Explain how a forensic psychology professional might use the results from this study in a correctional setting.

• Briefly summarize the research article you selected.
• Explain how the potential for violence was assessed and the important data, results, and “key takeaways” from the study.
• Explain how a forensic psychology professional might use the results from this study in a correctional setting.

The Suicide (SPI) and Violence Potential Indices (VPI) from the Personality Assessment Inventory: A Preliminary Exploration of Validity in an Outpatient Psychiatric Sample

Samuel Justin Sinclair & Iruma Bello & Maren Nyer & Jenelle Slavin-Mulford & Michelle B. Stein & Megan Renna & Daniel Antonius & Mark A. Blais

Published online: 9 February 2012 # Springer Science+Business Media, LLC 2012

Abstract Assessing risk of harm to self and others is an important component of psychological assessment, although methods for risk assessment vary considerably. The Person- ality Assessment Inventory (PAI) is frequently administered to evaluate general psychological functioning, as well as to provide information about suicide and violence risk. The purpose of this study was to evaluate the construct validity of the PAI Suicide (SPI) and Violence Potential (VPI) indi- ces in a sample of 158 psychiatric outpatients referred for psychological and neuropsychological assessment within a large northeastern academic medical center between 2007 and 2011. Results generally supported the convergent and divergent validity of both SPI and VPI when evaluating groups with and without a history of suicide and violence risk, and effect sizes were moderate to large even after controlling for other covariates. SPI and VPI scores were also found to vary significantly across different psychiatric groups in ways that would be expected. Finally, we explored the relationship between SPI and VPI, and executive

functioning impairment—a neuropsychological variable found to be associated with impulsive self and other- harming behaviors. Consistent with prior research, SPI and VPI were found to be significantly elevated in groups dem- onstrating executive dysfunction. The implications of these findings and specifically the utility of using SPI and VPI in the assessment of risk are discussed.

Keywords Personality assessment inventory . Risk assessment . Suicide potential . Violence potential

Psychologists and other healthcare providers are increasingly being asked to provide clinical judgments regarding a person’s level of risk of harm to self and others in both clinical and non- clinical settings. As some have noted previously, current frameworks for evaluating risk are varied and often imprecise, which when coupled with low base rates of suicide and aggression in the general population results in prediction models that are frequently unreliable (Monahan et al. 2001; Wenzel, et al. 2011). Further, research over the last decade has demonstrated that factors underlying risk for suicide and aggression are themselves complex, and include a wide array of affective, neuropsychological, biological, demographic, and contextual variables (Dougherty, et al. 2004; Jollant et al. 2005; Keilp, et al. 2001; Monahan et al. 2001; Monahan and Steadman 1994; Soloff, et al. 2000; van Heering and Marusic 2003; Wenzel et al. 2011). As a result, accurately predicting the near-term risk of suicidal and violent behaviors can prove to be challenging.

Despite these complexities, both suicide and violence con- tinue to be prominent public health issues. For example, The National Institute of Mental Health (NIMH) reported in 2007 that suicide was the tenth leading cause of death within the United States overall and the third leading cause of death for

An earlier version of this paper was presented at the annual meeting of the Society for Personality Assessment, Boston, MA, March 2011.

S. J. Sinclair (*) : I. Bello :M. Nyer : J. Slavin-Mulford : M. B. Stein :M. Renna :M. A. Blais Department of Psychiatry, Massachusetts General Hospital and Harvard Medical School, One Bowdoin Square, 7th Floor, Boston, MA 02114, USA e-mail: SJSinclair@Partners.org

D. Antonius University at Buffalo, State University of New York, Buffalo, NY, USA

D. Antonius New York University School of Medicine, New York, NY, USA

J Psychopathol Behav Assess (2012) 34:423–431 DOI 10.1007/s10862-012-9277-6

people ages 15 to 24, accounting for roughly 34,598 fatalities overall (NIMH 2011). Further, suicide rates have also been found to vary significantly across groups differing in terms of age, gender, and race/ethnicity. Men, those in adolescence and older age groups, and American Indians and Non-Hispanic Whites are disproportionately more likely to die by suicide. Those with mental illness (and primary affective/mood disor- ders specifically) and substance abuse disorders are also con- sidered to be at elevated risk (NIMH 2011).

With respect to the issue of violence, the US Federal Bureau of Investigation (FBI) reported 1,318,398 violent crimes in the United States in 2009, or 492.4 crimes per 100,000 citizens; the most frequent of these crimes was aggravated assault (FBI 2011). Within psychiatric popula- tions specifically, Monahan et al. (2001) found that those with substance abuse and prior violence histories, male gender, and psychopathy were risk factors for prospective violence direct- ed at others. Interestingly, Monahan et al. (2001) reported that major mental illness and especially psychotic illness were associated with lower rates of violence, although other re- search on this has been inconsistent.

While studies on this have been somewhat mixed, other cognitive/neuropsychological variables have also been found to be associated with both suicidal and violent behavior. For example, Voracek (2004, 2009) found a significant, positive correlation between intellectual functioning and rates of sui- cide in both men and women in an international study of 85 countries. Voracek discussed this relationship in the context of evolutionary theory, and specifically “that a threshold intelli- gence is necessary for suicidality” (Voracek 2004, p. 544). In contrast, other research has generally supported an inverse relationship between cognitive functioning on the one hand, and behavioral disinhibition and antisocial behaviors on the other hand (Farrington 2006; Neumann and Hare 2008). Some have argued that this relationship is accounted for by more focal neuropsychological deficits within the executive function domain, which has also been associated with both suicidal and violent behaviors (Harkavy-Friedman et al. 2006; Jollant et al. 2005; Marzuk, et al. 2005; Morgan and Lilienfeld 2000; Westheide et al. 2008). Although “Executive Function” refers to an array of cognitive abilities which implicate a network of cortical and sub-cortical structures in the brain, these findings have typically been explained in terms of how executive dysfunction specifically may predispose people to greater levels of disinhibition, impulsivity, cognitive rigidity, and re- duced capacity for complexity and effective decision-making.

