TABLE 2.

Principles for the Use of GRADE for Assessment of Evidence of the Risks of VTE and Bleeding Requiring Reintervention After Surgery

Domain	Criteria for Judgment in Our Study
RoB	We always rated down for RoB if most patients (>50%) came from studies at high RoB. We did not rate down for RoB if most patients (>50%) came from studies at low or very low RoB
Inconsistency	We rated down for inconsistency if more than 10% of the studies had at least a 3% difference from the median value of the VTE, or at least a 1.5% difference from the median value of the bleeding requiring reintervention. However, if removing outliers did not materially change the median estimate, we considered not to rate down for inconsistency
Indirectness	We did not usually rate down for indirectness, as the eligible studies measured relevant outcomes in representative populations
Imprecision	We rated down by 1 level if studies included <1000 patients and by 2 if they included <200 patients
Evidence certainty	Although certainty in a body of evidence from observational studies addressing a question of prognosis begins as high certainty, we rated it down to moderate owing to uncertainties in our models of the risk of VTE and bleeding over time and in our model of patient risk strata. We then further rated down as described for the other four categories