suicide by cop

Aggression and Violent Behavior 27 (2016) 107–120

Contents lists available at ScienceDirect

Aggression and Violent Behavior

Examining “suicide by cop”: A critical review of the literature

Christina L. Patton ⁎, William J. Fremouw Department of Psychology, West Virginia University, United States

⁎ Corresponding author at: Department of Psychology, E-mail address: [email protected] (C.L. Patton).

http://dx.doi.org/10.1016/j.avb.2016.03.003 1359-1789/© 2016 Elsevier Ltd. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history: Received 24 February 2015 Received in revised form 24 November 2015 Accepted 1 March 2016 Available online 7 March 2016

Suicide by cop (SbC), or evidencing intentionally life-threatening behaviors in order to coerce a law enforcement officer to respond with lethal force (American Association of Suicidology, 2013), is a phenomenon that has re- cently emerged as an area of scholarly interest, but little consensus has been reached regarding perpetrator de- mographics or intervention efficacy during SbC incidents. The present paper critically reviews the SbC research of the last 20 years with focus given to individual characteristics, situational variables, and legal intervention out- comes. Eighteen studies representing both early andmore recent empirical work from1994 to 2014were select- ed for review after meeting inclusionary criteria. Results indicated that the typical SbC perpetrator is a younger adult, Whitemale experiencing a romantic relationship conflict who has a significantmental health and criminal history and who often is intoxicated at the time of the offense. Typical legal interventions, including use of less- lethal means and verbal negotiation strategies, are not effective at preventing subject death due to officer response—unless the officer focuses verbal negotiation strategies on the perpetrator's problems. Common methodological limitations include inconsistent definitions of SbC, varying typologies of SbC intent, inadequate coding procedures, and reliance on convenience samples. Future research is needed to examine international trends in SbC perpetration, suicidal motivation in averted SbC cases, and to empirically validate SbC intervention efficacy.

© 2016 Elsevier Ltd. All rights reserved.

Keywords: Suicide by cop Police-assisted suicide Legal intervention Suicide

Contents

1. Challenges inherent in the examination of SbC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 2. Article selection and purpose . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108 3. Individual characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

3.1. Age . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 3.1.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

3.2. Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 3.2.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

3.3. Race/ethnicity, education and marital status . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 3.3.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

3.4. Subject intoxication and substance use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111 3.4.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

3.5. Mental health and criminal history . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 3.5.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

4. Situational variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113 4.1. Determining intent to die by SbC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

4.1.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 4.2. Duration and location . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

4.2.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 4.3. Weapon use . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115

4.3.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 4.4. Precipitating events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

4.4.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 5. Legal intervention efficacy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116

P.O. Box 6040, Morgantown, WV 26506, United States.

108 C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

5.1. Level and type of force used . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 5.1.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

5.2. Impact of intervention on outcome. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117 5.2.1. Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

6. Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 6.1. Integrative summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 6.2. Common methodological limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 6.3. Strengths and limitations of the current review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 6.4. Future directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

Recent research attention has begun to focus on the phenomena of “suicide by cop,” or when an individual desiring death engages in a set of intentionally life-threatening behaviors in order to compel a law en- forcement officer to respond with lethal force (American Association of Suicidology, 2013). Suicide by cop (SbC) subjects are thosewhowish to die butwho are not able to take their own lives. There are often second- ary victims of SbC, including the officer who engages the subject in a legal intervention shooting, and SbC perpetrators themselves repre- sent a small but meaningful group of individuals whose lives are often terminated as a result of these interactions. For these reasons, it is important to understand how an event of this type evolves. Rec- ognition and identification of individuals of high risk to perpetrate SbC may bolster prevention efforts, officer safety, psychological au- topsy procedures, legal proceedings, and community perception of safety (Mohandie & Meloy, 2000).

Although research on SbC is limited, some tentative estimates have emerged with regard to prevalence of SbC. It has been suggested that approximately 10–13% of cases with officer-involved shootings in the United States involve SbC (Hutson et al., 1998; Wilson, Davis, Bloom, Batten, & Kamara, 1998). Other estimates include 16–46% (Kennedy, Homant, & Hupp, 1998) and 28.5% (Lord, 2014) of all officer-involved shootings. Using a sample of North American officer-involved shootings (i.e., the United States and Canada), Mohandie, Meloy, and colleagues found estimates of SbC ranging from 76% of all hostage and barricade situations (Mohandie & Meloy, 2010) to 36% of all officer-involved shootings (Mohandie, Meloy, & Collins, 2009). Internationally, rates vary from 33.3% of all fatal police shootings in Victoria, Australia from 1980 to 2007 (Kesic, Thomas, & Ogloff, 2012) and 36.4% of all officer- involved shootings in England and Wales from 1998 to 2001 (Best, Quigley, & Bailey, 2004) to 48.2% of all incidentswhere Canadian officers had to respond to a potentially-lethal threat from 1980 to 1994 (Parent, 1998a). Because the study of SbC appears to be limited to these west- ernized countries, no other prevalence data were available.

1. Challenges inherent in the examination of SbC

Research on SbC is of interest tomany disciplines, including psychol- ogy, criminology, law enforcement, crisis response, and law. Because of this, there are many different approaches to understanding SbC with different research goals and methodologies. Additionally, this research relies on many different law enforcement agencies, which results in ju- risdictional differences in conceptualization of SbC, detection methods, and documentation standards. Despite source variability, there are common challenges to the study of SbC that are described here in order to make the following critique of the literature more salient and the repetitive nature of these issues more recognizable.

Definitional issues exist throughout the literature that prohibit an exact estimate of the nature of SbC—down to the very terminology used to classify SbC incidents. The use of terms like “police-assisted suicide,” “victim-precipitated homicide,” “legal intervention deaths,” and “law enforcement assisted suicide” has come to confound the iden- tification of SbC incidents. “Victim-precipitated homicide,” for example, is avoided because it too-generally describes an incident where victims

initiate a chain of events leading to their eventual death and may not capture variables specific to SbC (Mohandie & Meloy, 2000). This term, seen often in the early SbC literature, is not used in recent work because it places the law enforcement officer in the role of “suspect” and the SbC subject is viewed as a “victim.” Since SbC emerged as a re- search topic worthy of extended study, many researchers have adopted the colloquial expression “suicide by cop” to describe their work be- cause it is a term embraced by law enforcement officials, the media, and the general public (Mohandie & Meloy, 2000). The lack of a precise definition and unclear reporting procedures, however, lead to uncertain estimates from law enforcement personnel, and even when suicidal in- tent is documented by a note preceding the event, medical examiners often code such deaths as “justifiable homicide,” (Dewey et al., 2013) which prevents SbC frombeing conceptualized as a separate and unique form of suicide.

A greater understanding of the outcomes of SbC incidents is also dif- ficult given sample limitations. In many studies of SbC, the majority of SbC subjects are killed in the process of threatening lawenforcement of- ficers with bodily harm (Kennedy, Homant & Hupp, 1998; Hutson et al., 1998)—a rate that is significantly higher for SbC subjects than other in- dividuals involved in an officer-involved shooting (Mohandie et al., 2009). This makes accessing SbC cases for analysis a difficult process, and when subject information is made available, data is often limited to one type of offender (i.e., deceased). Often, these deaths occur within minutes of officer response to the scene. Some researchers place the av- erage time of death, from the time a law enforcement officer responds to the resolution of the incident, at 15–17 min (Arias et al., 2008; Hutson et al., 1998)—and some studies have groups of SbC incidents concluding in less than 10 min (Mohandie et al., 2009). When it comes to intervention efficacy, the comparatively short time to act, in addition to the reality that many SbC subjects present a legitimate threat warranting lethal force, may mean that typical law enforcement interventions, include use of less-lethal force options (e.g., verbal nego- tiation, physical restraint, batons, electro-conductive devices, and oleo- resin capsicum [OC] spray), may no longer be effective.

2. Article selection and purpose

Throughout the literature, conflicting perspectives exist pertaining to identification of individuals at-risk for SbC, as well as whether stan- dard legal interventions (e.g., the use of specialized tactics or hostage negotiators) are effective in these types of incidents. Classification issues have made the study of SbC difficult—thus limiting research to qualita- tive or case studies with very few statistical or empirical analyses (e.g., Arias et al., 2008; Bresler, Scalora, Elbogen, & Scott Moore, 2003; Falk, Riepert, & Rothschild, 2004; Parent, 1998a; Pinizzoto, Davis, & Miller, 2005). At present, a critical review of previous studies of SbC fo- cusing on more recent empirical research has not been accomplished. Intervention efficacy, or whether typical law enforcement resources are effective in averting the death of the SbC subject, has not been reviewed. Therefore, this study addresses an important gap by synthe- sizing the pertinent peer-reviewed literature of the past 20 years (1994–2014) related to the perpetration of SbC and the efficacy of

109C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

legal interventions of SbC. Extended focuswas given to the empirical re- search of the last 15 years, as other researchers have described the early research in great detail (see McKenzie, 2006, or Mohandie & Meloy, 2000). Methodological precision, as well as the strengths and weak- nesses, of each study were examined.

Studies were obtained through literature searches in PsycInfo, PsycArticles, and Academic Search Complete, as well as through exam- ination of the references of relevant articles. Search terms included combinations of the following: suicide by cop, suicide by police, law en- forcement assisted suicide, victim precipitated suicide, officer-involved shooting, legal intervention shooting, and police suicide. The initial search yielded 120 results. Inclusionary criteria were applied to stream- line these results and to ensure inclusion of appropriate studies. A study was included in the proposed critical review if it: (a) was published in a peer-reviewed journal after 1993, (b) was written in English, and (c) directly examined individuals who attempted or completed SbC. Studies not meeting these criteria, including case studies lacking de- scriptive analysis, were excluded.

Eighteen studies meeting these criteria were selected for review. Table 1 shows the study characteristics of each article, including infor- mation on how the authors of each study conceptualized intent to die by SbC. Articles came primarily from journals in criminal justice and law (e.g., Journal of Criminal Justice, Institutional Journal of the Sociology of Law), policing and police psychology (e.g., Journal of Police Crisis Negotiations, Police Quarterly), and forensic psychology (e.g., Journal of Forensic Sciences, Criminal Justice and Behavior).

These articles were summarized and grouped into the following domains: individual characteristics, situational variables, and legal intervention efficacy. These variables were of interest given the need to understandwhich personal variables characterize SbC subjects and con- textual variables of SbC incidents in order to inform prevention efforts. Individual characteristics included average age, gender, race/ethnicity, education, marital status, intoxication or substance use, mental health history, and criminal history of SbC perpetrators. Situational variables were described as the contextual elements of a SbC incident, including demonstration of intent to die by SbC, location, duration, weapon use, and precipitating events. Intervention efficacy focused on the level and type of force used and impact of existing interventions on final event outcome.1 Strengths and weaknesses were incorporated for each article throughout the review and summaries provided at the end of each con- tent area. Conclusions, including limitations of the current research and future directions, were provided after the research summarization.

3. Individual characteristics

The variables discussed in the following section pertain to individual characteristics of SbC perpetrators. Specifically, literature providing in- formation on perpetrator age, gender, race/ethnicity, marital status, mental health, and criminal background was reviewed and critiqued. Information on the highest education achieved was not discussed, as it is rarely examined and when it is collected, has yet to be available for more than 40% of perpetrators (Lord, 2012; Lord & Sloop, 2010).

