Study selection, data extraction and analysis
A two-stage screening process was conducted, identifying articles eligible for study involvement. Initially, titles and abstracts were reviewed independently by two reviewers and, if meeting inclusion criteria, articles were deemed eligible for the second phase of screening. Full texts were assessed against criteria to evaluate whether papers were suitable for study inclusion. Any discrepancies in screening or data extraction were discussed until a consensus was reached.
As part of a narrative synthesis a deductive thematic analysis (13) was conducted on studies to extract and categorise risk factors, in-line with the WHO Ecological Framework, used to guide violence prevention strategies (14). The model was generated to display the complexity of risk factors associated with interpersonal violence, and how this outcome is an interaction of factors across a variety of levels (14). The ecological model classifies risk factors into four main groups: individual, relationships, community, and societal – each category is further broken down by the WHO, to provide a list of key variables, and are displayed in Table 2(14). The categories within this model were used when interpreting the review’s findings. It was expected that the majority of risk factors identified within studies would align with the WHO framework. However, certain risk factors mentioned in the model were not captured within included studies (displayed in Table 2). This may be a result of the limited number of studies included within the review.
During the preliminary synthesis, the thematic analysis was conducted as papers were analysed by searching for risk factors that aligned with the themes developed from the existing WHO concepts(14) (Table 2). This was used to identify any patterns, similarities or differences of risk factors across the included studies. When identified, the primary author coded risk factors by colour and inputted into a spreadsheet. These were then tabulated to organise, present, and count findings following the WHO framework. Paragraphs were then drafted based on the frequency that a risk factor was mentioned across the studies. For example, seven studies highlighted ‘adverse childhood experiences’ as a risk factor, therefore these results were further explained. By using pre-determined themes, this research aims to guide recommendations based on a variety of studies, ensuring results can be utilised for future public health policies.
Meta-analysis was not conducted due to study heterogeneity in methodology and focus. Heterogeneity was assessed in the included studies for: clinical heterogeneity (the varied participant groups and outcomes assessed); and methodological heterogeneity (the differing WOE scores (Table 1) as well as study design, and statistical tests performed). Diversity of study is portrayed in Appendix B. This contains a summary table presenting key characteristics of included studies, providing information about study design, population, outcomes, and statistical tests when appropriate. This was reported prior to the preliminary synthesis.
To assess the robustness of the narrative synthesis, the primary research articles included were quality appraised, at study level, by the primary author using the Weight of Evidence approach (WOE) (15). This method was selected as a variety of different study designs were collected with a range of information available. This process allows for integration of evidence obtained from various results and methodologies when answering the proposed research question. This form of analysis assesses the overall quality of each paper depending on the following three criteria: WOE A (used to examine the clarity and accuracy of information); WOE B (assessing appropriateness of study methodology); and WOE C (how relevant study findings are to this systematic review). For each study, these three judgements were then combined to provide WOE D: an overall assessment of quality and relevance of evidence for risk factors associated with weapon usage and knife crime (Table 1). Research papers were ranked as either high (H), medium (M), or low (L) and were included only if overall quality (WOE D) rated as high or medium. These rankings were used to evaluate strength of key findings and results. Where there was uncertainty concerning study quality following the above criteria, the primary author liaised with co-authors to clarify ranking. This aimed to ensure reliability of the quality appraisal.
The Synthesis Without Meta-analysis (SWiM) guideline(16) was used to ensure clarity when reporting methods, including the narrative synthesis, and results of the review. A supplementary table outlining the checklist and where to find the reported information within the manuscript can be found in Appendix C.
A total of 2,335 articles were originally identified from the initial search (PRISMA flowchart (Figure C)), from which 622 duplicates were removed, resulting in 1,713 articles taken forward for abstract review. Stage one review excluded a further 1,665 articles and stage two, 31. Seventeen articles were included in the WOE quality assessment, where only one paper was excluded from the review with a quality ranked as low (L). Therefore, 16 articles were included in the systematic review. A quantitative analysis was not undertaken due to the heterogeneity of studies included (Appendix B), and therefore a narrative synthesis of risk factors influencing weapon-related crime was conducted.
Study characteristics
Of the 16 articles included, one was an intervention study evaluating current mitigation strategies for gang-related crime in London (8), while all other studies were observational. Included within the review were: three literature reviews (17–19); three cohort studies (20–22); three qualitative interview studies (8,23,24); three case-control studies (25–27) 2017; and four cross-sectional studies (29–32). All studies were conducted within the UK: four within the UK; one in Birmingham; one in Edinburgh; five in London; one in Wales; and one in England. Appendix B presents detail characteristics of each study, including study size and key findings.
