Sample

The sample comprised 280 children and adolescents presenting with sexually abusive behaviour to a community assessment and treatment service. The majority of referrals came through local authority social services. The sample was predominantly male (91%) and White (83%), with a mean age of 13.9 years (s.d.=3.0, range 5.5–21.1 years) at the time of the original assessment. The general characteristics of the sample are described in detail elsewhere (Home Office et al, 2006; Reference Vizard, Hickey and FrenchVizard et al, 2007).

Data collection

Psychosocial and behavioural data were gathered from the services' files which included reports from multiple informants across a range of domains. The file data were also used to score the Psychopathy Checklist–Youth Version (PCL–YV, Reference Forth, Kosson and HareForth et al, 2003), a 20-item rating scale for assessing psychopathy traits in 12–18-year-olds. Clinical items are usually scored following a combined interview and file review. However, for research purposes file review alone is acceptable as long as information is sourced from multiple informants and domains. Adequate internal consistency (Cronbach's alpha=0.94) has been reported for this measure in community samples (Reference Forth, Kosson and HareForth et al, 2003).

Data on convictions were obtained from the Offenders Index and covered the period up to December 2003. This is a database administered by the Home Office, containing details of convictions for standard list offences recorded in England and Wales. Prevalence rates for ‘lifetime’ convictions were calculated based on convictions that occurred between the individual's tenth birthday (the age of criminal responsibility in the UK) and December 2003. ‘Time at risk’ excluded any time spent in custody. Two mutually exclusive categories of offences (sex and violence) were examined, as well as the composite ‘any offences’.

Data analysis

There were three strands to the data analysis. Individuals were categorised as either early onset (n=93) or late onset (n=120), depending on whether their sexually abusive behaviour began before or after their 11th birthday. In 67 individuals the age at onset was not known and these were excluded from the analysis. Comparisons were made of psychosocial, sexual and non-sexual antisocial behaviour and conviction characteristics. It was hypothesised that the early-onset group would have higher rates of difficult temperament and maltreatment, show more indiscriminate sexually abusive behaviour, but during adolescence have similar antisocial behaviour and conviction profiles as the late onset group.

Individuals were categorised as presenting with (n=54) or without (n=149) emerging severe personality disorder traits. These traits were operationalised as scores above the sample mean for conduct disorder symptoms and on the PCL–YV. A total of 77 juveniles were excluded from the analysis either because they were too young (i.e. were under 12 years) or because of insufficient information. Comparisons were made of psychosocial, sexual and non-sexual antisocial behaviour and conviction characteristics. It was hypothesised that the group with emerging severe personality disorder would have higher rates of difficult temperament, engage in more predatory sexually abusive behaviour, abuse multiple types of victims and have higher conviction rates.

Individuals with emerging severe personality disorder traits on the early-onset trajectory (n=32) were compared with those on the late-onset trajectory (n=18) in relation to psychosocial and behavioural characteristics. Given the small sample sizes these comparisons were necessarily exploratory in nature.

Between-group comparisons were conducted using χ² or t-tests as appropriate.

Age at onset trajectories for sexually abusive behaviour

Psychosocial risk factors

As shown in Table 1, the early-onset group experienced significantly higher rates of psychosocial adversity than the late-onset group. As predicted, they had higher rates of difficult temperament and, apart from exposure to domestic violence, also had higher rates of maltreatment. In addition, the early-onset group were more likely to have been exposed to poorer parenting models, inappropriate sexualisation (childhood sexual abuse and/or inadequate family sexual boundaries) and were more likely to display behavioural problems. The only variable on which the late-onset group had a significantly higher rate than the early-onset group was substance misuse.

Table 1 Psychosocial characteristics of juveniles with sexually abusive behaviour according to age at onset and emerging severe personality disorder traits

