Overview of available evidence

Systematic literature searches identified a large body of literature that addresses behavioural and emotional characteristics in association with child maltreatment. Secondary screening identified systematic reviews for some of these characteristics. Where systematic reviews were not identified for particular aspects, individual studies were reported.

Narrative summary

Two systematic reviews synthesised data on the psychological effects in children of witnessing domestic violence.^84,85

In the first review (search date end 2000), the authors addressed six general categories of psychosocial adjustment (internalising (including somatic complaints), externalising, other psychological problems, total psychological problems and academic problems) and six types of specific responses to hypothetical episodes of interpersonal conflict (negative affect/distress, negative cognitions, withdrawal, intervention, aggression and positive coping).⁸⁴ The results of the meta-analyses are summarised in Box 7.1. The methodology of the review was found to be good but there was variation in the quality of studies used in the synthesis, particularly in the way non-witnesses of domestic violence were ascertained. [EL = 2+]

Box 7.1

Demeanours and behaviours of children who have experienced child sexual abuse (CSA) or who have witnessed domestic violence, as reported in the literature.

The second systematic review on domestic violence extracted data on 41 studies and found that 40 studies showed that children exposed to domestic violence had worse outcomes on internalising, externalising and post-traumatic stress disorder, although the pooled effect size was stated to be small.⁸⁵ The authors found that outcomes were similar in boys and girls and drew no conclusions about the effect of age on outcome. [EL = 2 −]

A narrative review of sexual abuse of boys (search dates 1985–1997) reported on the consequences of sexual abuse.⁸⁶ This review included some studies in adult males and a number of studies in specific populations such as chemical abusers. In studies that compared abused with non-abused males, rates of the following were statistically significantly higher in abused than non-abused males: major depression (four times), bulimia (three times), antisocial personality disorder, behaviour problems, low self-image, runaway behaviour and legal problems.

A review synthesised research on the impact of sexual abuse on children.⁸⁷ The authors extracted data from studies that compared CSA cases with non-clinical controls on the following demeanours: anxiety, fear, depressed, withdrawn, poor self-esteem and the composite symptoms of internalising and externalising behaviours. [EL = 1+] A summary of results is shown in Box 7.1.

Evidence statement

The systematic reviews indicate that abused children, regardless of the manner of abuse, display more emotional and behavioural problems than children who have not been maltreated. The heterogeneity of definitions, ascertainment and reporting in the studies should be taken into account when drawing conclusions.

See Section 7.1.11 for the GDG considerations and the recommendations.

Narrative summary

A narrative systematic review examined the link between child maltreatment and youth violence between the ages of 12 and 21 years.⁹⁰ No formal synthesis of results was conducted. The authors concluded that physical abuse is a predictor of youth violence but also that other forms of abuse of varying severity can lead to youth violence. [EL = 2+]

A study of young people showing fire-setting behaviour (n = 205, mean age 11.2 years, SD 3.1 years) investigated differences in fire-setting behaviour between maltreated and non-maltreated children.⁹¹ The children and their caregivers were recruited from an assessment and treatment centre for juvenile fire-setters. Maltreatment status was ascertained by asking the caregiver whether the child had ever been abused or neglected; suspected abuse cases were excluded. Forty-eight percent were found to have been maltreated. Fire-setting behaviour was recorded using a semi-structured interview. Maltreated children were found to have statistically significantly more frequent fire-setting episodes and to use a wider range of media. The differences between groups were small in both cases. [EL = 3]

See Section 7.1.11 for the GDG considerations and the recommendations.

Narrative summary

A case–control study aimed to determine how often sexually abused boys present with somatic and behavioural symptoms.⁹² One hundred and seven (sexually abused) schoolboys (cases) were compared with 107 schoolboys not sexually abused (controls). The results showed that somatic and behavioural symptoms were uncommon in both cases and controls: 83.6% of cases and 76.7% controls did not have symptoms. No statistically significant differences were found between the numbers of cases and controls who had presented with somatic and behavioural complaints (18 cases versus 25 controls). There were statistically significant differences between cases and controls with symptoms lasting over 1 year (P < 0.05). [EL = 2−]

A cohort study sought to explore the relationship between child abuse or neglect and school performance, mainly academic success, peer status and adaptive functioning.⁹³ The study found that the mean academic performance (100–500) at age 6 years was 260 (SD 85) and at age 8 years was 263 (SD 95). The mean peer status (1–5) at age 6 years was 3.5 (SD 0.85) and at age 8 years was 3.3 (SD 0.96). The total adaptive functioning (4–28) at age 6 years was 14.6 (SD 5.16) and at age 8 years was 14.6 (SD 5.28). Maltreatment was statistically significantly associated with poorer academic performance (P < 0.01) and poorer adaptive functioning (P < 0.001) but not with peer status. [EL = 2−]

Another study aimed to determine the relationship between child maltreatment and timing of learning difficulties.⁹⁴ Three-hundred maltreated children were compared with 300 non-maltreated children. The study found maltreated children were at higher risk of repeating kindergarten and first grade than non-maltreated children. There was no difference in the risk of repeating a grade for the first time. The absolute risk of receiving a poor English or mathematics grade changed across elementary years whereas the relative risk by maltreatment status did not. [EL = 3]

A comparative study was conducted in a community sample of 420 maltreated children to determine the relationship between child abuse and neglect, and academic performance, discipline referrals and suspensions.⁹⁵ The study found that maltreated children performed statistically significantly below non-maltreated children in standardised tests and grades, and were more likely to repeat a grade. Maltreated children also had statistically significantly more discipline referrals and suspensions. [EL = 3]

A descriptive study aimed to identify the predictors of attributions of self-blame and internalising behaviour problems in sexually abused children by using the Sexual Assault Profile, Child Behaviour Checklist (CBCL) and Social Adjustment Scale.⁹⁶ The study found that a child having a close relationship with the perpetrator, severe sexual abuse, perceiving sexual abuse as disgusting and coping with abuse by pretending it never happened led to increased attributions of self-blame. These factors did not predict internalising behaviour problems. [EL = 3]

Another descriptive study investigated the differences in achievement and related classroom behaviours among maltreated and non-maltreated children (receiving public assistance and lower middle class).⁹⁷ The Hahnemann Elementary School behaviour rating scale was used. The study found that maltreated children exhibited less classroom behaviour positively linked with academic achievement compared with non-maltreated children receiving public assistance and with non-maltreated children of lower middle class. [EL = 3]

A study tested the hypothesis that physically abused children are characterised by increased usage of immature defence mechanisms as compared with non-abused/non-neglected children.⁹⁸ The investigators used the Child Suicidal Potential Scales (CSPS), a clinician-administered interview schedule consisting of nine sections. The comparison group consisted of children neglected by their parents and children who were neither abused nor neglected The results showed statistically significant differences between the physically abused and the non-abused/non-neglected children for all ego defences except displacement. Statistically significant differences were found between physically abused and neglected children for regression, denial and splitting, projection, and introjection (high scores for the physically abused children) and for compensation and undoing (higher scores for the neglected) children). [EL = 3]

A study compared parent symptom reports from three prepubescent groups: non-abuse group (NA), sexual abuse with perpetrator confession (SA) and sexual abuse without perpetrator confession (AA).⁹⁹ The Structured Interview for Signs Associated with Sexual Abuse (SASA) was used. The results showed that both SA and AA groups reported increased sleep problems, fearfulness, emotional and behavioural changes, concentration problems, and sexual curiosity and knowledge than the NA group. [EL = 3]

See Section 7.1.11 for the GDG considerations and the recommendations.

Narrative summary

A small case–control study (n = 18 in each group) identified children who were selectively mute at school for at least 1 year and compared their maltreatment status with controls matched on age and sex from the same school class.¹⁰¹ The two control groups were children with speech or language problems and children with no speech or language problems. There were five definite abuse cases in the selectively mute children and three possible abuse cases; there was one possible abuse case in the group with speech or language problems and no abuse, either definite or suspected, in the normal controls. [EL = 2−]

See Section 7.1.11 for the GDG considerations and the recommendations.

Overview of available evidence

Two systematic reviews were found that reported on the association between insecure attachment and child maltreatment.^102,103 There was some overlap in the samples that were included in the accompanying meta-analyses.

