Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Nonsuicidal self-injury in street-involved adolescents : identification of risk and protective factors Laye-Gindhu, Aviva Mia 2011

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


24-ubc_2011_fall_layegindhu_aviva.pdf [ 1.44MB ]
JSON: 24-1.0072047.json
JSON-LD: 24-1.0072047-ld.json
RDF/XML (Pretty): 24-1.0072047-rdf.xml
RDF/JSON: 24-1.0072047-rdf.json
Turtle: 24-1.0072047-turtle.txt
N-Triples: 24-1.0072047-rdf-ntriples.txt
Original Record: 24-1.0072047-source.json
Full Text

Full Text

NONSUICIDAL SELF-INJURY IN STREET-INVOLVED ADOLESCENTS: IDENTIFICATION OF RISK AND PROTECTIVE FACTORS by Aviva Laye-Gindhu  B.A., McGill University, 1992 M.A., University of British Columbia, 2002  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (School Psychology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  August 2011 © Aviva Laye-Gindhu, 2011  ii ABSTRACT Nonsuicidal self-injury (NSSI), or the deliberate, direct, self-inflicted injury to body tissue that occurs in the absence of suicidal intent and developmental disabilities, is a serious and increasingly prevalent health risk among adolescents. Evidence suggests that vulnerable adolescents, such those that are street-involved, are at high risk for negative health outcomes, including NSSI. Using a theoretically and empirically derived model of risk and resilience, this study is the first to identify a broad range of risk and protective factors associated with NSSI. The study involved secondary analysis of data gathered using the Street-Involved Youth Health Survey (SYHS) with a sample of 762 adolescents aged 12-18 across British Columbia. Prevalence of NSSI was 56% and 34% for females and males, respectively, with sexual minority youth three times more likely to report this behaviour. Results from a series of logistic regression analyses revealed different models of risk and protection for males and females. At the multivariate level, the strongest risk factors for females were previous suicide attempt, risky behaviour, experiencing more consequences of substance use, sexual abuse by two or more perpetrators, and maternal problems. For males, the strongest risk factors were previous suicide attempt, risky behaviour, and being victim of relational aggression. The strongest modifiable protective factors for males and females were better emotional health and family connectedness, with school peer relations for males and subjective health status for females also showing significance. Probability profiles created from different combinations of the final set of salient factors highlight not only the multiplicative risk at play in these youths` lives but also the impact of protective factors to offset risk. For boys, with three risk factors (RF) and three protective factors (PF), the likelihood of NSSI ranged from 9% (0-RF, 3-PF) to 90% (3-RF, 0-PF). For girls, with three risk factors and two protective factors, the likelihood of NSSI ranged from 19% (0-RF, 2-PF) to 99% (3-RF, 0-PF). Profiles in this study underscore the value of risk and protection as  iii powerful tools for developing the knowledge base on NSSI and for guiding prevention and intervention efforts.  iv PREFACE Ethical approval for both the original McCreary Centre Society study (Smith et al., 2007) and for the present secondary analysis was granted by the University of British Columbia‘s Behavioural Research Ethics Board (UBC-BREB), certificate numbers H06-80235 and H07-03127, respectively.  v TABLE OF CONTENTS  Abstract.............................................................................................................................. ii Preface ..............................................................................................................................iv Table of Contents............................................................................................................... v List of Tables ................................................................................................................... viii List of Figures ....................................................................................................................ix Acknowledgements ............................................................................................................ x CHAPTER 1: INTRODUCTION ......................................................................................... 1 Purpose of the Study .................................................................................................... 2 Specific Research Questions .................................................................................. 3 Significance of the Study .............................................................................................. 4 Definition of Terms ....................................................................................................... 5 Overview of the Literature Review ................................................................................ 9 CHAPTER 2: REVIEW OF THE LITERATURE ............................................................... 11 Homeless and Street-Involved Youth ........................................................................ 11 Scope and Description ......................................................................................... 11 Street-Involved Youth, Family and Social Context ............................................... 12 Street-Involved Youth and Health ......................................................................... 15 Summary .............................................................................................................. 19 Nonsuicidal Self-Injury................................................................................................ 19 Conceptualizing Self-Injury ................................................................................... 20 Self-Injury versus Suicide ..................................................................................... 21 Diagnostic Issues ................................................................................................. 23 Prevalence and Sociodemographics of Self-Injury ................................................ 24 Nature of Self-Injury.............................................................................................. 26 Motivational Aspects and Function of Self-Injury .................................................. 27 Etiological and Theoretical Perspectives of Self-Injury.......................................... 30 Summary .............................................................................................................. 33 Street Youth and Self-Injury ....................................................................................... 34 Research on Self-Injury Among Street-Involved Youth ......................................... 35 Coping Among Street-Involved Youth: Links to Self-Injury .................................... 37 Resilience Theory: Links to Street Youth and Self-Injury ...................................... 38  vi Self-Injury in Adolescence: Associated Factors .......................................................... 41 Contextual Factors ............................................................................................... 42 Individual Factors ................................................................................................. 50 Cumulative Summary ................................................................................................. 57 CHAPTER 3: METHOD .................................................................................................. 59 Sample ....................................................................................................................... 59 Procedure .................................................................................................................. 60 Instrument and Variables ........................................................................................... 62 Street Youth Health Survey (SYHS) ..................................................................... 62 Outcome Variable: Self-Injury ............................................................................... 65 Potential Risk and Protective Factors ................................................................... 66 Data Management Issues .......................................................................................... 82 Data Analytic Plan ...................................................................................................... 83 CHAPTER 4: RESULTS ................................................................................................. 89 Descriptive Findings ................................................................................................... 89 Characteristics of the Self-Injury Subsample ........................................................ 89 Distribution of Risk and Protective Factors Against Self-Injury.............................. 93 Risk and Protective Factors Associated With Self-Injury ............................................ 97 Testing for Multicollinearity ....................................................................................... 101 Multivariate Models for Self-Injury ............................................................................ 103 Probability Profiling .................................................................................................. 107 Further Analyses for Females .................................................................................. 112 CHAPTER 5: DISCUSSION .......................................................................................... 117 Characteristics of Street-Involved Adolescents Who Self-Injure ............................... 118 Prevalence and Sociodemographics .................................................................. 118 Reasons for Self-Injury ....................................................................................... 121 Risk and Protective Factors Associated With Self-Injury .......................................... 122 Protective Factors .............................................................................................. 123 Risk Factors ....................................................................................................... 130 Probability Profiles .............................................................................................. 136 Strengths and Limitations of the Study ..................................................................... 138 Future Directions for Research ................................................................................. 142 Implications for Clinical Practice ............................................................................... 144  vii REFERENCES .............................................................................................................. 150 APPENDICES ................................................................................................................ 168 A: Source of Questions on the Street-Involved Youth Health Survey ............................. 168 B: Table of Hypothesized Risk and Protective Factors ................................................... 173  viii LIST OF TABLES  Table 1. Hypothesized Risk and Protective Factors for Self-Injury ................................. 43 Table 2. Outcome Variable: Self-Injury Measured ......................................................... 66 Table 3. Varimax Rotation of Three-Factor Solution for Family Connectedness ............ 74 Table 4. Zero-order Correlations for School Connectedness Scale ................................ 76 Table 5. Varimax Rotation of Two-Factor Solution for School Connectedness ............... 77 Table 6. Varimax Rotation of One-Factor Solution for Low Emotional Distress .............. 81 Table 7. Reasons Reported for Last Incident Self-Injury by Gender ............................... 91 Table 8. Risk and Protective Factors Associated with Self-Injury for Females ................ 94 Table 9. Risk and Protective Factors Associated with Self-Injury for Males .................... 96 Table 10. Bivariate Logistic Regression Analyses for Females: Risk and Protective Factors for Self-Injury ....................................................................................... 98 Table 11. Bivariate Logistic Regression Analyses for Males: Risk and Protective Factors for Self-Injury ....................................................................................... 99 Table 12. Age-Adjusted Partial Correlations Among Self-Injury and Protective Factors by Gender.......................................................................................... 101 Table 13. Age-Adjusted Partial Correlations Among Self-Injury and Risk Factors by Gender ...................................................................................................... 102 Table 14. Multivariate Logistic Regression Models of Risk and Protection for Females ......................................................................................................... 104 Table 15. Multivariate Logistic Regression Models of Risk and Protection for Males ...... 105 Table 16. Final Combined Multivariate Logistic Regression Model of Risk and Protection ....................................................................................................... 106 Table 17. Probability Profiles of Self-Injury for Females ................................................ 108 Table 18. Probability Profiles of Self-Injury for Males .................................................... 110 Table 19. Multivariate Logistic Regression Protection-Only Model for Females............. 114 Table 20. Final Combined Multivariate Logistic Regression Model for Females ........... 114 Table 21. Probability Profiles of Self-Injury for Females with Two-Protective Factors .... 115  ix LIST OF FIGURES  Figure 1. Overview of Data Analytic Plan ....................................................................... 88 Figure 2. Probability Profile for Self-Injury Among Females ......................................... 108 Figure 3. Probability Profile for Self-Injury Among Males ............................................. 112 Figure 4. Probability Profile for Self-Injury Among Females ......................................... 116  x ACKNOWLEDGEMENTS I would first like to acknowledge that this study was supported by a Doctoral Fellowship from the Social Sciences and Humanities Research Council (SSHRC) of Canada. My deepest gratitude to my Research Supervisor, Dr. Kim Schonert-Reichl, for her enduring support and leadership. I am indeed fortunate to have been guided by a trio of brilliant and wonderful researchers. I‘d like to take this opportunity to thank all three: Dr. Schonert-Reichl, along with Drs. Elizabeth Saewyc and Lynn Miller, for their time, encouragement, and thoughtful guidance. All three stimulated my curiosity while maintaining important focus on the practical elements of the job. I am deeply appreciative of their patience and support during this journey. I thank Dr. Saewyc, in her capacity as Research Director of the McCreary Centre Society, for granting me the permission to access the McCreary Centre survey data and for her ongoing guidance, exacting questions, and insightful feedback along the way. I would be remiss if I did not acknowledge the multitude of people who have played a role in my general development as a scholar, researcher, and clinician. I am grateful to my supervisors, colleagues, and to the children, youth, and families I‘ve had the privilege to meet and learn from. I am appreciative of the opportunities I‘ve had for research with the McCreary Centre Society, first with Aileen Murphy, and then under the direction of Dr. Elizabeth Saewyc, culminating in this undertaking. My long term in Dr. Lynn Miller‘s research lab not only taught me many important lessons about research with children and adolescents, but also about perseverance. And last, but not at all least, I am extremely fortunate to have had the unconditional support of my immediate and extended family and friends. Thank you.  1 CHAPTER 1: INTRODUCTION An estimated 150,000 Canadian and more than 1.3 million American youth are homeless on any given night (approximately 4% and 6% of the youth population aged 10-19 of each country as per population estimates by the US Census Bureau (2006) and by Statistics Canada (2006; DeMatteo et al., 1999; Hammer, Finkelhor, & Sedlak, 2002); a considerably greater number are street-involved and at risk of homelessness through precarious living conditions. Finding themselves precociously independent (Whitbeck et al., 1999; p. 276), street youths‘ lives are fraught with extreme stress and danger at a time when they are navigating their way through adolescence, a developmental period that can be challenging under the best of circumstances (Arnett, 1999; Schulenberg, Maggs, & Hurrelman, 1997). Marginalized and often disconnected from families and mainstream society, these youth lack important social resources and opportunities (Kidd, 2003b; Molnar, Shade, Kral, Booth, & Watters, 1998). Coping skills, critical to overcoming adversity, are acquired over the course of development; still adolescents, street youth‘s coping capacities are frequently overwhelmed (Ayerst, 1999; Cauce, 2000; Unger, Kipke, Simon, Montgomery, & Johnson, 1997; Whitbeck et al., 1999). Together, their individul vulnerabilities and the general lack of contextual factors to protect against the adversities of street life result in increased rates of mental health problems (Ayerst, 1999; Buckner & Bassuk, 1997; Cauce, 2000; Hughes et al., 2010; Ryan, Kilmer, Cauce, Wantanabe, & Hoyt., 2000; Stewart et al., 2004; Whitbeck et al., 1999; Whitbeck, Johnson, Hoyt, & Cauce, 2004) and a multitude of health-compromising behaviours that pose serious risks to well-being (Kidd, 2006; Laye, Murphy, & The McCreary Centre Society, 2002; Murphy & The McCreary Centre Society, 2001; Molnar et al., 1998; Rachlis, Wood, Zhang, Montaner, & Kerr, 2010; Smith et al., 2007; Whitbeck et al., 2000; Yoder, Hoyt, & Whitbeck, 1998).  2 The present study focused on nonsuicidal self-injury (NSSI), a perplexing healthcompromising behaviour that has been virtually ignored in the extant literature on streetinvolved adolescents, despite evidence suggesting significant prevalence in this developmental period. For the current purposes, self-injury is defined as deliberate, selfinflicted, physical injury that is direct and nonsuicidal in intent (e.g., cutting, carving, burning, self-hitting; Favazza, 1996; Walsh, 2006). This omission is especially salient if one considers the fact that NSSI is a serious public health issue with links to the spread of blood-borne disease through the sharing of tools used to injure (DeMatteo et al., 1999; DiClemente, Ponton, & Hartley, 1991), social contagion factors (Heath, Ross, Toste, Charlebois, & Nedecheva, 2009; Lloyd-Richardson et al., 2007; Walsh & Rosen, 1988), increased mortality rates (by suicide) and suicide attempts (Brausch & Gutierrez, 2010; Brunner et al., 2007; Muehlenkemp & Gutierrez, 2007; Whitlock & Knox, 2007), increased use of health and mental health care resources, and reduced personal productivity (Cloutier, Martin, Kennedy, Nixon, & Muehlenkamp, 2010; Nock & Prinstein, 2004; Walsh, 2006). As a high-risk behaviour that is strongly related to, but distinct from, suicide (Brausch & Gutierrez, 2010; Muehlenkamp & Gutierrez, 2004; Nock, Joiner, Gordon, Lloyd-Richardson, & Prinstein, 2006; Stanely, Gameroff, Michalson, & Mann, 2001; Whitlock & Knox, 2007) and indicative of serious distress and/or some level of psychopathology (Laye-Gindhu & Schonert-Reichl, 2005; LloydRichardson, Perrine, Dierker, & Kelley, 2007; Nixon, Cloutier, & Jansson, 2008; Nock et al., 2006; Patton et al., 1997; Ross & Heath, 2002), self-injury presents serious challenges to health and social service providers and to policy-makers seeking to help street-involved adolescents. In order to develop appropriate intervention and prevention efforts, it is critical to first gain an understanding of NSSI in this particularly vulnerable population. Purpose of the Study The current investigation examined nonsuicidal self-injury in an understudied, marginalized, and high-risk large systematic sample of street-involved adolescents (N = 762).  3 The study had three primary objectives: 1) to provide further information on the prevalence and demographics of self-injury and the reasons for the behaviour in this sample; 2) to empirically identify variables, risk and protective, that are independently associated with selfinjury; and 3) to provide insight into the relations among constellations of risk and protective factors and self-injury in order to elucidate evaluative or probability profiles of the likelihood of self-injury (as an outcome) given the presence of various combinations of risk and protection at both the contextual (i.e., environmental) and individual (i.e., personal) levels (Cichetti & Lynch, 2003; Luthar et al., 2000). Specific Research Questions The study addressed the following research questions: 1. a) What is the reported prevalence rate of self-injury across different demographic variables (e.g., ethnicity, sexual orientation)? Are some categories associated with increased risk of self-injury? It is expected that sexual minority youth, as a further marginalized and vulnerable population, will report higher rates of NSSI. b) What reasons for self-injuring are reported by the youth? Are different reasons reported across gender? 2. What is the relation between hypothetical, theoretically and empirically derived individual and contextual risk factors and self-injury? Protective factors and self-injury? Do these relations differ by gender? Which factors, when present, increase the likelihood of selfinjury as an outcome? And, conversely, which factors, when present, are potentially protective against self-injury (i.e., reduce likelihood of self-injury)? Given these various risk and protective factors, which are the strongest predictors of self-injury? 3. Given specific combinations of the strongest modifiable risk and protective factors, what is the probability or likelihood of self-injury as an outcome? It is expected that multiplicative risk, meaning more risk factors, will increase the likelihood of self-injury and this relation will be especially pronounced with the absence of protective factors. Conversely,  4 protective factors will decrease the likelihood of self-injury, and this will hold true even when risk factors are present. Significance of the Study Despite increasing prevalence rates and growing research and public interest in both street-involvement and in NSSI among adolescents, the systematic and empirical study of these areas, particularly the latter, is still relatively nascent. An exhaustive search found only four other studies involving homeless or street-involved youth (up to age 26) and any examination of self-injury (Kidd & Kral, 2002; Tyler, Melander, & Almazan, 2010; Tyler, Whitbeck, Hoyt & Johnson, 2003; Unger, Kipke, Simon, Montgomery, & Johnson, 1997). Only one study conducted by Tyler and colleagues (2003) provided an explicit examination of NSSI in a purely adolescent sample, albeit narrower in scope than the current investigation. In describing the health of street-involved youth in British Columbia, Smith et al. (2007) reported on a range of outcomes, including self-injury; the current study represents a secondary analysis of this data. Taken together, the limited evidence suggested that NSSI is a significant but overlooked health concern among this group of vulnerable young people. Notably, consistent with a paradigmatic shift in thinking about problems, the study is grounded in a theoretically and empirically derived model of risk and resilience that suggests both are worthy of our attention and that any best practice intervention will move beyond the common deficit perspective to include a dual focus on both reducing risk and promoting resilience (Bearinger, Pettingell, Resnick, & Potthoff, 2010; Blum, McNeely, & Nonnemaker, 2002; Luthar, 1993; Rutter, 1987; Zimmerman & Arunkumar, 1994). Gutierrez, a leading suicide researcher, proposed that ―…assessing risk and protective factors in combination has the best probability of informing the field‘s understanding…‖ (2006, p. 129). He suggests that by failing to do so, we are privy to only half the picture; given that both challenges and strengths are evident across the life-span, the integration of both risk and protection offers a  5 more realistic picture. Despite these clear clarion calls, research involving resilience factors, especially with high-risk youth, has been slow to emerge (for notable exceptions see Erdem & Slesnick, 2010; Kidd, 2003b; Rew & Horner, 2003; Smith et al., 2007). Moving beyond description by utilizing both bivariate and multivariate methods to examine the phenomenon among street-involved adolescents, results from this study augment the extant literature on NSSI and draw warranted attention to the clinical importance of this behaviour in the street youth population. If we hope to understand, intervene effectively, and ultimately prevent nonsuicidal self-injury, an important step lies in the identification of risk and protective factors associated with this behaviour. Acknowledgement and inclusion of protective factors in the midst of all the risk provides a more complete and holistic portrait of these youth. Definition of Terms Adolescence Adolescence refers to the critical period of physical, social, emotional, biological, and cognitive development that occurs between childhood and adulthood. According to the medical field, it is initiated with the onset of puberty, although it seems that no single event or line specifically denotes its beginning and its end is equally vague, sometimes referred to as ‗maturity.‘ The adolescent period is variously described as being between 10, 11, or 12 and 19, 20, or 21 (World Health Organization, 2004; Crockett & Peterson, 1993). In providing guidelines, the Canadian Pediatric Society (2003) suggests that the move from dependence to independence is crucial, stating, ―…[adolescence] terminates when an adult identity and behaviour are accepted.‖ Youth Whereas the term adolescence has been adopted by the medical field and many social science disciplines, the term youth is more commonly used on the ‗front lines‘ and in education, criminology, sociology, and in the child and youth care fields. Youth is sometimes  6 used to refer to a broader age range, extending into young adulthood, as compared to adolescence; for example, the UN refers to youth as ages 15 to 24 inclusive (UN for Youth, 2007, Much of the research on streetinvolvement and homelessness with this age group uses the term youth to describe the population. Therefore, for current purposes, the terms adolescence or adolescent and youth will be used interchangeably. The only exception will be in relating results from research in which the researchers are clear that they are referring to one or the other. Street-Involvement, Street-Involved, Street Youth Street-involved youth or adolescents are young people who: are precariously housed (e.g., temporary housing, shelter, or unsafe housing), and/or at risk of being homeless or without shelter, and/or using services designed for street youth, and/or whose identities, support networks, activities, and lifestyle are those in which the street figures prominently. Although some youth may live with a parent(s) or guardian(s) or with friends, a subgroup is homeless and many are homeless at some point during their street-involvement. Differentiations between street-involvement and homelessness are difficult and may be artificial in that these are cyclical circumstances with substantial fluidity between them. Homelessness Homelessness is used to refer to both having temporary or insecure shelter (ranging from living in a homeless shelter or safe-house to squatting or staying in substandard or unsafe housing) as well as sleeping outdoors ‗in the rough.‘ After concluding that operationalizing adolescent homelessness is a difficult task, Haber and Toro (2004) suggest that the presence or absence of a caregiver(s) is a key aspect in this determination. Some researchers refer to different levels of homelessness, including absolutely homeless, or lacking any physical shelter and literally living on the street, and relatively homeless, or having temporary or periodic, precarious, or substandard housing. The terms homeless youth and street (or street-involved) youth often describe the same population of young  7 people and therefore the terms will be used interchangeably. However, research that refers explicitly to either homeless, runaway, or street youth will be described accordingly. Nonsuicidal Self-Injury (NSSI), Self-Injury For the current study purposes, self-injury is deliberate and direct self-inflicted physical injury (i.e., alteration or destruction of one‘s body tissue such as cutting or burning) that is not socially sanctioned, is without conscious suicidal intent, and occurs in the absence of pervasive developmental delay (Favazza, 1996; Walsh, 2006). When reviewing literature that used either a broader or more specific definition of self-injury than stated here, an attempt was made to replicate the language used so as to be most precise. Given the lack of a standardized operational definition and conceptualization of self-injury and the plethora of terms in the literature for referring to self-injury and other related and overlapping behaviours, accuracy of terminology was deemed important. For example, European researchers tend to study deliberate self-harm which includes self-injurious behaviour that both is and is not suicidal in intent and frequently do not distinguish between the two; therefore when reviewing this research, the term deliberate self-harm was used. Risk Factor A risk factor is a variable (i.e., experience, circumstance, mechanism or process) that has a negative effect on outcome or is a ‗statistical correlate‘ of negative outcomes or maladjustment (Masten, Best, & Garmezy, 1990, p. 426). Risk factors are known to predict the increased likelihood or probability of a negative outcome or problem behaviour and decrease the likelihood of successful development (Blum et al., 2002; Fergus & Zimmerman, 2005; Masten et al., 1990; Zimmerman & Arunkumar, 1994). Variable across individuals, circumstances, and contexts, risk factors are multidimensional, dynamic, and interactive (Gore & Eckenrode, 1994; Rutter, 1993; Luthar & Zelazo, 2003; Schonert-Reichl, 2000). Further, these factors act cumulatively, having a multiplicative effect; in other words, more risk factors may predict greater probability of negative outcome (Jessor, 1991; Luthar, 1993;  8 Luthar et al., 2000; Masten & Obradovic, 2006; Rutter, 1993). Risk factors operate at the individual (e.g., behavioural, social, cognitive, emotional, physical) or contextual (e.g., experiential/historical, community, family, peer, school) levels (Luthar et al., 2000; Rutter, 1993). Risk factors confer vulnerability and can be moderated by the individual or by the environment (Luthar & Zelazo, 2003). Resilience Interest in resilience stems from the fact that some individuals exhibit successful adaptation despite facing substantial risk (Garmezy, Masten, & Tellegen, 1984; Masten. Best, & Garmezy, 1990; Rutter, 1987). The term resilience ―…refers to the process of, capacity for, or outcome of successful adaptation despite challenging or threatening circumstances‖ (Masten et al., 1990, p. 426). Resilience factors are those that interrupt the trajectory from risk to problem outcome, resulting in adaptive outcomes that are suggestive of ‗beating the odds.‘ As such, resilience is inferred based on the presence of both risk and positive adaptation or competence. Importantly, resilience is not a trait, rather the term refers to a process or phenomenon (Luthar & Zelazo, 2003; Masten & Obradovic, 2006). Resilience is described as dynamic, multidimensional, and context-specific, involving developmental change and examination of trajectories; in other words, an individual might be resilient to specific risk conditions but not to others (Blum, et al., 2002; Luthar et al., 2000; Luthar & Zelazo, 2003; Rutter, 1993). Protective Factor Research on resilience indicates the presence of factors and processes that buffer or mitigate risk, referred to as protective. Protective factors are variables that reduce or prevent the likelihood of dysfunction (negative outcome) and increase the likelihood of positive outcome in the presence of vulnerability (i.e., risk) (Blum et al., 2002; Fergus & Zimmerman, 2005; Luthar, 1993; Luthar & Zelazo, 2003; Rutter, 1987; Zimmerman & Arunkumar, 1994). These factors can be either individual or personal assets or contextual  9 such as family and environmental resources (Fergus & Zimmerman, 2005; Gore & Eckenrode, 1994). Importantly, protective factors interact with risk and although they may have a direct effect, the effect is frequently stronger in the presence of risk (Jessor et al., 1995; Luthar et al., 2000; Rutter, 1987). Protective factors have been found to vary across different contexts and across different risk factors or processes (i.e., a factor may be protective for one outcome but not for another; Blum et al., 2002; Luthar et al., 2000; Rutter, 2000). Like risk, protective factors operate multiplicatively and can potentiate each other, leading to resilience following exposure to a risk factor (Garmezy et al., 1984; Jessor et al., 1995; Luthar & Zelazo, 2003; Masten, et al., 1991; Rutter, 2000; Zimmerman & Arunkumar, 1994). According to Blum and colleagues (2002), in order to be effective, a policy or intervention must consider the links between risk and protective processes and outcomes. For the purposes of this investigation, variables or factors that have a positive effect on outcome (i.e., reduce the likelihood) will be referred to as protective. Overview of the Literature Review This section provided the background to this investigation. The literature review that follows will establish the empirical foundation and theoretical rationale for the study, situating it within the extant body of research. The review will highlight the links between and integrate the most relevant aspects of two disparate literatures, namely street-involvement and NSSI, with a specific focus on the adolescent population. A full appraisal of issues related to streetinvolved youth and/or to self-injury is beyond the scope of the current review. First, I will begin with a view into the condition of youth homelessness and street-involvement, focusing on the Canadian situation, before briefly reviewing the characteristics of street youth. Second, I will introduce NSSI, providing an overview of its conceptualization. Next, I will summarize prevalence and sociodemographic data and describe the nature and function of the behaviour, followed by a review of the most relevant etiological and theoretical perspectives. Subsequently, the links between NSSI and street youth will be explored  10 through a review of the few studies to include an examination of both and through a focused discussion of coping. Next, the importance of using a resilience paradigm will be highlighted. Finally, findings on NSSI in adolescence will be reviewed with attention to both individual (i.e., personal) and contextual (i.e., environmental – family, community) correlates and risk and protective factors, with a focus on how they may relate to adolescents on the street.  11 CHAPTER 2: REVIEW OF THE LITERATURE Homeless and Street-Involved Youth Scope and Description Youth, the single group most at risk for experiencing homelessness, represent the fastest growing segment of the homeless population (Haber & Toro, 2004). Canada is lacking a systematic collection of national statistics on youth homelessness and streetinvolvement; nevertheless, it has been estimated that 150,000 youth are homeless on any given night in Canada and approximately half of those are more chronically homeless (Boivin, Roy, Haley, & Galbaud du Fort, 2005; DeMatteo et al., 1999). An even greater number of adolescents are street-involved, although there is no estimate of the magnitude of this. According to McCormick of the B.C. Institute Against Family Violence (2004), more youth in British Columbia run away from home than any other Canadian province; over 15,000 B.C. adolescents were reported missing in 2002. Efforts to estimate the number of street-involved and homeless youth are confounded by the elusiveness and transience of the population (Berry, 2007; Wright, Allen, & Devine, 1995) and by lack of consensus around how to operationalize this social circumstance. Definitions vary from youth who are living in the absence of a caregiver to youth who live at home but who spend disproportionate time on the streets to youth who are living in temporary housing or who do not have shelter (Haber & Toro, 2004). Many youth remain invisible; they may not access any services or shelter, may live in the back alleys, in abandoned housing or in cars, move from couch to couch (―couch-surfing‖), or in other dangerous situations, and the fear of being forced to return to an untenable living situation combined with the stigma they experience may drive them even further underground (Kidd, 2007). As a result, researchers Haber and Toro (2004) suggested a continuum of street-involvement, with living on the street representing a ―marker of severity‖ (p. 128).  12 Heterogeneous in terms of pathways to homelessness, street experiences, and health profiles and needs, the street youth population is also heterogeneous in its composition. Studies tend to use convenience samples that may or may not be representative, feature a very small number of participants in a qualitative investigation, or report findings that are very specific to the geographic area. As such, sociodemographics emerging from the research may not be generalizable. Nevertheless, it is commonly believed that adolescent boys outnumber girls on the street, with ratios tending to differ across age with more young adolescent girls and more young adult men (DeMatteo et al., 1999; Public Health Agency of Canada-PHAC, 2006; Ringwalt, Greene, Robertson, & McPheeters, 1998). In Canada, youth with Aboriginal identity (i.e., First Nations, Métis, Inuit, and other Native peoples) are overrepresented in the street youth population (relative to their share of the total population), with overall findings ranging from 28% (Murphy & The McCreary Centre Society, 2001) to 36% (PHAC, 2006) to 72% in areas that have a larger Aboriginal population (Murphy & The McCreary Centre Society, 2001; Rachlis et al., 2009). Across studies, a disproportionate number of street youth, 7% to 35%, identify themselves as being gay, lesbian, or bisexual (Cauce, 2000; Cochran et al., 2002; PHAC, 2006; Murphy & The McCreary Centre Society, 2001; Rohde, Noell, Ochs, & Seeley, 2001; Smith et al., 2007; Unger et al., 1997). Street-Involved Youth, Family, and Social Context Antecedent experiences. An unequivocal finding across studies internationally has been that the majority of street youth have experienced trauma starting early in life, often with backgrounds featuring violence and conflict (Haber & Toro, 2004; Janus et al., 1995; Martijn & Sharpe, 2006; Raffaelli, 1997; Tyler, 2006; Whitbeck et al., 1997). Research with street youth highlights family histories of domestic violence, alcohol and substance abuse, criminality, and mental illness (Buckner & Bassuk, 1997; Cauce, 2000; Dadds, Braddock, Cuers, Elliot, & Kelly, 1993; Erdem & Slesnick, 2010; MacLean, Embry, & Cauce, 1999; Ringwalt, Greene, & Robertson, 1998; Ryan et al., 2000; Thrane, Whitbeck, Hoyt, & Yoder,  13 2006; Tyler, 2006; Whitbeck et al., 1997; Whitbeck et al., 1999). Alarmingly high rates of physical, sexual, and emotional abuse, and neglect have been documented, with typically at least 50% of street youth reporting a history of one or more forms of abuse (Haber & Toro, 2004; Janus et al., 1995; Murphy & The McCreary Centre Society, 2001; Molnar et al., 1998; Ringwalt et al., 1998; Ryan et al., 1997; Smith et al., 2007; Stewart et al., 2004; Tyler et al., 2001; Whitbeck et al., 1997). Often this abuse has been repetitive with both extra- and intrafamilial perpetrators, especially in the case of physical abuse (Baron, 2003; Janus et al., 1995; Ryan et al., 2000; Laye et al., 2002). Structural influences such as unemployment and poverty are prevalent in their childhood histories and are contributing factors (Haber & Toro, 2004; Ringwalt et al., 1998). These, along with the chaos, chronic instability, and disorganization that are common themes in street youths‘ families of origin (Martijn & Sharpe, 2006; Thrane et al., 2006; Tyler, 2006; Tyler et al., 2001), help set the stage for early independence. A recent Canadian study of 1,656 urban street youth found that 15% reported experiencing homelessness with their family prior to living on the street themselves (PHAC, 2006). Research has exposed childhoods filled with multiple transitions, including frequent residential moves, schooling disruptions, caregiver change, and interpersonal loss (Buckner & Bassuk, 1997; Cauce, 2000; Whitbeck et al., 1999). Studies typically find that at least 50% of the youth report removal from their homes by child protection authorities and placement in the system during childhood, most often in foster care or group homes; frequently, youth report multiple placements (PHAC, 2006; Haber & Toro, 2004; Murphy & The McCreary Centre Society, 2001; Ryan et al., 2000; Smith et al., 2007; Tyler, 2006). Research on family relationships has found that street youth report lower levels of satisfaction with their mothers and fathers and more feelings of being misunderstood relative to same-age non-street-involved peers; they also report a lack of parental attention, time, and support (MacLean et al., 1999; Murphy & The McCreary Centre Society, 2001; Whitbeck et  14 al., 1997). Despite these family histories, two studies of street-involved adolescents in British Columbia reported the youths‘ belief that their parents care about them ―quite a bit‖ (Murphy & The McCreary Centre Society, 2001; Smith et al., 2007), while another Canadian study has found that many youth maintain at least some contact with parents, most commonly with mothers (PHAC, 2006). In contrast, youth in other studies have relayed the belief that their parents do not care about them (Whitbeck et al., 1997). Limited empirical research has looked beyond victimization and family dysfunction to identify related areas that may be indicators of positive adjustment such as positive connections with family or significant adults or perception of being accepted or nurtured. Street experiences. In addition to negative antecedent experiences, these youths‘ street lives are wrought with danger and violence. Numerous studies have shown that they are prone to victimization on the street (Baron, 2003; Kipke et al., 1997; Hoyt, Ryan, & Cauce, 1999; Laye et al., 2002; Murphy & The McCreary Centre Society, 2001; Smith et al., 2007; Tyler, Hoyt, & Whitbeck, 2000; Tyler et al., 2001; Whitbeck et al., 1997; Whitbeck & Simons, 1990). Further, an extensive body of research on maltreatment suggests that victimization in childhood, an experience shared by many of these youth, significantly increases the chances of revictimization (e.g., Baron, 2003; Ryan et al., 2000; Stewart et al., 2004; Tyler et al., 2000, 2001; Whitbeck et al., 1999), a risk that increases exponentially for youth who live in the rough (i.e., homeless) and for those who are involved in subsistence strategies such as selling drugs and survival sex (i.e., exchanging sex for subsistence needs such as money, goods, drugs, or shelter, a form of sexual exploitation; Greene, Ennett, & Ringwalt, 1999; Kipke et al., 1997; Tyler et al., 2000, 2001; Whitbeck et al., 1999). In a Seattle sample of 374 homeless youth, 83% were victimized sexually and/or physically after they left home for the streets; boys were more likely to report physical assaults and threats whereas girls were more likely to report rape and sexual exploitation (Stewart et al., 2004). In another study, Kipke and colleagues (1997) found that 51% of 432 homeless youth in Los Angeles reported being  15 physically hurt and 15% reported sexual assault since living on the street. Further, these researchers found that 85% of the youth reported having witnessed sexual and/or physical violence. Clearly, victimization is a very real and imminent threat in these adolescents‘ lives, both in their homes and on the street. Street-Involved Youth and Health Extant research is in accord in suggesting that street-involved adolescents are at extreme risk for mental and physical health problems, including death (Boivin et al., 2005; Roy et al, 2004). Recent research attention has focused on substance abuse (Greene et al., 1999; Kerr et al., 2009; Ringwalt et al., 1998), posttraumatic stress disorder ( Stewart et al., 2004; Tyler et al., 2010), depression (Erdem & Slesnick, 2010), blood-borne and sexually transmitted infections (e.g., hepatitis, HIV; DeMatteo, 1999; PHAC, 2006; Roy et al., 2001), and suicidality (Kidd & Kral, 2002; Kidd, 2006; Molnar et al., 1998). Despite increased research interest as the population of disenfranchised youth continues to grow, these areas of inquiry are vast and logistical issues related to accessing and tracking the youth continue to pose barriers. Moreover, many youth do not access healthcare and therefore clinic-based data collection is inherently biased. Given the cross-sectional nature of much of this research, it is difficult to ascertain whether these health problems are caused by negative life events, an underlying mental disorder, the trauma of homelessness (Goodman, Saxe, & Harvey, 1991), serious substance use, or whether a combination of factors is at the root, the latter being most likely. Street youth report poor physical health and a greater incidence of illness and injury (Boivin, et al., 2005; DeMatteo et al., 1999; Ensign, 1998; Murphy & The McCreary Centre Society, 2001). Living without shelter, injection drug use, HIV infection, and engaging in sextrade activities have been found to increase the risk of poor health outcomes (Cheung & Hwang, 2004; Roy et al., 2004). Mortality rates are significantly higher among street-involved  16 youth, with most deaths attributable to suicide or drug overdose (Cheung & Hwang, 2004; Roy et al., 2004). A prospective study of 1,013 Montreal street youth over a five-year followup period revealed mortality rates that were eleven times higher than the rate observed for same-age youth in the general population (Roy et al., 1998; Roy et al., 2004). Given their increased levels of negative life experiences, marginalization, and streetrelated hazards, it is not surprising that homeless youth are generally believed to suffer from high rates of behavioural and emotional problems and have been found to manifest increased levels of psychopathology in comparison to ―normative‖ samples (Cauce, 2000; Dadds et al., 1993; Feitel, Margetson, Chamas, & Lipman, 1992; Hughes et al., 2010; MacLean et al., 1999; Rohde et al., 2001; Unger et al., 1997; Whitbeck et al., 2000; Whitbeck et al., 2004). In contrast with the 20% of school-based youth who experience any mental disorder (Arnett, 1999; Schonert-Reichl & Offer, 1992), prevalence rates among street youth range from 30% (MacCaskill et al., 1998) to 60% (Cauce, 2000) to 90% (Feitel et al., 1992), with a considerable number of youth suffering from comorbidities (i.e., coexisting disorders; Cauce et al., 2000; Unger et al., 1997; Whitbeck et al., 2000; Whitbeck et al., 2004). Research on early trauma and attachment as well as on exposure to violence suggests that homeless youth are an especially vulnerable group for mental health problems. Given the frequent presence of both antecedent and street-related trauma, posttraumatic stress (or PTSD) is particularly salient. In their study, Whitbeck et al. (2004) found that prevalence of PTSD was five times greater for homeless adolescent girls and twelve times greater for homeless adolescent boys than for same-age same-gender housed peers, with rates ranging from 16-24% of youth meeting diagnostic criteria for PTSD (Cauce, 2000; Stewart et al., 2004; Whitbeck et al., 2004). Depression is also common among homeless youth, with 26% to 34% meeting criteria for lifetime major depressive episode (Whitbeck et al., 2004). In their study of 602 runaway and homeless adolescents, Whitbeck, Hoyt, and Bao (2000) found that 23% and 39% of the young men and women, respectively,  17 demonstrated levels above the cut-off