|UFDC Home||myUFDC Home | Help|
This item has the following downloads:
1 ANXIETY, EXECUTIVE FUNCTIONING, AND QUALITY OF LIFE IN A PEDIATRIC CLINICAL POPULATION By ROBERT SCOTT MERRELL A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011
2 2011 Robert Scott Merrell
3 To Lisa, for her boundless compassion
4 ACKNOWLEDGMENTS I thank m y doctoral committee D r Gary Geffken, embodying compassion and keen mindedness, galvanized me into action and provided edifying feedback throughout the scientific process. Drs. S cott Miller and G reg Neimeyer blended nurturing reass urance with trenchant critique. By my side for the last 6 years, Dr. Ken Rice continues to disarm me with his perspicacity, dedication to the field, and rousing sense of humor ( not to mention his disdain for beginning a sentence with an introductory clause) I appreciate these folks more than they likely know. Several have assisted with the preparation of this doc ument : Bianca Augusto, Dr. Paulo Graziano, Laura Navia, and Adam Reid deserve speci fic mention. I thank Dr. Joe McNamara for his discernment and depth of caring. This study would not have been possible without Dr. Philip Nelson a true friend and immense source of inspiration and wisdom. Drs. Gustavo Benavides and Carlos Trujillo first invit ed me to t hink critically. Drs. Benjamin Bensadon, Matt Buman, W ayne Griffin, Marshall Knudson, Natasha MaynardPemba, Jennifer Sager, Paul Schauble, and Vincent Schroder have enriched m y graduate training substantially. I am indebted to Dr. Michael Murphy, a masterful clinician, for providing a model of compassion, solidity, and nonjudgmental curiosity Samuel Beam and Mylo Xyloto fueled the wri ting And, of course, my doctoral studies would not have happened without my friends and family. My mother father and brother have backed me without condition. Lisa, who handled the references and my dissertation isolation with her usual grace, made this all possible.
5 TABLE OF CONTENTS page ACKNOWLEDGMENTS .................................................................................................. 4 LIST OF TABLES ............................................................................................................ 7 LIST OF FIGURES .......................................................................................................... 8 LIST OF ABBREVIATIONS ............................................................................................. 9 ABSTRACT ................................................................................................................... 10 CHAPTER 1 INTRODUCTION .................................................................................................... 12 Background and Significance ................................................................................. 12 Anxiety .................................................................................................................... 15 Quality of Life .......................................................................................................... 17 Quality of Life and Anxiety ................................................................................ 18 Pediatric Quality of Life and Mental Illness ....................................................... 20 Executive Functioning ............................................................................................. 25 Executive Functioning and Mental Health ........................................................ 26 Executive Functioning and Pediatric Mental Health ......................................... 28 Attentional Control Theory: Anxiety Impairs Executive Functioning ........................ 30 Specific Aims and Hypotheses ............................................................................... 3 2 Specific Aim 1 ................................................................................................... 32 Hypothesis 1 (A and B) .................................................................................... 33 Specific Aim 2 ................................................................................................... 33 Hypothesis 2 (A through D) .............................................................................. 33 Post hoc Analyses ............................................................................................ 34 2 METHOD ................................................................................................................ 36 Participants ............................................................................................................. 36 Procedure ............................................................................................................... 37 Measures ................................................................................................................ 38 Measures Completed by the Parents/Primary Caregivers ................................ 39 Measures Completed by the Child/Adolescent Participant ............................... 43 Analyse s ................................................................................................................. 45 Multiple Imputation for Missing Values .................................................................... 46 Rationale for Mediation Analysis ............................................................................. 48 Hypothesis 1 ..................................................................................................... 52 Hypothesis 2 ..................................................................................................... 53 Post hoc Analyses ............................................................................................ 55
6 3 RESULTS ............................................................................................................... 65 Hypothesis 1 ........................................................................................................... 68 Hypothesis 1A: Parent Forms for BASC Anxiety and PedsQL, BRIEF as Mediator ........................................................................................................ 68 Hypothesis 1B: Child Forms for BASC Anxiety and PedsQL, BRIEF as Mediator ........................................................................................................ 71 Hypothesis 2 ........................................................................................................... 71 Hypothesis 2A.1: Parent Forms for BASC Anxiety and PedsQL, BRI and MI as Mediators .................................................................................................. 72 Hypothesis 2B.1: Child Forms for BASC Anxiety and PedsQL, BRI and MI as Mediators .................................................................................................. 73 Analysis 2A.2: Parent Forms for BASC Anxiety and PedsQL, ER and BR as Mediators ...................................................................................................... 74 Analysis 2B.2: Child Forms for BASC Anxiety and PedsQL, ER and BR as Mediators ...................................................................................................... 75 Hypothesis 2C: Parent Forms for BASC Anxiety and PedsQL, D KEFS Inhibit and Monitor as Mediators ................................................................... 75 Hypothesis 2D: Child Forms for BASC Anxiety and PedsQL, D KEFS Inhibit and Monitor as Mediators .............................................................................. 76 Post hoc Analyses: Possible Effects Related to Medication Status ........................ 77 Results Summary ................................................................................................... 80 4 DISCUSSION ......................................................................................................... 96 Findings Contextualized in Current Literature ......................................................... 96 Mediation with EF as a Unitary Construct ......................................................... 97 Mediation with EF subdomains ......................................................................... 99 Rater and Measurement Divergences ............................................................ 102 Additional Contributions to Pediatric Anxiety, EF, and QOL Research ........... 105 Limitat ions and Future Directions ......................................................................... 107 Summary and Conclusion ..................................................................................... 111 LIST OF REFERENCES ............................................................................................. 112 BIOGRAPHICAL SKETCH .......................................................................................... 122
7 LIST OF TABLES Table page 2 1 Steps in data handling and analyses .................................................................. 57 2 2 Age sex, and ethnicity of participants ................................................................ 58 2 3 P rimary diagnoses .............................................................................................. 59 2 4 Primary med ications ........................................................................................... 60 2 5 Specification of models ....................................................................................... 61 3 1 Means, SDs, ranges, alphas, and percent missing for measures ....................... 81 3 2 Correlations for measures and covariates .......................................................... 82 3 3 Estimates for hypothesis 1A model .................................................................... 84 3 4 Est im ates for hypothesis 1A model (selected covariates) .................................. 85 3 5 Estimates for hypothesis 1B model .................................................................... 86 3 6 Estimates for hypothesis 2A .1 model ................................................................. 87 3 7 Estima tes for hypothesis 2A.1 model (selected covariates) ............................... 88 3 8 Estimat es for hypothesis 2B.1 model ................................................................. 89 3 9 Estimates for analysis 2A.2 model ...................................................................... 90 3 10 Estimat es for hypothesis 2A.2 model (selected covariates) ............................... 91 3 11 Estimates for analysis 2B.2 model ...................................................................... 92 3 12 Estimates for hypothesis 2C model .................................................................... 93 3 13 Estimates for hypothesis 2D model .................................................................... 94 3 14 Summary of indirect effect estimates and effect sizes across models ................ 95
8 LIST OF FIGURES Figure page 2 1 Path diagram depicting a model in which the independent variable X has an effec t on the dependent variable Y ..................................................................... 62 2 2 Path diagram depicting a model in which the variable M mediates the effect that variable X has on variable Y.. ...................................................................... 62 2 3 Path diagram representing the EF (BRIEF GEC) mediatio n and the effects of covariates ........................................................................................................... 62 2 4 Path diagram representing simultaneous mediators (BRI and MI) and the effects of covariates ............................................................................................ 63 2 5 Path diagram representing simult aneous mediators (ER and BR) and the effects of covariates. ........................................................................................... 63 2 6 Path diagram representing simultaneous D KEFS mediators and the effects of covariates ....................................................................................................... 64 2 7 Path diagram representing the effects of anxiety on EF, as moderated by medication status ................................................................................................ 64
9 LIST OF ABBREVIATIONS ACT Attentional Control Theory BASC Behavior Assessment System for Children, Second Edition BR Behavioral Regulation (factor) BRI Behavioral R egulation I ndex BRIEF Behavior Rating Inventory of Executive Function D KEFS Delis Kaplan Executive Function System EF Executive functioning ER Emotional Regulation (factor) GEC Global E xecutive Composite MI Metacognition I ndex PedsQL Pediatric Quality of Life Inventory Version 4. PET Processing Efficiency T heory PRS Parent Rating Scale QOL Quality of life SRP Self Report of Personality
10 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy ANXIETY, EXECUTIVE FUNCTIONING, AND QUALITY OF LIFE IN A PEDIATRIC CLINICAL P OPULATION By Robert Scott Merrell Decem ber 2011 Chair: Kenneth G. Rice Major: Counseling Psychology Anxiety disorders are the most prevalent mental illness es in the United States Despite early age of onset, research in pediatric samples is limited. Available data suggest that anxiety signi ficantly impairs children's quality of life, spanning behavioral, emotional, and social domains. Threats to pediatric wellness may be diminished by developing a fuller understanding of the cognitive processes through which anxiety affects quality of life. Prior research suggests that executive functioning may represent one such construct. Based on the Attentional Control Theory prediction that anxiety disrupts executive functioning (Eysenck, Derakshan, Santos, & Calvo, 2007) th e current study examined the role of executive functioning in relation to anxiety and quality of life i n a mixed clinical sample (ages 4 to 18; M = 10.8, SD = 3.4). Measur ement included parent assessment for the three constructs of interest ( N = 108) with subsamples of child per formance on executive functioning tasks ( n = 81) and child self report s for anxiety and quality of life ( n = 42) Bootstrapped, bias corrected mediation analyses (5,000 resamples) provided evidence for executive functionings role as a mediator in the par ent assessed models Also c onsistent with theory, the i nhibition and shifting
11 subdomains of executive functioning were stronger mediators than other subdomains. Effect size for the shift subcomponent fell in the large range ( 2 = .273). Findings are discussed in relation to developing literature on pediatric anxiety, executive functioning, and quality of life, with emphasi s on rater concordance and measurement considerations.
12 CHAPTER 1 INTRODUCTION Background and Signifi cance Anxiety disorders are the most common mental illness es in the United States (Quilty, Van Ameringen, Mancini, Oakman, & Farvolden, 2003). Their impact on wellness is perhaps evidenced by the enormous amount of expenditure aimed at diminish ing their effect. In fact, a recent analysis reported that the annual true societal costs of anxiety disorders in the U.S. [are]more than $100 billion (Kessler & Greenberg, 2002, p. 990). With a median onset prior to the age of 12, anxiety disorders tend to dev elop early in life (Saddock, Saddock, & Ruiz, 2009). Even though other mental disorders (e.g., D epression, Oppositional Defiant Disorder, Conduct Disorder, S ubstance U se D isorders, and AttentionDeficit / Hyperactivity Disorder) also trace back to childhood, their onsets are often preceded by a diagnosable anxiety disorder (Kessler & Greenberg). Despite the 26.1% annual prevalence rate for anxiety in children and young adults ( Kim Cohen et al., 2003), research on the impact of psychiatric disorders in pedi atric populations is sparse at best. Available data suggest that mental disorders may impact younger individuals more adversely than older individuals (Castaneda, Henriksson, Marttunen, Suvisaari, & Lonnqvist, 2008). Thus, anxiety often significantly dec reases children's qual ity of life (Bastiaansen, Koot, Ferdinand, & Verhulst, 2004; Clark & Kirisci, 1996; Sawyer et al., 2002). Quality of life (QOL) is a multidimensional construct, including physical, behavioral, emotional, and social aspects of functioning (Ravens Sieberer, Erhart, Wille, Wetz el, Nickel, & Bullinger, 2006). The Diagnostic and Statistical Manual of Mental Disorders,
13 Fourth Edition, Text Revision (DSM IV TR; America n Psychiatric Association, 2000) enumerates a range of potential QOL dist ress areas related to mental illness. For instance, relational, educational, and occupational disturbances or problems in living (see Szasz, 1960) constitute Axis IV diagnostic considerations or P sychosocial and Environmental Problems. Such concerns also figure prominently in the V Codes subset of Other C onditions That M ay B e a F ocus of C linical Attention. Fittingly, the DSM IV TR and others ( e.g., Frisch, 2006) emphasized that QOL impairments often predate mental disorders or are products of the disorder. Mogotsi and colleagues (2000) highlighted this bi directional nature of effects, positing that QOL concerns are both causes and sequelae of anxiety disorders. Prospective studies are clearly needed to better elucidate the relationship between anxiety and QOL, yet it is equally apparent that anxiety disorders by their very nature significantly undermine QOL (Olatunji et al., 2007). R esearch on QOL and anxiety disorders has reliably portrayed a uniform picture of anxiety disorders as illnesses that markedly compromise QOL and psychosocial functioning (Mendl owicz & Stein, 2000, p. 669). Given that anxiety adversely affects behavioral and psychosocial functioning, it stands to reason that threats to pediatric QOL may be diminished by developin g a fuller understanding of the cognitive processes through which anxiety affects QOL Executive functioning (EF) likely represents one such construct. Executive functioning r efers to the cognitive processes involved in self regulation (Sarsour et al., 2011). Behavioral disturbances, interpersonal difficulties, and limited academic/vocational achievement have been associated with impaired EF (Baron, 2004; Lezak Howieson, & Loring, 2004). A n emerging literature demonstrates the links
14 between compromised EF and psychiatric disorders including Schizophrenia, Bipolar Disorder, AttentionDeficit/Hyperactivity Disorder Depression, and substance use disorders ( Banich, 2009). Nonetheless, r elatively little is understood about cognitive defici ts related to anxiety disorders, though pr eliminary findings suggest that adults with anxiety disorders exhibit decreased EF performance (Castaneda et al., 2008). These findings are consistent with Processing Efficiency Theory (PET; Eysenck & Calvo, 1992) and Attentional Control Theory (ACT; E ysenck, Derakshan, Santos, & Calvo, 2007). Whereas PET contends that anxiety disrupts global EF, ACT capitalizes on advances in EF research and maintains that anxiety specifically impairs two EF subcomponents: the ability to inhibi t a prepotent response and the ability to shift back and forth to meet situational demands. Thus, anxiety theoretically undermines EF, and a growing corpus of studies empirically supports this relationship (for a review, see Derakshan & Eysenck, 2009). Research on the role of EF in pediatric anxiety with regard to QOL is limited. The present study sought to expand the empirical knowledge base by addressing this critical gap in current literature. Using PET (Eysenck & Calvo, 1992) and ACT (Eysenck et al ., 2007) as a conceptual framework, t he study t est ed the hypothesis that EF mediates the relationship between anxiety and QOL. Specifically, it was h ypothesized that increased anxiety symptoms w ould be associated with increased executive dysfunction, whic h in turn would be associated with poorer QOL Desig n implications notwithstanding, results provided supporting evidence that EF may serve as part of the mechanism by which anxiety affects QOL.
15 T o set the context for this study, the ensuing literature rev iew will demonstrate the following key points: (a) anxiety disorders stand out relative to other psychopathology due to the significant personal and societal burdens they engender; (b) children and adolescents are particularly at risk to be adversely impac ted by anxiety; (c) the effects of anxiety extend well beyond the symptoms of the disorder itself and threaten QOL; and ( d) in consonance with PETand ACTbased predictions regarding how anxiety impacts cognitive functioning, EF may play a critical role i n how anxiety influences QOL. Anxiety Anxiety disorders are the most frequently occurring psychiatric disorders in the United States (Mendlowicz & Stein, 2000; Quilty et al., 2003). They have high chronicity and early age of onset (Kessler & Greenberg, 2002). With lifetime prevalence for the constellation of disorders estimated as high as 28.8%, they are associated with significant personal (i.e., role impairment) and societal strain (Kessler, Berglund, Demler, Jin, & Walters, 2005). A staggering economi c reality also underscores anxietys pernicious effects. In their exhaustive review of psychiatric epidemiologic surveys, Kessler and Greenberg reported that a 1996 estimate listed the annual societal cost of anxiety disorders at $47 billion, although a 1999 estimate reduced the figure to $42 billion. As indicated earlier, they argued that the true social costs exceed these numbers (surpassing $100 billion annually) because longterm opportunity costs (i.e., extended unemployment or underemployment) and c osts linked with comorbidity were excluded from prior analyses. Add to the financial burden factors such as severe distress, excessive worry, restlessnessin a word, misery and a more complete pictu re of anxiety begins to emerge.
