UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

New insights on psychosocial adjustment to pediatric cancer in caregivers Moseley-Giannelli, Janine V. 2011

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

Item Metadata

Download

Media
24-ubc_2011_fall_moseleygiannelli_janine.pdf [ 3.62MB ]
Metadata
JSON: 24-1.0071607.json
JSON-LD: 24-1.0071607-ld.json
RDF/XML (Pretty): 24-1.0071607-rdf.xml
RDF/JSON: 24-1.0071607-rdf.json
Turtle: 24-1.0071607-turtle.txt
N-Triples: 24-1.0071607-rdf-ntriples.txt
Original Record: 24-1.0071607-source.json
Full Text
24-1.0071607-fulltext.txt
Citation
24-1.0071607.ris

Full Text

NEW INSIGHTS ON PSYCHOSOCIAL ADJUSTMENT TO PEDIATRIC CANCER IN CAREGIVERS by Janine V. Moseley-Giannelli M.A., University of British Columbia, 2005 B.Sc., University of British Columbia, 2000  A DISSERTATION SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in The Faculty of Graduate Studies  (Psychology)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  January 2011  Janine V. Moseley-Giannelli, 2011  ii Abstract PURPOSE - While survival rates for pediatric cancer have dramatically improved, survivorship comes with its own stressors for this population and (especially) their caregivers, the most salient of which may be uncertainty. Answering a strong demand for modelling research, this study assessed the impact of numerous risk and resilience factors on caregiver emotional adjustment in the context of Mishel‟s „uncertainty in illness‟ (MUIT) framework. Of particular interest was examining the role of positive psychology variables. METHODS - One-hundred fifty-six caregivers of pediatric cancer patients/survivors were recruited. The sample was evenly divided across „on‟/„off‟ treatment groups and „urban‟/„rural‟ groups. Forty adolescent patients/survivors also participated, and 41 on-treatment caregivers were re-assessed longitudinally on some baseline questionnaires. Questionnaires sampled sociodemographic and cancer-specific variables, positive and negative thinking, school and family needs, and emotional adjustment. RESULTS - SEM analyses identified a version of the MUIT applicable to these caregivers in which hope, optimism, and benefit finding figured prominently. Hope was a strong predictor of positive and negative outcomes, across baseline and six months‟ follow-up, and it (along with perceived mastery and threat) was also a significant mediating variable. In descriptive analyses, school and family needs (especially medically-related) were considered important but often unmet. Surprisingly, proportion of met needs did not vary across urban/rural subgroups, and having needs met (broadly speaking) did not significantly predict any outcome variable. Adolescents did not report any areas of struggle. While caregivers of newly diagnosed patients reported elevated distress, this subsided regardless of treatment status or urban/rural status. Frequent levels of benefit finding were reported by caregivers and teens, and benefit finding significantly predicted hope. Explained variance for most caregiver outcome variables averaged 60%, with the exception of benefit finding (33%). DISCUSSION - This is the first study to explore a pediatric cancer caregiving model of emotional adjustment incorporating the roles of uncertainty, hope and benefit finding. The significant role of positive psychology variables here is suspected to reflect the „buffering capacity‟ of frequent positive affect. With this improved understanding of caregiver emotional adjustment, we are in a better position to design screening and intervention efforts for this population.  iii Preface UBC RESEARCH ETHICS BOARD: ▪ Behavioural Research Ethics Board CERTIFICATE NUMBERS: ▪ November 25th, 2005 – Original Certificate of Approval: B05 – 0732 ▪ January 31st, 2006 –Certificate of Approval: W05-0169 ▪ March 10th, 2007 - Minimal Risk Renewal: H05 – 80732 ▪ July 29th, 2008 - Minimal Risk Renewal: H05 – 80732 ▪ November 25th, 2008 – Minimal Risk Amendment: H05 – 80732 ▪ July 20th, 2009 - Minimal Risk Renewal: H05 – 80732 ▪ August 3rd, 2010 - Minimal Risk Renewal: H05 – 80732  iv  Table of Contents ABSTRACT .............................................................................................................................................................. ii PREFACE ............................................................................................................................................................... iii TABLE OF CONTENTS ............................................................................................................................................. iv LIST OF TABLES .................................................................................................................................................... vii LIST OF FIGURES................................................................................................................................................... vii ACKNOWLEDGEMENTS ......................................................................................................................................... iii INTRODUCTION ..................................................................................................................................................... 1 THE CANCER EXPERIENCE FOR CHILDREN............................................................................................................................ 1 THE CANCER EXPERIENCE FOR CAREGIVERS ........................................................................................................................ 3 CURRENT ESTIMATES OF DISTRESS IN PATIENTS AND CAREGIVERS ........................................................................................... 4 MODERATOR VARIABLES ASSOCIATED WITH CAREGIVER EMOTIONAL MALADJUSTMENT ............................................................. 5 Socio-Demographic Characteristics ..................................................................................................................... 6 Family Stress ........................................................................................................................................................ 7 Social Support ...................................................................................................................................................... 8 Communication with Healthcare Professionals ................................................................................................... 9 Coping Skills ....................................................................................................................................................... 10 Child Characteristics .......................................................................................................................................... 11 UNCERTAINTY AND THE ‘UNCERTAINTY IN ILLNESS THEORY’ ................................................................................................. 13 Research Related to the MUIT in Adult Patient Populations ............................................................................. 15 Research Related to the MUIT and Psychological Adjustment in Caregiver Populations .................................. 16 FIELD OF POSITIVE PSYCHOLOGY ..................................................................................................................................... 18 Positive Psychology Variables Relevant to Models of Adjustment and Uncertainty.......................................... 20 Benefit Finding ................................................................................................................................................... 20 Research on benefit finding in adults. ............................................................................................................................ 21 Research on benefit finding in children. ......................................................................................................................... 24 Hope................................................................................................................................................................... 25 Research on hope. .......................................................................................................................................................... 25 PRESENT STATE OF THE FIELD......................................................................................................................................... 27 CURRENT STUDY ......................................................................................................................................................... 28 Hypothesized Model .......................................................................................................................................... 30 Secondary Analyses ........................................................................................................................................... 32 Adolescent patients/survivors. ....................................................................................................................................... 32 Caregivers measured longitudinally................................................................................................................................ 32 Objectives of Study ............................................................................................................................................ 33 METHODS ............................................................................................................................................................ 34 RESEARCH DESIGN ....................................................................................................................................................... 34 INCLUSION AND EXCLUSION CRITERIA .............................................................................................................................. 36 MEASURES ................................................................................................................................................................. 37 Measures Pertaining to Variables in the MUIT .................................................................................................. 37 Mastery scale (MS). ........................................................................................................................................................ 37 Center for epidemiological studies depression scale (CES-D). ........................................................................................ 37 Cognitive appraisal of health scale – revised (CAHS-R). .................................................................................................. 38 Parent experience of child illness (PECI). ........................................................................................................................ 39 Positive Psychology Measures ........................................................................................................................... 40 Benefit finding scale (BFS). ............................................................................................................................................. 40 Benefit finding scale for children (BFSC). ........................................................................................................................ 40 Life orientation test - revised (LOT-R). ............................................................................................................................ 41  v Herth hope index (HHI). .................................................................................................................................................. 41 Subjective happiness scale (SHS). ................................................................................................................................... 42  Measures Pertaining to Family and Child-Specific Needs .................................................................................. 42 Family needs questionnaire (FNQ).................................................................................................................................. 43 School needs/resources questionnaire (SNRQ). ............................................................................................................. 44 Other Measures ................................................................................................................................................. 44 Pediatric quality of life inventory (PedsQL). ................................................................................................................... 44 Socio-demographic form. ............................................................................................................................................... 45 PROCEDURE ............................................................................................................................................................... 46 RESULTS ............................................................................................................................................................... 48 DATA SETS ................................................................................................................................................................. 48 PRELIMINARY DATA SCREENING ..................................................................................................................................... 48 PARTICIPANTS ............................................................................................................................................................. 50 Baseline Sample ................................................................................................................................................. 50 Caregivers. ...................................................................................................................................................................... 50 Pediatric patients/survivors. ........................................................................................................................................... 51 Adolescent Sample............................................................................................................................................. 52 Longitudinal Sample .......................................................................................................................................... 52 Caregivers. ...................................................................................................................................................................... 52 Pediatric patients/survivors. ........................................................................................................................................... 53 SOCIO-DEMOGRAPHIC DATA: DESCRIPTIVE STATISTICS AND GROUP COMPARISONS ................................................................. 53 Baseline Sample ................................................................................................................................................. 54 Adolescent Sample............................................................................................................................................. 54 QUESTIONNAIRE DATA: DESCRIPTIVE STATISTICS AND GROUP COMPARISONS ......................................................................... 54 Baseline (Caregiver) Sample .............................................................................................................................. 55 Adolescent Sample............................................................................................................................................. 58 Longitudinal (Caregiver) Sample ........................................................................................................................ 59 SAMPLE COMBINING .................................................................................................................................................... 60 CORRELATION ANALYSES............................................................................................................................................... 60 Variables to be Removed – Precursor Variable Data ......................................................................................... 61 Variables to be Removed – Questionnaire Data ................................................................................................ 61 Baseline (caregiver) dataset. ........................................................................................................................................... 61 Adolescent dataset. ........................................................................................................................................................ 62 Longitudinal (caregiver) dataset. .................................................................................................................................... 63 Inter-Correlations among Questionnaire Predictor Variables ........................................................................... 63 Baseline (caregiver) dataset. ........................................................................................................................................... 64 Adolescent dataset. ........................................................................................................................................................ 64 Longitudinal (caregiver) dataset. .................................................................................................................................... 64 Predictor-Outcome Variable Bivariate Correlations .......................................................................................... 64 Baseline dataset. ............................................................................................................................................................. 64 Adolescent dataset. ........................................................................................................................................................ 66 Longitudinal dataset. ...................................................................................................................................................... 66 Correlations between Corresponding Questionnaire Variables in Different Datasets ....................................... 67 Baseline (caregiver) - adolescent correlations. ............................................................................................................... 67 Baseline (caregiver) – longitudinal (caregiver) correlations............................................................................................ 68 REGRESSION ANALYSES................................................................................................................................................. 68 Baseline Dataset: ............................................................................................................................................... 69 Benefit finding predictions. ............................................................................................................................................ 69 Hope predictions. ........................................................................................................................................................... 71 Subjective happiness predictions.................................................................................................................................... 72 Distress predictions. ....................................................................................................................................................... 74 Mastery predictions. ....................................................................................................................................................... 75 Adolescent Dataset: ........................................................................................................................................... 76 Benefit finding predictions. ............................................................................................................................................ 76 Quality of life predictions. .............................................................................................................................................. 77  vi Longitudinal Dataset: ........................................................................................................................................ 77 Longitudinal distress predictions. ................................................................................................................................... 77 Longitudinal subjective happiness predictions. .............................................................................................................. 78 STRUCTURAL EQUATION MODELING ANALYSES ................................................................................................................. 80 Distress Model ................................................................................................................................................... 81 Subjective Happiness Model .............................................................................................................................. 83 Final Note........................................................................................................................................................... 85 DISCUSSION AND CONCLUSIONS ......................................................................................................................... 87 STRENGTHS OF STUDY .................................................................................................................................................. 87 MAIN FINDINGS .......................................................................................................................................................... 89 Caregiver Distress .............................................................................................................................................. 89 Caregiver Subjective Happiness ......................................................................................................................... 93 Adolescent Quality of Life .................................................................................................................................. 95 Important Needs (Caregivers and Adolescents)................................................................................................. 96 Benefit Finding (Caregivers and Adolescents).................................................................................................... 99 Caregiver Hope ................................................................................................................................................ 102 STUDY LIMITATIONS ................................................................................................................................................... 103 VALUE OF STUDY ....................................................................................................................................................... 105 Screening ......................................................................................................................................................... 106 Intervention ..................................................................................................................................................... 107 Service Delivery ................................................................................................................................................ 110 APPLICATION TO OTHER PEDIATRIC ILLNESSES ................................................................................................................. 111 FUTURE DIRECTIONS .................................................................................................................................................. 112 FOOTNOTES ....................................................................................................................................................... 117 TABLES ............................................................................................................................................................... 118 FIGURES ............................................................................................................................................................. 243 REFERENCES ...................................................................................................................................................... 263 APPENDICES ...................................................................................................................................................... 282 APPENDIX A. SAMPLES OF FORMS USED IN STUDY. ....................................................................................................... 283 Appendix A1. Socio-Demographic Form. ......................................................................................................... 284 Appendix A2. – Recruitment Brochure ............................................................................................................. 285 Appendix A3. Recruitment Letter ..................................................................................................................... 287 Appendix A4. – Verbal Consent Script .............................................................................................................. 289 Appendix A5. – Caregiver Consent Form .......................................................................................................... 290 Appendix A6. – Adolescent Assent Form .......................................................................................................... 284 Appendix A7. – Caregiver Baseline Package Cover Sheet ................................................................................ 298 Appendix A8. – Adolescent Package Cover Sheet ............................................................................................ 299 Appendix A9. – Caregiver Longitudinal Package Cover Sheet ......................................................................... 300  vii List of Tables METHODS TABLES TABLE 1 – LIST OF PRECURSOR VARIABLES .......................................................................................................................... 119 TABLE 2 - LIST OF MEASURES ACROSS DEPENDENT AND INDEPENDENT VARIABLES .................................................................... 120 SOCIO-DEMOGRAPHIC DATA: DESCRIPTIVE STATISTICS TABLE 3 – DESCRIPTIVE STATISTICS ON CAREGIVER SOCIO-DEMOGRAPHIC AND CANCER-SPECIFIC DATA, BASELINE SAMPLE ............... 121 TABLE 4 – DESCRIPTIVE STATISTICS ON PEDIATRIC PATIENT/SURVIVOR SOCIO-DEMOGRAPHIC AND CANCER-SPECIFIC DATA, BASELINE SAMPLE .............................................................................................................................................................. 125 TABLE 5 – DESCRIPTIVE STATISTICS ON PEDIATRIC PATIENT/SURVIVOR SOCIO-DEMOGRAPHIC AND CANCER-SPECIFIC DATA, ADOLESCENT SAMPLE .............................................................................................................................................................. 129 TABLE 6 – DESCRIPTIVE STATISTICS ON CAREGIVER SOCIO-DEMOGRAPHIC AND CANCER-SPECIFIC DATA, LONGITUDINAL SAMPLE ....... 133 TABLE 7 – DESCRIPTIVE STATISTICS ON PEDIATRIC PATIENT/SURVIVOR SOCIO-DEMOGRAPHIC AND CANCER-SPECIFIC DATA, LONGITUDINAL SAMPLE ......................................................................................................................................... 136 SOCIO-DEMOGRAPHIC DATA: CHI-SQUARE STATISTICS TABLE 8 – SIGNIFICANT CHI-SQUARE TESTS OF INDEPENDENCE ON SOCIO-DEMOGRAPHIC AND CANCER-SPECIFIC DATA, BASELINE SAMPLE ......................................................................................................................................................................... 139 TABLE 9 – SIGNIFICANT CHI-SQUARE TESTS OF INDEPENDENCE ON SOCIO-DEMOGRAPHIC AND CANCER-SPECIFIC DATA, ADOLESCENT SAMPLE .............................................................................................................................................................. 141 QUESTIONNAIRE DATA: DESCRIPTIVE STATISTICS AND GROUP COMPARISONS TABLE 10 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE BFS ......................................................................... 142 TABLE 11 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE CAHS-R ................................................................... 144 TABLE 12 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE CES-D ...................................................................... 145 TABLE 13 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE CAREGIVER FNQ ........................................................ 146 TABLE 14 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE HHI ......................................................................... 150 TABLE 15 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE LOT-R ...................................................................... 151 TABLE 16 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE MS .......................................................................... 152 TABLE 17 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE PECI ........................................................................ 153 TABLE 18 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE CAREGIVER PEDSQL .................................................... 155 TABLE 19 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE SHS ......................................................................... 157 TABLE 20 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE SNRQ ...................................................................... 158 TABLE 21 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE BFSC ....................................................................... 160 TABLE 22 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE TEEN FNQ ................................................................ 161 TABLE 23 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE TEEN PEDSQL ............................................................ 166 TABLE 24 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE LONGITUDINAL CES-D ................................................. 168 TABLE 25 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE LONGITUDINAL MS ..................................................... 169 TABLE 26 – DESCRIPTIVE STATISTICS AND GROUP COMPARISONS FOR THE LONGITUDINAL SHS .................................................... 170 QUOTES TABLE 27 – ADDITIONAL BENEFITS REALIZED DURING THE CANCER EXPERIENCE BY CAREGIVERS ................................................... 171 TABLE 28 – ADDITIONAL BENEFITS REALIZED DURING THE CANCER EXPERIENCE BY ADOLESCENT PATIENTS/SURVIVORS .................... 172 CORRELATION ANALYSES TABLE 29 – SOCIO-DEMOGRAPHIC AND CANCER-SPECIFIC VARIABLES (I.E., PRECURSOR VARIABLES) REMOVED FROM FURTHER INFERENTIAL ANALYSES .......................................................................................................................................... 173 TABLE 30 – INTER-CORRELATION SCORES OF CONCERN IN THE BASELINE (CAREGIVER) SAMPLE QUESTIONNAIRE DATA .................... 174 TABLE 31 – INTER-CORRELATION SCORES OF CONCERN IN THE ADOLESCENT SAMPLE QUESTIONNAIRE DATA .................................. 178 TABLE 32 – INTER-CORRELATION SCORES OF CONCERN IN THE LONGITUDINAL (CAREGIVER) SAMPLE QUESTIONNAIRE DATA ............. 181 TABLE 33 – CORRELATIONS AMONG REMAINING QUESTIONNAIRE PREDICTOR VARIABLES IN THE BASELINE (CAREGIVER) SAMPLE ....... 182 TABLE 34 – CORRELATIONS AMONG REMAINING QUESTIONNAIRE PREDICTOR VARIABLES IN THE ADOLESCENT SAMPLE .................... 185 TABLE 35 - CORRELATIONS AMONG REMAINING QUESTIONNAIRE PREDICTOR VARIABLES IN THE LONG. (CAREGIVER) SAMPLE ........... 186 TABLE 36 – CORRELATIONS BETWEEN PRECURSOR VARIABLES AND OUTCOME VARIABLES IN THE BASELINE SAMPLE ......................... 189  viii TABLE 37 - CORRELATIONS BETWEEN INDEPENDENT VARIABLES AND OUTCOME VARIABLES IN THE ADOLESCENT SAMPLE ................. 190 TABLE 38 – CORRELATIONS BETWEEN PRECURSOR VARIABLES AND OUTCOME VARIABLES IN THE LONGITUDINAL SAMPLE ................. 191 TABLE 39 – CORRELATIONS AMONG BASELINE (CAREGIVER) AND ADOLESCENT CORRESPONDING QUESTIONNAIRE DATA .................. 193 TABLE 40 – CORRELATIONS AMONG BASELINE (CAREGIVER) AND LONGITUDINAL (CAREGIVER) CORRESPONDING QUESTIONNAIRE DATA ......................................................................................................................................................................... 195 REGRESSION ANALYSES (NON-TRANSFORMED DATA) TABLE 41 – HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE BASELINE SAMPLE – PREDICTING BENEFIT FINDING (WITH NONTRANSFORMED DATA) ........................................................................................................................................... 196 TABLE 42 – HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE BASELINE SAMPLE – PREDICTING HOPE (WITH NON-TRANSFORMED DATA) ................................................................................................................................................................ 202 TABLE 43 – HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE BASELINE SAMPLE – PREDICTING SUBJECTIVE HAPPINESS (WITH NONTRANSFORMED DATA) ........................................................................................................................................... 205 TABLE 44 - HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE BASELINE SAMPLE – PREDICTING DISTRESS (WITH NON-TRANSFORMED DATA) ................................................................................................................................................................ 207 TABLE 45 – EXPLORATORY REGRESSION ANALYSES IN THE BASELINE SAMPLE – PREDICTING MASTERY (WITH NON-TRANSFORMED DATA) ......................................................................................................................................................................... 210 TABLE 46 – HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE LONGITUDINAL SAMPLE – PREDICTING LONGITUDINAL DISTRESS (WITH NON-TRANSFORMED DATA) ................................................................................................................................... 212 TABLE 47 - HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE LONGITUDINAL SAMPLE – PREDICTING LONGITUDINAL SUBJECTIVE HAPPINESS (WITH NON-TRANSFORMED DATA) .......................................................................................................... 214 REGRESSION ANALYSES (TRANSFORMED DATA) TABLE 48 – HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE BASELINE SAMPLE – PREDICTING BENEFIT FINDING (WITH TRANSFORMED DATA) ........................................................................................................................................... 216 TABLE 49 – HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE BASELINE SAMPLE – PREDICTING HOPE (WITH TRANSFORMED DATA) ......................................................................................................................................................................... 221 TABLE 50 – HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE BASELINE SAMPLE – PREDICTING SUBJECTIVE HAPPINESS (WITH TRANSFORMED DATA) ........................................................................................................................................... 226 TABLE 51 – HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE BASELINE SAMPLE – PREDICTING DISTRESS (WITH TRANSFORMED DATA) ......................................................................................................................................................................... 229 TABLE 52 – EXPLORATORY REGRESSION ANALYSES IN THE BASELINE SAMPLE – PREDICTING MASTERY (WITH TRANSFORMED DATA) ... 232 TABLE 53 – HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE ADOLESCENT SAMPLE – PREDICTING BENEFIT FINDING (WITH TRANSFORMED DATA) ........................................................................................................................................... 234 TABLE 54 - HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE ADOLESCENT SAMPLE – PREDICTING QUALITY OF LIFE (WITH TRANSFORMED DATA) ........................................................................................................................................... 236 TABLE 55 – HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE LONGITUDINAL SAMPLE – PREDICTING LONGITUDINAL DISTRESS (WITH TRANSFORMED DATA) ........................................................................................................................................... 238 TABLE 56 – HIERARCHICAL LINEAR REGRESSION ANALYSES IN THE LONGITUDINAL SAMPLE – PREDICTING LONGITUDINAL SUBJECTIVE HAPPINESS (WITH TRANSFORMED DATA) .................................................................................................................. 241  ix List of Figures THEORETICAL MODELS FIGURE 1 - MISHEL’S UNCERTAINTY IN ILLNESS THEORY......................................................................................................... 244 FIGURE 2 - PROPOSED CAREGIVER MODEL OF ADJUSTMENT WITH DISTRESS AS THE OUTCOME VARIABLE ...................................... 245 FIGURE 3 – PROPOSED DISTRESS MODEL IN SEM FORMAT .................................................................................................... 246 FIGURE 4 – PROPOSED CAREGIVER MODEL OF ADJUSTMENT WITH SUBJECTIVE HAPPINESS AS THE OUTCOME VARIABLE.................... 247 FIGURE 5 – PROPOSED SUBJECTIVE HAPPINESS MODEL IN SEM FORMAT ................................................................................. 248 DISTRESS SEM MODELS FIGURE 6 – DISTRESS MODEL #1 ...................................................................................................................................... 249 FIGURE 7 – DISTRESS MODEL #2 ...................................................................................................................................... 250 FIGURE 8 – DISTRESS MODEL # 3 ...................................................................................................................................... 251 FIGURE 9 – DISTRESS MODEL #4 ...................................................................................................................................... 252 FIGURE 10 – DISTRESS MODEL #5 .................................................................................................................................... 253 FIGURE 11 – DISTRESS MODEL # 6.................................................................................................................................... 254 FIGURE 12 – DISTRESS MODEL #7: FINAL MODEL................................................................................................................ 255 FIGURE 13 – FINAL DISTRESS MODEL: (STANDARDIZED) DIRECT EFFECTS ................................................................................. 256 FIGURE 14 – FINAL DISTRESS MODEL: SIGNIFICANT (STANDARDIZED) INDIRECT EFFECTS, HIGHLIGHTING MEDIATING FUNCTIONS OF MASTERY, THREAT, HOPE, AND CHILD EMOTIOANL STRUGGLES ..................................................................................... 257 SUBJECTIVE HAPPINESS SEM MODELS FIGURE 15 – SUBJECTIVE HAPPINESS MODEL #1 .................................................................................................................. 258 FIGURE 16 – SUBJECTIVE HAPPINESS MODEL #2 .................................................................................................................. 259 FIGURE 17 – SUBJECTIVE HAPPINESS MODEL #3: FINAL MODEL ............................................................................................. 260 FIGURE 18 – FINAL SUBJECTIVE HAPPINESS MODEL: (STANDARDIZED) DIRECT EFFECTS ............................................................... 261 FIGURE 19 – FINAL SUBJECTIVE HAPPINESS MODEL – SIGNIFICANT (STANDARDIZED) INDIRECT EFFECTS, HIGHLIGHTING MEDIATING FUNCTIONS OF MASTERY, THREAT, AND HOPE ........................................................................................................... 262  x Acknowledgements There are several people I would like to thank who have played an integral part in the production of this dissertation. First, I would like to thank my graduate supervisor, Wolfgang Linden, for the guidance, feedback, and support he has consistently provided. Wolfgang, your expertise and lightheartedness has always fostered an enjoyable and enriching working environment. I would also like to thank my research colleagues at BC Children‟s Hospital, Joanna Chung and Dina McConnell, for allowing me to carry out this research at the hospital and for helping me in this endeavour. Dina, I am particularly grateful for the mentoring you have provided throughout this process. I would not have completed this research in a timely manner without the excellent help of my research assistants: Melanie Phillips, Christopher Siu, and Helena Friesen. Helena, I would especially like to thank you for the long hours you spent recruiting our participants! I must also thank those funding agencies that enabled this research to run: The Canadian Association for Research in Pediatric Psycho-oncology, The Provincial Pediatric Oncology/ Hematology Network, and The Social Sciences and Humanities Research Council of Canada. To my dissertation committee, thank you for making time to absorb and critique this material. Susan Birch and Elizabeth Dunn, I am particularly thankful for your thoughtful designrelated suggestions during the early phase of this research. Lastly, thank you to my friends, family, and especially my husband, Mike, for motivating me, inspiring me, supporting me, and „understanding‟ the dedication required by this timeconsuming commitment. I am very grateful to have such loving people in my life.  1 Introduction The Cancer Experience for Children According to statistics from the Canadian Cancer Society (2008), the top three most common classes of childhood cancer are leukemia (26% of new cases), lymphomas (18% of new cases), and cancer of the central nervous system (16% of new cases). In terms of severity, 5-year survival estimates average 80% (Landier & Bhatia, 2008). They are highest for retinoblastoma (99%), renal tumors (91%), and malignant epithelial neoplasms and melanomas (91%), while they are poorest for hepatic tumors (68%) and malignant bone tumors (68%). Treatment for pediatric cancers tends to be multimodal, including some combination of chemotherapy, radiation therapy, surgery, and bone marrow transplant. Children with cancer face innumerable stressors secondary to their disease and its treatment. Fatigue, vomiting, low energy, and pain are all trying aspects of the acute cancer experience, as is restricted mobility and decreased pleasure from food (Eisenberg, Sutkin, & Jansen, 1984; Wu, Hsu, Zhang, Shen, Lu, & Li, 2009). As a result of these physical alterations, children experience drastic changes in their lifestyle. They are unable to engage in formerly enjoyable activities, such as playing outdoors, engaging in sport, going to school, and socializing with friends. This can lead to boredom, loneliness, and feelings of entrapment (Moody, Meyer, Mancuso, Charlson, & Robbins, 2006). Long-term hospitalizations, regular school absences, and changes in schooling (e.g., shift to home schooling, entrance into modified classes) further exacerbate this isolation. Separation from one‟s family can also foster feelings of abandonment (Sammarco, 2001), and time apart from peers can weaken social bonds and hinder social development (Moody et al., 2006). These children may fail to acquire important skills needed to maintain social relations with peers (Simms, 1995). Perspective-taking and empathy, for example, may not develop as quickly due to prolonged exposure to child-centered care. Moreover, with increasing time apart from friends, these children are less likely to share things  2 in common with their peers which can create anxiety (Kazak & Simms, 1996). Education disruption is also a concern with these children. In a study by Noll et al. (1999) that looked at school absenteeism in children with cancer aged 8 to 15 years, an annual average of 31.29 days of school were missed compared to the normative average of 6.17 days. Substantial school absenteeism can cause children to fall behind in their schoolwork, resulting in deteriorating achievement. Academic struggles are heightened by the declines in attention and information processing that often accompany the acute phases of illness (secondary to anxiety, fatigue, and pain) as well as by the physiological-based cognitive declines that accompany radiation treatment (Neville, 2003). These cognitive changes can be distressing to children, particularly those with a history of pre-morbid overachievement (Kazak & Simms, 1996). The stressors experienced by youth with cancer are likely to vary in nature and impact as the child passes through different phases of human development (Trask et al., 2003). For example, in grieving the loss of their former selves, younger children may develop a sense of „no control‟ over their health and future (Moody et al., 2006). In the absence of well developed problem solving and coping skills, as is often the case with chronically ill youth (Simms, 1995), poor emotion-regulation abilities can lead to irritability, resentment, and anger. Adolescent patients, in comparison, are confronted with even more stressors. They must come to terms with an altered body image, balance the influence of peers, family and societal expectations, internalize a personal value system, investigate their sexuality, and prepare for the workplace (Palmer, Mitchell, Thompson, & Sexton, 2007). Alongside these normal developmental stressors, additional disease-related stressors can be overwhelming. Adolescent patients may experience embarrassment due to changes in their appearance (e.g., hair loss, weight loss, growth stunting, scarring), frustration over restricted freedom and increased adult dependency, and anxiety over heightened awareness of their vulnerability to relapse and death (Ribi et al., 2005).  