Inconsistencies between ADNIMERGE.csv and MMSE.csv

[Tina Wang]

Hi,

I am trying to understand why the MMSE score does not always match the MMSE score from the same subject and visit, and which resource to trust. 

I merged ADNIMERGE.csv to MMSE.csv using RID, VISCODE (ADNIMERGE keys) and RID, VISCODE2 (MMSE keys). There are ~131 records where the MMSE scores are different, but the examdate (merged from REGISTRY.csv) is the same. 

For example, 

In ADNIMERGE.csv: RID==6432 and VISCODE=="m12" have an EXAMDATE=="2019-07-24" and MMSE==28

In MMSE.csv: RID==6432 and VISCODE2=="m12" have an EXAMDATE =="2019-07-24" and MMSE==25

Thank you,

Tina

[Mike Donohue]

Hi Tina,

The site entered the 25 and the 28 was scored algorithmically based on the item-level data. They missed one point on WORLD task and one point on recalling the day. The mmse table in the data package for R, SAS, etc has MMSE_EDC (as entered by the site) and MMSE (calculated). The issue has been forwarded for query at the site but not corrected.

Hope this helps,

Mike

[Helena Pereira]

Hi Mike and Tina,

Tina, thank you very much for your question. I had found the same incoherence when analysing the .csv from the 'Imaging download' tab. 

Mike, I found your answer very useful, but when you said 'They missed one point on WORLD task and one point on recalling the day' you mean that the MMSE (calculated) is the correct one? The site missed those points and entered the MMSE score wrongly? 

Can we trust on MMSE.csv?

Thank you both in advance,

Helena

--
You received this message because you are subscribed to the Google Groups "Alzheimer's Disease Neuroimaging Initiative (ADNI) Data" group.
To unsubscribe from this group and stop receiving emails from it, send an email to adni-data+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/adni-data/3ac56d14-ce89-4814-88b4-5ce32a3a7355n%40googlegroups.com.

[Tina Wang]

Hi Mike,

Thanks for your explanation for this instance, I was hoping to get more general advice with how to resolve inconsistencies between the two resources. I am using the downloaded data because I have found there are more rows in the CSVs than in the R-package. I am a new ADNI data user, and I was not sure how often the packages are updated with respect to the downloads resources.

I was also hoping to get advice with other inconsistencies when attempting to leftjoin ADNIMERGE.csv with MMSE.csv with the above keys, my specific question is underlined:

1. 131 rows merge between ADNIMERGE & MMSE, but the entries do not agree
   Similar to Helena, do you suggest I should use algorithmically calculated scores instead for all subjects from the MMSE resource?
   
   
2. 2427 rows do not merge between ADNIMERGE & MMSE, but the ADNIMERGE MMSE is not null
   Does ADNIMERGE have access to more data than what is provided in the MMSE.csv?
   
   
3. 24 rows merged, but MMSE has a score but ADNIMERGE is null.  
   Not sure how this happens given the above?
   
   

Thank you so much for your insights.

Kind regards,

Tina

[Mike Donohue]

We recently discovered the ADNIMERGE R package had not been updating, but it should updating daily again now.

I prefer to use algorithmically calculated scores in my work, mmse$MMSCORE or adnimerge$MMSE. This is because if there is a discrepancy between item-level data and the total score, a likely explanation is an arithmetic error by the site personnel entering the total score.

An exam date is required for a visit in order for the data to be merged into adnimerge, Some MMSE records may not make it into adnimerge if the registry table does not provide an exam date.

Mike