Soc Psychiatry Psychiatr Epidemiol (2003) 38: 644653

DOI 10.1007/s00127-003-0686-4

ORIGINAL PAPER

Ron Acierno Heidi Resnick Dean Kilpatrick Wendy Stark-Riemer

Assessing elder victimization

Demonstration of a methodology

Accepted: 15 May 2003

I Abstract Background Research on epidemiological assessment of interpersonal violence in younger adults indicates that there are essential components of interview-based queries necessary to maximize the likelihood that episodes of physical and sexual assault will be identified. These include phrasing to prime comprehensive disclosure, combined with extremely behaviorally specific descriptive questions about assault events. Such methodology permits simultaneous assessment of violence perpetrated by intimates or caregivers (i. e., abuse) as well as that perpetrated by strangers (i. e., assault), and is easily conducted via telephone. However, these strategies have not been used to simultaneously measure both abuse and assault in older adults, and some question exists as to the applicability of telephone interview techniques to geriatric populations. Methods A total of 106 elder respondents were surveyed, approximately half in person and half via the telephone, to evaluate the feasibility of using a telephone-based structured interview to measure both assault and abuse, as well as to screen for psychopathology in a geriatric population. In order to assure that a sufficient number of crime victims were available to assess the ability of dependent measures to detect a variety of abuse and assault types, approximately half of the group in each interview context was comprised of a victim oversample referred by local police departments, whereas the remainder of each group were randomly selected from local telephone directories. Results Data indicated that the methodology was successfully adapted to the geriatric population. Rates of lifetime and recent abuse and assault detected by in-person and telephone methods were comparable, and medical problems associated with agDr. R. Acierno () H. Resnick D. Kilpatrick W. Stark-Riemer National Crime Victims Research and Treatment Center 165 Cannon Street PO Box 250852 Charleston, SC 29425, USA Tel.: +1-843/792-2945 Fax: +1-843/792-3388 E-Mail: aciernre@musc.edu

ing posed no significant problem. Conclusions The methodology was acceptable to older adults and very effective in identifying both abuse and assault events, perpetrator status, and psychopathology. Telephone-based interviewing appeared to be a valid alternative to in-person interviews in identifying victimization and psychopathology. I Key words elder abuse violence prevalence epidemiology methodology

Introduction
Research on assessing interpersonal violence in adolescents and young adults is far more advanced than that on older adult victimization. The extent to which empirically based refinements to assessment methodology in younger populations are transferable to investigations with elders is not known. To date, studies of violence against young adults and adolescents indicate that prevalence estimates for criminal victimization may vary widely according to parameters of assessment methodology, including assessment context, assessment structure, assessor characteristics, and traumatic event definition (Breslau et al. 1991; Hanson et al. 1995; Kilpatrick et al. 1989; Koss 1993; Resnick et al. 1993; Resnick 1996). Overall, research with younger populations (National Womens Study: Kilpatrick et al. 1992; Kilpatrick et al. 2000) indicates that specific questions about violence must incorporate two crucial components in order to accurately detect victimization: (1) contextually orienting, empathetic preface statements, and (2) extremely specific behavioral descriptions of index events that elicit closed-ended responses. Traumatic events such as violent crime are associated with extremely aversive emotional and cognitive states. It is, therefore, important, both to respondent welfare and experimental integrity, to preface criminal victimization queries in such a way as to convey acceptance, empathy, and normalization. Obviously, victims will dis-

SPPE 686

645

close extremely personal and sensitive information only when they feel that such disclosure is worthwhile and relevant. Preface statements should establish this relevance. Of equal import, preface statements must also provide contextual orientation so that the likelihood of reporting that information sought by the investigator is maximized. For example, if questions regarding elder abuse follow a crime survey in which only police-reported crimes are investigated, and no preface statement is used to specifically direct respondents to disclose all assaults, including those not reported to authorities, then respondents might be biased toward disclosing only those events that have been reported to the police (Koss 1993). Similarly, if questions regarding assault follow a psychopathology survey, respondents might be biased toward disclosing only those assaults that are of a relatively bizarre nature (Koss 1993). Epidemiological researchers studying violence are typically interested in all experienced events. Thus, it must be made clear to the respondent that the individual collecting these data is interested in any assault perpetrated by any individual, at any time in their lives (e. g., assaults by family members years ago, as opposed to just assaults by strangers in the recent past). In successful studies of violence against younger populations, contextually orienting preface statements are followed by detailed behaviorally specific, closedended descriptions of trauma events under investigation (Kilpatrick et al.2000).Early criminal justice system (CJS) surveys of violence employed gateway screening questions characterized by very limited behavioral specificity (e. g., Have you ever been physically abused?). If respondents endorsed the gateway question, further questions about assault followed. Gateway questions shorten the overall interview process for those respondents not endorsing the gateway item. Unfortunately, gateway questions that lack specificity and that are not prefaced by contextually relevant background statements do not adequately orient respondents to the type of responses the assessor is seeking (i. e., they fail to state that one is interested in all abuse/assaults, not just those reported to the police or perpetrated by strangers). Most problematic, however, is that gateway questions are extremely subject to an individuals interpretation of queries (Koss 1993) (e. g.,Its not rape if my husband does it). Moreover, a respondents own victimization history will affect their personal definitions of elder abuse (Childs et al. 2000). Behaviorally specific, closed-ended (i. e., yes/no) questions are an alternative to gateway questions. Behaviorally specific descriptions of assault events minimize variance associated with cultural differences, personal differences in intellect, psychological stability, general willingness to disclose, or understanding of criminal justice terminology (e. g., rape, aggravated assault). These questions are highly detailed and require only yes or no answers in response to whether or not a specifically described event was experienced. In addition to removing definitional and cultural variance,

