.l. 1 . y . n 1. t . hrwu‘gffi m .4339 5.”. ‘ if}: akrrfluu‘ .- #2:»: 51“! it A (ll .1... .5 x: a. :3 V1,?! JLUH .i .fiwa. . uhflmkfiflaiflgtfl—NK This is to certify that the dissertation entitled Adolescent Health Risk Behavior: Parent and Peer Contributions and Health Outcomes presented by Allison Schettini Evans has been accepted towards fulfillment of the requirements for the PhD. degree in Clinical Psychology Major Ptotéssor’s Signature 06/15/2003 Date MSU is an Affirmative Action/Equal Opportunity Institution LIBRARY Michigan State University PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE 6/01 c:/ClRC/DateDue.p65-p.15 ADOLESCENT HEALTH RISK BEHAVIOR: PARENT AND PEER CONTRIBUTIONS AND HEALTH OUTCOMES By Allison Schettini Evans A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Psychology 2003 ABSTRACT ADOLESCENT HEALTH RISK BEHAVIOR: PARENT AND PEER CONTRIBUTIONS AND HEALTH OUTCOMES By Allison Schettini Evans Adolescence brings with it a sizeable increase in health risk behavior, most notably, alcohol and drug use, sexual risk taking, sensation seeking, and aggression. These behaviors are so widespread that some argue that involvement in health risk behavior is developmentally normative. However, a small group of adolescents engages in such high levels of health risk behavior that they put themselves at serious risk. Fully mediated, partially mediated, and moderated models were used to examine the combined effects of parent-adolescent relationships and peer factors on adolescent involvement in health risk behavior and the impact of all three on adolescent health status. One hundred and fifty six adolescent participants (95 boys and 61 girls) involved with a rural Midwestern county’s juvenile court were recruited over a two-year period. One parent or legal guardian of the adolescent also participated in the study. Adolescent participants ranged in age from 11 to 16 years (M = 14.7 years, S1; = 1.34). Items for the measurement scales were extracted from four questionnaires: (1) the Functional Impairment Scale for Children and Adolescents — Self Report, (2) the Youth on the Fringe Survey (3) the Child Health and Illness Profile: Adolescent Edition, and (4) the Social History Questionnaire. CFA procedures determined final items for each scale as well as established the structural integrity among the various factors. Parent-adolescent relationships were defined by conflictual dependency, mutuality, psychological control, and parent as well as child report of behavioral control; peer factors by peer health risk behavior and peer acceptance of deviance; adolescent health risk behavior by aggression /sensation seeking and substance use/sexual risk taking and adolescent health status by health risk consequences, health discomfort, and satisfaction with health. Results of this research generally supported a partially mediated model describing associations between aspects of the parent-adolescent relationship, peer health risk behavior, adolescent health risk behavior, and adolescent health status. More specifically, parent-adolescent relationships directly predicted peer factors, with parent-reported behavioral control linked to peer health risk behavior and child-reported behavioral control linked to peer acceptance of deviance. High levels of conflictual dependency were associated with greater adolescent involvement in aggression and sensation seeking. However, it was the two peer constructs that had the greatest effect on adolescent involvement in health risk behavior; both were linked to adolescent aggression/sensation seeking and peer health risk behavior was associated with adolescent substance use and sexual risk taking. Moreover, conflictual dependency had direct and indirect implications for the adolescent’s health status: directly linked to health discomfort and indirectly linked to health risk consequences via adolescent aggression/sensation seeking. Parent report of behavioral control had indirect implications for adolescent health risk consequences and health discomfort while child report of behavioral control had indirect implications for health risk consequences and health dissatisfaction. Unexpectedly, peer factors had direct implications for various health outcomes. ACKNOWLEDGEMENT I would first like to thank the probation officers, court referee, and secretary at the Clinton County Juvenile Justice System for all of their help in making this study possible during the past two years. Without their interest in this study, their desire to learn more about the adolescents who come in contact with the courts, and their patience with our presence in their work space this study would not be possible. I also want to specifically thank Fred Olmsted, a probation officer in the court. Fred has been more helpful with helping gather the data than he will ever know. But more importantly, Fred has been a wonderful friend, great inspiration, and an incredibly caring person. We should all be so lucky to know someone like him. I would also like to thank the members of my committee, Kelly Klump, Rick Deshon, and Tom Luster for their continuing guidance and enthusiasm for my endeavors. I have been privileged to know and to work with them. I would like to express my deepest gratitude to my advisor, Susan Frank, who has supported, encouraged, and cheered my academic, and intellectual growth for the last five years. Your knowledge, friendship, and belief in me has helped inspire me to continue along this path that I have chosen. However, none of this would be possible without the love of my family. Nonny: thank you for your novenas and love, which have helped guide me through all my journeys. Melissa, Chrissy, and Vicky: you are my greatest and truest friends. Thank you for your iv \ ears and hearts whenever I needed them! Mom and Dad: you have given me the courage to believe in myself and pursue my dreams. Your love, support, and determination to make me happy is incredible. I love you both dearly. And to my fiancé David, who has walked every step of the road with me: your joy in my accomplishments and your faith in my success, your dedication to me during my times of frustration and sadness, and your commitment to our future, has made everything brighter along the way. You have always been the light at the end of this tunnel. The long distance has finally come to an end and I look forward to the journey we will take together. The cabin in the woods at WFS gave me a place of solace, serenity and beauty that allowed me the time and space to think, write, and rewrite; Bruno was my faithful companion, always ready give me that much needed break. This investigation was supported by the National Institutes of Health, National Research Service Award 5 F31 MH64972—01A1from the National Institute of Mental Health. TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES INTRODUCTION Clarifying Terminology Background for this Study Parental Influences Mutuality Conflictual Dependency Psychological Control Behavioral Control Peer Influences Relative Influences of Parents and Peers Implications for Health Adolescent Health Predicted Relationships Model A: The Fully Mediated Model Model B: The Partially Mediation Model Model C: The Moderated Model METHOD Participants Measures A Family History Variable Procedures RESULTS Measurement Models The Parent-Adolescent Relationship Adolescent Health Risk Behavior Peer Acceptance of Deviance and Peer Health Risk Behavior Adolescent Health Status Tests of the Predicted Models Ruling Out Possible Confounds viii \OOOQVJIUJNr—A 10 14 17 21 22 24 25 26 27 28 28 30 35 36 38 38 39 42 46 50 52 52 Tests of the Full Mediation (Model A) and Partial Mediation (Model B) 53 Relationships Between Adolescent Health Risk Behavior and Adolescent Health Status 54 Relationship Between Peer and Adolescent Health Risk Behavior 54 Relationship Between the Parent-Adolescent Relationship Factors and Adolescent Health Risk Behavior. vi 56 Combined Effects of the Parent-Adolescent Relationship and Peer and Adolescent Health Risk Behavior on Adolescent Health Status Test of the Interaction Model (Model C) DISCUSSION Item Selection Individual Measurement Models for the Four Constructs Relationships Between Constructs Parent-Adolescent Relationships Predicting Peer Factors, Adolescent Health Risk Behavior, and Health Status Peer Factors Predicting Adolescent Health Risk Behavior and Health Status Adolescent Health Risk Behavior Predicting Health Status Linking Current Study Findings to Past Research Findings Implications Limitations and Conclusions APPENDIX REFERENCES vii 60 66 68 68 71 72 73 77 79 8O 81 82 84 92 Table 1 Table 2 Table 3 Table 4 Table 5 Table 6 Table 7 Table 8 Table 9 Table 10 Table 11 Table 12 Table 13 Table 14 LIST OF TABLES Constructs and Source of Initially Selected Scale Items Parent-Adolescent Relationship Items and Standardized Regression Weights Correlations for Parent-Report and Child-Report Behavioral Control Items Correlations Among the Five Parent-Adolescent Relationship Factors Adolescent Health Risk Behavior Items and Standardized Regression Weights Correlations Among the Alcohol and Drug Use Scales (Adolescent Health Risk Behavior) Correlations Among the Five Adolescent Health Risk Behavior Scales Peer Relationship Items and Standardized Regression Weights Correlations Among the Five Peer Factor Scales (Health Risk Behavior and Acceptance of Deviance) Health Items and Standardized Regression Weights Correlations Among the Three Health Constructs Standardized Regression Weights and Relevant Correlations for the Partial Mediation Model: Parent-Adolescent Relationship, Peer Relationships, and Adolescent Health Risk Behavior Standardized Regression Weights for the Full Partial Mediation Model: Parent-Adolescent Relationship, Peer Relationships, Adolescent Health Risk Behavior, and Health Status Standardized Direct, Indirect, and Total Effects for the “Trimmed” Partial Mediation Model Tables in Appendix Table A Table B Table C Lower Versus Higher Offenses Breakdown Of Offenses Into Six Categories Original Items Selected for Scales viii 34 39 40 41 43 45 45 48 49 51 52 59 62 64 85 86 87 Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 LIST OF FIGURES Model A: Fully Mediated Model Model B: Partially Mediated Model Model C: Moderated Model Adolescent Health Risk Behavior Implications on Health Peer Implications for Adolescent Health Risk Behaviors Partially Mediated Model for Parent-adolescent Relationship Factors Predicting Peer Factors Predicting Adolescent Health Risk Behavior Final Predicted Partial Mediation Model ix 25 27 27 55 56 58 61 INTRODUCTION Adolescence brings with it a sizeable increase in health risk behavior, most notably, alcohol and drug use, sexual risk taking, sensation seeking, and aggression (US Preventive Services Task Force, 1989). These behaviors are so widespread that some theorists argue that involvement in health risk behavior is developmentally normative and can actually be “growth producing” - facilitating achievement of necessary developmental tasks, most notably, separating from parents and developing more adult- like forms of personal autonomy (Jessor & Jessor, 1977; Maggs, Almeida, & Galambos, 1995; Shedler & Block, 1990; Sibereisen, Eyferth, & Rudinger, 1986). However, a small group of adolescents engages in such high levels of health risk behavior that they put themselves at serious risk, increasing their chances of short-and long-term health problems that can ultimately result in severe morbidity or mortality. This seems especially true of transition-prone adolescents - those adolescents who are in a hurry to obtain the privileges of adulthood. Some researchers suggest that frequent and persistent patterns of health risk behavior may be the result of a lack of opportunities for age- appropriate positive forms of autonomy-seeking. However, involvement in this behavior pattern may provide adolescents with a false sense of maturity, often at the cost of physical as well as psychosocial well-being. Previous research has looked mostly at either parent-adolescent relationship influences alone or peer influences alone on adolescent involvement in health risk behavior. This study, however, explores the combined influences of parent-adolescent relationships gig peer influences on adolescent involvement in health risk behavior. In the past, adolescent development was viewed as a unidirectional transition from parent influence to peer influence, with a normative decrease in the former associated with a normative increase in the latter. However, more recent research describes the process as being more mixed, with certain types of family relationships linked to over—involvement with deviant peers and other types of family relationships operating as protective factors that buffer or minimize the effects of negative peers on