3 These teenagers are consequently at increased risk for low self-esteem and withdrawal (Eisenberg et al., 1984). While the cancer experience is stressful to youth of all ages, it appears that adolescents are at heightened risk for emotional turmoil (Kazak & Simms, 1996). The Cancer experience for Caregivers Given the myriad of stressors that their child must cope with, together with their own unique stressors, it is no surprise that caregivers struggle through the cancer experience. Initial distress in caregivers is likely related to loss and fear. They grieve the loss of their former „previously taken-for-granted‟ lifestyle as well as the loss of their „healthy‟ child and the dreams they had for that child (Cohen, 1993; Rolland, 1994). Comparison with healthy families and among siblings and peers makes their reality even more painful (Melnyk, Feinstein, Moldenhouer, & Small, 2001). Caregivers also experience apprehension about the future. In addition to fearing the death of their child, they also worry about acute and delayed declines in a variety of their child‟s domains of functioning, including cognitive, adaptive, physical, academic, and emotional (Landier & Bhatia, 2008). Caregivers may also worry about their inability to care for their child in an appropriate manner (Wereszczak, Miles, & Holditch-Davis, 1997), and separation from their child during the treatment phase is often distressing. Feelings of helplessness are prone to arise in both mothers and fathers (Grootenhuis & Last, 1997b). Caregivers must also take on multiple roles throughout the cancer experience. In addition to normal family and occupational roles, they must now assume roles related to negotiating the healthcare system, patient transportation, and home care (Klassen, Raina, et al., 2007; Pai, Drotar, Zebrack, Moore, & Youngstrom, 2006). They must also become advocators for their child in an effort to increase the likelihood of receiving needed services (Bronfenbrenner, 1979). Neuropsychological assessments, occupational and physical therapy services, individualized educational plans, and teacher‟s aide assistance are some examples of needed resources, particularly for children with leukemia or brain cancer. Caregivers must often form collaborative  4 partnerships with their child‟s school as well as with other organizations to increase the likelihood that such resources are received. With new roles to fill, caregivers have decreased time for leisure, travel, and relationships with others (Kazak & Simms, 1996). Increased time demands may require parental work absences which can elevate stress depending on the flexibility of the parent‟s employer (Bronfenbrenner, 1979). High costs of treatment together with reductions in, or loss of, income can significantly hinder finances. While this financial strain is somewhat mitigated in wealthy countries with universal health care and strong social policies, it can be totally overwhelming in less wealthy countries. Limited time, energy, and money can disrupt family activities and negatively impact the quality of family relationships (Cohen, 1993). Spouses (particularly in younger couples) are less likely to spend time together and to communicate in a loving manner (Eisenberg et al., 1984). Siblings may also build resentment due to the extra attention their sick brother or sister is receiving, and this may create guilt in parents (Cayse, 1994). Overall, caregivers must juggle medical, family, personal, and work demands within a context of taxed resources. As time passes, many caregivers feel increasingly alone (Melnyk et al., 2001). Despite improved efforts by the health community to promote a sense of identity and community among these families, caregivers remain at increased risk for marginalization and alienation. With prolonged caregiving, feelings of burden are likely to increase (Kazak, 1987), and these tend to be heightened by concerns over future child dependency (Kazak & Simms, 1996). Together, these experiences of grief, fear, loneliness, and burden put caregivers at high risk for exhaustion and depression. Current Estimates of Distress in Patients and Caregivers Despite the many stressors that youth with cancer experience, reviews show these children to be adjusting quite well (Phipps, 2005). This is particularly true of younger children. In adolescents, signs of psychological distress are reported in approximately 10% to 20% of  5 cases (Jones, 2008). Combining all age groups, standardized effect sizes relative to healthy peers fall in the range of -.23 to -.68. Research by Phipps (2007) confirmed these self-reports are genuine and therefore not biased by denial, defensiveness, or socially desirable responding. Teacher and peer reports have also shown that youth with cancer tend to be appraised as less aggressive, more sociable, and more socially accepted by teachers and peers (Noll et al., 1999). The „resilient‟ nature of these children is believed to stem from strong social support systems as well as experiences of personal growth (Stam, Grootenhuis, Caron, & Last, 2006). Unfortunately, this resiliency is not paralleled in the caregiver population as heightened levels of psychological distress and depression are reported in these individuals (Grootenhuis & Last, 1997a; Kim, Duberstein, Sorensen, & Larson, 2005). Using meta-analysis, Pai and colleagues (2006) looked at psychological functioning in caregivers of children at various stages in the cancer course, from recent diagnosis to survivorship. They found a combined prevalence rate of 51% for anxiety, depression, and post-traumatic stress disorder. In a more recent study by Hardy and colleagues (2008) that assessed psychological functioning in caregivers of pediatric cancer survivors many years into survivorship (on average, 18.4 years post diagnosis), stress levels were found to be elevated and comparable to those of caregivers of pediatric cancer patients undergoing active treatment. According to Houts, Nezu, Zezu, & Bucher (1996, p.72), “social, moral, and economic demands of healthcare require that efforts to alleviate caregiver burden be forthcoming.” A better understanding of the forces impacting emotional adjustment in caregivers is therefore clearly needed as such knowledge will be crucial to the development of improved screening and intervention approaches. In the next section, those influential variables that have already received significant attention in the literature are reviewed. Moderator Variables Associated with Caregiver Emotional Maladjustment The literature has already identified several variables significantly associated with  6 emotional distress in caregivers of pediatric cancer patients. These are termed „moderator variables‟ because they influence the degree (i.e., the strength of the relationship) to which the cancer experience translates into emotional distress (Baron & Kenney, 1986). In the following sections, these are broken down into 6 over-arching domains: socio-demographic characteristics, family stress, social support, communication with healthcare professionals, coping skills, and child characteristics. Socio-Demographic Characteristics Socio-demographic predictors of caregiver emotional maladjustment have been shown to include female gender, African-American ethnicity, being single, and low socio-economic status (SES) or not being gainfully employed (Grootenhuis & Last, 1997b; Horton & Wallander, 2001; Klassen, Raina, et al., 2007; Van Dongen-Melman et al., 1995). Rural living has also been identified as a potential stressor worthy of further study (Silveria & Winstead-Fry, 1997). Being of African-American ethnicity may be maladaptive because it tends to be associated with greater life stressors, owing to its minority status and to its relationship with lower SES (Raina et al., 2004). Lower SES makes transportation challenging and access to appropriate care and support less likely (Guidry, Aday, Zhang, & Winn, 1997). It also tends to be associated with lower education which may, or may not, be maladaptive (Klassen, Raina, et al., 2007). While some believe that lower education is adaptive because it is associated with a poorer appreciation of the complexity of the problem (Geffken, Keeley, Kellison, Storch, & Rodrigue, 2006), others believe it is maladaptive because it is associated with inferior information processing skills and less acquired knowledge (Kim, Schulz, & Carver, 2007). The latter are believed to be helpful in communicating with health care professionals and in interpreting information (Mishel & Braden, 1988). The vulnerability of rural populations may be explained in several ways. In addition to being characterized by lower education, less health insurance, higher unemployment, and more  7 poverty (Elliott, Elliott, Renier, & Haller, 2004; Silveira & Winstead-Fry, 1997), rural living is associated with geographic isolation. Separation from metropolitan centers is likely to be associated with greater unmet needs. For example, rural families tend to have smaller social support networks and limited access to local physicians as well as specialized (diagnostic, treatment, and therapeutic) professional services (Elliott et al., 2004; Neville, 2003). Other needs that are often unmet relate to obtaining (appropriate) health information, being (sufficiently) involved in the child‟s care, and (actually) receiving guidance on how to help their child adjust to illness symptoms. Such unmet (medical) needs are a problem because they predispose family members to feelings of distress (Kaplan, 1982). Rural families are also at a disadvantage because they must travel considerable distances to receive specialized treatment. This is problematic given that travel is a significant psychological stressor (Payne, Jarrett, & Jeffs, 2000). Train, bus, and airline services are more limited in rural areas, making both short and long distance travel an inconvenience. Fatigue, discomfort, expense, and organizational difficulties due to being away from home are all reported as stressors of long distance travel (Junor, Macbeth, & Barrett, 1992). Lengthy travel also increases the probability that caregivers will suffer financial losses due to inability to hold down a job. Because of these concerns, rural families often attend fewer followup visits and experience more home care burden (Luft, Hershey, & Morrell, 1976). When they do travel, prolonged caregiver absence from home puts family organization at risk for difficulties (Klassen, Raina, et al., 2007), which represents another salient stressor for caregivers. Family Stress Concurrent family problems are likely to enhance psychological distress in caregivers. Examples of such stressors include chronic disease in a family member (Van Dongen-Melman et al., 1995), recent family death, marital dissatisfaction, family communication problems, and even positive stressors such as the birth of a child (Klassen, Raina, et al., 2007; Melnyk et al., 2001). Parenting style and parenting efficacy are also important variables to consider (Geffken et  8 al., 2006). Having flexibility and portraying a sense of normalcy through maintaining routines and expectations is considered adaptive (Knafl & Deatrick, 2006). In contrast, perceiving one‟s child as vulnerable and engaging in overprotective parenting can lead to dependency or resistance in the child and disturbed parent-child interactions (Bendell et al., 1994). This is problematic given that lower family cohesion has been found to be related to increased caregiver distress (Klassen, Raina, et al.). This finding reflects the importance of interpersonal support, a variable that has received much attention in the literature for its stress-buffering capacity (Hughes & Caliandro, 1996). Social Support Social support is defined as support available to an individual from their surroundings (Kim et al., 2007). It can be derived from anyone in a person‟s social environment, including friends, colleagues, neighbours, and family (Raina et al., 2004). It can also come in a variety of forms, including emotional (i.e., being cared for and loved), informational (i.e., receiving information and guidance), and instrumental (i.e., receiving aid materially or manually) (Cobb, 1976). Social support in caregivers of cancer patients is adaptive to the extent that it is perceived as available and satisfying (Klassen, Raina, et al., 2007). This importance of „perceived satisfaction‟ may explain rare findings where denser support networks are associated with worse emotional adjustment in chronically ill settings (Kazak, 1988). Social support is beneficial to caregivers because it fosters a sense of connectedness, provides a context in which emotions can be expressed (Manne et al., 2004), facilitates healthy living and access to medical care (Creagan, 1997), and allows for the exchange of ideas. Having one‟s beliefs affirmed and learning how to enhance coping and problem-solving skills are some ways this information exchange is useful (Cohen & Willis, 1985). The latter may be especially beneficial due to enhancing feelings of mastery or „control‟ (Janda, Eakin, Bailey, Walker, & Troy, 2006). In caregivers of pediatric cancer patients, emotional support is particularly effective  9 when provided by others who are experiencing, or have had previous experience with, the illness (Wortman, 1984). This is especially the case during hospitalization, as caregivers can exchange ideas on ways to monitor their child‟s clinical status, follow treatment plans, and reduce side effects. Support is also quite effective when provided by one‟s spouse. This is likely why being married has been associated with lower distress levels. Religious groups can also be helpful as they provide routine socialization and can organize fundraising efforts and „calming‟ spiritual ceremonies (Yeh, Lee, & Chen, 2000). Lastly, the healthcare system represents a source of support that can offer great comfort to caregivers. Unfortunately, this support is often not sufficient. Communication with Healthcare Professionals Limited contact with healthcare professionals has been highlighted by cancer patients and their families as one of the major stressors of the adjustment process (Burman & Weinert, 1997; McKenna, Collier, Hewitt, & Blake, 2009). Navigating through cancer treatment and the healthcare system can be overwhelming (Bronfenbrenner, 1979), and information received is often unfamiliar, overly complex, or inconsistent. In addition, unfamiliarity with hospital protocols and (frequently changing) staff can leave caregivers and patients feeling confused and intimidated (Geffken et al., 2006). When health professionals take time to establish rapport, discuss important topics, and provide guidance, caregiver distress often declines. These communications legitimize the caregiver‟s role as a member of the health care team, increase knowledge, enhance feelings of self-efficacy and control, and reduce feelings of isolation (Bronfenbrenner; Mishel & Braden, 1988). Examples of useful information include being told how the hospital and healthcare system operates, how to find help when needed, what the goals of treatment are, what alternative courses of action exist, what challenges lay ahead, and what caregiving responsibilities entail (Houts et al., 1996). Lastly, receiving direction from health professionals on how to express (and encourage) realistic optimism can be comforting and  10 empowering, as this may enhance a caregiver‟s coping ability. This represents an additional domain of moderating influence. Coping Skills Coping relates to cognitive and behavioral efforts to manage demands that are appraised as taxing or exceeding one‟s resources (Lazarus & Folkman, 1984). Coping styles associated with positive caregiver emotional adjustment tend to include problem-focused coping strategies in which actions are undertaken to remove the stressor (Lazarus, 1991). Examples of these include engaging with others, seeking information, and fulfilling personal needs and desires such as through work, hobby, and leisure activity (Grootenhuis & Last, 1997a; Houts et al., 1996; Klassen, Raina, et al., 2007). Alternatively, coping styles associated with caregiver emotional maladjustment tend to include emotion-focused strategies in which efforts are made to avoid or regulate the emotional distress (Lazarus, 1991). Examples of these include denial, behavioural withdrawal, and religious coping, in which individuals cognitively restructure stressful events to search for a spiritual meaning (Baskin, Forehand, & Saylor, 1985; Trask et al., 2003). Females tend to use these strategies more than men (Grootenhuis & Last, 1997a) which may explain heightened female vulnerability to caregiver distress. The distinction between „adaptive‟ problem-focused strategies and „maladaptive‟ emotion-focused strategies, however, is not entirely accurate. Equivocal findings have been reported, whereby problem-focused strategies (e.g., engagement coping) have failed to yield benefit to caregivers (Trask et al., 2003) and emotion-focused strategies (e.g., religious coping) have yielded significant benefit (Cayse, 1994; Van Dongen-Melman et al., 1995). The coping literature is plagued by the presence of numerous coping taxonomies, multiple operational definitions for each strategy, and an array of different measures assessing similar constructs. Many coping researchers fail to understand that coping can only be studied in the context of a specific stressor. These conceptual problems and inconsistencies have led to substantial  11 heterogeneity in the literature. In addition, moderating variables likely exist that influence the degree to which coping strategies are adaptive. In the area of religious coping, for example, degree of church-going and severity of (child‟s) illness are likely to be influential forces. Higher rates of church-going may be associated with more adaptive outcomes given that regularly attending church services makes this variable increasingly more social and active (i.e., problemfocused). Moreover, affiliation with religion has been found to be particularly adaptive in families experiencing advanced disease (Creagan, 1997), perhaps because the search for meaning and acceptance is stronger in individuals where dreaded events seem more likely to occur. This example highlights a variable from yet another domain of moderators likely to influence caregiver distress; that pertaining to the sick child. Child Characteristics Child influences on caregiver emotional adjustment include both socio-demographic variables and objective indicators of disability. In the realm of socio-demographic variables, (younger) child age as well as child emotional and behavioural problems have been identified as significant predictors of caregiver emotional maladjustment (Grootenhuis & Last, 1997b; Klassen, Raina et al., 2007). Younger child age may be a risk factor due to increased perceptions of emotional and physical (child) vulnerability as well as elevated worries surrounding school adjustment (Melnyk et al., 2001). Caregivers of elementary-aged cancer patients experience significant concern about teasing and academic struggles, failure to make friends, and responsibilities (of their own) to educate the school on the unique needs of their child. Emotional and behavioural problems are also distressing given that they require additional surveillance and control on the part of the caregiver (Klassen, Raina et al., 2007). While clinically significant emotional problems are not prevalent in pediatric patients, certain child characteristics are associated with higher levels of these. Female gender, minority status, and older age at diagnosis, for example, have all been  12 identified as vulnerability factors (Melnyk et al.; Stam et al., 2006). Meanwhile, pediatric patients fare better when they are well informed and included in decision making, when they possess effective problem-solving and decision making skills, when they keep busy and maintain socialization and school participation (Meijer, Sinnema, Bijstra, Mellenbergh, & Wolters, 2002), and when their classmates are well informed of their illness (Palmer et al., 2007). Interestingly, while low social support is usually associated with higher distress levels in these children (Trask et al., 2003), higher family cohesiveness has not always associated with better emotional adjustment (Rait et al., 1992). This may reflect the encroaching nature of overprotective parenting, which was highlighted earlier as a potential child stressor. With respect to child disability indicators, a shorter time since diagnosis, having a hospitalized child, and having a child currently receiving treatment are noted risk factors (Fife, Norton, & Groom, 1987), presumably due to increased demands on the caregiver as well as elevated perceptions of child distress and increased fear about child prognosis (Raina et al., 2004). In addition, greater number of hospitalizations (Mulhern, Fairclough, Smith, & Douglas, 1992), history of relapse (Grootenhuis & Last, 1997b), poorer prognosis (Klassen, Raina et al., 2007), greater functional impairment (Manne et al., 1995), and presence of long-term neurocognitive sequalae (Van Dongen-Melman et al., 1995) have all been associated with worse caregiver (and for the most part, patient) emotional adjustment. [Of note, greater functional impairment and more neurocognitive late effects are seen in young brain tumor and acute lymphatic leukaemia survivors, as these patients often sustain substantial damage to their central nervous system secondary to the toxic treatments they receive (Stam et al., 2006).] Despite these noted trends, inconsistencies prevail. Certain studies, for example, have failed to find negative outcomes associated with severity of physical disability or treatment status (Horton & Wallander, 2001; Trask et al., 2003; Van Dongen-Melman et al., 1995). In  13 addition, while caregiver distress is usually highest at periods closer to diagnosis, some studies have found significant emotional turmoil in caregivers of patients who are no longer on treatment, including those many years following transition into survivorship (Hardy et al. 2008). Once again, the inconsistency in the research emphasizes the need to consider the role of moderating variables. This is best accomplished in the context of a predictor model of caregiver distress – a framework that can incorporate a multitude of noted risk and resilience factors and allow testing of relations amongst these variables as well as between them and caregiver distress. In the next section, another powerful influence on caregiver distress will be introduced, and a well-known predictor model grounded in this variable will be explored. Uncertainty and the ‘Uncertainty in Illness Theory’ According to Mishel (1988), „uncertainty‟ is defined as being unable to form a cognitive framework for understanding one‟s situation or being unable to predict future outcomes. It has been described as the single greatest source of psychosocial stress for persons affected by lifethreatening illness (Koocher, 1985). In caregivers of oncology patients, uncertainty is consistently acknowledged as a major stressor throughout the experience, including years after treatment (Carpentier, Mullins, Chaney, & Wagner, 2006). While uncertainty may change in its focus and intensity as time passes (Mast, 1988), it seems to persist long after completion of treatment (Santacroce, 2002). Feelings of uncertainty are reported in relation to the nature of the illness and its treatment, the unpredictability of the disease course and treatment success, the impact of the disease on the child‟s cognitive and adaptive functioning, and the confusion surrounding the roles and responsibilities of the caregiver (Geffken et al., 2006; Green & Murton, 1996). In preliminary research using the Mishel Uncertainty in Illness Scale (MUIS; Mishel, 1981), Mishel observed that uncertainty in women with gynaecological cancer was a significant precursor of a „threat‟ appraisal (Mishel, Hostetter, King, & Graham, 1984). She therefore  14 applied her views on uncertainty to Lazarus & Folkman‟s (1984) „stress and coping model‟, in which stress levels are determined based on how a stressor is appraised and how one appraises their resources to cope with the stressor. Mishel‟s model was called the „Mishel Uncertainty in Illness Theory‟ („MUIT‟; Mishel, 1988). In her initial model, Mishel outlined how uncertainty arose within illness contexts and how its appraisal influenced psychological outcomes. The model addressed antecedents to uncertainty, types of cognitive appraisal of uncertainty, and coping pathways mediating the relationship between appraisal and emotional distress. Two years later, Mishel proposed a revised theory specifically tailored to chronic illness entitled the „Reconceptualization theory’ (Mishel, 1990). Another year later, the theory was further revised to account for the role of perceived mastery (Mishel & Sorenson, 1991; see Figure 1). Mishel conceptualized uncertainty as a neutral cognitive state that is neither desired nor dreaded until its implications are determined. She believed this to happen through the process of cognitive appraisal (see Figure 1, B) which, similar to Lazarus & Folkman‟s (1984) model, culminates in the perception of either a „danger‟ (i.e., threat) or an „opportunity‟ (i.e., challenge). While danger appraisals were presumed to occur in most uncertain (i.e., stressful) situations due to confusion and fear, Mishel hypothesized that opportunity appraisals could arise in situations perceived as controllable. Mishel therefore speculated that „perceived mastery‟, or beliefs about one‟s ability to be in control of events (Pearlin & Schooler, 1978), mediates the appraisal of uncertainty. Research with women receiving treatment for gynaecological cancer (Mishel & Sorenson, 1991) confirmed that perceived mastery acts as a situation-specific personality factor in this model. [This was is in agreement with Rosenbaum (1988) who also perceived mastery as a state variable, but it was in contrast to Folkman, Lazarus, Gruen, & DeLongis (1986) who classified mastery as a generalized (or trait) personality disposition.] Following the appraisal process, Mishel believed that coping efforts (see Figure 1, C)  15 were activated. She posited that these mediate the relationship between appraisal and emotional distress. Specifically, she hypothesized that „emotion-focused coping strategies‟ were implemented in the face of danger appraisals while „problem-focused coping strategies‟ were used in the face of opportunity appraisals, with either pathway capable of achieving distress reduction. Research Related to the MUIT in Adult Patient Populations In early research on the MUIT with adult patients, Mishel and Sorenson (1991), and later Mu et al. (2001), found that perceived mastery was a significant mediator in the relationship between uncertainty and stress appraisal, both for opportunity appraisals (R2=.20 = partial mediation) and danger appraisals (R2=.27=full mediation). The differing mediation strengths by mastery in these two appraisal pathways confirmed that these pathways are not exact opposites and, therefore, likely have unique antecedents. Mishel and Sorenson‟s research also revealed that coping strategies had minimal impact in mediating the relationship between appraisal and distress. Only two of seven coping strategies, namely „focusing on the positive‟ and „wishful thinking‟, were found to be significant mediators, and these mediation effects were small. In agreement with previous research (Christman et al., 1988, Webster & Christman, 1988), Mishel & Sorenson concluded that coping strategies have minimal impact in mediating the relationship between appraisal and distress. In later research by Padilla, Mishel, & Grant (1992) assessing health-related quality of life (QOL) as an outcome variable in women receiving treatment for gynaecological cancer, the insignificant mediating role of coping strategies was again confirmed. Meanwhile, uncertainty, mastery, danger appraisal, and positive mood state were found to account for the bulk of variance in psychosocial QOL (i.e., 56% of the total 59% variance explained). Several years later, a study looking at predictors of emotional distress in breast cancer survivors similarly found that symptom distress, fear of recurrence, concurrent illness problems, uncertainty, and  16 positive reappraisal (to be discussed) together explained 51% of variance in distress (Mast, 1998). In later studies on the MUIT with breast cancer survivors, both age and uncertainty were found to explain significant variance in threat appraisals (Wonghongkul, Moore, Musil, Schneider, & Deimling, 2000). Alternatively, age, education, and uncertainty all failed to explain significant variance in challenge appraisals. While the latter finding suggests uncertainty has little (negative) influence on challenge appraisals, it should be viewed with caution given that this sample was homogenous and biased towards low uncertainty and distress. The majority of participants had high levels of education and support, had been cancer-free for more than 5 years, and had been trained as support givers. In a further study by Sammarco (2001) assessing adult cancer survivors, regression analyses revealed that perceived social support entered in step 1 (R2=.17) and uncertainty entered in step 2 together accounted for a significant amount (R2=.27) of variance in overall QOL. Given that a large amount of variance in outcome remained unexplained, the authors encouraged future research to investigate new risk and resilience variables that may be impacting on the emotional adjustment process. Research Related to the MUIT and Psychological Adjustment in Caregiver Populations While most research pertaining to the MUIT and related constructs has been performed with adult patients, data in this regard are slowly accumulating on caregiver populations. Although the actual model has yet to be tested on caregivers, studies have assessed antecedents to (and outcomes of) certain constructs in the model. Uncertainty has been shown to predict distress in caregivers of chronically ill youth (Carpentier et al., 2006) as well as depression in mothers of children with cancer (Grootenhuis & Last, 1997b). Significant negative correlations between uncertainty and perceived mastery have also been documented in caregivers (Mu et al., 2001). Caregiver mastery tends to be lower in the  17 context of patient relapse (Mu et al., 2001) and higher in situations characterized by familiarity or clarity (Rosenbaum, 1988). Elevated mastery has been shown to lead to more active coping and it is repeatedly found to associate with better mental health in caregivers (Affleck & Tennen, 1991; Klassen, Raina, et al., 2007). In terms of strategies to deal with uncertainty, caregivers have reported relying on information management (i.e., gathering information about the illness) and avoiding social encounters or information that draws attention to negative aspects of uncertainty (Stewart & Mishel, 2000). In an attempt to better conceptualize the various forces influencing caregiver adjustment in pediatric cancer, Klassen, Raina et al. (2007) performed a review of the literature. Using the „caregiving process and caregiver burden model‟ („CPCBM‟; Raina et al., 2004) as a guiding framework, they examined the degree to which research findings on predictors of caregiver adjustment fit with this model. The model, which includes variables of uncertainty and perceived mastery among others, is based on several over-arching constructs gathered from research in both pediatric (child chronic physical illness and disability) and geriatric caregiving populations. From their analyses, Klassen, Raina et al. concluded that substantial evidence exists to support the influential roles of variables relating to child characteristics (e.g. child behaviour) and indicators of coping (e.g., social support) while little evidence exists to support the predictive role of variables such as mastery and perceived involvement in care. These null findings were suspected to reflect the dearth of research in this area rather than the insignificance of these variables. Klassen, Raina et al. emphasized the need for further research on risk and resilience variables as well as on mediator and moderator relations between these variables. [Of note, while Klassen, Dix, et al. (2007) subsequently developed and tested an oncology-specific stress process model that examined the impact of multiple predictor variables on caregiver healthrelated QOL, findings from this research have not been published. Regardless, their use of QOL  18 as an outcome variable is not ideal given the high degree of item overlap anticipated between predictor measures and the QOL measure.] To date, only a handful of models have been used to guide hypotheses in studies of emotional adjustment in the pediatric oncology caregiver population. These studies have shown explained variance in emotional distress ranging from 7% (Sloper, 2000) to 77% (Manne et al., 1995), with the average estimate hovering around 44% (Klassen, Raina et al., 2007). The models vary widely and most studies on them have had relatively small samples (i.e., approximately 90% have had sample sizes smaller than 100), thus limiting power to detect small but clinically significant relationships (Klassen, Raina et al., 2007). Moreover, only one study to date has reported results based on sophisticated statistical techniques such as structural equation modeling (Yeh, 2003). Worse still, these studies have paid little to no attention to the role of uncertainty. While correlation and (longitudinal) regression research has confirmed several antecedents to, and consequences of, caregiver uncertainty, additional factors need to be identified that mediate and moderate the impact of this variable (Stewart & Mishel, 2000). As uncertainty is a salient stressor in these caregivers, and one that appears to persist, it is essential that researchers start identifying variables that help individuals live comfortably with it. An improved model building off of the MUIT and incorporating new resilience variables is needed to guide future hypotheses and, ultimately, the development of enhanced screening and intervention approaches for these caregivers. To develop this model, we need to focus on concepts from the field of positive psychology. Field of Positive Psychology Positive psychology is an emerging scientific field that represents the study of positive emotion and positive character traits. Rather than focusing on what ails human beings, this field aims to understand personal fulfillment and optimal functioning. Paramount to the ideology of positive psychology is the belief that “frequent positive affect can be highly adaptive”.  19 According to Fredrickson‟s (2002) Broaden-and-build theory, experiencing positive emotions/thoughts broadens people's scopes of attention and cognition as well as their momentary thought-action repertoires. These changes are suspected to cultivate enduring personal resources, ranging from physical and intellectual resources to social and psychological resources. Novel and creative actions, ideas, and social bonds, for example, may be discovered through play, exploration, and structured goal-pursuit. Fredrickson believes that these experiences promote further positive affect thus propelling the „broaden-and-build‟ cycle forward. Positive emotion is also believed to be helpful in combating negative emotion. Through cultivating personal resources that instil a predisposition towards further positive affect across situations, positive emotion makes the escalation of stress less likely. This is referred to as the „buffering capacity‟ of positive emotion (Tugade, Fredrickson, & Barrett, 2004). It is hypothesized that individual positive emotions need not be intense to produce benefits given their power to initiate this upward spiral towards increasing well-being (Moskowitz, Folkman, & Acree, 2003). While past theorists equated positive beliefs and expectancies to denial in the context of chronic illness and consequently judged these maladaptive, there is now an appreciation for the need to consider the positive view of one‟s experience and to identify when these views arise and how they operate (Mishel, 1990). As previously mentioned, research with adult cancer patients has found positive mood to be one of the most significant predictors of health related QOL (Padilla et al., 1992). Similarly, in the area of pediatric caregiving, positive belief systems have been associated with lower psychological distress (Frey, Greenberg, & Fewell, 1989). However, it is relatively unknown which positive psychology variables are most influential in this context. We will now explore some potentially salient positive psychology constructs in this realm and review the applicable empirical research that has accumulated.  20 Positive Psychology Variables Relevant to Models of Adjustment and Uncertainty A „Classification of Character Strengths and Virtues‟ (CSV) similar to the DSM-IV classification of mental disorders was fairly recently created by Peterson & Seligman (2004). Of the six broad virtues highlighted in the CSV, three are particularly relevant to emotional adjustment in pediatric oncology: „humanity‟, „courage‟, and „transcendence‟. Humanity encompasses interpersonal strengths that involve tending and befriending others. Courage captures emotional strengths that involve determination to pursue goals in the face of external or internal resistance. Finally, transcendence taps into strengths that provide meaning. Two positive psychology constructs that tap into these virtues of humanity, courage, and transcendence are benefit finding and hope. In the following sections, these constructs will be defined and the literature supporting their potential role as moderators of uncertainty and distress in the pediatric oncology caregiving experience will be explored. Benefit Finding Benefit finding relates to finding the silver lining in adversity. Rather than being an active and intentional process, it involves passive acknowledgment of accumulated positive experiences in the aftermath of a negative event. The notion that life-threatening illness can stimulate self-reflection and provide meaning has long pervaded the health literature (Rapoport, 1962). Increasingly, research is showing that people with cancer and parents of children with severe health problems often identify positive ways in which their lives have changed as a result of the traumatic event (Helgeson, Reynolds, Tomich, 2006). Benefits often reported relate to acceptance (i.e., for things as they are), empathy (i.e., becoming aware of the vulnerability of self and others), appreciation (i.e., for the gift of life, for the need to increase life enjoyment, for how one has grown spirituality, behaviourally, and socially), family (i.e., closer family relations, greater support), positive self-view (i.e., stronger mental health and self-reliance), and reprioritization (i.e., of goals, attitudes, and purpose in life) (Carter, 1993; Kim et al., 2007;  21 Manne et al., 2004; Thornton, 2002). Over the years, several constructs have been studied that share similar features to benefit finding. In Mishel‟s original uncertainty in illness theory (Mishel, 1988), „focusing on the positive‟ was highlighted as a coping strategy utilized to block information that would otherwise threaten opportunity appraisals. This is similar to „positive reappraisal‟ or (more recently) „benefit reminding‟. All these strategies refer to „active coping styles‟ in which individuals intentionally bring to mind positive experiences as a way to combat negative emotions (Sears, Stanton, & Danoff-Burg, 2003). Benefit finding, in contrast, is viewed not as a coping strategy but rather as a passive „experience‟ that onsets randomly and serves no purpose. Research showing no relation between benefit finding and neuroticism is accepted as support for the position that benefit finding is not a coping strategy used to reduce distress (Helgeson et al., 2006). Though benefit finding has yet to be researched in the context of uncertainty, research on this construct in the cancer field is starting to accumulate. Research on benefit finding in adults.  In the only study to look at benefit finding in  caregivers (predominantly female spouses) of cancer patients (largely breast and prostate cancers diagnosed on average 2.2 years prior to study), Kim et al. (2007) found a moderate level of benefit finding overall. In order from most to least common, benefit finding was reported in areas of empathy („quite a bit‟), acceptance, reprioritization, family, positive self-view, and appreciation („moderate‟). Significant predictors of benefit finding were social support and religious coping. Older age, lower education, and male gender showed positive correlations with certain domains of benefit finding. Variables that have failed to show relations with caregiver benefit finding include income, spousal status, neuroticism, treatment status, and time since diagnosis (Helgeson et al., 2006; Kim et al., 2007). With respect to the latter, it may be that while increasing time since diagnosis provides one with a greater time span to reflect on  22 accumulated benefits, it is also associated with habituation to the stressor. Given that no other research exists on benefit finding in caregivers, a few details from the adult patient literature should be noted. In this realm, significant predictors of benefit finding have included optimism, female gender (in contrast to caregivers), recurrent cancer, problemsolving, and emotional expression (Guellati-salcedo, 2005; Helgeson et al., 2006; Lechner, Carver, & Antoni, 2006; Llewellyn et al., 2010). Benefit finding in adult cancer patients has also been positively correlated with minority status (Stanton, Bower, & Low, 2006) while findings have been mixed in regards to education, SES, spousal status, and age (Helgeson et al., 2006; Lechner et al., 2006; Linley & Joseph, 2004; Sears et al., 2003). Also, curvilinear relationships have been found in the area of disease severity such that benefit finding appears to be highest in medium-severity cases relative to low- and high-severity cases (Lechner, Zakowski, & Antoni, 2003). In this regard, it is speculated that „perceived severity‟ is more related to benefit finding than actual „disease severity‟ (Stanton et al., 2006). With respect to outcomes of benefit finding, Kim et al. (2007) found that benefit finding was significantly predictive of greater life satisfaction, but not depressive symptoms, in caregivers. When broken down across benefit finding sub-domains, both acceptance and appreciation predicted greater adaptation, while both acceptance and positive self-view predicted fewer depressive symptoms. Meanwhile, empathy and reprioritization surprisingly predicted greater symptoms of depression. Ambiguous findings like this have also been reported in the adult patient literature. For example, while benefit finding in this population has been linked to less distress (Urcuyo, Boyers, & Carver, 2005), greater well-being (Carpenter, Brockopp, & Andrykowski, 1999), increased perceptions of control (Tennen, Affleck, Urrows, Higgins, & Mendola, 1992), and more positive mood at up to 8 years post-diagnosis (Carver & Antoni, 2004), it has also been linked to more distress and poorer QOL (Tomich & Helgeson, 2004).  23 Some research has also failed to find an association between adult patient benefit finding and depression or quality of life (Llewellyn et al., 2010). Some of these inconsistent findings may be explained by moderating variables. Elevated hope (Stanton, Danoff-Burg, & Huggins, 2002) and minority ethnicity (Helgeson et al., 2006), for example, have been found to predict stronger relations between benefit finding and well-being. Another potential moderator may be time since diagnosis, as longer follow-up studies (Carver & Antoni, 2004) have shown better outcomes associated with benefit finding relative to shorter follow-up studies (Sears et al., 2003). In a meta-analytic review by Helgeson et al. (2006), benefit finding was found to associate more strongly with less depression and greater positive well-being when time since diagnosis exceeded two years. It is possible that benefit finding may not be adaptive until an individual has worked through their thoughts on constructs important to them, such as family, life purpose, and goal achievement (Sprangers & Schwartz, 1999). For many people, this „reinterpretation and acceptance‟ process may take considerable time to complete. Research methodology in studies on benefit finding varies widely which may also explain inconsistencies (Helgeson et al., 2006). Studies have used a variety of measures varying in degree of relation to actual „benefit finding‟. Instruments have ranged from „multi-domain‟ to „single item‟ questionnaires, and measures of positive reappraisal and post-traumatic growth have also been inaccurately used to assess this construct. Moreover, as most research has explored adult cancer patients, findings cannot be generalized to pediatric patients or their caregivers. And in the only study to look at benefit finding in caregivers, spouses were assessed rather than parental figures. Lastly, while current data suggest benefit finding may be more beneficial in certain individuals, at certain time periods, or in the context of certain variables, the failure of previous research to accurately assess moderator relations prevents firm conclusions  24 from being drawn. Clearly, more sophisticated research is needed to better delineate predictors and moderators of benefit finding. Research on benefit finding in children. While most research on benefit finding in the context of a serious illness has been conducted on adults, the fairly recent development of a benefit finding measure for children (the Benefit Finding Scale for Children, BFSC; Phipps, Ogden, & Long, 2007) has generated some pediatric data in this area. As distress is not alarmingly prevalent in this population, pediatric mental health research is increasingly moving towards a focus on resilience, making benefit finding a viable area of exploration. In research using the BFSC, Phipps et al. found that more benefit finding was reported in children with an older age at diagnosis and/or a shorter time since diagnosis (in contrast to caregivers/adult patients). In these kids, benefit finding was not related to age, gender, diagnostic category, or treatment status, but it was significantly more prevalent in African American children than Caucasian children. Benefit finding is theorized to represent one of the pathways towards resilience in children with cancer (Phipps et al., 2007). Lower distress levels have been observed in those accepting their fate or finding meaning in it (Meijer et al., 2002). Despite these findings, early research using the BFSC failed to find an association between benefit finding and health-related QOL (Phipps et al.). It is possible that by including children in the 7 to 12 year age range (i.e., those demonstrating „concrete operational thinking‟; Piaget, 1954), this study suffered dilution of effects that otherwise might have been found if only teenage children (i.e., those demonstrating „formal operational thinking‟; Piaget, 1954) were assessed. According to Lane and Schwartz‟s (1987) model of „emotional awareness‟, children at the concrete operational thinking level have not yet developed the cognitive and social skills (such as empathizing ability and logical and creative thought) required to engage in self-reflection, perspective-taking, and abstract  25 reasoning, which are abilities likely required in benefit finding. While Phipps et al. did not find benefit finding to vary as a function of age, the mean age of the sample was young at 12.4 years (i.e., likely still engaging in concrete operational thinking) and the authors did not report sample sizes (nor benefit finding scores) across different age groups. Given that older patients are more likely to remember and appreciate experiences related to their illness and treatment, further study pertaining to benefit finding in adolescent patients has been strongly encouraged (Phipps et al., 2007; Stewart & Mishel, 2000). Hope Another construct suspected to influence caregiver emotional adjustment is hope. Hope, for our purposes a state-dependent variable, is a dynamic life force characterized by a confident yet uncertain expectation of achieving future good (Dufault & Martocchio, 1985). It involves having an ability to generate plans, achieve goals, remain interconnected, and maintain positivity in the context of a life stressor (Herth, 1992; Snyder et al., 1996). It has been shown to be a powerful predictor of QOL during periods of stress and loss (Gottschalk, 1985), including in caregivers of chronically ill children (McKeever, 1981). The current literature echoes an increasing appreciation that there is no such thing as „false hope‟ in the cancer experience – just „hope‟ (Luttrell, 2009; Penson et al., 2007; Thompson, 2010; Wirga, 2010). Research on hope. In early research, Mishel et al. (1984) found that hope was unrelated to illness severity but was negatively associated with uncertainty. A few years later in a sample of adult cancer patients undergoing radiotherapy treatment, Christman (1990) reproduced the same finding and also found that hope was one of the significant predictors (along with uncertainty and symptom severity) of adjustment. Christman hypothesized that adaptive appraisals of uncertainty led to elevated hope. While no formal testing of this hypothesis occurred, research has shown that elevated hope is found in cancer patients that receive high incomes, have high education levels, and are married (Bunston, Mings, Mackie, & Jones, 1995).  26 Low hope, in contrast, has commonly been found in situations characterized by pain. Both „severity of pain‟ and „interference in daily life due to pain‟ have been negatively correlated with hope (Hsu, Lu, Tsou, & Lin, 2003). Meanwhile, research has been inconsistent regarding whether hope levels are associated with gender, ethnicity, and age. With respect to cancer severity and type, studies continue to find no correlation (Felder, 2004). Hope is one of the few inner resources that repeatedly predicts psychological well-being (Richardson, Gibson, & Parker, 2003). According to Horton & Wallander (2001), hope is most beneficial in situations where caregivers perceive themselves to be burdened by a lot of stress. In adult cancer populations, hope has also been associated with a greater quantity of, and more effective, coping responses (Bruhn, 1984; Felder, 2004). Wonghongkul et al. (2000), for example, found that hope significantly predicted positive reappraisal coping in a sample of breast cancer survivors. While hope was not found to explain significant variance in challenge appraisals, the authors acknowledged that their study population was generally biased towards elevated hope levels and challenge appraisals and therefore likely misrepresentative of typical (distressed) populations coping with cancer. Other research has found hope to moderate the relationship between positive reappraisal coping and positive outcomes (Stanton et al., 2002). In a study that assessed both benefit finding and positive reappraisal coping in women with early stage breast cancer (Sears et al., 2003), hope was found to predict the latter but not the former. Benefit finding, rather, was only predicted by optimism, a distinct construct from hope that taps into a stable personality disposition (rather than a state-dependent cognition). „Optimists‟ are believed to hold generalized expectancies for positive outcomes (Scheier & Carver, 1985). Like hope, optimism has been negatively correlated with distress and (in particular) positively correlated with better psychological adjustment and QOL in adult cancer patients (Aspinwall & MacNamara, 2005;  27 Phipps, 2007) as well as in caregivers of pediatric cancer patients (Grootenhuis & Last, 1997b). As with benefit finding, hope research in cancer patients is burgeoning but is still relatively sparse in the area of caregiving. There is also a clear need for more research on hope in the context of a guiding model. Previous research on hope in relation to the MUIT has only been regression-based, and hope has strictly been assessed for its impact on stress appraisal rather than for its development consequent to it. With a better understanding of how this variable relates to other variables in caregivers, superior interventions can be designed that target ways to enhance this construct and, in turn, overall caregiver emotional adjustment. Present State of the Field From the background literature, it is clear that caregivers of pediatric cancer patients are afflicted with debilitating uncertainty that persists during and beyond the acute illness phase. This is concerning due to the negative impact this anguish can have on caregivers as well as on their ill child. Distressed parents are less emotionally available to their children and therefore less able to alleviate their child‟s distress (Koocher, 1985). Moreover, elevated distress levels can undermine cognitive abilities at a time when caregivers need to be responsible for technical aspects of their child‟s care and are being asked to make life-altering decisions about their child‟s therapeutic course (Santacroce, 2002). Given the far-reaching implications of caregiver distress, it is no surprise that this population is increasingly the focus of research in the psychooncology field (Klassen, Raina, et al., 2007). To help these individuals better adjust, a thorough understanding is needed of the forces influencing caregiver emotional adjustment as well as of the inter-relations between these forces. While a handful of theoretical models have been created to guide hypotheses in this regard, tests of these reveal that significant variance in caregiver emotional adjustment remains unexplained. The majority of research in this area has also been under-powered and lacking in analytical (i.e., statistical) sophistication (Klassen, Raina, et al., 2007). In addition, certain pertinent variables  28 showing significant influence in other cancer populations have yet to be appropriately studied in these caregivers. With the growth of the positive psychology field in recent decades, there is potential to enhance our understanding of caregiver emotional adjustment and, ultimately, improve screening and intervention efforts aimed at helping this population. Current Study The aim of the current study was to accumulate more theoretical knowledge on forces impacting emotional adjustment in caregivers during the pediatric cancer experience, using modeling techniques. Focus here is on „emotional adjustment‟ rather than „rehabilitation‟ or „adaptation‟ given that these latter terms tend to encompass too broad a range of functioning. More specifically, chronic illness rehabilitation or adaptation refers to a process of change (initiated by the need to accommodate to a new illness environment) impacting physiological, psychological, social, environmental, occupational, independent living, and political realms (Livneh, Martz, & Bodner, 2006). While these areas are all pertinent to (particularly adult) patients with chronic illness or disability, they are less relevant to caregivers. In addition (and as previously mentioned), a more one-dimensional construct is desired as an outcome variable in modeling research to avoid overlap with predictor variables (which capture facets of functioning in many realms of life). Research on psychological adjustment has generated a large number of constructs that are known to be inter-related. For the sake of parsimony it is critical that we clarify how these constructs are inter-related and identify which constructs are (so overlapping that they be considered) redundant with each other. While this study will assess multiple predictors of emotional adjustment via regression analysis, its main goal is to test a new model of caregiver emotional adjustment using structural equation modeling (SEM). The intended purpose of this model is to increase our understanding of why certain individuals function relatively well during the cancer experience while others struggle greatly. The model builds on Mishel & Sorenson‟s  29 MUIT (1991; see Figure 1), which is one of the trademark models for conceptualizing adjustment in the context of serious physical illness. The MUIT is especially relevant to the cancer experience because it addresses the role of uncertainty, which is a hallmark of the cancer experience. While the MUIT has been utilized in several studies with adult cancer populations (Mishel & Braden, 1987; Mishel & Sorenson, 1991; Padilla, Mishel, & Grant, 1992; Sammarco, 2001; Wonghongkul, Dechamprom, Phymivichuvate, & Losawatkul, 2006), it has yet to be appropriately assessed in caregivers. Of note, the proposed model is based on only part of Mishel & Sorenson‟s MUIT model. It does not address aspects of the model relating to „antecedents to uncertainty‟ or „coping strategies following appraisal‟ given that (i) from a clinical perspective, we feel it is most important to identify variables that follow the experience of uncertainty (a given in the cancer experience) once it has arisen; (ii) research on coping strategies tends to be equivocal and inconsistent (as discussed); and (iii) a variety of studies have shown problem-focused and emotion-focused coping strategies to play a minimal role in mediating the relationship between appraisal and adjustment in the context of the MUIT (as discussed). One of the main goals of this modeling research is to identify variables that moderate and/or mediate pathways in this model; in particular, the pathway leading from uncertainty to mastery. Questions of interest relate to how caregivers living in a state of uncertainty gain a sense of mastery, and how a sense of mastery (or challenge) reduces caregiver distress. To attempt to answer these questions, we plan to expand the MUIT in several ways, assessing the impact of a variety of risk and resilience variables on this model. Many of the constructs identified as „moderator variables associated with caregiver emotional maladjustment‟ in the early section of this introduction will be assessed in this regard. Most of these variables have not been assessed in relation to the MUIT, and there is a strong demand for research on them  30 (Crowell & Strahlendorf, 2007). Of note, some of those „moderator variables‟ previously discussed will not be assessed in the model. For example, family stressors such as „parenting style‟ and „parenting efficacy‟ will not be measured but are considered worthy of attention in future research. While other family stressors such as „chronic disease in a family member‟, „recent family death‟, „marital dissatisfaction‟, „family communication problems‟, and „positive stressors‟ (e.g., birth of a child) will not be specifically assessed, „other stress‟ is a variable studied here and one that captures these stressors in one common variable. In addition, „child hospitalization status‟ will not be assessed, and only crude indicators of „child prognosis‟ (i.e., severity of cancer, as indicated by „low‟, „moderate‟, or „high‟ risk) and „SES‟ (i.e., number of income earners in household) will be used. Moreover, while various types of social support will be assessed, emotional support specifically provided by those who have had (or currently have) experience with pediatric cancer (i.e., „peer support‟) will not be assessed. Lastly (and as previously discussed), specific emotionfocused and problem-focused „coping strategies‟ will not be measured, so as to avoid the problems of ambiguity and inconsistency typically associated with coping research. Nonetheless, many of the variables that will be assessed (e.g., emotional support, informational support, instrumental support, hope, benefit finding) will tap into psychological processes that overlap with these coping constructs. Hypothesized Model Our proposed model in its entirety is displayed in Figures 2 (basic format) and 3 (SEM format). As one can see, in the absence of „effective‟ coping strategies we anticipate challenge appraisals to consistently predict positive emotional outcomes and threat appraisals to consistently predict negative emotional outcomes. According to Mishel & Sorenson (1991; see Figure 1), high perceptions of control (i.e., mastery) predict (the more advantageous) challenge appraisal. One of the questions we then sought to answer was: „what enhances perceptions of  31 control?