'OxMIV' violence risk assessment tool: external validation study. V1.0
Research type
Research Study
Full title
The ‘OxMIV’ violence risk assessment tool: an external validation study in patients referred to Early Intervention in Psychosis services using routine documentation in Electronic Patient Records.
IRAS ID
257332
Contact name
Daniel A Whiting
Contact email
Sponsor organisation
University of Oxford / Clinical Trials and Research Governance
Duration of Study in the UK
0 years, 3 months, 29 days
Research summary
Research Summary
Community mental health teams called Early Intervention in Psychosis (EIP) services assess and treat people with the possible onset of psychotic illnesses like schizophrenia. Although most people with schizophrenia never act violently, developing this type of illness does moderately increase the chances of this, especially if there are also other risk factors such as previous violence or drug use.
One thing EIP services must assess therefore is whether an individual might behave violently in the future. This can help guide what input someone might need. Prediction can be difficult though. Tools can help structure risk assessment, but currently these take hours and are expensive, so are unhelpful for busy frontline settings like EIP services.
A simpler, free assessment tool called OxMIV uses information that clinicians already routinely record, so could more helpfully support risk assessment in EIP services. It has so far been shown to perform well using data about patients in Sweden. More work is needed to see if it is helpful in UK EIP services.
The first stage is to see whether OxMIV is accurate when used to estimate violence in a new population, namely patients presenting to EIP services, or if it needs adjusting. It must first be tested without impacting how clinicians and patients act.
To achieve this, this study will use routine data from electronic records of patients previously assessed by EIP services to calculate OxMIV scores. Police and health records will be checked for subsequent violence. How well OxMIV predicted these events can then be calculated, and if necessary simple adjustments made to better fit the tool to the new setting without changing its core makeup. OxMIV will also be compared to the judgement the clinician made at the time of the first assessment, to see whether adding the tool could have helped.
Summary of Results
Whilst most people presenting to mental health services never behave violently, for a minority this is something that they might need help with as part of their treatment and support. This includes some people with the new onset of symptoms of psychosis, who are supported by specialist services called early intervention services. Whether this is a relevant need for someone is already routinely assessed, but currently clinicians do this by using their clinical judgement or experience alone. This is not always accurate or transparent, and might vary between clinicians.
"OxMIV" is a simple risk prediction tool or calculator that was developed using large sets of data from population registers. It weighs up information about different factors that increase the risk of violence to estimate the overall chances of this happening. This might help clinicians consider the relevant factors more accurately, and support their overall clinical assessments. However, it needs to have its accuracy tested in different populations. So we aimed to validate and update OxMIV in early intervention services, and consider its potential benefits as an addition to clinical assessment.
Data about individuals assessed by two UK early intervention services was included. Electronic health records were used to obtain the information needed for OxMIV from routine clinical assessments. The risk judgements made by assessing clinicians were also recorded, so that these could be compared with OxMIV. Police and healthcare data on violence in the 12 months after assessment was then checked to see how well the tool would have predicted this. This was all done using electronic records, and did not impact clinical assessment or practice.
Of 1,145 individuals presenting to early intervention services, 131 (11%) behaved violently in the next 12 months. OxMIV performed well on a measure of how it can tell apart those who do and do not go on to behave violently (telling them apart 75% of the time). The data was used to update the tool for this specific population, so that the average prediction made by the tool matched the average rate of violence. OxMIV had 71% sensitivity, which means 71% of those who behaved violently would have had an increased risk score with the tool. This compared favourably with the sensitivity of clinical assessment alone, which was 40%.
Taken together, these results tell us that OxMIV could potentially have a role in improving assessment processes in early intervention services. One way the tool could improve practice is to support the identification of more needs around violence risk so that extra, non-harmful support for things that might reduce risk (such as support for substance misuse) can be assigned more accurately to those most likely to benefit, in what is called a precision medicine approach. The next step for work with the tool is to understand how it can best be used in practice in these services.
REC name
London - Bromley Research Ethics Committee
REC reference
19/LO/0498
Date of REC Opinion
18 Apr 2019
REC opinion
Favourable Opinion