3.1. Age

Oneof the earlyfindings regarding demographic information unique to SbC is the relation between younger adults (ages 18 to 35) and per- petration of SbC. In an early multi-source qualitative examination of SbC incidents, Kennedy et al. (1998) collected information on officer-

1 Variables selected for analysis within each category (i.e., individual characteristics, sit- uational aspects, and intervention efficacy) represent those included inmost studies of SbC. The variables in the individual characteristics category, for example, were available inmost studies of SbC compiling subject information (e.g., age, gender, and ethnicity). Other var- iables that were not usually available, or only available for a small number of subjects, were excluded from analysis (e.g., education).

involved shootings across 18 metropolitan areas in the United States by canvassing local and regional newspapers. Utilizing an innovative coding system to aid in the conceptualization of SbC intent, the authors found 240 cases of officer-involved shootings and classified them as probable suicide, possible suicide, uncertain, suicide improbable, or suicide improbable. Probable or possible suicidal intent was found in 16% of the 240 cases, but the vast majority of cases were missing sufficient infor- mation to classify as anything but uncertain suicide. The majority of the sample was male (97%) and between 16 and 35 years old (68%). This sample was younger than expected, given that middle to older adults have traditionally accounted for more suicides than younger adults (American Foundation for Suicide Prevention, 2014).

Similarly, Homant, Kennedy, and Hupp (2000) found a mean age of 31.8 years in their study of SbC, and others have foundmean ages rang- ing from31.6 to 36 years (Dewey et al., 2013;Homant&Kennedy, 2000; Hutson et al., 1998;McLeod, Thomas, & Kesic, 2012;Mohandie &Meloy, 2010; Mohandie et al., 2009; Wilson et al., 1998). The only exception to this age effect was found in a female sample of SbC perpetrators, where average age was 40 years (Mohandie & Meloy, 2011). Lord and Sloop (2010) found that the age of SbC perpetrators was not significantly different from other acutely suicidal individuals involved with police. Other studies have described a relation between younger age and great- er likelihood that the SbC perpetrator engaged in criminal and/or aggressive behavior during commission of the act (though this associa- tion was based on very small cell sizes (e.g., n = 8, n = 9; Homant & Kennedy, 2000).

3.1.1. Summary SbC perpetrators are frequently younger adults withmean age 31 to

35 years, with female SbC perpetrators being somewhat older (mean age 40 years). Some have hypothesized that younger age may be asso- ciated with greater violence and/or criminal behavior during commis- sion of the act.

3.2. Gender

In an early descriptive study, Hutson et al. (1998) studied SbC by uti- lizing data from all officer-involved shooting cases investigated by the Los Angeles County Sheriff's Department in the years 1987–1997. The authors of this study had access to intervention data, which described the efficacy of officer-initiated protocols in reaction to the suicidal indi- vidual, and highlighted one important difference between SbC subjects and other officer-involved shooting victims—verbal de-escalation tac- tics were not effective and most incidents escalated within 30 min. The authors also described a slight increase in SbC fatalities over time (up 2% to 27% by 1997 compared to previous years) not yet suggested by research, which could reflect a genuine increase or an improvement in tracking or understanding SbC cases.

Within this ten-year period, there were 437 officer-involved shoot- ings. Of the 200 shootings resulting in death, 25 (12.5%) were considered to be SbC, and a mean of 4.2 cases of completed or attempted SbC oc- curred per year. Notably, 45 of the attempted SbC or completed SbC sub- jects (97.8%) in this sample were male, and only one of the subjects (2.2%) was female. This infrequency of SbC acts completed by females is a trend that is supported by several other researchers. The SbC sample distribution appears to vary from 90.4–100% male to 0–9.6% female (Best et al., 2004; Dewey et al., 2013; Homant & Kennedy, 2000; Wilson et al., 1998), with one study including a transgender individual (Mohandie & Meloy, 2010). Homant et al. (2000) described a sample of 109males (89%) and 14 females (11%), but because a significant propor- tion of these samples was comprised of cases collected from media sources (i.e., newspapers, internet sites, television news segments), it is likely that cases including female perpetrators,who could potentially gar- ner greater media attention, were over-represented.

In the only exploration of SbC gender differences, Mohandie and Meloy (2011) determined that female SbC perpetrators differed from

Table 1 Study characteristics.

Author(s) Sample size (# SbCs)

Geographic location Origin of sample materials Years included Description of coding procedures Variable(s) of interest

Best et al. (2004) n = 22 England and Wales All OIS investigated by the Police Complaints Authority

1998–2001 Absent; no information on coding procedures or IRR Demonstration of intent to die by SbC

Dewey et al. (2013)

n = 68 United States; 55 jurisdictions and 26 states

Closed state and local cases of OIS 1979–2005 Three graduate students reviewed files, then presented them to a three-person panel, who designated the case as SbC; graduate students coded all other variables; kappa across these variables ranged from .63–1.0

Criminal and mental health history

Homant and Kennedy (2000)

n = 143 Uncertain Cases taken from professional literature and newspaper accounts

Uncertain Counseling student coded independently; ratings compared to the authors' rating of each case; 96.5% agreement for SbC designation

Demonstration of intent to die by SbC

Homant et al. (2000)

n = 123 United States and Canada Media accounts of SbC cases, legal records, database search, cases described by previous researchers

Uncertain; collected from sources dated 1990–1998

Independent coding but not clear by whom; no description of IRR

Weapon

Hutson et al. (1998)

n = 46 Los Angeles, CA All officer-involved shootings investigated by Los Angeles County Sheriff's Department Homicide Bureau

1987–1997 All files independently coded by four trained individuals; three independent reviewers; no information provided on IRR

Gender

Kennedy et al. (1998)

n = 37 18 metropolitan areas in the United States

Newspaper articles about police shootings 1980–1995 Two independent raters; 74% agreement on all variables but no discussion of reasons why disagreement occurred

Age

Kesic et al. (2012) n = 15 Victoria, Australia All police shooting fatalities investigated by coroners

1980–2007 The first author coded all variables using database materials; the second author coded SbC intent variable only. Blinding process not used. Agreement reached on all cases.

Demonstration of intent to die by SbC

Lord (2000) n = 64 32 law enforcement agencies across North Carolina

Any cases meeting the researcher's definition of SbC

1991–1998 Police officers were told to select cases that met study's definition of SbC; no other description of coding or IRR

Criminal and mental health history

Lord (2001) n = 64 32 law enforcement agencies across North Carolina

Any cases meeting the researcher's definition of SbC

1991–1998 Police officers were told to select cases that met study's definition of SbC; no other description of coding or IRR

Intervention impact on outcome

Lord (2012) n = 293 United States; 17 states partic- ipating in NVDRS data collection

All violent deaths due to legal intervention 2003–2008 Absent; no description of coding procedures or IRR Race/ethnicity, marital status

Lord (2014) n = 262 United States; 17 states partic- ipating in NVDRS data collection

All violent deaths due to legal intervention 2004–2008 Independent coding by author and criminal justice practitioner; 92.4% rater agreement on subject and officer actions during SbC incident

Level and type of force used

Lord and Gigante (2004)

n = 8 “Large southeastern city” in United States

All H&B incidents 1998–2001 Absent; no description of coding procedures or IRR Precipitating events

Lord and Sloop (2010)

n = 47 All states participating in HOBAS data collection

All H&B, suicide, kidnapping, and attempted suicide incidents where special response teams were deployed

2003–2007 Two independent raters; comparison after procedure showed 100% agreement on four variables and 91.5% agreement on two others

Subject intoxication and substance use

McLeod et al. (2014)

n = 15 Victoria, Australia All police shooting fatalities investigated by coroners; sample as Kesic et al. (2012)

1980–2008 Absent; no description of coding procedures or IRR Criminal and mental health history

Mohandie and Meloy (2010)

n = 55 United States and Canada All H&B SbCs in sample of OIS investigated by participating police/law enforcement agencies

1998–2006 Blind and independent review and coding by both authors; IRR for overall variables = .88, ICC for SbC designation = .93

Level and type of force used

Mohandie and Meloy (2011)

n = 21 United States and Canada All female SbCs in sample of OIS investigated by participating police/law enforcement agencies

1998–2006 Blind and independent review and coding by both authors; IRR for overall variables = .88, ICC for SbC designation = .93

Gender

Mohandie et al. (2009)

n = 256 United States and Canada All SbCs in sample of OIS investigated by participating police/law enforcement agencies

1998–2006 Blind and independent review and coding by two authors; IRR for overall variables = .88, ICC for SbC designation = .93

Duration and location

Wilson et al. (1998)

n = 15 Portland, OR and Dade County, FL

Medical examiner records of suicidal individuals who provoked lethal police response

1969–1993 (FL); 1963–1995 (OR)

Death scene data, victim statements, mental health documentation, toxicology and autopsy reports, newspaper accounts, police reports

Ethnicity

Note: SbC = suicide by cop; OIS = officer-involved shootings; H&B = hostage/barricaded subject incidents; IRR = inter-rater reliability; ICC = intra-class correlation coefficient.

110 C.L.Patton,W

.J.Frem ouw

/A ggression

and V iolentBehavior

27 (2016)

107–120

111C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

males on several key variables. The authors used a multi-jurisdictional sample of 707 officer-involved shootings from several police depart- ments across theUnited States and Canada and conducted additional in- terviewswith arresting officials (i.e., detectives, police officers) involved in each casewhen othermaterials were lacking. In their sample, women represented 3% of officer-involved shootings (n = 21) and 5% of SbC cases. Fifty-seven percent of female subjects in officer-involved shoot- ings were identified as SbC—a rate nearly double that of male subjects (36%). Women were more likely than men to communicate suicidal in- tent before and during the act of SbC, but women were less likely than men to die during the attempt. More women (58%) than men (19%) were being treated for mental health concerns at the time of the inci- dent and women appeared to more often deliberately engineer an SbC situation (i.e., plan to attempt SbC before the police arrive). Their results suggest that though women attempt and complete SbC much less fre- quently than men, when women are involved in shootings with police, they are more likely than men to evidence suicidal intent, χ2 = 14.10, p b .001, andmore likely to plan the suicide attempt before the police ar- rive, χ2 = 5.207, p b .05. Despite the smaller sample of female SbC per- petrators (n = 21) precluding many statistical analyses, the authors bolstered their conclusions by discussing practical implications of their results, including potential intervention strategies focusing on relation- ship problems, mental health concerns, and greater attention paid to suicidal motivation in legal interventions with female subjects.

3.2.1. Summary SbC subjects are more frequently male than female, though female

SbC subjects may have significantly more mental health concerns and evidence suicidal intent more often than male subjects.

3.3. Race/ethnicity, education and marital status

Of the studies of SbC to date, when race/ethnicity and marital status are examined, they are most often briefly mentioned. For those few studies including information on education (Lord, 2012, 2014; Lord & Sloop, 2010), data was available for less than 40% of the sample. For these reasons, and because no studies focused on any of these demo- graphic variables exclusively, race/ethnicity and marital status were discussed together in the following section and conclusions regarding education omitted.

Wilson et al. (1998) examined the case characteristics of 15 SbC in- cidents in Dade County, Florida and the Portland, Oregon metropolitan area from 1969 to 1993 (Florida) and from 1963 to 1995 (Oregon). Al- though this sample may not have been representative of SbC in general, as subjects were non-randomly selected based on their inconsistently- coded manner of death and because they represented a cluster of such deaths in the Portland area in the early 1990s (Wilson et al., 1998), the authors took care to summarize subject/event commonalities, includ- ing contextual factors (crime in progress, recent relationship dissolution, primary weapon) and individual variables (mental illness, intoxication, presence of suicide notes)—making this study distinct from other early studies of SbC. In this sample, SbC subjects were predominantly younger adultswith amean age of 32 yearswhoweremale (93.3%), and Caucasian (86.6%). In fact, of the 15 SbC cases, only two were considered racial/ ethnic minorities (one Black and one Hispanic male). However, it is not clear whether this racial distribution was representative of the Portland area at this time, and as the authors did not list the counties they examined, there was no way to verify this in the present.

