I noticed there are discrepancies in counts of events for a given drug when querying OpenFDA for a given quarter (I tried q1 2014) and trying to get a matching answer from the downloadable FAERS dataset. What is the explanation for this? The error is typically small (maybe 5 %), but bothering nonetheless.


  • Can you please post the query you are running? – Hans Nelsen Jan 23 '16 at 21:01
  • For "PLAVIX", from 2014-01-01 to 2014-03-31. OpenFDA reports 1730 GASTROINTESTINAL HAEMORRHAGE as most frequent AE, whereas FAERS datafile downloaded as ascii for the same period reports 1707. Here it is a 1.3% difference actually, but it can be worse on some other PTs. – jymaze Jan 23 '16 at 21:19
  • The query is:api.fda.gov/drug/… – jymaze Jan 23 '16 at 21:36
  • Thanks for these details. A few followup questions: 1. Are you seeing only cases where the API is larger than the ASCII file? 2. Are you only looking at the MEDDRA PT field? If the answer to both of these is yes, then I believe what is happening is the API is normalizing the case of this MEDDRA PT field and I imagine your ASCII analysis code is not. Let us know if this isn't the case and we can dig into the details further. – Matt Mohebbi Jan 25 '16 at 1:25
  • Thanks for the follow-up. It is not case normalization. I have the same number of unique PTs before and after uppercasing the series. I tried with another drug (Avonex), and the difference is huge, which make forensics easier: After careful examination of the dataframe, it appears that many duplicate lines are generated during the join operations between ascii tables. I have no idea why. I will revisit my code and the source tables. Thanks for your help. My bad... – jymaze Jan 25 '16 at 2:10

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.