involvement in health risk behavior. This study will also examine several alternative models that assess the direct (and indirect) effects of the parent-adolescent relationship as well as interactive effects of relationships with parents and peers on health risk behavior. Health risk behavior during adolescence may partly stem from parents’ difficulties addressing adolescents’ autonomy needs, which in turn, encourages adolescents to seek out compensatory experiences in ‘ relationships with deviant peers. Factors associated with these deviant peer relationships potentially fuel adolescents’ involvement in health risk behavior which ultimately can lead to negative health consequences. Clarifying Terminology Although “deviance” and “health risk behavior” are often used interchangeably, there are important distinctions between the two. In fact, health risk behavior can be seen as a subset of deviant behavior that creates a unique type of risk to adolescents with regards to their physical well-being. The term delinquency also overlaps with notions of deviance and health risk behavior. However, delinquency is limited to acts of a criminal or legal nature (i.e., breaking and entering or stealing). All delinquent acts are deviant, but only a subset of deviant and/or delinquent behavior create health risks for the adolescent, i.e., drinking under age is deviant, delinquent, and a health risk behavior, but vandalism is deviant and delinquent only and lying is deviant but neither delinquent nor a health risk behavior. The present review will be drawing from the literature on deviant and delinquent behavior, in general, and health risk behavior, specifically, given that (a) these categories of behavior overlap and (b) the literature does not provide a way to distinguish predictors specific to any one. flckground for This Study Research consistently shows that adolescents, in general, are more likely to engage in health risk behaviors than any other age group. Some researchers suggest that normative involvement in health risk behavior serves particular functions in normal adolescent development and even addresses evolving needs for autonomy, mastery, and intimacy (Irwin & Millstein, 1986). First, since “normal” development is associated with “transition into adulthood,” which encompasses such things as increasing independence, sexual awareness, physiological and cognitive maturation, and greater peer affiliation, adolescents might believe that involvement in health risk behaviors is an expression of their adult status. In addition, normal cognitive processes during adolescence, often described under the rubric of adolescent egocentrism, may place adolescents at heightened risk for involvement in health risk behavior. David Elkind argues for two aspects of adolescent egocentrism: imaginary audience and personal fable (Elkind, 1976, 1985). The former involves self-conscious behavior; the sense of always being noticed and visible, in essence, being “on stage;” reinforcing social pressures to go along with peers. The latter, the personal fable, gives adolescents a sense of personal uniqueness and individuality. On the one hand, it leads them to feel that no one can understand how they really feel (Santrock, 2000), and on the other, it results in feelings of invincibility (Lawson & Lawson, 1992; Palmquist, 1992); for example, thinking, “What happens to others cannot happen to me” (Ryan & Kuczkowski, 1994). Warnings of health risks for smoking, drinking, and taking drugs seemingly have little personal relevance. Egocentrism, in general, and imaginary audience and personal fable, in particular, are normal aspects of early adolescent development that diminish over time, usually by late adolescence. However, in a cross-sectional study of 850 adolescents in the 7th, 8th, 9th, and 12th grades, Ryan and Kuczkowski (1994) found that egocentric thinking evolved into more mature thought in the context of secure parental relationships. If parent-adolescent relationships were characterized by insecurity, egocentric thoughts remained salient through late adolescence. Notably, societal influences, such as mass media and community norms, often present health risk behaviors in a positive light. The media regularly presents role models for unprotected sexual behavior, risky sexual activities, and alcohol and drug use, to name a few (Igra & Irwin, 1996), with adult role models depicted as young, good looking, and “having it all”. Different communities and neighborhoods provide adolescents with opportunities and motivations to engage in these behaviors as well. Local peer norms create expectancies of “typical adolescent behavior (Crockett & Petersen, 1993). In some peer contexts more than others, those who do not participate in alcohol or drug use, sexual activity, aggressive behaviors, or risk taking are perceived as “stuck” in childhood, ridiculed as “Sissies” or “babies.” Cultural expectations may also influence the onset of health risk behaviors. For example, despite similar ages of “sexual debut”, the US has the highest rates of adolescent childbearing and abortion in the developed world (Blum, 1991). This is thought to reflect the relatively liberal cultural attitudes toward adolescent sexuality in the United States (Perry, Kelder, & Komro, 1993). While several developmental processes during adolescence (i.e., egocentrism, social influences) contribute to “normative” involvement in health risk behavior, there are reasons why several health risk behaviors (i.e., sexual intercourse and alcohol use) are traditionally reserved for adults. These include, but are not limited to the inherent risk in such behaviors and the high degree of maturity needed for appropriate decision-making and behavioral constraint. Furthermore, research indicates that a subset of adolescents are involved in health risk behavior above gnd beyond so-called “normative” levels. Several studies suggest that the quality of parent and peer relationships may be particularly influential in predicting adolescents’ level of health risk behavior. This study examines several aspects of these relationships as well as the interrelationship between the two. _P_arent2_11 Influences Parents have a major part to play in transmitting cultural expectations to their children. In fact, the parent-child relationship, is the most immediate and significant socialization influence on the growing child (Galambos & Ehrenberg, 1997), with increasing peer influence in addition to, rather than a substitute for parent influence during the adolescent years. Research suggests that increased reciprocity, or the lack thereof, in parent-child relationships in large part, mediates the effect of other macro- farnily risk factors, such as single parenthood or household poverty, on the adolescent’s development and mastery of developmental tasks. Normal adolescent development encompasses such things as a move towards greater freedom, responsibility, and identity- formation (Baer & Bray, 1999; Schulenberg, Maggs, & Harrman, 1997). To a great extent, the family context is the appropriate medium in which developmental tasks are achieved. The challenge for parents during this period is to facilitate adolescent development, which includes greater self-directedness, while simultaneously minimizing the adolescent’s involvement in antisocial activity. Research indicates that when parents increasingly offer adolescents opportunities to exercise their developing competencies within the context of the family, the risk for involvement with antisocial peers, as well as a range of other problems, is minimized (Fuglini & Eccles, 1993). Alternatively, failure to provide these opportunities at home increases the likelihood that adolescents will seek opportunities to exercise autonomy in relationships with peers, especially deviant peers who, like themselves, are in a hurry to acquire the privileges associated with adult social status. Unfortunately, there is no agreed upon language to describe the different aspects of relationships with parents that either facilitate or impede reciprocity and balance within the relationship. In this study, four empirically related, but conceptually distinct variables, will be used to refer to and distinguish among various aspects of the parent- adolescent relationship associated with issues of reciprocity and control: mutuality, conflictuflependencv. psychological control, and behavioral control. All four of these variables refer to processes in the parent-adolescent relationship that have direct implications for adolescent development and outcomes, and together they take into account both interpersonal and intrapsychic dimensions of adolescent autonomy development within the family context. Examining these four factors extends prior work in the field, which too rarely considers multiple aspects of the parent-adolescent relationship in general, and both interpersonal and intrapsychic factors, in particular. This study will broaden and sharpen previous research, not only providing a greater understanding of the implications of adolescent-parent relationships for negative peer involvement and adolescent health risk behavior, but also specifying components of the adolescent-parent relationship that are most predictive of specific maladaptive outcomes. Mutuality Parenting that fails to provide adolescents with opportunities for mutuality, indicated by warm, engaging adolescent-parent interactions that call for competent contributions from the adolescent (Smetana, 1995; Allen, Moore, Kumerminc, & Moore, 1998), is known to contribute to maladaptive, antisocial behavior. Mutuality has an interpersonal emphasis, with adolescents reporting high levels of mutuality in daily interactions with parents presumably enjoying a healthy balance between autonomy and connection; high levels of perspective sharing and a certain degree of challenge within a context of support (Grotevant & Cooper, 1985). Described as the product of authoritative parenting, mutuality is achieved through increasing bilateral respect and reciprocity (compared to childhood) in the adolescent-parent relationship and correlates positively and strongly with connection and parental warmth and ability to foster competence. Adolescents experiencing high levels of mutuality in their relationships with their parents experience their parents as respecting of the adolescents’ need to individuate while attending to their continuing need for advice and support. Parents facilitate adolescent development by encouraging greater participation in farmly decision-making. Adolescents in families characterized by mutually respectful or “enabling” relationships with parents (Grotevant & Cooper, 1985; Hauser, 1991; Youniss, 1989) appear less likely to have affiliations with violent peers or become involved in individual violence and health risk behavior (Dishion, Patterson, & Griesler, 1994; Henry, Tolan, & Gorman- Smith, 2001). Conflictujal Dgrendencv In addition to mutuality in adolescent-parent interactions, current research has underscored the importance of intrapsychic autonomy for adolescent development and adjustment (Frank & Jackson, 1996; Fuglini & Eccles. 