Rune , . ‘Q‘I ‘v ”fie. nmfww $3.. «\ud n. a: . \u..¢\1 zs‘vwnfi. . file.» ”.95 3%“. ‘ . . vi . .. ‘15.. “um. .1 a mum . 3...». gnaw: .1329. #0.. I leO ‘6- . club. 3 uni ates? oivtns. x\v. O¢.I\tul. (.113: lull! 12mg ? i :4. 2-1. \ .Pttlnuvnq!£oh.! than .1.r...}......qh.u...nb..1&....ufi 1 3 l3 .3! ‘..l. 4.3.. 4.1.! 111:: . :1: .qu.‘ do» Rub... “a .I.‘ I1 VOA-.7. .rmm‘ mgwnmit .91 3 . :. v .x .. . $.52: . LlBfiARY Michigan State universnty -———.—..- _. This is to certify that the dissertation entitled FATIGUE, SELF-EFFICACY, AND PHYSICAL FUNCTIONAL STATUS IN PERSONS WITH LUNG CANCER presented by AMY JUDE HOFFMAN has been accepted towards fulfillment of the requirements for the Ph.D. degree in NURSING ‘7 %Pfifess6fs gignature {/6702 Date MSU is an affirmative-action, equal-opportunity employer .. —.-.-.--->-----o--u---o---- PLACE IN RETURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE NOV 0 4 2009 00} M23365 :1 093:!) 1919; flirt}? 9 H P ' .1 II all. g c7 '3 (ASEPLZSI I0” FEB 53 7 2“] it “at 1 cow-E ‘ 6/07 p:lCIRC/Date0ue.indd-p.1 FATIGUE, SELF -EFF ICACY, AND PHYSICAL FUNCTIONAL STATUS IN PERSONS WITH LUNG CANCER By Amy Jude Hoffman A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY College of Nursing 2007 ABSTRACT FATIGUE, SELF -EF F ICACY, AND PHYSICAL FUNCTIONAL STATUS IN PERSONS WITH LUNG CANCER By Amy Jude Hoffman Cancer-related fatigue (CRF) is a prevalent and severe symptom that is inadequately managed and accompanied by other unpleasant symptoms that negatively impact the physical functional status (PFS) of persons with cancer and especially those with lung cancer (LC). Symptom management occurs through self-directed action, with perceived self-efficacy (PSE) being a key factor. Existing theories and studies do not address the key role PSE plays in a person’s ability to manage symptoms and improve their PFS, making the current study unique in persons with LC and other cancer (0C) diagnoses. Secondary data analyses from baseline observation of two randomized control trials were performed on 63 persons with LC and 235 persons with 0C diagnoses who were undergoing a course of chemotherapy. For the total sample and in the LC and 0C groups separately, the hypothesis of mediation from CRF to PFS through PSE for fatigue management was tested showing significant support for partial mediation In the total sample, the magnitude of the relationship between CRF and PFS was reduced after PSE for fatigue management was controlled, with the mediation accounting for 12% of the variance (t = -2.59; p = .009). Consequently, CRF severity directly influences PFS and indirectly influences PFS by its effect on PSE for fatigue management Further, on a 0-10 scale (10 = most severe), similar levels of CRF severity were reported by persons with LC (M= 5.88; SD = 2.00) and OC (M= 5.83; SD = 2.29) diagnoses (t = -.161;df= 296; p = .872). However, through blockwise, hierarchical multiple regression, similar levels of CRF severity were found to significantly worsen the PFS of persons with LC as compared to OC diagnoses (t = -3.78). In addition to type of cancer diagnoses, five other factors in the total sample were identified through blockwise, multiple hierarchical regression as the most important factors accounting for 47.7% of the explained variance in PFS [F (28, 295) = 8.68, p = .000]. Specifically, higher levels of PSE for fatigue management (t = 3.55) were found to be one of the strongest predictors of greater PFS, while lower levels of PFS were predicted by greater total CRF severity (t = -5.39), greater number of co-morbid conditions (I = -4.20), greater total symptom severity (t = -2.46), and having surgery prior to chemotherapy (t = -2.31). Lower levels of PSE for fatigue management were identified through best of all subset regression to be a predictor of greater CRF severity in the total sample and in the LC and OC groups. Persons with LC (M = 4.99; SD = 1.43) as compared to OC (M = 4.54; SD = 1.60) diagnoses reported higher levels of total severity of the other unpleasant symptoms (I = ~1.99; df= 294; p = .047). Through path analyses, the CRF severity had a direct effect on increasing the total symptom severity of the other unpleasant symptoms (I = 9.69) which lowered the PFS (t = -2.7 l) for persons with LC and 0C diagnoses. The findings indicate that CRF is related to the presence of other symptoms, and PSE is an important factor in optimizing CRF management and PFS. This study provides the foundation for future intervention studies to increase PSE to achieve optimal symptom management and PFS in persons with cancer. Copyright by AMY JUDE HOFFMAN 2007 This work is dedicated to my Grandfather, William D. Lawton, who was a surgical technician overseas in World War II, and who died on February 19, 1976, after being diagnosed with lung cancer. I was only 10 years old at the time, and a Girl Scout, but I remember it as though it were yesterday. He was brave in his fight against the disease even though he had multiple occurring, severe symptoms of the disease and its treatment. As a little girl, I do remember the horror of the symptoms he experienced and how the symptoms robbed him of his dignity and quality of life. My hope is that this work will lead to improved symptom management and quality of life for persons with ltmg cancer so they will not have to suffer as my dear Grandfather did. O 0.. ACKNOWLEDGEMENTS Audrey G. Gift, PhD, RN, F AAN, Dissertation Committee Chairperson and Sponsor of my National Research Service Award is especially recognized for her outstanding mentorship throughout my tenure at Michigan State University. I am exceptionally appreciative for her expert teaching in the areas of respiratory disease particularly lung cancer, symptom experience and management, and theory. I was very fortunate to find a mentor for the past five years that took a clunce on and a special interest in me; challenged me to learn and do more; encouraged me when I felt like quitting; appealed to my highest level of thinking to attain the highest level of performance; cheered when I passed a milestone; inspired a fixture full of opportunity; and modeled the way to becoming a competent nurse scientist. Barbara A Given, PhD, RN, FAAN, is acknowledged for her guidance as a Dissertation Committee Member. I would also like to extend my deepest gratitude to her for serving as an exceptional mentor and providing me with research training opportunities spanning the research continuum during my Mary Margaret Walther Cancer Research Fellowship, Behavioral Cooperative Oncology Group, and the Walther Cancer Institute. Because of you, I learned more than I ever expected or knew possible! Also, I am very appreciative to her generosity in sharing the Family Home Care for Cancer: A Community-based Model for Symptom Management Project experience and data with me. I am greatly appreciative to Dr. Given for her sharing her knowledge, time, and expertise in the areas of symptom experience and management for persons with cancer. O 0.. Charles w. Given, PhD, is acknowledged for his guidance as a Dissertation Committee Member and Co-Sponsor of my National Research Service Award I am very grateful to his sharing of the Automated Telephone Information and Monitoring of Symptoms Project experience and data with me. I have appreciated his sharing of expertise in behavioral interventions for symptom management and functional status in persons with cancer and challenged me to think differently. Marilyn Rothert, PhD, RN, F AAN, is acknowledged for her substantial guidance and support as a Dissertation Committee Member. I have valued her direction in the area of health promotion, particularly in issues (self-efficacy) surrounding clinical decision-making so that at the end of the day I could know that what I was doing as a nurse researcher made a difference. Alexander von Eye, PhD, is most graciously acknowledged for his guidance and support as a Dissertation Committee Member. I lmve prized his outstanding leadership in psychology, research methodology, and statistical analysis. Dr. von Eye, you have a gift for empowering others to act Your gift of committed exemplary teaching paired with your ability to infuse confidence in others like myself, allowed me to take risks. I would never have made it without your ceaseless help. Ruth Ann Brintnall, DNSc(c), AOCN, HPCN, APRN-BC is acknowledged for her serving as a Clinical Consultant for my National Research Service Award and her ongoing support. Ruthann works tirelessly. You will find her caring for patients before the beginning and beyond the end of the day when any patient or family need has been identified. 1 am so gratefiil for Ruth Ann addressing my need by taking a vii O 0.. special interest in me and sharing with me her clinical oncology expertise and knowledge. Elise Lev, EdD, RN, CS, is recognized for serving as a Concept Consultant for my National Research Service AWard in the area of Self-Efficacy Theory. I am so very grateful for her self-less, spontaneous, and generous help at a time when I most needed it! I would also like to extend a very special thank you to the following: 0 v Linda Spence, PhD, RN, for support and guidance in Self-Efficacy Theory. 0 3° Manfred Stommel, PhD, for serving as a member of my Doctoral Program Committee and guidance in developing a measurement tool for self-efiicacy for cancer-related fatigue. 0 Tenko Raykov, PhD, for tutoring me to increase my statistical knowledge and ability in structural equation modeling. 03° Linda Scott, PhD, RN, Susan Bosold, PhD, RN, Cynthia Coviak, PhD, RN, CNE, and Katherine Kim, PhD, RN, for helping me see possibilities and encouraging me to purse my doctorate education. 0 0.. My great friends found in the doctoral program at Michigan State University: Ardith Doorenbos, PhD, RN; Susan Dunn, PhD, RN; Evelyn Gladney, MSN, RN; Anita J ablonski, PhD, RN; and, Jackie Keehne-Miron, PhD, RN. Thank You! Lastly, I don’t know where to begin to express the appreciation and love I have for my wonderful husband, Dave. His encouragement and support was limitless in the past five years during my doctoral education and training, from listening and cheering me on to going the extra mile in whatever needed to be done so I could attend class, viii write a paper, study, or travel to present scientific manuscripts, and rest a little more often to get through our surprise pregnancy, etc. Thank you, Dave! We made it, we finished the marathon together with our Sam and Nicholas and Sasha! I love you! Personal Funding Source: 0 0.0 O O 0.0 National Institutes of Health, National Institute of Nursing Research, Individual Ruth L. Kirschstein National Research Service Award, Grant 1F3l NR009621-01Al. Project Title: Fatigue, Self-Efficacy, and Functional Status in Persons with Lung Cancer. Awarded April 1, 2006 to present. Mary Margaret Walther Cancer Research Fellowship. Behavioral Cooperative Oncology Group. Walther Cancer Institute, Indianapolis, Indiana. Awarded July 1, 2005 to March 31, 2006. Blue Cross Blue Shield of Michigan Foundation, Grant Number 1044.SAP. Project Title: Fatigue, Self-Efficacy, and Functional Status in Persons with Lung Cancer. Awarded 2005 to 2007. Oncology Nursing Society Foundation 9th National Conference on Cancer Nursing Research Doctoral Scholarship. Awarded 2007. Oncology Nursing Society Foundation Congress Scholarship. Awarded 2006. Sigma Theta Tau International Honor Society of Nursing, Kappa Epsilon Chapter-At- Large, MI. Sigma Theta Tau. Research Grant. Awarded 2005. Alpha Psi Chapter Research Scholarship Award. Michigan State University. East Lansing, Michigan. Awarded 2005. ix 03° Graduate School Summer Acceleration Fellowship. Michigan State University. East Lansing, Michigan. Awarded 2005. 0:0 John F. Dunkel Scholarship Award, College of Nursing, Michigan State University, 2003 and 2004. .30 Graduate School Fellowship and Calder Scholarship. Michigan State University. East Lansing, Michigan Awarded 2002. Data Source: The Family Home Care for Cancer: A Community-based Model for Symptom Management” (FHCC) project (R01 CA-079280) sponsored by Barbara A Given, PhD., RN, FAAN, Principm Investigator, and “The Automated Telephone Monitoring for Symptom Management” (ATSM) project (R01 CA-30724) sponsored by Charles W. Given, PhD., Principal Investigator. TABLE OF CONTENTS CHAPTERONE CHAPTERTWO THEORETICAL FRAMEWORK AND REVIEW OF THE LITERATURE. Background... Significance... Theoretical Framework Reviewofthe Literature... 11 Patient Characteristics...... ll Perceived Self-Efficacy 15 Functional Status... l9 \lO‘Ji-h- CHAPTERTHREE PurposeandResearch Questions 24 Sample... 25 OperatronalDefinltlons 26 Instruments... 28 Protection ofHuman Subjects 38 DataAnalysisPlan... 43 CI-IAPTERFOUR PrelrmrnaryExammatlonoftheData 51 Descriptive Statistics for Physiological Patient Characteristics .... 52 Descriptive Statistics for Contextual Patient Characteristics... 58 Descriptive Statistics for Symptoms... 62 Descriptive Statistics for Perceived Self-Efficacy and Performance Outcome 65 ResearchQuestion#1... 66 ResearchQuestion#Z................. ResearchQuestion#3 ResearchQuestion#4................... ResearchQuestion#S ResearchQuestion#6................... ResearchQuestion#7 ResearchQuestion#S....... CHAPTER FIVE DISCUSSION. .. Relationship between the Patient Charactenstlcs Other Unpleasant Symptoms, PSE for Fatigue Management to the Prediction of CRF... .... The Symptom Experience and Relationship between CRF and the Other Unpleasant Symptoms from Cancer and Cancer Treatment... . . .... Relationship between CRF, PSE for Fatigue Management, and PF S . Relationship between Patient Characteristics, CRF, Other Unpleasant Symptoms, PSE for Fatigue Management, and PFS... .. .. Strengths... .. .. Limitations . . Implications for Study Significance for Clinical Practice Significance for Further Research... Closing........ APPENDICES... . ... Appendix A: Executive Summary for the Family Home Care for Cancer: A Community-Based Model (FHCC), Grant # R01 CA-79280... . Appendix A: Executive Summary for the Automated Telephone Monitoring for Symptom Management (ATSM), Grant # R01 CA-30724... . .. ... Appendix B: Collaborating Sites for the FHCC and ATSM Studies... Appendix C: Summary of Measures for Research... Appendix D: Symptom(s): Items from the Brief Fatigue Inventory Appendix E: Symptom(s): The Symptom Experience Inventory... . Appendix F: Self-Efficacy for Fatigue in Patients with Cancer Scale .............. xii 77 96 107 110 114 121 133 138 139 147 154 155 168 170 172 173 177 183 186 187 190 193 194 196 197 203 207 LIST OF TABLES Table 1. Physiological Patient Characteristics by Percentage of the Total SampleandbyGroup... 56 Table 2. Co—morbid Conditions by the Total Sample and by Group... 57 Table 3. Contextual Patient Characteristics by Percentage of the Total SampleandbyGroup... . . 60-61 Table 4. Comparison of Persons with Lung Cancer and Other Cancer Diagnoses Mean Scores on Symptom Severity Scores... .. 64 Table 5. Comparison of Persons with Lung Cancer and Other Cancer Diagnoses Mean Scores on Perceived Self-Efficacy for Fatigue Management and Physical Functional Status... .... .. 66 Table 6. Results of Step 1-6 Identifying the Best Model Predicting Total CRF Severity In Persons with Cancer Using a Theoretically Driven Backward Elimination Regression Model... . . .... 73-76 Table 7. Significance Testing of Predictors to Total CRF Severity In Persons Table 8. Results of Step 1-7 Identifying the Best Model Predicting Total CRF Severity In Persons with Other Cancer Diagnoses Using a Theoretically Driven Backward Elimination Regression Method... . 82-85 Table 9. Significance Testing of the Predictors to Total CRF Severity In PersonswithOtherCancerDiagnoses... 86 Table 10. Results of Step 1-8 Identifying the Best Model Predicting Total CRF Severity In Persons with Lung Cancer Using a Theoretically Driven Backward Elimination Regression Method... . .. .. 91-95 Table 11. Significance Testing of the Predictors to Total CRF Severity In PersonswithLungCancer." 96 Table 12. Rank, Frequency, and Percentage of Symptoms Reported by the Total Sample and by Group 1n the Past Seven Days from the BaselineInterview... ... . 