Thombs BD, Bernstein DP, Ziegelstein RC, Scher CD, Forde DR, Walker EA, Stein MB. An Evaluation of Screening Questions for Childhood Abuse in 2 Community SamplesImplications for Clinical Practice. Arch Intern Med. 2006;166(18):2020-2026. doi:10.1001/archinte.166.18.2020
Copyright 2006 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2006
A number of practice guidelines and recommendations call for the assessment of childhood abuse in adult medical patients, but none specifies how best to do this. The objective of this study was to use evidence from 2 community-based population samples to evaluate abuse-screening questions that are often asked in medical clinics and to identify a small set of questions to improve screening practices.
The Childhood Trauma Questionnaire–Short Form (CTQ-SF) was administered in 2 randomized telephone interview surveys with adults aged 18 to 65 years.
A total of 880 (2003 survey) and 998 (1997 survey) respondents completed the CTQ-SF in the 2 surveys. In both surveys, the rates of physical (16% and 15%), emotional (31% and 29%), and sexual (10% and 9%) abuse elicited using 3 behaviorally descriptive items in each abuse category were approximately twice the rates elicited using the explicit labeling terms physically abused (8% and 8%), emotionally abused (15% and 13%), or sexually abused (5% and 5%) (P<.001 for each). Inquiries explicitly using the labeling term abuse successfully identified a low percentage of respondents who reported behaviorally described abusive experiences for each type of abuse (34%-51%). In addition, after adjustment for the number and frequency of abusive experiences in both surveys, women were more likely than men to label themselves as explicitly abused for any abuse (odds ratio [OR], 1.7; P = .11 and OR, 2.8; P<.01), physical abuse (OR, 2.1; P = .14 and OR, 2.9; P<.01), emotional abuse (OR, 2.7; P<.01 and OR, 3.3; P<.01), and sexual abuse (OR, 3.5; P = .08 and OR, 1.5; P = .55).
Inquiries about childhood abuse that use broad labeling questions identify a substantially smaller number of patients than behaviorally specific questions and may be less effective in initial screening for a history of abuse.
A history of physical or sexual abuse in childhood is present in 20% to 50% of patients in primary care settings.1 Adult survivors of childhood abuse vary widely in their adjustment, and no single disorder or pattern of symptoms characterizes survivors of abuse.2 Nonetheless, an extensive body of literature documents a relationship between childhood abuse and adult psychiatric disorder,3,4 including posttraumatic stress disorder,5 depression,6 anxiety,4,5 eating disorders,7 substance abuse,4,5 and personality disorders.8 Childhood abuse also predicts medical problems in adults, including poorer overall health, even after controlling for psychiatric disorders.9 High rates of childhood physical and sexual abuse are found in patients with gastrointestinal disorders,10 headache,11 fibromyalgia,12 chronic pelvic pain,13 other chronic pain conditions,14 and chronic fatigue syndrome.12
The ability to screen for abuse history is clinically important. Although many medical patients with a history of childhood abuse do not improve with standard treatment of their medical disease, they may benefit from the integration of standard medical treatments and psychological treatments, such as psychotherapy or psychotropic medication, once their trauma history is recognized.15- 17 Practice guidelines and recommendations for a number of psychiatric and nonpsychiatric medical conditions call for an assessment of abuse history.10,18- 20 Recommendations, however, are not specific about the conditions under which screening is appropriate or the questions that most accurately elicit abuse history. In practice, inquiry about abuse is not part of routine care, even when physicians believe that it may be relevant to patient management.21 Numerous observer-rated and self-report instruments have been developed to assess childhood trauma in adults, but they tend to be too time-consuming to function as a brief screen.22 Thus, when screening does take place in medical clinics, it often involves embedding a few broad screening questions such as “Have you ever been physically, sexually, or emotionally abused?” into an initial intake form.
Evidence from longitudinal studies of adults with a documented history of childhood abuse demonstrates that retrospective reporting is prone to high false-negative rates.23 This problem may be exacerbated when the history is elicited using broad labeling terms such as abuse rather than with questions about behaviorally specific experiences.24,25 In addition, while both women and men underreport abusive experiences, men may be even less likely than women to explicitly label themselves as survivors of abuse.26,27
We conducted an exploratory study using questions about childhood abuse from the Childhood Trauma Questionnaire–Short Form (CTQ-SF)28 in a large community-based sample to assess the adequacy of broad screening questions about abuse that are often asked in medical clinics. Specifically, we sought to (1) evaluate the degree to which broad questions using the actual term abuse might result in the underidentification of patients who report having experienced one or more behaviorally defined abusive experiences in their childhood, (2) assess the degree to which this problem might be exacerbated among male patients, and (3) identify a small set of questions to improve screening for childhood abuse.
