Research and analysis

Violence Reduction Units year ending March 2024 evaluation report

Published 11 June 2025

Applies to England and Wales

Executive summary

Introduction

Violence Reduction Units (VRUs) operate in 20 police force areas (PFAs) with the aim to prevent and reduce serious violence (SV) by taking a whole-systems and early intervention approach. VRUs are part of the SV Fund, which also supports more policing focused Grip activity to tackle violence through activity in ‘hotspots’. This report evaluates the VRU programme for the 2023 to 2024 period, marking the fifth year of operation for most VRUs.

Methodological approach

The 2023 to 2024 evaluation updated the methodology from the 2022 to 2023 year, incorporating a new Theory of Change, systems mapping, and an ethnographic case-study approach to understand intervention implementation and outcomes. Quasi-experimental designs were used to estimate the effects of VRUs on key violence measures relative to non-VRU areas.

Impact on violence outcomes

Results from the primary quasi-experimental impact evaluation focused on all population measures, though not statistically significant at the 5% level, (this indicates a result that has a less than 5% chance of being a random occurrence), were encouraging:

• the estimated reduction in homicides and hospital admissions due to violent injury with a sharp object were 0.98 (per one million people) and 1.6 (per 100,000 people) respectively, the latter a larger reduction than reported last year (in the 2022 to 2023 evaluation report), and close to statistical significance with a p-value, which is the estimated likelihood of a result being due to chance, of 0.06 (6%)

• similarly, the estimated reduction for hospital admissions due to any violent injury (not just sharp object) was 10.39 (per 100,000 people), greater than last year’s 7.96, with a p-value of 0.08

• despite the 2023 to 2024 results not reaching statistical significance at the 5% level, the consistent reductions in hospital admissions across consecutive evaluation periods (and newly conducted subgroup analysis) indicate that SV funding is likely having an effect

• the subgroup analysis focused on people aged 24 and under, where statistically significant reductions were estimated for admissions for assaults by sharp objects (3.85 incidents per 100,000 persons) and for assaults by any violence (20.66 per 100,000 persons)

• estimates for police recorded crime (PRC) outcomes showed large confidence intervals (CI) and no statistical significance; this was likely because of lots of variation over time and between areas in this data, which reflects known issues with PRC data (for example, under-reporting or recording); such variability makes it challenging to isolate potential impacts

• wider evidence highlighted the contributions of VRU and Grip activity, with VRU partners reporting that their work to prevent violence locally had been supported and enhanced through the VRU

VRU changes and structures

• VRUs implemented changes, including recruitment, restructuring and team adjustments driven by the Serious Violence Duty (SVD) and local needs. These changes aimed to enhance capacity, partnership working and sustainability

Data-driven decision-making

A key focus of the evaluation included understanding which risk (and protective) factors for violence VRUs and partners considered most pressing to address. Common factors identified across VRUs included:   - education: school exclusions, absenteeism, low attainment and bullying

• adverse childhood experiences (ACEs) and trauma: for example, current or prior exposure to (domestic) violence or abuse

• deprivation: related with violence and other societal factors but recognised (relative to other factors) as difficult to address through the VRU

• unemployment: linked to deprivation but identified as an individual-level risk factor that the VRU and partners can work to address

• substance misuse: alcohol and drug misuse linked to poor mental health and family dysfunction

• mental health: both individual and familial aspects considered

• current involvement in violence and/or crime: though less frequently mentioned, a significant risk factor for further involvement in violence

VRUs continued to access and utilise diverse data sources to understand violence patterns and risk factors. This was evident in Strategic Needs Assessments (SNAs) with many trends and risk/protective factors identified. However, the factors identified as most pressing by VRUs and their partners (see above) were typically not highlighted as such in the SNAs. Progress varied, with challenges in accessing high-quality data from education and health sectors.

Whole-systems approach

While there is some variation in the balance between preventing and reacting to violence, and the extent to which communities and young people are involved, the VRUs’ whole-systems approach can be defined as follows (definition by evaluators):

The whole-systems approach for Violence Reduction Units integrates public health principles, focusing on data-driven early intervention in collaboration with diverse stakeholders. It aims to reduce and prevent violence by identifying and addressing its root causes, through co-ordinated efforts among partner agencies, communities and young people. The approach aims to share critical information, co-ordinate and streamline resources, and enhance protective measures for individuals and communities, ensuring a comprehensive and unified response to violence.

The VRUs’ role in the whole-systems approach is best described as convenors and facilitators in partnership working, aiming to enhance collaboration, provide strategic guidance, and promote a public health approach to violence prevention. A public health approach in this context focuses on early intervention and prevention through addressing the root causes of violence.

Key partners in the whole-systems approach include:

• police: provide data and intelligence, conduct proactive and reactive operations, engage in prevention and diversion activities, inform strategic direction and collaborate with other sectors

• health: share data on violence-related injuries, guide a public health approach, commission and deliver interventions, improve access to services and collaborate with other sectors

• education: deliver school-based interventions, develop trauma-informed practices, engage in awareness-building and contribute to strategic planning

• local authorities (LAs) and Community Safety Partnerships (CSPs): drive governance and delivery at the local level, engage with the community, share data, map provision and embed referral pathways

• communities and voluntary sector: engage with the community, inform VRU work and co-produce interventions​

All partners had clear links to the most pressing risk and protective factors identified. There were some (continued) issues with engagement levels from some partners, owing to competing priorities or, in the case of education, the absence of a single strategic point of contact.

Intervention commissioning and delivery

Across VRUs, 369 interventions were commissioned or funded, targeting various risk groups through initiatives like Hospital Navigators, cognitive behavioural therapy, sports programmes, after-school programmes and mentoring. The type of interventions aligned well to different levels of target groups (ranging from universal to those previously involved in violence). As to be expected, the cost of interventions related to the target groups’ level of need.

Monitoring processes included regular reporting on attendance, demographics and feedback. VRUs provided feedback and capacity-building workshops to support interventions. However, practices varied, indicating a need for standardised frameworks and clear Theories of Change (ToCs) to explain how activities lead to outcomes and contribute to violence reduction.

There was evidence of effective interventions, such as Hospital Navigators and mentoring programmes, being scaled up based on positive data. Some interventions facing implementation challenges were discontinued, such as low referrals or inadequate evidence of outcomes.

Interventions generally succeeded in engaging at-risk cohorts through formal and informal referrals and partnerships. However, challenges included the non-mandatory nature of interventions and varying referral appropriateness. Most strategic stakeholders agreed that VRU interventions effectively targeted and engaged at-risk groups. There was evidence from the case-study visits of early outcomes being achieved.

Recommendations

Based on the evaluation findings, key recommendations for VRUs, if considered appropriate and relevant locally, include: - ensuring the most pressing risk and protective factors for violence (and violence trends and patterns) are made clear in SNAs, Response Strategies, and more broadly when engaging partners; highlighting the most pressing factors, and where and how the VRU can make the biggest difference, could help with partner engagement and prioritisation of efforts

• linked to the above, engaging with key partners about the risk and protective factors that are most relevant to them, to harness and build upon their existing key functions (for example, by working with health to improve access to mental health services, or schools to identify those at risk of exclusion and work with the VRU to prevent this)

• continuing to convene partners, young people and communities to work in partnership across the ‘whole-system’, to include facilitating data sharing, developing or strengthening effective referral pathways, upskilling professionals, and more generally fostering relationships between key partners; the latter includes:

  • treating young people and community representatives as equal (to statutory agencies) partners to help shape the response to violence

  • working closely with LAs and CSPs to co-ordinate and operationalise local responses

  • continuing working with key partners to share and improve the quality of key data sources – where there have been longstanding challenges with the VRU directly accessing certain data sources (for example, health and education), consider if the VRU could focus on ensuring that existing data sharing between partners is sufficient and being used effectively for violence prevention; for example, if the police and health partners share data with each other already, explore whether the VRU could review and advise on how data is being used

  • working with delivery partners to ensure all interventions have at least a basic ToC and proportionate monitoring approach in place from the outset to help with articulating how commissioned interventions align to the most pressing risk and protective factors for violence locally, and to inform the evidence base and future commissioning decisions

Recommendations for the Home Office to support VRUs with the above include:

• encourage and support VRUs to periodically review their SNAs and Response Strategies to make clear the most pressing risk and protective factors for violence locally and how the whole-system approach aligns to these; this could be potentially covered in sustainability plans based on existing SNAs and the findings from this programme-level evaluation (rather than an additional or burdensome output)

• recognising the commonalities across VRUs around the most pressing risk and protective factors for violence identified, consider sharing existing research or guidance, or hosting cross-VRU learning events, focused on effective strategies to address each of these

• emphasise and support the embedding of whole-system approaches, which should include building on existing partnerships (for example, CSPs) and process to ensure they are functioning effectively, as well as new initiatives where there is a clear gap or need for these – this should include sharing systems leadership guidance developed by the evaluation team

• continue to fund, at a minimum, the whole-systems function of VRUs to ensure there is sufficient time to test a long-term approach to violence reduction; based on similar programmes, such as the Cardiff Model and Scottish VRU, 10 years (from VRUs establishment) provides a reasonable window to capture the longer-term effects

• continue to support VRUs with sustainability planning, for the core function and that of interventions

• continue to share resources, like the newly developed intervention monitoring toolkit, with VRUs to support the use of ToCs and proportionate monitoring and evaluation

1. Introduction

Violence Reduction Units (VRUs) aim to prevent and reduce serious violence (SV) in a police force area (PFA), particularly among young people aged 24 and under, by adopting a whole-systems approach. VRUs were established in 2019 in the 18 PFAs across England and Wales, which had the highest rates of SV as part of the Serious Violence Fund (which included enforcement focused hotspot policing ‘Grip’ programme activity). In 2022, the VRU programme was expanded to add 2 further PFAs in Cleveland and Humberside. Since 2019, the Home Office has invested a total of £391.7 million in the Serious Violence Fund. Funding allocations for each area were based on the number of (pre-VRU) hospital admissions resulting from violent injury from a sharp object (a key measure of SV), with areas with a higher number of admissions (typically those with a larger population) receiving more funding. Funding allocations by VRU area for 2023 to 2024 are provided in Annex A.

The Home Office commissioned Ecorys, Ipsos, and academic partners, Professor Iain Brennan and Professor Mark Kelson, to conduct an independent process and impact evaluation of the VRU programme. Since the programme began, this has included annual reports focused on VRU’s implementation and progress, and estimating the impact of VRUs and the wider Serious Violence Fund on key violence measures.

This report presents the findings from the 2023 to 2024 phase of the evaluation, which for most VRU areas was the fifth year of operation and Home Office funding. The evaluation focused on the following overarching questions:

• has violence been prevented or reduced as a result of the VRUs’ work?

• what is the whole-systems approach resulting from the VRUs’ work?

• to what extent is this whole-systems approach aligned to, and effectively addressing, identified local needs?

These questions represent a shift from previous years, where there was more focus on the progress made in implementation. The change in focus reflects the stage VRUs are at, where the whole-systems approaches developed and implemented can be assessed.

1.1 Report structure

The report is structured as follows:

• Chapter 2 summarises the methodological approach, including a revised programme-level Theory of Change (ToC)

• Chapter 3 discusses the impact of VRUs work on violence, which includes quantitative estimates of impact on key violence measures and, to support interpretation, qualitative insights

• Chapter 4 provides context for subsequent chapters with a brief overview of key changes to VRUs and discussion of VRU team structures

• Chapter 5 focuses on VRUs’ role as data-driven decision-makers, which includes a discussion of the risk and protective factors identified as most pressing to address

• Chapter 6 explores the whole-systems approaches developed and implemented by VRUs, including the aims of the approach, and the roles of the VRU and partners in the whole-systems approach

• Chapter 7 focuses on VRUs’ role as intervention commissioners, which includes their monitoring and evaluation of interventions

• Chapter 8 provides conclusions and recommendations

1.2 Acknowledgements

The evaluation would not be possible without the support and input of many people. We are thankful to the 20 VRUs across England and Wales for participating in the evaluation and sharing their valuable insights, without which the evaluation would not be possible. We are especially grateful to those delivering VRU-funded activities and young people who took part and shared their experiences with us.

We are grateful to the evaluation team: Amy Humphreys, Maya Hill-Newell, Sarah Corrigan, George Horton, Gabriela de Freitas, Kay Robinson, Olivia Petie, Polly Clowes, Robert Spence, Jaya Bagla, Shikha Chopra, Lilly Monk, Jude Maher, Lucy Newman, Daniel Latus, Andreas Culora, Jenni Barton-Crosby, Kate Smith (of Ecorys) and Nadia Badoui, Stephen Mallet, Zoe Williams, Catherine Fenton, Waseem Meghjee, Maddy Pickles, Kaviya Selvamanickam, Stephanie Gleeson, Heidi Hasbrouck, and Maryam Hamedi. The evaluation was led by Matthew Cutmore and Jo Llewellyn (Ecorys), in partnership with Caroline Paskell and Hattie Moyes (Ipsos).

We give special thanks to our academic partners, Professor Iain Brennan and Professor Mark Kelson, for their guidance and expertise throughout.

Finally, we thank colleagues at the Home Office, in particular James Mulcahy, Rebecca Channing, Katy Horrocks, Karen Lucas-Walker and John Harper for their ongoing support and management.

This report was written by Matthew Cutmore, Jo Llewellyn, Iain Brennan and Mark Kelson.

2. Methodological approach

Aligned to the new outcome-oriented focus of the evaluation in 2023 to 2024, we developed a new methodological approach. This included revising the programme-level ToC, introducing systems mapping, and an enhanced approach to understanding intervention implementation and outcomes. This chapter provides an overview of these new approaches, alongside the approach to impact evaluation.

2.1 Theory of Change

A revised programme-level ToC was developed by the evaluation team and is presented and described below. The ToC articulates the core functions, activities and outcomes of VRUs, as well as the assumptions and contextual factors that underpin them. The ToC was aligned with the overarching analytical focus and approach of the evaluation, which (as explained in the introduction) shifted from the previous years of the evaluation to address more outcome-oriented questions.

VRU inputs include Home Office funding and, increasingly, match funding from partners, and VRU teams to support/implement all activity. Other inputs include existing resources that support specific activities.

To maximise evaluation insights and value, activities (and associated outcomes) in the ToC are presented as 3 distinct but interconnected VRU core functions:

1. Data-driven decision-makers: The VRUs collect and analyse data and undertake research to improve the local understanding of violence patterns and trends, risk and protective factors, and the current system response. This supports a shared understanding of violence and its drivers amongst partners, identification of at-risk cohorts and system gaps/opportunities. With this improved understanding, it is assumed VRUs (and partners) will allocate resources (interventions or developing the whole-system approach) based on identified local needs and context, which in turn enables an efficient and adaptive response. Data-driven decision-making can be considered the foundation to other activities.

2. Whole-systems approach leads/co-ordinators: The VRUs engage, support and lead partners to develop and implement a whole-system approach to prevent violence. Importantly, partners include VRU core members (partner agencies) and local communities (representatives) to ensure all have a voice and are a part of the response.

   • engage: VRUs engage with a diverse range of partners to facilitate data sharing (to and from the VRU, and between partners) and co-develop the vision and objectives of the VRU; this enables a shared vision and objectives amongst partners, where their views have been considered and roles within the response recognised; as a result, there is shared ownership of the response across partners

   • support and equip: VRUs support and equip partners with the (shorter-term) resources (financial or otherwise) and (longer-term) capacity-building (skills, knowledge, capability); in the shorter-term, this provides partners with additional resources, (which further supports their engagement with the VRU), and capacity-building efforts are initiated; this leads to partners being better equipped to deliver on their roles within the whole-system response

   • lead and co-ordinate: VRUs lead, co-ordinate and monitor the implementation of the whole-systems response; in the shorter-term, VRUs are recognised by partners as system leaders and supporting the required changes; over time, effective strategies, processes and systems are embedded and upscaled

The longer-term outcome of the above activities is effective and self-sustaining inter-partner (systems) working to prevent violence.

3. Intervention commissioners: Aligned to identified local needs, VRUs develop and/or commission interventions. Interventions should be evidence-based and/or show promise in supporting identified local needs. In the short-term, commissioned interventions reach the identified at-risk cohorts and are monitored/evaluated to assess their effectiveness. This leads to risk factors being effectively supported, contributing to the evidence base, and effective interventions being upscaled.

The combined longer-term outcome/impact of all VRU activities is the prevention and reduction of violence.

Figure 2.1: Programme-level Theory of Change

Accessible version of Figure 2.1

The ToC provided the foundation for the evaluation framework, which mapped the research questions, indicators and methods for each component of the evaluation. The evaluation framework ensured that the evaluation was comprehensive, coherent and consistent across the different VRU areas and activities.

2.2 Systems mapping

The core function of VRUs is to lead and co-ordinate the whole-systems approach to prevent violence.

The approach to systems mapping focused on how the VRUs are addressing identified local needs and how partners fit into this. The overarching and sequential questions that were explored included:

1. What are the (key) local needs the VRU is seeking to address? Local needs can include risk/protective factors for violence, trends or patterns in violent/other offending, and current systems gaps.

2. To what extent does the VRU’s whole-systems approach align to identified (key) local needs?

3. What roles do (key) partners have in the whole-systems approach? How do these roles align to identify needs?

Note, that these questions mention ‘key’ local needs and partners. The desk review conducted highlighted that most VRUs identified many local needs (for example, risk factors for violence) in their SNAs, and VRUs are required to engage with many partners (core members stipulated by the Home Office). This 2023 to 2024 evaluation aimed to capture, through interviews with VRU teams and partners, the ‘core system’, which can be thought of as “What are the most pressing local needs for the VRU to address? How aligned is the VRU’s approach to these most pressing needs? And who are the key partners involved in the approach?” There is an intention to build on the core system mapping in the 2024 to 2025 evaluation.