Table 2 displays which risk factors - associated with youth violence - are described in papers and whether there is a positive, no, or unclear association . For papers including a quantitative analysis, risk factors with statistically significant results were categorised as positively associated. Whereas for qualitative studies, risk factors mentioned within interviews or literature reviews were identified as positively associated. The design of each study is further highlighted within the table. The count of studies mentioning each risk factor was used to determine the relevance for discussion.
To provide a rich reflection of evidence for the reader, we included both quantitative and qualitative studies as there is comparatively little evidence published on this topic within the inclusion parameters specified. Deductive thematic analysis was completed for qualitative papers using themes as set-out by the WHO. Therefore, if these themes/risk factors were identified within the paper, the paper was incorporated into Table 2 irrespective of study design.
As previously mentioned, risk factors were divided into four categories, and subcategories within these, following the Ecological Framework put forward by the WHO (3): individual; relationships; community; and societal. WHO violence risk factors not identified within included studies are also shown. Each subcategory of risk factor is discussed in the subsequent sections. As mentioned previously, a paper focusing on education and crime age profile was not included within the review due to ranking low quality (28).
Individual risk factors
Demography
Of the 16 studies included within the review, 10 investigated the association of demography with knife crime (Table 2). All six studies investigating age found a positive association between knife crime and adolescence (21,23,26,29,31,32). A 2006 cross-sectional study, using the Juvenile Attitudes Towards Weapon Scale (derived from the Attitudes Toward Guns and Violence Questionnaire), found that the prevalence of weapon carrying increases with age: 30% for individuals aged 11-13 years, 38.2% at 14-15, 47.4% at 16-17, and 52.6% at ages 18-19 (31) (WOE=H). Regarding gang violence, three studies found a young person, compared to other age groups, is positively associated with being in a gang (23,26,33). Densley et al (WOE=M) interviewed 69 self-described gang members, recruited from six London boroughs experiencing high levels of socioeconomic deprivation, with an age range of 13-34 (mean age 20). A participant aged 25 years described gang life to be a ‘young man’s game’ and ‘when you’re younger it’s about what you’ve got now and how fast…when you’re older you’ve got more to lose’ (23).
Results were mixed regarding the association with gender. One cohort and one cross-sectional study showed males were more likely to be associated with knife crime (22,29). However, a cohort study of in care homes suggested females were more likely to offend at a younger age (21)(WOE=M). A cross sectional-study looked more closely at the characteristics of weapon-carrying, e.g. type and use of weapon, and found no significant difference on the basis of gender: although 27% of males used their weapon to injure compared to 19% of females (31). No significant association between gang violence and gender was found in the three papers exploring this issue (24,26,32). For example, both males and females expressing the need to be the ‘biggest, baddest and the most untouchable’ (26) (WOE=M).
A cohort study found no association between knife crime and the ethnicity of the victim or perpetrator when controlling for confounders, including sex and family structure at an individual level and neighbourhood deprivation at a community level (17)(WOE=M). However, a literature review suggests that migrants and refugees may be at higher risk of victimisation of weapon-related crime (20) which may explain the overrepresentation within the media. A cross-sectional study of 797 school students and a case-control with 188 young offenders completed self-reported questionnaires and, through use of pre-determined criteria, were divided into categories depending on gang involvement. Comparisons of a variety of groups, rather than investigating one specific cohort of gang members, allows for differences to be evaluated. As neither studies identified a difference between the ethnicity of groups divided by level of gang involvement, this suggests no association with gang violence (20,25,32).
Adverse childhood experiences (ACE)
All seven studies investigating the association between teenagers with ACEs and weapon-related crime reported a positive association (Table 2). A cross-sectional study of 20 males convicted of homicide during their adolescence investigating adolescent homicide found that 25% of perpetrators had experienced either sexual or physical abuse and 90% were known to social services (30) (WOE=M), and all 20 had previously experienced neglect or parental separation (30); a higher prevalence compared to the general population. Within a study based in a care home, 91% of young people who had been convicted of a crime (the majority of which were violent or weapon-enabled) had experienced multiple placements (range of 1-30, mean of 8) (29) (WOE=M). Wood J et al (WOE=H) showed in a cross-sectional study that gang members and ‘gang affiliates’ self-reported more childhood traumatic events and were more likely to have been placed in local authority care compared to violent men not in a gang (27).
Education
Three studies investigated the impact of school exclusion on involvement in knife crime (19,21,30). However, one study did not show a clear association between education and gang membership. During qualitative interviews current and previous gang members expressed their opinions that school achievements and successful routes through education were unattainable (24). On the other hand, some gang members had obtained GCSEs and were still involved in criminality (24).