98 xiii Table 13. Table 14. Table 15. Table 16. Table 17. Table 18. Table 19. Table 20. Table 21. Table 22. Rank, Frequency, and Percentage of Symptoms Reported by the Total Sample and by Group On All Seven Days Prior to the Baseline Interview... . 100 Rank and Mean Symptom Severity" of the Total Sample and ”by Group... 103 Rank and Weighted Mean Symptom Severity Score for Persons with Lung Cancer and Other Cancer Diagnoses... 104 Correlations Among Total CRF Severity, Individual Symptom Severity, and Total Symptom Severity for the Total Sample and by CancerGroup... .. 106 Mediation Analysis for the Total Sample (11 = 298)... 109 Mediation Analysis for Persons with Other Cancer Diagnoses (n= 235).. 111 Mediation Analysis for Persons with Lung Cancer (11 = 63) ......... 113 Results of a Hierarchical Multiple Regression Analysis of the Tested Effects of Selected Variables on Physical Functional Status ofPersons with Cancer (n=296)... 118-120 Results of a Hierarchical Multiple Regression Analysis of the Tested Effects of Selected Variables on Physical Functional Status ofPersons with Other Cancer Diagnoses (n = 235)... 124-126 Results of a Hierarchical Multiple Regression Analysis of the Tested Effects of Selected Variables on Physical Functional Status ofPersons with Lung Cancer (11 = 63)... 130-132 xiv LIST OF FIGURES Figurel. TheoreticalFramework 10 Figure2. ConceptualModelforPathAnalysis................................. 136 Figure 3. Parsimonious Final Model... 137 CHAPTER ONE INTRODUCTION Persons with cancer report many troublesome symptoms. This is particularly true for persons with lung cancer (LC) who report even more symptoms than persons with other cancer (0C) diagnoses (A Doorenbos, C. Given, B. Given, & N. Verbitsky, 2006b; B Given, Given, Azzouz, & Stommel, 2001). Fatigue is an especially prevalent and distressing symptom in the cancer population (Chan, Richardson, & Richardson, 2005; Cooley, 2000; Cooley, Short, & Moriarty, 2003; de Jong, Kester, Schouten, Abu-Saad, & Courtens, 2006; Hickok et al., 2005; Irvine, Vincent, Graydon, Bubela, & Thompson, 1994; Oi-Ling, Man-Wall, & Kam-Hung, 2005; Okuyama et al., 2001; Sarna & Brecht, 1997; Schwartz et al., 2000; Visser et al., 2006). Fatigue is accompanied by many other severe symptoms that are poorly managed by patients and professionals (Degner & Sloan, 1995; Gift, Jablonski, Stommel, & Given, 2004; C Given, Given, Azzouz, Kozachik, & Stommel, 2001; McCorkle & Benoliel, 1983). In some populations, a positive relationship between a person’s perceived self-efficacy (PSE) (perception of ability) and his/her actual ability to manage symptoms has been shown (Barnason et al., 2003; Cheng & Boey, 2002; Federman, Amstein, & Caudill, 2002; Gardner et al., 2003; King, Wessel, Bharnbhani, Sholter, & Maksymowych, 2002; Lorig et al., 2001; Mathiowetz, Matuska, & Murphy, 2001; Pariser, O'I-Ianlon, & Espinoza, 2005; Wassem & Dudley, 2003; K. Wong, Wong, & Chan, 2005). Symptoms are one of the major determinants of physical firnctional status (PFS) (D Brown, McMillan, & Milroy, 2005 ; Byar, Berger, Bakken, & Cetak, 2006; Dodd, Miaskowski, & Paul, 2001; A. Doorenbos, B. Given, C. Given, & N. Verbitsky, 2006a; B. Given et al., 2001; Handy et al., 2002; Kurtz, Kurtz, Stommel, Given, & Given, 1999a, , 2000; Scott et al., 2003). Poorly managed symptoms impair daily functioning, interfere with cancer treatment, reduce quality of life, and jeopardize survival possibilities (Cleeland, 2000, , 2001; Vogelzang et al., 1997). However, to date, no research has been conducted studying the relationships between fatigue, PSE, and PFS in the cancer population. Demonstrating a relationship between increased PSE and the management of cancer-related fatigue (CRF), will support the designing of nursing interventions that help persons living with cancer increase their PSE so they can better manage their CRF and maintain their optimal PFS. The National Comprehensive Cancer Network (NCCN) together with other researchers substantiate that gaps in knowledge exist for the effective management of CRF (Ahlberg, Ekrnan, Gaston-Johansson, & Mock, 2003; Cleeland, 2001; Curt et al., 2000; Dean & Stahl, 2002; Ferrell, Grant, Dean, & Funk, 1996; V Mock, 2001; V Mock et al., 2001; Stricker, Drake, Hoyer, & Mock, 2004). The NCCN defines CRF as “a distressing persistent, subjective sense of tiredness or exhaustion related to cancer or cancer treatment that is not proportional to recent activity and interferes with usual functioning” (National Comprehensive Cancer Network, 2006). The inadequacy of CRF management continues to impact a person’s PFS (Lutz et al., 2001; V. Mock et al., 2000; Passik et al., 2002; Stone et al., 2003). This inconsistency between the state-of-the- science and the enormity of the problem was noted by the panel of experts at the “State- of-the-Science Conference on Symptom Management in Cancer: Pain, Depression, and Fatigue” held at the National Institutes of Health (NIH) in July 2002. The panel of experts recognized the limited scope of research and advocated that more study needs to be devoted to the definition, occurrence, assessment, and treatment of CRF (National Institutes of Health, 2003 ). This study is unique since few descriptive and interventional studies have been dedicated to the LC population regarding fatigue and its accompanying unpleasant symptoms (Ahlberg et al., 2003; Carr et al., 2002; Okuyama et al., 2001). Currently, most CRF management is carried out by patients via selfocare strategies (Curt et al., 2000; Stone et al., 2003). These strategies are often impacted by the person’s level of fatigue and PSE. In order to manage fatigue, it is critical to know what the person thinks of his/her ability to manage fatigue and how it impacts their self-directed action. Perceived self-efficacy forms the basis of any decision to act, the course of action selected, the degree of effort exerted, and the perseverance to continue in the face of obstacles and adversity (Bandura, 1997). Thus, the ability to exercise control over self- directed action is fundamental to symptom management and other actions which underpin the day-to-day responsibility of living with a life threatening chronic illness such as cancer. If the relationships between CRF, PSE, and PFS are found to be significant, they will provide the foundation for future intervention studies to increase PSE to achieve optimal symptom management and PFS in persons with cancer, and particularly for those with LC. CHAPTER TWO THEORETICAL FRAMEWORK AND REVIEW OF THE LITERATURE Background This year approximately 1.5 million Americans will learn they have cancer and 600,000 will die from the disease. The five-year survival rate for all cancers diagnosed between 1996 and 2002 is 66%, up fiom 51% from 1975 to 1977 due to earlier detection of cancer and advances in treatment. The National Cancer Institute estimates that approximately 10.5 million Americans with a history of cancer were alive in 2003 (American Cancer Society, 2007). Additionally, this year in the United States among men, solid tumors of the prostate, lung, and colon will account for approximately 54% of all newly diagnosed cancers. For women in 2007, the three most commonly diagnosed types of cancer will be breast, lung, and colon accounting for about 52% of all new estimated cases (Jemal et al., 2007). Consequently, many Americans in 2007 will learn they have been diagnosed with cancer, die from cancer, and will be living longer with the effects of the disease and its treatment. Regrettably, one of the major effects of cancer and its treatment is the burden of multiple concurrent symptoms (Miaskowski et al., 2006; Walsh & Rybicki, 2006). A panel of experts at the National Institutes of Health State-of-the—Science Conference on Symptoms Management in Cancer concluded that despite the fact that research is producing novel approaches to the causes and cures of cancer, research used to diagnose, treat, and manage even the most common symptoms such as pain, fatigue, and depression lag behind (National Institutes of Health, 2003). The effects of these symptoms and the inadequacy of symptom management are one of the major determinants of functional status and the inadequacy of symptom management affects a person’s functional status. In light of these facts, managing symptoms related to the effects of cancer and cancer treatment is important to optimize patient physical functional status (PFS). In the United States and throughout the world, lung cancer (LC) is the most common type of newly diagnosed cancer affecting both men and women (American Cancer Society, 2007; Jemal et al., 2007; Parkin, Bray, Ferlay, & Pisani, 2005; World Health Organization, 2003). Likewise, in the United States, LC is the most common cause of cancer-related mortality surpassing those of breast, prostate, and colorectal cancer deaths combined (American Cancer Society, 2007). The majority of persons with LC suffer fi'om multiple concurrent severe symptoms (Chan et al., 2005; Cooley, 2000; Cooley et al., 2003; Cooley, Short, & Moriarty, 2002; Fox & Lyon, 2006; Gift et al., 2004; Gift, Stommel, J ablonski, & Given, 2003). It has been noted that persons with LC may suffer a disproportionate symptom experience in comparison to persons with other cancer diagnoses (0C). Persons with LC have more symptoms than other patients with solid tumors who are newly diagnosed and at the end-of-life (Doorenbos et al., 2006b; B. Given et al., 2001). Moreover, the level of symptom severity and distress has been reported to rise until death in persons with LC (Degner & Sloan, 1995; Sarna, l993a, , 1998). Lung cancer is classified clinically as small cell (13%) and non-small cell (87%) (American Cancer Society, 2006). While significant advances have been made in LC treatment, the five-year survival rate for all stages of non-small cell LC is a dismal 15%, — and only 6%, for small cell LC (American Cancer Society, 2003). Life extending and palliative polychemotherapy regimens for persons with non-small cell LC have a one- year survival rate of 30% to 40% (Ramalingam & Belani, 2002). A recent, important study done by Winton et al. indicates greater length of survival for those with resected early stage non-small cell LC with adj uvant vinorelbine plus cisplatin (2005). However, greater survival length comes at a cost with the occurrence of multiple symptoms with fatigue being the most prevalent during chemotherapy which underscores the necessity for extending research to better manage the symptoms and improve functioning (W inton et al., 2005). Persons with small cell LC who present at diagnosis with extensive staged disease have a median life expectancy of 10 to 14 months (Ramalingam & Belani, 2002). The progressive decline for most persons with LC is due to the advanced stage of disease at diagnosis, the presence of pro-existing co-morbidities often associated with advancing age, and ineffective curative treatment. Significance The overall Healthy People 2010 objective for cancer supports this study stating its objective is: “To reduce the overall cancer death rate as well as illness, disability, and death by cancer ”(Healthy People 2010, 2001). This study parallels the current research endeavors of the National Institute of Nursing Research and the Oncology Nursing Society in the area of symptom management, a key nursing-sensitive patient outcome for the promotion of the delivery of high-quality cancer care (B Given & Sherwood, 2005). Moreover, this research project stems fi'om a larger randomized clinical trial that meets the core vision as articulated in the NIH Roadmap (National Institutes of Health, 2006). Consequently, given the enormity of the incidence of cancer, particularly LC, and the suffering of symptoms that corresponds with the disease trajectory, this research would greatly benefit those living with LC by optimizing his/her ability to manage CRF and to improve his/her functional status. Theoretical Framework The theoretical framework for this study draws upon a synthesis of the Theory of Unpleasant Symptoms (TOUS) with Self-Efficacy Theory to examine the relationships among the variables within this study. Introduction to the Theory of Unpleasant Symptoms The TOUS, developed by Lenz, Pugh, Milligan, Gift, and Suppe (1997), demonstrates the complexity of the symptom experience. Most models of symptoms focus on one symptom and specifically on the intensity of the symptom, not the quality, distress, or duration. The TOUS was the first to portray multiple symptoms occurring together and relating to each other in a multiplicative manner (Gift, 2003). Symptoms occurring together are depicted as catalyzing each other. Thus, this theory uniquely allows for the presence of multiple symptoms and implies that management of one symptom will contribute to the management of other symptoms. The current literature refers to these co—occurring symptoms as symptom clusters. The three components of the TOUS are the patient characteristics influencing the symptoms, the symptoms themselves, and the performance outcomes. The three components are interacting, and each component influences every other component The first component, the patient characteristics are categorized as physiological, psychological, and contextual. Physiological characteristics are commonly what describe the severity of the disease, such as co-morbidities, abnormal blood studies, or other pathological findings. Psychological characteristics affecting the symptom experience may include the person’s mood, affective reaction to disease, degree of uncertainty regarding the symptoms, meaning ascribed to the symptoms, and knowledge about the symptoms. Contextual characteristics refer to the social and physical environment that may affect the person’s symptom experience and their reporting of that experience, including social support, marital status, employment status, access to health care resources, lifestyle behaviors such as diet and exercise, and other resources. Symptoms, the second component of the TOUS, can be considered alone or in combination Symptoms have the dimensions of time (frequency and/or rate of occurrence; duration), severity (intensity), quality (description of qualifiers), and distress (bother). The quality dimension may be especially difficult, depending on the culture and language of the person, and the number of symptoms experienced at the same time. The final component of the TOUS is performance, the “outcome” or “consequences” of the symptom experience. Symptoms affect performance. Performance includes functional (e. g., physical activity; ADLs; social activities and interaction; and role performance including work and other role-related tasks) and cognitive (e.g., ability to concentrate; problem solve and/or think) activities. Performance in this study emphasizes the functional performance, specifically PFS. Introduction to Bandura ’s Self-Efiicacy Theory The addition of Bandura’s Self-Efficacy Theory proposes that PSE serves as a possible mediator between symptoms such as CRF and functional status. Parallel to Bandura’s definition, PSE is the person’s perception of ability to execute behavior(s) to manage his or her symptoms such as CRF (1986). Bandura has identified PSE as a powerful mediator linked to successful outcome attainment (1997). According to Baron and Kenny, mediators are defined most often as an internal property of a person that transforms the independent variable to explain how or why outcomes occur (1986). Bandura (1997) and Baron and Kenny (1986) stress the need to assess PSE for mediation rather than simply presuming such a link to develop a more precise understanding of the relationship between the independent and outcome variable. Like the TOUS, Self-efficacy Theory posits interacting relationships among the factors in a given phenomenon; but, interactive relationships are not a focus of this study and are not included in the model for this study. The theoretical framework of this research is presented in Figure 1. For persons with cancer (LC and OC diagnoses), the patient characteristics were examined as they relate to CRF severity in order to identify which characteristics relate to CRF. Next, the symptom experience of persons with cancer (LC and 0C diagnoses) was examined, specifically, the relationship between CRF and other unpleasant symptoms. Also, PSE for fatigue management was examined as a mediator between CRF and PFS. Lastly, the unique contribution of the physiological and contextual characteristics, CRF, other unpleasant symptoms, and PSE to the PFS of persons with cancer (LC and 0C diagnoses) were examined. seam accuceem easan can gnaw wean—umrcuoeuo 3233 Simon. a a dense 2 Essence 53533 3:83.. sea 8553 sec: . 