The long-term effects of child sexual abuse
- Determining the association between child sexual abuse and later outcomes
- A range of outcomes
- The impact of child sexual abuse on mental health
- Behavioural aspects of mental health functioning
- Interpersonal outcomes
- Physical health and overall developmental outcomes
- Gender differences in the long-term impacts of child sexual abuse and gaps in understandings of male victims/survivors
- A complex interplay
Determining the association between child sexual abuse and later outcomes
Research concerned with the links between child sexual abuse and later outcomes covers a broad range of areas and methodologies. It is important to be aware of the types of studies in which these findings have emerged, and to understand some of the methodological considerations and limitations of the research.
Determining the association between children's experiences of sexual abuse and later outcomes is not straightforward. Not least among the difficulties is the fact that child sexual abuse is usually hidden as a result of the very nature and underlying dynamics of this form of abuse (Priebe & Svedin, 2008). Many children who are sexually abused take years to disclose such abuse and some never do (Goodman-Brown, Edelstein, Goodman, Jones, & Gordon, 2003; London, Bruck, Ceci, & Shuman, 2005; Putnam, 2003).
Retrospective reporting of abuse
Most studies rely on the retrospective recall of adults about their childhood experience, often because of the ethical issues of asking children to answer questions about sexual abuse and sexual activity while they are still children. There is likely to be some bias in recall and error in these retrospective reports, which rely on the willingness of the respondent to report on them at that time. Fergusson, Horwood, and Woodward (2000), for example, found that when they asked the participants (aged 18-21 years) in their longitudinal study "whether, before the age of 16, anyone had ever attempted to involve them in any of a series of 15 sexual activities when they did not want this to happen",1 there was "considerable instability and change" between the responses given to this question at 18 years of age and 3 years later at age 21, several years after the cut-off age of 16. While there was no evidence to suggest that these reports were influenced by the psychological state of participants at the time, it does indicate that using different ages for retrospective reporting is likely to provide different results. Other research has also found evidence of a bias in recall in relation to false negatives that can lead to under-reporting of sexual abuse and other adversities in childhood (Hardt & Rutter, 2004). The level of reporting of sexual abuse in childhood is also significantly lower for people in older generations than for younger people (Green et al., 2010). The unwillingness of participants to disclose their childhood experiences or unreliability in doing so (Briere, 1992; Hardt & Rutter, 2004) means that any comparison between those allocated to the non-abused group may include some of those who were actually abused, and vice versa.
One way of overcoming the need for retrospective reporting by those potentially victimised as children is to use cases where the abuse was disclosed in childhood and recorded in administrative data or medical or forensic records. This of course means that any association with later outcomes is tied up with the disclosure experience and subsequent events connected with the formulation of these records. As Fergusson et al. (2008) pointed out, this introduces a sample selection bias as these "populations" differ from the general community base and do not include the unknown cases that have never been reported. Reported cases are more likely to involve perpetrators outside the family2 and therefore reflect different responses/reactions from carers/adults (Australian Bureau of Statistics [ABS], 2005; Lamont, 2011).
There is a range of "specialised populations" that provide evidence for a link between child sexual abuse and later outcomes. These special populations include those who have been referred for counselling, those attending specialised clinics, those seeking medical or psychiatric treatment, those in prison or detention, child victim-witnesses (child victims who have been through the legal process), and university or college students. Clearly findings might be expected to vary among these "populations", so generalising to the population at large from such specialised samples is risky (Boden, Horwood, & Fergusson, 2007; Frothingham et al., 2000). Those in detention or in prison or seeking psychiatric treatment are clearly sub-populations showing adverse outcomes, but the question is to what extent is this related to their experience of sexual abuse?
Other important methodological considerations concern the definitions of child sexual abuse and the measurement of outcomes, and the need to take into account other possibly "confounding" factors in disentangling the effects of related and other experiences on later outcomes. Child sexual abuse covers an "array of sexual activities" with children (Putnam, 2003, p. 269), and the experience and impact is also likely to vary in association with a number of factors. These include the relationship between the child and the perpetrator, the age and gender of both the child and perpetrator, and the frequency, duration and form of the abuse. In addition, the child's family circumstances and context are important background and possibly protective factors.