16 Pediatric anxiety. Worry figures prominently among the chief reasons for referral to childrens healthcare providers (March et al., 1999). It is perhaps not surprising that, similar to community studies of adults, anxiety disorders are also the most prevalent mental disorders in children and adolescents (Saddock et al., 2009). Consider those who suffer from ObsessiveCompulsive Disorder (OCD), an often debilitating anxiety disorder and one of the most common childhood psychiatric illnesses (Stewart et al., 2004). Epidemiologi c studies estimate prevalence rates of approximately 14% among children and adolescents (Douglass, Moffit, Dar, McGee, & Silva, 1995; Zohar, 1999). As delineated by DSM IV TR diagnostic criteria, the disorder is characterized by recurrent, timeconsuming obsessions or compulsions that cause marked distress or significantly interfere with normal functioning (American Psychiatric Association, 2000). In children, such disturbances often manifest in familial, social, and academic domains (Flament et al., 1988; Piacentini, Bergman, Keller, & McCracken, 2003). Thus, this multidimensional turbulence extends into various domains of functioning, and it also reaches out temporally. Flament, Koby, Rapoport, and Berg (1990) tracked clinically referred youth with OC D and found that 68% of those seen again still had the disorder 7 years later. Thomsen and Mikkelsen (1995) found remission and reemergence of OCD in their pediatric participants, and approximately one half of their sample met diagnostic criteria at 5 year followup. Taking OCD as illustrative of pathological anxietys significant and potentially durableif not treated effects on QOL for both patients and their families/peers (Barlow, 2000), it stands to reason that recent decades have seen a dramatic incr ease in researchers attention to
17 the link between anxiety and QOL (Hansson, 2002; Olatunji, Cisler, & Tolin, 2007; Quilty et al., 2003). Quality of Life Mendlowicz and Stein (2000) questioned the current validity of the idea that the cost of human sufferi ng cannot be measured, contending that various aspects of human sufferingas well as its absencecan be reliably assessed via the concept of QOL. Increased research focus on QOL originated from the shift in criteria for evaluating medical outcomes that has occurred in recent decades, stemming in part from the World Health Organizations (1948) invitation to look beyond symptom reduction/increased survival and toward a more patient centered consideration of wellness, including physical, emotional, and social wellbeing. More phenomenological in nature, QOL has biopsychosocial underpinnings that emphasize healthy living and health outcomes as products of the interplay between physical and psychological factors (Engel, 1977). Differences in QOL have been det ected between adult racial/ethnic groups (e.g., Utsey, Chae, Brown, & Kelly, 2002). As applied to children, QOL refers to a multidimensional construct involving physical, behavioral, psychological, and social aspects of functioning as perceived by either children themselves or parents/ other observers (e.g., primary caregivers, teachers) (Bullinger, 2002; Ravens Sieberer et al., 2006). Q uality of life was first applied in a medical context to assess how cancer treatments affected not only patients surviv al time but also their subjective sense of wellbeing. Dissatisfied by the available measures narrow focus on morbidity and mortality, Spitzer, Dobson, and Hall (1981) sought to develop an instrument more attuned to the social and emotional aspects of the patients life. Interestingly, they
18 found very few low scores (i.e., indicative of poor QOL) during Australian field testing trials. After reasoning that a cultural bias toward under reporting and methodological/psychometric shortcomings may plausibly explain their results, they alternatively questioned if they may be underestimating the ability of the human mind and spirit to compensate for major infirmity (p. 596). Accordingly, due to the inextricable links between ones experience of physical pain/disease, overall wellness, and psychological capacity to compensate for major infirmity assessing QOL within clinical psychiatric popu lations is of vital importance. Examination of QOL in the mental health setting may reveal compensatory cognitive processes that could be targeted to help people overcome psychiatric challenges, such as anxiety. Quality of Life and Anxiety The DSM IV TR ( America n Psychiatric Association, 2000) notes that impaired QOL is frequently a common cause or consequence of mental il lness, and thus should figure prominently in treatment planning. E ven though anxiety disorders are the most prevalent mental disorders, it would be erroneous to assume that they represent mild psychopathology (Barlow, 2000). On the contrary, research on QOL and anxiety disorders has reliably demonstrated anxietys association with compromised physical, behavioral, psychological, and social functioning (Hansson, 2002; Mendlowicz & Stein, 2000; Mogotsi, Kaminer, & Stein, 2000; Olatunji et al ., 2007; Quilty et al., 2003). Reflective of anxiety disorders high occurrence and pervasive influence on well being, research on the impact of anxiety on QOL has increased considerably in recent years yet is still in its nascence (Hansson, 2002). Principally, researchers have employed epidemiologic and clinical studies. Epidemiological surveys have been utilized to infer QOL from various indicators such as subjective assessment of physical
19 and emotional health, psychosocial functioning, and financial solvency (Markowit z et al., 1989). Clinical studies have relied on QOLspecific measures, often assessing aspects of physical, social, emotional, and vocational well being. Each methodological approach has its merits and drawbacks. What community based samples gain in descriptive value for the larger population, they sacrifice in terms of relevance to clinical practice. Turning now to the relationship between QOL and particular anxiety disorders, some evidence suggests that patients with panic disorder (PD) and post traumatic stress disorder (PTSD) report poorer QOL than individuals with other anxiety conditions (Hansson, 2002). For instance, community and clinical samples of individuals with PD report a high frequency of suicide attempts, severe vocational impairment, si gnificant psychological distress/constraints, and impaired emotional wellness (Quilty et al., 2003; Sherbourne, Wells, & Judd, 1996 ). P ost Traumatic Stress D isorder research has generally focused on community samples consisting of veterans (Jordan et al., 1992; Stein, Walker, Hazen, & Forde, 1997). Familial discord (ranging from nonspecific marital distress to increased violence) was more prevalent in families of veterans with PTSD than in families of veterans not suffering from the disorder. Additionall y, children of individuals with PTSD were more likely than their counterparts to demonstrate behavioral problems, findings that highlight how the impact of psychopathology extends beyond symptoms, transcending the intrapersonal and into the interpersonal ( Mendlowicz & Stein, 2000). Nevertheless, despite preliminary findings that PD and PTSD were associated with compromised QOL, a metaanalysis of 23 separate studies ( N = 2892) indicated that no
20 particular anxiety disorder was associated with significantly poorer QOL than any other particular anxiety disorder (Olatunji et al., 2007). Broadly stated, individuals with social phobia exhibited substantial deficits in educational, occupational, social, and romantic functioning (Stein & Kean, 2000). Despite scores indicative of robust physical health, individuals with obsessivecompulsive disorder (OCD) also demonstrated severe impairments in social functioning and role limitation due to emotional problems and mental health (Koran, Thienemann, & Davenport, 1996). Commonly proffering nosological concerns and comorbid disorders as an explanation, researchers noted that the paucity of data for QOL in individuals with noncomorbid generalized anxiety disorder (GAD) indicated substantial overall life impairment (Mas sio n, Warshaw, & Keller, 1993). In sum, Olatunji and colleagues (2007) metaanalysis demonstrated that anxiety disorders are associated wit h significant QOL impairment. Mogotsi and colleagues (2000) noted this earlier: Increasingly, the impact of anxiety disorders on QOL is being recognized and empirically documented, and current data indicate that both objective and subjective dimensions of QOL are significantly reduced in all of the anxiety disorders (p. 278). Taken collectively, these data bolster Mendlowicz and Steins (2 000) call to action: Anxiety disordersmarkedly compromise quality of life and psychosocial functioning in several domains[and] it is hoped that these findings will translate into a more accurate public (and health care polic y) view of anxiety disorders as serious mental disorders worthy of future research and appropriate health care expenditures. (p. 680) Pediatric Q uality of Life and Mental Illness Childhood and adolescence are regarded as periods of optimum health (Millstein, 1989), yet youth are clearly not immune from the ravages of illness. The sobering
21 reality is that children and adolescents are among the most medically underserved in the United States (McManus, Shejavali, & Fox, 2003). Because adolescents represent 15% of our nations population and 100% of our nations future (McManus et al., p. 1), recent literature suggests that researchers allocate energy toward better elucidating the psychological factors involved in the pediatric health promotion (Lerner, 2000; March et al., 1999; Tucker, 2002). Compared to otherwise psychologically healthy counterparts, children with psychiatric illness have a greater likelihood of developing another disorder post remission and they are more likely to experience social, educational, and occupational impairment as they age (Costello et al., 2003). Consistent with the dearth of research in this age group, measurement of pediatric QOL represents a longdisregarded topic of study (Bastiaansen et al., 2004; Kazdin, 2001; Sawyer et al., 2002). I mportant methodological considerations for this line of research regard how to best obtain meaningful child data. Researchers have offered mixed views on the advantages and disadvantage of c hild self report parent proxy report and informedother report (e.g., health care provider and teacher). M ethodological issues such as ageappropriateness of language, syntax, and response options have led some to view structured interviews and interviewer administered instruments as optimal ( Matza, Swensen, Flood, Secnik, & Leidy, 2004) Whereas inherent costs may be prohibitive (Ravens Sieberer et al., 2006), issues regarding potential l imits of childrens comprehension of more abstract psychological constructs appear to represent a more foundational constr aint. For instance, given that low concordance rates between adults and significant others have been detected in adults with psychiatric disorders (Saintfort, Becker, & Diamond, 1996) it is not surprising that
22 such disparities emerge in pedi atric researc h examining complex psychological constructs such as anxiety. In one study, findings revealed that all 8year old participants understood the word nervous compared to only 57% of 5year ol ds (Rebok et al., 2001) D evelopmental comprehension concerns notwithstanding, some r esearchers concluded that parent proxy report may offer greater reliability (Matza et al.) yet this benefit may come with a loss of validity in age groups better suited for self assessment G eneral consensus appears to point toward in corporating disparate points of view (Ravens Sieberer et al.), with design decisions ultimately deriving from the particular aims of a given study and the intended use of the data. In part due to the challenging methodological considerations, the scarc ity of pediatric QOL research is unsettling in light of Costello and colleagues (2003) longitudinal community study ( N = 1420) that investigated the prevalence and development of psychiatric disorders in children from ages 9 through 16 years. Findings su ggested that at least 1 in 3 children will suffer from 1 or more psychiatric disorders by the age of 16. Furthermore, children with significant emotional/behavior disorders have a higher likelihood of developing another disorder after remission than do their unaffected peers. The authors of this epidemiologic study reported that affected children are also more likely to experience functional impairment (i.e., interpersonal, educational, vocational) as they grow older. This enduring cascade of effects bey ond the symptoms seems especially relevant for the QOL implications for pediatric anxiety disorders, which frequently last for decades or an entire lifetime if left untreated (Barlow, 2000). Yet there is nearly a complete lack of research on QOL and pediatric anxiety. Matza and researchers (2004)
23 emphasized that current pediatric QOL literature examines almost exclusively medical diseases. This fact is surprising in light of observations that children with psychiatric disorders report more compromised QO L in many areas in comparison to children with physical disorders (Landgr af, Abetz, & Ware, 1996; Sawyer et al ., 2002). Only a few studies have focused on QOL and pediatric psychopath ology (Bastiaansen et al., 2004, 2005; Clark & Kirisci, 1996; Sawyer et al. ). The limited data suggest that children with psychiatric diso rders have a substantially poorer overall QOL relative to other children. Bastiaansen and colleagues (2005) investigated the context in which the effects of pediatric psychopathology on QOL differ. They found that girls experienced a greater detrimental impact of mental illness on QOL than boys. They argued that one possible explanation might be that boys tend to exhibit externalizing behavior, and thus may not experience their symptoms as adversely as girls, who may be less prone to externalizing. The authors also found that the negative impact of mental disorders on childrens QOL increased with age. Thus, beyond their demographic value, sex and age appear to be of interest with regard to st udies of pediatric QOL. Clark and Kirisci (1996) found that the effects on QOL differed by diagnosis for adolescents in their mixed clinical and community sample. Specifically, self report data on youth between the ages of 12 and 18 with Major Depres sive Disorder (MDD) indicated significant adverse effects on social and academic functioning. Whereas youth with PTSD exhibited a similar pattern, those who endorsed substance use had impoverished role functioning. Sawyer and researchers (2002) found that youth aged 6 to 17 years with MDD had more impaired QOL relative to children with ADHD and
24 Conduct Disorder. Compared to youth with purely physical disorders, youth with psychiatric disorders had substantially poorer QOL in multiple domains. In what appears to be the first study on QOL with respect to pediatric anxiety, Bastiaansen and colleagues (2004) examined QOL in a clinical sample of 310 children between the ages of 6 and18. They employed the Pediatric Quality of Life Inventory Version 4.0 (PedsQ L; Varni, Seid, & Kurtin, 2001) to measure QOL within four subdomains: physical, emotional, social, and school functioning. Assessing QOL differences across six diagnostic categories, they concluded that specific disorders were associated with different i mpacts across QOL subdomains. The authors cautioned that clinicians may consider anxiety disorders less severe than other psychiatric disorders (p. 228), which may reflect a clinical bias to underestimate the effects of pediatric anxiety. Nonetheless, their results indicated that youth with anxiety disorders had poorer emotional functioning in comparison to those with other disorders on both parent and clinician reports. These findings align with those of Mogotsi and colleagues (2000 ) who also detected compromised QOL in anxious adults. Thus, there are many compelling reasons to examine QOL and pediatric anxiety. Of course there is value in the treatment of symptoms, and outcomes may improve if interventions also target other factors that may affect Q OL (Bastiaansen et al., 2004). Given that QOL encompasses physical, emotional, and social well being (Bullinger, 2002; Ravens Sieberer et al., 2006), it is conceivable that threats to pediatric QOL may be affected by developing a richer understanding of the cognitive processes related to mental disorders. Executive functioning may represent one such construct. This construct refers to cognitive processes that govern self regulation (Sarsour et al., 2011).
25 Psychiatric disorders have been associated with c ompromised EF (Airaksinen et al., 2005; Boonstra, Oosterlaan, Sargeant, & Buitelaar, 2005; Emers on, Mollet, & Harrison, 2005; Francis, 1988; Julian & Arnett, 2009; Kendall & Chansky, 1991; Micco et al., 2009; Toren, 2002; Willcutt, Doyle, Nigg, Faraone, & Pennington, 2005). Furthermore, deficits in EF have been reliably associated with behavioral disturbance, restricted academic/occupational achievement, and social dysfunction (Baron, 2004; Lezak et al., 2004). Noting that these domains are integral to QO L, we now turn to relevant EF literature. Executive Functioning Although a neurobiological examination of EF exceeds the scope of this study, a brief explanation of terminology is appropriate. As indicated in Elliots (2003) review, EFs are commonly thought of as higher order cognitive processes carried out in neural networks including the prefrontal cortex, thalamus, and basal ganglia. Miyake and researchers (2000) explained that the terms frontal lobe tasks and executive tasks/functions are often employed synonymously in spite of their practical dissimilarity: studies on patients with frontal lobe lesions or other physical insults have revealed variable degrees of functional impairment or no impairment whatsoever (Reitan & Wolfson, 1994; Shallice & Burgess, 1991). Such contradictory findings suggest the imprecision of interchangeably using the anatomical term frontal lobe and the functional term executive. Usage of functional terminology (including the plural executive functions, or EFs) bett er fits the purpose of the present study. Even though a consensually agreed upon definition of the construct is lacking (Hughes & Ensor, 2008), EF generally encompasses planning, information updating and monitoring, mental set shifting, inhibition of prepotent responses, and
26 commencing/maintaining mental and physical activity ( Lehto et al., 2003; Lezak et al., 2004; Mikaye et al., 2000; Smitherman et al., 2007). E F refers to cognitive skills used to effortfully guide behavior toward a goal (Banich, 2009, p. 89) or an individuals efforts to modify her or his inner state and responses to contextual demands (Graziano, McNamara, Geffken, & Reid, 2011). E xecutive functioning is often personified as a business executive who lacks specialization in a specific domain and instead oversees and manages multiple subdomains (Salthouse, Atkinson, & Berish, 2003). Optimal living depends on ones capacity to assess options and choose from them. These choices often but not always pit proximal versus distal fulfillmen t: our actions are often directed toward achieving a positive outcome in a simulated future context and must therefore compete with alternative actions that might maximize initial benefits but have larger long term costs (Willcutt et al., 2005, p. 1336). In sum EF refers to the neurocognitive processes that enable us to sustain an appropriate problem solving set in order to move toward future goals. Despite the apparent lack of an integrated account of EF to guide research (Banich, 2009), factor analyti c studies have supported a three factor model: mental set shifting, information updating and monitoring, and inhibition (Lehto et al., 2003; Miyake et al., 2000). Executive Functioning and Mental Health It follows intuitively that researchers have emphasiz ed EFs centrality to attaining success in school, work, and life in general (Diamond, Barnett, Thomas, & Munro, 2007). An emerging literature demonstrates EFs associations with various health processes and outcomes. For instance, adult studies have link ed obesity to EF impairments (Ellis et al., 2004; Seeyave e t al., 2009; Wang et al., 2001). Studies in
27 pediatric samples have detected similar results (Braet, Claus, Verbeken, & V an Vlierberghe, 2007; Cserjesi, Lumi net, Molnar, & Lenard, 2007). Regarding EFs nexus with psychopathology, there is a growing corpus of studies demonstrat ing the links between executive dysfunction and Schizophrenia, Bipolar Disorder, AttentionDeficit/Hyperactivity Disorder Depression, and substance use disorders ( Banich, 2009). The majority of studies have examined ADHD ( for a review see Boonstra et al., 2005). Prevailing theoretical and neurobiological explanations maintain that EF deficits, chiefly as evidenced in behavioral dis inhibition, may be of etiological relevance for the disorder (Barkley, 1997; Durston, 2003). In their metaanalytic review, Wilcutt and colleagues (2005) concluded that although EF impairment alone is neither necessary nor sufficient to cause all cases of ADHD (p. 1343), significant executive dy sfunction in key domains (e.g., response inhibition and planning) is associated with the disorder. Not only does t his line of research illustrate EFs role within psychiatric disorders it also raises questions of practical, theoretical, and clinical impo rtance related to how EF affects and/or is affected by psychopathology. Due to the typically early onset of anxiety ( Kessler & Greenberg, 2002) and Eysenck and colleagues (2007) theoretical prediction that anxiety causes EF impairment, research on EF and anxiety appears warranted because of the plausible etiological relevance of executive dysfunction cited earlier in this paragraph. In comparison to studies examining EFs relation to ADHD, research on EF and clinical anxiety is scarce (Castaneda et al., 2008). Consequently, r esearchers have highlighted the need for inquiry in this area. Julian and Arnett (2009) studied the independent contributions of anxiety and depression to EF in a sample of adults with
28 M ultiple S clerosis (MS). They found that each disorder predicted EF dysfunction. They concluded that the more traditional assessment of only depressive symptoms in MS patients is insufficient and argued for anxiety assessment as well because t he treatment of anxiety may benefit patients with MS not o nly by alleviating psychiatric distress, but also by increasing the avai lability of cognitive resources (p. 802). Airaksinen and colleagues (2005) examined the EF/anxiety connection in their populationbased study of adult neuropsychological functioning in Sweden. The authors concluded that adults with anxiety disorders exhibited significant EF impairment relative to healthy controls. Specifically, EF deficits were noted in participants with OCD and Panic Disorder with and without agoraphobia. The auth ors asserted that compromised EFs associated with anxiety disorders may have a deleterious influence on QOL, and, in particular, social and occupational functioning. Before returning to this critical point of contact between anxiety, EF, and QOL, a review of studies on pediatric anxiety and EF is in order. Executive Functioning and P ediatric Mental H ealth Adult evidence to the contrary notwithstanding (S mitherman et al., 2007) a small yet compelling body of literature attests to the link between pediatric anxi ety and executive d ysfunction (Emerson et al., 2005; Francis 1988; Kendall & Chansky, 1991; Micco et al., 2009; Toren et al., 2000). Francis (1988) reported that children with higher anxiety levels also had significantly more task inhibiting thought s. Similarly, Kendall and Chansky (1991) found that children with anxiety disorders complained of frequent intrusive thoughts during cognitive tasks. Participants between the ages of 9 and 14 exhibited difficulties in shifting attention from internal to external stimuli. Toren and researchers (2000) studied neurocognitive correlates of anxiety disorders in a sample of
29 Hebrew youth ranging from 6 to 18 years in age. Compared to agematched nonanxious controls, the clinical participants exhibited decreas ed cognitive flexibility. Emerson and colleagues (2005) assessed anxious depressed boys between 9 and 11 years of age. They noted EF deficits in their clinical sample relative to their nonanxious, nondepressed peers. Specifically, the anxious depressed participants demonstrated significantly poorer ability on set shifting, hypothesis testing, and problem solving. Available research clearly demonstrates the association between anxiety and executive dysfunction. Methodological constraints, however, li mit conclusions regarding causality. Although the results are tempered by the studys design, Micco and colleagues (2009) research provided tentative data regarding directionality of effects. The authors examined 147 children of parents with Major Depressive Disorder (MDD), Panic Disorder (PD), the two comorbid disorders, and parents who did not meet diagnostic criteria for any mood or anxiety disorder. Their aim was to assess whether offspring (between the ages of 6 and 17) of parents with these mental disorders would show compromised EF relative to the children of the healthy control parents. Citing prior research supporting greater rates of psychopathology in the offspring of affected parents as compared to the offspring of healthy parents, the author s expected that if deficits in EF were markers for the development of MDD or PD, then they would detect more impaired EF in the children at risk for depression and anxiety. However, they found no association between offspring status and executive dysfunct ion. They concluded that compromised EF may not serve as a trait marker for developing anxiety or depression. Of additional value, their results indicated that the children with current
30 depressive and anxiety symptoms also exhibited EF impairment. Thus, executive dysfunction appeared to be symptomatic of the current disorder rather than a cause of the disorder. Caution must be exercised when interpreting these findings given the crosssectional and correlational nature of the data. Nevertheless, Micco and colleagues findings tentatively support the notion that anxiety leads to executive dysfunction rather than vice versa. These data coincide with Eysenck and colleagues (2007) theoretical stance that anxiety disrupts EF, as explicated by ACT. Attentio nal Control Theory: Anxiety Impairs Executive Functioning Within the cognitive psychology literature, a substantial body of studies examines the relationship between anxiety and cognitive performance (for a review, see Eysenck, 1992). Anxiety is defined as an aversive emotional and motivational state[and] individuals frequently worry about the threat to a current goal and try to develop effective strategies to reduce anxiety and achieve the goal (Eysenck et al., 2007, p. 336). Findings generally indic ate that anxiety detrimentally influences cognitive performance, and its disruptive effects tend to be more damaging with increasingly complex/cognitively demanding tasks. According to Deraksan and Eysencks (2009) review, empirical support for this gener alization has been detected when anxiety is regarded as a temporary mood state (i.e., state anxiety) and a relatively stable aspect of personality (i.e., trait anxiety). Theory as to how anxiety affects cognitive functioning is continually evolving. Eys enck and Calvo (1992) elaborated Processing Efficiency Theory (PET) to address gaps in extant theory (e.g., Sarason, 1988) on how anxiety diminishes cognitive performance. In brief, underlying the theory was the assumption that anxiety drained cognitive r esources by impairing EF and overall processing efficiency. Although the
31 theory generated considerable research supporting its assumptions (for a review, see Eysenck et al., 2007), precision regarding the specific nature of EF impairment was lacking (Dera kshan & Eysenck, 2009). Spurred by this imprecision and advances in EF research (Friedman & Miyake, 2004; Miyake et al., 2000), Eysenck and colleagues (2007) developed Attentional Control Theory (ACT). In its short existence ACT has already gained substantial empirical backing (Derakshan & Eysenck, 2009), perhaps due to the fact that it further articulates the already well supported PET (Eysenck & Calvo, 1994) An in depth account of ACT falls outside the scope of the present study. Of relevance is ACTs prediction that anxiety impairs two subcomponents of EF: the ability to inhibit a prepotent response and the ability to shift back and forth to meet situational demands. Whereas PET posited that anxiety compromises EF and impairs processing efficiency b ecause anxiety produces worry (Eysenck et al., p. 339), ACT contends that anxiety specifically leads to executive dysfunction in set shifting and inhibition. As the authors emphasized, a consensually agreed upon definition of EF is not existent, yet fact or analyses have supported three principal EF domains: mental set shifting, information updating and monitoring, and inhibition of prepotent responses (Lehto et al., 2003; Miyake et al., 2000). Unlike most PET and ACTinformed research, the present study d id n ot f ocus on narrowly defined areas of performance (e.g., continuous motor saccade, reading, visuospatial, and reaction time task s). It instead relie d o n their theoretical predictions that anxiety negatively influences EF in a less context specific manner. To date, PET and ACT have been utilized almost exclusively to predict the effects of anxiety on
32 EF/cognitive performance in nonclinical populations (Derakshan & Eysenck, 2009). The present study expands t he literature by beginning to investigat e anxietys impact on EF in a psychiatric population, which addresses the theorys authors call for increased research in nonstressful conditions (Eysenck et al., 2007, p. 349). In conclusion, prior research has demonstrated the association between an xiety and impaired EF. Eysenck and colleagues (2007) most recent theory of attentional control predicts that their relationship is more defined: anxiety disrupts EF, particularly in its shifting and inhibiting subcomponents. Executive dysfunction, in turn, has been linked with behavioral disturbance, constrained educational/vocational achievement, and interpersonal strife. These domains figure prominently in QOL. Consequently, it stands to reason that disorders linked with compromised EFwhich is crit ical for self directed behavior might be at increased risk for QOL erosion (Banich, 2009, p. 90) Similar relationships have been detected in studies of ADHD (Klassen, Miller, & Fine, 2004), traumatic brain injuries (Horneman, Folkesson, Sintonen, Von Wendt, & Emanuelson, 2005), and epilepsy (Sherman, Slick, & Eyrl, 2006). Research on the role of EF in pediatric anxiety with regard to QOL is sparse. Following Airaksinen and colleagues (2005) postulation that EF deficits would likely predict impaired QO L in anxious individuals, the present study address e d this vital gap in the literature using cognitive psychological theory as a conceptual framework from which to base its predictions. Specific Aims and Hypotheses Specific Aim 1 The first aim of the pres ent study was to investigate the nature of the association between anxiety, EF as a unitary construct, and QOL in children and adolescents.
33 Hypothesis 1 (A and B) In light of the reviewed literature (e.g., Airaksinen et al., 2005; Bastiaansen et al., 2004, 2005; Mendlowicz & Stein, 2000; Micco et al., 2009) and theory that anxiety disrupts EF (PET; Eysenck & Calvo, 1992), it was hypothesized that EF w ould mediate the relationship between anxiety and QOL. Specifically, increased scores on a measure of anxiety were hypothesized to be associated with poorer scores on a measure of EF, which in turn were hypothesized to be associated with reduced scores on a measure of QOL Hypothesis 1A examined parent reports on anxiety and QOL mediated by parent assessed EF. Hypothesis 1B evaluated childrens self reports of anxiety and QOL mediated by parent assessed EF. Based on prior research demonstrating the associations between age and EF (Diamond, 2007; Zelazo et al., 2003), sex and QOL (Bastiaansen, Koot, & Ferdinand, 2005), and raci al/ethnic status and QOL (Utsey et al., 2002), the model included age, sex and ethnicity as covariates. A primary diagnosis of ADHD was also covaried due to its substantial representation in the total sample (56%) and documented links w ith executive dysfunction (Barkley, 1997). Specific Aim 2 Whereas the first aim investigated a global EF construct as a mediator for the relation between pediatric anxiety and QOL, the second aim examined t he mediating f unctions of specific EF subcomponents. Hypothesis 2 ( A through D) H ypothesis 1 focused on EF as a uni fied construct. Capitalizing on theoretical advances in how anxiety impairs EF, Hypothesis 2 dismantled EF into various subcomponents. In keeping with ACTs prediction that anxiety speci fically compromises inhibit and shift domains of EF (Eysenck et al., 2007), it was h ypothesized that the
34 corresponding inhibit and shift scales of the utilized EF measure w ould demonstrate stronger mediation effects than the other nontheory r elevant EF subcomponent scales (e.g., updating) Six mediation models using different raters (i.e., parent or child) and alternate EF measurement (i.e., parent assessment or child performance) were tested. I nhibition/shifting were tested as mediators of parent assessed anxiety and QOL (Hypothesis 2A.1). In hibition/shifting were also evaluated as mediators of child self reported anxiety and QOL (Hypothesis 2B.1) Statistically significant results supporting the differing mediation roles of EF subdomains led to explor atory post hoc analyses ( A nalyses 2A.2 and 2B.2) of slightly different EF subdomains derived from prior empirical findings for the related EF measure (Gioa, Isquith, Retzlaff, & Espy, 2002). Finally, i nhibition and monitoring as measured on a child perfor mance task were evaluated as mediators of parent assessed anxiety and QOL (Hypothesis 2C) and child self reported anxiety and QOL (Hypothesis 2D) T hese analyses also covaried age, sex ethnicity and ADHD status. Post hoc Analyses Likely an artifact of the disparate emphases of the cited studies, the absence of analyses on pharmacological interventions is apparent in the reviewed l iterature. Medication effects are worthy of examination in the current sample due to the proportion of participants report ing u se of prescription medication (60% of total sample). Whereas a comprehensive assessment of this critical topic merits dedicated studies, followup analyses were conducted to assess potential effects related to participants medication use. S cores for the total sample parent assessed anxiety and QOL were analyzed w ith the BRIEF total score as the mediator variable. Medication status (i.e., on medication versus medicationnave) was covaried, and c ovariates from the prior best fitting models
35 were incl uded R esults were aggregated across 20 multiply imputed data sets. Lastly, m edication status was assessed as a potential moderator of anxiety's effects on EF, and then again as a moderator of EF's effects on QOL (Figure 27).