6% San— EES—afim o . cannon—cm no 33 o 58m .552 . 3 o hoe-U fie §§< gm Eggnog e58 3352 3283 e acoaoaoa ”m0: a seam sang... noses . e a 1 :3 accesses acme: . 3% 355m d 33:52.6 1 «:0qu 2982.00 . . 83.883 Banana. . Egg . .583 Bunch 330m 0 .850 no ouSm d 9%.“. c an 1333 r . . / .\\ . . . l ‘ .uoflowaeaz newton Sm sesameom Beacon. fieBoEEm 303285. A 059m 10 Review of the Literature The review of the literature follows the components of the theoretical framework A review of studies related to each component of the theoretical framework will be presentedfromlefitorightinthe model. Patient Characteristics The challenge to managing CRF is its multidimensionality. The current literature provides limited evidence of the etiology of CRF in the general cancer population and even less for persons with LC. In this study, physiological (i.e., type and stage of cancer, treatment modalities, co-morbid conditions, sex, age, & laboratory values) and contextual (i.e., race, marital status, level of education, employment and health insurance data) characteristics that previous research has proposed were included in the study to be related to the subjective experience of CRF. These characteristics affect the PPS of a person with cancer, particularly LC when faced with CRF and other unpleasant symptoms. Physiological Characteristics Given that persons with LC experience greater numbers of symptoms in comparison to other cancer (0C) diagnoses, one might expect a difference by the type of cancer diagnosis in the CRF experience. Additionally, among patients 65 years of age and older persons with LC were found to be at more risk of reporting pain and/or fatigue as compared with persons with breast cancer (C. Given et al., 2001). For stage of cancer, within a diverse type of cancer population, data indicate that advanced cancer is more likely to be related to both pain and fatigue when compared with pain alone (C. Given et al., 2001). Fatigue has been identified as a highly prevalent and most distressing ll symptom over time during various types of treatment for LC, such as chemotherapy, surgery, radiation therapy, and combined treatment (Cooley et al., 2003). Within a varied type of cancer population, data showed that patients with three or more co-morbid conditions reported increased fatigue (C. Given et al., 2001). For sex, when compared with men, women were more likely to report fatigue (C. Given et al., 2001). However, for both sex and age, women 60 years of age or greater with metastatic cancer found fatigue to be more disruptive than men. Recommendations for further investigation of the relationship between CRF and laboratory values have been advised (V Mock, 2001; National Institutes of Health, 2003). Since associations between cancer treatment (e.g., chemotherapy, radiation) and symptoms have been found (Cooley et al., 2003), this study examined the respective associations of absolute neutr0phil count (ANC) and hemoglobin levels with CRF. Hemoglobin was selected since it is a protein in the red blood cell that carries oxygen Low levels of hemoglobin cause a reduction in the amount of oxygen that can be carried to tissues in the body. A decreased delivery of oxygen has been related to fatigue (National Comprehensive Cancer Network, 2006). Absolute neutrophil count represents a subset of the white blood cell count and neutrophils are responsible for fighting infection. Neutropenia is an abnormal decrease in the number of neutrophils and has been established as a strong predictor of infection in persons with cancer (National Comprehemive Cancer Network, 2005). The Infectious Disease Society of America defines neutropenia as an ANC less than 500 cells/mm3 or an ANC of 500 to 1000 cells/mm3 in patients where fin'ther decline is expected (Hughes et al., 2002). It is important to note that neutropenia can occur when the total white blood cell count is within a normal range of (4000-10,000 cells/mm3). Consequently, 12 quantifying ANC is essential to achieving a correct assessment of neutrophil status. Pathophysiologicai processes leading to energy imbalance such as infection as indicated by the AN C level may be related to the report of fatigue (Fortner, Tauer, Okon, Houts, & Schwartzberg, 2005; Gutstein, 2001; Molassiotis & Chan, 2000). Contextual Characteristics There is a paucity of evidence concerning how the contextual characteristics, race, marital status, level of education, employment and health insurance data impact CRF in the general and lung cancer populations (Montazeri, Gillis, & McEwen, 1998). The burden associated with LC requires great amounts of support through contextual characteristics to sustain the patient and his/her family (Sarna & McCorkle, 1996). These need further investigation. Symptoms The management of unpleasant symptoms is a foremost priority for those suffering from LC. According to Gift et al., in a sample of 220 persons newly diagnosed with LC, a mean of 11 symptoms was reported per person (2004). These multiple distressing symptoms were related to severity and limitations. Additionally, B. Given et al. report among persons with cancer who were 65 years and older, those with LC reported a greater number of symptoms than those with other tumors of solid composition (2001). Symptoms place an additional burden on the lives of persons with LC (Cleeland & Reyes-Gibby, 2002). Cooley et al. found in a secondary longitudinal analysis that while many symptoms improve over time, some persons recently diagnosed with LC experience increasing levels of distress from pain frequency and fatigue (2003). Moreover, other studies report for persons with LC, the level of symptom severity and 13 distress rises until death (Degner & Sloan, 1995; Sarna, 1993a, , 1998). This increased symptom severity negatively impacts a person’s PFS. However, few studies have been conducted focusing on the relationship between symptoms and functional status for persons with LC. Symptoms are the “perceived red flags of threats to health” (Hegyvary, 1993). Symptoms are the experience of the person The TOUS highlights four dimensions that characterize the person’s symptom experience: timing (frequency and/or rate of occurrence; duration), severity (intensity), quality (description of qualifiers), and distress (bother). This study focuses on two dimensions of the person’s symptom experience: frequency or rate of occurrence and severity. Furthermore, the TOUS addresses not only one symptom, but multiple symptoms and their interactions. This is important to this study since during the course of their ilhtess trajectory, persons with cancer, particularly LC, not only present with CRF, but with other concurrent symptoms. The concurrence of symptoms are likely to catalyze each other worsening the overall level of symptom severity (Cleeland & Reyes-Gibby, 2002; Lenz et al., 1997). Therefore, the determination of the total symptom severity is critical to fully understand the symptom experience of the person with LC with fatigue and other concurrent symptoms. Given that persons with LC experience greater numbers of symptoms in comparison to other cancer (0C) diagnoses (B. Given et al., 2001), this study compared and described the other unpleasant symptoms associated with CRF. This is a vital step in portraying the true fatigue symptom experience of the person with cancer, particularly LC, which will serve as a foundation to design future interventional studies to enhance PSE and increase PFS. 14 Perceived Selfi-Eflicacy Although not studied in the LC population, a contributing factor to the achievement of symptom management of CRF to attain maximum PFS may be a person’s PSE. There is accumulating evidence that PSE expectations exert strong influence on behavior (Bandura, 1989). Self-efficacy contributes to health behavior change in chronic illness and is a key predictor of health promotion (Bandura, 1995, , 1997). Investigations substantiate that PSE plays a central role in producing positive health outcomes in symptom management and functional ability in people living with chronic conditions such as arthritis (Barlow, Turner, & Wright, 2000; Lorig, Ritter, & Plant, 2005; Lorig et al., 2001; Pariser et al., 2005), cardiac disease (Bamason et al., 2003; Cheng & Boey, 2002; Gardner et al., 2003; l-Iiltunen et al., 2005), chronic pain (Federman et al., 2002), COPD (K Wong et al., 2005), diabetes (Howells et al., 2002), fibromyalgia (King et al., 2002), and multiple sclerosis (Mathiowetz et al., 2001; Wassem & Dudley, 2003). Attention to PSE is important in enhancing a positive perception of control over the challenging situations faced by persons with cancer (Merluzzi & Martinez Sanchez, 1997; Telch & Telch, 1985, , 1986). Very few studies have been conducted relative to PSE and symptom management in the general cancer population and only one small study has been done concentrating solely on PSE in persons with LC (Porter et al., 2002). Using appropriate mesh terms, keywords and thesaurus terms via PubMed, MEDLINE, CINAHL, and PsycINFO (i.e., fatigue, cancer-relawd fatigue, symptoms, self-efficacy, and cancer), no studies have been conducted on the role PSE plays in the management of fatigue and other associated unpleasant symptoms to achieve optimal functional status for the LC population. 15 The few studies that have been conducted indicate that increased PSE has a positive impact on the lives of persons with cancer. For instance, Cunningham, Lockwood, and Cunningham reported in a convenience sample of 273 cancer patients (including 17 persons with LC) that an interventiOn involving implementation of coping skills led to a strong positive relationship between PSE, quality of life and mood state (1991). Likewise, Beckham, Burker, Lytle, Feldrnan, and Costakis found in a cross-sectional study of 42 male cancer patients (including 3 persons with LC) that PSE expectations relative to cancer symptoms accounted for a large proportion of the variance measures of cancer adjustment, psychological distress, positive and negative affect, and behavioral dysfunction (1997). Also, Lev, Paul, and Owen further substantiated these results through a longitudinal intervention study with a convenience sample of 307 general cancer patients (including 38 persons with LC) at baseline, 181 persons four months later, and 124 persons eight months later. In this study, patients reported that PSE affected their adjustment to cancer and without intervention their PSE and adjustment decreased over time (1999). In a later study, Lev et al. found in a randomized clinical trial of 56 women with breast cancer that PSE increased quality of life and decreased symptom distress (2001). Similar findings were found in a randomized clinical trial with 189 women with late stage breast cancer. Northouse et al. (2002) reported that PSE had positive effects on patients’ and family members’ quality of life. In a cross-sectional study with 63 men undergoing treatment for prostate cancer, Lev et al. (2004) found that psychosocial variables and physical symptoms were related to indicators of quality of life. Lev et al. (2004) advised that these data support Bandura’s (1997) assertion that psychosocial factors may determine quality of life, and using efficacy enhancing l6 interventions in persons with cancer may reduce a person’s perception of stress and reported symptoms and increase a positive perception of quality of life. Weber et al. (2004) reported similar results through a randomized clinical trial of 30 men undergoing radical prostatectomy for prostate cancer. Here the treatment group receiving a support intervention had increased PSE, decreased depression, and greater improvement in physical firnctioning in comparison to the control group receiving usual care. Parallel to these findings, in a cross-sectional study with 85 advanced cancer patients (including 18 persons with LC), Hirai et al. (2002) reported that patients in good physical condition had high PSE, and patients with high PSE were less emotionally distressed. Thereby, the researchers stated that the findings imply that psychological interventions which emphasize PSE would be effective for advanced cancer patients. Most recently in 2006, two studies conducted over time report on the positive effects of PSE on patient outcomes. The first study conducted by Eller et a1. evaluated the impact of PSE over time on the dimensions of quality of life in 159 men who were undergoing various treatments for prostate cancer (2006). Eller et al. reported that one subscale of PSE (positive attitude) was a significant predictor of two components of quality of life, social and functional well-being. The second study followed 95 women undergoing treatment for early stage breast cancer for one year. In this study, Manne et al. reported that cancer- related self-efficacy for activity management and self-satisfaction increased and remained relatively stable over the one year that participants were followed (2006). To conclude, the only study pertaining to PSE in persons with LC, Porter et al. found that for a small convenience sample of 30 subjects, those who rated their PSE as higher had lower levels of pain and other symptom severity (2002). Although some of these 17 studies include small sample sizes, have methodological problems, and are mainly focused on the general, breast, or prostate cancer population, they provide the evidence to support further investigation and the development of PSE strategies to manage CRF and optimize the PFS of persons with LC. Bandura posits that self-efficacy expectations are domain specific and are among the most effective mediators since they influence the initial decision to perform a behavior, the effort expended by the behavior, and the persistence of the behavior in the face of adversity (1997). Bandura distinguished outcome expectations, one’s perception that a behavior could produce a particular outcome, from self-efficacy expectations, the perception of one’s ability to execute this behavior successfully (1997). It is important to note that outcomes arise from actions, and the outcomes people anticipate depend largely on their perceptions of how well they will be able to perform in given situations (Bandura, 1997). In measuring PSE, people are presented with items describing different task demands and they rate the strength of their perception in their ability to execute behaviors. Items about PSE are phrased in terms of “can do” (perception of ability) rather than “will do” (perception of intention) (Bandura, 1997). Parallel to Bandura’s definition, PSE is the person’s perception of ability to execute behavior(s) to manage his or her CRF (1986). This involves an evaluative process of the meaning of CRF and other associated unpleasant symptom(s); CRF’s significance to a person’s well-being; as well as the person’s self-appraisal of his or her own PSE to manage fatigue. It is important to note that PSE is concerned with the judgments of what one can do with whatever knowledge and skills one possesses (Bandura, 1986). Perceived self-efficacy beliefs are developed and altered not only by direct mastery experiences, but 18 also by seeing people similar to oneself manage task demands successfully, social persuasion that one has the capabilities to succeed in given activities, and inferences from physiological and emotional states indicative of personal strengths and vulnerabilities (Bandura, 1997). Consequently, Self-Efficacy Theory is advantageous for use in research because