Re: Economic comparison of early intervention services with standard care: Flawed model of the costs of homicide

A. Pelosi

DOI: https://doi.org/10.1111/eip.12495

2018-06-01

Early Intervention in Psychiatry

Abstract:I am afraid that I have recently noticed a flaw—a very serious flaw—in the economic analysis by Park, McCrone, and Knapp (2016) that was published in this journal last Autumn. This has been available online since 2014 and has already been influential in writings about health care policy (Arango et al., 2017; Knapp et al., 2014; McDaid, Park, Iemmi, Adelaja, & Knapp, 2016). The decision analytic model in regard to homicide compares early intervention services with so-called standard care that “was assumed to be a specialized community mental health service which did not include any component particularly targeted at people with early psychosis”. The probability of homicide in any 1 year over a period of 10 years for those initially treated by early intervention services is estimated at 0.00011. Their equivalent figure for patients treated by a standard community mental health service is 0.0017—a staggering 15 times greater. It is obvious that this is entirely incorrect. There is nothing in psychiatry that is 15 times better than anything else. These figures appear to come from an important systematic review by Nielssen and Large (2010). Ten investigations within their meta-analysis allowed Nielssen and Large to estimate homicide rates in never treated people with psychotic illnesses and compare these with homicide rates in those who were receiving treatment. Treatment consisted of “standard care” of various types and in various parts of the world between the 1950s and the early 2000s. This comparison yielded the 15 times difference in estimations of homicide rates. These investigations did not compare different ways of organizing psychiatric services and had nothing whatsoever to do with the short-term approaches of the early intervention movement to the management of schizophrenia and other psychoses. There are other data that could have properly informed Park and colleagues’ model. They have partly based their study on the effectiveness and costs of the Lambeth Early Onset Team. The randomized controlled trial comparing the Lambeth team with standard care from community mental health services reported on rates of violence during an 18-month follow-period. Adverse incident records revealed that 17% of the early intervention treated patients and 19% of the standard care group were involved in a violent act towards a member of staff; 20% of participants in both of these groups were violent towards another patient or a member of the public (Garety et al., 2006). Why would there be near identical overall rates of violence but spectacularly different rates for the most severe and rarest type of violent behaviour? There is an even more important source of information about homicides by people with severe mental illness. If Ms. Park, Professor McCrone and Professor Knapp can bear it, they should learn more about these tragedies via a simple Google search for “HSG (94) 27.” This was the Department of Health circular that mandated a publicly available inquiry into each homicide carried out by a person under the care of psychiatric services in England. They should read just a selection of these inquiries and then ask themselves what it is about the few years of special work by early intervention teams that would make patients 15 times less likely to kill someone over a decade than patients treated from the start by community mental health teams. I would argue that certain features of generalist teams minimize the risks of extreme violence. Unlike some specialists, they do not exclude patients with psychotic illnesses who also have drug problems and/or alcohol problems and/or personality disorders (Sainsbury Centre for Mental Health, 2003); they build up experience of a wide range of mental states other than psychosis that can impact on the likelihood of serious violence; and they do not create an unnecessary “critical period” for patients by setting an arbitrary time frame for their clinical involvement. Of course they are never to know when they really have prevented an individual homicide by one of their patients—because they have prevented it. They can only know when they have failed to prevent this most catastrophic outcome in the whole of clinical practice. This is why it seems so wrong that this economic model builds in an assumption that rates of homicide in psychotic patients receiving ordinary multidisciplinary psychiatric care are similar to the rates in never treated people with psychoses. Park et al. (2016) inform us that the modelling in this paper was “developed through discussions with clinical experts in the mental health field, some of whom have been actively involved in the initiation, implementation and monitoring of [early intervention] services in England for some years”. They should also have spoken with ordinary mental health clinicians who could have provided a more realistic understanding of the rare instances of extreme violence by their patients. I do hope that our health economist colleagues will correct this flawed economic model of homicide by people with psychotic illnesses. Received: 26 April 2017 Revised: 24 July 2017 Accepted: 29 August 2017