2.3 Intervention case studies

An ethnographic approach to the case studies of VRU-commissioned interventions aimed to provide a rich and in-depth understanding of the implementation, and insights to the outcomes, of selected interventions. The case studies involved observations and interviews with intervention staff and participants, as well as a review of intervention documents and data. Interventions were selected based on sampling criteria which included the type and focus of the intervention and level of local evaluation activity (to mitigate research burden). The case studies focused on:

• the design and delivery of the interventions, including the alignment to local needs and evidence, the targeting and engagement of at-risk cohorts, and the adaptation to changing contexts

• the experiences and perceptions of the intervention staff and participants, including the challenges and facilitators of implementation, the outcomes and impacts of the interventions, and the feedback and suggestions for improvement

• the contribution of the interventions to the VRU work and the whole-system approach, including the links and co-ordination with other VRU activities and partners, and the potential for sustainability and scalability of the interventions

Researchers spent up to 2 days visiting each of the 11 interventions (across 7 VRU regions, (where the sampling criteria was met and the VRU/intervention was happy to take part) to observe the intervention and to carry out semi-structured interviews with intervention staff/volunteers (n=22), young people (n=55) and parents/carers (n=6). Researchers conducted filmed participant observation at the intervention visits across 3 VRU regions, and analysed footage and field notes based on:

• what people said, including their values, beliefs and perceived impact of the intervention

• what people did, including observing how they interacted with each other and their level of engagement

• the context including the physical space, the location in relation to the wider community, time and external factors that may influence engagement

• how the intervention related to others operating in the community

2.4 Quasi-experimental impact evaluation

The quasi-experimental impact evaluation aimed to estimate the impact of the SV funding, which supported the establishment and operation of VRUs, on the key outcome of reductions in violence. The evaluation involved the following methods:

• generalised synthetic control groups analysis (gsynth): where trends on key measures in VRU areas were compared to a statistically constructed synthetic control group, which is a weighted average of non-VRU areas; non-VRU areas with similar (pre-VRU) trends received a higher weighting to improve comparability

• meta-regression: explored the variation in impact estimates between VRUs, and the potential factors that could explain this variation, such as the level of funding, the type and intensity of activities, and the local context

The primary outcome measures included:

• NHS hospital admissions resulting from violent injury with a sharp object

• homicides (sourced from the Homicides Index)

Secondary outcome measures included:

• NHS hospital admissions resulting from any violent injury

• police recorded violence and weapon carrying offences

The gsynth is a valuable tool for evaluating the impact of programmes, especially when dealing with observational data where traditional methods such as difference-in-differences or simple time-series analysis may fall short. In particular, it relaxes the parallel trends assumption by explicitly constructing the synthetic control group for each treated unit (LAs) based on outcome trends leading up to the intervention (the establishment of VRUs). Some of the key considerations that can affect the reliability of the synthetic control group method, and how they were (as far as possible) addressed for this evaluation, included:

• how the synthetic control groups were constructed: in earlier years of the evaluation, different model specifications, which included constructing the synthetic control groups based on prior outcome trends and other factors, such as population and deprivation levels, were tested; it was determined that using only prior outcomes to construct the synthetic control group was sufficient and yielded similar and reliable results

• availability of suitable comparator groups: the evaluation leveraged the availability of suitable comparator groups by, where possible, focusing on LAs as the unit of analysis; VRU funding was allocated to PFAs with the highest levels of violence – while VRU areas tended to cover a greater number of LAs and/or the LAs had higher counts of violence, non-VRU areas still included LAs with similar levels of violence that could be drawn on to construct appropriate synthetic control groups

• volatility in the underlying data: outcomes were aggregated at the LA-level and quarterly time points to reduce volatility in the data, which created smoother trends, helping identify potential impacts amidst the natural fluctuations in violence rates; this proved successful for hospital admissions and homicide outcomes, but police recorded violence outcomes data remained relatively volatile – this volatility is likely explained by issues with consistency and quality of police data over time that are well recognised (OSR, 2024)

• low-count outcomes: these are homicides and hospital admissions resulting from violent injury from a sharp object; this can make it difficult to detect statistically significant changes

• anticipation and interference effects: the potential for anticipation effects, where individuals might change their behaviour in expectation of an intervention, is limited in the context of VRUs; the assumption of no interference, meaning that the intervention in one area does not influence outcomes in another, is addressed, and while spillover effects are possible, they would likely lead to an underestimation of the true impact of the VRUs, making the results more conservative

Although a technical note, the precise approach used was the gsynth package in R, which utilises a bootstrap procedure for statistical inference. This means that the reported confidence intervals (CIs) may not always align to the reported standard errors but are still valid.

Further detail on the approach is provided in Annex A.

2.5 Data collection

Data collection for the 2023 to 2024 evaluation included:

• a review of key VRU documents to provide background and context ahead of primary data collection (below) including: SNAs, which bring together data and evidence to provide insights to violence and local needs (risk and protective factors for violence); and Response Strategies and Delivery Plans that outline the VRUs response to violence; quarterly monitoring returns detailing progress, for example, in intervention delivery; and local evaluation reports

• 167 semi-structured interviews with all 20 VRU teams and their key partners (typically included core member representatives from police, health and education) focused on understanding views on the most pressing local needs for the VRU to address, the extent to which the VRU’s response aligned to these needs, and the roles of partners in the response, and also included role-specific questions (such as additional data sharing and analysis questions for the VRU data leads); as far as possible (depending on VRU structures and interviewee availability), VRU team interviews included directors, data and/or evaluation leads, partnership leads/co-ordinators, intervention commissioning leads, and community engagement leads

• an online survey with VRU partners, shared by VRUs with all their core members, to capture a wider range of views (that is, from partners that were not interviewed), focused on key local needs and the roles of partners, including inter-partner working, in the VRUs whole-system approaches; 117 responses were received from across 16 VRUs (no responses were received in 4 VRU areas)

• ethnographic case studies with 11 VRU-commissioned interventions, which were sampled (from VRU Delivery Plans) based on the following criteria:

• interventions defined by the Youth Endowment Fund as ‘high impact’ or ‘moderate impact’

• interventions aiming to support at least 50 young people

• interventions that were theoretically more amenable to ethnographic/observation work, such as group-based activities

• a good spread of interventions across VRU areas and intervention typologies

The precise data collection varied depending on the interventions context but was underpinned by a consistent framework (see Annex A) and involved: observing intervention delivery; semi-structured interviews with intervention delivery teams; and, where appropriate, interviews and a short questionnaire with intervention participants. Interviews were conducted with 55 young people and 5 parents/carers. Three interventions also included a video ethnography element. The aim of the case studies was to provide an in-depth understanding of intervention delivery and the perceived outcomes.

2.6 Limitations

The evaluation was designed to ensure as robust and comprehensive approach as possible, combining multiple methods of data collection and analysis. However, there are some limitations to consider when reading the evaluation findings in this report:

• the timeframe required for observing outcomes and impacts: the whole-system approach adopted by VRUs was intended to be long term; longer-term outcomes and impacts, as outlined in the programme-level ToC, are only likely to be fully observable beyond the timescales of this evaluation – as a reference point, similar programmes, like the Cardiff Model and Scottish VRU, took around 5 to 10 years to demonstrate impact; (the Cardiff Model focused on data sharing between health and the police to better understand and respond to patterns in violence; the Scottish VRU existed before VRUs in England and its success informed their roll-out)

• linked to the above, Cleveland and Humberside VRUs had only been operational for 2 years

• the primary evidence source for the systems mapping element was the interviews with VRU teams and partners to hone in on what these key stakeholders considered the most pressing local needs, and the ‘core’ whole-systems approach; as noted in Section 2.2 (above), VRU SNAs and Response Strategies cover a wider set of local needs and responses, which may not be covered in this report

• while not necessarily a limitation, the ethnographic case studies of interventions provided in-depth insights to a selection of VRU interventions but were not intended to represent all VRU commissioned or funded interventions

• PRC data for Greater Manchester, Devon and Cornwall and Dyfed-Powys PFAs were not included in the quasi-experimental design (QED) because of missing or inconsistent data submissions to the Home Office; as such, the results cannot be directly compared to those detailed in previous years’ evaluation reports

• a cost-benefit assessment was not conducted as it was not considered appropriate because the QED did not estimate a statistically significant reduction in violence, which the benefits would have been based on

3. Impact on violence outcomes

This chapter provides results from the QED that aimed to quantitatively estimate the impact of VRUs on (serious) violence outcomes. Note that VRU funding coincided with SV hotspot policing funding (hereafter Grip funding) as part of the wider Serious Violence Fund. The results from the primary QED presented in the next section estimated the impact of SV funding overall. Subsequent sections focused on exploring potential explanations for variation in impacts across VRUs, subgroup analysis of people aged 24 and under, and violence hotspots and qualitative insights aim to, as far as possible, draw out the VRU and Grip contributions.

Section 2.4 summarises the methodological approach and Annex A provides additional detail and outputs.

3.1 Quantitative impact on violence outcomes

As discussed in Section 2.4, the precise approach for estimating impacts was gsynth groups, the same approach implemented in previous years’ evaluations. A key consideration for the 2023 to 2024 (and 2022 to 2023) impact evaluation was the 2 new VRUs – Cleveland and Humberside – moving from the comparator group to the treatment group. To navigate this and provide maximum insights, the following treatment specifications were tested in the gsynth analysis:

Model 1

Cleveland and Humberside initially in the comparison group but then switching to treatment group in 2022 to 2023 financial year; this model aligned to VRU programme initiation dates in each PFA.

Model 2

Cleveland and Humberside in the comparison group only because the 2 new VRUs were still in an early phase, where direct interventions with young people, and wider activity, that could be expected to impact on violence outcomes were limited. This was consistent with the VRU Impact Feasibility Report, which highlighted the anticipated gradual and cumulative effect of violence prevention (Home Office, 2020).

Model 3

Cleveland and Humberside excluded from the data to test for any potential contamination of estimates resulting from either of the above models.

Prior to analysis, Model 2 was selected as the most appropriate and fair reflection of the current VRU programme status. For primary outcome measures, differences in impact estimates between models were limited. While some impact estimates resulting from Models 1 and/or 3 were smaller or different to Model 2, the overlapping confidence intervals (CIs) (see explanation below) indicate that these differences are not statistically significant. As such, and with the above in mind, this chapter only presents results for Model 2; however, Annex A contains the results from all models and discusses their differences.

Table 3.1 details the programme-level cumulative impact estimates for all outcomes of interest following 4 years and 9 months (6 months for homicides) of SV funding (the ‘treatment period’, which covered from April 2019 to December 2023). All outcomes were analysed as quarterly rates per 100,000 persons using LA or broadly equivalent CSP-level data. The exception to this was homicides and PFA-level data, which were quarterly rates per one million persons.

All outcome measures presented in Table 3.1 cover the total population. This focus provides a broader view of the impact of SV funding, which includes both VRU and Grip activity. VRUs often target people aged 24 and under as both potential perpetrators and victims of violence, but the impact of these activities is likely to extend beyond this age group. Grip activities, which concentrate on police enforcement in violence hotspots, are less age-specific, further reinforcing that the total population metric may capture the effects of these combined efforts more comprehensively. (Additionally, from a technical standpoint, the total population focus should provide a more stable dataset (particularly in previous evaluation years), while the smaller sample size in the people aged 24 and under may encounter issues common with low count outcomes). Section 3.3 provides post-hoc subgroup analysis of hospital admissions outcomes for people aged 24 and under.

In any statistics model, there is some uncertainty around the (impact) estimates due to variability in the measured outcome. CIs provided a range of plausible values for the sample mean that were consistent with the observed data. If the CI range included zero (admitting both positive and negative impacts), the team did not consider the impact estimate statistically significant (at the 5% level).

Table 3.1: Impact estimates

Outcome	Estimate	Standard error	Lower CI	Upper CI	P-value
Primary outcomes
Homicides	-0.96	2.97	-7.05	4.38	0.71
Hospital admissions - sharp object	-1.6	0.82	-3.16	0.04	0.06
Secondary outcomes
Hospital admissions - all violence	-10.39	4.75	-17.46	1.43	0.08
PRC: Violence with injury	43.09	116.50	-265.59	197.61	0.88
PRC: Violence without injury	-1,083.73	448.74	-1411.49	431.13	0.18
PRC: Possession of weapons offences	-62.31	31.26	-91.08	26.94	0.37

Key observations from Table 3.1 are:

• the impact estimate for homicides was a reduction of 0.96 (per one million persons) but this was not statistically significant and while this central estimate is less of a reduction than that reported in the 2022 to 2023 evaluation (1.5), the CIs overlap so it cannot be said the results are (statistically significantly) different – it is also worth noting that homicides is a low-count outcome, which can make it difficult to detect statistically significant effects; power analysis was conducted, which indicated a larger effect (a change of 5 per one million persons) would be detectable (at the 80% power level)

• the impact estimate for hospital admissions resulting from violent injury with a sharp object was a reduction of 1.6 (per 100,000 persons) but this was not statistically significant (at the 5% level), although more than the reduction reported in 2022 to 2023 (0.9) – similar to homicides, sharp object hospital admissions are a lower-count outcome; however, as indicated by the CIs and p-value (0.06), a slightly larger reduction would have likely been statistically significant

• there was an estimated reduction of 10.39 (per 100,000 persons) for the secondary outcome of hospital admissions resulting from any violent injury, again more of a reduction than reported in the 2022 to 2023 evaluation (7.96) but not statistically significant (at the 5% level) this time; however, the reduction should still be interpreted as encouraging in the context of a previously statistically significant reduction, low p-value (0.08), and a similarly positive result for the sharp object admissions (above)

• the impact estimates for PRC outcomes were not statistically significant and had large CIs; the latter was particularly the case for violence without injury, where caution is urged on interpreting this outcome

• results for homicides and hospital admission outcomes were similar across the different model specifications; results for PRC outcomes were more variable across models, reflecting the aforementioned data variability

While the above results were not statistically significant (at the 5% level), it is important to consider these in the wider context of the available evidence. Hospital admission data, as the most reliable and objective measure of violent injuries, consistently indicate a positive impact across the 2023 to 2024 and 2022 to 2023 evaluation periods. This suggests that SV funding was likely having an effect.

Focusing solely on statistical significance can overlook meaningful trends, especially with near-significant results like those seen this year. The consistent pattern of reductions over time suggests a likely effect, even when it does not meet the conventional 5% threshold, particularly for low-count events such as sharp object injuries.

It is also important to consider the relatively short duration of VRUs compared to similar programmes. For example, the Cardiff Model and the Scottish VRU took 5 to 10 years to demonstrate significant impacts. The early trends observed for VRUs are encouraging and should continue to be monitored.

The following graphs and discussion provide additional outputs from the synthetic control group analysis for the primary outcomes as well as all hospital admissions resulting from any violent injury (secondary outcome). Outputs for other outcomes, and further interpretation of results, are provided in Annex A. Note, all graphs show the trends over financial quarters and year ending. For example, Q1 2024 relates to April to June 2023 (which is in the 2023 to 2024 financial year).

Figure 3.1 shows the average quarterly homicide rates (per one million persons) for VRU areas, and the synthetic control group constructed. Relative to the (unweighted) average homicide rates in non-VRU areas, the synthetic control group provides a better counterfactual. Following the introduction of SV funding, homicide rates initially tended to track below the synthetic control group but returned to similar levels in more recent quarters.

Figure 3.1: Homicides synthetic control group analysis

Figure 3.2 shows the average quarterly hospital admissions for sharp object violent injury rates (per 100,000 persons) for VRU areas and the synthetic control group constructed. Relative to the (unweighted) average hospital admissions for sharp object violent injury rates in non-VRU areas, the synthetic control group provides a better counterfactual. The increase in quarter 2, 2021 (July to September) corresponds to restrictions lifting following the first national COVID-19 lockdown. Differences following SV funding in hospital admissions for sharp object violent injury rates between VRU areas and the synthetic control group (which would indicate an impact) initially appear limited but trends in VRU areas appear to track below the synthetic control group in more recent quarters.

Figure 3.2: Sharp object hospital admissions synthetic control group analysis

Figure 3.3 shows the average quarterly hospital admissions for any violent injury rates (per 100,000 persons) for VRU areas and the synthetic control group constructed. Relative to the (unweighted) average hospital admissions for any violent injury rates in non-VRU areas, the synthetic control group provides a better counterfactual. Following SV funding, trends in VRU areas tended to track below, or very close to, those in the synthetic control group. Differences were most apparent in the most recent quarters.

Figure 3.3: Any violence hospital admissions synthetic control analysis

In addition to the primary and secondary outcomes (see Table 3.1), exploratory analysis of the Crime Survey for England and Wales (CSEW) was undertaken for the first time. This initial analysis used regression techniques with individual-level data instead of synthetic control groups. The analysis did not identify a clear effect of the VRUs, and the overall explanatory power of the regression models was limited. While various factors were tested, they did not significantly improve the models’ ability to explain the outcomes. In future evaluations, alternative ways of using this data to better understand the impacts will be explored.

3.2 Exploring differences in impact and approaches

The team undertook meta-regression to determine whether there was any significant variation in the impact seen across the VRUs. The meta-regression explored whether different levels of VRU funding or approaches on how the funding was used could explain differences in impacts between VRUs. This used the impact estimates generated using synthetic controls for each VRU and then treating these as individual studies. Analysis focused on the impact estimates for hospital admissions outcomes as these are not prone to the limitations (for example, differences in recording) of police recorded outcomes, and the low counts associated with homicide outcomes.

Significant variation between VRU-level impact estimates were detected for the effect on hospital admissions, both from sharp objects and any violence. The team tested the following models for each outcome; Table 3.2 shows which showed a statistically significant (at the 95% level) result.

This shows that impact estimates for VRUs with higher funding per capita and VRU spending on interventions per capita were more likely to indicate a reduction in hospital admission from both sharp objects and any violence. VRUs with higher total intervention funding (including matched and SVD) funding) per capita were also more likely to indicate a reduction in hospital admissions from a sharp object.

