PPH week 10: data interpretation: bias

• What are the two main types of errors in epidemiologic research?

  Random errors and systematic errors (bias)
• What do random errors reflect in epidemiologic research?

  Fluctuations around a true value of a parameter due to sampling variability
• What is bias in the context of epidemiologic research?

  Systematic deviation of results or inferences from the truth
• What can lead to systematic errors in a study?

  Any error in the design, conduct, or analysis of a study
• What are the two broad categories of bias in epidemiologic research?

  • Selection bias
  • Information bias
• What is selection bias?

  Systematic differences in characteristics between those who are selected and those who are not
• What are common types of selection bias?

  • Prevalence/Incidence bias
  • Non-respondent or volunteer bias
  • Self-selection bias or membership bias
  • Loss to follow-up or withdrawal bias
  • Berkson’s bias
• What is prevalence/incidence bias relevant to?

  Case-control study design
• What happens when prevalent cases are used instead of incident cases?

  Prevalent cases over-represent long-duration cases and under-represent short-duration cases
• How does selective mortality affect case recruitment?

  Exposed cases with shorter survival times are less likely to be available for recruitment
• What is the effect of selective survival on the observed association?

  It results in an overestimation of the odds ratio (OR) or relative risk (RR)
• How can prevalence/incidence bias be avoided?


  By using incident cases
• What is non-respondent or volunteer bias?

  Differences between those who agree to participate and those who do not
• How can non-respondent or volunteer bias be minimized?

  By achieving high response rates
• What is the observed association among participants in a study?

  It refers to the relationship between cases and controls based on their exposure status
• What does loss to follow-up/withdrawal bias result from?

  Different rates of outcome among those who are followed compared to those who are lost
• How can loss to follow-up/withdrawal bias be avoided?

  By minimizing loss to follow-up
• What is Berkson’s bias?

  When hospital cases and/or controls differ systematically from the general population
• How can Berkson’s bias be assessed?

  Using multiple control groups
• What is self-selection or membership bias?

  Characteristics that affect an individual's membership in a certain group
• How can self-selection or membership bias be eliminated?

  By using randomized experimental design
• What is information bias?

  Inadequate means for obtaining information leading to incorrect data collection
• What are common types of information bias?

  • Recall Bias
  • Family Information Bias
  • Interviewer Bias or Abstractor Bias
  • Reporting Bias or Prevarication Bias
• What is recall bias?

  Differences in accuracy or completeness of recall of exposure between cases and controls
• How can recall bias be minimized?

  By obtaining exposure history from independent sources
• What is family information bias?

  When cases are more aware of family history due to disease occurrence
• How can family information bias be avoided?

  By validating disease and exposure status of family members
• What is interviewer bias?

  When interviewers probe more thoroughly for exposure in cases than in controls
• How can interviewer bias be avoided?

  By blinding interviewers to case/control status
• What is reporting bias?

  When a subject is reluctant to report an exposure due to attitudes or beliefs
• What can result from reporting bias?

  A bias may result if underreporting is more frequent among cases or controls