We open with some major and significant randomised controlled trials exploring interventions in resource-poor societies. Two papers by the same team writing in the Lancet explore the use of non-specialist ‘lay counsellor’ health workers to deliver brief psychological interventions for excessive alcohol consumption and depression in India. Male harmful drinkers in ten primary health centres were randomised to receive either enhanced usual care (EUC), or EUC and Counselling for Alcohol Problems (CAP). The active intervention produced significantly greater remission and abstinence rates, with an incremental cost per additional remission of $217. A Healthy Activity Program (HAP) for moderately to severely depressed men and women was evaluated in a different cohort, with a parallel design of EUC compared with EUC plus therapy. Once again, the active intervention produced significant improvements, with reductions in depression symptomatology and remission, as well as decreased illness consequences such as days out of work and intimate partner violence. Both interventions were reported to be acceptable to patients and practical for delivery in primary care settings.

We always hope you enjoy the free journal that comes with your Kaleidoscope subscription. A paper in this month's issue (pp. 429–436) showed that most scales assessing risk performed no better than the clinician/patient predictions following self-harm; this provoked a lively discussion on the journal's Twitter feed. A new paper by Seena Fazel's team explores their utility in criminal justice settings and forensic psychiatry. The authors note how such tools are used to inform critical aspects of patient management such as in-patient detention and discharge, custodial sentencing, parole, and post-release monitoring. This is despite a lack of reliable validation on predictive accuracy, especially in important groups such as women, ethnic minority populations, and those motivated by religious or political extremism. Furthermore, they find the literature is marred by significant publication and authorship bias, and suggest that better-quality data will allow better matching of relevant tools to clinical contexts. This is best exemplified by assessing the balance between optimising false positive v. false negative findings: highly sensitive tools (with low false negatives) may be optimal where ‘protecting the public’ is seen as key, whereas highly specific ones might best protect prisoner and patient rights and interests. Assessment tools have had accusations of implicit discrimination levelled against them, as they commonly capture sociodemographic data – age, gender, ethnicity, immigration status – that risk profiling and perpetuating stigma. But should this information be excluded, especially as some data may improve predictive accuracy? The analogy of racial profiling at airports is put forward: if this helps a limited, but highly contentious, screening resource prevent more atrocities, is it warranted? It's clearly a charged debate, and perhaps that is part of the problem, balancing emotion and fairness with science. In the absence of robust data, we walk the fine line between coarse variables that may perpetuate discrimination, and the risk of their politically driven removal.