The Relevance of Prior Offending to Risk and Need in Youth Family Violence: A Population Cohort Study

The present study examines differences in the characteristics and recidivism risk of young people reported to police for family violence (FV) with a history of prior offending (generalists) and those only known to police for using FV (family-only). A population-based cohort of youth aged 10–24 years (N = 5014) who were reported to police for using FV over a four-month period in 2019 was examined and FV-related risk and need data extracted, with a six-month follow-up period for further police-reported FV. All data was extracted from police databases. Logistic regression with odds ratios as a measure of effect size were used to compare generalist and family-only cohorts. Cox proportional hazards were used to assess time to FV recidivism among the two cohorts, and to assess whether diversity of prior offending was associated with risk of FV recidivism among generalist youth. Generalists were more likely than family-only youth to be recorded as using FV in a high severity FV incident, be abusive across multiple relationships, and breach court orders. Generalists experienced a greater level of need and were more likely to engage in FV recidivism, and do so more quickly, than family-only youth. Diversity of prior offending among generalists was positively associated with risk of FV recidivism. Compared to family-only youth, generalists represent a higher risk cohort with a greater level of need. History of prior offending among young people may be a simple and efficacious means of prioritising higher risk youth who use FV.

Youth family violence (FV) constitutes approximately 8-10% of police-reported FV incidents involving adolescents (aged 10-17 years; Phillips, & McGuinness, 2020;Snyder & McCurley, 2008), while 26% of all police-reported FV incidents involve those aged 20-29 years (Coghlan & Millsteed, 2017). Young people who engage in FV often do so as part of a broader pattern of antisocial behaviour (Moulds et al., 2019), yet limited research directly compares youth who engage in both FV and other offending (generalists) with young people who only engage in FV (family-only; see Ibabe and Jaureguizar, 2010;Moulds et al., 2019).
Broad definitions of FV are increasingly being adopted in recognition of its complex and multifaceted nature (Farrington & Ttofi, 2021;Jolliffe Simpson et al., 2021). These definitions conceptualise FV as involving abuse toward relatives (e.g. parents, siblings, other relatives) and abusive behaviour towards dating or intimate partners, with both physical (e.g. physical assault, sexual assault, etc.) and nonphysical (e.g. psychological abuse) behaviour recognised under the umbrella of FV. Additionally, in many jurisdictions (Jolliffe Simpson et al., 2021;Miles & Condry, 2016;Spivak et al., 2021), legislative definitions of FV include behaviours that are not otherwise criminal (e.g. verbal abuse such as swearing and screaming at the victim). For example, in the Australian state of Victoria, only half (50.8%) of all police-recorded FV incidents in Victoria between 2020-21 involved a criminal offence for which charges were laid (Crime Statistics Agency, 2021a).

3 The Interconnectedness of Family Violence and Antisocial Behaviour
Family violence and general antisocial behaviour are often examined separately, yet are interconnected (Farrington and Ttofi, 2021;Gottfredson & Hirschi, 1990). General theories of crime suggest that the many forms of offending and violence stem from similar underlying risk factors, traits, interactions within one's environment (Gottfredson & Hirschi, 1990;McCloud, 2021). Both the general and FV offending literatures highlight factors such as low levels of self-control (Gottfredson & Hirschi, 1990;McCloud, 2021), issues with social bonding (Hirschi, 1969;Foshee et al., 1999), social learning (Foshee et al., 1999;McCloud, 2021), and high levels of environmental strain (Agnew, 2006;McCloud, 2021) as relevant to engagement in offending and youth FV.
Additional risk factors common to both general offending and FV among young people and adults have been identified in the literature. These include an early-onset and diverse pattern of antisocial behaviour (Farrington and Ttofi, 2021;Piquero et al., 2012;Verbruggen et al., 2021), mental health issues (Kennedy et al., 2010;Simmons et al., 2018), substance use issues (Simmons et al., 2018;Sjödin et al., 2017), a history of victimisation (Kennedy et al., 2010;Verbruggen et al., 2021), school problems (Simmons et al., 2018), and unemployment (Sjödin et al., 2017). These risk factors are more common among young people who engage in FV and other offending behaviour, compared to those who only offend outside the family context (Ibabe and Jaureguizar, 2010;Kennedy et al., 2010).