This is a finding that has been supported by many researchers, with rates of SbC perpetrators identified as Caucasian or White ranging from 50.4% to 76.2% (Dewey et al., 2013; Hutson et al., 1998; Lord, 2000, 2012, 2014; Lord & Sloop, 2010). In larger international samples, this rate has been slightly lower—from 31% to 41% (Mohandie & Meloy, 2010; Mohandie et al., 2009).When the perpetrator is notWhite or Caucasian, the second most frequently endorsed race/ethnicity is Black/African- American (10.9%–45.9%), followed by Latino/Hispanic (1.6%–37%).

Lord (2014) found that SbC subjects were more likely to be White (76.7%) and non-SbC subjects equally as likely to be White (50.4%) or African-American (45.9%), though this relation was not significant. Like Wilson et al. (1998), however, previous authors have not remarked on whether their data on race distribution is representative of demographics in the area in general, or whether SbC perpetrators are significantly different from other violent perpetrators with regard to racial/ethnic category—making it difficult to evaluate the significance of these results.

In a much-later examination of all legal intervention deaths cap- tured by the CDC's National Violent Death Reporting System (NVDRS), Lord (2012) found similar results. Using a cross-national sample of 918 subjects who died during a legal intervention incident with police officers from 2004 to 2010, those demonstrating intent to complete SbCwere compared across varying degrees of intent (no intent, behavior intent only, behavior and verbal intent, behavior and planned intent, and behavior, verbal, and planned intent) on a number of individual and situ- ational variables. Intent was described as: behavior intent (life-threaten- ing behaviors with aweapon toward others or law enforcement), verbal intent (subject expressed to police, friends, family, or others a desire to be killed by police or that police would have to kill him/her rather than surrendering), or planned intent (subject left a note detailing intent to die by SbC, deliberately engineered police contact, or committed an “outrageous act” designed to bring about police response). Of those sub- jects killed during legal intervention, 57.8% wereWhite and 38.3% were African-American. Roughly half of the sample were single or widowed at the time of their deaths (54.2%), with others married/cohabitating (n = 236; 25.7%) or divorced/separated (n = 165; 17.9%). Of the per- sonal characteristics of those involved in legal intervention shootings, older White subjects were more significantly likely than younger or non-White subjects to exhibit a high degree of SbC intent. Married or widowed subjects were more likely to show behavior intent than un- married subjects, who more frequently showed no intent prior to the SbC attempt, χ2 = 9.601, p b .01. By comparing SbC to non-SbC perpe- trators across varying degrees of intent, Lord (2012) made clear the im- portance of multiple data points for prediction of SbC behaviors. However, there was no description of how the previously- documented idiosyncrasies of the NVDRS dataset, including missing case narratives, inconsistently coded manner of death, and duplicate case information (McNally, Patton, & Fremouw, 2015), were processed.

Although the demographic information in Lord's (2012) study was available only for individuals killed in shootings with the police, leaving readers to wonder about subject similarity to those in other SbC samples, other researchers have found similar patterns with regard to marital status. Mohandie et al. (2009) described a sample of SbC perpe- trators whowere either single (n=95; 37%), separated (n=25; 10%), divorced (n=15; 6%), cohabitating (n=35; 14%), or married (n=37; 13%). When collapsed into fewer categories, marital status has been found by others to be fairly similar for suicide only and SbC perpetra- tors, such that a large proportion of perpetrators (49.7% and 47.5%, re- spectively) fell into the “single” (rather than “married” or “divorced/ widowed”) category (Lord & Sloop, 2010). Lord (2014) found that sig- nificantly more SbC subjects than non-SbC subjects were currently or previously married.

3.3.1. Summary SbC subjects are usually of White race/ethnicity, with the second

most commonly endorsed race/ethnicity being African-American. Con- clusions aboutmarital status aremixed and depend on how researchers definemarital categories (e.g., single/widowed, single, or divorced/sep- arated). Not enough information is available on educational status to make conclusions at this time.

3.4. Subject intoxication and substance use

No researchers have focused exclusively on the role of substance use during commission of SbC, but a few have collected information on

2 MacCallum, Widaman, Zhang, and Hong (1999) recommend sample sizes of at least 300 cases whenmodels contain a small number of factors and just three or four indicators for each factor, whereas others (Tabachnick & Fidell, 2013) suggest that sample sizes “well below 100” (p. 618) are satisfactory when factor loadings are greater than .6 and factors are well-determined.

112 C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

suspected substance abuse and intoxication during SbC incidents. Lord and Sloop (2010) compared self-inflicted suicide (“suicide-only”) to SbC subjects on a number of individual and historical variables across levels of suicidal intent using data from the Federal Bureau of Investigation's Hostage Barricade Data System (HOBAS). Notably, HOBAS includes data on both completed and averted SbC cases, poten- tially allowing for generalization to other samples and setting this study apart from others using data from deceased subjects only. Of the 242 subjects involved in a hostage/barricade, suicide, kidnapping, or attempted suicide incident from an unknown period of time, only 18 cases (7.1%) were classified as SbC by the law enforcement agencies reporting their data. Using supplementary information from the inci- dent narratives, the authors concluded that 29 additional cases of SbC existed within the dataset. Sixty-nine of the total cases (36.1%) were flagged for alcohol abuse and 94 (49.2%) for abuse of “all other drugs” (p. 892). Eighty (43.2%) were using alcohol at the time of the incident and 57 (30.8%) were using all other drugs. Suicide-only subjects (n = 206) did not differ significantly from SbC subjects (n=47) with regard to drug addiction, though the vast majority of both groups were said to be addicted to drugs (suicide-only: n=127, 83.1%; SbC: 36, 94.8%). In- terestingly, the authors foundonly 47 SbC cases out of the total database of 5035 cases—an extremely low 0.9% of all incidents requiring special response teams. Given the shorter duration of SbC incidents relative to other officer-involved shootings, it is likely that many instances of SbC conclude before a special response team arrives, and utilizing a sample of SbCs taken exclusively from these types of events would lead to mis- representative conclusions about many aspects of the phenomenon, in- cluding intoxication or substance abuse.

Rates of intoxication during the commission of SbC from other stud- ies range from 36% to 75.5% (Dewey et al., 2013; Lord, 2000; Mohandie & Meloy, 2010; Wilson et al., 1998). Although Mohandie et al. (2009) found a much lower rate of intoxication during the event (36% of all SbCs), SbC subjects were more likely than non-SbC subjects to be under the influence of alcohol during the event, F = 9.9923, p b .005. When a SbC subject is intoxicated during the act, it is most often due to alcohol alone or alcohol combined with other drugs, with “hard drugs” (cocaine or methamphetamines) following second (Lord, 2000; Mohandie &Meloy, 2010). Substance abuse is also common for SbC sub- jects,with estimates of 53.8% (Lord, 2000) to 65.2% (Hutson et al., 1998), though data on long-term use is often limited by the incident-focused nature of many SbC samples. Kesic et al. (2012) described how, in their comparison of Australian SbC fatalities to non-SbC police shooting fatalities, both groups evidenced a history of substance abuse (60% and 43%, respectively), but the groups were not significantly different from one another. The study of the role of intoxicating substances' in the commission of SbC presents important implications, as the intoxicated subject has been found to be more impulsive and more lethal in these types of events (Beck, Weissman, & Kovacs, 1976).

3.4.1. Summary Many SbC subjects are intoxicated at the time of their attempts, with

alcohol described as primary substance used and “hard drugs” listed second. Substance abuse is also common, though it is not clear whether SbC subjects significantly differ from suicide-only subjects or non-SbC police shooting fatalities in this regard.

3.5. Mental health and criminal history

Many SbC perpetrators have mental health and criminal history characteristics that set them apart from other individuals, including those who die by suicide alone. Lord (2000) found that of the 64 SbC cases obtained from 32 state and local law enforcement departments in North Carolina, mental illness was identified either formally bymen- tal health providers or informally by familymembers for approximately 54% of SbC subjects. Of those individuals with a history of mental health commitment, 26.7% died by SbC and 18.2% did not, suggesting thatmore

serious mental health concerns may be associated with completion of SbC. When a diagnosis was known, it was commonly schizophrenia or bipolar disorder. With regard to suicidal behavior, more than half (58.5%) of all SbC subjects had made some type of pre-suicidal gesture, but the number of suicide attempts was not associated with either a completed SbC or an attempted SbC outcome. Slightly more than half of the sample had no criminal background (50.8%), but of those with criminal histories, domestic violence appeared to be most common (18%), followed by drug offenses (9.8%), and DUI charges (9.8%). When collecting information on relevant subject history, case files were read to the researcher by the law enforcement official at each department, meaning that decisions regarding the existence of a variable of interest (i.e., symptoms of mental illness) may have been largely up to the officer in charge. Because of this method of data collection, measurement of mental health and criminal histories should be viewed with caution.

Later research attempted to discern whether existing typologies of suicidal behavior would also apply to SbC perpetrators. Dewey et al. (2013) used data from 85 state and local law enforcement agencies from 26 different states on potential SbC incidents from the years 1979– 2005. To determine whether a case was classified as SbC, three graduate students reviewed files and presented their conclusions to other team members. Consensus with at least five of the seven teammembers was required to designate a case as a true SbC incident, and if agreement was not obtained, the team contacted the police department in charge of that investigation. Of the 85 officer-involved shootings across 55 juris- dictions, 58 were determined to be SbC, 10 had attempted SbC, and 17 were excluded because there was insufficient evidence of suicidal intent.

Exploratory factor analysis of 19 clinical risk and psychological vari- ables revealed a three-factor model of SbC: Mental Illness, Criminality, and Domestic Problems. A k-means factor analysis was conducted to de- termine if these factor loadings might represent different groups of SbC perpetrators. The Mental Illness cluster (n = 25) had the highest pre- dominance of clinical risk factors, with almost all group members hav- ing a history of depression (96%), prior suicidal ideation (88%), or diagnosed mental illness (80%). The second cluster, Criminality, was composed of SbC perpetrators who were recently involved in criminal activity (n = 24). Of this group, 100% had a history of arrests, 83.3% were facing imminent incarceration, and 75% had experienced a recent legal stressor. Many of these individuals were currently experiencing symptoms of depression or mental health concerns (54.2%), but at a rate much less than those in the first cluster. The last cluster, Not Other- wise Specified, included SbC perpetrators with somewhat high risks of depression and substance use (n = 19), but these risks were present for less than half of the group members (42.1%).

