Self-Report Measure of Financial Exploitation of Older Adults

Transcription

1 The Gerontologist The Author Published by Oxford University Press on behalf of The Gerontological Society of America. Vol. 50, No. 6, All rights reserved. For permissions, please doi: /geront/gnq054 Advance Access publication on July 28, 2010 Self-Report Measure of Financial Exploitation of Older Adults Kendon J. Conrad, PhD,*,1 Madelyn Iris, PhD, 2 John W. Ridings, PhD, 3 Kate Langley, MPH, 1 and Kathleen H. Wilber, PhD 4 1 School of Public Health (MC 923), University of Illinois at Chicago. 2 Leonard Schanfield Research Institute, CJE SeniorLife, Chicago, Illinois. 3 Metropolitan Family Services, Chicago, Illinois. 4 Andrus Gerontology Center (MC 0191), University of Southern California, University Park, Los Angeles. *Address correspondence to Kendon J. Conrad, PhD, School of Public Health (MC 923), University of Illinois at Chicago, 1603 West Taylor Street, Chicago, IL kjconrad@uic.edu Purpose: This study was designed to improve the measurement of financial exploitation (FE) by testing psychometric properties of the older adult financial exploitation measure (OAFEM), a client self-report instrument. Design and Methods: Rasch item response theory and traditional validation approaches were used. Questionnaires were administered by 22 adult protective services investigators from 7 agencies in Illinois to 227 substantiated abuse clients. Analyses included tests for dimensionality, model fit, and additional construct validation. Results from the OAFEM were also compared with the substantiation decision of abuse and with investigators assessments of FE using a staff report version. Hypotheses were generated to test hypothesized relationships. Results: The OAFEM, including the original 79-, 54-, and 30-item measures, met stringent Rasch analysis fit and unidimensionality criteria and had high internal consistency and item reliability. The validation results were supportive, while leading to reconsideration of aspects of the hypothesized theoretical hierarchy. Thresholds were suggested to demonstrate levels of severity. Implications: The measure is now available to aid in the assessment of FE of older adults by both clinicians and researchers. Theoretical refinements developed using the empirically generated item hierarchy may help to improve assessment and intervention. Key Words: Financial abuse, Mistreatment, Rasch measurement, Exploitation theory, Theoretical hierarchy Received February 22, 2010; Accepted June 21, 2010 Decision Editor: William J. McAuley, PhD In the last 30 years, because elder mistreatment began to be widely recognized by social scientists, strides have been made to identify, measure, and substantiate abuse. Building on a seminal report by the National Research Council (2003), research has focused on improving the measurement of elder mistreatment to understand better the scope of the problem and the associated risk and protective factors. Cooper, Selwood, and Livingston (2008) in a systematic review of elder mistreatment prevalence studies through October 2008 identified 49 such studies conducted in more than a dozen countries. However, they did not identify any studies specifically devoted to the conceptualization and measurement of financial exploitation (FE) of older adults. FE is a unique category of elder mistreatment that does not have previously developed scales to draw from as do other forms of abuse. As a result, empirical studies of FE have lagged. FE has been defined as the illegal or improper use of a vulnerable adult s funds or property for another person s profit or advantage (AARP International, 2006; National Center on Elder Abuse, 1998). In practice, FE may be difficult to detect for a variety of reasons. For example, the onset is often gradual and insidious and, lacking oversight, subtle deception may mimic legitimate transactions and escalate over time. Differentiating FE from legitimate transactions is challenging 758 The Gerontologist

2 in that there may be indications of consent by the older adult, for example, a signed document and an apparent gift, when in fact the perpetrator has used psychological manipulation or misrepresentation (Wilber & Reynolds, 1996). Cognitive impairment, sensory impairment, or lack of financial sophistication may also cloud the distinction between willing assent and FE. Risks for FE may increase for older adults with paid or unpaid caregivers who have access to their financial assets, such as bank accounts, investment funds, etc. (Anetzberger, 2000). Differentiating FE from legitimate resource sharing and gifting is especially difficult when the elder is not a reliable reporter because of cognitive impairment, coercion, or concern about what will happen to the suspected abuser. In addition, other factors such as different cultural perceptions of sharing wealth may blur the distinction between generosity and exploitation (Langan & Means, 1996). (For a more comprehensive review of the FE literature, see Conrad et al., in press; Hafemeister, 2003.) Conceptual Models Although conceptual development in elder abuse research is sparse, several authors have suggested theoretical approaches to guide data collection efforts and provide effective assessment of the risk factors for and the consequences of different types of abuse (Anetzberger, 2000; Godkin, Wolf, & Pillemer, 1989). For FE, a number of indicators of abuse have been identified, for example, suspicious signatures on checks, missing documentation about financial arrangements, and unusual banking activities (National Committee for the Prevention of Elder Abuse, 2008; Quinn & Tomita, 1997). Over the last decade, there have been several efforts to develop broader conceptual frameworks specific to FE. For example, Wilber and Reynolds (1996) identified four components of financial elder abuse: vulnerabilities of the elder; characteristics of the relationship between the older adult and the alleged perpetrator; an assessment of who benefits from the relationship and how; and consideration of the process and tactics used and whether or not these meet the standards of undue influence, deceit, coercion, or theft. Building on this framework, Kemp and Mosqueda (2005) developed and tested a model, that added several new areas, including the older adult or the transactions are kept isolated, controlled, or secret; a qualified expert assesses neither the elder s capacities nor whether the transaction was in the older adult s best interest; common business or personal ethics are not followed; and the perpetrator does not consider the effect on others including the victim, family, beneficiaries, or the public welfare system. Rabiner, O Keeffe, and Brown (2004) presented a comprehensive conceptual model that included micro processes such as power and exchange dynamics, characteristics of the relationship in addition to the victim and perpetrator, status inequality, and social networks. They also included the broader sociocultural and policy context to understand better the etiology of FE. These various models have several commonalities; primary among them is that they recognize the importance of including the perpetrator and his or her characteristics as well as the social network. In addition, the relationship itself must be assessed in terms of the (a) extent to which the perpetrator is in a position of trust, (b) status inequality between perpetrator and victim, (c) patterns of interaction over time, and (d) extent to which there is reciprocity versus highly skewed benefits and losses. Although the models help explain the etiology of general abuse and the nature of financial abuse, neither do they present examples of statements that represent individual components of FE nor do they indicate which components are most important or most severe. Understanding these issues is essential to obtaining accurate assessments of types and levels of exploitation. To address the latter issue, Conrad and colleagues (in press) developed a model using concept mapping that included five concepts in a severity hierarchy that resulted in a measure called the older adult financial exploitation measure (OAFEM). The initial OAFEM measure, upon which this study was based, was developed using three-dimensional concept mapping applying the Trochim (1989) mapping technique. Following this approach, concept maps for FE were developed using two sources. Statements were generated first from a literature review of all empirical and conceptual work on FE available. Subsequently local and national panels of experts (n = 16) brainstormed descriptive statements of key behaviors related to FE. In this way, any statements from the literature that were not suggested by the experts were included by the authors for consideration. The panel of experts then reviewed the list of behaviors, rating the severity of each, conceptually sorting the statements into groups, Vol. 50, No. 6,

