And much to enjoy!

Machine learning will shortly come to be used by psychiatrists and other physicians in clinical practice (Steele and Paulus, 2019). This prospect will dismay many of us, yet there is an obligation on our part to have some notion of what it is and what it may offer. Machine learning can be seen as a resource additional to our own brains. As one form of artificial intelligence, a computer takes in information, learns from that, identifies patterns and then makes decisions. So it is a prosthesis to use when faced with a difficult clinical situation. In an unusually challenging paper from Adelaide and Münster, Cearns et al. (this issue) begin with a clinical scenario where machine learning could help decide what treatment should be offered to a person with bipolar disorder who is at high risk. For the present writer, what follows in that paper would need an intensive tutorial to become comprehensible. But readers should examine what the authors have set out, so that the deeper principles can start to be appreciated. It is both admirable and reassuring that these authors urge clinicians using machine learning ‘to apply thought [sic] to clinical context’. This goes some way to assuage concerns that machine learning will lead to the birth of an Orwellian College of Psychiatric Robots.

An issue of close relevance to practice is whether persons on clozapine really do need their white cells counted every month. It is costly, an appreciable burden on service resources and often bothersome to patients. Nielssen (this issue) examines the question from every angle, particularly with a view to ensuring the best possible advice to our patients. Adroitly, he does not give an explicit answer, but persuasively calls for a reduction in barriers to use of the most effective antipsychotic to date. It would be a major advance if his contribution to this Debate were eventually acted on.

Could the diet and body mass index of patients with bipolar disorder influence their response to nutraceuticals? Ashton et al. (this issue) find only an uncertain association. They correctly acknowledge many limitations in the design and analysis of their trial. In short, one could say the answer to their original question remains unknown. Diet and body mass are central issues in one of the most difficult of all chronic diseases, anorexia nervosa. An internationally eminent clinician in this field (Russell, this issue) spells out the obstacles we currently face in its treatment. Readers will specially appreciate the last two sentences in her Commentary.

This issue carries several papers that are epidemiological in that they examine morbidity in relation to the environment, using either treated populations or community samples. From the original observation by Faris and Dunham (1939) in Chicago, we have known for some 80 years that psychosis has a higher prevalence in inner cities. Now, in a quite enthralling study from Amsterdam, Lemmers-Janssen et al. (this issue) have examined a component of social functioning in psychosis in relation to urban upbringing. But look closer: the authors measure the elusive entity of trust in their subjects, taking it as an indicator of social functioning, often impaired in psychosis. They then use functional magnetic resonance imaging (fMRI) to examine activation of the amygdala in relation to both urban upbringing and a measure of trust. Surely this is an example of progress in knowledge: relating brain function to environmental exposures. They find urban upbringing was not linked to baseline trust but to different pattern of activation in the amygdala. We should hope for more work of this type because it begins to throw light on brain function and environmental exposure in severe mental illness. These are big issues.

In a modestly sized sample of healthy young adults, themselves derived from a community-based prospective study of infectious diseases, Beilharz et al. (this issue) find that traumatic exposures in childhood were followed, presumably causally, by multiple impairments in physical and mental health during early adulthood. One impairment of particular note was autonomic dysregulation, as shown by heart rate after stress. Here is yet more evidence that adverse childhood exposures have an enduring toxic effect. If the ultimate gift of epidemiology is prevention, and if the attributable risk from childhood abuse is so pervasive, a much greater effort towards intervention is essential, even if it is perplexingly elusive.

A more readily achievable intervention is identified by Nielssen et al. (this issue). For 8 years, they have been examining some 2389 homeless people in Sydney, a sizeable series. As one would expect, about half were psychotic. The present study is a comparison of the latter with those who were not psychotic. Almost a third of the psychotic people had been released from prison into homelessness. Almost as many had lost tenancy of public housing. These findings are just so disturbing in our Clever Country, justifiably proud of its achievements elsewhere, such as in the Olympics. Nielssen and his colleagues end with the very modest assertion that there is a need for more supported housing. Many of us think this intervention, while itself straightforward, is unlikely to happen in the present social climate. We will continue to have very disabled and disadvantaged men and women on our city streets, people with brains severely damaged by long-standing psychosis and its sequelae.

A major difference between psychiatry and general practice is that subsyndromal disorders present mainly in the latter. We meet only a highly selected portion of all who are unwell. But distress and impaired function are substantial in states that fall short of being ‘a case’. Arithmetically, at the population level, the overall burden of morbidity is much greater than in the more severe cases who reach us. This epidemiological fact is under-recognised in the education of medical students. The contrast is well illustrated by Oh et al. (this issue) in their study of 6640 older persons in South Korea. Using reasonable diagnostic criteria, they found subsyndromal depression was 2.4 times more common than syndromal depression, while its incidence rate was 5 times higher. These authors were able to identify some of the attributes that differentiated the two groups. Such information is relevant in medical education and training.

In a remarkable longitudinal study of 2352 persons with anxiety disorders, Bokma et al. (this issue) from Amsterdam succeeded in examining their cohort on four occasions over 6 years, achieving a truly admirable final contact rate of 75%. These data enabled them to see how symptoms and disablement changed over the 6 years. Bokma and colleagues see merit in constructing a staging model for anxiety disorders. It allowed them to predict the course of the disorder. That could be translated into clinical practice where the prognosis is sought, an exercise that is now often overlooked by the clinician, patient and family. Figure 3 in that paper is most engaging. In common with many others, these authors found that comorbidity worsened the prognosis. Here is a finding that holds much attraction for further research. Why should this be and what does it mean?

Readers will enjoy the ideas on anti-psychiatry by McLaren (this issue). What does anti-psychiatry hold to be true and what is the evidence? Importantly, he asks if one can be critical of psychiatry but not align oneself with anti-psychiatry.

The fall-out continues from the United Nations provocative document on the rights of persons with disabilities (United Nations Human Rights Council, 2017). Newton-Howes and Gordon (this issue) offer a Viewpoint in which they draw attention to the neglect of the service users’ perspective. A second omission is that nothing is offered on how to translate these Rights into practice.

One of this Journal’s aspirations is that it can help make psychiatry enjoyable. The contributors to this issue have done just that.