In this editorial we define ‘productivity’ and ‘efficiency’ in a mental health service context, outlining the key challenges to measuring these in practice. We attempt to bring clarity of thought to this often debated, but rarely researched area.

Machine learning methods are being increasingly applied to physical healthcare. In this article we describe some of the potential benefits, challenges and limitations of this approach in a mental health context. We provide a number of examples where machine learning could add value beyond conventional statistical modelling.

Out-of-area (OOA) placements occur when patients cannot be admitted to local facilities, which can be extremely stressful for patients and families. Thus, the Department of Health aims to eliminate the need for OOA admissions. Using data from a UK mental health trust we developed a ‘virtual mental health ward’ to evaluate the potential impact of referral rates and length of stay (LOS) on OOA rates. The results indicated OOA rates were equally sensitive to LOS and referral rate. This suggests that investment in community services that reduce both LOS and referral rates are required to meaningfully reduce OOA admission rates.

The schizophrenias are uncommon before the age of 14 but incidence/prevalence figures are lacking. The 1-year incidence, clinical features and short-term outcomes in childhood-onset schizophrenia spectrum disorder were evaluated via the Child and Adolescent Psychiatry Surveillance System. Fifteen children with a provisional diagnosis were reported. Outcome data were obtained for 12 individuals, 8 of whom met the diagnostic criteria, equating to an estimated incidence of 0.21/100 000 (95% CI 0.08–0.34). Delusions and thought disorder were a more consistent predictor of ‘caseness' than hallucinations. Illness outcomes at 1 year were generally poor. Childhood-onset schizophrenia appears to be a rare but serious disorder.