based on factors related to the exposure or outcome of interest 1
- Sampling bias: when some individuals within a target population are more likely to be selected for inclusion than others 1
- Allocation bias: when there is a systematic difference between participants in exposed and unexposed groups 1
Causes of Selection Bias
- Non-response: when some individuals do not respond to a survey or study, potentially due to the exposure or outcome of interest 1
- Loss to follow-up: when some individuals are lost to follow-up, potentially due to the exposure or outcome of interest 1
- Pre-enrollment selective mortality: when participants who are selected into a study have survived to the age of enrollment only by virtue of their unusually effective detoxification genotype or cognitive acumen 1
Consequences of Selection Bias
- Spurious associations: when selection bias creates an association between the exposure and outcome that is not real 1
- Attenuation of effect estimates: when selection bias reduces the estimated effect of the exposure on the outcome 1
- Reversal of association: when selection bias reverses the direction of the association between the exposure and outcome 1 To minimize selection bias, researchers should use strategies such as complete enumeration and inclusion of all participants, concealment of treatment allocation, and sensitivity analyses to illuminate the robustness of the findings 1.
From the Research
Definition of Selection Bias
- Selection bias refers to the distortion of study results due to the selection of participants or samples in a way that does not accurately represent the population of interest 2, 3, 4.
- It can occur in various forms, including selection of representative subjects, selection of subjects to exposures, and selection of subjects at outcome 2.
- Selection bias can lead to biased estimates of causal effects and can be difficult to detect and correct 3, 4.
Types of Selection Bias
- Type 1 selection bias occurs when the sample is restricted to one or more levels of a collider (or a descendant of a collider) 4.
- Type 2 selection bias occurs when the sample is restricted to one or more levels of an effect measure modifier 4.
- Allocation bias is a type of selection bias that can occur in clinical trials, where knowledge of interventions prior to group allocation can result in systematic differences in important characteristics 5.
Causes and Consequences of Selection Bias
- Selection bias can be caused by conditioning on a collider variable, which can lead to spurious associations between variables 3.
- It can also be caused by restricting the sample to certain levels of a variable, which can lead to biased estimates of causal effects 4.
- Selection bias can have significant consequences, including overestimation of effect sizes and distorted study findings 2, 3, 5.
Minimizing Selection Bias
- Methods to minimize selection bias include using randomization, sequential numbering, and opaque sealed envelopes to allocate participants to groups 5.
- Increasing awareness and understanding of collider-conditioning bias can also help to reduce selection bias 6.
- Using causal diagrams and sensitivity analyses can help to identify and correct selection bias 3, 4.