Dear Sir,

Interpersonal violence occurring within the family and intimate relationships is an important challenge for public health globally and increasingly recognised in the context of mental health services in particular. The Covid-19 pandemic and national lockdowns have made family violence more visible to wider society and highlighted the care and support needs of all those affected. Labrum and Solomon (L&S)'s analysis of police data (‘Serious mental illness and incidents between adult children and parents responded to by police’, Psychological Medicine 1–10) highlights the significance of mental health symptoms in incidents of family violence where police are called to attend. There are important aspects to the interpretation of L&S's results which are relevant for UK mental health services, but also methodological queries which readers might helpfully consider in their interpretation of the findings.

We are pleased that this study is able to shed light on the relevance of verbal aggression for assessing the risk of later physically injurious harm. The paper restates the importance of clinical assessment and management of non-physical forms of violence and abuse, which remains a challenge for the delivery of mental health services and the training of professionals. The focus on the contextual characteristics of incidents in the study, driven by a qualitative methods approach, is also to be welcomed. In the study, the authors took all police attendances for family violence occurring in Philadelphia, comprising incidents of violence from an adult-child towards a parent and episodes of violence from a parent towards their adult child. All events were then coded for the presence of ‘serious mental illness’ based on police observations and written text entries suggesting the offender was affected by a mental health condition (e.g. ‘psych problems’), or where specific symptom, treatment or mental health status were mentioned (e.g. ‘paranoid’, ‘medication’, and ‘committed’).

We fear that this approach may correspond poorly to many widely accepted and/or current definitions of serious mental illness used in psychiatric research, including those based on ICD-10 or DSM-V criteria, or indeed previous research on family and partner violence, and studies cited in the paper itself. Therefore, the results may not necessarily allow for generalisation to patients with lived experience of severe and enduring mental health problems. This approach is also unable to untangle possible differences in family violence by diagnosis – especially relevant, given recent evidence of significant differences in associations with family violence between diagnostic groups (Yu et al., Reference Yu, Nevado-Holgado, Molero, D'Onofrio, Larsson, Howard and Fazel2019).

The authors suggest that risk assessment and management strategies should focus on the characteristics of offenders beyond serious mental illness. But by adopting an extremely broad definition of serious mental illness, this objective becomes difficult, because it is impossible to determine which groups of patients with ‘serious mental illness characteristics’ are most relevant, and which groups are not – this is important because a substantial proportion of perpetrators of partner violence in the population as a whole may experience symptoms of depression and anxiety, which may be less relevant to the occurrence of violence per se, compared to symptoms of irritability, agitation, anger, delusional jealousy, etc. which may present more risk. By conflating more and less severe symptoms, we fear that this study may lead to incomplete and unclear conclusions about this. This approach also complicates the understanding of the important distinction between incidents and patterns of violence in this population – it is recognised that in the general population, incidents occur against the backdrop of a pattern of abusive and harmful behaviour, which this study is unable to capture; it may be this pattern which is more relevant for later risks. It is unfortunate that the details of text-processing undertaken to generate the key variables examined in the study are left undefined – transparent reporting of measurement approaches become increasingly important as the technological complexity of the methods and the underlying data increase. A growing amount of research on mental health topics make use of police data – in this respect, clear description of any text-processing rules, diagnostic definitions which are used (or not used) where available, and descriptions of patterns of missing data, will be essential in integrating this literature with the broader psychiatric research on violence, in line with reporting statements such as STROBE and RECORD. Use of routinely collected text data requires close attention to the processes by which such information is collected – whether this is in health care or police/criminal justice settings, factors such as the gender and ethnicity of the recording professional, level of experience, degree of specialisation, socio-demography of family members are all likely to be highly relevant, and deserve investigation, particularly given concerns regarding unequal practice and differential and less positive experiences of minority ethnic groups in mental health services and the criminal justice system.

Finally, it is important to explicitly acknowledge the limitations of police informants in this study, and the impact this may have on the interpretation of L&S's results. Evidence suggests that family members caring for people affected by severe mental illness may be reluctant to involve the police in the event of violence, due to concerns about the consequences for their relative (Onwumere, Parkyn, Learmonth, & Kuipers, Reference Onwumere, Parkyn, Learmonth and Kuipers2019). This is consistent with evidence reported in L&S’ paper that only 5% of the study sample (327 individuals) actually met their criteria for serious mental illness, discrepant with their local population estimates. We think that incidents of family violence which are attended by police nevertheless reflect an important subset of the overall burden of family violence affecting the population as a whole. However, the focus on families dealing with violence and mental health problems and who remain in the same household, but who may not be on the radar of police services, must not be overlooked during this unprecedented moment.