This study was the first (and at the time of this review, the only) sta- tistical classification of SbC offenders into groups, and one of the only studies thus far to use multivariate techniques to better understand SbC in general. However, due to significant overlap between groups of of- fenders on varying risk factors (e.g., history of arrests), multicollinearity may have been an issue in this sample—especially when considering the authors used a dichotomous coding scheme for several of the core variables, including criminal history or mental illness (0 = not present, 1 = present). Mental health diagnoses were collected via a variety of sources, including family/friend statements and mental health notes, without anymethod of controlling for bias or lack of reliability across cli- nicians. Here, the authors had a smaller sample size (n = 58) and three factors (with one containing only two indicators) and factor loadings ranging from .52 to 83, perhaps limiting their ability tomake quantitative comparisons and to portray small groups within the larger sample.2

113C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

In their work on police interactions with the mentally ill, McLeod, Thomas, and Kesic (2014) compared SbC subjects to those suggested to have provoked police to shoot, subjects in acute psychiatric crises, and subjects who died during police interactions due to self-inflicted suicide. The authors were granted access to three different sources: the Victorian police's mental health transfer database, which docu- ments police interaction with the mentally ill and other referral agen- cies; the Law Enforcement Assistance Program (LEAP), which details all police contacts to those who are witnesses, victims, or perpetrators of a crime; and the Redevelopment of Acute Psychiatric Directions (RAPID) register, which is maintained by the Victorian Department of Health and documents all contacts with public mental health agencies in Victoria, Australia. Fifteen cases of SbC described in a previous study (Kesic et al., 2012) were used as the comparison group. The authors (2014) found that SbC subjects were significantly more likely than non-SbC police provoked shooting subjects, psychiatric crisis subjects, or self-inflicted suicide subjects to have previous criminal charges. Fur- ther, 73% (n = 11) of SbC subjects had a history of violent offending, compared to 29.3% (n = 27) of police-provoked shootings, 17.3% (n = 390) of psychiatric crisis subjects, and 15.9% (n = 58) of self- inflicted suicide subjects. SbC subjects were also more often male and slightly younger than police-provoked shooting subjects, and SbC sub- jects (53%) were approximately as likely as other groups to have a pre- vious Axis I diagnosis. Coding of SbC intent was described in this study as a “dynamic” process without a description of how the authors con- ducted it, leading to questions about inter-rater reliability and making it difficult to replicate results. More detailed information about age of first mental health contact, age of first criminal charges, history of non-violent offending, and number of criminal charges was present for other groups butmissing for SbC cases, perhaps preventing a greater understanding of the development and severity of mental illness and criminality in SbC perpetrators. Despite these limitations, the authors' conclusion that SbC perpetrators had a more entrenched history of vio- lent offending than any other group is innovative because it may high- light a specific risk factor easily identifiable by police upon immediate response to SbC incidents.

Despite the limitations of this work, others have also identified higher rates of mental illness and suicidal ideation in SbC samples (Homant et al., 2000; Mohandie & Meloy, 2010). Lord (2014) found that SbC subjects had a significantly greater frequency of suicide at- tempts (28.2% vs. 1.6%) and drug addiction/mental illness (27.5% vs. 11.8%) than non-SbC subjects. Lord and Sloop (2010) found that a higher percentage of suicide-only subjects had attempted suicide once, but SbC subjects were more likely to have attempted several times. Furthermore, though both groups had similar mental health treatment backgrounds, the suicide-only group was more likely to have been committed for inpatient treatment. Mohandie et al. (2009) described how SbC subjects were significantly less likely to be known gang members, χ2 = 24.993, p b .001, less likely to be on parole or probation, χ2 = 9.324, p b .005, and less likely to have had a prior parole or probation violation, χ2 = 10.552, p = .001. Fi- nally, Kesic et al. (2012) found that SbC subjects were significantly more likely to be experiencing chronic mental illness and to have previous suicide attempts than a non-SbC police shooting fatality group.

3.5.1. Summary SbC subjects often have significantmental health and criminal histo-

ries. Within groups of offenders, typologies of either significantly great- er mental health or greater criminality may emerge, such that groups are predominantly classified as mentally ill or criminally active. SbC subjects are often less mentally ill and more criminally active than suicide-only subjects and less criminally-active than non-SbC subjects, though variations in this pattern have emerged in international samples.

4. Situational variables

Situational variables related to perpetration of SbC are reviewed in this section and include the following: demonstration of intent to die by SbC, duration and location of the SbC incident, primaryweapon, sub- ject intoxication and substance use, and precipitating events.

4.1. Determining intent to die by SbC

Perhaps nothing is ofmore central interest to the study of SbC than a greater understanding of the subject's intent to die during interactions with police. Documentation of SbC intent has been undertaken in differ- ent ways by other researchers, but many focus on the requirement of four core criteria: 1) evidence of suicidality, 2) evidence of intent to die during a legal intervention with police, 3) evidence of possession of an operative or seemingly operative lethal weapon, and 4) intention- ally escalating police response by threatening police or others with that weapon (Hutson et al., 1998). When not relying on this criteria, previ- ous researchers have instead conceptualized SbC as: the result of a re- quest to be killed or to die followed by some action designed to elicit police response (Lord, 2000; Wilson et al., 1998); engaging in actual or apparent risk to others with the purpose being to provoke lethal force from law enforcement officials (Mohandie & Meloy, 2010, 2011); or a combination of primary, secondary, and irrational/disturbed indicators (Best et al., 2004). Because of the subtle differences in con- ceptualization across researchers, definitions of SbC intent are described in greater detail in Table 2.

In one of the first studies conceptualizing SbC intent, Homant and Kennedy (2000) started with 123 cases described in Homant et al. (2000) and added 22 additional cases obtained via newspapers, local prosecutors, and literature search of police shooting cases. Other than age and gender, no demographic information was provided, making comparison with other studies difficult. Non-lethal SbC attempts were included for analysis, which could have increased generalization, and cases not meeting their definition for inclusion were analyzed for com- mon patterns. Perpetratorswere divided into the following types:Direct Confrontation, Disturbed Intervention, or Criminal Intervention. Those in- dividuals in the Direct Confrontation group (n = 44, 30.8%) either used deadly force to suddenly attack police or bystanders (kamikaze attack; 3.5%), confronted police with a weapon and demanded to be killed (controlled attack; 4.2%), created a situation designed to elicit police re- sponse (manipulated confrontation; 15.4%), or committed a serious crime to bring about police action (dangerous confrontation; 7.7%). The Disturbed Intervention group (n = 82, 57.3%) was marked by SbC sub- jects who were either engaging in a suicide attempt but appearing am- bivalent (suicide intervention; 20.3%), involved in a domestic dispute and became suicidal upon police intervention (disturbed domestic; 16.8%), or were intoxicated, mentally ill, or acting strangely (disturbed person; 20.3%). Finally, the Criminal Intervention group (n = 17, 11.9%) was made of individuals who had committed either a major crime and were unwilling to return to prison (major crime; 6.3%) or a minor crime and were resentful of police intervention in general (minor crime; 5.6%).

Significantlymore successful interventionswere completed for indi- viduals in the suicide intervention subgroup ofDisturbed Intervention rel- ative to the other subgroups, χ2 = 6.60, p = .02—which the authors suggested may be related to a police officer's inclination to use less- lethal response force options with suicidal individuals. Explicit intent to attempt SbC seemed to be most evident in the Direct Confrontation group, while general suicidal ideation was explicitly evident for the Suicidal Intervention subgroup of Disturbed Intervention. Notably, the groups with more evident intent made up only 51% (n = 73) of the sample—illustrative of the difficulty inherent in research on suicidal in- tent in SbC incidents.

In a follow-up toHomant and Kennedy (2000); Best et al. (2004) ex- amined 22 police shootings referred for investigation by the Police

Table 2 Determination of intent to die by SbC by study.

Author(s) Determination of SbC intent

Best et al. (2004) Evidence of primary/secondary indicators, state based indicators of irrationality, or minimal evidence of suicidal intention Dewey et al. (2013) Decided by six-member team after records review; criterion uncertain Homant and Kennedy (2000) Cases taken from Hutson et al. (1998) cases used their definition; cases other than Hutson et al. (1998) defined as “individuals who, bent on

self-destruction, engage in threatening and criminal behavior in order to force police to shoot them (Geberth, 1990, p. 105) Homant et al. (2000) All cases previously identified as SbC by other sources Hutson et al. (1998) “Individuals stating outright that they wanted officers to shoot them, written or verbal communication to family or friends…or not dropping

their weapon when advised by officers to do so and then aiming their weapon at officers or civilians” (p. 666) Kennedy et al. (1998) Intent determined by rating as: “probable suicide, possible suicide, uncertain suicide, suicide improbable, or no suicidal evidence” (pp. 3–4) Kesic et al. (2012) All five of the following, based on Hutson et al. (1998): communication of suicidal intent, gestures of suicidal intent, person stated that he/she

wanted police to shoot them, possession/appearance of possession of a deadly weapon, and evidence that subject deliberately escalated the encounter to provoke police lethal response OR Three of the historical factors (e.g., chronic mental or physical health problems, suicide attempts) and five of the incident variables (e.g., possession of a deadly weapon, refusal to follow police instruction) described by Lindsay and Lester (2004, 2008)

Lord (2000) “Individuals who, after being confronted by law enforcement officers, either verbalized their desire to be killed by law enforcement officers and/or made gestures such as pointing weapons at officers or hostages, running at officers with weapons, or throwing weapons at officers” (p. 403)

Lord (2001) Individuals who, after being confronted by police, verbalized desire to die by SbC or who took action to enable lethal police response (e.g., pointed a weapon)

Lord (2012) Presence of at least one primary SbC indicator as defined in Lord and Sloop (2010) Lord (2014) “Individuals who possessed at least one primary indicator of SbC: verbal, behavioral, or planned intent to induce officers to shoot them” (p. 85) Lord and Gigante (2004) Incidents in which individuals confronted by police verbalize desire for SbC or make gestures to elicit police response Lord and Sloop (2010) One or more primary indicators of suicidal intent in incidents where the subject “attempts or completes suicide by inducing police officers to

shoot them” (p. 892) based on Best et al.'s (2004) typology McLeod et al. (2014) All five of the following, based on Hutson et al. (1998): 1) communication of suicidal intent, 2) gestures of suicidal intent, 3) person stated

he/she wanted police to shoot them, 4) possession/appearance of possession of a deadly weapon, and 5) evidence that subject deliberately escalated the encounter to provoke police lethal response

Mohandie and Meloy (2010) “Subject engaged in actual or apparent risk to others with the intent to precipitate the use of deadly force by law enforcement personnel” (p. 105)

Mohandie and Meloy (2011) “Subject engaged in actual or apparent risk to others with the intent to precipitate the use of deadly force by law enforcement personnel” (p. 665) Mohandie et al. (2009) “Subject engaged in actual or apparent risk to others with the intent to precipitate the use of deadly force by law enforcement personnel” (p. 457) Wilson et al. (1998) “A threat by the victim to kill the self, a request by the victim to be killed, an expressed desire to die, or the finding of some evidence of suicidal

ideation or intent…that was temporally related to the fatal incident” (p. 47)

114 C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

Complaints Authority in England and Wales from 1998 to 2001. The authors created their own typology of suicidal intent focusing on the presence of primary indicators (e.g., communicating intent to others or deliberately eliciting police response), secondary indicators (e.g., history of suicide attempts or articulating desire to die by SbC after being confronted during crime perpetration), irrational behavior, or minimal evidence of suicidal intent.

Of the 22 cases selected, four subjects (22.7%) engineered contact with the police and also communicated suicidal intent to others, evidenc- ing primary suicidal indicators. Therewas considerable evidence of intox- ication, mental health problems, and domestic disputes in this group—reminiscent of Homant and Kennedy's (2000)Disturbed Interven- tion type. Four individuals (18.1%) did not express suicidal intent but did show combinations of intoxication, mental health concerns, domestic disputes, and a history of suicidal attempts—evidencing secondary indi- cators of suicidal intent. Two cases (9%) showed irrational behavior or mental health problems, but for whom explicit suicidal intent could not be assumed. Three cases (13.6%) displayed inoperative weapons (i.e., unloaded or replica) and were classified as possessing minimal to no suicidal intent, as they appeared to have done so for a chance to es- cape. Only eight of the 22 cases (36.4%) in Best et al.'s (2004) sample demonstrated primary and secondary indicators of suicidal intent, and for eight additional cases, there were no risk factors identified to suggest suicidal motivation at all. This lack of apparent SbC intent may be reflec- tive of Best et al.'s (2004) definition for inclusion into the study—“a discharge of a police weapon resulting in an injury to a member of the public” (p. 353),which alsomakes comparison of thiswork to other stud- ies problematic.