on a measure of depression. Similarly, when comparing their sample of homeless youth to a community sample of youth, Rohde and colleagues (2001) found that the odds of major depressive disorder and dysthymic disorder were 5.51 and 13.08 times greater, respectively. Results from a 2010 study revealed that 48% of the homeless youth sample experienced clinically significant levels of both internalizing and externalizing symptoms (Hughes et al., 2010). Taken together, results on the psychological functioning of street youth suggest that they report significantly poorer psychological health compared to housed counterparts. Alarmingly high suicide ideation, attempt and completion rates are commonly reported among street youth (Cochran et al., 2002; Kidd, 2006; Murphy & The McCreary Centre Society, 2001; Molnar et al., 1998; Rohde et al., 2001; Roy et al., 2004; Smith et al., 2007; Unger et al., 1997; Yoder, Hoyt, & Whitbeck, 1998). Research has found rates of suicidal ideation, a powerful risk factor for suicide, to be 50% or higher in street youth samples (Molnar et al., 1998; Yoder et al., 1998). Studies with homeless youth have revealed suicide attempt rates between 26% and 46%, with significantly higher rates for females as compared to males (Feitel et al., 1992; Kidd, 2006; Murphy & The McCreary Centre Society, 2001; Molnar et al., 1998; Smith et al., 2007; Yoder et al., 1998). Of those who attempt suicide, many report making multiple attempts. Research suggests that street-involved adolescents also commonly engage in other high risk behaviours, including sexual risk taking. A disproportionate number of street youth report being sexually active compared to same-age, housed, school-based peers ( 81% vs. 24%; The McCreary Centre Society, 2001). Many youth report having multiple sexual partners (e.g., up to 200 in the past year, Cauce, 2000; DeMatteo et al., 1999; PHAC, 2006), high-risk sexual partners (i.e., injection drug users, sex trade workers, acquisition and/or carrier of sexually transmitted infection/STI), increased rates of unwanted pregnancy, failure to use protection and related STI‘s, substance use prior to sexual activity, and engaging in  18 survival sex (i.e., trading sex for goods, money, or shelter) or sex trade work (Bailey, Camlin, & Ennett, 1998; Cauce, 2000; DeMatteo et al., 1999; Greene et al., 1999; Haber & Toro, 2004; Murphy & The McCreary Centre Society, 2001; Rohde et al., 2001; Smith et al., 2007; Tyler et al., 2004). Age of sexual debut among street youth is considerably lower than that of their non-street-involved counterparts (Smith et al., 2007); however, this finding should be viewed with caution as it may be a function of abuse history (Cauce, 2000; Greene et al., 1999; Tyler et al., 2001). Having experienced abuse prior to street-involvement increases the likelihood of future sexual exploitation (e.g., survival sex, sex trade work) which then increases the likelihood of risky sexual behaviour and further victimization (Baron, 2003; Kipke et al., 1997; Tyler et al., 2001, 2004). Overall, research suggests that approximately one-third of street youth, regardless of gender, have traded sex (PHAC, 2006; Greene et al., 1999; Murphy & The McCreary Centre Society, 2001; Smith et al., 2007). Widespread research documents the extensive substance use in this population. Compared to same-age peers in the general population, rates of substance use are substantially higher among street youth (Boivin et al., 2005; Cauce, 2000; Greene et al., 1999; Kipke et al., 1997; MacLean et al., 1999; Murphy & The McCreary Centre Society, 2001, 2002; Ringwalt et al, 1998; Smith et al., 2007). For example, a study of homeless youth in California (Unger et al., 1997) found a prevalence of 56% and 60% for alcohol and drug use disorders, respectively. Two recent studies of street-involved youth in British Columbia revealed heavy marijuana and alcohol use, as well as the finding that 45% and 28% of street youth across BC reported having tried cocaine and heroin, respectively, with 24% and 8% reporting past month use (Murphy & The McCreary Centre Society, 2001; Smith et al., 2007). More than one in five reported ever injecting illegal drugs (Murphy & The McCreary Centre Society, 2001). Other Canadian studies found that about 1 in 4 street youth who injected drugs shared syringes (De Matteo et al., 1999; Lloyd-Smith et al., 2008). However, despite high prevalence of substance use, some youth report never or rarely using  19 alcohol or drugs (Cauce, 2000; Murphy & The McCreary Centre Society, 2001; Smith et al., 2007). Summary Adolescents who are street-involved are faced with a disproportionate number of challenges at a critical developmental juncture. Sadly, current statistics reveal that too many youth are finding their way to the streets (DeMatteo et al., 2002; Haber & Toro, 2004). A significant majority bear the emotional scars of childhood environments that impinged on their safety and well-being, a dynamic that too often continues to play out in the context of their street involvement. With traumatic backgrounds and marginalized status for reasons too numerous to mention (e.g., sexual minority, ethnic minority, victim of abuse, homeless, disability, and mental illness are all well-represented categories), these youth are vulnerable to a host of physical, mental health, and safety problems that have significant implications for their developmental trajectories. Previous research has extensively documented salient risk and risk factors in the lives of street-involved adolescents, but limited research exists to balance this portrait with information on strengths or protective factors. Despite the fact that nonsuicidal self-injury has been identified as a significant public health issue, little is known about how this behaviour is related to or interacts with other risks and protective factors within the street youth population. The next section of this paper focuses on nonsuicidal selfinjurious behaviour among adolescents. Nonsuicidal Self-Injury (NSSI) Self-injury refers to deliberate and direct self-inflicted physical injury that is without suicidal intent and occurs in the absence of pervasive developmental disorder (Favazza, 1996; Walsh, 2006). Further, NSSI describes harm to the body that is not socially recognized or sanctioned, as opposed to practices such as piercing and tattooing. Among adolescents, the most common self-injurious behaviours include cutting, burning, carving, scratching, and  20 hitting/punching or biting oneself (Alfonso & Dedrick, 2010; Jacobson & Gould, 2007; LayeGindhu & Schonert-Reichl, 2005; Lloyd-Richardson et al., 2007; Nock, 2010; Nixon, Cloutier, & Jansson, 2008; Ross & Heath, 2002). Recent evidence has suggested that the prevalence of NSSI among adolescents and young adults is increasing (Gratz, 2003; Jacobson & Gould, 2007; Lloyd-Richardson, Perrine, Dierker, & Kelley, 2007; Klonsky, Oltmanns, & Turkheimer, 2003; Whitlock, Eckenrode, & Silverman et al., 2006; Yates, 2004; Yates, Tracy, & Luthar, 2008), with some researchers, clinicians, and educators referring to it as an epidemic or the ―new anorexia‖ (Brumberg, 2006; Miller & Smith, 2008; Selekman, 2006). Conceptualizing Self-Injury Lack of consensus on how to define, operationalize, and classify self-injury abounds in the literature, limiting generalizability across studies and impeding the advancement of a unified research base. Extensive review of the literature reveals a myriad of idiosyncratic terms and definitions, introducing a great deal of confusion and requiring scrutiny of each study to determine exactly what is being studied. The behaviours examined fall across a broad continuum of self-harm, ranging from intentionally suicidal behaviour (i.e., attempt with intent to die) to deliberate and direct nonsuicidal self-injury (e.g., cutting, burning) to more indirect self-harming behaviour (e.g., substance abuse, disordered eating) to self-destructive behaviour and recklessness (e.g., physical risk-taking, risky sexual behaviour). In her attempt to define what she refers to as ‗self-harm‘, Skegg (2005) concludes, ―The difficulty is where to draw the line between self-harm and other potentially harmful behaviours (p. 1471).‖ Researchers have examined specific behaviours or behavioural subgroups using more than thirty terms, including self-mutilation (e.g., Ross & Heath, 2002; Tyler et al., 2002), deliberate self-harm (e.g., Gratz, 2003; Klonsky et al., 2003), parasuicide (e.g., Linehan, 1993), self-harm (e.g., Croyle & Waltz, 2007; Laye-Gindhu & Schonert-Reichl, 2005; Skegg, 2005), (delicate-) self-cutting (e.g., Rodham, Hawton, & Evans, 2004), nonsuicidal self-  21 damaging behaviour (e.g., Garrison et al., 1993), nonsuicidal self-injury (e.g., LloydRichardson, Perrine, Dierker, & Kelley, 2007), and self-injury (e.g., Osuch, Noll, & Putnam, 1999; Whitlock et al., 2006). Studies tend to be either over- or under-inclusive depending on the definition utilized. Frequently, differentiation is lacking and disparate behaviours are analyzed together (e.g., suicidal and nonsuicidal), obscuring important learning. In the UK and in Europe, deliberate self-harm is used to refer to both suicidal and nonsuicidal behaviour that is undifferentiated in terms of intent, whereas in North America there has been recent progress toward unified use of the term nonsuicidal self-injurious (NSSI) behaviour that, by definition, excludes intentionally suicidal behaviour and ingestion regardless of intent (Nock et al., 2006; Nock, 2010). Recent calls for clarity and uniformity in the field are generating increased interest and productivity. Within the last five years, researchers have started to come together to recognize and address the conceptual and methodological limitations inherent in the body of research and to coordinate research efforts under the rubric of NSSI. Self-Injury versus Suicide Although the idea that self-injury should be recognized as separate from suicide has found support in the last decade (Brausch & Gutierrez, 2010; Muehlenkamp & Gutierrez, 2004, 2007; Stanley, Gameroff, Michalsen, & Mann, 2001; Whitlock & Knox, 2007), both the literature and clinical accounts document a long history of treating self-injury as though it were a suicide attempt or gesture, leading to much confusion, misunderstanding, and invalidation (Nock & Kessler, 2006; O‘Carroll et al., 1996). Individuals who self-injure are themselves very clear that their behaviour is not suicidal (Lloyd-Richardson et al., 2007; Patton et al., 1997). However, evidence supports that a significant subgroup of individuals do also attempt and/or complete suicide (Brausch & Gutierrez, 2010; Brunner, Parzer, Haffner, Steen, Roos, et al., 2007; Guertin et al., 2001; Muehlenkamp & Gutierrez, 2004, 2007; Stanley et al., 2001) and therefore suicide risk is a major concern with this population. There has been speculation that self-injury may serve as ―practice‖ for later suicide, in particular through  22 habituation to the fear and/or physical pain involved as well as to the reinforcing effects of the behaviour (Jacobson & Gould, 2007; Joiner, 2005; Walsh, 2006; Whitlock & Knox, 2007). Poignant words from a participant in recent research on self-harm among criminalized women articulated the link with a reference to her behaviour as an ―installment plan for suicide‖ (C. A. Dell, personal communication, May 4, 2006). Importantly, research suggests that individuals who engage in NSSI who are also actively suicidal tend to use a different method for suicide attempt than for self-injuring (Stanley et al., 2001). Despite overlap between NSSI and suicide, findings have demonstrated that there are important phenomenological distinctions (Muehlenkamp & Gutierrez, 2004, 2007; Nock et al., 2006; Nock & Kessler, 2006). Suicidal and nonsuicidal self-injurious behaviour differ in several important ways. First, self-injury is much more common than suicidal behaviour (Cloutier et al., 2010; Laye-Gindhu & Schonert-Reichl, 2005; Muehlenkamp & Gutierrez, 2004; Patton et al., 1997). Second, a suicide attempt is considered an act of high lethality whereas self-injury typically is not (Rodham et al., 2004; Stanley et al., 2001). Third, research indicates that although individuals who self-injure may report suicidal ideation, they do not indicate a relation between their self-injurious behaviour and a wish or intent to die (Brausch & Gutierrez, 2010; Garrison et al., 1993; Muehlenkamp & Gutierrez, 2004, 2007; Patton et al., 1997). Fourth, individuals who engage in self-injurious behaviour often do so repeatedly and this chronicity is not typically present in suicide attempt (Favazza, 1996; Muehlenkamp & Gutierrez, 2004). Fifth, studies have demonstrated that a majority of individuals who selfinjure employ more than one method of injuring (Lloyd-Richardson et al., 1997; Nixon, Cloutier, & Jansson, 2008; Walsh, 2006). Finally, the cognitive orientation differs; by explanation, Favazza (1996) offers, ―Suicide is an exit into death, but self-[injury] is a reentrance into a state of normality. Suicide is an act of escape, but self-[injury] is a morbid act of regeneration. A person who attempts suicide seeks to end all feelings but a person who  23 self-[injures] seeks to feel better‖ (p. 271). Thus, self-injury is an attempt to cope by altering consciousness, whereas the suicidal individual is oriented toward terminating consciousness. Diagnostic Issues Discussion about whether self-injury is a symptom, syndrome, or a disorder in its own right has been inconclusive, although several separate diagnoses have been proposed (see Favazza, 1996; Pattison & Kahan, 1983, Muehlenkamp, 2005). Currently, in the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR, 2000), self-injurious behaviour is a key symptom of Borderline Personality Disorder and a proposal for inclusion in the next version, DSM-V, as either a Mood Disorder or as an Impulse Control Disorder is under review (Shaffer & Jacobson, 2010). Beyond that, NSSI is `diagnostically heterogeneous` (Nock et al., 2006, p. 65), with associations to a variety of disorders including dissociative disorders, antisocial personality disorder, post-traumatic stress disorder and other anxiety disorders, eating disorders, substance use disorders, depressive disorders, obsessive compulsive disorder, impulse control disorders, and adjustment disorders (Favazza et al., 1989; Jacobson & Gould, 2007; Klonsky et al., 2003; Muehlenkamp, 2005; Nock et al., 2006; Ross & Heath, 2002; Walsh, 2006; Walsh & Rosen, 1988). Recently, scholars and clinicians have noted that self-injurious behaviour is not limited to individuals with diagnosable disorders, and seems to be most rapidly increasing among youth who may not meet official criteria for a psychiatric diagnosis but who nonetheless are experiencing substantial stress and suffering (Deliberto & Nock, 2008; Walsh, 2006). However, much of the research that includes diagnostic data has been conducted with referred or inpatient samples and is therefore inherently biased; information on the diagnostic correlates of self-injury among community adolescents is notably lacking.  24 Prevalence and Sociodemographics of Self-injury The risk for engaging in NSSI is particularly high during adolescence, with prevalence estimates from surveys of school-based adolescents and undergraduate students revealing lifetime prevalence rates ranging from 13% to 56% (Alfonso & Moyer, 2010; Brunner et al., 2007; Gratz et al., 2002; Hilt, Cha, & Nolen-Hoeksema, 2008; Laye-Gindhu & SchonertReichl, 2005; Lloyd-Richardson et al., 2007; Muehlenkamp & Gutierrez, 2004, 2007; Polk & Liss, 2007; Ross & Heath, 2002; Whitlock, Eckenrode, & Silverman, 2006; Zoroglu et al., 2003). Two recent Canadian studies using samples of over 400 high-school based youth reported prevalence rates of 14-15% (Laye-Gindhu & Schonert-Reichl, 2005; Ross & Heath, 2002), and an internet-based 2006 study of over 8,000 university students in the U.S. reported a lifetime prevalence rate of 17% (Whitlock et al., 2006). In two U.S. based studies that included early adolescent participants (age 10-14), lifetime prevalence rates of 56% in a sample of girls (N = 94) and 28% in a large mixed gender sample (N = 1748) were reported (Alfonso & Dedrick, 2010; Hilt, Nock, et al., 2008). Estimates within the clinical population of youth (i.e., those receiving psychiatric services) have been significantly higher, with 40-80% reporting a history of self-injurious behaviour (Darche, 1990; DiClemente et al., 1991; Nixon, Cloutier, & Aggarwal, 2002; Nock et al., 2006; Nock & Prinstein, 2004; Schwartz, Cohen, Hoffman, & Meeks, 1989). Incarcerated youth have also been found to report elevated levels of self-injury, with 10-80% of youth reporting a history of this behaviour (Chowanec, Josephson, Coleman, & Davis, 1991; Murphy, Chittenden, & The McCreary Centre Society, 2005). Although it has been commonly suggested that NSSI is a predominately female phenomenon, this may be the artifact of a research base that has historically focused almost exclusively on females. Results on prevalence from studies of community adolescents and young adults that include both males and females have been mixed; whereas some studies have shown no difference in self-injury rates between males and females (Gratz et al., 2002;  25 Heath et al., 2008; Lloyd-Richardson et al., 2007; Muehlenkamp & Gutierrez, 2004; Zoroglu et al., 2003; Whitlock et al., 2006), others have found females significantly more likely to engage in the behaviour (Laye-Gindhu & Schonert-Reichl, 2005; Muehlenkamp & Gutierrez, 2007; Patton et al., 1997; Ross & Heath, 2002). A paucity of research has investigated other sociodemographic variables such as ethnicity, socioeconomic status, or sexual orientation. Limited research suggests that NSSI appears to be an international phenomenon with studies conducted beyond North America and English-speaking Europe in diverse places including Asia (e.g., Hong Kong, see Wong, Stewart, Ho, & Lam, 2007), the Middle East (e.g., Turkey, see Zoroglu, Tuzun, Sar, Tutkun, Savacs et al., 2003), and New Zealand (includes Maori youth, see Nada-Raja, Skegg, Langley, Morrison, & Sowerby, 2004). Studies conducted in the U.S. that report on ethnicity have shown variable results, with some findings reflecting higher rates of self-injury among Caucasian high-school students (Lloyd-Richardson et al., 2007; Muehlenkamp & Gutierrez, 2004, 2007) when compared with non-Caucasians and African-Americans, respectively. In contrast, in their sample of over 1000 affluent west-coast adolescents, Yates et al. (2007) found that students identifying as ‗Black‘ or ‗Other‘ (primarily Native American) were more likely to engage in a range of NSSI behaviours when compared to other groups. A Canadian study of 424 high-school students reported no differences in self-injury rates across ethnic groups (Laye-Gindhu & Schonert-Reichl, 2005). Some research suggests higher prevalence of self-harm more generally among indigenous people in countries with a history of colonization (see Nada-Raja et al., 2004) and among South Asian young women in England (see Marshall & Yazdani, 1999), but this has not been empirically investigated in Canada. Information on prevalence and phenomenology of self-injury among sexual minority adolescents is lacking, although a few recent studies with young adult samples in the U.S. have reported on sexual orientation or on attraction (Serras, Saules, Cranford, & Eisenberg, 2010; Tyler et al., 2010; Whitlock et al., 2006). In their studies of university students, both  26 Whitlock and colleagues (2006) and Serras et al. (2010) found that sexual minority status and, in particular, bisexuality was significantly related to NSSI. Similarly, in an internet survey of 128 young people, Murray, Warm, and Fox (2005) found that 23% and 4% reported bisexual and homosexual orientations, respectively. Findings from a study investigating deliberate self-harm (including both nonsuicidal and suicidal behaviour) with a young adult sample in New Zealand revealed that same-sex attraction was a significant predictor of NSSI, particularly for males (Skegg, Nada-Raja, Dickson, Paul, & Williams, 2003). Nature of Self-Injury Studies have consistently found mean age of onset for self-injury to be in early adolescence (Alfonso & Dedrick, 2010; Hilt, Nock, et al., 2008; Muehlenkamp & Gutierrez, 2004, 2007; Nixon et al., 2002; Nock & Prinstein, 2004; Ross & Heath, 2002; Walsh, 2006). However, the retrospective nature of the extant research represents a significant limitation. Very little is known about why some individuals engage in self-injurious behaviour only once or even twice, whereas others engage in the behaviour repetitively, with participants in some studies reporting in excess of 100 incidents (Brunner et al., 2007; Laye-Gindhu & SchonertReichl, 2005; Lloyd-Richardson et al., 2007; Nixon et al., 2002; Whitlock et al., 2006). Research on frequency is unreliable, however, given the retrospective and chronic nature of self-injury it is expected that most youth do not maintain an accurate count of their behaviours. Moreover, many studies ask youth to report on lifetime frequency, further complicating the results. Nevertheless, it is clear that for some youth the behaviour becomes entrenched and addiction-like (Brunner et al., 2007; Nixon et al., 2002). The most common methods of self-injury are cutting and related behaviours (cutting, scratching, carving words or numbers), burning, self-hitting or punching or hitting oneself with an object, self-biting, poking objects into the skin and drawing blood, and breaking bones on purpose (Laye-Gindhu & Schonert-Reichl, 2005; Lloyd-Richardson et al., 2007; Nock et al., 2006; Ross & Heath, 2002; Whitlock et al., 2006). Although there is consensus that self-  27 harmful behaviours fall along a continuum, there is ongoing debate in the literature as to which behaviours to include under the NSSI rubric. The relatively high base rates for some of the behaviours included in self-injury inventories, most notably picking at a wound, may reflect normative human behaviour that may be even more common in the adolescent age group. As a result, some researchers have included some or all of these behaviours, whereas others have excluded them, and still others have further classified behaviours as mildly or moderately/ severely injurious or as subclinical and clinical (e.g., Brunner et al., 2007; Croyle & Waltz, 2007; Lloyd-Richardson et al., 2007). Motivational Aspects and Function of Self-Injury Explanatory models of self-injury provide critical links to etiology and highlight the underlying motivations and functions that serve to initiate and maintain the behaviour. Although early research featured a wide variety of theoretical models (e.g., Simpson & Porter, 1981; Suyemoto, 1998) derived primarily from case studies among inpatient or incarcerated females, few models were supported empirically. Recent efforts focused on linking theory and empirical study have made significant strides in disentangling, integrating, and evaluating the various conceptualizations (e.g., Buckholdt, Parra, & Jobe-Shields, 2009; Hilt, Cha, & Nolen-Hoeksema, 2008; Klonsky, 2007; Klonsky, 2009; Klonsky & Glenn, 2009; Nock & Prinstein, 2004, 2005). This step is necessary and provides evidence to advance Suyemoto‘s (1998) assertion that the behaviour likely serves multiple functions simultaneously. More studies with representative community samples of adolescents are needed to empirically and systematically assess the contribution and generalizability of these models and to continue developing a cohesive knowledge base that can be applied to prevention and treatment. Whereas previous reports emphasized the attention-seeking aspects of NSSI despite limited empirical evidence, recent research has broadened to focus on intrapersonal  28 motivations and functions, with accumulating evidence for the affect regulation model (Klonsky, 2007, 2009; Nock, 2010; Rodham et al., 2004). This functional model suggests that the behaviour serves to provide quick and effective relief from aversive, distressing, and overwhelming emotional states (e.g., anxiety, depression, loneliness, dissociation, anger, fear). Evidence from studies of the psychophysiology of self-injury with adult samples has provided support for the model, demonstrating the increased arousal that accompanies NSSI and the negative reinforcement that maintains it and establishes it as a behavioural coping response (Brain, Haines, & Williams, 2002). Similarly, a recent study with a sample of selfinjuring adolescents revealed increased physiological reactivity (assessed via skin conductance) during a distressing task when compared to non-injuring peers (Nock & Medes, 2008). Most commonly endorsed by individuals who self-injure, intrapersonal functions include motivations such as feeling generation (e.g., relief from dissociative states) and emotional arousal reduction and release (i.e., relief from negative emotionality; Klonsky & Glenn, 2009; Laye, 2002; Nixon et al., 2002; Nock et al., 2006; Ross & Heath, 2003). In their psychometric evaluation of functions of NSSI in a college sample of young adults, Klonsky and Glenn (2009) found support for affect regulation and anti-dissociation functions and add other functions such as anti-suicide, marking distress, and self-derogation (i.e., to punish self), with the latter being the second most common intrapersonal motivation in that study, as in others (Glassman et al., 2007; Klonsky , 2007; Laye, 2002; Lloyd-Richardson et al., 2007; Nixon et al., 2002; Nock & Prinstein, 2004). Recent efforts to integrate the literature on functional models of NSSI, has culminated in a theoretically-based four-factor functional model that incorporates both self-focused and social factors, with growing evidence of its validity and reliability (see Nock & Prinstein, 2004, 2005, reviews by Klonsky, 2007; Nock, 2010). Importantly, previous findings relating to functions of self-injury also neatly map onto this model (e.g., Klonsky & Glenn, 2009). The model posits that behaviours are maintained through automatic (i.e., intrapersonal) and social  29 (i.e., interpersonal) contingencies as well as through reinforcement that is positive (i.e., followed by presentation of favourable stimulus) or negative (i.e., followed by removal of an aversive stimulus) categories that are not mutually exclusive. The most common function reported by adolescents for engaging in NSSI, automatic negative reinforcement (i.e., regulatory strategy to stop negative or bad feelings; Klonsky, 2007; Klonsky, 2009; Laye-Gindhu & Schonert-Reichl, 2005; Lloyd-Richardson et al., 2006; Nixon et al., 2002; Nock & Prinstein, 2004, 2005; Ross & Heath, 2002), is associated with history of suicidality, depression, and hopelessness (Klonsky & Glenn, 2009; Nock & Prinstein, 2005). Youth, especially those who are suffering from posttraumatic stress disorder or who have trauma histories, also report engaging in self-injury for automatic positive reinforcement (i.e., regulatory strategy to prompt feeling, e.g., dissociation, numbness; Hilt, Nock, et al., 2008; Jacobson & Gould, 2007; Lloyd-Richardson et al., 2007; Nock & Prinstein, 2004, 2005; Polk & Liss, 2007; Zoroglu et al., 2003). Along the social dimension, social negative reinforcement (i.e., to remove or avoid social responsibility) and social positive reinforcement functions (i.e., to get support, to bond with others, and/or to communicate or influence) are less frequently reported (Hilt, Nock, et al., 2008; Laye-Gindhu & SchonertReichl, 2005; Nixon et al., 2002; Nock & Prinstein, 2004, 2005). Although social factors would seem to play a smaller role relative to intrapersonal factors, extant research has pointed to the role of ‗priming‘ whereby NSSI is more likely to be initiated by an adolescent whose peer network already engages in the behaviour (Heath et al., 2009; Laye, 2002; Nock & Prinstein, 2005). Given the role of peers and interpersonal stressors in adolescence, social reinforcement factors may be especially salient during this developmental period. Together, these findings contradict the traditional view and long-maintained myth that self-injurious behaviour is primarily attention-seeking behaviour. Interestingly, evidence has suggested that there may be gender differences in function, with more males reporting interpersonal reasons (e.g., to make others angry, to demonstrate ‗toughness‘) and more girls  30 reporting intrapersonal motivations (e.g., to hurt or to punish oneself) (Andover, Primack, Gibb, & Pepper, 2010; Klonsky & Glenn, 2009; Laye-Gindhu & Schonert-Reichl, 2005; LloydRichardson et al., 2007; Sim, Adrian, Zeman, Cassano, & Friedrich, 2009). Given the dearth of studies that include males or systematically investigate gender differences, much remains unknown, including whether and how much our current understanding generalizes to males. Future research testing the four-factor model and other competing models in community samples and across genders will provide significant insight into this perplexing behaviour and inform clinical efforts. Etiological and Theoretical Perspectives on Self-Injury Nonsuicidal self-injury is a multi-determined behaviour, with theoretical and empirical links to individual (i.e., biological, developmental, cognitive, affective, behavioural) and contextual (i.e., social-environmental) dimensions. Accordingly, current evidence-based treatment for self-injury is conceptualized in terms of a biopsychosocial model (Crowell, Beauchaine, & Linehan, 2009; Linehan, 1993; Nock, 2010; Walsh, 2006), with interrelated dimensions and different combinations operating to initiate and maintain the behaviour in a way that is unique to the individual. A comprehensive description of etiology and theoretical explanations and perspectives for NSSI is beyond the scope of this review. However, a brief overview of those relevant to the current study will follow. Notably, the biological perspective, while offering significant contributions to the understanding of self-injurious behaviour, particularly in explaining physiological and neurochemical responses to trauma (e.g., serotonergic, opioid, noradrenergic, dopaminergic systems), will not be discussed here. Although self-injury has been viewed through the lens of differing theoretical perspectives, the notion of subjective trauma is a critical common element in both the early and contemporary literature. Early interest in self-injury took a psychoanalytic perspective, emphasizing intrapsychic drives and conflicts and defensive processes. This perspective proposed a traumagenic hypothesis whereby a traumatic event initiates the experience of  31 helplessness which then mobilizes the ego to avoid retraumatization and creates a condition of hypersensitivity acted out as aggression turned inward. Following that, neo-analytic adaptations such as object relations theory and attachment theory and the field of developmental psychopathology emphasized the role of early relational experience by way of the caregiver as a regulator for emotions and behaviours (see Eisenberg, Spinrad, & Eggum, 2010; Zeman, Cassano, Perry-Parrish, & Stegall, 2006). According to these theories, internalized early relationships with caregivers are responsible for the development of a cohesive sense of self and self-regulatory competencies. Self-injury represents the failure to develop a coherent and genuine sense of self and which is ‗played out‘ in a dysfunctional selfcare system whereby harming the body becomes a form of self-soothing (van der Kolk et al., 1991; Yates, 2004). Attachment theory posits that disorganized, disrupted, or insecure attachments (viewed as a form of trauma) in the relationship with the caregiver render a child more vulnerable to negative trajectories/outcomes, including subsequent engagement in selfinjury (Gratz et al., 2002; Yates, 2004). According to these theories and the evidence supporting them, the role of early childhood experiences is central in the development of selfinjury. Evidence has shown that a significant proportion of individuals who self-injure emerge from invalidating environments in which they did not learn how to regulate emotions, they have high negative affect and/or are emotionally reactive, and yet lack the tools to verbally express their emotions (Adrian, Zeman, Erdley, Lisa, & Sim, 2011; Chapman, Gratz, & Brown, 2006; Crowell et al., 2008; Crowell et al., 2009; Klonsky, 2009). Thus, maladaptive coping strategies that offer a ‗quick fix‘ in the face of unruly affect (or lack of affect) and selfdisturbance are employed, such as self-injury, disordered eating, and substance use. Behavioural perspectives of self-injury have predominated in the recent literature, with a growing evidence base. Social learning theory is hypothesized to play an important role in initiating the behaviour (i.e., through modeling, imitation, vicarious learning) and reinforcement contingencies in maintaining the behaviour (Lloyd-Richardson et al., 2007;  32 Nock, 2010; Nock & Prinstein, 2004, 2005; Suyemoto, 1998). Nock and Prinstein‘s fourfactor functional model of NSSI reviewed in the proceeding section is an example of this approach (2004, 2005). Another recent and related conceptualization of self-injury in the adult literature that is finding support is the experiential avoidance model (EAM; Chapman et al., 2006), a specific variant of the affect-regulation model. This model posits that self-injury is maintained by automatic negative reinforcement through escape from, or avoidance of, unwanted emotional experience or internal conditions (i.e., aversive affect/thoughts/ feelings/sensations, this could be intense emotion or absence of emotion). The EAM refers to a functional class of behaviours that features avoidant coping including thought suppressing, self-medication through substance use, and avoidance of feared objects, places, or situations (Chapman et al., 2006; Najmi, Wegner, & Nock, 2006). Research involving these two related behavioural models is in the early stages, but appears to manifest promise in expanding understanding of the determinants of self-injury. Applying these models to research involving community samples of mixed gender adolescents will be an important next step. Similar to aspects of the psychoanalytic and neo-analytic perspective, the behavioural perspective explains self-injury as a method of affect regulation and/or as a method of social influence or communication. The effectiveness of self-injurious behaviour in quickly altering affect is one important key to its repetition, with Nixon and colleagues (2002) finding some support for an addiction model with their sample of psychiatrically hospitalized youth. These researchers suggest that self-injury may play a self-medication role in which, over time, increasingly severe or frequent injury is required to achieve the affect-modulating effects. The aforementioned EAM model also highlights the power of escape conditioning whereby self-injury becomes an automatic escape response; however, these researchers also refer to the paradoxical effects of experiential avoidance in that although self-injury is an effective strategy in the short-term, in  33 the long term it leads to ―a rebound effect‖ of increased distress/negative affect and further likelihood of self-injury (Chapman et al., 2006, p. 385; Najmi et al., 2006). Research involving social factors has been less conclusive. Limited systematic empirical research has investigated either social learning as an instigator/maintainer of selfinjury or social communication as a function of self-harm. Interpersonal explanations for selfinjury refer to the behaviour as an outward expression of suffering meant to have an effect on others (Nock, 2010; Solomon & Farrand, 1996; Suyemoto, 1998; Walsh & Rosen, 1988). Similarly, Nock (2010) proposed a social signaling hypothesis whereby individuals use NSSI to effectively communicate either to signal distress when other weaker signals have failed (social positive reinforcement) or to signal strength or resilience (social negative reinforcement). Studies have found that adolescents who self-injure are more likely to have friends who self-injure or family histories of self-injury, providing some support for the social learning theory. However, these studies do not assess temporality, and some do not distinguish between suicidal and nonsuicidal behaviour (De Leo & Heller, 2004; Hawton et al., 2002; Heath et al., 2009; Laye, 2002; Nock & Prinstein, 2005; Whitlock et al., 2006). Given the increasing concern and discussion in the literature and in the public sphere regarding the social aspects of this behaviour among adolescents and young adults (Chowanec et al., 1991; Deliberto & Nock, 2008; Heath et al., 2009; Hilt, Nock, et al., 20008; Walsh & Rosen, 1988; Walsh, 2006), rigorous tests of these theoretical explanations with different samples of young people are needed to improve the current understanding. Summary Nonsuicidal self-injury is a perplexing behavioural phenomenon that contradicts all westernized notions of the sanctity and beauty of the body. Nevertheless, evidence suggests that its prevalence is increasing among young people and recent research documents its pervasiveness in both community, clinical, and forensic youth populations (Alfonso & Dedrick, 2010; Brunner et al., 2007; Cloutier et al., 2010; Hilt, Nock, et al., 2008; Nock et al., 2006;  34 Laye-Gindhu & Schonert-Reichl, 2005; Lloyd-Richardson et al., 2007; Nixon et al., 2009; Ross & Heath, 2002; Yates, Tracey, & Luthar et al., 2008). Until recently, NSSI was mostly overlooked in the empirical literature, with terminological and conceptual confusion contributing to it being obscured within the literature on suicide, despite clear indication that self-injury and suicide are distinct (Brausch & Gutierrez, 2010; Muehlenkamp & Gutierrez, 2004, 2007; Nock & Kessler, 2006; Stanley et al., 2001; Whitlock & Knox, 2007). Early conceptualization as a ‗female problem‘ has meant that we know little about the behaviour among males, although some recent findings suggest similar prevalence rates (Alfonso & Dedrick, 2010; Andover et al., 2010; Lloyd-Richardson et al., 2007; Muehlenkamp & Gutierrez, 2004; Zoroglu et al., 2003). Similarly, the association between other demographic characteristics and NSSI has not been well-established. Systematic research on the motivational and functional aspects of self-injury has provided a growing evidence base that highlights its role in effectively regulating affect (Adrian et al., 2011; Buckholdt et al., 2009; Chapman et al., 2006; Klonsky, 2007; Klonsky & Glenn, 2009; Laye, 2002; Lloyd-Richardson et al., 2007; Nock & Prinstein, 2004, 2005). As a complex and multidetermined behaviour associated with significant morbidity and mortality, it is imperative that more systematic empirical research be conducted on NSSI to address the gaps in the literature. Street-Involved Youth and Nonsuicidal Self-Injury With its long history of being understudied and misunderstood, NSSI has been almost entirely overlooked in the extant literature on street youth. To date, little is known about NSSI in this population, although there is evidence suggesting an increasing prevalence of the behaviour among community school-based youth (Alfonso & Dedrick, 2010; Hilt, Nock, et al., 2008; Lloyd-Richardson et al., 2007; Muehlenkamp & Gutierrez, 2007) and a significantly greater prevalence among clinical and incarcerated populations (e.g., Chowanec et al., 1991; DiClemente et al., 1991; Murphy et al., 2005; Nock et al., 2006; Schwartz et al., 1989).  35 Despite a focus in the literature on a number of risky and self-destructive behaviours engaged in by street-involved youth (e.g., substance abuse, sexual risk-taking, suicide), NSSI has been largely neglected. Street-involved youth represent an ideal population in which to examine NSSI for a variety of reasons: first, given that self-harming behaviours are conceptualized along a continuum and typically co-occur, one would expect prevalence rates of self-injury to be elevated in this high-risk group given higher rates of other risk behaviour; second, the preponderance of young men and the heterogeneity of the population on the street means that gender, ethnicity, and sexual orientation can be explored for their relation to NSSI; and third, the vulnerability imposed by these youths‘ social backgrounds and street-life exposure means that variables found to be associated with NSSI in previous research such as abuse, violence exposure, interpersonal loss/separation, and mental health and behavioural problems are prevalent (Lloyd-Richardson et al., 2007; Muehlenkamp & Gutierrez, 2004; Nock et al., 2006; Patton et al., 1997; Rodham et al., 2004, Ross & Heath, 2002; Wiederman et al., 1999; Zoroglu et al., 2003). Research on Self-Injury Among Street-Involved Youth Aside from the McCreary Centre data on which the present study is based and the corresponding report (Smith et al., 2007), extensive searching located only four studies that included any mention of self-injury among street-involved youth (Kidd & Kral, 2002; Tyler et al., 2010; Tyler et al., 2003; Unger et al., 1997). Of the four, only two focus on NSSI specifically, one in an adolescent sample (Tyler et al., 2003) and one in a sample of emerging adults (Tyler et al., 2010). When Tyler and colleagues (2003) asked 428 homeless youth aged 16-19 to complete a checklist indicating whether they had engaged in specific behaviours, 69% of the youth reported at least one occasion of self-injury in their lifetime, primarily cutting or carving (45% total) and burning the skin (29%). These researchers  36 reported that depression, staying on the street, sexual abuse, and external stressors including survival tactics (e.g., survival sex) were associated with an increase in the number of selfinjurious behaviours. Further, age (i.e., older youth) and non-heterosexual orientation were also significantly correlated with greater numbers of methods for self-injuring. Interestingly, Tyler et al. offer no explanation for why they did not examine the presence or absence or frequency of self-injury for their differential relation to correlates, and instead focus on the total number of methods of injuring. Although it is clear that some methods are associated with greater tissue damage, extant research has not established a firm marker of severity and therefore we do not know whether using a greater number of methods or whether more frequent NSSI is a better indicator of severity. The second study by Tyler and colleagues (2010) features an examination of NSSI from a social stress framework in a sample of 172 street youth aged 17-24. Consistent with their 2003 study, 62% of the males and 66% of the females in their sample reported NSSI. These researchers partially confirmed their hypothesis that marginalized social statuses, with their attendant stressors, confer additional risk for self-injury, above and beyond being homeless. Specifically, they found that GLB (gay, lesbian, bisexual) young adults reported more stressors and more NSSI whereas female gender status did not add risk in their sample. Further, these researchers found that correlates they tested from general population samples (e.g., sexual victimization, sexual abuse, neglect, substance abuse) were generalizable to their homeless sample. The third, a study by Unger and colleagues (1997) in Hollywood, CA, found that 49% of the 432 homeless youth aged 12 to 23 reported a history of self-injurious behaviour in response to a single dichotomous question. These researchers focused on a wide range of mental health problems in their sample, but reported that self-injurious behaviour was significantly correlated with younger age, female gender, ‗white‘ ethnicity, more than one-year of homelessness, depressive symptoms, suicidality, substance use disorder, low-self-esteem,  37 and sexual abuse history. The fourth, a qualitative study on suicide and sex trade work with a small sample of 17-24 year-old street youth, revealed that 11% of the 19 males and 30% of the 10 females in the sample reported a history of self-injury, primarily cutting (Kidd & Kral, 2002). No other NSSI findings were reported. These findings provide evidence to suggest that the prevalence of self-injury may be significantly elevated in this high-risk population and more like that reported among clinical samples. Interestingly, these researchers either didn‘t report on gender (Unger et al., 1997) or found no significant difference in prevalence between genders in their sample of homeless youth (Kidd & Kral, 2002; Tyler et al., 2010; Tyler et al., 2003). Coping Among Street-Involved Youth: Links to Self-Injury The extant literature provides abundant evidence of the relations among mental health problems, coping, and risky and/or self-harmful behaviour. For example, youth with mental health problems are more likely than psychologically healthy youth to use emotional or avoidance/escape coping strategies such as substance use and, similarly, substance use may both exacerbate and/or trigger emotional and behavioural problems (MacLean et al., 1999; Unger et al., 1997; Whitbeck et al., 2000). Researchers MacLean and colleagues (1999, 2001) proposed and found significant support for a self-medication and affectregulation (i.e., coping) model of substance use among homeless female adolescents. This model holds that street youth, due to their developmental disadvantage (i.e., maltreatment, trauma), premature independence, social isolation, and lack of adult supervision, ―…may not have had the time, opportunity, or role models to develop effective coping skills‖ (MacLean et al., 1999, 408). Instead, maladaptive and immature coping strategies that externalize, numb, or provide self-soothing against overwhelming and negative or painful affect or arousal frequently develop and provide short-term benefits that reinforce and increase the likelihood  38 that they will be repeated (Nixon et al., 2002; Nock & Prinstein, 2004, 2005; Votta & Manion, 2004). NSSI is another example of a behavioural outcome that fits this model. Scarce research has directly investigated coping models with street youth, however a few studies have found that emotion-focused coping strategies (i.e., efforts to reduce the emotional distress caused by the stressor, e.g., self-injury, substance use) increase risk for suicide (Kidd & Carroll, 2007; Unger et al., 1997). Research has shown that youth with emotion-focused coping style, high stress, and/or low social resources are at high risk for poor physical and mental health (Compas, Orosan, & Grant, 1993; Votta & Manion, 2004; Unger et al., 1997). Improved insight into street youths‘ coping processes and strategies, even if indirectly, is crucial step in developing effective prevention and intervention strategies. Resilience Theory: Links to Street Youth and Self-Injury Contemporary paradigm shifts have highlighted the importance of investigating positive adaptation and resilience and promoting health while simultaneously focusing on deficits or risk and risk reduction (Bearinger et al., 2010; Blum et al., 2002; Borowsky, Ireland, & Resnick, 2001; Fergus & Zimmerman, 2005; Luthar & Zelazo, 2003; Masten et al., 1990; Rutter, 1993). Important findings showing that some individuals demonstrate resilience, or positive development in spite of adversity, have led to increased emphasis on identifying resilience factors (i.e., protective factors that promote healthy development and adaptation and reduce the odds of problem outcome) as well as on better understanding the specificity of risk and protective factors (Luthar, 1993; Luthar et al., 2000; Luthar, Sawyer, & Brown, 2006; Masten et al., 1990; Rutter, 1987; Zimmerman & Arunkumar, 1994). Nowhere is this more important to acknowledge than in this population of extremely vulnerable, highly stigmatized young people (Kidd, 2007), and yet a significant limitation of the extant body of research on adolescent street-involvement is its heavy focus on risk and deficits to almost the exclusion of factors that contribute to health and well-being, adaptation, and reduced likelihood of negative  39 outcome (for notable exceptions regarding resilience, strengths and street youth, see Bender, Thompson, McManus, Lantry, & Flynn, 2007; Erdum & Slesnick, 2010; Kidd & Carroll, 2007; Rew & Horner, 2003). It is incumbent on social and behavioural scientists to attend to not only the risk and vulnerabilities of youth but also the strengths and protective factors that operate in combination with risk in their lives. Researchers have identified different kinds of protective factors that have been found to operate both directly and indirectly, independently and in combination. For the purposes of the present investigation, the work of seminal risk and resilience researchers Garmezy, Luthar, Rutter, and Masten, with reviews by Zimmerman and colleagues will be utilized to provide a heuristic framework for the examination of risk and protective factors related to NSSI among street-involved adolescents. This research focuses on factors at both the individual and contextual levels, the latter related to family, school, and community (Blum et al., 2002; Cicchetti & Lynch, 1993; Luthar et al., 2000). Notably, given the present focus on NSSI, a negative outcome, the current study does not examine resilience per se, but instead broadens traditional focus from risk-only to the inclusion of factors that may prove to be protective against self-injury (i.e., to reduce the likelihood) that are operating in the lives of the adolescents who are street-involved. The term protective is used as an umbrella term to indicate those internal assets or external resources that directly or indirectly decrease the likelihood of a negative outcome (in the current study, NSSI) and increase the likelihood of positive outcomes (no NSSI) (see Blum et al., 2002; Fergus & Zimmerman, 2005; Luthar, 1993; Luthar et al., 2000; Resnick et al., 1997; Sameroff, Guttman, & Peck, 2003; Zimmerman & Arunkumar, 1994). When examined in the context of risk exposure, protective factors buffer or reduce the effects and they may operate in varied nonmutually exclusive ways to influence outcomes. They may also serve to potentiate other protective factors, increasing strength by clustering (Fergus & Zimmerman, 2005; Luthar, 1993; Luthar & Zelazo, 2003).  