Paper: Evaluating Childhood Adversity

The psychosomatic approach, as outlined by Fava and based on Engel’s biopsychosocial model, involves the determination of psychological factors affecting individual vulnerability to develop, as well as influence the course and outcome of illness [1]. Various types of early life events seem able to alter vulnerability to illness later in life. In light of three decades of research, a comprehensive assessment of a person’s current and future risk for illness requires attention to childhood experiences, both adverse and positive [2‒4]. Fortunately, for clinicians and researchers, there is now an ample array of instruments to aid the relevant data collection. This chapter is divided into two parts. The first part includes a discussion of adverse childhood events, their relevance to psychopathology and methods designed to evaluate them. The second part reviews the instruments currently available for the assessment of adverse childhood events and provides recommendations for the best instruments to use for clinical and research purposes.

Abundant research has shown that adult survivors of early childhood trauma are at increased risk for a range of psychological and physical disorders. A selective list of publications shows links between childhood adversity and heightened anxiety and depression in adult mood [5‒8], suicidality and self-harm [9‒11], substance abuse disorders [12, 13], eating disorders [14‒16], personality disorders [17] and sexual disorders [18] as well as psychosomatic or medically unexplained syndromes [19‒22]. Abuse survivors also overutilize routine healthcare, exhibit impaired work functioning and parenting, and show traumatic revictimization in later years [18, 23, 24]. In this body of detailed empirical research, most emphasis has been placed on the study of physical and sexual abuse, although there is growing evidence that emotional abuse and neglect may be of even greater etiological importance [25‒27]. However, other writers identify violent and sexual traumas as conferring the greatest risk of severe adverse outcomes [24]. Other variables of the abuse-neglect paradigm may also be critical to a full understanding of psychopathology such as accumulation of impact, nature of the key players, the duration of the adverse experiences and the developmental stages at which the experiences occur.

Early trauma appears to compromise core psychobiological self-regulatory capacities, and be associated not only with psychiatric and behavioral morbidity, but also chronic stress-related gastrointestinal, metabolic, cardiovascular and immunological illness [2, 24]. Animal models indicate that events such as premature separation from the mother have consistently resulted in development of pathophysiological modifications, such as increased hypothalamic-pituitary-adrenal axis activation [28]. Extrapolation to humans suggests that similar events may render individuals more vulnerable to the effects of stress later in life [29]. McEwen and Stellar [30] proposed a formulation of the relationship between stress and the processes leading to disease based on the concept of allostasis and the ability of the organism to achieve stability through change. The concept of allostatic load refers to the wear and tear that results from either too much stress or from not turning off the stress response when it is no longer needed [31, 32]. Biological parameters of allostatic load include glycosylated proteins, coagulation/fibrinolysis markers and hormonal markers; these have been linked to cognitive and physical functioning and to mortality [1].

The effects of different types of adverse childhood events may be additive [26, 33, 34]. Polyvictimization (i.e. multiple types of victimization) appears to produce greater risk of internalizing and externalizing problems in children than does any single type of victimization [33, 34]. The Adverse Childhood Experiences study, amongst others, has revealed a dose relationship between adverse life events and important categories of emotional state, health risks, disease burden and healthcare costs [35]. Evidence from longitudinal studies suggests that the overall severity of cumulated experience is probably more important than the type of maltreatment [36, 37].

It is likely that there are different routes between maltreatment and psychopathology. For example, childhood abuse may lead to adult depression via abnormal self-concept and/or by abnormal self-regulation [38, 39]. At a physiological level, neurotransmitter system alterations and changes in brain size, structure and function have been proposed as mechanisms for long-term outcomes of childhood adversity [2, 29, 40]. There are also macro-level considerations; in the Dunedin longitudinal study, socioeconomic factors accounted for most of the relationship between child maltreatment and adult alcohol and drug dependence and cardiovascular risks [41]. None of the mechanisms hypothesized to account for poor outcomes, including impaired self-regulation, disrupted development and altered attachment working models, has yet been rigorously investigated [24].

The timing of adverse events in childhood may be a significant factor in the development of later psychopathology. Adverse events in early childhood are associated with greater pathology in late childhood, whereas those in late childhood are associated with externalizing pathology [24]. Because the transition to adulthood is a crucial developmental watershed, the mental health consequences of adverse childhood experiences are likely to disrupt the establishment of positive roles and relationships that set the course for adult occupational and social attainment [42]. There is a very strong association between childhood adversity and depressive symptoms, antisocial behavior and drug use during the early transition to adulthood [42].

Types of adverse childhood events may change over time, and new types of child maltreatment may become identified. One example of a ‘modern’ form of child abuse is bullying inflicted over the internet, ‘cyberbullying’ [43, 44]. A recent study of more than 200 Finnish adolescents found a prevalence of 4.8% for cyberbullying, 7.4% for cybervictimization and 5.4% for the combination status. Both cyberbullies and their victims were found to be more likely to experience psychosomatic and psychiatric problems [45].