36 CHAPTER 2 METHOD Parti cipants Data for participants in the present study was drawn from an archive of youth presenting for assessment at an outpatient psychiatric clinic during the period of January 2009 to May 2010. Participants were typically r eferred for assessment by pediatricians or child psychiatrists. The clinic was housed within the Department of Psychiatry in the College of Medicine at a large southeastern university. In addition to assessment services, the clinic provided ongoing cognitivebehavioral treatments for a variety of psychological disorders A licensed psychologist with more than twenty years of clinical experience supervised the predoctoral level psychology interns who performed the assessments. Table 21 summarizes the data handling and analyses steps t o facilitate comprehension of the sequence used to construct the subsamples and perform the analyses. Table 22 presents demographic information, such as age, sex, and racial/ethnic distribution. The total sample included data f or 108 individuals between the ages of 4 and 18, with a mean age of 10.8 years.1 1 Sample A contained two individuals (ages 4.3 and 4.6 years) that fell below the BASC age normative guideline of 5 years old. Some analyses utilized age restricted subsamples, given their examination of agespecific measures as follows: Sample A, the total sample, 108 individuals, all ages ( M = 10.8 years, SD = 3.4); Sampl e B, 42 individuals, participants ages 12 years and older ( M = 14.4 years, SD = 1.4); and Sample C, 81 individuals, participants ages 8 years and older ( M = 12.2 years, SD = 2.6). Individuals in the overall sample were 28.7% female and 29.6% of the
37 partici pants reported a nonWhite racial /e thnic cultural background (70.4% White, 14.8% African American, 9.3% Hispanic, 5.5% Other). Table 23 presents the primary diagnoses f or individuals in each sample, including AttentionDeficit/Hyperactivity Disorder Per vasive Developmental Disorders, Disruptive Behavior Disorders, Learning Disorders, Mood Disorders, Anxiety Disorders, Language Disorders, and other disorders. Note that most individuals had comorbid diagnoses: 76% in Sample A, 74% in Sample B, and 77% in Sample C. Table 24 presents the individuals primary classes of reported psycho tropic medications including stimulants, antidepressants, atypical antipsychotics, and other mood stabilizers Note that participants endorsed using m ultiple types of prescr iption psychotropic medications : 28%, 43%, and 33% in Samples A, B, and C, respectively. Procedure A ssessments were performed to evaluate the individuals psychological and emotional functioning. Referral questions typically addressed whether behavioral, learning, and/or emotional problems were present. Assessments also commonly evaluated individuals academic performance in relation to their behavioral/emotional disturbances and cognitive functioning in order to inform subsequent treatment and academic accommodations Prior to assessment, participants were informed that, upon their consent, their deidentified data would become part of a large HIPAA compliant clinical research database maintained by the colleges Division of Medical Psychology. Researc h staff obtained a signed informed consent from the legal guardian of all participants who agreed to have their data enter the database. Before completing the assessment battery, participants and their parents or primary caregivers engaged in a one hour semi structured clinical interview conducted
38 by a predoctoral level psychology intern and the licens ed psychologist. Evaluations lasted between three and six hours depending on the referral question, the length of the participant specific testing battery, and the participants and examiners joint pacing. Evaluations generally began at 9:00 a.m., ended around 3:00 p.m., and included a one hour break at midday for lunch. After the evaluation, the psychology intern composed an individualized report under the supervision of the licensed psychologist. The participant and parent or primary caregiver were invited to a feedback session with the intern and licensed psychologist approximately two to three weeks later in order to review the findings, discuss recomm endations, and address other questions. Measures Consistent with general consensus on responsibly conducting pediatric quality of life ( QOL ) research (Bastiaansen et al., 2005; Ravens Sieberer et al., 2006), data in the present study regarding the child and adolescent participants were collected from both parents/primary caregivers and the pediatric participants themselves when feasible. Because of the dependent nature of the parent child relationship and the centrality of the parents perspective in determining whether a child will seek treatment (Matza et al., 2004), parent/primary caregiver data figured prominently in the study. Researchers have argued that the perspective of significant others is vital in psychiatric research because mental health sym ptoms may distort self assessment (Saintfort et al., 1996). Younger childrens limitations in self asses sment of complex psychological symptoms such as anxiety have also been documented (Rebok et al., 2001). Additionally, due to e vidence that executive dysfunction may be detected better by family members who can assess children s performance in everyday life situations the most well known (Sherman et al., 2006, p. 1938) parent proxy assessment of EF was administered.
39 Measures Completed by the Parents/Primary Caregivers Demographics. Parents or primary caregivers completed a general demographic form, includi ng childs age, sex, ethnicity and medication status. Quality of life. The parent proxy version of the Pediatric Quality of Life Inventory V ersion 4.0 (PedsQL; Varni, Seid, & Kurtin, 2001) was administered to assess QOL. This measure was designed to capture the World Health Organizations (1948) three core dim ensions of functioning: Physical, Emotional and Social. The measure also includes item s evaluating academic functioning. The 23item inventory typically requires fewer than 4 minutes to complete. Parents or caregivers indicated how much of a problem their child has had on each item during the past month using a 5point Likert type sca le ranging from n ever to almost always (e.g., In the past ONE month how much of a problem has your child had with walking more than one block/feeling sad or blue/getting teased by other children/keeping up with schoolwork?). Raw scores per item are on a 0 4 scale and are reverse scored and linearly transformed to a 0 100 scale (0 = 100, 1 = 75, 2 = 50, 3 = 25, 4 = 0). The total score is the mean of the 23 items, with h igher scores represent ing b etter QOL. Age appropriate PedsQL versions were administ ered based on the following childrens age ranges: ages 2 4, 5 7, 8 12, and 13 18 years. The items for each version are essentially identical, differing only in the usage of developmentally appr opriate language. Varni and colleagues (2001) reported a hig h i nternal consistency coefficient for the total score Similarly, Cronbachs alpha for the current sample was .91. Correlations with other measures of disease burden and the significantly different scores for children with and without a chronic health condition provide evidence for the instrument s validity (Varni, Burwinkle, Seid, & Skarr, 2003).
40 Consistent with prior research (Varni, Seid, & Kurtin, 2001), construct validity was further evidenced by significant associations between scores indicative of poorer QOL and more school absences, inabi lity to play, and increased overall sickness. Anxiety. The Parent Rating Scale (PRS) of the Behavior Assessment System for Children, Second Edition ( BASC; Reynolds & Kamphaus, 2004) was completed by parents or primary caregivers to assess childrens anxiety The measure typically requires 10 to 20 minutes to complete. A commonly employed behavior checklist that assesses emotional and behavioral domains of children's functioning, the measure provides scores on broad internalizing, externalizing, and behav ior symptom domains as well as specific adaptive/soc ial functioning skills scales. As illustrated in these examples, anxiety items generally assess fear, worry, and nervousness. Different versions of the PRS correspond to ages 2 through 5 (Preschool form 134 items), 6 through 11 years (Child form, 160 items) and ages 12 through 21 (Adolescent form, 150 items). The present study examined the Anxiety scale of the PRS, which has 13 items on the Preschool form, 14 items on the Child form and 11 items on the Adolescent form. Items are rated on a four point Likert type scale of Never, Sometimes, Often, and Almost Always (e.g., caregivers assess the frequency of how often their child worries about making mistakes/is nervous/is fearful). Responses are associated with point values that are summed to form a raw score, which is then converted to a T score for the appropriate gender and age range. T scores greater than 65 indicate a significantly elevated level of anxiety ( i.e., scores 1.5 SD above the normative mean; Reynolds & Kamphaus, 2004).
41 According to a review by Tan (2007), the BASC is psychometrically sound, including high internal consistency reliability across domains and evidence of good construct validity. Factor analyses detected moderate to high loadings, and criterionrelated validity has been demonstrated by examining correlations between BASC composites/scales and other widely used child assessment instruments. The authors reported test retest reliabilities in the low .90s and internal consis tencies in the range of .80 .87. Alphas for parent forms in the current sample ranged from .84 to .90. Executive f unctioning. The Behavior Rating Inventory of Executive Function (BRIEF; Gioa, Isquith, Guy, & Kenworthy, 2000) was completed by parents or primary caregivers to assess childrens EF behaviors in a real world setting. The instrument contains 86 items that measure different aspects of EF in children between the ages of 5 and 18. I t typically requires 10 to 15 minutes to complete the measure. There are eight theoretically and empirically derived scales in total. Inhibit, Shift, and Emotional Control constitute the Behavioral Regulation Index (BRI). Initiate, Working Memory, Plan/Organize, Organization of Materials, and Monitor form the Metacognition Index (MI). Collectively, these indices form the Global Executive Composite (GEC). Within the Behavioral Regulation Index, the Inhibit scale assesses ones ability to control impulses and appropriately stop behavior. The Shift scale measures t he ability to transition freely from one task to another or to shift from one aspect of a problem to another in accordance with contextual demands. The Emotional Control scale measures the ability to control emotional responses appropriately. Within the Met acognition Index, the Initiate s cale measures the ability to begin a task and independently produce problem solving strategies. The Working Memory
42 scale assesses ones ability to retain information in order to complete a task. Whereas the Plan/Organi ze scale measures the ability to anticipate future events or perform tasks systematically, the Organization of Materials scale assesses the ability to maintain an orderly play area or workspace. Finally, the Monitor scale assesses ones ability to monitor progress on work or be attuned to ones behavioral impact on others. Items for the BRIEF are rated on a 3point scale (1 = nev er, 2 = sometimes, 3 = often) Raw scores for scales are converted to standard s cores by summing their items, then obtaining the corresponding T value for the appropriate gender and age range. Index scores are calculated by summing their respective raw scale scores, which has a related T value. For the total score (GEC), raw scores are summed for the BRI and MI. The total is m atched with a T value A standardized score equal to or greater than 65 (i.e., scores 1.5 SDs above the mean) indicate significant elevation in executive dysfunction. The BRIEF has established convergent reliability with related measures and samples ha ve demonstrated internal consistency reliabilities in the range of .88 .98 across all scales and indices (Gioa et al. 2000). Internal consistency alphas for the analogous indices used in the current sample ranged from .69 to .88. A ttesting to the ecolog ical validity of the BRIEF, Donders (2002) commented that the instrument assessed molar aspects of everyday behavior, and it ap p ears to measure something that is not routinely captured by other existing instruments, and that it may offer incremental knowl edge about the daily functioning of children and adolescents (p. 230). In relation to the instruments capacity to discern between EF subdomains, f indings from Gioia, Isquith, Retzlaff, and Espys (2002) factor analysis supported the discriminating
43 value of the B RI and MI They also provided evidence for a three factor solution structure for the BRIEF, which further supports a fractionated view of EF and was explored in post hoc analyses of the present study Measures Completed by the Child/Adolescent Participant Quality of life. T he self report version of the PedsQL ( Varni, Seid, & Kurtin, 2001) was administered to participants age 12 and older in order to assess QOL from the child/a dolescents perspective. In their 2003 study, Varni and colleagues reported an internal consi stency alpha of .8 9. Similarly, the alpha for the current sample was .90. Consistent with the findings for the parent proxy form in the 2003 study and prior research (Varni et al., 2001), construct validity for the child self report was supported by t he relationship between poorer self report scores and protracted school absences / general sickness. S elf report forms are scored in the same fashion as the parent forms indicated above. Anxiety The Self Report of Personality (SRP) of the B ASC (Reynolds & Kamphaus, 2004) was completed by children and adolescents ages 12 and older. The SRP includes items in True/False format and the four point Likert type scale used with the PRS. The SRP forms require 20 30 minut es to complete. Tan (2007) noted that the instrument has solid psychometric properties, with factor analytic studies yielding moderate to high loadings in support of construct va lidity. Test retest reliabilities were reported in the range of upper .70s to low .80s for SRP s cales for respondents below the college level (Reynolds & Kamphaus, 2004). Cronbachs alpha for the Anxiety scale in the current sample was .84. Scoring parallels that of the parent forms already discussed.
44 Executive functioning. The Delis Kaplan Executive Function System (D KEFS; Delis, Kaplan, & Kramer, 2001) was administered to child/adolescent participants ages 8 and older This instrument assesses critical EFs (e.g., in clud ing flexibility in thinking category switching, and the ability to inhibit automatic or dominant responses ) by measuring participants performance in a gamelike format in a controlled clinic setting. The present study utilized three specific D KEFS tests: (a) Trail Making, (b) Verbal Fluency, and (c) Color Word Interference. T he Trail Making Test measures flexibility of thinking on visual motor task s in five different conditions: Visual Scanning, Number Sequencing, Letter sequencing, Number Letter Switching, and Motor Speed. The Verbal Fluency Test measures ability to generate verbal responses according to set rules within a 60 second time period. The following three conditions were administered: Letter Fluency, Category Fluency, and Category Switching. Lastly, a version of the Stroop Test (Stroop, 1935), the Color Word Inter ference Test was given to measure inhibition of verbal responses through naming incongruent ink colors. All four increasingly more complex conditions were administered. Raw scores are converted to standardized scores ( M = 10; SD = 3) for each test. Infl uenced by prior factor analytic research (Latzman & Markon, 2010) composites for Inhibit and Monitoring EF subdomains were constructed using available archival data. Sufficient data were not available to assess the Shifting subdomain. It is noteworthy t hat whereas the I nhibition factor continues to denote the capacity to deliberately inhibit prepotent or automatic responses, the Monitoring factor is similar to Miyakes updating dimension, [which] reflects the abilities of actively monitoring and evaluat ing information (Latzman et al., p. 456).
45 Although internal consistency alphas in the current sample ranged from .75 to .87, t he D KEFS psychometric properties have generated considerable debate in the literature regarding the instruments limitations an d appropriate use. According to a review by Shunk, Davis, and Dean (2006), internal consistency coefficients ranged from low to high across the tests and age groups, which extend to an upper limit of 89 years. They added that this has been a popular cri ticism of the D KEFS system but does not pose serious concern because of the difficulties associated with measuring executive functioning (p. 277). Whereas Schmidt (2003) p ointed out the scant evidence supporting the measures validity, Delis, Kramer, Ka plan, and Holdnack (2004) retorted with a list of more than 25 studies demonstrating the D KEFS sensitivity to assess EF capacity in various clinical groups. Homack, Lee, and Riccios (2005) t est review further attested to the discriminant and convergent validity by citing several supporting studies. However, they also noted that it is unclear to what extent the instrument assesses EF in everyday functioning due to a lack of evidence demonstrating its ecological validity Taken collectively, proper caut ion is required when interpreting D KEFS scores, and it appears judicious to heed prior researchers suggestions that this measure might best be characterized as a research tool that can expand our knowledge of EF within more controlled (i.e., research lab or clinic) environments (Crawford, Sutherland, & G arthwaite, 2008; Homack et al.). Analyses After reviewing various analytical approaches to test the hypothesis that EF mediates the eff ects of anxiety on QOL in youth, mediation analyses w ere conducted u sing Preacher and Hayes (2008) indirect.sps m acro (SPSS versio n 18.0.3). The first
46 hypothesis position ed EF (BRIEF) as a potential mediator for the effect of anxiety (BASC) on QOL (PedsQL). The second hypothesis test ed a deconstructed EF model using the B RIEFs two primary indices (i.e., BRI and MI) as simultaneous mediators Significant differences in this model inspired an exploratory post hoc analysis of alternative empirically supported BRIEF factors (Gioa et al., 2002) as simultaneous mediators to more fully assess the mediation capacity of EF subdomains. Finally, a subsequent model investigated EFs multidimensional nature using two factors from the D KEFS child performance measure. All models tested the effects of age, sex, ethnicity, and ADHD as covariates as suggested by prior research (Bastiaansen et al., 2005; Barkley, 1997; Diamond et al., 2007; Utsey et al., 2002). A description of how missing data were handled is followed by the rati onale for conducting the mediation analysis. Finally a closer look at the analyses for each hypothesis concludes the chapter. Multiple Imputation for Missing Values Each variable or measure examined in this study demonstrated less than ten percent missing values. Step two of Table 21 summarizes the process followed for addressing missing values. As discussed and recommended by Schlomer, Bauman, and Card (2010), missing values were handled by utilizing the multiple imputation capabilities of the Amelia package (Honaker, King, & Blackwell, 2011) of R software ( R Development Core Team, 2010). Given that Amelia relies on the assumption of multivariate normality (i.e., imputed values are drawn from a normal distribution), the imputation w as conducted only after variable values were linearly transformed to achiev e univariate normality in effort to enhance multivariate normality. For example, a transformed PedsQL variable was calculated as follows: Transformed PedsQL =
47 [max(PedsQL) + 1 PedsQ L]^.8. Different transformations were utilized for each variable to bes t accommodate their distributions. Subsequently, multivariate normality was tested and supported with the energy test of multivariate normality ( E = 1.988, p = .593; mvnorm.e test function of ENERGY package for R; Rizzo & Szekely, 2011 ) even though multip le imputation with non normal data has proven effective/successful (Graham & Schafer, 1999). The Amelia function also requires the assumption that data are missing at random (MAR) or missing completely at random (MCAR). As suggested by Schlomer and colle agues (2010), dummy variables were calculated to indicate missing or nonmissing for each variable and were shown to be correlated with other variables used in the analyses and imputations, thereby suggesting a MAR pattern of missingness. For example, all dummy variables were significantly correlated with age. The only exception was the PedsQL dummy variable, which lacked t wo values Follow up inspection of the omitted values yielded no discernible pattern, and thus missingness fo r PedsQL was assumed to be MCAR. Overall, a lthough it appeared the data were MAR, it may not have been problematic if missingness were not missing at random (NMAR), given that multiple imputation has performed reasonably well with NMAR data (B uhi, Goodson, & Neilands, 2008). Following the 20 multiple imputations of missing values, variable values were reversetransformed to their original state. Samples were then created according to the measureappropriate ageranges indicated in step 3 of Table 21. Sample A consisted of all participants. Sample B consisted of child participants ages 12 years and older as appropriate for the self report form of the BASC. Sample C consisted of child
48 participants ages 8 years and older as appropriate for the D KEFS. To facilita te interpr etation and reduce possible effects of multicollinearity, continuous predictor variables were then centered on their withinsample means. Dummy variables for female, minority, and ADH D were left as originally coded (i.e., female, minority, ADHD diagnosis = 1; male, White, nonADHD diagnosis = 0). As reported in steps four and five of Table 21, all analyses were performed on each multiply imputed data set. Results from these analyses were aggregated using Rubins (1987) guidelines for combining results from multiply imputed data ( mi.inference function of NORM package for R; Novo, Schafer, & Fox, 2011). Given that Rubins methods rely on estimate values, their corresponding standard errors, and normal theory, this studys bootstrapbased estimates (not having standard error estimates) and confidence intervals (CIs) were aggregated simply by calculating the mean estimate and CI across results from multiply imputed data. Rationale for Mediation Analysis Zanna and Fazio (1982) described the generations through which a line of research evolves. A first generation research question for the line of research at hand would be whether anxiety and EF are related to QOL. A simple correlational analysis could examine whether these relations exist, and a multiple regression analysis could examine the size of the effect either anxiety or EF have on QOL, while controlling for the effects of each other. As reviewed herein, v arious studies have already demonstrated such correlations between anxiety and QOL (see reviews by Mendlowicz & Stein, 2000; Olatunji et al., 2007); anxiety and EF (e.g., Airaksinen et al., 2004; Micco et al., 2009); and EF and QOL (e.g., Klassen et al., 2004; Sherman et al., 2006 ).
49 A secondgeneration research question would be when, or under what conditions these relations exist or vary. For example, studies have demonstrated that variables such as age and sex have more than demographic value in pediatric QOL research. Bastiaansen and colleagues (2005) examined different factors that impact QOL i n a sample of children with psychiatric disorders. Results indicated that the deleterious impact of mental disorders on childrens QOL increased with age. Mental illness also affected girls QOL more than it did boys QOL. These findings help support th e inclusion of age and sex as cov ariates in the present study. The present study proposed to address what Zanna and Fazio (1982) called a third generation research question. It examined the process or mechanism underlying the relationship between anxiety and QOL. Castaneda and colleagues (2008) underscored the value in better elucidating this relationship specifically for these variables In accordance with ACT ( Eysenck et al., 2007), it was hypothesized that EF would be a mechanism though which anxiety affects QOL Specifically, the hypothesis was that higher anxiety would be associated with (or lead to) reduced EF, which in turn would be associated with (or lead to) reduced QOL. This is a basic mediation model, with EF mediating the effects of anxiet y on QOL. Using standard path label nomenclature (cf. Fritz & MacKinnon, 2007), Figure 21 presents a path diagram for a model in which the independent variable X (i.e., anxiety) has an effect on the dependent variable Y (i.e., QOL). The path coefficient c represents the total effect of X on Y Compare with Figure 2 2 which presents a path diagram for a model in which the effect of X on Y is mediated by a variable M (e.g., EF). The path coefficient a represents the effect of X on M and coefficient b re presents the subsequent
50 effect of M on Y The path coefficient c represents the direct effect of X on Y The indirect effect of X on Y could be represented by the product of coefficients a and b (i.e., ab) Thus, Figure 2 1s total effect c corresponds to the sum of Figure 22 s direct effect c and indirect effect ab In notation, c = c + ab To test the hypothesis that EF mediates the effect of anxiety on QOL (Figure 2 3) the analysis must test the process sequence of effects. The most substantive question is whether the combined effect of X on M and M on Y of anxiety on EF, and EF on QOL is significantly different from zero. More succinctly, the analysis must test the significance of the ab product the indirect effect of anxiety on QOL. Evaluat ing approaches to mediation. Over the past few decades, methodologists for mediation analysis have streamlined approaches to testing the significance of combined effects such as this. Baron and Kenny (1986) helped researchers better understand and utiliz e meditational analyses. Although the basic guidelines they offered appeared responsible for the subsequent proliferation of mediation analyses, many researchers have continued using them despite mediation specialists general consensus regarding their li mited accuracy and essential obsolescence (Fritz & MacKinnon, 2007; Preacher & Hayes, 2004). Some methodologists have proposed using only the second and third steps of the four steps Baron and Kenny suggested for assessing mediation (i.e., a is significan t and b is significant). Although this joint significance outperforms the four steps together and is more powerful than the Sobel test (Fritz & MacKinnon; MacKinnon, Loc kwood, Hoffman, West, & Sheets, 2002), other approaches described below provide a more accurate estimate of the CI for the ab product.