Table 3.2: Models that returned a statistically significant result for different types of hospital admissions

Explanatory variable	Hospital admissions from sharp object	Hospital admissions from any violence
Total VRU funding
Total VRU intervention funding
Total intervention funding
Population
VRU funding per capita	X	X
VRU intervention funding per capita	X	X
Total intervention funding per capita	X

3.3 People aged 24 and under analysis

Besides the primary analysis focused on the total population (see Section 3.1), the team conducted exploratory analysis specifically examining hospital admissions for people aged 24 and under. Since this analysis was done after completing and reviewing the primary analysis, it is considered post hoc. As a result, these findings should be interpreted with some caution, as performing additional analyses can increase the chance of finding results that seem significant just by chance. This risk is known as potential false positive bias, where the extra testing may lead to findings that appear stronger than they actually are.

As mentioned in Section 3.1, the decision to focus primarily on the total population reflects both theoretical and technical considerations.

Key insights from the people aged 24 and under analysis included:

• statistically significant reductions in hospital admissions among people aged 24 and under for both assaults by sharp objects and any violence, roughly double those seen in the total population:

  • admissions for assaults by sharp objects had an estimated reduction of 3.85 incidents per 100,000 persons

  • admission for assaults by any violence had an estimated reduction of 20.66 per 100,000 persons

• Figure 3.4 shows the results for the people aged 24 and under group against those for the ‘all population group’ (primary analysis) – the overlap between the central estimates supports the indication that SV funding may have positively impacted on reducing these types of admissions across both groups

Figure 3.4: Comparison of impact estimates (with 95% CIs) for hospital admissions outcomes, by population groups

To estimate the total number of admissions avoided/prevented since VRUs inception, the team adjusted the cumulative impact rate (per 100,000) to the total population of the target age group. Using the most recent population figure of 12.26 million, the team calculated that the intervention prevented an estimated 472 sharp object admissions and 2,534 any violence admission among individuals aged 24 and under.

While the effects in the people aged 24 and under group are promising and may reflect targeted efforts through both VRU and Grip activities, the primary emphasis remains on the total population due to the broader, theoretically justified nature of this measure. The findings in this subgroup should not be viewed as headline results but rather as a supportive indication of impact, consistent with the encouraging trends observed across the total population.

Continued evaluation, including further analysis in Year 6 of the VRU programme, will allow for a more nuanced understanding of these subgroup dynamics while keeping the main focus on population-wide outcomes.

3.4 Hotspot analysis

To understand the impact of SV funding in areas most affected by violence, the team undertook exploratory synthetic control group analysis using Lower Super Output Area (LSOAs)-level police recorded violence data. The areas most affected by violence (‘hot spots’) were defined as LSOAs with the highest levels (top 5% and top 10% within each PFA) of violence prior to SV funding. Recognising the analysis was exploratory, and that the analysis was limited to just 9 SV-funded areas, (SV-funded areas included: Bedfordshire, Essex, Hampshire, Leicestershire, Merseyside, Northumbria, Nottinghamshire, South Yorkshire and Sussex); and 13 non-SV-funded areas where LSOA information was recorded well and consistently over time, caution is advised with the results from this analysis.

It is also important to note the aforementioned limitations of police recorded data (see Section 2.5) and that the number and composition of SV-funded areas included in this year is different to those analysed in previous years.

The analysis was intended to primarily provide insights into the impact of ‘Grip’ (and its predecessor, ‘Surge’) funding, where there was an explicit focus on hot spot policing. While VRUs were expected to target activity at individuals and areas most in need, it was anticipated that their impact would be PFA-wide.

The LSOA-level estimates (over 4 years and 9 months) for violence with injury offences, relative to the synthetic control group, are presented in Table 3.3 (analysis of violence without injury outcomes was not included owing to model stability concerns arising from the variability in the underlying data). Similar to the analysis presented in Section 3.1, the wide (and overlapping) CIs mean the results are not statistically significant (or different from each other).

Table 3.3: LSOA-level violence with injury estimates from hotspot analysis

Hotspot definition	Estimate	Standard error	Lower CI	Upper CI	P-value
Top 5%	2.65	10.78	-20.86	20.2	0.94
Top 10%	-1.95	3.23	-8.25	4.18	0.57

The Home Office undertook a separate and bespoke analysis of Grip funding for hotspot policing in 2021 and 2022, focused on comparing instances of violence in areas on days with Grip patrols and days without. The analysis concluded that Grip resulted in an average 7% reduction in violence against the person and robbery. This statistically significant effect equated to an estimated 1,100 violence against the person and robbery crimes prevented, and £36 million in societal benefits (Jeffery and others, 2024).

3.5 VRU partners’ perception of impact

Figure 3.5 provides responses to impact-related questions from the online survey of VRU partners. Across all questions, around 80% of respondents strongly agreed or agreed their VRU had had a positive impact on their organisation and wider whole-systems approach to violence prevention. Note, for readability, percentage figures are only included for the options with larger responses. Further discussion of the VRU contribution to the whole-systems approach to prevent violence is provided in Section 6.5.

Figure 3.5: VRU partner responses to impact-related questions

4. VRU changes and structures

This chapter summarises changes to VRUs teams, structures and focus in 2023 to 2024, and a general overview of how VRU teams are structured. The main purpose of this chapter is to provide context for subsequent chapters.

4.1 Changes in VRU teams, structures and focus

Most VRUs experienced changes in 2023 to 2024. In many cases, the changes can be considered substantial, indicating that not all VRUs are in a ‘steady state’. These included some changes that have been common in previous years, such as the recruitment or replacement of staff, and others that have been less common, like restructuring the VRU team or governance model, and changes in the size and focus of teams.

The main common changes included:

• recruitment or replacement of staff – especially in data and evaluation, communications and education roles – which involved hiring new staff, filling vacant positions, or changing the roles or responsibilities of existing staff

• restructuring or reorganisation of the team or the governance model, often to better align with local priorities, the SVD, or more generally local needs and contexts; this type of change involved creating new sub-groups or workstreams, changing the reporting lines or accountability structures, or establishing new governance boards or committees

• expansion or reduction of the team size or scope, depending on the funding availability, and areas of focus; this type of change involved increasing or decreasing the number of staff or partners, expanding or narrowing the focus or remit of the VRU, or scaling up or down different activities

The main drivers for these changes included the introduction of the SVD and its legal obligations for partner organisations to co-operate and share information, capacity requirements, recruitment challenges and staff turnover due to the short-term funding and the competitive job market, and changes in funding and the pressure to spend the funding within the financial year.

The anticipated difference from these changes was to improve VRU teams’ capacity, partnership working and whole-systems approach, reach and impact, and sustainability.

4.2 VRU team responsibilities and structures

All VRUs have developed teams to deliver the VRU core functions (see programme-level ToC) based on their available resources and local contexts. As highlighted in the previous section, VRU team members and structures have continued to evolve over time. This section summarises the team structures and responsibilities within VRUs. These insights are based on the interviews with VRU data/evaluation, partnerships, intervention commissioning and community engagement leads. While interviews with all these roles were possible in most VRUs, there were some VRUs where this was not possible owing to staff availability, roles being vacant or not a specified role within the VRU.

4.2.1 Overall structures

At the start of the VRU programme in 2019, there were 2 distinct structures or models being considered and used by VRUs: centralised and ‘hub and spoke’. Most VRUs started with a centralised model, where decisions around strategy and commissioning were primarily made at a whole VRU-area level. Whereas hub and spoke models were structured for more local decision-making through VRU ‘spokes’ (aligned to LAs or CSPs) and central oversight and support provided through the ‘hub’. As the programme evolved, the distinction between these 2 models became less clear: with central models increasingly adapting to, and harnessing, existing local structures, and hub and spoke models requiring more than just hub-level oversight to develop and implement a co-ordinated response.

As a result, most VRUs can now be considered as having a hybrid model, which enables VRUs to effectively co-ordinate a whole-systems approach and respond to varying local needs by harnessing existing structures. For some VRUs, this transition was in response to (or accelerated by) the SVD, which emphasises a local-level response.

All VRUs have a director or lead. In some larger VRUs, the director may be supported by deputy directors. The directors oversee teams with specific functions, which are discussed in turn below, and sometimes cover certain additional roles (for example, partnerships working).

4.2.2 Data and evaluation teams

The Home Office required all VRUs to have specified data and evaluation leads. In some VRUs, the same person covered these roles; in others, they were separate roles, but with close working between them. Depending on the VRU size and structure, and working arrangements with existing or external data and evaluation partners, data and evaluation team size ranges from one to 8 individuals. Typical responsibilities of data and evaluation teams are discussed below.

• data focused roles were responsible for:

  • facilitating data sharing across the VRU partnership, ensuring that relevant data was collected, analysed, and used to inform the VRU’s strategy and interventions

  • leading or supporting the development of data products, such as dashboards, reports and SNAs, that provide insights into the patterns and drivers of SV in the area

  • specifically noted in some VRUs, a role in data ethics, information-sharing agreements and data improvements

• evaluation-focused roles were responsible for:

  • developing and implementing evaluation methodologies and frameworks for commissioned interventions

  • leading or supporting the VRU’s evidence base development by conducting or commissioning research on what works to prevent violence, and by sharing and disseminating the findings and learning across the VRU partnership and beyond

• specifically noted in some VRUs, a role in building the evaluation capacity and skills of the VRU staff and partners, and in ensuring that the VRU’s interventions are aligned with the best available evidence

4.2.3 Partnerships teams

The function of partnerships (whole-systems) working was covered through different teams and roles across VRUs, depending on the wider VRU (or existing local) structures. Some VRUs had specific roles or teams focused on building and managing partnerships. In other VRUs, partnership working was a shared responsibility among the VRU’s directors or other team members. For example, locality (aligned to the ‘hybrid’ VRU model) or thematic leads who focus on specific geographic or thematic (for example, education) areas, respectively.

The typical responsibilities of partnerships teams (or similar, where combined roles) included:

• leading and managing the VRU’s partnership working, including ensuring that the VRU’s partners are engaged, committed and accountable for the VRU’s strategy and delivery

• leading or supporting the development and implementation of the VRU’s governance structure, and for facilitating communication and collaboration across the VRU’s partners and stakeholders

• specifically noted in some VRUs, influencing and advocating for a whole-systems and preventative (a public health) approach to violence prevention, and in working with the VRU’s partners for the SVD

4.2.4 Intervention commissioning teams

Like partnerships teams, intervention commissioning was covered through different teams and roles across VRUs. In some VRUs, the intervention commissioning lead was a dedicated role within the VRU, and they had oversight of all the VRU’s interventions. In other VRUs, the intervention commissioning lead was a shared responsibility among the VRU’s directors or other team members, such as locality or thematic leads.

The typical responsibilities of intervention commissioning teams (or similar) included:

• commissioning, managing and monitoring the VRU’s interventions, and ensuring that they are delivered in accordance with the VRU’s strategy and objectives

• leading or supporting the co-design of interventions, with input from data and evaluation leads, partnership and community engagement leads, and the VRU’s partners and stakeholders

• specifically noted in some VRUs, identifying gaps and opportunities for intervention delivery, and ensuring the sustainability and scalability of the VRU’s interventions

4.2.5 Community engagement teams

The Home Office required all VRUs to have a specified community engagement lead. In some VRUs, the community engagement lead was a dedicated role within the VRU, and they had oversight of the VRU’s engagement across the VRU area. In other VRUs, community engagement was a shared responsibility among the VRU’s staff or partners.

The typical responsibilities of community engagement teams (or similar) included:

• leading and managing the VRU’s engagement with the communities affected by or at risk of SV, and ensuring that the VRU’s strategy and interventions are informed by and responsive to the community’s needs and preferences

• leading or supporting the development and implementation of the VRU’s communication and engagement strategy, and for organising and facilitating events, campaigns and consultations with the community

• specifically noted in some VRUs, supporting and empowering the community to participate in and co-produce the VRU’s interventions, and in building the community’s capacity and resilience to prevent violence

5. Data-driven decision-making

This chapter focuses on the VRUs’ insights to violence patterns and trends, and risk and protective factors identified as most pressing for the VRUs to address. A summary of the key data sources and analytical approaches enabling these insights is also provided. The chapter aligns to the data-driven decision-making function of VRUs described in the programme-level ToC (Figure 2.1). The aims of this chapter are to understand the commonalities and variations in local needs across the VRUs, and provide a reference point for subsequent chapters where it assesses alignment between identified needs and the whole-systems approach and intervention commissioning.

5.1 Violence patterns and trends

The definitions of violence and the relative focus on different types of violence across VRUs were increasingly based on analysis of local trends and patterns. This data-driven approach allowed VRUs to adapt their strategies to the unique challenges of their areas.

5.1.1 Definition and types of serious violence 

The VRUs were focused on reducing and preventing SV – broadly defined by the Home Office as violence causing serious injury in a public place, among people aged 24 and under (Home Office, 2018) – by considering the local context, and taking a public health approach that promoted early intervention and prevention.

The definition of SV varied across the VRUs, generally including violence involving weapons such as knives or firearms, and affecting young people in public settings. Some VRUs had a broader definition that also covered violence in domestic settings, violence affecting older age groups, and violence involving other weapons or methods.

The types of violence most identified by VRUs (through interviews and, to varying extents, in SNAs) as associated with SV were gang-related violence, drug-related violence, county lines, domestic abuse and sexual violence. These types of violence were prevalent across all VRU areas, although the extent and nature of each type varied depending on the local context and challenges. Some VRUs also identified other types of violence, such as violence in the night-time economy , robbery, and organised crime groups, as significant contributors to SV in their areas. 

5.1.2 Spatial, temporal and demographic distribution of violence 

Analysis of the spatial distribution of violence by VRUs showed that most violence was concentrated in urban areas, especially in cities and towns with high levels of deprivation and poverty. These areas often faced multiple and complex challenges, such as deprivation, lack of opportunities and social exclusion, that increased the risk of violence. However, some VRUs highlighted the challenges of rural and coastal isolation and the hidden nature of violence in these areas, where access to services, support and prevention could be limited, and where violence was often under-reported and less visible. 

The temporal distribution of violence showed it tended to peak at certain times of the day, week, month and year, depending on the type and context of violence. For example, some VRUs reported higher levels of violence during evenings and weekends (often linked to the night-time economy), school holidays and summer months, when more people were out and about and when tensions and conflicts were more likely to escalate. However, others noted seasonal variations in, for example, domestic abuse and sexual violence, which increased during winter months and festive periods. Stakeholders linked this to more time spent indoors and more stress and pressure. 

VRU’s analysis of the demographic distribution of violence showed that violence disproportionately affected males, especially those between 15 and 35 years old, who were more likely to be involved in violence as victims or perpetrators. Some VRUs have observed changes in the age profile of victims and perpetrators of violence, with some noting an increase in violence among older cohorts (aged 25 and over) or younger cohorts (aged 10 to 14). These changes may reflect the shifting patterns of demand and supply of drugs, the recruitment and exploitation of vulnerable young people by gangs and organised crime groups, the availability and use of weapons, and the impact of the COVID-19 pandemic and lockdown measures on young people (for example, the impact on their mental health). 

Many VRUs recognised the impact of violence on females, especially in relation to domestic abuse and sexual violence, where there was a recognition of the long-term and severe consequences for their physical and mental health and wellbeing.

The ethnicity and nationality of victims and perpetrators of violence also varied across the VRUs, reflecting the diversity of local populations. 

5.2 Risk and protective factors for violence

Consistent with the public health approach, VRUs have focused their analysis and research to identify the risk and protective factors (the root causes) of violence. The VRUs’ SNAs typically identified many factors (alongside violence patterns and trends), with some VRUs identifying around 20. As part of the interviews with VRU teams and partners, questions asked what the most pressing risk and protective factors for the VRU to address were. The aim of this was two-fold: to better understand each VRU’s priority risk and protective factors (and the reasons for this); and to explore the extent to which there was consensus on these factors within VRU teams and with partners.

Typically, SNAs did not highlight which factors were most pressing to address, which could be a potential area for improvement in future iterations (and for consideration in local ToCs).

5.2.1 Risk factors

Risk factors are characteristics or conditions that increase the probability of violence occurring or escalating. The World Health Organisation’s (WHO) ecological framework (WHO, nd) details 4 (interrelated) levels of risk (and protective) factors (individual, relationships, community and societal), which were reflected in discussions with VRU teams and partners. A general finding across VRUs was the recognition of the interrelation between different risk (and protective) factors. The VRUs and their partners identified the following risk factors as the most pressing to address through their work, considering them the most prevalent influential, and modifiable factors:

• education: almost all VRUs reported education as a risk (and/or as a protective) factor; typically, this centred on school exclusions and repeated absences, where the risk of young people being exploited or becoming involved with violence was heightened – some VRUs also mentioned the effect of low attainment, access to educational opportunities and bullying

• adverse childhood experiences (ACEs) and trauma: most VRUs reported ACEs as a key risk factor; the specific ACEs identified tended to focus on prior exposure to (often domestic) violence or abuse – some VRUs also mentioned children in care, who may have experienced other forms of abuse or neglect

• deprivation: most VRUs identified deprivation as a key risk factor (or driver) for violence; however, VRUs typically focused on the correlation between deprivation and violence, and its interrelationship with other risk factors (deprivation and other more societal factors are discussed below)

• unemployment: although noted as related with wider deprivation, some VRUs specifically mentioned unemployment as an individual-level risk factor

• substance misuse: some VRUs reported alcohol and drugs misuse as a key risk factor, which included its links to other factors such as poor mental health, and exposure to substance misuse (for example, family members)

• mental health: some VRUs mentioned mental health as a risk factor, which primarily focused on the individual but, like substance misuse, was also considered at a relationship-level (for example, as part of wider family dysfunction)

• current involvement in violence, crime or gangs: although there is potentially some crossover with ACEs, only some VRUs explicitly mentioned current (or recent) involvement in violence or crime, or gang affiliation, as a specific risk factor, which may be explained by the public health approach emphasising the need for early intervention (VRU and partners’ response to present day violence is discussed further in Chapter 6)

With the caveat of the interrelationship between levels, most of the risk factors described above can be considered individual- or relationship-level and aligned with existing research (WHO, 2025). Although not mentioned by (or explicitly asked of) all interviewees, some acknowledged limitations and challenges addressing societal-level factors, in particular, poverty and deprivation, through the VRU. Interviewees noted other challenging societal factors like systemic racism, education systems (specifically, exclusions policies), the night-time economy and alcohol use. VRUs recognised the importance of these factors (and sought to address them at an individual level, such as providing opportunities and activities in deprived areas) but recognised they could not fix entrenched social inequality, for example.