1993). Psychoanalytic theorists often discuss the importance of intrapsychic reorganization during adolescence. Primary tasks include (a) repudiation of parental identifications or “deidealization,” (b) greater emotional separateness, and (c) achievement of independence. These processes allow for a reorganization of the ego-superego balance for the adolescent: a shift from superego dominance (directed by parental introjects) to ego dominance (and self-direction; Blos, 1962; Blos, 1979). Intense conflicts associated with deidealization and emotional separateness, along with anxieties linked to increasing demands for competent behavior are subsumed under the notion of “conflictual dependency,”'often indicated by adolescents’ feelings of anger and shame in relation to their parents (Frank, Avery, & Laman, 1988; Frank, Poorman, & VanEgeren, 1997; Frank, Schettini, & Lower, 2002; Hoffman 1984; Hoffman & Weiss, 1987). As adolescents become increasingly aware of their parents’ (and by implication, their own) fallibility, they may become overwhelmed by the breakdown of childhood identifications and relevant ego ideal representations. This awareness results in feelings of vulnerability, which lead the adolescent to seek solace, guidance, or other affirmation elsewhere. At home, these adolescents often respond with feelings of anger or contempt for the parent. However, they simultaneously dread the loss of connection and support and ultimately feel ashamed of their inability to live up to the same parental expectations they profess to reject (Reimer, 1996; Frank, Pirsch, & Wright, 1990). Psychological Control Presumably, parents’ use of “psychological control” makes it more difficult for adolescents to become autonomous, primarily in the intrapsychic sense. While both psychological control and conflictual dependency can be linked to feelings of anger and shame in the adolescent, there is an important distinction between the two. Psychological control refers to parental attempts to dominate the psychological and emotional development of the child (by controlling the child’s emotions, thought processes, self- expression, and attachment to parents; Barber, 1996). Psychological control can be distinguished from conflictual dependency because whereas the former refers to the adolescent ’s perception of parental attempts to dominate the adolescent’s emotions and feelings, the latter refers to the adolescent’s feelings of shame and anger because of an inability to live up to (perceived) parental expectations. Guilt, expressions of disappointment or shame, and manipulation of the “love relationship” are the primary vehicles for “controlling” the adolescent (Baumrind, 1966). Psychological control has been implicated in undermining independent functioning and self-confidence while contributing to feelings of personal distress and inadequacy (Barber, 1996; Steinberg, 1990). Relationships high in psychological control are characterized by constraining interactions between parent and adolescent that evoke feelings of shame in the adolescent, inducing anger, guilt, and anxiety, rather than supporting the adolescent’s efforts to reevaluate parental identifications. Constraining interactions with parents devalue, distract, withhold or show indifference, judge, excessively gratify, and interfere in the development of individuality while undermining the adolescent’s self-esteem (Hauser, 1991). They also may undermine the adolescent’s participation in family interactions and discourage involvement and sharing of ideas and perceptions. Consequently, they are likely to diminish mutuality as well as increase intrapsychic conflict (Hauser et al., 1984). Bemioral Control Parental monitoring and supervision of the adolescent’s activities, usually referred to as behavioral control, is a distinct aspect of "parental control" separate from psychological control (Pettit, Laird, Dodge, Bates, & Criss, 2001; Steinberg, 1990). The conceptual and empirical distinction between so-called behavioral control and psychological control has been discussed extensively, with research showing that these qualitatively distinct aspects of control lead to different behavioral and emotional outcomes (Barber, 1996; Steinberg, 1990; Steinberg, Elmen, & Mounts, 1989). Steinberg (1990) argues that psychological control has negative implications for adolescent development (i.e., impeding autonomy-seeking and identity development) whereas behavioral control (when not too harsh or punitive) is often a positive mechanism, providing adolescents with needed support and guidance. Monitoring, a fundamental component of effective behavioral control, is usually measured in terms of parent’s reported supervision of their children’s whereabouts, activities and companions (Brown, Mounts, Lambom, & Steinberg, 1993). It is this component of behavioral control that most often has been linked to positive outcomes. 10 For example, in a prospective, longitudinal, multi-informant study of mothers and their 13-year-old children, Pettit et a1. (2001) found that more, as compared to less, monitoring was associated with fewer delinquent behavior problems. Others have found similar findings, with the risk of deviance and drug abuse increased, in part, by poor monitoring of adolescent behavior (Hawkins, Catalnao, & Miller, 1992). In fact, lack of parental monitoring is related to adolescent substance use not only directly, but indirectly as well via its influence on associations with deviant peers (Dishion, Capaldi, Spracklen, & Li, 1995). This study will measure and assess the effects of parent report as well as adolescent perceptions of parental supervision and monitoring on health risk behaviors. Research indicates the relative independence of the four aspects of autonomy in the adolescent-parent relationship studied here. Behavioral control is a strategy used by parents to modify and constrain behaviors in an attempt to socialize their children to societal standards of the larger community (Steinberg, 1990; Pettit et al., 2001). Monitoring is designed to prevent adolescents’ “drift” toward antisocial peers and to fend off related increases in risk behavior, delinquency, and other antisocial behaviors (Patterson, Crosby, & Vuchinich, 1992). In contrast, high levels of psychological control appear to stem from the parents’ own intrapsychic disturbances. Disturbances such as these cause parents to feel the need to protect their “psychological power” in the adolescent-parent relationship, and in so doing, they manipulate the emotional and psychological boundaries of the relationship, ultimately stunting the child’s emerging autonomy and self-development (Pettit et al., 2001). In fact, research has found, that high levels of psychological control are associated with both delinquent problems and greater anxiety/depression (e.g., Barber, Olsen, & Shagle, 1994; Patterson et al., 1992). ll Although psychological and behavioral control can occur simultaneously, this is not always the case. Parents can be “strict” monitors without forcing psychological dependency on the child (Baumrind, 1966, Schaefer, 1965; Steinberg, 1990; Steinberg, et al., 1989). Likewise, mutuality and conflictual dependency, although correlated, can occur together a1; independently of the other. They also can have distinct implications for outcomes. For example, a study of early adolescents in grades 4 to 8 found that mutuality and conflictual dependency had independent and somewhat different effects on personality adjustment and behavioral outcome. Path analysis findings were mostly consistent with a mediated model predicting personality from both conflictual dependency and mutuality and behavioral problems from personality. However, conflictual dependency directly predicted emotional difficulties for females; further, conflictual dependency had stronger and more consistent implications for adolescent personality and behavioral difficulties than mutuality (Frank, et al., 2002). In sum, differences in experiences of autonomy and connection that adolescents have with their parents may have direct implications for behavioral outcomes. Research shows that when adolescents experience a healthy balance between separation and connectedness, report low levels of psychological control, have appropriate levels of supervision, and experience high mutuality and low conflictual dependency, they are more likely to be psychosocially mature, well-adjusted, and self-reliant, in addition to having a healthy sense of identity (Grotevant & Cooper, 1985). Most pertinently, several studies suggest that these adolescents are less likely to engage in deviant activities, including activities that potentially compromise their physical health (Lambom, Mounts, Steinberg, & Dombusch, 1991; Steinberg, Lambom, Dombusch, & Darling, 1992). 12 V‘uflk “:1’" -- .1. In contrast, adolescents with little supervision and who experience a lack of mutuality, accompanied by high levels of conflictual dependency more than likely will find it difficult to individuate and express autonomy in prosocial ways, leaving them more vulnerable to the influence of transition-prone peers (Jessor, 1993). These same adolescents appear to be vulnerable to negative outcomes in general (including depression and anxiety, negative peer influence, alcohol and drug problems, eating problems, and involvement in antisocial activity) as well as negative health outcomes that often follow (such as personal injuries, illnesses, overdoes from drug abuse, and sexually transmitted diseases; Allen, et al., 1998; Baer & Bray, 1999; Biglan, Metzler, Wirtz, & Ary, 1990; Frank etal., 1997; Frank and Jackson, 1996; Fuligni & Eccles, 1993; Kratzer & Hodgins, 1997). The negative consequences of “negative” or “non-reciprocal” relationships with parents, in the interpersonal and intrapsychic sense, are fairly well documented. Some theorists argue that too much separation without connectedness can lead to adolescents’ detachment from parents, declining school achievement, reliance on peer “subcultures”, and poor school achievement (Eccles, Buchanan, Flanagan, & Fuglini, 1991; Lambom & Steinberg, 1993; Ryan & Lynch, 1989; Simmons & Blyth, 1987; Steinberg, 1990). Relatedly, studies of college students have found that excessive angry or resentful feelings or conflictual dependence with parents are negatively related to positive college adjustment — both academic and personal and to reported emotional problems (Hoffman 1984; Hoffman & Weiss, 1987; Rice, Fitzgerald, Whaley, & Gibbs, 1995). In addition, Frank and Jackson (1996) studied late adolescent women and found that those reporting less individuated relationships with their parents also were more likely to report certain 13 personality dysfunctions, namely, interoceptive confusion (difficulties identifying internal feelings and states), feelings of ineffectiveness, and maturity fears (anxieties associated with impending adulthood); making them vulnerable to severe eating problems. This study will explore how these “negative” relationships with parents are related to increased levels of involvement in health risk behaviors, with the most problematic relationships with parents leading to non-normative levels of involvement in health risk behaviors. Peer Influences Another outcome of difficulties in the parent-adolescent relationship is that feelings of inadequacy and dependency lead adolescents to find ways to prove their autonomy and individuation outside the family and with peers. Engaging in health risk behaviors with peers is one possible mechanism through which adolescents whose autonomy struggles are constrained by parental control. It is also hypothesized, given past research, that difficulties in the parent-adolescent relationship predict excessive peer involvement during adolescence, which in turn predicts high levels of health risk behavior. Although relationships with peers (as well as parents) are important well before the adolescent years, they appear to take on a different quality during adolescence (Brown, Dolcini, & Leventhal, 1997; Hartup, 1993). Major transformations in the peer environment occur alongside changes in the parent-adolescent relationship, independently and together motivating and shaping adolescents’ behaviors and attitudes (Bemdt, 1982; Hartup, 1993). 14 Brown et al., (1997) argue that there are three types of normative peer transformations that occur during adolescence. First, friendships shift from unstable, activity-based relationships to more stable, affectively oriented relationships. Second, sexual and romantic relationships become more acceptable and, to some extent, expected during this time. Third, adolescents become increasingly associated with peer “crowds”, with each member within a crowd having similar lifestyles and values. Many other researchers identify the same transformations and most support the idea that adolescent friendships increasingly are a stable source of emotional and instrumental support (Bemdt, 1982; Bigelow, 1977; Hartup, 1993; Sharabany, Gershoni, & Hofman, 1981). Adolescent friendships also are qualitatively distinct from childhood friendships because of the adolescent’s greater degree of independence from adult guidance and management. Adolescents typically place more importance on peers than younger children and feel that parental control over the types of friendships they form is no longer reasonable (Smetana & Asquith, 1994). In addition, as they enter their teens, they begin to spend a greater amount of time with peers and less time with family (Larson & Kleiber, 1993; Larson, Richards, Moneta, & Holmbeck, 1996). Given the increased time spent with peers, peer influence over the adolescent’s behavior and attitudes is likely to increase during adolescence as well. Peers can have both positive and negative influences on an adolescent, with the latter being the more frequently discussed. The influence of peers is often discussed in terms of “peer pressure,” most-often with a negative connotation. Peers are known to pressure others (implicitly or explicitly) into engaging in deviant behavior (Maggs, et al., 1995; Oeting & Beauvais, 1987). In fact, peers are seen as the most proximal risk for involvement in 15 deviant behavior (Brook, Whiteman, Gordon, & Cohen, 1986; Dishion, Patterson, & Reid, 1988; Dishion, Spracklen, Andrews, & Patterson, 1996; Klein, Forehand, Arrnistead, & Brody, 1994). Even without assessing for deviance among peers, correlational research consistently shows a relationship between amount of time spent with peers and involvement in drug use, including tobacco, alcohol, marijuana, and so on, as well as involvement in other delinquent activities, deviant sexual practices, etc. (Agnew & Petersen, 1989; Flannery, Williams, & Vazsonyi, 1999; Patterson, Dishion, & Yoerger, 2000). Nonetheless, other researchers have said that above and beyond time spent with peers, two factors that may be even more indicative of individual behavior are: deviancy among peers and peer approval of deviance. Presumably time spent with peers leads to identification with these peers as a source of approval and support. If individuals look up to these peers or identify with them more readily than negative peer influences on adolescent behavior are especially likely to “stick” or adhere — exerting influence in behavior above and beyond the time factor alone. A study of 96 young adolescents (mean age at first contact=11 years) over a three-year period, found that adolescents who engaged in more frequent risk behaviors were likely to report greater deviant involvement in the company of peers than those less involved in these behaviors. Changes in the levels of risk behaviors covaried with changes in the frequency of activities with peers (Maggs et al., 1995). The situation is exacerbated further when “peer characteristics” (i.e., positive versus negative types) are considered: the more serious the peer deviance, the more harmful the effect on adolescents’ behaviors (Tolan & Thomas, 1995). 