Important part of the data management job is to verify received data. Checking for inconsistencies and unexpected patterns. Verifying that the data is complete, legible, logical and plausible.

However, how to perform data verification?

You could regard the data verification job as completing a couple of puzzles. Each puzzle is one subject participating in the clinical trial or clinical study at stake. As such, the puzzles resemble each other a great deal. But they are not exact copies. Each subject, each puzzle, is (slightly) different, unique.

Pleasant and thoughtful team action:

Do you have a puzzle somewhere in a cupboard? More than one from the same series? At least 2 puzzles with > 100 pieces each? Open the boxes, drop their content in one pile on the table and start completing the puzzles/subjects. The more pieces in place of a puzzle, the more evident which pieces to expect.

1. Get the parts received, divide them per subject/puzzle and start making all the puzzles. The clinical information up on each subject is coming in pieces, per completed visit data, per available adverse event information. In the beginning you’ll thus work with lots of incomplete puzzles.

2. Any holes in any puzzle/subject, any missing parts, you need to look for/query. Note that holes are allowed if your puzzle/story is as such! However, leave no unexpected holes. Meaning that if an assessment took place, you want to have the corresponding result(s) completed.

3. Any duplicate pieces, get rid of them. Please query.

4. Any pieces not fitting your puzzle/subject story, you need to check up on. Maybe they belong to another puzzle/subject. Or they are incomplete and can therefore not fit (yet). They could even be wrong delivered and not belong to the study at all.

5. Any pieces fitting but rotated 90 or 180 degrees, please turn/query. Get the puzzle showing a logical story.

6. Any pieces damaged, please try to fix the damaged parts. E.g. spilled coffee over a paper CRF. Illegible text parts. Or unclear texts that can be interpreted differently.

7. Any pieces added at the wrong place, query and bring to their right position. E.g. an error in an assessment date.

In trial/study language, the more data for a subject received and in the database, the easier to get the subject’s story complete. However, the more care needed to get the true story. The logical, plausible subject story. Attention to medication given for an adverse event but missing in the concomitant medication list. Or laboratory shifts to worse results but missing corresponding adverse events listed.

Completing the holes in a puzzle is easy, for data management the edit checks help you tremendously with that. Getting a logical, plausible story for each patient, reflecting the truth, is the real data management challenge. Which takes more than just structuring pieces. It asks you to look and understand the pictures up on the pieces received.

Good luck with your data management puzzles,

Source:

“This is an article of ProCDM. Clinical data management training. Receive tips and the free e-book ‘Five strategies to get reliable, quality clinical data’ by subscribing via http://www.procdm.nl/pages/knowledgebase.asp.”

Comments? Join us at {EDC Developer}

Anayansi Gamboa, MPM, an EDC Developer Consultant and clinical programmer for the Pharmaceutical and Biotech industry with more than 13 years of experience.

Available for short-term contracts or ad-hoc requests. See my specialties section (Oracle, SQL Server, EDC Inform, EDC Rave, OpenClinica, SAS and other CDM tools)

As the 3 C’s of life states: Choices, Chances and Changes- you must make a choice to take a chance or your life will never change. I continually seek to implement means of improving processes to reduce cycle time and decrease work effort.

Subscribe to my blog’s RSS feed and email newsletter to get immediate updates on latest news, articles, and tips. I am available on LinkedIn. Connect with me there for technical discussions.

Fair Use Notice: This article/video contains some copyrighted material whose use has not been authorized by the copyright owners. We believe that this not-for-profit, educational, and/or criticism or commentary use on the Web constitutes a fair use of the copyrighted material (as provided for in section 107 of the US Copyright Law. If you wish to use this copyrighted material for purposes that go beyond fair use, you must obtain permission from the copyright owner. Fair Use notwithstanding we will immediately comply with any copyright owner who wants their material removed or modified, wants us to link to their website or wants us to add their photo.

Disclaimer: The legal entity on this blog is registered as Doing Business As (DBA) – Trade Name – Fictitious Name – Assumed Name as “GAMBOA”.

Advertisements