40 The well-established study of risk emerges from epidemiology, a field concerned with locating agents or conditions associated with the heightened probability of outcomes that compromise physical and emotional health and quality of life. Risk factors are conditions or variables (i.e., experiences, circumstances, mechanisms, or processes) that are associated with increased likelihood of negative outcomes (Jessor, 1991; Luthar, 1993; Masten et al., 1990). Risks are multidimensional, with individual, developmental, and contextual differences or variation (Gore & Eckenrode, 1994; Masten et al., 1990; Rutter, 1993). Further, they tend to co-occur, and are cumulative and multiplicative, with research showing that exposure to multiple risks increases the probability of one or more negative outcomes (Gore & Eckenrode, 1994; Masten et al., 1990; Masten & Obradovic, 2006; Sameroff et al., 2003). In describing and advocating for the simultaneous consideration of multiple risks, Luthar refers to the ―synergistic effect, wherein the effects of co-existing stressors far exceeds the effects of any single factor considered individually (1993, p. 444). Indeed, risk does not occur in a vacuum and youths‘ lives are testament to this, with a clustering of risks and recurring stressors. Researchers, including Garmezy and Rutter, and more recently Luthar, Masten, and Resnick have made significant contributions to the study of resilience and protection. By identifying individuals who have overcome adversity and thrived, they have identified salient protective factors that are common across demographic groups and risk factors (e.g., Garmezy et al., 1984; Luthar, 1993; Masten et al., 1990; Resnick et al., 1997; Rutter, 2000). Protective factors consistently identified by these researchers include individual personal characteristics (e.g., personality, locus of control, self-efficacy, intelligence), family environment characteristics (e.g., parental mental health, parenting style, connectedness and cohesion), and the availability of external community supports (e.g., school connectedness, supportive adult, involvement in community activities, healthy schools). Although protective factors are little examined in the street youth population, Kidd‘s recent research has highlighted interesting results revealing that maintaining hope and belief  41 for an improved future reduces suicide risk, whereas other results have been inconsistent in linking specific coping strategies such as valuing the self, spirituality, and realizing personal strength to suicidality (Kidd, 2003b; Kidd & Carroll, 2007; Kidd & Kral, 2002). Examination of protective factors is notably absent in the NSSI literature. In contrast, a growing research base has examined protective factors related to suicidality (e.g. Borowksy et al., 2001; Gutierrez, 2006; Skegg, 2005); given the overlap between suicide and self-injurious behaviour, this body of research can be used to inform the study of NSSI, but should be interpreted cautiously as there are very clear and significant differences between these behaviours (Brausch & Gutierrez, 2010; Muehlenkamp & Gutierrez, 2004, 2007; Whitlock & Knox, 2007). An improved understanding of factors that generate or increase risk and those that promote successful coping in this population will aid in the identification of intervention points and assist in optimizing development to overcome challenge and adversity. Self-Injury in Adolescence: Associated Factors Whereas many studies examine one risk and one protective factor at a time, the present study examines multiple hypothesized risk and protective factors both for their independent relation to self-injury and their combined relations. Using resilience theory as a framework, factors identified through the empirical and theoretical literature were hypothesized to act as either individual (i.e., personal) or contextual (i.e., environmental) risk or protective factors against NSSI are reviewed in this section as they pertain to streetinvolved youth (see Table 1). Attempts have been made to include protective factors that are amenable to change in order to yield direct links to intervention and clinical practice (Bearinger et al., 2010; Blum et al., 2002; Erickson, 1999; Luthar et al., 2006; Luthar & Zelazo, 2003). Although conceptually distinct from risk, some protective factors function as the polar end of risk (i.e., low family connectedness as a risk factor contrasted with high family connectedness as a protective factor; Luthar et al., 2006; Sameroff et al., 2003). Further research is necessary to clarify whether these factors are specific independent factors  42 or merely characteristics of more general and global problems or psychopathology (Gratz, 2003; Jacobson & Gould, 2007; Nock, 2010). The remainder of this section will be devoted to a discussion of these factors. Contextual Factors Trauma, abuse, and victimization. Theoretical literature on the development of selfinjury suggests that the role of childhood experiences that occur in the context of family are critical (Adrian et al., 2011; Buckholdt et al., 2009; Gratz, 2003; Jacobson & Gould, 2007; van der Kolk & Fisler, 1994; van der Kolk et al., 1991). Researchers have theorized about ―invalidating environments‖ (Linehan, 1993; Chapman et al., 2000; Crowell et al., 2009) in which children were not able to learn how to regulate their emotions, leading to high levels of negative affect in the absence of tools to verbally express and cope with the challenges. Research has been largely consistent with this theory and has highlighted the interpersonal deficits that can develop from these early environments (e.g., Crowell, Beauchaine, McCauley, & Smith, 2008; Hilt, Nock, et al., 2008; Nock & Mendes, 2008; Sim et al., 2009). Studies have provided ample evidence of a connection between both NSSI and trauma and between street-involved youth and trauma, however results on the exact nature of the trauma have not been consistent (e.g., Gratz et al., 2002; Heath et al., 2008). Contextual risk factors for self-injury identified in extant research include childhood physical, sexual, and emotional abuse, neglect, family violence, and parental psychopathology (Deliberto & Nock, 2008; Glassman, Weierich, Hooley, Deliberto, & Nock, 2007; Gratz et al., 2002; Hilt, Nock et al., 2008; Lloyd-Richardson et al., 2007; Tyler et al., 2003; van der Kolk et al., 1991; Wiederman et al., 1999; Yates, 2004; Yates et al., 2007; Zoroglu et al., 2003).  43 Table 1 Hypothesized Risk and Protective Factors for Self-Injury in the Current Study  Risk Factors  Protective Factors  Contextual:  Attachment disruption (placed in-care)  Trauma  Sexual abuse, sexual assault, multiple perpetrators  Physical abuse, physical assault, multiple perpetrators  Witness family violence  Kicked out of home  Parent problems (drug, alcohol, mental illness, criminal activity)  Sexually exploited  Street exposure  Victim of relational aggression  Overall victimization  Family history of suicide Individual:  Gender, age, sexual orientation  Substance use (early initiation, binge-drinking, injection drug use, problem use)  Disordered eating behaviours  Suicide attempt  Physical risk-taking (sharing needles, drug gear, razors, etc.)  Physical fighting  Contextual:  Family connectedness  School connectedness  Extracurricular activities  Extreme sports (sensation-seeking)  Individual:  Body satisfaction, positive body image  Emotional health/low emotional distress  Perception of life circumstances  Positive future outlook  Self-perceived health status  Getting along with school peers  Current school attendance  Educational aspirations  A history of sexual abuse has been found in a number of studies to be a risk factor for self-injury and there is some evidence that dissociation mediates this relationship, although much of this research has been conducted with adult clinical samples (Santa Mina & Gallop, 1998; van der Kolk et al., 1991; for notable exceptions, see Kiesel & Lyons, 1999; Zoruglu et al., 2003). No research has examined aspects of sexual abuse including frequency, severity, and perpetrator for their relation to self-injury, with the exception of a study by Santa Mina  44 and Gallop (1998) in which they found a unique association in adult women who were victims of incest. Previous research on other forms of sexual abuse, including involvement in the sex trade, has found an association between sexual exploitation and suicidality (Kidd, 2002; Kidd & Kral, 2002). However, in their study of street youth, Tyler and colleagues (2003) found that whereas ‗deviant subsistence strategies‘ (namely stealing, drug dealing, and shoplifting to survive) were significantly associated with self-injury in a multivariate model, sex trade involvement was not. A recent meta-analysis found that the relationship between child sexual abuse and NSSI was small, prompting the suggestion that sexual abuse functions as a proxy risk factor rather than as a unique or causal link (Klonsky & Moyer, 2008). Further research is necessary to better understand these relations. Findings regarding the relation between physical abuse and self-injury are inconsistent, with evidence of a significant relationship in a study with community adolescents (Zoruglu et al., 2003) and another with college students, especially among males (Gratz et al,. 2002), and yet a third with community adults (Wiederman et al., 1999), but insignificant findings with a small clinical adolescent sample (Glassman et al., 2007), a college sample (Whitlock et al., 2006), and with adult clinical samples (Zweig-Frank, Paris, & Guzder, 2004). Limited research on the association between other violence exposure and NSSI exists, however a 1999 study with 147 adult women in a primary care setting found that witnessing physical violence was independently related to self-injury (Wiederman et al., 1999). As aforementioned in this review, street youth are exposed to an alarming level of violence that, for many, starts prior to any street-involvement and continues with the dangers of the street. A caveat for much of this research is that because of its retrospective cross-sectional nature, temporality is unknown and it is possible that the onset of self-injury predates the potential risk factors. Further research is needed to explore the association between these traumatic experiences and self-injury and to determine the underlying factors or mechanisms.  45 Although a number of case studies have reported a link between childhood loss and separation from caregiver(s), limited empirical research has specifically investigated the relation between this experience and later NSSI (Gratz et al., 2002; Walsh & Rosen, 1988). However, studies with nonhuman primates have shown that disruption in early parental care (i.e., deprivation from caregiver) leads to attachment difficulties, impaired pain perception, and behavioural self-stimulatory problems including self-biting (e.g., Mineka & Suomi, 1978). Disrupted attachment is common among homeless youth, many of whom have been taken into care, often as a result of loss or maltreatment, parent substance misuse or psychopathology (Cauce, 2000; Haber & Toro, 2004; Ryan et al., 2000; Thrane et al., 2006; Tyler, 2006; Whitbeck et al., 1997; Whitbeck et al., 1999). Interruptions in caregiving impacts the child‘s ability to feel secure and nurtured and may result in feelings of perceived abandonment and traumatic loss that challenge coping abilities and/or prevent the acquisition of adaptive self-regulatory skills (Suyemoto, 1998; van der Kolk & Fisler, 1994; van der Kolk et al., 1991; Yates, 2004). A common circumstance among street youth that also entails loss and separation, being asked to leave or being ‗kicked out‘ of a family/guardian living situation, has not been explored for its relation to self-injury. It is unlikely to have a direct effect, but may be related through the mechanism of perceived or objective rejection which may confer additional vulnerability in situations of background risk. In fact, the literature on street-involved and homeless youth refers to these youth as ―throwaways‖ (Haber & Toro, 2004; Hammer et al., 2002; MacLean et al., 1999). Using current study data, Smith et al. (2007) found that 54% of the youth had been kicked out of their home at least once, with 60% by age 14. Research evidence suggests that patterns and combinations of cumulative and multiplicative risk and protection over time negatively impact adolescent adaptation (Fergus & Zimmerman, 2005; Hoyt et al., 1999; Masten et al., 1993; Luthar et al., 2000; Ryan et al., 2000; Sameroff et al., 2003; Whitbeck et al., 2000). As aforementioned throughout this  46 paper, street youth suffer from a disproportionate number of negative life events; many go from troubled families to navigating troubled streets. Research suggests that abuse experience is a predisposing factor for further victimization and certainly a plethora of studies provides evidence of multiple abuse and victimization experiences (e.g., Hoyt et al., 1999; Thrane et al., 2006; Tyler et al., 2000). The historical (and too often current) experiences that are common to both street youth and to adolescents who self-injure suggest that an index highlighting the multiplicative effect of these experiences may be of use in differentiating youth who self-injure from those who do not. Street exposure. Little is known about whether or how street exposure (i.e., living situation – housed/no shelter and length of time on the street) is associated with self-injury. According to Goodman and colleagues (1991) and their research with homeless adults, trauma theory suggests that living on the street or being homeless is a form of psychological trauma that acts to both cause new problems and exacerbate existing problems. In their model of risk amplification, researchers Whitbeck et al. (1999) suggest that the negative trajectory of homeless youth is amplified by their street experience, referring to the ―psychological impact of unprotected living‖ (p. 292). The salience of street exposure is further highlighted by research showing that despite different background experiences, exposure to the street (particularly in the form of being homeless) is a ―great equalizer‖ (MacLean et al., 1999, p. 186), meaning that the trauma of homelessness overrides any differential background factors. In a recent B.C. study with young adults, Rachlis and colleagues (2009) found that precariously housed or homeless youth were at risk for greater adverse health outcomes. Tyler and colleagues (2003) reported that sleeping on the street for at least one night was associated with an increase in the number of self-injurious behaviours, although unfortunately these researchers did not examine its relation to the presence or absence of self-injury. Alternatively, it may be that being housed provides some protection against self-  47 injury, although the type of living situation (e.g., living at home with parents, housed on own, living on streets) may interact with other factors to create either additional risk or protection. Tyler et al. also investigated the relation between the age at which the youth first ran away from home and self-injury and found a negative association. Further, these researchers found that what they refer to as street exposure, sleeping on the street and age at first run, accounted for 4% of the variance in self-injury in multivariate models, with the former being significant in all models tested. Thus, findings suggest that street exposure is of importance when examining risk behaviours or outcomes, however scarce research has examined the intersection of street exposure and self-injury. Family connectedness. A considerable body of research on risk, resilience, and protection and competence has found positive family relationships and/or connectedness to be a global protective factor (e.g., McNeely & Falci, 2004; Masten et al., 1991; Resnick et al., 1997; Sameroff et al., 2003; Sieving et al., 2001). Despite clear practical implications, a dearth of empirical research exploring the quality and security of attachment relationships to parents or primary caregivers and NSSI exists, with most research focusing on the negative aspects of family relationships. Recent research hypothesizing an association between perceived parental criticism and internalized criticism and self-injury (Glassman et al., 2007) found some support, although the specificity of the pathway was not substantiated. Findings from two recent studies examining family relationship dimensions (Bureau et al., 2010; Yates et al., 2007) highlight the salience of alienation from parents as an aspect of dysregulation between parent-child relationships that may be especially related to NSSI. Interestingly, Yates and colleagues found alienation to be significant especially among adolescent boys whereas Bureau et al. found this to be true among the young adult women in their sample. Research on adolescent suicide has demonstrated that positive parent relationships (i.e., family cohesion or connectedness) function as a protective factor against suicidal behaviour  48 (Borowsky et al., 2001; Erickson, 1999; Rubenstein, Heeren, Housman, Rubin, & Stechler, 1989; Saewyc, Wang, Chittenden, Murphy, The McCreary Centre Society, 2006). School connectedness. School connectedness refers to the qualitative aspects of the educational experience such as feeling cared for by teachers and positive relations with teachers and peers, rather than academic performance, and reflects a sense of belonging (MacKay, 2007; McNeely & Falci, 2004). Positive school relationships can provide an important source of support that may be especially important to marginalized youth, many of whom have troubled family backgrounds and/or a history of negative events. A growing body of research has found school connectedness to be protective against a variety of negative behavioural/health outcomes (Blum et al., 2002; Borowsky et al., 2001; Masten et al., 1991; McNeely & Falci, 2004; Resnick et al., 1997; Saewyc et al., 2006; Sieving et al., 2001). In describing this relation, Blum and colleagues (2002) reinforce the importance of the presence of caring adults and also highlight the benefit of access to school resources which they refer to as a form of social capital. No research has systematically examined school-related variables for their relation to NSSI. In their 2007 study of deliberate self-harm, Brunner and colleagues (2007) referred to a variable they called school performance, but did not operationalize it clearly; further, they collected data on whether the youth had trouble getting along with teachers, but did not state significance of their findings (showing that more frequent self-injury was associated with decreased quality of relationship). Moreover, although street-involved youth are notoriously ‗disengaged‘ from school and frequently have troubled educational histories, no research has specifically examined school connectedness or related school variables in this population. Given both the importance of education, supportive adults, and the fact that a considerable number of street-involved adolescents attend school (Smith et al., 2007), an improved understanding of school-related factors may have important implications.  49 Community involvement and activities. No research to date has investigated overall community involvement or various activities for their association to NSSI. Community involvement may act as a proxy for community connectedness or for personal competence and development. This type of involvement fosters positive socialization and interaction with others, connections to positive adult role models, and/or the development of skills (i.e., those involved in mastering a sport/recreational activity, arts, etc.). Researchers at The Search Institute ( developed a framework of external (contextual) and internal (individual) developmental assets derived from the empirical literature on resilience, protection, and adolescent development. These assets represent aspects of positive development that previous research as well as theory suggest may protect against or inhibit risk behaviours (Leffert et al., 1998). Within this framework, they have identified an asset category they refer to as ―constructive use of time‖, defined as spending time in structured youth programming or creative activities. This asset has been found to protect against a range of risk behaviours, including suicidal behaviour (Leffert et al., 1998). Identified not only as activities, but also as strategies for coping, these types of actions may also function to regulate mood. For example, Kidd‘s (2003) study with street youth showed that some rely on strategies such as hobbies, sports, and recreational activities to cope with their situations. Another study (Smith et al., 2007) revealed that street-involved adolescents engaged in considerable levels of extracurricular activities prior to street-involvement and continued to engage in activities while on the streets. These researchers found that although fewer youth accessed structured activities (e.g., coached sports), the number of youth reporting unstructured sports activity increased (from 41% to 64%, respectively). Extreme sports. Elements underlying extreme sports participation such as sensation-seeking, physical exertion, and corollary neurotransmitter (and tension) release (e.g., endorphin, adrenalin) may also be common to NSSI. Research on functional aspects of self-injury have identified sensation-seeking as one possible model (Klonsky, 2007; Klonsky &  50 Glenn, 2009; Yates, 2004). Previous research has situated NSSI within a continuum of selfdestructive and risky behaviour (Favazza, 2006; Laye, 2002; Nock, 2010; Skegg, 2005; Walsh, 2006). Walsh (2006) suggests that best practice assessment of self-injury must include investigation of a range of risk taking and reckless behaviours. Involvement in extreme sports requires taking calculated risks; by extension, this participation may be linked to self-injurious behaviour, with those who participate being less likely to self-injure. Further, pioneers in the field suggest that symptom substitution may be relevant (Favazza, 1996; Nock, 2010; Walsh, 2006), with evidence of alternate emotion-focused-type coping strategies or behaviours such as eating disorders and/or substance (Alfonso & Dedrick, 2010; Brunner et a.., 2007; Hilt, Nock et al., 2008; Nock et al., 2006; Ross et al., 2009; Serras et al., 2010; Svirko & Hawton, 2007). If extreme sports function as another form of effective mood/affect regulation, they may provide some protection against self-injury. Individual Factors Age and gender. Self-injury typically originates in adolescence, most often between 12 and 15 years of age (Alfonso & Dedrick, 2010; Hilt, Nock, et al., 2008; Muehlenkamp & Gutierrez, 2004, 2007; Nixon et al., 2002; Nock & Prinstein, 2004; Ross & Heath, 2002). Evidence is mixed regarding age. Results from some studies suggest that prevalence rates may decrease in late adolescence and again in adulthood (Nock & Prinstein, 2004, 2005; Whitlock et al., 2006), although studies with older adolescents and young adults have found equally high prevalence rates (Gratz et al., 2002; Polk & Liss, 2007; Whitlock et al., 2006). Findings should be interpreted cautiously as a majority of studies feature retrospective design and samples are often biased toward clinical and adult samples with few older adults. In their study of homeless youth, Unger et al. (1997) found a significant correlation between age and self-injury, with younger female adolescents being most likely to engage in the behaviour. In contrast, Tyler et al. (2003) found that older youth reported more self-injurious acts, but their  51 sample was restricted in terms of age, with only 16-19-year-olds participating. Increased understanding of age-related trends will provide further insight into this complex behaviour. Recent research with adolescent samples has called into question the commonly held belief that self-injury is a ―female problem‖ (Laye-Gindhu & Schonert-Reichl, 2005; Gratz et al., 2002; Hilt, Nock, et al., 2008; Muehlenkamp & Gutierrez, 2004, 2007; Tyler et al., 2003; Zoroglu et al., 2003). Given the lack of research involving males and the mixed findings examining gender and self-injury across different populations, further research on gender differences in prevalence, nature, and function of self-injury, as well as investigation of whether risk and protective factors operate differentially across genders is necessary. Sexual orientation. As a further marginalized and stigmatized group, adults and young adults indicating sexual minority status (i.e., gay, lesbian, bisexual) have been found to report higher rates of NSSI compared to their heterosexual counterparts (Tyler, Melander, & Almazan, 2010; Skegg et al., 2003; Whitlock et al., 2006), although limited research with adolescents specifically provides support for this relation (Deliberto & Nock, 2008; Tyler et al., 2003). A significant number of street youth identify themselves as belonging to a sexual minority group or as not exclusively heterosexual, more than 40% in a study of British Columbia street-involved adolescents (Smith et al., 2007). Ethnicity. To date, limited research has examined the association between ethnicity and self-injury and those few studies that have report conflicting findings. Minimal research on self-injury has been conducted in Canada and, to date, no statistics exist to identify the prevalence within Aboriginal/First Nations peoples. Youth of Aboriginal/First Nations backgrounds are disproportionately represented within the street-involved population (Murphy & The McCreary Centre Society, 2001; Smith et al., 2007). There exists a clear need for further consideration of ethnic differentials in self-injurious behaviour. Risky and risk-taking behaviour. Research has emphasized the interrelationships among health-risk behaviours, including self-harming behaviours. Findings from a number of  52 studies have suggested that individuals who engage in one self-destructive behaviour or health-risk behaviour are more likely to also engage in others (Alfonso & Dedrick, 2010; Brunner et al., 2007; Jessor, 1991; Patton et al., 1997; Skegg, 2005; Zweig, Lindberg, & McGinley, 2001). Behavioural characteristics found to be associated with both NSSI and with street-involvement include delinquent behaviour, general self-destructiveness, recklessness, disordered eating, substance use, risky sexual behaviours and, among a sample of clinical adolescents, sharing cutting tools (Brown et al., 2008; Brunner et al., 2007; Cauce, 2000; DiClemente et al., 1991; Hilt, Nock, et al., 2008; Laye-Gindhu & Schonert-Reichl, 2005; MacLean et al., 1999; Nock et al., 2006; Patton et al., 1997; Whitbeck et al., 2000), although the exact nature of the relations has not been adequately studied. Research has variously highlighted the co-occurrence of these behaviours and the similar functions they serve, whereas other research refers to symptom substitution with youth switching between different behaviours (Brunner et al., 2007; Favazza, 1996; Walsh, 2006). An unpublished study (Walsh & Frost, 2005, as cited in Walsh, 2006) with ―polydestructive‖ adolescents (p. 29), showed that the majority of the sample engaged in other more indirect self-harm; 94% reported physical risk-taking (i.e., ‗flirting‘ with death, e.g., walking in high speed traffic), 85% situational risk-taking (e.g., getting into a car with strangers), and many engaged in sexual risk-taking (e.g., many sexual partners, failure to use protection). In his comprehensive overview of self-injury, Walsh (2006) notes that lack of selfcare and physical risk taking should be a critical piece of a thorough self-injury assessment. In a study of school-based adolescents, Laye (2002) found that youth who reported self-injury were more likely to report concurrent risk-taking such as sensation- or thrill-seeking behaviour and recklessness and that this relation held true even in a multivariate model. These types of behaviours are infrequently measured in self-injury research. Street youth, whether because of limited opportunity to avoid risk, lack of self-care, self-destructiveness, lack of self-care, or  53 high-risk lifestyle, report engaging in significant risky and risk-taking behaviour (Lloyd-Smith et al., 2008; Murphy & The McCreary Centre Society, 2001; Smith et al., 2007). Physical fighting. Although no studies have investigated fighting specifically, review of research suggests that there may be links between self-injury and other self-harming or self-destructive behaviours that have not previously been considered. Moreover, these links may differ by gender (Laye-Gindhu & Schonert-Reichl, 2005; Walsh, 2006; Whitlock et al., 2006). For example, a recent study with college students in the U.S. found that almost 6% of their large sample reported engaging in physical fights or ―other aggressive activities‖ with the intent of injury (Whitlock et al., 2006, p. 1943). Similarly, anecdotal evidence from youth and family workers who regularly work with high-risk adolescents in the alternate education system suggests that boys may be engaging in physical fights with the goal of getting hurt, as suggested by several boys who reported going out on the weekends specifically aiming ―to get beat up‖ (M. Sullivan, personal communication, February 17, 2006). Given these findings, it may be that specific methods of injury are more unique to males or to certain individuals, but this has not been explored in the extant research. Substance use. The comorbidity of NSSI and substance use has been substantiated in previous research (Alfonso & Dedrick, 2010; DeLeo & Heller, 2004; Hilt, Nock, et al., 2008; Serras, Saules, Cranford, & Eisenberg, 2010; Yates, 2004), although few studies have provided further insight into this relation. Studies with clinical samples of adolescents have consistently found high rates of substance use disorder (Nixon et al., 2002; Nock et al., 2006; Schwartz et al., 1989). In contrast, results from a community sample of almost 6000 schoolbased adolescents revealed an elevated risk of NSSI for adolescents who ―occasionally‖ consumed illicit drugs, but frequent consumption was unrelated (Brunner et al., 2007). Both NSSI and substance abuse are effective means of mood regulation or coping, at least in the short-term, and are more common in people whose capacity to self-regulate emotions is impaired (Walsh, 2006). Given the prevalence of substance use among street youth, earlier  54 initiation may be more indicative of risk and entrenched coping deficits. Other potential risk factors that have not been specifically examined in the literature for their relation to self-injury include frequent binge-drinking and injection drug use. Parallels between actual behaviour – deliberately self-injuring with a ―tool‖ (i.e., needle, razor, lighter, etc.) and injecting a noxious substance (i.e., illegal drug) – both acting as effective short-term strategies to alter mood and ―feel better‖ and both fitting into a larger category of emotion-focused and/or avoidant coping, are interesting but have not been examined in the literature. Given the self-harming nature of these behaviours, we might expect a higher rate of each among self-injuring adolescents. Disordered eating and body image. Previous research has indicated a cooccurrence of eating disorders and self-injury among primarily clinical samples, but also in community samples (Alfonso & Dedrick, 2010; Darche, 1990; Nock et al., 2006; Ross, Heath, & Toste, 2009; Serras et al., 2010; Svirko & Hawton, 2007; Whitlock et al., 2006). In studies with high-school and university students, results showed links between eating pathology and NSSI, with more frequent injury associated with greater deficits in affect regulation and impulse control (see Ross et al., 2009; Whitlock et al., 2006). In their community sample of adolescents in Germany, Brunner and colleagues (2007) found that adolescents who perceived themselves as overweight demonstrated a 3-fold greater risk of self-injury, a risk that increased with low body mass. Researchers suggest that a negative attitude toward the body, body alienation, and poor body image are important in facilitating NSSI (Darche, 1990; Brunner et al., 2007). In their internet-based study of adolescents, Murray et al. (2005) found that 40% reported being never satisfied with their body shape and 23% rarely so. Further, 69% reported never or rarely feeling sexually attractive. Although no research exists to specifically support body satisfaction as protective against self-injury, it can be speculated that individuals who are satisfied with their bodies in terms of weight and who are not dieting might be less likely to engage in a physically damaging and self-punishing behaviour. Despite the food insecurities related to poverty and homelessness, evidence from a BC study  55 suggested that low body image and disordered eating behaviours such as binge-eating and purging were even more common among street-involved adolescents relative to their schoolbased peers (Murphy & The McCreary Centre Society, 2001). Suicidality. A theoretical relation has been presumed and explored empirically between suicidal behaviour and NSSI and overlap between the behaviours has been noted and reviewed earlier in this paper, with suicidal behaviour representing the most salient behavioural risk for NSSI to date (Brausch & Gutierrez, 2010; Brunner et al., 2007; Jacobson & Gould, 2007; Kidd & Kral, 2002; Muehlenkamp & Gutierrez, 2004, 2007; Stanley et al., 2001; Whitlock & Knox, 2007). In the absence of prospective longitudinal studies, the exact nature of the temporal relation between self-injury and suicide remains a mystery. The literature on street youth is replete with descriptive investigations of explicitly suicidal behaviour, citing a 20-40% attempt rate (Kidd, 2006; Molnar et al., 1998; Smith et al., 2007; Yoder et al., 1998), rising to 76% among youth involved in the sex-trade (Kid & Kral, 2002). Suicide is a leading cause of death among this group (Roy et al., 2004). Given the overlap between self-injury and suicide, these alarming rates, and the high-risk lifestyles that are typical of street youth (Cauce, 2000; Rachlis et al., 2009; Smith et al., 2007; Tyler et al., 2001; Tyler et al., 2003), further investigation of the relation between self-injury and suicide and the interrelations among risk and protective factors is critical. Emotional health and distress. Frequently comorbid with NSSI, negative or undesirable affect has been identified as an important diagnosis, risk factor, and correlate, as aforementioned in this review. Associations with emotional distress, including depression and anxiety, have been demonstrated for self-injury (Alfonso & Dedrick, 2010; Brunner et al., 2007; Garrison et al., 1993; Hilt, Cha, & Nolen-Hoeksema, 2008; Kiesel & Lyons, 1999; LayeGindhu & Schonert-Reichl, 2005; Muehlenkamp & Gutierrez, 2004; Nixon et al., 2002; Nixon et al., 2008; Nock et al., 2006; Ross & Heath, 2002, 2003), although findings are not consistent across the literature and these constructs are not specific to NSSI. Research on  56 adolescent suicide has identified emotional well-being and positive affect (as well as low levels of negative affect) as a protective factor at both the bivariate and multivariate levels (Borowsky et al., 2001; Erickson, 1999). Few studies on NSSI have explicitly examined the role of low levels of distress, positive affect, or emotional health, although it is likely they act as risk modifiers. Evidence for the link between internal distress symptoms and NSSI has shown the moderating effects of emotion regulation styles such as rumination (Hilt, Cha, & Nolen-Hoeksema, 2008). High levels of psychic distress (i.e., depression, anger, anxiety, emptiness, loneliness, hopelessness) reported by self-harming individuals prior to engaging in self-harm and increased levels of positive affect afterward (Jacobson & Gould, 2007; Laye-Gindhu & Schonert-Reichl, 2005; Murray et al., 2005; Nixon et al., 2002; Whitlock & Knox, 2007) lend support to the conceptualization of the behaviour as a mechanism for regulating and coping with aversive emotions (Lloyd-Richardson et al., 2007; Nixon et al., 2002; Nock & Prinstein, 2005; Rodham et al., 2004; Ross & Heath, 2003). In the absence of more adaptive selfsoothing strategies, self-harm effectively regulates overwhelming and uncontrollable negative affect and/or interrupts dissociative states (Laye-Gindhu & Schonert-Reichl, 2005; Rodham et al., 2004; Ross & Heath, 2002), enabling the individual to create a sense of control. Given the powerlessness experienced in the lives of homeless youth, the multiple stressors, and the concomitant psychological problems, these youth would be expected to have increased rates of anxiety, stress, depression, and hopelessness as manifested by this negative coping behaviour. Thus, having low levels of emotional distress may be protective against selfinjurious behaviour. This aligns with the treatment notion that the behaviour is not the problem, but instead is the ‗solution‘ and therefore intervention focuses on the underlying problems that are driving this behaviour, for example, significant emotional distress coupled with poor emotional regulation skills.  57 Optimism and positive attitude. In research that focuses on adults and young adults, Skegg (2005) refers to an ―optimistic outlook‖ (p. 1474) as a key personal protective factor against deliberate self-harm (i.e., injury regardless of intent, including suicidal behaviour). Similarly, Muehlenkamp and Gutierrez (2007) suggest that attraction to life, identifiable reasons for living, and strong self worth and efficacy increase resilience and decrease risk of suicide among adolescents who engage in self-injurious behaviour. Interestingly, results from these researchers indicate that those adolescents whose outlook on life is more negative or who have difficulty identifying positive goals are more likely to attempt suicide. The findings may have links to related constructs including future orientation and outlook, aspirations, and optimism. Leffert and colleagues (1998), in research on developmental assets and their relation to risk behaviours, identified the salience of what they categorize as personal identity and within that they refer to a similar construct, ―positive view of a personal future‖ (p. 216). Living with various kinds of historical and daily adversity, street youth may find it particularly difficult to maintain hope for the future. Although research has shown that some see themselves living a more mainstream existence in the future (e.g., having a home, legal job), others expect they will be dead or continue to live on the street (Smith et al., 2007). Additional research is needed to understand how these factors relate and impact risk or protection for self-injury. Cumulative Summary Both street-involvement and NSSI, affecting millions of adolescents each year, represent distinct concerns with significant implications for health and future functioning. The preceding sections have reviewed the extant literature regarding adolescent streetinvolvement, nonsuicidal self-injurious behaviour particularly among adolescents, and provided an integration relevant to the current investigation. Clinical and research interest is rapidly growing with new evidence of the pervasiveness of the behaviour and reminders of its strong link with suicide, the third leading cause of death for Canadian adolescents. Scant  58 research has systemically and empirically investigated NSSI, most typically conceptualized (and functioning) as a strategy for coping with overwhelming and uncontrollable negative affect, in the street youth population, despite evidence of its alarming prevalence and related health risks (Smith et al., 2007; Tyler et al., 2010; Tyler et al., 2003; Unger et al., 1997). The suggestion that studies be conducted to examine the interaction between and among individual and contextual risk and protective factors related to NSSI has also been made (Gratz, 2003; Jacobson & Gould, 2007), echoing the call from adolescent health researchers interested in risk and resilience who view this step as critical (e.g., Blum et al., 2002; Luthar et al., 2006; Masten et al., 1990). Review of the extant literature reveals substantial overlap in the histories and etiological pathways and risk factors for both self-injury and street-involvement. Although a majority of street-involved adolescents have adverse life experience, some engage in NSSI whereas others do not. Going beyond traditional focus on risk, resilience theory suggests the inclusion of factors that are protective against NSSI, increasing positive adaptation. An understanding of the factors that predict or influence the likelihood of self-injury is necessary for planning appropriate and effective prevention and intervention efforts. Findings from research considering the relation between and among risk and protection frequently reveal that more risk, in the absence of any protection, is most predictive of negative outcome, with increases in protective factors making a useful contribution to influencing the likelihood of outcome (Blum et al., 2002; Erickson, 1999; Rubenstein et al., 1989). To date, protective factors have been vastly understudied in the street youth literature and ignored in the nonsuicidal self-injury literature, despite having clear and practical implications. Research that involves both risk and protection in combination more accurately reflects reality, as humans manifest both vulnerabilities and strengths and face both challenges and opportunities.  59  CHAPTER 3: METHOD This secondary analysis used data of street-involved adolescents who participated in the McCreary Centre Society‘s (MCS) Street-Involved Youth Health Survey (SYHS, 2006) during fall and early winter 2006. Sample A total of 762 marginalized and street-involved adolescents aged 12 through 18 (M = 16.34, SD = 1.58) participated in the Street-Involved Youth Health Survey (SYHS; to be further described in a subsequent section). The youth represented a large, systematic sample drawn from nine different urban and suburban communities across British Columbia (BC), Canada, including: Vancouver, Surrey, Abbotsford/Mission, Victoria, Nanaimo, Kamloops, Kelowna, Prince George, and Prince Rupert. Nine communities were selected in order to maximize generalizability1. The nine communities selected to participate in the SYHS were identified both because they were geographically distributed throughout the province and because they reported having significant and measurable street-involvement or homelessness. Approximately 3% of youth approached by a research team member either declined to participate or did not fit the criteria, primarily because they were over 18 (1% refusal rate of those fitting criteria; E. Saewyc, personal communication, July 17, 2007). Of the 762 youth surveyed, 48% were male, 50% were female, 1% transgender, and 1% did not respond to the gender item. For the present purposes, the two latter categories of gender, youth with no response (n = 6) and youth who identified as transgender (n = 6), were  1  Efforts were made to capture a census sample of street-involved youth under age 19. Random or populationbased sampling of street-involved or homeless youth proves challenging as they tend to move in and out of homelessness and are an ‗invisible‘ group (Berry, 2007; Haber & Toro, 2004). Based on consultation with the participating communities, this sample is believed to be an excellent attempt at representation of street-involved youth across the province, with as many youth as possible being sampled (E. Saewyc, personal communication, July 17, 2007).  60 not included in analyses2. Aboriginal youth represented 54% of the youth sampled (40% exclusively Aboriginal), and 33% reported being of European background, with another 13% indicating ―multiethnic.‖ Other ethnic groups included Hispanic (1%), African (1%), East Asian (1%), South East Asian (1%), and South Asian (1%), and 8% of the youth reported not knowing their ethnic background. With regard to sexual orientation, almost half of the youth reported not being exclusively heterosexual and just over 20% reported being gay, lesbian, or bisexual. To provide greater detail about the circumstances of youth in the total sample, according to the 2007 MCS report (Smith et al., 2007), the youth reported significant instability in their lives with nearly half (40%) having been in the government care system in their lifetime, either in a foster or group home. A majority had a history of being kicked out of their home at least once or running away from home, and two-thirds reported experiencing both. In general, the youth reported highly unstable housing in the year prior to the survey, with many youth living in multiple different types of living situations. Procedure The McCreary Centre Society‘s unique procedures for survey design, participant recruitment, and data collection for the SYHS are noteworthy, and serve to increase reliability and validity of the findings. The MCS study utilized a participatory research design in which community research teams (i.e., a team for each of the nine communities) engaged collaboratively and interactively throughout the study process. Teams consisted of MCS researchers, experiential youth (i.e., youth who were then currently, or had been in the past, ―street-involved‖), and staff from community agencies, or university students studying the social sciences. All team members were required to complete the same in-person training  2  Transgender youth are a marginalized and vulnerable group who are frequently rendered invisible in scientific research (Grossman & D‘Augelli, 2006). Although the merits of investigating the health disparities, risk and protective factors and specific and unique needs of transgender youth are indisputable (Grossman & D‘Augelli, 2007; Russell, 2003), the small number of transgender youth in the current sample made further study impossible.  