As a clinical psychologist working in an in-patient psychiatric setting, I am fully supportive of the principles behind the recovery model (or recovery approach) in the treatment of severe mental health problems. As such, I am fully supportive of efforts to ensure that the recovery approach is at the heart of the service delivery.

I have observed that a number of services - including our own - have adopted the Recovery Star model. The model is ‘a tool for supporting and measuring change when working with adults of working age who are accessing mental health support services’ (www.mhpf.org.uk/recoveryStarApproach.asp). Although I am supportive of the aim to measure such change, I am concerned that the Recovery Star model itself does not appear to have been considered in any peer-reviewed publications. Furthermore, there do not appear to be any available normative data published alongside the instrument, or any statistics indicating its reliability and validity. Given that the authors specifically describe the tool as something to be used to measure change, this is a very notable omission. Without such data it is impossible to know whether, for example, two different scores on two different occasions represent genuine therapeutic change or simply arise out of error; nor is it possible to know the extent to which two different clinicians using the tool would be expected to concur with each other. Furthermore, the tool proposes that ten different factors of recovery exist, yet again there is no mention of a factor analysis suggesting how such factors were derived or how they interrelate.

Although the development of instruments to measure patients’ perceptions of engagement in the recovery model is very important, I am of the opinion that no psychometric instrument should be used clinically until it has been administered to a suitable sample and the results have been subject to the usual peer-review process. If these vital steps are abandoned as unnecessary, we have no idea what the instrument is measuring or whether results amount to positive therapeutic change. Given that numerous other freely available instruments have been published and validated within a range of clinical samples (see Campbell-Orde et al ^{Reference Cambell-Orde, Chamberlin, Carpenter and Leff1} and Burgess et al ^{Reference Burgess, Pirkis, Coombs and Rosen2} for two excellent reviews), it is puzzling that services are choosing to use an instrument where the basic statistical data are not available.