such closed-ended yes or no questions simplify the role of respondents and minimize the risk that anyone will overhear disclosure of highly personal events, particularly during telephone interviews, where at least the queries are unintelligible to others not on the phone. While this methodology has proven extremely effective with adolescents (Kilpatrick et al. 2000) and young adults (Kilpatrick et al. 1992), it has not yet been evaluated with older adults. (See Appendix for behaviorally specific questions used to define violence in the present study.)

I Telephone vs. in-person interview survey methodology

The previous discussion involved methodology used with younger adults in at least three prior populationbased studies conducted via telephone. To date, victimization interviews with older adults have almost exclusively been conducted in person. Several advantages exist for each format. In-person interviews permit visual contact between interviewers and respondents. In-person interviewers can also modulate their volume to a relatively greater extent than telephone interviewers. In addition, conducting in-person interviews allows surveyors to select an appropriate assessment location (i. e., a quiet, undisturbed room, as opposed to wherever the respondent happens to have answered his or her telephone). In-person interviews may also facilitate expressions of empathy, honesty, and respect, which then might encourage more complete self-disclosure. In-person interviews also allow interviewers to assess non-verbal communication not otherwise available to telephone interviewers. Finally, in-person interviews can be conducted in households that do not have telephones. (However, the advantage of in-person interviews over telephone interviews as far as telephone availability is concerned may be illusory. For example, according to the 1990 census, only 5 % of United States households did not have telephones.) By contrast, data indicate that telephone-based interviewing is an efficient method for collecting information from large representative samples of respondents at a relatively low cost with non-significant response bias in detection of critical variables of interest when compared to in-person interview approaches (Weeks et al. 1983). These issues have been examined specifically in terms of detection of rates of victimization using in-person vs. telephone interview methods (Catlin and Murray 1979), and police report data revealed that no differences in rates of detection of victimization were observed, supporting both the reliability and validity of the telephone method. One study (Paulsen et al. 1988) compared telephone and in-person assessment of DSM-III Axis I disorders, including anxiety disorders, affective disorders, alcoholism, and no mental disorder using a structured diagnostic interview. Kappas ranging from 0.69 to 0.84 were obtained, even with a delay between in-

646

person and telephone methods of 1219 months. To date, the efficacy of telephone-based assessment of interpersonal violence has not been compared to in-person interviewing with older adults.

I Assessment of abuse vs. assault

Past researchers made a distinction when studying victimization of older adults in that assessment efforts were confined to investigating either elder abuse/neglect by family members (including caregivers) or non-familial perpetrated criminal violence (i. e., assault), but not both in the same study with the same sample. This distinction may be artificial for three reasons. First, the physical and emotional effects of such events, particularly elder abuse and non-familial physical and sexual assault, are often very similar, or at least share a number of similarities (Acierno et al. 1997). Second, both forms of violence appear to have several risk factors in common (e. g., poverty, limited resources, previous victimization), indicating that victims of one type of assault may be more likely to experience the other type of assault relative to non-victims. Third, both forms of victimization are amenable to assessment through similar methodological strategies. Studies that endeavor to delineate risk factors for abuse or violence toward older adults should, therefore, simultaneously assess both forms of victimization when possible. Similarly, studies conducted to outline effects of these events on the elderly, and studies conducted to inform preventive interventions for both violence and effects of violence should use a methodology that assesses both forms of victimization. The primary purposes of the present study were to: (1) determine whether advances made in interpersonal violence assessment strategies with younger populations could be transferred to the older adult population; (2) simultaneously measure both sexual, physical, and emotional violence across the life span perpetrated by both strangers and family members/caregivers (previous studies have assessed only one or the other class of perpetrator); (3) quickly and unobtrusively screen for psychopathology according to DSM-IV criteria in relation to victimization; and (4) compare results obtained via telephone interviews with those obtained through in-person interviews.