Family-Only and Generalist Family Violence Types in Adult Samples
There is a high co-occurrence of FV and general offending behaviour among both adult (Dowling et al., 2021;Hilton & Eke, 2016) and youth samples (Moulds et al., 2019;Verbruggen et al., 2021). Adults who engage in both FV and other offending (i.e. generalists) comprise between 42 -96% of all FV users (Dowling et al., 2021;Goldstein et al., 2016;Hilton & Eke, 2016;Petersson et al., 2019), while family-only adults comprise approximately 47.5% of all adult FV users (Petersson & Strand, 2020). These results indicate two groups of FV-users, with prior research suggesting that the identification of individuals using the generalist/family-only typology is essential for assessment and intervention (Petersson & Strand, 2020).
Generalists display a greater level of violence-associated risk factors than family-only individuals, indicating an elevated level of risk and need (Goldstein et al., 2016;Petersson & Strand, 2020). The concepts of risk and need are drawn from the Risk-Need-Responsivity (RNR; Bonta & Andrews, 2016) framework. This framework highlights the importance of matching interventions to the individual's risk of recidivism (i.e. high risk offenders receive more intensive support; risk principle), and that criminogenic needs (e.g. substance abuse, history of antisocial behaviour, education/employment) should be targeted as part of any intervention (need principle;Bonta & Andrews, 2016).
The higher level of risk and need among generalists is demonstrated by their frequent and chronic offending Verbruggen et al., 2021), breach offences (Coghlan & Millsteed, 2017), and higher levels of FV and non-FV recidivism (Petersson & Strand, 2020). The elevated level of criminogenic need among generalists is demonstrated through their more extensive history of antisocial behaviour (Petersson & Strand, 2020), elevated level of substance abuse problems (Coghlan & Millsteed, 2017), higher levels of unemployment (Coghlan & Millsteed, 2017;Petersson & Strand, 2020), and greater association with antisocial peers (Petersson & Strand, 2020). Similarly, generalists display a higher level of non-criminogenic need -which is important for the purposes of risk management and intervention -compared to family-only individuals, including elevated levels of mental health and substance abuse problems (Coghlan & Millsteed, 2017) and more significant victimisation histories (Petersson & Strand, 2020;Verbruggen et al., 2021).
Youth-specific research examining the presence of generalist and family-only subtypes is needed as much of the adult literature is only applicable in a limited capacity. Much of the research pertains to physical intimate partner abuse by males, whereas most youth FV involves child-toparent abuse and is less gendered than adult FV (Phillips & McGuinness, 2020). Second, young people are undergoing significant developmental changes which can impact applicability of adult research to youth samples (Borum, 2000).

Family-Only and Generalist Subtypes in Youth Samples
The examination of family-only and generalist subtypes among youth is limited, despite being recognised as important for risk assessment and treatment (Boxall & Sabol, 2021;Petersson & Strand, 2020). There is emerging literature comparing generalists to non-FV offenders (Kennedy et al., 2010;Kuay et al., 2016;Sjödin et al., 2017), however direct comparisons of generalist and family-only youth are limited. Ibabe and Jaureguizar (2010) examined the files of Spanish youth aged 14-18 years charged with parent abuse. the authors found that family-only youth (i.e. those only charged with parent abuse) were more likely to be female, experienced better family economic status, displayed fewer disruptive behaviours in the school, and had a less extensive history of parent abuse than their generalist counterparts (i.e. those charged with both parent abuse and other non-FV offending; Ibabe and Jaureguizar, 2010). Moulds and colleagues (2019) analysed the data of Australian young people aged 10-17 years (n = 305) who had been apprehended by police for violence against a parent or stepparent. The sample was organised according to whether they had only ever been arrested for a single incident of parent abuse, both parent abuse and non-violent offences, or parent abuse and other violent offences. The authors found that parent abuse in isolation was rare, most youth who were arrested for abusing their parents went on to offend violently, and prior offending was a strong indicator of future offending (Moulds et al., 2019).
The results of these two studies (Ibabe and Jaureguizar, 2010;Moulds et al., 2019) suggest generalist and familyonly youth may represent two distinct cohorts. However, the generalisability of these findings is limited due to their specific focus on child-to-parent abuse among young people under the age of 18 years and their use of data drawn from official charges. Young people who use FV may engage in abusive behaviour toward multiple victims (Boxall & Sabol, 2021;Kuay et al., 2016) and research indicates less than half (48.8%) of police reports for FV in some jurisdictions result in criminal charges (Crime Statistics Agency, 2021a). As a result, these methodologies are significantly less likely to capture the broad range of behaviours which constitute FV. Similarly, police, youth service providers and justice agencies in Australia are often required to engage young people up to 25 years old (McGorry et al., 2022) who participate in a range of abusive and antisocial behaviour. As a result, additional research is needed to broaden both the scope of FV behaviour that is considered, as well as the age of FV users, to ensure a more comprehensive understanding of the generalist/family-only phenomenon.

The Present Study
This study adds to the relatively limited existing literature on general offending by young people who engage in FV. We sought to determine how the characteristics of young people reported to police for FV who have a history of prior non-FV offending (generalists) differed from those who are known to police only for engaging in FV (family-only). We conducted exploratory analyses comparing the individual characteristics and FV incident characteristics of generalist and family-only youth. We also aimed to test the following hypotheses informed by the extant research with adults and adolescents: Hypothesis 1: Generalist youth will have a more extensive lifetime history of FV perpetration compared to familyonly youth, as indicated by a greater likelihood of using FV in a high severity FV incident, increased likelihood of being reported as abusive across more than one relational dyad, and increased likelihood of breaching court orders related to FV. This hypothesis is drawn from literature suggesting both adult (Petersson & Strand, 2020) and youth (Moulds et al., 2019) generalists display a more extensive history of FV behaviour, and adult generalists display a greater number of breaches (Coghlan & Millsteed, 2017;Morgan et al., 2018) than family-only individuals. Hypothesis 2: Generalist youth will display a greater level of need than family-only youth, as indicated by higher levels of police-identified mental health issues, substance abuse, issues with school truancy/unemployment, and victimisation histories. Research shows generalists display more mental health issues (Kennedy et al., 2010) and school/employment issues (Ibabe & Jaureguizar, 2010) than non-FV offenders and family-only youth, while the adult literature shows a higher level of substance abuse and victimisation among generalists than family-only individuals (Petersson & Strand, 2020). Hypothesis 3: Generalist youth will display more FV recidivism compared to family-only youth, as indicated by being more likely to be identified as using abusive behaviour in a subsequent FV incident and in a high severity FV incident during the follow-up period. Generalist youth will also engage in FV recidivism more quickly than family-only youth. Adult generalists are significantly more likely to reoffend than family-only individuals, and to display more severe FV behaviour (Goldstein et al., 2016;Petersson & Strand, 2020). Hypothesis 4: A greater diversity of prior offending behaviour among generalist youth will be associated with shorter time to FV recidivism. Family violence research suggests greater diversity of offending is associated with higher frequency of FV (Boxall et al., 2015) which, in turn, is associated with shorter time to FV recidivism .

