finals: epi

Question 1

3 major sources of bias

Answer 1

1) confounding
2) selection bias
3) measurement error/misclassification (information bias)

Question 2

why DAGs useful (2 reasons)

Answer 2

• easy way to distinguish between confounding and selection bias
• Tool to identify which variables should
  be controlled for to block spurious
  associations between exposure and
  outcome

Question 3

4 stappen van DAG

Answer 3

C(onfounder) -> E(xposure) -> D(isease) -> (S)election into case control study

Question 4

Selection bias in DAG

Answer 4

conditioning on common effect (Aka Collider)

Question 5

Causes of bias: RCT

Question 6

Causes of bias: Cohort studies

Answer 6

[c] yes
[s-b] when selection of exposed and unexposed related to disease status. when LTFU different exposed and unexposed
[mc/me]: yes both diff and undiff

Question 7

4 ways to deal with confoudning in analysis

Answer 7

regression
stratification
propensity score matching
IV analysis

Question 8

Causes of bias: Case-control

Answer 8

[c] yes (un)measured covariates
[s-b] controls not selected independent of disease status
[mc/me]: ja. vooral recall bias.

Question 9

dealing with confoudning incase control:

Answer 9

design: restriction, matching (on disease status)
analysis: propensity scores, regression, stratifcication
NB: matching prevents you from assessing relation between match factors and outcome, doesnt remove bias

Question 10

before starting study: 3 ethical prinicples

Answer 10

beneficiency - minimize harm
justice - avoid exploitatins, minimzie risk
respect for persons - protect dignity and autonomy

Question 11

3 ways selection bias arises

Answer 11

1. The source of controls is not representative of the study base with respect to exposure
2. Subject selection
3. Losses of selected subjects

Question 12

Berksons bias?

Answer 12

Berkson’sbias may occur when hospital controls are used in a case-control study and hospitalization is related to the exposure under study
Downwards bias if controls are hospitalized due to the exposure under study

Question 13

3 eisen case crossover design

Answer 13

• crossover
• transcience of effect
• acute cases

Question 14

strenght case crossover design

potential for bias?

Answer 14

pro:Cost efficient (non cases drop out)
Bias (self matching)

con: confouding by time varying factors, recall bias (het event veroorzaakt anders herinneren