influencing the development of sources of PSE provides direction for effective alteration of behavior (Lev, 1997). According to Bandura, those persons with high levels of PSE are able to exert control over threats (1997) such as CRF and other unpleasant symptoms and strive to improve their PFS. Thus, it is both the fatigue and the person’s PSE related to management of the fatigue that determines the degree to which the person is vulnerable or empowered to meet the demands of the fatigue and other unpleasant symptom(s) derived from LC and its treatrhent. As previously described, there is accumulating empirical evidence of the effectiveness of PSE positively impacting the management of symptom(s) and in turn the functional abilities of those suffering from various chronic conditions. However, using appropriate mesh terms, keywords and thesaurus terms via PubMed, MEDLINE, CINAHL, and PsycINFO (i.e., fatigue, cancer-related fatigue, symptoms, self-efficacy, and cancer), there is little published research in the area of PSE and its impact on symptom management for persons with life threatening, chronic illnesses such as cancer. Functional Status Although limited research exists describing the relationship between fatigue and the functional status for persons within the general cancer population (including persons with LC), studies have found CRF to adversely impact the functional status of older and younger persons. A study comprised of 826 persons 65 years of age and older with a new 19 diagnosis of cancer revealed that fatigue was 1 of 3 independent predictors of their physical firnctioning (B. Given et al., 2001). C. Given, Given, Azzouz, Stommel, and Kozachik (2000) reported fiom a study consisting of 907 patients age 65 or older with a new diagnosis of breast, colon, prostate, or lung cancer that patients who reported neither pain nor fatigue scored 15 points higher in physical functioning than those with pain or fatigue and 25 to 30 points higher than those with both symptoms. Similarly, in a study consisting of 93 persons with cancer 18 years of age or older receiving chemotherapy, both fatigue and pain were found to be the largest contributors to change in functional status ofpersons with cancer (Dodd et al., 2001). Last, a study reported that fatigue was the most prevalent and severe symptom among 47 persons with cancer who were in the final month of their lives resulting in great fimctional decline, especially in the areas of physical and role functioning (Sahlberg-Blom, Ternestedt, & J ohansson, 2001). Also, for persons with LC specifically, fatigue has been demonstrated to negatively affect their fimctional status when undergoing different treatment modalities. Handy et al. compared fimctional status in 139 persons with LC undergoing surgery with age-matched healthy patients. The results indicated that persons with LC had significantly lower scores in their level of energy and physical and role-emotional functioning with subsequent deterioration in functional status six months post-operatively (2002). Likewise, Langendijk et al. reported that 164 persons receiving radical radiotherapy reported fatigue as the most prevalent and distressing symptom and it increased in intensity over time during treatment (2001). This study also reported deterioration in physical and role functioning over time during radiotherapy and in the 6 to 12 months period after treatment ended. 20 Furthermore, fatigue has been shown to interfere with functioning activities for those with advanced-stage LC. Sarna and Brecht reported that among 60 women with advanced stage LC, fatigue was the most frequent distressing symptom with severe fatigue lowering physical firnctioning scores (1997). In another study by Sarna it was found that 79% of 24 persons with advanced staged LC experienced serious fatigue with 44% having subsequent difficulty with household chores, 52% losing interest in recreational activities, and 61% changing their recreational activities due to CRF (1993b). Additionally, Tanaka, Tatsuo, Okuyama, Nishiwaki, and Uchitomi reported that 50% of 171 persons with advanced stage LC reporting a low severity level of fatigue stated that this level of fatigue interfered with at least one daily activity (2002). Okuyama et al. similarly found that roughly half of their 157 persons with advanced LC reported fatigue interfered with at least one daily life activity (2001). In this group nearly one-third had interference with physical activities, and one-fifth reported CRF impacted emotiorml activities such as enjoyment with life and mood ' In a later study, Brown, McMillan, and Milroy (2005) compared functional status in 38 persons with metastatic or locally advanced LC with age and gender-matched persons. Brown et al. reported that persons with LC as compared to the control group had greater levels of fatigue and lower firnctional performance, poorer grip strength, and longer chair-rise time. In summary, the prevalence and distress fi'om fatigue is high in persons with LC. Cancer-related fatigue alone or in conjunction with other symptoms heightens the total symptom impact adversely affecting the fimctional status of this population Hence, management of fiitigue and other associated symptoms of LC is a major element to optimal functioning. 21 Unpleasant symptoms such as CRF make it difficult for persons with cancer, particularly LC to maintain maximal functional status. The TOUS delineates the performance outcome component of the theoretical fiamework for this study (Lenz et al., 1997). According to the TOUS, performance outcomes are the consequences of symptom management which includes the PFS of the person with LC. Thus, the theoretical framework posits that multiple patient characteristics within the environment interact with each other and the symptom(s) allowing or restricting functional performance. In the face of the demands of CRF and other unpleasant symptoms, higher levels of functioning can be achieved by enhancing the PSE of the person with LC. Thus, PSE is an empowering mediator that enables the person with LC to achieve control over CRF and other unpleasant symptoms and enhance optimal functional performance. Summary Persons with LC have been found to experience more symptoms than those with DC diagnoses. Cancer-related fatigue is problematic and impacts the functional status of persons with LC. A contributing factor to symptom management of CRF is a person’s PSE. Limited research has been conducted regarding the impact of CRF on symptom management to improve the frmctional status of persons with cancer, particularly LC. This research seeks to measure a person’s fatigue self-efficacy and whether PSE relates to their ability to manage CRF and how this relates to their PFS. This study is important since it will lead to interventional research to improve functional status of persons with cancer, particularly LC when faced with fatigue. 22 CHAPTER THREE METHODOLOGY Design This study is a cross-sectional descriptive design that employs baseline measures obtained from patients who were undergoing a course of chemotherapy from two larger randomized control trials currently underway: “The Family Home Care for Cancer: A Community-based Model for Symptom Management” (FHCC) project (R01 CA-0792 80) sponsored by Barbara A. Given, Ph.D., R.N., FAAN, Principal Investigator, and “The Automated Telephone Monitoring for Symptom Management” (ATSM) project (R01 CA-30724) sponsored by Charles W. Given, Ph.D., Principal Investigator (see Executive Summaries in Appendix A). The baseline measures were completed prior to the beginning of the intervention This current study focuses on the baseline data and distinguishes itself from the larger study in that it focuses on the role perceived self- efficacy (PSE) plays in fatigue management and other unpleasant symptoms to achieve optimal physical functional status (PFS). Moreover, while the analyses of this study includes persons with other cancer (OC) diagnoses, the primary focus remains on persons with lung cancer (LC), a population that, when compared to other types of cancer diagnoses, has a greater number of symptoms (Doorenbos et al., 2006b; B. Given et al., 2001). 23 Purpose and Research Questions The foremost purpose of this descriptive study was to examine fatigue and PSE in persons with LC and analyze how PSE in managing fatigue impacts PFS. Utilizing persons with DC diagnoses as a comparison group to persons with LC as well as to increase sample size assisted in answering the following research questions for the study: 1. What are the patient characteristics that relate to CRF severity in persons with cancer (LC and OC diagnoses)? How does having a cancer diagnosis of LC compare with OC diagnoses as a predictor of CRF severity? What is the symptom experience of persons with LC and how does it compare to the symptom experience of OC diagnoses, including the relationships between CRF and other unpleasant symptoms? Does PSE for fatigue management mediate the relationship between CRF ' severity and PFS in persons with cancer (LC and 0C)? Does mediation differ in persons with LC as compared to OC when evaluating PSE for fatigue management as a mediator between the relationship of CRF severity and PFS? What is the unique contribution of physiological and contextual characteristics, CRF, other unpleasant symptoms, and PSE to the PFS of persons with cancer (LC and OC)? Considering the unique contribution of physiological and contextual characteristics, CRF, other unpleasant symptoms, and PSE for fatigue 24 management to PFS, how does having a cancer diagnosis of LC compare with OC diagnoses as a predictor of PFS? 8. Through the employment of a Path Model, is the PFS of persons with cancer (LC and OC) predicted through physiological and contextual patient characteristics, CRF, other unpleasant symptoms, and PSE for fatigue management? Sample The target sample for the study came from the baseline measures prior to the intervention of both the FHCC and ATSM studies (see Appendix A). Sample size included 63 persons with LC with a comparison group of 235 persons with DC diagnoses, which is composed of persons with breast cancer (n = 105); persons with colon cancer (n = 44); and persons with other sites of cancer (n = 86). As a result, the total sample size consists 298 persons with cancer, persons with LC (N = 63) and OC diagnoses (N = 235). Study participants were recruited who were cognitively intact, speak English, able to hear and speak for telephone interviews, have new or recurrent disease, not receiving hospice care, receiving chemotherapy for breast, colorectal, or lung cancer, other solid tumors, and non-Hodgkin’s lymphoma with at least two cycles remaining at time of enrollment, and may have been receiving concurrent radiation therapy. Thus, the baseline data for this study produced diversity in age, stage of cancer, and symptom severity. Exclusion criteria included those who were diagnosed with any hematological malignancy or whose treatment involved bone marrow transplant or stem cell rescue and for persons diagnosed with an emotional or psychological disorder for which they were currently under the care of a professional. 25 Setting The data for the FHCC and ATSM studies were collected from seven different collaborating cancer sites (see Appendix B). Operational Definitions Operational definitions of the study’s key variables are introduced below and are organized according to the theoretical fiamework guiding this study. The operationalization of the key variables is further discussed in the section that outlines the instruments used for this study. Physiological Patient Characteristics Physiological patient characteristics include type and stage of cancer; treatment information (radiation therapy and surgery); co-morbid conditions; sex; age; and laboratory values (hemoglobin and absolute neutrophil count). Type of cancer includes both LC (small cell or non-small cell) and OC (breast, colon, and other types) diagnoses. Stage of cancer is classified by the TNM system and a two-stage system. Treatment information included whether or not a patient was receiving radiation Treatment information also included two variables regarding surgery: surgery prior and surgery during. Surgery Prior means that the surgery occurred prior to chemotherapy and prior to consent. Surgery During means that the surgery occurred during chemotherapy and the audit period which is from consent to last interview. Contextual Patient Characteristics Contextual patient characteristics include race; marital status; level of education; employment data; and health insurance data. Employment data included whether a person was retired, receiving disability, was on a temporary leave from employment, whether the 26 person had to quit employment, and annual combined household income. Health insurance data included whether or not a subject had health insurance and if so, who held the policy (patient or spouse), and the type of health insurance policy held (private, Medicare, or Medicaid). Symptoms Symptoms are the “perceived red flags of threats to health,” (Hegyvary, 1993). Cancer-related fatigue and 15 other unpleasant symptoms were included in this study. The total CRF severity and total symptom severity from the other unpleasant symptoms were analyzed The total CRF severity score includes two items fi'om the Brief Fatigue Inventory, the patient’s current (now) and worst severity of CRF within the past 7 days. The total CRF severity score was calculated by summing each subject’s response to severity scores for both of the two CRF severity items and dividing by two (the two items assessing CRF severity) to standardize the score on an 11-point scale. The total symptom severity score from 15 other unpleasant symptoms (excludes CRF) associated with cancer and cancer treatment was calculated by summing each subject’s response to severity scores for each symptom reported (i.e., a reported symptom is a symptom with a severity score greater than zero) and dividing by the total number of symptoms reported to standardize the score on an 11-point scale. Perceived Self-Eflicacy for Fatigue Management Perceived self-efficacy for fatigue management is the person’s perception of ability to execute behavior(s) to manage his or her CRF. The Self-Efficacy for Fatigue in Patients 27 with Cancer Scale measured how certain persons with cancer were in performing specified behaviors/goals in managing fatigue. Performance Outcomes Performance outcomes are the Consequences of symptom management which includes the PFS of the persons with cancer which was measured via the Medical Outcomes Study Short Form-36 PFS subscale. Instruments Data for the study were collected in conj motion with the FHCC and ATSM studies at baseline via telephone interviews and medical record chart abstraction The telephone interview for the study took approximately 15 minutes to complete. The following instruments include measures of patient characteristics; CRF, other unpleasant symptoms, PSE, and PFS (see Summary of Measures for Research, Appendix C). Patient Characteristics (Physiological and Contextual) Demographic Questionnaire and Medical Record Chart Abstraction Patient characteristics were measured at baseline prior to the intervention using selected questions from a demographic questionnaire and medical record chart abstraction form developed for the FHCC and ATSM studies (see Appendix C). Most patient characteristics were self-report items. A demographic questionnaire was used for co-morbid conditions, sex, age, race, marital status, level of education, and employment data. Information on co-morbid conditions regards 15 chronic health conditions. Stage and type of cancer (LC or OC diagnoses), treatment information, hemoglobin, absolute neutrophil count, and health insurance data were obtained from the person’s chart via a chart abstraction form. Information on the stage of cancer was classified according to the 28 TNM staging system of the American Joint Committee on Cancer for non-small cell LC andOC diagnosesthat stagesthe canceronascale of0to IV. For small cell LC, atwo- staged system was used: limited or early stage and extensive or late stage. Symptoms (CRF and Other Unpleasant Symptoms) The Brief Fatigue Inventory (BFI) The BFI measures the severity from CRF (3 items), and the amount that CRF has interfered with aspects of the patient’s life (6 items) (Mendoza et al., 1999). The investigators designed the BFI based upon the Brief Pain Inventory that has demonstrated successful assessment (successfully utilized in telephone interviews) of the severity of cancer pain in the United States and other countries (Cleeland et al., 1989; G. Wong et al., 2004). Items on the BF I were derived from the data in the Wisconsin Fatigue Study that utilized normal volunteers, psychiatric patients receiving treatment for depression, and cancer patients. Appropriate psychometric properties were found using data on adult patients (N = 305) from the University of Texas MD. Anderson Cancer Center consisting of inpatients and outpatients with varying types of cancer (including persons with LC), and control subjects (N = 290) from the Houston area (Mendoza et al., 1999). The BFI achieved a high internal consistency level of 0.96. Cronbach’s coefficient alpha for each of the items (if deleted) was 0.95 for both interference in general activity and mood and 0.96 for the remaining items. Concurrent validity was established with two previously validated measures which are used for the assessment of fatigue, Profile of Mood States Fatigue Subscale (r = 0.84, p < 0.001) and the Functional Assessment of Cancer Therapy Fatigue Subscale (r = -0.88, p < 0.001) (I-Iwang, Chang, Cogswell, & Kasimis, 2002). 