51 The Sobel test (1982, 1986) performs the exact statistical test the proposed study aims to conduct the significance of the ab product but ultimately under delivers. Its tests and estimates of statistical s ignificance assume the ab sampling distribution is normal, whereas it is often skewed. MacKinnon and colleagues (2002) demonstrated the resulting relatively high rate of Type II errors. As a nonparametric alternative to the Sobel test, resampling or bootstrapping draws a large number of samples (e.g., 1000 samples or more, with replacement) from the data, each time estimating ab (MacKinnon, Lockwood, & Williams, 2004; Shrout & Bolger, 2002). Preacher and Hayes (2004) described a percentile approach in which they calculated a point estimate as the mean of these ab estimates. They suggested calculating a 95% CI by identifying estimates corresponding to the 2.5th and 97.6th percentiles when sorting the estimates by size. If this interval does not include the value zero, the implication is that ab is statistically significantly different from zero, at a 95% CI However, given that the distribution of ab estimates is often skewed, the percentilebased interval often does not center on the point estimate. Preacher and Hayes (2008) indirect.sps macro uses bias correction (Efron & Tibshirani, 1993) to overcome the effects of this skew. This approach has demonstrated better accuracy and power (Fritz & MacKinnon, 2007). In sum, having reviewed different approaches to mediation analysis, the present study utili zed Preacher and Hayes (2008) indirect.sps macro to generate bias corr ected and accelerated bootstrap CIs f or indirect effects (5, 000 resamples). This approach appears to be among the most well established and powerful (Fritz & MacKinnon, 2007; Woody, 2011), and it has been recommended for psychological research ( Mallinckrodt, Abraham, Wei, & Russell, 2006) Following Preacher and Kelleys (2011) guidelines,
52 kappasquared effect sizes ( 2) were estim ated using the MBESS package (Kelley & Lai, 2010) in R Software (R Development Core Team, 2010). The number of analyses outlined in the next sections in flate s experimentwise error rate (i.e., Type I error). Rather than manage this increase by adopting a lower alpha, the significance criterion was kept at p = .05 due to the potential to contribute to an understudied and critical topic ( Bastiaansen et al., 2004; Castaneda et al., 2008). Noting the debate over what level of significance to use, Howell (2002) recommended that the decision is most appropriately based on researchers thoughtful inspection of the imp lications of Type I and Type II errors in their studies. Given the exploratory nature of th is research, the decision was to accept a possibly higher Type I error rate in order to optimize the sensitivity of detecting theor ypredicted e ffects to inform future research. Hypothesis 1 Because of the theoretical reasons cited (Eysenck & Calvo, 1994; Eysenck et al., 2007), i t was hypothesized that EF would mediate the relationship between anxiety and QOL as illustrated in Figure 23 This hypothesis was tested in two iterations with (a) parent and (b) child raters for the anxiety and QOL variables. I n Hypothesis 1A, it was expected that increased scores on the parent report of the BASC Anxiety scale (indicating more distressing anxiety levels) w ould be associated with increased sc ores on the BRIEF EF measure (indicating greater executive dysfunction) which in turn w ould be associated with reduced scores on the parent report PedsQL (indicating poorer QOL) In Hypothesis 1B, the same mediation via the BRIEF was tested for pediatric self reported anxiety scores and pediatric self reported QOL scores. Table 25 presents a tabulated conceptualization of how this studys various analyses related and
53 differed. The table demonstrates that Hypothesis 1A and 1B each use the BRIEF total score as a mediator, but differ in using the parent v ersus child forms of the PedsQL and BASC Anxiety scale. M odels also covar ied age, sex ethnicity, and ADHD, as indicated by prior research ( Barkley, 1997; Bastiaansen et al., 2005; Diamond, 2007; Utsey et al., 2002; Zelazo et al., 2003). Preacher and Kelley (2011) described the challenges and underuse of effect size measures w ith mediation models, and their strategies were used to calculate effect sizes for indirect effects (i.e., 2) Hypothesis 2 Moving from a more global to granular level of analysis, Hypothesis 2 examined which specific aspects of EF mediated anxietys effects on QOL (Figure 2 4) According to ACT (Eysenck et al., 2007), anxiety compromises the inhibit and shif t domains of EF. In order to examine the specific EF domains, the B ehavioral Regulation Index (BRI) of the BRIEF was hypothesized to be the primary mediator because it chiefly assesses inhibition and shifting, as compared to the primary updating EF subdom ain measured by the Metacognition Index (MI). A nalyses were completed in two iterations, both using the BRI and MI as simultaneous mediators of the effect of anxiety on QOL. Hypothesis 2A tested the simultaneous mediation for the parent forms of the BASC Anxiety scale and PedsQL. T he second iteration tested the mediation us ing the child self reports of anxiety and QOL. Again, i t was hypothesized that the BRI (containing the i nhibit and shift EF subdomains) w ould exhibit a stronger mediation effect than the MI (containing updating). The v ariables were analyzed using the indirect.sps macro. A ge, sex ethnicity, and ADHD were c ovari ed. E ffect sizes were determined as earlier. Because results provided evidence for the differential mediation power of the BRIEFs BRI and MI indices, exploratory post hoc analyses were conducted to further
54 investigate the multidimensional nature of EF. Specifically, the Behavioral Regulation factor and the Emotional Regulation factor of Gioia and colleagues (2002) BRIEF confirmatory factor analysis were compared as simultaneous mediators (Figure 25) A note of caution is in order due to the semantic resemblance (e.g., BRI versus BR factor ) in the designation of these EF measures. Emerging from an initial exploratory factor analys is vital to the construction of the BRIEF as an everyday world (Gioia et al., p. 251) measure of EF, the Behavioral Regulation Index (BRI) is one of the two primary indices of the instrument, and it consists of the Inhibit, Shift, and Emotional C ontrol scales. Subsequent research by Gioia and Isquith (2002) examined the Monitor scale, which forms a part of the other principle BRIEF index, the Metacognition Index (MI). In keeping with their hypotheses, the Monitor scale contained items reflecting two distinct dimensions: monitoring of task related activities and monitoring of personal behavior. Not only did these dimensions demonstrate stability over time, but they also split their allegiance among the BRIEFs primary indices. Self monitoring loads strongly on the BRI, and task monitoring on the MI. Thus, follow up confirmatory factor analytic work was conducted to arrive at a more nuanced conceptualization of EF via BRIEF measurement. Gioia and colleagues concluded that the initial two factor model may be supplemented by consider ing the better fitting three factor solution, including the Metacognition factor, the Emotional Regulation factor (ER) and the Behavior Regulation factor (BR) Due to the current studys theoretical framework positing that the i nhibit and shift subdomains of EF are impacted by anxiety (Eysenck et al., 2007), exploratory analyses were performed to examine the Behavior Regulation factor (containing the Inhibit scale ) and the Emotional Regulation factor (containing the Shift
55 scale) in order to assess if there were diff erences in mediation strength. No h ypotheses were formed a priori. For clarification, Table 25 contextualizes Hypothesis 2 analyses and exploratory analyses in relation to other aspects of the study Whereas prior models relied on parent assessed EF, a final model investigated EFs fractionated nature using two factors from the Delis Kaplan Executive Function System (D KEFS; Delis, Kaplan, & Kramer, 2001). Guided by a factor analytic study (Latzman & Markon, 2010) and subsequent research (Latzman, Elkovitch, Young, & Clark, 2010), D KEFS data were used to form composites for inhibition and monitoring EF subdomains in order to further explore their potential mediating role in the relationship between anxiety and pediatric QOL (Figure 26). Potentially confusing terminology is again noteworthy. In the D KEFS context, the Inhibition factor continues to denote the ability to deliberately inhibit prepotent or automatic responses The Monitoring factor however is similar to Miyakes updating dimension, [which] reflects the abilities of actively monitoring and evaluating information (Latzman et al., p. 456). Post hoc Analyses The lack of analyses on pharmacological interventions is noteworthy in the reviewe d QOL literature (e.g., Bastiaansen et al., 2004, 2005; Varni et al., 2001, 2003) and likely stems from the divergent emphases of the cited studies. The absence of attention to medication effects does not imply the topic is unimportant. In fact, medicati on use could reasonabl y impact the predictor, mediator, and criterion variables in the current study Effects may be both direct and indirect, or relevant to the relationships among the variables. D ue to the proportion of participants reporting use of pr escription medication (60% of total sample) followup analyses were conducted to assess effects related to participants medication usage. W ith the BRIEF total score as
56 the mediator variable, s cores for 108 individuals on the parent versions of the BAS C Anxiety scale and the PedsQL were analyzed. Covariates from the prior best fitting model s were included. Results were aggregated across 20 multiply imputed data sets. Finally, m edication status was also assessed as a potential moderator of anxiety's eff ects on EF, and then again as a moderator of EF's effects on QOL (Figure 27).
57 Table 21 Steps in data handling and analyses Step Description 1 Archive a) Select cases from assessment period for PedsQL (dependent variable) b) Perform standard preparat ion of analysis va riables for correct format (e.g., dummy coding, convert age from months to years, etc.) c) Calculate internal consistency reliabilities 2 Missing values a) Transform variable values for univariate normality in order to improve multivariate normality b) Perform multiple imputation for missing values, using all analysis variables (including covariates; Amelia package, R software, 20 imputations) c) Reverse transform variable values 3 Sampling a) Determine samples by lower age limits for measures per analysis (upper limit: 18 years old) Sample A: all ages ( N = 108) Sample B: ages 12 and up ( n =42; lower age limit for BASC child self report form used) Sample C: ages 8 and up ( n = 81; lower age limit for D KEFS) b) Mean center continuous predictor variables within each sample 4 Mediation analyses a) Check significance of various combinations of mediators and covariates usin g SPSS indirect macro (Preacher & Hayes, 2008) b) Process models for each of 20 multiply imputed data sets c) Aggr egate each models results across imputations using mi.inference function (NORM package, R software), based on Rubins (1987) guidelines to aggregate multiply imputed data 5 Effect sizes a) Calculate effect size for each individual mediator variable using mediation function (MBESS package, R software), based on Preacher and Kelleys (2011) guidelines for kappasquared b) Process effect size mediation models for each of 20 multiply imputed data sets c) Calculate mean ab product and effect size across multipl y imputed data sets
58 Table 22. Age, sex, and ethnicity for samples Age N % Female % Minority M Mdn SD Range Sample A 108 28.7 29.6 10.8 10.5 3.4 4.3 18.8 Sample B 42 33.3 35.7 14.4 14.3 1.4 12.3 18.8 Sample C 81 29.6 29.6 12.2 12.6 2.6 8.0 18. 8 Note Sample A contained two individuals (ages 4.3 and 4.6 years) that fell below the standard BASC age normat ive guideline of 5 years old.
59 Table 23. Primary diagnoses for each sample Diagnosis Sample A ( n = 108) Samp le B ( n = 42) Sample C ( n = 81) Attention Deficit/Hyperactivity Disorder (ADHD) ADHD Predominantly Inattentive Type 15 9 14 ADHD Combined Type 42 14 25 ADHD Not Otherwise Specified (NOS) 3 0 2 Pervasive Developmental Disorders (PDD) Asperge r's Disorder 5 1 2 Autistic Disorder 1 0 1 PDD NOS 7 1 6 Disruptive Behavior Disorders Conduct Disorder 2 1 2 Oppositional Defiant Disorder 3 2 2 Learning Disorders Mathematics Disorder 3 3 3 Reading Disorder 2 0 1 Disorder of Written Expression 1 1 1 Learning Disorder NOS 1 1 1 Mood Disorders Major Depressive Disorder Depressive Disorder NOS Dysthymic Disorder 3 1 3 Mood Disorder NOS 3 2 3 Anxiety Disorders Generalized Anxiety Disorder Anxiety Disorder NOS 3 2 3 Obsessive Compulsive Disorder 2 1 2 Panic Disorder 1 0 1 Language Disorders Expressive Language Disorder 1 0 1 Mixed ReceptiveExpressive Language Disorder 2 0 1 Other Diagnoses Cognitive Disorder NOS 1 0 1 Psychotic Disorder NOS Schizophrenia 2 1 2 Tic Disorder NOS 1 1 1 Adjustment Disorder 1 1 1 Separation Anxiety Disorder 1 0 1 No Diagnosis 2 0 1 Note. Most participants had multiple diagnoses: 76% in Sample A, 74% in Sample B, 77% in Sample C.
6 0 Table 24 Primary medications reported in each sample Sample A ( n = 108 ) Sample B ( n = 42 ) Sample C ( n = 81 ) No medication s 43 10 32 Stimulant 33 15 25 S elective serotonin reuptake inhibitor (SSRIs) 12 8 10 Non stimulant/norepinephrine reuptake inhibitor (NRIs) 3 1 2 Atypical antipsychotic 6 3 4 Anti hypertensive 5 1 2 Anti convulsive 3 1 3 Norepinephrine dopamine reuptake inhibitor (NDRIs) 2 2 2 Noradrenergic and selective serotonergic anti depressant (NaSSAs) 1 1 1 Note. Percentages of participants report ing m ultiple med ications follow : 28% in Sample A, 43% in Sample B, 33% in Sample C.
61 Table 25 Specification of models analyzed Analysis index Mediator 1 Mediator 2 Sample Mediators based on BRIEF Parent form of PedsQL and BASC 1A Total A 2A.1 BRI MI A 2A.2 ER BR A Child form of PedsQL and BASC 1B Total B 2B.1 BRI MI B 2B.2 ER BR B Mediators based on D KEFS Parent form of PedsQL and BASC 2C Inhibit Monitor C Child form of PedsQL and BASC 2D Inhibit Mon itor B Note All analyses test ed age, sex, minority, and ADHD status covariates. BRI = Behavioral Regulation Index. MI = Metacognition Index. ER = Emotional Regulation factor. BR = Behavioral Regulation factor.
62 Figure 2 1 Path diagram depicting a model in which the independent variable X has an effect on the dependent variable Y. The path coefficient c represents the total effect of X on Y. Figure 2 2 Path diagram depicting a model in which the variable M mediates the effect that variable X has on variable Y. The path coefficient a represents the effect of X on M and coefficient b represents the effect of M on Y. The path coefficient c represents the direct effect of X on Y, whereas the ab product represent s the indirect effect of X on Y Figure 23 Path diagram representing the EF (BRIEF total score) mediation and the effects of covariates on the mediator and outcome.
63 Figure 24 Path diagram representing simultaneous mediators and the effects of covariates on mediators and outcome. BRI refers to Behavioral Regulation Index of BRIEF. MI refers to Metacognition Index of BRIEF. Figure 25. Path diagram representing simultaneous mediators and the effects of covariates on mediators and outcome. ER refers to Emotional Regul ation factor of BRIEF. BR refers to Behavioral Regulation factor of BRIEF.
64 Figure 26 Path diagram representing simultaneous mediators and the effects of covariates on mediators and outcome. Inhibit and Monitor refer to D KEFS factors Figure 2 7 Path diagram representing the effects of anxiety on EF, as moderated by medication status. The g path is an estimate of the OnMeds variable in interaction with the anxiety variable. The OnMeds variable is also included among the covariates depicted i n the model as contributing to predictions of EF and QOL.
65 CHAPTER 3 RESULTS Table 31 presents means, standard deviations, ranges, internal consistency alphas, and (initial) percentages of missing data for the measures used in Samples A, B, and C I nternal consistency reliabilities were generally above .70 and as high as .91. The slightly lower alpha of .69 for the B ehavior Rating Inventory of Executive Function (BRIEF; Gio i a et al., 2000) total score was derived from combining its two primary indices (B ehavioral R egulation I ndex [BRI] and M etacognition I ndex [MI]), and thus was somewhat expected in light of this s tudys second hypothesis assessing the multidim ensional nature of executive functioning ( EF ) The low alpha of .56 for the BRIEF Behavioral Regulation factor (BR; Gioia et al., 2002) corresponded to an empirically derived aspect of the BRIEF examined as a post hoc analysis. Archival constraints did not permit the omission of 4 specific items of the BR factor, which likely compromised the internal consistency. These items load on a third factor that was not of theoretical interest in this analysis. Given the exploratory nature of the post hoc analysis model using the BR factor, the lower alpha and slightly modified factor place constraints on interpretation of the analyses, yet do not jeopardize the validity of the studys main findings that were unaffected by this issue. Similarly, the rate of missing values less than ten percent for any variable within any samplelikely pose d little threat to the validity of the results, especially given the robust handling of missing values via the multiple imputation process R egression assumptions were checked and supported the viability of the analyses. Checks were made for unusual or influential data (e.g., cases who were outliers for size of residuals and whose data had unusual leverage to influence model estimates, as per
66 Cooks D), normality of residuals, heteroscedasticity, collinearity or multicollinearity, and nonlinearity of relationships between predictors and outcomes (UCLA: Academic Technology Services, 2011). There appeared to be no trends of multicollinearity (tolerance generally above .5 and VIF generally below 2.0) or nonlinear relationships (based on examination of partial regression pl ots). Slight heteroscedasticity was noted in the models, which indicates that the variance of residuals was not always homogeneous across levels of the predicted values (i.e., commonly less variability for the highest and lowest predicted values). This potentially affected the accuracy of inferences regarding p values or statistical significance of model parameters. In a very few models, the distribution of residuals barely met significance for nonnormality, which could also affect the accuracy of infer ences regarding statistical significance. However, given that heteroscedasticity and nonnormally distributed residuals do not affect parameter estimates themselves, the parameters of primary interest the ab products representing indirect effects and the associated effect sizes remained unaffected. Given that statistical significance of indirect effects was determined by nonparametric bootstrapping, it was not susceptible to heteroscedasticity or nonnormally distributed residuals. T able 32 presents the correlations among the various measures and covariates analyzed within each sample. Correlations between measures not used together within an analysis may of course be less relevant. Parent assessed anxiety ( M = 62.5, SD = 14.7) and child self assessed anxiety ( M = 52.31, SD = 10.90) were correlated ( r = .45, p = .004) in the current subsample (Sample B, n = 42) where measures were obtained
67 from both raters.1For EF, the m ean score o f the BRIEF total composite was elevated ( M = 70.1, SD = 1 0 5 ) in keeping with other pediatric mixed clinical samples (Gioia et al., 2000, 2002). Prior research suggests that parental of EF may not correlat e with child performance measures (Vriezen & Pigott, 2002), and a similar pattern was detected in the current study. Apart from the Delis Kaplan Executive Function System (D KEFS; Delis et al., 2001) factors inverse correlation with BRIEF Metacognition I ndex (MI; r = .34, p < .05), no other significant relationships emerged. M ean scores on the B ehavior Assessment System for Children, Second ( BASC; Reynolds & Kamphaus, 2004) A nxiety scale differed significantly between parent report and childreport, t ( 37) = 4.93, p < .001, indicati ng higher levels of parent observed childrens anx iety, which appears consistent with available literature on internalizing disorders ( Ravens Sieberer et al., 2006). Regarding quality of life ( QOL ) parent and child scores were significantly correlated ( r = .52, p < .001). P arent assessment ( M = 6 0.55, SD = 19.64) was significantly lower ( t ( 3 8) = 4.93, p < .001) th an child self reported QOL ( M = 6 9.33, SD = 15.75). Mean QOL in the current study was lower than that of other studies that used the Pediatric Quality of Life Inventory Version 4.0 (PedsQL; Varni et al. 2001) For example, Bastiaa nsen and colleagues (2004) clinical pediatric sample reported higher parent assessed QOL ( M = 71.98, SD = 1 3 03 ) The correlation between parent and child report resembled that of the current sample ( r = .51, p < .01). V arni and colleagues (2003) study described scores for healthy ( M = 82.29, SD = 15.55) and chronically ill (i.e., diagnoses of asthma, diabetes, depression, ADHD; M = 73.14, SD = 1 Correlations and t tests for the BASC and PedsQL described in this paragraph were pe rformed on the data prior to multiple imputation.