5.2.2 Protective factors

Protective factors are characteristics or conditions that decrease the probability of violence occurring or escalating. They aim to protect against or mitigate risk factors. Accordingly, VRU teams and partners tended to mention protective factors less, as these had already been mentioned in the context of risk factors, such as keeping young people in school.

Protective factors that were reported by VRUs and their partners included:

• supporting educational engagement, attainment and opportunities

• building positive and supportive family and peer relationships

• ensuring access to services, such as mental health, for support

• alternative pathways to crime and violence, such as training and employment

• engaging with and supporting communities

• public awareness about violence and exploitation

5.2.3 Level of consensus

Within VRU teams, there was consensus on the key risk and protective factors in most VRUs, but the emphasis on different factors often varied depending on job roles. For example, community engagement leads tended to prioritise local protective measures more. Divergences were most apparent in VRUs that were either still developing their SNAs or had recently changed their strategy or focus.

Alignment between VRU teams and their external partners was more variable. Although partners agreed on the importance of early intervention and prevention, particularly around education, there were generally differences in the risk and protective factors reported as most pressing. These divergences often related to the partners’ specific sectors, with health partners, for example, emphasising health-related risks.

Overall, there was typically consensus, at least at a high level, on the most pressing risk and protective factors for the VRUs to address, and differences can likely be explained by specific roles or sector representation, but there is potential to improve in many VRUs. For example, SNAs are rightfully comprehensive, detailing a wide range of risk and protective factors, but could highlight those that are most pressing or amenable for the VRU to address. Some VRUs had embedded (high-level) priority risk and protective factors into their overarching aims (see Chapter 6), which appeared to support better consensus.

5.2.4 Impact of VRU data insights

The value of VRUs’ analysis into violence patterns and trends, and the risk and protective factors for violence, was recognised by the wider VRU teams and their partners. Of the 117 VRU partners responding to the online survey, almost 60% (n=69) responded that the VRUs analysis had significantly enhanced their organisation’s understanding of violence, and 38% (n=44) said that it had somewhat enhanced their understanding. Based on the interviews with VRU partners and open-text survey responses, the value and impact of the VRUs’ insights can be summarised as:

• reinforcing or confirming the existing understanding of violence, based on professional or local knowledge, and providing evidence to support it; for example, some VRUs and partners reported that the analysis and research did not reveal any new or surprising insights, but rather validated what they already knew or suspected, and gave them more confidence and credibility to act on it

• enhancing or expanding the existing understanding of violence by providing more detail, depth, or nuance, and highlighting new or emerging issues; for example, some VRUs and partners reported that the analysis added more value and richness to their understanding of violence, and enabled them to identify and address some of the gaps or challenges in their knowledge or practice

• changing or challenging the existing understanding of violence, by providing different or contrasting perspectives, or questioning some of the assumptions or beliefs; for example, some VRUs and partners reported that the analysis provoked some critical reflection and debate among them

5.3 Data sources and analysis

All VRUs were using a wide range of data sources to identify violence patterns and trends, and better understand the at-risk cohorts. Table 5.1 details the data sources and the number of VRUs accessing these, by the level of granularity (individual, sub-area (such as LAs, neighbourhoods), or all VRU area population) of the data. Sub-area level data was the most commonly accessed and used level of data. A similar number of data sources were being accessed at a population and individual level. Crime/police data was most commonly available to VRUs at an individual level, recognising their closer links to the police.

Table 5.1: Data sources accessed by level of granularity

Data source/type	Level 1 (population level)	Level 2 (sub-area level)	Level 3 (individual level)
A&E/ISTV		7	2
Ambulance	1	14	1
Crime	4	7	14
Education	6	7	4
Fire	2	7
Health - other	5	3	3
Hospital admissions	9	9	4
Housing		3	4
Other	15	7	6
Prisons/probation	3	7	4
Social care	2	9	7
Socio-demographic	5	1
Youth justice/offending	2	5	2
Total	54	86	51

Based on the interviews with VRU data leads, the main data sources used to identify violence patterns and trends were police and health (ambulance, A&E and hospital admissions) data. These data sources provided information on the types, locations, times and characteristics of violent incidents and injuries, as well as the profiles of victims and perpetrators. However, as described in further detail towards the end of this chapter, not all VRUs were able to access granular and sufficient quality health data. Some VRUs also used other data sources, such as fire and rescue, probation, youth justice, education and public health data, to complement and enrich their understanding of violence in their areas.

The analysis approaches to understand violence patterns and trends varied across the VRUs, depending on the availability, quality and granularity of the data, as well as the capacity and skills of the analysts. Some VRUs used basic filtering and pivot tables, while others used more sophisticated tools, such as Power BI, mapping software, predictive modelling and data science techniques. Some VRUs also used external partners, such as academic institutions or consultancy firms, to conduct more in-depth analysis, such as supporting SNAs, geographical mapping and cohort analysis. Some VRUs also used existing frameworks and tools, such as the Cambridge Harm Index to enhance their analysis. 

The data sources used by the VRUs to identify the risk and protective factors for violence were similar to those used for violence patterns and trends, but with some variations and additions. For example, some VRUs used more public health data, such as Public Health England Fingertips, mental health indices, and substance misuse data, to understand the wider determinants of health and wellbeing that influence violence. Some VRUs also used more education data, such as exclusions, attendance, and attainment data, to understand the role of schools in preventing violence.

The analysis approaches for risk and protective factors were also similar to those for violence patterns and trends, but with some differences. For example, some VRUs used more qualitative and participatory methods, such as surveys, interviews, focus groups and co-production, to understand the views and experiences of the people and communities affected by violence. Some VRUs also used more overlaying and linking methods, such as vulnerability mapping, data dashboards and predictive modelling, to understand the interactions and correlations between different risk and protective factors. Some VRUs also used more evidence synthesis and benchmarking methods, such as literature reviews, national data and best practice guides, to understand the existing knowledge of and gaps in violence prevention. 

The progress, challenges and limitations of data access and sharing among VRU partners varied across the VRUs, depending on the local context, priorities and challenges. Some VRUs reported having good data access and sharing arrangements, facilitated by existing or new information-sharing agreements, data-sharing platforms and data working groups. VRUs also reported that having trusting relationships with partners supported data sharing.

However, some VRUs also reported having difficulties or barriers in data access and sharing, caused by various factors, such as: 

• information governance and data protection issues, such as GDPR, consent, anonymisation and security

• data quality and completeness issues, such as missing, inaccurate, inconsistent or outdated data

• data system and infrastructure issues, such as incompatible, outdated or unreliable data systems or platforms

• data capacity and resource issues, such as lack of staff, time, skills or funding to collect, share or analyse data

• data culture and attitude issues, such as lack of trust, awareness, understanding or willingness to share data among partners 

As per previous evaluation years’ findings, some VRUs continued to report difficulties in access high-quality and granular data from education and health partners, as they more often involved the issues outlined above. As to be expected, the VRUs with greater access to multiple (high-quality) data sources were typically more progressed with the detail and sophistication of their analysis.

Beyond differences in data access, other factors that may help explain different levels of progress in VRU’s data-driven decision-making include:

• limited in-house analytical capacity or resources, such as staff, time, funding, hardware and software, to perform data analysis at the scale and complexity required; few VRUs have also reported skill gaps or needs, such as qualitative analysis, cost-benefit analysis, forecasting, root cause analysis and data literacy 

• the VRUs have faced difficulties in using or communicating data analysis effectively, such as ensuring data quality, validity and reliability, translating data into actionable insights, presenting data in a clear and engaging way, and reaching different audiences and stakeholders with data products 

• the VRUs have lacked guidance or support at a programme level, such as a standardised template for the SNA, guidance on harnessing the SVD, or a forum for sharing best practice and learning across VRUs

6. Whole-systems approach

In this chapter, the VRUs’ whole-systems approaches are described and assessed for alignment to local needs. The intention of this assessment is to help understand how different partners fit into the whole-systems approach, and identify where there might be opportunities to better align and strengthen the approach. Before exploring the roles of partners, this chapter discusses the wider aims of the whole-systems approach and the VRUs’ role within this.

6.1 Whole-systems approach definition and aims

The ‘whole-systems approach’ terminology evolved from the public health approach, which focuses on data-driven early intervention and the multi-agency function of VRUs. The public health approach does not explicitly (or in detail) cover partnership working. The move away from multi-agency working terminology by the Home Office was in recognition that the VRU’s response should involve wider partners – specifically, young people and communities – rather than just agencies.

Broadly, whole-systems working can be defined as an approach of problem-solving that considers the entire structure, function and interrelations within a system, rather than focusing on its individual parts in isolation.

To help frame this chapter, the following definition of whole-systems approach in the context of VRUs was developed based on VRU teams’ views (see below):

The whole-systems approach for VRUs integrates public health principles, focusing on data-driven early intervention in collaboration with diverse stakeholders. It aims to reduce and prevent violence by identifying and addressing its root causes, through co-ordinated efforts among partner agencies, and communities and young people. The approach aims to share critical information, co-ordinate and streamline resources, and enhance protective measures for individuals and communities, ensuring a comprehensive and unified response to violence.

VRU teams typically described the overall aims of the whole-systems approach as follows:

• the overall aim of the whole-systems approach is to prevent violence and reduce its impact on communities

• this is achieved through a public health approach that focuses on identifying and addressing the root causes of violence

• the approach involves collaboration and partnership between different organisations and agencies, including police, local government, health, education, and the voluntary and community sector

• the whole-systems approach aims to align strategies, share data and avoid duplication of work, ensuring that resources are used effectively

• early intervention and prevention are key components of the approach, with a focus on building protective factors around young people and within communities

While the overall aims of the whole-systems approach are similar across VRUs, there were differences in the weight given to different aims:

• some VRUs placed more weight on aligning strategic aims and objectives across partners, while others prioritised training and development, and embedding community engagement

• only some VRUs explicitly mentioned responding to violence in the here and now (reactive) as part of their whole-system approach aims; however, as detailed in subsequent sections, the VRUs’ whole-system approaches do include more reactive elements of work (for example, working closely with the police, and Hospital Navigator programmes)

• the level of engagement with the community and voluntary sector also varies between VRU areas, with some placing a greater emphasis on co-production and community-led approaches

There were also differences in the extent to which specific risk and protective factors were explicitly referenced in VRUs’ overall aims. Some VRUs had priority areas embedded into their overall aims, which, although typically broad, supported a stronger consensus within VRU teams and, to an extent, partners, on the most pressing risk and protective factors.

With the above point in mind and noting that this was not a specific question in interviews, there was general alignment between the VRUs’ aims and objectives and the most pressing risk and protective factors identified (see Chapter 5).

6.2 VRUs’ role in the whole-systems approach

All VRUs recognised their central role in the whole-systems approach to violence. Across VRUs, their role can be summarised as:

• convenors, facilitators or co-ordinators of partnership working: the VRUs aim to break down silos, promote collaboration and avoid duplication of work among partners

• providing strategic oversight, direction and guidance for the whole-systems approach, based on a public health and trauma-informed perspective: the VRUs use data and evidence to inform decision-making, identify gaps and needs, and commission or support interventions that address the root causes and risk factors of violence

• communications and awareness-raising, both internally among partners and externally to the public and the media: the VRUs aim to keep violence prevention as a priority on the agenda, and to share good practice, learning and impact of their work

• supporting capacity-building and workforce development among partners and organisations, by providing training, resources, toolkits and funding: the VRUs also support innovation and experimentation, by encouraging partners to try new approaches, learn from failures and adapt to changing contexts

Key differences in VRUs’ description of their precise roles included:

• the extent to which they were leading the whole-systems approach to violence prevention: some VRUs are seen as leaders or drivers of the whole-systems response, while others are more focused on facilitating or supporting partners to take ownership and responsibility; this can be influenced by existing governance structures, partnership arrangements and local contexts of each area

• aligned to the insights from Chapter 5, some VRUs have a broader remit, covering other forms of violence such as domestic abuse, while others have a narrower focus on SV among young people

• reflected in the wider aims and objectives of the whole-systems approach, some VRUs specifically noted a more balanced approach, focusing both on prevention and reacting to violence, while others have a stronger emphasis on prevention and early intervention

• the levels of engagement and involvement with different partners – partner engagement is discussed in detail towards the end of this chapter

The SVD came into effect in January 2023 and was anticipated to affect the VRUs. For most VRUs, the SVD has not significantly changed or affected their role, as they were already doing the work required by the SVD, such as data sharing, needs assessments and partnership working. The SVD has reinforced and validated their existing work and vision. For other VRUs, the SVD has enhanced or improved their role by providing a legislative framework, a common purpose and a formal governance structure for their work. The SVD has also helped to increase the engagement, representation and accountability of partners, and to ensure compliance with the duty. For a few VRUs, the SVD has introduced some challenges or complexities for their role by creating confusion, duplication or fragmentation of their work. The SVD has also added some pressure, expectations or burdens on the VRUs and their partners, especially in terms of data sharing, reporting and monitoring.

6.3 Partner roles in the whole-systems approach

All VRUs engage with a range of partners to implement a whole-systems approach to violence. Key partners are the VRUs’ core membership, which collectively has responsibility for decisions relating to the VRUs’ operation, and includes 7 agencies; (These include the police, the Office of the Police and Crime Commissioner (or similar in areas where a Mayor is responsible), LAs, Integrated Care Boards, (regional) Office for Health Improvement and Disparities, Youth Offending Team, and education), representation from community and social enterprise sector (VCSE), and local communities and young people. Besides core members, there are other agencies (for example, probation and housing) that the Home Office encourages VRUs to engage with.

Aligned to the 2023 to 2024 evaluation aims, all VRU teams and partners interviewed were asked their views on who the key partners were in the whole-systems approach. This enabled the roles of key partners to be explored in detail, helping to ‘map’ the core whole-system approach. Importantly, this included understanding the alignment between different partners and the most pressing risk and protective factors for violence (see Chapter 5), and any inter-partner working (for example, between the police and health).

The precise roles of key partners in the whole-systems approach varies. The sections that follow aim to identify the commonalities across, and key differences between, VRUs. This VRU programme-level systems mapping is intended to bring together learning from across VRUs (rather than represent a particular VRU’s whole-systems approach).

6.3.1 Overview of key partners

Most VRUs reported that key partners included the police, health, education, LAs (including social care) and CSPs. Consistent with the whole-systems approach definition and aims, many VRUs also referenced communities as a key partner. All these partners have clear links to the risk and protective factors for violence, which are discussed in subsequent sections.

The roles of these partners vary according to their remit, expertise and resources, but generally involve sharing data and intelligence, strategic planning, providing referrals and support, and delivering interventions. The levels of engagement of these partners varies across VRUs, which affects their levels of involvement in the whole-systems approach. Partner engagement (and support) is discussed in Section 6.4.

Other partners that were noted as key in some VRUs included probation, youth justice, fire and rescue, and prisons. While not discussed in detail, these partners tended to have a role in data sharing, community engagement or rehabilitation. The latter may explain why some VRUs considered these partners as key and others did not. These partners can be considered more downstream of the risk and protective factors for violence and reacting to violence in the here and now. As described in Sections 6.1 and 6.2, there were variations in VRUs’ balance between preventing and reacting to violence.

6.3.2 Police

Police were a key partner for VRUs because of their core remit of violence reduction and prevention. While the role of the police was more focused on reacting to violence outcomes (enforcement), proactive and preventative roles were recognised across VRUs. The preventative focus ranged from being a key source of data and intelligence to inform the strategic direction of the VRU and/or targeting of activities, through to more direct involvement with VRUs’ and partners’ strategy planning and intervention delivery. The common roles or functions of the police in the whole-systems approach to prevent violence included:

• data and intelligence sharing: as detailed in Chapter 5, the police were a key source of data and intelligence for VRUs and partners, providing data and intelligence on violence trends and patterns – in particular, hotspot areas, the at-risk groups and individuals, and critical incident reports – to support the VRUs’ strategic direction, develop (local) action plans, and operational targeting

• reactive and proactive police operations: the police (within their core function, existing structures and resources) conducted proactive and reactive policing operations to prevent and reduce violence, such as hotspot policing (supported through Grip funding), specialist teams (for example, knife crime, domestic violence, exploitation) and operations (for example, Organised Crime Group disruption)

• prevention and diversion activities: the police engaged in various activities to prevent violence and divert young people from crime, such as education and awareness campaigns, custody navigators, school engagement officers, and community outreach; additionally, they referred those known to be already involved or at risk of SV to the appropriate support

• informing strategic direction: recognising the police’s core remit to reduce and prevent violence, they could provide the experience and expertise to help inform the strategic direction of the VRU

• partnership working and collaboration: as a more overarching role for the above functions, the police worked with a range of partners across different sectors and agencies, such as LAs (including social care), CSPs, health, education, youth justice, drug and alcohol services, housing, probation, prisons and VCSEs, to inform and support the whole-systems approach

The police’s role in the whole-systems approach was aligned with multiple risk and protective factors. These included: reducing exposure to (or working with other partners to ensure support provided for) ACEs and trauma; reducing violence, crime and gang activity; and improving access to support and alternative pathways (through direct intervention or referrals). The reported levels of alignment to specific factors varied by VRUs, in the context of the police being generally more ‘downstream’ and reacting to violence in the present day. The importance of early intervention and prevention was a common view across police stakeholders interviewed.

