There are two consistent themes in the current modernisation agenda for health and social care in England: the imperative to embrace change and abandon long-accepted traditional modes of working and the requirement to engage in evidence-based practice. Mental health, as one of the Government's key clinical priorities, is at the forefront of change. The difficulty for practitioners and policy makers alike is that little of what we have traditionally done in the mental health field and few of the prescriptions for change ordained by Government have been evaluated to currently accepted standards for evidence-based medicine (NHS Centre for Reviews and Dissemination, 2001). This partly reflects the generally poor standard of randomised controlled trials (RCTs) carried out within mental health and the methodological complexities surrounding mental health research (Reference Richardson, Baker and BurnsRichardson et al, 2000). Some very important issues may be difficult, if not impossible, to address using the RCT methodology. Others require the use of cluster-randomisation, a technique that is statistically complex, ill-understood by both researchers and funders, ethically challenging and potentially very expensive (Reference Ukoumunne, Gulliford and ChinnUkoumunne et al, 1999).

Thornicroft et al (Reference Thornicroft, Bindman and Goldberg2002, this issue) have produced 11 recommendations aimed at filling the palpable evidence gap within mental health policy and practice, drawing on a framework developed by the Medical Research Council (MRC) for the development and evaluation of RCTs for complex interventions to improve health. Their prescription is, predictably, comprehensive and intelligent and is firmly aimed at funders. It spans the development of research capacity, support for large-scale pragmatic RCTs that address real-world questions and the evaluation of the use of routine data-sets as an alternative to the rigours of the RCT (a highly controversial issue requiring very careful consideration). Particular gaps in the evidence base are emphasised in training, dissemination and organisational change. These issues can only be effectively addressed within an RCT at the patient level by cluster randomisation.

Adoption of these recommendations would go a long way towards the goal of supporting evidence-based change in mental health services. A few punches are pulled. The enormous importance of the pharmaceutical industry in funding research and the consequent impact on the choice of research careers adopted by trainee psychiatrists and the RCTs that are carried out, is not discussed. Thornicroft et al (Reference Thornicroft, Bindman and Goldberg2002, this issue) allude to (in recommendation 9) but fail to emphasise the importance of theory in mental health services research, which has become a theory-free zone. Not only do we need a social science capacity in mental health research, we need to ensure that psychiatrists are as literate in the social sciences as they are becoming in neurobiology. Mental health care is a multi-disciplinary activity: there is an urgent need to develop research capability among occupational therapists, social workers and nurses — professions that largely lack a research basis.

Finally, there is the big unanswered question within the modernised managed mental health service: how do we identify, support and fund service innovation? Without the capacity to innovate, evidence-based change cannot occur.