Characteristic	Age at onset			ESPD traits			Age at onset × ESPD
	Early, % (n=93)	Late, % (n=120)	With, % (n=54)	Without, % (n=149)	Early, % (n=32)	Late, % (n=18)
Parental and family factors
Parental criminality	33	29	35	34	41	22
Parental childhood abuse	44	33	54	34 **	53	56
Parental mental health problems	50	33 *	69	32 **	72	72
Parental time in care	27	15 *	24	22	28	17
Inconsistent parenting	77	53 **	76	58 *	81	72
Lack of parental supervision	65	30 **	46	46	63	17 **
Inadequate family sexual boundaries	59	25 **	39	42	50	17 **
Marital separation/divorce	77	70	70	74	72	67
Attachment related factors
Early difficult temperament	38	22 *	46	21 **	50	50
Removal to local authority care	83	73	93	74 **	94	89
6+ changes in home placement	50	30 **	62	30 **	77	39 **
Insecure attachment	68	33 **	72	44 **	78	56
Child factors
Peri-natal problems	32	21	35	27	44	17
Hyperactive/impulsive behaviour	75	61 *	87	62 **	88	89
Disruptive behaviour primary school	61	36 *	63	40 **	59	67
Excluded from school	51	40	67	40 **	69	67
Any sexual cruelty to animals	15	3 **	20	6	31	0 **
Any physical cruelty to animals	30	12 **	35	11	44	22
Sexual and physical cruelty to animals	10	0 **	15	1 **	22	0 *
Any substance misuse	15	27 *	32	20	22	50 *
Learning disability (IQ≤70)	31	21	35	22	41	17
Trauma factors
Childhood sexual abuse	83	58 **	69	72	81	44 **
Physical abuse	77	55 **	72	66	75	61
Emotional abuse	84	63 **	83	71	81	83
Physical neglect	72	41 **	54	58	66	28 *
Exposure to domestic violence	51	44	52	51	50	56

ESPD, emerging severe personality disorder

* P<0.05

** P<0.01

Sexually abusive behaviour

To ensure developmental comparability for the early- and late-onset groups, only sexually abusive behaviour displayed during adolescence (11–17 years) was investigated here. As shown in Table 2, during adolescence the early-onset group was more likely to abuse multiple types of victims, (males, females, children and adults). The late-onset group had higher rates of only abusing female victims, only abusing much younger children (i.e. those at least 5 years younger) and using verbal coercion. The late-onset group also appears to have higher rates of penetrating their victims.

Table 2 Sexually abusive behaviour according to age at onset and emerging severe personality disorder traits

Characteristic	Age at onset 1			ESPD traits 2			Age at onset ×ESPD 3
	Early, % (n=93)	Late, % (n=120)	With, % (n=54)	Without, % (n=149)	Early, % (n=32)	Late, % (n=18)
Victim characteristics
Female	80	82	93	86	97	83
Male	71	48 **	67	56	81	44 **
Male and female	55	33 **	61	44 **	78	33 **
Child and adult	31	19 *	44	24 **	47	44
Female only	25	49 **	32	42	19	50 *
Male only	16	16	6	12	3	11
Strangers	8	15	20	9 *	13	39 *
Adult women	4	9	9	4	3	22 **
≥ 5 years younger	63	72	56	68	66	39
≥ 5 years younger only	12	29 **	6	22 **	7	6
Abuse characteristics
Any penetration (anal or vaginal)	46	63 *	61	58	56	72
Predatory 3	–	–	76	57 *	78	72
Excessive force 3	–	–	17	7 *	16	22
Verbal coercion	25	45 **	52	36 *	41	72 *
Physical coercion	5	7	13	7	3	28 **
Prior grooming	25	35	56	34 **	59	57
With co-abusers	7	14	13	13	13	17

ESPD, emerging severe personality disorder

1. Comparisons are only in relation to abuse committed during adolescence

2. Comparisons are in relation to abuse committed at any time during childhood and adolescence

3. Data on abuse and excessive force not available for adolescent period

* P<0.05

** P<0.01

Developmental patterns of non-sexual antisocial behaviour

The results of the developmental stage analysis in relation to non-sexual antisocial behaviours are presented in Table 3. During infancy, early and middle childhood (0–10 years) the early-onset group tended to have higher rates of antisocial behaviour than the late-onset group, but as predicted, by adolescence both groups had generally similar rates. However, reckless behaviour and insecure attachment were significantly higher in the early-onset group at every stage of development. It is notable that physical aggression in the early-onset group begins early and increases consistently across every developmental stage. By contrast the level of physical aggression in the late-onset group presents at relatively modest levels but doubles between middle childhood and adolescence.

Table 3 Developmental patterns of non-sexual antisocial behaviour according to age at onset of sexually abusive behaviour

	0–3 years			4–6 years			7–10 years			11–17 years
	Early, % (n=93)	Late, % (n=120)	Early, % (n=93)	Late, % (n=120)	Early, % (n=93)	Late, % (n=120)	Early, % (n=93)	Late, % (n=120)
Difficult temperament 1	26	13 *	–	–	–	–	–	–
Oppositional	8	3	19	8 *	36	13 **	47	41
Insecure attachment	13	3 **	36	4 **	52	17 **	61	28 *
Physically aggressive	24	11 *	41	18 **	70	32 **	72	64
Physical cruelty to animals 2	–	–	5	0 *	14	1 **	18	9
Fire-setting 2	–	–	8	2 *	15	8	26	18
Stealing 2	–	–	5	4	29	13 **	48	55
Excluded from school 2	–	–	11	2 **	24	7 **	34	37
Impulsivity 2	–	–	19	9 *	32	11 **	51	39
Reckless behaviour 2	–	–	1	0	7	2	16	5 **
Socially isolated 2	–	–	9	5	26	15 *	63	55

