The flashcards below were created by user
on FreezingBlue Flashcards.
1. When is a study said to have internal validity? (2)
2. What is internal validity concerned with? (3)
3. What is external validity?
4. Why do we care about confounding and bias? What are they due to?
- 1. When there is (1) proper selection of study groups and (2) lack of error in measurements
- 2. Appropriate measurement of exposure, outcome, and association between exposure & disease.
3. Implies ability to generalize results beyond a set of observations to some universal statement.
4. Can make it difficult to discern causal relationship between association.
- Confounding - 3rd variable
- Bias - due to spurious observation
1. Define confounding
2. What is causally associated with what? Non-causally associated?
3. What does the confounder do?
4. What is the criteria for a confounder?
5. When does confounding occur?
1. Situation in which a noncausal association between an exposure and outcome is seen bc of the influence of a third variable (confounder)
2. Confounder is associated with outcome, confounder can be causally or non-causally associated with exposure.
3. Confounder distorts relationship between exposuer and outcome.
4. (1) be risk factor for disease (2) be associated with exposure (3) not be an intermediate step in the causal pathway between exposure and disease
5. When crude and adjusted measures of effect are not equal (difference >=10%)
1. Overall, what are the 2 overall types and 5 ways to control confounding?
1. Prevention strategies (randomization, restriction, matching) and analysis strategies (stratification, multivariate analysis)
1. Describe randomization
2. Advantages? (3)
3. Disadvantages? (2)
1. Randomization - attempts to ensure equal distributions of confounding variable in each exposure category
2. Convenient, inexpensive, permits straightforward data analysis
3. (1) need control over exposure and ability to assign subjects to study groups (2) Need large sample sizes
1. Define restriction
2. What does it provide?
3. What exactly does restriction do? How does this prevent confounding?
4. When is restriction particularly useful? (2)
5. Advantages? (3)
6. Disadvantages? (5)
1. Restriction - restricting participants' eligibility upon enrollment into study (i.e., by age, gender, race, etc)
2. Complete control of known (restricted) confounders
3. Restriction eliminates any possible association between the exposure and the confounder AND between the confounder and the disease. With restriction, there is no longer variability in the values of the confounding variable.
- 4. When confounder is quantitative in scale but difficult to measure
5. (1) conceptually straightforward (2) handles difficult to quantitate variables (difficulty in measurement) (3) can be used in analysis phase
6. (1) limits #
of eligible subjects (2) inefficient
to screen subjects then not enroll (3) residual confounding
may persist of restriction categories aren't sufficiently narrow (4) limits generalizability
(5) can't evaluate the relationship of interest at different levels of the restricted variable (e.g., can't assess interaction)
1. Define matching
2. Define the two types of matching
3. How does matching occur in cohort study? In case-control study?
4. Advantages? (2)
5. Disadvantages? (1)
1. Matching - matches subjects in study groups according to value of suspected/known confounders to ensure equal distributions
2. Frequency matching - the number of cases with particular match characteristics
Individual matching - pairing of one or more controls to each case based on similarity in sex, race, or other variables
3. Cohort study - unexposed subjects are "matched" to their exposed subjects according to their values for the potential confounder
Case-control - non-diseased controls are "matched" to diseased cases
4. Advantages (1) fewer subjects are required in unmatched studies of same hypothesis (2) may enhance validity of a follow-up study
5. Disadvantages - COSTLY, bc of extensive searching and recordkeeping are required to find matches.
1. Define stratification
3. Advantages? (3)
4. Disadvantages (4)
1. Analyses performed to evaluate the effect of an exposure within strata (levels) of the confounder
3. Advantages: (1) straightforward and logical approach (2) minimum assumptions must be satisfied in order for analysis to be appropriate (3) computational procedure is straightforward
4. Disadvantages: (1) small numbers of observations in some strata may result (2) inconsistency in ways to form strata with continuous variables (3) Difficulty in interpretation when several confounding factors must be evaluated (4) categorization produces loss of info as compared to original continuous variable
1. Define multivariate analysis
2. Advantages? (2)
3. Disadvantages? (1)
1. Use of computers to construct mathematical models that describe the influence of exposure and other factors that may be confounding the effect simultaneously
2. (1) continuous variables don't need to be converted to categorical variables (2) allow for simultaneous control of several exposure variables in single analysis