16 The Relative Influences of ParentsJand Peers The image of the family and peers as separate social worlds for teens is quickly being replaced by recognition of important links between these two worlds. For decades the common assumption was that there was a “trade-off” between parents and peers during adolescence, evidenced by an increased reliance on peers and a decreased reliance on parents; with peer influence swamping the influence of parents and/or schools (Brittain, 1963; Kandel & Andrews, 1987). Supporters of this view argued for qualitative differences between adolescent-peer and adolescent-parent relationships, with the former being more egalitarian and reciprocal, and the latter, more unilateral (Hartup, 1979). Greater opportunities for autonomy in relation to peers presumably results in greater distancing of adolescents from their parents, with both parents and adolescents reporting less closeness, cohesion, and engagement with each other (Collins, 1990; Collins & Russell, 1991). Researchers now argue that family and peer relationships are not two separate dimensions but rather that there is a continuity and overlap between functions of the two (Bukowski, Newcomb, & Hartup, 1998; Cooper & Ayers-Lopez, 1985). Focusing on one relationship without simultaneously assessing the other leaves the picture incomplete and skewed. Because both relationships are profoundly influential in organizing adolescent behavior, the combined influences will be a focus of this investigation. Therefore, this study will not only explore implications of peer acceptance of deviance and involvement of health risk behavior, but will do so while also considering the parent-child relationship. 17 Two alternative models for how relationships with parents and peers influence adolescent involvement in health risk behavior (and subsequent health outcomes) will be examined. One model suggests that peer factors mediate (or partially mediate) the relationship between adolescent-parent relationships and adolescent involvement in health risk behavior, the other model suggests that adolescent-parent relationships interact with peer factors to predict level of adolescent involvement in health risk behavior. Researchers find that healthy parent-adolescent relationships provide a foundation for healthy peer involvement. Parents who facilitate autonomy development via mutuality and behavioral monitoring presumably provide the adolescent with opportunities to learn important skills such as problem solving, social skills, emotional regulation, and conflict resolution, which can then be used in peer relationships (Asher, Renshaw, & Geraci, 1980; Carson, Burks, & Parke, 1987; MacDonald & Parke, 1984). Further, researchers suggest that certain parenting practices (i.e., those that promote mutuality) not only facilitate the learning of these types of skills, but more generally, are responsible for the development of certain characteristics during childhood that lay the foundations for associations with positive peers during adolescence (Brown, et al., 1993; Feldman & Wentzel, 1990). This study will not look at the reciprocal benefits of positive peer relationships on the quality of parent-adolescent relationships, but this type of reciprocal influence has been documented in a few studies. In particular, Youniss (1980) conducted a study with children from middle-childhood through adolescence and found that experiences of mutuality with peers helped adolescents coordinate their interests and beliefs in their relationship with their parents in a more egalitarian manner than during childhood. In 18 particular, Youniss (1980) argued that the negotiation and compromising skills learned in peer relationships increasingly were incorporated into the parent-adolescent relationships, thereby allowing parents to remain a major influence in the adolescents’ lives, albeit in a new, reciprocal way. This study will, however, examine whether certain types of parent-adolescent relationships are more likely to be associated with negative peer involvement, which in turn is associated with adolescent health risk behavior. More specifically, it is hypothesized that a lack of mutuality and monitoring in the face of high conflictual dependency and psychological control within the parent-adolescent relationship is likely to be associated with negative peer interactions (Fuglini & Eccles, 1993; Baumrind, 1991; Youniss, 1980). Research suggests that adolescents who are unable to obtain autonomy-enhancing experiences within their relationships with their parents are more susceptible to influence, especially negative influence, from their peers (Brown & Huang, 1995) and that negative peer influences, in turn, can result in an increase in health risk behavior For example, in a study of 1,771 children in 6’h and 7th grades using self-report questionnaires, Fuligni and Eclles (1993) found that early adolescents who perceived fewer opportunities for decision making with their parents were less likely to turn to their parents, and more likely to turn to peers (especially negative peers) for personal and instrumental support than those who perceived greater opportunities. Further, Devereux (1970) found that peer-oriented 6‘h graders (those who frequently endorsed being with and relying on peers) were most likely to come from an overcontrolling home environment. He suggests that the negative family environment “drove them into the peer environment (p. 106).” 19 Unfortunately, Shulman et al.’s (1995) findings suggest that when adolescents seek autonomy experiences from their deviant peers, they are often disappointed. Beyond a transitory sense of “we-ness” surrounding shared involvement in deviant activities, deviant peers are unable to compensate for unmet autonomy needs within the family. Deviant peers are often more tolerant of social deficits and antisociality and by jointly engaging in deviant activities, the peer group may provide a false sense of cohesion and support. Extrinsic rewards (i.e., acceptance by others and interpersonal gratification) that promote cohesion between the adolescent and his/her deviant peers, increase the potential level of deviant behavior by all, further strengthening the ties between the adolescent and peers (Edwards, 1996; Giordano, Cemkkovich, & Pugh, 1986). Research suggests that inadequate parental monitoring and inept discipline further promote involvement with deviant peers, oftentimes leading to greater adolescent involvement in risk behavior (Patterson & Stouthamer—Loeber, 1984; Steinberg, 1987). These studies lend support to Dishion et al.’s (1995) argument that poor parenting is a necessary component of risk but exerts its influence though increased association with deviant peers. Moreover, work by Brown and Huang (1995) extends these theories and findings, suggesting that peer and parent relationships may interact in predicting adolescent behavioral outcomes. As the basis to most of the work, they suggest that parental influences are “filtered” through adolescents’ experiences in peer social contexts. They agree with other theorists that negative aspects of parenting are related to increased associations with “negative” peers, which directly influences greater involvement in health risk behaviors. However, they argue that the outcomes of the 20 parent-adolescent relationship and the peer relationships may not be entirely straightforward. Rather, they propose that maladaptive peer contexts may have a debilitating effect, moderating in the direction of magr_rify_ing the negative outcomes associated with inhibitive parenting. Regardless of the types of peers, inhibitive parenting promoted deviance, but deviant behavior with peers may exacerbate the effects of constraining or non-supportive relationships with parents (Brown & Huang, 1995). Not only do adolescents in these situations receive compensatory interpersonal rewards (making up for conflicts in the parent-adolescent relationship), they also are enabled by their peers to engage in pseudo- mature but deviant acts; perhaps providing the basis for deviant peer group cohesion. Participation in deviance, including health risk behavior, with peers, therefore, helps to insulate the individual from rejection and ridicule within the peer group, rejection and ridicule that adolescents often feel within the family as well (Giordano et al., 1986). Implications for Harm In light of Brown and Huang’s research (1995), this study will examine both mediated and interactive models of parent and peer influences on adolescent health risk behavior. In addition, it will attempt to demonstrate if adolescent’s involved in high levels of health risk behavior will also be at greater risk for health morbidity. It should not be surprisingly that adolescents who engage in drinking, smoking, risky driving, etc., at drastically high levels are also likely to compromise their own health and well-being, hence the term, health risk behaviors. In fact, engaging in health risk behaviors is the most serious threat to an adolescent’s health and well being (US Preventive Services Task Force, 1989). Researchers have found that adolescent risk behaviors, most notably 21 those associated with violence, are a leading cause of morbidity and mortality in adolescence, identified by some as America’s most important public health and social problem (T olmas, 1998). Although these behaviors have profound implications for adolescents’ physical well being, there is surprisingly limited research in this domain. In addition to assessing the relative impact of various predictors of health risk behavior, this study will assess the extent to which these risk behaviors lead to generally poor adolescent health. Adolescent Heagh Adolescence is a paradox when it comes to physical health (Steinberg, 1999). On the one hand, adolescence is considered one of the healthiest periods in the life span. Some argue that many adolescents attain levels of health, strength, and energy never again experienced in their lives (Santrock, 2000). Steinberg points out the relatively low incidence of disabling or chronic illnesses, and fewer short-terrn hospital stays during adolescence. Over the past 50 years the rate of death and disability resulting from illness and disease during adolescence has decreased and new and better medical technology and health care delivery have improved the general well being of adolescence. Nonetheless, adolescence is also a period of high risk because of a rise during this period of unhealthy behaviors, violence, and risky activity. Although the “old” mortalities of adolescence have been rectified by improvements in medical practice there is a “new” mortality of adolescence that has emerged. Violence and injury are twice as likely to result in deaths among adolescents as compared to illness and disease - the converse of the trend fifty years ago. Death results more often from engagement in 22 health risk behaviors —- automobile accidents, suicide, homicide, substance use, and sexually transmitted disease (especially AIDS). The literature points to the potentially detrimental health effects of substance use, risky driving, and early sexual activity. In particular, the association between risky driving and morbidity/mortality is well known. Approximately, 78% of all unintentional injuries among youth are attributable to motor vehicle accidents