29 Furthermore, the BFI is sensitive in detecting severe from non-severe CRF at a cut score of 7 with a range of 7-10. Stability of the BFI to detect severity of CRF has been demonstrated (Hwang et al., 2002). For the study, two items from the BFI measuring severity of CRF were used to calculate a total CRF severity score (see Appendix D). On an 11-point scale (0-10), these two items evaluated the patient’s current (now) and worst severity of CRF within the past 7 days. The total CRF severity score was calculated by summing each subject’s response to severity scores for both of the two CRF severity items and dividing by two (the two items assessing CRF severity) to standardize the score on an 11-point scale. For this study, internal consistency reliability of the total CRF severity revealed a Cronbach’s alpha of 0.85 for the total sample, 0.81 for persons with LC, and 0.86 for persons with DC diagnoses. The symptom Experience Inventory Frequency and severity of the other unpleasant symptoms were assessed using the Symptom Experience Inventory which was developed and used in previous studies (successfully utilized in telephone interviews) by Dr. Barbara Given (Gift et al., 2004; B Given et al., 2002; C. Given et al., 2001) (see Appendix E). The inventory is a self-report measure containing 16 symptoms related to cancer and its treatment (i.e., pain, dyspnea, insomnia, nausea, difficulty remembering things, lack of appetite, dry mouth, vomiting, numbness or tingling, diarrhea, fever, cough, constipation, weakness, alopecia, fatigue). Frequency was evaluated by asking the patient to indicate the number of days in the past week that they experienced the symptoms. On an 11-point scale (0—10), patients were asked to rate their current severity of their symptoms. Frequency of the symptoms and 30 both the individual symptom severity and total symptom severity scores were calculated for persons with LC and OC diagnoses. Individual symptom severity scoring was rated from a 0 to 10, with 0 being “symptom not present” and 10 being “worst it can be”. The total symptom severity associated With each symptom was calculated by summing each subject’s response to severity scores for each symptom reported (i.e., a reported symptom is a symptom with a severity score greater than zero) and dividing by the total number of symptoms reported to standardize the score on an 11-point scale. The total symptom severity score did not include the symptom of fatigue. For this study, evaluation of the internal consistency reliability resulted in a Cronbach’s alpha for the total sample of 0.72, for persons with LC 0.77, and for persons with 0C 0.69. Perceived Self-Efiicacy for Fatigue Management Self-Eflicacy for Fatigue in Patients with Cancer Scale (SEFPCS) Review of the literature revealed that there was no existing tool to measure PSE for fatigue management in persons with cancer. In this study, PSE for fatigue management was measured using a six-item subscale adapted by the author from the Lorig Arthritis Self-Efficacy Scale (ASE). Lorig created the ASE to measure persons’ PSE to cope with the consequences of chronic arthritis. The ASE has a 3-factor solution accounting for 61% of the variance in the dimensions of PSE. Internal coefficient alphas for each of the subscales (N = 143) are 0.76 for pain mamgement, 0.89 for physical functioning, and 0.87 for coping with other symptoms (Lorig, Chastain, Ung, Shoor, & Holman, 1989). Of interest is the Coping with Other Symptoms Lorig subscale, which has been minimally altered for the cancer population by replacing the word fatigue for other identified symptoms to create the SEFPCS (see Appendix F). In a similar manner to the 31 SEFPCS, the Lorig subscale has been successfully adapted in the past for three chronically ill populations which includes patient populations with chronic pain, cancer pain and HIV disease (Anderson, Dowds, Pelletz, Edwards, & Peeters-Asdourian, 1995; Keefe et al., 2003; Shively, Smith, Bonnann, & Gifford, 2002), and successfully used to collect data via telephone interview (Keefe et al., 2003). These research studies with different chronically ill populations have demonstrated improved internal Cronbach’s alpha with adaptation of the Lori g Coping with Other Symptoms subscale as compared to the internal Cronbach’s alpha reported when used in the arthritis population Note that the degree of manipulation performed in all three past studies is greater than what was done for the SEFPCS. The SEFPCS is a self-report measure containing 6 items related to PSE for fatigue management. For this study, on an ll-point scale (0-10) with 0 being “very uncertain” and 10 “very certain”, persons with cancer were asked to rate how certain they were in performing specified behaviors/ goals in managing fatigue. The SEFPCS score was calculated by summing the responses for each item and dividing that sum by six, the number of items in the SEFPCS. Scores ranged from 0-10, with higher scores indicating greater PSE for fatigue management Content validity for the SEFPCS was reviewed by three nurse experts experienced in fatigue management of persons with cancer, and revisions were made prior to the SEFPCS use in the study (Nunnally & Berstein, 1994). For this study, evaluation of the internal consistency reliability resulted in a Cronbach’s alpha for the total sample of 0.92, for persons with LC 0.91, and persons with OC diagnoses 0.92. 32 Performance Outcomes (Functional Status) The Medical Outcomes Study Short F arm-3 6 (SF -3 6) The SF-36 is one of the most comprehensive, generic, multidimensional health- related quality of life measures for adults with chronic conditions which assesses various components of functional status (see Appendix G) (Maciejewski, 1997; Murdaugh, 1997; Ware, Snow, Kosinski, & Gandek, 1993 ). Eight subscales are contained within the instrument including physical functional status (10 items), role-physical functional status (4 items), bodily pain (2 items), general health perceptions (5 items), vitality (4 items), social functional status (2 items), role-emotional functional status (3 items), and mental health (5 items). Principle component analysis demonstrates that 80-85% of the variance in the eight subscales was accounted for by two factors, physical and mental health It is standardized, validated, (McHomey, Ware, & Raczek, 1993) and shows internal consistency of the subscales ranging fi'om .78 to .92 (Ware et al., 1993). Since symptoms are a major detemrinant of functional status, the physical functional status subscale fiom the SF-36 was used The SF-36 Health Survey Manual and Interpretation Guide was used for scoring (Ware et al., 1993). The subscale scores of the SF-36 are linearly transformed to range from 0 to 100, with higher scores representing better levels of functional status (Ware et al., 1993). For this study, evaluation of the internal consistency reliability resulted in a Cronbach’s alpha for the total sample of 0.91, for persons with LC 0.91, and persons with OC diagnoses 0.91. 33 Procedures Recruitment and Data Collection Data were obtained for the study fiom the baseline measures of the FHCC and ATSM studies. The baseline measures were collected prior to the beginning of the intervention. Recruiters for the FHCC and ATSM studies were employees of the participating sites which facilitated their ability to access patient records and information without breach of confidentiality. Recruiters were trained at Michigan State University (MSU) following a procedure developed by the FHCC and ATSM projects. The eligibility (i.e., see inclusion and exclusion criteria, p. 24) of the patient was determined at the participating sites. If eligible to participate in the study, the recruiters met with the patients to describe the studies, discuss roles and expectations, explain patients’ rights, obtain written consent, and enroll patients. Patient contact information was transmitted via a secure WEB-based server at the central site. Once the patient was enrolled, the patient’s symptoms were screened via completing twice weekly automated telephone calls, for up to six weeks. For the screening process of the FHCC study, symptoms were assessed until a symptom severity threshold of 2 out of 10 for both pain and fatigue was reached or a symptom severity threshold of 3 out of 10 on either pain or fatigue was reached For the screening process of the ATSM study, symptoms were assessed until a symptom severity threshold of a 2 or higher out of 10 for one or more symptoms was reached If after six weeks patients did not reach a symptom severity threshold level as designated by the FHCC or ATSM studies, they were sent a letter thanking them for participating. Patients reaching the symptom severity threshold were contacted by telephone by an interviewer to complete the baseline interview. The 34 interviewer received the patient’s name and their telephone number from the MSU Project Coordinator once patient’s fulfilled the eligibility, consent and enrollment process, and symptom assessment screening. The interviewer staff was trained and followed procedures developed by the research staff of the FHCC and ATSM studies. Interviews occurred based upon the convenience of the patient Instruments for the proposed project were part of the FHCC and ATSM study consent form and part of the interview materials and were approved by the MSU IRB. Once the baseline interviews were completed, patients were randomly assigned to either FHCC or ATSM study. Trained by the FHCC and ATSM research staff, auditors completed medical record chart abstraction. Inclusion of Women and Men A person’s sex was included in both the theoretical model and statistical models of the study. Eligible male and female adults age 21 years and older were enrolled in the study. Inclusion of both women and men were necessary because LC is the leading cause of cancer death in both women and men regardless of race (American Cancer Society, 2007). The LC incidence rate is declining significantly in men fi'om a high of 102.1 per 100,000 in 1984 to 77.8 in 2002 (American Cancer Society, 2006). While there has been a dramatic increase in the number of LC cases in women, it has leveled off at 52.8 per 100,000 in 1998 and continues to remain stable (American Cancer Society, 2004, , 2006). Inclusion of Minorities Race was included in both the theoretical and statistical models of the study. Non- Caucasian participants were differentiated in the study by the following groups: Hispanic or Iatino, Ammcan Indian or Alaska Native, Asian, Native Hawaiian or other Pacific 35 Islander, and Black or African American In past studies conducted by the Principal Investigators Drs. Given, accrual of minority subjects ranged between 8-10%. With the exception of the Detroit metropolitan area as a part of Wayne State University’s Karmanos Cancer Center which has a higher percentage of minority patients, Community Clinical Oncology Programs in the state of Michigan have approximately this same percentage of minority participation (B. Given, 1998-2007; C. Given, 2003-2007). Cooley reported in her systematic review of the empirical research examining symptoms in adults with LC the majority of studies were conducted with Caucasians and recommends special recruitment strategies to increase the likelihood of minority participation (2000). Nevertheless, recruitment efforts for minorities included strategies outlined in the literature as helpful to minority recruitment (D. R Brown, Fouad, Basen- Engquist, & Tortolero-Luna, 2000; Holcombe, Jacobson, Li, & Moinpour, 1999; Pinto, McCaskill-Stevens, Wolfe, & Marcus, 2000). Examples of efforts to augment minority participation included having recruiters at each site attend a training session consisting of a two-hour didactic session on how to recnrit minorities through role-playing opportunities and communication skills. Booster recruiter training sessions were planned annually with more scheduled when necessary. In addition, ethnically sensitive brochures about the benefits of participating in clinical trials were developed at a Flesch-Kincaid fifth grade level with a reading ease score above 70 on a 100 point scale with higher scores indicating greater ease of understanding. The pictures on the brochures and other participant information included a racial mix of people. 