Assessing Risk of Bias
- Bias: systematic error leading to underestimation or overestimation of the true intervention effect. Not to be confused with imprecision as bias is often a systematic error in design, meaning multiple replications of the same study would reach the same ‘wrong’ conclusion.
- Selection bias: systematic differences between baseline characteristics of the groups that are compared (prevent this with randomization and allocation concealment).
- Performance bias: systematic differences between groups in care provided or exposure to factors other than interventions of interest (prevent this by blinding or masking participants and personnel).
- Detection bias: differences between groups in how outcomes are determined (prevent this by blinding or masking of outcome assessors).
- Attrition bias: differences between groups in withdrawals from the study (prevent this by minimizing lost to follow up, and analyzing all available data without exclusions).
- Reporting bias: differences between reported and unreported findings (prevent by reporting all outcomes of interest even if not statistically significant or positive).