Sample
This study involved analysis of all police-reported FV incidents in the Australian state of Victoria between 1 September and 31 December 2019 (index period) in which a young person aged 10 to 24 (inclusive) was listed as the person responsible for aggression (the respondent, in police parlance) (N = 5014). Apart from young people being 10-24 years of age at the time of the index incident, no other inclusion or exclusion criteria were used. The age range was chosen because the minimum age of criminal responsibility in the jurisdiction from which these results were drawn is 10 years (Children, Youth, and Families Act, 2005, s.344) whilst the inclusion of young people up to, and including, 24 years old is in recognition that youth services are increasingly being required to engage with young adults (up to age 25 years; McGorry et al., 2022). The World Health Organisation (WHO) has formally recognized the parallels between those in adolescence and young adulthood, defining 'young people' as those aged 10-24 years (WHO, 2021).
The sample was drawn from all 24,419 FV incidents recorded by police during the same period, with the exception of 358 (1.50%) that were missing respondent age and so were excluded. Victoria Police are the sole policing agency for the Australian state of Victoria (a jurisdiction approximately the same geographic size as the United Kingdom with a population of 6.63 million at the time of the study; 66.99% of whom live in the capital city of Melbourne; Australian Bureau of Statistics (ABS), 2021). Victoria Police record all incidents of family violence as a matter of policy, regardless of whether charges were laid.
Young people who used FV were identified at the time of their index FV incidents and their data were linked to lifetime historical data and outcome data in the six-months following the index FV incidents. Over two thirds of FV users were male (n = 3528, 70.40%) with the remaining being female (n = 1484, 29.61%). There were two cases (0.04%) in which the sex of the FV user was not specified. The mean age of FV users was 19.19 years (SD = 3.51) at the time of the index incident. Victims were primarily female (n = 3660, 73.00%), with three cases (0.06%) not specifying victim sex. The mean age of victims was 33.49 years (SD = 15.45).

Definition of Family Violence
Victoria Police use a definition of FV from the Family Violence Protection Act 2008 to identify when to record an incident of FV. The Act defines FV as: "behaviour by a person towards a family member of that person if that behaviour is physically or sexually abusive; or is emotionally or psychologically abusive; or is economically abusive; or is threatening; or coercive; or in any other way controls or dominates the family member and causes that family member to fear for the safety and wellbeing of that family member or another person" (Family Violence Protection Act, 2008, s.5). Under the Act, the term family member refers to relatives, intimate partners, children who normally reside with the victim and/ or FV user as well as "any other person whom the relevant person regards or regarded as being like a family member" (Family Violence Protection Act, 2008, s.8), such as foster carers. Importantly, not all forms of FV recorded under the Act involve a criminal offence (e.g., there is no specific offense relating to psychological abuse or coercion in Victoria). In Victoria, only half (n = 47,468, 50.8%) of all FV incidents between July 2020 and June 2021 involved a criminal offence for which charges were laid (Crime Statistics Agency, 2021a).

Data Source
All data were extracted by Victoria Police staff from the Law Enforcement Assistance Program (LEAP), an electronic database used by Victoria Police to record all known offences and police involvements, regardless of outcome (e.g. arrested, charged, convicted). All data from the first FV incident involving a unique dyad during the index period was collected. Whenever Victoria Police members respond to an incident of FV, they record characteristics of the incident, the victim, the person using FV, and their relationship as part of a FV report (see the Victoria Police Code of Practice for the Investigation of Family Violence, s.3, for a description of the FV recording procedure employed at the time of this study). In addition to the demographic information of the young person who used FV and victim, the FV report contains 39 separate risk factors associated with future FV behaviour or lethal FV incidents which police score at the time of the index incident. This allows Victoria Police to collect information on a range of evidence-based factors related to future FV events (Spivak et al., 2021). All available data from the FV reports involving a unique relationship dyad during the index period was linked to historical and outcome data for each young person and victim.