In a more recent examination of SbC intent, Kesic et al. (2012) ana- lyzed the nature of SbC incidents in a sample of fatal police shootings from Australia between 1980 and 2007. The authors were granted access to cases of police fatalities investigated by coroners via the Victorian police's Use of Force (UoF) register, and, similar to prior re- search documented in McLeod et al. (2014), the Law Enforcement

Assistance Program (LEAP) and the Redevelopment of Acute Psychiatric Directions (RAPID) register. Subjects were classified as SbC when they demonstrated evidence of all five of the following, based on the work of Hutson et al. (1998): 1) communication of suicidal intent, 2) gesture of suicidal intent, 3) indication that the subject wanted to be shot, 4) possession of appearance of possession of a deadly weapon, and 5) evidence that the subject purposefully escalated the incident in order to have police shoot him/her. In addition to the first criteria, sub- jects could also be termed SbC if they possessed aminimumof three his- torical factors (e.g., chronic mental or physical illness, substance use disorder) and eight incident variables (e.g., forcing a confrontation, possessing a deadly weapon) associated with SbC perpetration accord- ing to previous researchers (see Lindsay & Lester, 2004, 2008, for more information).

The authors found that of the 45 police-shooting fatalities, 15 (33.3%) met the criteria described above to be considered SbC. All 15 subjects (100%) were in possession of a deadly weapon, threatened an officer with that weapon, and refused to follow police instructions. SbC subjects were significantly more likely than non-SbC subjects to have a chronic mental or physical health condition (χ2 = 8.76, p = .006), to have planned the incident (χ2 = 8.76, p = .006), and to have experienced a recent stressor (χ2 = 6.60, p = .003). This study is nota- ble because the authors used particularly stringent intent criteria for designation as SbC—perhaps excluding cases in which subjects were merely ambivalent about death by SbC and creating a cleaner portrayal of the SbC subject. The study is a useful contribution to the literature and illustrates that SbC is not a phenomena exclusive to the U.S. and United Kingdom. SbC subject demographic and incident variables were slightly different in this sample—for example, personality disorders were much more common for SbC subjects and the most frequent reason for police response was a mental health crisis. These differences from those docu- mented in North American samples may be related to cultural dissimi- larities, or could potentially be the result of the authors' less-inclusive intent criteria. The authors used appropriate statistical measures,

115C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

including the use of Fisher's exact test for small cell sizes and Bonferroni corrections for multiple comparisons. However, because cases could be labeled as SbC by meeting either the first or second set of criteria, com- parison of SbC frequency to that of other studies is difficult.

4.1.1. Summary Clear evidence of intent (i.e., verbal expression of wish to die by

SbC) is lacking for many subjects involved in SbC incidents. When evidence of SbC intent is presented, it is more often in the form of disturbed behavior or aggressive gestures, including intoxication, mental health concerns, behavior influenced by domestic disputes, or provoking police to shoot either by refusing to follow instructions or by advancing toward police with a deadly weapon. Researchers have yet to agree on a standard way to conceptualize SbC intent but approaches focusing explicitly on behavioral indicators, rather than on verbal expression of intent, appear to have the greatest utility.

4.2. Duration and location

Mohandie et al. (2009) was one of the first studies to utilize a large international SbC sample collected from 90 police departments and law enforcement agencies across North America. In the first use of their large international sample (n=707) of officer-involved shootings, they identified 256 SbC subjects (36% of their sample—a much higher estimate than in previous research) and compared them to non-SbC subjects using chi-square analyses—representing a subtle shift toward more empirically-based exploration of the phenomena. SbC subjects were significantly different than non-SbC subjects in a number of ways, including more likely to provoke and draw fire from police and more likely to hurt or kill officers or civilians during the course of the event. This implication was particularly important, as it dispelled the consistently-held belief that suicidal and homicidal ideation are nega- tively correlated (Mohandie et al., 2009). Most SbC incidents (n = 176; 72%) concluded in an hour or less. Sixty-two percent (n = 151) were finished within 30 min, 41% (n = 99) within 15 min, and 29% (n = 70) terminated within 10 min. In this sample, SbC incidents were not drawn-out sieges with hostages taken; rather, they were relatively short-lived events which rapidly evolved and concluded in a matter of minutes. The authors were not clear when events began to be timed (i.e., at time of police response, time of 911 call, etc.) and no in- formation was provided for duration of incidents for non-SbC incidents, making a comparison to other studies impossible. Approximately half (46%; n = 118) occurred at a residence and 38% (n = 97) at a public or open environment (again, this information was not provided for non-SbC incidents). SbC perpetrators were significantly less likely than non-SbC perpetrators to flee the police during the incident, χ2 = 64.789, p b .001, and significantly more likely to evidence inconsistent escape behavior, χ2 = 26.618, p b .001.

The suggestion that SbC incidents are often ofmuch shorter duration than typical law enforcement shootings has been supported by others (Lord, 2001).Mohandie andMeloy (2011) found that female SbCperpe- trators, who were more likely than male SbC perpetrators to be receiv- ing mental health treatment before the event, were killed in less than half the time males were (modes of 2 versus 10 min, respectively). Hutson et al. (1998) found a median time of 15 min in their sample of SbC incidents, with 20 SbC subjects (43.5%) killed in the first 5 min and the majority of incidents occurring at a residential location (71.7%). In Australia, more than a quarter of SbC events concluded in just 1min, but average incident duration for SbC incidentswas a lengthy 64 min and SbC incidents were significantly longer in duration than non-SbC police interventions (Kesic et al., 2012). In a study of SbC using hostage/barricaded subject incident reports, the mean duration of all hostage/barricade incidents was 2 h and 33 min, with a range from 3 min to 216 h and a mode of 10 min (Mohandie & Meloy,

2010). These statistics are interesting given estimates of typical resolu- tion of hostage/barricade incident of 4 h or less (Hammer, 2007).

4.2.1. Summary SbC incidents are relatively short in duration, with many cases ter-

minated within 1 h or less. International samples from Australia have a significantly longer duration, but may this may be related to the ten- dency for SbC calls in this area to begin as serious mental health crises instead of domestic violence situations. When SbC incidents include hostages or barricaded subjects, the average duration is longer, and at this time, it is not clear whether SbC incidents including hostages or a barricaded subject are significantly longer than other hostage/barricad- ed subject incidents. SbC cases often evolve in a domestic or residential setting.

4.3. Weapon use

It is a common misperception that individuals who wish to die by legal intervention frequently do not actually carry weapons, and if they have what appears to be a real weapon, it is often revealed to be a false or unloaded gun. In an examination designed to dispel these myths regarding actual versus perceived dangerousness, Homant et al. (2000) examined 123 cases of completed or averted SbC usingmultiple sources, including 65 described in previous research (Kennedy et al., 1998; Parent, 1998a; Wilson et al., 1998) and others from media or in- ternet sources, police departments, legal notes, database searches, and the authors' appearances as expert witness. “Real danger” was scored using a 6-point scale created by the authors (1 = one or more persons killed during the incident in a chain of events leading to the incident to 6 = subject bluffed having a weapon, reached for an absent gun, or used a replica or prop). “Perceived danger” was dichotomously coded (relatively lower perceived danger to relatively higher perceived danger). Their results indicated that SbC subjects used loaded firearms in 61 (49.6%) incidents. Knives were used in 24 (20% of cases) and other weapons, including broken glass, a broomstick, and a muffler pipe were used in other incidents (actual percentage not listed by the au- thors). The authors determined that only 22% of the time, police officers were not in real danger, using 27 cases where subjects carried an unloaded firearm (n = 11; 9%), a weapon with faulty ammunition (n = 1), or a prop/toy gun (n = 15; 12%) to draw this conclusion. The correlations between real and perceived danger (r= .19) and perceived danger and lethality (r = .22) were low, suggesting that use of deadly force was correlated with perceived danger but not real danger in SbC cases.

Others have found that most SbC subjects do threaten others with a lethal weapon, that this weapon is usually a loaded firearm (Lord & Gigante, 2004; Lord & Sloop, 2010), and that a significant proportion of subjects who have a firearm actually do fire their weapons at police or other individuals (Lord, 2014; Mohandie & Meloy, 2010). Knives tend to be the second most frequently used weapon (Hutson et al., 1998; Lord, 2000;Mohandie et al., 2009). When a prop or replicaweap- on is used, it is more often a non-powdered (BB/pellet), toy, or unloaded/inoperative firearm (Hutson et al., 1998; Mohandie & Meloy, 2010; Mohandie et al., 2009). Lord and Sloop (2010) concluded that SbC subjects were significantly more likely than suicide-only subjects to carry a firearm. Lord (2014) found similar results, concluding that SbC subjects were less likely than non-SbC subjects to carry “other weapons” and non-SbC subjects less likely to carry a knife. Potential gender differences in weapon choice were highlighted by Mohandie and Meloy (2011), who found that female SbC subjects were armed with a weapon 100% of the time, equally likely to be a firearm or knife. Similarly, SbC subjects from Australia were armed with a deadly weapon 100% of the time—but this might be due to more stringent SbC criteria leading to inclusion of a more resolute, violent perpetrator than in previous studies.

116 C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

4.3.1. Summary When SbC subjects use a weapon, they use a firearm as a primary

weapon in most cases, with knives or other sharp objects following second. Very few SbC incidents are characterized by an absent, unloaded, or inoperative firearm, suggesting a high level of danger to police and others. Rates of subject possession of a deadly weapon vary across studies, but most indicate that SbC subjects are of high risk of violence toward themselves or others in pursuit of their goals.

3 In this example, the exponential function of the regression coefficient (eb) is the odds ratio associated with higher odds that the subject is SbC when exposed to that variable.

4.4. Precipitating events

Often, SbC incidents occur after an individual, who is already experiencing symptoms of mental illness or interpersonal distress, faces some type of activating event. Using a sample of 30 hostage/barri- cade incidents requiring special response team intervention occurring in a large southeastern city from 1998 to 2001, Lord and Gigante (2004) compared SbC subjects to non-SbC subjects on a number of variables, including triggering or precipitating events leading up to the incident. The researchers compared their results to Amendola, Learning, and Martin (1996), who had examined similar characteristics in a hostage/barricade sample (potentially also including SbC subjects) years earlier. Like the earlier results of Amendola et al. (1996), domestic disturbances and the subject threatening suicidewere the twomost fre- quent precipitants to a special response team request. However, non- SbCs (n = 22) were much more diverse with regard to precipitating events, evidencing relationship, financial, family, criminal, and mental health concerns—whereas SbCs (n = 8) most often showed romantic relationship dissolution or family problems as antecedents.

Lord and Gigante (2004) concluded that their sample of hostage/ barricade subjects was very similar to Amendola et al.'s (1996) sample, with the main differences being a reflection of the differences between SbC and non-SbC subjects. In Lord and Gigante's (2004) sample, the in- cidents took longer to conclude, the subjects made greater demands of the police negotiators, and subjects threatened police more—things that were particularly true for SbC incidents. The authors collected a vast amount of information about these cases, including new informa- tion about typical responses of SbC subjects during hostage negotiation (e.g., “I′m not coming out” or making demands) and contextual vari- ables not yet studied by others. However, no statistical comparisons were made, perhaps due to the small sample size (n=30), so while re- sults appeared to be significant, no conclusions were offered regarding the magnitude of the differences between SbC and non-SbC subjects.