Clement et al identified in a Bristol-based study that 80% of younger offenders had previously been excluded from school, suggesting a link between school exclusion and involvement in violence(19). In contrast, Hayden et al found similar rates of school exclusion between offenders and non-offenders within a care home setting (40% for non-offenders and 44% for offenders)(21). However, as all individuals were removed from their family home this may affect the findings.
Mental health
Three studies investigated mental health associated with knife crime and both described poor mental health (suicide/depression/self-harm described by participants) as a risk factor(27,30,31). A cross-sectional study of 20 adolescents committing homicide revealed that all participants suffered from high levels of interpersonal conflict and psychological vulnerabilities (30). A case-control study of 1,539 men found that self-identified gang members and gang affiliates had a higher prevalence of psychological issues, including anxiety, psychosis and suicide attempt, than violent men not involved in gangs (27). For the particular study, gang affiliates and members both were involved in gang-related activity, however, categories differed as gang affiliates did not identify as a gang member.
Victimisation
Three studies investigated a link between previous victimisation of weapon-related crime and offending (20,27,31) that among young people self-reporting weapon possession within the last six months, ‘reactive weapon carriers’ - where weapon was used in conflict resolution, or user was a victim of threat, and/or injury - reported previously being a victim of threat or injury. Smith D et al (WOE=M) found that victims of bulling were also more likely to offend and be victims of weapon-related crime. Furthermore, gang members were more likely to be targeted as victims and self-reported more serious injuries compared to non-gang members (31). Gang affiliates also reported more incidents involving physical attacks compared to violent men who were not part of gangs, however unexpectedly more than gang members (27). Results found individuals who self-reported victimisation were more likely to offend and vice versa, therefore a bidirectional relationship may exist between being a perpetrator and victim of weapon-related crime.
Relationships risk factors
Seven studies reported peer influence as an important risk factor for knife crime (Table 2). For example, Barlas et al (WOE=H) described self-identified ‘offensive weapon carriers’ – those who carry weapons to injure or threaten – considered peer influence an important reason for weapon carrying. In conjunction with this, two studies showed peer influence as an important risk factor for gang membership(25,26). A majority (65%) of self-described gang members, and 57% of self-described members who also met Eurogang definition, identified one reason for joining was ‘because a friend was a member of the group’(25).
Two studies investigated parental relationships in association with knife crime and all identified that strong parental attachment acted as a protective factor (20,22). At age 15, a cohort study found conflicts with parents increased risk of victimisation and offending (20).
Community
Six studies included in this review investigated the impact of deprivation, all of which showed a positive association with knife crime. During interviews, Densley et al (WOE=M) concluded that areas with low socioeconomic status increase risk of gang involvement and with one interviewed gang member describing London communities as ‘built to encourage crime’ (23). According to a literature review, crime rates are highest in areas of economic deprivation, increasing the chance of adolescents’ involvement within violent crimes (18). Alongside this, individuals who identified as ‘defensive weapon carriers’ expressed the need to carry weapons for personal safety in high crime areas (31). Deprivation can further result in low social cohesion which has further been associated with offending behaviour of adolescents and gang members(32).
Societal
Three studies suggested a positive impact of economic deprivation and knife crime (Table 2). Densely et al explains how economic inequalities have forced young men into ‘self-destructive behaviour’ as the societal problems have left individuals with minimal options (23). Four studies showed a positive association between stigma and discrimination and weapon-related crime (8,23,25,32). A cross-sectional study of 797 secondary school students found negative perceptions of authority were highest in gang members, followed by peripheral youth (individuals involved within gang-related activity, but not classified as members), and lowest in non-gang youth (32). Furthermore, within interviews gang members described themselves as ‘urban outcasts’, explaining that ‘[they’re] automatically stereotyped, it’s like all black people are criminals… after a time you feel like ‘oh we a gang now? Ok we’ll show you gang’ (23).
Three studies described violence and weapon carrying as a method of gaining status, power, and masculinity (19,31,32). For example, Barlas et al (WOE=H) found young people explained the most common reasons for weapon carrying were: ‘for looking cool’, ‘other people’s respect’, ‘feeling powerful’, and ‘peer admiration’(31). Within five studies, gangs were described as providing identity, status, and companionship with membership proving as a method to build an individual’s reputation (19,24–26,32). For example, gang members have expressed the desire to ‘win approval from peers’ and two studies found that young gang members perceived social status as more important compared to non-gang-involved adolescents (19,31).