Demographic Variables
The sex and age of the victim and young person who used FV were recorded on the FV report at the time of the index incident, as well as accessibility needs and relationship type. Sex was coded in as either male or female whilst age was analysed as both a continuous variable and as a categorical variable with three levels: 10-14 years, 15-19 years, and 20-24 years. Accessibility needs are also recorded by police to identify whether the young person has issues relating to vision, hearing, mobility, understanding, communication, or memory. The relationship of abuse was recorded by the responding police officer according to the type of relationship between the victim and FV-user (i.e. child-to-parent abuse, intimate partner abuse, sibling abuse, other family abuse) at the time of the index incident. The category relating to abuse of other family members may include grandparents, uncles, aunts, cousins, carers, or the child of the young FV-user. Only a single relationship of abuse can be recorded on a FV report. In cases where a young person was abusive toward more than one family member (e.g. both a parent and a sibling), an additional FV report is created.
Socioeconomic status (SES) was approximated using the postcode recorded on the FV report and comparing it with the ABS (2018) Index of Relative Socio-Economic Advantage and Disadvantage (IRSAD; using Victoria-specific rankings). This resulted in the identification of the postcode's IRSAD decile, allowing researchers to code whether an incident occurred in an SES area identified as being in the lowest 20% (deciles 1 and 2), middle 60% (deciles 3-8), or highest 20% (deciles 9-10) of the state.

Diversity of Prior Non-Family Violence Offending
Comparison of the generalist and family-only youth is done using a dichotomous variable, in which a young person is classified as either a 'generalist' or a 'family-only' youth. The diversity of prior non-FV offending refers to the number of different offence classification types (excluding FVrelated offences) that the young person engaged in prior to the index FV incident. The Victorian Crime Statistics Agency offence classification scheme (Crime Statistics Agency, 2021b) was used, resulting in a score between zero and six. Offences categories included: crimes against the person, property and deception offences, drug offences, public order and security offences, justice procedures offences (e.g. public nuisance, breaches of orders), and other offences (e.g. regulatory driving offences, miscellaneous offences).

Lifetime History of Family Violence Variables
Variables relating to lifetime history of FV represent the combination of all historical FV-reports and index incident information (but not from the six-month follow-up period). These variables are coded dichotomously (i.e. Yes or No).
The variable 'ever abusive across more than one relational dyad' referred to whether a young person has ever (i.e. including both the index incident and any past FV incidents) been reported to police for using FV within two or more of the following relational dyads: child-to-parent abuse, intimate partner abuse, sibling abuse, other family abuse. For example, a young person who has historically been reported to police for being abusive toward their mother, and who then was abusive toward their father at the index incident, would not be identified as having used FV across more than one relational dyad, as both of these are representative of child-to-parent abuse. However, a young person who was previously abusive toward their mother and then used FV toward an intimate partner at the index incident would be identified as abusive across more than one relational dyad.
A high severity FV incident refers to an incident that was associated with a charge (including charges not subsequently authorised for prosecution) in any of the following categories: violent offences (indictable physical assault (excluding unlawful assault); homicide; armed robbery; robbery; aggravated burglary; false imprisonment/kidnap); stalking; threats to harm or kill; sexual offences (including rape and non-rape sexual offences, but excludes charges related to possession or online access, solicitation or distribution of child abuse material); arson causing death or endangering life; and driving offences causing death or endangering life. The rationale for classifying certain family violence incidents as high severity was to distinguish between incidents that involved physical harm, the threat of physical harm, and/or that had the potential for physical for substantial psychological harm (e.g., stalking, sexual offences) and those that did not. The charges that were included in the definition of a high severity family violence incident were determined by authors TM and BS in consultation with Victoria Police based on the above and the fact that the charges are indictable rather than summary offences.
The variables 'ever used physical abuse', 'ever used sexual abuse', and 'ever charged with stalking' refer to whether the young person used these forms of abusive behaviour in past FV incidents, or at the time of the index FV incident. The variables 'ever charged with breaching a court order' or 'ever charged with contravention of a restraining order' are dichotomous (i.e. Yes or No) and were extracted from the 39 questions asked as part of the FV report completed by responding police officers at the time of index incident.

Victimisation, Mental Health, Substance Abuse, and Unemployment/School Truancy
Each of the variables pertaining to the victimisation history of young people who use FV have been dichotomised (i.e. Yes or No) and represent whether a young person has ever been identified as the victim of a police-reported FV event at the time of index FV incident.
Binary dummy variables for each family relationship were created from the variable 'relational dyad in which young person has been victimised' to ascertain history of past FV victimisation. The variable 'victimised across > 1 relationship dyad' referred to whether a young person had ever been reported to police as the victim of abuse across at least two of the following: parent-to-child abuse, intimate partner abuse, sibling abuse, other family abuse.
Variables pertaining to the presence of mental health issues, substance abuse, and unemployment/school truancy were recorded in a binary manner (i.e. present or absent). These variables were derived from the 39 questions on the FV report, which police scored based on their questioning and discretion (i.e. asking the young person and the victim about whether the young person uses substances, noticing if the young person appears substance affected at the time of the incident). A formal diagnosis was not required for the presence of a mental health or substance abuse issue to be scored in the affirmative. As a result, prevalence of these issues in the present study are likely to represent relatively gross estimates.
The variables 'experiences substance abuse problems' and 'issues with school truancy/unemployment' were used in the present study as broad indicators of criminogenic need. The variables 'experiences mental health issues and 'historically been the victim of an FVI' are used as broad indicators of non-criminogenic need. Each of these variables were coded dichotomously. The broader offending literature recognises the domains of substance abuse and education/employment are identified as criminogenic needs, while mental health issues and history of victimisation are identified as non-criminogenic needs, or responsivity issues (Bonta & Andrews, 2016).

Family Violence Recidivism Variables
FV recidivism data was obtained from any FV reports uploaded to LEAP in the six months following the index incident in which the young person was reported for using FV again toward any person. All FV recidivism variables were coded dichotomously.