‟ Based on previous research (Janda et al., 2006; Tennen et al., 1992), perceiving important (family and child) needs as being met and thinking positively (in the form of dispositional optimism and/or benefit finding) is hypothesized to enhance perceptions of mastery in our model. In contrast, experiencing negative emotions (like sorrow and anger) and perceiving child struggles (in such areas as physical, emotional, and social functioning) is hypothesized to reduce perceptions of mastery. Based on the MUIT, it is presumed that perceptions of mastery in turn make one more likely to perceive their demanding situation as a personal challenge (rather than a threat). It is further hypothesized (based on research by Christman, 1990 and Stanton et al., 2002) that perceptions of challenge predict feelings of hope. And, based on Snyder‟s (1994) theory of hope-dependent „goal-pursuit‟ as well as Fredrickson‟s (2002) „broaden-and-build‟ theory of positive emotions, hope is hypothesized to reduce emotional distress. Alternatively, a lack of perceived mastery is anticipated to predict appraisals of „threat‟ (Mishel & Sorenson, 1991) and, in turn, heightened emotional distress. Emotional distress, for our purposes, represents depressive symtomatology. Consistent with other research related to adjustment to cancer and the field of positive psychology, „subjective happiness‟ has been added as a second dependent variable in this study. As such, a parallel model to that just described is proposed in which subjective happiness has been substituted in place of emotional distress. Figures 4 (basic format) and 5 (SEM format) display this model. Note that directions of influence are therefore changed in certain pathways to account for this positively valenced construct. Our proposed model differs from the MUIT in several ways. In addition to removing some variables and incorporating new ones, our model makes a clear distinction between the adaptive „challenge‟ pathway and the maladaptive „threat‟ pathway. [Recall, in the MUIT there  32 were several pathways to distress reduction due to consideration of post-appraisal coping strategies.] The proposed model also differs from the MUIT with respect to its exploration of „positive thinking‟. While Mishel & Sorenson (1991) explored the impact of „focusing on the positive‟ as a coping strategy, the proposed model examines the impact of more passive positive thinking, in the form of optimism and benefit finding. Secondary Analyses A secondary goal of this research is to explore (via basic correlation and regression analyses) relations among variables measured in a group of adolescent patients/survivors and in a group of on-treatment caregivers measured longitudinally. Adolescent patients/survivors.  Despite being in a vulnerable position, research shows  pediatric patients/survivors are adjusting quite well (Phipps, 2005). Analyses in the proposed research are therefore concerned with elucidating variables associated with superior adjustment. Such resilience-oriented research is highly encouraged in this population (Phipps, 2007; Stewart & Mishel, 2000), particularly that pertaining to benefit-finding. In the current study, benefit finding will be examined along with measures of „important needs‟ and QOL. Information from this type of research can be useful to the extent that it provides information on ways to optimize hardiness in pediatric cancer patients and survivors. Caregivers measured longitudinally.  Six-month follow-up data (on a subset of  baseline measures) will also be examined in caregivers who have children on treatment „for the first time‟ at baseline. The purpose of this additional research is to see how positive thinking and emotional adjustment change across the treatment trajectory in those vulnerable persons dealing with pediatric cancer for the first time. In addition, using baseline (i.e., time one) measures in regression analyses predicting longitudinal (i.e., time two) measures will allow for the opportunity to assess „change scores‟. Such information can also be used to support interpretations and conclusions about directionality of influences in our SEM analyses.  33 Objectives of Study In this study, the following research goals will be addressed: (i)  Through modeling analyses, identify new moderating and mediating variables influencing caregiver emotional adjustment. Of particular interest is: (1) identifying which resilience variables are able to reduce the negative impact of uncertainty on mastery; and (2) determining whether hope mediates the relationships that mastery and challenge appraisals have with each of the dependent variables (i.e., emotional distress and subjective happiness).  (ii)  Through descriptive analyses, determine prevalence levels of all variables studied in caregivers; in particular, for benefit finding and hope.  (iii)  Identify correlates and predictors of benefit finding and hope in caregivers. [Determining whether dispositional optimism is a predictor/moderator of benefit finding is of particular interest.]  (iv)  In regression analyses, determine: (1) the degree to which the various independent variables studied collectively explain variance in our main dependent variables (i.e., distress and subjective happiness); and (2) the relative importance of the independent variables in predicting these dependent variables. [Of particular interest here is elucidating the significance of treatment status and urban/rural status.]  (v)  As a secondary objective, explore relationships among the adolescent self-report variables studied as well as between corresponding variables measured in adolescents and caregivers.  (vi)  As a secondary objective, explore relationships among caregiver variables assessed at six months‟ follow-up, as well as between caregiver baseline and follow-up variables.  34 Methods Research Design The British Columbia Children‟s Hospital (BCCH) Oncology program serves children and families within all of British Columbia, including the (approximately) 120 patients newly diagnosed with cancer each year in this province (Canadian Cancer Society, 2008). For this study, caregivers with a child (2 to 12 years old) or teenager (13 to 17 years old) being treated or followed for cancer through BCCH were recruited. Adolescent patients/survivors were also recruited for some exploratory analyses. We chose to focus on teenage patients because they represent the youth demographic most at risk for emotional turmoil, and because they (possibly) are more able (relative to pre-teens) to comment on abstract constructs such as benefit finding. The caregiver sample was composed of equal numbers of individuals from „urban‟ (i.e., less than 90 minutes travel time to BCCH) and „rural‟ (i.e., 90 minutes or more of travel time) areas. Despite rural families representing a sizable portion of Canada‟s population (30.4% in Canada, and 37% in BC; Beshiri & Bollman, 2001), they are often neglected in psycho-oncology research. By assessing equal numbers of urban and rural caregivers, we aimed to give sufficient research attention to the impact of commuting for cancer treatment. Our sample was also composed of equal numbers of caregivers with children ontreatment and off-treatment. „Off-treatment‟ families had completed treatment at least one month prior but no more than 10 years prior. Some were cancer-free and some were not. By assessing equal numbers of caregivers at two different stages of the disease process, we hoped to shed light on how caregiver emotional adjustment (and relations among caregiver psychosocial variables) changes as a function of time along the disease trajectory. This research involves an observational design based on purposive sampling. Relations between variables were observed, and participants were selected based on a shared variable (i.e., the pediatric cancer experience) and on variables over which they vary (i.e., urban/rural status,  35 treatment status). [Utilizing these comparison groups allowed control over factors presumed to have a strong influence on caregiver distress.] Data were collected cross-sectionally (in caregivers and teens) and longitudinally (in a subset of caregivers). Emotional distress and subjective happiness served as the main dependent variables in the caregiver analyses. Emotional distress was chosen as a dependent variable because it was the dependent variable in Mishel & Sorenson‟s (1991) original model. Subjective happiness was utilized as an additional dependent variable to offer a positive outcome measure for comparison with emotional distress. Both these variables are viewed as superior to a QOL variable in the context of modeling research for reasons previously discussed. Caregiver independent variables were conceptualized to fall into four categories: precursor variables, negative predictor variables, positive predictor variables, and neutral variables. Precursor variables consisted of child and caregiver socio-demographic indicators as well as cancer-specific information (see Table 1). Negative predictors consisted of those variables anticipated to (positively) correlate with negative caregiver affect. These included „long-term uncertainty‟, „sorrow & anger‟, „guilt & worry‟, „threat appraisal‟, and „perceived struggles in child QOL‟ (in domains of physical, emotional, social, and school functioning). Positive predictors consisted of those variables anticipated to (positively) correlate with positive caregiver affect. These included „optimism‟, „mastery‟, „benefit finding‟ (and related subdomains), „hope‟, „proportion of (important) family needs met‟ – total score (and related subdomains), „proportion of (important) child school and resource needs met‟, and „challenge appraisal‟. Lastly, neutral variables consisted of those remaining variables that did not fall into any of the previous categories. These included „proportion of family needs considered important‟ and „proportion of child school and resource needs considered important‟. Independent variables in the adolescent dataset included „benefit finding‟ as well as  36 „proportion of (important) family needs met‟ (across domains of health information, emotional support, instrumental support, professional support, community support, and involvement with care). „QOL‟ was the lone dependent variable. This was viewed as an appropriate outcome variable for these supplementary teen analyses because (i) little item redundancy was anticipated between this measure and the independent variable measures being used; and (ii) a „positive‟ outcome variable is required in research focused on resilience. In the longitudinal (caregiver) dataset, independent variables consisted of all caregiver variables measured at baseline. „Mastery‟ measured at six-month follow-up also served as an additional independent variable. Meanwhile, the two dependent variables were the six-month follow-up measurements on „emotional distress‟ and „subjective happiness‟. Inclusion and Exclusion Criteria One caregiver per child was recruited for this study. Inclusion criteria required that caregivers were 18 years of age or older, able to read and write in English, and the primary caretaker (>50% effort) of their child. Their child could not be terminally ill and he/she must have been between the ages of 2 and 17 at the time of study participation. Diagnosis of cancer must also have occurred within the past ten years (1999 to 2009). These inclusion criteria derived largely from Chung et al.‟s (2005) Family Needs Study, which this study ran in parallel with. In addition to acquiring data from caregivers, data were also recruited from children between the ages of 13 to 17. These adolescents were required to be able to read and write in English and they were excluded if they had a disabling illness (unrelated to cancer), so as to avoid the confounding effect of physical comorbidity. Permission to contact families was sought from the child‟s treating oncologist, who was able to comment on exclusion criteria concerning English proficiency, palliative status, and presence of comorbid physical illness (in child).  37 Measures The following measures were carefully selected based on appropriate construct validity, acceptable instrument reliability, and with the aim of minimizing redundancy (and thus participant burden) in the questionnaire package. The large number of measures selected reflects the nature of SEM research. All measures are listed with their respective item totals in Table 2. Measures Pertaining to Variables in the MUIT Given that our proposed model is based on the MUIT, the following measures were selected for their ability to measure relevant variables in this regard. Mastery scale (MS).  The MS (Pearlin & Schooler, 1978) is a 7-item measure of  perceived mastery or „control‟ over life. It assesses the extent to which an individual regards their life as being under their control in contrast to being fatalistically determined. Responses are provided on a 4-point Likert scale ranging from 1 (Strongly disagree) to 4 (Strongly agree), with some items being reverse scored. A higher overall score reflects a greater sense of mastery. Sample items include “What happens to me in the future mostly depends on me” and “I often feel helpless in dealing with the problems in life”. The MS has been shown to exhibit reasonable internal reliability in the range of α=.75 (Folkman et al., 1986) and good construct validity. More specifically, it has been positively correlated with self-esteem (Pearlin & Schooler, 1978) and inversely related to a measure of (abnormal) psychological symptoms (Folkman et al.). Center for epidemiological studies depression scale (CES-D).  The CES–D (Radloff,  1977) is a widely used self-report scale designed to measure emotional distress over the past week. It contains 20 items that measure depressive symptoms across four domains: Somatic Retarded Activity (7 items), Depressed Affect (5 items), Positive Affect or Well Being (4 items), and Interpersonal Affect (2 items). Items are completed on a 4-point Likert scale ranging from 0 (Rarely or none of the time) to 3 (Most or all of the time), with some items being reverse scored. Higher scores indicate more distress, and a score of 16 indicates clinically elevated depressed  38 symptomatology. Examples of items include, “I could not get going” and “I felt depressed”. Space was provided at the bottom of this questionnaire so that participants could indicate if “anything especially good or especially bad happened this week”. The construct validity, testretest reliability (r=.45 to r=.70), and internal consistency (α=.85) of the CES–D are all good. The CES–D is appropriate for use in clinical research and it is also recognized as a useful screening measure (Yang, Soong, Kuo, Chang, & Chen, 2004). Cognitive appraisal of health scale – revised (CAHS-R). The CAHS-R (Ahmad, 2005) is a 13-item measure of stress appraisal associated with a health-related problem; in this case, the child‟s cancer experience. Items of the CAHS-R relate to 3 factors (or types of appraisal): Challenge (3 items), Threat (5 items), and Harm/Loss (5 items). Sample items include “There is a lot I can do to overcome this health problem” (challenge), “I have a lot to lose because of this health problem” (threat), and “This health problem has damaged my life” (harm/loss). A 5-point Likert scale is used to measure responses, which range from 1 (Strongly disagree) to 5 (Strongly agree). Scores are calculated for each factor rather than a total scale, and higher scores indicate greater agreement with that appraisal type. The CAHS (Kessler, 1998) was initially developed in a sample of breast cancer patients. While the original scale had four factors, psychometric evaluation by Ahmad (2005) in a sample of prostate cancer patients showed that a revised 3-factor version of the scale was better supported by exploratory and confirmatory factor analyses. Using a 19-item measure of these three factors, Ahmad, Musil, Zauszniewski, & Resnick (2005) reported alphas of .71, 74, and .84 for the challenge, threat, and harm/loss appraisals, respectively. Factor analysis validation by Ahmad (2005) of an even shorter 13-item measure concluded that this shorter scale was more parsimonious, equally valid, and acceptably reliable (challenge: α=.70; threat: α=.74; harm/loss: α=.79). While the proposed study collected data on all 13 items of the CAHS-R, the harm/loss  39 subscale was not analyzed due to redundancy with other measures. In addition, we sought consistency with the MUIT which incorporates just two appraisal factors. Parent experience of child illness (PECI).  The PECI (Bonner et al., 2005) is a 25-item  measure of parental adjustment related to caring for a child with a chronic illness. The questionnaire is composed of 4 factors, and validation has occurred with respect to these separately, rather than the full-scale score. Five of the items fall under the factor of Long-term uncertainty (α = .80) which assesses future implications of past decisions and ruminative concerns about the child‟s future well-being (e.g., “I worry about whether my child will be able to live independently as an adult”). Eleven of the items fall under the Guilt & Worry factor (α = .89) which taps into personal guilt as well as concerns about treatment decisions, the child‟s current well-being, and disease etiology (e.g., “I worry that I may be responsible for my child‟s illness in some way”). Eight of the items fall under the Sorrow & Anger factor (α = .86) which measures feelings of loss and anger over having to experience the chronic illness of a child (e.g., “I am jealous of parents who have healthy children”). Lastly, five of the items tap into the Emotional Resources factor (α = .72) which assesses feelings of competence and self-efficacy (e.g., “I feel ready to face challenges related to my child‟s well-being in the future”). While data were collected on all 25 items, the „emotional resources‟ subscale was not analyzed due to literal redundancy with other measures. Responses on the PECI are rated on a 5-point Likert scale ranging from 0 (Never) to 4 (Always), with higher scores on the first three subscales indicating greater angst. In addition to its acceptable reliability, construct validity of the individual factors has been demonstrated through positive correlations with several established measures of parent adjustment (Bonner et al., 2005). The PECI was validated in parents of pediatric oncology patients and it is applicable for caregivers of children of all ages as well as those both on and off treatment.  40 Positive Psychology Measures Given our goal of exploring resilience variables in caregiver and patient/survivor populations, the following measures were used to assess positive psychology variables of interest. Benefit finding scale (BFS).  The BFS (Kim et al., 2007) is a 17-item measure of the  degree to which caregivers of cancer patients find meaning and perceive personal growth through their caregiving experience. BFS items relate to six factors: Acceptance (3 items), Empathy (4 items), Appreciation (3 items), Family (2 items), Positive Self-View (3 items), and Reprioritization (2 items). The stem for each item is “Having provided care to someone with cancer has…” and sample items include “Taught me how to adjust to things I cannot change” and “Helped me become more focused on priorities, with a deeper sense of purpose of life”. A 5point Likert scale is used to measure responses, which range from 1 (Not at all) to 5 (Extremely). The BFS composite score has been shown to relate positively with life satisfaction (Kim et al., 2007). It has also shown good internal consistency (α = .95) as have the individual subscales (α = .78 - .90). In the proposed study, analyses were performed on the total score as well as on the factor domain scores. An „other‟ line added to the bottom of the measure gathered information on additional benefits gained through the caregiving experience, which were not assessed by the BFS items. Benefit finding scale for children (BFSC).  The BFSC (Phipps et al., 2007) is a  unidimensional 10-item measure of benefit finding in children that was piloted in a sample of children with cancer. Its items relate to potential benefits that children may gain during their illness experience. The stem for each item is “Having had my illness…” and sample items include “Has helped me become a stronger person” and “Has taught me to be more loving of others”. A 5-point Likert scale is used to measure responses, which range from 1 (Not at all true for me) to 5 (Very true for me). In addition to the 10 items, an „other‟ line was added to the  41 bottom of this measure to inquire about additional benefits gained through the cancer experience that were not assessed by the BFSC items. The BFSC has demonstrated good internal consistency (α = .83) and it has shown positive correlations with optimism (r=.29) and selfesteem (r=.20), a negative correlation with anxiety (r=-.16), and no significant relations to posttraumatic stress, social desirability, or defensiveness (Phipps et al.). Life orientation test - revised (LOT-R). The LOT-R (Scheier, Carver, & Bridges, 1994) is a 10-item scale that measures the stable personality variable, dispositional optimism. It consists of 4 filler items and 6 scale items. The total score is calculated by summing the three positively phrased items and the three negatively phrased (reverse-coded) items. Respondents are asked to indicate their level of agreement with each of the items on a 5-point Likert scale, ranging from 0 (Strongly disagree) to 4 (Strongly agree). Higher scores indicate more optimism. Sample items include, “I‟m always optimistic about my future” and „If something can go wrong for me, it will”. Initial validation work with the LOT-R in an undergraduate population found acceptable values for internal reliability (α = .78) as well as 28-month test-retest reliability (r=.79). Discriminant and convergent validity was also demonstrated by moderate correlations with measures assessing neuroticism, trait anxiety, self-esteem, and self-mastery. Of note, while optimism and self-mastery are conceptually similar, these constructs have been shown to be empirically distinct (Marshall & Lang, 1990). Herth hope index (HHI).  The HHI (Herth, 1992) is a 12-item measure of perceived  hope. It is a shortened version of the much longer Herth Hope Scale (HHS; Herth, 1991) that has been adapted for use with research and clinical populations. Its items relate to three subscales: Temporality and Future (4 items), Positive Readiness and Expectancy (4 items), and Interconnectedness (4 items). Responses are provided on a 4-point Likert scale, with answers ranging from 1 (Strongly disagree) to 4 (Strongly agree). Negative items are reversed scored so  42 that higher scores reflect a greater sense of hope. Sample items include, “I have short and/or long range goals” (temporality and future), “I believe that each day has potential” (positive readiness and expectancy), and “I have a faith that gives me comfort” (interconnectedness). In the validation study of the HHI, which was tested with a large convenience sample of ill adults, the alpha coefficient was 0.97 and a 2-week test-retest reliability of 0.91 was found. The HHI was also found to be positively correlated with the HHS (r = 0.92) and a measure of existential wellbeing (r = 0.84), and it negatively correlated with a measure of hopelessness (r = -0.73). The HHI has been used extensively with cancer populations, including patients from urban and rural populations (Howat, Veitch, & Cairns, 2006) and family caregivers of patients with advanced cancer (Duggleby et al., 2007). It has also been used in studies assessing uncertainty and appraisal (Wonghongkul et al., 2000). Subjective happiness scale (SHS). The SHS (Lyubomirsky & Lepper, 1999) is a 4item measure of global subjective happiness. Its items assess self-characterizations of how happy one is as well as happiness ratings (of oneself) relative to peers. Responses are provided on a 7point Likert scale, with higher scores indicating greater happiness. The SHS has been validated in 14 studies with a total of 2,732 participants, ranging from high school students to community adults. Results have indicated that the SHS has high internal consistency (ranging from α=.79 to α=.94). Test-retest correlations (ranging from r=.55 to r=.90 across time periods ranging from 3 weeks to 1 year) have suggested good to excellent reliability, and positive correlations with selfesteem (r=.53 to .58) and informant ratings of happiness (r=.65) among other positive variables attest to the validity of this scale as a measure of „subjective happiness‟. Measures Pertaining to Family and Child-Specific Needs Given the literature highlighting the significance to cancer families of personal needs, health information needs, and patient care and resource needs, we felt it important to assess the types of needs that caregivers and patients find most important, and the degree to which these are  43 perceived as being met. Family needs questionnaire (FNQ).  The FNQ (Kreutzer & Marwitz, 1989) is an  instrument that measures unique needs in families experiencing chronic illness. The caregiver version has 42 items while the adolescent patient version has 40 items. Factor analytically derived subscales include: Health Information (9 items), Emotional Support (8 items), Instrumental Support (6 items), Professional Support (5 items), Community Support Network (5 items), and Involvement with Care (3 items). There are also 4 to 6 filler items. Respondents are asked to make two independent ratings. In part I, they rate the importance of each need on a four-point Likert scale ranging from 1 (Not important) to 4 (Very important). In part II, they rate the extent to which each need has been met, with response options consisting of “Yes”, “No”, or “Partly”. Item examples include “I need to be told about all changes in my child‟s/my medical status” and “I need to have my significant other/parent understand how difficult it is for me”. Space is also left at the end of the questionnaire where respondents can list additional needs not assessed by the FNQ items. In the original factor analysis study, an acceptable reliability (α = .75) was found for the total score (Kreutzer, Serio, & Bergquist, 1994). Need factors were also found to correlate with dimensions of social support (Wortman & Silver, 1989). Studies of family adjustment have provided support for the FNQ‟s validity and clinical utility in assessing met and unmet needs as well as in assessing sensitivity to change over time (Witol, Sander, & Kreutzer, 1996). The FNQ subscale scores have consistently demonstrated acceptable internal consistency ranging from α =.78 to .89 (Serio, Kreutzer, & Witol, 1997). In the current study, data were gathered on two variables: proportion of needs considered „important‟ (i.e., „IMP‟, as indicated by responses of „3‟ or „4‟ on part 1 of the questionnaire) and proportion of important needs perceived to be „met‟ (i.e., „MET‟, as indicated by a response of „yes‟ in part 2 of the questionnaire). Both caregivers  44 and adolescent patients completed this measure, and analyses were performed on the total score and across factor domain scores. School needs/resources questionnaire (SNRQ).  The SNRQ (Chung et al., 2005) is a  23-item checklist that measures unique school needs and resources of children with a chronic medical illness. It was created for use in Chung et al.‟s (2005) Family Needs Study, the research project that this study ran in parallel with. Items relate to a variety of professional assessments and interventions, including resources related to neuropsychology, occupational therapy, physiotherapy, speech and language therapy, and vocational therapy. Family members are asked to make two independent ratings. First, an indication of the importance of each perceived need is rated on a four-point scale ranging from 1 (Not important) to 4 (Very important). Second, the family member rates the extent to which each need has been met, with response options consisting of “Yes”, “No”, “Partly”, or “Don‟t know”. Item examples include “My child needs consultation between occupational therapy and school staff or family” and “My child needs an Individualized Education Plan (IEP)”. Similar to the FNQ, data were gathered on two variables: proportion of needs considered „important‟ (i.e., „IMP‟, as indicated by responses of „3‟ or „4‟ on part 1 of the questionnaire) and proportion of important needs perceived to be „met‟ (i.e., „MET‟, as indicated by a response of „yes‟ in part 2 of the questionnaire). Other Measures Given that many child and adult characteristics (including socio-demographic and cancerspecific variables) have been identified as significant moderators of caregiver distress, we felt it essential to include the following measures in our study. Pediatric quality of life inventory (PedsQL).  The PedsQL 4.0 (Varni, Seid, & Kurtin,  2001) is a 23-item instrument that measures health-related QOL in youth aged 2 to 18 with acute or chronic health conditions. The total scale measures 4 health dimensions: Physical (8 items), Emotional (5 items), School (5 items), and Social (5 items). In the parent proxy form, caregivers  45 answer questions such as “How much of a problem has your child had with low energy level?” and “How much of a problem has your child had with getting along with other children?”. The measure specifically asks about functioning over the past month. A 5-point Likert scale is utilized, with responses ranging from 0 (Never a problem) to 4 (Almost always a problem). In addition to the caregivers completing the parent proxy form, the teenage patients completed a self-report version of this measure. In the original validation study of the PedsQL (Varni, Seid, & Kurtin, 2001), internal consistency of the total scale was shown to be strong (α =.80 for child self-report; α =.88 for parent proxy), and internal consistency reliabilities across the four subscales were acceptable (α‟s =.73 to .94). Test-retest reliability was shown to be in the range of .82 to .92. Both selfreport and parent proxy forms yielded significant correlations between the total scale and such outcome variables as „number of days child was too ill to pursue normal activities‟ and „number of days child needed someone to care for him or her‟. This total scale score also distinguished between healthy children and children with chronic health conditions, and the subscales were shown to be responsive to clinical change. Caregiver analyses in this study used domain scores, while the adolescent analyses assessed both the total score (23 items) and the „psychosocial health summary score‟ (15 items). This latter variable exclusively taps into the social, school, and emotional subscales of the PedsQL. It has shown acceptable internal consistency, both for the child self-report version (α = .83) and for the parent proxy (α = .86) (Varni, Seid, & Kurtin, 2001). In the adolescent analyses, items were reverse scored so that higher scores indicated fewer problems. Socio-demographic form. Social demographic and cancer-specific information was requested from caregivers using a form (see Appendix A1) specifically developed for this study. Information was gathered on each of the parent and child characteristics listed in Table 1.  46 Procedure Participants were recruited through poster/brochure advertisement (see Appendix A2) in the oncology clinic and on one of the oncology wards. In addition, eligible participants were identified through a data manager in the BCCH oncology clinic, after which a random selection process ensued. As previously mentioned, permission to contact „selected‟ caregivers was sought from the child‟s oncologist. When permission was received, a letter was sent to the family informing them of the study and alerting them that they would soon be receiving a phone call. This recruitment letter (see Appendix A3) clearly explained that participation was voluntary and that no negative consequence to the family would result from choosing not to participate or from deciding to withdraw from the study at a later time. Contact information and further details about study participation were also included in the recruitment letter. Two weeks following mailing of the recruitment letters, caregivers were called (by a research assistant) to inquire about their interest in participating and to answer any questions about the study (see sample script in Appendix A4). Regardless of the decision on whether or not to participate, all contacted families were entered into a grand prize draw. Once verbal consent (to participate) was obtained, questionnaire packets were mailed out. These included a written consent form for caregivers and an assent form for teens where applicable (see Appendix A5/A6), a socio-demographic data sheet (see Appendix A1), the questionnaires with respective cover sheets (see Appendix A7 to A9), and a pre-paid return envelope. A time commitment of 45 minutes or less was anticipated for caregivers while a commitment of 20 minutes or less was anticipated for teenage participants. Ordering of questionnaires was not randomized as we felt it necessary to have uplifting questionnaires at the end of the package. Caregivers filled out the demographic data sheet and caregiver questionnaire package, while adolescent participants filled out their own questionnaire package. Participant responding was anonymous. Packages were labelled with the participant‟s  47 study number as a way to monitor the database and to keep track of who had responded. Participants received two reminder calls spaced three weeks apart, starting at two weeks post mail-out. When completed packages were received, a remuneration gift was mailed to participants, thanking them for their participation. Caregivers receiving a longitudinal questionnaire package also received a phone call six months following completion of their first package, prompting them to complete the remaining package. They later received two reminder calls spaced three weeks apart, starting at two weeks post the telephone „prompt‟. When completed packages were received, another remuneration gift was mailed to participants, thanking them for their participation. Data collection for this study took one year to complete. All data were reviewed and analyzed at BCCH. Approval to run this study was obtained from both the UBC and BCCH institutional review boards.  48 Results Data Sets Analyses from this study are based on three datasets. The baseline dataset is the largest (N=156) and it contains information collected from our caregivers at the first time point. It includes caregiver self-report data as well as socio-demographic (and cancer-specific) information on both caregivers and their children. The adolescent dataset (N=40) refers to all information collected at the first time point relating to our teenage participants. It includes the teen self-report data as well as relevant caregiver and teen socio-demographic (and cancerspecific) information. Lastly, the longitudinal dataset (N=41) refers to all information we have on those caregivers who provided data at both the first and second (i.e., six-month follow-up) time points. Similar to the baseline dataset, it contains caregiver self-report data as well as sociodemographic (and cancer-specific) information on both caregivers and their children. Preliminary Data Screening All data were screened for cases of outliers, missing data, restricted range, and nonnormality. A handful of outliers were found across a few socio-demographic variables (i.e., time to hospital, number of children, and number of days school missed). Given that these outliers were suspected to have been accurately sampled from the target population, they were not deleted. Rather, outlying scores were changed to be less deviant than their original values, though still at the far end of the distribution. [This approach is considered a viable option (Tabachnick & Fidell, 2001) for dealing with occasional univariate outliers, defined as z-scores greater than 3.29 (p<.001, two-tailed test). Following changes made by this approach, all variable distributions had z-scores less than 3.29.] Missing data were minimized in this study through mandatory telephone follow-up calls to participants returning incomplete packages (as noted in the informed consent documents). Nonetheless, one questionnaire (i.e., caregiver PedsQL) had a substantial proportion of non-  49 random missing data. Given that the „school functioning‟ section of this questionnaire did not apply to caregivers of either very young children or long-term hospitalized children, it was often left blank. The „school functioning‟ domain pertaining to this (caregiver) measure was therefore removed from further analyses. Several socio-demographic dichotomous variables were observed to have extreme splits, defined by Rummel (1970) as equal to or greater than a 90-10 split between categories. Where extreme, these are highlighted in the descriptive statistics data tables (see Tables 3 to 7). Due to the extreme non-normality of these distributions, these variables were not used in further (statistical) analyses. Exceptions to this were made for two variables, caregiver gender and relapse status. Although these variables displayed slightly elevated splits, they were retained for consideration in further analyses due to their importance. Analyses involving these variables are interpreted with caution. To assess (non-) normality in our variables, histograms were reviewed and z-scores were calculated to determine the significance of kurtosis and skewness across all distributions. Significant z-scores (i.e., surpassing the critical +/- 3.091 threshold) indicating non-normality are highlighted in the descriptive statistics data tables (see Tables 3 to 7). The following categorical variables showed non-normal distributions: relationship to child, type of relationship (to child), mode of travel (to hospital), and type of cancer. These variables were removed from further statistical analyses as it was believed transformation of these variables would yield uninterpretable data. For the non-normal quantitative data, a combination of square root, cubed root, logarithm, inverse, and sine/arcsine transformations were attempted; 20 in the baseline dataset, 3 in the teen dataset, and 5 in the longitudinal dataset. Socio-demographic and cancerspecific variables that underwent transformation to reduce non-normality are identified in the descriptive statistics data tables (see Tables 3 to 7). Questionnaire variables that underwent  50 transformation to reduce non-normality are identified in the correlational and inferential statistics data tables (see Tables 30 to 56). Of the eight quantitative variables that remained non-normal following these transformations, only one (longitudinal cancer severity) showed a kurtosis/skewness score greater than 2, indicating extreme non-normality. While this variable was removed from further analyses, the remaining non-normal variables were retained (with caution noted). All of these variables were from the caregiver baseline dataset, and they included: caregiver education, household income, child age, family needs-health information MET, family needs-emotional support MET, family needs – involvement with care MET, and family needs – professional support MET. Participants Baseline Sample Caregivers.  Of the 248 caregivers who were approached during a 1-year period, 205  (82.7%) agreed to participate. There were no identifiable differences between those who participated and those who did not. The most common reasons for refusing to participate included poor English skills and lack of time. Of the 205 caregiver participants, 156 (76.1%) successfully returned completed packages (139 mothers, 14 fathers, 3 grandparents). Caregiver participants were primarily Caucasian, female, middle-aged, educated, and biologically related to the child with cancer. Most participants had more than one child in their care, and just over half of the sample identified themselves as spiritual/religious. The majority of participants were married, and 60.9% of the sample came from a dual income household. As selected for, the sample was evenly divided across urban (50%) and rural (50%) caregivers and according to treatment-status (50% on-treatment, 50% off treatment). Just under half (46.2%) of the sample had children on treatment for the first time, and only 26.9% of the sample had children currently receiving „active treatment‟ (i.e., aggressive treatment during the acutely ill phase). Transport time to the hospital ranged from 5 minutes to 630 minutes, with median time  51 for the urban subgroup being 45 minutes and median time for the rural subgroup being 240 minutes. The majority of commuters endorsed driving as their mode of travel, and just over half of the overall sample described the commute as either „sometimes difficult‟, „somewhat difficult‟, or „very difficult‟ (55.1%). The majority (69.9%) of caregivers experienced supplemental stress beyond the standard cancer experience, related to such things as physical/mental/family stress, child physical/emotional/learning struggles, progression of child‟s cancer, and previous experience with cancer. Descriptive statistics on these caregiver sociodemographic (and treatment-related) variables are displayed in Table 3. Pediatric patients/survivors.  Socio-demographic and cancer-specific data on the  pediatric patients/survivors of the caregivers just described are displayed in Table 4. The majority of these youth were Caucasian and just over half were identified as spiritual/religious. Median age was 10 years old and the sample was evenly divided according to gender. The primary cancer type was Leukemia (53.2%) and the majority of cancers (76.3%) were considered „moderate risk‟ according to 5-year prognosis estimates (National Cancer Institute Kramarova et al., 1996). Approximately half of these children had been diagnosed within the past 2 years and median age of diagnosis was 6 years old. Treatment methods included: chemotherapy (93.6%), surgery (47.4%), radiation (18.6%), and bone marrow transplant (7.7%), with exactly half of the sample receiving some combination of these. A small portion (3.8%) of patients received other forms of treatment (e.g., focal cryo/laser salvage therapy, antibody basedtherapy, Imantinib, Retinoic Acid). Time since last treatment ranged from <1 month to 120 months, with 87.2% of youth last being treated within the past 5 years. Physical sequelae from treatment were rare (9% visually impaired, 4.5% hearing impaired, 8.3% motor impaired), though the majority of these children missed some of their schooling during treatment (70.5%). Of those missing school, an average of 157 days were missed. Nonetheless, only four children in  52 this sample (2.6%) had to repeat a grade. Adolescent Sample Of the 54 teenagers who had parents participating in the study and thus were sent the adolescent questionnaires, 40 (74%) returned completed packages (24 females, 16 males). These adolescent participants were primarily Caucasian and nearly two-thirds were off-treatment. Most of the on-treatment patients were fighting cancer for the first time, though roughly just half were considered on active treatment. Median age was 15 years old and Leukemia was the most common type of cancer (42.5%). The majority of these teenagers had siblings as well as married parents. Most had not experienced supplemental stressors (i.e., metastasis, learning disability, mental health struggles, physical struggles). Half of these teenagers were identified as spiritual/religious and 60% were from an urban area. Just under 50% had been diagnosed within the past 3 years and at an age of 12 or older. The majority of these teens had been treated within the past two years, with chemotherapy being received by most and just over half also undergoing surgery. Descriptive statistics on these adolescent socio-demographic and cancer-specific variables are displayed in Table 5. Longitudinal Sample Caregivers.  Of the 72 caregivers in the baseline sample who had children on treatment  „for the first time‟, 41 (56.9%) returned completed longitudinal packages at six months followup (37 mothers, 3 fathers, 1 grandparent). These caregivers were primarily Caucasian, female, middle-aged, educated, and biologically related to the child with cancer. Most had more than one child in their care, and two–thirds (65.9%) identified themselves as spiritual/ religious. The majority of these caregivers were married and 61% came from a dual income household. The sample was fairly evenly divided across urban (48.8%) and rural (51.2%) status and just over two-thirds (68.3%) had children that were still receiving treatment at six months‟ follow-up. Just over half (56.1%) of these caregivers had children who had been on active treatment at baseline.  53 Transport time to the hospital ranged from 5 minutes to 420 minutes. The majority of commuters endorsed driving as their mode of travel, and 61% described the commute as either „sometimes difficult‟ or „somewhat difficult‟. Again, most (78%) had experienced supplemental stress beyond the standard cancer experience. Descriptive statistics on these caregiver sociodemographic (and treatment-related) variables are displayed in Table 6. Pediatric patients/survivors.  Socio-demographic and cancer-specific data on the  pediatric patients/survivors of the caregivers just described are shown in Table 7. The majority of these youth were Caucasian and median age was 9 years old. Just over half were male (56.1%) and identified as non-spiritual/religious (58.5%). The primary cancer type was Leukemia (63.4%) and the majority of cancers (85.4%) were considered „moderate risk‟ according to 5-year prognosis estimates (National Cancer Institute - Kramarova et al., 1996). Over half of the sample (61%) had been diagnosed within the past year and median age of diagnosis was 7 years old. In addition to chemotherapy (100%), treatment methods included surgery (48.8%), radiation (14.6%), and bone marrow transplant (2.4%). Physical sequelae from treatment were rare (4.9% visually impaired, 4.9% hearing impaired, 7.3% motor impaired), though the majority of these children missed some of their schooling during treatment (78.9%). Of those missing school, an average of 171.5 days were missed. None of these children had to repeat a grade. Socio-demographic Data: Descriptive Statistics and Group Comparisons In addition to the descriptive statistics presented in Tables 3 through 7, chi-square tests of independence were performed on original data to investigate whether subgroups of our baseline (caregiver and child data) and adolescent samples significantly differed on any sociodemographic or cancer-specific information. Analyses were performed across on-treatment versus off-treatment groups as well as urban versus rural groups. Tables 8 and 9 display significant results from these analyses.  