Confounding Factors (Cut off Exam 2) Flashcards
Confounding
- Extraneous factor that wholly or partially accounts for the observed effect
- Distortion of the association that occurs due to the effect of an extraneous factor
Confounding
Distortion of association due to a third variable
Confounder
- Variable causing the confounding
- Associated with BOTH exposure and outcome
Criteria for being a Confounder
- Risk factor for the outcome
- Associated with the exposure of interest
- Not an intermediate step between exposure and outcome
Covariate
- Potential confounder
- Select a list of these in your DB and determine whether they are confounders or not
Risk Factor
Variable positively associated with the outcome
Confounding by Indication
- Reason of prescription
- Underlying disease severity
- Worst prognosis are allocated preferentially to a particular treatment
Confounding by Co-medication
- Difficulty to isolate the effect of a specific drug
- Compliancy increasing their likelihood to outside interventions and improve their outcomes
Confounding by Dose/Drug Potency
- Analysis for a class or medications vs individual medications
- Dose-response (quantity/frequency)
- Potency
Controlling for Confounding: Study Design
- Restriction
- Matching
- Randomization: intervention studies only
Controlling for Confounding: Analysis
- Stratification
2. Multivariate Analysis
Restriction
- Restrict the study population to only one level of the confounding factor
- Subjects have the same level of a confounding factor
- Including only women or people in a certain age range for example
Restriction Advantages
- Simple
- Effective
Restriction Disadvantages
- Generalizability of results is limited
- Residual confounding if do not restrict too narrow
- Cannot evaluate the effects of factors that have been restricted for
- Reduces the number of eligible subjects (sample size problem)
Matching
- Ensure that study groups do not vary with respect to possible confounders
- Make groups artificially similar for potential confounders
- EX: match someone of the same age in each group
Matching Advantages
- Simple - when few confounders
- Efficient
Matching Disadvantages
- Very difficult if there are several confounders
- Limits sample size
- Cannot evaluate the effect of matched factors
- Required a special type of statistical analysis: McNemar, conditional logistic regression
Randomization
- Random distribution into study groups
- Confounders given “equal chance” to be in either treatment group
Randomization Advantage
- Very efficient
- Both known and unknown confounders are distributed equally
- Cannot be achieved with other techniques
Randomization Disadvantages
-Only feasible with interventional studies
Stratification
- Technique to control for confounding in the analysis of a study
- Evaluation of the association within homogeneous categories (strata) fo the confounding variable
- Form of post hoc restriction (during analysis)
- First derive statum-specific estimates and then calculate their weighted average
Stratification Advantages
- Easy to carry out
- Produces a single summary estimate that is unconfounded can be calculated
- Gives a fairly good picture of what’s going on
Stratification Disadvantages
- Inability to control simultaneously for multiple confounders
- Stratification may require a larger sample size
Multivariable Analysis
- Analytical technique that adjust for several variables simultaneously
- Efficient estimation of measures of association while controlling for a number of confounders
- Involves the construction of a mathematical model to describe most efficiently the association between exposure and disease and other factors
- Most common techniques: multiple linear regression, logistic regression, ANCOVA, Cox regression
Multivariate Analysis Advantages
- Very powerful technique
- Allows to control for multiple confounders in the same model (even when stratification would fail)
Multivariate Analysis Disadvantages
- Requires sophisticated skills in biostatistics and epidemiology
- Potential to over-adjust