61 which focused on relationship building and on the specifics of the research project. The community research teams and an Advisory Committee of community agency and government staff familiar with vulnerable youth helped to develop the sampling method and provided feedback on the questionnaire through an interactive and iterative process, facilitated by MCS. The experiential youth actively participated in an early pilot of the questionnaire, lending their expertise to guide the refinement of the survey. The community research teams recruited the youth and administered the survey. Beyond enhancing the legitimacy of the study, the inclusion of each community research team in all stages of the MCS study also provided the opportunity to maximize street youth recruitment and participation. Further benefit included the ownership felt by each community after having been involved throughout the process. From October through December 2006, street-involved youth were recruited by the community research teams through community agencies and through word-of-mouth and street outreach (i.e., direct on-street contact, meeting the youth where they were, e.g., parks, libraries, on street, bus station, community centres, street agencies). Due to challenges inherent in sampling street populations (Berry, 2007), the goal of wide representativeness was achieved through saturation of sampling of venues, with all street agencies and all public locations in each of the nine communities sampled. Survey administration locations, times, and days were varied in order to maximize access. The SYHS was administered either individually or in small groups of five or fewer youth at a time, and the research team members read aloud all questions and possible responses in order to overcome any potential literacy issues. Upon completion of the survey, to ensure anonymity and confidentiality, the youth were instructed to place their survey form in the envelope provided. All participating youth were given $10 in either coupons or cash as a gesture of appreciation from the research team.  62 Instrument and Variables Street Youth Health Survey (SYHS) Data were obtained via an adaptation of the provincial Adolescent Health Survey (AHS), developed by the McCreary Centre Society for use across the province in population health research with school-based youth, and its companion Street Youth Survey of 2000 (Murphy & The McCreary Centre Society, 2001). Close to 100,000 adolescents in British Columbia have participated in the AHS since its inception in 1992 (AHS-I, AHS-II, AHS-III, AHS-IV); further, over 500 street-involved youth in six communities across British Columbia completed the 2000 Street Youth Survey. The current study utilized a version of the AHS, entitled the Street-Involved Youth Health Survey (SYHS, 2006), consisting of a core set of questions from the school-based AHS III (2003), items included on the previous survey with street-involved samples (e.g., items related to living situation and street lifestyle, reasons for leaving home, sexual exploitation, substance use), and new items focusing on mental health and addictions. Like the AHS, the SYHS is a voluntary, anonymous, self-report survey. Consistent with current research that emphasizes risk reduction as well as resilience and health promotion, the SYHS focuses not only on risky behaviours and risk factors, but also on strengths and protective factors. Items on the Adolescent Health Surveys (AHS) are derived from several different highly reputable population health surveys [e.g., Adolescent Health Survey, University of Minnesota, (Blum et al., 1988), Centers for Disease Control and Prevention‘s Youth Risk Behaviour Survey (Kolbe, Kann, & Collins, 1993), World Health Organization‘s Health Behaviour in School-Aged Children (King, Wold, Tudor-Smith, & Harel, 1996), National Longitudinal Study of Adolescent Health (Resnick et al., 1997)], have undergone extensive field-testing, and were selected based on their psychometric properties. Other items were developed by MCS, in consultation with a cadre of adolescent health researchers, clinicians,  63 front-line workers, youth leaders, and policy makers (see Appendix A for Source of Survey Questions). Importantly, with McCreary‘s focus on youth development and engagement, youth were consulted in a variety of ways (e.g., focus groups with street-involved youth, Youth Advisory Council) both before and after their 2000 street youth study and during the development process for the SYHS. A paper-and-pencil questionnaire, the SYHS has 150 questions (147 forced choicedichotomous or Likert-scale and three open-ended), although given non-mutually exclusive options and questions within questions (e.g., timing, ever vs. yesterday vs. 30 days vs. 12 months all considered under one question number), there are actually closer to 500 different variables or individual response decisions. The questions address a range of topics including: demographics, family background, exposure to violence, educational experiences, physical and emotional health, health-risk behaviour (e.g., sexual behaviour, substance use, disordered eating, suicide, recklessness), health-protective behaviour (e.g., condom use), antisocial behaviour, housing/shelter issues, street experiences, income/employment, and personal attitudes and strengths. Some of these topics refer to individual characteristics, whereas others are concerned with contextual or environmental characteristics such as those of family, school, and larger community. The SYHS was designed to be user-friendly and completed within 45-60 minutes. SYHS items use dichotomous and categorical or ordinal response formats. Ordinal and categorical variables are commonly used in survey and health research and are a reliable and valid method of assessment. In the present study, when practical and meaningful, variables were dichotomized. Other constructs were examined through the use of established scales (e.g., connectedness scales) or through the creation of indices or composites (e.g., victimization index), with the final variable reflecting either categorical or ordinal/continuous data. All variables were standardized (0-1) prior to multivariate analyses, if  64 not already conforming to this scaling. This process will be further discussed in the data analysis section. Scales. Scales created and validated in previous research (see Resnick et al., 1997; Sieving et al., 2001; McNeely & Falci, 2004) include family connectedness (11 items), school connectedness (6 items), and emotional distress (or conversely in the present case, conceptualized as a protective factor, low emotional distress or emotional health; 4 items). In general, scale scores were computed by averaging responses to items for individuals with a minimum of 75% of completed items (Bearinger, Pettingell, Resnick, & Potthoff, 2010). However, where this was not appropriate (e.g., those youth without a parent or parents were not able to respond to parent specific items), cut-offs derived by MCS (e.g., youth must complete 3/11 variables in order to get a score for family connectedness; Saewyc & Homma, 2010) were adhered to. Although the scales have been found to be statistically reliable and valid measures with school-based youth on the surveys from which they are derived (e.g., National Longitudinal Study of Adolescent Health [Add-Health], University of Minnesota‘s Adolescent Health Survey) and they have been used by MCS in previous street youth surveys, they have not have been evaluated psychometrically with street-involved youth specifically. Therefore, basic psychometric analyses were conducted to assess the reliability (i.e., Cronbach‘s alpha) and validity of these scales for the current purposes. To evaluate the construct validity of the scales for the current study sample, principal components analysis (PCA) was performed using PASW 17.0 (SPSS, 2008) and, when indicated, confirmatory factor analyses (CFA) using the Lisrel 8.80 student version (Jöreskog & Sörbom, 2006). Development of composites and indices. After examining the distribution of individual variables, when possible and useful (e.g., feasible, conceptually meaningful, theoretically supported), related items from the SYHS were combined in psychometrically sound ways to produce composite or index variables. Base rates for some risk factors were  65 greatly elevated in this high-risk sample of youth, thereby limiting the usefulness of those factors as predictors of self-injury, or in distinguishing between youth who reported self-injury and those who did not. Use of composites or indices can help to extend the range of variation, adding greater sensitivity to measurement of some constructs. For example, a victimization index was created from several variables representing various direct and vicarious victimization experiences. Given that many street-involved youth have a history of victimization, it was hypothesized that a combined measure might tap into a greater range and could potentially provide more useful information without compromising rigor. However, as is common in population health surveys, many variables remained as single-items in the analyses. Outcome Variable: Self-Injury Three SYHS items relate to non-suicidal self-injury specifically (see Table 2). These items were developed by MCS for their survey purposes. Participants were coded as engaging in self-injury if they responded affirmatively, indicating that they had self-injured at least once in their lifetime; therefore, the self-injury outcome is a dichotomous variable. The second item, used only descriptively, asks youth about whether they sought medical attention for their injury. The third item, a question about self-injury function, asked youth to report reasons for last incident that were relevant for them from a list of eight options with a ninth ―other‖ category. Youth who indicated ―other‖ were asked for further description; if description matched a given category, responses were recoded into the appropriate category. The overall report published by MCS based on findings from the total street-involved youth sample states a lifetime self-injury prevalence rate of 56% for girls and 34% for boys (Smith et al., 2007).  66 Table 2 Outcome Variable: Self-Injury Measured Variable  Item #  Description of Variable(s)/ Survey Question(s)  Response Format  Self-injury  105  How often have you deliberately cut or injured yourself (but you weren’t trying to kill yourself)?  Never 1-2 times 3+ times  Self-injury medical Attention  106  The last time you deliberately cut or injured yourself did you seek medical attention?  Never injured Yes No  Self-injury function  107  Why did you deliberately cut or injure yourself the last time? …I was feeling lonely or depressed …I was bored …Problems w/drugs or alcohol …To get attention …To feel in control …Feeling rejected …Feeling stressed …Feeling angry …(Other: )  Potential Risk and Protective Factors Systematic review of the theoretical and empirical literature suggested variables that could be linked to self-injury. Therefore, the inclusion of some variables in the present study is supported by extant evidence of their association with self-injury, whereas others (for which previous research is equivocal or no known studies exist) were hypothesized to function as either risk or protective factors for self-injury in the sample. Factors were chosen for analysis according to an ecological framework, along both individual (i.e., personal) and contextual (i.e., environmental – family, school, community) dimensions. A thorough accounting of the risk and protective variables and factors selected for inclusion is presented, with specific items and description appearing in Appendix B. Sociodemographic variables. Participants were asked about demographic characteristics including age, gender, ethnicity, and sexual orientation. For the present study purposes, these variables were conceptualized as grouping variables that might be associated with disproportionate levels of risk rather than actual risk factors. Based on  67 empirical literature suggesting that self-injury may be more prevalent among females as well as evidence of gender differences in some of the hypothesized factors (e.g., sexual abuse, disordered eating behaviours, fighting), the present study generated separate models of risk and protection for adolescent boys and girls. Previous research supports that, in particular, a non-exclusively-heterosexual orientation (i.e., sexual minority status) may present additional risk for risk behaviours (Austin et al., 2004; Coker et al., 2010; Russell, 2003; Saewyc, 2011), including self-injury, although these findings have not been consistently replicated and have been primarily investigated in college and adult samples (Gratz, 2006; Tyler et al., 2010; Whitlock et al., 2006). After examining the distribution of the sexual orientation variable in the total sample, it was recoded in order to remove the ―not sure‖ responses and to collapse across the ―mostly homosexual‖ and ―100% homosexual‖ categories, thus leaving a categorical variable with four categories (100% heterosexual, mostly heterosexual, bisexual, and mostly or 100% homosexual; personal communication, E. Saewyc, personal communication, February 18, 2008; Austin, Conron, Patel, & Freedner, 2007; Saewyc et al., 2004; Smith et al., 2007). In later analyses, the ―mostly heterosexual‖ category was further examined against all hypothesized variables to determine if this group responded similarly (and therefore would be comparable) to other orientation categories such as the 100% heterosexual or the bisexual group. Inspection revealed that the ―mostly heterosexual‖ group most closely resembled the bisexual group, with higher levels of risk on most variables compared to the exclusively heterosexual group. Extant research examining other risk indicators and negative outcomes has reported similar findings, with ―mostly heterosexual‖ orientation (i.e., attraction) youth at elevated risk when compared to their exclusively heterosexual peers (e.g., see Austin et al., 2004; Austin et al., 2007; Coker et al., 2010; Thompson & Morgan, 2008). For the final logistic regression analyses, sexual orientation was dichotomized (100% heterosexual or sexual minority identification). This variable was then utilized as a grouping  68 variable to evaluate whether sexual minority youth were at disproportionate risk for NSSI as hypothesized. Extant research has provided some evidence that the relationship between sexual orientation and risk behaviour outcomes is at least partially mediated by general adolescent risk factors (e.g., victimization, depression, family suicide history). Researchers such as Russell (2003) suggest that sexual minority youth have disproportionately experienced normative risk factors and that risk disparities observed in research on suicidality, for example, could be accounted for by these risk factors as opposed to unique risk factors. Notably, the current study does not assess stigma and minority stress specifically, although these stressors may be qualitatively different and unique to this population (see Russell, 2003; Tyler et al., 2010). In contrast, minimal empirical evidence exists for the relation of ethnicity and selfinjury and findings from the few studies that include ethnicity are contradictory or equivocal (Muehlenkamp & Gutierrez, 2004, 2007; Yates et al., 2008). With half of the present study sample indicating any Aboriginal ethnicity, this variable was examined for its relation to selfinjury. Finally, whereas some research has found younger age is associated with self-injury (Unger et al., 1997), other research highlights the vulnerability of older youth (Tyler et al., 2003). Age, as a continuous variable, was examined as a possible demographic risk factor for self-injury. Further, given the considerable developmental differences that exist across adolescence, age effects were controlled for in analyses. Contextual Risk Factors Abuse. Two separate questions asked participants about any history of physical abuse or sexual abuse both within and outside of their families. Another question asked about experience of sexual coercion by an adult and/or by another youth. Responses indicating that youth had been forced to have sexual intercourse by an adult (―yes‖ to sexual abuse question and/or ―yes‖ to forced sex by an adult question) were used to create a final  69 dichotomous sexual abuse variable that included any experience of sexual abuse by an adult. A different dichotomous variable, sexually assaulted by a youth, was created from the sexual coercion question to indicate those youth who specifically reported being forced into sexual intercourse by another youth. Abuse perpetrators (multiple perpetrators). Research has revealed an association between abuse characteristics including type(s) or severity of abuse, frequency, duration, and perpetrator (e.g., intrafamilial, multiple) and impact or sequelae of abuse (Leserman et al., 1997; Nash, Zivney, & Hulsey, 1993). With the high rates of victimization in the study sample, the number of perpetrators was considered as potentially providing a more sensitive measure (E. Saewyc, personal communication, October 20, 2008). Youth were asked to indicate the perpetrator(s), if any, of physical and/or sexual abuse from a checklist of options (father, mother, step-parent, foster/group home parent, friend/acquaintance, romantic partner, trick/date, pimp/agency manager, police officer, stranger, other) and an index of the number of perpetrators was created, with separate indices for sexual abuse and for physical abuse. The distribution of the number of perpetrator variables was positively skewed (i.e., > 2.0) and leptokurtic (i.e., > 3) for both males and females, with empty cells (Tabachnick & Fidell, 2001). Although the subsequent logistic regressions did not require normally distributed variables, empty cells are problematic (Hosmer & Lemeshow, 2000; Peduzzi, Concato, Kemper, Holford, & Feinstein, 1996). Subsequent square root transformations to address distribution non-normality revealed that these new variables were neither stable nor added any further information regarding nature of abuse suffered. Thus, these variables did not prove statistically or theoretically useful. Instead, the distributions of perpetrators reported for sexual abuse and for physical abuse were examined separately by gender and dichotomized into two or more perpetrators (1) and less than two perpetrators (0) both for sexual abuse and for physical abuse, with the research-supported rationale that abuse by multiple perpetrators has been found to confer greater risk (Leserman et al., 1997; Nash et al., 1993).  70 Witnessed violence. One dichotomous item was used to assess whether the youth had ever witnessed a family member being assaulted or mistreated (0 = no, 1 = yes). Physically attacked/assaulted in past year. Youth were asked how many times, if any, they were physically attacked or assaulted within the previous twelve months. This variable was dichotomized to indicate never (0) or ever (1) being assaulted. Sexual exploitation. Youths‘ responses to questions about sexual exploitation demonstrated that they didn‘t always perceive the trading of sex for goods or shelter to be exploitive and therefore MCS researchers created a sexual exploitation indicator variable that utilized all possible responses to generate data indicating any history of trading sex for money, goods, shelter, or other (E. Saewyc, personal communication, November 12, 2007). The present study used this indicator variable, derived from eight different questions (representing 28 response choices). In-care status. Youth‘s living arrangement was assessed in one item with a matrix response format indicating time period (yesterday, past year, ever) and placed lived (15 possible responses and an ―other‖ category). For the present study purposes, youth who indicated ever living in a foster or group home were coded as having been in government care (E. Saewyc, personal communication, April 7, 2008). Parent problems. Four questions assessed whether youths‘ mother (4 items) and/or father (4 items) had a history of alcohol abuse, drug abuse, mental illness, or criminal record. Taken together, these items are indicative of parental mental health and risk behaviours, with implications for later adjustment through attachment disruption and social learning. The distribution of individual items was examined prior to the creation of two composite or index variables, reflecting the total number of problems dealt with by mother and by father, separately. Being kicked-out. Youth were asked, ―How old were you when you first got kicked out of home?‖ and were able to indicate at what age this first happened ranging from before  71 age 9 to age 17 or later. Responses were then dichotomized to reflect the experience of ever having been kicked out of home at any age (1) and never having been kicked out (0). Street exposure. Research suggests that street experiences can be very traumatic and that increased exposure to the street through unstable housing and absolute homelessness may predict more problems. In the current study, two questions related to street exposure were examined. One question (with a multiple response format) asked youth where they had been living. Youths‘ living arrangements during the previous 12 months were categorized as either precarious or not using an MCS-generated new variable (E. Saewyc, personal communication, April 8, 2008), with living in a shelter, squat, on the street, in abandoned building, ―nowhere‖, in a tent, and/or in a car identified as precarious (i.e., at imminent risk of literal or absolute homelessness). Another question that asked youth at what age they first became street-involved was examined and then dichotomized into 12 years old and younger (0) and age 13 and older (1) to represent early initiation. Victimization index. After examining relevant variables independently, a multidimensional index of the number of different types of victimization experiences was created. Items included in the index were: sexual abuse by an adult, sexual assault by another youth, sexual exploitation (indicator variable, as described), physical abuse, witnessing family abuse, and/or having been attacked or assaulted in the previous 12 months. These experiences were summed to create an index of victimization, with higher scores reflecting greater levels of victimization. Victim of relational aggression. Responses to survey questions asking about the number of times youth had experienced specific types of non-physical aggression in the past year were first dichotomized (0 = never; 1 = ever) and then summed to create a composite variable. Items reflected the number of times during the past year that someone: (a) made direct statements that led the youth to ―feel bad or extremely uncomfortable‖ (b) purposely socially excluded them, and/or (c) threatened to hurt them (―but did not actually hurt‖ them).  72 Individual Risk Factors Physical risk-taking. Responses to two questions reflective of objective physical risk-taking (i.e., sharing drug gear, sharing other equipment) were summed to create a combined physical risk-taking variable. It is impossible to ascertain the underlying motivation, for example, whether they are merely representative of context (for example, when physically on the street, gear may be more likely to be shared because of access issues), or whether they are representative of lack of self-care, lack of awareness of risks, disregard for perceived risks, or actual self-destructiveness. A composite variable (no risk behaviours, one risk behaviour, both risk behaviours) was created to reflect engagement in these risk behaviours. Past-year fighting. One question was used to indicate how many fights, if any, the adolescents were involved in for the previous 12 months. This item was dichotomized to reflect no fighting (0) and ever having been involved in a fight (1). Disordered eating. Two questions representing disordered eating were examined, one asking about frequency of binge-eating and the other about intentional vomiting. Each variable was examined independently for its relation to self-injury, and in combination. After it was revealed to have an association with self-injury, the question about vomiting was then dichotomized to indicate never (0) or ever (1) having engaged in this behaviour. Substance use. Given the high rate of substance use typically found among streetinvolved youth, substance use was examined in different ways to determine the most sensitive measure: (a) age of initiation to alcohol and marijuana was examined with the hypothesis that earlier initiation to both substances may confer higher risk, (b) binge-drinking was examined through the creation of a dichotomous variable that indicated whether the youth had engaged in binge drinking (typically defined in youth health surveys as five or more drinks in a row in a couple of hours) in the previous 30 days three or more times (Smith et al., 2007), (c) lifetime injection drug use was examined as one of the possible responses to the survey‘s substance use matrix, and (d) problem substance use was examined through a  73 series of questions in which participants were asked to indicate which of 13 problems or consequences, if any, had happened to them in the past year because of drinking alcohol or using drugs. Consequences that could be endorsed included: passed out, injured, car accident, trouble at school, poor marks, argued with family, physical fight, damaged property, lost friends, trouble with the police, break-up with boy/girlfriend, drug/alcohol treatment, and unwanted sex. Adolescents could also indicate ―I did not use alcohol.‖ An index of problems, consequences of substance use, was created from the 13 items and considered to be a proxy of the severity of substance use, with higher scores reflecting more problematic substance use. This index was thought to have the potential to be more useful given a majority of streetinvolved youth in the sample engaged in some degree of substance use. Suicidality. A single item on suicide attempt was dichotomized to indicate whether the youth had ever attempted suicide (1) or not (0). Contextual Protective Factors Family connectedness. An adolescent‘s overall feeling of connectedness to family was measured by a brief 11-item scale derived from the National Longitudinal Study of Adolescent Health (Resnick et al., 1997). For the purposes of the MCS study, the term family is considered to represent anyone the youth perceive to be their ―mother‖ and ―father‖ and the survey explicitly states that this could be biological or adoptive parents or other relatives or caregivers. Specifically, this scale measures the extent to which a youth feels (a) close to his or her mother, (b) close to his or her father, (c) his or her mother cares for him/her, (d) his or her father cares for him/her, (e) his/her mother is warm and loving, (f) satisfied with the relationship with his or her mother, (g) his or her father is warm and loving, (h) satisfied with the relationship with his or her father, (i) people in his or her family understand him/her, (j) his/her family has fun together, and (k) his/her family pays attention to him/her. The items were on a scale ranging from 1 to 3. Consistent with McCreary‘s protocol, scale scores were calculated as averages for those who completed a minimum of three of the eleven items (i.e.,  74 lower threshold to include youth who may not have a relationship with a parent/parents and/or family; see Saewyc & Homma, 2010). Research with this scale has found an inverse association between family connectedness and risk behaviours (Resnick et al., 1997; Saewyc & Homma, 2010). The current study found a Cronbach`s alpha for the total scale of rα = .87, consistent with previous evidence of internal consistency in this age group, with a Cronbach‘s alpha of .85 in a sample of BC young offenders (Viljoen et al., 2005) and .86 to .88 in population-based school samples (MacKay, 2007; Saewyc & Homma, 2010). In the core validation sample, Sieving et al. (2001) reported reliability r = .83 (for overall sample) that remained high across genders, grades, and ethnic groups. Principal components analysis (PCA) results with the family connectedness scale supported a one-factor or single component solution that explained 43.7% of the overall variance. As can be seen in Table 3, all item showed loadings on the factor at .61 to .72. Given that these findings are consistent with previous research with this scale (e.g., MacKay, 2007; Saewyc & Homma, 2010; Sieving et al., 2001), generate acceptable reliability, and items are moderately to highly intercorrelated with each other, use of the 11-item scale was supported. Table 3 One-Factor Solution for Family Connectedness Scale Item Father warm and loving Satisfied with relationship with father Close to father Father cares Mother warm and loving Satisfied with relationship with mother Close to mother Mother cares Family understands Family attention Family has fun together  Factor .643 .610 .651 .632 .718 .681 .664 .642 .667 .688 .671  75 School connectedness. The school connectedness scale, derived from the National Longitudinal Study of Adolescent Health (Resnick et al., 1997), consists of seven items that assess youths‘ connectedness to their schools (either in the present or retrospectively, depending on whether youth were attending school at the time of the survey). Specifically, previous studies have found this scale to measure both perceived teacher support and school belonging (MacKay, 2007; McNeely & Falci, 2004; McNeely, Nonnemaker, & Blum, 2002), by asking whether adolescents feel that their teachers care about them, feel that they are part of their schools, are happy to be at their schools, think that teachers at their schools treat students fairly, feel safe at their schools, and how often they have trouble getting along with their teachers, and have trouble getting along with other students. Youth were asked to rate how often they agreed with the statement (from ―never‖ to ―every day‖, from ―strongly agree‖ to ―strongly disagree‖, and from ―not at all‖ to ―very much‖, depending on the item). A study of B.C. young offenders found acceptable levels of internal consistency reliability (r = .73; Viljoen et al., 2005) and evidence from the original validation sample revealed a Cronbach‘s alpha of .75 that was stable across gender, grades, and ethnicity (Sieving et al., 2001). Recent research with this scale in a population-based school sample in B.C. showed that the final item on relations with other students (―trouble getting along with other students‖) did not effectively contribute to the scale (MacKay, 2007); therefore this item was dropped from the overall school connectedness scale in subsequent MCS research, including in the current study (Saewyc & Homma, 2010). In examining the six-item version of the scale, responses to five of the six items were reverse coded, with higher scores reflecting greater connectedness. Items were then averaged to create a final score of school connectedness. Scale scores were computed for those who responded to at least 5 of the 6 items (> 75%; Bearinger et al., 2005). The six-item school connectedness scale demonstrated acceptable internal consistency reliability, full scale: rα = .85. Investigation revealed, however, that reliability  76 could be increased to .87 by deleting the item, ―How often do/did you have trouble getting along with your teachers?‖ To investigate the factor structure (i.e., dimensionality) with this street-involved adolescent sample and to further justify the use of the items as a scale, a principal components analysis (PCA) was conducted forcing two factors and using varimax rotation. Although the scale has been validated in previous use, examination of its properties with high risk youth such as street-involved youth is new and therefore PCA represented an exploratory step. The bivariate zero-order correlations of the six items are presented in Table 4. Correlation between the items are generally moderate to high (.47 to .75), with the exception of the relatively lower correlations (.26 to .32) observed for the ―trouble getting along with teachers‖ item. Table 4 Zero-order Correlations for School Connectedness Scale School Connectedness Items 1. Teachers care 2. Feel like part of school 3. Happy to be at school 4. Teachers treat students fairly 5. Feel/felt safe at school 6. Trouble getting along with teachers  1 --.467 .507 .511 .470 .315  2  3  4  5  6  --.748 .571 .602 .274  --.596 .652 .295  --.608 .312  --.256  ---  Factor loadings did not provide evidence of the two subscales supported in previous research (e.g., MacKay, 2007; McNeely et al., 2002; Saewyc & Homma, 2010), school belonging and perceived teacher support (see Table 5). Instead, items loaded moderately to highly on Factor 1 (loadings .62 to .87, representing both teacher items and belonging items) with the exception of ―trouble getting along with teachers‖ which loaded by itself on a second factor. A second PCA was conducted to assess how the scale performed when one factor was forced (compared to the two-factor model). Results showed that again the first 5 items loaded highly (.71 to .86), whereas the relationship with teacher item did not (.47). Results were consistent even when filtering out only those youth who reported attending school at the  77 time of data collection (62.2% of the sample) and when examining by gender. Results provided further support for a single-factor 5-item version of the scale that did not include the ―trouble getting along with teachers‖ item. Table 5 Varimax Rotation of Two-Factor Solution for School Connectedness Scale Item Teachers care Feel like part of school Happy to be at school Teachers treat students fairly Feel/felt safe at school Trouble getting along with teachers  Rotated Components 1 2 .624 .371 .846 ..107 .868 .138 .771 ..237 .828 .101 .147 .966  Taken together, results from the PCA and from the reliability assessment suggested that the school connectedness scale behaved differently with the current street-involved sample when compared to extant findings with the ―general‖ (or normative) school-based adolescent population. Further, these results indicated the need for further investigation of dimensionality with the five-item version of the scale using confirmatory factor analysis (CFA) with Lisrel 8.80 (Jöreskog & Sörbom, 2006). CFA is a type of structural equation modeling concerned with measurement models that is used to assess the relationship between observable or measurable indicators (i.e., questionnaire items, predictors) and latent variables (e.g., hypothetical constructs such as school connectedness). CFA involves specifying the elements of the model a priori and then assessing (or confirming) whether the model fits the data adequately using various fit indicators (Brown & Cudeck, 1993; Hu & Bentler, 1999). The covariance matrices of the individual five items were analyzed using maximum likelihood estimation (MLE) and weighted least squares (WLS), as suggested for ordinal variables (Jöreskog & Sörbom, 2006). Various fit indices have been developed and are used to evaluate the goodness of fit between the hypothesized model and the data. The χ2 statistic is a common means of  78 evaluating fit, but is considered a biased index in the context of large samples (i.e., N > 200) and therefore it will not be reported here (Marsh, Balla, & McDonald, 1988). Researchers have recommended consulting multiple indices with different measurement properties to evaluate model fit (Hu & Bentler, 1999; Marsh, Hau, & Wen, 2004). Jackson and colleagues (2009) also suggest the use of fit measures that are robust (i.e., not dependent on sample size) and outperform others in detecting model misspecifications. As recommended by Hu and Bentler (1999), the Root Mean Square Error of Approximation (RMSEA), the Standardized Root Mean Square Residual (SRMR), the NonNormed Fit Index (NNFI), and the Comparative Fit Index (CFI) were used and examined with respect to recommended cut-offs (i.e., cut-off criteria for fit: RMSEA and SRMR < .08; NNFI and CFI > .95). The CFI is an incremental fit index that compares the lack of fit of a proposed model with the lack of fit of a null model to reveal the relative improvement in fit (Hu & Bentler, 1999). The NNFI, also referred to as the Tucker-Lewis Index (TLI), is a comparative fit index that is similar to the CFI but adjusts for parsimony (i.e., indexes discrepancy through degrees of freedom; Marsh et al., 1988). The SRMR, a residual-based fix index, is a measure of the standardized difference between the observed and predicted model covariances, with a value of zero indicating perfect fit (Hu & Bentler, 1999). The RMSEA estimates the degree to which the model reasonably fits in the population by evaluating the difference between the implied covariance matrix and the population covariance matrix (Brown & Cudeck, 1993). Results indicated that a one-factor (ordinal) model including five items demonstrated reasonable model fit to the data (n = 677, RMSEA = .065, 90% CI = .036, .097, SRMR = .039, CFI - .992, NNFI - .983). The RMSEA value falls within the .08 fit criteria for adequate fit, but is above the more recently recommended .06 criteria for excellence (Brown & Cudeck, 1993; Hu & Bentler, 1999) whereas the comparative incremental fit indices, NNFI and CFI, show good fit as they are above the recommended 0.95 criteria (Hu & Bentler, 1999). The SRMR is below .08 and is therefore considered to be an adequate fit (Hu & Bentler, 1999; Jackson et  79 al., 2009). All indicators also showed acceptable association with the latent construct of school connectedness (Lambda-X λx .67 to .91). In sum, in contrast to previous findings with the school connectedness scale in normative samples, psychometric analyses with the current study sample revealed a singlefactor five-item version was the most parsimonious fit with the data both in terms of reliability and validity. The sixth item, a question relating to adolescents‘ reports of trouble getting along with their teachers, was therefore eliminated from the scale and the five-item version was used for all subsequent analyses. Although previous research has supported its inclusion, it is clear from direct examination that whereas other items refer to perceptions or to feelings, this item refers specifically to behaviour, and has a different response format. Extracurricular community activities. Youth were asked to report on their participation in extracurricular activities prior to becoming involved in the street (―Before you were street-involved, did you do any of the following?‖). Questions related to both structured (and more often adult-led) and unstructured activities. For the current purposes, only structured activities that occurred outside of school were included: (a) sports with a coach or instructor, (b) dance or aerobics classes/lessons, (c) art/drama/singing/music lessons, clubs, or groups, and (d) club or group membership. Participation was rated on a dichotomous scale (0 = no, 1 = yes) for each activity. The distribution of each activity was examined against NSSI, and then combined and dichotomized so that youth who reported participating in any of the four categories were coded as ―yes‖ (1) and those who reported no participation were coded as 0. The original items were derived from the National Longitudinal Survey of Children and Youth (Human Resources Development Canada and Statistics Canada). Involvement in extreme sports. Adolescents were asked whether they had engaged in six different ―extreme‖ athletic activities prior to street-involvement (e.g., rock climbing, cliff and bridge jumping). After each of the activities was examined for its relation to  80 self-injury, a dichotomous variable reflecting any participation in extreme sports was created, with 0 = no participation, and 1 = any participation. Peer relations (less trouble getting along). One item taps how frequently the youth have had trouble getting along with their peers in the school environment (i.e., How often do you have trouble getting along with other students?). Although the question asks about trouble getting along, the variable was hypothesized to be a protective factor, with a higher score indicating less trouble getting along (i.e., better relations) with peers. Given that frequent trouble getting along with peers represents a chronic stressor, youth who reported getting along with their peers may possess greater social competence or be more skilled in managing social relationships in the school setting. Respondents indicated the frequency with which they had such troubles, from ―never‖ to ―daily.‖ Responses were recoded to indicate ―almost every day or daily,‖ ―about once a week,‖ or ―never/just a few times‖, with higher scores reflecting better peer relations (or less trouble getting along). For the one-third of students not attending school at the time of survey, response to this item would have been retrospective. Current school attendance. One item asked the youth if they were attending school at the time of the survey (yes = 1, no = 0). Individual Protective Factors Low emotional distress (emotional health). This scale, derived from the Adolescent Health Survey (University of Minnesota), consists of four items reflecting the youths‘ reports of their emotional distress across four dimensions in the 30 days prior to completing the survey. The youth reported whether they had been feeling: (a) any strain, stress or pressure, (b) any illness, physical problems, or health-related fears, (c) nervousness or ―nerves‖ (i.e., anxiety), and/or (d) so sad, discouraged, hopeless, or overwhelmed that they ―wondered if anything was worthwhile.‖ Items were summed and averaged for all youth who  81 responded to at least three of the four items. Higher scores indicate lower emotional distress (and therefore greater emotional health) and low scores were conceptualized as indicating higher emotional distress (and worse emotional health). Cronbach‘s alpha for this scale was acceptable, rα = .84. This psychometrics of this scale have not been investigated in samples with street-involved or high-risk adolescents. Principal components analysis results showed a one-factor solution, with all items loading on the single factor and explaining 67% of the variance. There was no cross-loading and all items demonstrated high loadings of .80 or greater (see Table 6). All items showed moderate intercorrelations as well (.51 to .61). Thus, the investigation of psychometric properties supported the use of the low emotional distress scale in this study. Table 6 Varimax Rotation of One-Factor Solution for Low Emotional Distress Scale Item Bothered by nerves Under strain, stress, or pressure Bothered by physical health, illness, or related fears Wondered if anything was worthwhile  Component 1 .847 .820 .809 .801  Body satisfaction. This composite represents two questions that reflect how an adolescent perceives his/her physical body in terms of body image (i.e., perceived shape or size) and whether s/he has any plans to change body weight (body weight plans). Each item was examined separately for its relation to self-injury. Positive future outlook. This variable reflects thoughts of youths‘ personal futures in terms of how they saw themselves in five years. Options included both positively valenced responses (e.g., in a job, in school, having a home, having a family) indicating a rosier or more optimistic outlook, and negatively valenced responses (e.g., dead, on the streets, in prison). Youth could also select ―I don‘t know.‖ The responses were recoded to reflect either a positive prediction (1) or a negative prediction (0) for themselves.  82 Educational aspirations. Youths‘ educational aspirations were examined and categorized to indicate no plans for future education, plans to finish high-school, and plans for further education beyond high-school. Feelings toward current circumstances. Youths‘ feelings toward their current circumstances was examined with one forced choice item and recoded to dichotomously indicate a more positive feeling (good/fair = 1) or more negative feeling (poor/awful = 0) toward current life circumstances. Subjective (self-perceived) health status. Youth were asked to rate their health status on a single-item with a 4-point scale ranging from poor to excellent. Responses were recoded to reflect dichotomy, excellent/good (1) and fair/poor (0), with the former category representing better perceived health (and therefore hypothesized to be protective). Data Management Issues Missing data In addition to questions that were clearly not responded to, other response categories were derived by MCS in order to denote missing data. For example, multiple responses to the same question (when only one response was required), ―don‘t know‖ or ―not sure‖ were coded as missing for the purpose of analyses (unless stated otherwise). Frequencies were examined for missing data for each variable. Responses across variables (e.g., gender, age, self-injury) were examined for statistically significant bias using a series of ANOVAs. In survey research, missing data can be the result of administration conditions, survey topics, or participant characteristics. Notably, response rates for items remained consistently high throughout the survey and rates of missing data were acceptably low (0% to 6.7%, with most in 0-3% range). An exception to these low rates of missingness is in the area of family connectedness; some items were not applicable to youth (―does not apply‖), for example, youth who did not have contact with either/both parent(s) and/or with other family members.  