Narrative summary

The more recent systematic review (search dates 1988–2005) identified eight studies (involving a total of 791 children) that investigated an association between child maltreatment and attachment difficulties.¹⁰² The inclusion criteria were that the maltreated children were younger than 48 months, the study included comparison groups, the Strange Situation procedure (a procedure that takes place under controlled conditions that is designed to assess infant attachment style) or an adaptation of it was used and data were reported in sufficient detail to warrant meta-analysis. By pooling data from the studies, the authors found that 80% of maltreated children had insecure attachment compared with 36% of the comparison group. Using meta-analytic techniques, the odds ratio for having insecure attachment and being maltreated compared with not being maltreated was 6.5 (95% CI 3.7 to 11.6). [EL = 2+]

The second review identified five studies that investigated the relationship between maltreatment and disorganised attachment.¹⁰³ These studies included a total of 323 children aged between 11 and 48 months. Using the study size to weight the effect from each individual study revealed a pooled correlation coefficient of 0.41 for disorganised attachment in maltreated children compared with non-maltreated children. The review reported that 48% of maltreated children had insecure attachment compared with 17% of the comparison groups. [EL = 2+]

See Section 7.1.11 for the GDG considerations and the recommendations.

Overview of available evidence

No suitable published literature was identified in relation to the question of whether repeated nightmares in the absence of an obvious cause are a reason to suspect child maltreatment. However, presence or absence of nightmares is an item on the Child Behaviour Checklist, so there are some studies that mention nightmares in relation to maltreatment but which were not designed to answer the question.

See Section 7.1.11 for the GDG considerations and the recommendations.

Overview of available evidence

Out of 21 retrieved papers, eight were found to be suitable for inclusion and addressed the question as to whether dissociation is a reason to suspect maltreatment.^104–111

The eight included papers comprised one prospective longitudinal study [EL = 2+], six case–control series [EL = 2−] and one questionnaire validation study [EL = 2−]. Six of the studies were from the USA and one each from Canada and Sweden. All but two of the studies recruited the participants from specialised setting such as social services and child maltreatment clinics, and many of participants in these studies were from low socio-economic groups. The Child Dissociative Checklist (CDC) and the Adolescent Dissociative Experiences (ADE) scales were the most frequently used although, in the majority of the studies, the primary outcome was not to determine an association between child maltreatment and the clinical feature of dissociation. The most frequent types of maltreatment investigated by these studies were sexual and physical maltreatment, usually both separately and together. Neglect was investigated in two studies.

Narrative summary

In a prospective longitudinal study, 585 children were randomly recruited from two cohorts starting at kindergarten in 1987 and 1988 in three public schools in the USA.¹⁰⁴ On recruitment, the developmental history of the child was taken by an interviewer (no details) in the family home and included details on child misbehaviour and discipline practices. At this point, the interviewer rated whether physical maltreatment had occurred or not. The follow-up for presence of dissociation symptoms was assessed in the 11th grade at school by the mothers completing the CBCL and the child completing the Youth Self-report Form of the CBCL. Both unadjusted analysis and analysis adjusted for covariates showed a statistically significant association with suspected child physical maltreatment and dissociation later in school life. The covariate-adjusted analysis of parental CBCL report was not maltreated 1.58 (SD 0.16) versus maltreated 2.8 (SD 0.37) (F = 10.01; P < 0.01). [EL = 2+]

In a case–control series, 198 pre-school children (mean age 5.5 years, SD 0.5 years) were recruited from families who had been referred to social services in the USA.¹⁰⁵ The children were classified as physical, sexual, neglected and no maltreatment (no numbers given per group) by social services records. The main outcome measure was the CDC and it was shown that there was a statistically significant overall effect for maltreatment sub-types on dissociation (P < 0.00001). All clinical groups (mean CDC values (no SD given) were physical abuse 8.91, sexual abuse 7.27, neglected group 5.52) demonstrated greater dissociation than the non-maltreated group (P < 0.001 for all). Further sub-analysis showed that between the three maltreatment groups, physical abuse and neglect were statistically significantly related to dissociation (P < 0.001) but sexual abuse was not (P > 0.1). [EL = 2−]

In a case–control series in the USA, 114 children and adolescents (age range 10–18 years) were recruited from social services and classified as no maltreatment (n = 27) sexual maltreatment (n = 25), physical maltreatment (n = 18) or sexual and physical maltreatment (n = 44), and were assessed using the ADE scale or the CDC scale according to age.¹⁰⁶ Results from the ADE scale showed that children with sexual abuse reported statistically significantly higher levels of dissociation (mean scores): no abuse 2.4 (SD 4.7), sexual abuse 3.4 (SD 2.6), physical abuse 2.4 (SD 1.8), sexual and physical abuse 3.7 (SD 2.1); P < 0.01. Results from the CDC scale showed that children with a history of sexual and physical abuse had higher levels of ‘perceived’ dissociation (mean scores): no abuse 4.7 (SD 2.0), sexual abuse 6.0 (SD 4.8), physical abuse 6.2 (SD 6.1), sexual and physical abuse 10.4 (SD 6.9); P < 0.05. [EL = 2−]

In a case–control series carried out in the USA, 189 children (age range 3–17 years) were recruited in a hospital-based child abuse evaluation unit.¹⁰⁷ The children took part in a 5 day physical and psychological assessment which included the Children's Perceptual Alteration Scale (CPAS), ADE and CDC. The results were presented in two ways: by age groups (3–5 years, 6–10 years and 11–17 years) and by abuse status (abused, neglected and control) but no statistical analysis was reported. The authors concluded that there was no statistically significant association between prior histories of abuse in any of the groups with any of the dissociation measures. [EL = 2−]

In a case–control series, 134 French-speaking girls were recruited either from referrals to a child protection clinic (n = 67, mean age 9.0 years, SD 1.4 years) or from one of three public schools (n = 67, mean age 9.2 years, SD 1.7 years) in Canada and assessed with the CDC in French.¹⁰⁸ The demographics of the two groups were broadly similar but differed in terms of family structure and parental level of education. In the sexually abused (SA) group, 65.6% were classified as very serious cases and 46.9% of the girls had experienced chronic abuse over months or years. The results were expressed in seven SA subgroups: no penetration, penetration, no intrafamilial, intrafamilial, no chronic abuse and chronic abuse. In the SA group 20 girls (29.9%) and in the control group three girls (4.5%) presented with clinical levels of dissociation (P < 0.01). After correcting for covariables, the odds of presenting with dissociative tendencies were presented as eight-fold in the SA group compared with the control group. The degree or type of sexual abuse did not prove to be predictive of dissociation symptoms. [EL = 2−]

In a case–control series of 57 adolescents (age range 11 years and 3 months to 17 years and 8 months) were recruited following admission into an acute adolescent inpatient unit in the USA and assessed using the ADE scale.¹⁰⁹ These children were of low socio-economic class and were categorised as sexually abused, physically abused or both sexually and physically abused. Their data were compared with a historical ‘control’ group of adolescents aged 13–17 years with a variety of diagnosis and abuse backgrounds. The mean ADE score of the total study group was 32 (no SD given) and this was compared with the mean ADE of the ‘control’ group 19.2 (SD 15.0) (P < 0.005). Individual ADE scores for the study subgroups showed sexually abused adolescents to have a greater score (34.7; SD 31.7) than physically abused adolescents (28.1; SD 25.1) but this was not statistically significant. [EL = 2−]

In a case–control series, 350 children (age range 7–18 years) were recruited from four different settings to form four study groups: non-psychiatric comparative (local schools) (n = 75, mean age 11.96 years, SD 2.25 years), psychiatric non-abused (from consecutive inpatient admission to a psychiatric unit) (n = 165, mean age 12.56 years, SD 2.74 years), psychiatric abused (consecutive children and adolescents seen in inpatient and outpatient settings with social services or police record of sexual abuse) (n = 72, mean age 12.05 years, SD 2.84 years), psychiatric suspected abuse (from inpatient and outpatient settings with unsubstantiated reports of sexual abuse) (n = 38, mean age 12.05 years, SD 2.84 years).¹¹⁰ The main outcome measures of interest were the dissociation subscale of the Trauma Symptom Checklist for Children (TSC-C) and the parent-reported CDC. The results showed ‘significant differences’ between the three clinical groups and the non-psychiatric control group but no differences between the three clinical groups in terms of the dissociation subscale of the TSC-C. There was no reporting of details of these statistical tests although means and standard deviations of the groups were given. The CDC results were also brief and the authors describe post hoc analysis of the data producing similar results to the dissociation subscale on the TSC-C. [EL = 2−]

