Who’s in a cohort? Identifying priority persons to tackle serious violence

This article is part of a series that make up Thames Valley Violence Reduction Unit’s Strategic Needs Assessment. To see the rest, visit tvvru.co.uk/strategic-needs-assessment

A key part of an effective public health approach is our ability to target our activities where they will have the most impact. It means that instead of waiting for those involved in serious violence to come to us, we must get better at identifying those individuals and groups that need support, and more quickly providing evidence-based interventions that can reduce the risk of them coming to or causing harm.

This is not a new approach for “policing” and the criminal justice response to tackling serious violence either. We are used to focussing our efforts where it will have the biggest impact. However, we need to do better.

This year we need to improve how we identify those individuals who are significantly more at risk from serious violence and those more likely to commit it. This module of our SNA will go only so far – much of what we need to do needs to be done at an individual level, making better use of our policing data and the data we share across our partnership.

Thames Valley Together will play a huge part of this, enabling us to spot even earlier when things are taking a wrong turn, and helping create evidence on the “what works” so we can work better with those at risk or whose lives have already been affected by serious violence.

This module of the SNA therefore provides an overview about what we know about those involved in serious violence. There’s much more to do, and so if you have an idea about something you think we should look at, or some feedback on what we’ve done so far, let us know.

Now, to the data…

A numbers game

Our data team in the Violence Reduction Unit is called the “Data and Targeting” team for an important reason – targeting (using data) is really important! Even if it were possible, it would be inefficient for our efforts to be spread equally across every single person across the Thames Valley. We can use data to help us focus much more on those individuals where focus is likely to make a difference.

One basic principal may be that we want to focus on those individuals who cause or experience the most “harm”. Harm is a subjective concept, but we have ways of viewing it through data such as through measures like the Cambridge Crime Harm Index, which aims to weight offences based on the time someone convicted of an offence is likely to spend in prison.

The more harmful the offence, the longer we may expect to see someone in prison. It aims to prevent us treating all crime the same – there is a huge difference in the harm caused due to a homicide compared to a shoplifting, let alone the difference in cost.

Applying the Cambridge Crime Harm Index to our data on serious violence between 2017-2021, we can see a dramatic concentration in the number of individuals who cause or experience harmful crime. Fewer than 4% of the total number of people who were recorded as a suspect in a serious violence offence between 2017 and 2021 were “responsible” for over 20% of the total harm that occurred during that time period. Nearly 75% of all crime harm occurs because of fewer than 40% of individuals.

This concentration is similarly observed when we switch our attention to victims. Around 20% of all of the serious violence harm that was experienced by individuals between 2017-2021 was experienced by fewer than 5% of the total number of individuals that were victims in a serious violence offence in that time period.

We also know that often some of the most harmful and harmed individuals are those who are both suspects and victims in serious violence. We looked at the phenomenon (often known as the “Victim/Offender” overlap) back in our 2020/21 SNA, and this year using a larger dataset we’ve identified that just 5% of the total number of individuals who were involved in serious violence between 2017-2021 were both a suspect and victim at least once during that time.

It may also help us to understand how many individuals have many individuals have been involved in multiple offences, or have been victimised or offended against multiple people. Between 2017-2021, 18% of all people who had been suspects and 5% of all people who had been a victim were involved in more than one offence. A small proportion (3%) of individuals had been suspects in four or more serious violence offences.

During the same time period, 19% of suspects committed violence against multiple victims, and 26% of victims were offended against by multiple suspects. This could be through multiple persons involved in the same offence (e.g. where multiple suspects committed a robbery against a single victim), or in multiple separate offences.

Finally, whilst we know from the overview section of the SNA that individuals are most commonly involved in serious violence aged 18-25, there are some slight differences in the age profile for serious violence involvement across the Thames Valley’s Community Safety Partnerships (CSP).

Milton Keynes, with its predominantly youthful population sees a higher proportion of suspects and victims in younger age groups, whereas Bracknell, Slough and Vale of White Horse have a more right-skewed distribution – particularly with higher numbers of older victims.

All of this information is helpful, and undoubtedly can be used to help us prioritise those we work with. However, it focusses on those who are already involved in serious violence. Part of an effective public health approach comes in our ability to focus on those at a secondary level – those individuals who are at risk of serious violence but who have not as yet been “exposed” to it.

To do that, we need to take a deeper look at our data to identify less visible risk factors and indicators that help us flag up those who would benefit from activity to prevent them becoming involved in serious violence, in simple terms, those with “risk factors”.

Factoring in risk factors

To help us understand what risk factors exist in our population, it helps first to understand how many individuals are known to police before becoming involved in serious violence.

For all individuals who became suspects or victims in serious violence during 2021, in most CSPs over half had been a suspect or victim in an offence in the two years previous to their serious violence offence, or had been subject of a risk management occurrence or child/adult protection occurrence.

There are CSPs where this is not true – particularly of victims, and even where it is true, a sizable minority do not appear to have come to police attention in the two years before their latest offence.

Where it is true, we can start to look at commonalities in the histories of suspects and victims that might help us understand promising intervention points that would help us proactively identify those at the greatest risk of serious violence and working to prevent that becoming a reality.

Using our policing data in increasingly advanced ways, we are able to start to look at this “person journey”. As in our previous SNA, police custody remains a leading commonality in both the histories of those involved in violence as a victim or suspect – over half of all suspects in serious violence during 2021 had previously been through police custody within the Thames Valley. Previous offending in violence was also relatively common, as were markers indicating risk of violence. Nearly 25% of all suspects in serious violence had a marker for drug use/supply, and over 20% had warnings for mental health.