were interviewed in person and 25 via telephone. Of the 59 randomly selected participants, 28 were interviewed in person and 31 via telephone. Sixty-eight participants were women, 38 were men. Of the total sample, 67 were White, non-Hispanic, 37 were African-American, non-Hispanic, 1 was Native American, and 1 refused to disclose. Among police-referred participants, 19 filed reports for physical assault, 12 filed reports for verbal assault or intimidation, 8 reports involved a domestic dispute, 4 participants filed reports for vandalism or burglary resulting in fear for their safety, the remaining 4 reported other crimes in which they feared for their safety, but were not assaulted, verbally abused, or robbed. I Measures A highly structured interview, modified slightly for older adults from that used in the National Womens Study (Kilpatrick et al. 1992), was used to collect information about a variety of topics, including demographic characteristics, interpersonal violence experiences, experience of other forms of trauma (e. g., natural disaster), and select forms of psychopathology and substance use problems. Interpersonal violence variables Sexual assault, molestation, physical assault, and emotional abuse were assessed. Sexual assault included forced or coerced vaginal or anal penetration by a penis or other object, or oral penetration by a penis. Physical assault was noted when participants reported being attacked by another person with a weapon or without a weapon, but with the intent to seriously injure or kill. Emotional abuse was defined as verbal attacks by others that resulted in strong feelings of threat, intimidation, humiliation, or worthlessness. Also included in this category were episodes in which participants were harassed or coerced into doing something against their will. (See Appendix for operational definitions of these study variables.) Psychopathology variables Screens for current major depressive disorder (MDD), panic disorder, generalized anxiety disorder (GAD), specific phobia, and post-traumatic stress disorder (PTSD) were conducted via a structured clinical interview based on DSM-IV and modified from the National Womens Study (NWS) (Kilpatrick et al. 1992). In addition, a variable lifetime substance use problems was created and considered positive if participants reported that they had ever experienced two or more DSMIV criteria for substance abuse (note that DSM-IV diagnosis of substance abuse requires participants to experience one of these problems as recurrent rather than ever). Categorical independent variables A major aim of this study was to determine the relative feasibility and usefulness of the structured interview delivered via telephone or inperson interviewers in terms of its ability to measure both assaultive and abusive forms of victimization, and screen for psychopathology in older adults. Therefore, participants were randomly assigned to either telephone or in-person interviews. Data were also considered in terms of the source of participants. Thus, individuals who necessarily had victimization backgrounds (e. g., police-referred victims) were considered relative to randomly sampled individuals. As mentioned, this victim oversample was used to insure that sufficient numbers of crime victims were present to evaluate the ability of the interview to identify both the victimization event and its psychological sequelae. I Procedure Participants were either randomly selected from the local population or referred by local police departments for participation in the study. For randomly selected participants, a random number generator was used to derive numbers that corresponded to the page number and

Subjects and methods

I Participants
= 66.5, sd = 8.5) Participants were 106 older adults aged 5585 years ( residing in a southeastern suburban region. Of the total, 47 comprised a victim oversample referred by local police departments and 59 were randomly selected from local telephone directories. The victim oversample assured that a sufficient number of crime victims were available to assess the ability of dependent measures to detect a variety of abuse and assault types. Interview context (in-person vs. telephone) was randomly assigned. Of the 47 police-referred participants, 22

647 listing number of participants from the telephone directory. Each randomly generated number was called by a project assistant, and the person answering the phone was asked if there were any individuals aged 55 and over residing in the home and available for interview. If more than one individual was available, the most recent birthday method was used to select a single respondent. Seventy-nine households were reached in which an eligible candidate resided, and 59 of the 79 older adults who were contacted agreed to participate in the project. (Forty-seven of 51 police-referred potential candidates agreed to participate in the project.) The project was described, the consent to participate was read aloud over the phone, and agreement for participation was obtained (note that respondents were told that participation might be either in person or via telephone). Subsequently, a witnessed coin toss determined whether or not the respondent was to be interviewed via telephone or in person, and a time convenient for the respondent was arranged. In-person interviews took place in the respondents home. Interviews followed the format of the National Womens Study in that victimization and psychopathology modules incorporated skipouts so that only questions relevant to each participants reported experiences and symptom presentation were asked. Because some participants were necessarily asked fewer questions than others, = 53 min, interview lengths varied considerably from 20 to 153 min ( sd = 20 min). In order to control for non-specific interviewer characteristics, interviewers administered approximately equal numbers of telephone and in-person interviews to police-referred and randomly selected participants. Interviewers were required to achieve 90 % accuracy relative to gold standard practice interviews prior to interviewing study participants for data collection. Two steps were taken in order to assure that participants could answer questions freely and without fear of negative consequences. First, interviewers asked whether participants could answer questions in relative privacy. If they were unable to answer privately, the interview was rescheduled. Second, the interview was structured nearly exclusively with closed-ended questions that required simple yes or no answers. Thus, if respondents were overheard, nothing they were saying would place them at risk. Participants received a $20 check as compensation for their time. The field period for the study was January 2000 to December 2001. I Participant protection The sensitive nature of the information collected from participants required strong steps to protect confidentiality. Therefore, in order to protect participant data from subpoena and other threats to privacy, a Certificate of Confidentiality was applied for and obtained from the National Institutes of Health. In addition, participant names were matched with arbitrarily assigned identification numbers and data files referenced by these numbers were stored on a secure computer. I Data Analysis Because this was a pilot feasibility demonstration project, the sample size was somewhat limited and the majority of analyses were descriptive, rather than inferential. Thus, usefulness of the method was illustrated by proportional responses to feasibility questions, actual rates of victimization, and descriptions of symptomatology. For select comparisons, chi square analyses were used to detect significant differences in categorical variables with an a priori alpha = p < 0.05.