36 Inclusion of Children Children under the age of 21 years were excluded from the study. It is extremely atypical for an individual under age 21 to suffer from LC. The sample would not contain enough participants of this age group to identify significant relationships. Instead, the majority of persons with LC were individuals in their sixth decade or older (National Cancer Institute, 2003). The focus of the research was a description of CRF and other unpleasant symptoms in adults with LC and how their PSE to manage their fatigue affected PFS. Data and Safety Monitoring Plan The Principal Investigator for this study received permission fi'om the Principal Investigators of the FHCC and ATSM projects for the data for analysis of the study. To protect confidentiality, the data were placed on a computer disc without any patient names or health care system identifiers. The FHHC and ATSM Principal Investigators hold the master list of patient identifying information and the Principal Investigator for thisstudyhasnoaccesstothismasterlistAcodebookdefiningeachvariableandthe range of permitted responses was obtained from the parent studies. The original copy of data from the disc was transferred to statistical analysis software and examined to ensure the prior editing of the variables were not missing, were within the permitted range, and were logically consistent with other variables. The codebook defined the basis for editing the data. Reliability analysis was performed to evaluate consistency of measures used in the target sample for the study as described for each instrument used in the study. The database will be completely edited and frozen before any final analyses for publication 37 begin. The disc containing the data for analysis for this study is stored in a secure area in a locked file cabinet. Protection of Human Subjects This study used a cross-sectional descriptive design that employed baseline measures taken from patients who were undergoing a course of chemotherapy prior to the beginning of the intervention from two randomized control trials: “The Family Home Care for Cancer: A Community-based Model for Symptom Management” (FHCC) project (R01 CA-079280) sponsored by Barbara A Given, Ph. D., RN, FAAN, Principal Investigator and “The Automated Telephone Monitoring for Symptom Management” (ATSM) project (R01 CA-30724) sponsored by Charles W. Given, Ph.D., Principd Investigator (see Executive Summaries in Appendix A). The FHCC and ATSM studies had Institutional Review Board (IRB) approval through the University Committee on Research Involving Human Subjects (UCHRIS) of MSU and collaborating sites, including approval for use of the measures being used in this study. Approval to conduct this study was obtained from MSU UCHRIS (February 24, 2006, July 26, 2006, and January 2007). The Principal Investigator for this study has adhered to all mechanisms for the protection of human subjects. Additionally, the Principal Investigator for this study will maintain the human subject certification administered by [RB at MSU throughout the study, analysis and publication of the data. The parent studies started data collection in December 2003 and completed in April 2006. The FHCC and ATSM studies obtained consent, enrolled patients, and collected data for this study. Recruiters, trained by the FHCC and ATSM studies, determined eligibility at the collaborating sites (see inclusion and exclusion criteria). The recruiters 38 met with the patients to provide a description of the study, discussed roles and expectations, explained patients’ rights, obtained written consent, and enrolled patients. Patients consented to this study as part of the parent studies consent Data were collected by interviewer staff who were trained by the research staff of the FHCC and ATSM studies. Trained auditors completed medical record chart abstraction Sources of Material Sources of research material obtained from individually identifiable living human subjects were acquired from self-report telephone questionnaires and medical record abstraction form. The data were obtained specifically for research purposes only. The sample for the study comes from the National Institutes of Health sponsored study, entitled FHCC (R01 CA-079280) and ATSM (R01 CA-30724) studies. The applicant obtained permission from Drs. Given to use the data. Institutional Review Board approval has been obtained Subjects did not incur any expense as a result of their participation in the study. Data were obtained using measures described earlier in the Methodology section which include: 1) Patient Characteristic Questions from the Demographic Questionnaire and Medical Record Chart Abstraction Form, 2) Items from the Brief Fatigue Inventory, 3) the Symptom Experience Inventory, 4) Self-Efficacy for Fatigue in Patients with Cancer Scale, and 5) the Medical Outcomes Study Short Form-36 PFS subscale. Additional information about the measures can be found in the appendices. The data for the FHCC and ATSM studies were collected from seven different collaborating sites which provide cancer care (see Appendix B). 39 Potential Risks and Protection Against Risks Since the study used a cross-sectional descriptive design not involving investigational intervention, patients were placed at very minimal risk by their participation in the study. Patients may have become fatigued while answering questions on the telephone for the baseline interview. The total respondent burden for instruments from the study took approximately 15 minutes. Great effort was made to limit the length of the interview. According to past research studies conducted by Drs. Given, it was found that older patients with cancer, even with persons who are quite ill, experience low respondent interview fatigue or burden and have little attrition However, if patients experienced fatigue or were too ill to complete the questionnaires, they could have withdrawn fi'om the study at any time. In the event a decision was made to skip and/or reschedule the interview, the Project Manager and Principal Investigators were notified Another potential risk may have involved patients feeling apprehensive in sharing personal information The interviewer was trained to introduce the patient to the interview questions, to recognize the patient’s feelings, to refer patients to their physician if needed, and to be adaptable and gracious to patients regardless of their responses. Breach of confidentiality was also a potential risk for patients. However, strict provisions to prevent this were in place. The patient was told by the recruiter and the interviewer that all demographic and health information conveyed in the interview was entirely confidential. None of the information provided to the phone interviewer would be shared with the nurse or their physicians. The research information would not be linked with them personally by name or other means that might identify them. However, all patients were told that clinical information obtained as part of the symptom 40 management content would be shared if this placed them at risk. Via an established protocol, this information would be shared with their oncologists to ensure optimal coordinated care. The patient was afforded the opportunity to ask questions about this process to ensure they understood the importance of this to their overall plan of care. The confidentiality of patients was safeguarded in the following ways: 1) by use of subject identification numbers, 2) by release of research data in aggregate form only, 3) by omission of agency names and/or identification in all presentations and reports, and 4) by not providing confidential interview research data given by subjects back to the agencies or participating stafl' members. Paper copies of all consent forms were transmitted via secure courier. These forms were retained in a locked file in the central MSU site. All computers that were used were double password protected A password would be required to open Windows 2000 and a second, different password would be required to log on to the network. Interview data were stored on a secure web server and were password protected Servers were scanned for viruses and systems were in place to detect attempts at unauthorized entry. Additional server logs record all persons accessing data files. Safety monitoring procedures were in place for interviews. All interviewers were carefully trained in correct interviewing procedures; received regular monitoring by the Quality Assurance Manager and Principal Investigators to assure ethics, scientific integrity, and confidentiality of the research Patients were frequently asked during the interview if they wanted to continue and were given a toll free number at the end of the interview to contact MSU if they had questions or concerns. The quality of the 41 interviewer and the interview process was monitored monthly by the Principal Investigators. Patients were informed that they could freely withdraw fiom the study at any time without loss of benefits they would otherwise be entitled to and without penalty to their health care. A request by any patient to withdraw their consent and to discontinue participation in the study was promptly and unconditionally honored Also, no financial costs to the patient resulted from their participation in the study. Patient and Other Anticipated Benefits It is unknown whether or not patients had directly benefited fi'om participation in the study. Patients had their symptoms assessed and the patient’s oncologist was notified of any symptoms reported by the patient that placed them at risk Sharing of their thoughts and feelings may have provided therapeutic benefits. Patients may have felt good inbeing able to contribute to the state-of-the-science which may help future patients. The patient’s oncologist was notified of any symptoms reported by the patient that may have placed them at risk IfCRF and other unpleasant symptoms along with PSE, physiological, and situational patient characteristics were found to impact functional status, strategies for treatment for CRF could be developed As a result, benefits to health care professionals include an increased understanding of the fatigue phenomenon in persons with LC as well as DC diagnoses. The risk is small compared to the potential gain of these anticipated benefits from the study. Importance of the Knowledge to be Gained The importance of the knowledge to be gained is that this is one of few studies identified that examined CRF and other associated unpleasant symptoms in persons with 42 cancer and specifically LC. This study is unique in that it focused on persons with LC and explored specific patient characteristics that predict higher levels of CRF. Moreover, this research measured a person’s PSE and whether PSE relates to their ability to manage their fatigue and to improve their PFS. Thus, this study is expected to lead to future interventional research to improve the PFS of persons with cancer, particularly LC when faced with fatigue. Data Analysis Plan The foremost purpose of this cross-sectional descriptive study was to examine fatigue in persons with LC and analyze how PSE in managing fatigue impacts PFS. Utilizing persons with OC diagnoses as a comparison group to persons with LC as well as to increase sample size assisted in answering the following research questions for the study. Data were analyzed using SPSS, Inc., 13.0 Version; LISREL 8.72 Version; and, SYSTAT, Inc., 11.0 Version (Point Richmond, CA). Tests were two-tailed with level of significance set at 0.05 for all tests. . Power As discussed by Cohen (1988), power analysis was determined based upon regression using predictor variables utilizing Research Question #6 of the study [i.e., What is the unique contribution of physiological and contextual characteristics, CRF, other unpleasant symptoms, and PSE to the PFS of persons with cancer (LC and OC)?]. Thus, power analysis of Research Question #6 was based on 28 possible predictors, and the selection was based on measures of tolerance, the independence of predictors (von Eye & Schuster, 1998). Given a sample size of 298 persons with cancer and utilizing all 28 predictors, a two-tailed test with a level of significance set at 0.05 and power set at 0.80 43 renders an effect size (ES) of 0.09 (Erdfelder, Faul, & Buchner, 1996). According to classification set by Cohen (1988), this value is within the small range of BS for multiple regression analysis (.02 - .15 small ES; .15 - .35 moderate ES; .35 > large ES). Rehse and Pukrop (2003) report a moderate BS is consistent in psychosocial studies provided to adult patients with cancer. Analyses of Research Questions The analyses of the research questions involved models in which the parameters are set equal for the comparison groups. For reason of sample size, the analysis plan included using the variable “type of cancer diagnoses” (LC and OC diagnoses) as a predictor in the models which incorporated the use of the total sample (N = 298). Research questions and plans for analyses of baseline data were as follows: 1. What are the patient characteristics that relate to CRF severity in persons with cancer (LC and OC diagnoses)? Regression was used to determine which patient characteristics predict variations of CRF severity for persons with cancer (LC and OC diagnoses), and whether having a cancer diagnosis of LC compared with OC diagnoses was a predictor of CRF severity. The model used was: Y (CRF) = be + b1 (type of cancer diagnoses) + I); (type of treatment variables requiring statistical control) + b; (stage of cancer) + b4 (co-morbid conditions) + b5 (sex) + b6 (age) + by (race) + b3 (marital status) + be (level of education) + bro (employment data) + bu (health insurance data) by von Eye and Schuster (von Eye & Schuster, 1998). A best of all subset regression using a backward elimination procedure was employed in a theoretically driven fashion with the goal of finding the most parsimonious model. In addition, all predictors were placed in the equation simultaneously in an unconstrained regression. It is important to note that prior to employing all regression techniques, univariate and bivariate analyses were performed to assess the quality of the data to check for missing data, outliers and distribution of variables. An estimation algorithm using the maximum likelihood method was used to handle missing values (von Eye & Schuster, 1998). Inclusion .of patient characteristics predicting CRF severity involved assessment for issues of multicollinearity employing such diagnostics as evaluation of tolerance levels, evaluation of effect sizes to ensure optimal use of power, and testing each predictor individually to ensure that the contribution made by that particular variable was statistically significant (von Eye & Schuster, 1998). Residual analysis was used to check the assumptions of normal distribution, homoscedasticity, and linear relationships between the statistically significant predictor variables and the criterion variable. 2. How does having a cancer diagnosis of LC compare with OC diagnoses as a predictor of CRF severity? Similar procedures were used for the separate group analysis of persons with LC and OC diagnoses as were used for the total sample of persons with cancer in Research Question #1. The exception being that “type of cancer diagnoses” (LC and OC diagnoses) was not entered into the initial model for group analysis. 3. What is the symptom experience of persons with LC and how does it compare to the symptom experience of OC diagnoses, including the relationships between CRF and other unpleasant symptoms? Descriptive statistics were used to analyze the frequency and severity of the individual symptoms. Correlation analysis was employed to examine the relationship 45 between CRF severity (total CRF Severity score) and each of the individual symptom’s severity level (Symptom Experience Inventory score) as well as the total symptom severity. The total symptom severity associated with each symptom was calculated by summing each subject’s response to severity scores for each symptom reported (i.e., a reported symptom is a symptom with a severity score greater than zero) and dividing by the total number of symptoms reported to standardize the score on an 11-point scale. The individual and total symptom weighted mean severity scores were calculated Individual symptom weighted mean severity scores were calculated by multiplying the mean score of each symptom by the number of persons who actually reported experiencing the symptom and dividing by the total number of persons in the sample. The total symptom weighted mean severity score was calculated by multiplying each symptom mean by the number of persons who reported the symptom. Next, the multiplied means were summed and divided by the total number of persons used for the multiplication 4. Does PSE for fatigue management mediate the relationship between CRF severity and PFS in persons with cancer (LC and OC diagnoses)? To test the mediation model, a series of three regression analyses specified by Baron and Kenny were performed (Baron & Kenny, 1986; Kenny, 2005). First, single order relationships among PSE, CRF and PFS (SF-36 PFS subscale) were established via Pearson correlations. Next, three regression analyses were estimated: 1) Y (PSE) = be + b1 (CRF); 2) Y (PFS) = bo+ b1(CRF);and, 3) Y (PFS) = bo+ b1(CRF)+ b2 (PSE)- Mediation was established when the following conditions were met: 1) CRF affects PSE in the first equation; 2) CRF affects PFS in the second equation; and, 3) PSE affects PFS in the third equation Mediation was established when these conditions all held in the 46 predicted direction, and the effects of CRF on PFS was less in the third equation than in the second Further analysis was performed which involved significance testing of the mediation pathway via Sobel Test (Baron & Kenny, 1986; Dudley, Benuzillo, & Carrico, 2004; Kenny, 2005). 