83 The nonlinear model that is part of the logistic regression assumptions requires a full set of data. The statistical program (SPSS) used for data analysis in the current study utilizes listwise deletion with missing data so that the remaining full dataset can then be used to calculate logistic parameters. Sampling Issues A minimum of 10 cases per variable is required to conduct multivariate logistic regressions (Hosmer & Lemeshow, 2000; Peduzzi et al., 1996). This has been debated and some researchers estimate that between 10-15 cases to the number of predictor variables is most reasonable, with attention to the smaller frequency of the dependent variable (Hosmer & Lemeshow, 2000). Whereas linear regression uses ordinary least squares (OLS) estimation, logistic regression uses maximum likelihood estimation (MLE) to derive parameters. If there are too few cases, the accuracy and precision of the regression coefficients is compromised. This large dataset provided sufficient power to address sampling; despite a considerable number of variables, the self-injury outcome is not rare and models are not subject to the problems of over- or under-fitting. Data Analytic Plan Data analysis aligned with the major study objectives in terms of: (a) describing the prevalence, demography, and reasons for self-injury in this sample of street-involved adolescents (b) assessing the independent bivariate and multivariate relationships between and among the hypothesized risk and protective factors, and (c) predicting the likelihood of self-injury as an outcome based on combinations of risk and protective factors. For second and third objectives stated, data analysis followed four distinct steps (see Figure 1). The first step in analysis focused on variable selection for subsequent analyses (Erickson, 1999; Hosmer & Lemeshow, 2000). Variables hypothesized to either confer risk or to provide protection against self-injury were selected from the SYHS to be further examined  84 first as individual items and, when theoretically and statistically feasible and meaningful, in combination (i.e., scales and indices or composites, as aforementioned). Variable distributions were examined through descriptive analyses (e.g., frequencies, M, SD, skew, kurtosis, SE). Bivariate analyses were used to compare the descriptive characteristics of youth who reported self-injurious behaviour (1 or p) to those who reported no self-injury (0 or 1-p) through crosstabulations (χ2 for categorical variables) and independent t-tests (for continuous variables), conducted separately for males and females. The contingency tables generated by crosstabulations inform whether the null hypothesis, that there is no relationship or no statistically significant difference between groups in frequency distributions, can be rejected; however, it is only by comparing the adjusted standardized residuals to the critical value that we can determine which cells were contributing the most to any differences revealed by the chi-square value. Variables that showed statistical significance in the first step, either significant chisquare and adjusted standardized residual of > 1.96 or significant t-test result (both using significance level of p < .05), were subjected to binary logistic regression analyses, conducted separately by gender. Analyses included age as a covariate. Logistic regression allows the most parsimonious prediction of a discrete outcome such as group membership for a variable set that may be dichotomous, categorical, and/or continuous. Importantly, compared to linear regression, logistic regression is relatively unrestricted as it does not assume linearity between independent and dependent variables, does not require normally distributed variables, does not require independent variables to be interval or unbounded, does not assume homescedasticity, and does not require random sampling as maximum likelihood reduces sampling error (Hosmer & Lemeshow, 2000; Tabachnick and Fidell, 2001). However, binary logistic regression assumptions include the following: the analysis requires a binary dependent variable (self-injury vs. no self-injury); only meaningful variables be included for accurate model fit; each observation must be independent; there is an absence of high  85 multicollinearity; sampling adequacy (all cells must be ≥ 1 and no more than 20% can be < 5 or the logistic model may be unstable); and a large sample size (or ratio of cases per independent variable as aforementioned) is required. Prior to logistic regression analyses, all variables were standardized to reflect a range of 0 to 1 in order to permit the odds ratios (OR) generated to represent the odds of reporting NSSI for those at the higher end of the variable/scale when compared to those at the lower end. For dichotomous and categorical predictors, the reference category of ―never‖ was chosen in order that other levels of the variable could be compared to determine whether they generated more, less, or equal likelihood for self-injury. Odds ratios are measures of association estimated directly from a logistic model and can range from zero to infinity. As a multiplicative measure of risk (or protection), OR report how much more likely it is (i.e., the odds) that an observation is a member of a target or reference group (in the present study, youth who self-injure) versus a member of the other group (youth who do not self-injure). OR less than 1.0 show a decrease in odds and those greater than 1.0 show an increase in odds. In step 2, all variables that achieved a statistical significance level of p < .05 and with an odds ratio of 0.5 or less for a protective factor and 2.0 or greater for a risk factor (i.e., representing clinical significance by the doubling or halving of the likelihood of occurrence of self-injury; see Bearinger et al., 2005; Poon, Chittenden, Saewyc, Murphy, The McCreary Centre Society, 2006) were considered for entry into multivariate logistic regression models. Multivariate logistic regression is an ideal analysis for providing insight into complex variable interrelationships when there are multiple independent variables and a dichotomous outcome (binomial or binary). This analysis allowed factors to be controlled for while measuring the unique independent effect of each factor on self-injury outcome. Before proceeding with the multivariate analyses, to assess multicollinearity (i.e., strong intercorrelations between independent variables), bivariate zero-order and partial  86 correlations (controlling for age) between all study variables considered for entry into multivariate models were examined separately by gender through an intercorrelation matrix. Collinear relations have been found to degrade the logistic regression parameters and can lead to incorrect conclusions about relationships between variables. Given that bivariate correlations do not assess interdependencies among variables, Variance Inflation Factor (VIF) and Tolerance, statistics derived from linear regression, were examined as well for their more sophisticated ability to detect multicollinearity (Hosmer & Lemeshow, 2000). Next, separate multivariate logistic regression analyses were conducted for risk and for protective factors and for males and for females, leading to two models for each gender (i.e., one risk only and one protective only). Multivariate analyses included age as a covariate. The overall model chi-square (also called the likelihood ratio chi-square test and similar to the global F statistic in ordinary least squares regression) described the degree of association between the variables entered as a group and the self-injury outcome, with the Wald chi-square associated with each variable then describing the degree of independent association between the variable and the outcome, with the other variables controlled. ORs (and 95% confidence intervals) were computed for each variable entered into the model. The third step featured multivariate logistic regression models combining statistically and clinically significant factors from the separate risk-only and protection-only models into one model (the top three of each; see Bearinger et al., 2005; Erickson, 1999; Poon et al., 2006), conducted separately for males and for females, with age as a covariate. These analyses provided the beta weights needed to form probability profiles (Bearinger et al., 2005; Erickson, 1999; Poon et al., 2006; Rubenstein et al., 1989). The final step featured probability profiling in which the estimated likelihood of the selfinjury outcome was computed with various combinations of the strongest risk and protective factors that emerged from the multivariate logistic regression models for each gender, with a  87 focus on including those that were empirically most salient, meaningful, clinically relevant, and amenable to intervention (Bearinger et al., 2005; Bearinger et al., 2010). Logistic models yielded a matrix of the likelihood or single probability of self-injury as the number of risk and protective factors was manipulated, with different combinations of risk factors (i.e., 0, 1, 2, or 3) and protective factors (i.e., 0, 1, 2, or 3) present. Unlike limitless OR, probability is represented as a number between 0 and 1 (Erickson, 1999; Poon et al., 2006). Probabilities for scale or composite variables were computed using scores representing the 10th and 90th percentiles for low and high levels of the factor, respectively (Bearinger et al., 2005; Borowsky et al., 2001; Borowsky et al., 1999; Rubenstein et al., 1989). The equation used to generate the probabilities is as follows: p=  1 (1 + exp (-bx)  As a methodology, probability profiling has a history in the public health field where it has most notably been applied to adolescent suicide and also to violence perpetration in both the general population and with American Indian youth samples (see Bearinger et al., 2005; Bearinger et al., 2010; Borowsky et al., 1999; Borowsky et al., 2001; Erickson, 1999; Poon et al., 2006; Resnick et al., 2004; Rubenstein et al., 1989). Probability profiles are useful for determining focus and priorities for developing a framework for prevention and intervention. As a statistical technique, probability profiling is unique in its practical emphasis on examining the interplay among variable combinations to provide what are essentially different profiles for each combination, offering up an estimate of the single probability of the likelihood of an outcome (for current purposes, self-injury, but can be used with any outcome) for an individual (adolescent) with that particular portrait or profile. By allowing a simultaneous focus on both risk and protection, this methodology can enhance understanding of factors that predict and protect against self-injury as well as mapping onto the critical ―dual strategies‖ of risk reduction and health or competence promotion (Bearinger et al., 2010, p. 48).  88  STEP 1  STEP 2  STEP 3  STEP 4  Figure 1: Overview of Data Analytic Plan  IDENTIFYING VARIABLES: VARIABLE SELECTION •PART 1: Descriptive statistics comparing SI vs. No-SI (independent t-tests and crosstabulations) by gender •PART 2: Bivariate logistic regression, examining odds ratios for all significant risk and protective factors from Part 1, separately by gender, age as covariate  IDENTIFYING COMBINATIONS OF VARIABLES •PART 1: Assess multicollinearity •PART 2: Multivariate logistic regression, separate models by gender, to identify top 2-3 factors , age as covariate •Risk factor only model •Protective factor only model  IDENTIFYING MOST SALIENT COMBINATIONS OF VARIABLES •Multivariate logistic regression by gender, top risk factors, top protective factors in combined model  GENERATING PROBABILITY PROFILES •Using beta weights derived from combined models to produce profiles of the likelihood of SI as outcome given different combinations of risk and protective factors by gender  89 CHAPTER 4: RESULTS Characteristics of the overall sample were previously described in the Method section and a more comprehensive detailing of the sample is presented in The McCreary Centre Society‘s research report, Against the odds: A profile of marginalized and street-involved youth in B.C (Smith et al., 2007). Therefore, the present chapter provides a focused overview of the results in alignment with the research study objectives. First revealed is a portrait of the youth who reported NSSI, with demographic and self-injury characteristics detailed. Next, findings from bivariate analyses are presented separately by gender to highlight association between potential individual and contextual risk and protective factors and the self-injury outcome. These findings formed the basis for the explicit selection of variables utilized in the multivariate logistic regression analyses. The subsequent section reports the multivariate results for combinations of risk and protective factors, with a focus on identifying a clinically and statistically salient set of modifiable factors. The final analyses report the profiles derived from probability profiling, the methodology used to predict the probability of self-injury as an outcome given various combinations of risk and protective factors (Bearinger et al., 2010; Borowsky et al., 2001; Erickson, 1999; Poon et al., 2006). Notably, the term predict is used statistically as the data are cross-sectional and therefore preclude the possibility of real prediction. Descriptive Findings Characteristics of the Self-Injury Sub-Sample Demographics. A total of 312 youth (45%) reported ever engaging in NSSI, with significantly more females than males doing so, 55.5% versus 33.6% respectively, 2 (1) = 26.63, p < .001. Youth ranged in age from 12 through 18 (M = 16.37, SD = 1.53), with increasing numbers of youth engaging in NSSI at every age, t (310) = 188.78, p < .001. Significant differences were noted between sexual orientation categories within the self-injury  90 sample, 2 (3) = 106.53, p < .001, with nearly 44%% of males and 70% of females identifying as a member of a sexual minority (i.e., nonexclusively heterosexual orientation). Overall, nearly 36% of the heterosexual youth reported engaging in NSSI, whereas 70% of ―mostly heterosexual‖ youth, 68% of bisexual youth, and 55% of homosexual and ―mostly homosexual‖ youth reported this behaviour. Ethnicity was explored in terms of the two primary (and overlapping) groups of self-injuring youth – Aboriginal (52%) and European (51%)– and no significant differences in frequency were noted, 2 (1) = 0.32, p = .570 and 2 (1) = 0.05, p = .820, respectively. Similar proportions of youth reporting Aboriginal or European ethnicity also reported NSSI (44% and 47%). Self-injury frequency and reasons. Of those youth reporting any history of selfinjury, a majority (59.9%) reported three or more incidents, whereas 40.1% reported engaging in this behaviour once or twice, 2 (1) = 12.32, p < .001. More self-injuring females than males reported engaging in self-injurious behaviour three or more times (66.2% females vs. 48.2% males), 2 (1) = 9.57, p = .002. A total of 24.5% youth reported seeking medical attention within the past year due to self-injurious behaviour. Youth were asked to indicate their reasons for their last incident of NSSI (see Table 7). Self-injuring youth reported varying numbers of reasons for engaging in the behaviour, with 8.9% of youth reporting no reasons, 33.7% reporting one reason, 26.4% reporting 2-3 reasons, and 31.0% reporting 4-9 reasons. Males reported 0-7 reasons (M = 1.64, SD = 1.59), with nearly half reporting one reason. In contrast, females reported significantly more reasons, range 0-9 (M = 3.06, SD = 1.92), t (300) = -6.49, p < .001. Two-thirds of the young women who reported nonsuicidal self-injury indicated three or more reasons why they did so at last incident. The following reasons for self-injuring were reported by youth: 50% lonely/depressed, 46% stressed, 46% angry, 27% rejected, 23% problems with alcohol and/or drugs, 22%  91 bored, 15% to feel in control, and 9% to get attention. Further, 19% reported other reasons, including: interpersonal problems (n = 6), loss or grief (n = 5), to externalize emotional pain (n = 5), to feel good (n = 4), to feel pain (n = 3), social context (e.g., ―carving in group home‖; n = 3), for physiological ―high‖ (n = 3), desire to die (n = 3), to feel alive (n = 2), and abuse/rape (n = 2). Examples of single case responses include: ―sports,‖ ―long storey‖ (sic), ―life,‖ ―problems of the past,‖ ―don‘t talk about it,‖ ―problems at school,‖ ―love for blood,‖ and ―feeling crazy.‖ Males and females were similar on their top reasons—feeling lonely/depressed, angry and/or stressed, however males also equally reported boredom. Reasoning among youth who reported three or more NSSI incidents were consistent with these findings. Further, compared with youth reporting 1-2 incidents, those with three or more were more likely to report NSSI to feel in control (6% vs. 21%, 2 (1) = 13.41, p < .001). The fourth most common reason for females and fifth for males, feeling rejected, can serve both emotion regulation functions (either overwhelming negative affect or numbing) and interpersonal functions (communication, social influence). More youth who reported at least three incidents reported feeling rejected (16% vs. 35%, 2 (1) = 13.66, p < .001) Table 7 Reasons Reported for Last Incident of Self-Injury by Gender Reason (last incident)  Boys %  Lonely/depressed Bored Angry Stressed Problems with drugs and/or alcohol Rejected To feel in control To get attention “Other”  25.0 25.0 29.6 23.1 15.7 13.9 9.3 8.3 14.8  Note. Males n = 108, Females n = 194.  Rank 2 2 1 3 4 6 7 8 5  Girls % 62.9 20.1 54.1 58.2 27.8 34.0 18.0 8.8 21.6  Rank 1 7 3 2 5 4 8 9 6  2 39.84 0.97 16.79 0.02 4.22 14.33 34.44 5.64 2.09  p < .001 .324 < .001 .899 .040 < .001 < .001 .020 .148  92 Background characteristics. The self-injuring adolescents reported significant trauma histories, with half reporting a history of sexual abuse, three-quarters reporting physical abuse, and 72% witnessing abuse within their family. Further victimization continued on the streets wherein half the youth reported being physically attacked/assaulted in the past year (27% two or more times), one-third reported having been sexually exploited, and 27% were sexually coerced by another youth. Close to three-quarters of the youth reported having been a victim of relational/verbal aggression during the previous year. Many youth reported families with multiple challenges including alcohol and substance abuse, mental illness, criminal involvement, and suicide. Half had experience with the government care system with either foster and/or group home placement. These adolescents also reported significant exposure to the street, with close to two-thirds reporting street-involvement by age 13 and nearly half reported living in unstable and/or unsafe conditions. Although over 60% of the adolescents reported attending school at the time of the survey, only 38% ―felt like part of their school‖, 44% ―agreed‖ or ―strongly agreed‖ that they felt happy to be at school, 51% agreed that they felt safe at school, and 57% believed that teachers treated students fairly. More than half of the youth believed that teachers generally cared about them (57% selected at least ―somewhat‖), although 56% reported having considerable trouble getting along with teachers. More than half of the adolescents reported getting along with other students on a regular basis while the others had consistent difficulties. Nearly 30% aspired to completing post-secondary education, and 21% to graduating high-school. The adolescents also reported engaging in a constellation of health-compromising or risk behaviours. Close to half of the youth endorsed sharing sharp tools (e.g., needles, razors, drug gear) and engaging in regular binge-drinking, and considerable number reported disordered eating practices and injection drug use (25.3% and 25.7%, respectively). More than 50% of the youth had tried both alcohol and marijuana by age 12 and many reported  93 experiencing multiple consequences to their substance use. Overlap between suicide and nonsuicidal self-injury was evident, with close to half (41%) of the self-injuring youth acknowledging a past year suicide attempt and 80% of youth who reported a past-year suicide attempt also self-injured. Just over one-quarter (26%) of the injuring youth reported two or more suicide attempts. Nevertheless, the youth reported feeling positive about some aspects of their lives. Close to half of the youth rated their health as ―excellent‖ or ―good‖ and expressed satisfaction with body image and weight practices, and a majority rated their life circumstances as ―fair‖ (42%) or ―good‖ (35%). In the area of emotional health or well-being, about one-third of the adolescents reported feeling only ―a little‖ or the ―usual amount‖ of stress or pressure, close to 40% reported not being bothered by nerves/nervousness, and approximately another third reported feeling sad, discouraged, and hopeless ―not at all‖ or only ―a little.‖ Many youth reported participation in structured community extracurricular activities (e.g., dance, art, drama, clubs, coached sports) and extreme sports prior to streetinvolvement. Distribution of Potential Risk and Protective Factors Against Self-Injury As the first step in identifying potential risk and protective factors for NSSI operating in these youths‘ lives, the distribution of hypothesized factors or variables was examined against this outcome. Findings are presented in summary tables separately by gender (see Tables 8 and 9). Notably, only variables found to exhibit significance at this descriptive level were subjected to logistic regression analyses in the subsequent step. Analyses revealed no difference in age between self-injuring and non-self-injuring youth, t (682) = 0.17, p = .869. Chi-square analyses revealed significant difference between sexual orientation categories, χ2 (3) = 55.74, p < .001, with half of self-harming youth reporting sexual minority status, compared with nearly 23% of the non-injuring youth.  94 Table 8 Risk and Protective Factors Associated with Self-Injury for Females Potential Risk and Protective Factors Potential contextual risk factors Mother problems index Father problems index Victimization index Victim of relational aggression Sexually abused Physically abused Two or more perpetrators of sexual abuse Two or more perpetrators of physical abuse Witnessed abuse Sexually assaulted (coercion) by youth Sexual exploitation (any) Physically assaulted (past year) Ever been in-care (foster/group home) Family suicide history (attempt/completion) Street exposure - precarious housing (past year) Ever kicked out of home Street-involved ≤ 12 years of age Potential individual risk factor Physical risk taking No –Never (sharing needle/razor, etc.) Yes – One Yes – Both Any suicide attempt (past yr) Physical fighting (past yr) Vomiting on purpose after eating (past year) Binge-eating Early debut alcohol and marijuana ≤ 12 yrs old Binge-drinking 3+ days in past month Ever injected drugs Consequences of substance use Potential contextual protective factors Family connectedness School connectedness Participation in 1+ extracurricular activities (prior) Involvement in 1+ extreme sports (prior) Less trouble getting along with students Potential individual protective factors Low emotional distress Educational aspirations Before highschool Graduate highschool Graduate post-sec Currently attending school  SI (n = 201) M (SD) / %  No SI (n = 161) M (SD) / %  t / χ2  P  .24 (.29) .26 (.30) .54 (.26) .72 (.35) 65.0% 76.4% 23.9% 50.7% 76.0% 31.8% 29.1% 47.7% 43.8% 77.6% 44.8% 61.4% 39.5%  .13 (.23) .17 (.24) .39 (.27) .50 (.39) 51.9% 61.4% 8.1% 26.1% 56.1% 15.5% 22.0% 26.5% 35.4% 42.3% 28.6% 55.1% 32.4%  -4.04 -2.98 -5.33 -5.62 6.20 9.32 16.26 24.54 15.76 12.83 2.29 16.39 7.61 33.42 10.01 1.46 1.80  <.001 .003 <.001 <.001 .017 .002 <.001 <.001 <.001 <.001 .145 <.001 .022 <.001 .002 .235 .210  41.5% 41.5% 17.0% 47.2% 70.6% 26.9% 42.3% 51.1% 49.5% 23.5% .33 (.26)  80.7% 13.8% 5.5% 9.0% 43.1% 13.8% 32.9% 34.8% 32.1% 27.5% .17 (.24)  50.61  <.001  60.48 26.61 9.03 3.30 4.91 10.95 0.39 -6.71  <.001 <.001 .003 .078 .086 .001 .228 <.001  .49 (.25) .52 (.27) 74.0% 35.8% 72.2%  .60 (.25) .57 (.26) 62.5% 29.8% 81.9%  4.05 1.741 -5.20 1.46 4.59  <.001 .083 .023 .261 .034  .45 (.27)  .65 (.28)  6.72  <.001  14.8% 31.1% 54.1% 69.5%  15.3% 40.8% 43.9% 74.5%  2.59  .274  0.70  .342  95  Table 8 continued... Potential Risk and Protective Factors Self-reported health status (good/excellent) Body image – perceive self as just right Body weight satisfaction – no plan to change Future outlook (5-yr) – positive outlook Perception of life circumstances – good or fair  SI (n = 201) M (SD) / % 49.7% 47.5% 39.5% 80.1% 75.7%  No SI (n = 161) M (SD) / % 66.5% 46.5% 44.0% 89.4% 90.8%  t / χ2 10.00 0.03 0.75 4.43 13.20  P .002 .915 .391 .045 .< .001  As is evident from Table 8, whereas numerous contextual and individual risk factors were associated with self-injury, fewer hypothesized protective factors showed this relationship. Findings from examination of the association between participation in structured, supervised extracurricular activities prior to street-involvement and NSSI revealed that this variable acted more like a risk factor, with significantly more self-injuring females reporting participation compared to non-self-injuring females. Notably, other hypothesized protective factors, including school-related variables (school connectedness, attending school, educational aspirations) and health attitudes and behaviours (body image, body weight plans, binge-eating) were not found to significantly differ between the groups (selfinjury vs. no-self-injury) for females. Two hypothesized risk factors that were not found to distinguish between the two groups were sexual exploitation, an indicator variable consisting of any acknowledgement of having ever traded sex, and injection drug use. Compared to their non-injuring counterparts, females who reported self-injury were not significantly more likely to report experiences such as early street involvement (before age 13), being ―kicked out‖ of home, or early initiation to alcohol and marijuana.  96 Table 9 Risk and Protective Factors Associated With Self-Injury for Males Potential Risk and Protective Factors Potential contextual risk factors Mother problems index Father problems index Victimization index Victim of relational aggression Sexually abused Physically abused Two or more perpetrators of physical abuse Witnessed abuse Physically assaulted (past year) Sexual assault (coercion) by another youth Any sexual exploitation Ever been in-care (foster/group home) Family suicide history (attempt/complete) Precarious housing (past year) Ever kicked out of home Street-involved ≤ 12 years of age Potential individual risk factors Physical risk taking .) No –Never (sharing needle/razor, etc.) Yes – One Yes - Both Suicide attempt (past year) Physical fighting (past year) Vomiting on purpose after eating (past year) Binge-eating (past year) Early debut alcohol and marijuana ≤ 12 yrs old Binge-drinking 3+ days in past month Ever injected drugs Consequences of substance use Potential contextual protective factors Family connectedness School connectedness Participation in 1+ extracurricular activities (prior) Involvement in 1+ extreme sports (prior) Less trouble getting along with students Potential individual protective factors Low emotional distress Educational aspirations Before highschool Graduate highschool Graduate post-sec Currently attending school Self-reported health status (good/excellent) Body image – perceive self as just right Body weight satisfaction – no plan to change  SI (n = 110) M (SD) / %  No SI (n = 217) M (SD) / %  t / χ2  P  .19 (.27) .24 (.30) .44 (.24) .60 (.38) 24.8% 85.4% 34.5% 65.4% 52.8% 14.5% 39.8% 41.0% 55.1% 43.6% 64.5% 44.3%  .13 (.23) .14 (.24) .30 (.25) .37 (.37) 13.9% 44.4% 21.7% 50.7% 36.6% 8.3% 29.9% 38.7% 33.8% 33.2% 51.7% 44.1%  -1.92 -2.88 -4.82 -5.17 5.66 12.45 6.45 6.15 7.52 3.06 3.12 .34 10.70 3.44 4.66 .002  .056 .004 <.001 <.001 .026 .001 .040 .016 .006 .087 .101 .846 <.001 .069 .041 1.00  54.1% 29.6% 16.3% 29.2% 66.4% 21.8% 27.3% 48.5% 43.5% 30.0% .22 (.21 )  81.8% 15.1% 3.1% 7.7% 66.3% 10.7% 21.0% 44.9% 39.0% 30.0% .17 (.19 )  28.58  <.001  25.43 < .001 7.28 1.62 3.22 .62 .99 -2.21  <.001 1.00 .012 .212 .200 .471 1.00 .028  .51 (.25) .52 (.25) 50.5% 54.5% 62.0%  .63 (.24) .57 (.24) 52.2% 53.0% 76.9%  3.99 1.70 .08 .07 7.81  <.001 .091 .774 .815 .006  .54 (.28) 13.0% 44.4% 42.6% 47.3% 30.0% 30.0% 53.3%  .69 (.28) 14.5% 42.7% 42.7% 55.8% 48.1% 36.6% 43.3%  4.63 0.09  <.001 .956  2.13 10.17 1.41 2.88  .160 .002 .267 .100  97  Table 9 continued... Potential Risk and Protective Factors Future outlook (5-yr) – positive outlook Perception of life circumstances – good or fair  SI (n = 110) M (SD) / % 74.7% 80.2%  No SI (n = 217) M (SD) / % 82.6% 82.4%  t / χ2  P  2.24 0.23  .141 .640  Similar to analyses for females, findings from descriptive analyses with males supported the use of most hypothesized risk factors; exceptions included sexual exploitation, history of being in-care, street exposure variables (precarious housing, street-involvement before age 13), and individual level variables including early alcohol and marijuana use, binge-drinking, binge-eating, fighting, and injection drug use. Findings failed to support nine of the thirteen hypothesized protective factors, leaving family connectedness, better peer relations at school, better self-rated health status (good/excellent), and greater emotional health (lower levels of distress) as protective factors for the next step of analysis. Again, as with the females, school-related variables including school connectedness, attending school, and educational aspirations did not differ significantly between the two groups. Risk and Protective Factors Associated with Self-Injury Variables that were determined to be significant at the descriptive level were then utilized in a series of bivariate logistic regression analyses. Due to developmental differences across adolescence in risk and protective factors, all analyses included age as covariate (adjusting for age). Each variable was examined for its independent association with selfinjury in order to determine the strongest statistically (p < .05) and clinically significant risk (OR > 2.0) and protective factors (OR ≤ 0.05) for both males and females. Results from these analyses are presented in Tables 10 and 11, as adjusted odds ratios (AOR).  98 Table 10 Bivariate Logistic Regression Analyses for Females: Risk and Protective Factors for Self-Injury Variable  Femalea AOR  AOR 95%CI Lower Upper  p  Risk Factors Sexual minority membership  2.75  1.71  4.42  <.001  Sexual abuse  1.85  1.19  2.87  .007  Sexual assault (coercion) by another youth  2.55  1.51  4.31  <.001  Physical abuse  2.14  1.33  3.43  .002  Witnessed family abuse  2.59  1.63  4.12  <.001  Attacked or assaulted in past yr  2.58  1.63  4.08  <.001  Victimization index (0-1)  9.78  4.13  23.18  <.001  Sexually abused by ≥ 2 perpetrators  4.38  2.18  8.79  <.001  Physically abused by ≥ 2 perpetrators  3.21  2.01  5.13  <.001  Victim of relational aggression (0-1)  5.25  2.88  9.53  <.001  Street exposure – past year precarious housing  2.21  1.39  3.49  <.001  Ever been in-care (foster/group home)  1.41  0.92  2.17  .069  Mother problems index (0-1)  5.79  2.34  14.30  <.001  Father problems index (0-1)  3.37  1.48  7.67  .004  Family history of suicide  3.16  2.03  4.91  <.001  Consequences for substance use (0-1)  27.32  9.32  80.10  <.001  Physical risk-taking (sharing needles/tattoo tools) (0-1)  13.97  6.04  32.39  <.001  Physical fighting - past yr  3.28  2.09  5.14  <.001  Vomiting on purpose after eating (past year)  2.26  1.30  3.91  .004  Binge drink 3+ times in past month  2.03  1.31  3.15  .002  Attempted suicide in past year  9.04  4.87  16.78  <.001  Low emotional distress (0-1)  0.06  0.03  0.15  <.001  Family connectedness (0-1)  0.18  0.08  0.43  <.001  Self-reported health status  0.51  0.33  0.78  .002  Outlook: perception of life circumstances  0.32  0.17  0.61  .001  5-year future outlook  0.48  0.24  0.97  .042  Less trouble getting along with students  0.58  .35  0.97  .309  Protective Factors  Note. AOR = adjusted odds ratios, controlling for age, CI = 95% confidence interval. aFemales n = 383  99 Table 11 Bivariate Logistic Regression Analyses for Males: Risk and Protective Factors for Self-Injury Variable/Factor  Malea AOR  AOR 95%CI Lower Upper  P  Risk Factors Sexual minority membership  3.18  1.66  6.11  <.001  Sexual abuse  2.10  1.15  3.84  .016  Physical abuse  2.42  1.48  3.94  < .001  Witnessed family abuse  1.88  1.15  3.05  .011  Attacked or assaulted in past yr  2.00  1.24  3.23  .005  Victimization index (0-1)  10.37  3.82  28.16  < .001  Physically abused by ≥ 2 perpetrators  1.93  1.14  3.27  .015  Victim of relational aggression (0-1)  5.04  2.62  9.70  <.001  Ever kicked out  1.76  1.08  2.88  .023  Father problems (0-1)  3.48  1.45  8.37  .005  Family suicide history Consequences for substance use (0-1)  2.45 3.22  1.49 1.01  4.01 10.32  < .001 .049  Physical risk-taking (sharing needles/tattoo tools) (0-1)  8.10  3.50  18.78  < .001  Vomiting on purpose after eating  2.40  1.27  4.51  .007  Attempted suicide in past year  5.18  2.65  10.12  <. 001  Low emotional distress (0-1)  0.17  0.07  0.39  < .001  Family connectedness (0-1)  0.16  0.06  0.42  < .001  Self-reported health status  0.45  0.28  0.73  .001  Less trouble getting along with peers  0.44  0.26  0.75  .003  Protective Factors  Note. AOR = adjusted odds ratio, controlling for age. CI = 95% confidence interval. aMales n = 367  Risk factors. Findings from bivariate logistic regression analyses revealed several risk factors that were shown to increase the odds of NSSI. The robustness of risk factors differed by gender. Results revealed an approximately three-fold increase in likelihood of NSSI among both males and females who reported sexual minority status. Notably, despite the strong association between this demographic factor and self-injury, it was not included in the multivariate models due to its aforementioned usage as a grouping variable.  100 The strongest risk factors to emerge for females (i.e., with at least a 9-fold increase in odds of self-injury) were as follows: previous year suicide attempt, sharing sharp tools (e.g., piercing/tattooing tools, needles or razors), overall history of being victimized (physical, sexual, relational), and experiencing the consequences of substance use. Other factors conferring risk included: history of parent problems, being abused by multiple perpetrators, living in precarious housing, witnessing abuse, physical fighting, intentional vomiting following eating, and family history of suicide. For males, the most significant risk factors, increasing the odds of NSSI between 5 to 10-fold, were: overall history of being victimized, engaging in physically risky behaviour (sharing sharp tools), previous suicide attempt, and being the victim of relational aggression. For example, males who reported high levels of victimization were more than ten times as likely to self-injure compared to males with the lowest levels of victimization. Other significant risk factors included: specific victimization variables (physical abuse, sexual abuse, assault), father problems, family suicide history, vomiting on purpose after eating, and experiencing consequences of substance use. Given the large number of risk factors that were found to be significantly and positively associated with NSSI, effort was made in subsequent multivariate analyses to use scales or indices and to use variables considered to be amenable to change whenever possible. Protective factors. In contrast with the risk factors, certain variables were linked to reduced likelihood that youth would self-injure. At this bivariate level, for both males and females, feelings of connection to family were associated with approximately half the odds of self-injury. Similarly, greater emotional health and a perception of health as good/excellent were both associated with significantly decreased likelihood of self-injury. Males were less likely to self-injure if they reported having less difficulty socially with school peers (i.e., better peer relations) whereas perception of life circumstances as good/fair, and a positive future outlook were protective for females.  101 Testing for Multicollinearity Prior to proceeding with multivariate logistic regression analyses, bivariate associations between the self-injury outcome and all significant risk and protective factors were examined for multicollinearity using point-biserial correlations for categorical by continuous variables, Pearson correlation coefficients for continuous by continuous variables, and phi for categorical by categorical variables (see Tables 12 and 13). Only those variables that would next be entered into separate multivariate models for each gender were scrutinized (therefore some cells in the Tables below are left blank as not applicable due to insignificant findings for males or small cell size in the previous step); whereas there were six protective factors for girls, only four factors were significant for boys. Examination of the correlations indicated that relationships between variables were in the expected direction and generally low to moderate, with all protective factor correlations at below .4 and all risk factor correlations below .5. Thus, there was no evidence of significant multicollinearity among either protective or risk factors. Table 12 Age-Adjusted Partial Correlations Among Self-Injury Outcome and Protective Factors by Gendera Variables  1  2  3  4  5  6  7  ---  -.350†  -.211†  -.122*  -.168**  -.129*  -.198†  1.  Self-injury  2.  Low emotional distress  -.244†  ---  .377†  .137**  -.271†  .262†  -.284†  3.  Family connectedness  -.214†  .323†  ---  .127*  -.183†  .203†  -.262†  4.  Get along with peers  -.169**  .161**  .143**  ---  -.080  .098  -.036  5.  Self-reported health status  .180†  -.174**  ..258†  -.113*  ---  -.213†  .237†  6.  Positive future projection  ---  ---  ---  ---  ---  ---  .236†  7.  Positive attitude current circumstances  ---  ---  ---  ---  ---  ---  ---  *p < .05 **p ≤ .01 † p ≤ .001 Note. Dashes were also used to denote correlations that were not tested for males. a Correlations for females are displayed above the diagonal; correlations for males are displayed below the diagonal.  102 Table 13 Age-Adjusted Partial Correlations Among Self-Injury Outcome and Risk Factors by Gendera  Variables  1  2  3  4  ---  .289†  .271†  .221  †  .265†  ---  .494  †  .415  † †  5  6  7  8  9  10  11  12  13  ..296†  .179†  .217†  .162**  .275†  .374  .478  †  .277†  .325†  .293†  .262  †  .300  †  .195  †  .293†  .211  †  .210  †  .062  .201  .141**  14  †  .159**  .276†  .341  †  .412†  .393  †  .186  †  .405  †  .343  †  .330†  †  .268  †  .152**  .314  †  .348  †  ..216 †  .208  †  .325  †  .060  .171  †  .219  †  .162**  1.  Self-injury  2.  Victimization  3.  ≥2 perpetrators – PA  ---  ---  ---  .444  4.  ≥2 perpetrators SA  ---  ---  ---  ---  .242  5.  Relational aggression victim  .287†  .397†  .230†  ---  ---  .126*  .148**  .219  †  .221  †  .291  †  .208  †  .375  †  .302  †  .258†  6.  Precarious housing  ---  ---  ---  ---  ---  ---  .153**  .076  .048  .242  †  .109*  .247  †  .276  †  .220†  7.  Mother problems  ---  ---  ---  ---  ---  ---  ---  .305  †  .184  †  .136*  .003  .144**  .173  †  .202  †  8.  Father problems  .168**  .221†  .353†  ---  .194†  .072  .356†  ---  .155**  .164**  .110*  .110*  .220  †  .179  †  9.  Family history of suicide  .195†  .254†  .200†  ---  .178**  .122*  .183†  .235†  ---  .200  †  -.022  .183  †  .263  †  .234  †  10. Physical risk-taking  .307†  .334†  .214†  ---  .179**  .258†  .200†  .261†  .090  ---  .126*  .274  †  .352  †  .350  †  11. Vomiting on purpose  .155**  .223†  .023  ---  .088  .139**  .172**  .071  .067  .154**  ---  .101  .061  .198  †  12. Physical fighting  ---  ---  ---  ---  ---  ---  ---  ---  ---  ---  ---  ---  .329  †  .281  †  13. Consequences - substance use  ---  ---  ---  ---  ---  ---  ---  ---  ---  ---  ---  ---  ---  .270  †  .292†  .247†  .023  ---  .226†  .075  .076  .139*  .162**  .269†  .180†  .066  .063  ---  14. Suicide attempt  †  *p < .05 **p ≤ .01 p ≤ .001 Note. SA = sexual abuse, PA = physical abuse. Dashes were also used to denote correlations that were not tested for males. a Correlations for females are displayed above the diagonal; correlations for males are displayed below the diagonal. †  103 To further investigate the results for multicollinearity beyond the bivariate level, the variables were subjected to linear regression analyses and collinearity statistics were examined. Tolerance and VIF, or Variance-inflation Factor, permit the examination of the regression of each independent variable on all of the others. Results revealed that interdependencies among variables were not significant, with Tolerance and VIF well within the recommended cut-offs for both males and females (Tolerance < .20 and VIF > 4.0; Menard, 1995; Tabachnick & Fidell, 2001). Specifically, for females tolerance across the models (risk, protection, combined) ranged from 0.50 to 0.97 and VIF ranged from 1.05 to 1.61. For males, tolerance ranged from 0.70 to 0.96 and VIF ranged from 1.04 to 1.44. As suggested by Tabachnick and Fidell (2001), condition indices were examined and found to be less than 30 (the largest is 9.35) and therefore assumed to be relatively stable (Hosmer & Lemeshow, 2000). Multivariate Models for Self-Injury Initial separate models of risk and protection. Findings from separate multivariate risk and protection models, with age as covariate, are presented in Tables 14 and 15 for females and males, respectively. When variables significant in the previous bivariate step were entered into the multivariate risk-only or protection-only models to determine if they would be significant even in the presence of other factors, not all were significant when considered in combination. Using the aforementioned criteria for statistical and clinical significance of OR (risk: OR ≥ 2.0, protective: OR ≤ 0.5, both p ≤ .05), the model revealed multiple risk factors (5 for females, 3 for males) and protective factors (1 for females, 3 for males). Taking into account all other factors in the models, these factors independently contributed to the self-injury outcome. For females, risk factors included: history of suicide attempt, experiencing consequences of substance use, physically risky behaviour, index of mother problems, and  104 being victim of multiple sexual abuse perpetrators. Only low emotional distress retained statistical and clinical significance; although a second protective factor, family connectedness, showed clinical significance in terms of the OR of 0.41, it was not statistically significant. For males, history of suicide attempt, physically risky behaviour, and being victim of relational aggression were significant risk factors. Three protective factors, connection to family, low emotional distress, and perception of getting along with peers influenced the odds of selfinjury, while a fourth protective factor, self-reported health status, showed statistical significance but was above the cut-off for clinical significance. Table 14 Multivariate Logistic Regression Models of Risk and Protection for Females Variable  B  S.E.  Wald  AOR ¤  Risk  AOR 95% CI Lower Upper  p  factorsa  ≥2 Perpetrators physical abuse  -0.03  0.38  0.01  0.97  0.46  2.05  .936  ≥2 Perpetrators sexual abuse  1.25  0.59  4.50  3.49  1.10  11.08  .034  Victimization index  -0.67  0.77  0.76  0.51  0.11  2.31  .382  Vomiting on purpose  0.41  0.41  0.98  1.50  0.67  3.36  .321  Father problems index  0.19  0.59  0.10  1.21  0.38  3.82  .748  Mother problems index  1.33  0.60  4.95  3.78  1.17  12.18  .026  Victim of relational aggression  0.66  0.45  2.20  1.94  0.81  4.64  .138  Family history of suicide  0.61  0.32  3.77  1.85  0.99  3.43  .052  Precarious housing  0.02  0.34  0.00  1.02  0.53  1.97  .958  Physical fighting  0.15  0.34  0.20  1.16  0.60  2.26  .653  Physically risky behaviour  1.08  0.55  3.88  2.95  1.01  8.64  .049  Consequences of substance use  1.48  0.76  3.82  4.37  1.00  19.19  .051  History of suicide attempt  1.78  0.42  17.88  5.92  2.60  13.49  < .001  Low emotional distress  -2.25  0.56  16.02  0.11  0.05  0.32  < .001  Family connectedness  -0.90  0.62  2.13  0.41  0.12  1.36  .144  Self-reported health status  -0.57  0.30  3.60  0.57  0.32  1.02  .058  Life circumstances  -0.28  0.44  0.41  0.75  0.32  1.79  .521  Protective factorsb  Future outlook -0.15 0.43 0.12 0.87 0.38 2.00 .735 a Risk-only model chi-square (df = 14) = 101.85, p < .001. b Protective-only model chi-square (df = 6) = 45.27, p < .001. Note. AOR = adjusted odds ratio, controlling for age. CI = 95% confidence interval.  105 Table 15 Multivariate Logistic Regression Models of Risk and Protection for Males Variable  B  S.E.  Wald  AOR  AOR 95% C.I. Lower Upper  p  Risk factorsa Victimization index 0.50 0.69 0.51 1.65 0.42 6.41 .473 Vomiting on purpose 0.76 0.49 2.36 2.13 0.81 5.60 .124 Index father problems 0.53 0.58 0.83 1.70 0.54 5.30 .362 Victim of relational aggression 0.89 0.42 4.43 2.43 1.06 5.53 .035 Family history of suicide 0.57 0.34 2.99 1.76 0.93 3.34 .084 Physically risky behaviour 1.28 0.53 6.12 3.60 1.31 9.92 .013 History of suicide attempt 1.29 0.45 8.65 3.63 1.54 8.57 .003 b Protective factors Low emotional distress -1.23 0.48 6.68 0.29 0.12 0.74 .010 Family connectedness -1.06 0.54 3.79 0.35 0.12 1.01 .051 Getting along with peers -0.64 0.29 4.99 0.53 0.30 0.92 .026 Self-reported health status -0.55 0.27 4.19 0.58 0.34 0.98 .041 a Risk-only model chi-square (df = 7) = 50.53, p < .001. b Protection-only model chi-square (df = 5) = 35.29, p < .001. Note. AOR = adjusted odds ratio, controlling for age. CI = 95% confidence interval.  Final combined models of risk and protection. Factors significant in the bivariate and initial risk- and protection-only models were combined into the final multivariate logistic regression models, covaried by age, and conducted separately by gender. Table 16 presents the final multivariate models for both males and females. As expected, findings revealed negative coefficients for the protective factors and positive coefficients for risk factors. The goal was to select a maximum of three factors each for risk and protection due to issues related to the complexity of constructing, presenting, and interpretability of probability profiles. With five clinically and statistically significant risk factors identified in the previous step for females and a goal of three maximum, two required elimination. Keeping in mind the goal of focussing on selecting a set of meaningful factors amenable to change through direct intervention in street-involved adolescents‘ lives, two contextual level variables, index of mother problems and a history of being sexually abused by more than two perpetrators (recent and/or historical), were not selected for use in the final model. The final multivariate logistic regression of protective and risk factors on self-injury among females included physically risky behaviour (e.g., sharing needles, razors), history of  106 suicide attempt, and experiencing the consequences of substance use as risk factors and low emotional distress as the sole protective factor. The overall test indicates that the four factors taken together were significantly related to self-injury. The ORs of the risk factors for females highlighted that each of the three risk factors increased the likelihood that a young woman would self-injure by approximately six times. Conversely, OR results showed that a young woman with a high score on the emotional health protective factor (indicating lower levels of distress) was less than half as likely to self-injure. For males, three risk factors were identified as significant in the previous risk only model and were therefore used in the final model. Similar to the females, history of suicide attempt and engaging in physically risky behaviour made it nearly four times as likely that a young man self-injured. The odds of NSSI were increased by 2.5 times among males who reported more victimization by relational aggression. Three protective factors, connections to family, low emotional distress, and getting along with school peers were utilized in the final model for males; each reduced the likelihood of self-injury by one-third to approximately onehalf. The overall model test indicates that the full combination of factors was significantly related to self-injury. Table 16 Final Combined Multivariate Logistic Regression Model of Risk and Protection Variable  B  S.E.  Wald  AOR  AOR 95% C.I. Lower Upper  Females Final Modela Physically risky behaviour 1.92 0.51 14.34 6.80 2.52 Consequences of substance use 1.75 0.67 6.80 5.76 1.55 History of suicide attempt 1.76 0.40 19.62 5.81 2.67 Low emotional distress -0.83 0.55 2.24 0.44 0.15 Males Final Modelb Victim of relational aggression 0.73 0.41 3.11 2.07 0.92 Physically risky behaviour 1.39 0.56 6.17 4.00 1.34 History of suicide attempt 1.18 0.42 7.87 3.26 1.43 Low emotional distress -0.82 0.58 2.01 0.44 0.14 Family connectedness -1.06 0.63 2.82 0.35 0.10 Get along with peers -0.52 0.34 2.36 0.59 0.31 aModel chi-square (df = 5) = 97.46, p < .001. bModel chi-square (df = 7) = 58.22, p < .001. Note. AOR = adjusted odds ratio, controlling for age. CI = 95% confidence interval.  p  18.34 21.45 12.67 1.29  < .001 .009 < .001 .134  4.66 11.93 7.46 1.37 1.19 1.15  .078 .013 .005 .156 .093 .124  107 Probability Profiling Beyond providing a portrait of the magnitude of risk and protection associated with each variable, the final combined multivariate logistic regression models also yielded beta weights necessary for constructing probability profiles. These profiles demonstrate the likelihood of self-injury as an outcome using different combinations of risk and protective factors identified in the previous step, and taking into account the impact of whether the youth scored high or low (90th vs. 10th percentiles, respectively, or present or absent, as per Rubenstein et al., 1989) on the factor. Probability profiling permits the prediction of the probabilities of self-injury, highlighting the increases or decreases related to particular profiles (or combinations). The possible combinations, 16 for females and 64 for males, and their related probabilities, are presented in Tables 17 and 18, respectively. Factors were selected from the multivariate logistic regression analyses to construct the probability profiles. For females, only one protective factor, low emotional distress, mitigated against the self-injury outcome whereas three risk factors were used: history of suicide attempt, consequences of substance use, and physically risky behaviour. The three strongest protective factors for males were connectedness to family, low emotional distress, and getting along with school peers. The three strongest risk factors for males were being a victim of relational aggression, history of suicide attempt, and physically risky behaviour. The probability profile for females shows decreasing probability of NSSI with the addition of the protective factor, especially in the presence of a single risk factor but this holds true with two risk factors present as well (see also Figure 2). With none of the identified risk or protective factors, the probability of self-injury is 35%. With the introduction of one protective factor, low emotional distress, the probability of self-injuring is reduced to nearly 22%. In contrast, in the worst case scenario with three risk factors and no protective factor, the probability of selfinjury is nearly a ‗sure thing‘, at 98%; in this scenario, the inclusion of the protective factor does little to offset risk with a decrease of only 1% (to 97%).  108 Table 17 Probability Profiles of Self-Injury for Females Number of Risk Factors Zero One  Risk Factors  Suicide attempt in past year Consequences for substance use Physical risk (sharing needles/tools)  Two  Three  Suicide attempt Consequences for substance use Suicide attempt Physical risk Consequences for substance use Physical risk Suicide attempt Consequences for substance use Physical risk  Protective Factor Low Emotional Distress Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No  Probability Of Self-injury 0.218 0.353 0.618 0.761 0.450 0.616 0.655 0.788 0.826 0.903 0.917 0.956 0.848 0.916 0.970 0.984  Note. BX = BConstant + BageXage + BsuicideattemptXsuicideattempt + BconsequencesXconsequences + BphysicalriskXphysicalrisk + BlowdistressXlowdistress Figure 2 Probability Profile for Self-Injury Among Females No Risk Factors (RF)  22  1 RF: Consequences of substance use  No Protective Factors 45  1 RF: Suicide attempt Risk Factors  35 62  62  1 RF: Physical risk  66  Low Emotional Distress  76 79  2 RF: Suicide attempt + Substance use  83  2 RF: Physical risk + Substance use  85  90 92 96 92  2 RF: Suicide attempt + Physical risk  98 97  All 3 RFs 0  20  40 60 Probability of Self-Injury  80  100  109 For males, the probability of NSSI increased as the number of risk factors grew, from nearly 40% with no risk or protective factors to 90% with three risk factors and no protective factors. The addition of protective factors was the most effective method to reduce likelihood of self-injury, even more so than reducing risk, and this held true at every level (see Figure 3). In particular, family connectedness appeared to be the strongest protective factor of the three. In general, the introduction of protective factors reduced the probability of self-injury by onethird with one protective factor and by close to one-half with two. Similarly, with no risk factors and all three protective factors, basically the ―best case scenario,‖ there was a 9% chance of self-injury; however, with the addition of each risk factor the probability increased, from 9% to 17-25% with one risk factor (depending on which one) to 30-41% with two, and to 58% with all three. In the profile with all three risk factors and 90% probability of self-injury, the addition of one protective factor offered little to mitigate, but with two protective factors, the probability reduced to 72-74% and, as aforementioned, with three protective factors, reduced further to 58%.  