3 and naming each group. This information was analyzed and visually presented as point and cluster maps using Concept Systems Software (Kane & Trochim, 2007) that illustrated relationships among the FE behaviors as well as their severity ratings (Conrad et al., in press). Statements were grouped into six clusters, visually depicted as maps, and ranked by the experts as follows in descending severity: (a) theft and scams, (b) financial victimization, (c) financial entitlement, (d) coercion, (e) signs of possible FE, and (f) money management difficulties (Conrad et al., in press). The statements developed for the concept map were subsequently made into questions, and questionnaires were developed for both staff observation and client self-report. Nine focus groups, six involving 44 staff from various agencies and three involving 20 older adult consumers, were conducted to review and refine the wording of the questions. Four cognitive interviews of the client measures were then conducted with substantiated elder abuse clients whose input was also incorporated. The resulting questionnaire consisted of 82 questions. Details of these focus groups and other qualitative work are beyond the scope of this article, but may be reviewed in Conrad, Iris, and Ridings (2009). Purpose of the Study This article reports on the next stage in the development of the OAFEM in which field tests were used to evaluate the instrument. Accordingly, we examined client self-report measures of FE using item response theory, including Rasch modeling, and traditional validation techniques. Our specific objectives were to: (a) test the fit of the items to the model; (b) test the construct dimensionality of the OAFEM by examining whether or not the items formed a single overarching FE construct; (c) assess internal consistency reliability for all forms using a standard of.80 person reliability; (d) develop short forms that would be user-friendly for clinical applications; (e) examine the appropriateness of the measure for the target population, that is, items centering on the sample as opposed to having floor and ceiling effects; (f) further test construct validity by positing a hierarchy of concept rankings that conforms to expectations developed in the prior research phase and then testing a set of hypothesized relationships using correlation analysis; and (g) propose a reasonable, although speculative given lack of external validation, cutoff to determine FE. Table 1. Demographic Characteristics of Staff Sample (n = 22) Percent Number a Gender Men Women Race African American Caucasian Hispanic Mixed/other Years of experience, M (SD) 5.46 (5.68) years a Numbers may not add up to 100% due to missing values Design and Methods Sample Data collection was supported through a research agreement with the Illinois Department on Aging (IDoA), which helped recruit elder abuse providers in the state. Seven adult protective services (APS) agencies in Chicago and its collar counties assisted in the development of a database to estimate the psychometric properties of two types of FE measures: APS client self-report and APS investigator (staff) assessments. We used data from client selfreport measures of FE administered via interview by 22 staff to 227 clients who were substantiated for at least one type of elder mistreatment. We included substantiated clients to ensure a target sample that was appropriate for the measures. To obtain the full range of the construct and include a substantial group in the floor, however, clients did not have to be substantiated for FE. In addition to administering the interview, the APS staff who participated completed a staff observation questionnaire on each of the clients they interviewed. Staff received training on the protocols, including interviewing clients, by the two lead authors. A key component of the interview was the assessment of cognitive status using the mini-mental state examination (MMSE; Folstein, Folstein, & McHugh, 1975). To participate in the study, the client had to score at least 17 on the MMSE or, in the judgment of the APS investigator, demonstrate adequate cognitive capacity to provide self-report. The human subjects research proposal and informed consent forms were approved by the University of Illinois at Chicago (UIC) internal review board via the human subjects subcommittee. All participating APS staff completed the human subjects committee online training program of UIC. As shown in Table 1, the staff sample (n = 22) was 760 The Gerontologist

4 Table 2. Demographic Characteristics of Client Sample (n = 227) Percent Number a Age M (SD): (9.95) <65 years years years >90 years Gender Men 29.6 Women 70.4 Race American Indian/Alaskan Native Asian/Pacific Islander African America Caucasian Mixed/other Refused Hispanic Yes No a Numbers may not add up to 100% due to missing values. predominantly women (86.36%). More than half was Caucasian (59.09%), a quarter was African American (27.27%), and the remainder Hispanic or mixed race. Average years of experience were 5.46 years. In Table 2, the sample of APS clients (n = 227) was also predominantly women (70.4%). The majority was African American (61.3%), more than one third was Caucasian (35.5%), and the remainder was of mixed race or other. Most were non-hispanic (92.9%). The majority was between 75 and 90 years (58.7%). Statistical Analysis The Rasch measurement model (Rasch, 1960) was chosen because of its desirable scaling properties of linear, interval measurement (Embretson & Reise, 2000). This was important to accurately establish item hierarchy and distances between items to support theory building and test construct validity. Therefore, the Rasch model was needed to test the theoretical hierarchy developed in prior work. This is a type of construct validation. The Rasch hierarchy was also useful in examining dimensionality, examining the rating scale, testing the fit of items to the model, and in suggesting possible cutoff scores. These are also aspects of construct validation that can be facilitated with the Rasch model. The Rasch rating scale model (Wright & Masters, 1982), used for this analysis, estimates the probability that a respondent will choose a particular response category for an item as Pnij ln = Bn Di Fj, P ni( j 1) where P nij is the probability of respondent n scoring in category j of item i, P ni (j 1) is the probability of respondent n scoring in category j 1 of item i, B n is the person measure of respondent n, D i is the difficulty of item i, and F j is the difficulty of category step j. Rating scale categories are ordered steps on the measurement scale. Completing the jth step can be thought of as choosing the jth alternative over the (j 1)th in the response to the item. Rasch analysis places persons (B n ) and items (D i ) on the same measurement scale where the unit of measurement is the logit (log odds unit). Although person reliability in Rasch is analogous to Cronbach s alpha in true score theory (Nunnally & Bernstein, 1994), it is more conservative (usually lower) because it estimates standard errors (SEs) for each person and each item (vs. test-wide SE for alpha), thereby indicating how reliably persons and items are placed on the scale. Alphas are provided for those who prefer them, but we note that they tend to give high estimates due to the inclusion of zero and extreme high scores in their calculation. Using the Winsteps Computer Program for these calculations (Linacre, 2009), reliability estimates were calculated from 0 to 1.00 on scales that are actually infinite in either direction (Linacre, 2002). Dimensionality. Because the Rasch model requires unidimensionality, principal component analysis of residuals is used to examine whether a substantial factor exists in the residuals after the primary measurement dimension has been estimated (Linacre, 1998a; Smith, 2002). Although there are no hard rules for interpreting principal components results, our rule of thumb for unidimensionality was variance explained of greater than 40% by the measurement dimension (Linacre, 2006). For comparison, Reckase (1979) used 20% to define a substantial factor. To be conservative in testing a second dimension, we set less than 15% (even lower than Reckase) as the criterion for variance explained by the first principal component of the residuals, that is, the second dimension. Simply put, using 40% and 15% variance as the criteria for the first and second dimensions is a rigorous test in that the measurement dimension must be Vol. 50, No. 6,