During 2022/23 we will work to create a “live” risk model through our Thames Valley Together data project, which will bring together data from across the partnership to generate daily insights as to the common histories of those involved in serious violence. This will help us ensure we are delivering interventions to the right groups, and in the right places.

However, even for some of the most common risk factors, there is significant variation across CSP areas.

In the Vale of White Horse, nearly 50% of all suspects in serious violence during 2021 had previously been a suspect in a violence with injury offence, but that figure falls to around 25% in Windsor and Maidenhead.

Overall fewer numbers of suspects in serious violence during 2021 had been victims of violence with injury, but again this proportion falls from over 30% in Slough to around 15% in Bracknell Forest.

And whilst over 40% of suspects in serious violence during 2021 in Oxford and Vale of White Horse had a drugs marker or a drug possession offence, in Wokingham that figure drops to around 23%.

These are just a selection of the thousands of potential events/data points we could use to help us understand those at risk of serious violence.

Luckily, there is a growing body of evidence that helps us prioritise which data we need to collect, and how we need to apply it. However, to look at risk factors properly we will need to bring together data wider than just police and crime information. This needs a secure, scalable mechanism to share information across the partnership, and analytical tools and resources that ensure we can apply this emerging evidence to our practice.

To finish off, let’s go back to where this all started…

The Serious Violence Strategy

In April 2018, the government published their Serious Violence Strategy. This comprehensive document brought together a range of evidence and insight around what was driving an increase in serious violence, and what could be done about it.

A year later, with the first funding awarded to the Thames Valley Violence Reduction Unit, we used the Serious Violence Strategy to create some basic mapping of where risk factors for serious violence existed in the Thames Valley, based on publicly available data, and a small amount of police data. The analysis was basic, but it helped us start to identify some places where we may want to focus our attention (and our funding) in that year.

Since then our team has grown (from one – not really a team!), and our capabilities have developed significantly. We now have access to a data science platform and a purpose built data environment to work on partnership data, which we aim to bring into live use during 2022/23.

To celebrate this, we thought we would go back to replicate some of the work we attempted to do in 2019, and see how much more we could do now.

We picked out six key risk factors from within the Serious Violence Strategy, which ended up as five risk factors that we then looked for across all individuals that had been involved in serious violence during 2021 in the Thames Valley.

Those risk factors were:

We found that nearly 60% of suspects had two or more risk factors from that list, with 45% of victims that had two or more risk factors before their latest victimisation in 2021.

As might be expected, younger people generally have fewer risk factors than those ages 25+. The lines below show how many individuals in each age category have the identified risk factors.

To really make this information helpful however, we need to be able to do more. To know how good these risk factors are at helping us identify those who are at risk of and from serious violence, we really need to know how many individuals have these risk factors but don’t go on to become involved in serious violence. To answer that question requires a significant amount more data, and the analytical tools and capability to handle that information. That’s what we will achieve through Thames Valley Together, and we are really excited about all it will enable us to do.

Delivering on the information we find will require us to have strong and effective relationships with those delivering services on the ground, plus the ability to influence and steer commissioning to tackle serious violence. We’ll need strong partnerships.

Why partnerships are important!

As the largest non-metropolitan police force area in England and Wales, the Thames Valley has a diverse geography, with urban areas spread out across a wide area. As a result, it is not uncommon for those involved in serious violence to have travelled from another area of the Thames Valley, or from further afield.

This directly affects our work, and requires our nine community safety partnerships to work closely together, as preventative activity that targets those committing serious violence in one CSP may need to be delivered in another.

Two CSP areas saw over half of those individuals who were suspects in serious violence in 2021 travelling from outside the CSP area, with Windsor and Maidenhead seeing nearly 60% of those identified travelling from outside the Royal Borough. This drops down to fewer than 20% in Milton Keynes.

There is work to do to improve our handling of spatial data, and in particular, the addresses of individuals involved in serious violence. However, preliminary analysis has identified some frequent home address CSP to offence location CSP patterns.

In 2022/23 we will continue to work at the centre of violence reduction activities across the Thames Valley, coordinating efforts and ensuring that we use data to target operations where they will make the greatest impact.

It’s not what you know

There is increasing interest in the opportunities that the use of social network analysis might play in helping us identify those at risk of serious violence. The VRU are working with academic partners to help us better understand the possibilities, including partnering with University College London as part of a STAR bid that is due to report this summer.

Initial research conducted by the team has observed significant clustering of violent offending and victimisation within the Thames Valley.

In a network created of everyone involved in a crime offence as a suspect between 2017 and 2021, individuals who were “connected” to one or more other people who had been a suspect in serious violence were over seven times more likely to have also been a suspect in serious violence. In individuals with three or more people linked to them who had been a suspect in serious violence, this rose to mean that they were ten times as likely to have been a suspect in serious violence as someone who had no connection to a serious violence suspect.

Victimisation was similarly clustered. Again, individuals who had one or more persons in their network who had been a victim were around seven times more likely to have been a victim themselves, and for those with three or more contacts who were victims, they were nearly eleven times more likely to have been a victim. A full findings paper will be released later in 2022.

Social Network Analysis is unlikely to be the single answer to identifying cohorts at risk, but combined with other approaches it will help us provide more contextual understandings of risk, and hopefully help us divert resources to those where we can make the biggest impacts.

Bringing it all together

This section of our SNA did not aim to explain our finished approach to identifying cohorts at risk. We will need to continue to develop our thinking and make the best use of available data throughout the next three years, and we aim to update our SNA with these new approaches as we are able.

Instead, we hope this section has given you an understanding of our approaches, and some idea of how they may fit together.

If you’ve spotted something that you think would be worthy of further analysis, or you’ve got some feedback on what we’ve done so far, get in touch.