54 Baseline Sample When the baseline sample was divided according to urban/rural status (see Table 8a), the urban subgroup displayed significantly higher numbers of male caregivers, participants from diverse ethnic backgrounds, and commuters who travel to the hospital via car (especially their own). This subgroup also tended to have older children and children who had been diagnosed at relatively older ages. Meanwhile, the rural subgroup showed significantly higher numbers of participants using bus, plane, and ferry as their mode of transportation to the hospital, and they tended to describe transportation as more difficult. When this sample was divided according to treatment status (see Table 8b), on-treatment caregivers reported significantly more „other‟ stressors and a shorter time since diagnosis. In addition, their children tended to have a later age at diagnosis and they were more likely to have received chemotherapy. Adolescent Sample When the adolescent dataset was divided into urban versus rural subgroups, no significant differences in socio-demographic variables were found. However, when it was divided according to treatment status, the off-treatment subgroup displayed a significantly younger age at diagnosis and a longer time since diagnosis (see Table 9). Questionnaire Data: Descriptive Statistics and Group Comparisons Descriptive statistics (including an index of internal consistency), independent samples ttests, and two-way analysis of variance (ANOVA) statistics on the questionnaire data pertaining to all three datasets are presented in Tables 10 to 26. Effect sizes are also listed in some cases. For each questionnaire, ANOVA analyses were first performed comparing means across our four subgroups (i.e., urban-off treatment, urban-on treatment, rural-off treatment, rural-on treatment). Only when significant effects were found were further t-tests performed. This approach was taken to reduce the number of comparisons and thus keep power high.  55 In the baseline analyses, there was n=40 across each of the four subgroups. As there were not equal sample sizes across the subgroups in the teen and longitudinal datasets, these subsample sizes (n‟s) are indicated in Tables 21 to 26. Of note, the central limit theorem reassures us that we have sufficiently large sample sizes (i.e., at least 20 degrees of freedom for error; Tabachnick & Fidell, 2001) across all three of our datasets to create normally distributed sampling distributions. This renders both the (two-tailed) F test and the (two-tailed) t-test robust to violations of normality (Glass & Hopkins, 1996). Consequently, original (non-transformed) data were used in these analyses as this was preferred for interpretation purposes. Baseline (Caregiver) Sample (see Tables 10 to 20) Most baseline measures showed good to excellent internal consistency (i.e., α‟s: .71 to .93). However, the challenge measure (α = .46) and some of the FNQ subscale scores (α‟s: .46 to .78, inclusive) were noted to display weaker reliability. Analyses involving these variables will be interpreted with caution. When this caregiver dataset was divided into urban versus rural subgroups, no significant differences in mean scores were found across any questionnaire. In contrast, when it was divided according to treatment status, on-treatment caregivers reported significantly more unresolved sorrow & anger, guilt & worry, distress, threat, and (perceived) child physical struggles relative to off-treatment caregivers. Meanwhile, off-treatment caregivers reported a significantly higher proportion of ‘important’ school needs & resources. Average item scores were also assessed to determine prevalence levels of the thinking styles and emotional outcomes studied. These analyses revealed that caregivers tended to „agree‟ with items on the threat, hope, and mastery measures but „disagree‟ with items on the optimism measure. On the challenge measure, caregivers tended to be „neutral‟. On the subjective happiness measure, caregivers tended to rate their happiness just slightly better than average  56 (i.e., score of 4.81 on a 1 to 7 scale). Interestingly, while this average was significantly lower than a US adult community normative population average (d = .75), it did not significantly differ from a US adult female community normative population average (d = -0.01). Meanwhile, whereas off-treatment caregivers described unresolved sorrow & anger as „rarely an issue‟, on-treatment caregivers tended to rate this as „sometimes‟ a problem. A similar treatment effect was observed for guilt & worry, as on-treatment caregivers considered this „sometimes‟ a problem (in contrast to off-treatment caregivers who considered this „rarely‟ a problem). Interestingly, long-term uncertainty demonstrated a significant interaction effect in two-way ANOVA analyses, indicating this type of uncertainty remained high for urban caregivers (regardless of treatment status) but declined for rural caregivers once off treatment. This was the only significant interaction effect observed. When it came to benefit finding, caregivers tended to agree „moderately‟ with items from the appreciation subscale, while agreeing „quite a bit‟ with items on all other subscales. Of note, nearly one-fifth (17.9%) of participants endorsed realizing „other benefits‟ in their cancer experience (i.e., besides those assessed by the BFS). A sample of these „additional benefits‟ is displayed in Table 27. On the distress measure, the average total score was 15.29, falling just below the „depression‟ threshold of 16. However, on-treatment caregivers averaged a total score of 17.97 which was significantly higher than the off-treatment group average (of 12.62). This suggests on-treatment caregivers would, on average, meet threshold criteria for „clinically depressive‟ symptoms. When these scores were compared to community adult norms, the total sample (d = .58) as well as both the on- and off-treatment subgroups (d = 0.82 and d = 0.35, respectively) showed higher scores. [Of note, just over a quarter of caregivers (27.6%) indicated that they had experienced a significant (positive or negative) stressor in the past week, which may have biased  57 their distress scores in a way that made them less representative of „typical‟ functioning.] On the measure assessing (perceived) struggles in child functioning, caregivers in general tended to perceive struggles in emotional and social functioning as „almost never‟ a problem. However, as previously highlighted by a significant treatment effect, on-treatment caregivers uniquely described struggles in physical functioning as „sometimes‟ a problem. With questionnaires pertaining to family needs and school needs & resources, the proportion of items considered „important‟ (i.e., IMP) were examined. On the family needs questionnaire, caregivers classified (on average) 81% of all needs as important. Broken down by subscales, importance levels were 98% for health information needs, 86% for community support needs, 86% for involvement with care needs, 82% for professional support needs, 73% for instrumental support needs, and 64% for emotional support needs. Meanwhile, on the school needs and resources questionnaire, caregivers classified (on average) just 25% of all needs as important. Of note, off-treatment caregivers reported an average of 10% more important needs in this area than on-treatment caregivers. Table 20b breaks down the range of desirability relating to these individual child school and resource needs. Another type of variable assessed on these needs questionnaires related to the proportion of important needs considered „met‟ (i.e., MET). On the family needs questionnaire, caregivers reported (on average) that 62% of their important needs had been met. Broken down by subscales, these percentages were 79% for health information, 79% for involvement with care, 61% for community support, 57% for professional support, 47% for emotional support, and 37% for instrumental support. On the school needs and resources questionnaire, just 41% of important needs were (on average) reported met. Lastly, when a specific item assessing the need for help in maintaining hope was examined on the family needs questionnaire, 73.7% of participants were found to consider this  58 need important. However, only 58.3% of participants reported this need had been met. Adolescent Sample (see Tables 21 to 23) Despite unequal sample sizes in the adolescent ANOVA and t-test analyses, the size of the total adolescent sample (N=40) and of the subsamples (n>6) suggests that these tests are robust to violations of normality (Moore, 2000). Across the adolescent questionnaires, total scores showed very good reliability (α‟s: .83 to .95). While the PedsQL domain scores also displayed strong internal consistency (α‟s: .84 to .92), this was not the case for some of the FNQ subscales (α‟s: .12 to .82, inclusive). Analyses involving these latter variables will be interpreted with caution. Similar to the baseline (caregiver) sample, no significant differences in mean scores were found on any adolescent questionnaire across urban versus rural subgroups. However, a treatment effect was observed on a domain of the family needs questionnaire (i.e., family needsinvolvement with care MET). Specifically, on-treatment teenagers reported that a significantly higher proportion of their important involvement with care needs had been met. On the quality of life measure, teenagers tended to rate struggles across all (physical, emotional, social, and school) domains as „almost never‟ a problem. In regards to benefit finding, they tended to agree „quite a bit‟ with items across the measure. Of note, one-fifth (20%) of participants endorsed realizing „other benefits‟ in their cancer experience (i.e., besides those assessed by the BFSC). A sample of these „additional benefits‟ is displayed in Table 28. On the family needs questionnaire, these adolescents classified (on average) 75% of all needs as „important‟. Broken down by subscales, importance levels were 84% for health information needs, 84% for instrumental support needs, 78% for involvement with care needs, 71% for community support needs, 69% for professional support needs, and 63% for emotional support needs. In regards to having their important needs „met‟, adolescents reported (on  59 average) that 71% of these had indeed been met. Broken down by subscales, percentages of important needs being met were 79% for health information, 78% for involvement with care, 71% for community support, 66% for emotional support, 61% for instrumental support, and 60% for professional support. Lastly, 65.9% of adolescents considered receiving help in maintaining hope an important need, and 68.3% of these teens reported this important need had been met. Longitudinal (Caregiver) Sample (see Tables 24 to 26) Despite unequal sample sizes in the longitudinal ANOVA analyses, the size of the total longitudinal sample (N=41) and of the subsamples (n>6) suggests that these tests are robust to violations of normality (Moore, 2000). All three longitudinal measures displayed high internal consistency (α‟s: .86 to .92). Interestingly, there were no significant differences in mean scores on any longitudinal questionnaire across either urban versus rural subgroups or on-treatment versus off-treatment subgroups. [Of note, as all participants in this longitudinal sample were on-treatment at baseline, „on-treatment‟ status in these analyses refers to caregivers who were still on-treatment at six months‟ follow-up.] Similar to the baseline sample, these caregivers tended to „agree‟ with items on the mastery scale and they rated their happiness levels just slightly better than average (score of 4.82 on a 1 to 7 scale). This happiness average was again significantly lower than the US adult community normative population average (d = .68) but it did not significantly differ from the US adult female community normative population average (d = -0.02). The average total score on the distress measure was 14.39 (again falling just below the „depression‟ threshold of 16). This time the F test was insignificant, and a medium effect size was again observed when this average was compared to the US adult community normative population average (d = .49). [Of note, only 9.6% of caregivers indicated that they had experienced a significant (positive or negative)  60 stressor in the past week, perhaps indicating less bias in results relative to the baseline sample.] Additional paired-samples t-tests were performed to compare baseline and longitudinal means on each of those three questionnaires measured at both time points. While no significant differences in mean scores were observed across the mastery and subjective happiness scales, the baseline distress measure was shown to yield a significantly higher score relative to the longitudinal distress measure (t40 = -2.718, p<.01; d=0.35). Sample Combining In each of our three datasets, questionnaire data did not significantly differ across urban and rural subgroups. Moreover, only a handful of variables significantly differed across ontreatment and off-treatment subgroups. Consequently, in each dataset we combined the subgroups into one large group for remaining inferential analyses. Nonetheless, treatment status and time to hospital (a reflection of urban/rural status) were examined as predictor variables. Correlation Analyses A variety of correlation coefficients (i.e., Pearson product-moment, point biserial, phi) were calculated to examine degrees of (linear) association among our predictor variables, between our predictor variables and outcome variables, and between corresponding variables in different datasets. Given the numerous correlations analyzed, both standard (i.e., p<.05) and more conservative (i.e., p<.01 and p<.001) type 1 error rates are reported. Owing to our large number of variables (and hence correlations), bivariate scatterplots were checked for linearity only in cases where non-normal variables were assessed. Invalid correlations due to non-linear relationships are highlighted in the correlation data tables (see Tables 29 to 40) where applicable. While the focus in these correlation analyses is on results from transformed data, all of the correlation data tables display correlations pertaining to both transformed and nontransformed data. Transformed variables in these tables are identified with a (T), and non-  61 transformed data correlations are displayed immediately below the correlations pertaining to transformed data. Variables to be Removed – Precursor Variable Data To avoid problems with multicollinearity in later regression and SEM analyses, independent variables were identified (and removed from further inferential analyses) that highly correlated with other independent variables. Tabachnick & Fidell (2001)‟s recommendation for a cut-off value of r =.7 (in two-tailed analyses) was applied. Table 29 displays elevated correlations that prevented certain variables from being retained. In the baseline dataset, the following precursor variables were removed from further consideration: child age at diagnosis, treatment for the first time, active treatment status, urban/rural status, child Caucasian status, child spirituality/religiosity, and time since treatment ended. Similarly, the following precursor variables were removed from the adolescent dataset because of high inter-correlations: child age at diagnosis, treatment for the first time, active treatment status, urban/rural status, and time since treatment ended. Lastly, both urban/rural status and child age were removed from the longitudinal dataset. Variables to be Removed – Questionnaire Data Tables 30 through 32 display elevated correlations that prevented several questionnaire variables from being retained in further inferential analyses, again to avoid problems with multicollinearity. Baseline (caregiver) dataset.  One set of variables displaying high inter-correlations in  the baseline sample was the subscales from the PECI. As all three of these subscales highly inter-correlated (see Table 30a), both the unresolved sorrow & anger subscale and the guilt & worry subscale were removed from further analyses. We decided to preferentially keep the longterm uncertainty subscale because this variable most strongly captures the „uncertainty‟ concept that our structural equation models (to be discussed) are based on.  62 Another set of elevated correlations in the baseline sample came from the BFS. Each of the six benefit finding subscales highly correlated with the BFS total score (see Table 30b). We decided to preferentially use the BFS total score because: (i) there were elevated subscale intercorrelations that prevented all subscales from being simultaneously examined; and (ii) we wanted to keep the number of predictor variables low in our statistical analyses to maximize power. Nonetheless, we did not rule out the possibility of examining some of the subscales (in place of the BFS total score) in supplemental analyses. Where this occurred, we remained cognizant of which scales could not be used together in the same analyses due to high intercorrelations (i.e., positive self-view with either acceptance or appreciation). Lastly, elevated inter-correlations were observed on the FNQ. Across both IMP and MET sets of variables, several of the family needs subscales highly correlated with their respective FNQ total score (see Tables 30c and 30d). We decided to preferentially use the total scores for the same reasons as just mentioned with the BFS and because of the stronger internal consistency of the total scores. Nonetheless, the possibility of examining some of the subscales (in place of the FNQ total scores) in supplemental analyses was not ruled out. Where this occurred, we again remained cognizant of which scales had weaker reliability and which scales could not be used together in the same analysis due to high inter-correlations (i.e., health information MET with either community support MET or involvement with care MET). Adolescent dataset.  One set of variables displaying elevated correlations in the  adolescent sample related to the PedsQL. All four subscales highly correlated with the PedsQL total score, and several of these subscales also highly inter-correlated (see Table 31a). The psychosocial health summary score also highly correlated with all subscales and with the total score. We decided to retain the PedsQL total score in place of the subscale scores because this was to be our main outcome variable in the adolescent analyses. However, we did not rule out  63 the possibility of examining the psychosocial health summary score and the individual subscale scores (in place of the PedsQL total score) in analyses where the PedsQL total score was not the outcome variable. Where this occurred, we remained cognizant of which scales could not be used together in the same analysis due to high inter-correlations (i.e., psychosocial health summary score with any subscale score, and the physical subscale score with either the social subscale or school subscale score). Similar to the baseline (caregiver) sample, high inter-correlations were observed on the adolescent FNQ (see Tables 31b and 31c). We decided to retain the total scores instead of the subscale scores given that (i) most of the FNQ subscale scores had low internal consistency, and (ii) several of the subscales (across both IMP and MET variables) highly correlated with their respective FNQ total score. Longitudinal (caregiver) dataset.  Similar to the baseline dataset, all three of the PECI  subscales highly inter-correlated (see Table 32a). For reasons previously discussed, both the unresolved sorrow & anger subscale and the guilt & worry subscale were removed from further analyses. Another variable strongly associated with other variables in this dataset was hope (see Table 32b). It highly correlated with several of the other positive predictor variables, including optimism, (baseline) mastery, and (baseline) subjective happiness. We decided to retain hope over these other positive predictors because of the poignancy of the „hope‟ construct in our hypotheses. Lastly, the BFS and FNQ total scores were kept in preference over their respective subscale scores given the limited number of predictor variables allowed (due to low power) in the longitudinal analyses (N=41). Inter-correlations among Questionnaire Predictor Variables Next, correlations among remaining (questionnaire) predictor variables were examined. This was done to assess degrees of association among our positive and negative predictor  64 variables and to aid in decision making for the later regression (and SEM) analyses. Baseline (caregiver) dataset.  Correlations among the baseline positive predictor  variables (see Table 33a) and negative predictor variables (see Table 33b) were assessed. To no surprise, many significant correlations (at the .001 level) were found. Significant correlations were also observed between certain positive, negative, and „neutral‟ variables (see Table 33c). Adolescent dataset.  When correlations were assessed among the adolescent predictor  variables, only a handful of significant correlations were found beyond those previously mentioned among the PedsQL subscales (see Table 34). Specifically, benefit finding positively correlated with family needs – total score IMP, emotional QOL positively correlated with family needs – total score MET, and family needs – hope item MET positively correlated with both family needs – hope item IMP and family needs – total score MET. The only correlation significant at the .01 level was that between family needs – hope item IMP and family needs – total score IMP. Longitudinal (caregiver) dataset.  Similar to the baseline dataset, many significant  correlations (at the .001 level) were found in the longitudinal dataset when associations were examined among positive predictor variables (see Table 35a) and negative predictor variables (see Table 35b). Several significant correlations were also observed between positive, negative, and neutral predictor variables (see Table 35c). Predictor-Outcome Variable Bivariate Correlations In preparation for regression analyses, we also examined degrees of association between our remaining predictor variables and our outcome variables. Only analyses based on transformed data will be reported here. Baseline dataset.  In this baseline dataset, the two main outcome variables were distress  and subjective happiness. However, benefit finding and hope were considered supplemental outcome variables given our interest in positive psychology variables.  65 Starting with our negative outcome variable, distress showed significant correlations at the .001 level with many questionnaire predictor variables (see Tables 33b and 33c), including: threat, long-term uncertainty, (perceived) child physical struggles, (perceived) child emotional struggles, (perceived) child social struggles, optimism, mastery, hope, subjective happiness, and family needs – total score MET. Challenge (at the .01 level) and family needs – hope item MET (at the .05 level) also significantly correlated with distress. With respect to precursor (i.e., sociodemographic and cancer-specific) variables, distress correlated at the .001 level with (caregiver) diagnosed mental health struggles and time post diagnosis, and it correlated at the .05 level with ease of transport, (caregiver) physical health struggles, „other‟ family stressors, and treatment status (see Table 36a). Moving onto our positive outcome variables, predictors that significantly correlated at the .001 level with subjective happiness included benefit finding, optimism, mastery, hope, challenge, family needs – total score MET, threat, distress, long-term uncertainty, and (perceived) child emotional struggles (see Table 33a and 33c). (Perceived) child social struggles also correlated with subjective happiness at the .05 level. In terms of precursor variables, (caregiver) physical health struggles correlated with subjective happiness at the .01 level, while (caregiver) diagnosed mental health struggles, caregiver age, caregiver gender, and child age all correlated with subjective happiness at the .05 level (see Table 36b). When associations with benefit finding were examined, significant correlations (at the .001 level) were found with hope, subjective happiness, and family needs - total score IMP (see Table 33a and 33c). Challenge also correlated with benefit finding at the .01 level, while school needs & resources – IMP, family needs - total score MET, and family needs - hope item MET correlated with benefit finding at the .05 level. Caucasian status was the only precursor variable that correlated significantly with benefit finding (see Table 36c).  66 Lastly, variables found to significantly correlate (at the .001 level) with hope included all seven other positive predictor variables, threat, distress, long-term uncertainty, (perceived) child emotional struggles, and (perceived) child social struggles (see Table 33a and 33c). At the .05 level, (perceived) child physical struggles and previous experience with cancer (see Table 36d) also correlated with hope. Adolescent dataset.  In the adolescent dataset the main outcome variable was quality of  life. Benefit finding was considered a supplemental outcome variable, given our interest in positive psychology variables. Household income was the only adolescent dataset variable found to significantly correlate with quality of life (at the .01 level) (see Tables 37a and 37b). Meanwhile, variables that significantly correlated (at the .05 level) with teen benefit finding included family needs – total score IMP (see Table 37a), child age, caregiver age, and caregiver education (see Table 37b). Of note, both of our adolescent outcome variables strongly associated with some caregiver questionnaire variables (see Table 37c). Specifically, adolescent quality of life significantly correlated with (caregiver) perceived child physical struggles and (caregiver) perceived child social struggles at the .001 level, and with (caregiver) perceived child emotional struggles at the .05 level. Meanwhile, adolescent benefit finding significantly correlated at the .05 level with (caregiver) benefit finding, (caregiver) family needs – instrumental support IMP, and (caregiver) family needs – professional support IMP. Longitudinal dataset.  In the longitudinal dataset the main outcome variables were  longitudinal distress and longitudinal subjective happiness. Independent questionnaire variables that significantly correlated at the .001 level with longitudinal distress included threat, (baseline) distress, long-term uncertainty, longitudinal  67 mastery, hope, challenge, family needs – total score MET, and longitudinal subjective happiness (see Table 38a). Meanwhile, (perceived) child emotional struggles correlated with longitudinal distress at the .01 level, while (perceived) child social struggles, family needs – hope item MET, and school needs & resources IMP correlated with longitudinal distress at the .05 level. With respect to precursor variables, (caregiver) diagnosed mental health struggles (at the .001 level) as well as both ease of transport and longitudinal treatment status (at the .05 level) also correlated with longitudinal distress (see Table 38b). Independent questionnaire variables that significantly correlated at the .001 level with longitudinal subjective happiness included threat, long-term uncertainty, longitudinal mastery, hope, family needs – total score MET, and (as previously mentioned) longitudinal distress (see Table 38c). (Baseline) distress and challenge also correlated with longitudinal happiness at the .01 level, while (perceived) child emotional struggles and family needs-hope item MET correlated at the .05 level. Precursor variables correlating with longitudinal subjective happiness included (caregiver) diagnosed mental health struggles (at the .01 level) and ease of transport, child age at diagnosis, and radiation (at the .05 level) (see Table 38d). Correlations between Corresponding Questionnaire Variables in Different Datasets Baseline (caregiver) - adolescent correlations.  When we examined the correlations  between measures given to both caregivers and their adolescent children, only a handful of significant correlations at the .001 level were observed (see Table 39). These related to family needs – health information IMP (r = .497), family needs – involvement with care IMP (r = .509), and child struggles in physical functioning (r = -.602). [The reason this latter correlation is negative is because the caregiver PedsQL was reverse scored so that higher scores represented greater „struggles‟. In contrast, reverse scoring was not used on the adolescent PedsQL as we wanted higher scores to represent greater „quality of life‟ (i.e., less struggles in functioning).] Correlations pertaining to family needs variables here should be interpreted with caution given  68 that several of the caregiver and adolescent family needs subscale scores exhibited low internal consistency (see Tables 13 and 22). Baseline (caregiver) – longitudinal (caregiver) correlations.  When we examined the  correlations between those three measures given to caregivers at both baseline and six months‟ follow-up, strong associations were consistently observed (see Table 40). The two distress measures (r = .553), mastery measures (r = .662), and subjective happiness measures (r = .716) all correlated at the .001 level. Regression Analyses Hierarchical linear regression analyses were performed to determine the degree to which our independent variables explained variance in our dependent variables. Across our regression models, variables we wanted to control for were entered in the first step (where applicable). These included those precursor variables (i.e., socio-demographic and cancer-specific data) that significantly correlated with our outcome variables. Questionnaire predictor variables of interest were then added into regression models at later steps. As one of the main goals of this study was to assess the degree to which positive predictor variables are adaptive to caregiver emotional adjustment, we tested both liberal models (i.e., in which positive predictor variables1 were entered into the analysis ahead of negative predictor variables) and conservative models (i.e., in which positive predictor variables were entered into the analysis after negative predictor variables). [While the liberal models were more likely to show significant predictive power in our positive predictor variables, the conservative models would better test the strength of this predictive power.] The standard type 1 error rate of p<.05 was used. Data were screened for linearity and homoscedasticity (via standardized residual regression plots) to assure the necessary assumption of multivariate normality was met. Collinearity diagnostics were also reviewed to confirm there were no problems with multicollinearity among our predictor variables.  69 The number of allowable predictor variables in each regression analysis was determined based on the conservative 10:1 rule (Halinski & Feldt, 1970), where one predictor variable is allowed per every ten participants. Based on our baseline dataset (N=156), this criterion suggested we had enough power for up to 16 predictor variables. In contrast, we only had enough power to support 4 predictor variables in analyses on the adolescent (N=40) and longitudinal (N=41) datasets. Correlation tables were consulted to identify which variables significantly correlated with our outcome variables (and hence which variables would serve as relevant predictor variables in each regression). A correlation of r =.70 or greater among predictor variables was also used as the threshold for rule out (to avoid problems with multicollinearity). Where inter-correlations were high, variables of greater theoretical interest to us were chosen. Only regression analyses using transformed data are reported here. Transformed variables are identified in the data tables with a (T). Regression analyses using non-transformed data can be found in Tables 41 to 47. [Of note, few differences were observed when regression data were compared across respective transformed data and non-transformed data tables.] Baseline Dataset: Benefit finding predictions.  According to correlation analyses (see Tables 33a, 33c,  36c, and 48a), relevant independent variables for predicting caregiver benefit finding included caregiver Caucasian status, family needs – total score MET, family needs – hope item MET, challenge, hope, subjective happiness, family needs – total score IMP, and school needs & resources – IMP. Given that we had enough power to include up to 16 predictor variables, and since all predictor inter-correlations were below threshold for rule-out, all of these predictor variables were used. [Of note, given that no negative predictor variables significantly correlated with benefit finding, contrasting liberal and conservative models could not be tested here.] Table 48b presents the results of the first regression analysis for predicting caregiver  70 benefit finding. Caucasian status was shown to predict a significant (p<.001) proportion of variance in benefit finding (9.8%) when entered in the first step; specifically, greater benefit finding was predicted by non-Caucasian status. The (19.6%) improvement to the model yielded by the questionnaire predictor variables in the second step was also significant at the .001 level. Perceiving the cancer experience as a challenge and recognizing important family needs as well as school and resource needs for one‟s child were shown to be significant predictors of (greater) benefit finding in this step. Collectively this model explained 29.4% of variance in benefit finding. Given that our sample size provided enough power to include more predictors, interaction terms (among significant precursor and questionnaire predictor variables) were tested as predictors in a third step (see Table 48c). The purpose of this was to assess for moderator variables impacting benefit finding. [Collinearity diagnostics confirmed there were no problems with elevated inter-correlations among this new set of predictor variables.] When this model was run, none of the interaction terms were found to be significant predictors and the improvement to the model (.8%) was negligible. These findings suggest that the significance of the questionnaire variables in predicting caregiver benefit finding does not vary according to Caucasian status. Owing again to our sample size being large enough to accommodate up to 16 predictor variables, supplementary regression analyses were performed in which family needs domain scores were substituted in place of total scores. Correlation analyses were consulted to determine which domain scores significantly correlated with benefit finding (see Table 48d) and to make sure there were no inter-correlations among predictor variables exceeding the .70 cut-off (see Table 48a). When this revised basic model was run (see Table 48e), it explained an additional 4% of variance in benefit finding (relative to the initial basic model, see Table 48b). Recognizing important family needs in the area of emotional support surfaced as a „new‟ significant predictor  71 of (greater) benefit finding. A similar model with added interaction terms (among significant precursor and questionnaire predictor variables) was next assessed (see Table 48f). [Collinearity diagnostics again confirmed there were no problems with elevated inter-correlations among this new set of predictor variables.] When this model was run, the interaction terms failed to be significant predictors (confirming the absence of moderator variables) and the improvement to the model relative to the initial interaction model (see Table 48c) was observed to be negligible. Hope predictions.  From the correlation analyses (see Tables 33a, 33c, and 36d),  relevant independent variables for predicting caregiver hope included optimism, mastery, subjective happiness, benefit finding, family needs – total score MET, family needs – hope item MET, challenge, (perceived) child physical struggles, threat, distress, long-term uncertainty, (perceived) child emotional struggles, (perceived) child social struggles, and caregiver previous experience with cancer. All these variables were used given the power provided by our large sample size and because all inter-correlations fell below r=.70 (see Tables 33a, 33c, and 49). Tables 49b and 49c present the results of the initial regression analyses for the predictions of caregiver hope. Both a liberal and conservative model was run. When the liberal model was first run, the sole precursor variable entered in step one (i.e., caregiver previous experience with cancer) was not found to yield a significant improvement (p=.984). Hence, the model was rerun dropping the first step. In this revised (liberal) model (see Table 49b), the positive predictor variables explained a significant amount of variance in hope (60.1%). Interestingly, the negative predictor variables did not significantly improve the model in the second step. Optimism, mastery, subjective happiness, and benefit finding were all found to be significant predictors of caregiver hope. Taking into consideration that hope is a (reflected) transformed variable, beta weights suggest that greater amounts of these thinking/feeling styles  72 are associated with more hopeful thinking. When the more conservative model (see Table 49c) was tested, both steps of the model were significant at the .001 level. Of note, the (31.5%) significant improvement to the model yielded by the positive predictors in the second step is greater than the (29.4%) improvement yielded by the negative predictors in the first step, attesting to the superiority of positive feeling and thinking styles in predicting hope. Interestingly, while (low) distress levels were now shown to significantly predict (more) hope in this model, benefit finding was no longer observed to be a significant predictor. A supplementary regression analysis was performed, this time specific to the liberal model (as we did not have enough power to test supplementary conservative models). Domain scores for family needs and benefit finding that significantly correlated with hope were assessed as predictor variables in place of total scores. Correlation analyses were consulted to determine which domain scores significantly correlated with hope and to make sure there were no intercorrelations among predictor variables exceeding r=.70 (see Table 49d). As 14 relevant positive predictor variables were identified and therefore entered in step one, it was not possible to enter negative predictor variables in a second step due to power limitations. When this model was run (see Table 49e), it was found to be significant at the .001 level. Benefit finding – acceptance surfaced as a „new‟ predictor variable in this model, indicating that higher levels of hope are predicted by higher levels of benefit finding in the realm of acceptance. Overall, however, this model showed a negligible improvement (1%) in total explained variance relative to the initial (liberal) model. Subjective happiness predictions.  From the correlation analyses (see Tables 33a, 33c,  and 36b), relevant independent variables for predicting caregiver subjective happiness included (perceived) child emotional struggles, (perceived) child social struggles, (caregiver) physical  73 health struggles, (caregiver) diagnosed mental health struggles, caregiver age, caregiver gender, child age, benefit finding, optimism, mastery, hope, challenge, family needs – total score MET, threat, long-term uncertainty, and distress. [In regards to the latter, we decided to use distress as a predictor in the subjective happiness regression analyses because distress and subjective happiness are not orthogonal constructs (Davidson, 1992; Lemonick, 2005; Watson, Clark, & Tellegen, 1988).] All these variables were used given the power provided by our large sample size and because all inter-correlations fell below the r=.70 cut-off (see Tables 33a, 33c, and 50a). Tables 50b and 50c present the results of the regression analyses for the predictions of caregiver subjective happiness. In the liberal model (see Table 50b), the positive predictor variables entered in step two explain the majority (45.1% of 66.1%) of total variance explained in subjective happiness. Nonetheless, all three steps significantly improved the model, with precursor variables and negative predictors contributing 13.4% and 7.5%, respectively, of explained variance. Examination of the beta weights indicates that greater levels of subjective happiness are significantly predicted by higher levels of optimism, hope, and perceived mastery, by lower levels of distress and perceived threat, and by older child age and a lack of (caregiver) current physical struggles. Surprisingly, the model also suggests greater happiness is significantly predicted by higher levels of (perceived) child social struggles. When the more conservative model (see Table 50c) was tested, all three steps of the model yielded significant improvements (p<.001). Of note, the second step in this model explained a less impressive 31.4% of variance in hope, when compared to the second step of the liberal model. This again attests to the superiority of positive feeling and thinking styles as predictors of positive outcomes. Interestingly, (perceived) child social struggles no longer significantly predicted hope when entered in the second step. [It is possible that the previous  74 significant finding related to this variable reflected a „suppressor variable effect‟, as this variable was found to significantly correlate (at the .001 level) with several other predictor variables but correlated with happiness at a relatively weak level (.05 level).] Distress predictions.  From the correlation analyses (see Tables 33b, 33c, and 36a),  relevant independent variables for predicting caregiver distress included (perceived) child physical struggles, (perceived) child emotional struggles, (perceived) child social struggles, (caregiver) diagnosed mental health struggles, time post diagnosis, ease of transport, (caregiver) physical health struggles, ‘other’ family stressors, treatment status, long-term uncertainty, threat, optimism, mastery, hope, challenge, family needs- total score MET, family needs – hope item MET, and subjective happiness. [In regards to the latter, we decided to use subjective happiness as a predictor in these analyses because (as previously mentioned) subjective happiness and distress are not orthogonal constructs.] Given that we only had enough power to accommodate 16 predictor variables, ‘other family stressors’ and family needs - hope item MET were not used given their relatively weak correlations with distress. Of note, none of the predictor variable inter-correlations exceeded the r=.70 cut-off (see Table 33b, 33c, and 51a). Tables 51b and 51c present the results of the regression analyses for the predictions of caregiver distress. Again, both a liberal and conservative model was tested. In the liberal model (see Table 51b), significant improvements were observed at all three steps and total explained variance (in distress) was 58.8%. The precursor variables explained 20.5% of variance in distress, the positive predictor variables explained another 27.7% of variance when entered in the second step, and the negative predictor variables explained another 10.6% of variance when entered in the third step. Specifically, elevated distress levels were predicted by a history of mental health struggles, a shorter time since diagnosis, more difficult transport experiences, lower levels of happiness, appraisals of threat, and greater amounts of long-term uncertainty.  75 When the more conservative model (see Table 51c) was tested, all three steps again yielded significant improvements. Not surprisingly, the positive predictor variables explained a less impressive (relative to the liberal model) 9.6% of variance in distress when entered in the third step. The negative predictors explained the most variance in distress, yielding a change in R2 of 28.7%. Greater (perceived) child emotional struggles were now shown to significantly predict greater distress in this model. Supplementary regression analyses examining domain scores (across family needs and benefit finding) and interaction effects (among our significant precursor and questionnaire variables) could not be performed for distress, given that regression models were already maxed out at 16 predictor variables. Mastery predictions.  Lastly, in preparation for SEM, some exploratory regression  analyses were performed in which mastery served as the outcome variable. The purpose of these regressions was to identify significant interaction terms (between long-term uncertainty and other independent variables) that could be tested as possible moderators (of mastery) in later structural equation models. Based on correlation analyses (see Tables 33a and 33c) as well as theoretical underpinnings of our proposed model (see Figure 2), relevant independent variables for predicting caregiver mastery included optimism, family needs – total score MET, family needs – hope item MET, long-term uncertainty, (perceived) child physical struggles, (perceived) child emotional struggles, and (perceived) child social struggles. Interaction terms between long-term uncertainty and these variables were therefore created and entered into regression equations (see Table 52a). [Collinearity diagnostics confirmed there were no problems with elevated intercorrelations among this set of predictor variables.] When the regression was run, significant interaction terms for predicting mastery included those involving optimism and (perceived) child  76 emotional struggles. A supplementary analysis was then performed in which relevant family needs and benefit finding domain scores were converted into interaction terms (with long-term uncertainty) and assessed as predictors of mastery. Correlation analyses were consulted to determine which domain scores significantly correlated with mastery (see Table 52b). All interaction terms significantly correlating with mastery were then entered into the regression. [Collinearity diagnostics again confirmed there were no problems with elevated inter-correlations.] When the regression was run (see Table 52c), significant interaction terms for predicting mastery included those involving optimism, benefit finding-acceptance, benefit finding-appreciation, family needs – professional support MET and family needs – health information MET. Of note, the interaction term with (perceived) child emotional struggles was no longer significantly predictive in this model. Adolescent Dataset: Benefit finding predictions.  According to correlation analyses (see Tables 37a, 37b,  and 37c), relevant independent variables for predicting adolescent benefit finding included family needs – total score IMP, child age, caregiver age, caregiver education, (caregiver) benefit finding, (caregiver) family needs – instrumental support IMP, and (caregiver) family needs – professional support IMP. No elevated inter-correlations amongst these predictor variables were found (see Table 53a). Table 53b presents the results of the regression analysis for the adolescent benefit finding predictions. Due to power limitations, only four predictor variables could be used. The three precursor variables along with the adolescent questionnaire predictor variable (i.e., family needs – total score IMP) were first tested as these variables were most interesting from a theoretical standpoint. While the precursor variables predicted a significant proportion of variance (26.8%) in benefit finding, the 2.7% improvement to the model yielded by family needs – total score IMP  77 was not statistically significant. Moreover, no predictor variables significantly predicted benefit finding in isolation. When (caregiver) benefit finding, (caregiver) family needs – professional support IMP, and (caregiver) family needs – instrumental support IMP were (one at a time) substituted for family needs – community support IMP in the final step of the regression model, the improvements (i.e., 6.4%, 3.2%, and 7.3%, respectively) were, again, not found to be statically significant (see Table 53b). Quality of life predictions.  According to correlation analyses (see Tables 37a, 37b, and  37c), relevant independent variables for predicting adolescent quality of life included household income, (caregiver) perceived child physical struggles, (caregiver) perceived child social struggles, and (caregiver) perceived child emotional struggles. Given that up to four predictor variables could be used, all four of these variables were entered into the regression. Table 54b presents the results of the regression analysis for the adolescent quality of life prediction. In this model, household income predicted a significant proportion of variance (20.8%) in quality of life. Moreover, the 23.2% improvement to the model yielded by the caregiver questionnaire variables was also statistically significant (p<.01). From the beta weights, greater adolescent quality of life was significantly predicted by greater household income (i.e., more earners) and by having caregivers who perceive fewer child social struggles. Longitudinal Dataset: Longitudinal distress predictions.  According to correlation analyses (see Tables 38a  and 38b), relevant independent variables for predicting longitudinal distress included (caregiver) diagnosed mental health struggles, ease of transport, longitudinal treatment status, threat, (baseline) distress, long-term uncertainty, (perceived) child emotional struggles, (perceived) child social struggles, longitudinal mastery, hope, challenge, family needs – total score MET, family needs – hope item MET, school needs & resources – IMP, and longitudinal subjective  78 happiness. Given that four variables could again only be used due to power limitations, variables were screened out based on high inter-correlations (see Tables 35a, 35b, 35c, and 55a), low bivariate correlations with longitudinal distress (see Tables 38a and 38b), and interest from a theoretical perspective. Predictor variables retained included (caregiver) diagnosed mental health struggles, long-term uncertainty, (baseline) distress, and hope. Tables 55b and 55c present the results of the regression analyses for the longitudinal distress predictions. Both a liberal and conservative model was tested. In the liberal model (see Table 55b), total variance explained (in longitudinal distress) was 52.8%. (Caregiver) diagnosed mental health struggles predicted a significant proportion of variance (33.4%) in the first step, as did hope in the second step (11.9%). The additional 7.6% of variance predicted by the negative predictors in the third step was not found to be a statistically significant improvement to the model. According to the beta weights, elevated distress levels at six-months‟ follow-up were significantly predicted by having a history of mental health struggles, perceiving lower levels of hope, and experiencing higher levels of long-term uncertainty. When the more conservative model (see Table 55c) was tested, the negligible improvement to the model yielded by hope in the third step (2.4%) was not statistically significant. Given that dropping the (non-significant) third step in this model would make room for an extra predictor variable, the interaction term between the remaining two significant predictors was tested in the third step in place of the positive predictors (see Table 55d). Collinearity diagnostics confirmed there were no problems with elevated inter-correlations among this new set of predictor variables. Interestingly, this interaction term was not significantly predictive, suggesting that the influence of long-term uncertainty (in predicting longitudinal distress) does not vary according to one‟s history of mental health struggles. Longitudinal subjective happiness predictions.  According to correlation analyses (see  79 Tables 38c and 38d), relevant independent variables for predicting longitudinal subjective happiness included ease of transport, (caregiver) diagnosed mental health struggles, child age at diagnosis, radiation, threat, (baseline) distress, long-term uncertainty, (perceived) child emotional struggles, longitudinal mastery, hope, challenge, family needs – total score MET, family needs – hope item MET, and longitudinal distress. Due to power limitations, variables were again screened out based on high inter-correlations (see Tables 35a, 35b, 35c, and 55a), low bivariate correlations with longitudinal subjective happiness (see Tables 38c and 38d), and interest from a theoretical perspective. Predictor variables retained included (caregiver) diagnosed mental health struggles, threat, hope, and family needs – total score MET. Tables 56a, 56b, and 56c present the results of the regression analyses for the longitudinal subjective happiness predictions. Again, both a liberal and conservative model was tested. In the liberal model (see Table 56a), total variance explained was 51.9%. This is lower compared to the baseline models, which is not surprising given the limited power in these longitudinal analyses. (Caregiver) diagnosed mental health struggles again predicted a significant proportion of variance (21%) in the first step. While the positive predictors yielded a significant improvement (of 28.3%) at the .001 level across the second step, threat failed to show a significant improvement in the third step. According to the beta weights, elevated subjective happiness at six-months‟ follow-up was predicted by lacking a history of mental health struggles and perceiving higher levels of hope. Given that dropping the (non-significant) third step in this model made room for an extra predictor variable, the interaction term between the remaining two significant predictors was tested (see Table 56b). [Collinearity diagnostics confirmed there were no problems with elevated inter-correlations among this new set of predictor variables.] This interaction term was not significant, suggesting that the predictive power of hope (in explaining significant variance in  80 longitudinal subjective happiness) does not vary according to one‟s history of mental health struggles. Lastly, when the conservative model (see Table 56c) was tested, threat entered in the second step now yielded a significant improvement (11.4%) at the .05 level. Of note, the 19.5% improvement to the model yielded by the positive predictors in the third step is also statistically significant, attesting to the superiority of these positive variables in predicting positive outcomes. Structural Equation Modeling Analyses Structural equation modelling techniques were used to identify new mediating and moderating pathways within the MUIT framework (see Figure 1) using the baseline (caregiver) dataset. SEM techniques could not be used with either the adolescent (N=40) or longitudinal (N=41) datasets due to power limitations. Due to the innovativeness of this research, these SEM analyses utilized a certain extent of „model fishing‟. They are therefore considered exploratory rather than confirmatory. EQS 6.1 was used to generate a covariance matrix of all measured variables in our models and to test the measurement and structural portions of these models. As it was desirable to use the non-transformed data to enable easier interpretation of results, Robust Maximum Likelihood estimation („ML Robust‟) was used to evaluate the fit of the measurement and structural models to our (non-normal) empirical data. As such, the SatorraBentler Scaled Chi-Square (χ2) value was used to assess goodness of model fit, along with other fit indices such as the Bentler-Bonett Non-normed Fit index (NNFI), Comparative Fit Index (CFI), and Root Mean-Square Error of Approximation (RMSEA). „Good fit‟ according to these indices breaks down as follows: Chi-Square >.05 (Barrett, 2007); RMSEA< .05 (Raykov & Marcoulides, 2006); and NNFI & CFI >.95 (Hu and Bentler, 1999). Squared multiple correlation (R2) values relating to our dependent variables were also assessed to evaluate the effectiveness of the model in explaining variance in these variables.  