and, interestingly, almost all motor vehicle deaths in youth ages 15 to 24 have been in rural areas (National Safety Council, 1993). Long-term use of tobacco, at least in the long-term, is clearly linked to lung cancer, atherosclerosis, coronary heart disease, chronic obstructive pulmonary disease, and other malignancies (Sells & Blum, 1996) Unprotected or early sexual activity can lead to pregnancy or sexually transmitted diseases. In fact, once every 104 seconds, a teenage female becomes pregnant (Children’s Defense Fund, 1994). Engaging in any or all of these health risk behaviors can also lead to a trajectory of unhappiness, overall decreased life-satisfaction, poor health, low self-esteem, poor school and work performance; the list goes on. Interestingly, adolescents often recognize that behaviors such as substance abuse and unprotected sex are potential health hazards. However, they also underestimate the potential negative outcomes that ensue (such as liver damage, sexually transmitted diseases, heart problems, general poor health, etc.). A denial of “mortality/illness” and feeling that “it will never happen to me” (as previously discussed) is potentially a function of adolescents’ sense of uniqueness and invulnerability (Elkind, 1976; Santrock, 2000). In addition, adolescents are at greater risk for poor health because they are less 23 likely than adults to seek and receive medical and dental care (Millstein, Peteren, & Nightingale, 1993). This study will examine three different aspects of general health status: overall health satisfaction, overall discomfort, and health disorders. Moreover, the model proposed here assumes that health disorders are primary and it is the presence (or absence) of these disorders that leads to health discomfort or health satisfaction. Predicted Relationships This study will examine and compare the utility of three models describing the effects of family and peer variables on adolescent health risk behavior and health outcomes. Distinctions among the three proposed models were suggested by work by Henry et al. (2001). Their research indicated that model in which family and peer factors had direct effects on adolescent behavior was far too simple; therefore, this model will not be tested in this study. Alternative models that are examined here include a full mediation model (Model A), a partial mediation model (Model B), and moderating model (Model C). In each of the models, parent-adolescent relationships are defined by: psychological control, conflictual dependency, behavioral control, and mutuality; and peer factors by peer engagement in health risk behavior (i.e., alcohol and drug use, sexual risk taking, sensation seeking, and aggression) and peer acceptance of involvement in health risk behavior. Adolescent health risk behavior is defined by four constructs (i.e., alcohol and drug use, sexual risk taking, sensation seeking, and aggression) and adolescent health status is defined by adolescent health risk consequences, satisfaction with health, and health discomfort. 24 Model A: The Fullv Mediafited Model As can be seen in Figure l, the full mediation model proposes that parent- adolescent relationships will directly “predict” peer relationships (arrow 1) which will “predict” adolescent health risk behavior (arrow 2) which will “predict” adolescent general health (arrow 3). More specifically, it is hypothesized that maladaptive parent relationships (as defined by high psychological control and high conflictual dependency together with low behavioral control and low mutuality) will lead to negative peer relationships (peers engaging in high levels of health risk behavior and peers that approve of health risk behavior). Negative peer relationships are expected to predict high levels of adolescent involvement in health risk behaviors which, in turn, will predict more negative health consequences, undermining health satisfaction and increasing health discomfort. Figure 1 Model A: Fully Mediated Model 25 Reign-Adolescent P_ec;r ACME“; edifices: Relationships Factors Health Risk Behavior Health Status Aggression _ Satisfaction geonfl‘gma' with Health pen ency Peer Health S l R’ k Risk Behavior exua. ‘5 A Mutuality Taking / l 2 3 Health Risk 5 . Consequences - Peer Approval Alcohol and 32:21:12“ of Health Risk Drug Use \ Behavior Health Psychological Sensation Discomfort COHUOI Seeking Model B: The PangHyMediated Model Model B proposes that there is also a direct relationship between parent- adolescent relationships and adolescent health risk behavior (arrow 4) and a direct relationship from parent-adolescent relationship problems to adolescent health status (arrow 5). This model is based on the assumption that adolescents who do not experience a balance between separation and individuation in their relationship with their parents not only will be more involved with deviant peers but also will seek a sense of maturity, adequacy, and/or mastery by directly engaging in health risk behaviors. These adolescents may be more prone to self-medicate, using alcohol or drugs or engage in premature sexual relationships in order to suppress negative feelings associated with conflicted parent-child relationships. A direct pathway from negative parent-adolescent relationships to adolescent health status is based on the assumption that constrictive, negative parenting can be highly stressful for the adolescent, which, in turn can compromise the adolescent’s health. Important research in the field of psychoneuroimmunology has found that stress can affect the immune system, which in turn, affects general health, behavior, and the central nervous system (Maier, Watkins, & Fleshner; 1994). The dotted line (arrow 6) in Figure 2 shows a final pathway from peer factors to adolescent health status. There is no theoretical reason to assume that such a relationship exists and hence it is not expected to be supported by the data collected for this study. 26 Figure 2 Model B: Partial] Mediated Model Peer Adolescent Factors ‘ 2 I Health Risk Behavror Parent- Adolescent Adolescent .......... Health Relationships 5 -7 Status Model C: The Moderated Model The moderated model shown in Figure 3 proposes that parent-adolescent relationships interact with peer relationships in predicting adolescent health behavior. This model is based on work by Brown and Huang (1995) suggesting that negative peer involvement magnifies the effect of negative parent-adolescent relationship on adolescent health risk behavior. The possibility that parent and peer factors interact in predicting adolescent health status also will be addressed. Figure 3 Model C: Moderated Model Peer Factors Parent x Adolescent Adolescent Peer ) Health Risk —p Health Interaction Behavior Status Parent- Adolescent Relationships 27 METHOD Participants One hundred and fifty six adolescent participants (95 boys and 61 girls) involved with a Midwestern county’s juvenile court were recruited over a two-year period. The rationale for selecting a juvenile justice sample was to ensure a wide range of involvement in health risk behaviors as well as a wide range of relationships with parents and peers. This study assessed adolescents who came in contact with the juvenile justice system (with or without official hearings). This sampling procedure maximized the chances of selecting adolescent participants involved across a range of health risk behavior, behaviors described as part of “typical” or “normative” adolescent experimentation (i.e., minor in possession of tobacco or alcohol, vandalism, curfew violation, etc.) as well as behaviors considered to be more “pathological” or “non- normative” (i.e., assault and battery, larceny, felonious destruction). The offenses of the adolescents ranged from first time to chronic, and minor to severe. In this sample, 89 (57%) of the adolescent offences were considered “lower” or minor offenses while 67 (43%) were considered “higher” or more severe offenses. See Appendix (Table A) for descriptions and crimes falling within the lower and higher categories. A more conceptual break down of the offenses (similar to the categories in this study) indicated that 65 (42 %) were classified as substance use offenses, 35 (22 %) as aggressive towards others, 5 (3 %) as non-aggressive towards others, 15 (10 %) as aggressive towards property, 11 (7 %) as non-aggressive towards property, 44 (28 %) as sensation seeking offenses. See Appendix (Table B) for clarification of categories. Unlike the higher-lower offenses, these categories were not mutually exclusive; those 28 with more than one offense were categorized as such (i.e., an adolescent with an assault and battery charge as well as a truancy charge was categorized as both aggressive - other aria sensation seeking). Approximately 17 (11 %) adolescents had multiple crimes. Most (N = 140) of the study participants were recruited when they appeared before a Hearing Officer in the County Juvenile Justice Court over a two-year period. An additional 16 were referred by probation officers and met inclusion criteria if they (I) committed an offense other than a Criminal Sexual Conduct (CSC) offense (research shows that CSC children experience a unique set of correlates; Office of Juvenile Justice and Delinquency Prevention Report, 2000), (2) had parental consent, (3) were within the same age range as the larger sample, and (4) had been released and been living in the community for a minimum of 3 months, if recently incarcerated. A parent or guardian (whoever was available at that time, or, if two were present - whoever knew the child best) also participated in the study. Parent participants (135156) included 117 mothers (75%), 29 fathers (18.6%), 5 grandmothers (3.2%), 4 step-parents (2.5%), and 1 “other” legal guardian. Participants ranged in age from 11 to 16 years M = 14.7 years, SD = 1.34), with the majority (92.8%) between the ages of 13 and 16, and most were in grades 8, 9, and 10 (20%, 25.7%, and 26.4%, respectively). Most were White (70.6%, p=108). Twenty-one (13.7%) of the adolescents classified themselves as being from mixed ethnicities, 4.6% (p = 7) were Hispanic, 4.6% (p = 7) were Native American, and less than 2% classified themselves as African American or Asian. Five percent were missing ethnicity data. Data on household composition available for 145 participants indicated that most (a = 102 or 70%) of the adolescents lived in a two-parent home: 49 (32%) with two 29 biological parents and 44 (30.3%) with their biological mother and a step- or adoptive- father. Forty-three (29.7%) adolescents lived in a single parent home and, of these, 33 lived with their biological mother. The average annual income for the sample was between $30,000 and $45,000. Less than 5% had incomes under $12,000 (p = 6) and approximately 14 percent (a = 19) earned greater than $70,000 per year. Parent report indicated that 19 adolescents were in part- or all-day special education, 22 had an ADI-{D diagnosis, 8 had a learning disability diagnosis, one had speech or language problems, and none had a history of mental retardation or brain injury. Ten parents reported “other” as a diagnosis (some wrote in “depression) Measures Four separate questionnaires were used to make up the relevant scales: (1) the Functional Impairment Scale for Children and Adolescents — Self Report (FISCA-SR), (2) the Youth on the Fringe Survey (YOFS), (3) the Child Health and Illness Profile: Adolescent Edition (CHIP-AB), and (4) the Social History Questionnaire. All questionnaires were chosen for their good psychometric properties, broad application, and relevance to the area of study. The Functionalimpairment Me for Children and Adolescents- Self Report (FISCA-SR) has 172 items, identical in content to the parent report FISCA (Frank & Paul, 1995; Frank, VanEgeren, Fortier, & Chase, 2000a; Frank, Paul, Marks, & Van Egeren, 2000b), and geared to a sixth grade reading level. Frank et al. (2000a) reported adequate internal reliability for seven of eight impairment domains assessed by this instrument (i.e., Delinquency, Undercontrolled Aggression, Self-Harm, Emotional Impairment, School, Thinking, and Alcohol and Drug Problems; the Home scale is the 30 exception). A recent study assessing agreement between parent and self-report forms of the FISCA, showed that a hypothesized three-factor model of child functional impairment fit the data for both sources. The model distinguishes among (a) undercontrolled (i.e., overt) aggression, (b) social role violations (covert deviances), and (c) self-focused impairment (i.e., emotional dysfunction and self-harm; Frank et al., 2000a; Frank et al., 2000b). Correlations between parents’ and adolescents’ functional impairment scores (ranging from .25 to .59) as well as paired comparisons of means, generally identified stronger agreement in “public” (e. g., school) than “private” (e.g., thinking) domains. Overall, Frank et al. (2000a) also found that parent-adolescent agreement coefficients for more than half of the FISCA domains exceeded the magnitude of association typically reported in the literature (Achanbach, McConaughy, & Howell, 1987). Because measures (especially youth-report) are relatively new, validity data are somewhat sparse. However, recent studies lend support to FISCA predictive validity in both inpatient and normal samples (Frank et al., 1998; Frank et al., 2002). Items from the FISCA-SR use several different formats (true-false, multiple choice, Likert scales); therefore, scales combining selected items from this and other measures, were translated into z—scores. The Child Health and Illness Profile — Adolescent Edition (CHIP-AB; Starfield et al., 1995; Starfield et al., 1999) is a general measure of health status, which consists of 6 domains (Discomfort, Disorders, Satisfaction with Health, Achievement, Risks, and Resilience) and 20 subdomains. This instrument assesses not only manifestations of ill health and current health, but characteristics likely to be related to health in the future, including health risk behaviors. In a school sample, Starfield and colleagues 31 (1995) showed that internal consistency reliability for the individual risk subdomains ranged from .77 to .87 with satisfactory test-retest reliability over a 3-month period (Pearson’s r: .87). Their research suggests that the CHIP-AB is suitable for the assessment of adolescent health status and it may be used to assess changes over time as well as in response to health services interventions. The Youth on the Fringe Surveyl (YOF) was developed by Frank and Schettini (2001) for an earlier study and is a compilation of various measures with sound psychometric properties (e.g., Cheung, 1997; Leffert et al., 1998; Offer, Ostrov, & Howard, 1989; Stutman & Lich, 1985). This adolescent self-report questionnaire has 94 items associated with a total of 14 different domains (including, mutuality, conflictual dependency, supervision, peer deviance, peer approval of deviance). Items for the various YOF scales, for the most part, were extracted from other instruments, including the Offer Self-Image Questionnaire (OSIQ; Offer at al., 1998) the Parental Separation Inventory (PSI; Hoffman, 1984), Parental Relationship Inventory (PR1; 44), the Children ’s Report of Parental Behavior Inventory (CRPBI-30; Schaefer, 1965; Schluderman & Schluderman, 1970), and Peer Support for Deviance Scale (Cheun g, 1997). The Social History Ouestionna_ir_e was developed by Susan J. Frank, Ph.D. as part of an intake procedure at a local adolescent inpatient facility. In addition to demographic information, questions on this form ask about supervision and biological family history and psychopathology. For the most part, items for individual scales associated with each of the four construct domains assessed in this study (adolescent-parent relationships, peer factors, 32 adolescent health risk behavior, and general health status) were extracted from the larger pool of items based on conceptual relevance. In a few instances, an already- established measure, included as a subscale in one of the larger questionnaires, was used in its entirety. For example, the YOF included a 7-item already-established scale assessing “conflictual dependency.” In every instance, the initial group of scale items identified as associated with each target construct was subjected to confirmatory factor analysis procedures to ensure the structural integrity of the scale while minimizing overlap among scales. Table 1 summarizes the scales associated with each of the construct domains as well as the source questionnaire for each scale. The table reports the original number of items selected for each scale. Note that the 27 items for the peer factor scales that were extracted from the FISCA-SR were modified to ask about peer, rather than respondent’s, behavior; for example, by asking “in the last three months, have your friends threatened to hurt someone” rather than “in the last three months have you threatened to hurt someone.” Originally, the study also intended to examine peer sensation seeking (similar to the one included in the adolescent health risk behavior construct); however because of 33 Table 1 Constructs agl Source of Initially Selected Scale Itema Construct Domaip Scalgnd Number of Original 1te_ms_ Parent-Adolescent Relationships Mutuality, 9 items Conflictual Dependency, 7 items Psychological Control, 3 items Behavioral Control: parent, 4 items Behavioral Control: child, 4 items Adolescent Health Risk Behavior Alcohol and Drug Use, 7 items Aggression, 16 items Sexual Risk Taking, 4 items Sensation Seeking, 7 items Peer Factors Peer Health Risk Behavior Alcohol and Drug Use, 12 items Aggression, 10 items Sexual Risk Taking, 5 items Peer Acceptance of Deviance, 5 items Adolescent Health Status Health Satisfaction, 7 items Discomfort, 32 items Health Risk Consequence, 27 items Source Questionnaire YOF YOF YOF Social History YOF FISCA-SR FISCA-SR FISCA-SR CHIP-AB CHIP-AB, FISCA-SR (m) FISCA-SR (m) CHIP-AB, YOF, FISCA-SR (m) YOF CHIP-AB CHIP-AE CHIP-AE Note. (m) indicates “modified” to ask about peer rather than respondent behavior. 34 a clerical error, this 7-item scale was not available. The number of items for each scale is the number of items originally selected. These original items are in Appendix (Table C). The actual number of items for each scale changed considerably as a result of CFA analyses; therefore, final items for each scale are reported in the Results section. In addition, scales for the three health status constructs, were completely revamped, again based on results of the CFA procedures. A Familv History Variablg A final scale was developed from the Social History questionnaire to assess family history of health risk behavior; this was a proxy measure of genetic transmission of health risk behavior. There has been ongoing debate and consideration regarding the role of genetic factors in antisocial and deviant behavior (DeLalla & Gottesman, 1989; Rodgers, Muster, & Rowe, 2001; Rowe, 1994). Arguments suggesting biological risk for adolescent health risk behavior suggest that genetic factors may account for many identified associations between adolescent-parent relationships and adolescent health risk taking behavior. For example, genetic contributions to parent drinking may produce negative parent-adolescent interactions an_d adolescent risk for alcohol abuse. Although a detailed discussion of this potential confound is beyond the scope of this study, a scale assessing family history of health risk taking behavior was developed as a proxy measure of genetic transmission. In particular, six items were used to tap adolescents’ family histories of engaging in risk taking behaviors: family history of alcohol use, drug use, criminal behavior, time spent in prison, extreme aggression, and spousal abuse. Each item was scored as “1” if one or more biological relatives (i.e., grandparents, aunts, uncles, cousins, nieces/nephews) was family history positive for 35 the item and “0” if no relative was family history negative for the item. The six items were summed to derive a single score ranging from 0 to 6. Procedures Participant recruitment for this study began after a determination was made regarding the dispensation of the adolescent’s case, i.e., warning and dismissal, consent probation, formal probation, or referral for a formal hearing. The Hearing Officer asked the family to speak with one of the research assistants from this study. Because dispensation was already determined, youth and parents could be reassured that their decision to participate (or not) would not affect the court’s response to the juvenile’s offense (also stated in the information form). Other assurances of confidentiality, as required by the university human subjects review board, were also provided. Youth and parents were informed of their opportunity to participate in this study and received monetary compensation for their time. After obtaining written consent from the parent and written assent from the adolescent, the family indicated whether they preferred to complete the measures: (1) immediately following the hearing, (2) at a later date at the court facility, or (3) at a later date in their home or some other facility (i.e., the study office). Those who were unsure of their availability, provided a phone number where they could be reached to set up a later appointment. Those without a telephone were asked to give the phone number of a friend or relative as well as to call the project offices at the university to set up a time to come to the courts to complete the testing. All participants (or their contact person) not tested at the court the day of their hearing, were called and reminded the night before their appointment. The various testing options combined with the reminder call reduced the possibility of losing hesitant, 36 situationally stressed, or forgetful families who might not otherwise keep a scheduled appointment. Out of all eligible adolescents coming through the juvenile court, a total of 76 adolescents (57 boys and 19 girls) ranging in age from 11 years, 5 months to 16 years, 4 months (M = 14.54, SD = 1.27) chose not to participate in this study. Most were White (p=53, 71.1%). Two of the adolescents classified themselves as being from mixed ethnicities, 3 were Hispanic, 1 was Native American, and 1 was African American. Approximately 20 percent (p = 15) were missing ethnicity data. 37 RESULTS Measurement Models Confirrnatory factor analysis procedures were used to establish the structural integrity and support distinctions among measures of four components of the parent- adolescent relationship: conflictual dependency, mutuality, psychological control, and behavioral control. Each construct was examined in a separate, rather than a single, multifactorial CFA analysis procedure because of constraints imposed by the size of the present sample. Loehlin (1992) conducted a monte carlo simulation study using confirmatory factory analysis models and concluded that for a model with two to four factors, investigators should have 200 subjects. Others have argued that 15 cases per predictor is a good rule of thumb (Stevens, 1996). Testing a multi-factorial measurement model in smaller samples may lead to inaccurate parameter estimates and standard errors. The Larent-Adolescent Relations—hip The 7-item measurement model for conflictual dependency held up well statistically (X2 = 21.39, pi = 14, p = .09; alpha = .74, GFI = .97, Tucker-Lewis Index = .94, RMSEA = .06, SRMR = .05 see Table 2, below). However, the 9 items for the mutuality scale did not cohere. Three items with low, statistically insignificant regression weights were eliminated, along with an additional item judged to be a conceptual outlier (“Most of the time, my parents are happy with me”). A follow-up CFA procedure on the remaining 5 items resulted in an adequate fit with the observed data (see Table 2; X2 = 8.63, d_f = 5, p = .12, alpha=.71, GFI=.98, Tucker-Lewis Index=.95, RMSEA=.07). 38 Table 2 P_arent-Adolescent Relationship Items and Standardized Regression Weights Standardized It_err§ Reggession Wei ghts* Conflictual Dependency My parents are ashamed of me. .50“ If I talk things over with my parents, I’m just asking for trouble. .62“ My mother expects too much of me. .52“ My parents make me angry. .64“ My father expects too much of me. .41“ I have been angry with my parents for years. .61“ Usually I feel that I am a bother at home. .46“ Mutuality I feel that I have a part in making family decisions. .45“ I usually go along with my parents advice. .33“ My parents respect my desire to be an independent person. .83“ My parents accept my need for privacy. .77“ It is all right with my parents if I disagree with them. .46“ Psychological Control My parents are people who are less friendly with me if I don’t see ** things their way. My parents are people who avoid looking at me when I have ** disappointed them. If I’ve hurt my parents’ feelings, they stop talking to me until I please ** them again. BehavioralQControl: Child Report Items During all or most school weeks, I’m supervised by a parent or adult .94“ over age 19.. During the weekend, I am supervised by a parent or another adult .60“ over age 19.. My parents know where I am when I am not home. .35“ Behmoral Control: Parent Report Item_s_ Adults in the home did not know the whereabouts of the child. .82“ Child was poorly supervised outside the home. .83“ Adults had to leave the child at home during the day without a .50“ responsible adult to supervise. Adults had to leave the child at home at night without a responsible .21’1 adult to supervise. a p < .05, bp < .01, “p < .001. “E. A confirmatory factor analysis could not be conducted on this scale because it only has 3 items; standardized regression weight estimates could not be calculated. 