Bias and Confounding

Bias

  • Occurs when there is a systematic difference between the results obtained from a study and the true population parameter – it can over- or under- estimate the true effect and thus create an incorrect association or conceal a real one
  • Bias can occur at any stage of the research process from planning to publication
  • Types of bias:
    • Selection bias
    • Information bias
    • Confounding
  • Selection bias
    • This arises when study participants are not representative of the population to which the results are going to be applied – it occurs due to lack of randomising selection
    • The best way of eliminating selection bias is proper randomisation
    • Ascertainment bias – Occurs when sample used is not randomly selected & differs in certain respects to the population
    • Attrition bias – Occurs when there is an unequal loss of participants from different groups in a longitudinal study – for example, when those who are lost to follow up differ in certain respects to those who are not lost to follow up
    • Healthy entrant effect – Occurs when mortality/morbidity rates are lower in initial stages of a longitudinal study in comparison to the population, if healthier people were included in the study
    • Response bias (or volunteer bias) – Occurs when participants who volunteer in the study differ in certain respects to the general population who did not volunteer to be in a study
    • Survivorship bias – Occurs when survival is compared in patients who do or do not receive a particular intervention where it only became available at some point after the start of the study 
  • Information bias
    • Occurs during data collection when measurements are incorrectly recorded in a systematic manner
    • Central tendency bias – occurs when responders tend to move towards the midpoint of a scale, for example when participating in a survey using a likert scale 
I like using a 5 point likert scale
Strongly disagree
Disagree
Neutral
Agree
Strongly agree
  • Lead-time bias – lead time is the period between early detection of a disease via screening and the time of  usual clinical detection/presentation. Lead time bias occurs when early detection of a disease via screening, incorrectly gives a overestimation of survival duration & makes it look like people with that condition are living longer.
  • Length bias – overestimation of survival duration among screen detected cases due to relative excess of slowly progressing cases detected
  • Measurement bias – occurs when error is introduced by inaccurate measurement
  • Observer bias – occurs when a researcher under or over reports a variable
  • Reporting bias – occurs when participants give answers which they perceive are of interest to the researcher. People can often behave or answer differently if they know they are being observed.
  • Publication bias – Occurs when only some trials are published and others are not, so conclusions are drawn only from those published. Negative studies tend to be harder to get published, so that evidence may be missing which falsely suggests a benefit
  • Recall bias – Occurs when participants answer a survey and are more likely to remember some events than others. This can an issue, as following a life changing diagnosis, patients are more likely to remember what they perceive as ‘toxic’ events and forget about other experiences that may have contributed to the diagnosis.
  • Confounding 
    • Confounding is when one of more variables may be related to the outcome and so it is difficult to assess the independent effect of each one on the outcome.
    • Confounding variable/confounder – another variable, whose presence, also has an effect on the dependant variable being studied
    • Example – A study investigates the effect of alcohol intake (the independent variable) on mortality (the dependant variable) and finds that higher alcohol intake is associated with higher mortality. However, alcohol intake and mortality may also be effected by other factors for example age, education, smoking and diet (the confounding variables). Patients who have a healthy diet may be less likely to drink alcohol and may have higher education which may influence mortality as well.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s