| (Observed) baseline incomparability . | |
|---|---|
| Covariate distribution shift | Systematic differences exist in the baseline characteristics of patients in external studies compared to those in SATs |
| (Unobserved) outcome mean non-exchangeability | |
| Unmeasured confounding | External control data might not capture the same detailed patient information as SATs, leading to biases from unknown or unmeasured factors |
| Lack of concurrency/ temporal bias | External control data and SATs may be collected in different time periods or under varying healthcare settings |
| Measurement error | There may exist potential inconsistencies in how patient information is collected and recorded, termed here as measurement errors in covariates |
| Outcome validity | Methods of measuring outcomes in external data sources might differ from those used in SATs, or the outcomes might not be clearly defined or reliable |
| Proper estimands | |
| Intercurrent events | Intercurrent events, such as stopping medication and/or adding rescue therapy, can confound the causal effects of the randomized treatment |
| (Observed) baseline incomparability . | |
|---|---|
| Covariate distribution shift | Systematic differences exist in the baseline characteristics of patients in external studies compared to those in SATs |
| (Unobserved) outcome mean non-exchangeability | |
| Unmeasured confounding | External control data might not capture the same detailed patient information as SATs, leading to biases from unknown or unmeasured factors |
| Lack of concurrency/ temporal bias | External control data and SATs may be collected in different time periods or under varying healthcare settings |
| Measurement error | There may exist potential inconsistencies in how patient information is collected and recorded, termed here as measurement errors in covariates |
| Outcome validity | Methods of measuring outcomes in external data sources might differ from those used in SATs, or the outcomes might not be clearly defined or reliable |
| Proper estimands | |
| Intercurrent events | Intercurrent events, such as stopping medication and/or adding rescue therapy, can confound the causal effects of the randomized treatment |
| (Observed) baseline incomparability . | |
|---|---|
| Covariate distribution shift | Systematic differences exist in the baseline characteristics of patients in external studies compared to those in SATs |
| (Unobserved) outcome mean non-exchangeability | |
| Unmeasured confounding | External control data might not capture the same detailed patient information as SATs, leading to biases from unknown or unmeasured factors |
| Lack of concurrency/ temporal bias | External control data and SATs may be collected in different time periods or under varying healthcare settings |
| Measurement error | There may exist potential inconsistencies in how patient information is collected and recorded, termed here as measurement errors in covariates |
| Outcome validity | Methods of measuring outcomes in external data sources might differ from those used in SATs, or the outcomes might not be clearly defined or reliable |
| Proper estimands | |
| Intercurrent events | Intercurrent events, such as stopping medication and/or adding rescue therapy, can confound the causal effects of the randomized treatment |
| (Observed) baseline incomparability . | |
|---|---|
| Covariate distribution shift | Systematic differences exist in the baseline characteristics of patients in external studies compared to those in SATs |
| (Unobserved) outcome mean non-exchangeability | |
| Unmeasured confounding | External control data might not capture the same detailed patient information as SATs, leading to biases from unknown or unmeasured factors |
| Lack of concurrency/ temporal bias | External control data and SATs may be collected in different time periods or under varying healthcare settings |
| Measurement error | There may exist potential inconsistencies in how patient information is collected and recorded, termed here as measurement errors in covariates |
| Outcome validity | Methods of measuring outcomes in external data sources might differ from those used in SATs, or the outcomes might not be clearly defined or reliable |
| Proper estimands | |
| Intercurrent events | Intercurrent events, such as stopping medication and/or adding rescue therapy, can confound the causal effects of the randomized treatment |
This PDF is available to Subscribers Only
View Article Abstract & Purchase OptionsFor full access to this pdf, sign in to an existing account, or purchase an annual subscription.