The CTQ-SF was administered as part of 2 large population surveys, the 2003 Mid-South Social Survey Research Program and the 1997 Memphis Area Study. Households were randomly sampled from telephone numbers found in the Memphis and surrounding Shelby County telephone directory. Eligible respondents were English-speaking residents aged 18 to 75 years for the 2003 survey and 18 to 65 years for the 1997 survey. For uniformity across samples, only respondents aged 18 to 65 years from the 2003 survey were included in this study. Households were randomly designated as male or female prior to telephone contact. If the person answering the telephone was of the specified sex, only that person could be interviewed. If the person was not of the specified sex, the interviewer asked the person to choose a household member of the specified sex. If a person of the specified sex did not live in the household, the person answering the telephone became the selected respondent. Both the 2003 and the 1997 studies were approved by the University of Memphis internal review board.
All interviews were conducted over the telephone between March 27 and May 23, 2003, for the 2003 study and between February 12 and April 25, 1997, for 1997 study. Each interviewer received a training manual and at least 6 hours of training as well as 3 to 4 practice interviews with supervision. Computer-assisted telephone interviewing files were downloaded automatically at the completion of each interview, and files were checked for reliability in data entry. Interviewers attempted to reach each selected household at least 10 to 12 times before listing it as a noncontact.
All data in this study were generated by responses to questions from the CTQ-SF, a 25-item self-report questionnaire that assesses both abuse and neglect and has 3 abuse scales (physical, emotional, and sexual). Each abuse scale includes 5 items, 1 of which inquires broadly as to whether the respondent was physically, emotionally, or sexually abused. At least 3 items on each abuse scale inquire about specific, behaviorally defined examples of childhood abuse (eg, for physical abuse, “People in my family hit me so hard that it left me with bruises or marks”). The broad items explicitly using the term abuse and the items that inquire about specific examples of abusive experiences are dispersed throughout the 25-item questionnaire. The item response options of the CTQ-SF define the frequency of maltreatment experiences (never, rarely, sometimes, often, or very often). Bernstein et al29 have reported good internal consistency of the CTQ-SF for each of the abuse scales across 4 heterogeneous samples (physical abuse, 0.83-0.86; emotional abuse, 0.84-0.89; and sexual abuse, 0.92-0.95).
Respondents were classified on 2 dimensions: (1) whether they explicitly labeled themselves as having been abused and (2) whether they indicated having experienced any of the behaviorally defined abusive events described in the CTQ-SF. Each type of abuse (physical, emotional, and sexual) was evaluated separately. For each type of abuse, respondents were considered to have endorsed the label of having been abused if they responded anything other than “never” to the single item of the appropriate CTQ-SF scale that explicitly used the term abused. Respondents were considered to have experienced abusive childhood treatment if they responded anything other than “never” on 1 or more of 3 items reflecting behaviorally specific abusive events from the respective CTQ-SF abuse scale. One item on each of the 5-item CTQ-SF abuse scales was not included in the analysis of reporting of abusive event history because it was subjective and not behaviorally defined (sexual abuse, “believe I was molested”) or because it was deemed to overlap with relatively common experiences and was endorsed by a very high rate of respondents (physical abuse, “punished with hard object,” endorsed by 61% in 2003 and 52% in 1997; emotional abuse, “family members said hurtful or insulting things,” endorsed by 30% in 2003 and 29% in 1997). Items from each abuse scale that were included in the analysis are listed in Table 1.
Demographic variables were reported, and tests for differences across sex were conducted using χ2 tests for categorical variables and 2-tailed t tests for continuous variables. To assess the degree to which screening with the label “abuse” might underidentify respondents who reported abusive experiences, the number of respondents in each survey who endorsed at least 1 abusive event and who endorsed having been explicitly abused was tabulated for each abuse category. These results were then used to calculate the rate at which queries explicitly including the term abuse identified respondents who endorsed at least 1 abusive event. To assess the extent to which underidentification might be exacerbated for male patients, the odds of endorsing having been explicitly abused were calculated for women who reported abusive experiences compared with men who reported abusive experiences for each type of abuse and for any abuse. Adjusted odds ratios (ORs) were also computed using logistic regression to control for scores on the 3 behaviorally defined items from each scale as well as for marital status and race or ethnicity, which differed across sex. All analyses were conducted using SPSS version 13.0 (SPSS Inc, Chicago, Ill).
