vigiMethods

vigiGrade

Multidimensional measure of the amount of information on reports. Highlights quality issues in collections of individual case safety reports. Gives a score for how complete a case report is. A number of fields have been selected for their usefulness in the case assessment process and a score is given to each field if it has information. The individual field scores are weighted and combined into a score for the whole ICSR. The maximum score is 1.0.

The main use of vigiGrade is as part of the communication with countries on data quality, but it has also proven to be an indicator of a true signal and is therefore one of the parameters included in the vigiRank method used by UMC in signal detection. 

vigiMatch

Probabilistic record matching method, a likelihood-based approach to identify unexpectedly similar record pairs in large databases. It computes a match score for each pair of records, where matching information is rewarded and mismatching information penalised. This match score reflects the probability that the two records relate to the same underlying entity or, in this setting, that they are duplicates. Record pairs with match scores that exceed a certain threshold are flagged as suspected duplicates. The threshold is derived from a comparison between the match scores of confirmed duplicates and of random record pairs in the database of interest. 

The vigiMatch method for detecting duplicates has been in use at the UMC in signal detection since 2014.

vigiRank

Predictive model that ranks pharmacovigilance safety signals according to multiple aspects of strength of evidence. Parameters that make up the vigiRank are Disproportionate reporting (IC025), Recent reporting, Geographic spread (NCountry), Informative reports (vigiGrade Completeness score as Ncomp09) and Narratives. The vigiRank has been successfully implemented in UMC’s signal detection since 2014.  

vigiPoint

Identifies and pinpoints key features of a data subset of reports (e.g. ages, sex, co-reported drugs and adverse reactions) in contrast to a broader dataset.

Read more

Last modified on: January 10, 2017
.