5. Does mediation differ in persons with LC as compared to OC diagnoses when evaluating PSE for fatigue management as a mediator between the relationship of CRF severity and PFS? Similar procedures were used for the separate group analyses of persons with LC and OC diagnoses as were used for the total sample of persons with cancer in Research Question #4. Similar to the total sample, the Sobel Test was employed for both persons with LC and OC diagnoses. However, while the Sobel Test yields greater description of the mediation pathway, Hoyle and Kenny (1999) and Kenny (2005) identified a limitation of the Sobel Test is that the power of the test is low and the test is very conservative. Consequently, Hoyle and Kenny (1999) and Kenny (2005) recommend a minimum sample size when using the Sobel Test of at least 200 cases. This limitation is noted later in the mediation analysis of persons with LC since the sample size of this group falls short of the recommended sample size of 200 cases. The Sobel Test serves as an informal test for persons with LC due to the low sample size of this group. Thus, formal evaluation of whether PSE for fatigue management mediates the relationship between CRF and PFS was formally examined using regression analyses procedures outlined by Baron and Kenny (1986) and Kenny (2005). 47 6. What is the unique contribution of physiological and contextual characteristics, CRF, other unpleasant symptoms, and PSE to the PF S of persons with cancer (LC and 0C diagnoses The analysis used incorporated hierarchical multiple regression model: Y (PFS) = b0 + b; (contextual patient characteristics) + b2 (physiological patient characteristics) + b3 (type of treatment variables requiring statistical control) + b4 (PSE) + b5 (other unpleasant symptoms) + be (CRF). In sequential order, the independent variables were entered in six separate blocks: contextual patient characteristics, physiological patient characteristics, type of treatment variables requiring statistical control, PSE, other unpleasant symptoms, and CRF. Consequently, after each block was entered, the contribution of each variable above and beyond the last was accounted for in the prediction of PFS of persons with cancer (LC and OC diagnoses). 7. Considering the unique contribution of physiological and contextual characteristics, CRF, other unpleasant symptoms, and PSE for fatigue management to PFS, how does having a cancer diagnosis of LC compare with DC diagnoses as a predictor of PFS? Hierarchical multiple regression analysis was performed separately on groups of cancer patients, those with LC and OC diagnoses. Similar procedures used in Research Question #6 for the total sample was also used for persons with LC and OC diagnoses. 8. Through the employment of a Path Model, is the PFS of persons with cancer (LC and OC) predicted through physiological and contextual patient characteristics, CRF, other unpleasant symptoms, and PSE for fatigue management? Final analysis included an exogenous-endogenous Path Model to test the hypothesis that PFS for persons with cancer (LC and OC diagnoses) is predicted through 48 physiological and contextual patient characteristics, CRF, other unpleasant symptoms, and PSE (Raykov & Marcoulides, 2006). This incorporated a sequence of predictions tested through a path model via LISREL Version 8.72 statistical software package. In a first step, exogenous variables were selected through bivariate analyses with an inclusion criterion set at p < 0.20. Multiple regression analyses was then conducted within each group of variables (physiological patient characteristics; contextual patient characteristics; type of treatment variables requiring statistical control) with an inclusion criterion set at p < 0.20. These criteria were chosen to maintain as many variables as possible, given that they could eventually be significant in the final analysis. Consequently 16 out of 25 exogenous variables to the prediction of CRF severity were retained in the initial model. All analyses were conducted using the Satorra-Bentler Robust Maximum Likelihood Method of parameter estimation to adjust model chi-square for non-normally distributed variables. Several model fitting measures were used to attain a parsimonious final solution which included evaluation of parameter estimates, modification indexes, theoretical considerations, and goodness-of-fit tests. Specific goodness-of-fit-tests included the Satorla-Bentler Scaled Chi-Square which reflects the degree of discrepancy between the observed covariance matrix derived from the data and that predicted by the model. A small, nonsignificant chi-square indicates that one cannot reject the null hypothesis that the tested model fits the (bra (Raykov & Marcoulides, 2006). Another goodness-of-fit test used was the Root-Mean Square Error of Approximation (RMSEA) which provides an estimate of the average absolute discrepancy between the model covariance estimates and the observed covariances (Raykov & Marcoulides, 2006). For 49 this index, values S 0.05 indicates close approximate fit with a value of zero indicating the best fit The 90% confidence intervals (CI) for population parameters estimated by the RMSEA reflects the degree of uncertainty associated with RMSEA as a point estimate at the 90% level of statistical confidence (Raykov & Marcoulides, 2006). Ifthe lower bound of a 90% CI is 5 0.05, the model has close approximate fit in the population Finally, the Comparative Fit Index (CFI) was used which indicates the amount of covariation in the data that can be reproduced by the given model (Raykov & Marcoulides, 2006). The CF I is more robust for deviations fi'om normality. A CFI value above 0.90 indicates reasonably good fit of the model. 50 CHAPTER FOUR RESULTS Preliminary Examination of the Data Normality Continuous variables, namely age, co-morbid conditions, individual symptoms, total symptom severity of the other unpleasant symptoms, worst and least cancer-related fatigue (CRF) severity, total CRF severity, perceived self-efficacy (PSE) for fatigue management, and physical functional status (PFS) were examined for normality of distribution Age was found to be normally distributed Univariate analysis revealed that some of the individual symptoms were not normally distributed When calculated, Fisher’s Measure of Skewness Test showed the variables of pain, nausea, cough, difficulty remembering, and weakness had values greater than +1.96. Likewise, Fisher’s Measure of Kurtosis showed that the individual symptoms of fatigue, nausea, insomnia, anorexia, dry mouth, constipation, weakness, and alopecia had values above +1.96 or below -1.96. However, the total symptom severity score of the other unpleasant symptoms which consists of a calculation utilizing the individual symptoms was normally distributed Co-morbid conditions and worst and least CRF had both skewed and kurtosed distributions with values above +1.96 or below -1.96. The skewness values for the summary scores of total CRF severity (-.759), PSE for fatigue management (-1.33), and PFS (-l .94) all had values indicating that that the distribution was not skewed However, the kurtosis values for the summary scores of total CRF severity (-2.90), PSE for fatigue management (-2.87), and PFS (-3.63) was platykurtic. The finding of non-normal 51 distributions among the symptom-related and other health-related variables was expected and therefore considered normal. Frequency and Missing Data The data were examined for fi'equency and patterns of missing data via the Missing Analysis Program in SYST AT Version 11.0 statistical software package (Point Richmond, CA). An estimation algorithm using the maximum likelihood factor of covariance was used for missing value analysis (von Eye & Schuster, 1998). A cut-off point determined at 50% or less missing cases was used to estimate incomplete data yielding 100% complete cases on data analyzed (von Eye & Schuster, 1998). All missing data were determined to be missing completely at random via The Little MCAR Test Statistic. Descriptive Statistics for Patient Characteristics: Physiological Type of Cancer The analysis was performed on a sample of 298 persons with cancer, with persons diagnosed with lung cancer (LC) comprising 21% of the sample. Of the persons with LC, 78% (N = 63) were diagnosed with non-small cell LC, with 22% diagnosed with small cell LC. Persons with other cancer (OC) diagnoses comprised 79% (N = 235) of the total sample with breast cancer (45%), colon cancer ( 19%), urological cancer (11%), gynecological cancer (9%), and other cancers (16%) making up the OC diagnoses category. Stage of Cancer For stage of cancer, 26% of persons with cancer were diagnosed in early stage (Stage I, II, and limited disease) while 74% (Stage III, IV, and extensive disease) were found to 52 be in a late stage cancer. Most persons with LC (81%) and 0C (72%) presented with late stage cancer (see Table 1). Among the total sample, there were no significant differences between early and late stage cancer between LC and OC diagnoses (x2 = 2.1; 1 df; p = .147). There were more persons diagnosed with late stage cancer as compared to early stage cancer for both LC (x3 = 24.1; 1 df,p = .000) and OC (x2 = 45.1; 1 dfip = .000) diagnoses. Treatment, Radiation Therapy Most persons with cancer (84%) did not report receiving radiation therapy at the baseline interview. However, persons with LC (25%) had a significantly higher frequency of reporting receiving radiation therapy as compared to persons with OC (13%) (x = 5.6; 1df,p = .018) (see Table 1). Treatment, Surgery Out of the total sample, most persons with cancer (62%) had a surgical procedure prior to receiving chemotherapy as compared to not having a surgical procedure prior to chemotherapy (x3 = 35.4; 1 of, p = .000). This included 71% of persons with DC, and 30% of persons with LC. The odds of having surgery prior to chemotherapy were 5.7 times higher in persons with OC as compared to LC (95% CI: 3.1 - 10.4) (see Table 1). Similarly, out of the total sample, 32% of persons with cancer reported having a surgical procedure during chemotherapy which included 36% of persons with DC and 16% of persons with LC (see Table 1). The odds of having a surgical procedure during chemotherapy were 2.94 times higher for persons with OC as compared to LC (95% CI: 1.4 -6.1). 53 Treatment, Chemotherapy Type at Consent Out of the total sample, first line initial chemotherapy treatment was reported by the largest percentage of persons with cancer (47%) at time of consent of the study, which included 65% of persons with LC and 42% of persons with OC. The next most frequently reported treatment at the time of consent of the study by the total sample (21%) was second line chemotherapy because first line did not work or the disease had progressed. This included 16% of persons with LC and 22% of persons with OC. Another treatment regimen reported by the total sample included adj uvant chemotherapy with radiation therapy (20%), which comprised 16% of persons with LC and 20% of persons with OC. Adj uvant chemotherapy is given after primary treatment to increase the rate of cure. Neoadj uvant chemotherapy was the least frequent treatment regimen reported by 12% of all persons with cancer. Neoadj uvant chemotherapy is administered prior to the primary treatment of cancer. Two percent of persons with LC and 15% of persons with OC reported receiving neoadj uvant chemotherapy at the time of consent into the study. C o-morbid Condition The total sample reported a mean number of two co-morbid conditions with hypertension (45%), emotional problems (28%), other major health problems (21%), and other cancer diagnosis (20%) accounting for the most commonly reported co-morbid conditions respectively. Persons with LC (M = 2.79; SD = 1.81) had a significantly higher mean number of co-morbid conditions as compared to persons with OC (M = 1.82; SD = 1.46) (t = -4.47; p = .000). Hypertension and emotional problems remained the top two ranked co-morbid conditions respectively in both persons with LC and OC (see Table 1 and Table 2). 54 Sex While not significantly different, women made up the majority (54%) of the LC group (x2 = .397; 1 df, p = .529). There were significantly more women, 75%, than men in the OC group (x2 = 56.3; 1 df; p = .000). As well, women significantly comprised the majority (70%) of the total sample (1’ = 9.97; 1 df; p = .002). The odds of being a woman were 1.7 times greater in the OC group as compared to the LC group (see Table 1). Age In the total sample, the range for age was 25 to 90 years (M= 57.10; SD = 11.88). Those persons with LC were significantly older with a mean age of 62 years (SD = 10) as compared to persons with OC with a mean age of 56 years (SD = 12) (t = -3.97; p = .000) (see Table 1). Laboratory Values This study proposed to analyze the absolute neutrophil count and hemoglobin level of persons with cancer to the prediction of CRF and PFS. However, this was not possible since up to 85% of the data regarding the absolute neutrophil count and hemoglobin were not collected. 55 Table 1 Physiological Patient Characteristics by Percentage of the Total Sample and by Group Characteristics % Total Sample % leg Cancer % Other Cancer (N = 298) (N = 63) (N = 235) Stage of Cancer Early stage 26% 19% 28% Late stage 74% 81% 72% Treatment Use Radiation therapy 16% 25% 13% Surgery priorto chanotherapy 62% 30% 71% Surgery dming chemotherapy 32% 16% 36% Sex Female 70% 54% 75% Male 30% 46% 25% Co-morbid Conditions M(SD) 2.02 (1.59) 2.79 (1.81) 1.82 (1.46) Minimrnn-Maximum 0-9 0-9 0-8 A86 (yam) M (SD) 57.10 (11.88) 62.30 (10.20) 55.72 (11.93) Minimum-Maximum 25-90 37-82 25-90 56 Table 2 Co—morbid Conditions by Percentage of the Total Sample and by Group Co-morbid Condition % Total Sample % Lung Cancer % Other Cancer (N = 298) (N = 63) (N = 235) Hypertension 45 60 40 Emotional problems 28 32 26 Other major health problems 21 21 20 Other cancer 20 16 20 Heart problem 17 25 15 Loss of mine beyond control 16 16 16 Diabetes 14 19 12 Cataract surgery 10 16 9 Arthritis, rheumatism 10 21 7 Emphysema 9 29 4 Wear a hearing aid 5 8 4 Smgical replacement of joint 4 6 3 Stroke 2 6 1 Angina 2 5 2 Fractured hip .30 — .40 57 Descriptive Statistics for Patient Characteristics: Contextual Race Racial composition of the total sample of persons with cancer was 87% Caucasian, 10% African American, and 3% other racial minorities. For persons with LC, 94% were Caucasian and 6% were Afiican American Similar findings were found for persons with OC with 85% reported being Caucasian, 11% Afiican American, and 4% another racial minority (see Table 3). Marital Status The majority of the total sample were married (68.8%) followed by being divorced or separated (15.4%), never married (9.1%), widowed (5.4%), or living together (1.3%). A similar trend of marital status was found among the LC and OC groups (see Table 3). Level of Education Educational achievement for the total sample consisted of 24.5% graduated from high school, 28.9% with some college/technical training, 20.1% completing college, and 16.4% with graduate/professional education Only 10% had some high school or less education (see Table 3). Employment Data For the total sample, 49.3% were retired, 21.8% were receiving disability, 20.1% were on a temporary leave from employment, and 13.8% reported that they had to quit employment. Similar findings concerning employment status were found by group for both persons with LC and OC cancer (see Table 3). Whether or not a change in employment status for the sample was due to cancer is not known. 58 Annual Combined Household Income Information on the annual combined household income showed that most persons with cancer reported earning within the range of $50,000 to $74,999. More specifically, 27% of persons with LC and 32% of persons with OC diagnoses reported producing $50,000 to $74,999 annually (see Table 3). Health Insurance Data For the total sample, most persons with cancer possessed health insurance (99%) with the insurance policy being held by themselves (70%) versus their spouse (25%). The majority of the total sample possessed private insurance (81%) either by itself or in addition to being insured with Medicare. The total sample reported being insured through Medicare (28%) and Medicaid (10%) (see Table 3). Most persons with LC (51%) held a Medicare policy in comparison to persons with OC (22%) diagnoses (x,2 = 20.2; df= 1; p = .000). The odds of a person with LC holding a Medicare policy was 3.6 times greater than for persons with OC (95% CI: 2.0 - 6.5). Persons with OC (85%) held a private insurance policy more often than persons with LC (68%) (x3 = 8.8; df= 1; p = .003). The odds of a person with OC diagnoses holding a private insurance policy was 2.6 times greater than persons with LC (95% CI: 2.0 - 6.5). 