Hi Nicola,
I'm not sure I have the stats answer you're looking for - it looks like
a complicated study so my first suggestion is to see if there's any way
you can simplify it by sounding out your research question, then
hopefully the stats method will present itself.
Are you using the patient experience questionnaire scores as the
outcome, as a way of measuring which intervention is better? You mention
looking to 'establish differences between groups' - do you mean
interventions, or the patient groups themselves? From the info you've
given it sounds like the allocation of patients to intervention(s) is
not random, but rather based on clinical need, which suggests that
looking for differences between patient groups after the intervention
could be somewhat circular, as you might just pick up the pre-exsiting
differences that led to the allocation of each intervention in the first
place. Likewise, if you're looking for differences between
interventions, you have to bear in mind that different patients were
involved. Without random allocation you have the confound of patient and
intervention being associated, which makes it difficult to tease out
where any differences might lie.
Because some patients will get more than one intervention, there's a
chance that experiencing one intervention might influence the other
(i.e. if intervention A doesn't work, they get B, then C etc.). One
option might be to just look at the patient's questionnaire of the first
intervention they experience, which will give you some data to compare
interventions across these naive patients (but still with the confound
that the intervention they received first was in part determined by
their patient status).
Another approach might be to look at effectiveness overall, regardless
of how many interventions they get - so how effective is the indiviual
program approach? I don't know how different these patient groups are,
but ideally you'd find a common metric across groups, and obtain a pre-
and post-intervention measurement so you can look at change scores. If
there's not a common metric as the patients have very different
symptoms, another approach might be to define an effect size of change
on core symptoms within each patient group that you could use to
determine if the intervention approach was effective.
If assessing effectiveness is not the purpose of the study, but it's
more concerned with the patients' experience, then perhaps a more
qualitative methodology might be better suited?
Sorry if that's more questions than answers, but hopefully might help
clarify your thinking!
Brian
Davies, Nicola wrote:
> Hi Everyone,
>
> I am working on a rather large project, which is certainly challenging my statistical expertise (which I didn't have much of in the first place).
>
> There will be four groups of people and four interventions. Whilst some will only be recieving one intervention, others might recieve two or three, or even possibly all. This is because the study is in a clinical setting and patients will be assigned to interventions according to their need, some of which might change.
>
> All patients will complete a patient experience questionnaire. How do I establish differences between groups, when there are so many conditions? Or is it just not possible?
>
>
> Kind Regards,
>
> Nicola Davies
> BSc (Hons); MSc Comm.; PhD Researcher
>
|