1. Only collected in the 0–3 years developmental periods

2. Only collected in the developmental periods covering 4–17 years

* P<0.05

** P<0.01

Emerging severe personality disorder traits in juveniles with sexually abusive behaviour

Developmental patterns of non-sexual antisocial behaviour

The results of the developmental stage analysis are presented in Table 4. Within each developmental period the group with emerging severe personality disorder traits was significantly more likely to display antisocial behaviour. Even in infancy a considerable proportion displayed difficult temperaments (33%) and physical aggression (30%). By middle childhood (7–10 years) over three-quarters were displaying physical aggression, whereas in adolescence over a quarter were physically cruel to animals.

Table 4 Developmental patterns of non-sexual antisocial behaviour according to the presence of emerging severe personality disorder traits

	0–3 years			4–6 years			7–10 years			11–17 years
	With ESPD, % (n=54)	Without ESPD, % (n=149)	With ESPD, % (n=54)	Without ESPD, % (n=149)	With ESPD, % (n=54)	Without ESPD, % (n=149)	With ESPD, % (n=54)	Without ESPD, % (n=149)
Difficult temperament 1	33	11 **	–	–	–	–	–	–
Oppositional	11	2 **	26	9 **	48	13 **	69	36 **
Insecure attachment	15	4 **	28	17	56	26 **	67	38 **
Physically aggressive	30	11 **	46	19 **	76	38 **	89	64 **
Physical cruelty to animals 2	–	–	7	0 **	13	2 **	28	9 **
Fire-setting 2	–	–	7	3	24	5 **	46	14 **
Stealing 2	–	–	13	2 **	32	14 **	80	47 **
Excluded from school 2	–	–	13	3 **	32	9 **	56	32 **
Impulsivity 2	–	–	26	9 **	43	13 **	74	34 **
Reckless behaviour 2	–	–	0	1	13	0 **	28	4 **
Fighting and stealing 2	–	–	9	1 **	24	7 **	69	34 **
Socially isolated 2	–	–	7	8	24	20	74	54 **

EPSD, emerging severe personality disorder

1. Only collected in the 0–3 years developmental period

2. Only collected in the developmental periods covering 4–17 years

** P<0.01

Age at onset trajectories and emerging severe personality disorder

Age at onset of sexually abusive behaviour

Moffitt (Reference Moffitt1993) proposed that those with an early onset of antisocial behaviour are impaired by the interaction of neuropsychological deficits and adverse environments. In support the current study found that those with an early onset of sexually abusive behaviour showed higher levels of early difficult temperament and adverse environmental experiences such as inadequate parenting, maltreatment, placement changes and insecure attachment. These factors also increase the risk of persistent antisocial behaviour throughout childhood and adolescence. Interestingly, the sexually harmful behaviour perpetrated by those on the early-onset trajectory tends to be generalised rather than targeted at specific victim groups. This suggests that their behaviour may not be primarily sexually motivated at this younger age but may be one feature of an externalising presentation.

By contrast, those with a late-onset of sexually abusive behaviour had different psychosocial and behaviour profiles consistent with Moffitt's (Reference Moffitt1993) hypothesis that late onset antisocial behaviour is less directly influenced by early developmental factors. The higher rates of substance misuse in this group perhaps reflect the greater influence exterted by the peer group. The sexually abusive behaviour of the late-onset group (for example, victimising females or younger children) is consistent with a greater influence of sexual arousal and an inability to achieve developmentally appropriate sexual relationships.

Emerging severe personality disorder traits

The current study used both behavioural and personality criteria to define emerging severe personality disorder. It was predicted that juveniles with such traits would show a characteristic set of developmental risk factors, as well as serious antisocial behaviours. The results supported these predictions. Those with such traits, although no more likely to experience maltreatment, were more likely to have had parents with mental health problems who had also been abused. These factors may have served to compromise their ability to provide good parenting and a secure attachment base.

Higher levels of poorer attachment, and impulsive and disruptive behaviour may contribute to feelings of social isolation from peers in adolescence, partly motivating subsequent aggressive and sexually abusive behaviour. It is striking that these young people present with difficulties, even before school, that persist across development. The sexually abusive behaviour of this group tended to be more predatory, more likely to entail excessive force and verbal coercion, and be targeted at a range of victims.