81 Distress Model The first distress model assessed for fit to the data was a path diagram of Mishel‟s basic framework (see Figure 6). This model did not fit the data (χ2(5, N=156)= 66.195, p<.001; NNFI = .239, CFI = .620; RMSEA = .281). As this model was run for replication purposes (of Mishel‟s work) rather than out of theoretical interest, we moved on to testing more complex models of interest. A latent-variable structural equation model of the same basic framework was next assessed (see Figure 7). Owing to our interest in detecting variables that moderate the impact of uncertainty on mastery, we also entered into this model those variables whose interaction terms (with long-term uncertainty) were found to be significant predictors of mastery in regression analyses (on non-transformed data; see Table 45).This model did not fit the data (χ2(55, N=156)= 151.518, p<.001; NNFI = .861; CFI = .916; RMSEA = .106). Given that none of the moderating pathways were statistically significant, they were removed from the model. As the direct pathway between optimism and mastery was statistically significant, it was retained. While some additional pathways were then added based on feedback from the residual output (see Figure 8), poor model fit remained (χ2(30, N=156)= 94.215, p<.001; NNFI = .765; CFI = .844; RMSEA = .118). Because our sample size allowed for an additional six predictor variables (before reaching the maximum number allowed of 16), further variables were entered into the model. These were selected based on significant findings from the distress and hope regression analyses (on non-transformed data; see Tables 44 and 42, respectively). [Of note, subjective happiness was not used as a predictor variable here given that it was incorporated into this study as a comparison (outcome) variable for distress in the modeling analyses. If we entered subjective happiness into this model, we would then have the comparison model sandwiched within our base model. In an effort to keep things simple and more easily interpretable, we decided not to use this (significant) predictor variable in the modeling analyses.]  82 This next model incorporated the (latent variable) hope mediating pathway, benefit finding (as a predictor of hope), and the two strongest „precursor variable‟ predictors of distress caregiver diagnosed mental health struggles and time post diagnosis (see Figure 9). This model did not fit the data (χ2(85, N=156) = 259.357, p<.001; NNFI = .638; CFI = .244; RMSEA = .115). While several pathway revisions were then introduced (see Figure 10) based on feedback from the residual output, good model fit was still not achieved across any fit index (χ2(83, N=156)= 121.531, p<.01; NNFI = .918; CFI = .943; RMSEA = .055). Given that challenge was now shown to lack significant relations with both hope and distress, this variable was removed. As we had power to introduce one more variable to the model, ease of transport (i.e., the remaining significant „precursor variable‟ predictor of distress) was incorporated. This model (see Figure 11) showed good model fit across all fit indices (χ2(76, N=156)=  84.25, p=.242; NNFI = .980; CFI = .988; RMSEA = .026). However, two pathways in the  model were insignificant. When these were removed and two new pathways were added based on the residual output (see Figures 12 and 13), excellent model fit was achieved (χ2(77, N=156)= 69.15, p=.726; NNFI = 1.018; CFI = 1.000; RMSEA = .000). This final model showed fairly quick convergence, taking only 10 steps to reach an acceptable fit function of .46587. All pathway coefficients in the model were found to be statistically significant at the .05 level, and all pathway coefficient signs described relations in anticipated directions. The overall model explained 58.9% of the variance in distress and 59.6%% of the variance in hope, both considered large effect sizes by Cohen‟s (1988) standards. These figures indicate the model was effective in representing a considerable proportion of variance in our variables of interest. Nonetheless, significant error variance terms across all of our dependent variables confirms the model is far from explaining all sources of influence. „Effect decomposition‟ was performed in EQS so that mediation pathways (i.e., indirect  83 effects) could be assessed and ultimately checked for significance via the Sobel Test. When nonstandardized direct and indirect effects were compared, ten partial mediation pathways of interest were detected (see Figure 14). Mastery, threat, hope, and (perceived) child emotional struggles were all identified as (partial) mediators in this model. More specifically, mastery partially mediated the significant indirect pathways between long-term uncertainty and hope (Sobel test statistic = -2.88, p<.01), long-term uncertainty and distress (Sobel test statistic = 2.113, p<.05), optimism and hope (Sobel test statistic = -4.902, p<.001), and optimism and distress (Sobel test statistic =2.620, p<.01). Meanwhile, threat was shown to partially mediate the significant indirect pathways between benefit finding and (perceived) child emotional struggles (Sobel test statistic = -2.209, p<.05), benefit finding and distress (Sobel test statistic = 2.369, p<.05), long-term uncertainty and (perceived) child emotional struggles (Sobel test statistic = 3.163, p<.01), and long-term uncertainty and distress (Sobel test statistic = 3.694, p<.001). Lastly, hope (as hypothesized) partially mediated the significant indirect pathway between mastery and distress (Sobel test statistic = 1.975, p<.05), and (perceived) child emotional struggles partially mediated the significant indirect pathway between threat and distress (Sobel test statistic = 2.574, p<.05). Subjective Happiness Model As subjective happiness was chosen to be a second outcome variable in SEM for comparison purposes with distress, the first subjective happiness model assessed for fit to the data was a replication of the final distress model. In this model, subjective happiness was substituted for distress as the final outcome variable (see Figure 15). While this model fit the data (χ2(42, N=156)= 46.108, p=.306; NNFI = .986, CFI = .992; RMSEA = .025), several pathways were insignificant. This was not surprising given that we know subjective happiness and distress are not orthogonal constructs. Statistically insignificant pathways were therefore removed from this model and further variables (i.e., perceived child social struggles and caregiver physical  84 struggles) were selected for entry based on significant findings from the subjective happiness regression analyses (on non-transformed data; see Table 43). [Note that distress was not used as a predictor variable here for the same reasons mentioned previously in regards to the exclusion of subjective happiness in the distress model.] When all significant predictors of subjective happiness were entered into the model (see Figure 16), acceptable model fit remained (χ2(28, N=156)= 33.795, p=.208; NNFI = .977, CFI = .988; RMSEA = .037). However, an insignificant pathway was again observed. When this pathway was removed and a new pathway was added based on the residual output, a model was identified that fit the data well (χ2(22, N=156)= 24.274, p=.333; NNFI = .989; CFI = .995; RMSEA = .026) and made sense conceptually (see Figures 17, 18, & 19). This final model showed fairly quick convergence, taking just 8 steps to reach an acceptable fit function of .16798. All pathway coefficients in the model were found to be statistically significant at the .05 level, and all pathway coefficient signs described relations in anticipated directions. All factor loadings were >.50, providing support for the validity of the measurement model. The overall model explained 62.7% of variance in subjective happiness and 61.2% of variance in hope, both considered large effect sizes by Cohen‟s (1988) standards. These figures indicate that the model was effective in representing a considerable proportion of variance in our variables of interest. Nonetheless, significant error variance terms across all of our dependent variables again confirms the model is far from explaining all sources of influence. „Effect decomposition‟ was again performed so that mediation pathways could be assessed and ultimately checked for significance via the Sobel Test. When non-standardized direct and indirect effects were compared, ten (partial) mediation pathways of interest were again detected (see Figure 19). As in the distress model (see Figure 14), mastery partially mediated the significant indirect relationships between long-term uncertainty and hope (Sobel test statistic  85 = -2.892, p<.01) and optimism and hope (Sobel test statistic = 4.883, p<.001). It also partially mediated the significant indirect relationships between long-term uncertainty and subjective happiness (Sobel test statistic = -2.235, p<.05) and optimism and subjective happiness (Sobel test statistic = 2.864, p<.01). Threat was again shown to partially mediate the significant indirect relationships between the emotional outcome variable (in this case, subjective happiness) and both benefit finding (Sobel test statistic = 2.414, p<.05) and long-term uncertainty (Sobel test statistic = -4.632, p<.001), and it also now partially mediated the significant indirect relationship between subjective happiness and caregiver physical struggles (Sobel test statistic = 2.787, p<.01). Lastly, hope was again shown to partially mediate the significant indirect relationship between the emotional outcome variable (i.e., subjective happiness) and mastery (Sobel test statistic = 4.196, p<.001), and it now also partially mediated the significant indirect relationships between subjective happiness and both benefit finding (Sobel test statistic = 2.907, p<.01) and optimism (3.282, p<.01). Final Note Across both models, Mardia‟s coefficient (distress model: 22.1334, z=5.7593; subjective happiness model: 9.7633, z=3.9357) confirmed the data were non-normal (owing in large part to the non-normal distributions of the distress and hope indicators), justifying use of the ML ROBUST estimation method. Moreover, the determinant of input matrix values (distress model: .66758D+11; subjective happiness model: .30526D+03) confirmed there were no problems with model redundancy (i.e., multicollinearity), and the degrees of freedom (distress model: df=77; subjective happiness model: df=22) confirmed the models were „properly identified‟ for parameter estimation. A lack of standardized residual covariance values larger than │2│also confirmed that the models did not over- or under-represent any parts of the data. In addition, the absence of outliers in the residual distributions (which were normal and symmetric across both models) indicated that no relations in either model needed to be altered to enhance model fit.  86 Indeed, our average absolute standardized residuals (AASR) values (equal to .04 in the distress model and .02 in the subjective happiness model) fell well below the recommended cut-off for good fit of <.06 (Raykov & Marcoulides, 2006). Lastly, it should be noted that while „bootstrapping‟ was considered as a technique for further analysis of the mediation pathways, we decided our small (N=156), non-normal dataset was not suitable for this approach. As demonstrated by Ichikawa and Konishi (1995) in a study on the effect of non-normality where bootstrapping consistently over-estimated empirical standard errors at N=150, small to moderate sample sizes can be a problem with this statistical technique.  87 Discussion and Conclusions The goal of this research was to accumulate more theoretical knowledge on forces influencing caregiver emotional adjustment to pediatric cancer. With three datasets covering two populations across numerous socio-demographic, cancer-specific, and questionnaire variables, new insights became available. In particular, prevalence levels of our positive psychology variables (e.g., hope, benefit finding) were quantified, and the significant role that these variables play in caregiver emotional adjustment in the pediatric cancer realm was confirmed. In this regard, the explanatory value of these positive psychology constructs was proven to be unique and not just a reflection of the absence of distress. This was an important achievement given the field‟s increasing appreciation for the adaptive nature of these positive psychology variables and the overwhelming demand for research on them (Crowell & Strahlendorf, 2007; Klassen, Raina, et al., 2007; Phipps, 2007; Stewart & Mishel, 2000). Through regression and SEM analyses, we also determined explained variance in all of our major outcome variables, highlighting how much we understand these (and how much we still need to discover). In this process, we deduced the (fairly insignificant) roles of urban/rural status and treatment status as predictors of adjustment in our populations. Lastly, and most importantly, the MUIT model was revised to be applicable in the pediatric cancer caregiving context, highlighting the roles that uncertainty, perceptions of mastery, appraisals of threat, and forms of positive thinking play in caregiver emotional adjustment. Partial mediation pathways highlighting how these variables influence each other were also identified. This study is the first to explore the roles of hope and benefit finding in this framework using modeling techniques. Strengths of Study There are several noteworthy strengths of this study. First, it assessed real subjects in their natural setting. Fairly high response rates were also observed across both caregivers (76.1% baseline, 56.9% longitudinal) and adolescents (74%), and samples were fairly well representative  88 of the diverse populations they came from. In particular, a variety of caregivers were sampled with respect to age, family size, spirituality/religiosity, child age, cancer type, and of course treatment status and urban/rural status. Another strength of this study was that it assessed a nice combination of disease specific experiences (e.g., family needs, school needs, benefit finding, long-term uncertainty) and common life experiences (e.g., happiness, optimism, distress). Important confounding variables were also controlled for, such as pre-existing distress and „other‟ life stressors. Moreover, the acquisition of multiple participant samples provided breadth in our analyses. Obtaining simultaneous data from teens and caregivers, for example, not only enabled outcome measurement in both populations but also allowed for the unique opportunity to compare across informants. In addition, our acquisition of longitudinal caregiver data enabled prospective measurement across the illness trajectory, research that is in high demand in the pediatric psycho-oncology field (Vrijmoet-Wiersma et al., 2008). Such analyses were critical in providing support for directionality assumptions in our modelling analyses. Moreover, by assessing emotional adjustment at two time points (and using the first measure as a predictor of the second), we were also able to assess „change‟ in our key variables of interest. Finally, the advanced statistical techniques used in this research were invaluable. The combination of SEM and hierarchical linear regression enabled multiple risk and resilience factors to be studied in a sensitive manner. These variables were scrutinized both for their influences on our outcome variables and for their inter-relations. Most importantly, we were able to test a theoretically grounded model that allowed for the identification of both direct and indirect relations among our variables. Compared to previous modeling studies in this population (most of which have been under powered, statistically unsophisticated, and neglectful of the role of uncertainty) which have averaged 44% explained variance in distress, this higher quality  89 study explained 59% of variance in distress. Main Findings There were many interesting findings in this research, some which confirmed hypotheses and others which did not. Some of the more exciting results are summarized below. Caregiver Distress From the descriptive analyses, we learned that these caregivers experience elevated distress levels relative to the normative population, including those caregivers who have children no longer receiving treatment. We also learned that elevated distress levels indicative of clinically significant depressive symptomatology were reported by on-treatment caregivers at baseline. This agrees with previous research showing that these caregivers do struggle emotionally (Pai et al., 2006), presumably due to the myriad of stressors mentioned in the introduction section. However, significantly lower distress levels in the (baseline) off-treatment caregivers (relative to those on-treatment) as well as in the entire longitudinal dataset at time 2 (relative to at time 1) suggest that feelings of distress tend to subside to non-clinical levels with time, regardless of stable long-term uncertainty levels. The longitudinal findings also suggest that this occurs regardless of treatment status. These „falling‟ distress levels may stem from the excellent medical treatment received as well as from the optimistic success rates associated with such treatment. While these caregivers are far from being distress-free, they no longer meet criteria for „mild depression‟. This rise and fall in caregiver affect is characteristic of an emotional stress reaction, which has previously been observed in this population (Trask et al., 2003). These findings contradict literature reporting that caregivers continue to experience clinically elevated levels of emotional stress in the years following treatment (Hardy et al., 2008). And while Trask et al. (2003) reported observing „falling‟ distress levels in the first month post diagnosis, our research suggests a more delayed onset of this, commencing no earlier than 2 months post diagnosis2. It is possible that use of different measures accounts for these  90 discrepancies between our findings and previous research. While Hardy et al. focused on „posttraumatic stress‟, Trask et al. measured distress from the Brief Symptom Inventory (Derogatis, 1992) which is confounded by its additional assessment of „anxiety‟ and „somatization‟. In regression analyses, a combination of positive, negative, and precursor variables were all found to significantly explain variance in caregiver distress, with many hypothesized relations confirmed and some new relations observed. In contrast to previous research, gender, ethnicity, marital status, employment, and „other’ current stressors did not surface as significant predictors. [It is possible that insufficient measurement and/or non-normality in our sociodemographic data accounts for some of these findings3.] Time to hospital (reflecting „urban‟ versus „rural‟ status) was also not a significant predictor of this outcome variable, or any outcome variable, which goes against our hypotheses but agrees with previous findings in adult cancer patients (Payne et al., 2000). In total, just over half of the total variance in distress was explained, indicating numerous other influential factors remain unaccounted for. In line with previous research, uncertainty, threat, caregiver diagnosed mental health struggles, elevated (perceived) child emotional struggles, shorter time since diagnosis, and difficulties in transportation all surfaced as strong influences on caregiver distress. [In regards to the latter variable, it seems that „perceptions‟ of difficulty may be more predictive of distress than actual levels of difficulty. The non-significance of such variables as time to hospital, household income, and perceived child physical struggles in predicting distress suggests that physical obstacles (i.e., lengthy travel, expense, handicap impositions) are not what make travel more distressing. It is likely that the perception of difficult travel is a broad proxy for numerous issues around transportation, including degree of traffic, inconvenience and cost of hospital parking, level of interruption of daily life caused by travel, or amount of other life stressors currently impacting the family. Further testing is needed to verify such more fine-grained hypotheses.]  91 Hope was also found to have significant influence on distress. In addition to the significant pathway leading from hope to distress in the SEM model, hope was also found to be a significant predictor of distress at six month‟s follow-up. These findings confirm previous research (Christman, 1990; Eapen, Mabrouk, & Bin-Othman, 2008) and likely reflect the „buffering capacity‟ of positive emotion discussed in the introduction. Other predictors of longterm distress included diagnosed mental health struggles and (in particular) uncertainty. All pathways in our distress model were significant and explained influence in anticipated directions, attesting to the strength and solid theoretical grounding of this model. Around 60% of variance in both distress and hope was explained (similar to regression analyses), and ten partial mediation pathways were confirmed. Perceived mastery was identified as a crucial variable in fostering hope, while (low) hope, mastery, and perceived child emotional struggles were all hallmarked as pivotal in predicting distress. These findings are the first to document the direct impact of hope on emotional adjustment in the pediatric psycho-oncology realm using modeling techniques. Threat was observed to exert significant influence on distress through its (partial) mediating roles (as hypothesized), and it was also shown to negatively influence aspects of hope. This latter finding is not surprising given that we know fear has a paralyzing effect on people, likely extending to thought processes involved in forms of positive thinking. In contrast to our success in finding partial mediation pathways, we failed to identify moderators (that minimize the negative influence of uncertainty on mastery). Not surprisingly, the positive psychology variables of optimism and benefit finding were observed to have direct positive effects on hope. Both variables also influenced distress through their impact on general well-being, and benefit finding was additionally found to significantly impact appraisals of threat. Meanwhile, optimism was also shown to influence mastery (as hypothesized). These findings suggest that while people with a predisposition to thinking more  92 positively (i.e., „natural optimists‟) are at an advantage in curbing distress, distress reduction is nonetheless possible in other positive ways independent of one‟s dispositional nature; in particular, through building hope and enhancing well-being. [The fact that caregivers tended to „agree‟ with items on the hope and mastery measures yet „disagreed‟ with items on the optimism measure further confirms it is possible to experience positivity despite lacking a natural propensity to it.] These positive psychology constructs are likely advantageous because they enhance the frequency of positive thinking (at the expense of negative thinking) and ultimately the frequency of „positive affect‟, which recall is one of the components of Clark & Watson‟s (1991) tripartite model of depression. Indeed, research has shown that the relative proportion of time that people experience positive emotions relative to negative emotions is a far greater predictor of emotional adjustment then is the intensity of positive emotion (Diener, Sandvik, & Pavot, 1991). Finally, it is noted that our data do not support Mishel & Sorenson‟s (1991) MUIT model (see Figure 1), and thus several of our hypotheses could not be confirmed. While our model shares certain features with their uncertainty framework, there are noteworthy differences. Most notably, our model does not support a role for challenge. Mastery was therefore shown to have a direct effect on hope (and distress) in our model (rather than an indirect effect through the challenge pathway as originally hypothesized). In addition, our model shows the mastery and threat pathways to be independent, with threat being directly influenced by uncertainty. Possible reasons for these unexpected results are (i) we used different measures (i.e., PECI4, CAHS-R5, Ces-D6); (ii) we used more advanced statistical techniques; (iii) the low internal consistency of our challenge measure undermined the role of „challenge appraisals‟ in this context; and (iv) Mishel & Sorenson‟s (adult) model is not applicable to caregivers of pediatric cancer patients. In regards to the latter, it is possible that pediatric cancer is viewed as less threatening/challenging  93 relative to adult cancer, owing in large part to the improved treatment success rates over the past few decades. Our data support this finding, as caregivers on average reported scores of 3.4 (on a 1 to 5 scale; „neutral‟ ranking) on the challenge measure and (only slightly higher) scores of 3.6 (on a 1 to 5 scale; „agree‟ ranking) on the threat measure, suggesting these perceptions do occur but are not extreme. Relatively low correlations between threat and variables such as treatment status (r=-.198, p<.05) and child physical struggles (r=.222, p<.01) further attest to the lessthreatening nature of the pediatric cancer experience. Of note, these findings support our use of distress as an outcome variable rather than „post-traumatic stress‟ (or „acute stress reaction‟), given that these latter variables are characterized by high levels of perceived threat. Furthermore, with threat being less intensely experienced in this population it makes sense that mastery (i.e., reflecting one‟s perceived ability to overcome „negative‟ circumstances) was not highly associated with the threat variable (r=-.246, p<.05). Caregiver Subjective Happiness Despite elevated distress levels in the on-treatment caregivers, these caregivers rated their happiness levels as slightly better than average (as did all caregivers). [This level of happiness was also maintained over six months‟ follow-up in our longitudinal sample, which simultaneously showed „falling‟ distress levels over this time period.] This agrees with previous physiological, neurological, and emotion-based research showing that negative and positive emotions can co-exist and therefore are not polar opposites (Davidson, 1992; Lemonick, 2005; Watson, Clark, & Tellegen, 1988). Further support of this comes from our SEM analyses, where the distress and subjective happiness models were shown to diverge. All pathways in our happiness model were significant and explained influence in anticipated directions, again a testament of the model‟s strength and solid theoretical grounding. Over 60% of variance in both happiness and hope was explained by this model (i.e. similar to regression analyses), and ten partial mediation pathways were again identified. As in the distress  94 model, uncertainty, threat, mastery, optimism, benefit finding, and hope all played significant roles. Mastery was again identified as a crucial partial mediating variable in fostering hope, while hope, mastery, and (low) threat were shown to be key partial mediators in enhancing subjective happiness. Not surprisingly, caregiver physical struggles were also shown to have direct negative influence on happiness. As many of these predictor variables represent experiences that can be altered by context, these results provide hope that people‟s happiness levels can be sustainably changed under the right conditions. This contrasts with early beliefs in a „happiness set point‟ („hedonic treadmill theory‟; Brickman & Campbell, 1971). As in the distress model, threat was again shown to negatively influence aspects of hope. Interestingly, comparison of standardized path coefficients across the two models suggests that (low) threat was also a more powerful predictor of subjective happiness relative to distress (subjective happiness model: β = -.328; distress model: β = .248). It is possible that the fear that accompanies threat appraisals primarily restricts people‟s scopes of attention, cognition, and behavioural action, which are the main features of functioning that get broadened in a happy state („Broaden-and-build theory‟; Fredrickson, 2001). Areas of restriction in distress likely span more extensive areas (e.g., energy, eating, sleeping) of which threat does not impact, thus explaining the relatively lower predictive power of threat on distress. In regression analyses, other significant predictors of happiness included (not surprisingly) distress, perceived child social struggles, and child age, with total explained variance reaching 66.1%. The significant predictive power of child age here is not surprising given the greater likelihood of an altered developmental course in younger patients along with their elevated risk of „late effects‟. This knowledge likely weighs heavily on caregivers‟ minds and perhaps limits their levels of excitability, bliss, and overall positive affect. As a whole, the significant predictors of subjective happiness reflect the uplifting quality  95 of positive thinking, the numbness imposed by fear (discussed above), and the negativity created by symptoms of distress/depression (such as despair, withdrawal, and anhedonia). Hope, along with caregiver diagnosed mental health struggles, was also a (strong) predictor of happiness at six-months‟ follow-up, attesting to the strength of this variable as a predictor of happiness. Conservative model testing further suggested that positive predictor variables rank superior (to negative predictive variables) in influencing happiness. These results highlight that there is clearly more to being happy than „not being distressed‟. It is likely that the endorphins, elevated dopamine levels, and enhanced left frontal cortical activity that uniquely characterizes positive emotion states (as opposed to the right frontal cortical activity and reductions in serotonin and norepinephrine that characterize negative emotion states) all play a crucial role in enabling the superior effects of the positive predictor variables. Adolescent Quality of Life In the adolescent analyses, teenagers described struggles across all (physical, emotional, social, and school) QOL domains as „almost never‟ a problem, regardless of treatment status. This is in parallel with findings that these children are adjusting well (Phipps, 2005, 2007; Trask et al., 2003). Unlike most pediatric research which has found older age to predict more problems (Kazak & Simms, 1996), as well as some which has found younger children to be more at-risk (Engelen et al., 2010), age was not significantly associated with quality of life in this study. Despite only looking at four predictor variables in our regression analyses (due to limited power), we were able to explain 44% of variance in teen quality of life. The fact that three of these predictors were caregiver variables and the remaining predictor was a precursor variable suggests a lot can be ascertained about the well-being of these children from outside sources (i.e., parent interview, chart review). Not surprisingly, vulnerable teens in our study were those who came from limited income households and had parents who perceived greater social struggles in them. As we failed to find significant positive predictors of teen quality of life,  96 pathways for patient/survivor resilience remain to be discovered. While these surely do exist, realizing benefits in one‟s cancer journey and having important needs met do not appear to be factors that enable these children to flourish, which goes against our study hypotheses. When we looked at concordance between caregiver-adolescent reports on similar variables, we observed only a few correlations at the .001 level. All of these variables related to medical constructs (i.e., child physical struggles, importance of health information, and importance of involvement with care). Although concordance findings pertaining to most of the FNQ subscales should be interpreted with great caution given the low alpha scores associated with these subscales, these findings nevertheless make sense. High concordance on „medical‟ variables is not surprising as these variables represent experiences that tend to be dependent on the type of cancer present, which is a shared (and often visible) experience among caregiveradolescent dyads. Meanwhile, perceptions of child emotional and social struggles across caregivers and teens correlated at the .01 level. These findings confirm previous literature showing that parents are good, but not perfect, reporters of child psychological experiences (Engelen et al., 2010; Sawyer, Antoniou, Toogood, & Rice, 1999). As with previous research, caregivers in our study tended to report more problems than their children did. Of note, benefit finding among caregivers and teens correlated at the .05 level, likely reflecting some effects of shared environment, religion, socio-economic status, and/or modeling influences (i.e., coping styles, outlook on life). Important Needs (Caregivers and Adolescents) When it came to the research on family needs, both caregivers and teens rated the majority of these as important, though on average only 62% and 71% (respectively) of these were reported to be „met‟. This confirms that family needs are an important aspect of life to this population, and that there is room for improvement in getting these needs met. Surprisingly, significant differences across urban versus rural populations were not found on any family needs  97 measure. Though the weak internal consistency of (some of) the FNQ subscales may shed doubt on these findings, similar conclusions from the (more reliable) total scales do reaffirm them. These findings do not support a hypothesis that rural families have greater unmet needs, and they help explain why we failed to observe a main effect for time to hospital in our (caregiver) distress and (teen) quality of life analyses. In line with previous reports that rural persons worry less about health concerns and are less likely to accept the illness role (Long & Weinert, 1989), our research suggests rural families are not overwhelmingly impacted by their inferior access to resources. Also surprising, and against our hypotheses, was the finding that having family needs met (broadly speaking) did not significantly predict any of our main caregiver or child outcome variables. However, when needs were broken down across sub-domains, having healthinformation needs and professional support needs ‘MET‟ did significantly predict (in interaction terms) caregiver perceptions of mastery. This agrees with research by Kaplan (1982) and suggests that having valued needs met does impact caregiver emotional adjustment. Moreover, it is possible that these (or other) domain-specific „met‟ needs would have predicted significant variance in our caregiver emotional outcome variables had we had enough power to test these individual domain scores as predictors. Treatment groups did not differ when it came to rating the importance of needs, highlighting that off-treatment families (an often „neglected‟ population) still seek attention. Again, while the weak internal consistency of (most of) the FNQ IMP subscales may shed doubt on these findings, similar conclusions from the (more reliable) total scale reaffirm them. Indeed, off-treatment adolescents reported a significantly higher amount of unmet involvement with care needs relative to on-treatment adolescents, attesting to this neglect in the post-treatment phase. Medically related needs exceeded personal needs in both populations, which is not surprising  98 given the increasingly well-documented desire for health-related information in caregivers (Knoderer, 2009; Poder & VON Essen, 2009; Trask, Welch, Manley, Jelalian, & Schwartz, 2009; McKenna et al., 2009) and patients (D‟Alton, O‟Malley, O‟Donnell, Gill, & Canny, 2010; Singer et al., 2010). Our research did show this type of information to be of particular benefit to caregivers, as having health information needs met was found to play a significant role in altering the impact of uncertainty on mastery. Important personal needs (such as emotional support) showed the lowest rates of being met across caregivers and teens. However, this is not overly concerning given that these types of needs were least desired and having them met was not shown to influence adjustment. There are several ways to explain the insignificance of emotional support in these populations. First, it is possible that this type of support is not in demand due to relatively low „fear‟ levels in these populations (compared to those seen in adult cancer patients and caregivers) stemming from more optimistic prognoses. Another possibility is that receiving medical information/ consultation trumps the need for emotional support from family and friends as medical staff do a sufficient job of alleviating concerns, reducing feelings of isolation, and cultivating empowerment (in caregivers, through legitimizing their important role). Lastly, it is possible that emotional support is ineffective in this context when provided by people who have not experienced a similar stressor. Beyond those areas assessed, qualitative analyses highlighted a strong demand for „complementary and alternative medicine‟ (CAM) resources in this population, a need that currently appears to be largely unmet. In a recent meta-analytic review of CAM in pediatric cancer, prevalence of CAM use ranged from 6% to 91%, with herbal remedies being the most popular (Bishop et al., 2010). Impressively, „benefits‟ from CAM use have been reported in up to 83% of cases (Clerici, Veneroni, Giacon, Mariani, & Fossati-Bellani, 2009).  99 In the area of school needs and resources, we observed off-treatment caregivers to have a significantly greater demand, likely reflecting the impact of late effects. Demand appeared to be greatest for „individual teacher support‟ and „educational assistant support‟. Not surprisingly, only around 41% of important needs were reported to be met. This speaks to the value of increasing access to neuropsychological assessments, so that difficulties in cognitive, motor, and sensory functioning can be appropriately diagnoses and attended to. Though missing data prevented us from assessing whether having these types of needs met significantly predicts caregiver emotional adjustment, it is highly likely that meeting these needs is important to the (current and long-term) functioning of these children. While history of learning disability could also not be assessed as a predictor variable in our analyses due to its significant dichotomous split, it was significantly associated with teen quality of life in our correlation analyses. Moreover, the detrimental impact of late effects, once they onset, on both caregivers and patients has been well documented (Edelstein et al., 2010). Benefit Finding (Caregivers and Adolescents) To date, this is the first study to assess „benefit finding’ in caregivers of pediatric cancer patients7 and only the sixth study to asses it in these adolescent patients/survivors. We found that both caregivers and teens reported „quite a bit‟ of positive outcomes from their cancer experiences, confirming that benefit finding is a reality in these populations (Currier, Hermes, & Phipps, 2009; Helgeson et al., 2006; Michel, Taylor, Absolom, & Eiser, 2010). Similar to previous research on caregivers (of adult cancer patients; Kim et al., 2007), benefit finding in domains of reprioritization and acceptance was most common while benefit finding in domains of positive self-view and appreciation was least common. Interestingly, benefit finding was not found to be predictive of (or for the most part even associated with) measures of immediate or future emotional adjustment in these populations. These null findings parallel previous reports in studies assessing the impact of benefit finding on caregiver distress (Kim et al., 2007) and  100 adolescent QOL (Castellano et al., 2010; Phipps et al.). Nonetheless, benefit finding was shown to significantly influence cognitive processes through its impact on both hope and threat (as observed in SEM analyses). In addition, certain domains of benefit finding (i.e., acceptance, appreciation) were shown to moderate the negative impact of uncertainty on mastery in regression analyses. Of note, it is possible that these benefit finding domain scores would have predicted significant variance in our caregiver emotional outcomes variables, had we had sufficient power to test them. It is possible that while perceiving benefits generates sufficient positivity to propel further positive thinking (i.e., hope) and minimize negative thinking (i.e., threat), the positivity is not intense enough to influence affective outcome. This lack of intensity may stem from minimal levels of meaning finding in this population, a reflection of the less-stressful nature of pediatric cancer (where prognosis tends to be good). Another way to explain the insufficient (emotional) driving force of benefit finding is that perhaps our participants did not previously realize their benefits (i.e., prior to completing the questionnaire). This would suggest our measurement of „benefit finding within the cancer experience‟ was invalid. A third possibility is that the benefit finding variable (i.e. total score) is composed of (emotionally) contrasting domain scores that cancel out each other‟s effects in the overall adjustment realm. This latter theory is supported by the ambiguous findings reported in research with caregivers (Kim et al., 2007) and adults (Tomich & Helgeson, 2004) that were discussed earlier in the introduction section. In regression analyses, caregiver benefit finding was significantly predicted by perceptions of challenge, valuing family needs (especially, emotional support), valuing school needs and resources, and being non-Caucasian. While the finding pertaining to challenge must be interpreted with caution given the low alpha score for this measure, it makes sense that „cognitive flexibility and positivity in the face of stress‟ (i.e., challenge) is predictive of further  101 positive thinking (i.e., benefit finding). The other significant predictors of benefit finding are hypothesized to additionally reflect the „positivity-generating‟ and „meaning-finding‟ capacities of emotional sensitivity, aware and pro-active parenting, and affiliation with religion and/or lower socio-economic status. [With respect to the latter, religion (but not spirituality) may be associated with benefit finding due to greater practice with positive reframing and finding deeper meaning in experiences. Meanwhile, lower socio-economic status may be associated with benefit finding due to greater experience with hardship and thus more practice in finding the silver lining in experiences.] The non-Caucasian effect replicates previous findings by Stanton et al. (2006). Time since diagnosis failed to associate with benefit finding in either of our populations, suggesting that levels of benefit finding remain stable across the cancer trajectory. This agrees with previous findings in caregivers (Kim et al., 2007) but not in pediatric patients (Phipps et al., 2007). In addition, while our significant (negative) correlation between teen benefit finding and child age also goes against the null finding reported by Phipps et al., this discrepancy is not overly meaningful given that (i) our correlation was weak (at the .05 level) and did not translate into a significant predictor relationship; and (ii) our sample was composed entirely of teenagers so comparison to Phipps et al.‟s younger sample is not appropriate. This latter point may also explain our contrasting results with respect to time since diagnosis. Of note, no moderator influences on benefit finding were identified and (in contrast to adult patient findings) optimism was not shown to be a significant predictor of benefit finding. The latter finding highlights that benefit finding is not restricted to just those people prone to positivity (i.e., „natural optimists‟). Of note, our insignificant correlation between benefit finding and uncertainty also disproves an early theory by Mishel & Grant (1992), stating that positive („probabilistic‟) thinking only occurs in the oncology realm once uncertainty levels have exceeded a certain threshold.  102 Overall, across both caregiver and adolescent models, approximately one-third of variance in benefit-finding was explained. Clearly more variables need to be studied for their influence on this form of positive thinking. This is particularly evident in the adolescent realm, as we failed to identify any significant predictors of teen benefit finding. Caregiver Hope While we discovered that help in maintaining hope is a highly desired need by both caregivers and teens, we observed this need to be greatly underachieved (recall, only 58.3% met in caregivers, 68.3% met in teens). Interestingly, valuing help in maintaining hope was not a significantly correlated construct across caregivers and teens, nor was receiving help in maintaining hope. This was a surprising result given the pivotal role of social support in hope (Herth, 1992; Snyder et al., 1996). Perhaps the support provided by peers, siblings, spouses, and medical staff plays a greater role in sustaining hope in the pediatric cancer experience than does the support offered in the context of the caregiver-patient dyad. In regression analyses, our predictor variables explained over 60% of variance in hope. Optimism, mastery, subjective happiness, and (to a lesser extent) benefit finding and distress were all observed to be significant predictors. These influences likely reflect the uplifting consequences of positive thinking and positive mood (previously discussed), as well as the incapacitating impact of symptoms associated with distress/depression (i.e., low energy, loss of pleasure, problems with sleep and concentration). The significance of benefit finding here is not surprising given previous findings in which positive reappraisal coping has been associated with elevated hope (Stanton et al., 2002; Wonghongkul et al., 2000). In contrast to previous research, however, we found no significant correlations between socio-demographic or cancer-specific variables and hope, suggesting that hope is observed (and possible) across a myriad of individuals and contexts.  103 Study Limitations There are several limitations to our study which should be noted. First, while our sample size was sufficient, it was not large enough to power thorough analyses. Particularly with the adolescent and longitudinal datasets, many variables could not be examined as predictors in regression analyses due to restrictions imposed by a limited sample size. SEM techniques could also not be used with these datasets given their small size. Similarly, predictor variables in (baseline) SEM analyses were limited by our number of participants. Additional reasons preventing variables from being analyzed in our study included inaccurate responding and nonrandom missing data. We also observed several cases of non-normality in our data, including extreme splits in many (socio-demographic and cancer-specific) dichotomous variables. While most of these variables were removed from inferential analyses, some were retained (with warnings noted advising caution in interpretation). Owing to weak internal consistency (α) scores, we also had to interpret findings pertaining to the challenge measure and some of the FNQ subscales with caution. Given that the alpha scores for several of our FNQ subscales (i.e., α =.46 to .67) were quite a bit lower than those reported in the literature (α=.78 to .89; Serio et al., 1997), we preferentially used the (more reliable) total scales in place of these subscales. When the subscales were analyzed in supplemental analyses, cautions were noted. Meanwhile, while our alpha score of .46 for the challenge scale is disappointingly low, it is not overly surprising given that previous literature has documented reliability for this 3-item scale only as high as α=.70 (Ahmad, 2005). While the general insignificance of challenge in the current study is presumed to reflect the fact that pediatric cancer is a (relatively) less threatening experience (relative to adult cancer, where prognosis is much worse), it is possible that these null findings resulted from an unreliable measure. Replication testing, perhaps with an improved measure of challenge appraisal, is  104 needed to determine whether this was the case. Another limitation to our study involves the makeup of our sample. First, caregivers were predominately female, educated, married, and Caucasian. Few families experienced relapse and none were in palliative situations. Consequently, findings may not generalize to a broader population of caregivers. Second, while heterogeneity with respect to cancer type, stage, and (actual) prognosis provided for a more representative sample of cancer families, it likely confounded our outcome data as distress (and happiness) levels presumably vary significantly across different values of these indicators. Unfortunately, we were unable to control for these confounds as data on staging and prognosis were unavailable, and data on cancer type could not be used in inferential analyses. Third, our „on-treatment‟ sub-group represented a heterogenous population as it included newly diagnosed patients just starting treatment as well as patients who had been on treatment for up to two years. [Power and time restrictions prevented the establishment of a more acutely ill sample here.] Fourth, selection bias became evident at the later stages of participant recruitment due to a shortage of on-treatment families. On-treatment families were eventually targeted for recruitment immediately following diagnosis, resulting in non-random sampling. Lastly, we did not have healthy control groups for our caregivers or adolescent patients, preventing normative comparisons from being made. Finally, some limitations are noted relating to our statistical analyses. First, the bulk of our analyses utilized cross-sectional data and thus „prediction‟ in these analyses represents relations of correlation rather than causation. Second, results from our SEM analyses should be interpreted with caution as they were exploratory and did involve some post hoc „data fishing‟. As extra analyses may have increased opportunities for chance effects, our findings require replication to enhance their reliability. In regards to these SEM analyses, it must also be stated that the „causal‟ relations discussed can only be considered „approximations‟ which can never be  105 proven true, only disproven. Moreover, while these „causal inferences‟ were consistent with our data, SEM does not specify directionality. Thus, it is possible that for each pathway identified, directions of influence occur in either or both directions. This research focused specifically on one direction of influence (i.e., from uncertainty to mastery/threat to hope to distress/subjective happiness) for three reasons: (i) this was the direction of influence previously discussed in the MUIT literature, and consistency with the literature was desired; (ii) our longitudinal analyses confirmed (the directionality of) some of these causal relations; and (iii) this direction of influence is most informative and conceptually valid from a clinical standpoint. When thinking about ways to screen for distress and intervene with those caregivers in need of help, it is far more important to understand those consequences that arise from high levels of uncertainty (which is a „given‟ in the cancer experience and something that is not likely to attenuate) and in turn hinder emotional adjustment, than it is to understand what causes uncertainty or what consequences arise from poor emotional adjustment. If our overall goal is to prevent (or reduce) the escalation of distress, we need to have explanations for what causes this phenomenon. Of course, experimental intervention will ultimately determine whether such causal inferences are true. Value of Study Given the level of uncertainty that pervades early treatment as well as the knowledge that two-thirds of survivors experience at least one late effect (Landier & Bhatia, 2008), it is not surprising that our research shows caregivers struggling, particularly in the early phase of treatment. The need for routine psychosocial care directed at caregivers is increasingly being voiced by researchers, professional bodies, policy makers, and advocacy groups (Hoven, Anclair, Samuelsson, Kogner, & Boman (2008); Klassen et al., 2008; Vrijmoet-Wiersma et al., 2008). While major gains in this area have been made in recent years in the adult patient sector (Holland, 2010), these have not been matched in the pediatric realm.  