Risk Assessment with the Personality Assessment Inventory

Although a myriad of risk assessment paradigms exist today, the Personality Assessment Inventory (PAI; Morey 1991,

2007) has been increasingly used to evaluate common forms of psychopathology and interpersonal style, as well as factors that impact the course of treatment and are associated with risk (e.g., aggressiveness, suicidal ideation). The PAI has been utilized in forensic settings to facilitate the assessment of dangerousness to self and others, treatment amenability, and classifying offenders (Edens, et al. 2001; Morey and Quigley 2002). Similarly, the PAI has also demonstrated predictive utility for determining risk of recidivism in inmates released from custody (Walters and Duncan 2005) and disciplinary issues during incarceration (Walters, et al. 2003).

Specifically, the PAI is a 344-item self report instrument assessing general psychological functioning and interper- sonal style (Morey 1991, 2007). Research has demonstrated its psychometric adequacy across a wide variety of clinical and non-clinical settings (Boone 1998; Braxton, et al. 2007; Deisinger 1995; Holden 2000; Morey 1991; Morey 2007; Siefert, et al. 2009; Sinclair et al. 2009; 2010). Its four validity scales have also been found to be effective in detecting invalid and inconsistent response styles (Sellbom and Bagby 2008). The PAI was developed to assess multiple relevant domains of psychopathology including: Somatic Complaints (SOM), Anxiety (ANX), Anxiety-Related Disorders (ARD), Depression (DEP), Mania (MAN), Paranoia (PAR), Schizophrenia (SCZ), Borderline Features (BOR), Antisocial Features (ANT), Alcohol Problems (ALC) andDrug Problems (DRG) (Morey 2007).

In addition to these self-report scales, a number of sup- plemental indices were also developed by Morey (1996) to assess for other clinical factors, including suicide and vio- lence potential specifically. The Suicide Potential Index (SPI) was constructed by identifying the 20 features of the PAI profile that have been found in the literature to be most associated with completed suicide. These include factors such as elevated affective distress, alcohol and drug abuse, mistrust, social withdrawal, insomnia, impulsivity, anger, mood fluctuations, among others. The SPI is scored by summing these 20 indicators and converting them into corresponding T-scores. Preliminary research by Morey (1996) suggests that SPI has been found to be associated with whether someone is on suicide precautions, has attemp- ted suicide in the past, and level of care. More recently, Breshears, et al. (2010) additionally reported that the SPI was a strong predictor of suicidal behavior among veterans with head injuries.

Similarly, the PAI Violence Potential Index (VPI) was developed by taking the 20 features of PAI profile that have been found to be most associated with violence and danger- ousness. These include variables such as anger, hostile control in relationships, sensation-seeking, impulsivity, agitation, an- tisocial behavior, grandiosity, and alcohol and drug abuse, among others. The VPI is scored by summing these 20 indi- cators and deriving corresponding T-scores, and has been

424 J Psychopathol Behav Assess (2012) 34:423–431

found to be associated with whether someone has a history of assault, has been convicted of violent crime, or is on assault precautions (Morey 1996). Similarly, research has shown small to moderate-sized effects for VPI in predicting violent nonsexual recidivism, nonvio- lent recidivism, and sex offender registry violations (Boccaccini, et al. 2010). That being said, other research has shown limited incremental validity for the VPI in differentiating those who reported interpersonal violence in the last year when compared to the Aggression scale, which performed somewhat better in terms of its effect (Crawford, et al. 2007).

A more recent study by Hopwood, et al. (2008) evaluated the construct validity of SPI and VPI (among other PAI scales) in a large sample of people who were court- mandated to attend a substance abuse treatment program in the United States. Of note, they found that while average VPI scores were not significantly different across groups with and without a history of assault or major rule infrac- tions within the treatment facility, mean SPI scores were significantly elevated in groups with a history of prior suicide attempts versus those without. However, the highly specific nature of the sample and possible other confounds (e.g., effects of demographic and other clinical variables) limit the extent to which these findings generalize to other clinical and non-clinical populations.

Study Purpose

Despite an emerging body of research on the validity and reliability of the PAI, the vast majority of this work has focused almost exclusively on the clinical scales, and much less is known regarding the properties of the supplemental indices including the SPI and VPI. Given the unique manner in which these indices were developed and potential added clinical utility in terms of evaluating for risk of harm to self and others, the specific purpose of this study was to inves- tigate the convergent and divergent validity of the SPI and VPI indices across groups varying in terms of their risk histories (i.e., those with histories of suicide attempts, vio- lence histories, etc). It was hypothesized that the SPI would yield greater effect sizes in differentiating groups with and without a history of suicidal ideation, suicide attempts, and inpatient psychiatric hospitalizations, whereas the VPI would better differentiate those with and without a history of violence and arrests.

In light of research that has also found varying rates of suicidal behaviors and violence across different psychiatric groups, the second purpose of this study was to evaluate whether SPI and VPI scores varied as a function of psychi- atric history. Although many forms of psychiatric diagnoses have been found to be associated with elevated risk of self-

harm and violence directed at others, primary mood (depres- sive, bipolar) disorders have been found to be strong pre- dictors of suicidal behaviors and primary substance abuse disorders have been found to be related to elevated rates of violence specifically (Dougherty et al. 2004; Jollant et al. 2005; Keilp et al. 2001; Monahan et al. 2001; Monahan and Steadman 1994; Soloff et al. 2000; van Heering and Marusic 2003; Wenzel et al. 2011). Given this literature, it was hypothesized here that SPI would yield larger effect sizes in differentiating those with and without primary mood disorders, while VPI would be more sensitive to those with and without a history of alcohol and drug abuse. Because the literature has been somewhat inconsistent with respect to the relationship between psychotic disorders and propensity for suicide and violence (Monahan et al. 2001; Monahan and Steadman 1994), it was expected that both SPI and VPI would yield small effect sizes in differentiating those with and without histories of psychotic symptoms.