Results
I Question 1: survey content and methodology: applicable to older adults?
This question is most directly answered by determining the extent to which we were able to detect victimization events, characteristics of these events and indicators of psychopathology in our sample. That is, this question can be answered by examining the characteristics of the data we obtained. Table 1 shows the prevalences of victimization variables for the total sample, and independently for police-referred and randomly selected subsamples for both lifetime events and events occurring since respondents were 55 years old. Overall, the methodology appeared quite useful in identifying victimization in older adults. Summed total sample prevalences are necessarily inflated because a large proportion of study participants were referred by the police. The fact that this inflation was detected in the police-referred relative to the randomly selected sample lends support to the validity of our victimization assessment. Data from the total sample indicated significant rates of recent physical assault and emotional abuse, but not sexual assault. Police-referred participants reported high rates of lifetime sexual assault (17 %) and physical assault (72 %). Emotional abuse appeared to be very prevalent in the total sample (about 85 % lifetime, 26 % recent). However, a differential finding was evident in terms of police/randomly obtained samples and time of occurrence. Specifically, lifetime emotional abuse was quite high in both samples, whereas recent emotional abuse was relatively lower in randomly sampled participants (51 % police vs. 7 % random). Information was also gathered on assault characteristics for each subsample. In the present study, victimization characteristics were consistent between policereferred and randomly selected participants who had been physically assaulted. These data are presented for the total sample, and for police and randomly selected subsamples in Table 2. Because the rate of sexual assault was so low,data are presented for only physical and emo-

Table 1 Prevalence of victimization events by referral source Victimization type Sexual assault Physical assault Emotional abuse Any crime

Lifetime Referral source Police Random Total % (n = 47) % (n = 59) % (n = 106) 17.0% (8) 72.3% (34) 83.0% (39) 87.2% (41) 0 7.5% (8) 22.0% (13) 44.3% (47) 86.4% (51) 84.9% (90) 91.5% (54) 89.6% (95)

Post age 55 Referral source Police Random Total % (n = 47) % (n = 59) % (n = 106) 0 48.9% (23) 51.1% (24) 68.1% (32) 0 1.7% (1) 6.8% (4) 8.5% (5) 0 22.6% (24) 26.4% (28) 34.9% (37)

648 Table 2 Assault and perpetrator characteristics Physical assault Referral source Characteristics Assault characteristics* Fear killed/injured Seriously injured Multiple events Reported to police Perpetrator characteristics** Know perp well Perp shares residence Perp substance abuse Perp police prob Perp in counseling Perp transport appt Perp obtains meds Can live w/o perp Police % (n) 60.7% (17) 10.7% (3) 35.7% (10) 82.1% (23) 83.3% (20) 66.7% (12) 55.6% (10) 58.8% (10) 33.3% (6) 11.1% (1) 0 86.7% (13) Random % (n) 53.8% (7) 7.7% (1) 25.0% (3) 38.5% (5) 60.0% (6) 50.0% (1) 50.0% (1) 0 100.0% (1) 100.0% (1) 100.0% (1) 50.0% (1) Total % (n) 58.5% (24) 9.8% (4) 32.5% (13) 68.3% (28) 76.5% (26) 55.0% (13) 55.0% (11) 52.6% (10) 36.8% (7) 20.0% (2) 11.1% (1) 82.4% (14) Emotional abuse Referral source Police % (n) 40.0% (12) 76.7% (23) 60.0% (18) 100.0% (29) 60.9% (14) 65.2% (15) 60.9% (14) 43.5% (10) 0 0 87.0% (20) Random % (n) 33.3% (3) 85.7% (12) 33.3% (3) 100.0% (11) 50.0% (3) 33.3% (2) 33.3% (2) 20.0% (1) 50.0% (1) 50.0% (1) 66.7% (2) Total % (n) 38.5% (15) 79.5% (35) 53.8% (21) 100.0% (40) 58.6% (17) 58.6% (17) 55.2% (16) 39.3% (11) 8.3% (1) 8.3% (1) 84.6% (22)

Note The n for each item varies because some participants for whom the item was irrelevant (e. g., perpetrator characteristics for individuals who reported that they did not know the perpetrator) were not asked about the item. Perp (perpetrator), Prob (problem), Appt (Appointment), Meds (medicine) * Of those reporting physical assault or emotional abuse, respectively ** Of those who knew the perpetrator

tional forms of victimization. Approximately 58 % of physically assaulted participants reported fear of being severely injured or killed during the incident. For those respondents reporting past emotional abuse, 38 % reported these fears. However, actual serious injury was sustained by 10 % of individuals during their physical assault, and by none during episodes of emotional abuse. Finally, 33 % and 80 % of those experiencing physical assault and emotional abuse, respectively, reported that these events were but one in a series, and many, but by no means all, incidents were reported to the authorities. The interview was also useful in gathering perpetrator characteristics for each form of assault (see Table 2). Perpetrators of physical and emotional violence were very similar in that most were well known to and had actually lived with their victims. About half had problems with substance abuse and the police, but few were in counseling. Many perpetrators appeared to provide some assistance to their victims, although most victims indicated that they could live without this aid. DSM-IV defined current psychopathology (with the exception of substance use problems) was also assessed and is illustrated in Table 3 for the total sample, and independently for police-referred and randomly recruited participants.As with victimization prevalences, the proportion of participants screening positive for psychopathology was inflated by the presence of a large group of recent crime victims, with lifetime rates of MDD ranging from 23 % in police-referred participants to 10 % in randomly selected individuals. Rates of positive screens for Panic (11 %) and GAD (8.5 %) were also high in both groups. PTSD was elevated in police-referred participants (15 %), but was also unexpectedly