68 16.46) participants between the ages of 2 and 16 ( N = 10,241). The Pearson product moment correlation for the entire sample was .61 ( p < .01). The authors calculated a minimal clinically important difference (MCID) score of 4.50 for the parent assessed PedsQL score. An MCID refers to the smallest difference in a score of a domain of interest that patients perc eive to be beneficial and that would mandate, in the absence of troublesome side effect and excessive costs, a change in the patients management (Varni et al., p. 332). Notably, the current samples mean parent assessed QOL falls more than two MCIDs bel ow Varni and colleagues sample mean. Hypothesis 1 This studys primary aim was to examine whether EF mediated the deleterious effect of anxiety on QOL in a pediatric clinical sample. It was hypothesi zed that the total score for EF (measured with the B R IEF ) would demonstrate statistical significance as a mediator of the effect of anxiety (measured with the Anxiety scale from the BASC on the total score of the PedsQL) T his hypothesis was tested with (a) parent form s for the PedsQL and the BASC Anxiety s cale and (b) child self report form s o f the same measures. Hypothesis 1A : Parent Forms for BASC Anxiety and PedsQL, BRIEF as Mediator The first iteration of Hypothesis 1 examined scores for 108 individuals on the parent versions of the BASC Anxiety scale and the PedsQL, with the BRIEF total score as the mediator variable. Table 33 provides parameter estimates for the model, including age, female, minority, and ADHD covariates (i.e., female, minority, ADHD diagnosis = 1; male, White, nonADHD diagnosis = 0 ) The first row of estimates in the table shows QOL regressed on anxiety (parameter c from Figure 21), before accounting for EF as a mediator, but controlling for the covariates ( p = .002)
69 The second portion of the table provides estimates for a model including the BRIEF total score as a mediator of the effect of anxiety on QOL. The first estimate is for the a path, or the BRIEF total score regressed on the BASC anxiety score. The estimate indicates that a onepoint increase on the anxiety scale corresponded to a .204 increase on the BRIEF, indicating increased executive dysfunction ( p = .011). T he estimate for the b path corresponds to Peds QL scores regressed on the BRIEF. The estimate indicates that a onepoint increase on the BRIEF ( i.e., increased executive dysfunction) corresponded to a 1.131 decrease on the QOL measure ( p < .001). The c parameter indicates the direct effect of anxiety on QOL after removing the indirect effect via the mediator variable. The c estimate suggests a similar negat ive relationship between anxiety and QOL a onepoint increase in the former corresponded to a .168 decrease in the latter but this relationship was not statistically significant ( p = .093) However, t he main estimate of interest in T able 3 3 is the indirect effect of anxiety through the BRIEF total mediator : the ab product. This estimate indicates that the indirect effect of a onepoint increase on the BASC anxiety measure through the EF mediator corresponded t o a .229 point decrease in QOL (95% confidenc e interval [CI] .417 to 070 ). Statistical significance does not address effect size. Preacher and Kelley (2011) developed the mediation function (MBESS package, R software) to calculate a number of effect size measures for indirect effects, ultimat ely recommending kappasquared ( 2) as the preferred statistic. Its values range from zero to one, indicating the ratio of the obtained indirect effect estimate in relation to how large it could have possibly been, given the data and model specification. However, Preacher and Kelleys mediation
70 function only estimates simple mediation models (i.e., a single mediator and no covariates).2Due to these software constraints, best practice was followed and effect sizes were calculated on the simple mediation model with only the BRIEF total s core as mediator, without inclusion of covariates. The indirect effect estimate was significant as shown in the lower portion of Table 33 (95% CI, .329 to .043). Preacher and Kelley (2011) recommended that 2 values can be evaluated with Cohen s (1988) guidelines of small ( .01), medium (. 09), and large (. 25). This effect was medium to large ( 2 = .150; 95% CI, 036 to .264). C ovariates in the model are listed in Table 33 as parameters d1 to d4 ( in relati on to the mediator ) and e1 to e4 ( in relation to t he dependent variable). Refer to Figure 2 3. T he only significant covariates were minority and ADHD in relation to the BRIEF and ADHD in relation to QOL. Individuals of minority (nonWhite) status had B RIEF scores that were 4.5 points lower on average ( i.e., less executive dysfunction), whereas individuals with a primary diagnosis of ADHD had BRIEF scores that were 4.388 points higher on average ( i.e., more executive dysfunction). Individuals with an AD HD diagnosis had QOL scores that were 10.303 poi nts higher, on average, than individuals with other diagnoses (e.g., internalizing disorders, learning disorders, Pervasive Developmental Disorders) in the mixed clinical sample. Given that the model report ed in Table 33 included a mix of parameters some significant and others not including covariates with no significant effects at all, a pared 2 Preacher confirmed that there is no existing prepared software function or macro to calculate 2 for more complex models, and that to create one would be a formidable challenge requiring matrix algebra and mathematical software programming far exceeding that used in the indirect macro or the MBESS mediation function (K. Preacher, personal communication, October 23, 2011).
71 down model was estimated. Table 34 presents estimates for a model omitting the age and female covariates, retaining the minority and ADHD covariates. The general trend of the estimates and their interpretation remains the same, and the adjusted R2 for the pared down model demonstrates superior model fit with the nonsignificant parameters removed. Hypothesis 1B: Ch ild Forms for BASC Anxiety and PedsQL, BRIEF as Mediator The second iteration of Hypothesis 1 differed from the first only in that it examined child self report forms of the BASC Anxiety scale and the PedsQL ( Table 3 5) The archive included fewer individuals in the proper age range (12 years and older) for the se lf report forms (n = 42; Sample B). The indirect effect for this model is nonsignificant, along with most other parameters in the model. The effect size estimate is low ( .053; 95% CI, .003 t o .191) but the CI allows for the possibility that the effect may be medium. Hypothesis 2 As a follow up to the first hypothesis examination of whether general EF appears to mediate the relationship between anxiety and QOL, Hypothesis 2 examined specific EF subcomponents for their relative strength in the mediation role. It was expected that in accordance with Attentional Control Theory (ACT; Eysenck et al., 2007), the Behavioral Regulation Index (BRI) of the BRIEF would demonstrate a stronger mediation effect than the Metacognition Index (MI) of the BRIEF, if the latter were to demonstrate any mediation effect at all. The BRI primarily assesses inhibition and shifting, whereas the MI evaluates updating, task initiation, and monitoring of performance. C ollectively, t hese two indices comprise the BRIEF total score.
72 A first set of Hypothesis 2 analyses used the BRI and MI indices as simultaneous mediators of the effect of anxiety on QOL. T he first iteration examined the parent forms of the BASC Anxiety s cale and PedsQL, and the second iteration assessed the child self report forms of each. The significant differences found between these mediators supported the hypothesis that particular subdomains of EF would demonstrate different mediation strength and spurred additional e xploratory analyses comparing empirically derived and theoretically relevant a lternative BRIEF factors (i.e., Behavior Regulation factor and Emotional Regulation factor ) as simultaneous mediators (Gioia et al., 2002). Finally, a set of analyses compared the mediation strength of two EF subdomains as assessed by the D KEFS child performance measure. Subdomain composition was informed by factor analy sis ( Latzman & Markon, 2010) and subsequent research (Latzman et al., 2010) Hypothesis 2A.1 : Parent F orms for BASC Anx iety and PedsQL, BRI and MI as Mediators The first iteration for Hypothesis 2 examined scores for 108 individuals on the parent versions of the BASC Anxiety scale and the PedsQL, with the BRIEF BRI and MI as simultaneous medi ator variables. Table 36 presents the estimates for t his simultaneous mediator model, including all covariates. T he BRI indirect effect (a1 b1) was statistically significant (95% CI, .388 to .095) and the MI indirect effect ( a2 b2) was not (95% CI .159 to .008) Before interpreting these results, the nonsignificant covariates of age and sex were eliminated from the model ( Table 37 ) and the adjusted R2 demonstrated better model fit. The model retained the minority and ADHD covariates due to si gnificant relationship s with EF and QOL, respectively After shedding the nonsignificant
73 parameters, the MI indirect effect demonstrated statistically significan ce (95% CI, .171 to .014). The BRI indirect effect estimate had a greater absolute value than that of the MI, and s ignificance testing for the contrast of indirect effects indicated that they di ffered significantly (95% CI, .295 to .026) Tested in single mediator models as per software limitations, t he BRI effect size appeared medium large ( 2 = .193; 95% CI, .073 to .310) and the MI effect size ( 2 = .072; 95% CI, .007 to .165) appeared smallmedium Compared to the effect size of .150 for the BRIEF total examined in Hypothesis 1A, th ese results may suggest that the BRI dimension of EF (i.e ., the inhibit and shift subdomains) functions as a purer mediator of anxietys negative effect on QOL. Estimates for the covariates in Table 37 indicate two statistically significant relationships Again, c ovariates were dummy coded (i.e., female, minority, ADHD diagnosis = 1 ; male, White, nonADHD diagnosis = 0). Minority was associated with lower BRIEF MI scores by 5.165 points, on average, suggesting less executive dysfunction in this domain ( p = .01). The presence of an ADHD diagnosis was associat ed with higher PedsQL scores ( p < .001), indicating that those with the externalizing disorder reported a 9.88 higher QOL score, on average, indicating better QOL than those with internalizing (e.g., anxiety and mood disorders) and other disorders i n this mixed clinical sample. Hypothesis 2B.1: Child F orms for BASC Anxiety and PedsQL, BRI and MI as Mediators The above analysis comparing simultaneous mediators of the BRIEF BRI and MI was repeated in a second iteration using child self report forms of the B ASC Anxiety scale and the PedsQL (Table 38). The archive included fewer individuals in the proper age range (12 years and older) for these forms ( n = 42; Sample B). Again, the smaller
74 sample size carried with it diminished power to detect effects The indirect effect and covariates were not statistically significant. The effect size of .015 for the BRI is rather small and indistinguishable from zero. The effect size for the MI ( 2 = .079; 95% CI, .005 to .265) is closer to medium and its wide CI ranges from small to large. Analysis 2A.2: Parent Forms for BASC Anx iety and PedsQL, ER and BR as Mediators The significant differences found between the BRIEF BRI and MI as simultaneous mediators in Hypothesis 2A.1 supported the idea that specific dimensions of EF would dem onstrate different strengths of mediation and spurred this additional exploratory analys is driven by a prior factor analytic study. G ioia and colleagues (2002) outlined a threefactor model of the BRIEF subscales. Two factors were of particular interest for the present study because resultant analyses could shed light on potential differences in mediation strengths between the inhibit and shift EF subdomains The Emotional Regulation (ER) factor was comprised of the Shift and Emotional Control subscales. The Behavioral Regulation (BR) factor was comprised of the Monitor and Inhibit subscales. (Note, as explained earlier, the Behavioral Regulation factor is linguistically similar though compositionally different than the Behavioral Regulation Index .) Archival limitations did not allow for the selection of specific items of the Monitor scale, thereby resulting in a BR factor that contained 4 additional task monitoring items. These 4 items load on the Metacognition factor, which was not of theoretical interest in this exploratory analysis. Exploratory Analysis 2A.2 used the parent forms of the BASC A nxiety scale and the PedsQL, with the BRIEF ER and BR factors as simultaneous mediators (Figure 2 5) These analyses examined data for the 108 individuals of Sample A.
75 Table 39 presents the estimates for this simultaneous mediator model, including all covariates. T he ER indirect effect ( a1 b1) was statistically significant (95% CI, .510 to .177), whereas the BR indirect effect (a2 b2) was not. N onsignificant parameters were eliminated to test a more parsimonious model ( Table 3 10). Model fit was not compromised (i.e., adjusted R2 of 539 decreased to .532). The model retained the ADHD covariate, which again demonstrated a positive association w ith QOL ( p = .008). The effect size for the ER indirect ef fect in a simple mediator model appeared large ( 2 = .273; 95% CI, .160 to.383) larger than both the effect sizes for BRIEF total in Hypothesis 1A and BR I in Hypothesis 2A.1, which suggest s that the shifting subdomain of EF functi oned as the most potent mediator of anxietys effec t on QOL observed in t his study, crosssectional design implication s notwithstanding. Analysis 2B.2: Child Forms for BASC Anx iety and PedsQL, ER and BR as Mediators The above analysis comparing simultaneous mediators of the BRIEF ER and BR was repeated in a second iteration u sing child self report forms of the BASC Anxiety scale and the PedsQL ( Table 3 11) The archive included fewer individuals in the proper age range (12 years and older) for these forms (n = 42; Sample B). Again, t he smaller sample size compromises statist ical power and influences interpretation of findings. Although the mediation indirect effect was not statistically significant, t he effect size of .091 for ER was medium to possibly large (95% CI .009 to .261) Hypot hesis 2C: Parent Forms for BASC Anxi ety and PedsQL, D KEFS Inhibit and Monitor as Mediators A final comparison of simultaneous child performance mediators using the D KEFS was influenced by prior factor analytic research ( Latzman & Markon, 2010) and subsequent research employing the factors ( Latzman et al., 2010). Composites for
76 I nhibit and M onitor ing EF subdomains were based on Latzman and colleagues design (Figure 26). As referenced earlier, although the Inhibition factor retains its usual meaning of the capacity to i nhibit dominant res ponses, the Monitoring factor resembles Miyake and colleagues (2000) updating subdomain, which refers to the capacity to evaluate and monitor information. T he Inhibit factor is the mean of two other composites: the m ean of the five Trail Making condition s and the mean of the Color Word Inhibit and Inhibit Switch con ditions. The Monitor factor is the mean of the three Verbal Fluency conditions. Archival limitations did not allow for strict adherence to Latzman and colleagues suggested incorporation of t he Design Fluency mean.3Table 312 presents the results of this model, demonstrating nonsignificance for essentially all parameters. Effect sizes for the D KEFS Inhibit and Monitor indirect effects are rather small at .002 and .010 respectively which seems to sugges t that the lack of significant findings was not simply due to low statistical power. The only significant covariates were minority (i.e., higher Inhibit scores ) and age (i.e., older individuals having higher Monitor scores ) Hypothesis C tested a mediation model with these simultaneous mediators using the parent forms of the PedsQL and BASC anxiety scale. Given that the D KEFS is appropriate for ages 8 and above, the resulting Sample C included 81 individuals Hypo thesis 2D: Child Forms for BASC Anxiety and PedsQL, D KEFS Inhibit and Monitor as Mediators The above analysis comparing simultaneous mediators of the D KEFS Inhibit and Monitor factor s was repeated in a second iteration using child self report forms of the 3 Attempted correspondence with the D KEFS lead author regarding this measurement c oncern was unsuccessful.
77 BASC Anxiety scale and the PedsQL ( Table 3 13) The archive included fewer individuals in the proper age range (12 years and older) with the self report forms (n = 42; Sample B). Size constraints and power concerns remained and may represent plausible explanations as to why th e indirect effect s for this model were not statistically significant However, because the Monitor indir ect effect had a 2 of .049, significant results may have been obtained in a larger sample. Post hoc Analyses: Possible Effects Related to Medication S tatus The absence of analyses on pharmacological interventions is somewhat surprising in the reviewed QOL literature (e.g., Varni et al., 2001, 2003), especially when considering how medications may influence the effects of this studys variables in clinical samples (e.g., Bastiaansen et al., 2004, 2005). Although few studies have examined QOL in relation to anxiety (Sherbourne et al., 1996), Mogotsi, Kaminer, and Stein (2000) reported that medications have been associated with enhanced QOL in anxiety dis order patients They added that further empirical evidence is needed, especially in relation to the links between treatment and Q OL change. Though neither a treatment nor an outcome study, the current research may partly illuminate the anxiety/medication relationship, at least insofar as it pertains to a mixed clinical sample. Child and adolescent participants medication status was gathered during the clinical interview as part of the assessment. Because 60% of the total sample reported current use of prescription medication, follow up analyses were conducted to assess effects related to participants medication usage. No hypotheses were formed a priori. Due to limited numbers of each particular medication type (Table 2 4), pediatric participants formed two medicationstatus groups: (a) a medication nave group, including participants not taking prescription medications for psychiatric conditions and
78 (b) a medication group, composed of children currently prescribed stimulants selective serotonin reuptake inhibitors nonstimulant/norepinephrine reuptake inhibitors norepinephrine dopamine reuptake inhibitor noradrenergic and selective serotonergic antidepressant anti convulsives anti hypertensives and atypical antipsychotics Worthy of note is the vast range of medicationtypes contained within the medicated subgroup. Accordingly, interpretation of the following post hoc analyses may be constrained by the dichotomizing of medicated versus nonmedicated participants as w ell as the wide variety of psychiatric diagnoses for which the medications were prescribed (Table 23). Hypothesis 1A was revisited adding m edication status as a covariate. W ith the BRIEF total score as the mediator variable, s cores for 108 individuals on the parent versions of the BAS C Anxiety scale and the PedsQL were analyzed. Covariates from the prior best fitting model were included (i.e., minority, ADHD diagnosis, on medication = 1; White, nonADHD diagnosis, medication nave = 0). Results were aggregated across 20 multiply imputed data sets. All trends in the results remained identical. Most notably, the mediation remained significant, suggesting that EF as a unitary construct mediated the anxiety and QOL relationship even in the presence of controls for the effects of medications on EF and QOL. Similarly, trends were also identical for Hypothesis 2A.1 and Exploratory Analysis 2A.2 when the medication status covariate was added to the models These findings s uggest t hat the particular EF subdomains demonstrated similar m ediation properties despite controls for effects of m edication status on EF and QOL
79 Whereas the post hoc analyses above covaried medication status in models to examine and control for its effect on EF and QOL, medication stat us was also examined as a potential moderator of anxiety's effects on EF, and then again as a moderator of EF's effects on QOL. A series of models tested the significance of the dummy coded medication status variable ( i.e., on medication = 1, medicationn ave = 0) in interaction with anxiety as a predictor of EF, and in interaction with the various EF measures (e.g., BRIEF total, BRI, MI, ER, and BR) as predictors of QOL. Anxiety and EF were meancenter to minimize multicollinearity problems and facilitate interpretation. Interaction terms were nonsignificant and thus were not retained in any models. However, tw o potential interactions may be of note (Figure 31). Estimates for the negative effect of anxiety on EF were larger for individuals taking p resc ription medications, yet the differences were not statistically significant. Specifically, anxiety's association with executive dysfunction measured by the BRIEF total score (as in Hypothesis 1A) appeared to occur almost exclusively among the group of individuals taking m edications. Their anxiety parameter estimate was .260 higher ( p = .089) than the medicationnave group estimate of .035 ( p = .784) Similarly, the medication group estimate for anxiety in relation to the BRIEF MI index (as in Hypothesis 2A.1) was .260 hig her ( p = .064) than the medicationnave group estimate of .141 ( p = 390). Whereas the p values did not meet the preestablished criterion for significance, some may be close enough to warrant attention in future studies of how medication usage may moderat e the effect of anxiety on EF. Extant studies on clinical levels of pediatric anxiety and EF tend to not examine medication status of participants (e.g., Emerson et al., 2005; Kendall & Chansky, 1991; Micco et al., 2009; Toren et al. 2000), perhaps in
80 part due to the nascence of this line of investigation and the consequent prioritization of other research questions within mixed clinical samples. Nevertheless, questions pertinent to medication use and its associated effects merit consideration. As noted, cautious interpretation of the current findings is warranted given the heterogeneity of the medications assessed in addition to the fact that they were prescribed for a range of psychiatric disorders rather than anxiety disorders or anxious symptomatology per se. Results Summary Table 314 provides a conceptual structure for how the various iterations of models and analyses fit together and build upon one another. In sum, Hypothesis 1 tested EF as a unitary construct whereas hypothesis 2 used three pairs of EF variables as simultaneous mediators: the first two pairs from the parent assessed BRIEF, and the third pair from the childperformed D KEFS. For the independent and dependent variables, all analyses were performed using (a) the parent forms and (b) the child self report forms of th e BASC Anxiety scale and PedsQL. There were two main findings. First, results from tests of Hypothesis 1 provided supporting evidence that EF may in fact function as a mediator of the relationship between anxiety and QOL. Secondly, results from tests of Hypothesis 2 support ed EF s multidim ensionality and the differing strength of specific subcomponents as m ediators of the relationship between anxiety and QOL. Table 314 also indicates which mediat ors demonstrated significant indirect effects a nd provides their effect sizes.
81 Table 31. Means, SDs, ranges, alphas, and percent missing for measures Variables M SD Range Alpha % missing Sample A ( n = 108) PedsQL Parent 6 3.9 17.7 13 99 91 1. 9 B ASC Anx Parent 56.8 13.7 28 90 .84 90 .9 BRIEF Total 70.1 10. 5 40 96 71 6.5 BRIEF BRI 66.9 13.3 35 96 .82 6.5 BRIEF MI 69.7 9.8 4 1 95 88 6.5 BRIEF ER 64.0 12.8 3 7 93 83 6.5 BRIEF BR 66. 6 10.2 39 91 70 6.5 Sample B ( n = 42) PedsQL Child 71. 3 15.6 12 99 90 2.4 B ASC Anx Child 5 0.9 11. 2 3 3 82 .84 9.5 BRIEF T otal 72.0 10.3 40 88 .69 .0 BRIEF BRI 68.5 14.0 41 96 80 .0 BRIEF MI 71.2 8.9 41 85 87 .0 BRIEF ER 65.5 13.1 42 93 85 .0 BRIEF BR 67.7 10.1 43 91 56 .0 D KEFS Inhibit 8.6 2.4 1.8 12.5 87 .0 D KEFS Monitor 9.4 3.3 3.0 18.3 79 2.4 Sample C ( n = 81) PedsQL Par ent 62. 2 17.5 13 99 91 2.5 B ASC Anx Par ent 58.9 13.3 35 90 84 .90 .0 D KEFS Inhibit 8.3 2.6 1.8 13.5 87 6.2 D KEFS Monitor 9.1 2.9 3.0 18.3 75 7.4 Note. Range of BASC alphas corresponds to ageappropriate forms.
