Gowers et al's (Reference Gowers, Wheetman and Shore2000) study of the impact of hospitalisation on the outcome of anorexia nervosa in adolescence is a useful and important contribution to a debate that is difficult to resolve meaningfully, because of lack of useful evidence. As the authors note, randomised controlled trials are both lacking and extremely difficult to perform, for both practical and ethical reasons. However, the significant mortality and morbidity associated with these disorders is such that this problem must not be ignored.

Although the paper raises some very important questions, we are concerned that the suggestion that in-patient treatment is associated with a poor outcome is premature, and may be taken by some to mean that in-patient treatment should not be considered. This view would be particularly worrying if adopted by cash-strapped health authorities that are already often reluctant to finance treatment of what is still sometimes seen as a trivial condition.

We believe that three questions need to be answered before making any general pronouncement on the appropriateness of in-patient treatment; (a) what factors lead to admission? (b) what is the relationship between these factors and outcome? and (c) what constitutes in-patient treatment, and is it a uniform concept?

Our experience of over 500 admissions of young people suffering from anorexia nervosa leads us to the view that many of the factors which lead to admission, but which are also predictive of poor outcome, are systemic. They will not therefore be measured by the Morgan-Russell Assessment Schedule (Reference Morgan and HaywardMorgan & Hayward, 1988) or other individual-based predictor variables. Such systemic variables include major psychosocial stresses within the family, and the health and strength of the professional network, but we have found it hard to find instruments that adequately measure these factors.

In other words, the measures used to assess severity in this study are all individual to the patient and do not sufficiently take account of the complex network of relationships within which anorexia nervosa takes root and either flourishes or dies. In our experience, the severity of symptoms such as weight loss does not bear a linear relationship to outcome because of highly complex intervening contextual variables, which need to be addressed by any outcome study.

We certainly share the view that inpatient treatment is not the only response, and that we need to be continually reflecting on the style and content of such treatment. However, we think it highly premature to conclude that it should be discouraged. It should be remembered that at present it is often a life-saver for many young people who are seriously ill.