Table 3 Prevalence of psychopathology Referral source Psychopathology MDD Panic GAD Social phobia Specific phobia PTSD Substance problems Police % (n = 47) 23.4% (11) 14.9% (7) 12.8% (6) 0 4.3% (2) 14.9% (7) 8.5% (4) Random % (n = 59) 10.2% (6) 8.5% (5) 5.1% (3) 3.4% (2) 5.1% (3) 8.5% (5) 15.3% (9) Total % (n = 106) 16.0% (17) 11.3% (12) 8.5% (9) 1.9% (2) 4.7% (5) 11.3% (12) 12.3% (13)

high in randomly sampled individuals (8 %). Finally, a large number of participants reported that they had two or more problems associated with substance abuse at some time in their lives. Rates of psychopathology-positive screens were also examined in terms of victimization (note that power for these analyses was extremely low due to the small sample size). Table 4 provides the correlation matrix between types of lifetime victimization and current psychopathology. Sexual assault, while infrequently experienced, was associated strongly with positive screens for MDD, Panic, GAD, and PTSD (specifically the arousal and avoidance symptom categories). By contrast, physical assault was significantly associated with avoidance and re-experiencing symptoms of PTSD, as well as substance use problems.

649 Table 4 Correlations between lifetime victimization and current psychopathology Psychopathology Sexual assault (n = 106) 0.459** 0.574** 0.297** 0.040 0.064 0.349** 0.271** 0.315** 0.111 0.111 Physical assault (n = 106) 0.024 0.041 0.067 0.124 0.019 0.161 0.138 0.241* 0.300** 0.187 Emotional abuse (n = 106) 0.041 0.067 0.034 0.058 0.155 0.067 0.113 0.133 0.101 0.110 Any crime (n = 106) 0.020 0.024 0.007 0.047 0.070 0.122 0.095 0.204* 0.132 0.127 Table 6 Prevalence of psychopathology by interview format Interview context Psychopathology MDD Panic GAD Social phobia Specific phobia PTSD Substance problem In-person % (n = 50) 16.0% (8) 10.0% (5) 8.0% (4) 0 8.0% (4) 14.0% (7) 10.0% (5) Telephone % (n = 56) 16.1% (9) 12.5% (7) 8.9% (5) 3.6% (2) 1.8% (1) 8.9% (5) 14.3% (8) 2 (df) 0.00 (1) 0.16 (1) 0.03 (1) 1.82 (1) 2.27 (1) 0.68 (1) 0.45 (1) p 0.992 0.685 0.864 0.177 0.132 0.411 0.502

MDD Panic disorder GAD Social phobia Specific phobia PTSD PTSD arousal PTSD avoidance PTSD re-experiencing Substance problems

Note There were no significant differences between prevalences of each psychopathology type detected by telephone vs. in-person interviews

* p < 0.05; ** p < 0.01 (2-tailed)

I Question 2: survey format: telephone or in-person

Table 5 shows lifetime and post-55 rates of victimization detected via telephone and in-person methods. There were no significant differences in observed rates of victimization, both recent and distant past, for either form of assessment. Comparisons between telephone and in-person interview formats were also made in terms of detected psychopathology. As is evident from Table 6, telephonebased interviewing was as effective as in-person interviewing in detecting psychological disorders in this population. Specifically, there were no significant differences in disorder prevalence between those participating in in-person vs. telephone interviews. The relative feasibility of conducting telephone and in-person interviews was also compared from the perspectives of both interviewees and interviewers and is given in Table 7. Specific feasibility questions are given in the Appendix. Very few participants in either format reported significant hearing problems, fatigue during the interview, or elevated distress from interview questions. Most individuals valued their efforts and thought that their participation would ultimately help others. Interviewers reported that very few participants in either
Table 5 Prevalence of victimization events by interview format Lifetime Interview context Victimization type Sexual assault Physical assault Emotional assault Any crime In-person % (n = 50) 8.0% (4) 42.0% (21) 82.0% (41) 86.0% (43) Telephone % (n = 56) 7.1% (4) 46.4% (26) 87.5% (49) 92.9% (52) 2 (df)

condition were overly fatigued by the interview, and very little interference secondary to medication or environmental intrusions was noted. Significant differences were only observed with respect to total interview time [shorter for telephone interviews 47.1 min vs. 58.6 min, F(1,101) = 9.15, p < 0.01] and interviewer safety selfrated on a 0 (totally unsafe) to 100 (perfectly safe) scale [safety was higher in telephone interviews, 100 vs. 93.9, F(1,98) = 11.82, p > 0.001]. Finally, 100 % of all interviews that were initiated were completed. That is, no participants had to stop the interview because of discomfort with the questions, fatigue, or physical problems.

