Skip to main content

Table 3 Criteria for appraising risk for bias in descriptive studies

From: Diabetes in pregnancy among indigenous women in Australia, Canada, New Zealand, and the United States: a method for systematic review of studies with different designs

Type of risk Low Moderate High
Selection (sampling of population) Consecutive unselected population
Sample from general population not a selected group
Rationale for case and control selection explained
Follow-up or assessment time appropriate
Sample selected from large population, but selection criteria not defined
Sample selection ambiguous, but may be representative
Eligibility criteria not described
Rationale for case and control selection not described
Follow-up assessment not appropriate
Analysis to adjust for sampling strategy bias
Sample selection ambiguous, and sample unlikely to be representative
Highly selected population, making it difficult to generalise findings
Selection (sampling size) if non-significant results; otherwise NA Sample size calculation conducted and adequate Sample size calculation not performed, but all eligible persons studied
Sample size calculation performed, and reasons for not meeting sample size given
Sample size estimation unclear or only subsample studied
Selection (participation rate) High participation rate (> 85%) Moderate participation rate (70-85%) Low participation rate (> 70%)
Performance bias (outcome assessment) Diagnosis on basis of consistent criteria and direct examination Assessment from hospital record or by questioning person
Assessment from administrative database or register
Assessment from non-validated data or generic estimate from overall population
Performance bias (confounding factors) Control for common confounders (e.g. risk factors) Only certain confounders adjusted (e.g. age) Not controlled for any confounders
Performance bias (blinding) when outcome and exposure relations being assessed (e.g. effect of GDM on offspring) Outcome and exposure assessment independently blinded Outcome and exposure assessment independently blinded  
Performance bias (methods to control for bias) Analysis appropriate for type of sample (e.g. subgroup analysis/regression etc.) Analysis did not account for common adjustment Data confusing
Attrition bias 0-10% attrition
All participants from initiation to final outcome assessment accounted for
Sensitivity analysis conducted for missing data
11-20% attrition
No sensitivity analysis for missing data
> 20% attrition and no sensitivity analysis for missing data
Other: for comparative studies Comparison groups similar at baseline Baseline characteristics of comparison groups not assessed Baseline characteristics dissimilar