In a retrospective questionnaire validation study, 523 adolescents were recruited to validate the Dissociation Questionnaire in Swedish (DIS-Q).¹¹¹ A clinical group of 74 adolescents (mean age 16.03 years) with a history of sexual and or physical maltreatment were recruited from a child and adolescent psychiatric clinic. A control group of 449 adolescents (mean age 15.07 years) was recruited from within schools in the same city. The main aim of the study was to validate the DIS-Q in Swedish but, in addition, the results showed that the prevalence of dissociation was 2.3% in the control group (mean score 1.42, SD 0.43) and 50% (2.52; SD 0.8) in the clinical group (P < 0.001). [EL = 2−]

Evidence statement

The type of evidence available to answer this question was low in terms of quality, i.e. mostly case–control studies, but it is important to note that this question could not be answered by an intervention study and therefore the design of the studies is appropriate and the grading less important. The choice of control group was not always appropriate and covariates not always controlled for. Numbers of participants were low. Overall, the evidence suggests there is a positive association of the presence of dissociation symptoms with previous and or current maltreatment. There was insufficient or no evidence to comment on the role of age or gender, or degree, type or chronicity of maltreatment in the development of dissociation symptoms.

Delphi consensus (see also Appendix C)

The GDG sought the opinions of the Delphi panel for its statement on dissociation. The following statement was used in the survey:

Overview of available evidence

A total of 4326 articles were identified and 32 articles were selected for detailed assessment. No relevant systematic reviews were identified. All the studies used an observational design, i.e. case–control, cohort or case series. A detailed description of each study is provided below.

Narrative summary

A prospective cohort study (n = 842) undertaken in the USA examined the relationship between behavioural and emotional problems and physical, sexual and emotional abuse (based on questionnaire responses) in a population of incarcerated adolescents (mean age 15.8 years, 84.2% male, 40% white).¹¹² The study found that, after adjusting for demographic variables (age, gender and ethnicity), emotional abuse was a predictor (P < 0.05) of internalising behaviour (including self-harm), and that physical and sexual abuse were predictors (P < 0.01) of externalising behaviour (including self-destructive behaviour). However, other variables such as age (P < 0.01), gender (P < 0.01)) and ethnicity (P < 0.001) were also significant factors in internalising and externalising behaviour. The study concluded that different forms of maltreatment have different behavioural impacts. [EL = 3]

A case–control study (n = 86) from the USA examined the relationship between maltreatment (based on questionnaire responses) and non-suicidal self-injury (NSSI) in a community sample of adolescents (aged 12–19 years, mean age 17.4 years, 78% female, 73% white).^113,114 Two groups were selected: Group 1 (n = 64) with a history of NSSI and Group 2 (n = 30) without a history of NSSI (94 total, only 86 completed all questionnaires). The results of univariate analysis showed that physical neglect (P < 0.05), emotional abuse (P < 0.01) and sexual abuse (P < 0.05) were all predictors of NSSI but emotional neglect and physical abuse were not. However, the study also found that a self-critical cognitive style was a mediating factor between emotional abuse and NSSI. The study concluded that not all types of maltreatment are associated with self-harm. [EL = 2−]

A case–control study (n = 2485) from Australia examined the relationship between sexual abuse (based on questionnaire responses) and suicidal behaviour in a community sample of schoolchildren (mean age 14 years, 55.5% males).¹¹⁵ The study found that 87 (3.6%) children had been sexually abused. Furthermore, the study found that 659 (27.1%) had suicidal ideation, 328 (13.7%) had plans for suicide, 253 (10.5%) threatened to commit suicide, 442 (18.4%) self-harmed, 139 (5.8%) had attempted suicide and 25 had required emergency treatment as a result of a suicide attempt. The study compared those who had been sexually abused with those who had not. The study found that 73% of abused compared with 25% of non-abused had had suicidal thoughts (P < 0.001), 30% versus 5% had injured themselves five or more times (P < 0.001), and 36% compared with 8% had been hospitalised as a result of a suicide attempt (P < 0.001). Using multivariate analysis the authors examined the mediating factor of distress (none, low, high) adjusting for depression, hopelessness and family functioning on suicidal behaviour in boys and girls. The study examined differences due to severity of abuse based on three categories: abused, low-level abuse and high-level abuse. The study found adjusted ORs of 5 (95% CI 1.5 to 16.8), not statistically significant, and 7.4 (95% CI 1.7 to 31.8) for suicidal ideas, respectively. For self-harm the adjusted ORs were 4.3 (95% CI 1.5 to 12.6), not statistically significant, and 4.8 (95% CI 1.4 to 16.6). For attempted suicide the adjusted ORs were 15.0 (95% CI 4.7 to 47.9), not statistically significant, and 18.7 (95% CI 5.0 to 70.1). For having planned suicide the adjusted ORs were 10.6 (95% CI 3.5 to 32.7), not statistically significant, and 13.3 (95% CI 3.6 to 49.6). For suicide threats the adjusted ORs were 10.9 (95% CI 3.9 to 30.4), 10.4 (95% CI 1.4 to 77.3), and 11.1 (95% CI 3.4 to 35.7). The study found that for girls the idea of suicide was statistically significantly higher among those who reported a high level of abuse compared with those who had not been abused (OR 3.3; 95% CI 1.1 to 10.2), but for self-harm and attempted suicide there was no difference between abused and non-abused. The study concluded that sexual abuse leads to increased risk of self-harm and suicide, especially in boys. [EL = 2−]

A cross-sectional survey (n = 489) undertaken in Hong Kong examined the psychological impact (self-harm and substance abuse) of physical maltreatment (diagnosed by responses to questionnaire) in adolescents from a school survey (aged 13 years or over).¹¹⁶ The study found that 4.5% had received corporal punishment from family members within the past 6 months, 10.9% had been beaten by a family member for no reason within the past 6 months and 10.4% reported being beaten to injury by a family member at some point. The study found an association between self-injury and ‘beaten to injury’, with an OR of 4.42 for ‘would hurt themselves when faced with difficulties’, an OR of 5.03 for ‘think of hurting themselves’ and an OR of 8.47 for ‘who have tried hurting themselves’ (all P < 0.01). Physical maltreatment was not associated with ‘tried hurting self’ (P = 0.054). The study concluded that physical maltreatment had an impact on psychological wellbeing. [EL = 3]

A case–control study (n = 405) undertaken in the USA examined factors associated with suicide attempts in children (aged 7–17 years, mean age 12.7 years, 54% male, 83% white) being treated for bipolar disorder.¹¹⁷ The study found that 128 of 405 had attempted suicide and that 41 (32%) of these children had been physically or sexually abused (based on responses to questionnaire) compared with 54 (20%) of the non-attempter group (P = 0.006). The study also found that psychiatric hospitalisation, self-injurious behaviour, mixed episodes, psychosis and age were statistically significant factors on suicide attempts. In addition, family factors such as depression, familial substance use and suicide attempts, and comorbid conditions, such as panic disorders and substance use were also predictors of suicide attempts. The regression model produced by the authors to explain maximum variance did not include either sexual or physical maltreatment. The study concluded that multiple clinical factors had to be taken into account when assessing suicide risk. [EL = 3]

A case–control study (n = 105) undertaken in the USA examined the relationship between physical and sexual abuse (based on any report to authorities) and psychological problems and suicide attempts in children (aged 12–18 years; 73 female) admitted to an inpatient psychiatric facility.¹¹⁸ There were four groups: no abuse (n = 35), sexual abuse (n = 17), physical abuse (n = 22), and sexual and physical abuse (n = 31). The study found no statistical difference between groups in terms of suicidal ideation (thoughts 60.0%, 82.4%, 59.1% and 74.2%, respectively; suicidal behaviour 37.1%, 29.4%, 40.9% and 29.0%, respectively; threats of suicide 32.3%, 31.3%, 26.3% and 43.3%, respectively; suicide attempts 48.6%, 47.1%, 45.5% and 61.3%, respectively). The study concluded that the symptoms of adolescents who are psychiatrically hospitalised do not differ with abuse history. [EL = 3]