Finally, given the multiple studies that have shown a relationship between executive dysfunction and elevated rates of impulsivity (including suicidal and violent behaviors), this study sought to test whether SPI and VPI scores varied across groups with and without exec- utive functioning impairment. In light of this research, it was hypothesized that SPI and VPI would both be sen- sitive in differentiating groups with and without execu- tive dysfunction.

Method

Participants

Study participants were 158 psychiatric outpatients referred for psychological and neuropsychological assessment within a large northeastern academic medical center between 2007 and 2011. The mean age for the sample was 42.0 years (SD014.6; range019 to 82 years). The majority of the sample was Caucasian (91%), male (55%), and right-handed (85%); and the average education in years was 14.3 (SD03.0). Just over one-quarter of the sample (29%) indicated they were married at the time of the assessment, while the majority reported never being married (57%). Roughly half of the sample was unemployed (53%); 42% reported they were working; 5% said they were retired; and 12% noted they were enrolled in an educational program of some kind.

All participants were engaged in psychiatric care at the time of the assessment, and were referred by their mental health providers for psychological and neuropsy- chological testing with the goal of providing greater diagnostic clarity and recommendations for treatment. As part of the evaluation process, diagnostic information was collected from the referring provider and medical

J Psychopathol Behav Assess (2012) 34:423–431 425

record prior to the evaluation (i.e., diagnoses made by referring providers) and the breakdown is as follows: Major Depressive Disorder (44%), Bipolar Disorder (22%), Anxiety Disorders (14%), Cognitive Disorders such as ADHD (7.2%), Psychotic Disorders (2%), Substance Abuse Disorders (4.6%), Somato- form Disorders (0.7%), Adjustment Disorders (0.7%), and other Axis I Disorders (4.8%).

Procedure

All assessments were conducted by licensed clinical psy- chologists in the clinic, or psychology post-doctoral fellows or pre-doctoral interns under their supervision. Assessment data from the clinic were entered into a de-identified data repository approved by the hospital’s Institutional Review Board (IRB). Included in the data repository are basic de- mographic information about the patients; clinical and diag- nostic information that was collected from the medical record, clinical interview with the patient, and interviews with other sources (e.g., spouses, friends, family members); and all psychological and neuropsychological test scores. The PAI and other clinical data reported in this study were drawn from this IRB-approved data repository.

Measures

The Personality Assessment Inventory The PAI is a 344- item self report measure of general psychological function- ing, which was developed to screen for various psychiatric disorders, alcohol and drug abuse, factors associated with treatment amenability, and interpersonal style (Morey 1991, 2007). The instrument contains 4 validity scales, 11 clinical scales (assessing constructs such as depression, anxiety, mania, schizophrenia, etc.), 5 treatment consideration scales (measuring factors such as lack of social support, amenabil- ity to treatment), and 2 interpersonal scales assessing dom- inance/assertiveness and warmth/affiliation. All items are rated along a 4-point Likert scale (False, Slightly True, Mainly True, Very True). In addition to these specific scales, a number of supplemental indices were also developed by Morey (1996) to assess for suicide and violence potential, among other factors. As noted above, the SPI and VPI indices were constructed using the 20 features of the profile that have been found to be most associated with suicidal and violent behavior (e.g., substance abuse, impulsivity, affec- tive dysregulation, etc.), and are scored by summing these indicators and deriving corresponding T-scores. Normative data were collected by the test developers to assist with interpretation using a US-census matched community sam- ple (N01,000) and a clinical sample (N01,246).

The Wechsler Abbreviated Scale of Intelligence (WASI) The WASI was developed in 1999 for purposes of providing a

brief yet reliable method for evaluating intelligence in pop- ulations ranging in age from 6 to 89 (The Psychological Corporation 1999). Similar to the longer Wechsler Adult Intelligence Scale (WAIS-III), the WASI derives estimates for Full-Scale IQ, Verbal IQ, and Performance IQ. The WASI consists of the four subtests from the WAIS-III which have been shown to load the strongest on g, or general intellectual functioning, as well as their relative relation- ships to both verbal and performance-based cognitive abil- ities: Vocabulary, Block Design, Similarities, and Matrix Reasoning (Wechsler 1997). The WASI takes approximately 30 min to administer.

Executive Functioning Measures In the present study, exec- utive functioning was evaluated using three commonly administered instruments: The Wisconsin Card Sorting Test (WCST; Heaton, et al. 1993), the Trail Making Test Part B (Reitan and Wolfson 1985), and the Stroop Neuropsycholog- ical Screening Test (Trenerry, et al. 1989). The WCST specif- ically requires examinees to sort cards according to the color, shape, and number of the objects represented, while also changing the sort criteria over the course of the test—which necessitates a capacity for set-shifting and cognitive flexibil- ity. In the current study, the number of sets established, as well as the number of total and perseverative errors were aggregat- ed (averaged) into a single WCST index score. Similarly, the Trail Making Test Part B requires examinees to draw lines through numbers and letters in alternating and ascending fashion, and is considered a test of cognitive flexibility and set-shifting. The total time to complete the task was used for purposes of deriving an overall score. Finally, the Stroop Neuropsychological Screening Test requires an individual to initially read 112 words written in different ink colors, and then on a second list of words they are asked to inhibit the natural reading response and give a dissonant response (name the color of the ink) instead. The number of correct responses within the 2-minute time limit was used to derive the score for the test. Patients’ scores on all three of these tests were converted into standard scores using existing normative data for each instrument to have a mean of 100 and standard deviation of 15.

