Use of risk assessment instruments to predict violence and antisocial behaviour in 73 samples involving 24 827 people: systematic review and meta-analysis
BMJ 2012; 345 doi: https://doi.org/10.1136/bmj.e4692 (Published 24 July 2012) Cite this as: BMJ 2012;345:e4692All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
We used the number needed to detain (NND) statistic (1) to describe the consequences of seeking to prevent violence by admitting people to psychiatric hospitals.(2) Because it is the inverse of positive predictive value (PPV), NND can also be used as a measure of the predictive validity of the various methods now used to assess the risk of psychiatric patient violence.
We reported an NND of 6. Fazel et al (BMJ 2012; 345:e4692), in their review of violence risk assessment instruments, report an NND of 2. The authors explain the difference in terms of the characteristics of the papers under review. For an explanation, there is no need to look beyond the properties of the NND statistic and the base rate used to calculate it. Fazel et al calculated their NND using a base rate of 32%. We calculated ours using a base rate of 9.5%. Because it is the inverse of PPV, for a given method of assessing risk NND rises as the base rate falls. The size of the change is also what one would expect given this difference in base rates.(3)
Two things follow. First, the data that Fazel et al report become more consistent with those reported by others. A fall in the NND that is not explained by base rates suggests an improvement in the predictive validity of violence risk assessment. In fact, the figures in Fazel et al for area under the ROC curve (AUC) reflect predictive validities similar to those reported in our paper and elsewhere. Second, and despite the authors’ explicit advice to the contrary, concern over psychiatric patient violence will lead some to conclude that an NND of 2 is a price worth paying, especially if detention leads to treatment. But 2 may not be the relevant number.
The implications of an NND depend on what the base rate refers to. The studies that Fazel et al used to generate an NND of 2 had follow-up periods (39 months) that are much longer than the periods over which the risk assessments of out-patient clinicians are usually intended to apply. Some of the studies used broad definitions of violence that did not require victim injury. By way of contrast, the NND using the 6-month violence rate (with a weapon or causing injury) in the clinical setting of the CATIE study (4)is 15.(3) If psychiatric detention is to be used to improve public safety by detaining people using present methods, someone is going to have to build more hospitals.
Alec Buchanan PhD MD FRCPsych
alec.buchanan@yale.edu
1. Fleminger S. Number needed to detain. Br J Psychiatry 1997;171:287.
2. Buchanan A, Leese M. Detention of people with dangerous severe personality disorders: a systematic review. Lancet 2001;358:1955-9.
3. Buchanan A. Risk of violence by psychiatric patients. Psychiatric Services 2008; 59: 184-190.
4. Swanson J, Swartz M, Van Dorn R, Elbogen E, Wagner H, Rosenheck R, Stroup T, McEvoy J, Lieberman J. A national study of violent behavior in persons with schizophrenia. Arch Gen Psych 2006; 63: 490-499.
Competing interests: No competing interests
Re: Use of risk assessment instruments to predict violence and antisocial behaviour in 73 samples involving 24 827 people: systematic review and meta-analysis
Dear Editor,
Fazel and associates report a meta-analysis of violence risk assessment that included 24 847 people, of whom 5879 (23.7%) re-offended over an average follow up of four years. They concluded that 2 to 4 people would need to be detained on the basis of risk assessments to prevent one act of future violence. However, this low number of people needed to detain is a result of the high base rate of 23.7% over the long period of follow up as much as the ability of the risk assessment tools to discriminate between those who will and will not be violent. Severe acts of violence, for example those that might attract 4 years in detention are actually quite rare. Hence, the number of people needed to detain to prevent rare more severe violence is very much greater than 2 to 4 suggested in the recent paper. For example, an analysis of the performance of optimal violence risk assessment estimated that the number of patients with schizophrenia that would be needed to be detained in order to prevent 1 homicide was 2,500 and was 35,000 if that homicide was of a stranger rather than an acquaintance or family member (1).
Moreover, the author’s comparison of violence risk assessment with prognostic tests for vascular disease is unhelpful. In no jurisdiction are patients with vascular risk factors involuntarily detained in health camps for their vascular health. Risk scales for vascular disease are used as a basis for reasoned discussion with the patient about the need for lifestyle change and further investigations. It would be more appropriate to compare the psychometric properties of violence risk assessment with the results of diagnostic tests, which can, like a violence risk assessment lead directly to invasive and significant changes in the patient’s management. In a comparison between diagnostic tests such as angiography, and violence risk assessment, the risk assessment tools would come a very poor second.
The literature on violence risk assessment, and the enthusiasm for its use continues to grow. However, it should be remembered that violence risk assessment and associated detention is a flawed method for preventing rare and severe violence.
1. Large MM, Ryan CJ, Singh SP, Paton MB, Nielssen OB.
The predictive value of risk categorization in schizophrenia.
Harv Rev Psychiatry. 2011 Jan-Feb;19(1):25-33.
Competing interests: I have received speakers fees from Astra-zenica to present my research on risk assessment.