CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Therapist variation within randomised trials of psychotherapy: A design of experiments perspective Authors:  Steven Gilmour - KCL (United Kingdom) [presenting]
Rebecca Walwyn - University of Leeds (United Kingdom)
Abstract: Clinical trials of psychiatric interventions present some particular challenges. One of these is that the therapist and/or other healthcare professional is an inseparable part of the treatment. Considering it as a treatment factor leads to a number of complex and non-standard factorial treatment structures. Often it is not possible for each therapist to implement each intervention of interest, so we get a nested, rather than crossed, treatment structure. Sometimes more than one type of healthcare professional is involved, when one can be crossed with interventions, but the other has to be nested. Very often, therapists can only work in one centre, so we get an unusual type of incomplete block design. In addition to this, it is often desirable to model the therapist effects as random, since the therapists in the trial can be considered a sample of a bigger population of therapists. A further complication is that it is often not possible to randomise therapists to centres. We use the principles of the design of experiments and, in particular, the link between randomisation and the derived model to untangle these different complicated structures. This has the advantage of providing models which can be justified by the trial design, which either confirms or improves on existing practice, which uses models without any clear theoretical justification.