Data Analyses

The construct validity of the PAI Suicide (SPI) and Violence Potential Indices (VPI) was evaluated in several different ways. First, mean SPI and VPI scores were estimated across groups with and without (i.e., dichotomous groups) a history of suicidal ideation, suicide attempts, inpatient psychiatric hospitalization, violence/assaultive behavior directed at others, and any form of arrest (including assault and other violent offenses).

426 J Psychopathol Behav Assess (2012) 34:423–431

Second, average SPI and VPI scores were estimated across different psychiatric groups, as defined by the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR, Amer- ican Psychiatric Association 2000). Groups included those with and without a history of a major depressive episode, manic episode, primary psychotic episode (e.g., hallucina- tions, delusions, paranoid ideation), alcohol abuse, and drug abuse. These groups were defined based on both a thorough review of the patient’s medical record prior to the clinical assessment, as well as a semi-structured clinical interview by the assessment psychologist to evaluate whether the patient had ever met DSM-IV criteria for any of these psychiatric conditions in the past. Dichotomous groups were again con- structed based on these criteria.

Finally, mean SPI and VPI scores were calculated for groups exhibiting any form of executive dysfunction versus those with no impairment. Dichotomous groups were created based on whether or not patients exhibited cognitive impair- ment on any of the three measures of executive function detailed above (i.e., the Wisconsin Card Sorting Test, Trail Making Test Part B, or the Stroop Neuropsychological Screening Test). For purposes of this study, impairment was defined as a standard score below 70, which is similar to how other studies have categorized cognitive deficit (e.g., see Harvey, et al. 2009). Such a score falls two standard deviations below the normative mean and is traditionally accepted as the definition of impairment (Lezak, et al. 2004). Subjects with any executive functioning score below a standard score of 70 were assigned to the executive dysfunction group (N047).

Univariate tests (i.e., independent samples t tests and Pearson correlations) were first conducted to evaluate the relationships between SPI and VPI, and other model variables (e.g., age, FSIQ, gender, race, and executive dysfunction). Second, least-squared adjusted means and effect sizes (i.e., Cohen’s d’s) were estimated for each of these group comparisons using the General Linear Modeling procedure in SPSS Version 16. Because re- search has demonstrated varying effects of gender, age, race, intellectual functioning, and other socioeconomic indicators on the potential for suicide and violence risk (detailed above), all Analysis of Covariance (ANCOVA) models were constructed to control for the effects of age, gender, education (dichotomized as a high school educa- tion or greater versus those with less than high school education), race (white versus non-white), and full scale IQ (derived from the WASI). Adjusted mean SPI and VPI scores and cohen’s d estimates were then estimated after accounting for these covariates to determine whether differences across groups were statistically significant, as well as to assess the magnitude of the effect. Cohen (1988) suggested d values of 0.2 as indicating a small effect, 0.5 as a moderate effect, and 0.8 and greater as a large effect.

Results

Descriptive analyses were first conducted to evaluate the linear relationships between SPI and VPI, and other varia- bles included in the models (presented below). First, SPI and VPI were found to correlate significantly (r00.740; p0 0.001). Second, both SPI (r0−0.239; p00.002) and VPI (r0−0.280; p<0.001) were found to be significantly corre- lated with Full Scale IQ, and VPI was significantly related to age (r0−0.221; p00.005) while SPI was not. Finally, inde- pendent samples t tests indicated that while mean SPI and VPI were not associated with gender or race, both were associated with education level and executive dysfunction. Specifically, mean SPI (M071.2; SD015.8) and VPI (M065.1; SD017.7) scores were both significantly (i.e., at the p<0.05 level) elevated in groups with a high school education or less, as compared to those who achieved higher levels of education (M062.5; SD013.4 & M054.8; SD0

Table 1 Adjusted Mean (SE)a PAI Suicide (SPI) & Violence Potential (VPI) index scores across groups with and without histories of suicide and violence risk

History of Suicide Attempt

Yes No F (df) p d

N 29 129

SPI 72.1 (2.7) 64.2 (1.3) 6.5 (1,151) 0.011 0.53

VPI 62.9 (2.8) 57.6 (1.3) 2.9 (1,151) 0.092 0.34

History of Suicidal Ideation

Yes No

N 69 89

SPI 71.6 (1.6) 61.1 (1.4) 23.9 (1,151) 0.001 0.70

VPI 61.6 (1.7) 56.2 (1.5) 5.3 (1,151) 0.022 0.34

History of Psych Hospitalization

Yes No

N 68 90

SPI 69.5 (1.8) 62.8 (1.5) 7.5 (1,151) 0.007 0.45

VPI 60.2 (1.8) 57.3 (1.6) 1.3 (1,151) 0.258 0.18

History of Violence/Assault

Yes No

N 24 134

SPI 70.6 (3.1) 64.8 (1.2) 2.9 (1,151) 0.091 0.39

VPI 66.6 (3.1) 57.1 (1.2) 7.7 (1,151) 0.006 0.61

History of Arrests

Yes No

N 32 126

SPI 75.3 (3.1) 64.1 (1.2) 10.6 (1,151) 0.001 0.75

VPI 69.5 (3.2) 56.8 (1.2) 13.3 (1,151) 0.001 0.81

a Adjusted Mean SPI and VPI (Standard Errors [SE]) scores after controlling for Age, Gender, Education, Race, and Full Scale IQ

J Psychopathol Behav Assess (2012) 34:423–431 427

13.0, respectively). Similarly, both mean SPI (M072.3; SD015.8) and VPI (M065.9; SD019.4) scores were signif- icantly elevated in groups demonstrating some form of executive functioning impairment, as compared to those groups without impairment (M062.9; SD013.6 & M055.4; SD012.6, respectively).