The nature and predictors of resilience among survivors of childhood psychological trauma need more research [46‒49]. A number of researchers are addressing this area. Examples of recent interesting studies include that of mothers and children in London [50], one in the USA with child welfare-identified maltreated children [51] and one with an epidemiological community sample in Dunedin, New Zealand [52]. These studies have specifically examined predictors of positive functioning in early/mid-adulthood by traumatized children. Protective factors included intelligence, female gender, secure attachment to a primary caregiver, living in a stable residence in a non-threatening neighborhood with prosocial parents and peer support as well as sub-traumatic life stressors and a primary partner in adulthood. Risk factors included parents with substance use problems, poor school functioning, living in neighborhoods with high crime rates and low cohesion, and informal social control. Risk and protective factors often had indirect moderating effects as well as direct effects on subsequent adjustment and functioning, and these may differ at different adolescent and adult developmental periods [51].

The pursuit of reliable and valid instruments for the assessment of adverse childhood events began with the pioneering, systematic work of David Finkelhor and Diana

Russell, both in the US [53, 54]. Each of these authors wrote popular books describing their careful studies of prevalence of childhood abuse, using in depth interviews. Each broadened the important questions about childhood from those raised by Kempe and colleagues in their work on battered children [55]. It became clear from those projects that briefer instruments would be necessary for use with random community samples and with linkage studies between childhood adversity and later health outcomes, both psychological and physical, for the research field to advance. A substantial and complex body of empirical work has emerged since then.

Adverse childhood events are most commonly evaluated clinically with a careful, semi-structured assessment interview conducted by an experienced clinician (psychiatrist, psychologist, nurse or allied health professional). However, a wide variety of self- and clinician-rated instruments is also available; some are questionnaires, some are interviews, some are designed for screening and some are designed for definitive appraisal. Several scales have been designed for use in adults for the specific evaluation of childhood maltreatment. These vary considerably in the types of abuse or neglect assessed and the amount of abuse-specific detail they gather. The reader is referred to four previous reviews of this area [56‒59], especially that by Roy and Perry, which we found to be the most comprehensive and useful [57].

The design of instruments that purport to measure adverse childhood experiences warrants some discussion. A number of currently used trauma instruments were developed in research contexts and do not meet current psychometric standards for clinical psychological tests [60, 61]. Although many of these tests are internally consistent, their actual clinical applicability and utility are often unclear. Some trauma impact measures do not have normative data from the general population. In this situation, clinicians cannot evaluate the meaning of a high score. In the case of diagnostic screening instruments, the absence of normative data is less of a problem because the only measurement issue is whether the participant endorses the required level of symptoms or not [62].

With regard to the question of which is better, questionnaire or interview, there is no straightforward answer. Brown advises that the choice should be made on scientific rather than economic grounds [36]. If questionnaires are used, he suggests that the best way forward might be to have a backup interview-based instrument for at least a subsample of the research population. In most research contexts, pragmatic considerations of funding issues and the acceptability of the instrument to the participant, especially the time commitment burden, are critical.

Retrospective assessment has a number of limitations. Some assert that only memories that are linked to self through emotional and motivational significance and form a coherent life narrative are truly autobiographical, in the sense of being replicable [63]. Others have shown that such detailed interviewing which aims to create a coherent story have proved surprisingly stable over a period as long as 10 years [64]. Brown has highlighted how different reports from the Dunedin longitudinal study showed poor agreement between adolescent retrospective reports of family life and measures collected during childhood [65]. He asserts that information about childhood maltreatment gathered without a deep contextualization of such adverse events produces superficial and non-replicable data which can mislead [36, 64]. This particularly applies, in his view, when cohort prediction effects between early experiences and later health parameters are being sought [66]. Other writers are more sanguine about replication and the value of short, retrospective questionnaires; this debate recalls the vicious disagreements between recovered and false memory factions which marked so much of the child abuse research in the 1980s [67]. Fortunately, with better knowledge about memory, a middle ground consensus has been reached. Fergusson et al. [68] have reported considerable instability in repeated reports from their cohort of 18- to 21-year-olds on childhood physical punishment and sexual abuse (test-retest values r = 0.45). Unreliability was explained by abuse victims who often provided false negative reports, rather than non-victims providing false positive accounts. Martin et al. [69] suggest that unstable reporting is a problem with less severe incidents of abuse. Fergusson et al. [68] state that the consequence of this instability is that single report measures may underestimate the true prevalence of abuse. However, the impact on statistical relative risk calculations of psychiatric adjustment is small. The use of a detailed semi-structured, leisurely assessment of context and personal meaning in research is considered optimal by Brown [36]. However, they too note that the impact of unstable reporting across time on statistical models linking child maltreatment and adult psychiatric disorder is minimal [64].

