groups » Data Management and Statistics » Approaches to resolving discrepancies

I'd like to hear which different approaches are being taken in terms of resolving discrepancies - especially when doing double data entry.

I have recently started in a new job and inherited two fairly large MS Access databases, one for first- and another for second-entry. I guess there is a variety of ways these two could be compared and the differences resolved - what do you guys suggest?
(The trick is to distinguish which differences are relevant - The difference between "Tom Jones" and "TOM JONES" would probably not be worth spending a lot of effort on correcting?)

Reply

data capture data management double data entry

  • hsosa hsosa 16 Aug 2010

    Thank you. I have read this discussion and it has helped us with our data management in our clinical trial in Lima.

  • wenny Ednah Wenwa 27 Jul 2010

    hallo Samuel,you are in quite a tricky situation but the contributions above are quite helpful.There is also an open source application available free of charge called openclinica that makes double data entry very simple.it complies to all the FDA rules and makes management of clinical data easy indeed.have a look at it and you could probably implement it in your next study. we use it here in kilifi and so far it is working perfectly. All the best.

  • David David Mwakazanga 8 Jul 2010

    I have always found the data comparison suite in Epi Info very efficient at resolving discrepancies in Access databases, though creating an audit trial therein is a big challenge. On issues of text, most statisticians prefer them to be in lower case, and embedded spaces to follow the normal rules of type setting. So, 'Tom Jones' would be entered, and appear in both databases as 'tom jones'. As such, the SOP used for the database clearly states such subtle rules.

  • cpmasesa Clemens Masesa 8 Jul 2010

    Samuel,

    With regards to the "Tom Jones" / "TOM JONES" issue.

    One possible approach would be to compare the text after converting it to the same case, upper or lower. In MS Access one could use the UCASE() or LCASE() functions within the comparision query / code to achieve that.

    Another problem that I frequently face with comparision of string variables is the presence of trailing / leading or multiple embedded spaces.

    An example would be "Tom Jones " / " Tom Jones" / "Tom Jones".

    Using the function TRIM() will get rid of the leading and trailing "spaces".

    Therefore one might want to use something similar to TRIM(UCASE(varname)) in the comparision query / code so that if the text is basically the same but is in a different case and / or has leading / trailing spaces then it is not flagged as an error (or flag as an error with low fix priority)

    How would one achieve the same for ultiple embedded spaces? eg "Tom Jones" and "Tom Jones" (First has 1 space, 2nd has 3 spaces)

    Kind regards,

    Clemens

  • romollo Raymond Omollo 7 Jul 2010

    There are several ways of sorting discrepancies depending on the database used. From my experience:
    In Acces platform, the third person who is senior to the data entry guys would compare the two entries in Epi Info and produce a report with discrepancies between the entries (which would also include the Tom Jones and TOM JONES difference). He/She will then make reconciliation by checking against the hard copy records (CRF's) in fields with discrepancies.
    In Open Clinica, the 2nd data entry person happens to be the one whose entries are eventually stored in the database. When discrepancies occurs between his entry and the 1st person, he/she can file a discrepancy note or change his entry to conform with the previous entry based on the actual record. We then have a third person, independent of the two who reviews the filed discrepancy notes for reconciliation.
    Generally, as a way of limiting discrepancies there is need to have field level and cross field checks inbuilt into the database prior to start of data entry for each trial.

  • hienhv hienhv 7 Jul 2010

    Hello,

    It's always a challenge to handle double data entry on MS Access databases.

    The most appropriate method I've ever used is to create two databases: one is for the first and another for the verification and those databases must be entered by two data entry staff. The verification database should have links to the first database's tables and a check will be executed whenever a new record are entered. by this way, every time the second pass entry differs from the first, a warning message will be generated and the second data entry staff will have a chance to look back what he has done and is able to correct his/her mistakes. Of course, a discrepancy should be genarated as an audit trail and for further review.

    Double data entry is a method to reduce the mistakes of data entry. with the second data entry staff having an opportunity to correct his/her mistakes, data entry errors should be minimized. However, how we trust the second data entry staff's decision if he/she want to hide his/her mistakes by leaving his/her mistakes in the database. To handle this, a third person (a supervisor) should review all discrepancy records, and make final decision on the data.

    The most challange on this is to maintain the two databases when we have to change the databases due to the changes of the CRFs.

  • herculekalonji Hercule Kalonji 6 Jul 2010

    Hello, thank you for your message. For us when we do double entry for Access databases, we then use software (Epi Info) to compare data from both databases. And then we correct the differences in the two databases.

  • mamulla Moses 6 Jul 2010

    Hello Moses. I shall tell you what we do and then I am sure others will comment. You are right, there are those that matter and those that do not. Before you start your data entry you need to set out your plan and agree what you are going to query and what you are not. You can also set down what the data manager can resolve (your Tom Jones TOM JONES example), what needs to go back to the trial team and which can only be resolved by the clinical team responsible for care. It is possible to predict most of what will arise - but not always. Good luck and interesting to now hear from others....

Please Sign in (or Register) to view further.