106 Through identifying direct and indirect influences on caregiver emotional adjustment in the context of pediatric cancer, our modeling research has provided a more thorough understanding of how this process might operate. Though more research is needed, feedback from studies like this should be distributed to local healthcare providers, administrators, and families. Armed with a better knowledge of what variables influence positive and negative adjustment, and how they do this, people will be in a better position to identify individuals at risk and guide evidence-based interventions. With distress screening recently being characterized as „the 6th vital sign‟ in cancer care (Howell et al., 2010), the current literature confirms a strong need for timely screening, information delivery, and evidence-based intervention efforts (Currie et al., 2010; Fitch, 2010; Trask et al., 2003). Screening Given that (front line) physicians and nurses are not formally trained in identifying atrisk individuals based on psychosocial presentation (Kiernan, Meyler, & Guerin, 2010), there is a need to develop user-friendly screening tools for their use. Research like the current study is ideally suited to direct the development of such tools. While data from this study confirm that most teens are adjusting well, it highlighted that „at-risk‟ teens can be identified, in large part, by factors such as household income and caregiver perceived child social struggles. As these indicators are accessible from parent interview and/or chart review, it seems quick and noninvasive screening to detect emotionally vulnerable children is possible. Also surfacing as an important aspect of child screening was the role of neuropsychological assessments, which can identify areas of struggle in learning, behavioural, and physical realms. These invaluable screens, particularly important in cases of leukemia and brain cancer, can ultimately assist with school reentry, access school accommodations for special needs, direct referrals for more specialized assessment, and recommend interventions to improve functioning. In regards to caregiver screening, the time efficiency, administrative simplicity, and  107 strong psychometric properties of the CES-D make it particularly useful as a screening measure for emotional distress in cancer populations (Hopko et al., 2008). However, analyses from this research have also revealed the importance of being attentive to variables such as uncertainty, mental health struggles, and perceived difficulties in transportation. As time post diagnosis, child age, and (perceived) child emotional struggles were also found to be significant predictors of adjustment, it is important that screening commence early in this population and, in particular, target caregivers of young children and emotionally vulnerable children. Both the 23-item PedsQL and the more sophisticated Behavioral Assessment System for Children (BASC-II; Reynolds & Kamphaus, 2004) are viable parent screening tools for child emotional vulnerability in this population (Banks, Barrowman, & Klassen, 2008; Wolfe-Christensen, Mullins, Stinnett, Carpentier, & Fedele, 2009). Given that happiness was found to be a particularly strong predictor of distress at baseline, the 4-item SHS may also be an efficient screening measure here. In addition, the superiority of long-term uncertainty as a predictor variable in longitudinal analyses suggests that the five items from the PECI (used to measure long-term uncertainty in this study) would be a good screening tool for long-term distress. Another area for screening relevant to both on- and off-treatment families relates to family needs. In particular, screening for needs in medical, personal, and complementary/alternative medicine areas should occur regularly. While such screening may not be crucial to caregiver emotional adjustment, it is in high demand and would greatly improve satisfaction with our healthcare system. Intervention In the realm of intervention, new efforts for these caregivers may focus on techniques that reduce perceptions of threat and enhance experiences of mastery, hope, and benefit finding, as these were found to be global experiences that significantly associated with better emotional adjustment. Based on our analyses as well as clinical insight, interventions targeting threat and mastery should perhaps focus on psycho-education as this has the potential to enhance  108 understanding, feelings of control, and acceptance of one‟s situation. Becoming more informed can be both comforting and empowering to caregivers of cancer patients, and it is therefore suggested that this type of intervention be given to all caregivers (not just those at risk for emotional distress). In this realm, information and resource provision should be made readily available as a routine service, offering general information on what to expect (in terms of symptoms, treatment, outcomes, sides effects), how to connect with others in similar situations, ways to reduce stress and enhance self-care, and types of CAM resources available (noting the benefits and risks of these; Clerici et al., 2009). Strategies to enhance benefit finding, in contrast, may want to utilize mindfulness-based techniques, which tend to promote lifestyles more conducive to positive thinking and meaning finding. This approach to stress reduction cultivates a propensity towards greater awareness, acceptance, and appreciation for things in the moment (Lawlor-Savage, Labelle, Campbell, & Carlson, 2010). While this is related to Folkman & Moskowitz‟s (2004) concept of „meaningfocused coping‟, we perceive this more as „meaning-focused being‟ as it involves changes in one‟s approach to life rather than in one‟s approach to combating stress. Exercises used in mindfulness-based approaches range from formal meditation (i.e., sitting quietly for up to 45 minutes while directing attention in specific ways) to less formal focus on daily life experiences (i.e., walking, bathing, driving, eating, breathing) while noting the what/where/why/when‟s of sensory experiences as they happen. A recent longitudinal study by Hirshberger, Peretz, and Mikulincer (2010) confirmed that adult cancer patients high in mindfulness were better able to reconstruct meaning and experience growth, outcomes likely associated with more benefit finding. Mindfulness also seems applicable to cancer patients and their caregivers given its focus on accepting things as they are (i.e., without change) – something important to the cancer experience, where stressors are often unchangeable.  109 Meanwhile, hope-focused work may center on goal-setting (and confiding in supportive others about these) given the ability of this technique to establish life priorities, foster motivation, and produce success. As hope is a dynamic process (Wirga, 2010) that varies across people and across time within any one person, it is important that caregivers revisit goal-setting on a regular basis to identify what their „hope statement‟ involves at that point in time. It is also imperative that goals are clear, specific, and appropriately suited to the individual‟s current situation (i.e., realistic). Art therapy, drama therapy, and music therapy are other types of interventions that can build hope, specifically through self-affirmation work that instills senses of achievement, worth, and purpose. Further strategies to enhance hope have been incorporated into a more standardized clinical approach, called „hope therapy‟ (Lopez, Floyd, Ulven, & Snyder, 2000). Regardless of the techniques used, it is important to realize that empowering hope does not mean denying suffering or raising „false hope‟. Rather, it involves motivating people to live in the present and appreciate that good things exist and that targeting small, shortterm goals is possible and constructive (Crowell & Strahlendorf, 2007; Wirga, 2010). In addition to these techniques, interventions promoting change in caregivers‟ „perceptions‟ of difficult travel may also be useful. This could be achieved through such things as normalizing (of difficulties all around), broadening one‟s time perspective (to include a healthy balance of past, present, and future focus), and „pleasant events scheduling‟ (for things to do when travel completed). Caregivers with diagnosed mental health struggles and/or chronic depression will likely require more intense intervention in the form of individual therapy that incorporates training in problem-solving and emotion-focused counselling. Given that this more family-focused approach requires significantly more time and resources, it should only be provided when other intervention efforts (i.e., psycho-education, hope building, promotion of benefit finding) have proven unsuccessful.  110 In regards to interventions for emotionally fragile children, our findings suggest that techniques aimed at alleviating social struggles may be most helpful. Encouraging increased independence (i.e., from parents) as well as regular social interaction with same-age peers would therefore be indicated. Where struggles are more pronounced, training in social skills and assertive behaviours (ideally in the context of group work) should be encouraged. Improving social relations is a desirable intervention target for any child given the many benefits that can follow from such an approach (i.e., improved self-esteem, enlarged social network, more frequent positive emotion, enhanced skill learning, etc.). Service Delivery Screening and subsequent intervention must be provided as early as possible in the cancer trajectory as prior research has shown initial adjustment to be predictive of later adjustment, among both children with chronic illness and their parents (Phipps, Dunavant, Lensing, & Rai, 2005). To increase the efficiency of this process, information collection and distribution could occur, at least in part, through such time-saving mediums as clinic handouts, hospital volunteers, CD-Rom‟s, annual seminars, wellness centre classes, or online newsletters and support groups. [Additional e-health interventions that make use of texting, twittering, blogging, Facebook, webcams, and chat rooms may be particularly attractive to adolescent patients/survivors as they likely find this familiar form of communication more interesting and less threatening.] Such communications should continue to occur following treatment as well as years into survivorship. The success of the following recent international initiatives aimed at addressing caregiver and patient psychosocial care proves that advanced technologies in this area are possible and promising: the „Children‟s Cancer Network‟ in Arizona (Luttrell, 2009), the „Cancer Helplines‟ and „Videotelephone-based supports‟ provided through the Centre for Online Health at the University of Queensland (Bensink et al., 2008; Chambers et al., 2010), the web-based education  111 program entitled „Coping with Cancer‟ at the University of Minnesota (O‟Conner-Von, 2009), the on-line psycho-educational group interventions and use of pre-appointment „digital QOL PROfiles‟ at Emma Children‟s Hospital in the Netherlands (Engelen, Detmar, Koopman, & Grootenhuis, 2010; Maurice-Stam, Silberbusch, Last, & Grootenhuis, 2009), and the „Dischargeplanning program‟ at the Ege University Pediatric Oncology Clinic in Turkey (Caliskan, Yilmaz, & Ozsoy, 2010). These types of services, which mainly incorporate features of distress screening, emotional support, and pscho-education, have been shown to yield substantial improvements in such areas as distress, anxiety, hope, and feelings of efficacy to cope (i.e., „mastery’) (Sinardo & Zupko, 2010). While Canadian programs such as the Canadian Cancer Society (2010), Cancer Care Nova Scotia (Fillion et al., 2010), and CancerviewCanada (Evans, Fitch, Greenberg, Kitchen-Clarke, & Rowland, 2010) have recently launched (emotional and informational) supportive services for adult patients (i.e., the Cancer Information (telephone) Service Follow-up Program, Cancer Patient Navigator, and www.cancerview.ca/thetruthofit, respectively), similar services are lacking in the (Canadian) pediatric realm. Moreover, interventions incorporating goal-setting and self-affirmation work (to build hope) as well as meditation and mindfulness (to promote benefit finding) are only in their infancy in international adult patient populations, and they have yet to be established in the pediatric realm worldwide. With new insights from this research, the potential exists to develop (and empirically test) advanced screening and intervention services specialized for use with these pediatric patients and their caregivers. Application to Other Pediatric Illnesses There are several critical reasons why the adjustment model identified in this research will likely not generalize perfectly to other pediatric illnesses. First, cancer has the potential to be a fatal disease. Second, in pediatric cancer there is often hope for remission (unlike many  112 pediatric illnesses associated with risk of death). Third, cancer treatment yields a likelihood of „late effects‟ evolving in which these children experience cognitive, behavioural, emotional, and adaptive declines in the years following treatment. This unique combination of characteristics sets pediatric cancer apart from most (if not all) other pediatric illnesses, and makes variables like uncertainty, threat, and hope particularly relevant to the pediatric cancer experience. Nonetheless, given that most pediatric chronic illnesses are associated with some level of uncertainty and negative impact on child functioning, aspects of the adjustment model (and some of the intervention efforts deduced from it) might be applicable to other populations. While this may be more the case with illnesses associated with a greater risk of death (e.g., cystic fibrosis, muscular dystrophy), some level of generalizability is anticipated to extend to other chronic illnesses such as autism spectrum disorders, rheumatoid arthritis, diabetes, epilepsy, and severe asthma. Future Directions Future research efforts should aim to replicate our current model findings, ideally in a sample with good representation of males, ethnic minorities, single parents, child learning disabilities, and cases of relapse. These replication studies should recruit larger samples to allow for the (eventual) inclusion of further predictor variables. In the realm of benefit finding, focus may want to shift to domain scores given the potential risk for null findings with the total score due to inter-scale cancellation effects. Researchers may also want to inquire about the degree to which benefits were previously realized (i.e., prior to questionnaire completion), as this may better tap into benefit finding during the cancer experience. In the realm of challenge appraisal, an improved measure of this may need to be used if poor internal consistency with the 3-item CAHS-R scale is again observed. Replication studies should also examine a more acutely ill sample for their „on-treatment‟ group. Lastly, longitudinal analyses should expand to include follow-up measurements of all baseline predictor variables, as latent change analysis could then  113 be used to examine how relationships within the model change as a function of time along the disease trajectory. While our regression and structural equation models were strong, explained variance percentages were nonetheless well shy of 100%. This indicates more variables need to be studied for their influence on this population. Based on findings from our study, it would seem advisable to assess the impact of CAM use (given strong desires for this resource). Assessing cancer stage also seems warranted, both because uncertainty was found to be a significant predictor variable, and because it would be ideal to have a more individualistic measure of prognosis. Using patient questionnaire variables as predictors in caregiver adjustment may also produce noteworthy findings8. In addition, a superior measure of financial strain (such as „household savings‟ or „net income‟) is recommended as this has been shown to be a significant risk factor for both caregivers (Klassen et al., 2008) and pediatric patients (Sung et al., 2009). Lastly, while the present study did not use subjective happiness as a predictor of distress (and vice versa) in modeling analyses due to reasons previously discussed, future modeling research should incorporate these significant influences. In an effort to identify further predictors of caregiver benefit finding, hope, distress, happiness, and mastery, new factors need to be assessed. Indicators of personality (such as extraversion, agreeableness, and conscientiousness) may be one interesting area of exploration here. According to recent literature, other variables significantly correlating with caregiver adjustment have included health behaviours, chemotherapy intensity, anxiety, and perceived „change‟ in child‟s functioning. More specifically, better eating, exercise, and sleep habits have all been associated with higher parental quality of life, and more intense chemotherapy has been highlighted as a risk factor for worse caregiver (and patient) adjustment (Klassen et al., 2008; Sung et al., 2009). Trait anxiety (Kuroda et al., 2010) and post-traumatic stress symptoms  114 (Rabineau, Mabe, & Vega, 2008) have also been linked to higher distress in adult patients and caregivers, respectively, and it is possible that anxiety underlies the predictive power of diagnosed mental health struggles in our study. Lastly, recent research has shown that the „discrepancy‟ between caregiver perceptions of their child‟s functioning prior to cancer diagnosis versus during cancer is more related to caregiver distress than is the actual extent of the disability (Hullman, Wolfe-Christensen, Meyer, McNall-Knapp, & Mullins, 2010). Further modeling research should examine the impact of these variables on caregiver emotional adjustment. The current literature also highlights variables deserving further attention for their impact on hope and mastery. For example, „peer‟ support (i.e., support provided by other cancer survivors/families) has been shown to be a powerful influence on hope (Taylor, 2010) and it is possible that this type of social support is what is most beneficial to caregivers (Wortman, 1984). Meaning-making is a further variable that should be assessed more thoroughly for its impact on benefit finding and hope. Meanwhile, in an effort to identify predictors of mastery, the positive psychology literature pertaining to „strengths‟ appraisal (Peterson & Seligman, 2004) should be examined, as cultivation of one‟s identified character strengths is believed to foster improved self-esteem, more frequent positive affect, and reductions in distress (Linley & Joseph, 2004). In the context of adolescent adjustment, we must continue searching for resilience pathways, ideally using modeling techniques. Hope, self-competence, and self-worth are variables that have been associated with positive outcomes and thus deserve further study in this population (Eapen et al., 2008; Maurice-Stam, Grootenhuis, Brons, Caron, & Last, 2007). Similar to caregivers, strengths appraisal and health behaviours should be assessed for influence on teen emotional outcomes (Carpentier, Mullins, Elkin, & Wolfe-Christensen, 2008). In addition, the utility of weekend retreats should be analyzed as these have been shown to yield  115 many benefits for adolescents, including improved quality of life and hope (Robinson & Rutledge, 2010a), with long-term gains being noted (Robinson & Rutledge, 2010b). Other variables recently receiving attention include family protective factors and perceived burden. More specifically, family cohesion and expression (Jobe-Shields, Alderfer, Barrera, Vannatta, Currier, & Phipps, 2009) as well as family conflict (Ach et al., 2010) have been shown to impact emotional and education outcomes. In addition, parenting styles such as „overprotection‟, „flexibility‟, and „maintaining normalcy‟ may be of interest here. Lastly, „cancer consequences‟ was recently found to explain 36% of variance in pediatric quality of life (Fonseca et al., 2010), attesting to the significant influence of perceived burden on this population. In this regard, the BFS has recently been revised to include indicators of burden (Currier et al., 2009), and recent research has revealed a more powerful impact of this variable relative to benefit finding (Grootenhuis et al., 2010). Perhaps the adaptive nature of benefit finding is overshadowed by feelings of burden, and thus perceptions of benefits are only helpful when perceptions of burden are reduced. Further study awaits such hypotheses. As more thorough models on caregiver emotional adjustment evolve, distress screening and intervention research should proceed (where not already done so) with randomized controlled experiments testing the effectiveness of such „change‟ efforts. By targeting change in „risk factors‟ in this way, we can more validly deduce whether these factors are indeed causally related to emotional outcomes. This research should ideally incorporate longitudinal time points to measure maintenance of gains over time. Through this research, the effectiveness of such strategies as neuropsychological and happiness screening, e-mental health informational technologies, peer supports, and (positive) cognitive exercises can be adequately assessed. Ultimately, change efforts yielding clinically significant improvements should be implemented into routine pediatric cancer care. These changes hold the promise for more efficient resource  116 allocation, less healthcare costs, and improved family care – end goals desired by everyone, including administrators, hospital staff, and our pediatric cancer families.  117 Footnotes 1  In the regression analyses, note that the term „positive predictors‟ incorporates both the „positive‟ predictor variables and the „neutral‟ predictor variables. 2  As „on-treatment‟ status in our study was defined as „currently receiving treatment‟ or „off treatment for no more than 1 month‟, and given that our baseline on-treatment sub-sample reported significantly elevated distress relative to the off-treatment sub-sample, our data suggest distress levels start subsiding no earlier than 2 months post diagnosis. 3  The failure of caregiver gender, household income (i.e., „employment‟), and marital status to predict significance variance in distress should be interpreted with caution given the extent of non-normality in these variables. Moreover, the insignificance of ethnicity as a predictor variable may be due to the lack of African American caregivers in our sample (n=1), as this tends to be the most at-risk sub-population. Lastly, while we observed a significant correlation (at the .05 level) between other current stressors and distress, the relationship was not strong enough to be predictive. Our use of open-ended questioning (rather than assessment of specific stressors) to measure this variable may account for this null finding. 4  Whereas we used the PECI to measure uncertainty, Mishel used the Mishel Uncertainty in Illness Scale (MUIS; Mishel, 1981) and, later, the Parent Perception of Uncertainty Scale (PPUS; Mishel, 1983). Unlike Mishel‟s scales, the PECI has been validated in parents of pediatric oncology patients in both acute and chronic phases of care. In addition to being poorly applicable to chronically ill populations, the PPUS has been accused of lacking sensitivity to current-day uncertainties in the health care context (Santacroce, 2002). With these limitations in mind, the PECI was used in place of the PPUS. 5  Whereas we used the CAHS-R to measure stress appraisal (threat, challenge), Mishel used the appraisal scale from Folkman & Lazarus‟ Ways of Coping Checklist (Folkman, 1982), collapsing the four subscales of threat, challenge, harm, and benefit into the two categories of danger and opportunity. We decided to use the CAHS-R over the latter method due to its shorter length and purer measurement of threat and challenge. [More recent research on the MUIT has also moved towards examining threat and challenge as indicators of appraisal, rather than danger and opportunity (Wonghongkul et al., 2000).] 6  Whereas we used the CES-D to measure distress, Mishel used the distress scale of the Profile of Mood States (POMS; McNair , Lorr, & Droppleman, 1971) in her early work. We decided to use the CES-D over the POMS due to its shorter length (and because Mishel made use of the CES-D in later work). 7  While researchers have purported to study „benefit finding‟ in caregivers of pediatric oncology patients (Michel et al., 2010), none to date have studied („true‟) benefit finding using the BFS. 8  We were unable to assess the impact of youth questionnaire variables on caregiver functioning in this study because this type of data were only gathered in N=40 cases. As ¾ of the sample would have had this data missing (non-randomly), this data was not used.  118  Tables  119  Table 1 List of Precursor Variables (to be obtained from the Socio-Demographic Form)  Caregiver Variables (Socio-demographic)  Child Variables (Socio-demographic & Cancer-specific)  _____________________________________________________________________________                 Age Gender Educational level Diagnosed mental health struggles (past/present)1 Physical health struggles Religiosity/spirituality Ethnic background Time to hospital (or urban/rural status) Mode of transport Ease of transport Marital status Relationship to child (mom, dad, etc.) Type of relationship to child (natural, adopted, foster)# of children Household Income (i.e., # of earners)2 Other family stressors (unrelated to cancer)  Additional variables4:  Mood medication status (on/off)  Previous experience with cancer 1             Age Gender Grade Diagnosed mental health struggles Physical health struggles (unrelated to cancer) Religiosity/spirituality Ethnic background Diagnosed learning disability Age at diagnosis Treatment received3 (past & present) # of hospitalizations # of relapses Duration of treatment (past & present) Treatment status (on/off) Time elapsed since last treatment Type of cancer Metastasis occurred (yes/no)    Severity of cancer5           Rather than excluding caregivers from the study with a history of diagnosed mental health struggles, they were retained due to interest in this population‟s adjustment experiences. „Prior diagnosis of mental health struggles‟ was then used as a predictor variable. 2 Household Income was used as a crude measure of socio-economic status (SES) in this study, with scores ranging from 1 to 3 (where 1=no income earners; 2=one income earner; and 3=two income earners). 3 Note that „bone marrow transplant‟ and „other‟ were added to the list of possible treatments following the study proposal, based on recommendations from the dissertation committee. 4 „Additional variables‟ include those added to the Socio-Demographic Form following the study proposal, based on recommendations from the dissertation committee. 5 Cancer severity was not assessed on the Socio-Demographic Form. Rather, it was determined based on the type of cancer identified, according to prognosis estimates from the National Cancer Institute (Kramarova et al., 1996).  120 Table 2 List of Measures across Dependent and Independent Variables Caregiver Self-Report Measures Adolescent Self-Report Measures _____________________________________________________________________________ Dependent Variables  CES-D (20): Emotional Distress SHS (4): Subjective Happiness  PedsQL* (23): Self-reported Quality of Life Psychosocial Health Summary Score (15)  Independent Variables  CAHS-R (13): Threat (5) Challenge (3) PECI (25): Long-term Uncertainty (5) Guilt & Worry (11) Sorrow & Anger (8) PedsQL*a (23): Physical QOL Struggles (8) Emotional QOL Struggles (5) Social QOL Struggles (5) School QOL Struggles (5) MS (7): Perceived Mastery BFS (17): Benefit Finding Acceptance (3) Empathy (4) Appreciation (3) Family (2) Positive Self-View (3) Reprioritization (2) LOT-R (10): Dispositional Optimism HHI (12): Hope FNQ*(42): IMP/MET Health Information (9) Emotional Support (8) Instrumental Support (6) Professional Support (5) Community Support Network (5) Involvement with Care (3) SNRQ*(23): IMP/MET  BFSC (10): Benefit Finding FNQ* (40): IMP/MET Health Information (9) Emotional Support (8) Instrumental Support (6) Professional Support (5) Community Support Network (5) Involvement with Care (3)  Total Items:  (196 Items)  (73 Items)  Note. Measures are indicated in bold, with relevant scales listed beside and relevant subscales listed below. Item totals pertaining to scales and subscales are indicated in brackets. a Measures identified with an asterisk were used in Chung et al.‟s (2005) Family Needs Study.  121 Table 3 Descriptive Statistics on Caregiver Socio-Demographic and Cancer-Specific Data, Baseline Sample Variables  N (%)  Age 20y to 29y 30y to 39y 40y to 49y 50y to 59y 60y to 69y  11 (7.1) 50 (32.1) 80 (51.3) 14 (9.0) 1 (0.6) Mode= 41y / Median = 41y Range = 23y to 61y  Gender Male Female  15 (9.6) 141 (90.4) Split: 1 : 9.4  Relationship to child Mother Father Grandparent  139 (89.1) 14 (9.0) 3 (1.9) Skewness: 3.186, z = 16.42* a Kurtosis: 10.127, z = 26.24*  Type of relationship Natural Adopted Step Foster  151 (96.8) 2 (1.3) 2 (1.3) 1 (0.6) Skewness: 6.643, z = 5.16* Kurtosis: 45.418, z = 117.66*  Education**b / Primary & Secondary School / Extra Occupational/Vocational training / Pre-Univ. Education, College &University  32 (20.5) 34 (21.8) 90 (57.7) Skewness: -0.777, z = -4.005* Kurtosis: -1.015, z = -2.630 After Transformation: Skewness: 0.558, z = 2.876 Kurtosis: -1.439, z = -3.728*  122 Variables Ethnicity/Race White/Caucasian Asian/Pacific Islander Indo Canadian Hispanic/Latino Native American Indian African American  N (%) 119 (76.3) 15 (9.6) 12 (7.7) 6 (3.8) 3 (1.9) 1 (0.6) Skewness: 2.526, z = 13.021* c Kurtosis: 5.157, z = 13.360*  Caucasian status Caucasian Non-Caucasian  119 (76.3) 37 (23.7)  Marital status Married Common law Single Divorced Widow Other  124 (79.5) 13 (8.3) 10 (6.4) 7 (4.5) 1 (0.6) 1 (0.6)  Number of children** 1 2 3 4 5  24 (15.4) 78 (50) 39 (25) 9 (5.8) 6 (3.8) Skewness: 0.870, z = 4.480* Kurtosis: 0.930, z = 2.410 After Transformation: Skewness: 0.273, z = 1.407 Kurtosis: 0.119, z = 0.308  Spirituality Spiritual/Religious Non-spiritual/Non-religious  87 (55.8) 69 (44.2)  Supplemental stressors: Previous experience with cancer Diagnosed mental health struggles Current physical health struggles  64 (41.0) 30 (19.2) 16 (10.3)  Other „current stressors‟ (non-cancer related)  51 (32.7)  Child‟s cancer metastasized  14 (9.0) d  123 Variables Supplemental stressors cont’d: Child‟s cancer relapsed  N (%) 14 (9.0) Split: 1 : 10.1  Child learning disability  5 (3.2) Split: 1 : 30.2  Child (diagnosed) mental health struggles  10 (6.4)  Child physical struggles (non-cancer related)  26 (16.7)  None of the above experienced  47 (30.1) e  Split: 1 : 14.6  Household income** Caregiver(s) not employed One caregiver employed Both caregivers employed  5 (3.2) 56 (35.9) 95 (60.9) Skewness: -0.876, z = -4.515* Kurtosis: -0.259, z = -0.671 After Transformation: Skewness: 0.599, z = 3.088 Kurtosis: -1.352, z = -3.503*  Proximity of residence & Child treatment status Urban living Rural living Child off-treatment Child on-treatment Child on treatment for 1st time Child on active treatment f Subgroup Urban on-treatment Urban off-treatment Rural on-treatment Rural off-treatment Mode of travel Drive Plane Ferry Friend/Family Taxi Other Bus  78 (50) 78 (50) 78 (50) 78 (50) 72 (46.2) 42 (26.9)  39 (25) 39 (25) 39 (25) 39 (25) 122 (78.2) 13 (8.3) 9 (5.8) 7 (4.5) 2 (1.3) 2 (1.3) 1 (0.6) Skewness: 2.142, z = 11.041* Kurtosis: 3.530, z = 9.145*  124 Variables Time to get to hospital (minutes)** Total sample Urban sample Rural sample  N (%) Range = 5 min to 630 min g Median h = 45 min Median i = 240 min Skewness: 1.194, z = 6.155* Kurtosis: 0.761, z = 1.972 After Transformation: Skewness: -0.266, z = -1.371 Kurtosis: -0.693, z = -1.795  Ease of transport Very easy Somewhat easy Sometimes difficult Somewhat difficult Very difficult a  27 (17.3) 43 (27.6) 68 (43.6) 17 (10.9) 1 (0.6)  The asterisk following the z-score signifies statistical significance at the (conservative) ρ<.001 level. Significant skewness and kurtosis values reflect non-normal distributions. b The double asterisk identifies a variable that underwent transformation to reduce non-normality. c Given our interest in ethnicity as a predictor variable, we dichotomized this (non-normal) categorical distribution (i.e., into „Caucasian‟ vs. „non-Caucasian‟) for the inferential analyses. d This value may represent an underestimation given that 43 (27.6%) participants did not respond to this item. Due to the extent of incomplete data, this variable will not be used in further (inferential) statistical analyses. e This value may represent an overestimation given the non-response rates to the metastasis item. f Active treatment is defined as treatment during the acutely ill phase. This is distinguished from „maintenance treatment‟ which is a less aggressive approach for when the disease is in remission. g Recall, this is a corrected range (following revision of extreme outliers in rural sample). h/i Median scores are provided (in place of mean scores), given large standard deviation values.  125 Table 4 Descriptive Statistics on Pediatric Patient/Survivor Socio-Demographic and Cancer-Specific Data, Baseline Sample Age** a 2y to 5y 6y to 9y 10y to 13y 14y to 17y  Variables  N (%) 33 (21.2) 33 (21.2) 44 (28.2) 46 (29.4) Median= 10y Skewness: -0.157, z = -0.809 Kurtosis: -1.220, z = -3.161*b After Transformation: Skewness: 0.607, z = -3.129* Kurtosis: -0.781, z = -2.023  Gender Male Female Ethnicity/Race White/Caucasian Asian/Pacific Islander Indo Canadian Hispanic/Latino Native American Indian Mixed  79 (50.6) 77 (49.4)  111 (71.2) 17 (10.9) 11 (7.1) 8 (5.1) 5 (3.2) 4 (2.6) Skewness: 2.501, z = 12.892* c Kurtosis: 5.554, z = 14.389*  Caucasian status Caucasian Non-Caucasian  111 (71.2) 45 (28.8)  Spirituality Spiritual/Religious Non-spiritual/Non-religious  71 (45.5) 85 (54.5)  Type of cancer Leukemia Lymphomas Soft tissue sarcomas CNS neoplasms Sympathetic nervous system tumors Malignant bone tumors Renal tumors  83 (53.2) 17 (10.9) 11 (7.1) 10 (6.4) 9 (5.8) 8 (5.1) 7 (4.5)  126 Variables Type of cancer cont’d: Hepatic tumors Germ-cell neoplasms Retinoblastoma Carcinomas & other malignant epithelial neoplasms  N (%) 4 4 2 1  (2.6) (2.6) (1.3) (0.6)  Skewness: 1.226, z = 6.320* Kurtosis: .024, z = .062  Severity of cancer d Low risk Moderate risk High risk Treatment received Surgery Bone marrow transplant  14 (9.0) 119 (76.3) 23 (14.7)  74 (47.4) 12 (7.7) Split: 1 : 12  Chemotherapy  146 (93.6) Split: 1 : 14.6  Radiation Other Time since diagnosis** < 12m (1yr) 13m to 24m (2yr) 25m to 36m (3yr) 37m to 48m (4yr) 49m to 60m (5yr) 61m to 72m (6yr) 73m to 84m (7yr) 85m to 96m (8yr) 97m to 108m (9yr) 109m to 120m (10yr)  29 (18.6) 6 (3.8)  54 (34.6) 21 (13.5) 19 (12.2) 10 (6.4) 6 (3.8) 11 (7.1) 10 (6.4) 9 (5.8) 7 (4.5) 9 (5.8) Median = 25m Skewness: 0.833, z = 4.294* Kurtosis: -0.681, z = -1.764 After Transformation: Skewness: -0.402, z = -2.072 Kurtosis: -0.707, z = -1.832  Age at diagnosis <1y to 2y 3y to 5y 6y to 9y 10y to 13y 14y to 17y  33 (21.2) 43 (27.6) 30 (19.2) 28 (14.1) 22 (14.1)  127 Variables Age at diagnosis cont’d: Time since treatment ended <1m e 2m to 6m 7m to 12m 13m to 24m 25m to 36m 37m to 48m 49m to 60m 61m to 72m 73m to 84m 85m to 96m 97m to 108m 109m to 120m  N (%) Median = 6y  78 (50) 4 (2.6) 11 (7.1) 19 (12.2) 8 (5.1) 6 (3.8) 10 (6.4) 9 (5.8) 6 (3.8) 2 (1.3) 1 (0.6) 2 (1.3) Median = 1.5m  Child missed school Average # days absent from school g: Just for those missing school: Grade missed Kindergarten  110 (70.5) f Mean = 132.2dys (SD=130.48) Mean = 156.9dys (SD=126.8)  23 (14.7)  Grade 1  14 (9.0)  Grade 2  10 (6.4)  Grade 3  7 (4.5)  Grade 4  15 (9.6)  Grade 5  16 (10.3)  Grade 6  11 (7.1)  Grade 7  8 (5.1)  Split: 1 : 10.1 Split: 1 : 14.6 Split: 1 : 21.3 Split: 1 : 9.4 Split: 1 : 8.8 Split: 1 : 13.2 Split: 1 : 18.5  Grade 8  9 (5.8) Split: 1 : 16.3  Grade 9  13 (8.3) Split: 1 : 11  Grade 10  10 (6.4) Split: 1 : 14.6  Grade 11  5 (3.2) Split: 1 : 30.2  Grade 12  4 (2.6) Split: 1 : 38  128 Variables Child missed school cont’d: Child repeated a grade  N (%) 4 (2.6) Split: 1 : 38  Late effects / Sequelae Child visual impairment  14 (9.0)  Child hearing impairment  7 (4.5)  Child motor impairment  13 (8.3)  Split: 1 : 10 Split: 1 : 21 Split: 1 : 11 a  The double asterisk identifies a variable that underwent transformation to reduce non-normality. The asterisk following the z-score signifies statistical significance at the (conservative) ρ<.001 level. Significant skewness and kurtosis scores reflect non-normal distributions. c As with the caregiver data in this baseline sample, we dichotomized this (non-normal) categorical distribution (i.e., into „Caucasian‟ vs. „non-Caucasian‟) for the inferential analyses. d This is an approximate measure of disease severity, based on 5-year survival rates as published by the National Cancer Institute (Kramarova et al., 1996). e This group represents patients „on-treatment‟, as on-treatment status relates to those currently receiving treatment (N=73) as well as those treated within the past month (N=5). f This value may represent an underestimation given that 12 (7.7%) participants did not respond to this item as it was not applicable to their „younger child‟. Due to the extent of this (non-random) incomplete data, this variable will not be used in further (inferential) statistical analyses. g This variable is based on the N=127 school-age children in our sample. b  129 Table 5 Descriptive Statistics on Pediatric Patient/Survivor Socio-Demographic and Cancer-Specific Data, Adolescent Sample Variables  N (%)  Age 13y 14y 15y 16y 17y  8 (20) 8 (20) 9 (22.5) 10 (25) 5 (12.5) Median = 15y  Gender Male Female Ethnicity/Race White/Caucasian Asian/Pacific Islander Indo Canadian Hispanic/Latino Mixed  16 (40) 24 (60)  24 (60) 7 (17.5) 4 (10) 4 (10) 1 (2.5) Skewness: 2.195, z = 5.869* a/b Kurtosis: 4.128, z = 5.632*  Caucasian status Caucasian Non-Caucasian Spirituality Spiritual/Religious Non-spiritual/Non-religious  24 (60) 16 (40)  20 (50) 20 (50)  Family Presence of siblings Parents married  34 (85) 36 (90) Split: 9 : 1  Household income Caregiver (s) not employed One caregiver employed Both caregivers employed Caregiver age 20y to 29y 30y to 39y 40y to 49y 50y to 59y 60y to 69y  0 (0) 15 (37.5) 25 (62.5) 1 (2.5) 3 (7.5) 28 (70) 7 (17.5) 1 (2.5)  130 Variables  N (%)  Family cont‟d: Caregiver education / Primary & Secondary School / Extra Occupational/Vocational training / Pre-Univ. Education, College &University  10 (25) 8 (20) 22 (55)  Proximity of residence & Child treatment status Urban living Rural living  24 (60) 16 (40)  Child off-treatment Child on-treatment Child on treatment for 1st time Child on active treatment c Subgroup Urban on-treatment Urban off-treatment Rural on-treatment Rural off-treatment Type of cancer Leukemia Lymphomas Soft tissue sarcomas CNS neoplasms Malignant bone tumors Germ-cell neoplasms Hepatic tumors Severity of cancer d Low risk Moderate risk High risk Treatment received Surgery Bone marrow transplant  25 (62.5) 15 (37.5) 12 (30) 8 (20)  9 (22.5) 15 (37.5) 6 (15) 10 (25)  17 (42.5) 7 (17.5) 7 (17.5) 3 (7.5) 3 (7.5) 2 (5) 1 (2.5)  2 (5) 34 (85) 4 (10) 23 (57.5) 4 (10) Split: 1 : 9  Chemotherapy  36  (90) Split: 9 : 1  Radiation Other  12 (30) 1 (2.5) Split: 1: 39  131 Variables Time since diagnosis < 12m (1yr) 13m to 24m (2yr) 25m to 36m (3yr) 37m to 48m (4yr) 49m to 60m (5yr) 61m to 72m (6yr) 73m to 84m (7yr) 85m to 96m (8yr) 97m to 108m (9yr) 109m to 120m (10yr)  N (%) 9 (22.5) 4 (10) 6 (15) 3 (7.5) 1 (2.5) 4 (10) 4 (10) 2 (5) 1 (2.5) 6 (15) Median = 37.5m  Child age at diagnosis 3y to 5y 6y to 9y 10y to 13y 14y to 17y  4 (10) 8 (20) 16 (40) 12 (30) Median = 12y  Time since treatment <1m 2m to 6m 7m to 12m 13m to 24m 25m to 36m 37m to 48m 49m to 60m 61m to 72m 73m to 84m 97m to 108m 109m to 120m  15 (37.5) 1 (2.5) 1 (2.5) 8 (20) 3 (7.5) 1 (2.5) 3 (7.5) 2 (5) 4 (10) 1 (2.5) 1 (2.5) Median = 20m  Supplemental stressors: Child’s cancer metastasized  3 (7.5) e Split: 1 : 12.3  Child’s cancer relapsed  4 (10) Split: 1 : 9  Child learning disability  1 (2.5) Split: 1 : 39  Child (diagnosed) mental health struggles  2  (5) Split: 1 : 19  Child physical struggles (non-cancer related)  8 (20)  132 Variables Supplemental stressors cont‟d: None of the above experienced  N (%) 27 (67.5) f  Late effects / Sequelae Child visual impairment  5 (12.5)  Child hearing impairment  0 (0)  Child motor impairment a  4 (10)  The asterisk following the z-score signifies statistical significance at the (conservative) ρ<.001 level. Significant skewness and kurtosis scores reflect non-normal distributions. b As with the baseline sample, we dichotomized this non-normal categorical distribution (i.e., into „Caucasian‟ vs. „non-Caucasian‟) for the inferential analyses. c Active treatment is defined as treatment during the acutely ill phase. This is distinguished from „maintenance treatment‟ which is a less aggressive approach for when the disease is in remission. d This is an approximate measure of disease severity, based on 5-year survival rates as published by the National Cancer Institute (Kramarova et al., 1996). e This value may represent an underestimation given that 11 (27.5%) participants did not respond to this item. f This value may represent an overestimation given the elevated non-response rate to the metastasis item.  133 Table 6 Descriptive Statistics on Caregiver Socio-Demographic and Cancer-Specific Data, Longitudinal Sample Variables  N (%)  Age 20y to 29y 30y to 39y 40y to 49y 50y to 59y  7 (17.1) 14 (34.1) 16 (39.0) 4 (9.8) Mode= 41y / Median = 39y Range = 24y to 53y  Gender Male Female  3 (7.3) 38 (92.7) Split: 1 : 12.6  Relationship to child Mother Father Grandparent  37 (90.2) 3 (7.3) 1 (2.4) Skewness: 3.535, z = 9.58* a Kurtosis: 12.959, z = 17.90*  Type of relationship Natural Adopted Foster  39 (95.1) 1 (2.4) 1 (2.4) Skewness: 5.047, z = 13.678* Kurtosis: 26.297, z = 26.32*  Education Primary & Secondary School Extra Occupational/Vocational training Pre-University Education, College &University  Ethnicity/Race White/Caucasian Asian/Pacific Islander Indo Canadian Hispanic/Latino Native American Indian African American  7 (17.1) 10 (24.4) 24 (58.5) 35 (85.4) 2 (4.9) 1 (2.4) 1 (2.4) 1 (2.4) 1 (2.4) Skewness: 3.431, z = 9.298* b Kurtosis: 12.159, z = 16.794 *  134 Variables Ethnicity/Race cont’d: Caucasian status Caucasian Non-Caucasian  N (%)  35 (85.4) 6 (14.6)  Marital status Married Common law Single Divorced  36 (87.8) 1 (2.4) 2 (4.9) 2 (4.9)  Number of children 1 2 3 4 5  7 (17.1) 22 (53.7) 8 19.5) 2 (4.9) 2 (4.9)  Spirituality Spiritual/Religious Non-spiritual/Non-religious  27 (65.9) 14 (34.1)  Supplemental stressors: Previous experience with cancer Diagnosed mental health struggles Current physical health struggles  19 (46.3) 9 (22.0) 6 (14.6)  /  ‘Other’ current stressors (non-cancer related)  20 (48.8)  Child’s cancer metastasized Child’s cancer relapsed (i.e., „progressed‟)  4 (13.8) c 3 (7.3)  /  Child learning disability Child (diagnosed) mental health struggles  Split: 1 : 12.7  1 (2.4) Split: 1 : 40  3 (7.3) Split: 1 : 12.7  Child physical struggles (non-cancer related)  8 (19.5)  None of the above experienced  9 (22.0) d  Household income Caregiver(s) not employed One caregiver employed Both caregivers employed  2 (4.9) 14 (34.1) 25 (61.0)  135 Variables Proximity of residence & Child treatment status Urban living Rural living  N (%) 20 (48.8) 21 (51.2)  Child still on-treatment Child now off-treatment  28 (68.3) 13 (31.7)  Child had been on active treatment e  23 (56.1)  Subgroup Urban on-treatment Urban off-treatment Rural on-treatment Rural off-treatment  14 (34.1) 6 (14.6) 14 (34.1) 7 (17.1)  Mode of travel Drive Plane Ferry Other  31 (75.6) 6 (14.6) 3 (7.3) 1 (2.4) Skewness: 1.984, z = 5.377* Kurtosis: 2.864, z = 3.956 *  Time to get to hospital (minutes) Total sample Urban sample Rural sample  Ease of transport Very easy Somewhat easy Sometimes difficult Somewhat difficult a  Range = 5 min to 420 min Medianf =45 min Mediang =240 min  9 (22.0) 7 (17.1) 18 (43.9) 7 (17.1)  The asterisk following the z-score signifies statistical significance at the (conservative) ρ<.001 level. Significant skewness and kurtosis scores reflect non-normal distributions. b As with the baseline sample, we dichotomized this (non-normal) categorical distribution (i.e., into „Caucasian‟ vs. „non-Caucasian‟). c This value may represent an underestimation given that 12 (29.3%) participants did not respond to this item. Due to the extent of incomplete data, this variable was removed from further (inferential) statistical analyses. d This value may represent an overestimation given the non-response rates to the metastasis item. e Active treatment is defined as treatment during the acutely ill phase. This is distinguished from „maintenance treatment‟ which is a less aggressive approach for when the disease is in remission. f/g Median scores are provided (in place of mean scores), given large standard deviation values.  136 Table 7 Descriptive Statistics on Pediatric Patient/Survivor Socio-Demographic and Cancer-Specific Data, Longitudinal Sample Variables  N (%)  Age 2y to 5y 6y to 9y 10y to 13y 14y to 17y  12 (29.3) 10 (24.4) 9 (22.0) 10 (24.4) Median= 9y  Gender Male Female Ethnicity/Race White/Caucasian Asian/Pacific Islander Indo Canadian Hispanic/Latino Native American Indian Mixed  23 (56.1) 18 (43.9)  33 (80.5) 2 (4.9) 1 (2.4) 2 (4.9) 2 (4.9) 1 (2.4) Skewness: 3.494, z = 9.469 * a/b Kurtosis: 13.211, z = 18.247 *  Caucasian status Caucasian Non-Caucasian  33 (80.5) 8 (19.5)  Spirituality Spiritual/Religious Non-spiritual/Non-religious  17 (41.5) 24 (58.5)  Type of cancer Leukemia Lymphomas Sympathetic nervous system tumors CNS neoplasms Soft tissue sarcomas Malignant bone tumors Renal tumors Germ-cell neoplasms  26 (63.4) 4 (9.8) 3 (7.3) 1 (2.4) 3 (7.3) 2 (4.9) 1 (2.4) 1 (2.4) Skewness: 1.631, z = 4.421* Kurtosis: 1.191, z = 1.645  137 Variables Type of cancer cont’d: Severity of cancer c Low risk Moderate risk High risk  N (%) 2 (4.9) 35 (85.4) 4 (9.8) Skewness: .525, z = 1.423 Kurtosis: 4.358, z = 6.019* d  Treatment received Surgery Bone marrow transplant  20 (48.8) 1 (2.4) Split: 1 : 40  /  Chemotherapy Radiation Other  Time since diagnosis**e < 12m (1yr) 13m to 24m (2yr) 25m to 36m (3yr) 37m to 48m (4yr) 49m to 60m (5yr)  41 (100) 6 (14.6) 1 (2.4)  25 (61.0) 9 (22.0) 5 (12.2) 1 (2.4) 1 (2.4) Median = 10m Skewness: 1.544, z = 4.184* Kurtosis: 2.787, z= 3.849* After Transformation: Skewness: 0.695, z = 1.883 Kurtosis: 0.812, z= 0.251  Age at diagnosis <1y to 2y 3y to 5y 6y to 9y 10y to 13y 14y to 17y  5 (12.2) 12 (29.3) 10 (24.4) 5 (12.2) 9 (22.0) Median = 7y  Child missed school Average # days absent from school (for those missing school)  30 (78.9) f Mean = 171.5dys (SD=109.2)  138 Variables Child missed school cont’d: Grade Missed: Kindergarten  N (%)  5 (12.2)  Grade 1  4 (9.8)  Grade 2  4 (9.8)  Grade 3  3 (7.3)  Grade 4  3 (7.3)  Grade 5  7 (17.1)  Grade 6  4 (9.8)  Grade 7  2 (4.9)  Grade 8  1 (2.4)  Split: 1 : 9.3 Split: 1 : 9.3 Split: 1 : 12.7 Split: 1 : 12.7  Split: 1 : 9.3 Split: 1 : 19.5 Split: 1 : 40  Grade 9  4 (9.8) Split: 1 : 9.3  Grade 10  3 (7.3) Split: 1 : 12.7  Grade 11  1 (2.4) Split: 1 : 40  Grade 12  2 (4.9) Split: 1 : 19.5  Child repeated a grade  0 (0.0)  Late effects / Sequelae Child visual impairment  2 (4.9)  Child hearing impairment  2 (4.9)  Child motor impairment  3 (7.3)  Split: 1 : 19.5 Split: 1 : 19.5 Split: 1 : 12.7 a  The asterisk following the z-score signifies statistical significance at the (conventional but conservative) ρ<.001 level. Significant skewness and kurtosis scores reflect non-normal distributions. b As with the baseline sample, this (non-normal) categorical distribution was dichotomized (i.e., into „Caucasian‟ vs. „non-Caucasian‟). c This is an approximate measure of disease severity, based on 5-year survival rates as published by the National Cancer Institute (Kramarova et al., 1996). d While this variable diverges from normality due to its leptokurtic distribution, it was retained for further analyses (with caution noted). No successful transformations could be performed. e The double asterisk identifies a variable that underwent transformation to reduce non-normality. f This value may represent an underestimation given that 3 (7.3%) participants did not respond to this item as it was not applicable to their „younger child‟. Due to the extent of this (non-random) incomplete data, this variable was not examined in further (inferential) statistical analyses.  139 Table 8 Significant Chi-square Tests of Independence on Socio-Demographic and Cancer-Specific Data, Baseline Sample 8a) Chi-square Tests across Urban vs. Rural Groups Variables Across Subgroups Gender Urban-Gender (male/female) Rural –Gender (male/female)  Ethnicity Urban-Ethnicity (White/Asian/Latino/Native/ Rural  African/Indo) -Ethnicity (White/Asian/Latino/Native/ African/Indo)  Model of Travel Urban-Mode of Travel (Drive/Taxi/Bus/Plane/Friend/ Rural  Ferry/Other) -Mode of Travel (Drive/Taxi/Bus/Plane/Friend/ Ferry/Other)  N (%)  /12 (15.4) / 66 (84.6) /3 (3.8) / 75 (96.2) χ2=5.974, ρ<.05  /48(61.5)/ 13(16.7)/ 5(6.4)/ 1(1.3)/ 1(1.3)/ 10(12.8)  /71(91.0)/ 2(2.6)/ 1(1.3)/ 2(2.6)/ 0(0)/ 2(2.6)  χ2=21.845, ρ<.01  /71(91)/ 2(2.6)/ 0(0)/ 0(0)/ 5(6.4)/ 0(0)/ 0(0) /51(65.4)/ 0(0)/ 1(1.3)/ 13(16.7)/ 2(2.6)/ 9(11.5)/ 2(2.6)  χ2=31.564, ρ<0.001 Ease of Transport Urban-Ease of Transport / (Very Difficult/Somewhat Easy/Sometimes Easy/  /0(0)/ 3(3.8)/ 24(30.8)/  Somewhat Easy/Very Easy)  28(35.9)/ 23(29.5)  Rural – Ease of Transport /  (Very Difficult/Somewhat Easy/Sometimes Easy/ Somewhat Easy/Very Easy)  /1(1.3)/ 14(10.9)/ 44(56.4)/ 15(19.2)/ 4(5.1)  χ2=31.301, ρ<0.001  Child Age Urban – Child Age (2 to 5/ 6 to 9/ 10 to 13/ 14 to 17) Rural – Child Age (2 to 5/ 6 to 9/ 10 to 13/ 14 to 17)  Child Ethnicity Urban-Child Ethnicity (White/Asian/Latino/Native/ Rural  Indo/Mixed) - Child Ethnicity (White/Asian/Latino/Native/ Indo/Mixed)  /15(19.2)/ 18(23.1)/ 15(19.2)/ 30(38.5) /18(23.1)/ 15(19.2)/ 29(37.2)/ 16(20.5) χ2=9.261, ρ<.05  /42(53.8)/ 14(17.9)/ 7(9.0)/ 2(2.6)/ 9(11.5)/ 4(5.1) /69(88.5)/ 3(3.8)/ 1(1.3)/ 3(3.8)/ 2(2.6)/ 0(0)  χ2=26.840, ρ<0.001  140 Variables Across Subgroups  N (%)  Age at Diagnosis Urban – Age at Diagnosis Rural  (<1yr to 2y/ 3 to 5y/ 6 to 9y/ 10 to 13y/ 14 to 17y) – Age at Diagnosis (<1yr to 2y/ 3 to 5y/ 6 to 9y/ 10 to 13y/ 14 to 17y)  13(16.7)/ 20(25.6)/ 15(19.2)/ 12(15.4)/ 18(23.1)  / 20(25.6)/ 23(29.5)/ 15(19.2)/ 16(20.5)/ 4(5.1)  χ2=11.175, ρ<.05  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. Note. These chi-square tests should be interpreted with caution as many involved small expected count values.  