i. Service level – these are typically provided by more than one individual, each providing a complex package of care in a specific context and inter- acting with others in a complex way. Examples, might include stroke unit interventions [12] or early supported discharge teams [13]. It is interesting to note that some of the most robust stroke rehabilita- tion evidence comes from trials of such complex interventions. However, there is often difficulty in interpreting and implementing such evidence.
ii. Operator level – these interventions are typically provided by a single operator such as the therapist or nurse, who provides a complex package of care that could incorporate both the personal interaction between the therapist and patient plus the therapy they provide. A good example of this level of inter- vention is occupational therapy for stroke patients living at home [14, 15] or stroke family support workers [16].
iii.Treatment level – at this level of complexity, the impact of an individual intervention is evaluated. This may arguably provide the most useful evidence for a clinician. Ideally the potential impact of the therapist should be removed from the evaluation of an individual reproducible intervention but in prac- tice this can be difficult to achieve. Examples of such treatment decisions include functional electrical stimulation for upper limb recovery [17] and treadmill gait retraining [18].