Background: The prescription of medicines above recommended dosages, is one of the most common medication errors responsible for causing preventable harms. Not all medication error reports are coded during submission to VigiBase, the WHO global database of individual case safety reports.
Objectives: To explore the possibility to perform quantitative signal detection of possible adverse drug reactions related to medication errors in VigiBase based on high-dose reports identified by statistical outlier detection, which were not coded as medication errors.
Methods: A robust Z-score method, which identifies outlying deviations from the median reported dose of each drug, was used. The threshold was identified using the maximum recommended daily dose according to UK electronic Medicines Compendium (UKeMC) for a subset of 12 drugs with diverse ATC-codes.
Case series associated with outlying high doses were identified based on the odds ratio for the strength of association between the drug-adverse event pair and the outlying high drug dose. Case series were selected for review from the list based on clinical judgment and were manually reviewed for preventable cause of the high dose and potential medication error based on predefined questions.
The dataset consisted of reports entered in VigiBase up to January 2020. The study was limited to reports with an official ATC-code, the drug being reported as either suspected or interacting, and with a computable daily dose. Since route and age might influence the dose, the scope was further limited to reports concerning adult patients aged >= 18 and < 65 with oral administration.
Results: In total, 62 case series were clinically evaluated, comprising 898 suspected high dose reports according to the Z-score method, of which 502 (56%) included narratives. 372 (41%) were considered high dose cases according to the maximum daily dose specified in the UKeMC. During manual review, at least one report with a dose higher than the UKeMC maximum was identified in 51 (82%) out of the 62 case series. In 26 (42%) case series, all outlying high-dose reports identified by the Z-score method were confirmed to be high according to the UKeMC.
Conclusions: Current pharmacovigilance infrastructures have the potential to address medication errors and other drug-related problems in large databases using statistical methods. The Z-score method enabled the identification of high-dose reports which were not explicitly coded as medication errors in VigiBase. However, the identified reports did not contain enough information to identify root causes of the high doses, which may be needed for implementing risk minimization actions.
.jpg "Eva-Lisa Meldau, MSc photo")
