How to measuring child sexual abuse

How to measuring child sexual abuse Once the criteria for the definition of child sexual abuse are determined, the next
issue is how best to determine whether the respondent has been sexually abused. The
issue concerning accurate disclosure can probably be subsumed within two
categories—methodological and ideological concerns. From an ideological standpoint,
the issue of accurate recall of abuse is heated and controversial. Because the focus of
this chapter is methodology, the ideological nature of the debate is not addressed.
Methodologically, the issue is how best to garner accurate disclosures of abuse
histories. Several methodological factors must be considered, including (a) the type
of measurement used (e.g., interview format, self-administered questionnaire, or telephone survey), (b) the number of screen questions asked, (c) how behaviorally
specific these questions are, and (d) characteristics of the interviewer. All may have
a considerable impact on the accuracy of the disclosed abuse rate for that population.
Needless to say, the accuracy of the disclosed abuse rate is often fundamental to the
ability to infer findings from the study.
Mode of administration: Until the 1980s, self-administered questionnaires were
assumed to be the mode of choice when asking questions about sensitive material.
Yet in 1986, Peters, Wyatt, and Finkelhor found that studies employing selfadministered
questionnaires had lower prevalence rates of child sexual abuse than
studies employing face-to-face interviews.
Two multivariate analyses of the effect of methodology of prevalence rates of
child sexual abuse have now been conducted (Bolen & Scannapieco, 1999; Gorey &
Leslie, 1997). In both, after controlling for other factors, no significant relationship
between the mode of administration and prevalence was found. An Australian study
(Martin, Anderson, Romans, Mullen, & O’Shea, 1993) that employed both a postal
survey and a face-to-face interview format, however, found that the relationship
between type of format and rate of disclosure was more complicated. While 25% of
female respondents reported a history of contact sexual abuse prior to the age of 16,
10% of these cases were not reported during the face-to-face interview. Furthermore,
12% of women who reported no abuse on the postal questionnaire disclosed abuse
during the interview. This study suggests that both methods have limitations and that
multiple methods of assessment may be needed.
The mode of administering the measure may also be important in clinical
studies. In reviewing studies of inpatient adult psychiatric or emergency room
patients in which prevalence was noted, chart review was associated with the lowest
prevalence rates of reported abuse. Prevalence ranged from 2% to 10% of all patients
whose charts were reviewed (Briere & Zaidi, 1989; Carmen, Rieker, & Mills, 1984;
Jacobsen, Koehler, & Jones-Brown, 1987). When the respondents were later asked
directly about a past history of abuse, the prevalence of abuse reported in one of
these studies went up to 70% (Briere & Zaidi, 1989). In studies using a questionnaire
format, however, between 19% (Metcalfe, Oppenheimer, Dignon, & Palmer, 1990)
and 24% (Goff, Brotman, Kindlon, Waites, & Amico, 1991) of all psychiatric
inpatients, and 44% (Bryer, Nelson, Miller, & Krol, 1987) of female psychiatric
inpatients reported a history of child sexual abuse. In other studies using face-to-face
interviews, 46% (Beck & van der Kolk, 1987), 50% (Goodwin, Attias, McCarty,
Chandler, & Romanik, 1988), and 51% (Craine et al., 1988) of patients reported
histories of child sexual abuse. Finally, one study compared the disclosure of sexual
abuse by adult female psychiatric patients during regular intake procedures to
disclosure on a self-administered questionnaire (Dill, Chu, Grob, & Eisen, 1991).
While 35% of patients disclosed at intake, 52% disclosed on the questionnaire;
overall, 58% disclosed a history of child sexual abuse. These studies suggest that the
method of eliciting disclosure affects the willingness of individuals to disclose a
previous history of abuse.
Type of questions: Another crucial methodological characteristic is the type of
questions used to elicit abuse disclosures (Peters et al., 1986). Studies typically use
one or more screen questions to ask about the possibility of abuse. Screen questions
can be either broad funnel questions or inverted funnel questions. In the former, if respondents answer “yes” to a broad question, it is followed by more restrictive
questions. In the latter, a number of questions are asked, each pertaining to a
narrowly delimited type of child sexual abuse. By clarifying the nature of the
incident being solicited, this behaviorally specific format is thought to trigger
memories that facilitate recollection of abuse incidents. This latter type appears to be
the more effective method both in studies of rape (Koss, 1993) and child sexual
abuse (Martin et al., 1993; Wyatt & Peters, 1986).
Number of questions: The number of screen questions potentially accounts for
a great deal of the variation in reported abuse rates across studies. In an early review
of community prevalence studies, Peters, Wyatt, and Finkelhor (1986) found that
those using only one screen question had prevalence rates ranging from 6% to 22%.
Prevalence rates in studies using two to four screen questions ranged from 11% to
34%, and studies using more than four questions had a range of 54% to 62%
(including noncontact abuse). Peters et al. concluded that the number of screen
questions was an important predictor of prevalence rate. Indeed, in a multivariate
analysis of community, state, and national random prevalence studies done in North
America (Bolen & Scannapieco, 1999), the number of screen questions was the
strongest predictor of prevalence.
In another study, Williams, Siegel, and Jackson Pomeroy (2000) asked a series
of 19 screen questions to a group of 136 women who, before the age of 13, had a
documented case of child sexual abuse. This study reviewed the capability of the
screen questions to elicit the respondents’ histories of child sexual abuse. Although
80 percent of disclosing women were identified after four general gate questions,
only after eight questions did 90 percent of the disclosing women report an initial
incident of abuse. Fourteen questions were required to elicit all disclosures that were
forthcoming in the interview. Even after 14 questions, however, more than a third of
the index incidents were never disclosed and 12% of the women never disclosed any
abuse incidents. These findings corroborate the findings by Bolen and Scannapieco
(1999) that a large number of screen questions are necessary to elicit disclosures of
child sexual abuse and that underdisclosure remains a significant issue even with the
use of multiple screen questions.
Peters, Wyatt, and Finkelhor (1986) discuss factors involved in why multiple
screen questions of an inverted funnel type (i.e., behaviorally specific format) coincide
with higher prevalence rates. First, the way a survivor remembers the actual abuse
might not match the manner in which the question is asked. Consequently, more
questions are more likely to trigger a specific memory of abuse. Second, multiple
screen questions may work better because they provide a longer period of time
during which disclosure occurs, a factor that seems particularly important when
dealing with sensitive material.
Characteristics of the interviewer and respondent: A final issue is the
characteristics of interviewers and respondents. In their early review of methodologies
of prevalence studies, Peters, Wyatt, and Finkelhor (1986) found that attention to the
training of the interviewers might be an important methodological feature. They state:
It is probably not coincidence that in both of the higher prevalence sexual abuse [faceto-
face interviewing] studies (Russell, 1983, Wyatt, 1985), interviewers were specially
selectedfor the study and received extensive training designed to sensitize them to the issues involved in asking about sexual abuse. If [face-to-face interviewing] allows for
the possibility of better reporting, it may be because of the possibility of using wellselected
and trained interviewers to enhance candor. (p. 40)
Finally, matching the race or ethnicity of the interviewer with that of the
respondent may be an important feature. In a secondary analysis (Bolen, 1998b) of
Russell’s (1983) community prevalence study, an interaction effect between race of
the respondent and level of comfort with the interview was noted. Women of Asian
descent who were uncomfortable with the interview disclosed far less abuse
proportionally than those who felt more comfortable with the interview. This finding
suggests that certain cohorts, especially if they do not feel comfortable with the
interview, may be reticent to disclose actual abuse. 1