39 Statistical constraints associated with the CFA procedure made it impossible to examine the three—item psychological control scale by itself. Therefore, the psychological integrity of the psychological control scale as well as the distinctions among the conflictual dependency, mutuality, and psychological control constructs (all of which describe emotionally laden aspects of the adolescent-parent relationship) were assessed in a single CFA analysis. This results indicated a good fit with the observed and predicted model (x2 = 104.96, at = 87, p = .09, RMSEA = .04, Tucker-Lewis Index = .95, SRMR = .07, GFI = .92). Correlations among the eight items initially selected to measure behavioral g)n_tro_l (4 based on parent report and 4 based on child report) showed stronger associations within, as compared to between, reporters (see Table 3). Table 3 Correlations for Parent-Report; and Child-Report Behavioral Control Item_s Childl Child2 Child3 Parentl Parent2 Parent3 Parent4 M 1 .57b .20“ .33 b .26b .15 .14 gag; 1 .23 b .17 a .04 .03 .02 _c_m1_d_3 1 .13 .01 - .01 .05 am 1 .68 b .37 b .15 my 1 .43 b .14 23311; 1 .34b .P_ar_gn_tA 1 a p < .05, b p < .01. Note. Individual items can be seen on page 40. 4O Seemingly, children’s and parent’s subjective experiences of monitoring and supervision were independent of one another. A CFA analysis testing a 2-construct (parent and child) model of monitoring led to the removal of one insignificant child- report item. The final model based on a 3-item child and a 4-item parent behavioral control scale provided a fairly adequate fit to the data (Table 2). Although the chi square for the observed versus the hypothesized model was significant (X2 = 22.634, _d_f_= 13, p = .05), other indices of model fit were respectable (GFI = .96, Tucker-Lewis Index = .93, and RMSEA = .07; parent supervision scale alpha = .70; child supervision scale alpha = .65). In addition, the two behavioral control factors were distinct from (a) psychological control (X2 = 42.51, d_f = 32, p = .10; Tucker-Lewis Index = .96; RMSEA = .05; GFI = .95); (b) mutuality (x2 = 88.73, at = 51, p = .05; Tucker-Lewis Index = .95; RMSEA = .05; GFI = .94); and (c) conflictual dependency (X2 = 77.41, d_f = 62, p = .09; Tucker- Lewis Index = .95; RMSEA = .04; GFI = .94). Moreover, as can be seen from Table 4, correlations among the unit-weight scales for each variable are fairly low, supporting distinctions among these conceptually related constructs. Table 4 Correlations Among the Five Parent-Adolescent Relationship Fgors .1. 2 3 fl 5 1. Mutuality 1 -.14 -35" .25b .10 2. Psychologicagiontrol - 1 .41b -.04 -.11 3. Conflictuial Dependency 1 -.06 -.21a 4. Behafiral Control: Child 1 .17"| 5. Behamral Control: Parent 1 41 Adolescent qulth Risk Behavior Confirmatory factor analysis was used to examine the integrity and independence of the four adolescent health risk behavior scales. As a first step, items selected as indicators of each of the four adolescent health risk behavior variables were subjected to confirmatory fact analysis procedures. Final scales for each construct are shown along with their regression weights in Table 5. The original seven items defining smation seekiyg did not cohere together as a scale, but after dropping one non-significant item, a six-item scale proved adequate. Later results indicated that one of the six items loaded highly on another risk taking construct. Therefore, a confirmatory factor analysis was conducted on a 5-item construct which not only met stringent criteria for fit (X2 = 3.72, g = 5, p = .59; alpha = .70; Tucker-Lewis Index = 1.02; RMSEA = .00; GFI = .99) but also proved to be statistically superior to the 6-item construct or2 difference = 10.99, d_f change = 4, p < .05). or the 16 items originally selected as indicators of aggression, 6 were dropped because of inadequate variability. These items generally measured very severe acts of aggression (i.e., “During the past three months, how often did you kill someone?”). A CFA procedure on the remaining 10 items indicated an inadequate fit. After eliminating non-significant indicators, a follow-up confirmatory factor analysis supported the structural integrity of a 5-item scale (X2 = 3.58, _c_l_f_= 6, p = .61; alpha = .74; Tucker-Lewis Index = 1.02; RMSEA = .00; GFI = .99). 42 Table 5 Adolescent Health Risk Behavior Itema_ and Stzmdardized Remssion Weights Standardized m Reggession Weights* Aggression During the past 3 months, how often did you. . .argue or fight with .57 other kids? bully, threaten, or shove other kids? .67 tease, ridicule, or pick on other kids? .73 Annoy others on purpose, or damage other people’s things on .41 purpose? say really mean or cruel things to others? .62 Sensation Seeking During the past 3 months, how often did you. . .race on a bike? .45 do something risky? .78 Break a rule for the thrill of it? .51 steal or shoplift? .47 slip out at night when your parents thought you were still asleep? .63 Alcohol Use During the past 3 months, how often did you drink alcohol? .94 During the past 3 months, how often did you drink enough alcohol to .97 get intoxicated or drunk? Drug Use During the past 3 months, how often did you use drugs or inhalants? .97 During the past 3 months, how often did you use enough drugs or .96 inhalants to get high? During the past 3 months, how often did you sell drugs? .68 Seml Risk Takrlrg During the past 3 months, how often have you had sex with a lot of -- different people, or have had unprotected sex? Latent Constructs Adolescent Health Risk Behavior I Aggression .22 Sensation Seeking .98 Adolescent Health Risk Behavior 11 Alcohol Use .57 Drug Use .77 Sexual Risk Taking .50 *Note. All standardized regression weights p < .001 43 Seven items targeting alcohol 2m drug use also did not provide a good fit to the data. However, previous research suggests that alcohol and drug use during adolescence have different correlates and might best be captured as two distinct constructs (Kandel & Yamaguchi, 1993). Therefore, CFA procedures were used to assess the integrity and independence of separate drug and alcohol scales (2 and 3 items, respectively). For these procedures, two items that did not distinguish between alcohol and drug use were eliminated, (e.g., “in the last three months, did you drink alcohol or use drugs before d1iving?”). Correlations among items targeting the same substance were higher (r’s ranged from .66 to .91) than items targeting different substances (r’s ranged from .30 to .46; see Table 6). A confirmatory factor analysis supported the structural integrity of the 2-construct model (X2 = 5.83, _d_f = 4, p = .21; Tucker-Lewis Index = .99; RMSEA = .06; GP] = .99; alpha for the alcohol scale = .95, alpha for the drug scale = .85). Finally, among the four items originally selected to measure sexual risk taking only one item had adequate variability, “I have sex with a lot of different people or have unprotected sex.” Therefore, this construct was measured by the single item. 44 Table 6 Correlations Among Alcohol Use and Drug Use (Adolescent Health Risk Behavior) Alcohol 1 Alcohol 2 Drug 1 Drug 2 Drug 3 Alcohol 1 l .91 a .46 a .44 a .30 3 Alcohol 2 - 1 .46 a .47 a .31 D_ru_g_1_ 1 .93 “ .66a M 1 .663 Drug; 1 ap < .01. _N_o_ta. Individual items can be seen on page 40. Table 7 shows the correlations among the final sensation seeking, aggression, alcohol use, drug use, and sexual risk taking scales. Although sensation seeking and aggression, like alcohol and drug use, were fairly highly correlated, a CFA analysis supported the distinction between the sensation seeking and aggression constructs (X2 = 41.47, if = 34, p = .18; Tucker-Lewis Index = .91; RMSEA = .06; GFI = .93). Table 7 CorrelaLions Among the Five Adolescent Health Risk Behavior Scalg SS* Aggression Alcohol Drug Sexual Risk* as 1 .45 a .10 .25 a .12 ggr_ession - l .06 .21 a .12 Alcohol Use 1 .44 a .30 a Drug Use 1 .37 a Sexual Risk 1 a p < .01. *Note. SS indicates sensation seeking and sexual risk indicates sexual risk taking. 45 Nonetheless, when the overall pattern of correlations was considered as a whole, it suggested that relationships among the adolescent health risk behavior variables might be best summarized by a two-factor model: with aggressiorflnd sensation seekflgas indicators of one latent factor and sexu4al risk-taking and subsgnce (alcohol and drug) use as indicators of another. Results of a follow up confirmatory factor analysis supported the two factor model (the relationship between the two factors was constrained to .001 to conserve degrees of freedom; X2 = 8.07, d_f = 5, p = .15; Tucker-Lewis Index = .94; RMSEA = .06; GFI = .98). Regression estimates for items associated with each factor are shown in Table 5. Peer Acceptance of Deviance and Peer Heahi Risk Behavior The procedure for developing scales measuring peer acceptance of deviance, peer alcohol and drug use, peer sexual risk taking, and peer aggression were essentially the same as those used to develop scales for the adolescent health risk behavior variables. A CFA on the 5 items selected to measure peer acceptance of deviam resulted in an adequate fit (X2 = 7.72, d_f = 5, p = .172; alpha = .884; Tucker-Lewis Index = .99; RMSEA = .06; GFI = .98; see Table 8 for items). As stated previously, these items assessed the degree of embarrassment the respondents experience when peers observe them engaging in health risk behavior (low embarrassment suggests high peer acceptance). As might be expected given the findings for the adolescent health risk behavior variables, peer substance use was best defined by two separate alcohol and drug use constructs (a CFA using all items resulted in a poor fit). Correlations indicated that marijuana use was more strongly associated with the alcohol as compared to the “hard” drug use variables (although related to both). However, the marijuana item was 46 dropped because the distinction between latent constructs describing alcohol (and marijuana) versus (other) drug use was less clear when it was included in the analysis. CFA analysis assessing the structural integrity of a 3-item alcohol use scale and an 3 item hard drug use scale (see Table 8) generally supported the assumption of two correlated but distinct constructs even though the correlation between the two was quite high (g = .80; x2 = 17.07, a: = 8, 12 = .03; Tucker-Lewis Index = .98; RMSEA = .08; OF] = .96). The two factor model was significantly superior to a one-factor model for substance use (X2 difference = 37.39, at difference = 1, p < .01). The original 5-item peer sexual risk_takl_'pg scale was reduced to 4 items (see Table 8) because one item had very little variance (“. . .how often did a peer sexually abuse or molest someone?”). A confirmatory factor analysis on the 4-item peer sexual risk taking scale supported its structural integrity (X2 = 4.87, d_f = 2, p = .09; Tucker- Lewis Index = .92; RMSEA = .09; GFI = .99). The original 10 items defining peg aggression did not cohere together in a confirmatory factor analysis. Hence, items that were poor indicators of the latent construct or later proved to cross-load on other peer health risk behavior factors were deleted from the next analysis. Fit statistics for a 6- item, 5—item, and 4-item aggression scale were all adequate; however, the 4-item scale (see Table 8) provided the best fit when evaluated in combination with the other peer health risk behavior scales (X2 = 1.10, a: = 2, p = .58; alpha = .81; Tucker-Lewis Index = 1.01; RMSEA = .00; GFI = .99). 47 Table 8 Peer Relationship Items and Standardized Regression Weights Items Peer Aggression As far as you know, have your friends done the following in the past three months. . .argued or fought with someone. lie, con, manipulate, or take advantage of others. annoy others on purpose. threaten to hurt someone. Peer Sexual Risk Behwr have sexual intercourse. act inappropriately seductive, or do unusual or inappropriate things of a sexual nature in public. act sexually promiscuous or loose, engage in high risk sexual behaviors or have unprotected sex. have any of your friends ever been pregnant (girls) or gotten someone pregnant (boys). Peer Alcohol Use drank beer or wine. drank hard alcohol or mixed drinks. had five or more drinks in a row. Peer DraaUse cocaine, ice, or crack. injected other drugs. carry or sell drugs. Peer Acceptance of Deviamca If your peers saw you smoking cigarettes, how embarrassed would you be? If your peers saw you getting drunk, how embarrassed would you be? If your peers saw you destroying other people’s things, how embarrassed would you be? If your peers heard you swearing, how embarrassed would you be? If your peers saw you or heard about you having sex, how embarrassed would you be? _L_atent Peer Health Risk Behavior Scale Peer Aggression Peer Sexual Risk Behavior Peer Alcohol Use Peer Drug Use *Ngta. All standardized regression weights p < .01. 