Discussion
This study demonstrates that advancements made in assessing interpersonal violence with adolescents and young adults, namely the use of explicit, behaviorally specific closed-ended event questions, contextually orienting preface statements, and simultaneous assessment of both assault by strangers and abuse by family members/caregivers, are applicable to older adults. This methodology was effective in detecting various types of interpersonal violence, as well as in outlining violencerelated assault and perpetrator characteristics. In contrast to existing studies of elder abuse and elder assault, this assessment protocol is the first to simultaneously

Post age 55 Interview context p 0.868 0.647 0.430 0.248 In-person % (n = 50) 0 18.0% (9) 28.0% (14) 32.0% (16) Telephone % (n = 56) 0 26.8% (15) 25.0% (14) 37.5% (21) 2 (df) 1.16 (1) 0.12 (1) 0.35 (1) p 0.281 0.727 0.553

0.028 (1) 0.210 (1) 0.623 (1) 1.33 (1)

Note There were no significant differences between prevalences of each victimization type detected by telephone vs. in-person interviews

650 Table 7 Feasibility ratings: interviewers and participants Feasibility questions Significant distress Made you tired $20 is enough Problems with interview Participation will help Conducive environment Respondent overly fatigued Physical limitations Medication interfered Memory problems Hearing difficulties Interview context In-person % (n) 8.0% (4) 8.0% (4) 86.0% (43) 0 84.8% (39) 78.0% (39) 12.2% (6) 8.2% (4) 2.0% (1) 10.2% (5) 10.0% (5) Telephone % (n) 5.5% (3) 12.7% (7) 83.6% (46) 7.3% (4) 85.2% (46) 86.5% (45) 7.4% (4) 1.9% (1) 3.8% (2) 11.3% (6) 27.8% (15) 2 (df) 0.27 0.62 0.11 3.78 0.00 1.29 0.69 2.15 0.27 0.03 5.28 n 105 105 105 105 100 102 103 102 102 102 104 p 0.602 0.430 0.736 0.052 0.955 0.258 0.408 0.142 0.605 0.856 0.022

Note Sample n varied for these comparisons because data were missing for some participants. There were no significant differences in these feasibility ratings detected by telephone vs. in-person interviews

measure both assault by strangers and abuse by family members or caretakers using the same event-based questions. This is also the first study to directly compare telephone and in-person formats when assessing victimization and screening for psychopathology in older adults. However, an important limitation of this study should be noted: our methodology is biased toward validating telephone-based interviewing because a significant proportion of participants were recruited over the telephone to participate in the study. Hence, these individuals would be unlikely to object to telephone interviewing, having just acquiesced to telephone study recruitment. Moreover, individuals with significant disability, who feared telemarketing fraud, or who found telephone interviewing distasteful would not have been captured by the present methodology in that we would have been unable to contact them via telephone in the first place. This is problematic if these factors correlate with victimization or psychopathology status (e. g., disability). Along these lines, older adults recruited from police reports had already been interviewed at least once with respect to their victimization, and thus may differ (i. e., be more willing to participate in victimization interviews) from those who are victimized, but do not make reports to the police. With these caveats in mind, we found that telephone interviews produced similar detection rates of each form of recent and distant past victimization and psychopathology. Moreover, subsequent feasibility analyses indicated that telephone interviewing was more easily accomplished, required less time, and was safer for interviewers than face-to-face interviewing. Notwithstanding the aforementioned recruitment bias, physical problems and medication effects did not seriously impede telephone interviewing procedures or diminish the validity of collected data. Finally, informal queries revealed that elderly participants were far more comfortable with telephone interviews.

These data also demonstrate that the survey content and method of asking very personal questions about painful experiences and emotional problems were acceptable to older adult participants. Moreover, the finding of higher rates of victimization in the police-referred relative to randomly sampled participants supports the accuracy of the assessment instrument. With respect to content, no older adult participant found the survey questions too objectionable to continue. Further, few were fatigued by the interview, and none terminated the interview due to impaired hearing or other physical problems. In addition to findings regarding feasibility of the method, several specific preliminary observations were notable. Of particular interest was the high rate of lifetime sexual assault (17 %) and physical assault (72 %) among police-referred subjects. Many of these participants were identified for this study from police reports that did not involve these forms of assault. Indeed, recall that, in the police-referred group, only 40 % filed police reports for physical assault and none filed reports for sexual assault. This indicates that these individuals comprise a multiply victimized group, and that this victimization is quite chronic and may change in character (i. e., from sexual and physical to emotional) over the life span. Moreover, this finding demonstrates that the interview was sensitive to changes over time, and specifically to retrospective report, with more intense and frequent events reported in the distant past relative to recent past. Lifetime emotional abuse was quite high (i. e., in excess of 80 %). This engendered concern that the criteria for this type of event (see Appendix) may have been too liberally defined. However, high rates of emotional abuse in randomly sampled participants occurring before age 55 (86 %) contrasted with much lower (7 %) rates of more recent post-age 55 emotional abuse in the same subsample. This differential finding further supports the ability of the instrument to detect this form of