Table 1 presents adjusted mean SPI and VPI scores across groups with and without a history of suicidal idea- tion, suicide attempts, inpatient psychiatric hospitalization, violence/assaultive behavior directed at others, and arrests. As hypothesized, differences in SPI were found to be sta- tistically significant at the p<0.05 level when comparing those with and without a history of suicidal ideation (F(1,151)023.9, p00.001), suicide attempts (F(1,151)0 6.5, p00.011), and inpatient psychiatric hospitalization (F(1,151)07.5, p00.007) after controlling for age, gender, race, education, and full-scale IQ. Similarly, VPI scores were also found to be statistically significant when compar- ing those with and without a history of violence/assault

(F(1,151)07.7, p00.006) and arrests (F(1,151)013.3, p00.001).

The pattern of effect sizes also supported the convergent and divergent validity of SPI and VPI. Specifically, SPI was found to be more sensitive in differentiating groups with and without a history of suicidal ideation, suicide attempts, and inpatient psychiatric hospitalization (range of Cohen’s d’s0 0.45 to 0.70), as compared to VPI (range of Cohen’s d’s0 0.18 to 0.34). In contrast, VPI was found to better differen- tiate those with and without a history of violence towards others and arrests (range of Cohen’s d’s00.61 to 0.81), as compared to SPI (range of Cohen’s d’s00.39 to 0.75).

Table 2 presents adjusted mean SPI and VPI scores across groups with and without a history of a major depres- sive episode, manic episode, psychotic episode, alcohol abuse, and drug abuse. As hypothesized, SPI scores were found to be significantly higher among those with a history of a major depressive episode (F(1,151)043.5, p00.001), manic episode (F(1,151)08.7, p00.004), alcohol abuse (F(1,151)0 16.5, p00.001), and drug abuse (F(1,151)038.6, p00.001). No significant differences in SPI were found between groups with and without a history of psychosis. Similarly, VPI scores were also significantly different between groups with and without a history of a major depressive episode (F(1,151)0 10.2, p00.002), manic episode (F(1,151)08.7, p00.004), alcohol abuse (F(1,151)020.8, p00.001), and drug abuse (F(1,151)021.8, p00.001). VPI differences were not found to be statistically significant across groups with and without a history of psychosis. Of note, while SPI was more sensitive than VPI in differentiating groups with and without a history of a major depressive episode (Cohen’s d’s01.28 and 0.66, respectively) and drug abuse (Cohen’s d’s00.95 and 0.73, respectively), both were comparable in terms of differentiating those with and without a history of mania (Cohen’s d’s00.52 and 0.48, respectively), psychosis (Cohen’s d’s00.00 and 0.08, respectively) and alcohol abuse (Cohen’s d’s00.61 and 0.66, respectively).

Finally, Table 3 presents adjusted mean SPI and VPI scores across groups with and without executive functioning

Table 2 Adjusted Mean (SE)a PAI Suicide (SPI) & Violence Potential (VPI) index scores across psychiatric groups

History of Depression

Yes No F (df) p d

N 132 26

SPI 68.6 (1.1) 49.5 (2.7) 43.5 (1,151) 0.001 1.28

VPI 60.2 (1.2) 49.8 (3.0) 10.2 (1,151) 0.002 0.66

History of Mania

Yes No

N 45 113

SPI 70.9 (2.2) 63.2 (1.2) 8.7 (1,151) 0.004 0.52

VPI 63.6 (2.2) 56.0 (1.3) 8.7 (1,151) 0.004 0.48

History of Psychosis

Yes No

N 29 129

SPI 65.7 (2.8) 65.7 (1.3) 0.00 (1,151) 0.987 0.00

VPI 59.6 (2.9) 58.3 (1.3) 0.16 (1,151) 0.688 0.08

History of Alcohol Abuse

Yes No

N 67 91

SPI 70.9 (1.7) 61.8 (1.4) 16.5 (1,151) 0.001 0.61

VPI 64.5 (1.7) 54.2 (1.5) 20.8 (1,151) 0.001 0.66

History of Drug Abuse

Yes No

N 51 107

SPI 75.3 (1.9) 61.1 (1.3) 38.6 (1,151) 0.001 0.95

VPI 66.3 (2.0) 54.9 (1.3) 21.8 (1,151) 0.001 0.73

a Adjusted Mean SPI and VPI (Standard Errors [SE]) scores after controlling for Age, Gender, Education, Race, and Full Scale IQ

Table 3 Adjusted Mean (SE)a PAI Suicide (SPI) & Violence Potential (VPI) index scores across groups with and without executive function- ing impairment

Executive Functioning Impairment

Yes No F (df) p d

N 47 111

SPI 71.0 (2.4) 63.4 (1.4) 6.3 (1,151) 0.013 0.48

VPI 63.2 (2.4) 56.6 (1.5) 4.7 (1,151) 0.032 0.44

a Adjusted Mean SPI and VPI (Standard Errors [SE]) scores after controlling for Age, Gender, Education, Race, and Full Scale IQ

428 J Psychopathol Behav Assess (2012) 34:423–431

impairment. As hypothesized, both SPI (F(1,151)06.3, p0 0.013) and VPI (F(1,151)04.7, p00.032) scores were found to be significantly higher among those demonstrating impair- ment on at least one measure of executive function after controlling for age, gender, race, education, and full-scale IQ. Further, both were comparable in terms of their sensitivity in differentiating these groups, as is evidenced by their similar effect size estimates (Cohen’s d’s00.48 and 0.44, respectively).

Discussion

Evaluating risk of harm to self and others is often an impor- tant component of psychological assessment, and psychol- ogists and other health care providers are increasingly being asked to provide clinical judgments regarding a person’s level of risk across a variety of clinical and non-clinical (e.g., occupational, screening, etc.) contexts. Despite much research in this area, existing methods for assessing risk are varied (e.g., self-report psychometric instruments, actuarial tools, patient interviews, using collateral information, etc.) and often imprecise. Given this and the low base rates of suicidal and violent behaviors in the general population, making these determinations can be challenging.