Contextualization and the search for meaning of events have not been explored as much in children as in adults who have experienced trauma, either as adults or as children. This is probably because of the assumption that a child's fundamental needs, particularly for security, will differ little across settings. However, contextualization may be necessary to reflect the individual's culture properly, especially when discovering whether a particular behavior was experienced as abusive. Against this body of reasoning, are the detailed efforts of several groups of researchers, aiming to develop brief, reliable and valid tools which can be robust enough to detect longitudinal pathways [70‒72]. Interestingly, Brown's own group has also been involved in the search for good, brief, self-report measures [73]. The tension between the two approaches is healthy when it spurs the careful evaluation of available tools and a sophisticated discussion about ‘horses for courses’, i.e. which instrument is best for which research question or which clinical presentation.

In addition to all of the above issues, it can be noted that it is not always possible to distinguish between the specific and general effects of trauma; comorbid sources of trauma-related distress can confound interpretation of data. Being victimized is a risk factor for further victimization [74, 75]. Multiple traumatic events experienced by the same individual may be also referred to as ‘event comorbidity’. Multiple and potentially interactive psychological effects of these various traumas result in a complex symptom picture. Emotional or cognitive suppression, denial, dissociation, memory distortion or involvement in activities that numb or distract may produce

avoidance, denial and underreporting. Elliott and Briere [79] described a subsample of children for whom there was direct evidence of sexual abuse (such as unambiguous medical findings, explicit photographs or abuser confession) who nonetheless denied they had been abused and scored lower than non-abused subjects on the Trauma Symptom Checklist for Children. Widom et al. [78] and Williams [77] also independently showed underreporting of childhood sexual and physical abuse by 30-40% of victims later in adulthood.

Overreporting and malingering may also skew the data gathering. Some individuals may consciously or unconsciously magnify their symptoms as a ‘cry for help’ or an attention-getting device [79‒81]. There may also be financial motives for symptom endorsement. Clinical experience suggests that false reports of victimization-related symptomatology are relatively rare in non-forensic clinical settings. However, even when accurately reported, misidentification and distortion of the sequelae of childhood trauma may occur. Such effects may be mislabeled as personality disorders to the extent that they involve interpersonal difficulties, chaotic internal states, tension-reduction behaviors or other affect-avoidance activities [79].

Failure to identify early maltreatment is determined in part by practical and legal difficulties of collecting such information during childhood. This can explain why most research has been done either in adults or on children who have been legally or clinically defined as maltreated. Ideally, it would be good to develop more measures of maltreatment that can be used in childhood itself so that a clear picture of a high-risk constellation can be detected early and effective intervention offered, if needed.

Rigorous searches were conducted using Medline, EMbase, PsycINFO, PsycBOOKS, and EBSCO's CINAHL and Psychology and Behavioral Sciences Collection databases. For Medline, EMbase, PsycINFO and PsycBOOKS, the searches were carried out using a combination of subject headings and keywords (and in the case of PsycINFO, classification codes) integrated into the strategy to describe the three concepts of childhood abuse or neglect, adult psychopathology, and tests/measures. Both OvidSP Advanced and OvidSP Basic search modes were used, and the Find Similar feature was run on the most relevant of the results. In the EBSCO databases, searches were carried out using any available subject headings plus keywords. No limitation on year of publication was set, and it is of interest that the earliest begins in 1983 [82]. A total of 145 articles were identified using the search strategy described above. A number of articles used the Minnesota Multiphasic Personality Inventory as a proxy assessment of childhood abuse without an additional specific abuse instrument being used. When these articles were eliminated from consideration, the number of potentially relevant articles reduced to 123. Abstracts of these articles were subsequently reviewed by the authors for relevance, with the requirement that the reports address tool development and testing, with some description of psychometric properties. This further reduced our list to 38 instruments. Finally, some instruments not captured by our search strategy, but mentioned in the four previous review articles on this area and a few others identified by cross-referencing available articles, were added to the shortlist and resulted in a total of 43 instruments for consideration as listed in table 1 (see also online suppl. appendix). We checked the frequency of citations for each instrument by its name in PsycInfo using their Tests and Measures search limitation function; this search was conducted on 3.3.2011.OSM

Streiner [60] created a valuable checklist of criteria for assessing such instruments. This included four main areas to be considered: the items themselves, reliability, validity and utility. Item criteria of interest include how they were developed and assessed. Reliability indices include internal consistency, test-retest reliability and inter-rater reliability. Validity criteria include face validity, content validity, criterion validity and construct validity. Utility criteria include whether the scale can be completed in a reasonable amount of time, the amount of training required for administration and the ease of scoring of the instrument. Although we initially aimed to evaluate all child abuse measurement instruments using these criteria, it quickly became apparent that the majority of articles did not provide adequate information for us to do so.

