In addition, as our patients and their carers were not aware that they were in a randomised trial to assess our stroke family care worker and that our follow up was attempting to assess her effectiveness, they were in effect partially blinded. We might imagine that loyalty to the care worker might have biased their responses had they known the precise purpose of our follow up. Fourthly, our approach allowed patients or carers to decide to see the stroke family care worker when it was relevant to them. Some patients might not consent to randomisation shortly after their stroke, when they are unlikely to foresee the possible psychosocial impact of the stroke on them and their families. They might then regret the decision not to be randomised when the potential benefits of the intervention become more evident. Lastly, as our intervention was applied to patients and their families it was unclear who might most appropriately give consent.