Table 2.

Requirements to identify disproportionately reported subgroup DECs

Requirements by Sandberg et al. [29]	Adaptations for sensitivity analyses
1. Subgroup IC0005 > 0 and unadjusted entire database IC025 ≤ 0 or for covariates with > 2 subgroups: IC005 > 0 simultaneously in ≥ 2 subgroups and unadjusted entire database IC025 ≤ 0a	1. Unadjusted entire database IC025 ≤ 0 ignored 2. Subgroup IC0005 > 0 changed to subgroup IC005 > 0
2. IC value computed by dividing the subgroup O/E by the adjusted O/E for the remainder of the databaseb with log adjustment and shrinkage (ICΔ) > 1	3. ICΔ > 1 changed to ICΔ > 0 or ICΔ not calculablec

DECs drug–event combinations, IC information component, IC₀₀₀₅, IC₀₀₅ and IC₀₂₅ lower limit of the 99.9%, 99% and 95% credibility interval for the IC, respectively, O/E observed-to-expected ratio

^aThis alternative allows for detecting patterns that would exist in several but not all subgroups with a less strict credibility interval for each individual subgroup

^bObserved (O) is the number of observed reports for the subgroup DEC and expected (E) is the number of expected reports for the subgroup DEC obtained by multiplying the number of reports for the product of interest in the subgroup by the number of reports for the event of interest in the subgroup divided by the total number of reports in the subgroup. The adjusted O/E ratio for the remainder of the database is computed as a weighted average of the O/E ratios in the other subgroups for the same covariate

^cIC_Δ cannot be calculated when the product or event is not reported with other subgroups (i.e. the expected number of reports from the remainder of the database is 0); thus, the related adjusted O/E ratio cannot be calculated