In terms of variation in the role of police across the VRUs, this was typically explained by local structures and approaches. For example, in some VRUs, the police were more embedded as they hosted the VRU and/or provided staff, funding and other resources, which meant they had greater input to the strategic direction and governance of the VRU. As noted above, there was also some variation in the balance between preventing and reacting to violence.

6.3.3 Health

Health agencies were another key partner reported across VRUs. Similar to the police, health partners had a role in reacting to violence outcomes and preventing them. However, a key difference was that ‘violence prevention’ was not explicitly noted in interviews as the primary focus of health’s core remit. Health partners include Integrated Care Boards (ICBs) and (regional) Office for Health Improvement and Disparities (OHID). Reflecting the origins of the public health approach (data-driven early intervention), health partners’ role in the whole-systems approach was more focused on prevention, and aligned with multiple risk and protective factors. The common roles or functions of health included:

• data and intelligence sharing: health partners shared data on violence-related injuries and hospital admissions with VRUs and other agencies to inform the analysis of local needs and trends; as noted in Chapter 5, this could include identifying violence that is not reported to the police, and better understanding the risk and protective factors

• promoting and guiding a public health approach: health partners could provide a public health perspective that focuses on the underlying causes and risk factors, and support the VRUs (and partners) to adopt and implement this into their strategies and activities; this was similar to the police’s role to inform the strategic direction of the VRU based on experience and expertise

• commissioning and delivering interventions: health partners were involved in commissioning or delivering interventions that aimed to reduce violence or support victims and perpetrators, such as Hospital Navigators, trauma-informed training, forensic Child and Adolescent Mental Health Services (CAMHS), domestic abuse services and recovery support

• improving access to services: health partners facilitated access to services and support for people who are affected by violence or at risk of involvement, such as mental health, substance misuse, safeguarding and primary care services; this could include working with VRUs to reduce duplication in the system

• partnership working and collaboration: health partners could work with a range of other partners, such as police, education, criminal justice, voluntary and community sector, and LAs, to support the above roles and functions; this could involve co-commissioning, co-delivering or co-evaluating interventions, as well as sharing best practices

These aspects strongly aligned with the most pressing risk and protective factors for violence identified by VRUs and their partners. These included reducing ACEs and trauma, improving mental health, reducing substance misuse and increasing access to services.

There was more variation in health’s role across VRUs owing to differing levels of engagement and capacity to meaningfully engage alongside other priorities (see Section 6.4).

6.3.4 Education

A key theme across VRUs was the potential pivotal, but often constrained, role of education partners in the whole-systems approach. As noted in Chapter 5, education was highlighted as a key risk (exclusions, poor attainment and attendance) and a protective factor, and was arguably the earliest point at which emerging needs can be identified and addressed. However, even where there was senior representation from education within VRUs, the fragmented nature of schools (in particular, academies) and wider factors (see Section 6.4) have, in many cases, prevented education’s potential role being fully realised. Some of the roles or functions of education partners included:

• intervention delivery within schools: education partners delivered various interventions in schools to prevent violence and support vulnerable young people, such as mentoring, sport and nurturing programmes

• trauma-informed practice: education partners worked with the VRUs to develop trauma-informed practice across schools, which involves training staff, creating safe spaces and addressing the root causes of violence

• engagement and awareness-building: the VRUs and other partners (such as the police) worked with schools to facilitate engagement and building awareness of SV and the risks (for example, through school assemblies)

• strategy development and implementation: education partners contributed to the strategic planning and policy guidance of the VRUs, including the development of inclusion charters and embedding of education focused positions within VRU teams, that aimed to emphasise and encourage the protective role of schools and education; relative to police and health partners, education partners were not typically informing the strategic direction to the same extent

6.3.5 Local authorities and Community Safety Partnerships

LAs and CSPs were key partners in the whole-systems approach to violence prevention. They played an important role in driving governance and delivery at the local level, ensuring that the response to violence was relevant to local needs and appropriate for the communities they serve. LAs and CSPs lead on VRU activities, which are based on local area action plans, and played a key role in community engagement. They also worked closely with other partners, such as the police, health and education, to share data, map and streamline provision, share good practice, and embed referral pathways. Linked to the most pressing risk factors identified, LAs (specifically, children’s social care) engaged with and supported young people who experienced trauma and childhood adversity (such as ACEs).

Note that LAs and CSPs were only interviewed directly in some VRUs. As such, the above insights are less detailed and based on the views of wider stakeholders (principally, the VRUs leads and partnerships leads). A research priority for the 2024 to 2025 evaluation will include exploring the whole-systems approach at a local level, with LAs and CSPs.

6.3.6 Communities

Many VRUs specifically noted communities and young people (and/or VCSEs) as key partners in the whole-systems approach to violence prevention, but recognised their distinction from other partners. They played a vital role in community engagement, informing the work of the VRU and helping to steer interventions to address evidence-based local needs. In some VRUs, activity included community engagement, seconding community representatives into the VRU, co-production or delivering at the grassroots level. This was often in hotspot areas with targeted pieces of work.

6.4 Partner engagement and support

This section focuses on the levels of, and factors that influenced, engagement between the VRUs and partners, and the support provided by the VRU to partners.

6.4.1 Partner engagement

Generally, partners from various sectors demonstrate positive engagement with the VRUs, evident in the whole-systems approaches that have been developed. Partners recognised the value and benefits of early intervention and working together to address risk and protective factors. However, partners faced challenges and pressures that limited their capacity and resources to engage with the VRU, such as competing demands, funding constraints, staff turnover and structural barriers. VRUs used various strategies to overcome these challenges and improve engagement, such as demonstrating the impact and relevance of the VRU work, building relationships and trust, providing support and training, and aligning the VRU objectives with existing priorities and initiatives.

Some sectors and areas were more engaged and committed than others, depending on their level of involvement, interest and influence in the VRU work. For example, police and LAs continued to be more engaged than education and health, and urban areas (where violence tends to be concentrated) were generally more engaged than rural areas.

Effective partner engagement was supported by effective governance structures and mechanisms, such as strategic boards, operational groups, thematic delivery groups and partnership agreements. Strong leadership and communication, demonstrating mutual understanding and respect, and sharing a common vision and direction were also noted as important.

The introduction of the SVD, which placed a statutory duty on partners to work together and share data to prevent SV, had a mixed but generally positive impact on engagement levels. Some partners felt that the SVD helped to clarify their roles and responsibilities, increase their representation and commitment, and focus their attention and resources on the issue of violence. Others felt the SVD did not affect their engagement levels, as they were already working towards these obligations prior to the SVD, or that the SVD added more pressure and expectations without sufficient funding and guidance.

Some of the challenges and successes of partner engagement that were specific to some VRUs or sectors are summarised in Table 6.1.

Table 6.1: Partner engagement – specific challenges and successes

Challenges	Successes
Varying levels of engagement with education partners, especially schools and academies, due to their complex and fragmented structures, their reluctance to be associated with violence, and their lack of awareness or recognition of their role in violence prevention.	Improving engagement with education partners by providing training and resources, developing protocols and policies, specialist education roles within the VRU, and involving student champions and mentors.
Difficulties in engaging with health partners, especially NHS trusts and A&E departments, due to their competing priorities, data protection issues, and lack of central leverage or mandate.	Improving engagement with health partners by establishing Hospital Navigators, embedding violence prevention into business as usual, and involving directors of public health in strategic planning.
Challenges in engaging with VCSE and community partners, due to their limited capacity and resources, their consultation fatigue, and their lack of representation or influence at the strategic level.	Building engagement with VCSE and community partners, such as providing funding and support, involving lived experience groups, and co-producing interventions and solutions.

In summary, partner engagement with VRUs was a complex and dynamic process, influenced by various factors and contexts. The VRUs made significant progress in engaging with partners from different sectors and areas, and in taking ownership of their role in the whole-system response to violence. However, there were also challenges and gaps that need to be addressed. The introduction of the SVD provided an opportunity and a challenge for the VRUs and their partners, as it has increased the expectations and obligations for preventing SV, but also highlighted the need for more funding, guidance and co-ordination.

6.4.2 Partner support

The most common types of support provided by the VRUs to their partners were funding, data and analysis, resources and training, and relationship building and facilitation. These types of support were reported by most VRUs, although the extent and nature of the support varied depending on the local context, needs and capacity.

Funding support

The funding support was seen as essential to enable partners to develop and deliver interventions, initiatives and services that align with the VRU’s whole-systems approach and aims. The VRUs allocated funding to a range of partners, including LAs, CSPs, police, health, education and VCSEs. The funding was mainly used to support existing or new projects, and staff that focused on prevention, early intervention, diversion and support for those at risk of, or affected by, violence.

However, the short-term and uncertain nature of the funding was a major challenge for partners, as it affected their ability to plan, recruit and sustain work. Many partners reported difficulties managing the funding cycle, which was often delayed, inconsistent or misaligned with their financial year. Some partners also expressed concerns about the lack of clarity, transparency and criteria for the funding allocation and evaluation. Moreover, some partners highlighted the need for more flexibility and autonomy in the use of the funding, as well as more recognition and support for core costs and overheads.

Data and analysis support

As detailed in Chapter 5, partners valued the data and analysis support as it helped them to understand the trends in, and risk and protective factors for, violence in their areas, and to target their resources and efforts accordingly. The VRUs provided data and analysis support to partners in various ways, such as conducting or commissioning research and evaluation, developing or accessing data dashboards and systems, and sharing or presenting data and evidence. The data and analysis support was mainly used to inform the strategic direction, prioritisation and design of the VRU’s and partners’ work, as well as to monitor and evaluate the outcomes and impact of their activities.

However, some partners faced barriers in accessing, sharing and using data, due to issues such as information governance, data quality and data literacy. Some partners reported continued challenges obtaining or sharing data from or with other agencies, especially those in the health and education sectors, due to legal, ethical or technical constraints. Some partners also mentioned issues with the reliability, validity or timeliness of the data, which affected their confidence and trust in the data. Furthermore, some partners indicated a need for more support or training in analysing, interpreting or applying the data, as well as more feedback or communication on the data and analysis from the VRU.

Resources and training support

The resources and training support were useful for partners to enhance their knowledge, skills and capacity to adopt a public health and trauma-informed approach to violence prevention. The VRUs provided resources and training support to their partners in various ways, such as developing or delivering training courses and workshops, providing guidance and toolkits, and facilitating learning or peer support events. The resources and training support were mainly used to increase the awareness, understanding and competence of the VRU’s and partners’ staff and volunteers, as well as to promote the VRU’s principles and practices across the system.

However, some partners also expressed a need for more tailored and consistent training and resources, as well as more guidance and tools to implement the VRU’s approach. Some partners reported gaps or variations in the availability, quality or accessibility of the training and resources, which affected their uptake and effectiveness. Some partners also mentioned a lack of clear or practical guidance or tools on how to apply the public health and trauma-informed approach to their specific contexts, roles or sectors. Additionally, some partners suggested more opportunities or mechanisms for sharing learning, good practice or feedback across the VRU areas and partners.

Relationship building and facilitation support

Reflecting the VRUs’ core function to co-ordinate the whole-systems approach, relationship building and facilitation support was important for partners to establish and maintain trust, communication and collaboration across different sectors and agencies. The VRUs provided relationship building and facilitation support to their partners in various ways, such as creating or leading strategic or operational groups or boards, organising or hosting meetings, events or consultations, and brokering or mediating relationships or agreements between partners. The relationship building and facilitation support was mainly used to co-ordinate, align and integrate the VRUs’ and partners’ work, as well as to foster a shared vision, culture and ownership of the violence prevention agenda.

The value of VRU support was further evident in the survey of core members where over 80% of the 117 that responded either agreed or strongly agreed that the VRU had effectively helped define or shape their role, recognised their role by harnessing their knowledge and expertise, and supported them (financially or otherwise) to fulfil their role effectively, in the whole-systems approach to prevent violence.

6.5 VRU contribution to the whole-systems approach

To conclude this chapter, this section summarises VRU teams’ and partners’ views on the overall VRU contribution to the whole-systems approach, the extent to which VRUs were effectively leading/co-ordinating the approach, what was working well and what could be improved.

Generally, VRUs made a significant contribution to the whole-systems approach to prevent violence. They have brought a strategic focus to violence prevention, established formal processes, and raised awareness among partners about their role in violence prevention. The VRUs have also facilitated partnership working, supported the building of local-level relationships, and helped to address emerging issues. Variations in the perceived VRU contribution or impact of the whole-systems approaches tended to be explained by the different levels of engagement from partners, funding or resource constraints, VRU maturity, changes in teams, or it being considered too soon (particularly in the newer VRU areas) to realise the impact of the VRU and the whole-systems approach.

Partners generally felt that the VRUs were effectively leading or co-ordinating the whole-systems approach to prevent SV. They saw VRUs as key facilitators of partnership working, providing a central team to co-ordinate and link up partners, and bringing partners together to focus on SV and the specific contextual challenges in their areas. Partners also praised VRUs for their effective and strategic engagement, their data-driven approach, and their ability to bring together a broad range of partners at a regional level.

In terms of potential areas for improvement, some common examples included:

• securing longer-term and consistent funding and resources that support the sustainability (and credibility) of the VRU

• strengthening the engagement and representation of health and education partners, as well as communities and young people

• improving the data sharing and analysis across partners, especially in relation to health data and outcomes

• scaling up and spreading the good practices and learning from the VRUs across different areas and regions  

7. Intervention commissioning and delivery

This chapter details findings about VRU’s commissioning and delivery of interventions. Types of interventions and intended target groups are detailed as well as the influence of risks, protective factors and evidence on decision-making. Intervention monitoring processes are explained, alongside how interventions were selected for more in-depth local evaluation activity and how evaluations were designed. Finally, the delivery of interventions and indications of outcomes are discussed, in addition to any evidence around co-development or co-design, and building the evidence base around ‘what works’.

7.1 Intervention commissioning and target groups

Analysis of monitoring data submitted to the Home Office shows the VRU programme commissioned 368 interventions, with common approaches including A&E navigators, social skills training, mentoring, sports programmes and cognitive behavioural therapy. Other approaches – such as after-school programmes, focused deterrence, interventions to prevent school exclusion, and parenting programmes – were also implemented, along with a range of additional targeted interventions and ‘other’ approaches.

Table 7.1 shows the target group of interventions by the intervention delivery model. As to be expected, prevention and early intervention models were mostly targeting universal and potentially higher-risk groups. Whereas therapeutic and desistence models tended to be concentrated on those with known risk factors and/or already involved in violence or crime.

Table 7.1: Interventions by target group and delivery model (year ending March 2024)

Target group	Prevention	Early intervention	Therapeutic	Desistence	Total
Universal	72	21	16	2	111
Potentially high risk	26	50	10	6	92
Known risk	11	52	35	17	115
Involved in violence/crime	4	6	5	35	50
Total	113	129	66	60	368

Notes: The delivery model for one intervention was listed as ‘Other’ and was not included in the above table for readability.

Table 7.2 shows the number of individuals (excluding professionals) supported by VRU-commissioned interventions, and the average funding per individual, by intervention target group. Universal interventions reached the greatest number of individuals and had the lowest average costs. However, it should be noted that this included several projects focused on awareness-raising and/or chaperoning, where support reached many individuals but could be considered a more indirect form of support. Two such interventions reached approximately 152,000 individuals. Generally, as the level of risk increases (indicated by the target group definitions) so numbers supported are fewer but costs higher.

Table 7.2: Number supported and average costs by target group (year ending March 2024)

Target group	Number of interventions	Number of individuals supported	Mean funding	Median funding
Universal	111	341,451	£211	£60
Potential high risk	92	92,463	£458	£136
Known risk	116	78,132	£1,384	£450
Involved in violence/crime	50	14,011	£1,173	£753
Total	369	526,057

The 2022 to 2023 VRU programme evaluation found that the 3-year funding was seen as a catalyst for progress around strategic commissioning. The 3-year funding enabled VRUs to commission and award grants through more structured and collaborative processes and to take a longer-term view on impact measurement and what the system can achieve. Across the 2023 to 2024 year, evidence suggests this was beginning to be realised, with many VRUs offering multi-year funding to interventions, conditional on performance. Offering multi-year funding to interventions supported asks of the community for longer-term access to activities, demonstrating VRUs considering the whole-system.

There was a strong focus on commissioning being informed by the Youth Endowment Fund (YEF) toolkit as a result of the Home Office grant requirement for at least 30% of each VRU’s intervention budget to be spent on interventions the YEF consider ‘high impact’ (YEF, 2019). VRUs felt strongly directed by the YEF evidence and the Home Office, but acknowledged that commissioning was therefore more evidence-based. Alongside what they felt was a more ‘directive’ push from the Home Office, VRUs used local evaluation findings to inform the commissioning decisions, as well as insights from SNAs and Response Strategies, and a growth this year in using the voice of young people and communities. VRUs discussed the balance of 3 key elements in their commissioning decisions:

• using existing evidence base, such as YEF Toolkit endorsed intervention types

• recommissioning on evidence of something working (management information, evaluation, intelligence or insights – including from SNAs and Response Strategies)

• testing new interventions and building their evidence base

• not discounting smaller community-based interventions, where the ability to develop the evidence base further is more limited (owing to sample sizes, and typically smaller funding amounts) but are underpinned by understandings of what works in these contexts)

Across VRU areas, the extent and way key risk and protective factors have translated to the at-risk groups identified and being targeted varied. Generally, VRUs were at least focusing some of their interventions in urban/inner-city areas where violence and deprivation was more concentrated (see Section 5.1.2). Some VRU areas have been successful in using data and analysis of key risks and protective factors more explicitly to align the targeting of interventions to at-risk groups. However, other VRUs have faced challenges in targeting interventions due to limitations in the tools and data available to them.

