LO: Explain how chance, bias, confounding and random error can influence research findings and describe approaches used in study design and analysis to minimise these threats to validity

Bias

• Systematic deviation from the underlying true effect of an intervention on an outcome of interest in a given population
• Caused by a flaw in the study design/conduct/analysis - leads to incorrect estimate

Selection bias

• Subjects studied are not representative of the target population
• Relationship between exposure and outcome differs from those who participate and those who do not
• Mitigation - at design, draw controls from a population as close as possible to the cases
• Attrition bias:
  • Type of selection bias
  • Participants drop-out, those that leave have systemic differences from those who remain
  • Issue if drop-out is related to outcome/exposure
  • Mitigation - maximise patient follow-up, aiming for consistency across outcome/exposure

Information bias

• Distortion of results due to errors in measurement or classification of exposures/outcomes
• Recall bias:
  • Type of information bias
  • Differences in the accuracy/completeness of participants' recollections of past events or experiences
  • Issue if differences relate to outcome/exposure
  • Common in case-control
  • Mitigation - minimise recall periods, use objective measures
• Detection bias:
  • Systematic differences between groups in how outcomes are determined
  • Type of information bias
  • Mitigations - blinding/masking, standardised protocols, objective measures

Confounding

• Third variable that is associated with both the exposure and outcome, distorting the true relationship between them
• Mitigation - restriction of participants, 'matching' (match subjects on confounder)