Difference between revisions of "Performance of probabilistic method to detect duplicate individual case safety reports"

Revision as of 21:31, 30 September 2015

Background

The duplication of case reports is problematic and misleading for effective pharmacovigilance. The sources of these duplicates are multifactorial and there different approaches to detecting these duplicates but this article evaluated the probabilistic recording or patient matching approach ^[1]

Method

The authors used a likelihood-based detection algorithm (VigiMatch®) that computes a match score (the probability that the two likely records relate same entity) for each pair of likely duplicated records was utilized. VigiMatch®was applied on the WHO global individual case safety reports database (vigiBase®) that contain over 8 million reports of suspected adverse effects of drugs reactions from 112 countries but this study focused on three countries (UK, Spain and Denmark).

Results

The vigiMatch achieved a very high predictive value for confirmed duplicates for each data set that range from 82% to 32%. The use of rule-based duplicates detection failed in significant proportion. The sources of duplication varied by country but similarities were observed.

Conclusion

The vigiMatch hit-miss approach detected duplicates that were missed by rule-based methodology resulting in reduced suspected duplicates and improved manual reviews accuracy.

Remarks about the article

My concern is that the authors did not consider the contextual differences among the countries selected for the study. There could have been that the data collection, analysis and reporting plan was different for each country and this could have affected the reported data to vigiBase®. While this article will serve as a base for my project, the context of Nigeria is very different from the European countries.

Reference

1. ↑ Tregunno, 2014. Performance of probabilistic method to detect duplicate individual case safety reports. http://www.ncbi.nlm.nih.gov/pubmed/24627310