110 Table 18 Probability Profiles of Self-Injury for Males Number of Risk Factors  Risk Factors  Zero  One  Physical risk  Suicide attempt  Victim of relational aggression  Two  Physical risk Suicide attempt  Protective Factors Low Emotional Family Distress Connectedness Yes Yes Yes Yes Yes No Yes No No Yes No Yes No No No No Yes Yes Yes Yes Yes No Yes No No Yes No Yes No No No No Yes Yes Yes Yes Yes No Yes No No Yes No Yes No No No No Yes Yes Yes Yes Yes No Yes No No Yes No Yes No No No No Yes Yes Yes Yes Yes No Yes No No Yes No Yes No No No No  Probability Getting Along With Peers Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No Yes No  of Self-injury 0.094 0.148 0.175 0.264 0.160 0.243 0.282 0.398 0.171 0.258 0.298 0.417 0.276 0.391 0.440 0.569 0.252 0.362 0.410 0.539 0.384 0.512 0.562 0.683 0.176 0.265 0.306 0.426 0.283 0.400 0.449 0.578 0.403 0.532 0.581 0.700 0.554 0.677 0.719 0.812  111 Table 18 cont’d... Probability Protective Factors Risk Factors Low Emotional Family Getting Along of Self-injury Distress Connectedness With Peers Two cont’d... Physical risk Yes Yes Yes 0.300 Victim of Yes Yes No 0.419 relational Yes No Yes 0.469 aggression Yes No No 0.598 No Yes Yes 0.441 No Yes No 0.571 No No Yes 0.619 No No No 0.732 Suicide attempt Yes Yes Yes 0.412 Victim of Yes Yes No 0.541 relational Yes No Yes 0.590 aggression Yes No No 0.708 No Yes Yes 0.563 No Yes No 0.685 No No Yes 0.726 No No No 0.817 Three Physical risk Yes Yes Yes 0.583 Yes Yes No 0.702 Suicide attempt Yes No Yes 0.742 Yes No No 0.829 Victim of No Yes Yes 0.721 relational No Yes No 0.813 aggression No No Yes 0.841 No No No 0.899 Note. BX = BConstant + BageXage + BphysicalriskXphysicalrisk + BsuicideattemptpastyrXsuicideattemptpastyr + BrelationalaggressionXrelationalaggression + BlowdistressXlowdistress + BfamilyconnectednessXfamilyconnectedness + BgetalongXgetalong Number of Risk Factors  112 Figure 3 Probability Profile for Self-Injury Among Males  No Risk Factor (RF) 1 RF: Physical risk  68  25  2 RF Physical risk + Relational aggression  All 3 PF  58  18  1 RF: Suicide attempt  No PF 57  17  1 RF: Victim of relational aggression Risk Factors  40  9  73  30  2 RF: Suicide attempt + Physical risk  81  40  2 RF: Suicide attempt + Relational aggression  82  41  All 3 RF  90  58 0  20  40  60  80  100  Probability of Self-Injury  Further Analyses for Females Given the influence of connections to family at the bivariate level (recall, OR = 0.18, p < .001, CI95 0.08, 0.43), the clinical significance of family connectedness in the protectiveonly multivariate logistic regression model (OR = .41, p = .144, CI95 0.12, 1.36) and the large body of research highlighting the centrality of relationships in protecting against negative outcomes (and in promoting positive outcomes) for vulnerable adolescents, it was surprising, and even counterintuitive, that this variable did not show statistical significance in the protective-only model for females, as it did for males. Correlation between low emotional distress and family connectedness, the two clinically significant variables in the protectiononly model, was .38 for females, not indicative of multicollinearity and, as aforementioned, specific collinearity statistics did not reveal any problems. However, it is worth noting that this  113 correlation coefficient was the highest found among protective factors. Finally, estimated standard errors in the reported protective-only model were not large, regression coefficients were not large, and their signs did not change, all indicators of numerical problems in the model (Hosmer & Lemeshow, 2000). Nevertheless, it was hypothesized that the relationship between these two variables and among the group of protective factors presented possible confounding. Therefore, further examination of protective factors was undertaken in order to better understand the relation of the variables to each other and to self-injury. When a multivariate logistic regression analysis for the protection-only model was conducted with an interaction term (Family Connectedness x Low Emotional Distress), results revealed that the interaction was not significant (β = -2.63, S.E. = 2.05, Wald = 1.65, OR = .07, p = .199), indicating that the assumption of linearity of the logit had been met (Hosmer & Lemeshow, 2000) and finding no evidence of suppressor effects. A scatterplot with these variables showed a positive relationship with equal scatter on both sides of the fit line. Next, a multivariate logistic regression analysis was conducted for all protective factors found to be significant for females at the bivariate level, this time excluding low emotional distress, in order to determine whether family connectedness would influence the self-injury outcome in the absence of the emotion variable (see Table 19). Results from this model showed that both connection to family and subjective health status independently contributed to the model by at least halving the odds of NSSI. Interestingly, in the absence of the emotion variable, both factors achieved clinical and statistical significance even in the presence of the other protective factors. These two protective factors were then combined with the three risk factors from the risky-only model (history of suicide attempt, physical risk-taking, and experiencing the consequences of substance use) in a final multivariate logistic regression model (see Table 20). The overall test indicates that the five factors together as a set were significantly related to the self-injury outcome. The ORs of the risk factors for females highlighted that each of the  114 three risk factors influenced the odds that a young woman would self-injure by five or six-fold. Conversely, results showed females who were more connected to their family were significantly less likely to self-injure. Table 19 Multivariate Logistic Regression Protection-Only Model for Females Variable  B  S.E.  Wald  AOR  P  AOR 95% CI Lower  Upper  Family connectedness  -1.56  0.57  7.49  0.21  0.07  0.64  .006  Self-reported health status  -0.71  0.28  6.24  0.49  0.28  1.02  .012  Life circumstances  -0.58  0.42  1.92  0.56  0.25  1.27  .166  Future outlook -0.31 0.40 0.61 0.73 Protective-only model chi-square (df = 5) = 27.91, p < .001. Note. AOR = adjusted odds ratio, controlling for age. CI = 95% confidence interval.  0.33  1.60  .435  Table 20 Final Combined Multivariate Logistic Regression Model for Females Variable  B  Physically risky behaviour 1.91 Consequences of substance use 1.63 History of suicide attempt 1.76 Family connectedness -1.18 Self-reported health status -0.34 Model chi-square (df = 5) = 96.59, p < .001. Note. AOR = adjusted odds ratio, controlling for age.  S.E. 0.50 0.66 0.39 0.57 0.29  Wald 14.45 6.15 20.43 4.21 1.42  AOR 6.72 5.11 5.82 0.31 0.71  AOR 95% C.I. Lower Upper 2.52 17.95 1.41 18.56 2.71 12.48 0.10 0.95 0.41 1.25  P < .001 .013 < .001 .040 .233  CI = 95% confidence interval.  Beta weights from the combined model were utilized along with scale scores representing the 10th and 90th percentiles (or absence/presence) to construct probability profiles highlighting the interaction of risk and protective factors in influencing NSSI. The predicted probabilities based on this combination of three risk factors and two protective factors revealed profiles presented in Table 21 and represented graphically in Figure 4. For the profiles characterized by both protective factors, family connectedness and self-reported health status, and no risk factors, 19% of the females were ―predicted‖ to self-injure. With no protective factors and in the absence of risk factors, the estimate jumps to 43%. As soon as a single risk factor is added, with no protective factors the likelihood of NSSI ranges from 67% to 83% depending on the variable, however the introduction of both protective factors reduces the odds by almost half, ranging from 39% to 62%. At higher levels of risk, with two and three  115 risk factors, the buffering effect of the protective factors is far less evident; with three risk factors and both protective factors the likelihood of self-injury is almost 100%. Table 21 Probability Profiles of Self-Injury for Females with Two Protective Factors Number Of  Protective Factor Risk Factors  Risk Factors Zero  One  Three  Self-reported  Connectedness  Health Status  Of Self-Injury  Yes  Yes  0.192  Yes  No  0.250  No  Yes  0.346  No  No  0.427  Yes  Yes  0.580  Yes  No  0.660  No  Yes  0.755  No  No  0.812  Consequences of  Yes  Yes  0.393  substance use  Yes  No  0.477  No  Yes  0.590  No  No  0.670  Yes  Yes  0.615  Yes  No  0.692  No  Yes  0.780  No  No  0.833  Suicide attempt  Yes  Yes  0.790  Consequences of  Yes  No  0.841  substance use  No  Yes  0.893  No  No  0.922  Suicide attempt  Yes  Yes  0.903  Physical risk  Yes  No  0.929  No  Yes  0.954  No  No  0.967  Consequences of  Yes  Yes  0.813  substance use  Yes  No  0.860  Physical risk  No  Yes  0.906  Suicide attempt  No Yes  No Yes  0.932 0.962  Consequences of  Yes  No  0.973  substance use  No  Yes  0.983  Suicide attempt  Physical risk  Two  Family  Probability  Physical risk No No 0.988 Note. BX = BConstant + BageXage + BsuicideattemptpastyrXsuicideattemptpastyr + BphysicalriskXphysicalrisk + BconsequencesofsubuseXconsequencesofsubuse + BfconnectXfconnect + BhealthstatusXhealthstatus  116 Figure 4 Probability Profile for Self-Injury Among Females with Two Protective Factors  No RF  43  19  1 RF: Consequences of substance use  No PF  1 RF: Physical risk Risk Factors  Both PF  67  39  83  62  1 RF: Suicide attempt  81  58  2 RF Suicide attempt + Substance use  79  2 RF: Physical risk + Substance use  81  2 RF: Suicide attempt + Physical risk  92 93  90  97 99 96  All 3 RF 0  20  40  60  Probability of Self-Injury  80  100  117 CHAPTER 5: DISCUSSION Reports of increasing prevalence of nonsuicidal self-injury among school-based adolescents and young adults have prompted a burgeoning clinical and research interest in this counterintuitive behaviour in recent years. Emphasis on conducting rigorous empirical studies with community samples has contributed to our growing understanding of NSSI, but many gaps remain. With the traditional focus on risk and deficits, protective factors have been little explored for their relation to NSSI. This study is the first to comprehensively examine and identify multiple risk and multiple protective factors for NSSI among streetinvolved adolescents. Consistent with a risk and resilience framework, this study uses a novel methodology, probability profiling, to highlight the key role of protection in offsetting risk, applicable even in this sample of highly vulnerable youth. Results of this study add weight to the argument for a dual approach encompassing both risk reduction and protection promotion in our prevention and intervention policies and efforts with this population. The present study provides a theoretically and empirically guided examination of risk and protective factors for NSSI among a street-involved sample of 762 adolescents across nine communities in British Columbia. The study involved secondary analyses of data collected using the McCreary Centre Society‘s Street-Involved Youth Health Survey (SYHS). Findings obtained through a series of bivariate and then multivariate logistic regression analyses became the foundation for the development of probability profiles highlighting the likelihood of NSSI given various combinations of risk and protective factors (Bearinger et al., 2010; Bearinger et al., 2005; Erickson, 1999). With extant evidence suggestive of gender and age differences in risk behaviour, risk factors, and mitigating protective factors, separate analyses were conducted by gender and age was controlled for in the present study. This chapter begins with a review of the purpose and specific aims of the current study. This is followed by a discussion of the results that is contextualized in terms of  118 previous research and suggestions for future research. Next, strengths and limitations of the study are described, with a brief follow up on future research directions. Finally, implications for practice are presented. The current research had three primary aims: 1) To describe the characteristics including sociodemographic distribution and reasons for NSSI among street-involved adolescents. 2) To examine relations between and among individual and contextual factors and NSSI and to identify those that increased the likelihood of self-injury and, conversely, those that decreased the likelihood. 3) To estimate the probability of NSSI given various combinations of the strongest risk and protective factors. Findings from each of these objectives will be discussed in the subsequent section. Characteristics of Self-Injury Among Street-Involved Adolescents Prevalence and Sociodemographics With respect to the first research objective, estimates of prevalence lead to improved understanding of the scope of this problem and the related needs for resource allocation, prevention, and for determining who would benefit from being targeted for specific intervention. Previous, albeit limited, research has suggested that certain groups may be at disproportionate risk for NSSI. Extant research findings on gender differences have been inconsistent. Results showed that NSSI is prevalent, with more females than males reporting this history (56% versus 34%). Of these, a significant number, 66% females and 48% males, reported engaging in the behaviour three or more times. These rates are consistent with those found in the few studies of North American street youth (49-69%, e.g., Kidd, 2002; Tyler et al., 2003; Unger et al., 1997), but are higher than those recently cited in studies with community school-based adolescents (most typically 13-28%; Alfonso & Dedrick, 2010; LayeGindhu & Schonert-Reichl, 2005; Ross & Heath, 2002). Accumulating evidence particularly with community samples of adolescents and young adults suggests similar prevalence across  119 genders when males are adequately sampled (Alfonso & Dedrick, 2010; Hilt, Nock, et al., 2008; Lloyd-Richardson et al., 2007; Muehlenkamp & Gutierrez, 2004; Whitlock et al., 2006). Notably, there are vast differences in how NSSI is operationalized and assessed, rendering cross-study comparison challenging. The current study utilized a single question to determine the history of self-injury, rather than soliciting information about specific behaviours. In their studies of homeless older adolescents and young adults in the United States using checklists with numerous specific self-injurious behaviours to endorse, Tyler and colleagues (Tyler et al., 2010; Tyler et al., 2003) cited a prevalence rate of 60-70% for both males and females. It is possible that using a prescriptive list of concrete behaviour reduces subjectivity and reminds youth of the range of behaviours they may have engaged in, leading to greater endorsement. Thus, the current study prevalence may underestimate the extent of this behaviour. A study utilizing both methods to collect data followed by a debriefing or interview with the youth may help us to better understand the costs and benefits of using specific questions and/or formats and to provide an opportunity to expand on responses provided. Congruent with expectations and with previous research showing that gay, lesbian, and bisexual (GLB) youth are at higher risk for a variety of negative outcomes (Austin et al., 2007; Coker et al., 2010; Garofalo et al., 1999; Saewyc, 2011; Saewyc et al., 1997), in the current study identifying as a sexual minority youth (i.e., non-exclusively heterosexual) increased the likelihood of NSSI approximately three-fold for both males and females, with half to three-quarters of the GLB adolescents reporting the behaviour. Similarly, research on NSSI in university students has shown increased prevalence among sexual minority youth and bisexual-identified individuals, in particular (Deliberto & Nock, 2008; Skegg et al., 2003; Whitlock et al., 2006). Sexual minority youth who self-injure may face additional risk due to poor social support, social exclusion, secrecy, shame, and stigma, all aspects of marginalized social status (Tyler et al., 2010). In fact, sexual minority self-injuring street youth could be  120 thought to be triply stigmatized. Given that negotiating emerging sexuality is an important developmental task during adolescence, youth who do not fit the dominant binary mode would be expected to experience increased normative stress as well as minority stress. Researchers studying suicide among sexual minority youth have found that general adolescent risk factors in the lives of these youth (e.g., victimization, depression) may mediate the relation between sexual orientation and suicidality (Russell, 2003; Russell & Joiner, 2001), leading to a call for increased focus on studying risk and protective factors unique to sexual minority status. Further research efforts to clarify the relations between sexual minority status and outcomes such as NSSI, and to better understand the degree to which risk varies between minority statuses (mostly heterosexual, bisexual, lesbian, gay, and questioning), risk and protective factors and outcomes, is warranted and has significant implications for intervention. Results showed similar rates across the two majority ethnic groups represented, Aboriginal and European. Studies have infrequently examined ethnic differences or included sufficiently large minority groups to be able to detect differences (e.g., Hilt, Nock et al., 2008, reported on sample that was 87% White), and ethnic groups are not generalizable geographically. Nevertheless, consistent with this finding, some studies report no difference (Alfonso & Dedrick, 2010; Hilt, Nock et al., 2008; Whitlock et al., 2006) whereas others report higher prevalence among Caucasian (or White) adolescents (Lloyd-Richardson et al., 2007; Ross & Heath, 2002). In the United Kingdom, research has identified that South Asian young women are at increased risk for NSSI (Marshall & Yasdani, 1999). In Canada, rates of suicide among Aboriginal youth are alarmingly high, but a search found no data specific to nonsuicidal self-injury. To my knowledge, no other Canadian research has examined ethnic differences in NSSI. Although given the statistics on suicide, we might expect disproportionately high rates of NSSI among Aboriginal youth on the street, results revealed no difference. It may be that the high level of risk universally experienced by the sample of  121 youth regardless of ethnicity both prior to street-involvement and related to the unique social circumstance of street life is the ―great equalizer.‖ Reasons for Self-Injury Consistent with the view that self-injury is an overdetermined behaviour (i.e., serving multiple functions simultaneously), both males and females endorsed multiple reasons for injuring at last incident, with the most frequent being feelings of loneliness or depression, stress, and anger. A recent theoretically and empirically supported conceptualization of the functions of NSSI categorizes them as being either automatic (i.e., intrapersonal) or social (i.e., interpersonal), and providing either negative (i.e., remove or stop an aversive cognitive or emotional state) or positive (i.e., generate a desired cognitive or emotional state) reinforcement (Klonsky, 2009; Klonsky & Glenn, 2009a; Nock, 2009, 2010; Nock & Prinstein, 2004, 2005). The most commonly reported present sample reasons are consistent with an automatic negative reinforcement function in which intolerable or uncomfortable internal states are regulated through self-injury (e.g., feeling stressed, engage in self-injury, infer that stress is reduced at least temporarily although youth were not explicitly asked). In contrast, the social or interpersonal function (e.g., attention-seeking, communication) was less frequently endorsed. Previous studies have found gender differences in NSSI method and motivation among adolescents (Laye-Gindhu & Schonert-Reichl, 2005) and young adults (Andover et al., 2010; Klonsky & Glenn, 2009; Whitlock et al., 2006). Differences in method and in function may reflect gender differences in underlying psychopathology and coping strategies, with pressure to conform to masculine gender norms such as toughness, aggression, emotional control, and sensation-seeking falling more typically into the male domain (Andover et al., 2010; Laye, 2002). With only eight reasons listed in the current study survey, some of which are ambiguous (e.g., I was angry or I felt rejected could be either automatic negative  122 reinforcement or social positive reinforcement), results provide only a glimpse into function. For example, self-derogation, frequently nominated as an important aspect of NSSI in the literature, is absent from this list. Future research would benefit from using a psychometrically robust measure to examine generalizability of this functional model to the street youth population and to both males and females. Another point worthy of investigation is how the functions may be similar or different relative to other self-harming or risk behaviour engaged in by the youth. An important question that has not yet been answered in the research is: why this particular behaviour, at this particular time, for this particular individual? Further, given their early independence, marginalized status, specific social needs, and the issues of belonging and peer group affiliation, an improved understanding of the role of social functions in this population is of keen interest. Risk and Protective Factors Associated with Self-Injury In response to the second objective, reflecting the reality that youth have a combination of vulnerability and positive protective factors at play in their lives, a unique contribution of the current study was to identify those factors that either increased or decreased the likelihood of NSSI as an outcome. Results revealed a wide range of individual and contextual risk and protective factors that were associated with NSSI and suggested that these factors were different for males and females. A dearth of studies have examined variables that protect against NSSI, and although the literature is rife with studies of risk factors, rarely are these factors looked at in combination or by gender. A recent book chapter explicitly on psychosocial risk and protective factors by Klonsky and Glenn (2009) mentions only two protective factors for NSSI: what these researchers refer to as ―effective management and expression of negative emotions‖ (p. 51; i.e., less intense and less frequent negative emotional reactions; Klonsky et al., 2003) and family cohesion, as borrowed from the suicide literature (Rubenstein et al., 1998).  123 Protective Factors Based on the literature on protective factors for suicide and for other risky behaviour, specific variables were hypothesized to reduce the likelihood of NSSI and to mitigate the risk. These findings varied by gender. For males in this sample, although thirteen variables were examined as potential protective factors at the bivariate level, only four were significantly associated with NSSI, and three of the four independently contributed to reducing the odds of NSSI outcome by at least 50% (p ≤ .05) in multivariate models, with other factors controlled. The three clinically and statistically significant protective factors for males were: feeling connected to family, better peer relations at school (i.e., relatively infrequent trouble getting along with peers), and greater emotional health (i.e., lower distress). In particular, feeling connected to family and reporting better emotional health reduced the odds of NSSI by twothirds for males. In the multivariate model, only good/excellent subjective health status, a quality of life indicator that has been found to correlate well with mental health and well-being in other studies (Topolski et al., 2004), was statistically but not clinically significant and was therefore not used in the final model. When the same thirteen potential protective factors were examined at the bivariate level for females, five demonstrated significant association with NSSI. As for males, greater emotional health, higher levels of connection to family, and good/excellent subjective health status reduced the odds of NSSI. Additionally, two further factors were protective: having a positive perception of life circumstances and positive five-year future outlook. In the initial multivariate model only one of the five, low emotional distress (i.e., better emotional health) remained significant. This finding was counterintuitive, given the significance of family connectedness for males and the salience of attachment and family relationships in the literature on self-harm generally and on NSSI. Studies of adolescent suicide have highlighted the protective role of family connection and its relation to emotional health for both boys and girls (Borowsky et al., 2001; Resnick et al., 1997). However, when the distress variable was  124 removed due to potential confounding and the four remaining factors identified as protective at the bivariate level were included in a new multivariate model, both having better connections with family and reporting better subjective health status were protective against NSSI. It would appear that by including both scale variables, family connectedness and low distress, in the same model, the effects of the former were suppressed for females only. Research on the development of emotional regulation, family emotional climate, and gender socialization of emotion may help to provide insight into the relations among family connectedness, distress, NSSI, and gender, while placing these findings in a larger developmental context (see reviews by Crowell et al., 2009; Eisenberg et al., 2010; Zeman et al., 2006). Children‘s emotion regulation is both directly and indirectly impacted by parenting style, parenting practices and behaviours, observation, and emotional climate of the family, all of which are associated with perceived connectedness to family (Eisenberg et al., 2010; Morris, Silk, Steinberg, Myers, & Robinson, 2007; Sim et al., 2009). Studies have found gender-typic socialization of emotional behaviours; for example, parents reinforce displays of sadness in girls and anger in boys, increasing the likelihood of these emotional responses. Further, girls are socialized to be more relationship-oriented in emotion regulation strategies and to take on more familial and mediational roles (Zeman et al., 2006). Girls‘ emotional states are more closely linked to their emotional connections with others (Adrian et al., 2009) as they are socialized to have greater emotional investment in their families. They are also thought to be more susceptible to parental modelling and reinforcement of maladaptive emotion regulation strategies and to the negative effects of family discord (Adrian et al., 2009; Eisenberg et al., 2010). This may mean that girls‘ distress and relationship with parents is more complicated and confounded than that of boys. In a further study focusing on parent-child relations and emotion, with a sample of young adult students (75% female), Buckholdt et al. (2009) found that parenting responses in relation to negative emotion, specifically punishment and neglect of sadness, were directly  125 related to NSSI (as a maladaptive emotion regulation strategy) and that the associations were mediated by emotion regulation. They theorize that it is not only adverse events or trauma that confer risk for NSSI, but also the reaction of family members, similar to the notion of an invalidating environment (Crowell et al., 2009; Linehan, 1993). Similarly, Sim and colleagues (2009) found further support for a model whereby emotional regulation partially mediated the relationship between invalidating family climate and NSSI, but only for girls. These researchers suggest pathways between unsupportive social context, emotion regulation, and NSSI may differ by gender. Correspondingly, they suggested that family cohesion, as a positive factor, would be associated with adaptive emotion regulation behaviour and therefore with less distress or greater emotional health. Bureau and colleagues (2010) also found gender differences in their examination of perceived dimensions of parenting in a sample of young adult students. In this study, poor communication with parents and feelings of fear and alienation in relationship were only predictive of NSSI for females. Across gender, reporting better emotional health was the most powerful protector against NSSI. That is, those adolescents who reported less frequent stress/pressure, nervousness, sadness/hopelessness/discouragement, and fewer pains and/or health-related fears over the past month, in other words, those with a greater sense of well-being, were far less likely to have engaged in NSSI. This finding lends support to the supposition that selfinjury functions as an emotional regulation or experiential avoidance strategy (i.e., to avoid/escape, reduce aversive or overwhelming negative affect). Despite the fact that the survey items ask about 30-day emotional status and is therefore limited by temporal quality, it is likely that the youths‘ responses reflect a more enduring pattern of emotional functioning. Recent studies with both adolescent and adult samples investigating emotional valence, affective arousal, and function, have shown that NSSI temporarily reduces negative affect (e.g., stress, anxiety, loneliness, depression) and increases positive affect (e.g., relief, focussed, calm) (Klonsky, 2007; Klonsky, 2009; Laye-Gindhu & Schonert-Reichl, 2005; Nixon  126 et al,. 2002; Nock, 2010). The common reasons identified by the sample youth for their last incident of NSSI (e.g., loneliness, depression, stress, anger) further underscore this finding. More research is needed to elucidate the relations among aspects of emotional health, positive affect, and NSSI. The second most powerful protective factor for both males and females, connections to family, represents feelings of being cared for by family/caregivers, feeling close, having fun together, satisfaction, and perceived quality of relationship. Previous research and theory have suggested that family relationships and specific dimensions within them may influence the development and maintenance of NSSI, but this has had limited direct empirical evaluation (see Bureau et al., 2010; Crowell et al., 2008; Hilt, Nock, et al., 2008; Yates et al., 2008). Little research has specifically examined the positive functioning of families (e.g. connectedness), but a significant body of research has examined family dysfunction and suicide and a smaller body has looked at family dysfunction and more general self-harming behaviours. Findings from this first study to explicitly examine connectedness as a protective factor lend further support to the notion that improving relationships with caregivers may be a key avenue for prevention and intervention (Bureau et al., 2010; Hilt, Nock, et al., 2008). Especially given the developmental status of this particular group of young people, the power of positive relationships to influence the developmental trajectory must not be underestimated. In their observational study, Crowell and colleagues (2008) used an ecological framework to assess the interplay between biology and environment and NSSI and suicide. Their findings suggested a relation between the neurotransmitter serotonin and the expression of affect in parent-adolescent dyads such that protection offered by higher levels of peripheral serotonin is moderated by the quality of parent-child interaction. According to these researchers, this relation may be bidirectional, hinting at a more general dysregulation within the parent-child relationship. The quality of attachment relationships with parents has  127 been found to relate to NSSI, although most reports have been retrospective (Gratz et al., 2002; Yates et al., 2007; Walsh, 2006). Two recent longitudinal studies (Yates et al., 2007; Hilt et al., 2008) found that adolescents who self-injured reported poorer quality of relationship with parents and increased alienation. Moreover, in a study that examined differential relations with caregivers, Hilt, Nock, and colleagues (2008) further found that adolescents who self-injured reported an increase in positive relationship quality with fathers. However, this study was not prospective and therefore it is difficult to interpret this finding. A majority of street-involved youth report poor relationships with their caregivers, regardless of whether they self-injure (Haber & Toro, 2004; McMorris et al., 2002; Ryan et al., 2000). Together with previous research, results from the current study suggest the value of examining the nature and quality of relationships between adolescents and various attachment figures. Further research disentangling dimensions of interpersonal relationships to better understand their role in NSSI is recommended. Consistent with the present study results regarding positive orientation toward future and current circumstances for females specifically, in their recent study of NSSI among over 1700 early adolescents, Alfonso and Dedrick (2010) found that maintaining an optimistic belief that the future was full of possibilities showed a small protective effect against selfinjury, reducing the odds by one-third. They did not report on these findings by gender. Relatedly, in investigating adaptive attitudes and beliefs, other researchers have highlighted the construct ‗reasons for living‘ as a factor that differentiates youth who are at higher risk for suicide. Muehlenkamp and Gutierrez (2004, 2007) compared youth with a history of NSSI to youth who also had a history of suicide (NSSI+suicide) and found that the former reported more positive attitudes toward life and a greater number of reasons for living relative to youth reporting suicidality. Consistent with the findings for females in the present study, these researchers suggest that maintaining a more positive outlook on life as well as being able to identify future positive goals may be protective against NSSI and that the converse may be a  128 risk factor, similar to hopelessness. Research examining the influence of positive present and future attitudes and beliefs on NSSI may help to clarify gender differences and provide more direction for preventive intervention. It is intriguing that despite the fact that 60% of the adolescents were attending school at the time of the survey, none of the school variables differentiated between youth who selfinjured and those who did not. Neither future educational aspirations, feelings of connection to school (i.e., feelings of belonging, being supported, positive relations with teacher), nor continued attendance at the time of survey offered any protection. Even a closer look at only the youth who were attending school failed to find any difference in overall feelings of connection to school and NSSI. Notably, the full six-item school connectedness scale, as validated with the typical school-based population in previous adolescent health surveys (e.g., MacKay, 2007; Saewyc & Homma, 2010), did not show good fit for the street-involved sample; rather, a five-item version that did not include an item about difficulty getting along with teachers showed better fit. The school connectedness scale, devised and validated previously for normative school-based population health research did not perform the same with this nonnormative sample. As a primary socialization agent along with caregivers, school can be an important locus of identification and/or intervention and it may be that as yet untapped aspects are protective against NSSI for this particular group of young people. For example, other potential benefits or protective aspects of school for street-involved youth could include access to survival resources (e.g., showers, food, bathrooms), peer factors (e.g., support, belonging/affiliation, status), adult support (e.g., relationship with significant adult), and/or simply the predictability, stability, or conventionality involved. Further it is unclear whether attending actually means participating and going to class or whether youth really are merely visiting the building for whatever benefits it provides. Given that many of the youth were attending school (or had started the year attending), it may prove useful in future to understand the youth perspective on why and what kept them coming back.  129 No studies have investigated the role of school variables and school relationships on the development and maintenance of NSSI. Future research should seek to better understand the nature of the educational experience, both positives and negatives, for streetinvolved youth who self-injure, with the goal of locating dimensions that may prove protective (e.g., adult support, programming options, school characteristics). These youth show remarkable strength for managing to participate in normal developmental experiences such as formal schooling and it behooves us to understand how to maximize this opportunity to connect, engage, and alter the trajectory toward better health and well-being. Current study findings did not support the hypothesized relations between body image and weight satisfaction and NSSI. Studies have reported a positive association between poor body image and maladaptive eating habits and NSSI (e.g., Brunner et al., 2007; Hilt et al., 2008; Murray & Fox, 2005). However, studies have not examined body satisfaction as a protective factor. In this street-involved sample, these variables may be complicated by survival issues, food security, and by other weight and body image factors (e.g., influence of substance use). Moreover, the survey questions may not have been explicitly reflective of true body satisfaction. Similarly, results diverged from literature by finding no protective effect for participation in supervised extracurricular activities or what Leffert and colleagues (1998) refer to as ‗constructive use of time.‘ Compared to youth who did not self-injure, males who did were just as likely to participate in these activities and females were even more likely to report extracurricular involvement. The current study did not examine specific types of activities (e.g., sports, arts, organized groups) or duration (e.g., Leffert et al. suggest 13 wks per year) for their independent associations. Further, the survey question referred to prestreet involvement specifically, requiring adolescents to distinguish between their current and previous situations, but the item did not provide any time markers, meaning that youth may have responded affirmatively even if they were very young when they participated.  130 Risk Factors In contrast to the relatively fewer significant protective factors for NSSI, many factors showed themselves to increase the odds of self-injury. Some of the associations are wellsupported in the literature whereas others represent an extension or new direction or do not conform to expectation. For males, twelve out of fifteen risk factors were clinically and statistically significant (at least doubled the odds of NSSI, p ≤ .05) at the bivariate level. Risk factors that distinguished between those who self-injured and those who did not included the following: a range of victimization variables (history of sexual abuse, physical abuse, being physically assaulted); having a father with multiple problems (mental health, substance abuse, criminal activity); family history of suicide; experiencing higher level of relational victimization; engaging in physically risky behaviour such as sharing drug/tattoo gear/razors; intentional vomiting after eating; experiencing more consequences of substance use; and attempting suicide in the past year. In particular, the three strongest and most salient risk factors that were linked to NSSI for males even in the presence of other risk factors were history of suicide attempt within the past year, physically risky behaviours such as sharing needles or other tools, and experiencing a higher level of relational victimization. When present, each of the three risk factors increased the probability of self-injury as an outcome by approximately three-fold. For females, nineteen out of twenty hypothesized risk factors at least doubled the likelihood of NSSI at the bivariate level. Variables included all victimization experiences, with the exception of sexual abuse (OR < 2.0), as well as living in precarious housing, mother and father having multiple problems, family history of suicide, experiencing more consequences of substance use, physically risky behaviour, vomiting after eating, frequent binge-drinking, physical fighting, and having attempted suicide in the past year. Only five of these nineteen factors retained significance when examined in a multivariate model in which other variables were controlled for. This suggests that there may be a common or indirect relationship  131 among variables rather than a unique relationship. The most robust factors were past year suicide attempt, physically risky behaviour (sharing drug needles or sharp implements), experiencing consequences for substance abuse, having a mother with multiple problems (substance, mental illness, criminal behaviour), and being the victim of sexual abuse by two or more perpetrators. All five risk factors increased the odds of self-injury by at least threefold. Across genders, the most powerful risk factor for NSSI was past year history of suicide attempt, increasing the odds of self-injury by six-fold for females and nearly four-fold for males after controlling for other factors. Consistent with previous research highlighting the strong correspondence between nonsuicidal injury and suicidality (e.g., Brausch & Gutierrez, 2010; Muehlenkamp & Gutierrez, 2004, 2007), nearly half of the self-injuring youth reported a past-year suicide attempt (compared to 8% of the non-injuring youth) and the vast majority of youth (80%) with a suicide attempt reported NSSI. These numbers are higher than those typically found in less vulnerable groups of adolescents and are reflective of the high morbidity among street-involved and homeless youth. Although self-injury is far more common than suicide, predictive of future suicide, and indicative of serious distress and coping deficits, it has not been given the serious attention in the literature that it deserves (Brausch & Gutierrez, 2010; Muehlenkamp & Gutierrez, 2007; Nock et al., 2006; Schwartz et al., 1989; Stanley et al., 2001). Extant research suggests the presence of common risk factors for NSSI and suicide; future research should work to identify and disentangle the unique effects. The present study showed alarming rates of overall victimization as well as a range of victimization experiences that independently contributed to increased risk for NSSI. Given the cross-sectional nature of the study, the temporal relations among different victimization experiences, the sequence of perpetrators (e.g., peers versus family members versus stranger adults), and NSSI is unknown. Researchers have posited that an invalidating  132 environment during childhood, including traumatic experiences such as neglect, abuse, and attachment disruption, is a strong risk factor for NSSI (Chapman et al.,. 2000; Crowell et al., 2009; Linehan, 1993; Van der Kolk et al., 1991, 1994; Yates, 2004). In the current study, a majority of individual victimization experiences were significant at the bivariate level for both males and females (with the exception of sexual exploitation), however when considered in multivariate models, fewer remained clinically and statistically significant. Given the disproportionately high level of victimization in this street-involved sample in contrast to ―normative‖ adolescent samples, these experiences may not be as salient as risk factors or we may need to assess more sensitive or specific aspects or pathways. In their studies with street-involved young adults, Tyler and colleagues (2010) found that childhood sexual abuse was indirectly associated with NSSI through on-street sexual victimization. Past research has focused on maltreatment history as a risk factor for NSSI but recent research (e.g., Gratz et al., 2002; Van der Kolk, 1991, 1994; Yates, 2004), including a metaanalysis of 43 studies (Klonsky & Moyer, 2008), has been equivocal about the relations between various types of abuse and NSSI. Whereas some research has found that sexual abuse was associated with increased risk for NSSI, other studies have shown greater association with physical or emotional abuse (Gratz et al., 2002; Whitlock et al., 2009; Yates et al., 2008). Klonsky and Moyer (2008) found that the association between sexual abuse and self-injury was attenuated when other relevant variables were controlled (e.g., psychological risk factors), however the relationship with physical abuse was stable. This led them to propose that sexual abuse specifically could be conceptualized as a general proxy risk factor for NSSI, meaning that both sexual abuse and NSSI are associated with the same risk factors, and that childhood sexual abuse might be related to self-injury through mediating variables rather than having a primary role in the initiation and maintenance of the behaviour (Kraemer, Stice, Kazdin, Offord, & Kupfer, 2001).  