The warlike events resulting from terrorism in London on 7 July 2005 have again shown the importance of enhancing human resilience and give special relevance to June's issue of the Journal. In a marvellous overview, Professor Wessely (Reference Wessely2005) gave us his thoughts concerning psychological trauma, modern psychiatric trauma concepts, and the emergence of new syndromes, especially in military settings.

Contrary to Professor Wessely, we are convinced that longitudinal selection provides considerable advantages for psychiatric risk management. Despite the unsatisfactory American experience with personality testing during the Second World War (Reference Jones, Hyams and WesselyJones et al, 2003) our main field of activities is cohort-based psychometric screening and prediction models. In 2002, the Swiss Armed Forces assigned us to investigate new methods to predict psychiatric disorders in servicemen. At first we were sceptical that such a task could be fulfilled. However, we found prediction models to forecast outcome in emergency patients in the medical literature (Reference Tuhrim, Dambrosia and PriceTuhrim et al, 1988). Furthermore, personality seemed to play some part in the outcome of somatic disorders (Reference EysenckEysenck, 1988) and suicide seemed predictable from demographic variables (Reference Holinger, Offer and ZolaHolinger et al, 1988). Consequently we investigated how these techniques could be transferred to psychiatry.

In a small preliminary (2002) study we screened 3000 recruits on their first day of basic training and followed their medical records for psychiatric problems. Based on clinical-epidemiological knowledge, logistic regression helped us to create a robust multivariable model. Since 2003 the model has been used by the Swiss Armed Forces for recruitment. The model compares each conscript with about 30 000 servicemen. As a result, subsequent psychiatric discharge on the grounds of receiving an ICD-10 (World Health Organization, 1992) diagnosis was significantly lowered by a factor of 3 (or 72%) compared with unscreened recruits. The personality trait of the conscripts did not have any impact.

We are convinced that our prediction model can be successfully adapted to any military service model and operational setting. Therefore, we believe it is too early to bid farewell to psychiatric screening systems in medical risk management.