Ethics
The study was approved by the Blind for Review and the Blind for Review Human Research Ethics Committee.

Data Analysis
Univariate analyses (i.e. descriptive statistics, chi-square tests, and t-tests) were conducted using IBM SPSS Statistics, version 28 (2020), whilst logistic regressions and cox regression were undertaken using R (R Core Team, 2020). Statistical packages used to analyse data in R include rms (Harrell, 2018), Hmisc (Harrell and Dupont, 2018), and dyplr (Wickham et al., 2018). Binary logistic regression with odds ratios as a measure of effect size were used to compare generalist and family-only cohorts in terms of lifetime history of FV behaviour (hypothesis one), level of need (hypothesis two), and recidivism characteristics (hypothesis three). Lifetime history variables combined data from both historical incidents and the index incident. Logistic regressions were initially run with a single predictor variable (i.e. generalist vs family-only). However, adjusted odds ratios controlling for age and sex of the young person who used FV were also produced for each model given generalist youth (N = 2609, M = 20.05, SD = 2.97) were significantly older than family-only youth (N = 2405, M = 18.27, SD = 3.81, indicating a statistically significant difference M = 1.78, 95% CI [1.59-1.97], t(5012) = 18.51, p < 0.001. Generalists were also significantly more likely to be male (χ 2 (1, N = 5012) = 70.42, p < 0.001, OR = 1.69, 95% CI [1.49-1.91]).
Logistic regression assumes a linear relationship between continuous independent variables (e.g. age of young person who used FV) and the log odds of the dependent variable. A Box-Tidwell test indicated a significant result for both the main term (p < 0.001) and the interaction term (p < 0.001) when age of the young person was examined, suggesting a non-linear relationship. Restricted cubic splines were fitted to the age variable to address non-linearity. A cubic spline is a piecewise cubic function used to estimate nonlinear associations in regression analyses, with piecewise functions referring to statistical techniques in which separate slopes are fitted to model various areas of the outcome. This allows a flexible means of modelling relationships that are not adequately accounted for by polynomial transformations (e.g. squaring the predictor values). The separation points for each slope in the piecewise function are referred to as knots, with three knots applied in the present paper.
Post-hoc analyses examined differences between generalist and family-only youth for individual and index incident variables, with Bonferroni corrections applied and provided at the bottom of the relevant results tables. Analyses related to age of the young person (10-14 years, 15-19 years, and 20-24 years) and the relational dyads in which abuse occurred (child-to-parent abuse, intimate partner abuse, sibling abuse, child maltreatment, and other family abuse) used binary dummy variables. For example, when attempting to determine whether child-to-parent abuse was more common among generalist or family-only youth, child-to-parent abuse was coded as one and all other relational dyads were coded as zero.
Two cox proportional hazards models were constructed with significance tests conducted on the log hazard ratios. The first model examined the relationship between being a generalist or family-only youth and time to FV recidivism (in days), whilst controlling for the age and sex of the young person who used FV (hypothesis three). The second model examined the relationship between diversity of prior offending among generalists and time to FV recidivism (in days), whilst controlling for the age and sex of the generalist youth (hypothesis four). Proportional hazards assumptions were examined by plotting Schoenfeld residuals and conducting significance tests examining the independence between residuals and time. Visual inspection of the residuals plot did not indicate any obvious relationship between residuals and time and tests were not significant at an alpha level of 0.05. Table 1 provides the descriptive statistics of generalist and family-only youth based on characteristics of the index FV incident. Generalist youth were more likely than familyonly youth to be male (χ 2 (1, N = 1972) = 70.42, p < 0.001, OR = 1.69, 95% CI [1.49-1.91]), aged 20-24 years (χ 2 (2, N = 1544) = 155.82, p < 0.001, OR = 2.04, 95% CI [1.82-2.28]), and to come from a regional/remote area (χ 2 (1, N = 960) = 28.77, p < 0.001, OR = 1.38, 95% CI [1.23-1.56]) compared to family-only youth. Generalist youth were significantly more likely to be identified as being in the lowest socioeconomic quintile compared to familyonly youth (χ 2 (1, N = 960) = 28.77, p < 0.001, OR = 1.51, 95% CI [1.34-1.70]).

Demographic Characteristics of Sample the Index Incident
Most generalist youth had engaged in both violent and non-violent offences (n = 1486, 56.96%), while over one third had engaged exclusively in non-violent offences (n = 1008, 38.64%), and a small proportion had exclusively engaged in violent offences (n = 115, 4.41%). Of those who had been charged with a violent offence, nearly two-thirds (n = 1596, 61.17%,) had been charged with more than one. Sexual offences were examined separately, with 8.59% (n = 224) of generalist youth ever having been charged with a non-FV sexual offence.

Lifetime Prevalence of Family Violence by Family-Only and Generalist Youth
The lifetime history of FV behaviour among family-only and generalist youth are provided in Table 2. The adjusted odds of a generalist youth being abusive across more than one relational dyad were 6.04 times greater than for familyonly youth. The adjusted odds of generalist youth using FV in a high severity FV incident or engaging in physical abuse during a FV incident were between two to three times greater than for family-only youth, representing small to medium effect sizes (Chen et al., 2010). Large effect sizes were observed regarding the likelihood of breaching a court order and contravening a restraining order, however results should be interpreted with the understanding that only a small proportion of family-only youth ever breached a court order (n = 30, 1.25%) or a restraining order (n = 99, 4.13%). The adjusted odds of breaching a court order were 47.68 times higher for generalist youth than family-only youth, whilst their odds of contravening a restraining order were 8.15 times greater. As an example, the model estimated a 91.60% probability of a 20-year-old male who contravened a restraining order being a generalist, compared to only a 57.22% probability of being a generalist if they had not contravened a restraining order.

