Skip to content
Visit our COVID-19 Response Centre for all our updates Read more

Robin Williams, B.S., CMC, Senior Safety Scientist, Ashfield Pharmacovigilance

Data mining can be described as a statistical technique, by which databases are searched (or mined) to detect strong, consistent associations of drug-event pairs that occur at higher than expected frequencies.

Data mining challenges 

Data mining in large databases, although typically fairly easy to maneuver, pose challenges which must be taken into consideration when performing data mining.
One particular challenge is selection bias, when the reporters are in control of what actually gets reported. This could lead to an over-representation, or just as easily, an under-representation which would skew the true data.

Adverse Event (AE) reporting can also be influenced by media attention. This can cause a sudden influx of reporting that may or may not be a true representation of a drug event association. A bandwagon effect can occur – the novelty of the drug can affect the rate of reporting events.

Another challenge is that source data is typically unverified, especially if they are post-marketing sources. Consumers self-diagnosing can also skew the data.

Download the whitepaper to find out more about the role of data mining in signal detection.

Receive all of our latest blog entries straight to your inbox!

Contact us