8b) Chi-square Tests across On-Treatment vs. Off-Treatment Groups Variables Across Subgroups ‘Other’ Current Stressors On-Tmt- „Other‟ current stressors (yes/no) Off-Tmt- „Other‟ current stressors (yes/no) Chemotherapy Received On-Tmt- Chemotherapy Received (yes/no) Off-Tmt- Chemotherapy Received (yes/no)  Time Since Diagnosis On-Tmt- Time Since Diagnosis (<12m/ 13m-24m/ 25m-36m/37m-48m/49m-60m/ 61m-72m/ 73m-84m/ 85m-96m/97m-108m/ 109m-120m) Off-Tmt- Time Since Diagnosis (<12m/ 13m-24m/ 25m-36m/37m-48m/49m-60m/ 61m-72m/ 73m-84m/ 85m-96m/97m-108m/ 109m-120m)  N (%)  /33(42.3)/ 45(57.7) /18(23.1)/ 60(76.9) χ2=6.555, ρ<.05 /77(98.7)/ 1(1.3) /69(88.5)/ 9(11.5) χ2=6.838, ρ<.01 Fisher’s exact test a: ρ<.01 /48(61.5)/ 14(17.9)/ 8(10.3)/ 2(2.6)// 3(3.8)/ 0(0)/ 1(1.3)/ 2(2.6)/ 0(0)/ 0(0) /6(7.7)/ 7(9.0)/ 11(14.1)/ 8(10.3)/3(3.8)/ 11(14.1)/ 9(11.5)/ 7(9.0)/ 7(9.0)/ 9(11.5)  χ2=75.251, ρ<0.001  Age at Diagnosis On-Tmt- Age at Diagnosis (<1y-2y/ 3y-5y/ 6y-9y/ 10y-13y/ 14y-17y) Off-Tmt- Age at Diagnosis (<1y-2y/ 3y-5y/ 6y-9y/ 10y-13y/ 14y-17y)  10(12.8)/ 22(28.2)/ 16(20.5)/ 13(16.7)/ 17(21.8) 23(29.5)/ 21(26.9)/ 14(17.9)/ 15(19.2)/ 5(6.4)  χ2=11.966, ρ<.05 Note. Significant findings (at the .05 level or smaller) are indicated in bold font. Note. These chi-square tests should be interpreted with caution as many involved small expected count values. a Fisher‟s Exact Test provides a second opinion about the data in a 2 X 2 table of counts where expected frequencies are very low. Despite observing small expected counts across other analyses, Fisher‟s test could not be used elsewhere as these other analyses involved too many categories (i.e., they were not 2X2 tables).  141  Table 9 Significant Chi-square Tests of Independence on Socio-Demographic and Cancer-Specific Data, Adolescent Sample Variables Across Subgroups  N (%)  On-Treatment vs. Off-Treatment Groups Time Since Diagnosis On-Tmt- Time Since Diagnosis (<12m/ 13m-24m/ 25m-36m/ 37m-48m/ 49m-60m/ 61m-72m/ 73m-84m/ 85m-96m/ 97m-108m/ 109m-120m) Off-Tmt- Time Since Diagnosis (<12m/ 13m-24m/ 25m-36m/ 37m-48m/ 49m-60m/ 61m-72m/ 73m-84m/ 85m-96m/ 97m-108m/ 109m-120m)  /9(60.0)/ 2(13.3)/ 1(6.7)/ 0(0)/ 1(6.7)/ 0(0)/ 1(6.7)/ 1(6.7)/ 0(0)/ 0(0) /0(0)/ 2(8.0)/ 5(20.0)/ 3(12.0)/ 0(0)/ 4(16.0)/ 3(12.0)/ 1(4.0)/ 1(4.0)/ 6(24.0)  χ2=26.844, ρ<.01  Age at Diagnosis On-Tmt-Age at Diagnosis (3y-5y/ 6y-9y/ 10y-13y/ 14y-17y) Off-Tmt-Age at Diagnosis (3y-5y/ 6y-9y/ 10y-13y/ 14y-17y)  1(6.7)/ 0(0)/ 5(33.3)/ 9(60.0) 3(12.0)/8(32.0)/ 11(44.0)/ 3(12.0)  χ2=12.533, ρ<.01 Note. Significant findings (at the .05 level or smaller) are indicated in bold font. Note. These chi-square tests should be interpreted with caution as many involved small expected count values.  142 Table 10 Descriptive Statistics and Group Comparisons for the BFS (‘Benefit Finding’ scale)  Overall Sample Total Score: / Avg item score: Cronbach‟s Alpha:  Acceptance Scale (ACC): Avg item score: Empathy Scale (EMP): Avg item score: Appreciation Scale (APP): Avg item score: /Family Scale (FAM): Avg item score: Positive Self-View Scale (PSV): Avg item score: Reprioritization Scale (REP): Avg item score:  Mean & N (%)  SD & Variance Statistics  Range & Significance Test Scores  59.99  13.65 /  Range = 22 - 85 /Range = 1.29 – 5.00  2.75  Range = 3 – 15  4.10  Range = 4 - 20  2.94  Range = 3 - 15  2.05  Range = 2 - 10  2.96  Range = 3 - 15  2.00  Range = 2 - 10  /3.53 (“Quite a bit”) α = .93  11.27 3.76 (“Quite a bit”) α = .89 13.96 3.49 (“Quite a bit”) α =.82 9.49 3.16 (“Moderately”) α =.73 7.27 3.64 (“Quite a bit”) α =.75 10.37 3.46 (“Quite a bit”) α =.80 7.63 3.82 (“Quite a bit”) α =.78  Caregiver realized additional benefits (besides those on questionnaire):  / 28 (17.9%)  Total Score - Across four subgroups: Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  60.33 61.41 58.13 60.08  Levene’s HOV test:  / /  /  /F (3, 152) =.444,  F (3, 152) =.387, p > .05  p>.05  ACC- Across four subgroups: Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test:  11.28 11.21 11.13 11.46 /F (3, 152) =.125, p>.05  F (3, 152) =.103, p > .05  143 Mean & N (%) EMP- Across four subgroups: Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  SD & Variance Statistics  Range & Significance Test Scores  /F (3, 152) =.306,  F (3, 152) =1.056, p > .05  14.85 14.13 13.46 13.41  Levene’s HOV test:  p>.05  APP- Across four subgroups: Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  9.44 9.87 9.03 9.62  Levene’s HOV test:  /F (3, 152) =.402,  F (3, 152) =.566, p > .05  p>.05  FAM- Across four subgroups: Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  7.21 7.67 7.03 7.18 /F (3, 152) =.126,  Levene’s HOV test:  PSV- Across four subgroups: Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  p>.05  F (3, 152) =.709, p > .05  10.03 10.79 9.79 10.85  Levene’s HOV test:  /F (3, 152) =.523,  F (3, 152) =1.274, p > .05  p>.05  REP- Across four subgroups: Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test:  7.54 7.74 7.69 7.56 /F (3, 152) =.756, p>.05  F (3, 152) =.094, p > .05  144 Table 11 Descriptive Statistics and Group Comparisons for the CAHS-R (‘Challenge’ and ‘Threat’ scale)  Overall Sample Challenge  Mean  SD & Variance Statistics  10.31  2.15  Avg item score: Cronbach‟s Alpha:  /3.4 (“Neutral”) α = .46  Avg item score: Cronbach‟s Alpha:  3.6 (“Agree”) α = .71  Threat  18.1  Challenge - Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  Threat - Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  Range = 4 - 15 /Range = 1.33 – 5.00  3.86  Range = 7 - 25 /Range = 1.40 – 5.00  10.62 9.79 10.46 10.38 /F (3, 152) =.233,  Levene’s HOV test:  Range & Significance Test Scores  F (3, 152) =1.086, p >.05  p>.05  18.59 17.56 19.15 17.13 /F (3, 152) = .532,  Levene’s HOV test:  p>.05  F (3, 152) =2.306, p >.05 Interaction Effect = NS Urb/Rur Main Effect = NS Tmt Main Effect = (p<.05)  On-Tmt vs. Off-Tmt Subgroups On-Tmt Sample: Off-Tmt Sample:  18.87 17.35  3.69 3.90  t (154)=-.508, p<.05  145 Table 12 Descriptive Statistics and Group Comparisons for the CES-D (‘Distress’ scale) Mean & SD & Range & N (%) Variance Significance Test Statistics Scores Overall Sample Total Score: 15.29 12.15 Range = 0 - 54 / /(“Not Depressed”) a/ // Cronbach‟s Alpha: // /α = .93 /  Comparison to normative sample b  Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  /9.25  18.82 13.85 17.13 11.38  Levene’s HOV test:  On-Tmt vs. Off-Tmt Subgroups On-Tmt Sample:  /F (3, 152) =.768, /p>.05  17.97 (“Depressed”)  /  Comparison to normative sample b  8.58  12.62  10.90  (“Not Depressed”) Comparison to normative sample b  /9.25  d = 0.58  F (3, 152) =3.037, p<.05 /Interaction Effect = NS /Urb/Rur Main Effect = NS /Tmt Main Effect = (p<.01)  12.80  9.25  / Off-Tmt Sample:  /8.58  /8.58  d = 0.82  d = 0.35  t (154)= -2.815, p<.01 /# Endorsing ‘+/- Stressor’ in Past Week:  43 (27.6)  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. a Note that a total score of 16 or greater receives a classification of „Depressed‟ on this measure, indicating „depression symptomatology‟ or „clinically significant level of psychological distress‟ (Radloff, 1977). b This normative sample is represented by a group of US community adults (n=2514) (Radloff, 1977).  146 Table 13 Descriptive Statistics and Group Comparisons for the Caregiver FNQ (‘Family Needs’ scale) Mean & SD & Range & N (%) Variance Significance Test Statistics Scores Overall Sample Total Score - Proportion of needs/„important‟ 0.81 0.14 Range = 0.17 – 1.00 (IMP):  Cronbach‟s Alpha:  α = .89  /Health Information (HI):  0.98  /Emotional Support (ES): /Community Support (COM): /Instrumental Support (INS): /Involvement with Care (INV): /Professional Support (PRO):  Total Score – Proportion of IMP needs „met‟ (MET):  Cronbach‟s Alpha:  /Health Information (HI): /Emotional Support (ES): /Community Support (COM): /Instrumental Support (INS): /Involvement with Care (INV): /Professional Support (PRO):  Hope item /Considered an IMP need: /Considered this need MET:  /α = .52 0.64 /α = .78 0.86 /α = .52 0.73 /α = .67 0.86 /α = .46 0.82 /α = .74 0.62  0.69  Range = 0.44 – 1.00  0.30  Range = 0.00 – 1.00  0.20  Range = 0.20 – 1.00  0.27  Range = 0.00 – 1.00  0.22  Range = 0.00 – 1.00  0.26  Range = 0.00 – 1.00  0.25  Range = 0.00 – 1.00  0.24  Range = 0.00 – 1.00  0.35  Range = 0.00 – 1.00  0.36  Range = 0.00 – 1.00  0.38  Range = 0.00 – 1.00  0.31  Range = 0.00 – 1.00  0.38  Range = 0.00 – 1.00  α = .92  0.79 /α = .76 0.47a /α = .73 0.61 /α = .71 0.37b /α = .74 0.79c /α = .47 0.57d /α = .76  115 (73.7) 91 (58.3)  147 Mean & N (%) IMP - Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  SD & Variance Statistics  Range & Significance Test Scores  /F (3, 152) =.463,  F (3, 152) =.192, p>.05  .83 .80 .81 .82  Levene’s HOV test:  p>.05  IMP-HI- Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  .98 .96 .98 .99  Levene’s HOV test:  .06 .10 .06 .04 F (3, 152)=4.557, p<.01e  F (3, 152) =1.392, p>.05  IMP-ES- Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  .67 .64 .65 .61  Levene’s HOV test:  /F (3, 152) =.242, p>.05  F (3, 152) =.255, p > .05  IMP-COM- Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  .90 .84 .84 .85  Levene’s HOV test:  /F (3, 152) =2.282,  F (3, 152) =.990, p > .05  p>.05  IMP-INS - Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test:  .71 .74 .75 .71 /F (3, 152) =.603, p>.05  F (3, 152) =.227, p > .05  148 Mean & N (%)  SD & Variance Statistics  Range & Significance Test Scores  /F (3, 152) =.503,  F (3, 152) =.037, p > .05  IMP-INV- Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  .85 .86 .87 .86  Levene’s HOV test:  p>.05  IMP-PRO- Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  .84 .79 .79 .87  Levene’s HOV test: /F (3, 152) =2.488,  F (3, 152) =.995, p > .05  p>.05  MET- Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  .65 .61 .62 .60  / F (3, 152) =.855, p>.05  Levene’s HOV test:  F (3, 152) =.360, p > .05  MET-HI- Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  .83 .76 .81 .77  Levene’s HOV test:  /F (3, 152) =2.132,  F (3, 152) =.850, p > .05  p>.05  MET-ES- Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test:  .44 .54 .43 .47  .32 .36 .30 .43 F (3,152)=5.149, p<.01f  F (3, 152) =.785, p > .05  149 Mean & N (%)  SD & Variance Statistics  Range & Significance Test Scores  /F (3, 152) =2.094,  F (3, 152) =1.409, p > .05  MET-COM- Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  .71 .61 .58 .55  Levene’s HOV test:  p>.05  MET-INS - Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  .40 .34 .38 .35  Levene’s HOV test:  /F (3, 152) =.427, p>.05  F (3, 152) =.201, p > .05  MET-INV- Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  .84 .79 .80 .74  Levene’s HOV test:  /F (3, 152) =2.453, p>.05  F (3, 152) =.808, p > .05  MET-PRO- Across four subgroups  Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test:  .62 .57 .57 .52 /F (3, 152) =.743, p>.05  F (3, 152) =.457, p > .05  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. a/b/c/d Note that several participants did not identify any items in these areas as „important‟ or „very important‟. As such, a proportion for these „MET‟ values could not be calculated due to a denominator of 0. To avoid missing values, the mean score for each respective variable was substituted in for those participants where a value could not be calculated (Substitutions: 7 ES, 4 INS, 1 INV, 1 PRO). e/f While Levene‟s test in these two cases suggests that the assumption of homogeneity of variance (HOV) across distributions has been violated, it is noted that such violations have negligible consequences (on type-1 error statements or power) when n‟s are equal (Moore, 2000).  150 Table 14 Descriptive Statistics and Group Comparisons for the HHI (‘Hope’ scale) Mean SD & Variance Statistics Overall Sample Total Score: 37.79 4.94 Average item score: Cronbach‟s Alpha:  Across four subgroups: Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test:  Range & Significance Test Scores Range = 15-48  3.15 (“Agree”) α = .88  37.90 38.13 37.41 37.74 /F (3, 152) =.818, p>.05  F (3, 152) =.143, p > .05  151 Table 15 Descriptive Statistics and Group Comparisons for the LOT-R (‘Optimism’ scale) Mean SD & Range & Variance Significance Test Statistics Scores Overall Sample Total Score: 14.44 4.24 Range = 0-24 Average item scorea: Cronbach‟s Alpha:  Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test:  a  1.44 (“Disagree”) α = .82  14.64 14.56 13.79 14.74 F (3, 152) =.458, p>.05  F (3, 152) =.404, p > .05  Note that this average item score has used reverse scoring to account for negatively worded items.  152 Table 16 Descriptive Statistics and Group Comparisons for the MS (‘Mastery’ scale) Mean SD & Variance Statistics Overall Sample Total Score: 20.36 3.42 Average item score: Cronbach‟s Alpha:  Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test:  Range & Significance Test Scores Range = 9-28  2.91 (“Agree”) α = .82  20.13 20.23 20.41 20.67 F (3, 152) =.664, p>.05  F (3, 152) =.183, p > .05  153 Table 17 Descriptive Statistics and Group Comparisons for the PECI (‘Long-term Uncertainty’, ‘Unresolved Sorrow & Anger’, and ‘Grief & Worry’ scale) Mean  SD & Variance Statistics  Range & Significance Test Scores  Overall Sample – Average item scores: /Long-Term Uncertainty (LTU):  /1.50 (“Sometimes”)  /0.86  /Range = 0.00-3.80  Cronbach‟s Alpha: /Unresolved Sorrow/Anger (USA): Cronbach‟s Alpha: /Guilt & Worry (GW): Cronbach‟s Alpha:  α = .82 /1.42 (“Rarely”) α = .88 /1.69 (“Sometimes”) α = .88  /0.89  / Range = 0.00-3.88  /0.80  Range = 0.27-3.45  /F (3, 152) =.872,  F (3, 152) =2.866, p<.05  LTU - Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  1.55 1.57 1.71 1.17  Levene’s HOV test:  p>.05  USA - Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  /F (3, 152) =.256, p>.05  GW - Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test:  Interaction Effect = (p<.05) Urb/Rur Main Effect = NS Tmt Main Effect = NS  1.56 1.26 1.83 1.04  Levene’s HOV test:  On-Tmt vs. Off-Tmt Subgroups On-Tmt Sample: Off-Tmt Sample:  /  1.70 (“Sometimes”) 1.15 (“Rarely”)  0.84 0.87  1.80 1.59 2.02 1.35 /F (3, 152) =.594, p>.05  F (3, 152) =6.452, p<.001 Interaction Effect = NS Urb/Rur Main Effect = NS Tmt Main Effect = (p<.001)  t (154) =-3.996, p<.001  F (3, 152) =9.777, p<.001 Interaction Effect = NS Urb/Rur Main Effect = NS Tmt Main Effect = (p<.001)  154  On-Tmt vs. Off-Tmt Subgroups On-Tmt Sample: Off-Tmt Sample:  Mean  SD & Variance Statistics  1.91(“Sometimes”) 1.47 (“Rarely”)  0.77 0.77  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  Range & Significance Test Scores  t (154)= -3.607, p<.001  155 Table 18 Descriptive Statistics and Group Comparisons for the Caregiver PedsQL (‘Perceived Child Physical/Emotional/Social Struggles’ scale a) Mean  SD & Variance Statistics  Range & Significance Test Scores  34.42 b  27.32  Range = 0 - 100  20.56  Range = 0 - 95  20.10  Range = 0 - 90  /F (3, 152) =.983,  F (3, 152) =7.517, p<.001  Overall Sample Physical Functioning Struggles (PF): Average item score: Cronbach‟s Alpha:  Emotional Functioning Struggles (EF): Average item score: Cronbach‟s Alpha:  Social Functioning Struggles (SoF):  (“Almost Never”) α = .93  35.52 (“Almost Never”) α = .84  25.87  Average item score: Cronbach‟s Alpha:  (“Almost Never”) α = .78  PF- Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  45.67 25.80 42.55 23.64  Levene’s HOV test:  On-Tmt vs. Off-Tmt Subgroups On-Tmt Sample:  p>.05  44.11 (“Sometimes”)  Off-Tmt Sample:  24.72 (“Almost Never”)  EF- Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test:  Interaction Effect = NS Urb/Rur Main Effect = NS Tmt Main Effect = (p<.001)  27.35 23.74  t (154) = -4.728, p<.001  37.82 32.56 41.54 30.13 /F (3, 152) =1.025, p>.05  F (3, 152) =2.511, p>.05  156 Mean  SoF- Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test:  SD & Variance Statistics  Range & Significance Test Scores  /F (3, 152) =1.034,  F (3, 152) =0.905, p>.05  26.28 21.41 27.95 27.82  p>.05 Note. Significant findings (at the .05 level or smaller) are indicated in bold font. a Due to many patients being too young or sick to attend school, there was a large set of non-random missing data on the School Functioning domain of this measure. As such, this domain was removed from statistical analyses. b Note that scoring for the PedsQL linearly transforms item scores onto a 0-100 scale (0=0, 1=25, 2=50, 3=75, 4=100).  157 Table 19 Descriptive Statistics and Group Comparisons for the SHS (‘Subjective Happiness’ scale) Mean SD & Range & Variance Significance Test Statistics Scores Overall Sample Avg item scorea: 4.81 1.23 Range = 1-7 Cronbach‟s Alpha: / /  Comparison to normative sample 1b Comparison to normative sample 2c  Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt: Levene’s HOV test: a  α = .88 5.62 4.80  0.93 1.12  4.87 4.79 4.64 4.95  d = 0.75 d = -0.01  F (3, 152) =.444, p > .05 F (3, 152) =.056, p>.05  Note that this average item score has used reverse scoring to account for negatively worded items. This normative sample is represented by a group of US community adults (n=198) (Lyubomirsky & Lepper, 1999). c This normative sample is represented by a group of US female community adults (n=92) (Lyubomirsky & Lepper, 1999). b  158 Table 20 Descriptive Statistics and Group Comparisons for the SNRQ (‘School Needs & Resources’ scale) 20a) Descriptive Statistics, T-tests, and ANOVA analyses for the SNRQ Mean SD & Variance Statistics Overall Sample Total Score – proportion of needs .25 0.30 considered „important/very important‟ (IMP): Cronbach‟s Alpha:  Total Score – proportion of imp. needs met  /  .41a α = .79  IMP - Across four subgroups Urban On-Tmt: Urban Off-Tmt: Rural On-Tmt: Rural Off-Tmt:  .23 .31 .17 .29  Levene’s HOV test:  / /  0.40  Range = 0 – 1.00  /  /F (3, 152) =2.303,  Levene’s HOV test:  MET - Across four subgroups b Urban On-Tmt (n=30): Urban Off-Tmt (n=30): Rural On-Tmt (n=21): Rural Off-Tmt (n=28):  Range = 0 – 1.00  / /  α = .95  (MET): Cronbach‟s Alpha:  On-Tmt vs. Off-Tmt Subgroups On-Tmt Sample: Off-Tmt Sample:  Range & Significance Test Scores  p>.05  .20 .30  .33 .26  .42 .37 .42 .45  .36 .41 .45 .40 /F (3, 152) =1.437,  F (3, 152) =1.680, p > .05 Interaction Effect = NS Urb/Rur Main Effect = NS Tmt Main Effect = (p<.05)  t (154)= 2.063, p<.05  F (3, 152) =.200, p > .05  p>.05 Note. Significant findings (at the .05 level or smaller) are indicated in bold font. a This percentage is based on data from only 109 participants (i.e., 70% of the total data set), as values for this variable could not be calculated for those 47 participants who did not endorse any school needs or resources as „important‟ (and thus a denominator of „0‟ resulted for this variable). Given the extent of missing data for this variable, it will not be used in further inferential analyses. b Note that sample sizes are unequal in this ANOVA because of the problem mentioned in the above footnote. This is not overly concerning. When larger sample sizes are paired with smaller standard deviations (as is the case here), results of ANOVA are considered liberal [i.e., the likelihood of making a type 1 error (alpha) is higher; Glass & Hopkins, 1996]. However, given our insignificant F test, this is not a concern here.  159 20b) Frequency Percentages for ‘Importance’ scores (i.e., described as ‘important’ or ‘very important’) across individual SNRQ items „Important‟ Child School Need or Resource Frequency (%) Individual support from the classroom teacher  46  Individual support from an educational assistant  39  An Individualised Educational Plan (IEP) Supervision to ensure safety  31  Provision of accommodations (e.g., extra time for tests, oral tests, etc.)  30  A psycho-educational assessment Withdrawal support (e.g., small class, group, or one-to-one instruction)  29  Individual physiotherapy sessions Consultation with psychologist/psychological associate regarding learning difficulties  28  Provision of modifications to academic curriculum (e.g., working at lower grade lvl)  27  A neuropsychological assessment Consultation between occupational therapist (OT) and school staff or family  26  A formal identification as having exceptional learning needs (e.g., IPRC)  25  Individual OT sessions  21  Consultation between physiotherapist and school staff or family Transition support from elementary school to high school  20  Transition support from high school to college/university/work force Vocational counselling  19  Individual speech/language therapy sessions Consultation between speech/language therapist and school staff or family  18  Cooperative/vocational training programmes  17  Access to adaptive technology in the classroom (e.g., for visual, hearing or motor impairment)  16  Placement in a specialized classroom  14  160 Table 21 Descriptive Statistics and Group Comparisons for the BFSC (‘Adolescent Benefit Finding’ scale) Mean & SD & Variance Range & N (%) Statistics Significance Test Scores Overall Sample Total Score: 38.03 7.08 Range = 19-48 Average item score: Cronbach‟s Alpha:  # of teenagers realizing „additional‟ benefits (i.e., beyond those assessed in questionnaire):  Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10): Levene’s HOV test:  3.80 (“Quite a bit”) α = .83  8 (20 %)  34.56 39.67 36.33 39.70 /F (3, 36) =1.170, p>.05  F (3, 36) =1.320, p>.05  161 Table 22 Descriptive Statistics and Group Comparisons for the Teen FNQ (‘Adolescent Family Needs’ scale) Mean & SD & Variance Range & N (%) Statistics Significance Test Scores Overall Sample Total Score - Proportion of needs 0.75 0.17 Range = 0.35 – 1.00 „important‟ (IMP): Cronbach‟s Alpha:  /  Health Information (HI):  //  Emotional Support (ES):  //  Community Support (COM):  //  Instrumental Support (INS):  //  Involvement with Care (INV):  /  /  Professional Support (PRO):  Total Score – Proportion of IMP needs /  „met‟ (MET): Cronbach‟s Alpha:  //  Health Information (HI):  /  Emotional Support (ES):  /  Community Support (COM):  /  Instrumental Support (INS):  /  Involvement with Care (INV):  /  Professional Support (PRO):  Hope item /Considered an IMP need: /Considered this need MET:  /α = .88  0.84 /α = .82 0.63 /α = .82 0.71 /α = .50 0.84 /α = .37 0.78 /α = .12 0.69 /α = .54 0.71  0.23  Range = 0.11 – 1.00  0.32  Range = 0.00 – 1.00  0.29  Range = 0.00 – 1.00  0.16  Range = 0.50 – 1.00  0.24  Range = 0.33 – 1.00  0.29  Range = 0.00 – 1.00  0.25  Range = 0.16 – 1.00  0.26  Range = 0.22 – 1.00  0.30  Range = 0.00 – 1.00  0.35  Range = 0.00 – 1.00  0.33  Range = 0.00 – 1.00  0.35  Range = 0.00 – 1.00  0.43  Range = 0.00 – 1.00  α = .90  0.79 /α =.80 0.66a /α =.72 0.71b /α = .59 0.61 /α =.57 0.78 /α =.57 0.60c /α =.70  27 (65.9) 28 (68.3)  162 Mean & N (%) IMP- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  F (3, 36) =.380, p>.05 /F (3, 36) =.024, p>.05  .93 .85 .78 .80  Levene’s HOV test:  IMP-ES- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  .72 .61 .63 .60  Levene’s HOV test:  F (3, 36) =.286, p>.05 /F (3, 36) =.181, p>.05  .72 .77 .75 .60  Levene’s HOV test:  IMP-INS- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  F (3, 36) =.671, p>.05 /F (3, 36) =1.337, p>.05  Levene’s HOV test:  IMP-COM- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  Range & Significance Test Scores  .80 .76 .75 .71  Levene’s HOV test: IMP-HI- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  SD & Variance Statistics  /F (3, 36) =.765, p>.05  F (3, 36) =.716, p>.05  .89 .81 .89 .82 F (3, 36) =.683, p>.05 /F (3, 36) =.457, p>.05  163 Mean & N (%) IMP-INV- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  F (3, 36) =.748, p>.05 /F (3, 36) =.128, p>.05  .69 .70 .71 .65  Levene’s HOV test:  MET- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  .75 .68 .79 .66 /F (3, 36) =.365, p>.05  Levene’s HOV test:  F (3, 36) =.500, p>.05  .89 .80 .82 .68  Levene’s HOV test:  MET-ES- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  F (3, 36) =.073, p>.05 /F (3, 36) =.891, p>.05  Levene’s HOV test:  MET-HI- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  Range & Significance Test Scores  .85 .76 .83 .70  Levene’s HOV test:  IMP-PRO- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  SD & Variance Statistics  /F (3, 36) =1.458, p>.05  F (3, 36) =1.078, p>.05  .68 .62 .71 .66 F (3, 36) =.155, p>.05 /F (3, 36) =.597, p>.05  164 Mean & N (%) MET-COM- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  F (3, 36) =.296, p>.05 /F (3, 36) =.942, p>.05  .66 .55 .72 .58  Levene’s HOV test:  MET-INV- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  Range & Significance Test Scores  .72 .64 .72 .78  Levene’s HOV test:  MET-INS- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  SD & Variance Statistics  /F (3, 36) =.254, p>.05  .96 .68 .94 .67  .11 .41 .14 .41  F (3, 36) =.499, p>.05  F (3, 36) =2.249, p>.05  F (3, 36)=7.656, p<.01 d  Levene’s HOV test:  Across URB-RUR subgroups Urban (n=24): Rural (n=16):  .78 .77  .35 .35  Across ON-OFF subgroups On-Tmt (n=15): Off-Tmt (n=25):  .96 .67  .12 .40  t (38)= -.122 , p>.05  t (38)= -2.664, p<.05  Welch’s t(30) = 3.380, p<.05 e  165 Mean & N (%)  MET-PRO- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10): Levene’s HOV test:  SD & Variance Statistics  Range & Significance Test Scores  .58 .51 .67 .73 F (3, 36) =.555, p>.05 /F (3, 36) =.549, p>.05  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. a/b/c Note that several participants did not identify any items in these areas as „important‟ or „very important‟. As such, a proportion for these values could not be calculated due to a denominator of 0. To avoid missing values, the mean score for each respective variable was substituted in for those participants where a value could not be calculated (Substitutions: 3 ES, 2 COM, 1 PRO). d This violation of the homogeneity of variance assumption, in combination with unequal sample sizes among subgroups, is concerning as it significantly impacts our type-1 error. T-tests (displayed below the ANOVA results) were therefore performed for exploratory purposes to examine group differences. e Given the unequal variances in the t-test examining the treatment effect , Welch‟s t-test was also performed. This is a (less powerful) adaptation of the student‟s t-test intended for use in two samples displaying unequal variance (Glass & Hopkins, 1996). Welch‟s t-value of 3.380 surpasses the (two-sided) critical value of 95t30= 2.042.  166 Table 23 Descriptive Statistics and Group Comparisons for the Teen PedsQL (‘Adolescent Quality of Life’ scale) Mean SD & Variance Range & Statistics Significance Test Scores Quality of life - Teens (PedsQL) Overall Sample Total Scale (TOT): Avg item score: Cronbach‟s Alpha:  Psychosocial Health Summary (PHS):  73.86a  73.08  /  Avg item score: Cronbach‟s Alpha:  (“Almost Never”) α = .92  /  Emotional Functioning Struggles (EF): Avg item score: Cronbach‟s Alpha:  (“Almost Never”) α = .84  Social Functioning Struggles (SoF): Avg item score: Cronbach‟s Alpha:  (“Almost Never”) α = .89  School Functioning Struggles (ScF): Avg item score: Cronbach‟s Alpha:  (“Almost Never”) α = .85  Physical Functioning Struggles (PF): Avg item score: Cronbach‟s Alpha:  (“Almost Never”) α =.92  /  / /  / /  / /  TOT- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  72.38  83.13  63.75  75.31  Levene’s HOV test:  Range = 21.74 - 100  18.16  Range = 20 - 100  19.81  Range = 25 - 100  19.89  Range = 5 - 100  24.09  Range = 0 - 100  24.84  Range = 25 - 100  69.69 75.29 65.40 80.54 F (3, 36) =.955, p>.05  Levene’s HOV test: PHS- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  19.24  (“Almost Never”) α = .95  /F (3, 36) =.918, p>.05 69.81 73.44 67.22 79.00 F (3, 36) =.643, p>.05 /F (3, 36) =1.137, p>.05  167 Mean  EF- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  F (3, 36) =.347, p>.05 /F (3, 36) =1.109, p>.05  84.44 81.00 77.50 88.50 F (3, 36) =.454, p>.05  Levene’s HOV test:  ScF- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  /F (3, 36) =2.755, p>.05  55.00 66.00 57.50 72.00 F (3, 36) =.962, p>.05  Levene’s HOV test:  PF- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10): Levene’s HOV test: a  Range & Significance Test Scores  70.00 73.33 66.67 76.50  Levene’s HOV test:  SoF- Across four subgroups Urban On-Tmt (n=9): Urban Off-Tmt (n=15): Rural On-Tmt (n=6): Rural Off-Tmt (n=10):  SD & Variance Statistics  /F (3, 36) =.797, p>.05  69.44 78.75 61.98 83.44 F (3, 36) =1.216, p>.05 /F (3, 36) =.540, p>.05  Note that scoring for the PedsQL linearly transforms item scores onto a 0-100 scale (0=100, 1=75, 2=50, 3=25, 4=0).  168 Table 24 Descriptive Statistics and Group Comparisons for the Longitudinal CES-D (‘Longitudinal Distress’ scale) 24a) Descriptive Statistics and ANOVA for the Longitudinal CES-D Mean & SD & Variance N (%) Statistics Overall Sample Total Score: Cronbach‟s Alpha: /  Comparison to normative sample b  Across four subgroups Urban On-Tmt (n=14): Urban Off-Tmt c/(n=6): Rural On-Tmt (n=14): Rural Off-Tmt (n=7):  14.39  12.36  (“Not Depressed”)a/ α = .88  /  9.25  8.58  Range = 0 - 53 /  d = .49  13.64 11.67 18.29 10.43 F (3, 37) =.806, p>.05  Levene’s HOV test:  /# Endorsing ‘+/- Stressor’ in Past Week:  Range & Significance Test Scores  F (3, 37) =.227, p>.05  15 (9.6)  a  Note that a total score of 16 or greater receives a classification of „Depressed‟ on this measure, indicating „depression symptomatology‟ or „clinically significant level of psychological distress‟ (Radloff, 1977). b This normative sample is represented by a group of US community adults (n=2514) (Radloff, 1977). c „Off-treatment‟ status in this longitudinal dataset refers to those families who (had been on-treatment at baseline but) were off-treatment at six-months‟ follow-up.  24b) Descriptive Statistics and Paired-Samples T-Test, Comparing means across Baseline CES-D and Longitudinal CES-D Mean SD & Variance Range & Statistics Significance Test Scores Baseline CES-D (n=41): Longitudinal CES-D (n=41):  18.85 14.39  13.33 12.36  Range = 1 - 54 Range = 0 - 53 t (40) = -2.718, p<.01 d = 0.35  169 Table 25 Descriptive Statistics and Group Comparisons for the Longitudinal MS (‘Longitudinal Mastery’ scale) 25a) Descriptive Statistics and ANOVA for the Longitudinal MS Mean SD & Variance Statistics Overall Sample Total Score:  20.61  Average item score: Cronbach‟s Alpha:  2.94 (“Agree”) α = .86  Across four subgroups Urban On-Tmt (n=14): Urban Off-Tmt a (n=6): Rural On-Tmt (n=14): Rural Off-Tmt (n=7):  20.36 21.67 20.00 21.43  Levene’s HOV test:  4.18  Range & Significance Test Scores Range = 10-28  F (3, 37) =.316, p>.05 F (3, 37) =.504, p>.05  a  „Off-treatment‟ status in this longitudinal dataset refers to those families who (had been on-treatment at baseline but) were off-treatment at six-months‟ follow-up.  25b) Descriptive Statistics and Paired-Samples T-Test, Comparing means across Baseline MS and Longitudinal MS Mean SD & Variance Range & Statistics Significance Test Scores Baseline MS (n=41): Longitudinal MS (n=41):  20.49 20.61  3.21 4.18  Range = 12 - 26 Range = 10 - 28 t (40) = -.247, p>.05 d = -0.03  170 Table 26 Descriptive Statistics and Group Comparisons for the Longitudinal SHS (‘Longitudinal Subjective Happiness’ scale) 26a) Descriptive Statistics and ANOVA for the Longitudinal SHS Mean SD & Variance Statistics Overall Sample Total Score (Avg item score): Cronbach‟s Alpha: / /  Comparison to normative sample 1a Comparison to normative sample 2b  Across four subgroups Urban On-Tmt (n=14): Urban Off-Tmt c (n=6): Rural On-Tmt (n=14): Rural Off-Tmt (n=7):  4.82  Range & Significance Test Scores  1.42  Range = 1-7  0.93 1.12  d = 0.68 d = -0.02  α = .92 5.62 4.80  5.16 4.58 4.25 5.46  F (3, 37) =1.645, p>.05 F (3, 37) =.868, p>.05  Levene’s HOV test: a  This normative sample is represented by a group of US community adults (n=198) (Lyubomirsky & Lepper, 1999). This normative sample is represented by a group of US female community adults (n=92) (Lyubomirsky & Lepper, 1999). c „Off-treatment‟ status in this longitudinal dataset refers to those families who (had been on-treatment at baseline but) were off-treatment at six-months‟ follow-up. b  26b) Descriptive Statistics and Paired-Samples T-Test, Comparing means across Baseline SHS and Longitudinal SHS Mean SD & Variance Range & Statistics Significance Test Scores Baseline SHS (n=41): Longitudinal SHS (n=41):  4.88 4.82  1.32 1.41  Range = 1 - 7 Range = 1 - 7 t (40) = -.377, p>.05 d = 0.04  171 Table 27 Additional Benefits Realized during the Cancer Experience by Caregivers “I have learned not to waste time and to sometimes waste time. I have learned to appreciate the importance of being a good parent; being there for them and listening and being honest with your child - being a team. Never getting emotional in front of your child when things are difficult …You need to show them stability.” “I have learned that children are so full of wisdom - I love mine.” “I learned to love, live and laugh everyday with no holding back.” “I have become impatient to things that take time away from home life.” “I no longer fear cancer or my own mortality.” “My increased reliance on God has been significant.” “My accepting help from others has been challenged and very humbling.” “The experience was life changing! Money and things don‟t mean anything but relationships are the most valuable thing in life.” “Even though my marriage ended; my ex-husband and I have become closer and we have realized that life is too short to bicker.” “I have learnt more about how wonderful, amazingly strong, and funny my daughter is. The list goes on and on….” “I appreciate more the value of positive energy and attitudes. I have opened up more fully to the possibilities and power of energy healing – It is especially empowering as a parent.” “I find that having others be responsible for my child‟s medical care has promoted volunteerism in me. I perform daily acts of kindness to the environment or to strangers.” “A benefit for me is learning the importance of time. How much or little time we are given to live life. Time is precious and not to be taken for granted.” “I am more aware of judgment passed on to others by strangers. I am not so quick to judge someone and I have a stronger sense of understanding for why someone may look like they do or act a certain way.” “The experience taught me to be more grateful for life‟s blessings.”  172 Table 28 Additional Benefits Realized during the Cancer Experience by Adolescent Patients/Survivors “I‟ve had exposure in local papers and media and have also volunteered for the Canadian cancer society. I have also met awesome nurses and doctors at the Children‟s Hospital.” “Camp Good Times is such an amazing place.” “It taught me what my future goals are; it has taught me to enjoy every second of everyday and everything in it; it has brought out some talents that I never knew I had; it raised a lot of awareness for me, my family and friends about serious childhood illness and cancer.” “I experienced benefits with the teen oncology group – I got to talk to people my age about my experiences - I made good friends. Also, I have become much more loved in the community in relation to the children‟s hospital ward because of the people I have met.” “I have been able to meet other teens who have gone through what I have in terms of treatment and surgery. I have also been exposed to much in the way of medical personnel, doctors and equipment.” “I saw more relatives, got a dog, and now I enjoy good things more.” “I‟ve learned not to take anyone for granted. I also learned that some things happen for a reason. I honestly look at my illness as a blessing; though it has not been easy. I never thought it would happen to me. So expect the unexpected!”  173 Table 29 Socio-Demographic and Cancer-Specific Variables (i.e., Precursor Variables) Removed from Further Inferential Analyses Variable to be Removed:  Elevated Inter-Correlation:  Baseline (Caregiver) Dataset: Child age at diagnosis  r = .761, ρ<.001 with Child age T r = .800, ρ<.001 with Child age  Treatment for the first time  rΦ = 1.000, ρ<.001 with On/off treatment status  Active treatment status  rΦ = 1.000, ρ<.001 with On/off treatment status  Urban/rural status  rpb = .850, ρ<.001 with Time to hospital T rpb = .766, ρ<.001 with Time to hospital  Child Caucasian status  rΦ = .842, ρ<.001 with Caregiver Caucasian status  Child spirituality/religiosity  rΦ =.762, ρ<.001 with Caregiver spirituality/religiosity  Time since treatment ended  r = .851, ρ<.001 with Time since cancer diagnosis  Adolescent Dataset: Child age at diagnosis  r = -.876, ρ<.001 with Time since diagnosis  Treatment for the first time  rΦ = 1.000, ρ<.001 with On/off treatment status  Active treatment status  rΦ = 1.000, ρ<.001 with On/off treatment status  Urban/rural status  rpb = .831, ρ<.001 with Time to hospital T rpb = .757, ρ<.001 with Time to hospital  Time since treatment ended  r = .800, ρ<.001 with Time since diagnosis  Longitudinal (Caregiver) Dataset:  T  Urban/rural status  rpb = .753, ρ<.001 with Time to hospital  Child age  r = .978, ρ<.001 with Child age at diagnosis  „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  174 Table 30 Inter-Correlation (r) scores of Concern in the Baseline (Caregiver) Sample Questionnaire Data 30a) PECI Subscale Inter-correlations (r): Long-term uncertainty T ---  Unresolved sorrow & anger  Guilt & worry  .804** .820**  .759** .771**  Unresolved sorrow & anger  ---  ---  .796**  Guilt & worry  ---  ---  ---  Long-term uncertainty T  ** Significant at p<.001. Correlations above the recommended cut-off of r=.7 are indicated in dark bold font. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  175  30b) BFS Subscale Inter-correlations (r): Acceptance T  Empathy  Appreciation  Family  Positive self-view  Reprioritization T  Benefit finding -total  Acceptance T  ---  .564** .573**  -.542** .535**  -.513** .516**  -.705** .717**  .526** .523**  -.792** .799**  Empathy  ---  ---  .521**  .583**  .607**  -.628** .631**  .840**  Appreciation  ---  ---  ---  .574**  .704**  -.525** .521**  .795**  Family  ---  ---  ---  ---  .570**  -.529** .528**  .754**  Positive self-view  ---  ---  ---  ---  ---  -.619** .620**  .872**  Reprioritization T  ---  ---  ---  ---  ---  ---  -.767** .768**  Benefit finding – total  ---  ---  ---  ---  ---  ---  ---  ** Significant at p<.001. Correlations above the recommended cut-off of r=.7 are indicated in dark bold font. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  176 30c) FNQ Subscale Inter-correlations (r) – Needs IMP Health information  Emotional support  Instrumental support  Professional support  Community support  Involvement with care  Family needstotal IMP T  Health information  ---  .305**  NS  .458**  .244*  .293**  .427** .486**  Emotional support  ---  ---  .462**  .456**  .584**  .409**  .901** .889**  Instrumental support  ---  ---  ---  .235*  .230*  .327**  .630** .635**  Professional support  ---  ---  ---  ---  .248*  .410**  .659** .664**  Community support  ---  ---  ---  ---  ---  .303**  .618** .631**  Involvement with care  ---  ---  ---  ---  ---  ---  .556** .581**  Family needs – total IMP T  ---  ---  ---  ---  ---  ---  ---  ** Significant at p<.001. Correlations above the recommended cut-off of r=.7 are indicated in dark bold font. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  177 30d) FNQ Subscale Inter-correlations (r) – Needs MET Health information  Emotional support  Instrumental support  Professional support  Community support T  Involvement with care  Family needs – total MET  Health information 1  ---  .386**  .313**  .575**  .984** .452**  .718**  .783**  Emotional support 2  ---  ---  .544**  .492**  .381** .550**  .318**  .713**  Instrumental support  ---  ---  ---  .465**  .319** .419**  .217* (Non-linear) 5  .644**  Professional support 3  ---  ---  ---  ---  .597** .508**  .407**  .802**  Community support T  ---  ---  ---  ---  ---  .693** .347**  .782** .727**  Involvement with care 4  ---  ---  ---  ---  ---  ---  .623**  ---  ---  ---  ---  ---  ---  ---  Family needs – total MET ** Significant at p<.001. Correlations above the recommended cut-off of r=.7 are indicated in dark bold font. * Significant at p<.01. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2nd correlation relating to non-transformed data. 1/2/3/4 Caution should be used in interpreting correlations pertaining to these variables because (as previously mentioned) their distributions were found to diverge from normality and transformations were not possible to improve this. 5 Note that this is an invalid correlation as the relationship between these two variables was found to be non-linear.  178 Table 31 Inter-Correlation (r) scores of Concern in the Adolescent Sample Questionnaire Data 31a) PedsQL Subscale Inter-correlations (r): Physical QOL  Emotional QOL  Social QOL T  School QOL  PedsQL TOTAL  Psychosocial health summary  Physical QOL  ---  .502*  -.727** .709**  .718**  .916**  Emotional QOL  ---  ---  -.501** .496**  .564**  .714**  .794**  Social QOL T  ---  ---  ---  -.681** .698**  -.838** .844**  -.832** .855**  School QOL  ---  ---  ---  ---  .878**  .902**  PedsQL TOTAL  ---  ---  ---  ---  ---  .956**  Psychosocial health summary  ---  ---  ---  ---  ---  ---  .759**  ** Significant at p<.001. (Note that all correlations are significant at the p<.001 level). Note that correlations above the recommended cut-off of r=.7 are indicated in dark bold font. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2nd correlation relating to non-transformed data.  179 31b) FNQ Subscale Inter-correlations (r) – Needs IMP: Health Emotional Instrumental information support support Health information  Professional support  Community support  Involvement with care  Family needs – total IMP  ---  NS  .556**  NS  NS  .432*  .653**  ---  ---  .429*  .445*  .646**  NS  .777**  ---  ---  ---  .430*  .404*  NS  .539**  ---  ---  ---  ---  .380  NS  .767**  ---  ---  ---  ---  ---  NS  .736**  ---  ---  ---  ---  ---  ---  .372  ---  ---  ---  ---  ---  ---  ---  Emotional support Instrumental support Professional support Community support Involvement with care Family needs – total IMP ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. Note that correlations above the recommended cut-off of r=.7 are indicated in dark bold font.  180 31c) FNQ Subscale Inter-correlations (r) – Needs MET: Health information  Emotional support  Instrumental support  Professional support  Community support  Involvement with care T  Family needstotal MET  Health information  ---  .412*  .490**  .544**  .547**  .598** .495**  .825**  Emotional support  ---  ---  .595**  .522**  .613**  NS NS  .698**  Instrumental support  ---  ---  ---  .550**  .451*  NS NS  .772**  Professional support  ---  ---  ---  ---  .537**  .477* .412*  .802**  Community support  ---  ---  ---  ---  ---  NS NS  .687**  Involvement with care T  ---  ---  ---  ---  ---  ---  .591** .508**  Family needstotal MET  ---  ---  ---  ---  ---  ---  ---  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. Note that correlations above the recommended cut-off of r=.7 are indicated in dark bold font. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2nd correlation relating to non-transformed data.  181 Table 32 Inter-Correlation (r) scores of Concern in the Longitudinal (Caregiver) Sample Questionnaire Data 32a) PECI Subscale Inter-correlations (r): Long-term uncertainty  Unresolved sorrow & anger  Guilt & worry  Long-term uncertainty  ---  .824**  .818**  Unresolved sorrow & anger  ---  ---  .812**  Guilt & worry  ---  ---  ---  ** Significant at p<.001. Correlations above the recommended cut-off of r=.7 are indicated in dark bold font. Note that these variables were all measured at baseline (i.e., not at six-month follow-up).  32b) HHI Correlations (r) with other Predictor Variables Hope T  Optimism  (Baseline) mastery  (Baseline) subjective happiness  Hope T  ---  -.700** .732**  -.773** .775**  -.701** .753**  Optimism  ---  ---  .703**  .738**  (Baseline) mastery  ---  ---  ---  .620**  (Baseline) subjective happiness  ---  ---  ---  ---  ** Significant at p<.001. Correlations above the recommended cut-off of r=.7 are indicated in dark bold font. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data. Note that these variables were all measured at baseline (i.e., not at six-month follow-up).  182 Table 33 Correlations (r, rpb, rΦ ) among Remaining Questionnaire Predictor Variables in the Baseline (Caregiver) Sample 33a) Correlations among Positive Predictor Variables Benefit finding  Optimism  Mastery  Hope T  Subjective happiness  Challenge  Family needstotal MET  Family needshope MET  Benefit finding  ---  NS  NS  -.305** .314**  .273**  .231*  .160  .166  Optimism  ---  ---  .601**  -.588** .615**  .594**  .299**  .283**  NS  Mastery  ---  ---  ---  -.659** .671**  .624**  .434**  .279**  .196  Hope T  ---  ---  ---  ---  -.664** .695**  -.367** .375**  -.394** .398**  -.284** .295**  Subjective happiness 1  ---  ---  ---  ---  ---  .352**  .339**  NS  Challenge  ---  ---  ---  ---  ---  ---  .275**  .196  Family needs-total MET  ---  ---  ---  ---  ---  ---  ---  .543**  Family needs-hope MET  ---  ---  ---  ---  ---  ---  ---  ---  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data. 1 Subjective happiness (one of our outcome variables) is in this table of „predictor variables‟ because it is a potential predictor variable for the regression analyses where „distress‟ will serve as the outcome variable.  183  33b) Correlations among Negative Predictor Variables Threat  Distress T  Long-term uncertainty T  (Perceived) child physical struggles T  (Perceived) child emotional struggles  (Perceived) child social struggles  Threat  ---  .517** .496**  .495** .480**  .222* .223*  .417**  .246*  Distress1 T  ---  ---  .504** .509**  .378** .335**  .484** .497**  .302** .294**  Long-term uncertainty T  ---  ---  ---  .351** NS  .403** 422**  .395** .404**  ---  ---  ---  ---  .476** .471**  .531** .509**  (Perceived) child emotional struggles  ---  ---  ---  ---  ---  .516**  (Perceived) child social struggles  ---  ---  ---  ---  ---  ---  (Perceived) child physical struggles  T  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data. 1 Distress (one of our outcome variables) is in this table of „predictor variables‟ because it is a potential predictor variable for the regression analyses where „subjective happiness‟ will serve as the outcome variable.  184 33c) Other Correlations among Predictor Variables Threat  Distress T  Long-term uncertainty T  (Perceived) child physical struggles T  (Perceived) child emotional struggles  (Perceived) child social struggles  Family needstotal IMP T N  Family needs - hope IMP N  School needs & resources IMP T N  NS  NS -.166  NS NS  NS NS  NS  NS  .264** .256**  NS  .229* .242*  Optimism  -.322**  -.446** -.502**  -.422** -.426**  -.178 -.162  -.318**  -.251*  NS NS  NS  NS NS  Mastery  -.246*  -.483** -.537**  -.407** -.427**  -.203 -.171  -.287**  -.260**  NS NS  NS  NS NS  Hope T  .375** -.385**  .508** -.577**  .354** -.384**  .196 NS  .313** -.314**  .269** -.259**  NS NS  NS NS  NS NS  Subjective happiness  -.515**  -.572** -.581**  -.397** -.395**  NS NS  -.312**  -.165  NS NS  NS  NS NS  Challenge  -.255**  -.210* -.286**  -.223* -.236*  NS NS  -.197  -.171  NS NS  NS  NS NS  Family needstotal MET  -.324**  -.280** -.279**  -.395** -.396**  -.196 NS  -.357**  -.415**  NS NS  NS  NS NS  Family needshope MET  -.239*  -.177 -.192  -.294** -.334**  NS NS  -.172  -.272**  NS NS  NS  NS -.181  Family needstotal IMP T N  NS NS  NS NS  .174 .162  NS NS  NS  NS  ---  .636** .641**  .308** .286**  Family needshope IMP N  NS  NS NS  .251* .261**  NS NS  NS  NS  ---  ---  .277** .246*  School needs & resources IMP T N  NS  NS Ns  .255** .286**  .198 NS  .157 .169  .333** .345**  ---  ---  ---  Benefit finding  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data. N „N‟ indicates a neutral variable. These variables were perceived to be neither „positive‟ nor „negative‟.  185 Table 34 Correlations (r, rpb , rΦ ) among Remaining Questionnaire Predictor Variables in the Adolescent Sample Benefit finding  Family needs – total IMP  Family needs hope IMP  Family needs – total MET  Family needs – hope MET  Physical QOL  Emotional QOL  Social QOLT  School QOL  PedsQL TOTAL  Psychosocial health summary  Benefit finding  ---  .313  NS  NS  NS  NS  NS  NS NS  NS  NS  NS  Family needs – total IMP  ---  ---  .433*  NS  NS  NS  NS  NS NS  NS  NS  NS  Family needs - hope IMP  ---  ---  ---  NS  .361  NS  NS  NS NS  NS  NS  NS  Family needs – total MET  ---  ---  ---  ---  .351  NS  .330  NS NS  NS  NS  NS  Family needs - hope MET  ---  ---  ---  ---  ---  NS  NS  NS NS  NS  NS  NS  Physical QOL  ---  ---  ---  ---  ---  ---  .502*  -.707** .709**  .718**  .916**  .759**  Emotional QOL  ---  ---  ---  ---  ---  ---  ---  -.510** .496**  .564**  .714**  .794**  Social QOLT  ---  ---  ---  ---  ---  ---  ---  ---  -.681** .698**  -.838** .844**  -.832** .855**  School QOL  ---  ---  ---  ---  ---  ---  ---  ---  ---  .878**  .902**  PedsQL TOTAL  ---  ---  ---  ---  ---  ---  ---  ---  ---  ---  .956**  Psychosocial health summary  ---  ---  ---  ---  ---  ---  ---  ---  ---  ---  ---  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2nd correlation relating to non-transformed data.  186 Table 35 Correlations (r, rpb , rΦ ) among Remaining Questionnaire Predictor Variables in the Longitudinal (Caregiver) Sample 35a) Correlations among Positive Predictor Variables (Longitudinal) mastery  Benefit finding  Hope T  Family needs – total MET  Challenge  Family needs - hope MET  (Longitudinal) subjective happiness1  (Longitudinal) mastery  ---  NS  -.585** .595**  .604**  .613**  .421*  .649**  Benefit finding  ---  ---  -.322 .365  NS  NS  NS  NS  Hope T  ---  ---  ---  -.641** .654**  -.600** .577**  -.509** .483**  -.661** .678**  Family needs – total MET  ---  ---  ---  ---  .527**  .492**  .593**  Challenge  ---  ---  ---  ---  ---  .439*  .436*  Family needs - hope MET  ---  ---  ---  ---  ---  ---  .343  (Longitudinal) subjective happiness  ---  ---  ---  ---  ---  ---  ---  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2nd correlation relating to non-transformed data. 1 Longitudinal subjective happiness (one of our outcome variables) is in this table of „predictor variables‟ because it is a potential predictor variable for the regression analyses where „longitudinal distress‟ will serve as the outcome variable.  187  35b) Correlations among Negative Predictor Variables Threat  Distress  Long-term uncertainty  (Perceived) child physical struggles  (Perceived) child emotional struggles  (Perceived) child social struggles  (Longitudinal) distress 1 T  Threat  ---  .582**  .500**  NS  .368  NS  .533** .511**  Distress  ---  ---  .481**  NS  NS  .327  .553** .667**  Long-term uncertainty  ---  ---  ---  NS  .374  .397*  .622** .698**  (Perceived) child physical struggles  ---  ---  ---  ---  .544**  .585**  NS NS  (Perceived) child emotional struggles  ---  ---  ---  ---  ---  .638**  .421* .528**  (Perceived) child social struggles  ---  ---  ---  ---  ---  ---  .381 .434*  (Longitudinal) Distress T  ---  ---  ---  ---  ---  ---  ---  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. 1 Longitudinal distress (one of our outcome variables) is in this table of „predictor variables‟ because it is a potential predictor variable for the regression analyses where „Longitudinal subjective happiness‟ will serve as the outcome variable. