confounding?
- systematic difference between the groups that distort the true association between exposure and disease (want to control it)
- bias toward or away from the null
what are the sources for confounding?
- Experimental and cohort studies occur when the exposed and unexposed groups differ by more than just exposure—they differ on some other variable.
- Case-control study: occurs when cases and controls have different characteristics
- This can occur in all types of epidemiological studies
- Unlike bias, it is an inherent characteristic of the population
what effect can confounding have ?
- This results in a distortion of the true association between E and D and can bias either towards or away from the null.
- Can be adjusted or fixed to a point
what can confounding be thought of as?
- mixing of effects
- The estimate of the effect of exposure on disease is distorted because it is mixed with the effect of other factors associated with E and D.
- Sometimes referred to as the third variable problem
counterfactual ideal?
- comparison group would be the exact same people who are exposed in the group had they not been exposed
how do epidemiologist select for people that are similar as possible with respect to?
- With respect to other factors that could influence the outcome
- With respect to the collection of comparable and accurate information
How are confounder and counterfactual ideals related?
- Confounding can be thought of as a failure to come close to the counterfactual ideal.
- he difference in disease risk between the exposed and unexposed is because of factors other than just the exposure.
When is a variable a confounder? Three criteria
- Independent predictor of the outcome - The confounder is a risk factor for disease among unexposed people
- Associated with the exposure
- The confounder occurs more or less often among the exposed than unexposed - It cannot be an intermediate on the causal pathway between exposure and disease
- The confounder cannot be caused by exposure
How do I know what might be a potential confounder for my research question?
- Know your subject area
- Complete a comprehensive literature review and read - Previous research will help you identify known confounders
- Historical confounders - Some variables are always considered potential confounders (age, sex, race)
mediator?
a variable that is a step in the causal chain
- is in between exposure and disease
what is in the design and analysis phase?
- design: randomization, restriction, matching
- analysis: standardization, stratified. multivariate
What is randomization?
Randomly allocates study subjects to treatment groups so each subject has an equal chance of being assigned to the treatment or comparison group.
when does randomization work?
- study is large enough
- The investigator does not influence the treatment assignment
What kind of characteristics should be similar in randomization?
- baseline numbers should be similar, and the only different numbers should be what is the treatment
what are the strengths and limitations of randomization?
- There is no limit on the number of confounders that can be controlled for
- Do not need information about unknown confounders
-You do not need to know what they are or measure them
limitation
- Limited to experimental studies
- Less efficient with a smaller sample size
what is restriction
Limits the study to people who are within one category of the confounder
- sex is a confounder then only men or women in study
Strengths and limitations of restriction?
- Simple, convenient, low-expense, and effectiveness
- conceptually and practical
- Effective control of characteristics being restricted
limitation - Only possible for known, measured confounders
- Incomplete control for confounding (residual confounding) if the restriction is not narrow enough
- Cannot evaluate restricted variables
- Limits sample size
- Limits generalizability of results
what is matching
- selects study subjects so that confounders are distributed identically among exposure and unexposed (cohort study) or cases and controls (case-control study)
- Used more often in case-control studies than in cohort studies
two types of matching
- Individual matching - identifies individual subjects for comparison that resemble a matched subject (Cases: 65-year-old male with lung cancer/ Control: 65-year-old male withouy lung cancer )
- Frequency matching - balances the proportion of people with a confounding factor in the compared group (20% were 40-49 (exposed) makes sure that the unexposed who were 40-49 also make up 20% )
What are the strengths and limitations of matching?
- Simple and effective control of characteristics being matched
- Good for when case series are very small
- Useful for variables that are complex or difficult to capture (ex, neighborhood or occupation )
limitation
- Only possible for known measured confounders
- It can be difficult, expensive, and time-consuming to find appropriate responses.
- Cannot evaluate matched variables
- It can overmatch on a confounder - which is problematic because it makes the matched analysis inefficient.
what is stratification?
- stratify (separate) your study population into subgroups where one group has the confounder characteristic and one group does not, then calculate a measure of association for each subgroup
strengths and limitations of startifcation?
- Straightforward and easy to perform
- Effective control of characteristics being separated
limitation - It is difficult to control for many confounders simultaneously because of sparse data problems.
- Difficult presentations, especially if many confounders
- Continuous variables not easily stratified
what are the five steps startfication?
- Calculate crude measures of association
- Divide subjects into strata of the confounder
- Calculate stratum-specific measures of association
- Calculate adjusted measures of association
- Determine whether the crude measure of association and by how much (magnitude)
how can you determine if a confounding has occurred?
- ompare a crude (unadjusted) measure of association to a measure that has been adjusted for confounding