Instead, we have developed a shortlist of good quality instruments for clinicians and researchers using the following set of principles. Firstly, we searched for the use of a proper ‘gold standard’ or more properly, a criterion standard, during the development and testing of the instrument. In accord with long standing tradition, we deemed this to be the in-depth sophisticated clinical interview with an empathic and experienced interviewer. As a secondary consideration, we were interested in instruments which were validated against other instruments that had been properly validated against such a gold standard. Secondly, we counted the types of abuse and neglect covered (sexual abuse, physical abuse, emotional and physical neglect, losses, etc.) and the range of items under each domain. Thirdly, we critically evaluated the psychometric properties of each instrument, being particularly interested in good reliability and validity for both the whole instrument and any subscales which were described. Fourthly, we assessed the utility of each instrument, considering the time taken to administer it, the cost of its use, the need for training and/or availability of a manual, its acceptability and social desirability. Finally, we reviewed the interpretive value of each instrument, taking into consideration its comparability, robustness and the number of times it had been used by other researchers. The following instruments proved themselves worthy of further consideration.

The Computer Assisted Maltreatment Inventory is a web-based adult self-report measure of child maltreatment, which includes sexual, physical and psychological abuse, neglect, and exposure to interparental violence. It takes 10-30 min to complete and produces a total score and scores for each abuse type. It is reported to be easy to use with minimal training and time required of administering staff. It was tested with 1,398 undergraduate students using the Childhood Trauma Questionnaire (CTQ), which itself has respectable psychometric properties, as a substitute gold standard [72]. Internal consistency for subscales was reported as good, with a Cronbach’s a coefficient for psychological abuse scales of 0.91 and an a coefficient for neglect scales of 0.88. Test-retest reliability (2-4 weeks between repeat administration) ranged from 0.54 to 0.80 with a mean of 0.70, showing ‘good agreement’. Further discussion of psychological abuse and neglect scales suggests that these subscales are difficult to interpret [122]. PsycInfo citations n = 9.

Comment. This instrument is of great interest given the utility offered by its online nature. It was preferred by participants and considered more confidential than an interview or self-report questionnaire format [85].

The Childhood Experience of Care and Abuse Questionnaire (CECA-Q) [73] is the self-report version of the CECA (see below under clinician-administered instruments), which itself is well validated and was used as the gold standard to test the CECA-Q with 179 London UK women. It covers lack of parental care (neglect and antipathy), physical and sexual abuse. It is in the public domain and no training is needed; time for completion is unclear. There are 16 parental care items, one screen item for physical abuse with four supplementary items and three screen items with eight additional items for sexual abuse. Good internal consistency (antipathy 0.81, neglect 0.80) and test-retest reliability after 5 years (0.51-0.84, all p < 0.0001) were reported. Against the CECA, all correlations were significant (r = 0.48-0.66 with and without women with depression, sensitivity 73% specificity 78%. Criteria validity indices were good against depression scales and the Parental Bonding Instrument. It has also been validated in a clinical population of depressed patients with good results [123]. PsycInfo citations n = 10.

Comment. A first choice for a research or clinical screen with a broad scope, with the additional back up for the interview CECA if more objective and detailed (and expensive) data are required.

These three linked retrospective measures assess CSA (1) Frequency, (2) Type and (3) Characteristics (duration, age and relationship to abuser, victim's age at beginning and end of abuse, and methods of coercion used), respectively [58]. Thirteen types of sexual activity on a 6-point frequency scale were assessed. Two modest-sized samples of women were used for instrument testing - a family practice clinic sample (n = 132) and a community sample (n = 19). Physical and psychosocial symptomatology and depression were chosen to assess criterion validity. The CSA Frequency Scale had internal consistency = 0.90, test-retest reliability = 0.94, criterion-related validity r = 0.36, and depression r = 0.38. The CSA Count Index had good test-retest reliability = 0.92, criterion-related validity against symptomatology r = 0.41, p = 0.01, and depression r = 0.40, p ? 0.01. The CSA Multiple Characteristics Index had internal consistency = N/A, test-retest reliability = 0.94, criterion-related validity for symptomatology r = 0.40, p ? 0.01, and depression r = 0.39, p ? 0.01). External validity was not available for the CSA Frequency Scale, and for both the CSA Count Index and CSA Multiple Characteristics Index, external validity was low. Each of these three distinct approaches to measurement of childhood sexual abuse met most, if not all, of the criteria for satisfactory quality, and no one measure was found to be superior to the others. PsycInfo citations n = 1 for each tool.

Comment. These instruments would benefit by having their construct validity assessed against an acceptable gold standard before the reader can have confidence in their overall utility.