48 Standardized Reggession Weights* .62 .78 .82 .68 .53 .84 .32 .66 .94 .98 .91 .71 .62 .84 .86 .94 .65 .78 .68 .72 .80 .85 .80 Correlations among the peer health risk behavior scales, shown in Table 9, suggested a high degree of overlap among the four scales describing peer health risk behavior. These findings suggested a simpler 2-factor model, with one latent factor identified by peer health riskfibemvior and a second, by perceptions of peer acceptance of Mg. Findings from a confirmatory factor analysis were consistent with the notion of a single “peer health risk behavior” construct (X2 = .48, fl = 2, p = .79; Tucker-Lewis Index = 1.02; RMSEA = .00; GFI = .99; see Table 8 for each item’s weighting on the latent construct) and a follow-up analysis supported the distinction between the p_ee_r haalth risk behavior construct and the peerJacceptance of deviance construct (X2 = .3187, £1! = 26, p = .20; Tucker-Lewis Index = .99; RMSEA = .04; GFI = .96). Table 9 Correlations Among the Five Peer Factor Scales LHealth Risk Behavior and Acceptance of Deviaggfi M1 _Dpug _Sflual Acceptance of Aggression & E M. Deviance Agggession 1 .61a .57 a .69 a .31 a Alcohol Use 1 .68 a .68 a .41 a Drug Use 1 .63 a .37 a Sexual Risk * 1 .41 Accept. DCVIM l 3‘ p < .01. *Note. All scales in table refer to peer behavior. Sexual risk refers to sexual risk taking and accept. deviance indicates acceptance of deviance. 49 Adolescent Health Sprtya Three scales from the CHIP-AB purportedly assessing Satisfaction with Health, Health Discomfort, and Health Disorders initially were selected as indicators of the three adolescent’s health status variables included in the predicted models (health consequences, satisfaction, and discomfort). However, confirmatory factor analyses assessing the integrity of each of these scales resulted in a poor fit between the observed and predicted measurement models. In fact, closer inspection of the items associated with each of the dimensions suggested that several items, at face value, were unrelated to the constructs of interest to the present study. Moreover, a number of items in the Health Discomfort scale overlapped conceptually with the Health Disorders scale, with the Health Discomfort items describing the direct consequences of items in the Health Disorders scale (e. g. “having a cough,” a health discomfort item and “having bronchitis,” a health disorder item). Therefore, a subgroup of items from the three CHIP-AB dimensions that, at face value, appeared to measure each of three (presumably) distinct constructs of interest to this study, i.e., health satisfaction, health discomfort, and health risk consequences, were selected from the larger pool of CHIP-AB items. Both conceptual relevance and statistical significance (based on CFA procedures) were used to select the final sets of items that best captured each of these three constructs (see Table 10). 50 Table 10 Health Items and Standardized Regression Weights W229! Ileana Regzession Weights* Haalth Satisfaction I feel full of energy. .62 I resist illness well. .74 I recover quickly from illness. .73 I am happy with my health in general. .53 Haalth Discomfort In the past month, how often did you have .....stomach aches? .56 Pain that really bothered you? .49 aches, pains or soreness in your muscles or joints? .57 a headache? .46 trouble eating or have a poor appetite? .27 tire easily or feel like you had no energy? .30 feel really sick? .38 Risk Consequence In the past 12 months, .24 did you have any of the following injuries. . ..bad cut or scrape? bad sprain or torn ligament? .27 broken bone? .07 bad head injury or concussion? .07 gun shot wound or stab wound? .07 bite from another person or animal? .33 bad burn? .20 did anyone physically hurt you on purpose? .25 * Nata. All standardized regression weights p < .01. A confirmatory factor analysis supported the structural integrity of a 4—item @111 satisfaction scale (X2 = 5.32, d_f = 2, p = .07; Tucker-Lewis Index = .94; RMSEA = .08; GFI = .98) and a similar procedure also pointed to the structural integrity of a 7-item filth discomfort scale (X2 = 16.85, Qf = 14, p = .26; Tucker-Lewis Index = .99; RMSEA = .04; GFI = .97). Additionally, a confirmatory factor analysis indicated respectable structural integrity for an 8-item health consequences scale (X2 = 22.94, g = 20, p = .29; Tucker-Lewis Index = .95; RMSEA = .03; GFI = .96). As can be seen in Table 11, 51 correlations among the three health constructs do not exceed .29, supporting the uniqueness of each variable. Table 11 Correlations Among the Three Health Constructs Haalth Discomfort Health Satisfaction Risk Confluence Health Discomfort 1 .25 a .29 a Health Satisfaction l .08 _Ri_sk Consequence 1 a p < .01 Tests of the Predicted Models Ruling Out Possible Confounds Before testing the three predicted models (Model A, Model B, and Model C), correlational analyses were used to rule out the potential confounds of demographic variables and family history of deviant behavior (with family history “standing in” for possible genetic influences). Correlations examined the relationships between family income and a composite variable for family history of health risk behavior (i.e. family history of alcohol use, drug use, criminal behavior, time spent in prison, extreme aggression, and spousal abuse) on the one hand and parent-adolescent relationships, peer health risk behavior, adolescent health risk behavior, and health status on the other hand. Household income was associated with only one of the parent-adolescent variables (i.e., mutuality, g = .32, p < .01). However, income also correlated with peer health risk behavior (a = .20, p < .05), adolescent substance use/sexual risk taking (a = .18, p < .05), and adolescent health discomfort (g = .25, p < .01). However, later analysis showed that 52 these associations could not explain significant relationships among the variables of interest to this study. None of the correlations for family history of health risk behavior correlated with parent-adolescent relationships, peer health risk behavior/acceptance of deviance, adolescent substance use and sexual risk taking, adolescent aggression/sensation seeking, or health status. To further rule out any significant associations, family history was converted to a simple yes (1) or no (0) scale, indicating whether or not there was a family history for one or more of any of the six variables. With the exception of health risk consequences (5 = -. 16, p < .05), family history scored as a dichotomous variable did not correlate significantly with any of the variables in this study, ruling out the possibility of confounds with this variable. Tests of the Full Mediation (Model A) and Partial Mediation (ModelB) As stated previously, two alternative models (Model A and Model B) proposed in this research describe adolescent health risk behavior as mediating the effects of peer health risk behavior and perceived relationships with parents on adolescent health status. Model A assumes that peer health risk behavior and peer acceptance of deviance mediate the effects of parent-adolescent relationships on adolescent health risk behavior and that adolescent health risk behavior mediates the effects of the peer variables on health outcomes. Alternatively, Model B assumes that in addition to the proposed relationships in Model A, the parent-adolescent relationship variables can be expected to directly predicted adolescent health risk behavior and health status. Only the direct paths from the peer variables to the adolescent health status variables are expected to be insignificant. Because of the complexity of the two models and in particular, the large number of indicators and factors, manifest as opposed to latent variables were used to test 53 predicted relationships. Scales were computed and standardized for each latent construct using the items identified in the measurement models derived from the CFA’s. Because of constraints associated with sample size, predicted relationships within each component of the overall model were tested first: specifically, relationships. between (a) adolescent health risk behavior and adolescent health status, (b) peer acceptance of deviance/peer health risk behavior and adolescent health risk behavior and (c) parent—adolescent relationships and adolescent health risk behavior). These procedures were followed by a final test of the overall model which included significant pathways identified in the simpler analyses. _Ralationalmrs Between Adolescent Health Risk Be_havior and Adolescent Healta _S_tat_as. Both Model A and Model B propose that adolescent health risk behavior is directly related to adolescent health risk consequences, which in turn are linked to health satisfaction. This model provided an adequate fit to the data (X2 = 8.86, d_f = 5, p = .115; Tucker—Lewis Index = .86; RMSEA = .07; GFI = .98, see Figure 4). In contrast, a “Full Model” in which adolescent health risk behavior directly as well as indirectly predicted pay; of the three health outcomes did not provide an adequate fit (X2 = 7.59, at = 1, p = .006; T ucker-Lewis Index = .23; RMSEA = .21; GFI = .98). Relationship Between Peer and Adolescent Health Risk Behavior. Figure 5 shows direct relationships between the peer and adolescent health risk behavior constructs predicted in Model A and Model B. A confirmatory factor analysis indicated 54 Figure 4 Adolescent Health Risk Behavior Implications on Hea_1m* .72 .58 Adolescent Adolescent Aggression and Alcohol and Drug ensation Use and Sexual Seeking Risk Taking .29 .23 Risk Consequence 17 .28 L Health Health J: Satisfaction _30 Discomfort *Note. All beta weights are significant p < .01, except, risk consequence to health satisfaction: p < .05. good model fit (x2 = .77, a = 1, 2 = .382; Tucker-Lewis Index = 1.02; RMSEA = .00; GFI = .99). Three of the four predicted pathways were significant (p < .001). The path from peer acceptance of deviaag: to adolescent substance use/sexual risk taking was the notable exception. A follow-up CFA was conducted on a model that included only the significant pathways, yielding a model with good statistical fit (X2 = .4.33, d_f = 2, p = .12; Tucker-Lewis Index = .95; RMSEA = .08; GFI = .98). 55 Figure 5 Peer Implications for Adolescent Health Risk Behaviors’t .35 .99 .72 Peer Acceptance Peer Health of Devrance Risk Behavior o Adolescent mespenné . c o a Aggressron and Drug Use Sensation . . and Sexual Risk Seeking Taking *Note. With the exception of the insignificant path (dotted line), all beta weights are significant p < .01. Ralartionship Between the Parent-Adolescent Relationship Factogrnd Adolescent filth Risk Behayior. Separate chi square analyses tested fully mediated relationships (i.e., Model A) and partially mediated relationships (Model B) from the parent-adolescent relationship variables to adolescent health risk behavior. Predicted associations between the peer and adolescent health risk behavior variables were modeled on the basis of the previous findings. Results for Model A resulted in a significant chi square (X2 = 26.15, _df = 12, p = .01) with other indices also suggesting a poor fit to the data (Tucker-Lewis Index = .80; RMSEA = .09; GFI = .96). In contrast, a test of Model B (the partially 56 mediated model) resulted in an adequate fit (X2 = 4.47, _cfi = 2, p = .11; Tucker-Lewis Index = .89; RMSEA = .09; GFI = .99; see Figure 6; see Table 12 for regression weights and relevant correlations for Model B), with Model A providing a significantly worse fit than Model B (X2 difference = 21.68, g1_f difference = 6, p < .01). However, as can be seen in Figure 6, not all predicted paths were statistically significant. A follow-up CFA analysis tested a “trimmed” model in which several non-significant paths in Model B from the parent-adolescent relationship variables to the peer and adolescent health risk behavior variables were eliminated from the model. This trimmed model provided a respectable fit (X2 = 27.91, if. = 19, p = .09; Tucker-Lewis Index = .93; RMSEA = .07; GFI = .96). As was seen in Figure 6, child and parent reports of parent behavioral control were linked to adolescent health risk behavior via peer factors, a finding consistent with Model A. Consistent with Model B, conflictual dependency had direct implications for adolescent health risk behavior, specifically with regards to aggression/sensation seeking. 57 Figure 6 Partially Mediated Model for Parent-adolescent Relationship Factors Predicting Peer Factors Predicting Adolescent Health Riak Behav_ior - behavior behavior 3:32:52 33x32] mutuality control - control - child rep. par. rep. / xxfi~llax;/a/ I.“ 4 /\