In some areas, restructuring the VRU led to improved identification and targeting of at-risk groups. In a few VRUs areas, the link between risk and protective factors and the at-risk groups identified was not always clear, and with some apparent misalignment between intervention commissioning and SNAs or Response Strategies. Translating the identified risk and protective factors to the at-risk groups identified and targeted via the interventions commissioned was more assumed rather than explicit. Overall, the VRU areas are working towards identifying and targeting the most at-risk groups, but the effectiveness of their efforts varied.

7.1.1 Scaling up or scaling down interventions

There was evidence of interventions being upscaled and expanded as well as amended or discontinued. Upscaling and expansion were noted where there was evidence of effectiveness, usually following a mixture of triangulating monitoring information data analysis (of effective performance), evaluation (of effectiveness of delivery and indication of positive achievement of short-term outcomes), qualitative feedback from stakeholders including intervention participants (feedback surveys, insights and intelligence). Examples included Hospital Navigators (where there was evidence of reduced repeated visits), mentoring programmes or where demand for certain interventions was high (and therefore a perceived unmet need).

There was also evidence of VRUs discontinuing interventions based on monitoring and evaluation evidence. Discontinuing interventions typically occurred where there were implementation challenges which could not be resolved, such as an insufficient generation of referrals into interventions; for YEF funded interventions, where there were insufficient numbers of participants to their evaluation requirements; insufficient evidence of delivery or insufficient quality monitoring information; or limited evidence of outcomes. Although a minor theme, some stakeholders discussed the need for a ‘tighter focus’ on interventions that should not continue where there was a lack of evidence of effectiveness.

7.2 Intervention monitoring and evaluation

VRUs and interventions established the monitoring and evaluation processes, depending on the requirements and the resources. VRUs required the interventions to submit regular monitoring reports, which included data on attendance, demographics and feedback. VRUs also provided feedback and support to the interventions, and helped them to understand the monitoring requirements, such as through capacity-building workshops and partnership working.

Interventions conducted their own monitoring and evaluation as required by VRUs, using different methods and tools, such as questionnaires, surveys, interviews, case studies and outcome measures. Interventions used the data and feedback to improve their services and respond to individual needs. Although monitoring and evaluation processes were not consistent or standardised across most interventions at most VRUs, and some interventions lacked a clear ToC or a logic model to explain how their activities lead to their outcomes and how these contributed to violence reduction.

7.2.1 Intervention monitoring processes

VRUs used a range of processes to monitor intervention delivery, including standardised monitoring spreadsheets covering reach and demographic information, self-report questionnaires from interventions and regular performance management meetings. Typically, data was submitted monthly and/or quarterly, with some quarterly reports being more in depth and involving additional qualitative information and case studies. Qualitative data collection was also administered at the intervention level, collating feedback through questionnaires or informal feedback gathering. Monitoring information was then discussed in regular performance management meetings (termed in a range of ways to support partnerships and engagement).

In many cases, a ToC underpinned intervention delivery to demonstrate the logic of an intervention and to map the outputs to intended outcomes. However, it was still the case across most VRUs that smaller interventions did not have a ToC. VRUs were beginning to receive some data on outcomes measures but often what was submitted could be of varying quality or completeness. In a few cases, VRUs had established outcomes frameworks (of varying designs) to align the intended outcomes of all funded interventions. In one case, a VRU had standardised key outcomes metrics to be administered across all eligible interventions. In another case, a VRU had outlined monitoring minimum standards for all its commissioned interventions to adhere to from the outset of their delivery.

7.2.2 Types of data collected

VRU monitoring information processes were established to collect reach and demographic data from almost all interventions to understand the scale of delivery in line with intended outputs. Flexible community funds were where the monitoring information collection was less clear or varied, because of the often small-scale nature of the grants. Both individual-level and aggregated data were collected, determined by the scale and nature of the intervention, and in some cases, the capacity and capability of intervention providers.

In some cases, larger-scale interventions provided individual-level data and aggregated data were available from smaller-scale interventions. Although VRUs noted that the level of data available depended on the intervention type (for example, aggregated data was more attainable where there may be a rotating group of young people participating in interventions, such as detached youth work or in-community sports-based interventions). Whether VRUs were able to obtain individual level also depended on the setup of information or data-sharing agreements. Where agreements were not in place, VRUs could access aggregate-level data.

A few VRUs across the programme used dashboards, which collated monitoring returns data. Dashboards provided breakdowns of key areas to inform performance management and to monitor progress, such as the scale and reach of delivery across a local authority area. Dashboards could support the VRU and partners to see whether commissioned interventions were active in hotspots or areas of need, although resource and technical or administrative capacity were required to operationalise.

Some commissioned interventions were requested to capture outcomes for participants. However, stakeholders discussed the subjective nature of some outcomes, which means they were difficult to validate, such as changes to an individual’s ‘level of risk’. Another key challenge for outcomes measurement was discussed around the difficulty of measurement of preventative activities. That is, capturing the avoidance of negative outcomes. Further challenges included the lack of administrative capacity (and capability) of some, often smaller interventions to collect outcomes measures, or that data collection was deprioritised, where interventions needed to be more focused on inputs to generate sufficient referrals. While a minor theme in the data, interventions could be reluctant to share data, particularly at the individual level (where there were concerns about how the data would be used or about identifying individuals), even where appropriate data-sharing agreements were in place.

Stakeholders, typically from VRUs, discussed the desire for further standardised guidance or direction from the Home Office, or YEF, to better support quality monitoring information and outcomes data collection and collation. The appetite for co-ordination of evidence had been co-ordinated by the Learning and Evaluation Network, (the VRU Learning and Evaluation Network is a cross-programme VRU initiative for staff working in data and evaluation within VRUs to exchange knowledge and learning); where opportunities for knowledge exchange had been provided through regular meetings and guest speakers.

7.2.3 In-depth evaluation selection and design

VRUs described how interventions were selected for in-depth evaluation (that is, beyond analysis of monitoring data) to meet the Home Office grant requirement to spend 10% of their total intervention funding on evaluation. Principally, interventions were selected based on their scale, timing and capacity, or because they were co-funded by YEF, which brought specific evaluation requirements attached to the grant. VRUs often excluded smaller interventions where there was limited amenability, capacity, capability and resources for in-depth evaluation. Yet, in a few cases, where the VRU commissioned an independent evaluation partner, a subset of interventions were included in the evaluation activity, within which smaller interventions could be included in the sample.

“[Our evaluation partners] sit down with us and we try and co-produce [evaluation plans] with them to say what’s achievable, what isn’t achievable. We tend to avoid, if we can, the RCT [randomised controlled trial] route. But the caveat I’ll put on that is YEF love the RCT route, and so as a result we have to balance that out a little bit. So, some of the design is because we want to build the evidence base to be part of the national evidence base that YEF are delivering for us… But some of it is because we need to make sure we’ve got a methodology that’s deliverable and for example, [the] ethics.”

(Strategic stakeholder)

VRU commissions of independent evaluators ranged from commissioning specific evaluations of an intervention to an evaluation partner role, where a sample of interventions would be selected for evaluation activity. Often, interventions were not new or had been mobilised before evaluation partners were in place, meaning that there was an absence of a baseline, and insights were mostly qualitative with some pre/post metrics or retrospective outcomes measurement. Additionally, locally commissioned independent evaluators advised on which interventions should be selected for more in-depth evaluation by conducting impact evaluation feasibility assessments. Available evidence suggests that many impact evaluation designs (such as randomised control trials or quasi-experimental designs) were considered infeasible following the assessment study phase.

Available local evaluation evidence to date concentrated on process evaluation and shorter-term outcomes evaluations, theory-based impact evaluations, and impact feasibility studies or pilot studies. A shorter-term focus on process evaluation helped VRUs and partners to understand the effectiveness of implementation. Impact evaluations of primary and secondary interventions were considered tricky in the short-term as impacts would be realised much further in the future. Impact evaluations (not theory-based), where feasible, were expected to be able to produce findings over the longer-term. There was some feeling, however, that the prioritisation of certain evaluation methods was not appropriate for some interventions, and validated metrics were not appropriate for certain communities or experiences (for example, inappropriate or negative phrasing of questions given participants’ likely experiences). Formal evaluation, particular evaluation designs, and certain metrics could be a barrier to generating insights from some interventions.

The Learning and Evaluation Network was discussing where VRU-commissioned evaluation reports (evaluating commissioned interventions or VRU operations) could be stored online to access evidence across the VRU programme.

Hospital navigator interventions

All VRUs commissioned navigator interventions. The interventions are designed to support individuals who present at A&E departments with injuries related to violence or exploitation. These interventions involved health professionals or other support staff (the ‘navigators’) who engaged with patients to identify risk factors and provide referrals to appropriate services. The aim was to offer a holistic approach to care, addressing not just the immediate health needs but also underlying issues such as mental health, social support and safety. By doing so, Hospital Navigators sought to prevent re-attendance and support long-term recovery and wellbeing.

Stakeholder interviews and documentary review of local evaluation reports uncovered insights into the effectiveness of hospital navigator interventions. Mostly, stakeholders discussed that it was too soon to comment on the effectiveness of these types of interventions, or that data was insufficient (for example, not recording or being able to access re-attendance at hospital information). However, there were a couple of cases where local evidence had found effectiveness and these interventions had been upscaled. In a few cases, these interventions had not been recommissioned because of implementation challenges, where there was positive feedback from stakeholders, but monitoring data did not reflect successful engagement or delivery.

There were several key implementation challenges for these types of interventions, including low numbers of people presenting with injuries, higher than anticipated levels of mental health needs, or people presenting with self-harm injuries. More persistent or systemic challenges included youth worker shortages, churn of hospital staff, which created a need for continuous training, and the disengagement rate of young people.

“We know from the interim evaluation report, that [the intervention] actually is identifying lots of the risk factors… The one I think that stuck out for me was mental health… But then that helps us to informing the integrated care board, for planning for improved mental health services. Because children shouldn’t be ending up in A&E, they should be getting better support earlier.”

(Strategic stakeholder)

Interviews and evaluation reports suggested awareness raising, development and training of hospital staff were critical to identify risk factors, as well as to be aware of their biases, to ensure referrals were made into the intervention. It was suggested that referrals could be improved by having a ‘champion’ or dedicated clinician in relevant services who could co-ordinate the intervention. Although, the ability to do this could be limited due to capacity challenges and constraints on NHS services. VRUs were working together to better understand how to identify clinicians who could become champions.

Moreover, evidence sources indicate the need for quality over quantity of hospital navigator intervention delivery. As a result, some VRUs were working on prioritising a greater focus on reaching a smaller number of people with more intensive intervention where there was need, or focusing on ensuring there was a wider multi-agency response for those who did not want to engage.

“The biggest element would be about model of delivery, that it’s about quality not quantity and about your hotspots in that hospital. So, your hotspots might be through a given time period… Thursday, Friday, Saturday, is the best time to have your ED Navigators in, because your data tells you that.”

(Strategic stakeholder)

Overall, there were common factors which influenced VRUs’ understandings of the effectiveness of hospital navigator interventions. These included a lack of, or delay of, data-sharing agreements, preventing access to key hospital and police records, inconsistency or incompleteness of data, small sample sizes, the absence of comparison groups, and short timeframes for analysis where outcomes required testing over a longer period. To overcome key challenges, stakeholders suggested that if this intervention was of interest, there should be a multi-site trial across the programme to test its effectiveness. Although, a multi-site A&E navigator trial funded by YEF did not fund a further evaluation beyond a feasibility study (Sutherland and others, 2023). Alongside this, VRU stakeholders suggested that VRUs, the Home Office and health representatives could work together to support the identification of clinicians to act as champions for these interventions.

“It’d be hard to put a nail in it to say it’s effective and everybody is experiencing the same with that, from what I understand. We need to look nationally with one evaluation as to the whole project, because while it is appearing effective on a local level, it’s not statistically significant, I wouldn’t say, with the numbers that we’re reaching.”

(Strategic stakeholder)

7.3 Intervention delivery and outcomes

7.3.1 Co-development and co-design

VRUs consulted with young people and communities about the commissioning of interventions or, more widely, the approach to reduce and prevent SV through a range of methods (see Chapter 6). VRUs engaged co-development with local communities and young people through surveys, consultations, focus groups, events and youth assemblies. The level and extent of co-development varied across different VRUs and interventions, depending on the local context, needs and resources. Holding ‘Hope Hacks’ (run by the Hope Collective) were cited across many VRUs, where young people aged 15 to 24 participated in workshops to discuss their hopes for the future, and groups to share thoughts and solutions on issues that affect their lives and communities. These were highly regarded amongst those who had participated in them, and there were wider positives around increasing young people’s awareness of VRUs.

“Off the back of the event we had really good feedback from the young people themselves, from saying that they’re never normally listened to, or they’re asked their opinion and then nothing is ever done of it. So the fact that even the projects are in the works now, we’ve had really positive feedback. We’ve had loads of them saying that they’d want to come back to an event again, it felt like a safe space, they felt listened to. In turn, that’s created not only awareness of [the VRU] to these young people, but also trust in it, trust that we’re an organisation that are taking them seriously and want to hear from them.”

(Strategic stakeholder)

Alongside, there were various other forms of ‘listening’ and consultation projects that VRUs carried out in collaboration with wider system partners (for example, with LAs, youth panels and youth councils). Consultation and co-development processes highlighted community needs and VRUs tried to translate these findings into their strategy as far as possible. Overall, co-development occurred at different stages of the intervention cycle, such as design, implementation and evaluation. However, most of the co-development activities were focused on the commissioning or design stage, where communities and young people were consulted or involved in shaping the intervention ideas, objectives and methods.

VRU-funded interventions used a secondary stage of co-development, taking inputs from young people participating in interventions to shape activities by asking young people what they wanted to do and how they wanted to do it. This could be by choosing selected elements from an overall set programme, or by suggesting the different activities the intervention delivered week-on-week. Evidence showed that some interventions did not have structured processes, but considered the feedback and suggestions of intervention participants to influence and adapt current and future activities.

These various forms of co-development reportedly contributed to making interventions more relevant, engaging and empowering for the target groups, as well as building trust with communities that they were being listened to. Stakeholders believed it helped to identify gaps, challenges and opportunities for improving the VRU’s approach and intervention delivery. Some examples of co-developed interventions included a knife crime media campaign, a community-led alliance, a radio station and a mentoring programme.

Yet, not all VRUs were sufficiently engaging young people and communities about strategy or interventions, and there was a keenness to do more in the future. Furthermore, co-development opportunities were not without challenges and limitations, such as time constraints, resource limitations, trust issues, representation gaps and sustainability concerns. Some VRUs reported they had to balance the need for co-development with the urgency of delivering interventions within a short timeframe and with limited funding. Stakeholders discussed a key constraint being the timings of funding cycles (April to March), which limited the possibility of co-development and co-design. The commissioning and mobilisation process was thought to already shorten the timeframe available for delivery and adding further stages of co-development or design would further shorten delivery timeframes.

VRUs suggested it was important to be focused on delivery, and there was some indication of a pressure to be delivering on outcomes, which could influence the ability to engage young people and communities. Co-development and co-design activities could also take notable resource to do them sufficiently and to avoid tokenism. Overall, stakeholders suggested that incorporating feedback loops enabled activities to adapt across an intervention’s life cycle.

Other stakeholders mentioned they faced difficulties in building trust and rapport with some communities and young people, especially for marginalised or excluded groups. Moreover, some VRUs acknowledged they could not reach or involve all the relevant or affected groups and individuals in their co-development and engagement processes, and that they needed to be more inclusive. Finally, some VRUs expressed concerns about the sustainability and scalability of co-developed interventions, especially when they relied on external or short-term funding or support. Particularly where this was discussed, VRUs felt it was misleading to co-develop an intervention or activity, for it to only be available for a short time. They considered this could potentially act to erode trust from these communities.

7.3.2 Effectiveness of target group engagement and implementation

Evidence indicated that interventions were generally quite effectively engaging at-risk cohorts. VRUs used formal and informal methods of engagement to identify and reach potential participants through referral processes and partnership working across a range of agencies. Intervention and VRU staff demonstrated how eligibility criteria had been established and how they had adapted referral processes to ensure they were reaching intended cohorts. This varied based on intervention type, and all visited interventions demonstrated reaching those who had a range of risk factors for being affected by violence, either now or in the future (that is, primary, secondary and tertiary levels of intervention) (Conaglen and Gallimore, 2014). In a survey of strategic VRU stakeholders (responses n=117), 31% strongly agreed that VRU interventions were effectively targeting and engaging at-risk groups, with 51% of respondents agreeing; however, 13% of respondents responded neutrally, and 3% of respondents disagreed and less than 1% strongly disagreed.

However, there were some challenges in engaging at-risk cohorts. Some stakeholders linked challenges to the non-mandatory nature of VRU-funded interventions, which could mean that some of those at risk chose not to participate. A wider challenge could be the appropriateness of referrals against the eligibility criteria, which was more pronounced where there was a wide range of referring agencies. That is, where it was more difficult to ensure shared understandings of those ‘at risk’, or where VRU interventions were perhaps seen to be available for a wider range of needs, where other services had capacity challenges to meet the needs of those with lower levels of risk/need (for example, children’s social care).

Across all evidence sources, analysis suggested that most commissioned interventions were perceived to be implemented effectively. Interventions were reportedly and demonstrably flexible and responsive to the needs and interests of the young people and communities within which they operated. Interventions were delivered by qualified and experienced staff, who had the skills and the passion to work with young people, often overseen by organisations with many years of experience in their communities or of violence reduction initiatives. Interventions provided activities to suit the preferences and motivations of the young people, such as sports, arts, education and mentoring. Fieldwork visits to interventions observed staff creating a safe, supportive and inclusive environment for young people, where they were accepted and understood. Staff built positive and trusting relationships with young people and, where relevant, their families, and provided them with ongoing support and guidance.