The Personality Assessment Inventory (PAI) specifically has seen increased use in clinical, forensic, and non-clinical settings as a means of assessing risk of harm to self and others (Boone 1998; Braxton, et al. 2007; Deisinger 1995; Holden 2000; Morey 1991; Morey 2007). However, despite a growing body of research focusing on the clinical and treatment consideration scales of the PAI, less is known regarding the PAI supplemental indices, particularly those assessing Suicide (SPI) and Violence Potential (VPI). The specific purpose of this study was to evaluate the construct validity of SPI and VPI in a sample of 158 psychiatric outpatients referred for psychological and neuropsychological assessment within a large northeastern academic medical center between 2007 and 2011.

Overall, results generally supported the convergent and divergent validity of both SPI and VPI. SPI was specifically found to be significantly elevated in populations with a history of suicidal ideation, suicide attempts, and inpatient psychiatric hospitalization as compared to those without. Further, these differences were observed even after control- ling for other factors known to be associated with each. While the former two are intuitively obvious, the latter is also expected given that the vast majority of these psychi- atric admissions are accounted for by people who were admitted in the context of primary mood difficulties (often with secondary suicidal ideation), as well as some following an actual suicide attempt. Only a small minority of these

admissions are accounted for by people expressing ideation about hurting others.

These findings are consistent with the study by Hopwood et al. (2008), who reported similar effect sizes for SPI across groups with and without a history of suicidal behaviors, and also extends this research in several important ways. First, the current study generalizes these findings further using a differ- ent sample of psychiatric outpatients, while also utilizing a more complex array of criterion variables to assess construct validity (e.g., history of suicidal ideation, inpatient hospitali- zation, executive functioning impairment, etc.). Second, the current study also controlled for a variety of other factors that have been known to be associated with suicidal behaviors, providing for more robust evidence of this relationship.

Similarly, VPI was found to be significantly elevated in groups with a history of violence towards others and arrests as compared to those without. Although history of arrests was used here as a global category, these patterns also make intu- itive sense given that many of these people were arrested for violent crimes, the most common of which was assault. Inter- estingly, these results contradict the findings byHopwood et al. (2008), who found no significant differences in VPI across groups with and without a history of assault, and provide preliminary support for the use of the index in outpatient psychiatric settings.

SPI and VPI scores were also found to vary significantly across different psychiatric groups in ways that would be expected. Specifically, while SPI better differentiated those with and without a history of a major depressive episode, both SPI and VPI were comparable in terms of discriminat- ing those with and without a history of mania. This finding is generally consistent with the literature, and specifically the effects of affective instability and mood lability on the potential for impulsive behaviors that are destructive in quality (Dougherty et al. 2004; Jollant et al. 2005; Keilp et al. 2001; Monahan et al. 2001; Monahan and Steadman 1994; Soloff et al. 2000; van Heering and Marusic 2003; Wenzel et al. 2011).

While both SPI and VPI were also found to be compara- ble in differentiating those with and without a history of alcohol abuse, SPI was slightly more sensitive in discrimi- nating between groups with and without a history of drug abuse, despite both effect sizes being large in magnitude. Overall, these findings also make intuitive sense given that both alcohol and drug overuse are variables used in deriving both SPI and VPI. This also generally fits with the literature, which has suggested a relationship between substance abuse and elevated risk of harm to self and others (Monahan et al. 2001; Monahan and Steadman 1994; NIMH 2011).

Neither SPI nor VPI were found to differentiate between groups with and without a history of psychotic symptoms. Given that the literature has been somewhat inconsistent with respect to the relationship between psychotic experience and

J Psychopathol Behav Assess (2012) 34:423–431 429

propensity for suicide and violence (Junginger 1996; Monahan et al. 2001; Monahan and Steadman 1994), this finding is not surprising. Within the context of the current study, it may also be explained by the fact that most of the patients being assessed were not psychotic at the time of the assessment, and only 3 out of the 129 were diagnosed with primary psychotic disorders. Future research is necessary to better examine the relationship between SPI/VPI and primary psychotic disorders.

Finally, both SPI and VPI scores were found to be ele- vated in groups manifesting impairment on at least one of the three executive functioning measures administered as part of the assessment process. This finding is also generally consistent with the literature, which has demonstrated a relationship between executive dysfunction and propensity for suicidal and violent behaviors (Harkavy-Friedman et al. 2006; Jollant et al. 2005; Marzuk, et al. 2005; Morgan and Lilienfeld 2000; Westheide et al. 2008). Perhaps most sig- nificant, effect sizes were still in the moderate range after accounting for general intellectual functioning (full-scale IQ), which provides additional evidence for the unique contribution of executive functioning impairment specifically (as opposed to more global cognitive functioning) on the potential for impulsive behaviors that are destructive in qual- ity. These findings have typically been explained in terms of how executive deficits may predispose people to greater levels of disinhibition and impulsivity, which may increase the risk of harm to self and/or others.

Overall, these results are promising in terms of the con- vergent and divergent validity of SPI and VPI in evaluating risk of harm to self and others, and contribute to the litera- ture in several ways. Perhaps most importantly, the current study extends the current research on SPI and VPI (e.g., Hopwood et al. 2008) by using multiple criterion variables for each index (i.e., not just a history of suicide attempts or violence), including psychiatric diagnoses and neuropsy- chological variables which may be hypothesized to yield different kinds of relationships. Following on this point, all analyses controlled for other demographic and clinical var- iables that have also been shown to relate to risk, providing further evidence for the validity of both indices. Finally, the current study also extends prior research that has often used highly specific samples (e.g., forensic patients, incarcerated populations, etc.) through the use of a more generalized outpatient psychiatric sample.

