A quick post here regarding our most recent paper on a brief version of the Neck Disability Index, now known as the NDI-5. I’ll not go into detail on the processes we used, as the paper is open access and anyone interested can take a read. I’ll make a few comments though on how the project came to be and what I think is the impact.
The NDI is, according to a scoping review from one of its developers Dr. Howard Vernon, the most commonly-used neck-specific disability scale in the world. And I don’t doubt that, I can’t remember the last time I read a paper that used a scale other than the NDI for assessing neck-specific disability. So it bears saying ‘well done’ to Drs. Vernon and Mior for creating such a ubiquitous scale. And it was this fact that actually gave my co-author Dr. Joy MacDermid and I some pause when deciding whether to publish our work on the scale. At the end of the day, our paper (if you read even slightly between the lines) doesn’t cast the NDI in a particularly positive light. There are conceptual and statistical problems throughout that could, if we’re not careful, render many research findings that have used the NDI as the primary outcome (and there are a lot of them!) invalid or at least potentially inaccurate. Did we want to potentially open that Pandora’s box? Not really.
At the end of the day though you could argue that all is generally well, in that the NDI-5 correlates very strongly with the original NDI, so on balance, results you got using the original NDI would likely be about the same as those using the ordinal version of the NDI-5. That said, using Rasch analysis we were able to create a linear transformation matrix that, in a nutshell, indicates that research outcomes using the original NDI may in fact be quite different if the linearly-transformed NDI-5 had been used instead.
Of course, as with anything in research, we shouldn’t jump to any conclusions. Until our results are replicated by another, independent group (anyone game?), it’s possible that our results are somehow, well, wrong. I’m not saying we weren’t careful, far from it, this was probably the most careful I’ve ever been with such an analysis, recognizing the potential implications. But one never knows in post-positivistic quantitative research – could there have been something about our sample that might lead others to find different results? Don’t know, and while I personally feel comfortable using the NDI-5 going forward, others may wish to wait until you get some independent consensus. That’s ultimately up to you.
I will close with this – if you do choose the NDI-5 as a primary outcome going forward (and I suggest you strongly consider it), then you’ll also need to be sure and include a standalone symptom scale of some kind. That could be something as simple as a Numeric Rating Scale, or any number of more complex pain and other symptom-type scales currently available. I would also highly recommend the use of a sleep-quality scale as we removed the sleep item in the NDI-5 and it is by most accounts an important domain to measure, adequately important to warrant its own scale actually. If you’re wondering why we removed that item, be sure to check out the paper.
You can grab a copy of the NDI-5 and the scoring and transformation matrix under the Clinician Resources section of the page in the right navigation bar.