Criminogenic and Non-Criminogenic Needs
The victimisation, mental health, substance abuse, and unemployment/truancy characteristics of the family-only and generalist groups are provided in Table 3. The adjusted odds of a generalist youth having historically been a victim of any FV incident was 3.55 times greater than for familyonly youth, whilst the adjusted odds of them being a victim of a high severity FV incident were 3.70 times greater. Their odds of being physically and sexually victimised were between two to three times greater than for family-only youth, whilst their odds of experiencing abuse across more than one relational dyad was 4.39 times greater. Generalists were significantly less likely than family-only youth to be aged 0-11 years when first reported to police as the victim of FV. While generalist youth showed significantly more mental health issues, substance abuse issues and school truancy/ unemployment, the magnitude of the effect sizes differed. The odds of generalists experiencing mental health issues were only 1.40 times greater than for family-only youth, whereas the odds of generalists experiencing substance abuse issues and unemployment issues were 3.68 and 2.53 times greater, respectively.

Family Violence Recidivism
Over one third (n = 1767, 35.24%) of all young people who used FV at the index incident engaged in FV recidivism during the 6-month follow-up period. As displayed in Table 4, 43.27% (n = 1129) of generalists engaged in FV recidivism, compared to 26.53% (n = 638) of family-only youth. The adjusted odds of generalists engaging in FV recidivism were 2.25 times greater than for family-only youth. As an example, the model estimated a 76.16% probability of a 20-year-old male who engaged in FV recidivism being a generalist, compared to a 58.65% probability of being a generalist if they had not engaged in FV recidivism. The adjusted odds of a generalist youth engaging in a high severity FV recidivism were 2.11 times greater than for family-only youth, and their odds of engaging in FV-based stalking during the follow-up period were 3.06 times greater. There were no significant differences in the likelihood of family-only and generalist youth engaging in sexual abuse as part of a FV incident during the follow-up period, however the adjusted odds of a generalist youth using physical abuse during a subsequent FV incident were 2.19 times greater. Figure 1 depicts Cox regressions examining: 1) the time to FV recidivism for generalist and family-only youth,  controlling for age and sex; and 2) time to FV recidivism according to diversity of prior offending among generalist youth, whilst controlling for the age (using a cubic spline) and sex of the young person. The model examining time to FV recidivism for generalist and family-only youth was significant (Likelihood ratio test = 168.41, df = 4, p < 0.001), indicating generalists engaged in a subsequent FV incident more quickly than family-only youth. The hazard ratio for generalist/family-only status (HR = 1.91, 95% CI = 1.72-2.11, p < 0.001) suggested the risk of FV recidivism was 91% greater for generalist youth. The model examining time to FV recidivism according to diversity of prior offending among generalist youth was significant (Likelihood ratio test = 218.40, df = 5, p < 0.001), indicating that a higher diversity of offending is associated with shorter time to FV recidivism. Similarly, the hazard ratio for offending diversity (HR = 1.48, 95% CI [1.32-1.57], p < 0.001) suggests that the risk of FV recidivism was 1.48 times greater for those who had previously engaged in just one type of (non-FV) offence compared to those who had not ever engaged in prior offending (i.e. family-only youth). Similarly, the difference in recidivism risk between having one type of (non-FV) offence and having engaged in three different offence types gives a hazard ratio of 2.11 [95% CI = 1.88-2.38], suggesting that the risk for FV recidivism more than doubles. However, it appears that there is no substantial increase in risk with each new offence type after engaging in three or more non-FV offence types. The difference in recidivism risk between having two types of offences to three types of offences indicates a 15% increase (HR = 1.15, 95% CI [1.11-1.19]), whilst the increase from three to four offences showed a 9% increased risk (HR = 1.09, 95% CI [1.04-1.15]), and the rise from four to five offences showed an 8% increased risk of recidivism (HR = 1.08, 95% CI [1.02-1.15]).