5 large at 40%, whereas the second dimension must be quite small at under 15%. We also tested dimensionality using the procedure of Linacre (1998b). We extracted two subsets of items representing the opposite poles of the factor. We then measured each subject on each subset of items. We cross-plotted the subject measures and obtained correlation coefficients. Additional criteria for unidimensionality were employed using item fit statistics discussed next. Quality Control With Fit Statistics. Rasch analysis provides fit statistics to test assumptions of fundamental measurement (Wright & Stone, 1979). Fitting the model simply means meeting basic assumptions of measurement, for example, high scorers should endorse or get right almost all of the easy items. Once identified, persons and items that misfit can be examined qualitatively to determine the causes of the problems, which may include items with confusing wording or items that assess a construct that is different from the principal one being measured, that is, multidimensionality. Understanding poor fit can lead to improving or dropping items. A guide to interpreting fit statistics can be found at The Rasch model provides two indicators of misfit: infit and outfit for both persons and items (Wright & Stone). Person fit indicates the extent to which each person s performance is consistent with the way items are used by other respondents. Item fit indicates the extent to which the use of a particular item is consistent with the way sample respondents have answered other items. For this type of analysis, values under 1.33 are considered acceptable (Smith, 2000; Wilson, 2005). Low fit values, for example, <0.75 provide less motivation for item editing than do high values (Wilson), and they do not disturb the meaning of a measure though they may reduce precision (Linacre & Wright, 1994). We also used statistical significance (p <.05) as a criterion to examine items that should be dropped. Person fit statistics were examined to inform the clinical interpretation of the person measures but no persons were dropped. Rating Scale. The proper functioning of the rating scale was examined using: (a) fit statistics where outfit mean squares should be <2.0, (b) average measures advance monotonically with each category, and (c) step calibrations increase monotonically (Linacre, 1999, 2002; Zhu, 2002; Zhu, Updike, & Lewandowski, 1997). We did not expect the suspected category to perform as a typical rating scale category. We expected it to be used very rarely, but, based on qualitative input, it was important to include. A not applicable/don t know category was coded as missing data. The results tables are modified from Winsteps 3.67 (Linacre, 2009) with annotated explanations and interpretations. For an overview of Rasch analysis, see Conrad and Smith (2004); for a complete treatment, see Bond and Fox (2007); and Rasch Measurement Transactions at Construct Validation. In Rasch analysis, the item hierarchy that is created by the item difficulty estimates provides an indication of construct validity (Smith, 2001). Items should form a ladder with low severity symptoms on the bottom and high severity symptoms on the top. In our prior work developing the FE measure (Conrad et al., in press), 16 experts grouped the items into six categories and rated the severity of the items on a scale from 1 to 5. These item severities were then averaged within each category. The result was a theoretical hierarchy of six conceptual components of FE arranged in descending severity as follows (expert rating in parentheses): Theft and scams (4.31), financial victimization (4.20), financial entitlement (4.04), coercion (3.92), signs of possible abuse (3.27), and money management difficulties (1.94). We tested whether the hierarchy developed by the experts was validated compared with the client ratings given by respondents in the present study using the Rasch calibration on each item and averaging those within each group of items. Multitrait, Multimethod Analysis. Construct validation also may be tested by setting up a pattern of theoretical expectations and testing whether those expectations are supported by the data (Campbell & Fiske, 1959). As Campbell and Fiske pointed out, measures of the same construct should be highly correlated especially if they use the same method of observation. Measures Used in Construct Validation. The IDoA uses a questionnaire for elder abuse investigations that covers many forms of elder abuse, including FE. The IDoA form contains several sections that ask investigators to circle specific indicators of each type of suspected abuse. At the bottom of each 762 The Gerontologist