A prospective cohort study (n = 140) undertaken in the USA examined the relationship between childhood sexual abuse in females (abuse reported by child to have happened before the age of 14 years) and re-victimisation and self-harm in children who had been sexually abused (mean age 18.81 years).¹¹⁹ The study found that in the sexually abused group (n = 70) 32.3% had self-harmed compared with 8.8% in the comparison group (n = 70, P = 0.02). In addition, the study found no relationship between physical abuse, neglect or emotional abuse and self-harm. The results from multiple regression found an OR of 5.64 for those who had been sexually abused and self-harmed (P < 0.01), but an OR of 2.26 for physical, 0.74 for neglect and 0.57 for emotional (all not statistically significant). The study concluded that people who had been sexually abused were more likely to self-harm than those who had not been sexually abused. [EL = 2+]

A case–control study (n = 188) undertaken in the USA examined the relationship between physical abuse and suicidal behaviour in adolescents (aged 12–18 years) who had either been physically abused (n = 99, based on social service register) or not (n = 99, randomly identified via telephone interview, age 15 or 16, sexually abused excluded).¹²⁰ The study found a difference between groups for suicide ideation (P = 0.014) but not for probability of suicide. Multivariate analysis found that physical abuse was not a predictor of suicide probability (P = 0.099), while other factors were: family cohesion (P = 0.004), adult disruptive disorder (P = 0.0003) and adolescent unipolar depression (P = 0.003). The study concluded that abused adolescent had higher suicide probability scores than non-abused, but the link between the two was not direct. [EL = 2+]

A case–control study (n = 71) undertaken in the USA examined the relationship between abuse and neglect (based on childhood trauma questionnaire) and suicidal behaviour in children (52.2% girls, mean age 14.8 years, 5% white) admitted to an acute medical facility over a 1 year period.¹²¹ The study found that sexual abuse (P < 0.001), physical abuse P < 0.01), emotional abuse (P < 0.01), emotional neglect (P < 0.001) but not physical neglect (not statistically significant) were linked with suicide attempts. Multivariate analysis showed that sexual abuse (P < 0.01) and emotional neglect (P < 0.05) but not physical abuse, emotional abuse or physical neglect were linked to attempted suicide. Furthermore, the analysis showed that sexual abuse (P < 0.01) and emotional neglect (P < 0.05) but not physical abuse, emotional abuse or physical neglect were linked to self-harm. When gender was added into the model, female gender (P = 0.001) and sexual abuse (P = 0.05) were predictors of attempted suicide. The study concluded that emotional neglect was an unrecognised predictor of attempted suicide. [EL = 2−]

A case–control study (n = 3416) undertaken in the USA examined the factors associated with suicide attempts in female adolescents involved in a twins cohort study (mean age 15.5 years, 13% non-white).¹²² The study found that 143 (4.2%) had attempted suicide. The study found using multiple regression that physical abuse (based on questionnaire; 2.2% versus 15.7%) was associated with attempted suicide (OR 3.5; 95% CI 1.6 to 7.3). It also found that alcohol dependence, conduct disorder, major depression, social phobia, and African-American ethnicity were statistically significant markers, but alcohol abuse, any specific phobia and generalised anxiety were not. Furthermore the study found that suicide within the family was a statistically significant predictor for attempting suicide. The study concluded that familial factors and possibly genetics played a role in suicide attempts. [EL = 2−]

A case–control study (n = 292) undertaken in New Zealand examined the risk factors for suicide attempts in adolescents (aged 13–24 years).¹²³ The study compared those who had attempted suicide requiring medical treatment (n = 129) against a randomly selected group of people who had not (n = 153, age and gender stratified). The study found that sexual abuse (adjusted OR 3.7; 95% CI 1.6 to 8.3; P < 0.005) was a marker for suicide attempts. However, it also found that poor parental relationship, affective disorder, substance use, antisocial behaviour, age, low education outcome, low income and residence changed within 6 months were also statistically significant predictors (P < 0.001 to 0.05). The study concluded that risk of suicide increased as social adversity increased. [EL = 2−]

A case–control study (n = 88) undertaken in Australia examined the risk-factors associated with self-harm in adolescents (mean age 16.4 years).¹²⁴ The study compared those who had self-harmed (n = 52, 69% female) against a reference group (n = 36, 61% female) being treated for medical conditions or undergoing surgery with a no history of self-harm or psychological illness. The study found that physical abuse (based on responses to a questionnaire; 13 versus two; OR 6.5; 95% CI 1.5 to 29), but not sexual abuse (six versus three; OR 2.0; 95% CI 0.5 to 8) was a predictor of self-harm. The study also found that family structure and substance use were statistically significant predictors of self-harm. The study concluded that self-harm was linked to serious personal and interpersonal problems and a multidisciplinary approach was required to identify and treat it. [EL = 2−]

A cross-sectional survey (n = 352) undertaken in the USA examined the relationship between sexual and/or physical abuse (reported by questionnaire) and substance use and suicide among pregnant teenagers.¹²⁵ The study found that 39 had been physically abused, 52 had been sexually abused, 11 had been sexually and physically abused and 272 had not been abused. Of these groups, 46%, 33%, 83% and 12%, respectively, reported suicidal ideation (P < 0.0001). The study concluded that pregnant teenagers should be screened for abuse and suicidal ideation. [EL = 3]

A case–control study (n = 114) undertaken in Israel examined the relationship between depression and suicide in abused children (aged 6–12 years, 61.4% males).¹²⁶ There were three groups: Group 1 (n = 41) had been physically abused (based on questionnaire responses); Group 2 (n = 38) had been neglected; and Group 3 (n = 35) had been neither abused nor neglected. The study reported that suicidal ideation was found in 22 of Group 1, two of Group 2 and two of Group 3 (r² = 33.63; P < 0.001). Suicidal expression was found in 23 of Group 1, two of Group 2 and two of Group 2 (r² = 37.21; P < 0.001). Risk-taking behaviour was found in 31 of Group 1, two of Group 2 and three of Group 3 (r² = 57.54; P < 0.001). The study concluded that the physically abused group had higher suicidality than the others. [EL = 2−]

A case–control study (n = 117) undertaken in the USA examined the relationship between maltreatment and suicide in adolescents (aged 13–18 years, mean age 14.6 years, 66 females, 82.4% white) admitted to a psychiatric facility.¹²⁷ The group was split between those who had attempted suicide, suicidal ideators and those who were not. The study found that those reporting having been abused (based on questionnaire, n = 55) were statistically significantly more likely to have attempted suicide or have suicidal ideation than those who were not (n = 62) (P < 0.05). Furthermore, the study found that frequency of abuse was related to number of suicide attempts and suicidal ideation for both sexual and physical abuse (P < 0.05). The study found that duration of abuse was related to number of suicide attempts and suicidal ideation for sexual abuse (P < 0.05) but not for physical abuse. The study concluded that history of abuse was related to number of suicide attempts. [EL = 3]

A case–control study (n = 157) undertaken in the Netherlands examined the relationship between life events in childhood (younger than 12 years) and suicidal behaviour in adolescents (aged 12 years or over) in a group aged 14–21 years (mean age 17.5 years, 41 females).¹²⁸ The study compared three groups: Group 1 (n = 48) were people who had attempted suicide (selected within mental health services); Group 2 (n = 66) were depressed (selected within mental health services); and Group 3 (n = 43) were non-depressed people who had never attempted suicide (selected at random from a student population). The study found statistically significant differences (P < 0.05) between the rate of physical abuse before the age of 12 years between the three groups: on average, people who had attempted suicide reported 0.19 (SD 0.49) sexual abuse events, depressed adolescents reported 0.14 (SD 0.43) events and normal controls reported 0.00 events per person. The study found no statistically significant differences between the three groups in the number of episodes of sexual abuse before the age of 12 years: 0.17 (SD 0.48) versus 0.05 (SD 0.21) versus 0.05 (SD 0.21). The study found 0.23 (SD 0.42), 0.29 (SD 0.46) and 0.07 (SD 0.26), respectively, episodes of physical abuse after the age of 12 years (P < 0.05 for difference between depressed and normal controls). The study found on average in each group 0.44 (SD 0.68), 0.26 (SD 0.54) and 0.05 (SD 0.21), respectively, episodes of sexual abuse after the age of 12 years (P < 0.05 for difference between attempters and normal controls). The study found on average in each group 0.13 (SD 0.33), 0.09 (SD 0.29) and 0.00, respectively, episodes of physical abuse within the past year (not statistically significant). The study found on average for each group 0.10 (SD 0.31), 0.05 (SD 0.27) and 0.00, respectively, episodes of sexual abuse within the past year (P < 0.05 for difference between attempters and normal controls). However, change in living situation, change in caregiver, separation of parents and total number of life events experienced were all associated with differences between groups (P < 0.05). The study concluded that the number of life events was linked to suicidal behaviour. [EL = 2−]