Despite evidence of much successful implementation, there could be key challenges communicating eligibility criteria and ensuring that appropriate referrals were made to engage the right target groups. Wider challenges for implementation included high levels of unmet need, such as mental health needs, where interventions were bridging a gap for a lack of mental health services. Additionally, some stakeholders discussed the challenges implementing high-impact interventions in line with YEF and Home Office guidelines. For example, key issues implementing the focused deterrence model were identifying the ‘right’ target cohort and ensuring they actively engaged, rather than seeing it as something that may support them with future sentencing.

Stakeholders also felt they were developing these interventions in very different contexts to historic-focused deterrence implementation, with most of the evidence of implementation being from the US, with limited implementation in the UK (Braga and others, 2019).

7.3.3 Outcomes

Much like the 2022 to 2023 evaluation findings, immediate - and short-term outcomes varied across VRUs and interventions. Overall, there was evidence of VRU-commissioned interventions delivering on intended early outcomes for young people. VRU monitoring and evaluation activity provided data that reinforced qualitative insights. Most VRUs could point to ever-growing volumes of evidence (quantitative and qualitative) about outcomes for young people participating in interventions (or any wider target group). Evidence included evaluation reports, monitoring data, observations and feedback from young people, parents/carers or communities as relevant.

The intervention case-study visits and (more in-depth) review of local evaluation reports focused on a specific selection and set of intervention types. With this caveat in mind, immediate and short-term outcomes being achieved, continued to relate to several key domains:

• connection: trust in services and trusted adults, being signposted or referred to other services

• education, employment and training: increased educational engagement or reduced risk of school exclusion, access to or engagement with employment, technical or other skills, achievements or awards

• social: attitudes to violence, self-confidence, leadership and teamwork skills, feelings of safety and belonging, respect for others, sense of resilience, self-regulation

• wellbeing: mental health, drug use, housing, including access to support

• offending: reduced engagement in re-offending and/or antisocial behaviour (ASB), proximity to gang involvement, risk of criminal exploitation and engagement in risk-taking behaviours

These outcomes aligned well with the risk and protective factors for violence identified as most pressing to address through the VRU (see Chapter 5).

The mix of reported outcomes indicated that VRUs had commissioned interventions that could have more durable impacts on young people’s lives – reducing engagement with offending alongside developing broader life skills. This was corroborated through research with interventions, which found that young people were experiencing a range of the above positive outcomes while engaging in interventions. Interventions also helped young people to make positive lifestyle choices, such as staying in education, avoiding criminality, and pursuing their goals and aspirations. Interventions demonstrated building positive relationships between young and staff, who acted as mentors and role models, providing ongoing support and guidance. Young people were very positive about the engagement they had with intervention staff.

“[The mentor] is like my best mate. You know they really care about you.”

(Young person)

“It feels like they are a safe person to talk to. They’ll help you through and they’ll look out for you.”

(Young person)

These impacts, more directly linked to SV, were also found across other interventions, for young people attending interventions and being supported by staff. For example, one young person avoided a custodial prison sentence, and for another, they had been protected from increased risks of exploitation.

“When I started coming down here, my life was going down the criminal side, and this place has basically changed it, changed me as a person.”

(Young person)

“[The intervention] is the only thing that’s made a difference… it’s helped him not get into the gang that he was on the edge of getting into… It’s preventative and has also massively improved his mental health and his confidence.”

(Referrer)

Some interventions, where relevant and offering group-based activities, created a sense of community by bringing together young people from a range of backgrounds and circumstances, and fostered a culture of respect and friendship. Where interventions engaged parents/carers or parents/carers were the intended target group, there similarly were positive reports from those interviewed.

“It has been useful to be honest. You feel like you are on your own but when you come to these groups, the staff are brilliant, the parents are great and everyone just gives you that bit of a boost, gives you tips on, like, how to help your child sleep a bit better. Any question you have for them they have the answer straight away for you. Even school advice, like looking for schools.”

(Parent/carer)

Overall, the sample of young people (n=55) and parents/carers (n=5) interviewed showed clear evidence that interventions had taught or supported them with something, even when engagement had been for a short time. There was no evidence to suggest interventions were failing to deliver on outcomes. Although for some young people, it was too early in their engagement for them to comment on outcomes that would be expected to be realised over consistent engagement over a longer period.

Staff delivering and managing interventions self-reported various outcomes. These included greater awareness of community needs, risks and protective factors, trauma, as well as enhanced understandings of neurodivergence, mental health needs, and the wider support system around a young person or their family. Working with the VRU could contribute to capacity-building of organisations commissioned to deliver interventions, through support with monitoring, data, evaluation and access to training. That said, training opportunities were not always offered by the VRU, but the VRU would support interventions to access training. But not all interventions felt so positively about the relationship with the VRU. Some stakeholders suggested the relationship with VRUs could feel ‘extractive’ as VRUs required a high level of monitoring in comparison with other funding sources or for the scale of grant they were receiving.

While a more minor theme, interventions and VRU suggested that positive influences on the wider community were beginning to be observed. These included a reduction in perceptions or reporting of ASB, littering, shoplifting, and increasing the use and safety of public spaces.

Case study: A model of focused deterrence

A focused deterrence intervention involved identifying young people likely to be involved in violence as either a perpetrator or a victim and supporting them to desist. This support took the forms of inputs that emphasise the consequences of violence as well as providing any support young people may need. A team of a police officer and a social worker provided activities and built positive relationship with participating young people. Multi-agency working was important to create referral pathways for the young people to access additional support from services and the community.

The intervention functioned through informal community mapping of networks including social work, community networks, schools, and young people. The police officer listened to intelligence from a range of stakeholders and shared this with youth workers, so that they could undertake targeted intervention and prevention activity.

Much of this intervention relied on relationship building. There was indication of how the police officers forged positive relationships with the young people. For example, schools could report a pupil’s absence, and the officer would drop around to pick them up to take them to school. Young people reflected the positive relationships built:

“It’s a good relationship. He talks to me about how to keep safe and stuff because obviously stuff has happened me in the past. So he gives me advice which I sometimes take on board when I’m out and about.”

(Young person)

Another young person spoke about how they could speak to the police officer more easily than their father:

“I can talk to [the police officer] about things more than I can talk to my dad.”

(Young person)

The intervention had only been running for 2 months so it was too early for stakeholders to comment on most outcomes. Delivery staff were cautious about suggesting whether positive longer-term outcomes would be realised but they had faith in the potential of the intervention.

Multi-agency working appeared to be effective. Focused deterrence (a tertiary model) identifies participants when the risk level is already quite high. Therefore stakeholders perceived that the outcomes were harder to achieve. The main outcomes related to preventing the cohort of becoming victims or perpetrators of serious violence. Practitioners referred to intended outcomes of keeping young people in education, reducing their risks of exploitation and being involved in crime.

7.3.4 Intervention sustainability

Sustainability was not the primary focus of the evaluation. Yet, from evaluation data, it was difficult to ascertain the level of sustainability of VRU-funded interventions. Several stakeholders suggested that their intervention, or some interventions the VRU commissioned, would take place regardless of VRU funding. For example, where they were part of established community organisations and where interventions predated the VRU, stakeholders suggested they would continue into the future. However, many interventions noted that the VRU was their primary or only funding source and that beyond the life cycle of the VRU grant, it was difficult to know whether the intervention could continue. More widely than funding sustainability, there were many concerns across stakeholders that there would be limited options of where to refer young people to go to if interventions did not continue, as there was limited free local provision.

Analysis of management information showed that while VRU funding formed most of the financial support for interventions, match funding played a significant supplemental role across the programme. On average, match funding accounted for roughly 22% of total intervention funding, highlighting the importance of additional resources to expand the programme’s reach and impact.

However, the reliance on match funding varies considerably across VRUs. Some areas exhibit a strong reliance on VRU funding alone, while others have leveraged additional local or third-party resources. This variation may reflect differences in VRU-funding strategies, where, for example, match funding has been used to fund specific roles or other activities (rather than interventions).

8. Conclusions

Conclusions are provided against the key overarching research questions.

Has violence been prevented or reduced as a result of the VRUs work?

Results from the quasi-experimental impact evaluation, though not statistically significant at the 5% level, were encouraging. The estimated reduction in the primary outcomes of homicides and hospital admissions due to violent injury with a sharp object were 0.98 (per one million people) and 1.6 (per 100,000 people), respectively. The latter was a larger reduction than reported last year (in the 2022 to 2023 evaluation report), and was close to statistical significance with a p-value of 0.06. Similarly, the estimated reduction for the secondary outcome of hospital admissions due to any violent injury was 10.39 (per 100,000), better than last year’s 7.96, with a p-value of 0.08. Estimates for police recorded crime (PRC) outcomes showed large CIs and no statistical significance, owing to variability in the underlying data.

While the exploratory analysis of hotspots did not provide an indication of the VRU/Grip contribution to impacts, separate analysis by the Home Office and qualitative insights from VRU teams and partners highlight the contribution of each.

What is the whole-systems approach resulting from the VRUs’ work?

All VRUs carried out data analysis and research to better understand, and strengthen the local evidence base around, violence patterns and trends, and the risk and protective factors for violence. Most VRUs identified the same or similar set of risk (and protective) factors that are most pressing to address, such as education, ACEs and trauma, substance misuse, mental health and, to varying extents, involvement in violence or crime. Deprivation was also a key risk factor, but VRUs recognised the challenges of addressing societal issues. Generally, individual- and relationship-level factors were the ones considered most amenable to VRU activity. These aforementioned factors are well-evidenced by, for example, the World Health Organisation (WHO), and many of the individuals or groups experiencing them should be known to existing services and partners (for example, social services and ACEs).

While there is some variation in the balance between preventing and reacting to violence, and the extent to which communities and young people are involved, the VRUs’ whole-systems approach can be defined as follows:

The whole-systems approach for VRUs integrates public health principles, focusing on data-driven early intervention in collaboration with diverse stakeholders. It aims to reduce and prevent violence by identifying and addressing its root causes, through co-ordinated efforts among partner agencies, and communities and young people. The approach aims to share critical information, co-ordinate and streamline resources, and enhance protective measures for individuals and communities, ensuring a comprehensive and unified response to violence.

The VRUs’ role in the whole-systems approach can be described as convenors and facilitators in partnership working, aiming to enhance collaboration, provide strategic guidance, and promote a public health approach to violence prevention. Few VRUs described themselves explicitly as ‘leaders’ of the whole-systems approach, which is perhaps important and intentional when a core aim of a whole-systems approach is shared ownership, and potentially reflects the relative power VRUs have.

The typical key partners in VRUs whole-systems approaches included the police, health, education, LAs and CSPs, and, to varying extents, communities and young people. The police and health have substantial roles at an overall strategic level and operationally preventing and reacting to violence. Reflecting on VRU structures, LAs and/or CSPs are important for driving governance and delivery of the whole-systems approach at a local level. While all partner engagement and precise roles varied across VRUs, education was seemingly most variable. Some VRUs had good strategic buy-in from senior education representatives and had made good progress with strategic and operational activities.

However, even with this buy-in, the fragmented nature of the education system was problematic for many VRUs. This potentially questions the role of education in the whole-systems approach. While they should be (substantially) engaged and supported, a more embedded role like that of the police and health might not be feasible.

The whole-system response included a combination of strategic and operational activity. Common examples of strategic activity included the co-development of strategies and local action plans, which drew on the expertise of partners, relationship building between partners, data and intelligence sharing, and aligning and supporting existing partner functions and priorities to embed violence prevention. Operational activity focused more on intervention commissioning, delivery and monitoring.

To what extent is this whole-systems approach aligned to, and effectively addressing, identified local needs?

Generally, the whole-systems approach was aligned to the most pressing risk and protective factors (and wider local needs), but the extent to which this was consistently articulated by VRU teams and partners varied. While stakeholders emphasised the focus on data-driven early intervention (public health approach), the linking between the roles of partners and specific risk and protective factors was developed through a critical assessment across evidence by the research team. Some of the links made are obvious ones and may be detailed in, for example, some VRUs Response Strategies. A clearer ‘connecting the dots’ of VRU’s whole-systems approach could strengthen a shared vision across teams and partners, and potentially the overall alignment of the approach.

Recommendations

Based on the evaluation findings, key recommendations for VRUs if considered appropriate and relevant locally include:

• ensure the most pressing risk and protective factors for violence (and violence trends and patterns) are made clear in SNAs, Response Strategies, and more broadly when engaging partners; highlighting the most pressing factors and where and how the VRU can make the biggest difference could help with partner engagement and prioritisation of efforts

• linked to the above, engage with key partners about the risk and protective factors that are most relevant to them, to harness and build upon their existing key functions (for example, by working with health to improve access to mental health services, or schools to identify those at risk of exclusion and work with the VRU to prevent this)

• continue to convene partners, young people and communities to work in partnership across the ‘whole-system’, which includes facilitating data sharing, developing or strengthening effective referral pathways, upskilling professionals, and more generally fostering relationships between key partners – the latter includes:

  • treating young people and community representatives as equal (to statutory agencies) partners to help shape the response to violence

  • working closely with LAs and CSPs to co-ordinate and operationalise local responses

• keep working with key partners to share and improve the quality of key data sources; where there have been longstanding challenges with the VRU directly accessing certain data sources (for example, health and education), consider if the VRU could focus on ensuring that existing data sharing between partners is sufficient and being used effectively for violence prevention – for example, if the police and health partners share data with each other already, explore whether the VRU could review and advise on how data is being used

• work with delivery partners to ensure that all interventions have at least a basic ToC and proportionate monitoring approach in place from the outset (including to collect baseline data), which will help with articulating how commissioned interventions align to the most pressing risk and protective factors for violence locally, and to inform the evidence base and future commissioning decisions

Recommendations for the Home Office to support VRUs with the above include:

• encourage and support VRUs to review their SNAs and Response Strategies to make clear the most pressing risk and protective factors for violence locally and how the whole-system approach aligns to these; this could be potentially covered in sustainability plans based on existing SNAs and the findings from this programme-level evaluation (rather than an additional or burdensome output)

• recognising the commonalities across VRUs around the most pressing risk and protective factors for violence identified, consider sharing existing research or guidance, or hosting cross-VRU learning events, focused on effective strategies to address each of these

• emphasise and support the embedding of whole-system approaches, which should include building on existing partnerships (for example, CSPs) and process to ensure they are functioning effectively, as well as new initiatives where there is a clear gap or need for these – this should include sharing systems leadership guidance developed by the evaluation team

• continue to fund, at a minimum, the whole-systems function of VRUs to ensure there is sufficient time to test a long-term approach to violence reduction; based on similar programmes, such as the Cardiff Model and Scottish VRU, 10 years (from VRUs establishment) provides a reasonable window to capture the longer-term effects

• continue to support VRUs with sustainability planning, for the core function and that of interventions

• continue to share resources, like the newly developed intervention monitoring toolkit, with VRUs to support the use of ToCs and proportionate monitoring and evaluation

9. References

Braga AA, Weisburd D and Turchan B (2019) ‘Focused deterrence strategies effects on crime: A systematic review’ Campbell Systematic Reviews, vol. 15(3), e1051.

Conaglen P and Gallimore A (2014) ‘Violence prevention: a public health priority’. Glasgow: NHS Scotland.

Home Office (2018) ‘Serious violence strategy’ (viewed on 21 April 2025).

Jeffery O, Bloom J, Eales T, Morgan N, Patman R, Gould W, Balachandran R and Pancheva P (2024) ‘Evaluation report on Grip and bespoke-funded hot spot policing’. Home Office Research and Analysis (viewed on 21 April 2025).

OSR (2024) ‘The quality of police recorded crime statistics for England and Wales’. Office for Statistics Regulation (viewed on 20 April 2025).

Sutherland A Dr, Makinson L, Bisserbe C and Farrington J (2023) ‘Hospital Navigators: multi-site evaluation of practices’. Youth Endowment Fund (viewed on 21 April 2025).

WHO (nd) ‘The VPA Approach’. World Health Organization (viewed on 21 April 2025).

YEF (2019) ‘YEF Toolkit’. Youth Endowment Fund (viewed on 21 April 2025).

Annex A: Quasi-experimental impact evaluation – additional detail

This technical annex provides detail on the data sources and analytical approach for the quasi-experimental designs implemented as part of the 2023 to 2024 evaluation of VRUs. Also included are additional outputs from the synthetic control group analysis and a comparison of hospital admission results across the evaluation years.

Data sources and analytical approach

Data sources

This section details the outcomes data accessed for the evaluation. This includes the geographical coverage and level of data, and key decisions taken when preparing data for analysis.

Hospital admissions data

Hospital admissions represent the most serious non-fatal violence, which is characterised by a violent injury that could not be treated in an emergency department. One study has found that approximately 30% of attendances to emergency departments for violent injury are admitted to hospital (Quigg and others, 2012). When a patient is admitted, their reason for admittance is recorded using the ICD-10 set of indicators. These appear in the NHS Digital records in several forms; ‘finished admissions episode’ is the most suitable way to identify these admissions. The admission is recorded against the month in which it concluded (as opposed to the admission date or the incident that led to the admission).

Therefore, it may not capture the small number of prolonged admissions that may have begun but not concluded during the data collection period and some admissions may have begun before introducing funding but concluded afterwards.

In the NHS Digital dataset, admissions are attributed to hospitals located within a PFA and LA and the month in which the admission concluded is recorded. Accordingly, admissions for violence in those hospitals are attributed to that PFA/LA, which allows the data to be used to examine differences in admission for violence in VRU and non-VRU areas and the timing of the admission closure (before or after SV funding) allows pre-post trends and a synthetic control group approach to be employed.

The analysis incorporated all hospital admissions for violent injury (ICD-10 codes: X92-Y09) and a subgroup, hospital admission for violent injury with a sharp object (ICD-10 code: X99).