The significantly higher levels of convictions, particularly violent convictions, suggests that the construct of emerging severe personality disorder may prove useful in helping to identify those young people most at risk of later serious offending.

Age at onset and emerging severe personality disorder

The exploratory group comparisons highlight that young people presenting with such traits are not an homogenous population. Those with traits on the early-onset trajectory tended to target a range of victims whereas those with such traits on the late-onset trajectory perpetrated higher rates of more worrying sexually abusive behaviour (e.g. higher levels of rape, abuse of strangers and physical coercion). On this basis, it could be predicted that juveniles at highest risk for continuing to sexually offend in adulthood could come from this group.

Prevention

The established cost benefits of preventing childhood conduct disorder (Reference Scott, Knapp and HendersonScott et al, 2001) and investing in multi-systemic treatment for juvenile sexual offenders (Reference Borduin and SchaefferBorduin & Schaeffer, 2001) indicate the economic viability of primary prevention. Therefore, there is a clear case for investment in prevention and treatment resources that enable local services to identify and intervene early with vulnerable children.

Early identification

Early assessment and intervention is highly recommended if there is evidence of an early onset of sexually abusive behaviour or emerging severe personality disorder traits. The findings presented here indicate that an assessment of emerging severe personality disorder traits may need to become part of the assessment portfolio of child and adolescent mental health services (CAMHS) which will require appropriate measures to be developed and resources to provide preventative input for children identified as high risk (Royal College of Psychiatrists, 1999: pp. 34–46). Furthermore, the early identification of children under 10 years of age displaying sexually abusive behaviour may help prevent a trajectory of development leading to contact with the criminal justice system, with significant benefits for the child and society. However, any risk assessment should be coupled with a needs assessment (Reference Kroll, Woodham and RothwellKroll et al, 1999) which includes planning for the input of appropriate resources. Concerns about the labelling of children at an early age should be set firmly against the lifelong preventative benefits to the child and society of the early identification of need.

Service provision

Juvenile sexually abusive behaviour and emerging severe personality disorder traits in children are complex, multi-dimensional problems requiring coordinated responses from a range of community-based agencies including local authorities and CAMHS. At present, few local services are willing to accept such cases, and many are not convinced that the needs of these children fall within their remit.

There is a pressing need for more active involvement of CAMHS in the assessment and treatment of these children (Royal College of Psychiatrists, 1999: pp. 34–46). It has been noted that particular attention should be given to the delivery of services to children whose complex needs span different specialties, such as juvenile sexual offenders and people with learning disabilities, since care planning may require close cooperation between a wide range of services (Royal College of Psychiatrists, 1999: pp. 34–36). These children require a range of services, including non-residential services for the majority and specialist residential services for the few with a more disturbed presentation. Strategic thinking is needed to clarify how these will be achieved.

Research strategies

Further research is required to improve early identification of sexually abusive behaviour and emerging severe personality disorder traits, to develop appropriate interventions, and to determine long-term outcome. The sample in this study may not be representative of all children with sexually abusive behaviour and there was a limited follow-up for antisocial and sexual conviction. From a practical perspective research is required to develop and evaluate developmentally sensitive measurement tools for assessing such traits in young people.

Retrospective studies of adults with severe personality disorders, and not just antisocial personality disorder, would prove helpful in identifying predisposing childhood and adolescent developmental characteristics. Such retrospective studies would build on an existing evidence base that has already explored developmental models of antisocial personality and has identified childhood-onset conduct disorder as a factor that increases the relative risk of developing adult antisocial personality disorder (Reference Loeber, Green, Lahey, Farrington and CoidLoeber et al, 2003). However, the limitations of both longitudinal (Reference RobinsRobins, 1966) and adult retrospective studies (Reference Zoccolillo, Pickles and QuintonZoccolillo et al, 1992) have been discussed and the dearth of prospective, longitudinal studies of relevant antecedents to antisocial personality disorder has been noted (Reference Loeber, Green, Lahey, Farrington and CoidLoeber et al, 2003). The results of the present study strongly support the case for prospective, longitudinal research. Such prospective studies with high-risk children and adolescents would improve understanding of factors that allow some children to move off a severe personality disorder trajectory.

Policy development

Given the complexity and diversity of need within populations of young people with sexually abusive behaviour and emerging severe personality disorder, such as those reported here, it is essential that government policy addresses service provision within community services and residential care services. A dedicated interdepartmental government committee may be required to coordinate and monitor progress, and to facilitate inter-agency liaison.