Validity, Bias, Confounding Flashcards
External validity
Meaning of the study to the intended population
- are results of study applicable to the real population?
- are subjects comparable to the population?
- is environment/exposure comparable or achievable in nature?
- do findings relate to animals in the larger population?
Internal validity
Correct assessment of exposure and outcome in study groups
- concerns the formation of the comparison groups and measurement of exposure and disease
______ validity requires ____ validity, but not vice versa
External; internal
- ex: lab studies may have high internal validity, but limited external validity
Bias
Systematic error that results in an incorrect estimate in the relationship between exposure and disease
What is an example of observation bias?
- scale that always reads 15 lbs underweight
- doing a careful physical exam on exposed, and an abbreviated exam on nonexposed
- in assessing response to 1 of 2 abx, you know which animals received your preferred abx
What is an example of selection bias?
Assigning the sickest animals to receive the new abx and the less sick your favorite abx
Bias renders the results ___
Worthless!
Selection bias
Concerns the animals who participate and those that do not
- if the study involves treatment, who receives what treatment
- do the animals reflect the target population -> external validity
Information bias
Assessment of exposures and outcomes
- how is it measured
Comparisons must be drawn from the same population
Selection bias
- loss to follow up
Any bias that relates to data collection
Information bias
- observer bias
- recall bias
- misclassification
Differences in way information collected from E+ vs E-, or D+/-
Observer bias
- observers are different between groups
- observers have knowledge of group, treatment, or prior findings
- observers rely on self training or experience
- observers have a conflict of interest
Recall bias
Those D+ may tend to recall exposures in greater detail than those D-
Misclassification
When either D or E is not properly classified
- disease is diagnosed by producers, not veterinarians
Non-differential misclassification bias
Misclassification affects both D+ and D- to the same relative degree
- biases toward OR=1 (no effect)
When does differential misclassification happen?
When misclassification errors are uneven between outcome groups
- unequal misclassification could produce an exaggerated RR (further from 1)
Confounding
Mixing of effects
Confounding variable
Extraneous variable that can wholly or partly account for an apparent association between an exposure and outcome
- produces spurious association or masks real association
- correlated with explanatory and response variables
Confounders are related to both _____ and _____
Exposure and the outcome
- mixing up of things effecting outcome cannot be separated and results cannot be interpreted
Restriction
Restrict study subjects to avoid known confounding factors
- vaccination is prevention of disease –> restrict to cases occurring after immunity could be operational
- get rid of animals that are at extremely high risk for the outcome, but not necessarily anything to do with exposure
Matching
Match pairs on a confounding variable so that confounding is distributed evenly between comparands
Analysis
- stratification
- multivariable techniques
A large difference in crude RR and strata specific RR indicates ____
Confounding