This self-report instrument [110, 124] appears to be the most widely used and has had the most psychometric analysis. The authors created it from the Childhood Trauma Interview (CTI), which they had also developed [110]. Initially, the CTQ was a 70-item instrument assessing abuse and neglect experiences in childhood, rated on a 5-point Likert scale and requiring 10-15 min to complete. The CTQ (twice, on average 3.6 months apart) and the CTI were tested with inpatients with alcohol/drug dependence. Internal consistency for the entire scale was high, 0.95, as it was for the four factors - physical and emotional abuse, emotional neglect, sexual abuse, and physical neglect (0.79-0.94). The CTQ test-retest reliability for the entire scale was 0.88, and for individual factors between 0.80 and 0.83. The CTQ showed good convergent validity with the CTI. The CTQ does not cover trauma characteristics (age at onset and relationship of perpetrator to victim). The authors suggest that ease of administration and relative noninvasiveness make the CTQ a good brief screening tool. Paivio and Cramer [90] added psychometric data using a student, nonclinical sample. Internal consistency ranged from 0.75 to 0.97 (0.75, the lowest being for physical neglect, was considered fair) and test-retest reliabilities ranged from 0.85 to 0.97, which were considered acceptable for the total scale and all five-factor scales (emotional, physical and sexual abuse, and emotional and physical neglect). PsycInfo citations n = 600.

Comment. A valuable choice where it is important to be able to compare results with a range of previously published studies using the same instrument, e.g. in an exploration of cross-cultural differences in prevalence or consequences of a particular type of abuse. It also has a broad scope.

The Childhood Trauma Questionnaire-Short Form (CTQ-SF) [111] is a 28-item (25 clinical items, 3 validity items) self-rated questionnaire that also identifies physical abuse, sexual abuse, emotional abuse and neglect. It was developed through factor analysis from the CTQ. It takes only 5 min to complete and does not require formal training. The CTQ-SF was initially tested in four population samples including that used earlier for the CTQ alcohol or drug disorder inpatients, 296 adolescent psychiatric inpatients and 579 controls), a total of 1,978 people. Results were compared with scores on the CTQ and the therapist's maltreatment ratings (Child Maltreatment Ascertainment Interview) for the adolescent group. The instrument was found to have good internal consistency (Cronbach’s a for emotional abuse = 0.84-0.89, physical abuse = 0.81-0.86, sexual abuse = 0.92-0.95, emotional neglect = 0.88-0.91, and physical neglect = 0.61-0.78). Good correlation was found between subsets of abuse on CTQ SF and therapist maltreatment ascertainment interview ratings in the adolescent group, suggesting criterion validity. Further work with the short-form 28-item CTQ in a racially mixed US community sample (n = 1,007, males and females aged 18-65) generated acceptable internal consistency coefficients for the entire measure (0.91) with physical neglect subscale the lowest (58), and sexual abuse subscale the highest (94) [91]. Its wide use is reflected in the number of translations which have been produced, including German, Dutch, Italian, French, Spanish, Portuguese, Turkish, Norwegian, and Haitian Creole. PsycInfo citations n = 35.

Comment. Robust psychometric data and the extensive use of this instrument make it a worthy candidate for consideration in research projects not requiring contextualization. These features also make it a good choice for clinical situations also although its overall utility has not been discussed in detail.

The Early Trauma Inventory-Self Report (ETI-SR) [94] is another instrument for the assessment of childhood physical, emotional and sexual abuse, and general traumas (e.g. death of a parent), which measures frequency, onset, emotional impact and other variables. It is a 62-item modification of the Early Trauma Inventory (ETI; see interview instruments below) a semi-structured interview; the ETI-SR takes about 30 min to complete. It can be located at: http://userwww.service.emory.edu/~jdbremn/instruments/ETISR-SF.pdf.

The ETI-SR was administered to 288 subjects of mixed diagnoses, including some with no diagnosis. Validity (criterion) was assessed by correlating the ETI-SR score with PTSD symptom severity measured with the Clinician Administered PTSD Scale (CAPS).Individual domains showed good internal consistency (a = 0.78-0.90). In terms of validity, most items correlated >0.2 with the CAPS. A subset of 27 items was selected for a shorter form. The authors concluded that the ETI-SR can be used for clinical and research purposes. PsycInfo citations n = 58.

Comment. This may be a good choice for a screen for a wide range of adverse experiences including abuses, given its broad design. We did not find construct validity reports for it.

