research-grade

GRADE Evidence Assessment

Rate evidence quality using the GRADE (Grading of Recommendations, Assessment, Development, and Evaluation) methodology.

GRADE Levels

Grade A — High Quality

• Consistent findings from multiple high-quality sources
• Large effect sizes (d > 0.8) with narrow CIs
• Direct applicability to research question
• No major methodological limitations
• Further research unlikely to change confidence

Grade B — Moderate Quality

• Consistent findings from adequate sources
• Meaningful effect sizes (d 0.5-0.8) with reasonable CIs
• Mostly applicable to research question
• Minor methodological limitations
• Further research may change the estimate

Grade C — Low Quality

• Inconsistent findings or limited sources
• Small or uncertain effect sizes (d 0.2-0.5) or wide CIs
• Indirect applicability
• Notable methodological limitations
• Further research likely to change the estimate

Grade D — Very Low Quality

• Single source or expert opinion only
• Negligible effect (d < 0.2) or no quantifiable effect
• Largely indirect or speculative
• Major methodological concerns
• Any estimate is very uncertain

Upgrade Factors (+1 level)

• Large effect magnitude (d > 0.8 with narrow CI)
• Dose-response gradient
• All plausible confounders would reduce the observed effect

Downgrade Factors (-1 level)

• Risk of bias in study design
• Inconsistency across sources (I^2 > 50%)
• Indirectness of evidence
• Imprecision (CI crosses null/zero)
• Publication bias suspected

Quality Criteria

• Every claim has a grade with rationale
• Upgrade/downgrade factors are explicitly documented
• Bayesian posteriors are updated based on evidence grades
• Grade distribution is reported