82 Table 32. Correlations for measures and covariates Variables 1 2 3 4 5 6 7 8 9 10 11 12 Sample A ( n = 108 ) 1 PedsQL Par 2 B A SC Anx Par ent 35** 3 BRIEF Total 66** 22* 4 BRIEF BRI 66** 29** 86** 5 BRIEF MI 53** 13 .91 ** 57** 6 BRIEF ER 71** 37** 77** 93** 49** 7 BRIEF BR 49** 11 87** 81** 75** 6 1 ** 8 Age 07 28** 06 02 07 02 01 9 Female 12 21* 09 04 16 05 12 03 10 Minority 09 05 18 08 24* 08 10 14 04 11 ADHD 24* 28** 13 06 14 05 20* 09 13 03 Sample B ( n = 42) 1 PedsQL Child 2 B ASC Anx Child 57** 3 BRIEF total 26 09 4 BRIEF BRI 27 05 87** 5 BRIEF MI 21 15 90** 58** 6 BRIEF ER 46** 21 79** 91** 52** 7 BRIEF BR 07 15 84** 84** 68** 60** 8 D KEFS Inhibit 08 09 27 11 34* 11 18 9 D KEFS Monitor 04 27 25 14 29 20 17 60** 10 Age 12 12 23 18 22 24 04 19 23 11 Female 28 31 0 5 08 11 12 06 21 09 20 12 Minority 03 17 22 12 25 10 11 14 19 34* 11 13 ADHD 22 25 17 04 22 08 16 17 06 01 27 02 Note. p < .05. ** p < .01
83 Table 32 (cont.). Correlations for measures and covariates Variables 1 2 3 4 5 6 7 8 9 10 11 12 Sample C ( n = 81) 1 PedsQL Par ent 2 B ASC Anx Par ent 27* 3 D KEFS Inhibit 03 05 4 D KEFS Monitor 04 12 62** 5 Age 06 15 14 18 6 Female 16 27* 23* 14 01 7 Minority 05 08 06 16 23* 05 8 ADHD 26* 19 05 04 08 17 06 Note. p < .05. ** p < .01
84 Table 33. Estimates for h ypothesis 1A mediator m odel (all covariates) Model Parameter Estimate SE p CI (lower) CI (upper) Model without mediator (total effect of IV on DV) BASC Anx Par c .398 .129 .002 .652 .145 Model with mediator Predictors for mediator (BRIEF total) BASC Anx Par a .204 .080 .011 .047 .361 Age d1 .112 .333 .737 .542 .765 Female d2 1.487 2.277 .514 2.978 5.951 Minority d3 4.500 2.166 .038 8.747 .254 ADHD d4 4.388 2.056 .033 .358 8.418 Predictors for outcome (PedsQL Par) BRIEF total b 1.131 .122 < .001 1.370 .893 BASC Anx Par c' .168 .100 .093 .363 .028 Age e1 .191 .396 .63 0 .587 .968 Female e2 .143 2.749 .958 5.246 5.533 Minority e3 .830 2.698 .759 6.119 4.460 ADHD e4 10.303 2.582 < .001 5.242 15.364 Adjusted R2 for DV model R2 Y,MX .527 < .001 Indirect effect (BRIEF total) a b .229 .417 .070 Model with mediator and no covariates Indirect effect (BRIEF total) a b .175 .329 .043 Effect size for indirect e ffect .150 .036 .264
85 Table 34. Estimates for h ypothesis 1A mediator m odel (selected covariat es) Model Parameter Estimate SE p CI (lower) CI (upper) Model without mediator (total effect of IV on DV) BASC Anx Par c .404 .122 .001 .643 .165 Model with mediator Predictors for mediator (BRIEF total) BASC Anx Par a 221 .0 74 .0 03 .0 75 366 Minority d1 4.362 2.141 .042 8.559 .166 ADHD d2 4.270 2.037 .036 .278 8.263 Predictors for outcome (PedsQL Par) BRIEF total b 1.126 .121 < .001 1.363 .889 BASC An x Par c' .155 .093 .096 .339 .028 Minority e1 .631 2.664 .813 5.854 4.591 ADHD e2 10.253 2.548 < .001 5.258 15.249 Adjusted R2 for DV model R2 Y,MX .534 < .001 Indirect effect (BRIEF total) a b .248 .421 .105 Model with m ediator and no covariates Indirect effect (BRIEF total) a b .175 .329 .043 Effect size for indirect e ffect .150 .036 .264
86 Table 35. Estimates for hypothesis 1B mediator model (all covariates) Model Parameter Estimate SE p CI (lower) CI (upper) Model without mediator (total effect of IV on DV) BASC An x Child Child c .800 .218 < .001 1.228 .372 Model with mediator Predictors for mediator (BRIEF total) BASC An x Child a .016 .1 71 .927 .352 .320 Age d1 1.277 1.291 .323 3.807 1.254 Female d2 1.289 3.827 .736 6.212 8.790 Minority d3 3.234 3.618 .371 10.325 3.857 ADHD d4 3.630 3.412 .287 3.057 10.317 Predictors for outcome (PedsQL Child ) BRIEF total b .453 .206 .028 .856 .050 BASC An x Child c' .809 .206 < .001 1.213 .405 Age e1 1.172 1.576 .457 1.918 4.261 Female e2 .754 4.636 .871 9.841 8.334 Minority e3 1.118 4.445 .801 7.595 9.831 ADHD e4 3.835 4.216 .363 4.430 12.099 Adjusted R2 for DV model R2 Y,MX .365 .002 Indirect effect (BRIEF total) a b .007 .145 .192 Model with mediator and no covariates Indirect effect (BRIEF total) a b .039 .087 .201 Effect size for indir ect e ffect .053 .003 .191
87 Table 36. Estimates for hypothesis 2A .1 mediator model ( all covariates ) Model Parameter Estimate SE p CI (lower) CI (upper) Model without mediator (total effect of IV on DV) BASC Anx Par c .398 .129 002 .652 .145 Model with mediator Predictors for mediator 1 (BRIEF BRI) BASC Anx Par a1 .338 .102 .001 .139 .537 Age d1 .198 .428 .643 1.038 .642 Female d2 .338 2.872 .906 5.968 5.292 Minority d3 2.392 2.750 .384 7.782 2.998 ADHD d4 3.980 2.620 .129 1.156 9.116 Predictors for mediator 2 (BRIEF MI) BASC Anx Par a2 .104 .074 .160 .041 .249 Age f1 .233 .306 .447 .367 .833 Female f2 3.424 2.097 .103 .687 7.536 Minority f3 5.463 2.000 .006 9.382 1.543 ADHD f4 3.848 1.901 .043 .122 7.574 Predictors for outcome ( PedsQL Par ) BRIEF BRI b1 .671 .113 < .001 .893 .449 BRIEF MI b2 .486 .158 .002 .795 .177 BASC Anx Par c' .121 .100 .227 .317 .075 Age e1 .043 .395 .913 .732 .818 Female e2 .104 2.731 .970 5.458 5.250 Minority e3 .005 2.687 .998 5.261 5.271 ADHD e4 9.887 2.532 < .001 4.924 14.850 Adjusted R2 for DV model R2 Y,MX .547 < .001 Indirect e ffect 1 (BRIEF BRI) a1 b1 .225 .388 .095 Indirect e ffect 2 (BRIEF MI) a2 b2 .050 .159 .008 Contrast of indirect effects 1 and 2 .175 .338 .045 Model s with mediator s and no covariates Indirect e ffect 1 (BRIEF BRI) a1 b1 .227 389 .086 Effect size for indirect e ffect 1 1 .193 .073 .310 Indirect e ffect 2 (BRIEF MI) a2 b2 .086 .213 .014 Effect size for indirect e ffect 2 2 .072 .007 .165
88 Table 37. Estimates for hypothesis 2A.1 mediator model ( selected covariates ) Model Parameter Estimate SE p CI (lower) CI (upper) Model without mediator (total effect of IV on DV) BASC Anx Par c .404 .122 .001 .643 .165 Model with mediator Predictors for mediator 1 (BRIEF BRI) BASC Anx Par a1 .323 .094 .001 .13 8 .508 Minority d1 2.584 2.705 .339 7.886 2.718 ADHD d2 4.020 2.595 .121 1.066 9.106 Predictors for mediator 2 (BRIEF MI) BASC Anx Par a2 .141 .070 .043 .004 .278 Minority f1 5.165 2.005 .010 9.095 1.235 ADHD f2 3.582 1.909 .061 .160 7.323 Predictors for outcome ( PedsQL Par ) BRIEF BRI b1 .671 .111 < .001 .888 .453 BRIEF MI b2 .484 .153 .002 .784 .184 BASC Anx Par c' .119 .092 .198 .300 .062 Minority e1 .051 2.635 .985 5.114 5.215 ADHD e2 9.880 2.495 < .001 4.989 14.770 Adjusted R2 for DV model R2 Y,MX .554 < .001 Indirect e ffect 1 (BRIEF BRI) a1 b1 .216 .364 .098 Indirect e ffect 2 (BRIEF MI) a2 b2 .068 .171 .014 Contrast of indirect effec ts 1 and 2 .149 .295 .026 Model s with mediator s and no covariates Indirect e ffect 1 (BRIEF BRI) a1 b1 .227 .389 .086 Effect size for indirect e ffect 1 1 .193 .073 .310 Indirect e ffect 2 (BRIEF MI) a2 b2 .086 .213 .014 Effect size for indirect e ffect 2 2 .072 .007 .165
89 Table 38. Estimates for hypothesis 2B .1 mediator model ( all covariates ) Model Parameter Estimate SE p C I (lower) CI (upper) Model without mediator (total effect of IV on DV) BASC Anx Child Child c .800 .218 < .001 1.228 .372 Model with mediator Predictors for mediator 1 (BRIEF BRI) BASC Anx Child a1 .108 .232 .64 1 .346 .562 Age d1 1.606 1.807 .374 5.148 1.936 Female d2 .943 5.334 .860 9.511 11.398 Minority d3 2.307 5.062 .649 12.229 7.615 ADHD d4 1.838 4.775 .700 7.520 11.196 Predictors for mediator 2 (BRIEF MI) BASC Anx Chil d a2 .077 .146 .599 .363 .210 Age f1 .788 1.085 .468 2.915 1.339 Female f2 3.193 3.228 .323 3.134 9.521 Minority f3 3.232 3.040 .288 9.190 2.727 ADHD f4 4.260 2.867 .137 1.360 9.879 Predictors for outcome ( PedsQL Child) BRIEF BRI b1 .081 .185 .659 .443 .281 BRIEF MI b2 .446 .309 .149 1.052 .160 BASC Anx Child c' .829 .214 < .001 1.249 .409 Age e1 1.268 1.596 .427 1.861 4.397 Female e2 .181 4.816 .970 9.260 9.622 Minority e3 .958 4.545 .833 7.9 51 9.868 ADHD e4 4.237 4.385 .334 4.359 12.834 Adjusted R2 for DV model R2 Y,MX .347 .005 Indirect e ffect 1 (BRIEF BRI) a1 b1 .007 .293 .060 Indirect e ffect 2 (BRIEF MI) a2 b2 .045 .096 .290 Contrast of indirect ef fects 1 and 2 .038 .333 .134 Model s with mediator s and no covariates Indirect e ffect 1 (BRIEF BRI) a1 b1 .016 .161 .085 Effect size for indirect e ffect 1 1 .015 .000 .059 Indirect e ffect 2 (BRIEF MI) a2 b2 .063 .050 .243 Effect size for indirect e ffect 2 2 .079 .005 .265
90 Table 39. Estimates for analysis 2A .2 mediator model ( all covariates ) Model Parameter Estimate SE p CI (lower) CI (upper) Model without mediator (total effect of IV on DV) BASC Anx Par c .398 .129 .002 .652 .145 Model with mediator Predictors for mediator 1 (BRIEF ER) BASC Anx Par a1 .391 .096 < .001 .204 .578 Age d1 .298 .396 .452 1.075 .480 Female d2 .768 2.716 .777 6.092 4.556 Minority d3 2.421 2.586 .349 7.490 2.648 ADHD d4 1.300 2.457 .597 3.516 6.115 Predictors for mediator 2 (BRIEF BR) BASC Anx Par a2 .121 .080 .129 .035 .277 Age f1 .000 .337 .999 .662 .662 Female f2 2.764 2 .244 .218 1.634 7.162 Minority f3 2.381 2.155 .269 6.605 1.843 ADHD f4 5.309 2.041 .009 1.310 9.309 Predictors for outcome ( PedsQL Par ) BRIEF ER b1 .793 .137 < .001 1.061 .524 BRIEF BR b2 .296 .163 .071 .616 .025 BASC Anx Par c' .052 .105 .618 .259 .154 Age e1 .173 .396 .663 .951 .605 Female e2 1.339 2.750 .626 6.729 4.052 Minority e3 1.652 2.613 .527 3.469 6.773 ADHD e4 7.938 2.602 .002 2.837 13.039 Adjusted R2 for DV model R2 Y,MX .539 < .001 Indirect e ffect 1 (BRIEF ER) a1 b1 .305 .510 .177 Indirect e ffect 2 (BRIEF BR) a2 b2 .036 .125 .006 Contrast of indirect effects 1 and 2 .269 .494 .146 Model s with mediator s and no covariates In direct e ffect 1 (BRIEF ER) a1 b1 .324 .502 .177 Effect size for indirect e ffect 1 1 .273 .160 .383 Indirect e ffect 2 (BRIEF BR) a2 b2 .066 .187 .037 Effect size for indirect e ffect 2 2 .054 .003 .146
91 Table 310. Estimat es for hypothesis 2A .2 mediator model ( selected covariates ) Model Parameter Estimate SE p CI (lower) CI (upper) Model without mediator (total effect of IV on DV) BASC Anx Par c .398 0.122 .001 .637 .159 Model with m ediator Predictors for mediator (BRIEF ER) BASC Anx Par a1 .363 .089 < .001 .189 .537 ADHD d1 1.430 2.440 .558 3.351 6.212 Predictors for outcome ( PedsQL Par ) BRIEF ER b1 .946 .101 < .001 1.145 .748 BASC Anx Pa r c' .055 .097 .573 .245 .135 ADHD e1 6.721 2.517 .008 1.787 11.655 Adjusted R2 for DV model R2 Y,MX .532 < .001 Indirect e ffect 1 (BRIEF ER) a1 b1 .343 .525 .190 Model s with mediator s and no covariates In direct e ffect (BRIEF ER) a b .324 .502 .177 Effect size for i ndirect e ffect .273 .160 .383
92 Table 311. Estimates for analys is 2B.2 mediator model (all covariates) Model Parameter Estimate SE p CI (lower) CI (upper) Model wit hout mediator (total effect of IV on DV) BASC Anx Child Child c .800 .218 < .001 1.228 .372 Model with mediator Predictors for mediator 1 (BRIEF ER) BASC Anx Child a1 .295 .209 .160 .116 .705 Age d1 2.472 1. 635 .130 5.676 .732 Female d2 .645 4.819 .894 10.090 8.800 Minority d3 1.560 4.584 .734 10.545 7.425 ADHD d4 .572 4.324 .895 9.047 7.904 Predictors for mediator 2 (BRIEF BR) BASC Anx Child a2 .129 .171 .450 .463 .206 Age f1 .205 1.294 .874 2.331 2.741 Female f2 3.021 3.845 .432 4.516 10.558 Minority f3 1.649 3.625 .649 8.754 5.455 ADHD f4 3.153 3.417 .356 3.544 9.849 Predictors for outcome ( PedsQL Child) BRIEF ER b1 .475 .224 .034 .913 .036 BRIEF BR b2 .157 .279 .574 .390 .703 BASC Anx Child c' .643 .227 .005 1.088 .197 Age e1 .543 1.657 .743 2.705 3.791 Female e2 2.114 4.673 .651 11.272 7.044 Minority e3 2.108 4.363 .629 6.444 10.660 ADHD e4 1.431 4.232 735 6.864 9.726 Adjusted R2 for DV model R2 Y,MX .373 .002 Indirect e ffect 1 (BRIEF ER) a1 b1 .144 .529 .028 Indirect e ffect 2 (BRIEF BR) a2 b2 .010 .231 .046 Contrast of indirect effects 1 and 2 .135 .519 .1 10 Model s with mediator s and no covariates Indirect e ffect 1 (BRIEF ER) a1 b1 .105 .428 .015 Effect size for indirect e ffect 1 1 .091 .009 .261 Indirect e ffect 2 (BRIEF BR) a2 b2 .035 .018 .234 Effect size for in direct e ffect 2 2 .039 .002 .205
93 Table 312. Estimates for hypothesis 2C (all covariates) Model Parameter Estimate SE p CI (lower) CI (upper) Model without mediator (total effect of IV on DV) BASC Anx Par c 8.278 3.940 036 .556 15.999 Model with mediator Predictors for mediator 1 (D KEFS Inhibit) BASC Anx Par a1 4.156 2.922 .155 9.882 1.571 Age d1 .250 .519 .630 .768 1.268 Female d2 .761 .580 .189 .375 1.898 Minority d3 6.797 3.207 .034 .511 13.084 ADHD d4 .525 3.256 .872 5.856 6.907 Predictors for mediator 2 (D KEFS Monitor) BASC Anx Par a2 .047 .489 .923 1.007 .913 Age f1 .538 .083 < .001 .375 .701 Female f2 .046 .096 .630 .142 .234 Minority f3 814 .531 .125 .227 1.855 ADHD f4 .336 .548 .539 1.411 .739 Predictors for outcome ( PedsQL Par ) D KEFS Inhibit b1 .301 .154 .050 .603 .000 D KEFS Monitor b2 .147 .981 .881 1.777 2.070 BASC Anx Par c' 7.041 3.946 .074 .693 14.774 Age e1 .310 .871 .722 1.398 2.018 Female e2 .386 .780 .620 1.142 1.915 Minority e3 2.972 4.479 .507 11.751 5.808 ADHD e4 2.340 4.352 .591 6.191 10.870 Adjusted R2 for DV model R2 Y,MX .053 .140 Indirect e ffect 1 (D KEFS Inhibit) a1 b1 1.246 .250 4.964 Indirect e ffect 2 (D KEFS Monitor) a2 b2 .051 1.192 1.076 Contrast of indirect effects 1 and 2 1.297 .552 4.900 Model s with mediator s and no covariates Indirect e ffect 1 (D KEFS Inhibit) a1 b1 .002 .021 .068 Effect size for indirect e ffect 1 1 .002 .000 .013 Indirect e ffect 2 (D KEFS Monitor) a2 b2 .012 .015 .123 Effect size for indirect e ffect 2 2 .010 .000 .077
94 Table 313. Estimates for hypothesis 2D (all covariates) Model Parameter Estimate SE p CI (lower) C I (upper) Model without mediator (total effect of IV on DV) BASC Anx Child Child c .800 .218 < .001 1.228 .372 Model with mediator Predictors for mediator 1 (D KEFS Inhibit) BASC Anx Child a1 .012 .038 .746 .0 86 .062 Age d1 .377 .301 .210 .213 .966 Female d2 1.248 .887 .159 .489 2.986 Minority d3 .526 .841 .532 1.124 2.175 ADHD d4 .530 .794 .504 2.086 1.025 Predictors for mediator 2 (D KEFS Monitor) BASC Anx Child a2 .062 .0 52 .239 .041 .164 Age f1 .453 .407 .266 .345 1.251 Female f2 .538 1.206 .655 1.825 2.902 Minority f3 .682 1.143 .550 1.557 2.922 ADHD f4 .114 1.076 .916 1.996 2.224 Predictors for outcome ( PedsQL Child) D KEFS Inhibit b1 1.113 1.156 .336 3.377 1.152 D KEFS Monitor b2 .898 .857 .295 .783 2.578 BASC Anx Child c' .871 .232 < .001 1.327 .415 Age e1 1.761 1.698 .300 1.567 5.090 Female e2 .434 5.073 .932 10.378 9.510 Minority e3 2.556 4.712 .588 6.680 11.792 ADHD e4 1.505 4.464 .736 7.244 10.254 Adjusted R2 for DV model R2 Y,MX .279 .021 Indirect e ffect 1 (D KEFS Inhibit) a1 b1 .017 .071 .240 Indirect e ffect 2 (D KEFS Monitor) a2 b2 .053 .031 .439 Contrast of in direct effects 1 and 2 .036 .446 .117 Model s with mediator s and no covariates Indirect e ffect 1 (D KEFS Inhibit) a1 b1 .004 .098 .026 Effect size for indirect e ffect 1 1 .004 .000 .026 Indirect e ffect 2 (D KEFS Mon itor) a2 b2 .049 .009 .331 Effect size for indirect e ffect 2 2 .049 .003 .261
95 Table 314. Summary of indirect effect estimates and effect sizes across models Analysis index Mediator 1 ( 2) Mediator 2 ( 2) Difference in mediators Signif icant covariates Sample Mediators based on BRIEF Parent form of PedsQL and BASC 1A Total* (.15) Minority, ADHD A 2A.1 BRI* (.19) MI* (.07) (BRI > MI)* Minority, ADHD A 2A.2 ER* (.27) BR (.05) (ER > BR)* ADHD A C hild form of PedsQL and BASC 1B Total (.05) B 2B.1 BRI (.02) MI (.08) B 2B.2 ER (.09) BR (.04) B Mediators based on D KEFS Parent form of PedsQL and BASC 2C Inhibit (.00) Monitor (.01) Age, M inority C Child form of PedsQL and BASC 2D Inhibit (.00) Monitor (.05) B Note. p < .05 for mediators indirect effects or for difference between mediators indirect effects. BRI = Behavioral Regulation Index. MI = Metacognition Index. ER = Emotional Regulation factor. BR = Behavioral Regulation factor.
96 CHAPTER 4 DISCUSSION The current study assessed executive functionings (EF) hypothesized mediation of the relation between anxiety and quality of life (QOL) in a pediatric clinical sample. With design constraints (i.e., cross sectional data) as the interpretational backdrop for the findings, support for EFs mediating role was detected. In keeping with a Processing Efficiency Theory based prediction that anxiety compromises EF as a unitary construct (PET; Eysenck & Calvo, 1994), current findings supported this disruption for the parent assessed models. Anxiety related executive dysfunction was in turn associated with decreased pediatric QOL. Further analyses assessed a prominent m odel of EF as a fractionated construct (Miyake et al., 2000) and concomitant theoretical advances regarding anxietys effect on particular subdomains of EF. Attentional Control Theory (ACT; Eysenck et al., 2007) posits that anxiety impacts the inhibit and shift subdomains of EF, and the present findings generally aligned with this theoretical prediction as well. In fact, relative to a monolithic EF construct, inhibition and shifting more strongly mediated the anxiety QOL relationship. Furthermore, explor atory analyses demonstrated the even greater mediation capacity for additional empirically derived and theoretically sound EF subdomains (Gioia at al., 2002). This chapter will contextualize current findings within the extant literature, offering limitati ons and future directions, followed by a brief summary and conclusions. Findings Contextualized in Current Literature This section discusses current findings related to EF as a global then fragmented mediator. Parent versus child r ater results are examined in light of the available re search, followed by a consideration of the implication s of EF mode of measurement
97 (i.e., parent assessment versus child performance) Finally, additional overarching findings related to pediatric anxiety and QOL are discussed. Mediation with EF as a Unitary Construct Design implications notwithstanding, findings provide preliminary support for Airaksinen and colleagues (2005) postulation that EF deficits would p redict eroded QOL related to anxious symptomatology. This study fits well within the extant literature on pediatric anxi ety and executive dysfunction (Emerson et al., 2005; Francis 1988; Kendall & Chansky, 1991; Micco et al., 2009; Toren et al., 2000). Just as Francis (1988) found that children with higher anxiety levels also had significantly more task inhibiting thoughts this study provided preliminary evidence that EF serves as a mechanism for how anxiety impacts QOL. The impairments in cognitive flexibility detected in Toren and colleagues (2000) s ample resemble the current findings supporting the plausibly adverse effects of anxiety on EF, and specifically, the shifting and inhibiting subdomains. T he current results fit well with Micco and colleagues (2009) findings. As referenced earlier, their study pr ovided tentative data regarding direction of effects by examining children of parents with anxiety and depressive disorders. Due to prior research indicating higher prevalence of psychopathology in the offspring of affected parents as compared to the offs pring of healthy parents, Micco and colleagues expected to find EF impairments in children at risk for depression and anxiety. Their r esults however, revealed no relationship between offspring status and executive dysfunction, leading them to conclude that impaired EF may not serve as a trait marker for developing anxiety or depression. Conversely, they found that children with current mood and anxiety symptoms also exhibited EF impairment, which suggested that
98 executive dysfunction may be symptomatic of depression and anxiety ra ther than causal Micco and colleagues conclusions align ed with Eysenck and colleagues (1992, 2007) theor ies on anxiety and disrupted EF, which collectively set the stage for the current study. Driven by Eysenck and Calvos (1992) theory on anxietys impairing effects on general EF, the current studys initial model using the parent assessed Behavior Rating Inventory of Executive Function (BRIEF; Gioa et al., 2000) as the mediator which has been noted to possess greater ecologi cal validity than narrowly focused performance measures, as described later demonstrated statistical significance as well as a medium to large effect size for the mediation. Due to mathematical limitations (K. Preacher, personal communication, October 23, 2011), recall that the effect sizes were only calculated in simple mediation models, and thus did not account for simultaneous mediators or the effect of covariates. Interpretation is not as clear in the child self reported models analyzing the hypothesi zed BRIEF mediation. All self report analyses for the Pediatric Quality of Life Inventory Version 4.0 (PedsQ L; Varni et al., 2001) and the Behavior Assessment System for Children, Second Edition ( BASC; Reynolds & Kamphaus, 2004) Anxiety scale examined Sam ple B, consisting of 42 individuals ages 12 and older. Essentially all parameters in these models were nonsignificant. Sometimes the effect sizes were large enough to suggest that with larger samples significant effects w ould potentially be detected. Co nsider Hypothesis 1B The low sample size and resulting diminished power to detect effects are o bvious potential considerations for why the indirect effect of anxiety on QOL is nonsignificant along with most other parameters in the model.