A case–control study (n = 597) undertaken in the USA examined the relationship between sexual abuse and psychological problems (suicide and self-harm) in females (mean age 15.6 years) being treated for substance abuse.¹²⁹ The girls were divided into four groups: Group 1 were non-victims (n = 383); Group 2 experienced extra-familial abuse (based on questionnaire, n = 120); Group 3 experienced intra-familial abuse (n = 47); and Group 4 experience both extra-familial and intra-familial abuse (n = 43). The study found that suicidal behaviour was statistically significantly more likely in the abused girls than non-abused (P < 0.0001). There was no difference between groups for suicide attempts (20.4%, 35.7%, 56.5% and 44.2%, respectively). Suicidal thoughts were more likely in the abused than non-abused (52.4%, 64.1%, 65.2% and 74.4%, respectively; P < 0.05) and eating problems were also more prevalent (P < 0.05). However, nervousness (P < 0.01), sleeplessness (P < 0.001) and sexual problems (P < 0.001) were also linked to suicidal behaviour. The study concluded that within a group who already had multiple problems, sexual abuse leads to different and more serious psychopathology. [EL = 2−]

A case–control study (n = 570) undertaken in the Netherlands examined the characteristics of children (aged 15 or 16 years) who did or did not have a history of suicidal behaviour.¹³⁰ The sample was taken from a larger school survey of 13 400 children. Group 1 had a history of suicidal behaviour (n = 185 females, 100 males) and Group 2 did not (n = 185 females, 100 males). Analysis was undertaken by gender. For females the study found that physical abuse (based on questionnaire) (51% versus 24%; P < 0.001) and sexual abuse (32% versus 7%; P < 0.001) were related to attempting suicide. In addition, depression, suicidal thoughts, low self-esteem, feeling of failure, negative future achievements and substance abuse were all statistically significantly related to suicide attempts. For males the study found that physical abuse was not statistically significant (37% versus 32%) and sexual abuse (22% versus 2%' P < 0.001) was statistically significantly related to attempting suicide. In addition, depression, suicidal thoughts, low academic achievement and substance abuse were statistically significantly related to attempted suicide. The study concluded that, in addition to other variables, sexual and physical abuse need to be taken into account when dealing with youngsters demonstrating suicidal behaviour. [EL = 2+]

A cross-sectional survey (n = 775) undertaken in the USA examined the relationship between sexual/physical abuse and suicidal behaviour in children (aged 12–19 years, 65% male, 46% white) who were homeless.¹³¹ The study found that 451 (58%) had thought about suicide (195 of 272 females and 256 of 505 males) and 266 of 775 (34%) had attempted suicide (130 of 272 females and 136 of 505 males). There were statistically significant differences between genders in suicidal thoughts and suicide attempts (P < 0.05). The study found that 119 of 503 males and 189 of 272 females had been sexually abused (based on questionnaire), and of these, for 96 males and 167 females it had happened before they left home. The study reported that 175 of 503 males and 153 of 272 females had been physically abused before leaving home. The study found that 225 of 503 males and 217 of 272 females had been sexually and/or physically abused. In all cases, females were statistically significantly (P < 0.05) more likely to have been abused than males. Logistic regression found that for females being sexually abused before leaving home (OR 3.2; 95% CI 1.8 to 5.6) and being physically abused at home (OR 1.9; 95% CI 1.1 to 3.3) was associated with suicidal behaviour. For males it found that being sexually abused at home (OR 4.3; 95% CI 2.5 to 7.1) and being physically abused at home (OR 4.2; 95% CI 2.6 to 6.5) was associated with suicidal behaviour. The study concluded that interventions on homeless children must take account of physical and sexual abuse. [EL = 3]

A cross-sectional survey (n = 1051) undertaken in the USA examined the relationship between suicidal ideation and maltreatment or risk of maltreatment in a group of children (52.5% female, 55.1% white) who were 8 years old.¹³² The study found that 9.9% of the sample had thought about suicide. The study found that white ethnicity (OR 0.55; 95% CI 0.32 to 0.84), maltreatment (OR 1.91; 95% CI 1.14 to 3.20) and witnessed violence (OR 1.68; 95% CI 1.34 to 2.06) were markers of suicidal ideation (P < 0.05). The study also found that psychological problems and substance use were statistically significant predictors of suicide ideation (P < 0.05), but that maltreatment was not (OR 1.49; 95% CI 0.74 to 2.78). Subgroup analysis on children who had been maltreated (rather than those at high risk) found that severity of physical abuse (OR 1.24; 95% CI 1.04 to 1.48), chronicity of maltreatment (OR 1.19; 95% CI 1.02 to 1.39) and multiple types of maltreatment (OR 1.81; 95% CI 1.11 to 2.95) were markers of suicide ideation. The study concluded that the risk factors of ethnicity, maltreatment and witnessed violence were all mediated by a child's psychological and behavioural variables. [EL = 3]

A survey of secondary school students (n = 839, aged 14–17 years, mean age 15.9 years) in Turkey investigated the relationship between child maltreatment (physical, emotional and sexual abuse, and neglect) and attempted suicide, self-mutilation and dissociation.¹³³ Thirty-four percent of the cohort reported at least one type of maltreatment. Suicide attempt was reported by 10% of the cohort and self-mutilation (including banging head, hitting, cutting, hair-pulling and burning) was reported by 20%. A statistically significant relationship was found between ever having been maltreated and both attempting suicide and self-mutilation. Dissociation scores according to the Turkish version of the Dissociative Experiences Scale were statistically significantly higher in maltreated children than non-maltreated children. [EL = 3]

A case–control study (n = 352) undertaken in Australia examined the relationship between family functioning, sexual abuse and suicidal behaviour in children (aged 14–18 years, mean age 15.2 years, 99% white) from a single high school.¹³⁴ The study found that 20 females (13.2%) and nine males (4.5%) claimed to have been sexually abused. Of those who claimed to be abused, 24.1% had no suicidal behaviour, 13.8% had suicidal thoughts, 10.3% had made plans, 1% had self-harmed, 13.8% had made a single attempt, and 10% had made multiple attempts. Of the non-abused, 32 (9.1%) had thought of suicide, 16 had planned suicide, 15 had self-harmed, 20 had made a single attempt, and 16 (4.6%) had made multiple attempts. The study found that, of 161 children from dysfunctional families, abused children (53% of 19) were more likely than non-abused (8.5% or 142) (χ² = 24.1; P < 0.001). In functional families with abuse, the RR of suicidal behaviour was 7.1, in abused children in dysfunctional families the RR was 6.2, and in abused children in dysfunction families the RR was 9.4, compared with normal children. The study concluded that sexual abuse was more important to suicidal behaviour than family dysfunction. [EL = 3]

A case–control study (n = 127) undertaken in the USA examined the correlates between child abuse (based on questionnaire responses) and risk of suicide in children (aged 12–18 years, mean age 15.8 years, 38 males, 109 white) admitted to a psychiatric unit.¹³⁵ Group 1 were children who reported abuse based (on the Millon Adolescent Clinical Inventory (MACI) abuse scale, n = 74, mean age 16.0 years) and those who reported depression (on the DSM-III-R criteria and Beck depression scale, n = 53, mean age 15.6 years). The study found no difference in reported suicidal behaviour (mean suicide risk scale score 9.1 (SD 2.6) versus 8.3 (SD 2.6)) between abused or not. The study found that self-criticism (P = 0.02) on a depressive experience questionnaire for adolescents, alcohol abuse (P = 0.02) on an alcohol abuse involvement scale and previous feelings or acts of violence (P = 0.08) on a past feelings and acts of violence scale were associated with suicidal behaviour. The study concluded that abused children at risk of suicide report different psychological profiles from those who have not been abused. [EL = 2−]