133 Results for females from the present study showed that being sexually abused by multiple perpetrators increased the odds of NSSI even in the presence of other risk factors. Recommendations for future research include consideration of other abuse characteristics that may be more sensitive and salient in predicting self-injury, for example the frequency, role of the perpetrator, number of perpetrators, nature of the abuse, developmental timing, temporal relation, or sequelae or potentially mediating variables such as dissociation, posttraumatic stress, or other psychological characteristics. Systematic studies focusing on the developmental progression from abuse to NSSI as well as the temporal relations among different victimization experiences, for example those prior to the street-involvement and those subsequent to it, should further our understanding and inform theory. Additionally, various abuse types or parameters of abuse may differentially relate to different parameters of NSSI (e.g., frequency, recurrent versus intermittent, methods, locations, intrapersonal/automatic versus interpersonal/social functions). Contrary to expectation, the experience of having traded sex for money, shelter, or goods failed to distinguish between adolescents who self-injured and those who did not. Studies with homeless youth have found an association between suicide and sexual exploitation or involvement in the sex trade (Greene et al., 1999; Kidd & Kral, 2002; Kidd, 2006). However, the present study finding mirrors the results from a study by Tyler and colleagues (2003) who found that trading sex was not significantly associated with NSSI in their sample of older homeless adolescents, although they refer to other ―deviant subsistence strategies‖ that predicted the behaviour, namely shoplifting, drug-dealing, and criminal activity. Evidence from extant research indicates that youth with previous histories of abuse face increased likelihood of revictimization through sexual exploitation (Greene et al., 1999; Kipke et al., 1997). It may be that in the context of other strong risk factors, including victimization experiences, trading sex is heavily confounded or does not confer additional risk for NSSI.  134 For males, having experienced a greater level of relational victimization in the year prior was a strong risk factor for NSSI, even when other factors were controlled. Of males who reported engaging in NSSI, 14% indicated that their last incident was due to feeling rejected. Unfortunately, the survey items do not differentiate between peers or others. However, in the present study, having better peer relations (i.e., infrequent or no trouble getting along with peers at school) was found to protect against NSSI in the multivariate model for males only, reducing the likelihood by 50%. A paucity of research has examined social functioning of adolescents who engage in NSSI, although issues of social contagion and social reinforcement have been discussed (e.g., Heath et al., 2009; Hilt, Cha, et al., 2008; Nock & Prinstein, 2004, 2005). Relational victimization has been linked as a risk factor to adjustment difficulties in children and adolescents (Cullerton-Sen & Crick, 2005; McLaughlin, Hatzenbueler, & Hilt , 2009), and to increased emotion dysregulation among both adolescent girls and boys (McLaughlin et al., Hilt, 2009). Future research focusing on social influence and social contexts including social relationships and experiences of adolescents who selfinjure may help deepen our understanding of pathways and provide links to prevention. Street-involved youth, often marginalized and victimized, and who tend to maintain unconventional identities and social groupings, may have much to teach us about social affiliation, social exclusion, and NSSI. The two strongest risk factors for both males and females – engaging in suicidal behaviour and sharing needles, razors, or tools – are alarming because of their serious public health implications. The risk involved in explicit suicide attempts is clear and there is significant evidence in the extant literature to support the strong link between suicide and NSSI (Brausch & Gutierrez, 2010; Muehlenkamp & Gutierrez, 2004, 2007; Stanley et al., 2001). In contrast, physically risky behaviour, in the form of sharing needles or gear for drugs and/or sharing tattoo/piercing equipment or razors, increased the likelihood for NSSI threefold in multivariate analysis for both males and females, but has not been previously  135 examined in the context of NSSI. Indirect self-harm behaviour such as physical or situational risk-taking has been shown to be associated with NSSI (Laye, 2002; Walsh, 2006; Zweig et al., 1991). The physical risk involved in sharing tools in the present study may be related to these types of risk taking behaviour. Results from a study with adolescents in a psychiatric inpatient program found that over one-quarter of those who engaged in NSSI also shared cutting tools (DiClemente et al., 1991) and other research with street youth has found that about one-third share needles (Lloyd-Smith et al., 2008). In their study, Lloyd-Smith et al. identified relevant social and structural barriers to obtaining clean needles. Further research to examine issues of location and access, self-care and beliefs and attitudes about health, social influence and social meanings of sharing, and related health risks would inform understanding of this finding and provide some context. Interestingly, injection drug use, hypothesized to be a risk factor for NSSI in this study, was no more likely among youth who self-injured. This is in contrast with findings of McBride and colleagues (2001) in their exploratory qualitative study of 24 injection drug using adults who reported similarities between the two behaviours in terms of pain and relief mechanisms. These researchers were interested in examining ―needle fixation‖, a psychological construct whereby the administration of the drug becomes reinforcing through the ritual aspects of injection, as well as secondary gain through a conditioning process. No other studies have examined shared drug gear and/or injection drug use and self-injury. As expected and consistent with prior research demonstrating the co-occurrence of problem or health-compromising behaviours (Jessor, 1991; Jessor et al., 1995) and the overlap in NSSI and indirect self-harming behaviours specifically (Alfonso & Dedrick, 2010; Hilt, Nock, et al., 2008; Laye-Gindhu & Schonert-Reichl, 2005; Nock, 2010; Serras et al., 2009), current study results revealed that several behavioural correlates increased the odds of NSSI. For females, experiencing more consequences of substance use, a psychosocial risk factor, increased the risk of NSSI more than four-fold, even at the multivariate level.  136 Other risk behaviours associated with NSSI in the present study included: vomiting on purpose, frequent binge-drinking, and fighting, particularly among females, although these relations did not remain at the multivariate level. This suggests that the variables may have an indirect relationship with NSSI or may be related through a third variable. Interestingly, a novel study using real-time data collection (electronic diaries) found that adolescents experiencing self-injuring thoughts reported simultaneously having thoughts of using substances and engaging in bingeing and purging behaviours 15-35% of the time (Nock, Prinstein, & Serba , 2009). These behaviours share common elements; in addition to having the potential to cause physical harm, they also represent efforts to modify the affective, social, and/or cognitive experience, and they are highly reinforced. Probability Profiles Consistent with risk and resilience theory, present study findings demonstrated not only the powerful influence of risk, but also the ability of protective factors to mitigate risk. These results revealed that although risk is endemic in the lives of street-involved adolescents, and they are at higher risk for self-injury with lower levels of protective factors, when the latter are present they clearly reduce the odds of NSSI. In the current study, this was particularly true at all levels of risk for males and at low to medium levels of risk for females. Although no other studies on NSSI have used this methodology, the findings are consistent with those found in studies of risk and protective factors for suicide that demonstrate the decrease in risk for that outcome through the introduction of protection (e.g., Borowsky et al., 2001; Erickson, 1999; Poon et al., 2006; Rubenstein et al., 1989). For males, I combined the three most predictive risk factors (physical risk-taking, suicide attempt, and experiencing greater level of relational victimization) and the three strongest protective factors (family connectedness, better emotional health, and having less trouble getting along with peers) plus age into profiles of self-injury. When the young men  137 had all three risk factors and none of the protective factors, the probability of NSSI was a high 90%. With no risk factors and all three protective factors, the probability of NSSI decreased ten-fold to 9%. In contrast, in the profile with no risks and no protective factors, the chances of NSSI increased to 39%, four times greater than in the presence of protective factors. The influence of protective factors was stronger at lower levels of risk (one or two factors compared to three). Even with two risk factors, with the addition of protective factors the probability dropped by half, from 81% to 40%, or from 73% to 30%, or from 82% to 41%, depending on which combination of risk was involved in the profile. For females, I first combined the strongest three risk factors (suicide attempt, physical risk-taking, and experiencing more consequences of substance use) with the only protective factor (better emotional health or low distress) that emerged in multivariate models. At high levels of risk (all three risk factors) this protective factor offered no protection from the likelihood of self-injury (98% to 97%). At lower levels of risk (one risk factor), the introduction of just one protective factor reduced the likelihood of NSSI by 15% for all three combinations. At two risk factors, the impact of the protective factor is already small (4-8%). Even among females without any of the three risk factors from the model, the presence of the protective factor decreased the risk of NSSI, from 35.3% (with no protective factors) to 21.8% (with one protective factor). Given the possibility of confounds in the protective factor model for females, the restriction on temporality (limited by 30-day distress variable), and the evidence-supported hypothesis of the importance of family connectedness in youths‘ lives, a further multivariate model was examined with the combination of the same three risk factors and two protective factors (family connectedness and subjective health status) that emerged when the low distress variable was not used. When the young women had all three risk factors and none of the protective factors, the likelihood of self-injury was almost certain (99%). With no risk factors and no protective factors, there was still a 43% chance that the youth would self-  138 injure. Notably, with the introduction of both protective factors, this probability is more than halved (19%). In this model, again the introduction of protective factors at high level of risk (two or more factors) did less to buffer against the risk, reducing likelihood of NSSI by 7-13%, depending on the profile. However, at lower levels of risk with one risk factor, the likelihood of NSSI is significantly reduced by the addition of protective factors, from 67% to 39%, from 83% to 62%, and from 81% to 58, depending on the specific profile. Strengths and Limitations of the Study The present study extends previous literature on NSSI in four major ways: (a) by incorporating a theoretical perspective and specifically one that emphasizes a dual focus on both risk and protection within an ecological framework (i.e., both individual and contextual dimensions considered); (b) by exploring both bivariate and multivariate relations among variables for independent contributions as well as interplay among and between protection and risk in the prediction of self-injury; and (c) by using a unique methodology that highlights the interaction of risk and protection and provides useful information to inform clinical practice; and (d) by examining NSSI in a large sample of vulnerable street-involved adolescent boys and girls, with separate attention to factors by gender. These strengths will be further explicated. In contrast to a traditional deficit model with an emphasis on risk, a dual approach derived from the study of resilience features an in-depth focus not only on risk but also on factors that promote health. Further, using an ecological framework (Cichetti & Lynch, 1993; Fergus & Zimmerman, 2005) to examine factors across multiple domains, namely individual and contextual, including personal, family, school, peer, and community, more closely approximates the multiple influences at work in youths‘ lives and supports the evidence for the multi-determined and etiologically heterogeneous nature of NSSI. This approach is grounded in a more realistic and holistic view of adjustment. Although street-involved youth  139 typically live high-risk lives in terms of background and street-level factors, not all self-injure. While furthering our understanding about the risks, it is also important to broaden and deepen our understanding of the multiple factors that reduce the likelihood of this outcome. Risk and protection represent powerful tools for developing the knowledge base on NSSI and for guiding the development of effective identification and intervention. A thorough search of the literature suggests that, to my knowledge, this is the only study to include an explicit examination of multiple protective factors for NSSI in a community sample of adolescents. Overall, scant research has been conducted on protective factors for NSSI; in fact, much of the knowledge base is derived from the study of suicidal behaviour or anecdotal accounts and case studies. Extant research on other health-compromising behaviours and negative outcomes has demonstrated the power of protective factors to mitigate risk and/or to promote health (e.g., Bearinger et al., 2005; Bearinger et al., 2010; Blum et al., 2002; Borowsky et al., 1999; Erickson, 1999; Poon et al., 2006; Rutter, 2000). Research has suggested that the presence of general protective factors such as connectedness and social support may buffer against a range of risks to positively influence outcomes (e.g., Klonsky & Glenn, 2009b; McNeely et al., 2002; Resnick et al., 1997), but we know little about how they are related to NSSI specifically, nor do we know about how they hold up in relation to other factors. We have limited understanding of any specific or unique predictors that protect against NSSI. An improved understanding of protection is particularly salient in this high risk sample of young people. Whereas many common factors or risk correlates of NSSI have been identified, much less is known about unique predictors. We know that all factors are not equal in altering the developmental course. This study represents an attempt to identify the risk and protective factors that are especially influential and to generate estimates of the magnitude of the relationship among these factors and self-injury. Information about which factors remain salient even when considered in combination with others yields understanding that can be  140 utilized to inform our approach in developing effective prevention and intervention efforts and in deciding who should be targeted. Examination of indirect effects, while a very worthy endeavour, is beyond the scope of the present investigation. A methodological strength of the current study lies in the use of probability profiling to analyze various combinations of robust risk and protective factors and history of NSSI in the sample of street-involved adolescents. Through the use of a systematic sequence of steps involving logistic regression analyses, probability profiles, providing a single probability of the likelihood of NSSI for an individual youth with that particular profile or combination of factors, were developed. This statistical technique permits the consideration of multiple predictors or influences simultaneously. In this study, the methodology highlights and graphically represents the multiplicative effect of risk and the accumulated impact of protective factors, at least to a certain level of risk. Given the high risk status of these street-involved youth, the fact that the protective factors actually function in interaction to reduce the risk is notable. As such, this practical approach can improve the assessment of vulnerability and guide efforts toward risk reduction and protection promotion. The relatively large sample size and the systematic sampling of a considerable number of young men allowed for testing of a greater number of variables as well as examination of gender variation. Recent studies with school- and university-based samples of adolescents and young adults have included males and found some gender differences, however NSSI has a long history of reporting on small clinical or inpatient female-only samples. Gender differences in psychopathology, risk behaviour, victimization, suicidality, and street-involvement, together with limited evidence from these community samples and recommendations from the literature (e.g., Gratz, 2003), suggested the value of examining separate gender models of NSSI.  141 There are a number of limitations to this study that must be noted. First, the crosssectional nature of the study makes it impossible to assess temporality and therefore precludes any determination of direction of influence. With this correlational design, all variables were measured at a single point in time and therefore, although the term ―predictive‖ is used, there can be no causal inference. The survey is retrospective, asking about feelings, behaviours, and experiences within specific time-frames, including the past 30-days, past 12 months, and lifetime (―ever‖). The self-injury item reflects ever having injured and, as such, it is impossible to determine the chronology of these different factors in the adolescents‘ lives. Prospective and longitudinal studies that examine NSSI among street-involved or other samples of high-risk adolescents are needed to uncover the temporal relations and causal associations. As well, information obtained on onset (and offset, if applicable) of behaviours and chronology of events would prove useful in furthering understanding of the pathways to NSSI and the trajectory of the phenomenon. Second, this research is limited by the validity of single item measures, including reliance on a single question to assess NSSI. However, as in other adolescent health surveys and other research on self-injury, single item measures are frequently a necessary method to efficiently collect information about a variety of health-related factors. Findings based on single item measures, including asking about NSSI, may under-represent the construct. Also, a single item response is insufficient to capture the true complexity of constructs. It is also possible that single-item responses, while increasing face validity, introduce measurement error and reduce reliability, although public health surveys are well respected and their use has been well-supported in psychometric investigations. The third limitation concerns the reliance on self-report data. With collection of data form a single source, the veracity of the responding must be questioned and common method variance may result in either over- or under-reporting as the youth may either intentionally or unintentionally provide inaccurate responses. On the other hand, for sensitive questions or  142 individual feelings, opinions, and perceptions, self-report is often the only method possible and the best method for accessing this information, especially during adolescence. The anonymous nature of the survey also likely increased participation. Moreover, street-involved youth may be less affected by a social desirability response bias as they typically live a marginalized and unconventional lifestyle, they may not be as inclined as their non-involved counterparts to either minimize or embellish their behaviour. Fourth and fifth refer to the use of the survey specifically. The use of survey data frequently entails attenuation or reducing categorical variables to fewer categories. When variables are dichotomized, information is loss and error is increased. The risk involved in doing so is referring to as ―binning‖ whereby the variability of a variable is limited. For the current study analyses, this was a necessary step. A final and related limitation refers to the challenges inherent in using secondary data whereby variables must be derived from the existing survey and sample. Although the survey features items along both risk and resilience dimensions, the measures of protection were fewer and not as strong, especially in the face of such high risk. The number of items included on a survey is typically fixed or limited and McCreary Centre Society‘s surveys go through a rigorous selection process. In evaluating NSSI, if would be a useful adjunct to include multiple items to measure both the presence of NSSI and characteristics. Future research would benefit from a focused and systematic approach that is not constrained by the data collection tool. The addition of items measuring a wider range of solid protective factors (e.g., peer support, help-seeking, selfesteem, cultural connection) is recommended. Finally, the findings may not be generalizable to the non-street-involved population of adolescents. Future Directions for Research Future research may seek to clarify the relations among risk and protective factors and NSSI found in the study and to evaluate the generalizability of the findings with other  143 community samples of youth and with gender-mixed samples. Much of the current knowledge comes from biased sampling with females and therefore we are only beginning to examine gender differences in NSSI. Current findings suggest the value of ongoing examination of gender differences in parameters of NSSI and risk and protective factors. Sociodemographic results revealed heightened risk of NSSI among sexual minority youth, but further examination of this relation was beyond the scope of the current investigation. Taken together with other research on negative outcomes for GLB youth (e.g., Austin et al., 2004; Austin et al., 2007; Garofalo et al., 1999; Russell & Joyner, 2001), it is clear that more research is needed to better understand the complexity of the relations between and within sexual minority status, risk and protective factors, and NSSI. Results suggest the value of continuing to identify protective factors for NSSI. Limited research has investigated protective factors specifically and findings from the current study suggest that the resilience and positive psychology knowledge bases may have something to offer to the study of NSSI. Future studies may seek to identify dimensions of interpersonal relationships that offer protection from NSSI both within the family, school, and community settings. Given that not all street-involved youth engage in NSSI despite their challenging background and the trauma of street-life, attention may be directed toward identifying intrapersonal resilience factors and interpersonal resources among youth who do not selfinjure. Prospective and longitudinal studies are needed to identify the temporal relations among risk factors and NSSI. Additionally, studies using methodologies such as structural equation modelling would allow examination and clarification of the nature of relations among variables, including those that exhibit indirect influence on NSSI. These modelling techniques could be used to improve understanding of the relations among risk and protective factors, including family connectedness, emotional health, emotional distress, and NSSI both for males and for females. Due to the concordance of risk factors in street-involved youths‘  144 background and in those found in the literature for NSSI, modelling these relations in street youth or other high risk adolescent samples may elucidate important relationships that would translate to preventive interventions. Current study findings of covariation of risk behaviour (see Jessor, 1991) highlights the need for further research examining the specific mechanisms involved with the function of these behaviours. Examination of their associations and temporal relations may help to improve understanding of why one is enacted at one particular time and why another behaviour may be enacted under different circumstances. Longitudinal studies examining the onset of these behaviours, their relation to NSSI, and their differential functions may prove useful in delineating the relations especially for high-risk adolescents who may engage in many risk behaviours. Findings highlighting common functions would suggest implications for assessment and intervention. Studies using methods such as structural equation modelling to investigate the nature and structure of co-occurrence may help us answer questions such as: Do the behaviours serve different functions or do they work through different mechanisms? Why does an individual choose one behaviour over another? What is the role, if any, of symptom substitution? Can studying these related risk behaviours or indirect harming methods (those where the harm may be secondary) provide insight into why adolescents engage in NSSI? More research is needed to clarify the relations, mechanisms, and functions among these different behaviours both for males and for females. Implications for Clinical Practice This study represents the first attempt to identify a wide range of risk and protective factors for nonsuicidal self-injurious behaviour within a large sample of street-involved adolescents. The clinical and public health significance of NSSI among adolescents, and especially among those whose unique social experience puts them at disproportionate risk for a variety of negative outcomes, is indisputable. The results of the current study suggest that  145 a considerable number of street-involved adolescents engage in NSSI and that their health needs are complex, with a range of personal and contextual factors that serve to either increase or decrease the risk of self-injury. Current findings further demonstrate that even within this group of vulnerable and marginalized youth, some groups, namely GLB-identified adolescents, are at even greater risk for NSSI, while other groups, such as Aboriginal youth, are not. Females in the study were also more likely than males to report NSSI, but in general research findings have been mixed on gender differences and further research is needed. Results provide some direction for avenues of prevention and targeted intervention, and suggest that both males and females should be included in any efforts, with an awareness of gender differences in risk and protective factors. Findings of increased likelihood of NSSI for specific groups such as females or sexual minority youth beg the question of whether we need to tailor interventions to meet the specific needs of these demographic groups. Results highlight the importance of assessing adolescents who are street-involved for both NSSI and suicidality as well as the need for primary prevention efforts among groups found to be at elevated risk for the behaviour. Despite the prevalence of nonsuicidal self-injurious behaviour, early identification and assessment occur infrequently. Results help to counter the longstanding myth that the behaviour is primarily attention-seeking, with adolescents indicating a range of reasons, most being related to attempts to regulate negative affect (e.g., lonely/depressed, stressed). An enhanced understanding of the factors that may increase the likelihood of NSSI and those that might buffer against risk or decrease the likelihood may aid in the identification, assessment, case conceptualization, and treatment planning of adolescents who self-injure. Results also highlight the need for more resources for children with parents who struggle with mental illness, substance/alcohol abuse, and/or who are criminally involved, all experiences that generate significant family stress and additional risk, tax coping resources, highlight parent skills deficits, and can disrupt the attachment process with implications for the  146 development of emotion regulation and coping skills. Similarly, given the high level of victimization in the sample and the salience for females of being sexually victimized by multiple perpetrators, efforts must be devoted to primary prevention and early intervention for abuse and exploitation as well as to building protective factors to mitigate the vulnerability conferred by these experiences. With overlap between parental psychopathology, substance abuse, criminal involvement, and child maltreatment, programming that supports families and enhances family resources is critical. Further, particularly for families or children/youth ‗atrisk‘, fostering connections and help-seeking behaviour may be other components in the prevention and early intervention arsenal. While nonsuicidal self-injury and suicide differ in important ways, there is considerable overlap between the behaviours. Consistent with the literature, NSSI and suicide attempt were strongly associated, with many adolescents reporting both. A history of suicide attempt markedly increased the likelihood of NSSI. Although positive present and future orientation did not remain a significant protective factor in the multivariate model as it was at the bivariate level, research has shown that compared to suicide, NSSI is associated with more attraction to life and belief in possibilities (Alfonso & Dedrick, 2010; Muehlenkamp & Gutierrez, 2004). Further research should help delineate the relation of this positive psychology variable to NSSI. Cultivating a positive orientation may be an important component of intervention for street-involved youth. Taking this one step further, we can speculate that by reducing NSSI, the suicide rate will also decrease. The finding of an association between NSSI and sharing needles, tattoo/piercing tools, or other sharp implements has serious health implications. It is not known whether this behaviour is due to access issues, lack of self-care or motivation, need or desire for additional risk, or peer norms. However, both males and females who reported more sharing had significantly increased likelihood of self-injury compared to those who did not share. In collaboration with youth, a harm reduction approach may be helpful in preventing any  147 physical health complications related to sharing. Although not explicitly assessed in the current study, in assessment with youth, it may be worthwhile to ask about sharing implements used specifically to injure as well, above and beyond any questions about sharing needs or other tools. Results substantiate the value of assessment, programming and service delivery that includes a focus on protective factors. Two of the most influential protective factors among both males and females, greater emotional health and stronger connections to family, highlight the importance of attachment and childhood socialization of emotion in the context of family or the caregiving relationship (Eisenberg et al., 2010; Morris et al., 2006; Zeman et al., 2006) and link to the functional understanding of NSSI as an emotional regulation or communication strategy in the absence of more adaptive skills. Results suggest that interventions aimed at building distress tolerance and emotion regulation skills as well as strengthening relationships with important caregivers may be critical. Although few prevention or intervention approaches for reducing NSSI have been empirically assessed and none have been established as effective, present findings are consistent with evidence supporting the use of dialectical behaviour therapy (DBT) with adolescents who engage in NSSI (Chapman et al., 2006; Crowell et al., 2009; Linehan, 1993). An important but flexible component of DBT involves skills training groups that are meant to have adolescent-caregiver participation, thereby strengthening family connectedness, family emotional functioning and emotional well-being, including developing adaptive regulation strategies. However, this model, if delivered with all components intact, is intensive and requires significant commitment from both youth and their caregivers, making it more challenging to provide such services to street-involved youth. Importantly, DBT intervention can be implemented with youth both in individual and in group modalities, regardless of caregiver participation. Effectiveness trials of DBT with both community samples of adolescents and with street-involved youth will be an important next step.  148 Similarly, there is no evidence-based ‗gold standard‘ for intervention with streetinvolved youth and there is a dearth of outcome research in general. Little research with family-based or ecological intervention has been conducted. Because family dysfunction is highly correlated with street-involvement and family connectedness has been found to prevent and buffer against risks, including NSSI, this may be a fruitful avenue of intervention for some youth. A careful and thorough assessment can help to determine which youth may benefit from family intervention (Kidd, 2003b). Other researchers (Bender et al., 2007; Kidd, 2003b) have focused on the individual strengths, abilities, and resources of street-involved youth (e.g., positive future outlook, resourcefulness, survival skills), suggesting that they are effective targets of programming and represent important points of engagement with these youth. Slesnick and colleagues suggest the value of comprehensive integrated interventions that address multiple overlapping needs rather than those focused on a single risk behaviour (Slesnick, Dashora, Letcher, Erdem, & Serovich, 2009). Intervention studies that strategically strive to bring together risk and protective factors evaluated in the current study may have clinical implications for NSSI intervention. For example, a recent quasi-experimental intervention study with a one-year follow-up period shows promising results for promoting healthy development among high risk sexually exploited adolescent runaway girls (Saewyc & Edinburgh, 2010). Specifically, researchers found that girls participating in the strength-based intervention showed significant and sustained improvement in protective factors including family connectedness, school connectedness, perception of being cared for by other adults, and self-esteem, and significantly reduced risk behaviours including suicidality, substance use, risky sexual behaviour, and lower emotional distress. Despite high risk and low protection initially, after intervention, the participants were similar to nonabused peers on these variables. Most interestingly, results showed that the intervention had the strongest positive effects for the girls with the highest risk and the lowest connectedness prior to intervention. This study  149 offers important insights for working with street youth from an ecological strength-based perspective, underscoring the salience of a dual focus (i.e., fostering protection and reducing risk) and highlighting the value of both interpersonal connections and intrapersonal strengths in altering the developmental trajectory of vulnerable youth. Systemic and longitudinal intervention and outcome studies are clearly important to advancing the knowledge base and providing evidence of what works.  150 REFERENCES Adrian, M., Zeman, J., Erdley, C., Lisa, L., & Sim, L. (2011). Emotion dysreguation and interpersonal difficulties as risk factors for nonsuicidal self-injury in adolescent girls. Journal of Abnormal Child Psychology, 39, 389-400. doi:10.1007/s10802-0109465-3 Alfonso, M. & Dedrick, R. F. (2010). Self-injury among early adolescents. American Journal of Health Education, 41(2), 74-84. Retrieved from Andover, M. S., Primack, J. M., Gibb, B. E., & Pepper, C. M. (2010). An examination of nonsuicidal self-injury in men: Do men differ from women in basic NSSI characteristics? Archives of Suicide Research, 14(1), 78-88. doi:10.1080/13811110903479086 American Psychiatric Association. (2000). Diagnostic and statistical manual of mental disorders, 4th edition, text revision (DSM-IV-TR). Washington, DC: American Psychiatric Press, Inc. Arnett, J. J. (1999). Adolescent storm and stress, reconsidered. American Psychologist, 54(5), 17-326. doi:10.1037/0003-066X.54.5.317 Austin, S. B., Conron, K., Patel, A., & Freedner, N. (2007). Making sense of sexual orientation measures: Findings from a cognitive processing study with adolescents on health survey questions. Journal of LGBT Health Research, 3(1):55-65. doi:10.1300/J463v03n01_07 Austin, S. B., Ziyadeh, N., Kahn, J. A., Camargo, C. A. Jr., Colditz, G. A., & Field, A. E. (2004). Sexual orientation, weight concerns, and eating-disordered behaviors in adolescent girls and boys. Journal of the American Academy of Child and Adolescent Psychiatry, 43(9):1115-1123. doi:10.1097/01.chi.0000131139.93862.10 Ayerst, S. L. (1999). Depression and stress in street youth. Adolescence, 34(135), 567-575. Retrieved from Baron, S. W. (1997). Risky lifestyles and the link between offending and victimization. Studies on Crime and Crime Prevention, 6, 53-72. Baron, S. W. (2003). Street youth violence and victimization. Trauma, Violence, & Abuse, 4(1), 22-44. doi:10.1177/1524838002238944 Bearinger, L. H., Pettingell, S. L., Resnick, M. D., & Potthoff, S. J. (2010). Reducing weaponcarrying among urban American Indian young people. Journal of Adolescent Health, 43-50. doi:10.1016/j.jadohealth.2009.12.033 Bearinger, L. H., Pettingell, S., Resnick, M. D., Skay, C. L., Potthoff, S. J., & Eichhorn, J. (2005). Violence perpetration among urban American Indian youth. Archives of Pediatric and Adolescent Medicine, 159, 270-277. Retrieved from  151 Bender, K., Thompson, S. J., McManus, H., Lantry, J., & Flynn, P. M. (2007). Capacity for survival: Exploring strengths of homeless street youth. Child and Youth Care Forum, 36, 25-42. doi:10.1007/s10566-006-9029-4 Berry, B. (2007). A repeated observation approach for estimating the street homeless population. Evaluation Review, 31(2), 166-199. doi:10.1177/0193841X06296947 Blum, R., Geer, L., Hutton, L,. McKay, C., Resnick, M. D., Rosenwinkel, K. et al. (1988). The Minnesota Adolescent Health Survey: Implications for physicians. Minnesota Medicine, 71, 143-146. Blum, R. W., NcNeely, C., & Nonnemaker, J. (2002). Vulnerability, risk, and protection. Journal of Adolescent Health, 31S, 28-39. Retrieved from Boivin, J., Roy, E., Haley, N., & Galbaud du Fort, G. (2005). The health of street youth: A Canadian perspective. Canadian Journal of Public Health, 96(6), 432-437. Retrieved from Borowsky, I. W., Ireland, M., & Resnick, M. D. (2001). Adolescent suicide attempts: Risks and protectors. Pediatrics, 107(3), 485-492. doi:10.1542/peds.107.3.485 Borowsky, I. W., Resnick, M. D., Ireland, M., & Blum, R. W. (1999). Suicide attempts among Indian and Alaska Native youth: Risk and protective factors. Archives of Pediatrics and Adolescent Medicine, 153, 575-580. Retrieved from Brain, K. L., Haines, J., & Williams, C. L. (2002). The psychophysiology of repetitive selfmutilation. Archives of Suicide Research, 6, 199-210. doi:10.1080/13811110214140 Brausch, A. M. & Gutierrez, P. M. (2010). Differences in nonsuicidal self-injury and suicide attempts in adolescence. Journal of Youth and Adolescence, 39(3), 233-242. doi:10.1007/s10964-009-9482-0 Browne, M.W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136–162). Newbury Park, CA: Sage. Brumberg, J. J. (2006). Are we facing an epidemic of self-injury? Chronicle of Higher Education, 53(16), B6-B8. Retrieved from Brunner, R., Parzer, P., Haffner, J., Steen, R., Roos, J., Klett, M. et al. (2007). Prevalence and psychological correlates of occasional and repetitive deliberate self-harm ing adolescents. Archives of Pediatric and Adolescent Medicine, 161(7), 641-649. Retrieved from Buckholdt, K. E., Parra, G. R., & Jobe-Shields, L. (2009). Emotion regulation as a mediator of the relation between emotion socialization and deliberate self-harm. American Journal of Orthopsychiatry, 79(4), 482-490. doi:10.1037/a0016735 Buckner, J. C. & Bassuk, E. L. (1997). Mental disorders and service utilization among youths from homeless and low-income household families. Journal of the American Academy of Child and Adolescent Psychiatry, 36, 890-900. doi:10.1097/00004583-199707000-00010  152 Bureau, J., Martin, J., Freynet, N., Poirier, A. A., Lafontaine, M. & Cloutier, P. (2010). Perceived dimensions of parenting and nonsuicidal self-injury in young adults. Journal of Youth and Adolescence, 39, 484-494. doi:10.1007/s10964-009-9470-4 Cauce, A. M. (2000). The characteristics and mental health of homeless adolescents: Age and gender differences. Journal of Emotional and Behavioral Disorders, 8(4), 230-239. doi:10.1177/106342660000800403 Chapman, A. L., Gratz, K. L., & Brown, M. Z. (2006). Solving the puzzle of deliberate selfharm: The experiential avoidance model. Behaviour Research and Therapy, 44, 371-394. doi:10.1016/j.brat.2005.03.005 Cheung, A. M. & Hwang, S. W. (2004). Risk of death among homeless women: A cohort study and review of the literature. Canadian Medical Association Journal, 170(8), 1243-1247. Chowanec, G. D., Josephson, A. M., Coleman, C. & Davis, H. (1991). Self-harming behaviour in incarcerated male delinquent adolescents. Journal of the American Academy of Child and Adolescent Psychiatry, 30(2), 202-207. doi:10.1097/00004583-199103000-00007 Cicchetti, D. & Lynch, M. (1993). Toward an ecological/transactional model of community violence and child maltreatment: Consequences for children‘s development. Psychiatry, 56, 96-118. Cloutier, P., Martin, J., Kennedy, A., Nixon, M. K., & Muehlenkamp, J. J. (2010). Characteristics and co-occurrence of adolescent non-suicidal self-injury and suicidal behaviours in pediatric emergency. Journal of Youth and Adolescence, 39, 259-269. doi:10.1007/s10964-009-9465-1 Coker, T. R., Austin, S. B., & Schuster, M. A. (2010). The health and health care of lesbian, gay and bisexual adolescents. Annual Review of Public Health, 31, 457 – 477. doi:10.1146/annurev.publhealth.012809.103636 Crowell, S. E., Beauchaine, T. P., McCauley, E., Smith, C. J., Vasilev, C. A., & Stevens, A. L. (2008). Parent-child interactions, peripheral serotonin, and self-inflicted injury in adolescents. Journal of Consulting and Clinical Psychology, 76(1), 15-21. doi: 10.1037/0022-006X.76.1.15 Crowell, S. E., Beauchaine, T. P., & Linehan, M. M. (2009). A biosocial developmental model of borderline personality: Elaborating and extending Linehan‘s theory. Psychological Bulletin, 135(3), 495-510. doi: 10.1037/a0015616. Cullerton-Sen, C. & Crick, N. R. (2005). Understanding the effects of physical and relational victimization: The utility of multiple perspectives in predicting social-emotional adjustment. School Psychology Review, 34(2), 147-160. Retrieved from Dadds, M., Braddock, D., Cuers, S., Elliott, A., & Kelly, A. (1993). Personal and family distress in homeless adolescents. Community Mental Health Journal, 29, 413-422. doi:10.1007/BF00754409  153 Darche, M. A. (1990). Psychological factors differentiating self-mutilating and non-selfmutilating adolescent female inpatients. Psychiatric Hospital, 21, 31-35. De Leo, D. & Heller, T. S. (2004). Who are the kids who self-harm? An Australian self-report school survey. Medical Journal of Australia, 181(3), 140-144. Retrieved from Deliberto, T. L. & Nock, M. K. (2008). An exploratory study of correlates, onset, and offset of non-suicidal self-injury. Archives of Suicide Research, 12, 219-231. doi:10.1080/13811110802101096 DeMatteo, D., Major, C., Block, B., Coates, R., Fearon, M., Goldberg, E., et al. (1999). Toronto street youth and HIV/AIDS: Prevalence, demographics, and risks. Journal of Adolescent Health, 25(5), 358-366. DiClemente, R. J., Ponton, L. E., & Hartley, D. (1991). Prevalence and correlates of cutting behaviour: Risk for HIV transmission. Journal of the American Academy of Child and Adolescent Psychiatry, 30, 735-739. doi:10.1016/S0890-8567(10)80007-3 Eisenberg, N., Spinnard, T. L., & Eggum, N. D. (2011). Emotion-related self-regulation and its relation to children‘s maladjustment. Annual Review of Clinical Psychology, 6, 495-525. doi:10.1146/annurev.clinpsy.121208.131208 Ensign, J. (1998). Health issues of homeless youth. Journal of Social Distress and the Homeless, 7, 159-174. doi:10.1023/A:1022931628497 Erdem, G. & Slesnick, N. (2010). That which does not kill you makes you stronger: Runaway youth‘s resilience to depression in the family context. American Journal of Orthopsychiatry, 80(2), 195-203. doi:10.1111/j.1939-0025.2010.01023.x Erickson, S. L. (1999). Probability profiling with urban American Indian youth: Determining priorities for suicide prevention. (Unpublished doctoral dissertation). University of Minnesota, Minneapolis, MN. Favazza, A. R. (1996). Bodies under siege: Self-mutilation and body modification in culture and psychiatry. Baltimore, MD: Johns Hopkins University Press. Feitel, B., Margetson, N., Chamas, J., & Lipman, C. (1992). Psychosocial background and behavioural and emotional disorders of homeless and runaway youth. Hospital and Community Psychiatry, 43(2), 155-159. Retrieved from Fergus, S. & Zimmerman, M. A. (2005). Adolescent resilience: A framework for understanding healthy development in the face of risk. Annual Review of Public Health, 26, 399-419. doi:10.1146/annurev.publhealth.26.021304.144357 Garmezy, N., Masten, A. S., & Tellegen, A. (1984). The study of stress and competence in children: A building block for developmental psychopathology. Child Development, 55, 97-111. Retrieved from  154 Garofalo, R., Wolf, C., Wissow, L. S., Woods, E. R., & Goodman, E. (1999). Sexual orientation and risk of suicide attempts among a representative sample of youth. Archives of Pediatric and Adolescent Medicine, 153, 487-493. Retrieved from Garrison, C. A., Cheryl, L. A., McKeown, R. E., Cuffe, S. P., Jackson, K. L, & Waller, J. L. (1993). Nonsuicidal physically self-damaging acts in adolescents. Journal of Child and Family Studies, 2, 339-352. doi:10.1007/BF01321230 Glassman, L. H., Weierich, M. R., Hooley, J. M., Deliberto, T. L., & Nock, M. K. (2007). Child maltreatment, nonsuicidal self-injury, and the mediating role of self-criticism, Behaviour Research and Therapy, doi:10.1016/j.brat.2007.04.002. Goodman, L., Saxe, L., Harvey, M. (1991). Homelessness as psychological trauma. American Psychologist, 46(11), 1219-1225. doi:10.1037/0003-066X.46.11.1219 Gore, S. & Eckenrode, J. (1994). Context and process in research on risk and resilience. In R. J. Haggerty, L. R. Sherod, N. Garmezy, & M. Rutter (Eds.), Stress, risk, and resilience in children and adolescents: Processes, mechanisms, and interventions (pp. 19-63). New York: Cambridge University Press. Gratz, K. L. (2003). Risk factors for and function of deliberate self-harm: An empirical and conceptual review. Clinical Psychology, 10(2), 192-207. doi:10.1093/clipsy.bpg022 Gratz, K. L., Conrad, S. D., & Roemer, L. (2002). Risk factors for deliberate self-harm among college students. American Journal of Orthopsychiatry, 72, 128-140. doi:10.1037/0002-9432.72.1.128 Greene, J. M., Ennett, S. T., & Ringwalt, C. L. (1999). Prevalence and correlates of survival sex among runaway and homeless youth. American Journal of Public Health, 89(9), 1406-1409. Grossman, A. H. & D‘Augelli, A. R. (2006). Transgendered youth: Invisible and vulnerable. Journal of Homosexuality, 51(1), 111-128. Grossman, A. H. & D‘Augelli, A. R. (2007). Transgendered youth and life-threatening behavior. Suicide and Life-Threatening Behavior, 37(5), 527-537. Guertin, T., Lloyd-Richardson, E., Spirito, A., Donaldson, D., & Boergers, J. (2001). Selfmutilative behaviour in adolescents who attempt suicide by overdose. Journal of the American Academy of Child and Adolescent Psychiatry, 40(9), 1062-1069. doi:10.1097/00004583-200109000-00015 Gutierrez, P. M. (2006). Integratively assessing risk and protective factors for adolescent suicide. Suicide and Life-Threatening Behaviour, 36(2), 129-135. doi:10.1521/suli.2006.36.2.129 Haber, M. G. & Toro, P. A. (2004). Homelessness among families, children, and adolescents: An ecological-developmental perspective. Child and Family Psychology Review, 7(3), 124-164. doi:10.1023/B:CCFP.0000045124.09503.f1  155 Hammer, H., Finkelhor, D., & Sedlak, A. (2002). Runaway / thrownaway children: National estimates and characteristics. National Incidence Studies of Missing, Abducted, Runaway, and Thrownaway Children. Office of Juvenile Justice and Delinquency Prevention. Hawton, K., Rodham, K., Evans, E., & Weatherall, R. (2002). Deliberate self-harm in adolescents: Self-report survey in schools in England. British Medical Journal, 325, 1207-1211. doi:10.1136/bmj.325.7374.1207 Heath, N. L., Ross, S., Toste, J. R., Charlebois, A., Nedecheva, T. (2009). Retrospective analysis of social factors and nonsuicidal self-injury among young adults. Canadian Journal of Behavioral Science, 41(3), 180-186. doi:10.1037/a0015732 Hilt, L. M., Cha, C. B., & Nolen-Hoeksema, S. (2008). Nonsuicidal self-injury in young adolescent girls: Moderators of the distress-function relationship. Journal of Consulting and Clinical Psychology, 76(1), 63-71. doi:10.1037/0022-006X.76.1.63. Hilt, L. M., Nock, M. K., Lloyd-Richardson, E. E., & Prinstein, M. J. (2008). Longitudinal study of nonsuicidal self-injury among young adolescents: Rates, correlates, and preliminary test of an interpersonal model. Journal of Early Adolescence, 28(3), 455-469. doi: 10.1177/0272431608316604 Hoyt, D. R., Ryan, K., & Cauce, A. (1999). Personal victimization in a high-risk environment: Homeless and runaway adolescents. Journal of Research in Crime and Delinquency, 36(4), 371-392. doi:10.1177/0022427899036004002 Hu, Li-tze, & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling, 6(1), 1–55. doi:10.1080/10705519909540118 Hughes, J. R., Clark, S. E., Wood, W., Cakmak, S., Cox, A., MacInnis, M.,…Broom, B. (2010). Youth homelessness: The relationships among mental health, hope, and service satisfaction. Journal of the Canadian Academy of Child and Adolescent Psychiatry, 19(4), 274-283. Retrieved from Jackson, D. L., Gillaspy, J. A. Jr. & Purc-Stephenson, R. (2009). Reporting practices in confirmatory factor analysis: An overview and some recommendations. Psychological Methods, 14(1), 6-23. doi:10.1037/a0014694 Jacobson, C. M. & Gould, M. (2007). The epidemiology and phenomenology of non-suicidal self-injurious behaviour among adolescents: A critical review of the literature. Archives of Suicide Research, 11, 129-147. doi:10.1080/13811110701247602 Jessor, R. (1991). Risk behaviour in adolescence: A psychosocial framework for understanding and action. Journal of Adolescent Health, 12, 597-605. Kerr T., Marshall, B. D. L., Miller, C., Shannon, K., Zhang, R., Montaner, J. S. G., Wood, E. (2009). Injection drug use among street-involved youth in a Canadian setting. BMC Public Health; 9(1): 171. doi:10.1186/1471-2458-9-171  156 Kidd, S. A. (2003a). The need for