The definition of sexual abuse
It is important to understand how child sexual abuse is defined and how that may vary across a range of studies since this affects the way findings may be compared and the conclusions that can be drawn. Studies on the prevalence and impact of sexual abuse vary in terms of the cut-off age used for "defining" childhood sexual abuse and the various characteristics of the abuse such as the types of sexual behaviours that are included. In addition, studies vary as to how the relationship between the victim and the perpetrator is categorised (often as "within the family" and "outside the family"). For example, in an early Australian prevalence study, Goldman and Goldman (1988) defined child sexual abuse as "some form of sexual abuse or exploitation by age 18 years by a person five or more years older" that included behaviours such as: "being hugged in a sexual way", "adult showing genitals" and being "invited to do something sexual" (p. 99). Such acts were reported to have been experienced by 28% of the girls and 9% of the boys. A telephone survey asked a random sample of 4,449 adults aged 18-59 on the electoral roll whether "someone" had engaged them in a range of both penetrative and non-penetrative sexual experiences when they "did not want them to" before they were 16 years of age (Dunne, Purdie, Cook, Boyle, & Najman, 2003).3 The reported prevalence rates for females were 34% for non-penetrative and 12% for penetrative experiences, and for males, 16% and 5% respectively. In a more recent Australian study, Moore et al. (2010) asked the 24 year-old participants in their 10 year follow-up of a cohort of 14-15 year old Victorian students, whether "any adult or older person involved" them in a range of similar unwanted incidents of sexual activity before the age of 16.4 This study reported that the prevalence of "any sexual abuse with/without contact" was 17% for girls and 7% for boys.
The importance of the definitional issue is that the types of sexual abuse that are included will affect the findings in relation to outcomes, not just prevalence. As Briere (1992) pointed out, it is likely that studies which use a definition of sexual abuse that is restricted to more intrusive and severe forms of abuse involving penetration will report more severe adverse outcomes than those using broader definitions.
There are a range of outcomes associated with child sexual abuse, related to mental health, behaviour and interpersonal relationships. Different studies may use different measures for similar outcomes. For example, studies on mental health outcomes may include clinical frameworks such as the DSM-IV (Green et al., 2010; McLaughlin, Conron, Koenen, & Gilman, 2010; Scott, Smith, & Ellis, 2010; Ystgaard, Hestetun, Loeb, & Mehlum, 2004), or standardised measures such as the Trauma Symptom Inventory (Briere & Elliott, 2003) or the Symptom Checklist 90-Revised (Gold, Lucenko, Elhai, Swingle, & Sellers, 1999), whereas others have used their own study-specific measures. Clearly, it is easier and more reliable to compare the findings from studies that use commensurate measures. Other studies have relied upon official records for hospitalisation or psychiatric admissions, imprisonment, referral to child protection agencies for parenting problems or positive measures of educational achievement. There is a similar picture of diversity with the use of some standardised measures in relation to other types of outcomes as well.
Taking account of other factors
An important methodological issue is the need to disentangle the effects of abuse from other influences by taking account of a range of individual, family and social factors that might affect or contribute to adverse long-term outcomes. The most significant of these are various aspects of the family environment in which the child was living, including the quality of parenting, parental mental health and possible substance abuse, as well as socio-economic status (parental education and employment), and the possibility that the child was exposed to other forms of abuse and adversity, not sexual abuse alone (Higgins & McCabe, 2001; Noll, 2008). Green et al. (2010), for example, found that while 6% of people in their large-scale community survey reported sexual abuse in their childhood, 72% reported having at least one other childhood adversity such as family dysfunction, parental death or divorce, physical illness or economic adversity. In a critical review of 29 earlier studies in which adult retrospective reports of more than one form of child maltreatment (sexual abuse, physical abuse, psychological maltreatment, neglect, or witnessing family violence) were assessed, Higgins and McCabe (2001) found that "the specific impact of multi-type maltreatment … was associated with greater impairment than single forms of abuse or neglect" (p. 547).
Other likely significant contributing and possibly confounding factors for long-term outcomes include the person's more recent and current circumstances and life experiences. However, there is also evidence that suggests some of these life experiences are likely to have been influenced by childhood sexual abuse itself or by the particular vulnerabilities of the child at the time. As outlined later, for example, children who have been sexually abused are more likely than other children to be re-victimised both as adolescents and adults. They are also more likely to have been targeted by the perpetrator specifically because of their particular vulnerabilities including having socially isolated parents who lack partners and other supports (Conte, Wolf, & Smith, 1989; Elliot, Browne, & Kilcoyne, 1995).