6 section, the staff member is asked to indicate if abuse is substantiated. The form also asks investigators for closing status on the case, identifying which types of abuse are substantiated. We examined the correlation of this closing status substantiation decision on FE with results from the OAFEM questionnaires. The following coding was used: 1) Client gender: men = 0, women = 1. 2) FE substantiation decision: we considered FE substantiated if the staff member coded it verified or some indication. For cases marked no indication or unable to verify, FE was considered not substantiated. 3) OAFEM staff: the staff reported OAFEM person reliability on 227 clients reported by the 22 staff was very high at 0.94 with a Cronbach s alpha of The Rasch item reliability was also very high at The final 82 items of staff-reported FE (similar but not the same as the 79 clientreported items) met stringent Rasch analysis fit and unidimensionality criteria. 4) OAFEM client: details are described in the Results section. The direction and strength of construct pairs depends on method and theoretical expectations. We set up a pattern of expected correlations roughly corresponding to the guidelines of Cohen (1988, 1992) as follows: NS = nonsignificant, <.1 = low, >.3 = moderate, and >.5 = high. Others have suggested lower values based on reviews of research, for example, >.2 = moderate and >.3 = high (Hemphill, 2003), so there are no absolute guidelines available. This hypothesized pattern and resulting correlations are in the upper right half of Table 4. The diagonal entries are the person reliabilities. The hypothesized correlations are stated above each correlation coefficient and are bulleted below: 1) Client gender: we had no reason to expect differential exploitation by gender so all gender correlations were expected to be NS. 2) FE substantiation decision: Moderate correlation with OAFEM and High correlation (because staff complete both) with Older Adult Mistreatment Assessment (OAMA) staff FE. 3) OAMA staff FE: High correlation (because similar method and questions) with OAFEM. In the multitrait, multimethod analyses, the most complete versions of OAMA measures were used. Results Of all 227 clients who completed the OAFEM self-report questionnaires via interview, 164 (72%) had at least some indication of FE based on IDoA verification decision criteria. 1. Test the fit of the items: items were dropped because they did not meet criteria for fit, that is, they had both infit and outfit >1.33, and/or point measure correlation, <.2. Because this was an iterative analysis, 3 of the original 82 items were dropped because they misfit, then the analysis was rerun. The remaining 79 items fit on the second and final run. 2. Test construct dimensionality: the variance explained in the remaining 79 items was 44.3%. This large amount, beyond the 40% criterion, was supportive of a strong principal measurement dimension. Moreover, unidimensionality was supported because the residual variance explained by the first contrast was very small 7.0% indicating no substantial rival dimension. The corresponding percentages for the measurement dimension and first factor of residuals respectively were 42.3% and 7.7% for the 54-item version and 45.2% and 10% for the 30-item version. The resulting correlations using the procedure of Linacre (1998b) were.79,.78, and.78 for the 79-, 54- and 30-item forms respectively, that is, strongly supportive of unidimensionality. 3. Assess internal consistency reliability using a standard of.80: the Rasch person reliability for the 79-item version was very high at 0.92 (alpha = 0.96). Similarly, Rasch item reliability was very high at Develop short forms that would be userfriendly for clinical applications: to test if a more parsimonious model would also function well, we developed two shorter forms containing 54 items and 30 items respectively (Appendix contains items by form information). Because both met stringent Rasch analysis fit and unidimensionality criteria, we report on the final 30-item instrument. The short form is viewed as most useful, but the longer forms provide a bank of items that may be useful in future development of alternative forms or computerized adaptive tests. Vol. 50, No. 6,

7 For the 54-item version, person reliability was 0.88 (alpha = 0.95). Rasch item reliability was The Rasch person reliability for the 30-item form remained high at 0.85 (corresponding to Cronbach s alpha of 0.93). The Rasch item reliability was also very high at Examine the appropriateness of the measure for the target population: in Figure 1, the Rasch ruler for the 30-item version is displayed. On the far left is the measurement scale in logits ranging from 3 to +3. Persons (n = 227) are arrayed on the left of the dashed line (representing the ruler) and the 30 items on the right. By convention, the item mean is the 0 point on the ruler. The items form a hierarchy of severity based on frequency of endorsement with lower severity items (more frequently endorsed) at the bottom and higher severity items (less frequently endorsed) at the top. The persons are displayed according to their measures on the OAFEM scale with low scorers (low FE) at the bottom and high scorers (high FE) at the top. There is a substantial floor of persons at the bottom who are not registering any FE. This was expected because all staff substantiated elder abuse clients were accepted, whether or not they were substantiated specifically for FE. Although the persons in the floor were included on the map, they were not included in the calculation of the person mean (M = 0.79 on the left side of the ruler) and standard deviation (SD = 1.02). We interpreted this as reasonably well targeted because the person mean was within one logit and about one item SD of the item mean of zero. 6. Test construct validity by positing a hierarchy of concept rankings and a set of hypothesized relationships: looking at Table 3, Original concept group, the ordering of the FE conceptual components was the same in four out of six cases for experts, averaging their concept map ratings, and clients, averaging their Rasch measurement calibrations. The first difference between experts and clients was coercion that jumped in rank from fourth most severe as rated by experts to second most severe based on client endorsements. Clients ranked abuse of trust fifth whereas experts had ranked it second. However, these differences are negated by the fact that the model SE was Therefore, the differences in the rankings were not statistically significantly different among the four concepts in the middle of the expert hierarchy. Specifically, they ranged from abuse of trust at 0.13 to signs of possible abuse at 0.07 to financial entitlement at 0.03 to coercion at In other words, using the client rankings, there was a statistical four-way tie among the concepts for second place. Rasch results indicated that 79-, 54- and 30-item versions formed a unidimensional overarching measure of client-reported FE. Therefore, rather than separating FE into several separate dimensions, the Rasch results suggested a single hierarchy that could be conceptualized according to its severity levels. The structure of the client data as seen in the Wright map presented a simpler picture of FE than the expert groups and ratings. Four groups were identified consisting, in descending order of severity, of major theft and scams (MT), lesser theft and scams (LT), risk (dropped from the 30-item form), and entitlement and expectations (EE). Examination of the Wright map (Figure 1) indicates these severity groups using the two-letter suffixes above. These were similar to the experts concept groups except that signs and risk factors were excluded from the short forms because they do not connote actual exploitation, and the coercion items were incorporated into the other groups because they were dispersed throughout the hierarchy. The validity of these groups is supported because, on average, they are located over one SE (SE = 0.36) from each other, that is, average item calibrations for MT = 0.48, LT = 0.20, and EE = We can see that MT and EE are nearly 3 SEs apart. Regarding correlational validation, we hypothesized that all gender correlations would be NS and they were (Table 4). The other three correlations, two high and one moderate, were as hypothesized that was supportive of the OAFEM s validity. 7. Identify an appropriate cutoff to determine FE: because there is no solely empirical way to determine a cut-point, we discuss the logic of our cut-points below. Discussion This study used Rasch item response theory and traditional validation approaches to examine the psychometric properties of the OAFEM. Previous work with expert-developed concept maps of FE, focus groups, and cognitive interviews resulted in an 82-item 764 The Gerontologist

8 Figure 1. Ruler of Rasch measurement person and item hierarchies (item numbers keyed to Appendix). Persons n = 227 and items n = 30. questionnaire administered in the present study by APS staff (n = 22) to substantiated clients (n = 227) in seven agencies in Illinois. Because no distinct empirically validated measure of FE exists, the OAFEM offers an important tool for elder abuse research as well as to practitioners working in the field. The OAFEM met stringent Rasch model criteria for item fit and unidimensionality; it had high internal Vol. 50, No. 6,

