Table 5.
The ten highest-scoring predicted ADRs that were not present in the drug knowledge graph and were validated in EHRs. The number of reports of each drug-ADR pair (“Drug + ADR”) and the total number of reports of all ADRs for each drug (“Drug (all)”) are shown for both the EHR used for validation and the EudraVigilance database. The total ADR reports for each drug in the EHR only includes the 10 ADRs used for validation. The EudraVigilance reports include all cases for all ADRs reported in the dataset up to August 2017 (accessed October 2017). Note that the ratio of “Drug + ADR” to “Drug (all)” is expected to be much larger in the EHR as only 10 ADRs are considered, vs all ADRs for EudraVigilance.
| Drug | ADR | EHR | EudraVigilance | ||
|---|---|---|---|---|---|
| Drug + ADR | Drug (all) | Drug + ADR | Drug (all) | ||
| Imipramine | Akathisia | 2 | 4 | 5 | 1,465 |
| Trimipramine | Akathisia | 1 | 2 | 3 | 931 |
| Amitriptyline | Akathisia | 2 | 13 | 16 | 8,832 |
| Quetiapine | Alopecia | 18 | 20 | 74 | 34,010 |
| Mirtazapine | Neuroleptic Malignant Syndrome | 2 | 81 | 81 | 10,215 |
| Clomipramine | Pulmonary Embolism | 1 | 8 | 15 | 3,676 |
| Lamotrigine | Pulmonary Embolism | 8 | 63 | 29 | 21,168 |
| Donepezil | Pulmonary Embolism | 7 | 12 | 15 | 5,129 |
| Haloperidol | Pulmonary Embolism | 6 | 53 | 99 | 9,532 |
| Aripiprazole | Stevens-Johnson Syndrome | 3 | 564 | 29 | 17,758 |