There are various ways of trying to control for the likely influence of other adversities in childhood as well as family background, and these include using matched or comparison groups and taking account of these factors statistically. Matching groups and using statistical means of equivalence are by no means foolproof, however, as Briere and Elliott (2003) pointed out:
there are significant statistical issues associated with controlling for abuse-correlated variables when abuse is antecedent to such variables (Davis, 1985; Pedhazur, 1982) or when the abuse variable is, itself, logically inseparable from the controlled variable (Briere & Elliott, 1993). For example, in the case of family environment, child abuse may further disrupt an already dysfunctional family, and a dysfunctional family may be an important aspect of child abuse (especially intra-familial sexual and physical abuse). As a result, controlling for family environment when examining the relationship between abuse and later psychological symptoms may be a highly conservative, or even nonsensical procedure (e.g., examining the effects of incest after removing variance associated with living in a disturbed or dysfunctional family environment). (p. 1206)
Twin studies, involving one twin known to have been abused and the other not, provide a particularly strong research design because they provide a comparison that controls for family background for twins raised together, and also genetic make-up for identical versus mono-zygotic twins (Dinwiddie et al., 2000; Kendler et al., 2000; Nelson et al., 2002).
Causality and design issues
Many studies in this field have relied on cross-sectional designs whereby participants simultaneously provide information about their experience of abuse (or not) as well as their current outcomes such as their mental health or functioning. This can be problematic for several reasons. First, causality cannot be inferred from correlational analyses and, as Briere (1992) pointed out, cause and effect can become blurred in correlational and retrospective designs. It is possible, for example, that "current distress or symptomatology" may affect the way earlier experiences are perceived and reported. It is also possible that "abuse-related symptomatology can wax and wane across the life span", particularly in relation to current life experiences and developmental stages with the emergence, for example, of intimacy and sexual problems in adolescence or "sleeper effects" later in adulthood with the birth of a child (Briere, 1992, pp. 196-197). More importantly, however, cross-sectional studies cannot differentiate "abuse-specific from abuse-concurrent or abuse-antecedent events" (Briere, 1992, p. 197). To what extent are the problems that sexually abused adolescent and adults exhibit a direct or indirect result of the sexual abuse, of other forms of co-existing abuse and family dysfunction, of pre-existing vulnerabilities and problems, or later problems that are not related to the sexual abuse?
The most effective design to overcome these problems is a longitudinal prospective design in which a random selection of children is followed from birth. This allows a comparison between those who experience various forms of abuse, including sexual abuse, and those who were not abused at all. Of course, sexual abuse may not be revealed until later, and perhaps not at all, so weakening any comparison concerning the effects of sexual abuse. Attrition or the loss of participants to the study also reduces the power of the analyses, and there is still the need to take account of other factors apart from abuse that may contribute to adverse outcomes.
Box 1: Towards better methodology
The Christchurch Health and Development Study - A large prospective longitudinal study
The best known substantial prospective longitudinal study which provides useful findings in this field is the Christchurch Health and Development Study (CHDS), a longitudinal study of a birth cohort of 1,265 children born in the Christchurch (New Zealand) urban region in mid-1977, with information from a variety of sources including: parental interviews, teacher reports, self-reports, psychometric assessments, medical, and other record data.
Sexual abuse was assessed by asking whether, before the age of 16, anyone had ever attempted to involve them in any of a series of 15 unwanted sexual activities, including: (a) non-contact episodes involving indecent exposure, public masturbation, or unwanted sexual propositions; (b) episodes involving sexual contact in the form of sexual fondling, genital contact, or attempts to undress the respondent; or (c) episodes involving attempted or completed vaginal, oral, or anal intercourse (Fergusson, Horwood, & Lynskey, 1996; Fergusson, Lynskey, & Horwood, 1996). Using these data, participants were classified into one of four exposure groups reflecting the extent/severity of child sexual abuse reports: (a) no sexual abuse (85.9% of the sample); (b) non-contact sexual abuse only (2.7% of the sample); (c) contact sexual abuse not involving attempted or completed sexual penetration (5.1% of the sample); and (d) attempted or completed sexual penetration including vaginal, oral and anal intercourse (6.3% of the sample) (Boden et al., 2007, p. 1104).