In the 2003 and 1997 surveys, CTQ-SF interviews were completed in 880 of 1266 eligible households (69.5% response rate) and 998 of 1303 eligible households (76.6%), respectively. Of 775 respondents aged 18 to 65 years in the 2003 sample, 756 provided data for all abuse items, 771 for all physical abuse items, 768 for all emotional abuse items, and 765 for all sexual abuse items. Of 998 respondents in the 1997 sample, 982 respondents provided data for all abuse items, 994 for all physical abuse items, 990 for all emotional abuse items, and 992 for all sexual abuse items. As summarized in Table 2, age distribution and education levels were similar for women and men in both samples. In the 2003 sample, men were significantly more likely to be European American and never married. There were similar, albeit nonsignificant, trends in the 1997 sample.
The endorsement rates of each abuse-related item of the CTQ-SF are listed in Table 3 for both samples. The rates of report of abusive events using the 3 behaviorally specific items were approximately double the rates of endorsement of having been explicitly abused for both the 2003 and 1997 samples for physical (16% vs 8% and 15% vs 8% for 2003 and 1997, respectively), emotional (31% vs 15% and 29% vs 13%), and sexual abuse (10% vs 5% and 9% vs 5%) (P<.001 for all comparisons). As detailed in Table 4, a very low number of respondents across samples and scales indicated explicitly that they had been abused without endorsing any specific behaviorally defined abusive events (<1%-2% for physical and sexual abuse; 3%-4% for emotional abuse). For the 2003 sample, the single item asking explicitly about abuse detected a low percentage of respondents who reported specific types of physical (39%), emotional (36%), and sexual abuse (44%). Results were similar for the 1997 sample.
Among respondents who reported at least 1 behaviorally defined abusive event, women in both samples were more likely than men to report having been explicitly abused across all abuse categories (OR, >1), even after adjustment for levels of behaviorally defined abusive events (P<.01). Adjusted ORs were not significant for all categories of abuse, but this may have been a function of statistical power. For the 2003 sample, adjusted ORs were significant for emotional abuse (2.7; 95% confidence interval [CI] 1.3-5.6) (P≤.01) and showed nonsignificant trends in the other categories: any abuse (1.7; 95% CI, 0.9-3.3) (P = .11), physical abuse (2.1; 95% CI, 0.8-5.6) (P = .14), and sexual abuse (3.5; 95% CI, 0.9-14.4) (P = .08). For the 1997 sample, adjusted ORs were large and significant (P<.01) for any (2.8; 95% CI, 1.5-5.1), physical (2.9; 95% CI, 1.3-6.7), and emotional abuse (3.3; 95% CI, 1.6-6.5) but nonsignificant for sexual abuse (1.5; 95% CI, 0.4-6.2) (Table 5).
For each type of abuse, we identified a single item and a pair of items that were most likely to successfully identify respondents who endorsed at least 1 behaviorally defined abusive experience. Using 2003 and 1997 data, respectively, for physical abuse, the single item “People in my family hit me so hard that it left me with bruises or marks” identified 80% (102/127) and 88% (135/154) of respondents alone and 92% (117/127) and 92% (141/154) in combination with the item “I got hit or beaten so badly that it was noticed by someone like a teacher, neighbor, or doctor.” For emotional abuse, the item “People in my family called me things like ‘stupid,’ ‘lazy,’ or ‘ugly’” identified 80% (187/235) and 83% (243/292) of respondents who reported emotionally abusive experiences alone and 95% (224/235) and 89% (261/292) in combination with the item “I felt that someone in my family hated me.” For sexual abuse, the item “Someone tried to touch me in a sexual way, or tried to make me touch them” alone correctly classified 86% (68/79) and 98% (89/91) of respondents who endorsed sexually abusive experiences and 94% (74/79) and 99% (90/91) in combination with the item, “Someone threatened to hurt me or tell lies about me unless I did something sexual with them.”
This is the first study, to our knowledge, to use a large community sample to investigate the effect of question selection on the reported rate of childhood abuse among adults as well as to demonstrate sex-related response biases when broad labeling terms such as abuse are used in the questions. In addition, this study identified a small number of questions that may be useful in screening adults for childhood abuse in medical settings.