99 Although t h e effect size estimate was l ow the CIs expanse allowed for the possibility of a medium effect. The likelihood of whether a larger sample would have detected a significant effect is unclear though probable. Fritz and MacKinnon (2007) suggested that analyses such as the bias corrected, bootstrap mediation used in this study require at least 71 individuals in order to achieve a statistical power of .80 for detecting medium effects. Detection of small effects requires 148 to 462 individuals. The current sample, however, contained only 42 children and adolescents. In conjunction with the effect size and CI yielded herein, detection of significant findings in larger samples appears probable, at least sufficient ly enough to justify future research. Although t here is a paucity of r esearch on the role of EF in pediatric anxiety with regard to QOL, studies have demonstrated links between executive dysfunction and eroded QOL related to ADHD (Klassen et al., 2004), tra umatic brain injuries (Horneman et al., 2005), and epilepsy (Sherman et al., 2006). The current findings add to this literature, now with an emphasis on anxious symptomatology. Castaneda and colleagues (2008) highlighted the importance of exploring third variables influencing the a nxiety and QOL rela tionship, and the current study shed light on EFs potential ly mediating role as a unitary and fragmented construct. Findings suggested that particular subdomains of EF may exert even greater mediation strength. Mediation with EF subdomains Eysenck and colleagues (2007) current theory of the relationship between anxiety and EF capitalized on research advances related to what is often called the unity and disunity of EF. Investigating latent variables with a confirmatory factor analysis, Miyake and colleagues (2000) detected a threefactor model of EF that has become prominent in the field and continues to garner attention (Latzman & Markon, 2010). Their results
100 indicated that three separable yet related subdomains constitute EF: inhibition, or the ability to exert control over automatic responses; shifting, or the capacity to perform multiple operations and switch from one task to another while experiencing interference; and updating, or the capacity to monitor and assess the relevance of novel inform ation while performing the task at hand. Basing their predictions on t his deconstructed model of EF, Eysenck and colleagues further specified that anxiety impacts shiftin g and inhibiting subcomponents of EF. This studys findings supported EFs multidim ensionality. As mediators, the BRIEF Behavioral Regulation Index (BRI) containing the inhibition and shifting subcomponents of EF outperformed the updating subdomain as represented by the BRIEF Metacognition Index (MI). Although both indices demonstrated significant mediation effects for parent assessed anxiety and QOL, contrast estimates suggested the primacy of inhibition and shifting as mediating subdomains. Additionally, in Hypothesis 2A.1, a greater effect size was detected for the inhibition/shifti ng index than the updating analog. The larger effect is somewhat expected in light of the BRIs advantage derived from containing two EF subdomains in comparison to the single subdomain occupying the MI. Notably the strength of effect for the inhibit/shift composite exceeded the effect size for the BRIEF total as examined earlier, perhaps suggesting that the inhibit/shift subdomains function as a purer mediator of anxietys effect on QOL. More precisely, the shifting subdomainor the ability to move fre ely between tasks or to t ransition from one aspect of a problem to another in response to contextual demands may exhibit greater potential as a mediator in the relationship between anxiety and QOL. Promising results for the differential mediation power inspired
101 e xploratory post hoc analyses based on findings from Gioia and colleagues (2002) BRIEF confirmatory factor analysis. Whereas initial analyses showed some support for stronger mediation properties in the inhibit/shift subdomains in comparison to the updating domain, there was interest in determining how inhibition and shifting would operate independent of one another. Thus, the inhibit versus shift subdomains competed as s imultaneous mediators, and t he Emotional Regulation (ER) factor corresponding to the shift subdomain significantly mediated anxietys effects on QOL. It had a large effect size, exceeding the effect sizes of unitary EF and the inhibit/shift subdomain composite. The shifting subdomain of EF appeared to function as the strongest me diator of anxietys negative effect on QOL observed in this study. Child self reports, again, presented ambiguity in relation to assessing varying strengths of EF subdomains. In the model examining the differential power of the inhibit/shift subcomponent s (BRI) versus the updating subcomponent (MI) the updating dimension had a greater effect size than inhibit/shift. Considered independently, this inconsistency is difficult to interpret, and it may represent an artifact of a model with limited significan t parameters. The finding may also stem from differences related to self report versus parent report. However, interpretation of this difference should be tempered with caution given that the respective indirect effects did not differ at a level of stati stical significance. Although the model was not significant, power was again compromised due to the small sample size. Recall that the sample size of 42 individuals for this analysis f ell below Fritz and MacKinnons (2007) .80 threshold to detect medium sized effect s Clearly, additional research is needed to shed light on the disparities between parent and child assessment in relation to EFs potential mediation
102 of anxiety and QOL. Divergences among raters and measurement instruments are not novel in t he pediatric QOL research, and the discuss ion now turns to this topic. Rater and Measurement Divergences In the models using the parent forms for the PedsQL and the BASC Anxiety scale, the EF mediation appeared consistently significant with the BRIEF (global EF composite and subdomains as mediators), but not with the D KEFS factors. These findings raise at least two important methodological questions related to raters and measures. As recommended for pediatric QOL research (Bastiaansen et al., 2005), dat a from child/adolescent participants and parents/primary caregivers were examined. Previous research suggested that parent assessed QOL was r eliable in a sample of asthmatic children (Le Coq, Boeke, Bezemer, Bruil, & Van Eijk 2000) yet overall findings for parent child agreement are mixed (Ravens Sieberer et al., 2006). R eliability and validity are challenging iss ues in pediatric QOL assessment (Connolly & Johnson, 1999; Matza et al., 2004) p articularly with abstract psychologi cal constructs such as anxiety. Ravens Sieberer and colleagues (2006) noted that lower concordance rates predominate for internalizing or emotional content areas, and in the current study, parent assessed anxiety were significantly greater than child self reported anxiety. Generally, parent proxy report is thought to offer greater reliability perhaps at t he cost of decreased validity (Matza et al. 2004) It is difficult to assess the extent to which skewed parental or child reporting is operating. O n the one hand, the mere presence of mental health symptoms may confound self assessment (Saintfort et al., 1996), especially for children and complex mental healthrelated concepts (Rebok et al., 2001). On the other hand, the individuals unique perspective takes on even greater re levance when speaking of internalizing disorders such as anxiety, given the
103 experience is often one of private misery: undetected, misconstrued, or deliberately avoided by even the m ost well intentioned observer. The tension between the parent and child perceptions is hard to reconcile and extends beyond the scope to the current study, though it is worthy to note that it is often the parents perspective that determines whether a child will seek treatment. Accordingly, the current mixed overall findings f it well within extant literature, given that low concordance rates between children and significant others have been noted. Future studies involving multiple objective raters (e.g., health care providers, teachers) may help clarify rater discrepancies i n this line of anxiety, EF, and QOL research. Even though additional perspectives are also vulnerable to reporting bias, additional raters enrich prospect s for cross validation. M easurement concerns also pertain to EF assessment. Evidence suggests that e xecutive dysfunction may be detected better by family members who can assess childrens performance in everyday life s ituations (Sherman et al., 2006). Yet, as noted, the possibility of parental misreporting has been documented in the literature, especial ly in relation to childrens internalizing mental disorders and neuropsychological dysfunction (Matza et al., 2004) The present study detected statistically significant findings in the EF mediation models employing the parent forms for the PedsQL and the BASC Anxiety scale. Child performance on the D KEFS, however, yielded nonsignificant findings. In conjunction with disparate rater related findings, another rivaland perhaps more likely explanation for the discrepancy in EF measurement stems from the nature of the measures themselves. For instance, Vriezen, Pigott, and Pelletier (2001) found that behavioral ratings and performancebased assessment were not correlated, which may imply that children can excel on restricted testing in controlled
104 clinical environments and still experience significant impairment in everyday living. The current findings were consonant with this lack of correlation between behavioral ratings and performancebased assessment Of course, there are proponents of clinical EF ass essment ( e.g., Kalinian, 2003; Latzman et al., 2010), as it appears to discriminate between different groups in relation to executive dysfunction. O thers have argued that clinical instruments such as the D KEFS may not adequately detect compromised EF due to the highly controlled, oneonone testing environment ( e.g., Schmidt, 2003; Strauss, Sherman, & Spreen, 2006) Participants are given explicit rules and timeframes. Although a white coat effect may threaten optimal performance, the structured, qui et ambience sets the stage for effective problem solving and task performance. Acco rdingly, a general theme in the literature is that c linical e valuation may not be as ecological ly valid as an instrument like the BRIEF, which appears more attuned to EF d eficits in daily activities and thus better suited to capture real world behavior and difficulties (Denckla, 2002; Donders, 2002; Sherman et al., 2006). Banic h summarized the challenges of clinical EF measurement well, noting that the very nature of ex ecutive functioning makes it difficult to measure in the clinic or laboratory; it involves an individual guiding his or her behavior, especially in novel, unstructured, and nonroutine situations that require some degree of judgment (p. 89, 2009). It is perhaps for this reason that current findings with only the BRIEF pr ovided preliminary evidence for E Fs mediation of anxiety and QOL. That is not to say that EF assessm ent by thirdparty observers is i mmune from bias and subsequent critical interpretatio n. Denckla (2002) noted two important
105 interpretational caveats. Not unlike other modes of assessment, a level of linguistic competence is required in the parent, and such an understanding is largely influenced by divergences in colloquial versus professi onal denotations of words. Whereas Denckla highlighted the common tendency to erroneously pathologize organized and meticulous though otherwise functionally normal behavior as Obsessive Compulsive Disorder, she also noted that semantic misunderstandings are possible though less likely to occur with the BRIEF. A greater threat to validity comes in the form of emotional biases of the thirdparty observer, who may knowingly or unknowingly alter r esponses in effort to exaggerate or minimize child rens dys functional behavior. Such questions of perception and the m any influences that inform it are not unique to parent assessed EF, though responsible interpretation of results requires awareness of these factors. Additional Contributions to Pediatric Anxiety EF, and QOL Research Implicit within the mediation analyses, findings were consistent with prior research demonstrating the association between anxiety and impaired EF (Emerson et al., 2005; Francis, 1988; Kendall & Chansky, 1991; Micco et al., 2009; Toren et al., 2000). Of note, however, were the relatively low parent assessed and chi ld self reported anxiety scores in the current sample. Anxious symptomatology, when evaluated by either rater, did not meet the level of clinical significance according t o the measures guidelines Given the studys aim of broadly assessing anxious symptoms in a naturally occurring clinical setting, th is sample characteristic was somewhat unexpected in light of the high prevalence of pediatric anxiety ( Saddock et al., 2009 ). The diagnostic c omposition of the sample also made the low level of anxiety surprising, given that anxiety is one of the most frequent comorbid disorders in children with ADHD (Tannock, 2000).
106 The preponderance of ADHD diagnoses in the sample was not anticipated, nor was it uncommon due to the generally high frequency of children referred to outpati ent clinics for ADHD assessment Sixty of the 108 participants had ADHD diagnoses, which influenced the decision to control for ADHD status in order to m inimize interpretational confounds. G reater executive dysfunction a ppeared associated with ADHD in keeping with prior research. Somewhat counterintuitive, perhaps, was that individuals with an ADHD diagnosis had QOL scores that were 10.3 points higher on average than those without an ADHD diagnosis in the overall sample when examined in Hypothesis 1. Although this finding was incongruent with at least one other study that reveal ed no statistically significant QOL difference across diagnostic categor ies ( Bastiaansen et al., 2004), t his finding may be reasonable within the context of a mixed clinical sample, considering that individuals with ADHD were compared to individuals diagnosed with a mental disorder. In fact, as illustrated in Table 23, only 2 of the 108 individuals in Sample A had no diagnosis, and perhaps the nonADHD diagnoses (e.g., Pervasive Developmental, Disruptive Behavior, Learning, Mood, and Anxiety Disorders) were s imply associated with lower QOL. T he current sample reported relatively poorer QOL than others in the literature. Bastiaansen and colle agues (2004) clinical sample exhibited a mean score approximately half a standard deviation higher than what was observed in the current study Speculations could abound as to the m any unexam ined third variables or other betweensample differences, though an obvious potential confound is cultural relativism. That is, their study was conducted in the Netherlands, and cross cultural comparison of QOL in the absence of culturespecific norms is challenging at best, and the interested
107 reader can pursue this topic further. Analyses closer to home help contextualize the impoverished QOL within the current sample. Evaluating a U.S. based pediatric sample, V arni and colleagues (2003) also detected a mean QOL score that exceeded the current sample by half a standard deviation. Their sample consisted of chronically ill youth (i.e., diagnoses of asthma, diabetes, depression, and ADHD) between the ages of 2 and 16. In relation to t heir minimal clinica lly important difference (MCID)1Limitations and Future Directio ns the current sample average fell two units below Varni and colleagues sample mean, suggesting an overall poor report for childrens psychosocial, physical, and emotional well being. Prior research has documented the impai red QOL of children with psychiatric disorders in comparison to healthy and physically ill children (Sawyer et al., 2002), and the current study further attests to the compromised Q OL in a child psychiatric clinical sample. This s tudy has a number of strengths and limitations. In terms of diversity, some analyses demonstrated significantly better EF in racial ethnic participants as compared to White counterparts. Standardization of the utilized measures and a search of t he available literature revealed no documented effects for racial ethnic status (Delis et al., 2001; Gioia et al., 2000; Tan, 2007). Future studies might examine this effect more deeply, perhaps with increased multiculturally heterogeneous samples. Regar ding sample composition, the mixed clinical sample examined here is a considerable strength in that it may provide practicing psychologists with preliminary data relevant to clinical settings, though future replication and further elaboration of this work is needed. 1 As noted earlier, an MCID refers to the smallest difference in a score of a domain of interest that patients perceive to be beneficial and that would mandate, in the absence of troublesome side effect and excessive costs, a change in the patients management (Varni et al., 2003, p. 332).
108 This study adhered to the aspirational yet pragmatically challenging practice of incorporating disparate points of view (Matza et al., 2004; Ravens Sieberer et al., 2006). As referenced, f uture studies might also benefit from additional thirdparty raters (e.g., teac her or health care professionals) in order to cross validate parent reports Additionally, this study gathered self report data for children ages 12 years and older, an age range that seemed appropriate for self report on a complex psychological variable such as anxiety. Debate on age appropriateness for child self report of internalizing psychiatric symptomatology is unresolved (Bastiaansen et al., 2004, 2005; Matza et al.), and future work may consider deeper investigation of thi s topic or incorporate larger samples than the self report subsample examined herein. In keeping with third party involvement, some variables that may have been relevant and ideal to control for statistically (Kessler, 1987) were not measured. Perhaps parents or caregivers overcompensation for childrens lower EF resulted in children with very low EF experiencing somewhat higher QOL, in which case it would have been ideal to attempt to measure this effect and include it in the analyses. Even though measurement of this phenomenon may be challenging, future research may benefit from integrating parental level of accommodation. The choice to keep the significance criterion at p = .05 is relevant for interpretation of current results. I nflat ion of experimentwise error rate (i.e., Type I error ) was deemed acceptable and responsible given the limited research in pediatric QOL ( Bastiaansen et al., 2004) and the anxiety/EF relationship ( Castaneda et al., 2008). Howell (2002) underscored the debate as to what s ignificance criterion is appropriate, concluding that researchers informed views on the tradeoff between Type I and Type II errors may be
109 the wisest guide t o set an alpha level. The exploratory nature of this research provided a rationale for the decisi on, and results may best be interpreted in light of this analytic approach. Wellrespected and psychometrically solid instruments were utilized, apart from some reservations about the D KEFs as noted (Crawford et al., 2008; Homack et al., 2005) T he B RIEF i s widely used due to its ecological validity and clinical utility. Future studies may benefit from utilizing item level data for this instrument which would allow for assessment of the BR factor (tested here in a post hoc exploratory analysis) with the 4 task monitoring items omitted as prescribed by a prior confirmatory factor analysis (Gioia et al., 2002). In light of the stronger ER results, this followup work appears less critical, though this line of research is in its incipient phases and re plication is warranted. Similarly, the DKEFS factors could also be enriched by a nonrestrictive archive to better replicate Latzman and colleagues (2010) composite structure. Nonetheless, the earlier caveats related to childperformance assessment may re nder this a moot point. Additionally, future studies may be enriched by utilizing a dedicated anxiety measure (e.g., Multidimensional Anxiety Scale for Children; March, 1997) Now that preliminary evidence supporting the EF mediation hypothesis has been detected, additional contributions can be made regarding wh ich specific domains of QOL (i.e., physical, emotional, social or academic) are most involved in the relationship between anxiety and EF. Driven by Eysenck and Calvo s (1992) and Eysenck and colleagues (2007) theor ies results provided evidence for the hypothesis that EF is a m echanism through which anxiety conveys its effect on QOL. However, any study of mediation is up against
110 serious challenges. Psychologists are appropriately concerned about causality in mediation models ( Spencer, Zanna, & Fong, 2005). According to Woody (2011), one challenge is the relative ease of obtaining apparently significant tests of mediation even when the models are nonsensical He ca utioned well, reminding that establishing mediation is difficult, requiring multiple converging approaches in a creative program of research. It is not, and never will be, reducible to any formulaic statistical tes t, no matter how sophisticated (p. 243). One limitation is the cros ssectional nature of these data (i.e., for each individual, data were collected for all measures on a single occasion). Thus, t he data are especially susceptible to uncertainties related to causality and temporal ordering of the effects. Maxwell and Col e (2007) pointed out how cross sectional data insufficiently address the causal ef fects hypothesized by mediation. However, the present studys use of an empirically supported theoretical framework (see Derakshan & Eysenck, 2009) clearly bolsters the findings, which in turn invite replication The current study sets the stage for related future work in an obviously important area of research. C areful interpretation is imperative so as to not extend too far beyond the design implications. Although t he present studys theory driven approach yielded supporting evidence for EFs role as a mechanism through which anxiety affects QOL, the dynamic interplay between this set of variables appeals to reason. The likely reciprocity of influence between anxiety and QOL has been noted (Bastiaansen et al., 2004; Mogotsi et al., 2000). Some evidence points toward executive dysfunction as symptomatic of anxiety (Micco et al., 2009) yet bi directional effects c annot be ruled out by any means (Emerson et al., 2005; Francis, 1988; Kendall & Chansky, 1991; Toren et al., 2000).
111 Again, t his area of inquiry is clearly in its nascence, and t he current findings may catalyze future investigation and substantiate the investment associated with more advanced designs. Ideal mediation studies overcome the shortcomings of cross sectional data by using lon gitudinal designs, which appear well founded in light of the current findings Dwyer (1983) for instance, described the benefits of longitudinal data for evaluati ng reciprocal causation between variables. As noted by Mac K innon and Luecken ( 2008) a diversity of approaches is of substantial value in research that examines third variables (i .e., mediators and moderators) such as EF in relation to anxiety and QOL. Summary and Conclusion In light of the s carcity of r esearch on the role of EF in pediatric anxiety and QOL, the present study sought to expand the empirical knowledge base by addressing this critical gap in the literature. Using theory based predictions, the study t est ed the hypothesis that EF would m ediate the relationship between anxiety and QOL. For parent assessment of EF, results supported the predictions for (a) EF as a unitary mediator and (b) EFs multidimensional nature, including the differing mediation strengt hs of the inhibit and shift subdomains. These preliminary findings offer empirical support for conducting more strategic ally designed studies to corroborate, c ontradict or expand on the current findings. Evidence supporting the EF mediation may provoke further investigation into the potential for anxiety focused clinical interventions that emphasize bolstering EF Because Daily EF exercise appears to enhance EF development much as physical exercise builds bodies (Diamond et al., 2007, p. 1388), poss ible therapeutic approaches to anxiety the nations most prevalent mental disorder may be enriched by future investigation of the EF/anxiety/treatment nexus.
112 LIST OF REFERENCES Altemeier, L., Jones, J., Abbott, R. D., & Berninger, V. W. (2006). Executive functions in becoming writing readers and reading writers: Note taking and report writing in third and fifth graders. Developmental Neuropsychology, 29(1), 161173. American Psychiatric Association (2000). Diagnostic and statistical manual of mental disor ders (4th ed., text rev ). Washington, DC: Author. Banich, M. T. (2009). Executive function: The search for an integrated account. Current Directions in Psychological Science, 18(2), 8994. Barkley, R. A. (1997). Behavioral inhibition, sustained attention, and executive function: Constructing a unified theory of ADHD. Psychological Bulletin, 121, 6594. Baron, I. S. (2004). Neuropsychological evaluation of the child. New York: University Press. Baron, R. M., & Kenny, D. A. (1986). The moderator mediator v ariable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1 173 1182. Bastiaansen, D., Koot, H. M., & Ferdinand, R.F. (2005). Determinants of quality of lif e in children with psychiatric disorders. Quality of Life Research, 14, 15991612. Bastiaansen, D., Koot, H. M., Ferdinand, R.F., & Verhulst, F. C. (2004). Quality of life in children w ith psychiatric disorders: Self parent, and clinical report. Journal of the American Academy of Child and Adolescent Psychiatry, 43(2), 221230. Boonstra, A. M., Oosterlaan, J., Sarg eant, J. A., & Buitelaar, J. K. (2005). Executive functioning in adult ADHD: A metaanalytic review. Psychological Medicine, 35(8), 10971108. B raet, C., Claus, L., Verbeken, S., & Vlierberghe, L. (2007). Impulsivity in overweight children. European Child & Adolescent Psychiatry, 16, 473 483. Buhi, E. R., Goodson, P., & Neilands, T. B. (2008). Out of sight, not out of mind: Strategies for handling missing data. American Journal of Health Behavior, 32, 83 92. Bullinger, M. (2002). Assessing health related quality of life in medicine: An overview over concepts, methods and applications in international research. Restorative Neurology and Neuroscienc e, 20 (3 4), 93101. Calhoun, S. L., & Dickerson Mayes, S. (2005). Processing speed in children with clinical disorders. Psychology in the Schools, 42(4), 333343.