Other longitudinal studies have followed a cohort of children who were identified at some point following the abuse - for example, after coming to the attention of the police or statutory child protection agency (Brown, Cohen, Johnson, & Smailes, 1999; Scott et al., 2010; Trickett, Noll, Reiffman, & Putnam, 2001), attending a child sexual assault unit at a hospital (Swanston et al., 2002), or appearing as witnesses in child sexual assault prosecutions (Quas et al., 2005). The main comparisons in these studies are within group comparisons, looking at the factors that differentiate between better versus poorer later outcomes. Other studies, including a reputable large-scale study in Victoria Australia (Cutajar et al., 2010a, 2010b), have matched official or administrative records indicating the disclosure or investigation of sexual abuse during childhood with later records for various outcomes (for example, a public psychiatric database or corrective service/prison records). This allows an analysis of those cases where the abuse precedes the mental health outcome and overcomes the problem of retrospective recall but of course misses all those cases that do not come to official attention.
Another important type of study is meta-analysis, a systematic review of a body of empirical studies that "looks at the results within each study, and then calculates a weighted average" of the effect size across a number of studies (Cochrane Collaboration, 2002, p. 2). Several meta-analyses have been conducted in relation to the association between child sexual abuse and various adverse outcomes. These include: a meta-analysis of the relationship of child sexual abuse to HIV risk behavior among women (Arriola, Louden, Doldren, & Fontenberry, 2005); a review of meta-analyses on the association between child sexual abuse and adult mental health outcomes (Hillberg, Hamilton-Giachritisis, & Dixon, 2011); a meta-analysis of the risk-factors for perpetration of child sexual abuse (Whitaker et al., 2008); and an earlier meta-analysis on 6 different outcomes (Paolucci, Genuis, & Violato, 2001; see Box 2).
Box 2: A meta-analysis of the published research on the effects of child sexual abuse
Paolucci, E. O, Genuis, M. L., & Violato C. (2001). Journal of Psychology, 135(1), 17-36.
A meta-analysis of published research on the effects of child sexual abuse for 6 outcomes: post-traumatic stress disorder (PTSD), depression, suicide, sexual promiscuity, victim-perpetrator cycle, and poor academic performance. Thirty-seven studies published between 1981 and 1995, mostly US studies, involving 25,367 people were included. The meta-analyses found a significant effect of child sexual abuse on depression (d = 0.44), suicide (d = 0.44), PTSD (d = 0.40), and sexual promiscuity (d = 0.29).
Paolucci et al. concluded that the analyses provide clear evidence confirming the link between child sexual abuse and subsequent negative short- and long-term effects on development. There were no statistically significant differences or effect size when various potentially mediating variables such as gender, socioeconomic status, type of abuse, age when abused, relationship to perpetrator, and number of abuse incidents were assessed. The results support the multi-faceted model of traumatisation rather than a specific sexual abuse syndrome of child sexual abuse.
In summary, research in this area has utilised a range of study designs but recent research increasingly has used more rigorous designs that take into account possible confounding factors and use more standardised measures.
1 These activities spanned: (a) non-contact episodes involving indecent exposure, public masturbation or unwanted sexual propositions; (b) episodes involving sexual contact in the form of sexual fondling, genital contact or attempts to undress the respondent; and (c) episodes involving attempted or completed vaginal, oral or anal intercourse (Fergusson, Boden, & Horwood, 2008, p. 610).
2 With the likely exception of clergy-related abuse (Fogler et al., 2008).
3 The age of consent to sexual activity is 16 in all states and territories in Australia except South Australia and Tasmania, where it is 17 for both males and females (unless the other person is in a position of "care, supervision or authority"). The Dunne et al. (2003) study was criticised by Stanley and Kovacs (2004) on several grounds, including the likely absence of Indigenous respondents and the use of the term "when you did not want them to" in the question because of their concern that children who are sexually abused may not have objected and may feel responsible and ashamed of their involvement.
4 These activities included: inviting/requesting you to do something sexual; kissing or hugging you in a sexual way; touching or fondling private parts; showing their sex organs to you; and making you touch them in a sexual way.