Given the unique manner in which both SPI and VPI were developed (i.e., using multiple indicators from the PAI profile found to be associated with each), these findings would suggest that both may be useful ancillary measures for evaluating the potential for suicidal and violent behavior. Of course, such interpretation should be conducted within the context of a thorough clinical interview and review of the person’s history, as well as utilizing other corroborating

sources for evaluating prospective threat. However, using measures such as SPI and VPI may provide health care providers with additional, methodologically varied techni- ques to be used within a larger risk assessment context.

Limitations and Future Directions

Despite these promising results, there are a number of issues with respect to the current study design that should be addressed in future research. First, the retrospective nature of this study (i.e., evaluating SPI and VPI in relation to past behaviors) provides more limited data in terms of prospec- tive risk. It will be important for future research to further explore the predictive utility of SPI and VPI in determining risk of harm to self and others, particularly given that this is what psychological evaluators are frequently being asked to do in their work. Second, while the overall sample used in this study was moderate, some of the specific group cells were smaller than hoped, which renders the models more unstable. Additional research on larger samples would allow for more covariates to be used, and greater stability in the score estimates. Finally, the psychiatric conditions coded in this study were historical and not all patients met criteria for these conditions at the time of assessment. Future research should evaluate SPI and VPI in relation to current psychiatric conditions to determine whether these patterns change. This is particularly true of psychotic experiences, given the small effects observed in the current study.

References

American PsychiatricAssociation. (2000).Diagnostic and statisticalmanual of mental disorders-TR (4th ed.). Washington DC: American Psychi- atric Association.

Boccaccini, M. T., Murrie, D. C., Hawes, S. W., Simpler, A., & Johnson, J. (2010). Predicting recidivism with the personality assessment inventory in a sample of sex offenders screened for civil commit- ment as sexually violent predators. Psychological Assessment, 22, 142–148.

Boone, D. (1998). Internal consistency reliability of the personality assessment inventory with psychiatric inpatients. Journal of Clinical Psychology, 54, 839–843.

Braxton, L. E., Calhoun, P. S., Williams, J. E., & Boggs, C. D. (2007). Validity rates of the personality assessment inventory and the minnesota multiphasic personality inventory-2 in a VA medical center setting. Journal of Personality Assessment, 88, 5–15.

Breshears, R. E., Brenner, L. A., Harwood, J. E. F., & Gutierrez, P. (2010). Predicting suicidal behavior in veterans with traumatic brain injury: the utility of the personality assessment inventory. Journal of Personality Assessment, 92, 349–355.

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). New Jersey: Lawrence Erlbaum.

Crawford, E. F., Calhoun, P. S., Braxton, L. E., & Beckham, J. C. (2007). Validity of the personality assessment inventory aggres- sion scales and violence potential index in veterans with PTSD. Journal of Personality Assessment, 88, 90–98.

430 J Psychopathol Behav Assess (2012) 34:423–431

Deisinger, J. A. (1995). Exploring the factor structure of the personality assessment inventory. Assessment, 2, 173–179.

Dougherty, D.M.,Mathias, C.W.,Marsh, D.M.,Moeller, F. G., & Swann, A. C. (2004). Suicidal behaviors and drug abuse: impulsivity and its assessment. Drug and alcohol dependence. Special Issue: Drug Abuse and Suicidal Behavior, 76, S93–S105.

Edens, J. F., Cruise, K. R., & Buffington-Vollum, J. K. (2001). Forensic and correctional applications of the personality assessment invento- ry. Behavioral Science and the Law, 19, 519–543.

Farrington, D. P. (2006). Family background and psychopathy. In C. Patrick (Ed.), Handbook of psychopathy (pp. 229–250). New York: Guilford Press.

Federal Bureau of Investigation. (2011). Crime in the United States. Retrieved July 20, 2011 at http://www2.fbi.gov/ucr/cius2009/ offenses/violent_crime/index.html.

Harkavy-Friedman, J. M., Keilp, J. G., Grunebaum, M. F., Sher, L., Printz, D., Burke, A. K., & Oquendo, M. (2006). Are BPI and BPII suicide attempters distinct neuropsychologically? Journal of Affective Disorders, 94, 255–259.

Harvey, P. D., Keefe, R. S. E., Patterson, T. L., Heaton, R. K., & Bowie, C. R. (2009). Abbreviated neuropsychological assessment in schizophrenia: prediction of different aspects of outcome. Jour- nal of Clinical and Experimental Neuropsychology, 31, 462–471.

Heaton, S. K., Chelune, G. J., Talley, J. L., Kay, G. G., & Curtiss, G. (1993).Wisconsin card sorting test manual: Revised and expanded. Odessa, FL: Psychological Assessment Resources.

Holden, R. (2000). Are there promising MMPI substitutes for assessing psychopathology and personality? Review and prospect. In R. H. Dana (Ed.), Handbook of cross-cultural and multicultural person- ality assessment (pp. 267–292). Mahwah, NJ: Lawrence Erlbaum Associates, Inc.

Hopwood, C. J., Baker, K. L., & Morey, L. C. (2008). Extratest validity of selected personality assessment inventory scales and indicators in an inpatient substance abuse setting. Journal of Personality Assessment, 90, 574–577.

Jollant, F., Bellivier, F., Leboyer, M., Astruc, B., Torres, S., Verdier, R., Castelnau, D., Malafosse, A., & Courtet, P. (2005). Impaired decision making in suicide attempters. The American Journal of Psychiatry, 162, 304–310.

Junginger, J. (1996). Psychosis and violence: the case for a content analysis of psychotic experience. Schizophrenia Bulletin, 22(1), 91–103.

Keilp, J. G., Sackeim, H. A., Brodsky, B., Oquendo, M., Malone, K., Sackeim, H., & Mann, J. J. (2001). Neuropsychological dysfunc- tion in depressed suicide attempters. The American Journal of Psychiatry, 158(5), 735–741.

Lezak, M. D., Howieson, D. B., & Loring, D. W. (2004). Neuropsy- chological assessment (4th ed.). New York: New York, Oxford University Press.