The Familial Experiences Interview (FEI) and Familial Experiences Questionnaire (FEQ) [95, 115, 125] are closely related retrospective measures assessing the frequency, severity and duration of a wide range of adverse childhood experiences including physical and sexual abuse, neglect, losses, frequent moves, school difficulties, parental unemployment and others. The questionnaire was adapted later [125] from the interview [115]. Interviewers were clinical psychologists and were given 2 months extensive training with manuals to study. Completion time for the FEI and FEQ are not clear. Preliminary internal consistency FEI data were good [126]. Interview inter-rater reliabilities ranged from 0.47 to 1.0 (only 6/150 <0.70). Durrett et al. [95] compared the FEI with the FEQ in undergraduate university students (n = 421). Correlation between the FEQ and FEI was high for sexual abuse by any perpetrator (0.91) and less strong (0.59) for physical abuse by a parent. Internal consistencies were good for severity and duration of sexual abuse (0.58-0.75), strong for severity of physical abuse (0.81), and low for agreement on duration of physical abuse (0.32). FEQ test-retest reliability for sexual abuse was 0.52 and for physical abuse was 0.48. The 2-year test-retest reliability of the assessments was only fair for both the FEI and the FEQ. PsycInfo citations FEI n = 2, FEQ n = 1.

Comment. These instruments have not been properly validated and require lengthy training. However, they have a wide scope and may be useful where validity considerations are not prime, e.g. where cross sample comparison is not relevant.

Discussing the problem of ‘neglect of neglect’, Harrington et al. [96] reported on reliability and validity of their Neglect Scale, a self-rated, retrospective measure previously developed in 1995 by Straus and colleagues, using a low income sample of 151 maternal caregivers from families at risk of neglect. A 40-item version was assessed for internal consistency and factor analyzed. It was self-administered using a computer format; the time for completion was not stated. Child abuse researchers then coded these items into one of the four subscales developed by Straus's team: Emotional, Cognitive, Supervisory, and Physical. The researchers agreed on 24/40; the other 16 items were dropped. Psychometrics: internal consistency for 40 items was very high (a = 0.96) and moderate for subscales (Emotional 0.85, Physical 0.82, Cognitive 0.78, Supervisory 0.81). Contrary to their implication in the abstract, the authors did not assess validity against any gold standard. Rather, they used the factor analyses to eliminate items which loaded on more than one factor, thus reducing the Neglect Scale further down to 18 items (Emotional 5, Cognitive 4, Supervisory 3, Physical 6). They commented that further research will be necessary to determine the criterion and construct validities particularly against criteria used by Child Protection Agencies. They recommended use of the full 40-item version; if a shorter one is needed, they recommended using only the 11 items fitting both Straus group's and their own data. PsycInfo citations n = 28.

Comment. This instrument awaits validation, but may be worth considering for studies of neglect without other forms of adverse events.

Engels and Moisan [98] were interested in developing a self-report scale to tap psychological maltreatment, which they suggest is more prevalent than other forms of abuse/neglect. The Psychological Maltreatment Inventory is a 25-item scale which was psychometrically assessed on adult outpatients (n = 118). No training is described; time for completion was not noted. Three highly intercorrelated factors emerged, Emotional neglect, Hostile rejection, and Isolation. Internal consistency was high (0.94). Test-retest coefficients after 18-24 months (with low response) were good (overall 0.81, Neglect 0.78, Hostile rejection 0.78 and Isolation 0.75). They criterion-validated the Psychological Maltreatment Inventory against two symptoms scales, a self-efficacy scale and a diagnosis of personality disorder using DSM-III-R, and used a modification of the Adult Parental Acceptance and Rejection Questionnaire (APARQ) of Rohner and Rohner to test construct validity [127]. The APARQ has good psychometric properties. PsycInfo citations n = 7.

Comment. Another measure of parental psychological maltreatment with indirectly validated qualities only.

This questionnaire was developed in Binghamton, New York State as a brief screening device to aid taking a child sexual abuse history. In its final form, the Sexual Abuse Questionnaire [101] has 45 true-false items and takes about 5 min to complete. No training is required, the scoring is simple; certain items need to be reversed, and then all items are summed. The authors included items which they call ‘non face valid’ designed to elicit symptoms often associated with childhood trauma. Its psychometric parameters were tested in two samples of university undergraduate students mostly Caucasian, n = 533, 58% females, 10.3% of whom reported CSA, with retesting one month later. Participants also completed the Trauma Symptom Checklist 40 (TCS-40) and a PTSD subscale, to assess criterion validity. Good test-retest reliability, internal consistency, and convergent and discriminative validity (TSC-40 r = 0.71, and the PTSD scale r = 0.70) were found. PsycInfo citations n = 27.

Comment. The authors called the Sexual Abuse Questionnaire a time- and cost-efficient method, commenting that its performance was unknown in other populations, e.g. non-university samples. Given its wide use, it is worth considering for studies of sexual abuse.

