The findings of Flynn and colleagues^{Reference Flynn, Ibrahim, Kapur, Appleby and Shaw1} are important, but certain interpretations may be incorrect. It is potentially misleading to draw any conclusions on homicides by people with schizophrenia based on percentages of the total if overall rates show changes. The authors demonstrate that overall age-standardised homicide rates initially rose then have fallen over time in England and Wales. But there is no corresponding figure which shows rates over time among people diagnosed with schizophrenia. This is an important omission. An early study showed remarkably stable rates for mentally disordered homicide offenders in England and Wales over time, based on court adjudications.^{Reference Coid2} Flynn and colleagues have better data to test this possibility. If rates for schizophrenia remained the same while the overall rates fell, it cannot then be concluded that substance misuse or any other unknown, complex factors had ‘driven’ anything. It would then mean their findings reflected, first, the failure of mental health services despite their changing configuration to have any impact whatsoever on homicide by persons with schizophrenia; and, second, decreasing willingness over time by psychiatrists to offer a bed when the offender appears in court, or to conclude that their responsibility is diminished. More persons with schizophrenia would then inevitably be sent to prison.

Unfortunately, persons with schizophrenia who have killed do not suit prevailing UK service provision of home treatment or early intervention. Courts require robust assurances that the public will be protected from them in future. There is also a worrying trend for secure services to return their patients to prison, where they are lost to follow-up. The authors rightly point out that it is difficult to obtain a secure bed in the first place. It could be that secondary diagnoses of substance misuse and personality disorder have increasingly provided the convenient excuses necessary to reject these patients for psychiatric treatment. Research findings on the impact of schizophrenia on violence and the role of substance misuse have been unhelpful to clinicians and are inconsistent, confusing multiple associations of an unspecified nature with causation.

The authors are to be complimented for not relying solely on diagnostic labels with doubtful temporal proximity to the homicide and, most importantly, for demonstrating that the large majority of persons with schizophrenia who killed had active psychotic symptoms at the time. These can be causative factors for violence,^{Reference Ullrich, Keers and Coid3} whereas with diagnostic labels, it is impossible to tell.

Could they provide comparative population rates over time for persons with schizophrenia compared with all others? It might then be possible to infer whether population risk and protective factors for homicide perpetration are likely to be similar or very different.