Hospital admissions for any injury incurred in violence were not a primary outcome of the VRU programme, but given the range of VRU activities, it is theoretically plausible and very likely that the VRU activity could have impacted all violence. There are 16 categories of violent admissions in the ICD-10 catalogue. This includes varied mechanisms such as assault by bodily force; assault by different types of firearm; assault by drowning; assault by smoke, fire and flames; and 2 groups of ‘Other’ (specified and unspecified) assault mechanisms. We considered limiting the categories to assault mechanisms that could reasonably fit within a theoretical framework of VRU activity. However, because the range of activities by VRUs tends to aim to affect behaviour rather than the mechanism of violence, we could not confidently limit the data via theory.

We also considered using objective statistical techniques, identifying the point of maximum curvature using the dataset of admissions before SV funding (2012 to 2018), to limit the data set to the most common types of violence. Using this analysis, we identified a set of 6 admission injury mechanisms, which included ‘Other unspecified’ but excluded theoretically relevant mechanisms such as handguns and rifles. This was unsatisfying, as it was not theoretically sensible. In addition, such a statistical approach would create limits for the generalisability of findings beyond England and Wales. Accordingly, we opted to include all admissions for violent injury within the ICD-10 catalogue.

Homicides data

The Home Office provided homicides data at the PFA level. The data sources were:

• quarterly Homicide Index (HI) data covering fiscal Q1 April to June 2015 to Q2 July to September 2023

• monthly PRC homicides covering August 2013 to December 2023

The HI is continually updated (as police investigations and court cases progress) and is the primary data source for national publications on homicides. The data was used for the descriptive statistics and the programme-level synthetic control group analysis (using quarterly data) presented.

Recognising the low number of quarterly homicides, the evaluation team also conducted synthetic control group analysis on annual homicides. To meet the minimum number of pre-intervention periods required for the analysis, they aggregated PRC monthly homicides data to annual homicides.

To avoid biasing the impact estimates, the team removed homicides related to the following events from both data sources:

• in the financial year ending March 2017, 96 victims of the Hillsborough disaster and 4 victims of the Westminster Bridge attack

• in the financial year ending March 2018, 31 victims of the terrorist attacks that involved multiple victims, including the Manchester Arena bombing and the London Bridge attack, and 11 victims of the Shoreham air crash

• in the financial year ending March 2020, 39 victims found in a lorry in Essex

Police recorded crime (PRC) data

The Home Office provided monthly data of police recorded violence at the CSP level. Each PFA comprises multiple CSPs, which are broadly aligned to LAs. At the time of data collection, all police forces submit these data to the Home Office. Monthly counts for the following offence types were provided:

• violence with injury

• violence without injury

• possession of weapons offences

Data was also provided at the Lower Super Output Area (LSOA) level to facilitate exploratory analysis to understand the impact of SV funding in sub-areas with the highest (historical) levels of violence.

While Greater Manchester Police provided data for the first time since their systems change, the evaluators did not include these in the analysis. This was because of known under-reporting, which was clearly visible in the data and risked compromising the models.

Constructing the counterfactual

To estimate the impact of SV funding, a counterfactual (what would have likely happened in the absence of funding) was constructed through:

• where available data allowed, conducting analysis at the LA or CSP-level, where treated (VRU) and untreated (non-VRU) areas are more comparable in population size and outcomes of interest

• population adjusting outcomes data to make treated (VRU) and untreated (non-VRU) areas to improve comparability

• forming synthetic control groups (a weighted average of CSPs in non-VRU areas) that followed a similar pre-SV funding trend in (population adjusted) police recorded violence to CSPs in VRU areas

The first 2 points are discussed and illustrated below using PRC data. Focusing on LA-level data and population adjustment improved the comparability of hospital admissions data. Homicides data was only available at PFA level, but population adjustment was still able to improve comparability.

Analysis at CSP-level

PFAs were selected for SV funding based on levels of SV. The 18 PFAs experiencing the highest levels of SV received SV funding and established VRUs (this was later extended to 2 more PFAs). However, these higher levels of violence are driven by the larger populations in VRU areas.

Funded PFAs typically cover large geographical areas and/or have more densely populated urban centres (relative to non-funded PFAs). The box-and-whisker diagram below (Figure A.1) shows that the population distributions in funded (VRU) and non-funded (non-VRU) PFAs are not very comparable (limited overlap between the distributions).

Figure A.1: Distribution of PFA populations (excludes outlier Metropolitan Police (population 8.9 million))

However, if smaller and more similar-sized funded and non-funded geographical units are used (in this case, CSPs), comparability is improved. This is evident in Figure A.2 where there is greater overlap between VRU and non-VRU CSP population distributions than was observed in Figure A.1. It is still worth noting that CSP populations within funded areas were, on average, larger (reflecting a concentration of cities/large towns) than those in non-funded areas.

Figure A.2: Distribution of CSP populations

Population adjustment

Figure A.3 shows how population adjusting the outcomes can further support the construction of an appropriate counterfactual. The left-hand panel shows the average monthly count of police recorded violence with injury offences in funded and non-funded CSPs. Note that the left-hand panel very closely mirrors Figure A.2, highlighting the association between population size and outcomes of interest. The right-hand panel shows the same outcome but after population adjustment (average monthly rate per 100,000 persons), where comparability between funded and non-funded CSPs is substantially improved.

Figure A.3: Comparison of average violence and violence rates in VRU and non-VRU CSPs

Synthetic control method (SCM)

SCMs seek to construct an appropriate counterfactual by creating a synthetic control group, which is a weighted average of potential comparator areas. Comparator (non-VRU) areas (PFAs/LAs/CSPs) that are more similar to the treated (VRU) areas on pre-intervention (population adjusted) outcome trends receive a heavier weighting than those that are less similar.

As well as providing reliable impact estimates (O’Neill and others, 2020), SCMs can provide overall/average impact estimates for the entire treatment period, and average/cumulative impact estimates for each individual treatment period (such as months, quarters). The latter is particularly useful for VRUs where it is anticipated impacts will be gradual and accumulate over time.

The precise approach was generalised synthetic control groups (Xu, 2017), which allows for the inclusion of multiple treated areas (and constructing a synthetic control group for each) and staggered treatment starts. Diagnostic plots were assessed to ensure there was sufficient overlap between treated and comparator areas for impact estimates to be considered reliable – this is referred to as the ‘common support’ assumption.

Consideration for the new VRUs

A key consideration for the 2022 to 2023 impact evaluation, expanded upon in the 2023 to 2024 evaluation, was the 2 new VRUs, Cleveland and Humberside, moving from the comparator group to the treatment group. To navigate this and provide maximum insights, the evaluators tested the following treatment specifications in the synthetic control group analysis:

Model 1

Cleveland and Humberside initially in the comparison group but then switching to treatment group in 2022 to 2023 financial year; this model aligns to the VRU programme initiation dates in each PFA.

Model 2

Cleveland and Humberside in the comparison group only, because the 2 new VRUs were in a setup phase in 2022 to 2023, where direct interventions with young people and wider activity that could be expected to impact on violence outcomes were limited. This is consistent with the VRU Impact Feasibility Report, which recommended only testing for impacts on the original 18 VRUs after 18 months of programme delivery to allow sufficient time for the anticipated longer-term preventative effects of VRUs to materialise.

Model 3

Cleveland and Humberside excluded from the data to test for any potential contamination of estimates resulting from either of the above models.

Assumptions and limitations

A key consideration for the synthetic control group analysis was the impact of COVID-19 on violence. Specifically, the impact of lockdowns/restrictions on violence outcomes. It was theorised that restrictions on people’s movement reduced the opportunity for violence. To test this theory, Google mobility data was examined.

Figure A.4 shows that, on average, the LAs in VRU and non-VRU areas followed a very similar trend in residential movement (time spent at home) between February 2020 and December 2021. Regression analysis, which tested the interaction between VRU status and time on movement, confirmed there was no statistically significant difference between LAs in VRU and non-VRU areas (estimate = 0.00009, p-value = 0.89).

Figure A.4: Percentage change in residential movement in VRU and non-VRU areas

Recognising the above, the impact estimates (which are the differences between VRU areas and the synthetic control group constructed from non-VRU areas) can be considered net of any effects on violence resulting from changes in movement related to COVID-19. In other words, it was not necessary to adjust the synthetic control group analysis for changes in movement because VRU and non-VRU areas experienced very similar (and not statistically significantly different) trends in movement.

While the above showed the differences in movement between LAs in VRU and non-VRU areas were, on average, similarly affected by COVID-19, Google mobility data was not available at more granular levels to test this at the LSOA level. As such, some caution is advised with the analysis of ‘hot spots’, which used LSOA-level PRC data. It is possible that COVID-19 affected movement differently in LSOAs.

Another consideration was the level of analysis conducted to understand changes in violence trends. While monthly data was provided, the main analysis presented included aggregating the data to fiscal quarters. The team took this decision to better capture underlying trends in violence, which were less clear (owing to volatility) in the monthly data.

It is important to note the 2 main potential limitations of the police recorded violence data:

• not all violence is reported to the police by the public and the police do not detect all violence, and this is not captured in the data; increased levels of reporting and/or detection do not necessarily reflect a real increase in the levels of violence an area is experiencing – it is possible that VRU activity (for example, building relationships with communities affected by violence, better use of data) and Grip activity (additional enforcement) could affect the reporting and/or detection of violence

• changes in how the police record violence; this can include changes to and/or improvements over time in data capture systems

The police recorded violence data results should be interpreted with the above points in mind.

There are some limitations to using hospital admissions data as an indicator of SV in a PFA. Firstly, the catchment area of emergency departments is not necessarily co-terminus with PFAs. As patients typically travel to their closest emergency department for treatment (Haas and others, 2015), the hospital in which they were treated may not be in the PFA where the violent incident occurred. Similarly, patients injured in more rural areas – which are also likely to have lower levels of violence and therefore less likely to receive SV funding – have to travel to population centres such as towns and cities for treatment meaning that violence captured by the hospital(s) in a PFA may not have occurred in that area. Hospital admissions are also susceptible to resourcing issues that may affect their validity.

For example, hospitals with fewer resources and staff may be less likely to admit a patient for a violent injury than a hospital with more resources. In addition, the availability of a minor injury treatment centre within the hospital’s catchment area may reduce the number of patients attending (Rudge and others, 2013).

Additional outputs

The table below shows the estimates and inference across all outcomes and model specifications.

Model	Outcome	Estimate	Standard error	Lower CI	Upper CI	P-value
Model 1	Homicides	0.38	2.73	-5.56	5.38	0.98
Model 2	Homicides	-0.96	2.97	-7.03	4.4	0.71
Model 3	Homicides	-0.4	2.82	-6.08	4.46	0.7
Model 1	Hospital admissions sharp object	-0.75	0.72	-2.07	0.75	0.39
Model 2	Hospital admissions sharp object	-1.6	0.82	-3.16	0.04	0.06
Model 3	Hospital admissions sharp object	-0.93	0.70	-2.08	0.66	0.28
Model 1	Hospital admissions any violence	-9.12	5.06	-17.84	1.84	0.09
Model 2	Hospital admissions any violence	-10.39	4.75	-17.46	1.43	0.08
Model 3	Hospital admissions any violence	-10.05	5.05	-19.18	0.66	0.08
Model 1	PRC: Violence with injury	-24.87	111.26	-299.27	140.74	0.44
Model 1	PRC: Violence without injury	-721.73	470.34	1044.34	926.31	0.42
Model 1	PRC: Possession of weapons offences	-7.35	20.17	-46.03	32.71	0.65
Model 2	PRC: Violence with injury	43.09	116.5	-265.59	197.61	0.88
Model 2	PRC: Violence without injury	-1083.73	448.74	1411.49	431.3	0.18
Model 2	PRC: Possession of weapons offences	-62.31	31.26	-91.08	26.94	0.37
Model 3	PRC: Violence with injury	-23.18	115.75	-310.56	145.05	0.42
Model 3	PRC: Violence without injury	-728.63	479.78	1025.12	983.92	0.45
Model 3	PRC: Possession of weapons offences	-7.4	19.85	-45.29	31.86	0.63

The figures below show the synthetic control group analysis outputs for PRC outcomes.

Figure A.5: VRU and synthetic control group trends – Violence with injury – Model 2

Figure A.6: VRU and synthetic control group trends – Violence without injury – Model 2

Figure A.7: VRU and synthetic control group trends – Possession of weapons – Model 2

Comparison and interpretation of hospital admission outcomes

Explanation of statistical significance and 95% confidence intervals

A 95% CI is constructed from a procedure that produces intervals that contain the true effect size 95% of the time. If the CI does not cross zero, it indicates that the result is statistically significant at the 5% confidence level. This means that there is only a 5% chance that the observed effect (or more extreme) would have been observed if there were truly no effect. The decision to set CIs at 95% was made prior to analysis and sets a high and defendable benchmark, like standards used in medicine.

Comparison of 2022 to 2023 evaluation and 2023 to 2024 evaluation results

In evaluating the impact of VRUs, we consider reductions in hospital admissions due to sharp objects and any violence as favourable outcomes.

• primary outcome: hospital admissions resulting from violent injury with a sharp object

  • 2023 to 2024 evaluation: the estimate is -1.6 with a CI from -3.16 to 0.04

  • 2022 to 2023 evaluation: the estimate is -0.9 with a CI from -2.14 to 0.18

Figure A.8: Hospital admissions- Sharp object: Impact estimates by evaluation year

For hospital admissions due to sharp objects, the results from both evaluation periods suggest a reduction. While the CIs include zero (indicating that the results are not statistically significant at the 95% level), the direction of the estimates (negative values) is encouraging. The data suggest that VRUs may be having a positive effect, even though we cannot definitively rule out random variation. Importantly, the estimate for 2023 to 2024 is very close to statistical significance, which is particularly encouraging given that this is our primary outcome. Furthermore, sharp object incidents are low-count events, making impacts harder to detect statistically.

• secondary outcome: hospital admissions resulting from any violent injury

  • 2023 to 2024: the estimate is -10.39 with a CI from -17.46 to 1.43

  • 2022 to 2023: the estimate is -7.96 with a CI from -13.99 to -0.002

Figure A.9: Hospital admissions- Any violence: Impact estimates by evaluation year

For hospital admissions due to any violence, the 2022 to 2023 evaluation period shows a statistically significant reduction, as the CI does not include zero. This indicates that the intervention had a clear positive effect during this period. In the 2023 to 2024 period, while the central estimate suggests an even larger reduction, the CI includes zero, meaning the result is not statistically significant. It is clear here that there is considerable overlap between the 95% CIs, indicating that there is a large range of positive impact estimates consistent with both years. We should avoid a reductive interpretation based solely on the dichotomy of whether a p-value lies below a threshold. Taken in the round, it appears well supported that VRUs are having a positive impact.

Data coverage and context

The post-intervention data covers 3 years and 9 months for the 2022 to 2023 evaluation and 4 years and 9 months for the 2023 to 2024 evaluation. This duration is important because similar programmes, like the Cardiff Model and Scottish VRU, took around 5 to 10 years to demonstrate impact. Therefore, our findings within a shorter time frame are encouraging and suggest VRUs are on the right track.

Variation and additional data

It is also worth noting that as more data is added to the model, variation can increase (as well as decrease). This is a normal occurrence in statistical analyses and reflects the natural fluctuations in the data over time.

Interpretation and recommendations

Overall, the central estimates for both outcomes indicate reductions in hospital admissions, which is promising. The addition of an extra year’s worth of post-intervention data for the 2023 to 2024 period shows results that are consistent with previous findings but are not statistically significant. This consistency suggests that our intervention is likely effective, even if we cannot conclusively prove it with the current data.

It is important to remain grounded and not overstate the results. While the estimates and lower ends of the CIs are encouraging, they do not provide definitive proof of effectiveness. However, the similar pattern of results over 2 evaluation periods is a positive sign and helps make the argument for effectiveness more compelling.

Given that hospital admissions due to sharp objects are our primary outcome, and we see consistent reductions, this is particularly encouraging. The data for all violence, while not statistically significant in the latest period, still suggest positive trends.

In summary:

• the reduction in sharp object-related admissions is encouraging and close to statistical significance, which is notable given it is a primary outcome and a low-count event where impacts are harder to detect

• the statistically significant reduction in any violence-related admissions in 2022 to 2023 is promising and broadly consistent with that observed in this most recent year – we should continue monitoring these trends and consider these results as indicative of potential positive impacts, while remaining cautious in our conclusions

• the intervention is relatively young compared to similar successful programmes, and it is not uncommon for such programmes to take 5 to 10 years to demonstrate a clear impact

Funding allocations by VRU area

The table below details the funding allocations for each VRU area. These figures were used in the meta-regression to explore potential differences in impacts between area.

Area	Grip police force allocation	VRU allocation	Serious Violence Fund (total)
Metropolitan Police/ London	£8,907,532	£9,497,400	£18,404,932
West Midlands	£3,371,824	£4388,080	£7,759,904
Greater Manchester	£2,355,620	£4,388,080	£6,743,700
Merseyside	£1,746,172	£4,388,080	£6,230,120
West Yorkshire	£1,842,040	£4,388,080	£6,230,120
South Yorkshire	£1,217,527	2,163,720	£3,381,247
Northumbria	£1,079,203	£2,163,720	£3,242,923
Thames Valley	£861,445	£1,618,667	£2,480,112
Lancashire	£890,205	£1,528,667	£2,418,872
Essex	£795,707	£1,528,667	£2,324,374
Avon & Somerset	£799,815	£1,528,667	£2,324,374
Kent	£709,425	£1,528,667	£2,238,092
Nottinghamshire	£660,122	£1,058,313	£1,718,435
Leicestershire	£590,275	£1,058,313	£1,639,001
Bedfordshire	£580,688	£1,058,313	£1,639,001
Sussex	£545,080	£1,058,313	£1,603,393
Hampshire	£566,992	£1,058,313	£1,603,393
South Wales	£535,493	£1,058,313	£1,593,806
Cleveland	£524,536	£1,058,313	£1,582,849
Humberside	£490,298	£1,058,313	£1,548,611
Totals	£29,069,999	£47,577,000	£76,646,999