TABLE 2.

General definitions of select terms

Term	Definition
Target trial	Ideal randomized clinical trial one would conduct if possible [16]
Counterfactual	Potential outcomes [17]
Exchangeability	Assumption that all individuals in a study population are exchangeable with each other on all relevant factors except exposure status [19]
Causal diagram (e.g. DAG)	Graphical tool used to represent a causal model and assess for potential violations of exchangeability [20,21]
Antecedent	Factor that is a cause of another factor [23]
Descendant	Factor that is an effect of another factor [23]
Minimally sufficient adjustment set	Set of factors identified by a DAG that, if adjusted for, can efficiently close off biasing noncausal paths [23]
Confounding bias	Factor that is associated with the outcome (in the causal direction towards the outcome) regardless of exposure status, distributed unequally across exposure levels, and is not on the causal path between the exposure and outcome [24]
Selection bias	Selecting or restricting the study sample based on a common antecedent factor of the exposure and outcome (i.e. a confounding factor), which violates exchangeability [27]
Collider bias	Selecting or restricting the study sample based on, or adjusting for or stratifying on, a common descendent factor of the exposure and outcome, which can induce a spurious association or distort the magnitude or direction of an association between the exposure and outcome [26]
Information bias	Error in obtaining or documenting information used to inform, characterize, or define a factor; related to misclassification bias [29]
Index time	Start of study period and/or assessment of exposure status
Time-dependent bias	Failure to account appropriately for index time or time during follow up, which erroneously favors one exposure group over another [30]
Immortal time bias	Participant’s inability to reach an outcome over a set period of time because of study design and/or analytic plan restrictions [31]
Immeasurable time bias	Exposure groups are assigned erroneously the same exposure time because of lack of information about how exposure status changes over time [32]
Propensity score	Conditional probability of exposure (i.e. treatment) assignment based on observed covariates [36]
Table 2 fallacy	Presentation of multiple effect estimates for exposures and covariates from the same multivariable model that confounds direct-effect and total-effect estimates and encourages misinterpretation of results [43]
Competing risk	A secondary event that precludes a participant’s ability to reach the primary outcome, which biases survival estimates [49]
Sensitivity analysis	Quantification of a model’s uncertainty because of bias related to the assumptions that one makes in the primary analyses [52]
Negative control	Surrogate marker for the actual exposure or outcome that can help identify and address various sources of bias [64]
Effect modification	Magnitude or direction of the exposure-outcome association differs across levels of a third factor [74]
Mediation	Third factor that lies on the causal and temporal paths between an exposure and outcome, leading to indirect effects of the exposure on the outcome via the third factor, in addition to the direct effects of an exposure on an outcome [80]

DAG, directed acyclic graph.