Marzuk, P. M., Hartwell, N., Leon, A. C., & Portera, L. (2005). Executive functioning in depressed patients with suicidal ideation. Acta Psychiatrica Scandinavia, 112, 294–301.

Monahan, J., Steadman, H. J., Silver, E., Appelbaum, P. S., Robbins, P. C., Mulvey, E. P., Roth, L. H., Grisso, T., & Banks, S. (2001). Rethinking risk assessment. Oxford: Oxford University Press.

Monahan, J., & Steadman, H. J. (1994). Violence and mental disorder. Chicago: University of Chicago Press.

Morgan, A. B., & Lilienfeld, S. O. (2000). A meta-analytic review of the relation between antisocial behavior and neuropsychological measures of executive function. Clinical Psychology Review, 20, 113–136.

Morey, L. C. (1991). Classification of mental disorder as a collection of hypothetical constructs. Journal of Abnormal Psychology, 100, 289–293.

Morey, L. C. (1996). An interpretive guide to the personality assessment inventory. Odessa, FL: Psychological Assessment Resources.

Morey, L. C. (2007).Personality assessment inventory: Professionalmanual (2nd ed.). Odessa, FL: Psychological Assessment Resources.

Morey, L. C., & Quigley, B. D. (2002). The use of the personality assessment inventory in assessing offenders. International Journal of Offender Therapy and Comparative Criminology, 46, 333–349.

National Institute of Mental Health. (2011). Suicide in the US: Statistics and Prevention. Retrieved July 15, 2011 at http://www.nimh.nih. gov/health/publications/suicide-in-the-us-statistics-and-prevention/ index.shtml.

Neumann, C. S., & Hare, R. D. (2008). Psychopathic traits in a large community sample: links to violence, alcohol use, and intelligence. Journal of Consulting and Clinical Psychology, 76, 893–899.

Reitan, R. M., & Wolfson, D. (1985). The Halstead-Reimn neuropsycho- logical test battery: Theory and clinical interpretation. Tucson, AZ: Neuropsychology Press.

Sellbom, M., & Bagby, R. M. (2008). Response Styles on Multi-Scale Inventories. In R. Rogers (Ed.), Clinical assessment of malingering and deception (3rd ed.). New York: Guilford Press.

Sinclair, S. J., Siefert, C. J., Shorey, H., Antonius, D., Shiva, A., Kehl-Fie, K., & Blais, M. A. (2009). A psychometric evaluation of the personality assessment inventory—short form (PAI-SF) clinical scales in an inpatient psychiatric sample. Psychiatry Research, 170, 262–266.

Siefert, C. J., Sinclair, S. J., Kehl-Fie, K. A., & Blais, M. A. (2009). An item level psychometric analysis of the personality assessment inventory clinical scales. Assessment, 16, 373–383.

Sinclair, S.J., Antonius, D., Shiva, A., Siefert, C.J., Kehl-Fie, K., Lama, S., Shorey, H.S., & Blais M.A. (2010). The psychometric properties of the personality assessment inventory-short form (PAI-SF) in inpatient forensic and civil samples. Journal of Psychopathology and Behavioral Assessment, 406–415.

Soloff, P. H., Lynch, K. G., Kelly, T. M., Malone, K. M., & Mann, J. J. (2000). Characteristics of suicide attempts of patients with major depressive episode and borderline personality disorder: a compara- tive study. The American Journal of Psychiatry, 157, 601–608.

The Psychological Corporation (1999). Wechsler Abbreviated Scale of Intelligence. New York, NY: The Psychological Corporation: Harcourt Brace & Company.

Trenerry, M. R., Crosson, B., Deboe, J., & Leber, W. (1989). Stroop Neuropsychological Test (SNST). Odessa, FL: Psychological Resources Inc.

Van Heering, C., & Marusic, A. (2003). Understanding the suicidal brain. The British Journal of Psychiatry, 183, 282–284.

Voracek, M. (2004). National intelligence and suicide rate: an ecological study of 85 countries. Personality and Individual Differences, 37, 543–553.

Voracek, M. (2009). National intelligence, suicide rate, and subjective well-being. Perceptual and Motor Skills, 109, 718–720.

Walters, G. D., Duncan, S. A., & Geyer, M. D. (2003). Predicting disciplinary adjustment in inmates undergoing forensic evalua- tion: a direct comparison of the PCL-R and the PAI. The Journal of Forensic Psychiatry and Psychology, 14, 382–393.

Walters, G.D., & Duncan, S.A. (2005). Use of the PCL-R and PAI to predict release outcome in inmates undergoing forensic evaluation. The Journal of Forensic Psychiatry and Psychology, 16, 459–476.

Wechsler, D. (1997). Wechsler adult intelligence scale—3rd edition (WAIS-3®). San Antonio, TX: Harcourt Assessment.

Westheide, J., Quednow, B. B., Kuhn, K., Hoppe, C., Cooper- Mahkorn, D., Hawellek, B., Eichler, P., Maier, W., & Wagner, M. (2008). Executive performance of depressed suicide attemp- ters: the role of suicidal ideation. European Archives of Psychiatry and Clinical Neuroscience, 258(7), 414–421.

Wenzel, A., Berchick, E. R., Tenhave, T., Halberstadt, S., Brown, G. K., & Beck, A. T. (2011). Predictors of suicide relative to other deaths in patients with suicide attempts and suicide ideation: a 30-year pro- spective study. Journal of Affective Disorders, 132, 375–382.

J Psychopathol Behav Assess (2012) 34:423–431 431

Copyright of Journal of Psychopathology & Behavioral Assessment is the property of Springer Science &

Business Media B.V. and its content may not be copied or emailed to multiple sites or posted to a listserv

without the copyright holder’s express written permission. However, users may print, download, or email

articles for individual use.