You don’t need me to tell you that SEND is the standard for representing nonclinical data, but what would you think if I told you that just by looking at the data, it’s quite easy tell which software and organization produced the SEND datasets? How ‘standardized’ does that sound?
In my last post, I touched on the idea that interpretation of the SEND Standard is quite subjective. SEND veterans are well aware of this, but it often comes as a surprise to the newbie. This subjectivity leads to organizations interpreting SEND differently, meaning that SEND is a standard that everyone does slightly differently.
The result is that SEND is a ‘Subjective Standard’, which itself sounds like an oxymoron.
There are various things I can point at to illustrate this subjectivity, however, the easiest example, is that the SEND Implementation Guide contains many, many statements that either imply, or outright state the fact that there is a choice in how data are represented in SEND. It does this with statements along the lines of “In this case, the sponsor chose to represent the data like so…”. A quick search shows over fifty examples where such a choice is either stated or implied.
So how did we get here? Those defining the early versions of the SEND Standard, set out with the best of intentions and needed to address the particular landscape before them. That landscape included the various ways different organizations represented, what were essentially the same data; and often the only source of data were the legacy studies in PDF format. Also, it was intended to mean that no organization would need to change their internal practices – but hindsight has taught us something very different. I’ve discussed in this blog before, the naivety of thinking that SEND would not change data collection practices, but back then, that was the intention. Therefore, the standard needed to be flexible. It couldn’t be too prescriptive. We couldn’t have a standard that required information, data or metadata, that were never collected. Therefore, many SEND variables offered some level of choice as to whether or not they would need to be populated.
This resulted in a standard where some concepts are defined very loosely. As any SEND expert knows, probably the biggest offender is the Clinical Signs domain. The only variable really standardized here is the Category. The other variables offer a considerable amount of choice in how data are represented and which variables to populate.
There are plenty of other areas we could also point at for examples of SEND requiring a subjective opinion. Timing variables, for example. Just this week, we’ve again been debating population of the Nominal Label. Yet, we could also be debating which combination of timing variables could and should be populated.
We could also say so much about the subjective nature of Trial Design, but I think I’ve said enough to convey my general point:
We are all grappling with how we choose to represent our data in a standard that can be very subjective.
Maybe in a future blog post I’ll discuss the advantages and disadvantages of having such a flexible standard.
Till next time,