Based on 2 large population samples, the rate of explicit report of abuse was approximately half the rate obtained from asking behaviorally defined questions. Only a small percentage of respondents who reported behaviorally defined abusive events explicitly reported having been abused. In addition, we found that women were more likely than men to report explicit abuse, after controlling for levels of reported behaviorally specific abusive events. Thus, screening questions that use broad terms such as abuse yielded comparatively low response rates in all cases and particularly when respondents were male. Broad labeling questions about abuse require that the patient define the meaning of the term abuse behaviorally and then determine if her or his own experiences fit that definition. Low rates of endorsement of broad labeling questions may occur because (1) some survivors of what clinicians would clearly label abuse do not label their own experiences in that way,30 or (2) some survivors desire to avoid the stigma associated with the term abuse.24
Many patients with abuse history go from one treatment provider to another and receive various medical diagnoses, yet their underlying psychological condition is never recognized. Accurate identification of these patients with appropriate triage to mental health services could potentially reduce suffering for these patients while also decreasing their use of expensive medical resources. One strategy for doing this would be to screen for abuse in the initial medical history.21 This has the advantages of normalizing the screening process and of allowing the clinician to intervene, if necessary, at an early point in the course of treatment—prior to the development of treatment roadblocks or patient dropout. A disadvantage of this strategy is that patients are unlikely to have developed a relationship of trust at this point. Thus, an alternative is to wait to inquire about abuse until clinical data are suggestive—for instance, when symptoms are refractory and disabling or when patients have difficulties with gynecological examinations or other medical procedures.10
In either case, the clinician should provide the patient a rationale for asking about abuse history. Clinicians can explain to patients, for instance, that things that have happened in the past sometimes result in ongoing distress that impacts how patients cope with medical illness or that exacerbate medical symptoms or the experience of these symptoms. A reasonable strategy would be to use the items identified in this study, beginning with the most sensitive single item followed by the second item in the pair, if necessary. For instance, to screen for physical abuse, clinicians could ask a patient if, when growing up, family members had ever hit her or him so hard that it left bruises or marks. If the patient responds in the negative, based on clinical judgment, a second question may be asked: “Did you ever have the experience of being hit or beaten to the extent that it was noticed by someone like a teacher, neighbor, or doctor?” A similar process is recommended in screening for sexual and emotional abuse. An alternative strategy would be to ask a single question for each type of abuse, resulting in a 3-item screen: (1) “Growing up, did family members ever hit you so hard that it left bruises or marks?” (2) “Did anybody try to touch you in a sexual way, or try to make you touch them?” and (3) “Did people in your family call you names like ‘stupid,’ ‘lazy,’ or ‘ugly?’”
There are limitations that should be taken into consideration in interpreting results from this study. The actual community prevalence of childhood abuse among adults is unknown because most cases are never reported to authorities.31 As is the case in most retrospective studies of childhood maltreatment, data in this study were limited to self-report. However, cutoff scores derived from CTQ-SF items used in this study have produced satisfactory sensitivity compared with more extensive abuse information solicited by structural interview32 and with verified cases of abuse when corroborative evidence was available.29 Also, this study was exploratory, and both the general inquiries explicitly mentioning the term abuse and the questions about behaviorally defined abusive experiences were embedded within the same questionnaire. Although the questions were dispersed throughout the questionnaire rather than being clustered by abuse type or type of question (labeling vs behaviorally defined), the order of these questions was the same for all respondents. Whether this would produce different results compared with an experimental study in which some respondents are asked explicitly about abuse and others only about abusive experiences, for instance, is an empirical question for future research. Finally, this study did not address implications for the use and scoring of the CTQ-SF. Thus, the degree to which sex biases may affect scoring or necessitate separate female and male cutoff scores, for instance, remains an open question.
In summary, this study makes an important contribution by providing evidence that the use of screening questions that inquire generally whether or not a patient has been physically, sexually, or emotionally abused identify low numbers of patients who report specific abusive events. Screening by physicians or other clinical personnel in the medical clinic setting provides an opportunity to recognize and manage both the emotional and physical sequelae of childhood abuse. It is important that effective, efficient screening methods be developed for this purpose.
Correspondence: Brett D. Thombs, PhD, Institute of Community and Family Psychiatry, Sir Mortimer B. Davis–Jewish General Hospital, 4333 Cote Ste Catherine Rd, Montreal, Quebec, Canada H3T 1E4 (firstname.lastname@example.org).
Accepted for Publication: March 9, 2006.
Author Contributions:Study concept and design: Thombs, Bernstein, Ziegelstein, Scher, Forde, Walker, and Stein. Acquisition of data: Scher, Forde, and Stein. Analysis and interpretation of data: Thombs. Drafting of the manuscript: Thombs. Critical revision of the manuscript for important intellectual content: Bernstein, Ziegelstein, Scher, Forde, Walker, and Stein. Statistical analysis: Thombs. Obtained funding: Forde.
Financial Disclosure: None reported.
Funding/Support: This study was funded in part by the Memphis Shelby Crime Commission, Memphis, Tenn (Dr Forde), and the Miller Family Scholar Program, Baltimore, Md (Dr Ziegelstein).
Additional Information: Dr Bernstein is a coauthor of the CTQ-SF instrument used in this study.