Bias
On this page
Bias refers to systematic errors that can distort the findings of a study, reducing its validity and leading to incorrect conclusions. Unlike random error, bias consistently pushes results in a specific direction. Bias can occur at any stage of a study, from design and data collection to analysis and reporting.
Types of bias
| Type | description |
|---|---|
| Selection bias | Occurs when the method of selecting participants leads to differences between groups that affect the outcome. Example: If healthier individuals are more likely to be included in one group, the results may favour that group unfairly. |
| Performance bias | Happens when participants or investigators behave differently due to knowledge of group assignment. Example: Participants in the control group may seek additional treatments if they know they are not receiving the intervention. Prevented or reduced by: Double blinding of participants and staff. |
| Detection bias (observer bias) | Occurs when the way outcomes are assessed is influenced by the knowledge of group allocation. Example: A researcher may interpret symptoms more favourably in the treatment group. Prevented by: Blinding outcome assessors. |
| Measurement bias (systematic error) | Arises when tools or methods used to collect data consistently lead to inaccurate measurements. Example: A poorly calibrated blood pressure monitor always giving low readings. Sometimes referred to as systematic error, in contrast to random error. |
| Attrition bias | Results from systematic differences in withdrawals or loss to follow-up between groups. Example: If more participants drop out of the treatment group due to side effects, the group may appear more effective than it truly is. Reduced by: Complete follow-up and intention-to-treat analysis. |
| Reporting bias | Occurs when only certain results (usually favourable or significant ones) are selectively reported. Example: A study may publish only positive outcomes while ignoring adverse events. Addressed by: Study registration and adherence to reporting guidelines (e.g., CONSORT). |
Reducing Bias
- Randomisation: Reduces selection bias by ensuring equal probability of group assignment.
- Blinding (Single or Double): Reduces performance and detection bias.
- Standardised Procedures: Reduces measurement bias.
- Complete Follow-up: Minimises attrition bias.
- Pre-registration of protocols: Helps prevent reporting bias.
Key point
Systematic reviews and well-conducted randomised controlled trials (RCTs) aim to reduce, but cannot entirely eliminate bias. Understanding and mitigating different types of bias is crucial for producing reliable medical evidence.
Systematic reviews and well-conducted randomised controlled trials (RCTs) aim to reduce, but cannot entirely eliminate bias. Understanding and mitigating different types of bias is crucial for producing reliable medical evidence.