improved operational definition of suicide attempts: Illustrations from the case of street youth. Death Studies, 27, 449-455. doi:10.1080/07481180302877 Kidd, S. A. (2003b). Street youth coping and interventions. Child and Adolescent Social Work Journal, 20(4), 235-261. doi:10.1023/A:1024552808179 Kidd, S. A. (2006). Factors precipitating suicidality among homeless youth: A quantitative follow-up. Youth & Society, 37(4), 393-422. doi:10.1177/0044118X05282763 Kidd, S. A. (2007). Youth homelessness and social stigma. Journal of Youth and Adolescence, 36, 291-299. doi:10.1007/s10964-006-9100-3 Kidd, S. A. & Carroll, M. (2007). Coping and suicidality among homeless youth. Journal of Adolescence, 30, 283-296. doi:10.1016/j.adolescence.2006.03.002 Kidd, S. A. & Kral, M. J. (2002). Suicide and prostitution among street youth: A qualitative analysis. Adolescence, 37(146), 411-430. King, A., Wold, B., Tudor-Smith, C., & Harel, Y. (1996). The health of youth: A cross-national survey. World Health Organization regional publications, European Series No. 69. Copenhagen: World Health Organization. Kipke, M. D., Montgomery, S. B., Simon, T. R., & Iverson, E. F. (1997). Substance abuse disorders among runaway and homeless youth. Substance Use and Misuse, 32, 969-986. Kipke, M. D., Simon, T. R., Montgomery, S. B., Unger, J. B. & Iverson, E. F. (1997). Homeless youth and their exposure to and involvement in violence while living on the streets. Journal of Adolescent Health, 20(5), 360-367. Retrieved from Klonsky, E. D. (2009). The functions of self-injury in young adults who cut themselves: Clarifying the evidence for affect-regulation. Psychiatry Research, 166, 260-268. doi:10.1016/j.psychres.2008.02.008 Klonsky, E. D. (2007). The functions of deliberate self-injury. A review of the evidence. Clinical Psychology Review, 27, 226-239. doi:10.1016/j.cpr.2006.08.002 Klonsky, E. D. & Glenn, C. R. (2009a). Assessing the functions of non-suicidal self-injury: Psychometric properties of the Inventory of Statements About Self-injury (ISAS). Journal of Psychopathology and Behavioral Assessment, 31, 215–219. doi:10.1007/s10862-008-9107-z Klonsky, E. D. & Glenn, C. R. (2009b). Psychosocial risk and protective factors. In M. K. Nixon & N. L. Heath (Eds.), Self-injury in youth: The essential guide to assessment and intervention (pp. 45-58). New York: Routledge. Klonsky, E. D. & Moyer, A. M. (2008). Childhood sexual abuse and nonsuicidal self-injury: Meta-analysis. British Journal of Psychiatry, 192, 166-172. doi:10.1192/bjp.bp.106.030650  157 Klonsky , E. D., Oltmanns, T. F., & Turkheimer, E. (2003). Deliberate self-harm in a nonclinical population: Prevalence and psychological correlates. American Journal of Psychiatry, 160(8), 1501-1508. Retrieved from Kolbe, L. J., Kann, L., & Collins, J. L. (1993). Overview of the youth risk behaviour surveillance system. Public Health Reports, 108, 10. Retrieved from Kraemer, H. C., Stice, E., Kazdin, A., Offord, D., & Kupfer, D. (2001). How do risk factors work together? Mediators, moderators, and independent, overlapping proxy risk factors. American Journal of Psychiatry, 158(6), 848-856. Retrieved from Laye, A. (2002). Adolescent self-harm in a school-based population: Prevalence and Correlates. (Unpublished master‘s thesis). The University of British Columbia, Vancouver, BC. Laye, A., Murphy, A., Katzenstein, D., & The McCreary Centre Society (2002). Between the cracks: Homeless youth in Vancouver. Burnaby, BC: The McCreary Centre Society. Laye-Gindhu, A. & Schonert-Reichl, K. A. (2005). Nonsuicidal self-harm among community adolescents: Understanding the ‗whats‘ and ‗whys‘ of self-harm. Journal of Youth and Adolescence, 34, 447-457. doi:10.1007/s10964-005-7262-z Leffert, N., Benson, P. L., Scales, P. C., Sharma,A. R., Drake, D. R., Blyth, D A. (1998). Developmental assets: Measurement and prediction of risk behaviours among adolescents. Applied Developmental Science, 2(4), 209-230. doi:10.1207/s1532480xads0204_4 Leserman, J., Li, Z., Drossman, D. A., Toomey, T. C., Nachman, G. & Glogau, L. (1997). Impact of sexual and physical abuse dimensions on health status: Development of an abuse severity measure. Psychosomatic Medicine, 59, 152-160. Retrieved from Linehan, M. (1993). Cognitive-behavioural therapy for borderline personality disorder. New York: Guilford Press. Lloyd-Richardson, E., Perrine, N., Dierker, L., & Kelley, M. (2007). Characteristics and functions of non-suicidal self-injury in a community sample of adolescents. Psychological Medicine, 37(8), 1183-1192. doi:10.1017/S003329170700027X Lloyd-Smith, E., Kerr, T., Zhang, R., Montaner, J. S. G. & Wood, E. (2008). High prevalence of syringe sharing among street involved youth. Addiction Research and Theory, 16(4), 353-35. doi:10.1080/16066350701823094 Luthar, S. S. (1993). Annotation: Methodological and conceptual issues in research on childhood resilience. Journal of Clinical Child Psychology and Psychiatry, 34(4), 441-453. doi:10.1111/j.1469-7610.1993.tb01030.x  158 Luthar, S. S., Cicchetti, D., & Becker, B. (2000). The construct of resilience: A critical evaluation and guidelines for future work. Child Development, 71(3), 543-562. doi:10.1111/1467-8624.00164 Luthar, S. S., Sawyer, J. A., & Brown, P. J. (2006). Conceptual issues in studies of resilience. Annals of the NY Academy of Sciences, 1094, 105-115. doi:10.1196/annals.1376.009 Luthar, S. S. & Zelazo, L. B. (2003). Research on resilience: An integrative review. In S. S. Luthar (Ed.), Resilience and Vulnerability: Adaptation in the Context of Childhood Adversities (pp. 510-549). New York: Cambridge University Press. MacKay, L. (2007). Healthy development: The role of youth assets. (Unpublished doctoral dissertation). Simon Fraser University, Burnaby, BC. MacLean, M. G., Embry, E. L., & Cauce, A. (1999). Homeless adolescents‘ paths to separation from family: Comparison of family characteristics, psychological adjustment, and victimization. Journal of Community Psychology, 27, 179-187. doi:10.1002/(SICI)1520-6629(199903)27:2<179::AID-JCOP5>3.0.CO;2-S MacLean, M. G., Paradise, M. J., & Cauce, A. M. (1999). Substance use and psychological adjustment in homeless adolescents: A test of three models. American Journal of Community Psychology, 27(3), 405-427. doi:10.1023/A:1022286111098 Marsh, H. W., Balla, J. R., & McDonald, R. P. (1988). Goodness of fit indexes in confirmatory factor analysis: The effect of sample size. Psychological Bulletin, 103, 391-410. doi:10.1037/0033-2909.103.3.391 Marsh, H. W., Hau, Kit-Tai, & Wen, Z. (2004). In search of golden rules: comment on hypothesis-testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing Hu and Bentler‘s (1999) findings. Structural Equation Modeling, 11(3), 320–341. Retrieved from Marshall, H. & Yazdani, A. (1999). Locating culture in accounting for self-harm amongst Asian young women. Journal of Community & Applied Social Psychology, 9, 413-433. doi:10.1002/(SICI)1099-1298(199911/12)9:6<413::AID-CASP543>3.0.CO;2 Martijn, C. & Sharpe, L. (2006). Pathways to youth homelessness. Social Science & Medicine, 62, 1-12. doi:10.1016/j.socscimed.2005.05.007 Masten, A., S., Best, K. M., & Garmezy, N. (1990). Resilience and development: Contributions from the study of children who overcome adversity. Development and Psychopathology, 2, 425-444. doi:10.1017/S0954579400005812 Masten. A. S. & Obradovic, J. (2006). Competence and resilience in development. Annals of the American Academy of Sciences, 1094, 13-27. doi:10.1196/annals.1376.003 McBride, A. J., Pates, R. M., Arnold, K., & Ball, N. (2001). Needle fixation, the drug user`s perspective: A qualitative study. Addiction, 96(7), 1049-1058. doi:10.1046/j.1360-0443.2001.967104914.x  159 McLaughlin, K. A., Hatzenbuehler, M. L., & Hilt, L. M. (2009). Emotion dysregulation and adolescent psychopathology: A mechanism linking peer victimization to internalizing symptoms. Journal of Consulting and Clinical Psychology, 77(5), 894-904. doi:10.1037/90015760 McNeely, C. & Falci, C. (2004). School connectdness and the transition into and out of health-risk behaviour among adolescents: A comparison of social belonging and teacher support. Journal of School Health, 74(7), 284-292. doi:10.1111/j.1746-1561.2004.tb08285.x McNeely, C., Nonnemaker, J. M., & Blum, R. W. (2002). Promoting school connectedness: Evidence from the National Longitudinal Study of Adolescent Health. Journal of School Health, 72(4), 138-146. doi:10.1111/j.1746-1561.2002.tb06533.x Menard, S. (1995). Applied logistic regression analysis. Sage university paper series on quantitative applications in the social sciences. Thousand Oaks, CA: Sage. Miller, A. L. & Smith, H. L. (2008). Adolescent non-suicidal self-injurious behaviour: The latest epidemic to assess and treat. Applied and Preventive Psychology, 12, 178-188. doi:10.1016/j.appsy.2008.05.003 Mineka, S. & Suomi, S. J. (1978). Social separation in monkeys. Psychological Bulletin, 85, 1376-1400. Molnar, B. E., Shade, S. B., Kral, A. H., Booth, R. E., & Watters, J. K. (1998). Suicidal behaviour and sexual/physical abuse among street youth. Child Abuse & Neglect, 21(3), 213-222. doi:10.1016/S0145-2134(97)00137-3 Morris, A. S., Silk, J. S., Steinberg, L., Myers, S. S., & Robinson, L. R. (2007). The role of the family context in the development of emotion regulation. Social Development, 16(2), 361-388. doi:10.1111/j.1467-9507.2007.00389.x Muehlenkamp, J. J. (2005). Self-injurious behavior as a separate clinical syndrome. American Journal of Orthopsychiatry, 75(2), 324–333. doi:10.1037/0002-9432.75.2.324 Muehlenkamp, J. J. & Gutierrez, P. M. (2004). An investigation of differences between selfinjurious behavior and suicide attempts in a sample of adolescents. Suicide and Life-Threatening Behavior, 34(1), 12-23. doi:10.1521/suli. Muehlenkamp, J. J. & Gutierrez, P. M. (2007). Risk for suicide attempts among adolescents who engage in non-suicidal self-injury. Archives of Suicide Research, 11(1), 69-82. doi:10.1080/13811110600992902 Murphy, A., Chittenden, M., & The McCreary Centre Society (2005). Time Out II: A Profile of BC Youth in Custody. Vancouver, BC: The McCreary Centre Society. Murphy, A. & The McCreary Centre Society (2001). No place to call home: A profile of street youth in British Columbia. Burnaby, BC: The McCreary Centre Society. Murray, C. D., Warm, A., & Fox, J. (2005). An internet survey of adolescent self-injurers. Australian e-Journal for the Advancement of Mental Health, 4(1), 1-9. Retrieved from  160 Nada-Raja, S., Skegg, K., Langley, J., Morrison, D., & Sowerby, P. (2004). Self-harmful behaviours in a population-based sample of young adults. Suicide and LifeThreatening Behaviour, 34(2), 177-186. doi:10.1521/suli. Najmi, S., Wegner, D. M., & Nock, M. K. (2006). Thought suppression and self-injurious thoughts and behaviours. Behaviour Research and Therapy, doi:10.1016/j.brat.2006.09.014. Nash, M. R., Zivney, O. A., & Hulsey, Y. (1993). Characterisitcs of sexual abuse associated with greater psychological impairment among children. Child Abuse and Neglect, 17, 401-408. doi:10.1016/0145-2134(93)90063-B Nixon, M. K., Cloutier, P. F., & Aggarwal, S. (2002). Affect regulation and addictive aspects of repetitive self-injury in hospitalized adolescents. Journal of the American Academy of Child and Adolescent Psychiatry, 41(11), 1333-1341. doi:10.1503/cmaj.061693 Nixon, M. K., Cloutier, P. F., & Jansson, S. M. (2008). Nonsuicidal self-harm in youth: A population-based study. CMAJ, 178(3), 306-312. doi:10.1503/cmaj.061693 Nock, M. K. (2010). Self-injury. Annual Review of Clinical Psychology, 6, 339-363. doi:10.1J46/annurev.clinpsy.121208.131258 Nock, M. K., Joiner, T. E. Gordon, K. H., Lloyd-Richardson, E. E., & Prinstein, M. J. (2006). Non-suicidal self-injury among adolescents: Diagnostic correlates and relation to suicide attempts. Psychiatry Research, 144, 65-72. doi:10.1016/j.psychres.2006.05.010 Nock, M. K. & Kessler, R. C. (2006). Prevalence of and risk factors for suicide attempt versus suicide gestures: Analsyis of the National Comorbidity Survey. Journal of Abnormal Psychology, 115(3), 616-623. Retrieved from Nock, M. K. & Mendes, W. B. (2008). Physiological arousal, distress tolerance, and social problem-solving deficits among adolescent self-injurers. Journal of Consulting and Clinical Psychology, 76(1), 28-38. doi:10.1038/0022-006X.76.1.28 Nock, M. K. & Prinstein, M. J. (2004). A functional approach to the assessment of selfmutilative behaviour. Journal of Consulting and Clinical Psychology, 72(5), 885-890. doi:10.1037/0022-006X.72.5.885 Nock, M. K. & Prinstein, M. J. (2005). Contextual features and behavioural functions of self mutilation among adolescents. Journal of Abnormal Psychology, 114(1), 140-146. doi:10.1037/0021-843X.114.1.140 Nock, M.K., Prinstsein, M. J., & Serba, S. K. (2009). Revealing the form and function of selfinjurious thoughts and behaviours: A real-time ecological assessment study among adolescents and young adults. Journal of Abnormal Psychology, 118(4), 816-827. doi:10.1037/a0016948 Pattison, E. M., & Kahan, J. (1983). The deliberate self-harm syndrome. American Journal of Psychiatry, 140, 867–872. Retrieved from  161 Peduzzi, P., Concato, J., Kemper, E., Holford, T. R., & Feinstein, A. (1996). A simulation of the number of events per variable in logistic regression analyses. Journal of Clinical Epidemiology, 99, 1373-1399. Retrieved from Patton, G. C., Harris, R., Carlin, J. B., Hibbert, M. E., Coffey, C., Schwartz, M., & Bowes, G. (1997). Adolescent suicidal behaviours: A population-based study of risk. Psychological Medicine, 27, 715-724. doi:10.1017/S003329179600462X Poon, C., Chittenden, M., Saewyc, E., Murphy, A., & The McCreary Centre Society. (2006). Promoting healthy bodies: Physical activity, weight, and tobacco use among B.C. youth. Vancouver, BC: The McCreary Centre Society. Public Health Agency of Canada. (2006). Street youth in Canada: Findings from the Enhanced Surveillance of Canadian Street Youth [ESCSY], 1999-2003. (Report Cat No. HP5-15/2006) Retrieved from Rachlis, B. S., Wood, E., Zhang, R., Montaner, J. S. G., & Kerr, T. (2009). High rates of homelessness among a cohort of street-involved youth. Health Place, 15(1), 10-17. doi:10.1016/j.healthplace.2008.01.008 Resnick, M. D., Bearman, P. S., Blum, R. W., Bauman, K. E., Harris, K. M., Jones, J…& Udry, J. R. (1997). Protecting adolescents from harm: Findings from the National Longitudinal Study on Adolescent Health. Journal of the American Medical Association, 278, 823-832. Retrieved from Rew, L. & Horner, S. D. (2003). Personal strengths of homeless adolescents living in a highrisk environment. Advances in Nursing Science, 26(2), 90-101. Retrieved from Ringwalt, C. L., Greene, J. M., & Robertson, M. J. (1998). Familial backgrounds and risk behaviours of youth with thrownaway experiences. Journal of Adolescence, 21, 241-252. doi:10.1006/jado.1998.0150 Ringwalt, C. L., Greene, J. M., Robertson, M. J., & McPheeters, M. (1998). The prevalence of homelessness among adolescents in the United States. American Journal of Public Health, 88(9), 1325-1329. Retrieved from Rodham, K., Hawton, K., & Evans, E. (2004). Reasons for deliberate self-harm: Comparison of self-poisoners and self-cutters in a community sample of adolescents. Journal of the American Academy of Child and Adolescent Psychiatry, 43(1), 80-87. doi:10.1097/00004583-200401000-00017 Rohde, P., Noell, J., Ochs, L., & Seeley, J. R. (2001). Depression, suicidal ideation, and STD-related risk in homeless older adolescents. Journal of Adolescence, 24, 447-460. doi:10.1006/jado.2001.0382 Ross, S. & Heath, N. (2002). A study of the frequency of self-mutilation in a community sample of adolescents. Journal of Youth and Adolescence, 31(1), 67-77. doi:10.1023/A:1014089117419  162 Ross, S. & Heath, N. (2003). Two models of adolescent self-mutilation. Suicide and LifeThreatening Behaviour, 33(3), 277-287. doi:10.1521/suli. Ross, S., Heath, N. L., & Toste, J. R. (2009). Non-suicidal self-injury and eating pathology in high school students. American Journal of Orthopsychiatry, 79(1), 83-92. doi:10.1037/90014826 Roy, E., Haley, N., Leclerc, P., Sochanski, B., Boudreau, J., & Boivin, J. (2004). Mortality in a cohort of street youth in Montreal. JAMA, 292(5), 569-574. doi:10.1001/jama.292.5.569 Rubenstein, J. L., Heeren, T., Housman, D., Rubin, C., & Stechler, G. (1989). Suicidal behaviour in ―normal‖ adolescents: Risk and protective factors. American Journal of Orthopsychiatry, 59(1), 59-71. doi:10.1111/j.1939-0025.1989.tb01635.x Russell, S. T. (2003). Sexual minority youth and suicide risk. American Behavioral Scientist, 46(9), 1241-1257. doi:10.1177/0002764202250667 Russell, S. T. & Joyner, C. (2001). Adolescent sexual orientation and suicide risk: Evidence from a national study. American Journal of Public Health, 91, 1276-1281. Retrieved from Rutter, M. (1987). Psychosocial resilience and protective mechanisms. American Journal of Orthopsychiatry, 57, 316-331. doi:10.1111/j.1939-0025.1987.tb03541.x Rutter, M. (1993). Resilience: Some conceptual considerations. Journal of Adolescent Health, 14, 626-631. doi:10.1016/1054-139X(93)90196-V Rutter, M. (2000). Resilience reconsidered: Conceptual considerations, empirical findings, and policy implications. In J. P. Shonkoff & S. Meisels (Eds.), Handbook of early childhood intervention 2 (pp. 651-682). New York: Cambridge University Press. Ryan, K. D., Kilmer, R. P., Cauce, A. M., Watanabe, H., & Hoyt, D. R. (2000). Psychological consequences of child maltreatment in homeless adolescents: Untangling the unique effect of maltreatment and family environment. Child Abuse & Neglect, 34, 333-352. doi:10.1016/S0145-2134(99)00156-8 Saewyc, E. M. (2011). Research on adolescent sexual orientation: Development, health disparities, stigma, and resilience. Journal of Research on Adolescence, 21(1), 256-272. doi:10.1111/j.532-7795.2010.00727.x Saewyc, E. M., Bauer, G. R., Skay, C. L., Bearinger, L. H., Resnick, M. D., Reise, E., & Murphy, A. (2004). Measuring sexual orientation in adolescent health surveys: Evaluation of school-based surveys. Journal of Adolescent Health, 35, 345.e1-e15. Saewyc, E. M. & Edinburgh, L. D. (2010). Restoring healthy developmental trajectories for sexually exploited young runaway girls: Fostering protective factors and reducing risk behaviors. Journal of Adolescent Health, 46, 180-188. doi:10.1016/j.jadohealth.2009.06.010  163 Saewyc, E. M. & Homma, Y. (2010). Psychometric properties of scales in the BC Adolescent Health Surveys of McCreary Centre Society: 1998, 2003, and 2008. Vancouver, BC: McCreary Centre Society. Saewyc, E., Wang, N., Chittenden, M., Murphy, A., & The McCreary Centre Society. (2006). Building resilience in vulnerable youth. Vancouver, BC: The McCreary Centre Society. Sameroff, A., Gutman, L. M., & Peck, S. C. (2003). Adaptation among youth facing multiple risks: Prospective research findings. In S. S. Luthar (Ed.), Resilience and Vulnerability: Adaptation in the Context of Childhood Adversities (pp. 365-391). New York: Cambridge University Press. Santa Mina, E. E. & Gallop, R. M. (1998). Childhood sexual and physical abuse and adult self-harm and suicidal behaviour: A literature review. Canadian Journal of Psychiatry, 43, 793-800. Retrieved from Schonert-Reichl, K. A. (2000, April). Children and youth at risk: Some conceptual considerations. Paper prepared for the Pan-Canadian Education Research Agenda Symposium, Children and Youth at Risk, Ottawa, Canada. Retrieved from Schonert-Reichl, K. A. & Offer, D. (1992). Gender differences in adolescent symptoms. In B. B. Lahey & A. E. Kazdin (Eds.), Advances in clinical psychology (vol 14; pp. 27-60). New York: Plenum Press. Schulenberg, J., Maggs, J. L., & Hurrelman, K. (1997). Negotiating developmental transitions during adolescence and young adulthood: Health risks and opportunities. In J. Schulenberg, J. L. Maggs, & K. Hurrelman (Eds.), Health risks and developmental transitions during adolescence (pp. 1-22). New York: Cambridge University Press. Schwartz, R. H., Cohen, P., Hoffman, N. G., & Meeks, J. E. (1989). Self-harm behaviours (carving) in female adolescent drug abusers. Clinical Pediatrics, 28, 340-346. doi:10.1177/000992288902800801 Selekman, M. D. (2006). Working with self-harming adolescents: A collaborative strengthsbased therapy approach. New York: W. W. Norton & Company. Serras, A., Saules, K. K., Cranford, J. A., & Eisenberg, D. (2010). Self-injury, substance use, and associated risk factors in a multi-campus probability sample of college students. Psychology of Addictive Behaviour, 24(1), 119-128. doi:10.1037/a0017210 Shaffer, D. & Jacobson, C. (2010). Proposal to the DSM-V Childhood Disorder and Mood Disorder Work Groups to include nonsuicidal self-injury (NSSI) as a DSM-V Disorder. American Psychiatric Association. Retrieved from  164 Sieving, R. E., Beuhring, T., Resnick, M. C., Bearinger, L. H., Shew, M. et al. (2001). Development of adolescent self-report measures from the National Longitudinal Study of Adolescent Health. Journal of Adolescent Health, 28, 73-81. Retrieved from Sim, L., Adrian, M., Zeman, J., Cassano, M., & Friedrich, W. N. (2009). Adolescent deliberate self-harm: Linkages to emotion regulation and family emotional climate. Journal of Research on Adolescence, 19(1), 75-91. doi:10.1111/j.1532-7795.2009.00582.x Skegg, K. (2005). Self-harm. Lancet, 366(9495), 1471-1483. doi:10.1016/S0140-6736(05)67600-3 Skegg, K., Nada-Raja, S., Dickson, N., Paul, C., &Williams, S. (2003). Sexual orientation and self-harm in men and women. American Journal of Psychiatry, 160(3), 541-546. Retrieved from Smith, A., Saewyc, E., Albert, M., MacKay, L., Northcott, M., & The McCreary Centre Society. (2007). Against the odds: A profile of marginalized and street-involved youth in B.C. Vancouver, BC: The McCreary Centre Society. Social Planning and Research Council of British Columbia. (2005, September). On our streets and in our shelters: Results of the 2005 Greater Vancouver Homeless Count. Vancouver, BC: Author. Stanley, B., Gameroff, M .J., Michalsen, V., & Mann, J. J. (2001). Are suicide attempters who self-mutilate a unique population? American Journal of Psychiatry, 158(3), 427-432. Retrieved from Statistics Canada. Table 051-0001, Population by Age and Sex for Canada, Provinces, and Territories. Canadian Statistics, Version updated October 26, 2006. Retrieved from Stewart, A. J., Steiman, M., Cauce, A. M., Cochran, B. N., Whitbeck, L. B., & Hoyt, D. R. (2004). Victimization and posttraumatic stress disorder among homeless adolescents. Journal of the American Academy of Child and Adolescent Psychiatry, 43(3), 324-331. doi:10.1097/00004583-200403000-00015 Suyemoto, K. (1998). The functions of self-mutilation. Clinical Psychology Review, 18(5), 531-554. doi:10.1016/S0272-7358(97)00105-0 Svirko, E. & Hawton, K. (2007). Self-injurious behaviour and eating disorders: The extent and nature of association. Suicide and Life-Threatening Behaviour, 37(4), 409-421. doi:10.1521/suli.2007.37.4.409 Tabachnick, B. G. & Fidell, L. S. (2001). Using Multivariate Statistics , 4th ed. Boston: Allyn & Bacon. Thompson, E. M. & Morgan, E. M. (2008). ―Mostly straight‖ young women: Variations in sexual behaviour and identity development. Developmental Psychology, 44(1), 15-21.  165 Thrane, L. E., Hoyt, D. R., Whitbeck, L. B., & Yoder, K. A. (2006). Impact of family abuse on running away, deviance, and street victimization among homeless rural and urban youth. Child Abuse & Neglect, 30, 1117-1128. doi:10.1016/j.chiabu.2006.03.008 Topolski, T. D., Edwards, T. C., & Patrick, D. L. (2004). Toward youth self-report of health and quality of life in population monitoring. Academic Pediatrics, 4(4), 387-394. doi:10.1367/A03-131R.1 Tyler, K. A. (2006). A qualitative study of early family histories and transitions of homeless youth. Journal of Interpersonal Violence, 21(10), 1385-1393. doi 10.1177/0886260506291650 Tyler, K. A., Hoyt, D. R., Whitbeck, L. B. (2000). The effects of early sexual abuse on later sexual victimization among female homeless and runaway youth. Violence and Victims, 16, 441-455. doi:10.1177/088626000015003001 Tyler, K.A., Hoyt, D. R., Whitbeck, L. B., & Cauce, A. (2001). The impact of childhood sexual abuse on later sexual victimization among runaway youth. Journal of Research on Adolescence, 11, 151-176. doi:10.1111/1532-7795.00008 Tyler, K., Melander, L., & Almazan, E. (2010). Self-injurious behaviour among homeless young adults: A social stress analysis. Social Science & Medicine, 70(2), 269-276. doi:10.1016/j.socscimed.2009.10.008 Tyler, K. A., Whitbeck, L. B., Hoyt, D. R., & Johnson, K. D. (2003). Self-mutilation and homeless youth: The role of family abuse, street experiences, and mental disorders. Journal of Research on Adolescence, 13(4), 457-474. doi:10.1046/j.1532-7795.2003.01304003.x Unger, J. B., Kipke, M. D., Simon, T. R., Montgomery, S. B., & Johson, C. J. (1997). Homeless youth and young adults in Los Angeles: Prevalence of mental health problems and relationship between mental health and substance abuse disorders. American Journal of Community Psychology, 25(3), 371-394. doi:10.1023/A:1024680727864 U.S. Census Bureau. Table: Annual estimates of the population by five-year age groups and sex for the United States: April 1, 2000 to July 1, 2006 (NC-EST2006-01), Population Division, U.S. Census Bureau. Retrieved from van der Kolk, B. A. & Fisler, R. E. (1994). Childhood abuse and neglect and loss of selfregulation. Bulletin of the Menninger Clinic, 58(2), 145-169. Retrieved from van der Kolk, B. A., Perry, J. C., & Herman, J. L. (1991). Childhood origins of self-destructive behaviour. American Journal of Psychiatry, 148(12), 1665-1671. Retrieved from Viljoen, J. L., O‘Neill, M. L., & Sidhu, A. (2005). Bullying behaviours in female and male adolescent offenders: Prevalence, types, and association with psychosocial adjustment. Aggressive Behaviour, 31, 521-536. doi:10.1002/ab.20036  166 Votta, E. & Manion, I. (2004). Suicide, high-risk behaviours, and coping style in homeless adolescent males‘ adjustment. Journal of Adolescent Health, 34, 237-243. doi:10.1016/j.jadohealth.2003.06.002) Walsh, B.W. (2006). Treating self-injury: A practical guide. New York: The Guildford Press. Walsh, B. W. & Rosen, P. (1988). Self-mutilation: Theory, research, and treatment. New York: Guilford Books, Inc. Whitbeck, L. B., Hoyt, D. R., & Ackley, K. A. (1997). Families of homeless and runaway adolescents: A comparison of parent / caretaker and adolescent perspectives on parenting, family violence, and adolescent conduct. Child Abuse & Neglect, 21, 517-528. doi:10.1016/S0145-2134(97)00010-0 Whitbeck, L. B., Hoyt, D. R., & Bao, W. (2000). Depressive symptoms and co-occurring depressive symptoms, substance abuse, and conduct problems among runaway and homeless adolescents. Child Development, 71(3), 721-732. doi:10.1111/1467-8624.00181 Whitbeck, L. B., Hoyt, D. R., & Yoder, K. A. (1999). A risk-amplification model of victimization and depressive symptoms among runaway and homeless adolescents. American Journal of Community Psychology, 27(2), 273-296. doi:10.1023/A:1022891802943 Whitbeck, L. B., Johnson, K. D., Hoyt, D. R., & Cauce, A. M. (2004). Mental disorder and comorbidity among runaway and homeless adolescents. Journal of Adolescent Health, 35, 132-140. doi: 10.1016/j.jadohealth.2003.08.011 Whitbeck, L. B. & Simons, R. L. (1990). Life on the streets: The victimization of runaway and homeless adolescents. Youth & Society, 22, 108-125. doi:10.1177/0044118X90022001007 Whitlock, J., Eckenrode, J., & Silverman, D. (2006). Self-injurious behaviour in a college population. Pediatrics, 117(6), 1939-1948. doi:10.1542/peds.2005-2543 Whitlock, J. & Knox, K. L. (2007). The relationship between self-injurious behaviour and suicide in a young adult population. Archives of Pediatrics and Adolescent Medicine, 161(7), 634-640. Retrieved from Wiederman, M. W., Sansone, R. A., & Sansone, L. A. (1999). Bodily self-harm and its relationship to childhood abuse among women in a primary care setting. Violence Against Women, 5(2), 155-163. doi:10.1177/107780129952004 Wong, J. P. S., Stewart, S. M., Ho, S. Y., & Lam, T. H. (2007). Risk factors associated with suicide attempts and other self-injury among Hong Kong adolescents. Suicide and Life-Threatening Behaviour, 37(4), 453-466. doi:10.1521/suli.2007.37.4.453 World Health Organization. (2004). Adolescent health and development. Retrieved from  167 Wright, J. D., Allen, T. L., & Devine, J. A. (1995), Tracking non-traditional populations in longitudinal studies. Evaluation and Program Planning, 18(3), 267-277. doi.10.1016/S0149-7189(95)00020-8 Yates, T. M. (2004). The developmental psychopathology of self-injurious behaviour: Compensatory regulation in posttraumatic adaptation. Clinical Psychology Review, 24, 35-74. doi:10.1016/j.cpr.2003.10.001 Yates, T. M., Tracy, A. J., Luthar, S. S. (2008). Nonsuicidal self-injury among ―privileged‖ youth: Longitudinal and cross-sectional approaches to developmental process. Journal of Consulting and Clinical Psychology, 76, 52-62. doi:10.1037/0022-006X.76.1.52 Yoder, K. A., Hoyt, D. R., & Whitbeck, L. B. (1998). Suicidal behaviour among homeless and runaway adolescents. Journal of Youth and Adolescence, 27(6), 753-771. doi:10.1023/A:1022813916476 Zeman, J., Cassano, M., Perry-Parrish, C., & Stegall, S. (2006). Emotion regulation in children and adolescents. Developmental and Behavioral Pediatrics, 27(2), 155-168. doi:10.1097/00004703-200604000-00014 Zoroglu, S. S., Tuzun, U., Sar, V., Tutkun, H., Savacs, H. A., Ozturk, M. et al. (2003). Suicide attempt and self-mutilation among Turkish high-school students in relation with abuse, neglect, and dissociation. Psychiatry and Clinical Neurosciences, 57, 119-126. doi:10.1046/j.1440-1819.2003.01088.x Zweig, J. M., Lindberg, L. D., & McGinley, K. A. (1991). Adolescent health risk profiles: The co-occurrence of health risks among females and males. Journal of Youth and Adolescence, 30(6), 707-726. doi:10.1023/A:1012281628792  168  Appendix A:  Source of Questions on the Street-Involved Youth Health Survey (SYHS) (The McCreary Centre Society, 2006)  169  Appendix A Source of Questions on the Street-Involved Youth Health Survey (SYHS) Variable Self-injury Self-injury medical attention Self-injury function  Survey Item # 105 106 107  Descriptor of Survey Questions  Source  Self-injury and frequency Medical attention Reasons for self-injuring  The McCreary Centre Society (MCS) (1st time these items have been used)  Ever: sexual abuse Sexual coercion (forced to have sex – by adult, by youth) Perpetrator: who (options) Ever: physical abuse Perpetrator: who (options) Witnessing family member being assaulted or mistreated Where lived last 12 months, ever, where live now (yesterday) (options) Obtain money from child welfare/ward of the courts Mother or father with alcohol and/or drug problem , with mental illness, or involved in criminal behaviour Age at first being kicked out of the home  Adolescent Health Survey (AHS – Minnesota) Adolescent Health Program, University of Minnesota  Sexual activity with a male in exchange for money/goods Sexual activity with a female in exchange for money/goods In past 30 days, obtained money from In last year, frequency of trading sex for food, clothing, shelter, transportation, money, drugs or alcohol, or other Engaged in sexual activity for pimp, escort agency, to support a friend-partner or relative, other Locations traded sex (options) Age at first trading sex Where living when first traded sex  Specific to MCS street surveys – derivation unknown  Hypothesized Contextual Risk Factors Sexual abuse  Witness family violence Ever been in-care  124 94 125 122 123 121 14  Parent problems  26 134-137  Kicked out of home  21  Sexual exploitation  96 97 26 98  Physical abuse  99 100 101 102  Specific to MCS street surveys – derivation unknown  170  Variable Street exposure  Survey Item # 13 14  Descriptor of Survey Questions Age at first street-involvement Where live now (yesterday), past 12 months, ever  Source Specific to MCS street surveys – derivation unknown  Hypothesized Individual Risk Factors Gender Age Sexual orientation Ethnicity Physical risk-taking  2 1 95 3 80 81  Gender Age Feelings about being attracted to others (options) Ethnic background (options) Re-using drug paraphernalia/gear Re-using tattoo or body piercing equipment, shaving razors  MCS MCS AHS, Minnesota MCS Specific to MCS street surveys – derivation unknown  Fighting, being assaulted (12months)  129 130  Frequency of physical fights in past 12 months Frequency of physical injury requiring medical treatment from fighting in past 12 months  132  Frequency of being physically attacked/assaulted in past 12 months  67 68 69 77  Frequency of binge-eating or gorging Frequency of intentional vomiting after eating Age at initiation to marijuana Age at first drink of alcohol (other than few sips)  Youth Risk Behavior Survey (YRBS), Centre for Disease Control and Prevention (CDC), Atlanta, Georgia National Longitudinal Survey of Children and Youth (NLSCY), Statistics Canada and Human Resource and Development Canada (HRDC) AHS, Minnesota  75 77 72  Frequency of binge drinking in past 30 days  116  Suicide attempt in past 12 months  Disordered eating behaviour Early initiation of substance use Binge-drinking, past month Injection drug use Suicidal ideation and behaviour (12 months)  YRBS, CDC AHS, Minnesota YRBS, CDC  Injected illegal drug ever YRBS, CDC  171  Variable  Survey Item #  Descriptor of Survey Questions  Source  Contextual Protective Factors Family connectedness  27-34  Feel understand by people in family Have fun together with family Family pays attention Feelings of closeness to mother Feel cared for by mother Feelings of closeness to father Feel cared for by father Perceive mother as warm and loving toward self Satisfaction with relationship with mother Perceive father as warm and loving toward self Satisfaction with relationship with father  National Longitudinal Study of Adolescent Health (Add-Health), The Carolina Population Center, University of North Carolina  School connectedness  45-48  Feel cared about by teacher(s) Feel/felt like part of the school Happy to be at school Feel that teachers treat/treated students fairly Feel/felt safe at school Frequency of trouble getting along with teachers Frequency of trouble getting along with other students  Add-Health, University of North Carolina  Community activities  50  Involvement in extracurricular activities before street involvement (sports/physical activities either with coach/instructor or not, dance/aerobic classes/lessons, art, drama, singing, music groups/clubs/lessons, clubs/groups  NLSCY, Statistics Canada and HRDC  Extreme sports  52  Involvement in extreme sports before street-involvement (Backcountry skiing/snowboarding, high-speed motorized sport, cliff and bridge jumping, rock climbing, white water activities, downhill mountain biking, other)  MCS  172  Variable  Survey Item #  Descriptor of Survey Questions  Source  Individual Protective Factors Past month emotional distress scale  109 110 111 112  Under strain, stress, or pressure past 30 days Bothered by any illness, physical problems, pains, or fears about health in past 30 days Bothered by nervousness or “nerves” in past 30 days Felt so sad, discouraged, hopeless, or had so many problems wondered if anything was worthwhile  AHS, University of Minnesota  Positive outlook  148 43  Where will I be in 5 years Educational aspirations  Specific to MCS street surveys – derivation unknown  Positive attitude  139 149 54  Feeling about current life circumstances Best thing about life Happiness  Specific to MCS street surveys – derivation unknown  Body satisfaction  65 66  Think of body as under, over or just right Trying to lose, gain weight or stay the same  YRBS, CDC  Note: Minor changes have been made to adjust for Canadian content or language-appropriateness (The McCreary Centre Society, 2007)  173  Appendix B: List of Hypothesized Risk and Protective Assessed by The Street-Involved Youth Health Survey (SYHS)3  3  Survey available upon request.  174 Appendix B List of Hypothesized Risk and Protective Assessed by the Street-Involved Youth Health Survey (SYHS) Variable  Item #  Item Stem from Survey  Response Format  Sexual abuse (including being forced by adult)  124 94  Have you ever been sexually abused? Have you ever been forced to have sexual intercourse when you did not want to?  Yes / No  Sexual assault by a youth  94  Have you ever been forced to have sexual intercourse when you did not want to?  No / Yes, by a youth / Yes, by an adult  Physical abuse  122  Have you ever been physically abused or mistreated by anyone in your family or by anyone else?  Yes / No  Witness family violence  121  Have you ever witnessed someone in your family being assaulted or mistreated?  Yes / No  Sexual exploitation indicator  96  Yes / No  99  Have you ever engaged in any sexual activity with a male in exchange for money or goods? Have you ever engaged in any sexual activity with a female in exchange for money or goods? In the past 30 days, did you obtain money from any of the following sources? (Mark all that apply) In the last 12 months, how often have you traded sex for: food, clothing, shelter, transportation, money, drugs or alcohol, other: Have you ever engaged in sexual activity for the following…  100  In the last 12 months, where have to traded sex:  On the street /in a trick pad/house / in a hotel / in a club/bar / on the internet / in a bathhouse / in a massage parlour / other  101  How old were you when you first traded sex?  <= 10 yrs through 18 yrs old  102  Where were you living when you first traded sex?  With my family / in a foster home / in a group home / in a shelter or safehouse / hostel-hotel-motel / with a friend(s) or boyfriend-girlfriend / in my own place / on the streets / living nowhere in particular-living all over (couch-surfing), other  97 26 98  No / Yes – by youth / Yes – by adult  Yes / No Sex trade, prostitution / Phone sex / Internet sex Never / once / 2 or more times Pimp / escort agency / to support a friend-partner or relative / other  175 Variable  Item #  Item Stem from Survey  Response Format  Multiple perpetrators of physical abuse  123  Who has physically abused you? (Index dichotomized to reflect ≥ 2 perpetrators as risk factor)  mother, father, step-parent, foster/group home parent, other relative, friend/acquaintance, romantic partner, trick/date, pimp/agency manager, police officer, stranger, other (specified)  Multiple perpetrators of sexual abuse  125  Who has sexually abused you? (Index dichotomized to reflect ≥ 2 perpetrators as risk factor)  mother, father, step-parent, foster/group home parent, other relative, friend/acquaintance, romantic partner, trick/date, pimp/agency manager, police officer, stranger, other (specified)  Attacked or assaulted (past year)  132  During the past 12 months, how many times did someone physically attack or assault you?  Never / Once / 2 times / 3-4 times / 5+ times  Ever been in-care  14  People can live in a lot of difference places for short or long periods of time. Please tell us where you live now (yesterday), and all the places in the past 12 months, and any place you’ve ever lived  House-apartment / Hotel / Parent’s home / Sister(s) or brother(s) home (step/adoptive) / Other relative’s home / Foster home / Group home / Safe house-shelter / Transition house / Squat / On the street / Abandoned building / Living nowhere, living all over (couchsurfing) / Tent / Car / Other: Child welfare-ward of the courts /  Do you think anyone in your family has a problem with… …alcohol (e.g., addicted to alcohol or is an alcoholic) …drugs (e.g., addicted to drugs) Is there anyone in your family who has ever had a mental illness? Is there anyone in your family who has a criminal record?  No / Yes, my mother / Yes, my father / Yes, another member(s) of my family / Don’t know  Index mother problems Index father problems  134 135 137 138  Family suicide history  113  Has someone in your family ever tried to kill themselves (attempted or committed suicide)?  No / Yes, within the last 12 months / Yes more than a year ago / I don’t know  Kicked out of home  23  How old were you when you first got kicked out of the home?  Range from <9 to 17+ years old  Street exposure (past 12 months) Early debut Precarious housing  13  How old were you when you first became street-involved?  Options for <9 through 17+ years old  14  People can live in a lot of different places for short or long periods of time. Please tell us where you live now (yesterday), all the places in the past 12 months, and any place you’ve ever lived.  House-apartment / Hotel / Parent’s home / Sister(s) or brother(s) home (step/adoptive) / Other relative’s home / Foster home / Group home / Safe house-shelter / Transition house / Squat / On the street / Abandoned building / Living nowhere, living all over (couchsurfing) / Tent / Car / Other  176 Variable  Item #  Item Stem from Survey  Response Format  Victim of relational aggression  132  During the past 12 months, how many times did someone: …say something personal about you that made you feel bad or extremely uncomfortable? …keep you out of things on purpose, exclude you from their group of friends or completely ignore you? …threaten to hurt you but did not actually hurt you?  Never /Once / 2 times / 3 or 4 times / 5 or more times  Victimization index  Combine 6 variables for index: (1) sexual abuse (including forced by adult), (2) physical abuse, (3) witnessed abuse, (4) sexually exploited, (5) attacked/assaulted, and (6) sexual assault by another youth.  Gender  2  What is your gender?  Male / Female / Transgender (M-F)/ Transgender (F-M)  Age  1  How old are you?  Options: 12 years old through 18 years old  Sexual orientation  95  People have different feelings about themselves when it comes to questions of being attracted to other people. Which of the following best describes your feelings?  100% heterosexual / Mostly heterosexual / Bisexual / Mostly homosexual / 100% homosexual  Ethnicity  3  What is your background? (Mark all that apply)  Aboriginal-First Nations / African / European / East Asian / South Asian / Southeast Asian / West Asian / Other: / Don’t know  Physical risk-taking  80  Have you ever used needles or other gear (crack pipes) for drugs after someone else had used it? Have you ever used tattoo equipment, body piercing equipment or shaving razors after someone else used it?  Yes / No / Not sure  0 times / 1 time / 2-3 times / 4-5 times / 6-11 times / 12+ times  81 Physical fighting (past 12-months)  129  During the past 12-months, how many times were you in a physical fight?  Disordered eating behaviour: Bingeing Vomiting on purpose  67  How often do you eat so much food in a short period of time that you feel out of control and would be embarrassed others saw you.(binge-eating or gorging)? How often do you vomit (throw-up) on purpose after eating?  68  Never / Once a month / 2-3 x - month / Once a week / 2+ times a week  177 Variable  Item #  Item Stem from Survey  Response Format  Substance use  71  Early initiation to alcohol and marijuana  77  How old were you when you tried marijuana (pot, grash, hash, weed) for the first time? How old were you when you had your first drink of alcohol other than a few sips?  Less than 9 years old / 9 or10 years old / 11 or 12 years old / 13 or 14 years old / 15 yrs old / 16 yrs old / 17 or more years old  30-day binge-drinking  75  Frequency of binge drinking: During the past 30 days, on how many days did you have 5 or more drinks of alcohol in a row, that is, within a couple of hours?  0 days / 1 day / 2 days / 3-5 days / 6-9 days / 10-19 days / 20+ days  Injection drug use recent  72  Have you ever used any of the following drugs (yesterday, past month?  Injected an illegal drug (shot up with a needle) (one item on matrix listing various drugs)  Consequences for substance use  78  During the past 12 months, have any of the following happened to you because you were drinking or using drugs?  Mark all that apply. Passed out / Got into a car accident / Got injured Got in trouble at school / Poor school work or marks / Argued with family members / Got into a physical fight / Damaged property / Lost friends/ Got in trouble with police / Broke up with boyfriend or girlfriend / Had to get treatment for alcohol or drug abuse / Had sex when I didn’t want to  Suicide attempt (past 12 months)  114  During the past 12 months, how many times did you actually attempt suicide  0 times / 1 time / 2-3 times / 4-5 times / 6+ times  Family connectedness  27-34  How much do you feel that… …people in your family understand you? …you and your family have fun together …your family pays attention to you? How close do you feel to your mother? How much do you think your mother cares about you? How close do you feel to your father? How much do you think your father cares about you? How true are the following statements: Most of the time my mother is warm and loving toward me Overall, I am satisfied with my relationship with my mother Most of the time my father is warm and loving toward me Overall, I am satisfied with my relationship with my father  Not at all / some / a lot / don’t know or doesn’t apply  Often or very true / sometimes or somewhat true / never or not true / does not apply  178 Variable  Item #  Item Stem from Survey  Response Format  School connectedness  45  How much do/did you feel that your teachers care/cared about you?  Not at all / very little / somewhat / quite a bit / very much Strongly agree / agree / neither agree nor disagree / disagree / strongly disagree  47  How much do you agree/disagree with the following: I feel/felt like I was part of my school I am/was happy to be at my school The teachers at my school treat/treated students fairly I feel/felt safe at my school How often do/did you have trouble getting along with your teachers  Community activities  50  Before you were street-involved, did you do any of the following:  Mark all that apply. Play sports or do physical activities… …without a coach or instructor …with a coach or instructor (other than in gym class) Take part in dance/aerobic classes or lessons (other than gym class) Take part in art, drama, singing or music (groups, clubs, lessons) outside of class Take part in clubs or groups such as Guides, Scouts, 4H, community, church, religious group  Extreme sports  52  Before you were street-involved, did you do any of the following extreme sports?  Mark all that apply. I didn’t do extreme sports / Backcountry skiing/snowboarding / High-speed motorized sport (e.g., motorcross) / Cliff and bridge jumping / Rock climbing / White water activities / Downhill mountain biking / Other  Current school attendance  40  Are you currently attending school?  No / Yes, regular school / Yes, alternate school  46  Never / just a few times / about once a week / almost every day / every day  179 Variable  Item #  Item Stem from Survey  Response Format  Low emotional distress  109  During the past 30 days, have you felt you were under any strain, stress, or pressure? (stress) During the past 30 days, have you been bothered by any illness, physical problems, pains, or fears about your health? During the past 30 days have you been bothered by nervousness or “nerves”? (anxiety) During the past 30 days, have you felt so sad, discouraged, hopeless, or had so many problems that you wondered if anything was worthwhile? (depression/hopelessness)  Yes, extremely so Yes, quite a bit Some (more than usual, enough to bother me) A little (usual) Not at all  110 111 112  Positive future outlook  148  Where do you see yourself in 5-yrs?  In a job / In prison / In school / Dead / Having a home of my own / On the street / Having a family / Don’t know / Other  Educational aspirations  43  When do you expect to finish your education?  Before I graduate from high-school / When I graduate from high-school / When I graduate from community college or technical institute / When I graduate from university  Reasonably positive feelings toward circumstances  139  How do you feel about your current life circumstances?  Good / Fair / Poor / Awful  Body satisfaction  65 66  How do you think of your body? Which of the following are you trying to do?  Underweight / Just right / Overweight Lose / Gain / Stay the same / Nothing  Self-reported health  55  In general, how would you describe your health  Excellent / Good / Fair / Poor  Get along with peers  Q48  How often do/did you have trouble getting along with other students?  Never / Just a few times / About once a week Almost every day / Every day  Note: Bold = responses that are relevant to the study for that variable, e.g., if hypothesized protective factor is body satisfaction, responses indicating that the individual thinks of his/her body as “just right” and is not trying to change are bolded (aligned with hypothesis of acting as risk or protective factor  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items