No formal name was given by the authors [71], although another instrument with very similar provenance, called the Child Maltreatment History Self-Report [128] is regarded as identical by Roy and Perry [57]. Both contain sexual abuse items adapted from a 1984 Canadian government survey of sexual abuse of children [129] and differ only in minor wording and frequency counts. Both add physical abuse items, either from Briere's work [71] or from the Conflict Tactics Scale of Straus [128]. Only the paper by Leserman et al. [71] gives psychometric properties; these authors do an excellent job of assessing the instrument using classical psychometric approaches. The gold standard used was an in-depth, structured abuse interview conducted by one trained female psychologist, with uncertain responses discussed with senior team researchers. Research participants (n = 139) were outpatients in a gastroenterology clinic. The questionnaire, published fully in the 1995 article has six sexual (yes/no) and five physical abuse items (never, seldom, occasionally, often) answered each for child and adult years. Sexual abuse at first administration: test-retest reliability after a median of 2 months (range 0-16 months) was 0.63, with 81% overall agreement. Other parameters reported were: sensitivity 71%, specificity 91%, positive predictive value 90%, negative predictive value 74%. Test-retest reliability was 0.54. Also reported were: sensitivity 68%, specificity 74%, positive predictive value 73%, negative predictive value 68%. The authors also give psychometric values at the readministration, which in general were a little better. They provide an in-depth discussion of possible causes of discrepancies. They conclude that the sexual abuse questionnaire has acceptable test-retest reliability and criterion validity, as does the physical abuse questionnaire, although to a lesser extent. PsycInfo citations n = 6.

Comment. This instrument has been directly content validated against the best gold standard, and will be a first choice for a screen for studies addressing sexual and physical abuse alone. This questionnaire has been further developed into a structured interview, the Structured Sexual and Physical Abuse Interview (SSPAQ), see below.

The Childhood Experiences of Care and Abuse (CECA) [130, 131] is a clinician-rated instrument that detects physical abuse, sexual abuse, neglect and antipathy occurring before the age of 17 years. Some items are scored on a Likert scale and others are reported as direct answers. It was initially administered to a sample of 179 women (18-51 years). Participants were also rated on the Parental Bonding Instrument and the Present State Examination. Satisfactory internal consistency was found for antipathy (0.81) and neglect (0.80), but not for other subscales. Satisfactory test-retest reliability was found for most subscales. A type of validation was undertaken with 87 pairs of sisters, to see what concordance existed between individually completed scales assessing parental neglect, physical abuse and sexual abuse. The authors viewed the mean correlation for the three experiences of 0.60 as satisfactory.

The interview takes an average of 1.5 h or so to administer for a medium risk case and around three times as long to transcribe and score. The full scale is provided with the original article, and a 220-page manual with definitions and examples is available. In accord with the Bedford College's group philosophy of interview use, the CECA uses a conversational style, encouraging detail; they consider it important to move beyond standard questions to develop a coherent narrative [64]. The length of the interview is variable depending on the complexity of the childhood circumstances. Full training takes 2 days and currently costs GBP 350. PsycInfo citations n = 10.

Comment. The most detailed and contextualizing approach, with the strong research provenance of its developers. Although the CECA is too time consuming for most projects, it can be considered for use in a subsample, to set the criterion evaluation against which other instruments can be validated. It will have a role where the personal meaning of the abuse experiences is salient to the study, rather than the simpler epidemiological questions of abuse prevalence.

This instrument [108, 109] assesses physical, sexual and verbal abuse, witnessing violence, emotional neglect, separation from and loss of caregiver. A short form (11 items, with multiple sub-questions) is freely available online at http://www.johnbriere.com/ Unpublishedtests.htm and seems to have replaced the long version. It takes 5-10 min to complete, with no training required. Both the long and short form are written as an interview, but can also be administered as a self-report questionnaire [109]. No validity data are available for either version; internal consistency of the psychological abuse sub-scale is good for both, e.g. for the long version a = 0.87 [132]. PsycInfo citations n = 12. Comment. This has a wide scope and may be useful where validity considerations are not prime, e.g. where cross-sample comparison is not relevant. Unlike the FEI and FEQ, the absence of the need for training may make it attractive to researchers.