Discussion
The present study sought to characterise young FV users with a history of prior offending (generalists) and those who were known to police only for engaging in FV (family-only) to determine whether these groups were unique. As anticipated, generalist youth had a more significant history of FV incidents and were a higher-risk group in terms of the likelihood and severity of FV, as well as exhibiting a shorter time-to-recidivism. Generalist youth displayed a diversity of violent and non-violent offending external to the family context, which has previously been observed in both the adult (Coghlan & Millsteed, 2017) and youth (Moulds et al., 2019;Verbruggen et al., 2021) FV literatures. They were also significantly more likely to be in the lowest 20% of socioeconomic status, come from a regional/remote area, and display a greater level of need, as assessed by higher levels of mental health issues, substance abuse problems, victimisation histories, and unemployment/truancy from school.
The first hypothesis was supported, as generalists were significantly more likely than family-only youth to have engaged in a high severity FV incident, to display abusive behaviour across more than one relational dyad, and were more likely to breach court orders than family-only youth. These results are broadly consistent with the adult FV literature, which suggests that generalists display a greater severity of FV (Goldstein et al., 2016) and are more likely to breach court orders than family-only youth (Coghlan & Millsteed, 2017;Morgan et al., 2018). The adjusted odds of a generalist youth breaching a court order were found to be 47.68 times those of family-only youth, whilst their adjusted odds of contravening a restraining order were nearly ten times greater than that of family-only youth. The substantial difference observed between the groups for breaching court orders and restraining orders may be due to a more generalists ever having been placed on these orders, thus resulting in more opportunity for breaches to occur. However, generalists also displayed a greater severity and diversity of behaviour which, coupled with their propensity for breaching legal orders, supports the adult literature indicating generalists are a higher risk cohort (Cantos et al., 2015;Petersson & Strand, 2020;Verbruggen et al., 2021) with elevated levels of antisociality (Petersson & Strand, 2020).
Consistent with the second hypothesis, generalist youth displayed a higher level of police-recorded criminogenic and non-criminogenic need than family-only youth, including more issues with mental health, substance abuse, unemployment and school truancy, and more significant policereported victimisation histories. The adjusted odds of generalist youth experiencing substance abuse problems were nearly four times that of family-only youth, whilst their adjusted odds of experiencing issues with unemployment or school truancy were 2.53 times that of family-only youth. High rates of substance abuse (Phillips & McGuinness, 2020), and academic issues (Ibabe and Jaureguizar, 2010) have previously been reported for young people who use FV, as have elevated rates of mental health issues (Kennedy et al., 2010;Phillips &McGuinness, 2020), andvictimisation (Simmons et al., 2018).
Generalists were more likely to have been the victim of police-reported FV, including high severity FV, were more likely to be abused across more than one dyadic relationship, and were more likely to have experienced physical and sexual abuse than family-only youth. These results suggest a higher level of poly-victimisation and abuse severity among generalists, indicating a greater degree of dysfunction within the family context of generalist youth. These findings are broadly consistent with research suggesting that individuals who experience poly-victimisation are at greater risk of engaging in antisocial behaviour (Papalia et al., 2020) and offending of all types (Papalia et al., 2020) than their nonpolyvictimised peers. The findings also support previous research identifying different types of offending and violent behaviour likely stem from similar underlying risk factors (Farrington & Ttofi, 2021). It is therefore possible that factors identified in general theories of crime, such as self-control (Gottfredson & Hirschi, 1990), lack of social bonding (Hirschi, 1969), social learning (Bandura & Walters, 1977), and high levels of environmental strain (Agnew, 2006) may also be more common among generalists and provide some explanation for the higher prevalence of criminogenic and non-criminogenic need within this group.
In contrast, the lower level of risk and need among familyonly youth is indicative of "a more socially well-adjusted" (Petersson & Strand, 2020, 378) individual with more prosocial traits. It is possible that the engagement in FV by familyonly youth is related to their early FV victimisation. Familyonly youth scored significantly higher than generalists in only one domain in the present study, and this was related to whether they had ever been the victim of police-reported FV as a child (aged 0-11). This could indicate the relevance of social learning and/or the presence of dysfunctional family dynamics (Simmons et al., 2018) in the development of FV behaviour among family-only youth (with these factors also likely present among generalist youth), however additional research in this area is needed.
Consistent with the third hypothesis, generalists were more likely to be reported for a subsequent FV incident, and did so more quickly, than family-only youth, similar to research in the adult literature (Petersson & Strand, 2017). The proportion of recidivists in the current sample is higher than that observed in youth FV research with a similar follow-up period . This may be the result of the inclusion of youth aged 20-24 years, who are more likely to be generalists, in the present analyses, whereas previous research typically focused on the adolescent period Phillips & McGuinness, 2020). Shorter time to FV recidivism, as displayed by generalist youth, has been associated with more severe offending (Petersson et al., 2019), higher risk of recidivism Petersson & Strand, 2017), and higher frequency of FV incidents  among adult and youth FV users, suggesting generalist youth are likely to come into contact with the justice system more often than their family-only counterparts.
Consistent with the fourth hypothesis, diversity of prior offending was significantly associated with a shorter time to FV recidivism, which is broadly consistent with Gottfredson's and Hirschi's (1990) general theory of crime. It is possible that generalists who engage in more diverse patterns of offending display a higher level of antisociality and use a similar repertoire of skills (i.e. aggression, other antisocial behaviour) both outside and within the family home. This may result in young people accruing a greater range of different offence types whilst also a behavioural repertoire which increases the likelihood that they will respond to familial issues in an aggressive and/or antisocial manner (Boxall et al., 2015).
Yet, the lack of a dose-response relationship once an individual comes to police attention for three or more different offence categories may be indicative of a ceiling effect in level of criminogenic need and/or antisocial tendencies. There may be more variability in the characteristics (e.g. age, SES) and needs (e.g. substance abuse) among those who engage in one or two different offence types, however it is possible that there is a threshold at which level of need and antisociality no longer cumulatively add to an individual's risk of future FV.