113 Castaneda, A.E., TuulioHenriksson, A., Marttunen, M., Suvisaari, J., & Lonnqvist, J. (2008). A review on cognitive impairments in depressive and anxiety disorders with a focus on young adults. Journal of Affective Disorders, 106, 1 27. Cohen, J. (1988). Statistical power analysis fo r the behavioral sciences (2nd ed.). Hillsdale, NJ: Lawrence Erl baum Associates. Crawford, J. R., Sutherland, D., & Garthwaite, P. H. (2008). On the reliability and standard errors of measurement of contrast measures from the DKEFS. Journal of the International Neuropsychological Society, 14, 10691073. Cserjesi, R., L uminet, O., Molnar, D., & Lenard, L. (2007). Is there any relationship between obesity and mental flexibility in children? Appetite, 49, 675 678. Darakshan, N., & Eysenck, M. W. (2009). Anxiety, processing efficiency, and cognitive performance. European Ps ychologist, 14(2), 168176. Delis, D. C., Kaplan, E., & Kramer, J. H. (2001). Delis K aplan Executive Funct ion S ystem San Antonio, TX: The Psychological Corporation. Denckla, M. B. (2002). The behavior rating inventory of executive function: Commentary. C hild Neuropsychology, 8(4), 304306. Diamond, A., Barnett, W. S., Thomas, J., & Munro, S. (2007). Preschool program improves cognitive control. Science, 318(5855), 13871388. Donders, J. (2002). The Behavior Rating Inventory of Executive Function: Introduction. Child Neuropsychology, 8(4), 229230. Durston, S. (2003). A review of the biological basis of ADHD: What have we learned from imaging studies? Mental Retardation and Disabilities Research Reviews, 9 (3), 184195. Dwyer, J. H. (1983). Statistical mod els for the social and behavioral sciences New York: Oxford University Press. Efron, B., & Tibshirani, R. (1993). An introduction to the bootstrap. Boca Raton, FL: CRC Press. Ellis, L. K., Rothbart, M. K., & Posner, M. I. (2004). Individual differences in executive attention predict self regulation and adolescent psychosocial behaviors. Annals of the New York Academy of Sciences 1021, 337340. Emerson, C. S., Mollet, G. A., & Harrison, D. W. (2005). Anxious depression in boys: An evaluation of executive f unctioning. Archives of Clinical Neuropsychology, 20, 539546.
114 Engel, G. L. (1977). The need for a new medical model: A challenge for biomedicine. Science, 196, 129136. Eysenck, M. W., & Calvo, M. G. (1992). Anxiety and performance: The processing effici ency theory. Cognition and Emotion, 6, 409434. Eysenck, M. W., Derakshan, N., Santos, R., & Calvo, M. G. (2007). Anxiety and cognitive performance: Attentional control theory. Emotion, 7(2), 336353. Flament, M.F., Koby, E., Rapoport, J.L., & Berg, C.J. (1990). Childhood obsessive compulsive disorder: A prospective follow up study. Journal of Child Psychology and Psychiatry, 31, 363380. Flament, M.F., Whitacker, A., Rapoport, J.L., Davies, M., Berg, C., Kalikow, K., Sceery, W., & Shaffer, D. (1988). Obses sive compulsive disorder in adolescence: An epidemiological study. Journal of the American Academy Children and Adolescent Psychiatry, 27 764771. Francis, G. (1988). Assessing cognitions in anxious children. Behavior Modification, 12, 267 280. Frisch, M. B. (2006). Quality of life therapy: Applying a life satisfaction approach to positive psychology and cognitive therapy Hobok en, NJ: John Wiley & Sons, Inc. Fritz, M. S., & MacKinnon, D. P. (2007). Required sample size to detect the mediated effect. Psycho logical Science, 18 233 239. Gioa, G. A., Isquith, P. K., Guy, S.C., & Kenworthy, L. (2000). Behavior Rating Inventory of Executive Function. Lutz, FL: Psychological Assessment Resources. Gioia, G. A., Isquith, P. K., Retzlaff, P. D., & Espy, K. A. (2002). Confirmatory factor analysis of the Behavior Rating Inventory for Executive Function (BRIEF) in a clinical sample. Child Neuropsychology, 8(4), 249 257. Graham, J. W., & Schafer, J. L. (1999). On the performance of multiple imputation for multivariate data with small sample size. In R. Hoyle (Ed.), Statistical strategies for small sample research (pp. 1 29). Thousand Oaks, CA: Sage. Hirschfeld, R. M. A. (2001). The comorbidity of major depression and anxiety disorders: Recognition and management in primar y care. The Primary Care Companion to the Journal of Clinical Psychiatry, 3(6), 244 254. Homack, S., Lee, D., & Riccio, C. A. (2005). Test review: Delis K aplan executive function system. Journal of Clinical and Experimental Neuropsychology, 27, 599609.
115 Ho naker, J., King, G., & Blackwell, M. (2011). Amelia (Version 1.54) [Computer software and manual]. Retrieved from http://cran.rproject.org/web/packages/Amelia Horneman, G., Folkesson, P., Sintonen, H., Von Wendt, L., & Emanuelson, I. (2005). Health rela ted quality of life of adolescents and young adults 10 years after serious traumatic brain injury. International Journal of Rehabilitation Research, 28, 245249. Howell, D. C. (2002). Statistical methods for psychology (5th ed.). Pacific Grove, CA: Wadswor th. Hughes, C., & Ensor, R. (2008). Does executive functioning matter for preschoolers problem behaviors? Journal of Abnormal Child Psychology, 36(1), 1 14. Jarek, S. (2011). MVNORMTEST (Version 0.17) [Computer software and manual]. Retrieved from http:/ /cran.rproject.org/web/packages/mvnormtest Jordan, B. K., Schlenger, W. E., Hough, R.L., Kulka, R. A., Weiss, D. S., Fairbank, J. A., Marmar, C. R. (1991). Lifetime and current prevalence of specific psychiatric disorders among Vietnam veterans and controls. Archives of General Psychiatry, 48, 207 215 Kalinian, H. (2006). Executive function performance as assessed by the D KEFS in female psychopaths. Archives of Clinical Neuropsychology, 18, 760761. Kazdin, A. E. (2001). Almost clinically significant (p <.10): Current measures may only approach clinical significance. Clinical Psychology: Science and Practice, 8, 455 462. Kelley, K., & Lai, K. (201 1 ). MBESS (Version 3.2. 1 ) [Computer software and manual]. Retrieved from http://www.cran.rproject.org/ MBESS K endall, P. C., & Chansky, T. E. (1991). Considering cognition in anxiety disordered children. Journal of Anxiety Disorders, 5 167 185. Kessler, R. C. (1987). The interplay of research design strategies and data analysis procedures in evaluating the effect s of stress on health. In S. V. Kasl & C. L. Cooper (Eds.), Stress and health: Issues in research methodology (pp. 113140). New York: Wiley. Kessler, R. C., Berglund, P. A., Demler, O., Jin, R., & Walters, E. E. (2005). Lifetime prevalence and ageof onset distributions of DSM IV disorders in the national comorbidity survey r eplication. Archives of General Psychiatry, 62, 593602. Kim Cohen, J., Caspi, A., Moffit, T. E., Harrington, H., Milne, B. J., & Poulton, R. (2003). Prior juvenile diagnoses in adults with mental disorder. Developmental follow back of a prospectivelongitudinal cohort. Archives of General Psychiatry, 60, 709717.
116 Klassen, A. F., Miller, A., & Fine, S. (2004) Healthrelated quality of life in children and adolescents who have a diagnosis of attention deficit/hyperactivity disorder. Pediatrics 114, 541 547. Koran, L. M., Thienemann, M. L., & Davenport, R. (1996). Quality of life for patients with obsessive compulsive disorder. American Journal of Psychiatry, 153, 783 788. Landgraf, J. M., Abetz, L., & Ware, J. E. (1996). The CHQs users manual. Boston, Health Institute: New England Center. Latzman, R. D., Elkovitch, N., Young, J., & Clark, L. A. (2010). The contribution of executive functioning to academic achievement among male adol escents. Journal of Clinical and Experimental Neuropsychology, 32(5), 455462. Latzman, R. D., & Markon, K. E. (2010). The factor structure and agerelated factorial invariance of the Delis Kaplan Executive Function System (D KEFS). Assessment, 17(2), 172 184. Le Coq, E. M., Boeke, A. J. P., Bezemer, P. D., B ruil, J., & Van Eijk, J. T. M. (2000). Clinimetric properties of a parent report on their offsprings quality of life. Journal of Clinical Epidemiology, 53 (2), 139146. Lehto, J. E., Juujaervi, P., Kooistra, L., & Pulkkinen, L. (2003). Dimensions of executive functioning: Evidence from children. British Journal of Developmental Psychology, 21, 5980. Lezak, M. D., Howieson, D. B., & Loring, D. W. (2004). Neuropsychological assessment (4th ed.). New Yo rk, NY: Oxford University Press. MacKinnon, D. P. (2008). Introduction to statistical mediation analysis Mahwah, NJ: Erlbaum. MacKinnon, D. P., Fritz, M. S., Williams, J., & Lockwood, C. M. (2007). Distribution of the product confidence limits for the indirect effect: Program PRODCLIN. Behavioral Research Methods, 39(3), 384 389. MacKinnon, D. P., Lockwood, C. M., Hoffman, J. M., West, S. G., & Sheets, V. (2002). A comparison of methods to test mediation and other intervening variable effects. Psychologic al Methods, 7(1), 83104. MacKinnon, D. P., Lockwood, C. M., & Williams, J. (2004). Confidence limits for the indirect effect: Distribution of the product and the resampling methods. Multivariate Behavioral Research, 39, 99128. MacKinnon, D. P., & Luec ken, L. J. (2008). How and for whom? Mediation and m oderation in health psychology. Health Psychology, 27 (2), 99 100.
117 Mallinckrodt, B., Abraham, W. T., Wei, M., & Russell, D. W. (2006). Advances in testing the statistical sign ificance of mediation effects. Journal of Counseling Psychology, 53(3), 372378. March, J. S. (1997). Multidimensional A nxiet y S cale for C hildren. San Antonio, TX: The Psychological Corporation. Markowitz, J. S., Weissman, M. M., Ouellette, R., Lish, J. D., & Klerman, G. L. (1989). Qu ality of life in panic disorder. Archives of General Psychiatry, 46, 984 992 Massion, A. O., Warshaw, M. G., & Keller, M. B. (1993). Quality of life and psychiatric morbidity in panic disorder and generalized anxiety disorder. American Journal of Psychia try, 150 600 607. Matza, L. S., Swensen, A. R., Flood, A. M., Secnik, K., & Leidy, N. K. (2004). Assessment of healthrelated quality of life in children: A review of conceptual, methodological, and regulatory issues. Value in Health, 7(1), 7992. Maxwel l, S. E., & Cole, D. A. (2007). Bias in cross sectional analyses of longitudinal mediation. Psychological Methods 12, 2344. Mendlowicz, M. V., & Stein, M. B. (2000). Quality of life in individuals with anxiety disorders. American Journal of Psychiatry, 1 57(5), 669682. Micco, J. A., Henin, A., Biederman, J., Rosenbaum, J. F., Petty, C., Rindlaub, L. A. Hirshfeld Becker, D. R. (2009). Executive functioning in offspring at risk for depression and anxiety. Depression and Anxiety, 26, 780 790. Novo, A. A., & Schafer, J. (2011). NORM (Version 1.09.2) [Computer software and manual]. Retrieved from http://cran.rproject.org/web/packages/norm Olatunji, B. O., Cisler, J. M., & Tolin, D. F. (2007). Quality of life in the anxiety disorders: A meta analytic review. Clinical Psychology Review, 27 572581. Olley, A., Malhi, G., & Sachdev, P. (2007). Memory and executive functioning in obsessive compulsive disorder: A selective review. Journal of Affective Disorders, 104(1 3), 15 23. Piacentini, J., Bergman, R. L., Keller, M., & McCracken, J. (2003). Functional impairment in children and adolescents with obsessivecompulsive disorder. Journal of Child and Adolescent Psychopharmacology, 13(1) 6169. Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects in simple mediation models Behavior Research Methods, Instruments, and Computers, 36, 717731.
118 Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling procedures for assessing and comparing indirect effects in multiple mediator models. Behavior Research Methods, 40, 879891. Preacher, K. J., & Kelley, K. (2011). Effect size measures for mediation models: Quantitative strategies for communicating indirect effects. Psychological Methods, 16(2), 93 115. Quilty, L. C., Van Ameringen, M., Mancini, C., Oakman, J., & Farvolden, P. (2003). Quality of life and the anxiety disorders. Journal of Anxiety Disorders, 17, 405426. R Development Core Team (2011 ). R: A language and environment for statistical computing. Vienna: R Foundation for Statistical Computing. ISBN3900051070, URL http://www/Rproject.org. Ravens Sieberer, U., Erhart, M., Wille, N., Wetzel, R., Nickel, J. & Bullinger, M. (2006). Generic healthrelated quality of life assessment in children and adolescents: Methodological considerations. Pharmacoeconomics, 24(12), 11991220. Rebok, G., Riley, A., Forrest, C., Starfield, B., Green, B., Robertson, J., & Tambor, E. (2001). Elementary school aged childrens reports of their health: A cognitive interviewing study. Quality of Life Research, 10, 5970. Reitan, R. M., & Wolfson, D. (1994). A selective and critical review of neuropsychological deficits and the frontal lobes. Neuropsychology Review, 4 161 197. Reynolds, C.R., & Kamphaus, R.W. (2004). BASC: Behavior Ass essment System for C hildren (2nd ed.). Circle Pines, MN: American Guidance Service. Reynolds, C. R., & Richmond, B. O. (2008). Revised Childrens Manifest Anxiety S cale (2nd ed.). USA: Western Psychological Services. Rizzo, M. L., & Szekely, G. J. (2011). ENERGY (Version 1.40) [Computer software and manual]. Retrieved from http://cran.rproject.org/web/packages/ energy Rubin, D. B. (1987). Multiple imputation for nonresponse in surveys. New York, NY: John Wiley. Sadock B. J., Sadock, V. A., & Ruiz, P. (Eds.). (2009). Kaplan and Sadocks comprehensive textbook of psychiat ry (9th ed.). Philadelphia, PA: Lippencott, Williams, & Wilkens. Saintfort, F., Becker, M., & Diamond, R. (1996). Judgments of quality of life of indi viduals with severe mental disorders: Patient self report versus provider perspectives. American Journal of Psychiatry, 153, 497502.
119 Salthouse, T. A., Atkinson, T. M., & Berish, D. E. (2003). Executive functioning as a potential mediator of agerelated cognitive decline in normal adults. Journal of Experimental Psychology, 132(4), 566594. Sarason, I. G. (1988). Anxiety, selfpreoccupation and attention. Anxiety Research, 1, 3 7. Sarsour, K., Sheridan, M., Jutte, D., Nuru Jeter, A., Hinshaw, S., & Boyce W. T. (2011). Family socioeconomic status and child executive functions: The roles of language, home environment, and single parenthood. Journal of the International Neuropsychological Society, 17, 120132. Sawyer, M. G., Whaites, L., Rey, J. M., Hazell, P. L., Graetz, B. W., & Baghurst, P. (2002). Health related quality of life of children and adolescents with mental disorders. Journal of the American Academy of Child and Adolescent Psychiatry, 41, 530 537. Schlomer, G. L., Bauman, S., & Card, N. A. (20 10). Best practices for missing data management in counseling psychology. Journal of Counseling Psychology, 57(1) 1 10. Seeyave, D. M., Coleman, S., Appugliese, D., Corwyn, R. F., Bradley, R. H., Davidson, N. S., Kaciroti, N., & Lumeng, J. C. (2009). Abi lity to delay gratification at age 4 years and risk of overweight at age 11 years. Archives of Pediatric and Adolescent Medicine, 163(4), 303308. Shaffer, D., Fisher, P., Dulcan, M. K., Davies, M., Piacentini, J., Schwab Stone, M. E., Reqier, D. A. (1996). The NIMH diagnostic interview schedule for children version 2.3 (DISC 2.3): Description, acceptability, prevalence rates, and performance in the MECA Study. Methods for the epidemiology of child and adol escent mental disorders study. Journal of the American Academy of Child and Adolescent Psychiatry, 35, 865 77. Shallice, T., & Burgess, P. W. (1991). Deficits in strategy application follow ing frontal lobe damage in man. Brain, 114, 727 741. Sherbourne, C. D., Wells, K. B., & Judd, L. L. (1996). Func tioning and well being of patients with Panic Disorder. American Journal of Psychiatry, 153, 213 218. Shunk, A.W., Davis, A.S., & Dean, R.S. (2006). Delis Kaplan executive function system (D KEFS) (Test Review). Applied Neuropsychology, 13, 275279. Smithe rman, T. A., Huerkamp, J. K., Miller, B. I., Houle, T. T., & OJile, J. R. (2007). The relations of depression and anxiety to measures of executive functioning in a mixed psychiatric sample. Archives of Clinical Neuropsychology, 22, 647654.
120 Sobel, M. E. (1982). Asymptotic confidence intervals for indirect effects in structural equation models. In S. Leinhardt (Ed.), Sociological methodology (pp. 290312). Washington, DC: American Sociological Association. Sobel, M. E. (1986). Some new results on indirect effects and their standard errors in covariance structure models. Sociological Methodology, 13, 290312. Spitzer, W. O., Dobson, A.J., & Hall, J. (1981). Measuring the quality of life of cancer patients: A concise QLindex for use by physicians. Journal of Chronic Diseases, 34, 585 597. Stein, M. B., & Kean, Y. M. (2000). Disability and quality of life in social phobia: Epidemiologic findings. American Journal of Psychiatry, 157, 16061613. Stein, M. B., Walker, J. R., Hazen, A. L., & Forde, D. R. (1997). Full and partial posttraumatic stress disorder: Findings from a community survey. American Journal of Psychiatry, 154, 1114 1119 Stewart, S.E., Geller, D.A., Jenike, M., Pauls, D., Shaw, D., Mullen, B., & Faraone, F. B. (2004). Longterm outcome of pediatr ic obsessive compulsive disorder: A meta analysis and quantitative review of the literature. Acta Psychiatrica Scandanavia, 110, 4 13. Strauss, E., Sherman, E. M. S., Spreen, O. (2006). A compendium of neuropsychological tests (3rd e d.). New York: Oxford University Press. Stroop, J. R. (1935). Studies of interference in serial verbal reactions. Journal of Experimental Psychology, 18, 643662. Szasz, T. S. (1960). The myth of mental illness. American Psychologist, 15(2), 113118. Tan, C. S. (2007). Test review: Behavior Assessment System for C hildren (2nd ed.). Assessment for Effective Intervention, 32, 121 124. Tannock, R. (2000). Attentiondeficit/hyperactivity disorder with anxiety disorders. In T. E. Brown (Ed.), Attentiondeficit disorders and comorbidities in children, adolescents, and adults (pp. 125170). Washington, DC: American Psychiatric Press. Thomsen, P.H., & Mikkelsen, H.U. (1995). Course of obsessivecompulsive disorder in children and adolescents: A prospective follow up study of 23 Danis h cases. Journal of the American Academy of Children and Adolescent Psychiatry, 34, 14321440. Toren, P., Sadeh, M., Wolmer, L., Eldar, S., Koren, S., Weizman, R., & Laor, N. (2000). Neurocognitive correlates of anxiety disorders in children: A preliminary report. Journal of Anxiety Disorders, 14(3), 239247.
121 UCLA: Academic Technology Services, Statistical Consulting Group. (2011). Reg ression with SPSS: Chapter 2 R egression diagnostics. Retrieved from h ttp://www.ats.ucla.edu/stat/spss/webbooks/reg/chapter 2/spssreg2.htm Utsey, S. O., Chae, M. H., Brown, C. F., & Kelly, D. (2002). Effect of ethnic group membership on ethnic identity, racerelated stress, and quality of life. Cultural Diversity and Ethnic Minority Psychology, 8 (4), 366377. Varni, J. W., Burwinkle, T. M., Seid, M., & Skarr, D. (2003). The PedsQL 4.0 as a pediatric population health measure: Feasibility, reliability, and validity. Ambulatory Pediatrics, 3, 329341. Varni, J. W., Seid, M., & Kurtin, P. S. (2001). PedsQL 4.0: Reliability and val idity of the Pediatric Quality of Life Inventory version 4.0 generic core scales in healthy and patient populations Medical Care, 39, 800812. Vriezen, E. R., & Pigott, S. E. (2002). The relationship between parental report on the BRIEF and performancebased measures of executive function in children with moderate to severe traumatic brain injury. Child Neuropsychology, 8(4), 296303. Vr i e zen, E. R., Pigott, S. E., & Pelletier, P. M. (2001). Developmental implications of early frontal lobe damage: A case study. Brian and Cognition, 47, 222225. Wang, G. J., Volkow, N. D., Logan, J., Pappas, N. R., Wong, C. T., Zhu, W., Fowler, J. S. (2001). Brain dopamine and obesity. Lancet, 357(9253), 354357. Willcutt, E. G., Doyle, A. E., Nigg, J. T., Faraone, S V., & Pennington, B. F. (2005).Validity of the executive function theory of attentiondeficit/hyperactivity disorder: A metaanalytic review. Biological Psychiatry, 57, 13361346. Woody, E. (2011). An SEM perspective on evaluating mediation: What every clinical researcher needs to know. Journal of Experimental Psychopathology, 2, 210251. World Health Organization. (1948). Constitution of the World Health Organization: Basic document Geneva, Switzerland: World Health Organization. Zanna, M. P., & Fazi o, R. H. (1982). The attitudebehavior relation: Moving toward a third generation of research. In M. P. Zanna, E. T. Higgins, & C. P. Herman (Eds.), Consistency in social behavior: The Ontario Symposium ( Vol. 2 ) ( pp. 283301). Hillsdale, NJ: Erlbaum. Zelazo, P. D., Muller, U., Frye, D., & Marcovitch, S. (2003). The development of executive function in early childhood. Monographs of the Society for Research in Child Development, 68 (3), 93119.
122 BIOGRAPHICAL SKETCH Born and raised in New Jersey, Robert Merr ell spent his sophomore year in Andaluca, Spain, at the Universidad de Sevilla. A scholarship later allowed him to study in South America at the Universidad de Concepcin in southern Chile. A Phi Beta Kappa member, he earned his Bachelor of Arts ( summa cum laude) and Master of Arts in Spanish Literature from Villanova Universi ty outside of Philadelphia, Pennsylvania. Influenced by his minor related studies in philosophy and passion for helping others move toward optimizing personal growth, he began studying psychology and received his Master of Science degree from t he University of Florida. After completing his doctoral studies in December 2011, he plans to remain in his current position within the University of Floridas Department of Psychiatry H is interests range from clinical supervision of psychiatry residents to the provision of exposure and response prevention treatment for ObsessiveCompulsive Disorder for children and adults. Future career plans may involve further specialization in the treatment of anxiety and mood disorders.