Recently I was invited to present at an event organized by the PSDM (Pharmaceutische Statistiek en Data Management). The PSDM, and I quote from their website, is a Dutch network group of statisticians, clinical programmers and data managers working in and for the pharmaceutical industry.
The title for the event was “Implementation of CDISC standardization: Pragmatic Approaches” and I spoke about using the new idea of Biomedical Concepts but making them work in a pragmatic way within a current process.
I won’t write up the complete presentation but there is one section I created for the talk that I think might be worth sharing. I was wanting to illustrate some of the issues I see within the current CDISC standards; the incomplete nature and where Biomedical (Research) concepts can help us.
It was about 6.30am on the morning of the presentation and I was unhappy with the slides. I was wanting a practical example of some of the current issues. I was skimming through therapeutic area guides looking for something that would fit the message I was trying to get across and I came across the Glasgow Coma Scale within the Traumatic Brain Injury TA User Guide. So I produced these slides and put them into the presentation to illustrate some of the issues we as an industry face. Note this is not the whole PSDM presentation, just a few slides focused on the issues we face.
In the spec was a nice Concept Map diagram (I like these, these are good thing) that gave me a quick overview of the data being collected and why (slide 2) along with a CRF (slide 3). So I had a nice overview. Along with this was the normal example data we see in an implementation guides (slide 4).
So I picked these apart and created a quick spreadsheet setting up some simple Biomedical Concepts (BCs) and put the sheet onto a couple of slides (slides 5 and 6) and diagrammed one of the concepts out in a tree form (to match those earlier in the presentation).
Looking at slides 5 and 6 in particular I can start to see issues. There are coded values in the example data (e.g. the test codes) that don’t exist within the current CDISC terminology. Neither does the category code. Should I be using a subcategory code? Normally with a scale like this it might be used to distinguish the total score from the questions. What about the evaluator, what are the possible values? Investigator is used within the example but are there others? None of the result codes are specified in terminology, the example CRF suggest values but it allows people to use their own.
So, in the words of Freddie Mercury and Queen, I want it all. I want the full picture. I want full and complete Biomedical Concepts. We need to fill in all the holes. If not, how do we expect people to submit consistent data within an organisation never mind across organisations?