651

elder mistreatment, and allays the concern that this variable was too liberally defined. Observed rates of lifetime and recent violence and abuse were somewhat higher than those obtained by previous investigators. The 7 % prevalence of post-age 55 emotional abuse detected in randomly sampled participants was higher than reported by Pillemer and Finkelhor (1.1 %); however, rates of physical assault were 1.7 %, or very close to the 1.1 % detected by these investigators. Pillemer and Finkelhors rates might be lower because they assessed abuse only by caregivers, whereas we assessed emotional abuse by intimates, caregivers, and strangers. The interview also effectively obtained information about assault and perpetrator characteristics, which are potentially important data in risk factor research. Consistent with findings on younger adults, most perpetrators were known to their victims and had substance use problems. Particularly interesting were the high rates of shared residence between perpetrators and victims. Even though we assessed both stranger and family member perpetrated violence, more than half of all victims lived with the perpetrator. Overall, the assessment methodology was effective in detailing a variety of potentially relevant event and perpetrator characteristics for risk factor research. As with younger adults, such data are integral in structuring primary and secondary preventive interventions that are uniquely tailored to specific experienced crime characteristics. Informal debriefing of our interviewers for the pilot study indicated that there were several advantages in addition to those noted earlier for the telephone-based interview format. Many participants who were scheduled for in-person interviews indicated that they had some hesitation allowing a stranger into their homes. For many participants, the telephone format also appeared to be perceived as relatively more anonymous and less intimidating than in-person disclosures of personal victimization, particularly when perpetrators were family members. Indeed, this anonymity may facilitate disclosure of embarrassing or potentially problematic material. Moreover, this anonymity may reduce the risk of negative outcomes upon disclosure of abuse events. That is, if an interview is conducted in person, the interviewer is present in the house and clearly noticed by the abuser. The abuser may even overhear the interview questions and be aware of the older adults responses. This was not a problem during our telephone-based interviews, in which the interview itself was subtly completed, and interviewees were required only to give yes,no,or number responses, which, if overheard, would not be incriminating. In addition, older adults may be more likely to disclose assault events to faceless voices over the telephone than to in-person interviewers, leading to greater sensitivity of the measure. That is, the potential for embarrassment upon disclosure of personal victim-

ization and emotional problems is greater via the inperson interview format. Telephone-based interviewing also has the advantage of improving access to participants from across the socio-economic status range. Thus, the very rich, rich, middle class, lower class, and poor are equally approachable if they have a telephone. It is unlikely that the upper and lower ends of the socio-economic spectrum would be available for in-person interviews. Another important concern and advantage of telephone assessment is interviewer safety. In order to achieve national representativeness, all geographic and economic areas must be surveyed and interviewers must enter highcrime areas where they will be at increased risk for victimization. Telephone-based assessment overcomes the risk of victimization that interviewers would certainly face in conducting a large-scale study. Moreover, telephone interviews are far less expensive than in-person interviews, and generally require significantly less time to complete. Finally, logistic factors such as scheduling, dealing with mandatory reporting issues, overcoming participant hesitation at having strangers in the house, etc. are relatively less problematic for telephone-based interviews. Advantages of in-person interviewing in establishing occurrence of interpersonal violence and deriving psychological diagnoses, however, cannot be dismissed. Inperson interviews permit visual assessment of both the respondents physical presentation and his or her reactions to questioning. Moreover, in-person interviewing permits some flexibility to interviewers in terms of reflecting empathy, offering non-verbal indications of support, and pursuing relevant areas of assessment to a greater degree. In-person interviewing is also advantageous in terms of verifying that instructions have been understood through body language (e. g., a puzzled look would indicate that they had not, and would be missed by telephone methodology). Finally, the validity of clinical diagnosis made on the basis of in-person interviews is higher than that of diagnosis made via telephone simply because more convergent (or divergent) lines of data are available to interviewers in the former format (e. g., visual and non-verbal cues in addition to simple self-report). Overall, the assessment methodology employed and evaluated in this pilot study proved effective in identifying both abuse and assault in older adults. In addition, this methodology was useful in outlining characteristics of interpersonal violence events, and the perpetrators of this violence. Moreover, potential outcomes of violence in the form of psychopathological symptoms were also successfully measured. Finally, conducting interviews via telephone did not diminish the validity or feasibility of the assessment method among those older adults without cognitive or physical disabilities that would preclude telephone use.

