GRADE Framework

Grading of Recommendations, Assessment, Development and Evaluation (GRADE) provides a systematic approach to rating evidence certainty.

What is GRADE?

GRADE assesses:

Certainty in evidence (confidence in estimates)
Strength of recommendations
Quality across outcomes

Certainty Levels

High Certainty

Very confident in estimate
Further research unlikely to change
Strong basis for recommendations

Moderate Certainty

Moderately confident
Further research may change estimate
Generally reliable for recommendations

Low Certainty

Limited confidence
Further research likely to change
Recommendations made cautiously

Very Low Certainty

Very little confidence
Estimate very uncertain
Recommendations uncertain

Factors Affecting Certainty

Downgrading Factors

1. Risk of Bias

Study quality issues
Domain-specific concerns
Overall bias assessment

2. Inconsistency

Heterogeneity in results
Unexplained variation
Conflicting findings

3. Indirectness

Population differences
Intervention variations
Outcome substitutions
Indirect comparisons

4. Imprecision

Wide confidence intervals
Small sample sizes
Crossing decision thresholds

5. Publication Bias

Funnel plot asymmetry
Missing studies
Selective reporting

Upgrading Factors

Large Effect

Very large magnitude
Dose-response relationship

Plausible Confounding

Would reduce effect (but effect remains)
Strengthens conclusions

EvidAI GRADE Features

Automated Assessment

The AI evaluates:

Risk of bias summary
Statistical heterogeneity
Indirectness indicators
Precision measures
Publication bias tests

Suggested Ratings

For each outcome:

Preliminary certainty rating
Factors for downgrading/upgrading
Supporting evidence
Your final judgment

Evidence Profiles

Automatic generation of:

Summary of Findings tables
Evidence profiles
GRADE summary tables

Summary of Findings Tables

Auto-Generated Content

Tables include:

Outcome descriptions
Relative effects (RR, OR)
Absolute effects
Number of participants
Number of studies
Certainty rating
Comments

Customization

Adjust:

Outcomes included
Effect measures displayed
Baseline risk assumptions
Plain language summaries

Best Practices

During Assessment

Consider all domains systematically
Document reasoning
Be consistent across outcomes
Engage clinical expertise

Reporting

Be transparent about judgments
Explain uncertainty clearly
Provide plain language summaries
Include all domains assessed