9 Table 3. Expert Item Groups and Rankings Compared With Client Rankings Original concept group (expert order) Average measure a (client order) Theft and scams (1) 0.42 (1) Abuse of trust (2) 0.13 b (5) Financial entitlement (3) 0.03 (3) Coercion (4) 0.03 b (2) Signs of possible abuse (5) 0.07 (4) Risk factors (6) 0.34 (6) a Based on the client endorsement of the items with a model SE of b Abuse of trust and coercion were the only concepts whose Rasch average measure was out of order with the original expert ranking, but stability of the rankings is in question because their average calibrations are so close together, that is, within 1 SE. consistency and item reliability. As a unidimensional measure of FE, it was found to have levels of severity rather than distinct subdimensions. These ranged from major theft at the high end to lesser theft in the middle to expectations and entitlement at the low end. This severity hierarchy helps us to develop a suggested cutoff score, though it is admittedly speculative at this early stage. If we look at Figure 1, the person item map, 173 persons endorsed at least one item on the 30-item OAFEM. Note that above 1.0 on this ruler, the item meanings and locations indicate that this may be a useful cutoff score. In other words, this level indicates more serious violations such as unexplained disappearances of the elder s possessions and alleged abuser lying about spending the elder s money. Above this 1.0 level were 102 persons. These persons had a score of 12 or more of a possible 60 raw score points on the measure. If we use a higher severity of 0 on the ruler as the criterion for serious FE, there were 41 persons above this level having even more severe symptomatology, that is, mostly major theft and scams. These symptoms included alleged abuser forcing the older adult into signing legal documents and pressuring the older adult to modify their will. Therefore, using cutoffs suggested by the client-reported empirical hierarchy, there were 102 persons who reported clinically significant FE. Of these, 41 were suffering from very severe FE. Study Strengths and Limitations This study developed the largest known database of substantiated clients of elder abuse to test the validity of the OAFEM and to suggest cutpoints for judgments of severity. Although it had well-targeted clients, expert interviewers, and modern measurement techniques, it was limited to the small geographic area of Cook and surrounding counties in Illinois. Although several validity tests were applied, many more can be imagined in other areas and populations. The groups and cutpoints suggested here were based on the logic of the Wright map hierarchy, but are otherwise speculative, and will require further replication and validation with external criteria to refine them, for example, using bank records and other financial documents, and sensitivity specificity analysis once a cut-point is defined. Conclusions Table 4. Hypothesized and Actual Correlations a These measures, used appropriately as long and short forms, should help to open the neglected area of FE of older adults for improved services and Client gender FE substantiation decision (IDoA) OAMA staff FE OAFEM Client gender NS NS NS M = 0, F = FE substantiation decision High Moderate (IDoA).544 **.310 ** OAMA staff FE.94 b High.734 ** OAFEM.92 Note: FE = financial exploitation; IDoA = Illinois Department on Aging; OAFEM = older adult financial exploitation measure; OAMA = Older Adult Mistreatment Assessment. a Hypothesized correlations: NS = non-significant, >.1 = low, >.3 = moderate, and >.5 = high are listed above the actual correlations. b Person reliabilities of OAMA scales are located on the diagonal. ** Correlation is significant at the 0.01 level (two tailed). 766 The Gerontologist