Etiologic factors precipitating SbC are some of the most-frequently studied aspects of the SbC phenomenon. Hutson et al. (1998) found that domestic violence or despondence over relationship dissolution ap- peared to precede 58.7% of cases of SbC, with impending incarceration (8.7%) or loss of employment following thereafter (4.3%). Lord (2014) found much higher rates of imminent incarceration as a precipitant (30.5%), with non-SbC subjects more likely than SbC subjects to be caught in a criminal act (75%) and to have a criminal history (62.6%). Similarly, Lord and Sloop (2010) found that perpetration of a criminal act was significantly more likely to precede a lethal contact with police for SbC rather than suicide only subjects. Lord (2014) found that SbC subjects were significantly more likely than non-SbC subjects to be involved in a domestic dispute but non-SbC subjects significantly more likely to be interrupted while committing a crime proximal to police arrival on scene. Kesic et al. (2012) found that in their Australian sample, the most common SbC precipitants were a mental health crisis call (33%) and police intervention during subject com- mission of a crime (33%), followed by domestic abuse calls (20%), ar- rest raids (6.7%), and traffic stops (6.7%). SbC subjects experienced significantly more stressors than non-SbC subjects in the day leading up to the SbC incident (p = .003), including interpersonal problems (66.7%), grief and loss (33.3%), and legal (26.7%) and medical (20%) problems.

4.4.1. Summary Many SbC incidents are unplanned police operations and typically

are preceded by domestic or interpersonal disputes, with threat of impending incarceration or arrest often described as another common precipitant. Mental health crises, though less frequently documented, are also described as a precipitant for many SbC incidents. Stressors common for SbC subjects in the 24 h prior to the SbC attempt are inter- personal problems, loss/bereavement, physical health concerns, and legal issues.

5. Legal intervention efficacy

Police intervention strategies are discussed in this section and will focus on the following topics: level of and type of force utilized and in- tervention impact on outcomes of SbC incidents.

5.1. Level and type of force used

The authority to employ force and to use situational variables to determine the appropriate level of force is one that differentiates law enforcement officers from many others. Law enforcement officers are granted the ability to use lethal or deadly force (resulting in death or great bodily harm) and non-lethal or less-lethal force options (designed to incapacitate a subject) and the U.S. Supreme Court has detailed the context in which either force option can be used by law enforcement (Lord, 2014). As described in Robinson (2011), the International Associ- ation of Chiefs of Police (IACP) standard for use of force response op- tions includes passive interference (police presence on scene), commands (verbal orders to the public), physical coercion (grabs but no physical strikes or kicks to the subject), incapacitation (body strikes, TASERs, OC sprays, or blunt objects), and deadly force.

Lord (2014) examined the influence of characteristics of SbC and non-SbC subjects on level of force used during legal intervention shoot- ings using the same NVDRS dataset from Lord (2012). SbC intent was determined by the presence of the primary indicators of suicidal intent criteria from Lord and Sloop (2010)—that is, indicators of verbal, behav- ioral, or planned intent. Of the officer-involved shooting death cases which provided enough information for consideration (n = 508), 262 (28.5% of the total sample) showed evidence of at least one primary in- dicator of suicidal intent. Use of forcewas broken down into the catego- ries described previously (passive interference, physical coercion, incapacitation, and deadly force), but physical coercion and incapacitation were combined into a category called use of low-lethal or physical restraints.

Bivariate analyses revealed that police officers used significantly different levels of force for SbC incidents than non-SbC incidents. SbC in- cidentsweremarked by greater use of commands and negotiation strat- egies than non-SbC incidents, and non-SbC incidents by significantly greater use of lethal force. Multivariate logistic regression revealed sev- eral significant predictors of police response, with the model including independent variables significantly improving the intercept-only model (p b .001, Nagelkerke = .413). For SbC subjects, police were sig- nificantly more likely to use commands and warnings instead of lethal action if the subjects were not experiencing interpersonal crises (eb3 = .426, p b .05), were intoxicated during the incident (eb = 1.93, p b .05), refused to surrender (eb =17.22, p b .01), were less aggressive (eb = .527, p b .01), and did not have a criminal history (eb = .461, p b .05). Negotiation was more likely than lethal action to be used if there was a domestic dispute in progress (eb = 2.72, p b .05) and if the SbC subject was intoxicated (eb = 5.32, p b .05). For the non-SbC

117C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

subjects, police were more likely to use low-lethal or physical restraints when subjects evidenced substance addiction or mental health prob- lems (eb=5.43, p b .001) butwere less likely to use themwhen subjects were aggressive (eb = .086, p b .001). Across SbC and non-SbC models, the command only and low-lethal or physical restraint force options pro- duced significantly different actions between SbC and non-SbC subjects—specifically, that SbC subjects were less likely to surrender during use of commands only and to be aggressive during cases using low-lethal or physical restraints. These results suggested that police are significantly more likely to use less-lethal response options with SbC subjects than non-SbC subjects, but because the author did not elucidate the nature of significant differences during bivariate analysis, themech- anism of the significance of this main effect goes unrecognized. This could have been remedied by converting multi-level categorical vari- ables into dichotomous ones to facilitate direct comparison.

Mohandie and Meloy (2010) examined a subset of hostage/ barricade (H&B) cases (n = 84) within their larger officer-involved shooting sample (n = 707) to understand not only the prevalence of SbC and subject descriptive characteristics, but also whether interven- tion efficacy differed significantly between SbC and H&B subjects. Fifty-five (66%) of theH&B caseswere classified as SbC,whichwas near- ly twice the likelihood of the larger sample. Eight (10%) were consid- ered to be completed or attempted suicides during a police encounter, and 21 (25%) as pure H&B (officer-involved shooting with hostages and/or a barricaded subject). Seventy-one percent of the H&B subjects (n=15), 25% of suicide-only subjects, and 32% (n=18) of the SbC sub- jects survived the encounter—representing a significant difference in fa- talities across groups, F = 5.708, p b .005. Though small cell sizes prevented the use of inferential statistics on many variables, there were 32 SbC fatalities (58%) and four H&B fatalities (19%) among those who received verbal interventions. Verbal efforts to de-escalate appeared to have little to no effect, as they successfully concluded only one of the H&B incidents and none of the SbC incidents. The effica- cy of less-lethal interventions between groups was not significantly different, F=.332, norwas there a significant difference in subject fatal- ity between incidents of longer (N1 h) and shorter (b1 h) duration, χ2 = .629.

These data suggest that when a subject in a hostage/barricade inci- dent appears to have SbC motivation, the odds that the subject survives are low. The researchers provide important information about how SbC subjects may be different than others also involved in H&B incidents, and present interesting conclusions regarding the real efficacy of inter- ventions long thought to be key in prevention of death during high-risk incidents. Further, they present innovative recommendations for police and policymaker action, including developing innovative suicide inter- vention programs while also teaching standard suicide protocols to all police officers—not just members of special response teams. However, the authors selected only H&B incidents where shots were fired by police and drew from a non-random sample of officer-involved shootings—potentially resulting in sampling bias. Therewasno validation of howwell negotiators in these incidentswere trained and noway to as- sess intervention quality, which could also have skewed results.

5.1.1. Summary Negotiation and/or physical restraint are more frequently used in

SbC incidents than non-SbC incidents, especially when the subject appears mentally ill or intoxicated. SbC subjects are not affected by many legal interventions, including verbal negotiation, physical re- straints, or less-lethal response options, even though these options are often used in SbC incidents. SbC subjects may actually become more aggressive during use of restraints than non-SbC subjects.

5.2. Impact of intervention on outcome.

Questions regarding whether standard law enforcement interven- tions can be used successfully with SbC subjects have led to greater

exploration of the ultimate outcome of SbC incidents. Using the same 64 cases of SbC obtained fromNorth Carolina law enforcement agencies described in Lord (2000, 2001) directly examined the impact of the legal intervention strategy on completed or averted SbC incidents. Rather than analyzing the tools used in these interventions (i.e., police used a TASER, physical restraint, etc.), Lord reviewed how police officers used tactical strategies (e.g., establishing a safe perimeter around the subject, using distraction devices) or verbal interventions (e.g., focusing on sub- ject weapon, talking about subject's problems) to de-escalate a SbC situation.

Establishing a perimeter was the intervention strategymost used by police (n=26, 40.6%). Other tactical strategies, including distraction de- vices (1.6%), physical restraint (7.8%) and OC gas (12.5%), were used much less often and did not appear to impact lethality of the outcome. Of all the strategies employed, discussing the subject's problems result- ed in the highest percentage of averted SbC (n = 26; 66.7%). These re- sults suggest that when no strategies are used at all, SbC events are likely to be short-lived with greater subject lethality—but when verbal negotiation strategies focusing on resolving the subject's problems are used after establishing a safe perimeter, the incidentmay be quickly re- solved and the subject physically restrained.

Lord's (2001)work is notable because it is among the few to focus on not just which legal interventions are used with SbC subjects, but also on how these tools are utilized. This study is also useful because it di- rectly relates intervention strategy to incident outcome, where other studies have simply detailed strategies used in their samples of SbC. As this research represents early forays into the study of SbC, the data is descriptive in nature, and sample size relatively small (n = 64). Al- though some potentially meaningful differences emerged between averted and completed SbC cases with regard to intervention strategy, others are unable to determine whether these differences are signifi- cantly different due to sample limitations. The inability to run inferential statistics on these results makes comparison of this work to other stud- ies of intervention efficacy difficult, if not impossible.

Estimating the ratio of lethal to less-lethal force used in SbC inci- dents is difficult for a number of reasons. First, many SbC samples in- clude only deceased subjects, thereby restricting analyses to only those subjects killed using lethal force. Second, of the sample including living subjects, very fewhave directly examined police response—and of thosewhich have, the level of force information specific to SbC cases has rarely been provided (Mohandie & Meloy, 2010). For these reasons, a summary of other research providing clarification on use of force is de- scribed here instead of with each section above. Many others have found that SbC incidents often result in subject death, with estimates of subject lethality ranging from 40.4%–92% (Lord & Sloop, 2010; Mohandie & Meloy, 2011). Like Mohandie and Meloy (2010) and Lord (2001), others have found that when less-lethal strategies are used, they have little effect (Hutson et al., 1998). Kesic et al. (2012) found that police weremore likely to use less-lethal methods, including nego- tiation, tactical disengagement, and cordon and containment, in SbC fa- talities than in non-SbC fatalities. Mohandie et al., (2009) also found that less-lethal force was more likely to be used with SbC subjects than other officer-involved shooting subjects, χ2 = 17.715, p b .001, but SbC subjects were more likely to be killed during the incident than other subjects, F = 25.458, p b .001—suggesting that less-lethal force may be initially deployed but is ultimately ineffective in SbC incidents. McLeod et al. (2014) explain that this may be the case not only because SbC subjects often have histories of aggressive behavior andnegative in- teractions with law enforcement, but also because the use of less-lethal force options may lead SbC subjects who are ambivalent about death to escalate their behavior to force a more lethal police response.

5.2.1. Summary Subject lethality in SbC incidents is high, especially when no law en-

forcement strategies are employed due to time or logistical constraints. Many legal interventions, including use of less-lethal devices, tactical

118 C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

disengagement, negotiation focused on dropping the weapon, and use of a cordon to separate from the threat, are largely ineffective with SbC subjects. When verbal negotiation strategies focusing on the subject's problems are used, the subject may be more amenable to event resolution, enabling use of physical restraints until he/she may be transported.