Practical Implications
These findings provide a unique contribution to the youth FV field, with this being the first study directly comparing the needs, diversity of offending, and victimisation history among generalist and family-only youth across a broad age range. The findings suggest that generalist youth represent a higher risk cohort with a greater level of need than family-only youth, with implications for risk assessment and intervention.
Classifying young people who use FV as generalist or family-only youth at the assessment stage may assist in the early identification of those most at risk of future FV. The higher prevalence and severity of FV recidivism, coupled with their more rapid rate of recidivism, among generalist youth indicates that the identification of this higher risk cohort at the time of the index incident may be an efficacious means of identifying those most likely to come to police attention in future. In particular, generalists with a greater diversity of prior offending are at significantly greater risk of using FV in future FV incidents. However, it must be noted that over one quarter (26.53%) of family-only youth engaged in FV recidivism within six months, so generality of violence should be only one of many factors used to inform an assessment of a young person's risk for future FV.
The high level of risk and need exhibited by generalist youth indicates that this cohort should also be prioritised for intervention more readily than family-only youth (Bonta & Andrews, 2016), with extra emphasis placed upon their broader criminogenic and non-criminogenic needs, including mental health, substance abuse, unemployment/school truancy, and victimisation experiences. Results of the present study indicate that adherence to court orders for generalists is less than that of family-only youth, indicating that it may be necessary to engage generalists in pre-treatment interventions addressing their individual needs (e.g. substance abuse, mental health issues, school truancy; Cantos & O'Leary, 2014). Intensive assertive outreach programs may be one way of reducing the high rates of attrition which have been observed among this cohort (Cantos et al., 2015). Specialised case management approaches with assertive outreach addressing individual needs and managing ongoing risk issues have been used internationally with highrisk youth with mental health and substance abuse issues (McGorry et al., 2022) and may be useful for higher risk generalist youth.
Given that greater diversity of offending (i.e. 3 + offence types) among generalists is associated with more rapid FV recidivism, and that shorter time to recidivism is associated with more frequent use of FV , there is a need for intervention to be delivered in a timely manner. Similarly, intervention should be prioritised for young people who use FV (Campbell et al., 2020;Purcell et al., 2014), particularly those aged 14 years or younger (regardless of whether they are generalist or family-only), who come to the attention of police or other services, as this early-onset cohort is at an elevated risk of chronic offending (Piquero et al., 2012). This suggests a need to extend evidence-based youth FV treatment programs to include those as young as 10 years old to ensure early intervention occurs.

Limitations and Directions for Future Research
The present study is limited in several respects. First, the use of official police records meant that only FV incidents and prior offending that were reported to police were captured, with the results very likely underestimating the prevalence of FV, general offending and recidivism. Family violence recidivism data was limited to a six-month follow-up period, preventing extrapolation of results beyond six months. However, reduced follow-up time is less of a problem in designs measuring recidivism when using police reports as re-reporting typically happens quickly. For example, Morgan et al. (2018) showed that more than half of all FV recidivists were reported to police within sixty days of their index incident, whilst  showed that the probability of repeat FV by young people declines sharply approximately one month following the index incident.
Given some family-only youth may have engaged in offending behaviour unknown to police, it would be beneficial for future research to examine whether the characteristics of generalist and family-only youth observed here are sustained when using self-reported offending. While these are important limitations, they do not significantly impact upon the import of these results for criminal justice system responses. Police and courts are only able to respond to known incidents and the research presented here is important for informing police responses to FV incidents for which they are called to.
Second, the results are based on police-reported incidents of FV, not necessarily offence or arrest data. While this allows for a broader scope of FV behaviour to be captured, it limits comparison with other studies which do use offence or arrest data and may capture incidents in which FV did not occur. However, the use of incident-based statistics is also a strength as it allows for consideration of a broader range of FV behaviours, including those not considered criminal offences in the state of Victoria (e.g. psychological abuse, coercive control).
Third, only data pertaining to FV recidivism was available, with information related to general recidivism (i.e. non-FV recidivism) either being unavailable or the variable was found to contain 80% or more missing data and so was excluded from the analysis. This reduces the comprehensiveness of the present study's findings, with future research needed to examine non-FV recidivism (i.e. violent and nonviolent offending outside the family context) among generalist and family-only youth. Providing a more complete picture of other offending outcomes would help support resource allocation among police and service providers.
An additional area for future research would be to determine whether young FV-users differ from the average violent young offender (i.e. those who are violent outside the family context). Whilst Sjodin and colleagues (2017) have conducted some comparisons, the study was limited by their focus on male dating violence and sampling from imprisoned young adults aged 18-25 years. Research employing a broader age range, includes both male and female FVusers, and which examines multiple relationships of abuse will assist in identifying the key risks and needs differentiating the groups which can then be used to develop tailored assessment and intervention strategies. Similarly, further research identifying whether generalists and family-only youth reoffended within the same relationship dyad may assist to further understand diversity of behaviour among these cohorts.

Conclusion
Generalist and family-only youth represent distinct subgroups of young FV-users with generalists displaying a significantly higher level of risk and need. Results provide support for generalist youth to be prioritised for assessment and intervention and indicate the need to reconceptualise how youth FV is addressed and managed, however also indicates the presence of a discernibly different group of youth who engage in FV despite a lower level of need. This research suggests potential avenues for further research such as the self-reported motivations for FV behaviour between the two cohorts and situational and contextual antecedents of abusive incidents associated with the use of violence by generalist versus family-only youth.

Declarations
Conflict of interest The authors have no conflicts of interest to disclose.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.