10 research. They should improve the understanding of prevalence by offering researchers a tested approach to the measurement of FE as well as enabling more accurate self- and third-party reporting. Using the Wright map, the measures provided theoretically supportable gradations along the continuum of abuse severity that can enable better decision making. Improved measurement will also enable practitioners to screen clients more efficiently, systematically, and precisely, so that, with the development of cutoff scores, cases may be triaged more effectively into appropriate interventions. Funding National Institute of Justice (2006-MU-MU-0004). Acknowledgments Points of view are those of the authors and do not necessarily represent the position of the U.S. Department of Justice. We appreciate cooperation for data collection by the Illinois Department on Aging and the participating APS agencies in the Chicago area. References AARP International. (2006). Keeping the wolves from grandma s door: Financial exploitation of the elderly. New York: International Network for the Prevention of Elder Abuse, World Elder Abuse Awareness Day, United Nations. Anetzberger, G. (2000). Caregiving: Primary cause of elder abuse? Generations, 24(2), Bond, T. G., & Fox, C. M. (2007). Applying the Rasch Model: Fundamental measurement in the human sciences (2nd ed.). Mahwah, NJ: Lawrence Erlbaum Associates. Campbell, D. T., & Fiske, D. W. (1959). Convergent and discriminant validation by the multitrait-multimethod matrix. Psychological Bulletin, 56, Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Mahwah, NJ: Lawrence Erlbaum Associates. Cohen, J. (1992). A power primer. Psychological Bulletin, 112, Conrad, K. J., Iris, M., & Ridings, J. W. (2009). Conceptualizing and measuring financial exploitation and psychological abuse of elderly individuals. Final report to the National Institute of Justice. Project #2006-MU-MU Washington, DC: National Criminal Justice Reference Service # Conrad, K. J., Ridings, J. W., Iris, M., Fairman, K. P., Rosen, A., & Wilber, K. H. (in press). Concept maps and models for measuring financial exploitation of older adults. Journal of Elder Abuse and Neglect. Conrad, K. J., & Smith, E. V. (2004). International conference on objective measurement: Applications of Rasch analysis in health care. Medical Care, 42(Suppl. 1), 1 6. Cooper, C., Selwood, A., & Livingston, G. (2008). The prevalence of elder abuse and neglect: A systematic review. Age and Ageing, 37, Embretson, S. E., & Reise, S. P. (2000). Item response theory for psychologists. Mahwah, NJ: Lawrence Erlbaum Associates. Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). Mini-mental state. A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 12, Godkin, M., Wolf, R., & Pillemer, K. (1989). A case-comparison analysis of elder abuse and neglect. International Journal of Aging and Human Development, 28, Hafemeister, T. L. (2003). Financial abuse of the elderly in domestic settings. In Panel to Review Risk and Prevalence of Elder Abuse and Neglect, Richard J. Bonnie, & Robert B. Wallace (Eds.), Elder mistreatment: Abuse, neglect, and exploitation in an aging America. Washington, DC: National Research Council of the National Academies, The National Academies Press. Hemphill, J. F. (2003). Interpreting the magnitudes of correlation coefficients. American Psychologist, 58, Kane, M., & Trochim, W. (2007). Concept mapping for planning and evaluation. Thousand Oaks, CA: Sage Publications. Kemp, B. R., & Mosqueda, L. A. (2005). Elder financial abuse: An evaluation framework and supporting evidence. Journal of the American Geriatrics Society, 53, Langan, J., & Means, R. (1996). Financial management and elderly people with dementia in the U.K.: As much a question of confusion as abuse? Ageing and Society, 16, Linacre, J. M. (1998a). Detecting multidimensionality: Which residual data-type works best? Journal of Outcome Measurement, 2, Linacre, J. M. (1998b). Structure in Rasch residuals: Why principal components analysis (PCA)? Rasch Measurement Transactions, 1, 636. Linacre, J. M. (1999). Investigating rating scale category utility. Journal of Outcome Measurement, 3, Linacre, J. M. (2002). Optimizing rating scale category effectiveness. Journal of Applied Measurement, 3, Linacre, J. M. (2006). Data variance explained by measures. Rasch Measurement Transactions, 20, Linacre, J. M. (2009). Winsteps. Chicago, IL: University of Chicago, MESA Press. Linacre, J. M., & Wright, B. D. (1994). Chi-square fit statistics. Rasch Measurement Transactions, 8, 350. National Committee for the Prevention of Elder Abuse. (2008). Financial abuse. Retrieved February 18, 2008, from use.org/elderabuse/fin_abuse.html National Elder Abuse Incidence Study. (1998). The national elder abuse inci dence study. Washington, DC: National Center on Elder Abuse at American Public Human Services. Retrieved February 21, 2008, from Abuse/docs/ABuseReport_Full.pdf National Research Council & Panel to Review Risk and Prevalence of Elder Abuse and Neglect. (2003). Elder mistreatment, abuse, neglect, and exploitation in aging America. R. Bonnie, & R. Wallace (Eds.). Washington, DC: National Academies Press. Nunnally, J., & Bernstein, I. (1994). Psychometric theory. New York: McGraw-Hill. Quinn, M. J., & Tomita, S. K. (1997). Elder abuse and neglect: Causes, diagnosis, and intervention strategies. New York: Springer Publishing Company. Rabiner, D., O Keeffe, J., & Brown, D. (2004). Financial exploitation of older person: Policy issues and recommendations for addressing them. Journal of Elder Abuse and Neglect, 16, Rasch, G. (1960). Probabilistic models for some intelligence and attainment tests. Copenhagen, Denmark: Danmarks Paedogogiske Institut. (Republished Chicago: The University of Chicago Press, 1980). Reckase, M. (1979). Unifactor latent trait model applied to multifactor tests: Results and implications. Journal of Educational and Behavioral Statistics, 4, Smith, E. V. (2001). Evidence for the reliability of measures and validity of measure interpretation: A Rasch measurement perspective. Journal of Applied Measurement, 2, Smith, E. V. (2002). Detecting and evaluating the impact of multidimensionality using item fit statistics and principal component analysis of residuals. Journal of Applied Measurement, 3, Smith, R. M. (2000). Common oversights in Rasch studies. MESA note 9. Retrieved March 15, 2008, from Trochim, W. (1989). An duction to concept mapping for planning and evaluation. Evaluation and Program Planning, 12, Wilber, K., & Reynolds, S. (1996). Introducing a framework for defining financial abuse of the elderly. Journal of Elder Abuse and Neglect, 8, Wilson, M. (2005). Constructing measures: An item response modeling approach. Mahwah, NJ: Lawrence Erlbaum Associates. Wright, B. D., & Masters, G. N. (1982). Rating scale analysis. Chicago: MESA Press. Wright, B. D., & Stone, M. H. (1979). Best test design. Chicago: University of Chicago, MESA Press. Zhu, W. (2002). A confirmatory study of Rasch-based optimal categorization of a rating scale. Journal of Applied Measurement, 3, Zhu, W., Updike, W. F., & Lewandowski, C. (1997). Post-hoc Rasch analysis of optimal categorization of an ordered response scale. Journal of Outcome Measurement, 1, Vol. 50, No. 6,

11 Appendix Scale and Item Information for Client Financial Exploitation. Subscale Name and Item Stem Concept Name Measure/Severity Group Results of Analyses In the past 12 months: (Responses: Yes=1, No=2, Suspected=3. Unknown=4) 1. Has felt entitled to use your money for themselves? AAFeltEntitledToUseEld$ForThemselves 2. Was name put on your bank account? AANameOnEldBankAccount 3. Have you been financially dependent on others for meeting your basic needs (food, rent, utilities, etc.)? EldFinanDepOnOthersForMeetingBasicNeeds 4. Have you had serious problems due to poor money management? EldHadSeriousProbsDueToPoor$Management 5. Have you felt pressured to stay in your current living arrangement because your money is needed for support of others? EPressureStayLivArrang/$NeedSuppOths 6. Has been financially dependent on you? AABeenFinanciallyDependentOnEld 7. Have you had to rely on someone else to cash your checks? EldRelyOnSomeoneElseToCashTheirChecks 8. Have you trusted someone else with your money? EldTrustedSomeoneElseWithEld s$ 9. Have you had a trusted other who might have reasons to exploit you (for example, someone who gambles, is unemployed, has substance abuse problems)? AAReasonsToExploit(Gambles/Unemployed) 10. Have you lacked information about financial affairs (for example, do not know where checkbook, ATM, or credit card is. Do not know ATM pin number, name of bank, direct deposit amount, or pension information)? ELackInfoFinanAff(CBook/ATM/CredCd/PIN) 11. Have you had mail piled up, including unpaid bills? EldHadMailPiledUp,IncludingUnpaidBills 12. Have you given access to your financial accounts? EGaveAAaccessTo$Acct 13. Have you received calls from bill collectors without knowing why? EReceivCallsFrmBillCollectNotKnowingWhy 14. Have you had trouble managing your money on your own? EldHasHadTroubleManaging$OnTheirOwn Measures based on Blank means included in 79 item analysis a 79 item analysis 3, Financial Entitlement 1.36/Entitlement Expectation Included in 54 Item Analysis 5, Signs of Possible Abuse.46/Risk Included in 54 Item Analysis 5, Signs of Possible Abuse.31/Risk 6, Risk Factors.26/Entitlement Expectation 5, Signs of Possible Abuse.36/Entitlement Expectation Included in 54 Item Analysis 5, Signs of Possible Abuse 1.56/Entitlement Expectation 6, Risk Factors.30/Risk Misfitting item Deleted from all analyses 5, Signs of Possible Abuse.63/Entitlement Expectation Included in 54 Item Analysis 6, Risk Factors.31/Risk 6, Risk Factors.02/Entitlement Expectation Misfitting item Deleted from all analyses 5, Signs of Possible Abuse.04/Lesser Theft & Scams Included in 54 Item Analysis 6, Risk Factors.58/Risk Appendix continued 768 The Gerontologist

