II. Background
- Structured approach to healthcare recommendations established in 2000
III. Criteria: Quality of Evidence (Certainty)
- Certainty and confidence
- High
- Authors have high confidence that the true effect is similar to the estimated effect
- Moderate
- Authors have moderate confidence that the true effect is similar to the estimated effect
- Low
- True effect may be significantly different than the estimated effect
- Very Low
- True effect is likely significantly different than the estimated effect
- High
- Factors that decrease certainty
- Risk of Study Bias
- Imprecision (in relation to the 95% Confidence Interval)
- Inconsistency (e.g. overlapping Confidence Intervals, heterogeneity)
- Indirectness (e.g. patient population, environment and other factors differ from that of the applied recommendations)
- Publication Bias (e.g. missing evidence, e.g. in observational studies, industry funded studies)
- Factors that increase certainty
- Large magnitude of effect
- Consistent dose response gradient
- Confounders are likely to increase effect magnitude
IV. Criteria: Strength of Recommendation
- Problem
- Importance and frequency of the healthcare condition
- Values and Preferences
- How important are the health outcomes to the affected population?
- Quality of Evidence
- Quality of evidence (see above)
- Benefits and Harms
- Degree to which the net benefit exceeds the net harm
- Resource Implications
- Cost effectiveness and its associated incremental benefit
- Equity
- Does the recommendation reduce health disparities?
- Acceptability
- Is the recommendation acceptable to most stakeholders (esp. the target population)
- Feasibility
- Is the implementation practical given resources and acceptability?
V. Resources
- GRADE Approach (Wikipedia)
- What is GRADE? (BMJ)