Confounding

Question 1

How to control confounding (Design)

Answer 1

Restriction: Restrict study to one confounder
- Limits generalizability
- May lose power to detect true difference

Matching: Match study groups based on confounders
- Overmatching can hide association
- Can not determine relationship between confounders and outcomes

Propensity Score Matching

Question 2

How to control confounding (Analysis)

Answer 2

Stratification:
- Analyze association between exposure and outcomes at different levels of the confounding variable
- Loses power to detect a true difference

Multivariate Analysis:
- Mathematical model that simultaneously controls the effects of multiple variables

Question 3

Kinds of Multivariate Analysis

Answer 3

Logistic Regression Model

Cox Proportional Hazard Model

Poisson Regression Model
Linear Regression Model

Question 4

Logistic Regression Model

Answer 4

Most Common

Dependent Variable: Dichotomous

Independent Variable: Dichotomous, Categorical, Continuous

Question 5

Cox Proportional Hazard Model

Answer 5

Dependent: Combination of time and if the outcome has occured or not

Independent: Dichotomous, Categorial, Continuous

Question 6

Confounding Variables

Answer 6

A third variable related to both exposure and outcome
- Is not related to casual pathway
- Distorts the association between exposure and outcome

Question 7

Effect Modification

Answer 7

A third variable that affects the magnitude by which exposure affects the outcome