12 Appendix (continued) Subscale Name and Item Stem Concept Name Measure/Severity Group Results of Analyses 15. Have you trusted someone new with your money? EldTrustedSomeoneNewWith$ 5, Signs of Possible Abuse.11/Risk 16. Have you made changes in your will or trust in favor of any family members or individuals? EChangeWill/TrustInFavorOfFamMemb/Indiv 17. Did give poor reasons for spending your money? PoorReas4SpndE$ 18. Were family members fighting over your money? EldFamMembersFightOverE$ 19. Have you written out any checks for someone else to cash? EWrittenOutChecksForOthToCash 20. Have there been any sudden changes in your financial management (titles are changed, quit claim deeds, retirements or investments cashed in)? SudChangeInE sfinanman(titles,deeds) 21. Has convinced you to turn the title of your home over to them? ConvTrnrOvrTitl 22. Even though you should have enough money, have you had unpaid bills, eviction notice, or utilities shut off? EldShdHaveEnough$ButUnpaidBillsEvictNoticeUtilShutOff 23. Have you thought was lying about how they were spending your money? AALyingAbtSpdE$ 24. Has refused to give you an accounting of spending your money? Refus2GivAcct$ 25. Have there been unexplained disappearances of your money or possessions? UnexplDisappPoss 26. Have you signed documents without understanding what you were signing? EldSignedDocsWithoutUnderstanding 27. Have your legal or financial documents been frequently changed? Leg$DoxFreqChng 28. Has persuaded you to sign any documents even though it was not in your best interest? SignDxNotBstInt 5, Signs of Possible Abuse.27/Risk 3, Financial Entitlement.58/Entitlement Expectation Short Form Item 3, Financial Entitlement.26/Entitlement Expectation Included in 54 Item Analysis 5, Signs of Possible Abuse.30/Risk 5, Signs of Possible Abuse.01/Major Theft & Scams 2, Abuse of Trust.53/Major Theft & Scams Short Form Item 5, Signs of Possible Abuse.14/Entitlement Expectation Included in 54 Item Analysis 4, Coercion.70/Entitlement Expectation Short Form Item 5, Signs of Possible Abuse.55/Entitlement Expectation Short Form Item 5, Signs of Possible Abuse.75/Lesser Theft & Scams Short Form Item 5, Signs of Possible Abuse.09/Lesser Theft & Scams 5, Signs of Possible Abuse.37/Major Theft & Scams Short Form Item 2, Abuse of Trust.03/Lesser Theft & Scams Short Form Item Appendix continued Vol. 50, No. 6,

13 Appendix (continued) Subscale Name and Item Stem Concept Name Measure/Severity Group Results of Analyses 29. Has obtained or changed a power of attorney? AAObtain/ChangedPOA 30. Has become the payee on your benefit check and used the money for themselves? PayeeOnChkUse4Sf 31. Has kept you from seeing your own mail? AAKeptEldFromSeeingOwnMail 32. Has anyone had a new interest in your assets? NewInterestInEldAssets 33. Has changed the direct deposit destination so as to benefit themselves? ChngDirDpDestin 34. Has had your bills sent to a different address (for example, the trusted other s address) so as to benefit themselves? AASentEBillsToDiffAddrBenefitSelf 35. Has used your bank pin or account number for their own gain? AAUsedEldBankPINOrAcctNumForOwnGain 36. Have you suspected someone of signing your name without permission? EldSuspectsSomeoneSignNameNoPermission 37. Has anyone been frequently asking you for money? EldFreqAskedFor$ 38. Has anyone kept asking about your money even though it made you feel uncomfortable? EldAskedAbout$FeltUncomfort 39. Did anyone put pressure on you to get a reverse mortgage? EldPressureToGetReverseMortgage 40. Have you not been getting the kind of care that you can afford? EldNotGettingCareTheyCanAfford 41. Has anyone misused your ATM or credit card? MisuseEldATM/CredCard 42. Was there unusual activity in your bank accounts, for example, large withdrawals, frequent withdrawals? UnusActvEBnkAcc 43. Did mix their funds with your funds? AAMixFundsWithEld 2, Abuse of Trust.06/Risk 1, Theft & Scams.49/Major Theft & Scams Short Form Item 4, Coercion.27/Major Theft & Scams Included in 54 Item Analysis 5, Signs of Possible Abuse.23/Major Theft & Scams 3, Financial Entitlement 1.24/Major Theft & Scams Short Form Item 3, Financial Entitlement.73/Major Theft & Scams 1, Theft & Scams.07/Lesser Theft & Scams 1, Theft & Scams.31/Lesser Theft & Scams Included in 54 Item Analysis 5, Signs of Possible Abuse.92/Entitlement Expectation Included in 54 Item Analysis 5, Signs of Possible Abuse.55/Entitlement Expectation Included in 54 Item Analysis 4, Coercion 1.11/Major Theft & Scams 1, Theft & Scams.04/Lesser Theft & Scams Included in 54 Item Analysis 1, Theft & Scams.02/Major Theft & Scams Included in 54 Item Analysis 5, Signs of Possible Abuse.24/Lesser Theft & Scams Short Form Item 5, Signs of Possible Abuse.24/Major Theft & Scams Included in 54 Item Analysis Appendix continued 770 The Gerontologist

