10 - 11 Random Error, Bias - Confounding

Question 1

What is accuracy (epidemiological term)

Answer 1

Accuracy = Precision and Validity

Accuracy -> absence of errors
Precision -> absence of random error
Validity -> absence of systemic error

Question 2

What is bias?

Answer 2

Bias is any systemic error in the design or analysis of a study that leads to mistaken results

Question 3

What is selection bias?

Answer 3

Bias that results from the selection process of the study population from the population of interest

Question 4

What is response bias?

Answer 4

Bias by systemic non-participation

example: low participation of unexposed healthy persons -> underestimation (W<1) or high participation of exposed diseased persons (W>1) both can happen in the same study

Question 5

What is W for?

Answer 5

W is a correction factor to account for over (W>1) or under estimation (W<1)

Question 6

What is admission rate bias?

Answer 6

• chance of exposed cases is different to exposed controls
  -> exposure and disease are both factors to go to the hospital
• typical for case-control studies

examples: lung cancer and asbestos (exposure in cases to high), lung cancer and smoking (exposure in controls to high; no cancer but other disease caused from smoking)

Question 7

What is migration bias?

Answer 7

Bias due to systemic migration between comparison groups

example: moving from are with high air pollution to low pollution because of respiratory diseases;
Healthy worker bias: change work to avoid exposure that causes problems

Question 8

Name reasons for information bias

Answer 8

• participant (can‘t articulate, recall, central tendency on questionaire, intentional misinformation)
• data collector (unclear questions, result expectation, lack of neutrality, inaccurate transcription)
• data managers (inaccurate transcription, misreading, miscueing, programming errors
• data analyst (coding or programming errors, inappropriate statistical method)
• data interpreter

Question 9

(Differential) Misclassification

Answer 9

• can cause drastic changes in RR

rate of misclassification differs:
- measurement of exposure depends on disease status or
- measurement of disease depends on exposure status
-> tendency to put disease and unexposed into disease and exposed

Question 10

What is recall bias?

Answer 10

important information is likely better recalled by a „case“ then by a „control“

Question 11

What is interviewer bias?

Answer 11

• different questions/interviews for cases and controls
• differential misclassification of exposure status

Question 12

What is detection bias?

Answer 12

• better diagnosis in exposed people
• differential misclassification of disease status

Question 13

What is diagnostic suspicion bias?

Answer 13

• overestimation of the exposure effect leads to more thorough examination of exposed

Question 14

What is confounding?

Answer 14

• a confounded causes spurious associations between exposure and outcome
• confounder has effect on disease
• confounder has effect on exposure
• confounder is not only a result of exposure
• confounder is not only a step between exposure and disease
• common confounders: age, gender, socioeconomic status, smoking

Question 15

How to avoid confounding?

Answer 15

Stratification:
- data analysis is repeated for each subgroup (example: age)

Regression analysis:
- adjustment for several covariants simultaneously
-> control confounders in the analysis to get an estimate of the net exposure effect

Question 16

Main difference between bias and confounding?

Answer 16

Bias: data is biased/incorrect
Confounding: data is correct, but interpretation is incorrect