6. Discussion

6.1. Integrative summary

The results of the studies described in this review lendweight to the idea that SbC is an event worthy of extended study—one conceptually distinct from other types of officer-involved shootings. Perpetrators of SbC are often young adult White males who are single, divorced or widowed and who have significant mental health and criminal backgrounds. Many SbC incidents are short in duration and occur at a residential location, with the perpetrator frequently intoxicated. Precip- itating events are often romantic conflict, with many incidents indica- tive of recent or impending relationship termination. SbC perpetrators demonstrate suicidal intent through a unique combination of verbal, be- havioral, and planned indicators, with some perpetrators committing “outrageous acts” designed to elicit police response and others showing little, if any, suicidal ideation prior to the SbC incident. Deadly force is used in the majority of SbC incidents and many SbC subjects are killed by police after threatening to harm police or others. Inmany legal inter- ventions where the subject is an attempted or completed SbC, less- lethal force options (e.g., verbal negotiation strategies, physical re- straints or incapacitating devices like TASERs or batons) generally do not appear effective.

Comparison studies have shown that SbC perpetrators are signifi- cantly different from other subjects also involved in legal interventions. SbC subjects are more criminally active, attempt suicide less frequently, and receive less inpatient therapy than suicide-only subjects.When an- alyzed against other officer-involved shooting subjects, SbC perpetra- tors are more likely to display current symptoms of substance abuse and mental illness and to have attempted suicide more than once, but less likely to have a lengthy criminal history. Thus, if consideringmental health concerns and criminality on separate spectrums, the research seems to indicate that SbC perpetrators fall in the middle of both suicide-only and officer-involved shooting only subjects—such that they have significantly more mental health concerns and less criminal activity than non-SbC subjects, but less mental health concerns and greater criminal activity than suicide-only subjects. Like suicide-only subjects, they present with a host of psychological problems, including drug addiction and/or an inability to process an interpersonal crisis, but SbC subjects are more likely to refuse to surrender, more likely to be ag- gressive to others, and more likely to die during the event—suggesting that though they have actionable targets for mental health intervention, they are likely to be so resolute in their pursuit of suicide that they will not avoid hurting others to get it.

SbC research focusing on intervention efficacy has revealed interest- ing yet worrisome results. SbC incidents are often concluded by the subject's death, and when less-lethal use of force options are used, they are rarely effective. Even verbal negotiation strategies, which have been effective at resolving hostage/barricade incidents without SbC subject intent, fall short of resolving SbC incidents. Purposefully drawing out incident duration, which has long been thought to lead to positive outcomes in hostage/barricade situations (Mohandie & Meloy, 2010), did not lead to significant differences in subject fatality between SbC incidents lasting 1 h or more than 1 h. These results could suggest that SbC subjects are not diverted from their goals by police officers or special response teams because they may view police officers as a means to an end, or a tool used to achieve their goal to die by suicide—thus limiting the impact negotiators have on their survival. The only study to find an intervention efficacy of greater than 50% was

Lord (2001), where verbal interventions focused on discussions of the subject's problems resulted in averted SbC for over 60% of subjects. This may present important implications for police training, as typical verbal intervention strategies, including focusing on getting the subject to lower a weapon, may actually lead to increased likelihood that the SbC subject pursues with his/her agenda.

6.2. Common methodological limitations

Over the last 20 years, SbC research has informed clinicians and law enforcement officers about an event once thought to be one of the rarest forms of officer-involved shootings, but several common limitations to this body of research exist. First, SbC researchers have yet to agree on a conceptualization that may be used to classify offenders based on a common understanding of the necessary elements of SbC. For example, some researchers categorized cases as SbC when subjects harmed or threatened to harmotherswith the intent to elicit lethal police response (Mohandie & Meloy, 2010). However, some studies also include sub- jects who took their own lives during a standoff with police (Lord, 2001) or sometime after law enforcement interaction (Lord & Sloop, 2010). Other samples exclude incidents that resolvedwithout discharge of a lawenforcementweapon (Mohandie &Meloy, 2010, 2011), thereby eliminating those SbC incidents from consideration which are resolved without the use of lethal or less-lethal force.Whatmakes determination of suicidal intent even more difficult for cases of SbC is that its subjects do not communicate suicidal intent to anyone, andwhen they do, it very rarely reflects a desire to die explicitly by cop (Mohandie et al., 2009). Without a clear definition of what SbC is—for example, whether SbC also includes suicides completed during law enforcement interventions—it will be difficult for future researchers to determine what SbC is not.

A second common methodological issue evident in SbC research is how researchers evaluate and determine intent to die by SbC. Histori- cally, researchers have disagreed on the demonstrated level of intent re- quired for SbC case designation, with some utilizing distinct hierarchies of primary and secondary indicators of intent (Best et al., 2004; Homant & Kennedy, 2000) and others examining contextual and individual- level variables only to arrive at a “yes or no” decision regarding SbC clas- sification (Mohandie &Meloy, 2011; Mohandie et al., 2009). Still others have utilized separate hierarchies conceptualizing SbC as probable, pos- sible, uncertain, or improbable to describe degrees of likelihood that a subject desires death by SbC (Kennedy et al., 1998), and other re- searchers have focused explicitly on explaining SbC via a typology of of- fenders (Homant & Kennedy, 2000). Together, these studies do not arrive at a firm consensus on the best way to identify suicidal intent in SbC incidents, nor does there seem to be any agreement on which sys- tem to use until the “best way” is found.

Third, flawed coding procedures are evident in many of the SbC stud- ies reviewed here. In many earlier SbC studies, the person(s) responsible for coding variables like SbC intent are not listed. In others, no informa- tion is given about how disagreements between raters were resolved, and estimates of interrater reliability are not given (Best et al., 2004; Hutson et al., 1998), are provided as an estimate of percent agreement between raters (Kennedy et al., 1998; Homant & Kennedy, 2000) or meet a standard described only as “fair to good” (Dewey et al., 2013; Homant & Kennedy, 2000) by other researchers (Fleiss, Levin, & Paik, 2003). Later studies of SbC (Lord, 2012; Lord & Sloop, 2010) describe at- tempts for raters to code independent of one other, and some studies pro- vide stronger estimates of interrater reliability like kappa, which corrects for chance agreement between raters (Howell, 2010). However, the ma- jority of SbC research suffers from inconsistent, vague, or absent descrip- tions of data coding processes, with one exception being the research of Mohandie and Meloy, who provide clear and detailed descriptions of their coding procedure and have even made a copy of their codebook available to other researchers (Mohandie & Meloy, 2010, 2011; Mohandie et al., 2009).

119C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

A fourth limitation focuses on whether appropriate statistical analyses are included or described in sufficient detail. Across the SbC literature, practices recommended by the American Psychologi- cal Association (2010) for reporting the results of statistical analyses, including the provision of degrees of freedom (Lord, 2012;Mohandie & Meloy, 2011), confidence intervals (Lord & Sloop, 2010), standard deviations (Homant & Kennedy, 2000; Mohandie & Meloy, 2011), and effect sizes (Lord & Sloop, 2010; Mohandie et al., 2009), are often ignored. When effect sizes are absent, some researchers rely simply on p-values to determine the significance of results—a meth- od which is discouraged because it assumes that statistically signifi- cant results are also clinically significant (Howell, 2010). Many of the studies included in this review (e.g. Mohandie & Meloy, 2011, Mohandie et al., 2009) compare two or more groups of individuals on a number of variables without mention of an adjustment to ac- count for Type I error. Without this, it becomes likely that over time and with more comparisons, the groups will differ on at least one variable due to chance alone, resulting in a false positive or in- creased Type 1 error. A Bonferroni correction, which is the most con- servative option but is free of dependence and distributional assumptions, may have been an appropriate choice (Dunn, 1961). It is also unclear whether any of the authors utilizing multiple comparisons used a Fisher's exact test for categorical variables with small or uneven cell size. Finally, when there is missing data, some researchers (e.g. Hutson et al., 1998, Lord, 2014) do not discuss why this might be so, or what methods are used to address the problem.

One final limitation of the current research focuses on sampling pro- cedures. Like many other topics of interest subsumed within the “vio- lence and violent behavior” body of literature, the study of SbC is reliant on archival data. Information on suicidal subjects desiring death by SbC has been, unfortunately, hard to come by, so many early SbC researchers instead utilized non-random, convenience samples from state or local law enforcement agencies (Best et al., 2004; Lord, 2000), media sources (Kennedy et al., 1998), case files from medical examiners/attorneys (Wilson et al., 1998), or some combination of all of these (Homant & Kennedy, 2000). Conclusions drawn from these samples may not be representative, as they often emerge from one geographical location, or suspect, as with reliance on the media's portrayal of a SbC case. Many studies were conducted by the same researchers using the same sample but focusing on different variables of interest—thus lessening the impact of significant results and raising questions of validity (Lord, 2000; Mohandie & Meloy, 2011; Mohandie et al., 2009). Other researchers relied on data from areas documented to have higher violent crime rates than the national average (e.g., Los Angeles)—perhaps leading to conclusions that SbC in- cidents typical of this region are not representative of others in smaller, less densely populated cities (Los Angeles Times, 2011).

Later studies used SbC data from large national databases of either hostage/barricade incidents (HOBAS; Lord & Sloop, 2010; Mohandie & Meloy, 2010) or violent deaths (NVDRS; Lord, 2012, 2014). These stud- ies illustrate a departure from local or composite data and are potential- ly more representative of SbC than earlier datasets, but still are restricted because they capture only those SbC cases where special re- sponse teams are deployed or in which there are no survivors. Lipetsker (2004) cautioned against conclusions that HOBAS data is rep- resentative of offenders in general, as data is collected only from those police departments who volunteer to provide it, and also described how HOBAS data is only as good as the information police officers choose to provide in each case. Mohandie and Meloy (2010) mention that cases from HOBAS may likely reflect a sampling bias toward positively-resolved cases due to data submission from well-established special response teams. The most recent work from Mohandie and Meloy (Mohandie & Meloy, 2011; Mohandie et al., 2009) utilizes a sam- ple of officer-involved shootings with a small subset of cases identified as SbC, and is, again, representative of what may be the “gold standard” with regard to SbC research—a large international sample including

men and women and attempted and completed SbC, with the option to correspond with the responsible law enforcement agency on cases if needed. Still, this sample ismade up primarily of SbC caseswith fatalities, which may imply a selection bias for cases with more violent behavior, less ambivalence, or greater media attention—again highlighting the need for large, diverse datasets from other geographical areas.

6.3. Strengths and limitations of the current review

This review of the literature captures early descriptive work on SbC, aswell as later empirical studies of the phenomenon. Priority is given to later empirical work, which has yet to be reviewed and summarized for the purposes of directing researchers toward future aims. The stud- ies here are not just reviewed, but critiqued for methodological limitations—a process necessary to elevate consistent problems and to improve the quality of research emerging within this field. Al- though others have reviewed the SbC research prior to 2000 (Mohandie & Meloy, 2000) or have more generally summarized a few qualitative studies of SbC (McKenzie, 2006), none but this study have described more recent advances in the study of SbC which employ statistical analyses to understand event characteris- tics. Finally, the results of this review may be used to provide recom- mendations for law enforcement professionals on how to respond in potential SbC incidents—for example, how to adapt existing verbal negotiation tactics to focus on subject problem resolution rather than on subject separation from his/her weapon.