14 Appendix (continued) Subscale Name and Item Stem Concept Name Measure/Severity Group Results of Analyses 44. Have there been unauthorized withdrawals from your bank account? UnathW/dEBnkAcc 45. Has tricked or pressured you into buying something that you now regret buying? PresBuyItmRgrtBuy 46. Have you felt cheated after sold something to you? EFeltCheatedAfterAASoldSomethingToE 47. Have you been a victim of a scam that involved giving to bogus charities? EVictimOfScamBogusCharities 48. Have you been tricked or pressured into selling a valuable possession for less than its worth? AATrick/PressureSellValuPossessLessThanItsWorth 49. Has an institution misused your funds? InstitutionMisusedEldFunds 50. Has a health care provider used your name to file false claims? EldHealthcareProvFiledFalseClaims 51. Were you talked into making investments that were not in your best interest? ETalkedIntoInvestmtNotInBestInterest 52. Have you been pressured to modify your will? EPrssr2ModWill 53. Has forced you to sign legal or financial documents? AAFrcESigLeg$Dx 54. Has coerced you to give them larger than usual gifts (money, cars, homes)? CoerceE2GivBigGft 55. Have you been pressured to co-sign any loans? EPrsrCoSigLoans 56. Have services been withheld by anyone, because they wanted more money? ServWithheldForMore$ 57. Have you been convinced to contribute beyond your means to churches or charities? EldContributeBeyondMeansToChurch/Charity 58. Has obtained a new credit card in your name? AAObtainedNewCredCardInEldName 59. Has used your money on themselves instead of for you? AAUsd$OnSlfNotE 1, Theft & Scams.05/Lesser Theft & Scams Short Form Item 1, Theft & Scams.32/Major Theft & Scams Short Form Item 1, Theft & Scams.95/Major Theft & Scams 1, Theft & Scams 1.20/Major Theft & Scams Included in 54 Item Analysis 2, Abuse of Trust.89/Major Theft & Scams Included in 54 Item Analysis 1, Theft & Scams 2.21/Major Theft & Scams Misfitting item Deleted from all analyses 3, Financial Entitlement.77/Major Theft & Scams Included in 54 Item Analysis 5, Signs of Possible Abuse.77/Major Theft & Scams Short Form Item 4, Coercion.11/Major Theft & Scams Short Form Item 4, Coercion.10/Major Theft & Scams Short Form Item 4, Coercion.57/Major Theft & Scams Short Form Item 4, Coercion.71/Major Theft & Scams Included in 54 Item Analysis 4, Coercion.92/Major Theft & Scams Included in 54 Item Analysis 1, Theft & Scams.63/Major Theft & Scams 2, Abuse of Trust.94/Entitlement Expectation Short Form Item Appendix continued Vol. 50, No. 6,

15 Appendix (continued) Subscale Name and Item Stem Concept Name Measure/Severity Group Results of Analyses 60. Has bought things for you but not given you back your change? AABoughtThingsEldNeverGaveBackChange 61. Has borrowed money and not paid it back? AABrrw$NotPdBck 62. Has said they were buying something for you, but it was really for their own use? AAPrch4EUsdSlf 63. Has taken your prized belongings (for example, jewelry) without permission? EldPrizedBelongingsTakenWithoutPermiss 64. Has anyone switched some of your expensive items for cheaper ones? ExpnsvSwtc4Cheap 65. Has overcharged you for work or services that were done poorly or never done? EOvch4WkSrvPoor 66. Did misuse their power of attorney or guardianship? AAMisusePOAOrGuardianship 67. Were you forced into making financial decisions you would not normally make? EForceIntoFinanDecNotNormMade 68. Did you think that has taken advantage of you to get a hold of your resources such as a house, car, or money? TkAdvntE4Resors 69. Has tried to prevent you from spending your money in order to maximize their inheritance? PrvntESpnd2MaxInhrt 70. Has lived with you, but refused to pay their share of expenses? AALivedWithEldRefusePayExpenses 71. Has taken your money to do something for you but never did? Tk$2DoSmthgNvrDid 72. Has handled your money irresponsibly, for example, gambling, illegal activities? HndlE$Irrspnsbly 73. Has used love (sex, or intimacy, if applicable) to gain control of your money? AAUsedLove(Sex/Intimacy)ToGainContOfE$ 74. Has promised companionship in exchange for your money? AAPromisECompanionshipExchFor$ 2, Abuse of Trust.45/Entitlement Expectation Included in 54 Item Analysis 2, Abuse of Trust 1.13/Entitlement Expectation Short Form Item 2, Abuse of Trust.06/Lesser Theft & Scams Short Form Item 1, Theft & Scams.06/Major Theft & Scams Included in 54 Item Analysis 1, Theft & Scams.90/Major Theft & Scams Short Form Item 1, Theft & Scams.46/Major Theft & Scams Short Form Item 2, Abuse of Trust.19/Major Theft & Scams Included in 54 Item Analysis 4, Coercion.30/Lesser Theft & Scams 2, Abuse of Trust.59/Entitlement Expectation Short Form Item 2, Abuse of Trust.14/Major Theft & Scams Short Form Item 3, Financial Entitlement.59/Entitlement Expectation Included in 54 Item Analysis 1, Theft & Scams.35/Entitlement Expectation Short Form Item 2, Abuse of Trust.06/Lesser Theft & Scams Short Form Item 4, Coercion.12/Major Theft & Scams Included in 54 Item Analysis 4, Coercion.27/Major Theft & Scams Appendix continued 772 The Gerontologist

16 Appendix (continued) Subscale Name and Item Stem Concept Name Measure/Severity Group Results of Analyses 75. Has demanded money from you? AADemanded$FromE 4, Coercion.66/Entitlement Expectation Short Form Item 76. Has used pressure, intimidation, or punishment to 4, Coercion.49/Entitlement Expectation try to get your money? AAPressure/Intimid/PunishToGetE$ 77. Did you let spend your money on themselves 4, Coercion.12/Entitlement Expectation Short Form Item because you were afraid of them? Sp$SfBcEAfraid 78. Has taken advantage of cultural or family expectations to get your resources? TkAdvtCult/FamExp 79. Has promised you lifetime care but then did not provide it? PrmLifCarNvrPrvd 80. Has not met your basic needs even though you had enough income? EldBasicNeedsNotMetThoughEnoughIncome 81. Has felt entitled to use your money for him/herself? AAFltEntitl2E$ 82. Have caregivers overcharged you for their services? CaregiversOverchargeEldForServ 4, Coercion.04/Entitlement Expectation Short Form Item 2, Abuse of Trust.20/Major Theft & Scams Short Form Item 1, Theft & Scams 32/Lesser Theft & Scams 3, Financial Entitlement 1.00/Entitlement Expectation Short Form Item 1, Theft & Scams.51/Major Theft & Scams Vol. 50, No. 6,