Table 3.
Factors that lead to increasing evidence quality in the GRADE framework
| FACTORS | EXPLANATION | EXAMPLES |
|---|---|---|
| Large magnitude of effect and narrow CIs | When an intervention has a large effect on an outcome | Large effects are considered to be a risk ratio of at least 2 or a relative risk reduction of at least 50%. Having infants sleep on their backs has been associated with a 50% to 70% reduction in the risk of sudden infant death syndrome28 |
| Dose-response gradient | The presence of a dose-response relationship between intervention and outcome | When 50% of the population is immunized, it results in a 20% lower risk of disease; when 70% of the population is immunized, it results in a 40% lower risk; and when 90% of population is immunized, the risk is lowered by 80%29 |
| All plausible confounding | When inclusion of all plausible confounders or biases in observational studies that are unaccounted for in the analysis would result in an underestimate of an apparent treatment effect, or would increase the effect when no effect was found | Although an early study showed a positive association between vaccines and autism, further studies did not support this association, despite the fact that parents with autistic children might have been more likely to recall and report their vaccination experience than those whose children did not have autism28 |