A prospective cohort study (n = 144) undertaken in the UK examined the relationship between sexual abuse and psychological disturbance in children (aged 16 or younger, 75% females) where alleged or suspected sexual abuse had taken place.¹³⁶ All were investigated then followed-up at 4 weeks, 9 months and 2 years. The study found that by 4 weeks there were no self-mutilation or suicide attempts (n = 99), by 9 months there were five and five (n = 91), respectively, and by 2 years (n = 66) there five and eight, respectively. The study found no statistically significant change in the frequency of events over time. The study made no conclusions in relation to maltreatment and psychological problems, but highlighted that the level of problems did not change with time. [EL = 3]

A retrospective case series (n = 112) undertaken in Australia examined factors associated with repeat suicide attempts in adolescents (aged 13–20 years, 36 males of mean age 18.6 years, 76 females of mean age 17.5 years).¹³⁷ Multivariate analysis found that chronic medical conditions (OR 3.29; 95% CI 1.11 to 9.78), non-affective psychotic disorder (OR 3.81; 95% CI 1.05 to 13.89), alcohol abuse (OR 3.56; 95% CI 1.02 to 12.42) and drug abuse (OR 4.22; 95% CI 1.29 to 13.84), but not sexual abuse (OR 3.03; 95% CI 0.95 to 9.71), were statistically significantly associated with repeat suicide attempts. The study concluded that a multidisciplinary approach was required to investigate and treat adolescents who have attempted suicide. The study further concluded that sexual abuse was likely to be under-reported in the retrospective sample, so was likely to be a more important factor than the results suggest. [EL = 3]

A cross-sectional survey (n = 7241) undertaken in the USA examined the risk factors associated with suicide among Navajo adolescents (mean age 14.4 years) as part of a community survey.¹³⁸ Multiple regression analysis adjusted for age and gender found that physical abuse (OR 1.9; 95% CI 1.5 to 2.4), sexual abuse (OR 1.5; 95% CI 1.2 to 1.9), being female (OR 1.7; 95% CI 1.4 to 2.0), a family history of suicidal behaviour (OR 2.3; 95% CI 1.6 to 3.2), friend attempt (OR 2.8; 95% CI 2.3 to 3.4), poor health (OR 2.2; 95% CI 1.3 to 3.8), mental health problems requiring professional help (OR 3.2; 95% CI 2.2 to 4.5), extreme alienation from family (OR 3.2; 95% CI 2.1 to 4.4) and alcohol abuse (OR 2.7; 95% CI 1.9 to 3.9) were all associated with suicide attempts. The study concluded that prevention of suicide needs to target certain risk factors. [EL = 3]

A cohort study (n = 659, 91% white) undertaken in the USA examined the relationship between childhood adversity and suicide attempts during late adolescence and early adulthood (mean age 22 years) from a community sample of families surveyed four times over 18 years.¹³⁹ The study reported that physical childhood abuse (16/587 versus 5/36; OR 5.10; 95% CI 1.78 to 14.64) and sexual abuse (19/602 versus 4/21; OR 7.22; 95% CI 2.22 to 23.53), controlling for age, sex, psychiatric symptoms and parental psychiatric disorders, were statistically significantly related to suicide attempts during late adolescence and early adulthood. However, the study found statistically significant relationships on a further 20 variables. The study found that the effects of childhood maltreatment and adversity were mediated by interpersonal problems during middle adolescence. The study concluded that maladaptive parenting and childhood maltreatment may be associated with severe interpersonal difficulties during adolescence. [EL = 3]

A case–control study (n = 664) undertaken in Canada examined the relationship between sexual abuse and delinquent and self-destructive behaviour in girls.¹⁴⁰ Three groups were compared: Group 1 (n = 140, mean age 14.8 years) who had recently disclosed sexual abuse to authorities; Group 2 (n = 94, mean age 15.05 years) who reported sexual abuse in a survey; and Group 3 (n = 430, mean age 14.97 years) who had not reported sexual abuse. The study found that victims of sexual abuse were more likely than the non-abused to report the following: self mutilation, eating disorders, resisting help and dangerous acting-out (all P < 0.001). Those that had disclosed abuse were statistically significantly more likely (P < 0.01) than those who had not reported abuse to open veins (OR 4.96), to bang head (OR 1.73), to refuse medication (OR 1.94), to not ask for help (OR 1.72), to refuse to eat (OR 2.08), to display daredevil behaviour (OR 1.72), to induce vomiting (OR 2.24) and to scratch till bleeding (OR 1.29), but not to burn skin, punch walls, throw self from vehicle, cut self, strangle self, swallow poison, hit/prick self or use laxatives. The study examined the family structure correlates for maltreatment, and a model containing family adversity, economic problems, violence during abuse, relation with mother and depression explained 48% of the variance of self-injury. The study reported statistically significant differences between abused and non-abused children. [EL = 2−]

A cross-sectional survey (n = 661 males and n = 1323 females) undertaken in the USA examined the risk factors for attempting suicide among Native Alaskan youths (aged 12–18 years) who responded to a survey that they had attempted suicide.¹⁴¹ The study found that sexual abuse was linked to attempted suicide in males (OR 2.17; 95% CI 1.39 to 3.39) and in females (OR 1.46; 95% CI 1.21 to 1.77). The study found that physical abuse was linked to attempted suicide in males (OR 1.60; 95% CI 1.16 to 2.19) and in females (OR 1.73; 95% CI 1.44 to 2.08). However, age, substance misuse, friend or family suicide, mental health and family structure were also found to relate to suicide. [EL = 3]

A cohort study (n = 3017) undertaken in Canada examined the correlates with suicide attempts.¹⁴² Surveys were undertaken at three points in the individual's life: aged 6–12 years, then 15–18 years, then 19–24 years. The study included a random selection of 2000 (999 females) children and a second sample of 1017 (424 females) children who showed disruptive behaviour. Multiple regression analysis identified sexual abuse (OR 1.2; 95% CI 1.1 to 1.3) as being linked with suicide attempts. However, persistent ideation, insecure attachment, disruptive disorders and female gender were also statistically significant. Physical abuse was not statistically significant on univariate analysis and thus not included in the model. A regression model stratified by gender found that sexual abuse was statistically significant for suicide attempts in females (OR 1.22; 95% CI 1.06 to 1.41) but not males, and that different sets of variables were related to suicidal ideation in both groups. The study concluded that suicide ideation changes with persistence of ideation and gender. [EL = 3]

A case–control study (n = 134) undertaken in the USA examined the familial risk factors for suicide in adolescents.¹⁴³ Two groups were assessed: Group 1 (n = 67, mean age 17 years, 95% white, 85% male) were adolescents who had committed suicide, relatives of whom were interviewed; Group 2 (n = 67) were randomly identified and demographically matched adolescents. The study found that physical abuse within the past year was statistically significantly related to suicide (P = 0.06) but physical abuse before the past year was (P < 0.01). Sexual abuse was not statistically significantly related to suicide. Parent/carer–child conflict, parental unemployment, parent somatic illness, parent legal trouble, move from neighbourhood and parental mental disorders were found to be related to suicide. A multiple regression model showed that family history of depression, family history of substance abuse and lifetime history of parent–child discord were statistically significantly related to suicide. The study concluded that children of people with depression and/or substance abusers should be screened for suicidal behaviour. However, the study was based on relatives' recall and was thus liable to bias. [EL = 2−]

Additional evidence

In addition to the evidence on the relationship between maltreatment and self-harm in children there is a larger body of work examining the long-term impact of child maltreatment in adults. This evidence has not been reviewed here but points to a relationship between childhood maltreatment, particularly sexual abuse, and later self-harm (suicide, self-destructive behaviour and self-harm).

Evidence statement

Evidence from 16 studies found a statistical link (P < 0.05) between sexual abuse and suicidal behaviour compared with five studies that showed no association. Evidence from ten studies found a statistical link (P < 0.05) between physical abuse and suicidal behaviour compared with five studies that found no association. Evidence from four studies showed a statistical link (P < 0.05) between sexual abuse and self-harm compared with one that did not, and two studies found a statistical link (P < 0.05) between sexual abuse and self-destructive behaviour. Evidence from two studies found a statistical link (P < 0.05) between physical abuse and self-harm compared with two that did not, and one study found a link between physical abuse and self-destructive behaviour. Few studies examined emotional abuse or neglect.

