Bias

Non-random deviation from the truth

systematic error; so it can reduces accuracy (validity) of the test;

can occur when

• recruiting participants
• performing study
• interpreting results

Intention-to-treat vs As-treated Analyses

When the treatment regimen selected for a patient depends on the severity of the patient's condition, a form of selection bias known as susceptibility bias (confounding by indication) can result. To avoid selection bias in studies, patients are randomly assigned to treatments to minimize potential confounding variables.

Many studies also perform an intention-to-treat (ITT) analysis to deal with selection bias. An ITT analysis compares the initial randomized treatment groups (the original intention) regardless of the eventual treatment to avoid counting crossover patients.

Conversely, as-treated analysis compares the groups based on the actual treatment received. An as-treated analysis is performed to gauge the effectiveness of the treatment itself, with less regard for potential confounders.

Selection Biases

Inappropriate selection or poor retention of study subjects

• Ascertainment (sampling) bias - study population differs from target population due to nonrandom selection methods

• Healthy worker effect: The working population is healthier on average than the general population and often exhibit lower mortality rates → Any sample consisting of only working individuals does not represent the general population.

• Volunteer bias: Individuals who volunteer to participate in a study have different characteristics than the general population. (healthier people more likely to volunteer for a study)

• Participation bias

  • individuals who choose to participate may differ from those who do not (e.g., more health-conscious)

• Recruitment bias

  • Strategies in recruitment that mean individuals with certain characteristics are more likely to take part (e.g., only advertising in gyms)

  Nonrespondent bias: High nonresponse rate to surveys/questionnaires can cause errors if nonresponders differ in some way from responders

  Berkson bias: Disease studied using only hospital-based patients may lead to results not applicable to target population

  Prevalence (Neyman) bias: Exposures that happen long before disease assessment can cause study to miss diseased patients that die early or recover

  Attrition bias: Significant loss of study participants may cause bias if those lost to follow-up differ significantly from remaining subjects