Responding to Robinson's article on avoiding hospital deaths from anorexia nervosa, ^{Reference Robinson1} the most helpful context to consider this in relation to teenage patients is to place it within a broader concern about risk. Robinson states that a ‘very unwell’ patient should be admitted, but crucially, the definition of that is still not sufficiently clear. How risk is perceived, including what is severely disabling as well as what may be ‘life-threatening’, is a key issue.

Using death certificate data provided by the Office for National Statistics about 18 years ago, I observed 112 certified deaths in England and Wales over a 5-year period; however, only 7 of these individuals had been below their 18th birthday. Notwithstanding the uncertainty of death certificate methodology, ^{Reference Gowers, Clark, Roberts, Griffiths, Edwards and Bryan2} in this instance, suggested by the observation that a third of the 112 deaths had occurred after the person's 65th birthday, these 7 deaths approximate to only around 1 in 5000 adolescents with anorexia – an important finding to set in context fears about these young patients.

That death-data enquiry had been to establish a better empirical understanding about risk following our team's decision (which I supported) to recommend the de-commissioning of a psychiatric in-patient unit that had often provided long-term treatment for teenagers with anorexia. It had previously participated in the UK's first prospective multicentre study of adolescent psychiatric admissions, which demonstrated disappointing treatment effects for those with anorexia nervosa. ^{Reference Rothery, Wrate, McCabe, Aspin and Bryce3} But without such a facility, might there be a local increased risk of fatal outcomes for this condition? Reassured that the probability of death was unlikely to be significantly increased by closing the unit, a substantial change in practice was possible, relocating therapeutic skills to enhance out-patient treatment capacity. Gower et al's subsequent treatment study ^{Reference Gowers, Clark, Roberts, Griffiths, Edwards and Bryan2} confirmed our view that without hospitalisation the disorder should not usually be regarded as hard to treat, untreatable or life-threatening.

Declining death rates observed for anorexia nervosa over the past two decades have been attributed to its more effective and earlier introduced treatment, but not necessarily because the treatment was hospital based. ^{Reference Reas, Kjelsås, Heggestad, Eriksen, Nielsen and Gjertsen4} A careful review of the literature provides two lessons less prone to grab media headlines than premature deaths. First, in adolescence at least, chronicity rather than death is by far the more likely adverse outcome of failing to effectively treat the condition. In comparison with adults, in whom medical complications are not uncommon and excess mortality rates have been observed compared with the normal population, the only significant medical complication (as opposed to biological adaptation to starvation) during adolescence is progressive loss of bone mineralisation. Yet published studies on adolescent admission imply that hospitalisation was most often considered essential to avoid a youngster's possible death, not to divert them from a pathway into chronicity. The COSI-CAPS multicentre study of adolescent psychiatric hospitalisation is particularly instructive in throwing light on how risk in these patients is constructed. ^{Reference Tulloch, Lelliott, Bannister, Andiappan, O'Herlihy and Beecham5} Anorexia nervosa was the single most frequent diagnosis at admission (108/403 patients); only a sixth of those patients were detained but two-thirds nevertheless were considered at risk to themselves. The cohort was disproportionally White, female, aged 15-17, living at home, and with an overrepresentation of single parents. The body mass index (BMI) of all patients with anorexia on admission was within the ICD-10 diagnostic threshold (of 16, for adults), but most were not far below it (14.8; s.d. = 1.8, n = 108, 95% CI 14.3-15.4). Since the normal range of BMI for adolescents aged 15-17 is also less than for adults, it seemed that a relatively low threshold for admission was occurring.

This study had usefully included a number of independently provided units (private hospitals), accounting for a third of their non-eating disorder cases. Such youngsters were significantly less likely to have been receiving any psychiatric treatment before admission (P<0.001), emphasising the part community concerns play in hastening hospitalisation. In short, the second lesson taught me that risk often seems to have been ‘socially constructed’ rather than medically evidenced, a concept developed by Mary Douglas, the distinguished anthropologist who died last year. This concept has also been important for the support I provide to clinical practice in remote and rural communities.

Robinson posed questions for further research, for example: (1) how to manage severely physically ill patients who resist nutritional treatment; and (2) what is the best model of cooperative care between medical and specialist psychiatric services. In my experience, any request for medical care of these patients must be very carefully defined, usually circumscribed to stabilising metabolic problems. Nasogastric refeeding is not required for that, however self-evident the case might seem for rapidly improving poor nutritional state (it does not directly stabilise a patient's illness and might instead produce other medical problems, as I have observed and Robinson has indicated, as well as to adversely affect the therapeutic alliance).

Addressing his question on ‘how to manage severely physically ill patients who resist nutritional treatment’, my experience suggests that it is important to distinguish between what is being ‘resisted’: normalising metabolism, restoration of metabolic rate in particular (since this directly affects cognition, mood and exercise intolerance), or the additional caloric requirement to improve absolute weight gain or BMI, which frighten these patients. Teenagers often develop anorexia nervosa in response to otherwise unaddressed, perhaps previously unrecognised, psychological distress (problems that might have first resulted in compensatory overeating and excessive weight gain). So nutritional treatment addressing metabolic rate, and thus general well-being, is a far more readily agreed first treatment goal between the patient and their professional carer. Securing collaborative care is an unarguable vital step towards eventual recovery.