3. Disadvantages: potential for misuse
1. What can confounding in a study lead to? (2)
2. Is it an error in the study? What does it reflect?
3. What would be an error in the case of confounding?
1. Can lead us to conclude that there is (or isn't) a causal relationship when there in fact isn't (or is)
2. No, it reflects a true underlying association although the association may not be causal.
3. To not control for confounding
1. What is bias?
2. What can it be an issue in?
3. What are types of selection bias? Define
4. What are types of information bias? Define
1. Systematic error in the design, conduct, or analysis of study resulting in mistaken estimate of an exposure's effect on disease.
2. Can be an issue in virtually any epid study design
3. Selection bias - arises when relation bt exposure and disease is different for those who participate and those who theoretically would be eligible for study but do not participate (cohort, exclusion, non-response)
4. Systematic error in obtaining information regarding subjects in a study (abstracting records, interviewing, surrogate interviews, surveillance bias, recall bias, abstractor bias, lying bias, reporting/publication bias)
1. What is differential misclassification?
2. What is non-response bias? What would it result in?
3. How do you deal with non-response bias? (2)
1. When misclassification of disease as non-diseased is different in exposed and unexposed persons (information bias)
2. Non-response bias - form of selection bias in which response rate may be higher in diseased subjects who were exposed, resulting in apparent association when there is none
3. By minimizing non-resopnse and characterizing the non-responders as much as possible
1. What is exclusion bias? What type of bias is it?
2. Recall bias - what type of bias is it?
3. Misclassification bias? Differential vs. non-differential? What type of bias is it?
4. What can differential/non-differential misclassification result in?
1. If you apply different eligibility criteria to cases vs. controls. Selection bias.
2. Cases may recall info to a greater degree than the controls - DIFFERENTIAL RECALL that may artifactually suggest a relationship. Information bias.
3. Can occur with regards to disease status or exposure status.
Differential - rate of misclassification differs in two groups - i.e., misclassification of exposure in cases not controls
Non-differential - inaccuracy that is not related to exposure/disease status
4. Diff can lead to apparent association where there is none or an apparent lack of association where there is one.
Non-diff dilutes RR or OR and shifts it to 1.
How can you prevent bias? (5)
- 1. Careful attention to sampling
- 2. Minimize non-resopnse
- 3. Standardization of measurements
- 4. Training and quality control
- 5. Blinding
2. Synonyms (3)
3. Positive interaction? negative interaction? akas
4. What questions should you ask to determine if there is an interaction? (3)
1. Interaction - when the magnitude of association (bt exposure and disease) meaningfully differs bc of some third variable.
DEF #2: When incidence rate of disease in the presence of two or more risk factors differs from the incidence rate expected to result from their individual effects!!!
2. Effect modification, effect-measure modification, heterogeneity of effect
3. Synergism, antagonism
4. (1) is there association? (2) if so, is it due to confounding? (3) is the association equally strong in strata formed on the basis of the 3rd variable? IF YES --> NO INTERACTION. IF NO --> THERE IS AN INTERACTION PRESENT
Create an additive and multiplicative model for incidence rates and relative risks for the following example:
12.0+9.0 = 21.0 (instead of 15.0 + 9.0 which would be counting background risk of 3.0 twice)!
For multiplicative model, since factor A triples baseline risk by 3 (3.0 x 3 = 9.0), we sould expect factor A to triple the risk of 15.0 observed when exposure to factor B is present.
Then for relative risk, there is a 3fold risk for factor A, 5fold risk for factor B, leading to a total of 15.
1. In looking at a graph for risk of disease vs. unexposed/exposed, how can you tell if there is a multiplicative interaction?
2. What is the difference between confounding and interaction?
1. If the lines for third variable are non-parallel
2. YOu may discover interaction by performing stratification as a means to get rid of confounding
- Confounding = extraneous/nusicance pathway that an investigator hopes to rule out
- Interaction - more detailed description of relationship between exposure & disease; richer description of biologic/behavioral system under study (a finding to be reported, not a bias to be eliminated!)