Bias and Measurement Error

Question 1

Types of error

Answer 1

• Due to chance
  o Random error
  o Minimised by increased sample size and tighter confidence intervals
• Due to bias
  o Systematic error
  o Minimised at design phase

Question 2

Random error pushes findings towards

Answer 2

H0 - decreases chance of observing effect

Question 3

Random error can be from

Answer 3

o Poor instruments or tests
o Subject error
o Data-entry error

Question 4

What is bias

Answer 4

Systematic deviation from the truth

Question 5

Does bias increase or decrease effect

Answer 5

Can do either

Question 6

2 main types of bias

Answer 6

Selection - who is in study
Information - information from subjects about exposure/outcome

Question 7

Sources of selection bias

Answer 7

• Volunteer – more likely to volunteer answers for study if posses desired outcome
• Referral – such as in hospital, chance of exposed cases admitted to study is different from controls
• Healthy worker effect – usually occupational exposure, as unwell people not at work so do not take part
• Non-participation bias – non response or loss to follow up

Question 8

Sources of information bias

Answer 8

• Observer – questions one group to greater detail than other
• Surveillance – one group more likely to have outcome detected due to increased surveillance
• Attention – gives you answers you want to hear
• Recall – forgets correct information
• Misclassification

Question 9

How to minimise bias

Answer 9

• Aware of potential sources
• Equal treatment all groups
• Standardised questionnaires
• Blinding where possible
• Minimise recall period
• Objective measures where possible

Question 10

Bias in case control studies

Answer 10

• Selection due to separate sampling of cases and controls
  o Take all cases
  o Controls sampled from same population as cases and similar in all respects other than disease
• Information due to retrospective measurement of exposure variables
  o Especially recall bias -> minimise recall period
  o Objective measurement of exposure
  o Consider doing prospective study

Question 11

Bias in cohort study

Answer 11

• Attrition bias (loss to follow up), kind of selection bias
  o Treat both groups the same in regard to reminding to respond
  o If identify bias examine the non-responders by age/sex/occupation etc

Question 12

Types of validity

Answer 12

• Internal
  o Extent to which findings are true within the population studied
• External validity
  o Findings are generalisable to other populations

Question 13

How does bias compromise validity

Answer 13

• Selection
  o Both internal and external validity
• Information
  o Internal validity