652

Appendix
Operational definitions of select study variables. I Victimization questions Sexual assault was scored as present when participants answered affirmatively to any of the following questions. 1. Now I want to ask you about unwanted sexual advances that you may have experienced over your lifetime. People do not always report such experiences to the police or discuss them with family or friends. The person making the advances isnt always a stranger, but can be a friend, romantic partner, or even a family member. Such experiences can occur any time in a persons life even as a child. Regardless of how long ago it happened or who made the advances, has anyone ever made you have sex by using force or threatening to harm you or someone close to you? (For women, also read below) Just so there is no mistake, by sex we mean putting a penis in your vagina. 2. Has anyone ever made you have oral sex by force or threat of harm? (For women, also read below) Just so there is no mistake, by oral sex we mean that a man or a boy put his penis in your mouth, or someone penetrated your vagina or anus with their mouth or tongue. (For men, also read below) Just so there is no mistake, by oral sex we mean that a man or a boy put his penis in your mouth, or someone penetrated your anus with their mouth or tongue. 3. Has anyone ever made you have anal sex by force or threat of harm? (For women, also read below) Has anyone ever put fingers or objects in your vagina or anus against your will by using force or threats? (For men, also read below) Has anyone ever put fingers or objects in your anus against your will by using force or threats? Physical assault was scored as present when participants answered affirmatively to any of the following questions. 1. Another type of stressful event that people sometimes experience is being physically attacked by another person. Has anyone including romantic partners, family members, or friends ever attacked you with a gun, knife, or some other weapon, regardless of when it happened or whether you ever reported it or not? 2. Has anyone including romantic partners, family members, or friends ever attacked you without a weapon, but with the intent to kill or seriously injure you? Again, do not include any incidents you already reported to me. 3. Has anyone including romantic partners, family members, or friends ever physically attacked you so that you suffered some degree of injury, including cuts, bruises, or other marks? Emotional abuse was scored as present when participants answered affirmatively to any of the following questions. 1. Has anyone including romantic partners, family members, and friends ever verbally attacked you so that you felt afraid for your safety or verbally attacked you so that you felt threatened or intimidated? 2. Has anyone including romantic partners, family members, and friends ever verbally attacked you so that you felt humiliated or very embarrassed, for example by calling you names such as stupid, or telling you that your opinion was worthless. 3. Has anyone including romantic partners, family members, and friends ever forcefully and repeatedly asked or told you to do something so much that you felt harassed or coerced into doing something against your will?

I Feasibility questions The following questions were asked of participants to determine the feasibility of the methods used. 1. Did you have any trouble hearing me or understanding the questions I was asking you? 2. Did any of the questions we asked you cause you significant distress or make you very upset? 3. Did you get very tired as a result of participating in this interview? 4. Do you think $20 is enough payment for your time with this interview? 5. Do you think your participation in this study will help other older adults? The following questions were asked of interviewers to determine the feasibility of the methods used. 1. Did respondents hearing difficulties interfere with interview? 2. Was respondents home environment conducive to interview, or were there frequent interruptions? 3. Did the respondent seem overly fatigued or tired during any part of the interview? 4. Did the respondent have any obvious physical limitations, other than hearing problems, that interfered with the interview? 5. Did any medication the respondent was taking interfere with his or her ability to engage in the interview? Answer yes if respondent voiced this or if interviewer was under this impression. 6. Did the respondent have any obvious memory problems? That is, in addition to memory questions asked, did it appear respondent could not remember specific questions long enough to respond? I Acknowledgement Funding for this research was provided by the National Institute on Aging pilot research program #AG1811501.

References
1. Acierno R, Resnick HS, Kilpatrick DG (1997) Health impact of interpersonal violence: I. Prevalence rates, case identification, and risk factors for sexual assault, physical assault, and domestic violence in men and women. Behav Med 23(2):5364 2. Breslau N, Davis GC, Andreski P, Peterson E (1991) Traumatic events and posttraumatic stress disorder in an urban population of young adults. Arch Gen Psychiatry 48(3):216222 3. Catlin G, Murray S (1979) Report on Canadian victimization survey methodology pretests. Statistics Canada, Ottawa 4. Childs HW, Hayslip B, Radika LM, Reinberg JA (2000) Young and middle-aged adults perceptions of elder abuse. The Gerontologist 40(1):7585 5. Hanson RF, Kilpatric DG, Falsetti SA, Resnick HS (1995) Violent crime and mental health. Traumatic stress: from theory to practice. Plenum Press, New York, pp. 129161 6. Kilpatrick DG, Acierno R, Resnick H, Saunders B, Best C (2000) Risk factors for adolescent substance abuse: data from a national sample. J Consult Clin Psychol 68:1930 7. Kilpatrick DG, Edmunds CN, Seymour AK (1992) Rape in America: a report to the nation. National Victim Center & Medical University of South Carolina, Arlington, VA 8. Kilpatrick DG, Saunders BE, Amick-McMullan A, Best CL, Veronen LJ, Resnick HS (1989) Victim and crime factors associated with the development of crime-related post-traumatic stress disorder. Behav Therapy 20:199214 9. Koss MP (1993) Detecting the scope of rape: a review of prevalence research methods. J Interpers Violence 8(2):198222 10. Paulsen AS, Crowe RR, Noyes R, Pfohl B (1988) Reliability of the telephone interview in diagnosing anxiety disorders. Arch Gen Psychiatry 45(1):6263 11. Pillemer K, Finkelhor D (1988) The prevalence of elder abuse: a random sample survey. The Gerontol Soc Am 28(1):5157

653 12. Resnick HS, Dansky BS, Saunders BE, Best CL (1993) Prevalence of civilian trauma and post traumatic stress disorder in a representative national sample of women. J Consult Clin Psychol 61(6):984991 13. Resnick HS, Falsetti SA, Kilpatrick DG, Reedy J (1996) Assessment of rape and other civilian trauma-related PTSD: emphasis on assessment of potentially traumatic events. In: Miller TW (ed) Theory and assessment of stressful life events. International University Press, Inc, Madison, pp. 235271 14. Weeks MF, Kulka RA, Lessler JT, Whitmore RW (1983) Personal versus telephone surveys for collecting household health data at the local level. Am J Public Health 73(12):13891394