There were general problems in the research due to self-reporting of maltreatment (28 of 31 studies) and varying definitions used for maltreatment and self-harm. This makes comparison of studies and reporting of figures unreliable.

In addition, maltreatment is usually found in association with a set of other personal, familial and wider social problems. Therefore, the causal pathway of any statistical association may not be direct.

GDG considerations

While many activities undertaken by children and young people may be harmful (for example, ingesting alcohol or illicit drugs), the GDG believes it is important to focus on the issue of intent to harm the self and for healthcare professionals to be alert to the deliberate nature of self-harm in some children and young people and its link to child maltreatment. The GDG wishes to raise awareness of the clinical evidence for pre-teenage children to present with deliberate self-harm even though traditionally such behaviour might be thought to be restricted to teenagers.

There was consensus within the GDG about the recommendation in this section and thus the views of the Delphi panel were not sought.

Narrative summary

One case–control study was found that reported the differences in somatic and emotional reactions of girls who had reported sexual abuse and those who had not.¹⁴⁵ Seventy-two children who had attended a referral centre for sexual abuse were identified for inclusion in the study and controls of similar age and initial clinic visit date and no history of physical abuse were selected from admission records to a general clinic. Data were extracted from medical records on a number of reported symptoms including gastrointestinal irritability and chronic abdominal pain. Children who had been sexually abused were more likely to have reported chronic abdominal pain than those in the control group (P < 0.01). [EL = 2−]

GDG considerations

The GDG did not identify a good evidence base for whether a history of recurrent abdominal pain is a reason to suspect child maltreatment. The GDG believes that, in the absence of an obvious medical cause, recurrent abdominal pain can be caused by emotional disturbances resulting from child maltreatment. However, as recurrent abdominal pain is common and often unexplained, the GDG was not able to make a recommendation.

Overview of available evidence

Five studies were identified that looked at disordered eating in association with maltreatment. No suitable published literature was identified that looked specifically at hoarding or stealing behaviours.

Narrative summary

A US-based case–control study (n = 40, aged 10–15 years) investigated whether sexually abused (defined as unwanted sexual activity or sexual activity that involved a person more than 5 years older) girls in treatment for abuse showed more eating disorder behaviours than non-abused girls and whether multiple forms of abuse increased the severity of the eating disturbance.¹⁴⁶ Girls in both groups were asked to fill in the Childhood Trauma Questionnaire (CTQ), the Body Rating Scale for Adolescents, the McKnight Risk Factor Survey and the Kids' Eating Disorder Survey (KEDS). Fifteen items were reported on and the sexually abused girls had statistically significantly greater weight dissatisfaction, reported eating less when they were bored, upset or trying to feel better about themselves, had a lower score on perfectionism and chose a thinner figure that represented how they would like to look than non-abused girls. [EL = 2−]

A number of studies in this area have arisen out of a large US survey of secondary school students in Minnesota conducted in 1987. The first paper reviewed here selected females who reported that they had ever been sexually abused and had discussed the problem with someone (n = 1011, mean age 15.28 years).¹⁴⁷ They were compared with a group selected randomly from the survey cohort who had not been sexually abused according to the survey questions (n = 1011, mean age 14.92 years). Prevalence of evaluating oneself as overweight (55.6% versus 43.7%), binge-eating (40.3% versus 31.7%), non-stop eating (24.6% versus 16.7%), more than ten dieting episodes in the preceding year (17.9% versus 12.3%), self-induced vomiting more than once a week (4.4% versus 2.7%), use of diuretics (4.4% versus 2.7%) and use of laxatives (3.7% versus 2.2%) were found to be statistically significantly higher in the girls who reported abuse than those who did not. [EL = 3]

A 10% subsample (n = 6224) from the Minnesota study was used to investigate associations between abuse history and disordered eating in 9th and 12th graders only.¹⁴⁸ Adolescents were said to have disordered eating if they reported two of out-of-control eating, using laxatives and vomiting. There were 318 females and 84 males who met these criteria and reported at least one type of abuse. Some participants reported more than one type of abuse but this was not accounted for in the analysis. Approximately twice as many abused females had disordered eating than non-abused females; in males, approximately ten times as many had disordered eating in the abused group compared with the non-abused group. [EL = 3]

Another study compared eating behaviours and weight perception of males (n = 370, mean age 15.26 years, SD 1.7 years) and females (n = 2681, mean age 15.37 years, SD 1.7 years) who reported past sexual abuse (defined as ‘someone in your family, or someone else, touches you in a place you did not want to be touched, or does something to you sexually which they shouldn't have done’).¹⁴⁹ More abused girls than boys thought of themselves as overweight (52% versus 21%), reported binge-eating episodes (41% versus 22%), reported being afraid of not being able to stop eating (23% versus 8%), had dieted in the preceding year (70% versus 27%), had induced vomiting in themselves (20% versus 10%) and had used diuretics to lose weight (3.7% versus 1.4%). More boys than girls were satisfied with their body weight and proud of their body. There were no statistically significant differences between males and females in the use of laxatives (1.6% versus 3%) or ipecac (1.4% versus 1.1%). [EL = 2−]

Another large survey of adolescent females in the USA (n = 7903, mean age 14.5 years, SD 1.6 years) investigated whether increasing numbers of episodes of physical or sexual abuse led to increasing numbers of purging episodes.¹⁵⁰ The study found an association between physical abuse and purging behaviour (OR 1.81; P = 0.0014) after adjusting for some confounders but found no relationship between sexual abuse and purging behaviour. [EL = 3]

Evidence statement

A number of surveys have investigated eating behaviours and attitudes to body weight and their relationship with maltreatment. The studies are generally of poor quality but suggest that children who had been maltreated reported more bingeing than those who had not.

GDG considerations

There is a range of disturbance in eating behaviour in children, including hoarding, hiding and stealing food, bingeing, pica and disturbed feeding patterns. It is the GDG's view that these can be associated with various forms of maltreatment because they may be a manifestation of underlying distress, of a lack of physical and emotional nurturing, or of disturbed parent/carer–child interactions focused around feeding. The GDG believes that once medical causes such as bulimia and problems in the autistic spectrum have been ruled out, these behaviours are concerning. The GDG is also of the view that eating disorders, seen more commonly in older children and adolescents, which include anorexia nervosa, bulimia and obesity, may also be associated with a past history of maltreatment. The strength of association varies according to the type of disorder. The GDG chose not to make a recommendation about eating disorders in relation to current abuse.

There was consensus within the GDG about the recommendation in this section and thus the views of the Delphi panel were not sought.

Overview of available evidence

One cross-sectional study was identified.

Narrative summary

A German study of children (n = 140, aged 10 months to 11 years) in residential care homes asked caregivers to rate the occurrence of 15 stereotyped behaviours in non-handicapped children in their care.¹⁵¹ Of the children included in the study, 45 had a history of suspected child abuse; this was not defined in the paper. In the questionnaire, caregivers were asked to rate how often each child performed each behaviour. The results cited were based on daily occurrences. Body rocking was observed in 11.1% of suspected abuse cases and in 6.3% of the remaining children, head nodding or shaking was observed in 4.4% of the suspected abuse cases and in 4.2% of the remaining children and head-banging was observed in 4.4% of the suspected abuse cases and in 1.1% of the other children. None of these proportions were statistically significantly different between groups. This result could be due to the reasons that the children are in residential care. [EL = 2−]

Evidence statement

The retrieved study indicates that head-banging and body rocking are uncommon behaviours in children who have a history of suspected abuse and are no longer living with their families.

Delphi consensus (see also Appendix C)

The lack of literature in this subject caused the GDG to seek external validation for their opinions. The following statement on body rocking was put into the Delphi survey:

GDG considerations

The GDG believes that body rocking is associated with emotional neglect and that it is a sign of inadequate stimulation. Body rocking is common in children and young people with learning disabilities and, while it is important to exclude neurodevelopmental disorders as the cause of the rocking, it is imperative to recognise that abuse may be the cause. The GDG sought the opinions of the Delphi panel on Statement 37a about body rocking and sufficient agreement was reached (see above and Section C.2.3). The GDG incorporated this statement into the above recommendation on behaviours and emotional states (see Section 7.1).