The ETI [114] is a 56-item clinician-rated questionnaire that identifies sexual abuse, physical abuse, emotional abuse and general trauma (such as witnessing family violence), but not neglect. It asks about the types and frequency of abuse, age at which the abuse started and stopped), the perpetrator and the impact of the event(s). It takes about 45 min to administer, and the authors wish an experienced clinician to be involved, either to conduct the interview themselves or to supervise directly the interviewers. Most answers are rated on a 7-point Likert scale. An ETI childhood trauma severity index has also been developed to indicate the ‘total burden’ of abuse over a person's childhood. In addition, the authors have developed a self-report version that takes 45 min to complete (see above). The ETI was validated in a population of 137 adults with mixed diagnoses. The ETI showed good inter-rater reliability (0.99 overall, 0.97 for physical abuse, 0.97 for emotional abuse, 0.99 for sexual abuse, and 0.94 for general trauma). Test-retest reliability was also very good, 0.91 (0.97 for physical abuse, 0.98 for emotional abuse, 0.99 for sexual abuse and 0.51 for general trauma). Internal consistency was high, 0.95 (0.86 for physical abuse, 0.92 for emotional abuse, 0.92 for sexual abuse, and 0.74 for general trauma). Convergent validity was found to be satisfactory with a correlation of 0.63 between the ETI and a CLTE. ETI scores were higher in PTSD patients than healthy controls. PsycInfo citations n = 58.

Comment. Another useful instrument, with some sound psychometric results. However, there is not a full construct validation available yet.

This 38 item seeks information about sexual abuse and physical abuse (with ‘intent to kill or seriously injure’) with details on duration, number of incidents, identity of perpetrator, separately for before 13 years and subsequently. The Chapel Hill group used data from this interview to develop an Abuse Severity Measure, summing a score for each of: invasiveness of sexual abuse, severity of physical injury, total number of life threatening, physical abuse incidents [118] .

Comparing the instruments detailed above, our recommendations are as follows. The CECA (full interview) is the most detailed interview developed and has broad scope. It is likely to be too time consuming for most research projects. However, as it is closely yoked with its short questionnaire version, the CECA-Q, the CECA could be the criterion standard, against which the questionnaire is validated. One can envisage the CECA being given to a subset of participants in a study to validate the CECA-Q, which would be given to all participants. The use of either clinically is not yet clear. The questionnaire developed by Leserman and colleagues (called the Sexual and Physical Abuse History Questionnaire by Roy and Perry) has excellent psychometric properties and would be the ideal questionnaire where the focus is on childhood physical and sexual abuse only.

As the CTQ has received the most attention and publications, it will be a good choice where comparison with many other studies is desirable, and multiple abuse/neglect processes are being studied. It has the best pedigree for cross cultural and other language research to date, although this may change. The Computer Assisted Maltreatment Inventory can be chosen where the participants (and researchers) are internet savvy, and good validity is important. This relatively new instrument is one to watch to see if its uptake confirms its early promise of high utility and good psychometric properties. We consider that the other instruments individually described in detail all may have a place depending on the research questions, the population being studied and time.

It is clear from our review, and those of others, that there have been major advances in the past three decades, with steady improvement in the range of instruments available and data supporting their utility. Researchers have continued to improve existing instruments as well as creating new ones. It is encouraging to see the recent appearance of tools to assess neglect and emotional abuse, both areas of concern previously thought unmeasurable. Further work on developing clear definitions of abuse is needed. We recommend that authors of future scales produce a complete description of their instruments according to the guidelines suggested by Streiner in order that a comparison can be made with existing instruments. Most workers concur with the general principle in abuse prevalence research that a precise behavioral description is best, rather than leaving the participant to make a subjective judgment about whether their own experience was ‘abusive’ or ‘traumatic’. This guiding principle should continue. It will be necessary to do more work on cross cultural notions of abuse and the accurate translation of these instruments into other languages. Our review has only dealt with English language publications. There is also an obvious need to develop a nuanced understanding of the use of the main instruments with special groups; men, visible minorities, elderly and the young immediately come to mind. As a generalization, validation is specific to the population in which it was carried out, and cannot be presumed to apply to other sociodemographic groups. The extension of scales for use with children is urgently needed, assuming that earlier detection will lead to more timely and effective interventions. We refer the authors of such scales to guidelines by Hamby and Finkelhor [56], which set out recommendations for the sensitive interviewing of children. A few instruments have already been used with adolescents, e.g. CTQ-SF, but more studies are also needed for this age group.

We would contend that a thorough psychosomatic assessment needs to address child abuse and neglect in order to provide a deep understanding of the risk factors for anxiety and hyperarousal states which may lead to adverse outcomes. One-off events may set in train alterations within the family which have adverse effects on the child's development, and so they too are important to record. We would agree with Ford that prospective studies with high- and low-risk samples are needed to assess at repeated intervals, beginning in early childhood and continuing through adulthood using psychometrically sound measures of trauma exposure and psychobiological outcomes. Temporal trajectories of risk and resilience must be mapped in order to move beyond static formulations of the relationships among traumatic victimization, psychobiological mediators and moderators and adult outcomes [24]. Although the link between childhood trauma and adult psychopathology is well established in many demographic groups, the elucidation of the mechanisms through which this relationship develops needs to be understood further. A full explanation will include the moderating effects of temperament or other biological and environmental variables, the nature of resilience and vulnerability. Only then will we be in a sound position to develop preventative interventions.