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  188  35c) Other Correlations among Predictor Variables Threat  Distress  Long-term uncertainty  (Perceived) child physical struggles  (Perceived) child emotional struggles  (Perceived) child social struggles  Family needs – total IMP N  Family needs - hope IMP N  School needs & resources IMP T N  (Longitudinal) distress T  -.394  -.541**  -.564**  NS  -.453*  -.312  -.442*  NS  NS NS  -.690** -.779** 1  Benefit finding  NS  -.349  NS  NS  NS  NS  NS  NS  NS NS  NS NS  Hope T  .485** -.442*  .635** -.660**  .548** -.522**  NS NS  .377 -.337  NS NS  NS NS  NS NS  NS NS  .567** -.616**  Challenge  -.314  -.627**  -.498**  NS  -.386  -.382  NS  NS  .321 -.311  -.500** -.622**  Family needs – total MET  -.421*  -.511**  -.502**  NS  -.553**  -.389  -.495**  -.351  NS NS  -.527** -.623**  Family needs - hope MET  -.358  -.389  -.531**  NS  -.327  NS  NS  NS  NS NS  -.320 -.359  Family needs – total IMP N  NS  NS  .366  NS  NS  NS  ---  .632**  NS NS  NS NS  Family needs - hope IMP N  NS  NS  .343  NS  NS  NS  ---  ---  NS NS  NS NS  School needs & resources IMP T N  NS NS  NS NS  -.480** .474*  -.374 .345  -.377 .396*  -.414* .393  ---  ---  ---  .319 .371  (Longitudinal) subjective happiness  -.507**  -.444*  -.487**  NS  -.364  NS  NS  NS  NS NS  -.611** -.683**  (Longitudinal) mastery  ** Significant at p<.001. Correlations above the recommended cut-off of r=.7 are indicated in dark bold font. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data. N „N‟ indicates neutral variable. These variables were perceived to be neither „positive‟ nor „negative‟. 1 This elevated correlation between Longitudinal Mastery and (non-transformed) Longitudinal Distress prevents these two variables from both being used as predictor variables in regression analyses on non-transformed data.  189 Table 36 Correlations (r, rpb) between Precursor Variables and Outcome Variables in the Baseline Sample 36a) Correlations with ‘Distress’ Outcome Variable (Caregiver) diagnosed mental health struggles  Time post diagnosis T  Ease of transport  (Caregiver) physical health struggles  „Other‟ family stressors  Treatment status  -.295** -.316**  -.260** -.260**  -.217* -.255**  -.200 NS  -.158 NS  -.240* NS  Distress T  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  36b) Correlations with ‘Subjective Happiness’ Outcome Variable (Caregiver) diagnosed mental health struggles  Caregiver age  Caregiver gender  (Caregiver) physical health struggles  Child age T  .164  .169  -.188  .216*  .196 .225*  Subjective happiness  * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  36c) Correlations with ‘Benefit finding’ Outcome Variable Caucasian status Benefit finding  .313**  ** Significant at p<.001.  36d) Correlations with ‘Hope’ Outcome Variable (Caregiver) previous experience with cancer Hope T  -.176 NS  * Significant at p<.05. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  190 Table 37 Correlations (r, rpb) between Independent Variables and Outcome Variables in the Adolescent Sample 37a) Correlations between Precursor Variables and Outcome Variables Household income  Child age  Caregiver age  Caregiver education  .456*  NS  NS  NS  NS  -.338  -.353  -.332  Quality of life (PedsQL TOTAL) Benefit finding  * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level.  37b) Correlations between Questionnaire Independent Variables and Outcome Variables Benefit finding  Family needs – total IMP  Family needs hope IMP  Family needs – total MET  Family needs – hope MET  Physical QOL  Emotional QOL  Social QOLT  School QOL  Quality of Life (PedsQL TOTAL)  Psychosocial health summary  Quality of Life (PedsQL TOTAL)  NS  NS  NS  NS  NS  N/A  N/A  N/A  N/A  ---  N/A  Benefit finding  ---  .313  NS  NS  NS  NS  NS  NS NS  NS  NS  NS  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. „N/A‟ means not applicable. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  37c) Correlations between Caregiver Independent Variables and Outcome Variables  Quality of life (PedsQL TOTAL) Benefit finding  (Caregiver) perceived child physical struggles  (Caregiver) perceived child social struggles  (Caregiver) perceived child emotional struggles  (Caregiver) benefit finding  (Caregiver) family needs - instrumental support IMP  (Caregiver) family needs - professional support IMP  -.491**  -.613**  -.430*  NS  NS  NS  NS  NS  NS  .332  .373  .367  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level.  191 Table 38 Correlations (r, rpb) between Precursor Variables and Outcome Variables in the Longitudinal Sample 38a) Correlations between Longitudinal distress and Questionnaire Independent Variables  Longitudinal distress T  Longitudinal distress T (cont‟d)  Threat  (Baseline) distress  Long-term uncertainty  (Perceived) child physical struggles  (Perceived) child emotional struggles  (Perceived) child social struggles  Longitudinal mastery  Benefit finding  .533** .511**  .553** .667**  .622** .698**  NS NS  .421* .528**  .381 .434*  -.690** -.779**  NS NS  Hope T  Challenge  Family needs – total MET  Family needs hope MET  Family needs – total IMP  Family needs hope IMP  School needs & resources IMP T  Longitudinal subjective happiness  -.567** -.616**  -.500** -.622**  -.527** -.623**  -.320 -.359  NS NS  NS NS  .319 .371  -.611** -.683**  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  38b) Correlations between Precursor Variables and Longitudinal Distress  Longitudinal distress T  (Caregiver) diagnosed mental health struggles  Ease of transport  Radiation  Longitudinal treatment status  -.578** -.736**  -.369 -.471*  NS -.422*  -.330 NS  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  192  38c) Correlations between Longitudinal Subjective Happiness and Questionnaire Independent Variables  Longitudinal subjective happiness  Longitudinal subjective happiness (cont‟d)  Threat  (Baseline) distress  Long-term uncertainty  (Perceived) child physical struggles  (Perceived) child emotional struggles  (Perceived) child social struggles  Longitudinal mastery  Benefit finding  -.507**  -.444*  -.487**  NS  -.364  NS  .649**  NS  Hope T  Challenge  Family needs – total MET  Family needs hope MET  Family needs – total IMP  Family needs hope IMP  School needs & resources IMP T  Longitudinal distress  -.661** .678**  .436*  .593**  .343  NS  NS  NS NS  -.611** -.683**  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2nd correlation relating to non-transformed data.  38d) Correlations between Precursor Variables and Longitudinal Subjective Happiness  Longitudinal subjective happiness * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level.  Ease of transport  (Caregiver) diagnosed mental health struggles  Child age at diagnosis  Radiation  .367  .458*  .369  .341  193 Table 39 Correlations (r, rΦ ) among Baseline (Caregiver) and Adolescent Corresponding Questionnaire Data Significant at .001 level Benefit finding (BFS/BFSC)  Significant at .05 level  Not Significant at .05 level  .332  Child struggles in physical functioning (pedsQL proxy/pedsQL)  -.602 1  Child struggles in emotional functioning (pedsQL proxy/pedsQL)  -.481*2  Child struggles in social functioning (pedsQL proxy/pedsQL T)  .403* -.406*3  Family needs – total IMP (Caregiver FNQ /Teen FNQ) Family needs – health information IMP (Caregiver FNQ/Teen FNQ)  .354  .497  Family needs – emotional support IMP (Caregiver FNQ/Teen FNQ)  .318  Family needs – informational support IMP (Caregiver FNQ/Teen FNQ)  -.024  Family needs – professional support IMP (Caregiver FNQ/Teen FNQ)  .423*  Family needs – community support IMP (Caregiver FNQ/Teen FNQ) Family needs – involvement with care IMP (Caregiver FNQ/Teen FNQ)  .005  .509  Family needs – total MET (Caregiver FNQ/Teen FNQ) Family needs – health information MET (Caregiver FNQ /Teen FNQ)  .265  .445*  194 Significant at .001 level  Significant at .05 level  Not Significant at .05 level  Family needs – emotional support MET (Caregiver FNQ /Teen FNQ)  .024  Family needs – informational support MET (Caregiver FNQ/Teen FNQ)  .012  Family needs – professional support MET (Caregiver FNQ /Teen FNQ)  .180  Family needs – community support MET (Caregiver FNQ T /Teen FNQ)  -.052  Family needs – involvement with care MET (Caregiver FNQ /Teen FNQ T)  .351 .363  Family needs – hope IMP (Caregiver FNQ /Teen FNQ)  .274  Family needs – hope MET (Caregiver FNQ/Teen FNQ)  -.071  * Significant at p<.01. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data. 1/2/3 The reason these correlations are negative is because the caregiver PedsQL measure was reverse scored so that higher scores represent greater „struggles‟ in functioning. Recall that this same measure in adolescents is scored such that higher scores represent greater „quality of life‟ (i.e., less struggles in functioning).  195 Table 40 Correlations (r) between Baseline (Caregiver) and Longitudinal (Caregiver) Corresponding Questionnaire Data Distress /Longitudinal distress T  .553** .667**  Mastery / Longitudinal mastery  .662**  Subjective happiness / Longitudinal subjective happiness  .716**  ** Significant at p<.001. Note that correlations above r=.7 are indicated in dark bold font. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  196 Table 41 Hierarchical Linear Regression Analyses in the Baseline Sample – Predicting Benefit Finding (with Non-transformed Data) 41a) Predictor Variable Inter-correlations (r, rpb, rΦ) of Interest to the Regression Analyses Predicting Benefit Finding  Distress T  Caucasian status  Family needs – instrumental support MET  Family needs – professional support MET  Family needs – involvement with care MET  Family needs – emotional support IMP  Family needs – professional support IMP  NS NS  -.225* -.239*  -.201 -.184  -.178 (non-linear) 1 -.209*  .223* .233*  NS NS  * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data. 1 Note that this is an invalid correlation as the relationship between these two variables was found to be non-linear.  197 41b) Results of Hierarchical Liner Regression Analysis Predicting Benefit Finding -Liberal Model  Benefit finding (1) Caucasian Status (2) Positive Predictor Variables: Hope Subjective happiness Family needs-total score IMP School needs & resources - IMP Challenge Family needs-total score MET Family needs-hope item MET (3) Negative Predictor Variable – Distress  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  10.0 ---.39 .66 20.5 8.5 1.0 5.7 1.4 .01  .313 ---.14 .06 .22 .19 .16 .10 .05 .01  4.1 (.000) ---1.4 (.172) .60 (.549) 3.1 (.003) 2.4 (.016) 2.1 (.035) 1.1 (.266) .61 (.540) .09 (.929)  .098 .302 ---------------------.302  16.7 (.000) 8.0 (.000) ---------------------7.0 (.000)  .098 .204 ---------------------.000  16.7 (.000) 6.1 (.000) ---------------------.01 (.929)  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  41c) Results of Hierarchical Liner Regression Analysis Predicting Benefit Finding - Conservative Model B  Benefit finding (1) Caucasian Status (2) Negative Predictor Variable – CES-D (3) Positive Predictor Variables Hope Subjective happiness Family needs-total score IMP School needs & resources - IMP Challenge Family needs-total score MET Family needs-hope item MET  β  t and ρ Values  Not showing data as model is not supported. Two-step model (in above Table) appears best.  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  .098 .120 .302 ----------------------  16.7 (.000) 10.4 (.000) 7.0 (.000) ----------------------  .098 .022 .182 ----------------------  16.7 (.000) 3.8 (.054) 5.5 (.000) ----------------------  198 41d) Results of Hierarchical Liner Regression Analysis Predicting Benefit Finding, using added interaction terms -Liberal Model  Benefit finding (1) Caucasian Status (2) Positive Predictor Variables: Hope Subjective happiness Family needs-total score IMP School needs & resources - IMP Challenge Family needs-total score MET Family needs-hope item MET (3) Interaction Terms: Caucasian X FN – total score IMP Caucasian X SNRQ - IMP Caucasian X Challenge  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  10.0 ---.39 .66 20.5 8.5 1.0 5.7 1.4 ----17.6 4.7 -.80  .313 ---.14 .06 .22 .19 .16 .10 .05 ----.54 .17 -.30  4.1 (.000) ---1.4 (.172) .60 (.549) 3.1 (.003) 2.4 (.016) 2.1 (.035) 1.1 (.266) .61 (.540) ----1.2 (.238) .68 (.499) -.81 (.422)  .098 .302 ---------------------.313 ----------  16.7 (.000) 8.0 (.000) ---------------------6.0 (.000) ----------  .098 .204 ---------------------.011 ----------  16.7 (.000) 6.1 (.000) ---------------------.74 (.528) ----------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. Note. Collinearity diagnostics show that this regression has a high degree of multicollinearity. As such, unique effects of each predictor should be interpreted with caution.  199 41e) Results of Hierarchical Liner Regression Analysis Predicting Benefit Finding, using Family needs Domain Scores in place of Total Scores -Liberal Model  Benefit finding (1) Caucasian Status (2) Positive Predictor Variables: Hope Subjective happiness Family needs – ES IMP Family needs – PS IMP School needs & resources - IMP Challenge Family needs – IS MET Family needs – PS MET Family needs – IC MET Family needs – hope item MET (3) Negative Predictor Variable – Distress  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  10.0 ---.38 .58 10.1 5.5 8.3 .95 .84 2.9 5.3 1.3 -.01  .313 ---.14 .05 .22 .10 .18 .15 .02 .08 .12 .05 -.01  4.1 (.000) ---1.4 (.173) .54 (.591) 2.8 (.006) 1.3 (.197) 2.4 (.016) 2.0 (.050) .29 (.774) .90 (.369) 1.6 (.119) .57 (.568) -.07 (.942)  .098 .342 ------------------------------.342  16.7 (.000) 6.8 (.000) ------------------------------6.2 (.000)  .098 .244 ------------------------------.000  16.7 (.000) 5.3 (.000) ------------------------------.01 (.942)  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  200 41f) Results of Hierarchical Liner Regression Analysis Predicting Benefit Finding, using Family needs Domain Scores in place of Total Scores –Conservative Model B  Benefit finding (1) Caucasian Status (2) Negative Predictor Variable – Distress (3) Positive Predictor Variables Hope Subjective happiness Family needs – ES IMP Family needs – PS IMP School needs & resources - IMP Challenge Family needs – IS MET Family needs – PS MET Family needs – IC MET Family needs – hope item MET  β  t and ρ Values  Not showing data as model is not supported. Two-step model (in above Table) appears best.  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  .098 .120 .342 -------------------------------  16.7 (.000) 10.4 (.000) 6.2 (.000) -------------------------------  .098 .022 .223 -------------------------------  16.7 (.000) 3.8 (.054) 4.8 (.000) -------------------------------  201  41g) Results of Hierarchical Liner Regression Analysis Predicting Benefit Finding, with Interaction Effects, using Family needs Domain Scores in place of Total Scores –Liberal Model  Benefit finding (1) Caucasian Status (2) Positive Predictor Variables: Hope Subjective happiness Family needs – ES IMP Family needs – PS IMP School needs & resources - IMP Challenge Family needs – IS MET Family needs – PS MET Family needs – IC MET Family needs – hope item MET (3) Interaction Effects: Caucasian X Family needs – ES IMP Caucasian X SNRQ - IMP Caucasian X Challenge  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  10.0 ---.38 .58 10.1 5.5 8.3 .95 .84 2.9 5.3 1.3 ----5.2 4.0 -.93  .313 ---.14 .05 .22 .10 .18 .15 .02 .08 .12 .05 ----.19 .14 -.35  4.1 (.000) ---1.4 (.173) .54 (.591) 2.8 (.006) 1.3 (.197) 2.4 (.016) 2.0 (.050) .29 (.774) .90 (.369) 1.6 (.119) .57 (.568) ----.62 (.535) .60 (.550) -.94 (.349)  .098 .342 ------------------------------.349 ----------  16.7 (.000) 6.8 (.000) ------------------------------5.4 (.000) ----------  .098 .244 ------------------------------.007 ----------  16.7 (.000) 5.3 (.000) ------------------------------.51 (.673) ----------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  202 Table 42 Hierarchical Linear Regression Analyses in the Baseline Sample – Predicting Hope (with Non-transformed Data) 42a) Results of Hierarchical Liner Regression Analysis Predicting Hope - Liberal Model  Hope (1) Positive Predictor Variables: Optimism Mastery Subjective happiness Benefit finding Family needs – total score MET Family needs – hope item MET Challenge (2) Negative Predictor Variables: (Perceived) child physical struggles Threat Distress Long-term uncertainty (Perceived) child emotional struggles (Perceived) child social struggles  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ---.24 2.9 1.3 .05 1.4 1.1 .02 ---.01 -.01 -.06 .27 .01 -.01  ---.21 .29 .32 .13 .07 .11 .01 ---.04 -.01 -.16 .05 .03 -.04  ---3.1 (.002) 4.0 (.000) 4.5 (.000) 2.4 (.017) 1.1 (.268) 1.9 (.062) .13 (.900) ---.56 (.574) -.11 (.914) -2.1 (.034) .66 (.510) .49 (.627) -.53 (.595)  .642 ---------------------.654 -------------------  37.9 (.000) ---------------------20.6 (.000) -------------------  .642 ---------------------.012 -------------------  37.9 (.000) ---------------------.83 (.549) -------------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  203 42b) Results of Hierarchical Liner Regression Analysis Predicting Hope - Conservative Model  Hope (1) Negative Predictor Variables: (Perceived) child physical struggles Threat Distress Long-term uncertainty (Perceived) child emotional struggles (Perceived) child social struggles (2) Positive Predictor Variables: Optimism Mastery Subjective happiness Benefit finding Family needs – total score MET Family needs – hope item MET Challenge  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ---.03 -.13 -.20 -.46 .00 -.03 ---.22 2.7 1.1 .04 1.5 1.1 .00  ---.15 -.10 -.50 -.08 .01 -.14 ---.19 .27 .27 .12 .08 .11 .00  ---1.9 (.061) -1.3 (.201) -6.0 (.000) -.97 (.335) -.14 (.890) -1.6 (.102) ---2.8 (.006) 3.5 (.001) 3.4 (.001) 2.1 (.035) 1.1 (.262) 1.7 (.087) .01 (.994)  .372 ------------------.654 ----------------------  14.7 (.000) ------------------20.6 (.000) ----------------------  .372 ------------------.282 ----------------------  14.7 (.000) ------------------16.6 (.000) ----------------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  204 42c) Results of Hierarchical Liner Regression Analysis Predicting Hope, using Domain scores of Family Needs and Benefit Finding in place of Total Scores Hope (1) Positive Predictor Variables: Optimism Mastery Subjective happiness Family needs – hope item MET Challenge Family needs – emotional support MET Family needs – instrumental support MET Family needs – professional support MET Family needs – community support MET Family needs – involvement with care MET Benefit finding - acceptance Benefit finding - empathy Benefit finding - appreciation Benefit finding - family Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  B  β  t and ρ Values  Multiple R2  F and ρ Values  ---.24 2.9 1.3 1.3 -.02 -.30 .53 1.5 -.69 -.38 .17 .00 .05 .08  ---.21 .28 .33 .13 -.01 -.02 .04 .12 -.05 -.03 .10 .00 .03 .03  ---3.1 (.003) 3.8 (.000) 4.5 (.000) 1.9 (.061) -.14 (.890) -.27 (.788) .64 (.527) 1.7 .086) -.77 (.446) -.41 (.684) 1.4 (.155) .02 (.985) .40 (.692) .44 (.660)  .653 -------------------------------------------  18.9 (.000) -------------------------------------------  205 Table 43 Hierarchical Linear Regression Analyses in the Baseline Sample – Predicting Subjective Happiness (with Non-transformed Data) 43a) Results of Hierarchical Liner Regression Analysis Predicting Subjective Happiness -Liberal Model  Subjective happiness (1) Precursor Variables: Diagnosed mental health struggles Caregiver age Caregiver gender Physical health struggles Child age (2) Positive Predictor Variables: Challenge Optimism Mastery Benefit finding Family needs – total score MET Hope (3) Negative Predictor Variables: Threat (Perceived) child emotional struggles (Perceived) child social struggles Long-term uncertainty Distress  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ---.38 .01 -.57 .76 .04 ---.02 .05 .54 .01 .12 .10 ----.07 .00 .01 -.02 -.02  ---.12 .06 -.14 .19 .15 ---.03 .17 .21 .06 .02 .39 ----.23 .01 .13 .01 -.15  ---1.6 (.114) .57 (.571) -1.7 (.088) 2.4 (.017) 1.6 (.110) ---.53 (.599) 2.3 (.023) 2.7 (.008) 1.1 (.273) .39 (.700) 4.6 (.000) ----3.5 (.001) .11 (.915) 2.0 (.044) .20 (.844) -1.9 (.058)  .126 ---------------.595 ------------------.662 ----------------  4.3 (.001) ---------------19.2 (.000) ------------------17.0 (.000) ----------------  .126 ---------------.469 ------------------.067 ----------------  4.3 (.001) ---------------27.8 (.000) ------------------5.5 (.000) ----------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  206 43b) Results of Hierarchical Liner Regression Analysis Predicting Subjective Happiness -Conservative Model  Subjective happiness (1) Precursor Variables: Diagnosed mental health struggles Caregiver age Caregiver gender Physical health struggles Child age (2) Negative Predictor Variables: Threat (Perceived) child emotional struggles (Perceived) child social struggles Long-term uncertainty Distress (3) Positive Predictor Variables: Challenge Optimism Mastery Benefit finding Family needs – total score MET Hope  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ---.38 .01 -.57 .76 .04 ----.07 .00 .00 -.10 -.05 ---.00 .04 .59 .01 .27 .07  ---.12 .06 -.14 .19 .15 ----.23 .01 .06 -.08 -.45 ---.01 .13 .23 .06 .05 .29  ---1.6 (.114) .57 (.571) -1.7 (.088) 2.4 (.017) 1.6 (.110) ----3.0 (.003) .17 (.867) .76 (.451) -.98 (.327) -5.5 (.000) ---.09 (.931) 1.9 (.064) 3.0 (.003) .99 (.325) .82 (.415) 3.5 (.001)  .126 ---------------.460 ---------------.662 -------------------  4.3 (.001) ---------------12.3 (.000) ---------------17.0 (.000) -------------------  .126 ---------------.334 ---------------.202 -------------------  4.3 (.001) ---------------17.9 (.000) ---------------13.8 (.000) -------------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  207 Table 44 Hierarchical Linear Regression Analyses in the Baseline Sample – Predicting Distress (with Non-transformed Data) 44a) Remaining Predictor Variable Inter-correlations (r) of Interest to the Regression Analyses Predicting Distress (Caregiver) diagnosed mental health struggles  Time post diagnosis T  Ease of transport  (Caregiver) physical health struggles  Treatment status  Benefit finding  NS  NS NS  NS  NS  NS  Family needs – hope item MET  NS  NS NS  NS  NS  NS  „NS‟ means not (statistically) significant at p<.05 level. „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  T  208  44b) Results of Hierarchical Liner Regression Analysis Predicting Distress - Liberal Model  Distress (1) Precursor Variables: Diagnosed mental health struggles Ease of transport Time post diagnosis (2) Positive Predictor Variables: Optimism Mastery Subjective happiness Challenge Family needs – total score MET Family needs – hope item MET Hope Benefit finding (3) Negative Predictor Variables: (Perceived) child emotional struggles (Perceived) child social struggles Threat Long-term uncertainty (Perceived) child physical struggles  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ----8.1 -3.0 -.08 ----.23 -3.3 -2.7 -.34 1.4 -1.7 -.42 .03 ---.11 -.01 1.7 1.7 .03  ----.26 -.23 -.25 ----.08 -.13 -.27 -.06 .03 -.07 -.17 .03 ---.18 -.01 .11 .12 .07  ----3.6 (.000) -3.1 (.002) -3.4 (.001) ----.98 (.330) -1.5 (.129) -3.1 (.002) -.90 (.369) .38 (.703) -.97 (.332) -1.8 (.078) .53 (.595) ---2.6 (.010) -.17 (.867) 1.6 (.123) 1.6 (.103) 1.1 (.287)  .205 ---------.534 ------------------------.616 ----------------  13.1 (.000) ---------15.0 (.000) ------------------------14.0 (.000) ----------------  .205 ---------.329 ------------------------.082 ----------------  13.1 (.000) ---------12.7 (.000) ------------------------6.0 (.000) ----------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  209  44c) Results of Hierarchical Liner Regression Analysis Predicting Distress -Conservative Model  Distress (1) Precursor Variables: Diagnosed mental health struggles Ease of transport Time post diagnosis (2) Negative Predictor Variables: (Perceived) child emotional struggles (Perceived) child social struggles Threat Long-term uncertainty (Perceived) child physical struggles (3) Positive Predictor Variables: Optimism Mastery Subjective happiness Challenge Family needs – total score MET Family needs – hope item MET Hope Benefit finding  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ----8.1 -3.0 -.08 ---.13 0.0 3.5 3.3 .02 ----.06 -3.0 -1.8 -.23 5.5 -.76 -.43 .01  ----.26 -.23 -.25 ---.22 -.01 .22 .23 .03 ----.02 -.12 -.19 -.04 .11 -.03 -.17 .01  ----3.6 (.000) -3.1 (.002) -3.4 (.001) ---2.9 (.005) -.07 (.941) 3.1 (.002) 3.2 (.002) .45 (.657) ----.28 (.781) -1.4 (.156) -2.1 (.038) -.65 (.519) 1.6 (.123) -.47 (.643) -2.0 (.052) .09 (.928)  .205 ---------.480 ---------------.616 -------------------------  13.1 (.000) ---------17.0 (.000) ---------------14.0 (.000) -------------------------  .205 ---------.275 ---------------.136 -------------------------  13.1 (.000) ---------15.6 (.000) ---------------6.2 (.000) -------------------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  210 Table 45 Exploratory Regression Analyses in the Baseline Sample – Predicting Mastery (with Non-transformed Data) 45a) Regression Analysis Predicting Mastery with Long-Term Uncertainty Interaction Terms Mastery (1) Interaction Terms: Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty  X Optimism X Family needs –total score MET X Family needs –hope item MET X (Perceived) child emotional struggles X (Perceived) child social struggles X (Perceived) child physical struggles  B  β  t and ρ Values  Multiple R2  F and ρ Values  ---.02 -.20 .02 .00 .00 .00  ---.38 -.22 .03 -.43 -.22 .10  ---3.7 (.000) -2.0 (.046) .33 (.741) -3.5 (.001) -1.8 (.074) .81 (.419)  .242 -------------------  7.9 (.000) -------------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  45b) Additional Domain scores of Family Needs that significantly correlate (r) with Mastery Family needs – community support MET Mastery  * Significant at p<.01.  .244*  211  45c) Regression Analysis Predicting Mastery with Long-Term Uncertainty Interaction Terms, using Domain scores of Family Needs and Benefit Finding in place of Total Scores Mastery (1) Interaction Terms: Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty Long-term uncertainty  X Optimism X Benefit finding – acceptance X Benefit finding – appreciation X Family needs –emotional support MET X Family needs –professional support MET X Family needs –instrumental support MET X Family needs –health information MET X Family needs –involvement with care MET X Family needs –community support MET X (Perceived) child emotional struggles X (Perceived) child social struggles X (Perceived) child physical struggles  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  B  β  t and ρ Values  Multiple R2  F and ρ Values  ---.03 -.01 -.01 -.01 .13 -.08 -.41 .11 .11 .00 .00 .00  ---.59 -.19 -.15 -.01 .19 -.11 -.59 .18 .16 -.44 -.05 .11  ---5.2 (.000) -1.2 (.222) -.88 (.379) -.07 (.947) 1.6 (.106) -1.1 (.259) -3.3 (.001) 1.2 (.209) 1.5 (.129) -3.6 (.000) -.37 (.713) .85 (.398)  .341 -------------------------------------  6.2 (.000) -------------------------------------  212 Table 46 Hierarchical Linear Regression Analyses in the Longitudinal Sample – Predicting Longitudinal Distress (with Non-transformed Data) 46a) Results of Hierarchical Liner Regression Analysis Predicting Longitudinal Distress -Liberal Model Longitudinal Distress (1) (Caregiver) diagnosed mental health struggles (2) Positive Predictor Variable: Hope (3) Negative Predictor Variable: Long-term uncertainty (Baseline) distress  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  -21.7 -.80 ---4.9 .08  -.74 -.36 ---.36 .09  -6.8 (.000) -3.3 (.002) ---3.4 (.002) .64 (.526)  .542 .642 .730 -------  46.1 (.000) 34.0 (.000) 24.4 (.000) -------  .542 .100 .089 -------  46.1 (.000) 10.6 (.002) 5.9 (.006) -------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  46b) Results of Hierarchical Liner Regression Analysis Predicting Longitudinal Distress -Conservative Model Longitudinal Distress (1) (Caregiver) diagnosed mental health struggles (2) Negative Predictor Variables: Long-term uncertainty (Baseline) distress (3) Positive Predictor Variable - Hope  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  -21.7 ---5.5 .18 -.40  -.74 ---.41 .19 -.18  -6.8 (.000) ---4.0 (.000) 1.6 (.129) -1.5 (.153)  .542 .714 ------.730  46.1 (.000) 30.8 (.000) ------24.4 (.000)  .542 .172 ------.016  46.1 (.000) 11.2 (.000) ------2.1 (.153)  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  213  46c) Results of Hierarchical Liner Regression Analysis Predicting Longitudinal Distress, with added interaction term -Conservative Model  Longitudinal Distress (1) (Caregiver) diagnosed mental health struggles (2) Negative Predictor Variable – Long-term uncertainty (3) Interaction Term (Long-term uncertainty X Caregiver previous mental health struggles)  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  -21.7 6.0 -2.8  -.74 .45 -.30  -6.8 (.000) 4.4 (.000) -.99 (.328)  .542 .696 .704  46.1 (.000) 43.4 (.000) 29.3 (.000)  .542 .154 .008  46.1 (.000) 19.2 (.000) .98 (.328)  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  214 Table 47 Hierarchical Linear Regression Analyses in the Longitudinal Sample – Predicting Longitudinal Subjective Happiness (with Nontransformed Data) 47a) Results of Hierarchical Liner Regression Analysis Predicting Longitudinal Subjective Happiness -Liberal Model  Longitudinal Subjective happiness (1) (Caregiver) diagnosed mental health struggles (2) Positive Predictor Variables: Hope Family needs – total score MET (3) Negative Predictor Variable – Threat  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  1.5 ---.13 1.1 -.08  .46 ---.50 .20 -.21  3.2 (.003) ---3.2 (.002) 1.2 (.257) -1.6 (.117)  .210 .505 ------.538  10.3 (.003) 12.6 (.000) ------10.5 (.000)  210 .296 ------.033  10.3 (.003) 11.1 (.000) ------2.6 (.117)  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  47b) Results of Hierarchical Liner Regression Analysis Predicting Longitudinal Subjective Happiness, with added interaction term Liberal Model B β t and ρ Multiple F and ρ Change Change in F 2 Values R Values in R2 and ρ Longitudinal Subjective Happiness (1) (Caregiver) diagnosed mental health struggles (2) Positive Predictor Variables: Hope Family needs – total score MET (3) Interaction Term (Hope X Caregiver previous mental health struggles)  1.5 ---.13 1.1 -.01  .46 ---.50 .20 -.02  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  3.2 (.003) ---3.2 (.002) 1.2 (.257) -.05 (.958)  .210 .505 ------.505  10.3 (.003) 12.6 (.000) ------9.2 (.000)  210 .296 ------.000  10.3 (.003) 11.1 (.000) ------.00 (.958)  215  47c) Results of Hierarchical Liner Regression Analysis Predicting Longitudinal Subjective Happiness -Conservative Model  Longitudinal Subjective Happiness (1) (Caregiver) diagnosed mental health struggles (2) Negative Predictor Variable – Threat (3) Positive Predictor Variables Hope Family needs – total score MET  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  1.5 -.15 ---.11 1.0  .46 -.38 ---.44 .19  3.2 (.003) -2.5 (.015) ---2.9 (.007) 1.1 (.284)  .210 .324 .538 -------  10.3 (.003) 9.1 (.001) 10.5 (.000) -------  .210 .114 .214 -------  10.3 (.003) 6.4 (.015) 8.4 (.001) -------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font.  216 Table 48 Hierarchical Linear Regression Analyses in the Baseline Sample – Predicting Benefit Finding (with Transformed Data) 48a) Remaining Predictor Variable Inter-correlations (r, rpb, rΦ) of Interest to the Regression Analyses Predicting Benefit Finding Caucasian status  Family needs – instrumental support MET  Family needs – professional support MET 1  Family needs – involvement with care MET2  Family needs – emotional support IMP  Family needs – professional support IMP  Family needs – total score MET  NS  N/A  N/A  N/A  -.244*  NS  Family needs – hope item MET  NS  .405**  .395**  .256**  NS  NS  Challenge  NS  .207  .295**  .181  NS  NS  Hope T  NS NS  -.294** .299**  -.353** .347**  -.220* .235*  NS NS  NS NS  Subjective happiness  NS  .199  .270**  .256**  NS  NS  Family needs – total score IMP T  NS NS  NS NS  NS (non-linear) 3 NS (non-linear) 4  NS NS  N/A N/A  N/A N/A  School needs & resources - IMP T  NS .192  NS NS  -.231* -.270** (non-linear) 5  NS NS  .249* .244*  .242* .240*  Family needs – instrumental support MET  NS  ---  ---  ---  ---  ---  Family needs – professional support MET  NS  .465**  ---  ---  ---  ---  Family needs – involvement with care MET  NS  .217*(Non-linear) 6  .407**  ---  ---  ---  Family needs – emotional support IMP  NS  NS  NS  -.185  ---  ---  Family needs – professional support IMP  NS  NS  NS  NS  .456**  ---  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data. 1/2 Caution should be used in interpreting correlations pertaining to these variables because (as previously mentioned) their distributions were found to diverge from normality and transformations were not possible to improve this. 3/4/5/6 Note that these are invalid correlations as the relationships between the respective variables were found to be non-linear.  217 48b) Results of Hierarchical Liner Regression Analysis Predicting Benefit Finding  Benefit finding (1) Caucasian Status (2) Positive Predictor Variables: Hope T Subjective happiness Family needs-total score IMP T School needs & resources - IMP T Challenge Family needs-total score MET Family needs-hope item MET  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  10.0 ----2.1 .80 30.8 6.9 1.1 5.1 1.3  .31 ----.12 .07 .22 .17 .17 .09 .05  4.1 (.000) ----1.2 (.236) .75 (.454) 3.0 (.003) 2.2 (.033) 2.2 (.028) 1.0 (.320) .55 (.585)  .098 .294  16.7 (.000) 7.6 (.000)  .098 .196  16.7 (.000) 5.8 (.000)  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. T „T‟ indicates transformed variable.  48c) Results of Hierarchical Liner Regression Analysis Predicting Benefit Finding, with added interaction terms  Benefit finding (1) Caucasian Status (2) Positive Predictor Variables: Hope T Subjective happiness Family needs-total score IMP T School needs & resources - IMP T Challenge Family needs-total score MET Family needs-hope item MET (3) Interaction Terms: Caucasian X FN-total score IMP T Caucasian X SNRQ - IMP T Caucasian X Challenge  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  10.0 ----2.1 .80 30.8 6.9 1.1 5.1 1.3 ----21.3 3.4 -.86  .31 ----.12 .07 .22 .17 .17 .09 .05 ----.62 .13 -.32  4.1 (.000) ----1.2 (.236) .75 (.454) 3.0 (.003) 2.2 (.033) 2.2 (.028) 1.0 (.320) .55 (.585) ----.90 (.368) .50 (.617) -.85 (.395)  .098 .294 ---------------------.302 ----------  16.7 (.000) 7.6 (.000) ---------------------5.7 (.000) ----------  .098 .196 ---------------------.008 ----------  16.7 (.000) 5.8 (.000) ---------------------.56 (.639) ----------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. T „T‟ indicates transformed variable.  218 48d) Correlations between Family needs Subscales and Benefit Finding (r) Benefit finding Family needs – health information (HI) IMP  NS  Family needs – emotional support (ES) IMP  .264**  Family needs – instrumental support (IS) IMP  NS  Family needs – professional support (PS) IMP  .294**  Family needs – community support (CS) IMP  NS  Family needs – involvement with care (IC) IMP  NS  Family needs – health information (HI) MET 1  NS  Family needs – emotional support (ES) MET 2  NS  Family needs – instrumental support (IS) MET  .167  Family needs – professional support (PS) MET 3  .192 (non-linear) 5  Family needs – community support (CS) MET  NS NS  Family needs – involvement with care (IC) MET 4  .164  ** Significant at p<.001. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data. 1/2/3/4 Caution should be used in interpreting correlations pertaining to these variables because (as previously mentioned) their distributions were found to diverge from normality and transformations were not possible to improve this. 5 Note that this is an invalid correlation as the relationship between these two variables was found to be non-linear.  219 48e) Results of Hierarchical Liner Regression Analysis Predicting Benefit Finding, using Family needs Domain Scores in place of Total Scores  Benefit finding (1) Caucasian Status (2) Positive Predictor Variables: Hope T Subjective happiness Family needs – ES IMP Family needs – PS IMP School needs & resources - IMP T Challenge Family needs – IS MET Family needs – PS MET Family needs – IC MET Family needs – hope item MET  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  10.0 ----2.1 .71 10.0 5.7 6.7 1.1 .91 2.5 5.2 1.1  .31 ----.12 .06 .22 .11 .16 .16 .03 .07 .12 .04  4.1 (.000) ----1.2 (.233) .67 (.502) 2.7 (.007) 1.3 (.184) 2.1 (.034) 2.1 (.038) .31 (.757) .78 (.435) 1.5 (.130) .50 (.620)  .098 .334  16.7 (.000) 6.6 (.000)  .098 .236  16.7 (.000) 5.1 (.000)  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. T „T‟ indicates transformed variable.  220  48f) Results of Hierarchical Liner Regression Analysis Predicting Benefit Finding, with Interaction Effects, using Family needs Domain Scores in place of Total Scores  Benefit finding (1) Caucasian Status (2) Positive Predictor Variables: Hope T Subjective happiness Family needs – ES IMP Family needs – PS IMP School needs & resources - IMP T Challenge Family needs – IS MET Family needs – PS MET Family needs – IC MET Family needs – hope item MET (3) Interaction Effects: Caucasian X Family needs – ES IMP Caucasian X SNRQ - IMP T Caucasian X Challenge  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  10.0 ----2.1 .71 10.0 5.7 6.7 1.1 .91 2.5 5.2 1.1 ----.97 2.9 -.40  .31 ----.12 .06 .22 .11 .16 .16 .03 .07 .12 .04 ----.36 .11 -.15  4.1 (.000) ----1.2 (.233) .67 (.502) 2.7 (.007) 1.3 (.184) 2.1 (.034) 2.1 (.038) .31 (.757) .78 (.435) 1.5 (.130) .50 (.620) ----.97 (.334) .44 (.662) -.475 (.635)  .098 .334 ------------------------------.340 ----------  16.7 (.000) 6.6 (.000) ------------------------------5.2 (.000) ----------  .098 .236 ------------------------------.006 ----------  16.7 (.000) 5.1 (.000) ------------------------------.44 (.724) ----------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. T „T‟ indicates transformed variable.  221 Table 49 Hierarchical Linear Regression Analyses in the Baseline Sample – Predicting Hope (with Transformed Data) 49a) Remaining Predictor Variable Inter-correlations (rpb, rΦ) of Interest to the Regression Analyses Predicting Hope Previous Experience with Cancer Optimism  NS  Mastery  NS  Subjective happiness  NS  Benefit finding  NS  Family needs – total score MET  NS  Family needs – hope item MET  NS  Challenge  -.187  (Perceived) child physical struggles T  NS NS  Threat  NS  Distress T  NS NS  Long-term uncertainty T  NS NS  (Perceived) child emotional struggles  NS  (Perceived) child social struggles  NS  „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  222  49b) Results of Hierarchical Liner Regression Analysis Predicting Hope - Liberal Model  Hope T (1) Positive Predictor Variables: Optimism Mastery Subjective happiness Benefit finding Family needs – total score MET Family needs – hope item MET Challenge (2) Negative Predictor Variables: (Perceived) child physical struggles T Threat Distress T Long-term uncertainty T (Perceived) child emotional struggles (Perceived) child social struggles  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ----.03 -.49 -.18 -.01 -.27 -.15 .00 ---.01 .01 .05 -.14 .00 .00  ----.18 -.31 -.28 -.13 -.09 -.10 -.01 ---.02 .03 .10 -.07 -.03 .03  ----2.6 (.010) -4.1 (.000) -3.8 (.000) -2.3 (.023) -1.3 (.194) -1.5 (.131) -.12 (.903) ---.28 (.281) .38 (.708) 1.3 (.201) -.88 (.383) -.43 (.670) .42 (.677)  .601 ---------------------.609 -------------------  31.8 (.000) ---------------------17.0 (.000) -------------------  .601 ---------------------.008 -------------------  31.8 (.000) ---------------------.47 (.829) -------------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. T „T‟ indicates transformed variable. Note. In reviewing the (direction of the) beta weights, keep in mind that hope is a (reflected) transformed variable.  223  49c) Results of Hierarchical Liner Regression Analysis Predicting Hope - Conservative Model  Hope T (1) Negative Predictor Variables: (Perceived) child physical struggles T Threat Distress T Long-term uncertainty T (Perceived) child emotional struggles (Perceived) child social struggles (2) Positive Predictor Variables: Optimism Mastery Subjective happiness Benefit finding Family needs – total score MET Family needs – hope item MET Challenge  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ----.02 .02 .19 .14 .00 .01 ----.03 -.47 -.16 -.01 -.27 -.15 -.01  ----.08 .12 .40 .07 .02 .13 ----.18 -.30 -.25 -.11 -.09 -.09 -.02  ----.92 (.359) 1.3 (.183) 4.4 (.000) .76 (.451) .19 (.852) 1.4 (.152) ----2.5 (.013) -3.7 (.000) -2.9 (.005) -1.9 (.060) -1.2 (.228) -1.4 (.151) -.26 (.797)  .294 ------------------.609 ----------------------  10.3 (.000) ------------------17.0 (.000) ----------------------  .294 ------------------.315 ----------------------  10.3 (.000) ------------------16.3 (.000) ----------------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. T „T‟ indicates transformed variable. Note. In reviewing the (direction of the) beta weights, keep in mind that hope is a (reflected) transformed variable.  224 49d) Correlations between Hope and Domains of Family Needs and Benefit Finding that significantly correlate with Hope (r, rpb) Hope T  Optimism  Mastery  Subjective happiness  Family needshope MET  Challenge  Family needsES MET1  Family needsIS MET  Family needsPS MET 2  Family needsCS MET T  Family needsIC MET 3  Benefit finding – acceptance  T  Benefit findingempathy  Benefit findingappreciation  Benefit finding family  Family needs – ES MET  -.312** .320**  .205*  .228*  .260**  .619**  NS  ---  .544**  .492**  .381** .550**  .318**  NS non-linear 4 NS non-linear 5  NS  .225*  .271**  Family needs – IS MET  -.294** .299**  .224*  .200  .199  .405**  .207  ---  ---  .465**  .319** .419**  .217*  NS NS  .160  .181  .258**  Family needs – PS MET  -.353** .347**  .172  .223*  .270**  .395**  .295**  ---  ---  ---  .597** .508**  .407**  NS NS  NS  .270**  .217*  Family needs – CS MET T  -.254** .312**  .227* .244*  NS .302**  .237* .327**  .334** .382**  .209* .207*  ---  ---  ---  ---  .693** .347**  NS NS  NS NS  .252* .198  NS NS  Family needs – IC MET  -.220* .235*  .185  .159  .256**  .256**  .181  ---  ---  ---  ---  ---  -.173 .158  NS  .246*  .194  Benefit finding -  .332** .314**  -.162 .163  -.207* .203  -.240* .226*  NS NS  -.228* .225*  ---  ---  ---  ---  ---  ---  .564** .573**  -.542** .535**  -.513** .516**  Benefit finding – empathy  -.190 .190  NS  NS  NS  .164  .166  ---  ---  ---  ---  ---  ---  ---  .521**  .583**  Benefit finding-  -.354** .366**  .249*  .184  .369**  .203  .235*  ---  ---  ---  ---  ---  ---  ---  ---  .574**  -.223* .240*  NS  NS  .217*  NS  NS  ---  ---  ---  ---  ---  ---  ---  ---  ---  (Nonlinear) 6  acceptance T  appreciation  Benefit finding family  ** Significant at p<.001. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data. 1/2/3 Caution should be used in interpreting correlations pertaining to these variables because (as previously mentioned) their distributions were found to diverge from normality and transformations were not possible to improve this. 4/5/6 Note that these are invalid correlations as the relationships between the respective variables were found to be non-linear. Note that benefit finding – positive self-view and family needs – health information MET were not assessed here because we know from Tables 1b and 1c, respectively, that these variables correlate above the .70 threshold with other variables included in this table.  225 49e) Results of Hierarchical Liner Regression Analysis Predicting Hope, using Domain scores of Family Needs and Benefit Finding in place of Total Scores B  β  t and ρ Values  Multiple R2  F and ρ Values  ----.03 -.46 -.18 -.18 .01 .07 -.08 -.27 -.02 .12 .17 .00 -.01 .00  ----.19 -.29 -.29 -.11 .01 .03 -.04 -.13 .00 .05 .14 .00 -.02 .00  ----2.6 (.011) -3.8 (.000) -3.8 (.000) -1.6 (.116) .23 (.820) .40 (.687) -.57 (.569) -1.8 (.080) -.05 (.959) .63 (.530) 2.0 (.046) -.05 (.960) -.29 (.771) -.02 (.987)  .618 -------------------------------------------  16.3 (.000) -------------------------------------------  T  Hope (1) Positive Predictor Variables: Optimism Mastery Subjective happiness Family needs – hope item MET Challenge Family needs – emotional support MET Family needs – instrumental support MET Family needs – professional support MET Family needs – community support MET T Family needs – involvement with care MET Benefit finding - acceptance T Benefit finding - empathy Benefit finding - appreciation Benefit finding - family  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. T „T‟ indicates transformed variable. Note. In reviewing the (direction of the) beta weights, keep in mind that both hope are benefit finding – acceptance are (reflected) transformed variables.  226 Table 50 Hierarchical Linear Regression Analyses in the Baseline Sample – Predicting Subjective Happiness (with Transformed Data) 50a) Remaining Predictor Variable Inter-correlations (r, rpb) of Interest to the Regression Analyses Predicting Subjective Happiness (Caregiver) diagnosed mental health struggles  Caregiver age  Caregiver gender  (Caregiver) physical health struggles  Child age T  Hope T  -.176 .208*  NS .162  NS NS  NS NS  NS NS  Optimism  .174  .223*  -.213*  NS  NS NS  Mastery  .190  NS  NS  NS  NS NS  Benefit finding  NS  NS  NS  NS  NS NS  Challenge  NS  NS  NS  NS  NS NS  .207*  NS  NS  NS  .234* .240*  NS NS  NS NS  NS NS  NS NS  NS NS  Distress T  -.295** -.316**  NS NS  NS NS  -.200 -.217*  NS NS  (Perceived) child emotional struggles  -.210*  NS  NS  NS  NS NS  (Perceived) child social struggles  NS  NS  NS  NS  NS NS  Threat  NS  NS  NS  -.237*  NS NS  Family needs – total score MET Long-term uncertainty T  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  227 50b) Results of Hierarchical Liner Regression Analysis Predicting Subjective Happiness -Liberal Model  Subjective happiness (1) Precursor Variables: Diagnosed mental health struggles Caregiver age Caregiver gender Physical health struggles Child age T (2) Positive Predictor Variables: Challenge Optimism Mastery Benefit finding Family needs – total score MET Hope T (3) Negative Predictor Variables: Threat (Perceived) child emotional struggles (Perceived) child social struggles Long-term uncertainty T Distress T  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ---.37 .01 -.57 .75 .62 ---.03 .06 .60 .01 .12 -.51 ----.07 .00 .01 .02 -.14  ---.12 .04 -.14 .19 .19 ---.04 .20 .24 .08 .03 -.32 ----.21 .02 .13 .01 -.19  ---1.6 (.122) .37 (.711) -1.7 (.083) 2.4 (.018) 2.0 (.044) ---.69 (.492) 2.7 (.008) 3.0 (.003) 1.3 (.196) .40 (.693) -3.9 (.000) ----3.1 (.002) .30 (.763) 2.0 (.044) .07 (.948) -2.6 (.010)  .134 ---------------.585 ------------------.661 ----------------  4.7 (.001) ---------------18.5 (.000) ------------------16.9 (.000) ----------------  .134 ---------------.451 ------------------.075 ----------------  4.7 (.001) ---------------26.1 (.000) ------------------6.2 (.000) ----------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. T „T‟ indicates transformed variable. Note. In reviewing the (direction of the) beta weights, keep in mind that hope is a (reflected) transformed variable.  228 50c) Results of Hierarchical Liner Regression Analysis Predicting Subjective Happiness -Conservative Model  Subjective happiness (1) Precursor Variables: Diagnosed mental health struggles Caregiver age Caregiver gender Physical health struggles Child age T (2) Negative Predictor Variables: Threat (Perceived) child emotional struggles (Perceived) child social struggles Long-term uncertainty T Distress T (3) Positive Predictor Variables: Challenge Optimism Mastery Benefit finding Family needs – total score MET Hope T  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ---.37 .01 -.57 .75 .62 ----.07 .00 .00 -.30 -.31 ---.02 .05 .61 .01 .26 -.36  ---.12 .04 -.14 .19 .19 ----.22 -.01 .06 -.09 -.41 ---.03 .16 .24 .07 .05 -.23  ---1.6 (.122) .37 (.711) -1.7 (.083) 2.4 (.018) 2.0 (.044) ----2.7 (.007) -.09 (.926) .78 (.439) -1.2 (.252) -5.0 (.000) ---.46 (.649) 2.2 (.027) 3.2 (.002) 1.2 (.232) .82 (.416) -2.9 (.004)  .134 ---------------.448 ---------------.661 -------------------  4.7 (.001) ---------------11.8 (.000) ---------------16.9 (.000) -------------------  .134 ---------------.314 ---------------.213 -------------------  4.7 (.001) ---------------16.5 (.000) ---------------14.5 (.000) -------------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. T „T‟ indicates transformed variable. Note. In reviewing the (direction of the) beta weights, keep in mind that hope is a (reflected) transformed variable.  229 Table 51 Hierarchical Linear Regression Analyses in the Baseline Sample – Predicting Distress (with Transformed Data) 51a) Remaining Predictor Variable Inter-correlations (r, rpb) of Interest to the Regression Analyses Predicting Distress (Caregiver) diagnosed mental health struggles  Time post diagnosis T  Ease of transport  (Caregiver) physical health struggles  Treatment status  Hope T  -.176 .208*  NS NS  NS NS  NS NS  NS NS  Optimism  .174  NS NS  .204  NS  NS  Mastery  .190  NS NS  NS  NS  NS  Subjective happiness  .164  NS NS  NS  .216*  NS  Challenge  NS  NS NS  NS  NS  NS  .207*  NS NS  .171  NS  NS  Long-term uncertainty T  NS NS  NS NS  -.183 -.168  NS NS  -.167 NS  (Perceived) child physical struggles T  NS NS  -.303** -.257**  NS NS  NS NS  -.344** -.356**  (Perceived) child emotional struggles  -.210*  NS NS  NS  NS  -.203  (Perceived) child social struggles  NS  NS NS  -.197  NS  NS  Threat  NS  NS NS  -.158  -.237*  -.198  Family needs – total score MET  ** Significant at p<.001. * Significant at p<.01. „NS‟ means not (statistically) significant at p<.05 level. T „T‟ indicates transformed variable. Two correlations are reported for these variables, with the 2 nd correlation relating to non-transformed data.  230  51b) Results of Hierarchical Liner Regression Analysis Predicting Distress -Liberal Model  Distress T (1) Precursor Variables: Diagnosed mental health struggles Ease of transport Physical health struggles Treatment status 1 Time post diagnosis T (2) Positive Predictor Variables: Optimism Mastery Subjective happiness Challenge Family needs – total score MET Hope T (3) Negative Predictor Variables: (Perceived) child emotional struggles (Perceived) child social struggles Long-term uncertainty T (Perceived) child physical struggles T Threat  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ----1.0 -.32 -.78 -.05 -.80 ----.02 -.40 -.44 .03 -.27 .30 ---.01 .00 .79 .06 .07  ----.24 -.18 -.14 -.02 -.25 ----.05 -.12 -.32 .03 -.04 .14 ---.14 .03 .18 .09 .15  ----3.2 (.002) -2.4 (.017) -1.9 (.059) -.16 (.880) -2.5 (.013) ----.54 (.591) -1.3 (-.200) -3.6 (.000) .49 (.628) -.59 (.556) 1.5 (.145) ---1.9 (.058) .39 (.696) 2.4 (.017) 1.2 (.220) 2.0 (.044)  .205 ---------------.482 ------------------.588 ----------------  7.8 (.000) ---------------12.2 (.000) ------------------12.4 (.000) ----------------  .205 ---------------.277 ------------------.106 ----------------  7.8 (.000) ---------------12.8 (.000) ------------------7.2 (.000) ----------------  Note. Significant findings (at the .05 level or smaller) are indicated in bold font. T „T‟ indicates transformed variable. 1 For qualitative purposes, this model was re-run replacing treatment status with active treatment status. No significant differences in the model were observed. The t-score for the active treatment status beta weight was t=-.23, p=.817. Recall that this variable highly correlated (r>.70) with treatment status and was therefore removed from consideration in further inferential analyses.  231  51c) Results of Hierarchical Liner Regression Analysis Predicting Distress -Conservative Model  Distress T (1) Precursor Variables: Diagnosed mental health struggles Ease of transport Physical health struggles Treatment status Time post diagnosis T (2) Negative Predictor Variables: (Perceived) child emotional struggles (Perceived) child social struggles Long-term uncertainty T (Perceived) child physical struggles T Threat (3) Positive Predictor Variables: Optimism Mastery Subjective happiness Challenge Family needs – total score MET Hope T  B  β  t and ρ Values  Multiple R2  F and ρ Values  Change in R2  Change in F and ρ  ----1.0 -.32 -.78 -.05 -.80 ---.01 .00 1.1 .06 .11 ---.01 -.31 -.33 .04 .65 .25  ----.24 -.18 -.14 -.02 -.25 ---.17 .01 .24 .10 .25 ---.01 -.09 -.24 .05 .10 .12  ----3.2 (.002) -2.4 (.017) -1.9 (.059) -.15 (.880) -2.5 (.013) ---2.1 (.036) .13 (.893) 3.3 (.001) 1.2 (.227) 3