Habitual head-banging can be distinguished from that associated with an outburst of anger. While habitual head-banging is a relatively uncommon clinical finding, there is no general prevalence data. The data linking it with child maltreatment are weak. Therefore, the GDG, having sought the opinion of the Delphi panel, chose not to make a recommendation about head-banging (see above and Section C.2.8).

Narrative summary

One systematic review pooled comparative data on the effects of CSA,⁸⁷ acknowledging that source materials were heterogeneous. This review found eight studies which compared sexualised behaviours in sexually abused children and controls from the community. In all eight studies, sexually abused children showed more sexualised behaviour than the children who had not been sexually abused. [EL = 2−]

One descriptive systematic review on the sexual abuse of boys concluded that abused males (younger than 19 years) showed more sexualised behaviours, such as difficulty controlling sexual feelings, hypersexuality, coercive behaviour towards others, engagement in prostitution and unprotected sexual intercourse, than non-abused boys.⁸⁶ [EL = 2−]

A comparative study of girls who were being treated after sexual abuse within a 2 year period of reporting abuse reported scores on the CSBI in 20 CSA cases, 20 psychiatric controls and 20 non-psychiatric controls.¹⁵⁹ Mean CBSI scores were found to be 30.6 (SD 20.3), 15.2 (SD 9.9) and 10.8 (SD 9.6), respectively, and the groups were found to be statistically significantly different. [EL = 2+]

A retrospective study matched children who had been sexually abused (n = 22, 13 girls, age range 2–7 years) with controls recruited from a paediatric practice and a public health centre.¹⁶⁰ The children were interviewed with a questionnaire about sexual knowledge. No differences were found in the sexual knowledge of the two groups. [EL = 2−]

One case–control study compared children (n = 17, age range 5–15 years) who had been sexually abused and were protected from the perpetrator at the time of investigation with a group of controls (n = 17) matched on age, sex, socio-economic status and current living situation (single parent, divorced parents, etc).¹⁶¹ A number of validated questionnaires were applied to all children in the study or their caregivers as appropriate, including the CBCL, on which the six sex problem items were combined to give a sex problem score. On this measure, the abused children scored higher than those in the control group (P = 0.05). In the abused group, the alleged abuse had happened within the year before the study and a wide range of abuses was reported. [EL = 2−]

A longitudinal survey of children who had either been maltreated early in life or who were at risk of early maltreatment investigated the effects of maltreatment other than sexual abuse on sexualised behaviours (n = 690, age approximately 8 years at data collection, 53% male).¹⁶² A modified version of the CSBI was used to measure sexualised behaviours; maltreatment reports to child protective services were classified as early if they occurred before age 4 years and late if they occurred between age 4 years and the time of the survey. Children who had reports of sexual abuse were excluded. Late physical abuse was associated with boundary problems (OR 1.9; 95% CI 1.1 to 3.5), displaying private parts (OR 2.4; 95% CI 1.0 to 5.6) and sexual intrusiveness (OR 2.6; 95% CI 1.3 to 5.2). Late emotional abuse was associated with sexual knowledge (OR 2.0; 95% CI 1.2 to 3.4). Early physical abuse was associated with displaying private parts (OR 2.4; 95% CI 1.1 to 5.4). Early emotional abuse was protective against displaying private parts (OR 0.3; 95% CI 0.1 to 0.8) and early neglect was protective against sexual intrusiveness (OR 0.4; 95% CI 0.2 to 0.9). There was no normative sample in this study. [EL = 2−]

A survey of sexually active African-American females (n = 725, mean age 16.6 years, SD 1.6 years) attending an adolescent primary care and prevention clinic investigated associations between reports of sexual abuse and attitudes towards condom use.¹⁶³ Participants were asked whether they had ever been sexually abused or molested and at what age. Those who said they had (n = 167) reported a greater number of sexual partners in their lifetime (6.5 versus 4.4; P < 0.05) and a greater frequency of unprotected vaginal sex in the preceding 90 days (5.7 versus 4.5; P < 0.05) than those who had not (n = 558). There were no differences between the groups in frequency of protected vaginal sex in the preceding 90 days or condom use consistency. [EL = 2−]

Evidence statement

The comparative studies cited here show that, for the most part, sexualised behaviour occurs more often in children who have been sexually abused than those who have not. One small study showed that sexual knowledge did not differ between the two groups.

GDG considerations

Based on the GDG's clinical experience and studies of normative behaviour, the GDG believes that certain sexualised behaviours that are uncommonly encountered are a cause for concern and that the explanation of the behaviours should be sought; sexualised behaviours can be associated with sexual exposure, which may be a part of sexual grooming behaviour or contact sexual abuse, both of which form the definition of sexual abuse adopted in this document (see Section 2.6 on definitions of child maltreatment).

The GDG believes that sexualised behaviours as a result of maltreatment become different in nature as children move into adolescence; these include promiscuity, sexually precocious behaviour and risk-taking sexual behaviours. Risk-taking sexual behaviours may be recognised as such or their results come to light when a child or young person has an STI or is pregnant (see Section 4.2.3 on STIs and Section 5.1 on pregnancy). The GDG's clinical experience is that sexual behaviours due to maltreatment are often resistant to limits or distractions set by the parents/carers. However, difficulties in the autistic spectrum should be taken into account.

The GDG believes that children and young people involved in prostitution and sexual exploitation are in need of protection but recognises that the decision to initiate child protection proceedings should not deter the young person from seeking and receiving medical attention.

There was consensus within the GDG about the recommendations in this section and thus the views of the Delphi panel were not sought.

Overview of available evidence

A number of surveys of homeless and runaway youth were identified. Given the low quality of the evidence, a small number of studies have been reviewed in detail and some others have been presented in Table 7.1.

Table 7.1

Surveys of homeless and runaway youth identified.

Narrative summary

In a US study of homeless female adolescents (n = 216, mean age 17.7 years, range 13–20 years), sexual abuse (defined as prepubertal sexual contact with an older person) was reported by 38% of study participants.¹⁷⁴ The mean age of the first incident of abuse was 6.7 years (SD 2.9 years) and the mean age of becoming homeless was 14.3 years (SD 2.5 years).

A survey of homeless and runaway youth (n = 372, median age 17 years, range 13–21 years) found that 47% of responders (n = 326) had been physically abused before they left home and 29% of responders had been sexually abused.¹⁷⁵ There was no difference between males and females in the rates of physical abuse but more females than males had been sexually abused.

A survey of runaways at a shelter (n = 187, median age 18 years, range 16–21 years) reported the reasons why the young people had left home for the first time and the most recent time.¹⁷⁶ Respondents were asked to rate a list of given reasons using a Likert-like scale of importance. Reasons for leaving home the first time, rated as somewhat important, important or very important, were physical abuse (40%), sexual abuse (12%), being thrown out (38%), conflict with a male adult (57%), conflict with a female adult (57%) and feeling unloved (56%). Seventy-four percent of the people surveyed had run away from home more than once; the important reasons for running away the most recent time was physical abuse (33%), sexual abuse (9%), being thrown out (55%), conflict with a male adult (56%), conflict with a female adult (55%) and feeling unloved (48%). The median age of onset of physical abuse was reported to be 12 years. [EL = 3]

Evidence statement

A number of surveys of young people who are either homeless or have run away from home indicate that up to 62% have suffered some form of abuse in the past. Definitions of homelessness and runaway behaviour differ between studies; maltreatment is measured in different ways and is not substantiated in any of the studies. Many studies asked questions about physical or sexual abuse but few reported on neglect or emotional abuse.

GDG considerations

Many of the reasons given by children and young people for leaving home are to do with a negative atmosphere in the home; either conflict or abuse, or fear of conflict or abuse. Although the literature does not indicate clearly that young people who exhibit runaway behaviour are currently in need of protection, the GDG is of the opinion that running away from home implies that the young person perceives the home to be a place that is unsafe or intolerable. The GDG believes that it is important to establish whether parental/carer consent has been given if a child or young person is found not to be living at home, but notes that maltreatment is less of a concern in 16- and 17-year-olds. Refer to the national guidelines on runaways.¹⁶⁴

There was consensus within the GDG about the recommendation in this section and thus the views of the Delphi panel were not sought.