Despite these strengths, limitations exist which may affect the im- pact of this review. Early qualitative work was excluded to enable a focus on studies using statistical analysis. This may limit the breadth and utility of this review when it is used to explore the progression of SbC research. Additionally, this review includes only 18 articles, which is less than desirable butwhichmay be reflective of the dearth of empir- ical studies of SbC in the last 20 years. Though the variables included for conceptualization were ones most commonly described by other re- searchers, there may be other individual, situational, or intervention- type variables of interest that are absent in this review. For example, sit- uational variables including time of day, season, and/or environmental stimuli present during the SbC attempt, were excluded due to their mention in only one study (Lord & Gigante, 2004), but these might be of interest to other researchers. Other variables yet to be studied in de- tail, including history of aggressive behavior or motivation for SbC, may be studied in later reviews once they become available.

6.4. Future directions

As we approach the twentieth year of empirical research on SbC, there are still a few things we do not know. SbC has been investigated in the United States, Canada, England, Australia, andWales, but no stud- ies have emerged from other countries. Researchers should take care to understand how andwhy SbCmay be different in nations outsidewest- ern influence. Comparisons of completed to averted SbC incidents have so far led to the conclusion that these two groups are not very different from one another (outside of greater suicidal intent in the former), but this conclusion was formed using a small sample with SbC cases from a concentrated area in the southeastern United States (Lord, 2000, 2001). Newer studies may benefit from an analysis of what makes averted SbC subjects amenable to legal interventions, with the potential for the re- sults to change the way police respond to and intervene in SbC inci- dents. Similarly, no researcher has yet studied suicidal motivations in a sample of SbC attempters. Doing so may yield information about why existing legal interventions are ineffective, and could speak tomo- tivations behind SbC only alluded to in previous research (Homant et al., 2000). The provision of additional legal intervention tools, like a crisis response unit from a local mental health referral agency, may lead to not only a reduction in completed SbCs, but also a potential increase in police confidence in mitigating aggressive behavior in individuals

120 C.L. Patton, W.J. Fremouw / Aggression and Violent Behavior 27 (2016) 107–120

with mental illness. Future researchers may choose to investigate the impact of improving service delivery from such agencies, as suggested by Ogloff et al. (2013) and Kesic and Thomas (2014). Finally, others may choose to conduct an empirical investigation of intervention effica- cy to determine which individual-level variables (e.g., age, gender, in- toxication) influence intervention success.

References

Amendola, K., Learning, M., & Martin, J. (1996). Analyzing characteristics of police–citizen encounters in high-risk search warrant issuances, domestic disturbances, hostage and barricaded persons incidents, and encounters with fleeing felony suspects (Grant 92-IJ- CX-K019). Washington D.C.: Police Foundation.

American Association of Suicidology (2013). Suicide by cop fact sheet. Retrieved from http://www.suicidology.org/Portals/14/docs/Resources/FactSheets/2011/ SuicideCop2014.pdf

American Foundation for Suicide Prevention (2014). Facts and figures. Retrieved from https://www.afsp.org/understanding-suicide/facts-and-figures.

American Psychological Association (2010). Publication manual of the American Psycholog- ical Association. Washington, D. C.: American Psychological Association.

Arias, E. A., Schlesinger, L. B., Pinizzotto, A. J., Davis, E. F., Fava, J. L., & Dewey, L. M. (2008). Police officers who commit suicide by cop: A clinical study with analysis. Journal of Forensic Sciences, 53, 1455–1457. http://dx.doi.org/10.1111/j.1556-4029.2008.00861.x.

Beck, A. T., Weissman, M. A., & Kovacs, M. (1976). Alcoholism, hopelessness and suicidal behavior. Journal of Studies on Alcohol, 37, 66–77.

Best, D., Quigley, A., & Bailey, A. (2004). Police shooting as a method of self-harming: A review of the evidence for “suicide by cop” in England and Wales between 1998 and 2001. Institutional Journal of the Sociology of Law, 32, 349–361. http://dx.doi. org/10.1016/j.ijsl.2004.09.002.

Bresler, S., Scalora, M. J., Elbogen, E. B., & Scott Moore, Y. (2003). Attempted suicide by cop: A case study of traumatic brain injury and the insanity defense. Journal of Forensic Sciences, 48, 1–6.

Dewey, L., Allwood, M., Fava, J., Arias, E., Pinizzotto, A., & Scheslinger, L. (2013). Suicide by cop: Clinical risks and subtypes. Archives of Suicide Research, 17, 448–461. http://dx. doi.org/10.1080/13811118.2013.801810.

Dunn, O. J. (1961). Multiple comparisons among means. Journal of the American Statistical Association, 56(293), 52–64. http://dx.doi.org/10.1080/01621459.1961.10482090.

Falk, J., Riepert, T., & Rothschild, M. A. (2004). A case of suicide by cop. Legal Medicine, 6, 194–196. http://dx.doi.org/10.1016/j.legalmed.2004.04.001.

Fleiss, J. L., Levin, B., & Paik, M. C. (2003). Statistical methods for rates and proportions. New York, NY: John Wiley & Son, Inc.

Geberth, V. J. (1990). Suicide by cop: Inviting death from the hands of a police officer. Law and Order, 105-108.

Hammer, M. R. (2007). Saving lives: The S.A.F.E. model for resolving hostage and crisis inci- dents. Westport, CT: Greenwood Publishing Group, Inc.

Homant, R. J., & Kennedy, D. B. (2000). Suicide by police: A proposed typology of law en- forcement officer assisted suicide. Policing: An International Journal of Police Strategies and Management, 23, 339–355.

Homant, R. J., Kennedy, D. B., & Hupp, R. T. (2000). Real and perceived threat in police officer-assisted suicide. Journal of Criminal Justice, 28, 43–52.

Howell, D. C. (2010). Statistical methods for psychology. Belmont, CA: CengageWadsworth. Hutson, H. R., Anglin, D., Yarbrough, J., Hardaway, K., Russell, M., ... Blum, B. (1998). Sui-

cide by cop. Annals of Emergency Medicine, 32, 665–669. Kennedy, D. B., Homant, R. J., & Hupp, R. T. (1998). Suicide by cop. FBI Law Enforcement

Bulletin, 67, 21–28. Kesic, D., & Thomas, S. D. M. (2014). Do prior histories of violence and mental disorders

impact on violence behavior during encounters with police? International Journal of Law and Psychiatry, 37, 409–414.

Kesic, D. K., Thomas, S. D. M., & Ogloff, J. R. P. (2012). Analysis of fatal police shootings: Time, space, and suicide by police. Criminal Justice and Behavior, 39, 1107–1125. http://dx.doi.org/10.1177/0093854812440084.

Lindsay, M., & Lester, D. (2004). Suicide by cop: Committing suicide by provoking the police to shoot you. Amityville, NY: Baywood.

Lindsay, M., & Lester, D. (2008). Criteria for suicide by cop incidents. Psychological Reports, 102, 603–605.

Lipetsker, A. (2004). Evaluating the Hostage Barricade Data System (HOBAS). Journal of Police Crisis Negotiations, 4, 3–27. http://dx.doi.org/10.1300/J173v04n02_02.

Lord, V. (2001). Comparison of situation and tactical strategies of successful and unsuc- cessful SbC incidents. In D. C. Sheehan, & J. I. Warren (Eds.), Suicide and law enforce- ment (pp. 607–626). Washington D.C.: Behavioral Science Unit, FBI Academy.

Lord, V. B. (2000). Law enforcement assisted suicide. Criminal Justice and Behavior: An International Journal, 27, 401–418. http://dx.doi.org/10.1177/0093854800027003007.

Lord, V. B. (2012). Factors influencing subjects' observed level of suicide by cop in- tent. Criminal Justice and Behavior, 39, 1633–1646. http://dx.doi.org/10.1177/ 0093854812456689.

Lord, V. B. (2014). Police responses in officer-involved violent deaths: Comparison of sui- cide by cop and non-suicide by cop incidents. Police Quarterly, 17, 79–100. http://dx. doi.org/10.1177/1098611114522040.

Lord, V. B., & Gigante, L. (2004). Comparison of strategies used in barricaded situations: SbC and non-SbC subjects. In V. Lord (Ed.), Suicide by cop: Inducing officers to shoot (pp. 203–230). Flushing: Looseleaf Law Publications.

Lord, V. B., & Sloop, M. W. (2010). Suicide by cop: Police shooting as a method of self- harming. Journal of Criminal Justice, 38, 889–895. http://dx.doi.org/10.1016/j. jcrimjus.2010.06.004.

Los Angeles Times (2011, November). L.A.’s triumph over crime. Retrieved from http://articles.latimes.com/2011/nov/26/opinion/la-ed-crime-20111126.

MacCallum, R. C., Widaman, K. F., Zhang, S., & Hong, S. (1999). Sample size in factor anal- ysis. Psychometric Methods, 14, 84–99.

McKenzie, I. (2006). Forcing the police to open fire: A cross-cultural/international exam- ination of police-involved, victim-provoked shootings. Journal of Police Crisis Negotiations, 6, 5–25.

McLeod, L. C., Thomas, S. D. M., & Kesic, D. (2014). The frequency and nature of resolution of potential police provoked shooting encounters. International Journal of Law and Psychiatry, 37, 383–389. http://dx.doi.org/10.1016/j.ijlp.2014.02.2009.

McNally, M. R., Patton, C. L., & Fremouw, W. J. (2015). Mining for murder-suicide: An ap- proach to identifying cases of murder-suicide in the National Death Reporting System (NVDRS) Restricted Access Database (RAD). Journal of Forensic Sciences, 1-4. http://dx. doi.org/10.1111/1556-4029.12887.

Mohandie, K., & Meloy, J. R. (2000). Clinical and forensic indicators of “suicide by cop.”. Journal of Forensic Sciences, 45, 384–389.

Mohandie, K., & Meloy, J. R. (2010). Hostage and barricade incidents within an officer- involved shooting sample: Suicide by cop, intervention efficacy, and descriptive char- acteristics. Journal of Police Crisis Negotiations, 10, 101–115. http://dx.doi.org/10.1080/ 15332581003799737.

Mohandie, K., & Meloy, J. R. (2011). Suicide by cop among female subjects in officer- involved shooting cases. Journal of Forensic Sciences, 56, 664–668. http://dx.doi.org/ 10.1111/j.1556-4029.2010.01686.x.

Mohandie, K., Meloy, J. R., & Collins, P. I. (2009). Suicide by cop among officer-involved shooting cases. Journal of Forensic Sciences, 54, 456–562. http://dx.doi.org/10.1111/j. 1556-4029.2008.00981.x.

Ogloff, J. R. P., Thomas, S. D. M., Luebbers, S., Baksheev, G., Elliott, I., ... Moore, E. (2013). Policing services with mentally ill people: Developing greater understanding and best practice. Australian Psychologist, 48, 57–68.

Parent, R. B. (1998a). Suicide by cop: Victim-precipitated homicide. The Police Chief, 65, 111–114.

Pinizzoto, A. J., Davis, E. F., & Miller, C. E. (2005). Suicide by cop: Defining a devastating dilemma. FBI Law Enforcement Bulletin, 8–20.

Robinson, J. (2011). The IACP use-of-force continuum. Retrieved from http://www.ehow. com/facts_7214778_iacp-use-of-force-continuum.html

Tabachnick, B. G., & Fidell, L. S. (2013). Using multivariate statistics. New Jersey: Pearson Education, Inc.

Wilson, E. F., Davis, J. H., Bloom, J. D., Batten, P. J., & Kamara, S. G. (1998). Homicide or sui- cide: The killing of suicidal persons by law enforcement officers. Journal of Forensic Sciences, 43, 46–52.