59 Table 3 Contextual Patient Characteristics by Percentage of the Total Sample and by Group Characteristics % Total Sample % erg Cancer % Other Cancer (N = 298) (N = 63) (N = 235) Race Caucasian 86.9 93.7 85 Afiican American 9.7 6.3 10.6 Native American Alaskan 2.0 - 2.6 Mexican American Hispanic .70 - .90 Oriental Asian Pacific Islander .70 -- .90 Marital Status Married 68.8 66.7 69.4 Divorced/Separated 15.4 17.5 14.9 Never married 9.] 4.8 10.2 Widowed 5.4 7.9 4.7 Living together 1.3 3.2 .90 Level of Education High school or less 10 15.9 8.6 High school 24.5 23.8 24.7 Some college/Technical training 28.9 36.5 26.8 College 20.1 12.7 22.1 Graduate/Professional 16.4 1 1.1 17.9 60 Table 3 Continued Characteristics % Total Sample % Lung Cancer % Other Cancer (N = 298) (N = 63) (N = 235) Employment Information Retired 49.3 58.7 46.8 Receiving disability 21.8 25.4 20.9 On a temporary leave fiom work 20.1 17.5 20.9 Quit work 13.8 11.1 14.5 Annual Combined Income < $24,999 14.4 20.6 12.8 $25,000 to $49,999 25.8 33.3 23.8 $50,000 to $74,999 28.2 31.7 27.2 $75,000 to $99,999 11.4 4.8 13.2 $100,000 to > $200,000 20.1 9.5 23.0 Type of Health Insurance Plan Private 81.2 68.3 84.7 Medicare 28.2 50.8 22.1 Medicaid 9.7 12.7 8.9 61 Descriptive Statistics for Symptoms: Cancer-Related Fatigue (CRF) Fatigue Frequency The total sample of persons with cancer reported experiencing fatigue on at least one of the seven days from their baseline interview. The mean number of days that all persons with cancer reported having fatigue was 5.14 (SD = 2.02), with 47% stating fatigue on all seven days. No significant differences (t = -l.470; df= 296; p = .143) were found in the frequency of the mean report of fatigue by persons with LC (M = 5.48; SD = 1.81) and OC diagnoses (M = 5.10; SD = 2.07). The most frequently rated number of days in which persons with LC (52%) and 0C (45%) experienced fatigue was seven days. Lastly, the median number of days that persons with LC experienced fatigue was higher at 7 days as compared to 5 days for persons with DC diagnoses. Current Fatigue Severity During the bweline interview, 99% of the total sample of persons with cancer reported currently experiencing fatigue, with a mean fatigue severity score of 5.24 (SD = 2.33) for those who reported fatigue. There was no significant difference (t = -. 885; af= 292; p = .377) in the current mean report of fatigue severity by persons with LC (M = 5.47; SD = 2.19) and OC (M = 5.17; SD = 2.37) diagnoses (see Table 4). Worst Fatigue Severity In addition, during the baseline interview, the total sample of persons with cancer rated their worst fatigue severity in the past seven days as a mean score of 6.52 (SD = 2.39). However, there was no significant difference (t = .560; df= 113; p = .577) in the worst fatigue severity score by persons with OC (M = 6.55; SD = 2.46) and LC (M = 6.38; SD = 2.08) (see Table 4). 62 Total CRF Severity The total CRF severity score was calculated by summing each subject’s response to the current and worst severity scores and dividing by two to standardize the score on an ll-point scale. The total CRF severity score reported by the total sample of persons with cancer (N = 298) was a mean of 5.84 (SD = 2.23). No significant difference was found (t = -. 161; df= 296; p = .872) in the mean total CRF severity score report for persons with LC (M = 5.88; SD = 2.00) and OC (M = 5.83; SD = 2.29) (see Table 4). Descriptive Statistics for Symptoms: Other Unpleasant Symptoms Associated with Cancer and Cancer Treatment Frequency and severity of the other unpleasant symptoms associated with cancer and cancer treatment will be discussed in greater detail under research question #2. Of those persons (N = 296) who reported other unpleasant symptoms associated with cancer excluding fatigue, the total mean symptom severity score was 4.64 (SD = 1.58). Persons with LC (M = 4.99; SD = 1.43) had a significantly higher total symptom severity score as compared to persons with DC (M = 4.54; SD = 1.60) diagnoses (t = -1.99; df= 294; p = .047) (see Table 4). To assist in the interpretation of the significance of the findings, an effect size for the difference of the means between the two groups (LC and OC diagnoses) concerning the total symptom severity was small (d = .30). Cohen’s (1988) thresholds for small, moderate, and large are respectively .20, .50, and .80. It should be remembered that Cohen (1988) defined these thresholds to reflect the typical effect sizes encountered in the behavioral sciences as a whole and that many effects involving clinical and psychological research as well as new areas of research are likely to be small. 63 Thus, interpretations of effect sizes are based upon not only the thresholds prescribed by Cohen, but also judgment of the clinical-research phenomenon. Table 4 Comparison of Persons with Lung Cancer and Other Cancer Diagnoses Mean Scores on Symptom Severity Scores Symptom Severity Total Sample Lung Cancer Other Cancer Scores M (SD) M (SD) M (SD) t df p Current Fatigue 5.24 (2.33) 5.47 (2.19) 5.17 (2.37) -.855 292 .377 Score N=294 N=62 N=232 Worst Fatigue 6.52 (2.39) 6.38 (2.08) 6.55 (2.46) .560 113 .577 Score N = 298 N = 63 N = 235 Total CRF Severity 5.84 (2.23) 5.88 (2.00) 5.83 (2.29) -. 161 296 .872 Score N= 298 N= 63 N= 235 Total Symptom 4.64 (1.58) 4.99 (1.43) 4.54 (1.60) -l.99 294 .047 Severity Score N = 296 N = 63 N = 233 64 Descriptive Statistics for Perceived Self-Efficacy for Fatigue Management in Persons with Cancer The mean score for the total sample for persons with cancer (N = 298) for PSE to manage fatigue was 6.43 (SD = 2.25). While the mean PSE to manage fatigue score was slightly higher in persons with LC (M = 6.66; SD = 1.96) as compmed to persons with OC (M = 6.37; SD = 2.33), the difference was not statistically significant (t = -1.028; af= 113; p = .306) (see Table 5). Descriptive Statistics for Performance Outcome: Physical Functional Status The PFS mean score reported by the total sample of persons with cancer was 58.10 (SD = 27.20). Persons with LC (M = 44.29; SD = 26.73) had a statistically significantly lower PFS score when compared to persons with DC (M = 61.81; SD = 26.16) (t = 4.70; df= 296; p = 0.000) (see Table 5). An effect size for the difference of the means regarding PFS in persons with LC and OC was moderate (d = .662). 65 Table 5 Comparison of Persons with Lung Cancer and Other Cancer Diagnoses Mean Scores on Perceived Self-Efficacy for Fatigue Management and Physical Functional Status Total Sample Lung Cancer Other Cancer M (SD) M (SD) M (SD) 1 df p PSE for Fatigue 6.43 (2.25) 6.66 (1.96) 6.37 (2.33) -1.028 113 .306 Management N= 298 N= 63 N= 235 Score Physical 58.10 (27.10) 44.29 (26.73) 61.81 (26.16) 4.70 296 .000 Functional N= 298 N = 63 N= 235 Status Score Results of Research Questions Research Question #1 : What are the patient characteristics that relate to CRF severity in persons with cancer (LC and OC diagnoses)? Prior to conducting analysis of Research Question #1, a total CRF severity score was calculated by summing each subject’s response to current and worst severity of CRF items and dividing by two to standardize the score on an 11-point scale. This score was used in all regression analyses with total CRF severity as the criterion variable. Correlations between predictor variables and total CRF severity were calculated The predictor variables that significantly correlated with total CRF severity were the patient’s 66 stage of cancer (r = .149; p = .01), total number of co-morbid conditions (r = .139; p = .016), and sex (r = .149;p = .010). A total of 25 patient characteristics were identified as important to the initial best of all subset regression model (T able 6). The following eight categories of Physiological Patient Characteristics included 11 out of 25 possible variables to the prediction of total CRF severity: 1) type of cancer, 2) stage of cancer; 3) receiving radiation therapy; 4) surgery prior to chemotherapy which includes four groups [yes had surgery; don’t know if had surgery; this response choice was not selected; no surgery and this group served as the reference group when the groups were dummy coded (i.e., group left out)]; 5) surgery during chemotherapy which includes three groups [yes had surgery; this response choice was not selected; no surgery and this group served as the reference group when the groups were dummy coded (i.e., group left out)]; 6) co-morbid conditions; 7) age; and 8) sex The six categories of Contextual Patient Characteristics included 14 out of 25 possible variables to the prediction of total CRF severity: 1) race; 2) marital status; 3) level of education achieved; 4) employment data including whether a person was retired, receiving disability, was on a temporary leave from employment, and whether they had to quit employment; 5) annual combined household income; and 6) health insurance data including whether or not a person with cancer had health insurance and if so, who held the policy (patient or spouse), and the type of health insurance policy held (private, Medicare, or Medicaid). A backward elimination procedure was used in a theoretically driven fashion using the theoretical framework for the study to identify the patient characteristics that predict total CRF severity in persons with cancer. The first step began with computing the initial 67 model that included all 25 predictors. Having computed the first model, goodness-of-fit criterion was used in model selection that included examination of tests on individual regression coefficients (t-values) to remove the variables that contribute the least to the model (assuming that its contribution is not significant), and the model’s multiple correlation coefficient alongside the F -statistic with its associated significance level (p- value). Additionally, all predictor variables entered into the model had to reach a tolerance level of 0.4 and a level of significance = 0.05. Step] The first step, the calculation of the initial model to find the best of all subset model to predict total CRF severity in persons with cancer, resulted in a model that was not statistically significant [R2 = .105; F (25, 272) = 1.274; p = .177]. To improve the model, nine predictors with t-values < -/+.50 were eliminated in Step 2 (see Table 6). Step 2 The elimination of 9 out of 25 predictors from the initial model resulted in a statistically significant model in Step 2 explaining 10.2% of the variance [F (16, 281) = 2.0; p = .013]. The model in Step 2 had only one statistically significant patient predictor to total CRF severity, the total number of co-morbid conditions. However, in this model, 9 out of 16 predictors had t-values 3 -/+ 1.00. In an attempt to improve the model in Step 3, all predictors with t-values > -/+ 1.00 were retained which meant that seven predictors were eliminated (see Table 6). Step 3 With the elimination of seven predictors, the explained variance in Step 3 decreased from 10.2% to 8.4%, but the model became more statistically significant in predicting 68 total CRF severity in persons with cancer [F (9, 288) = 2.932; p = .002]. In addition to the total number of co-morbid conditions, age became a statistically significant predictor to total CRF severity. Subsequently, to improve the model predicting total CRF severity, all predictors with t-values > -/+ 1.50 Were retained (stage of cancer, total number of co- morbid conditions, age, and sex) and five predictors were eliminated in Step 4 (see Table 6). Step 4 The elimination of five predictors identified in Step 3 resulted in a further decrease in the explained variance to 7.3%. Like the model in Step 3, the model in Step 4 became more statistically significant in predicting the total CRF severity in persons with cancer [F (4, 293) = 5.759; p =.000]. The total number of co-morbid conditions, age, and sex were all statistically significant predictors in the total CRF severity of persons with cancer. As in Step 3, stage of cancer remained not statistically significant in the prediction of total CRF severity in persons with cancer (I < -/+ 2.00). One more attempt to improve the model predicting total CRF severity in persons with cancer included eliminating the predictor, stage of cancer, from the model in Step 5 (see Table 6). Step 5 The elimination of stage of cancer resulted in a slight reduction in the explained variance to the prediction of total CRF severity in persons with cancer, from 7.3% in Step 4% to 6.2% in Step 5 with a slight increase in the F-statistic. Additionally, the t-values and corresponding p—values for the statistically significant predictors (total number of co- morbid conditions, age, and sex) in Step 5 as compared to Step 4 increased in value. Consequently, as compared to the other four models, the model in Step 5 was selected as 69 the best model to represent the prediction of total CRF severity in persons with cancer (see Table 6). Further Analysis, Step 6 The Theory of Unpleasant Symptoms (TOUS) predicts that patient characteristics, symptom dimensions, and interactions between symptoms have an effect on a person’s performance outcome, PFS. Consequently, to depict a more comprehensive and truer picture of the fatigue phenomenon in persons with cancer, further analysis was conducted This step added two components of the theoretical framework guiding this dissertation project that was not accounted for in the original prediction equation of total CRF severity in persons with cancer. Consequently, the patient characteristics of PSE for fatigue management and the other unpleasant symptoms from cancer and cancer treatment were accounted for in the final model predicting total CRF severity in persons with cancer as identified in Step 6. The PSE for fatigue management score and the total symptom severity score were the values added to the model in Step 5 to calculate the more comprehensive description of the fatigue phenomenon in persons with cancer as guided by the TOUS. Once added, the explained variance in the prediction of the total CRF severity in persons with cancer increased fi'om 6.2% to 39.2% [F (5, 290) = 37.41; p = .000]. The predictors of total number of co-morbid conditions and sex remained statistically significant (t > 2.0); but age no longer remained a predictor of total CRF Severity (t = -1.67). Both PSE for fatigue management (I = -6.49) and the other unpleasant symptoms associated with cancer and its treatment (t = 9.77) were statistically significant in the prediction of total CRF severity in persons with cancer (see Table 6). 7O The Final Model The initial model as identified in Step 1 demonstrated good apriori power at 0.80 to detect relationships with small effects (.085) while utilizing all 25 predictors with a two- tailed test and a level of significance set at 0.05 (Erdfelder et al., 1996). Therefore, at better-than-chance-levels, the final model as depicted in Step 6 for the prediction of total CRF severity in persons with cancer accounts for 39.2% of the variance [F (5, 290) = 37.41; p = .000]. This prediction of the actual value of total CRF severity was made with small residual variance. The standard error of the estimate shows that on average, the actual value will be +/- 1.76 of the predicted value 68% of the time, and will be within +/- (2) (1.76) 95% of the time. Tolerance levels for the predictors ran from a low of .810 for co-morbid conditions, .811 for age, and .957 and greater for all others. Significance testing. In the final model depicted in Step 6 each predictor was tested to ensure that the contribution made by that particular variable was statistically significant (von Eye & Schuster, 1998). This involved comparing two (nested models, the constrained model (c) and the unconstrained model (u) to calculate the F-statistic. As depicted in Table 7, the significance testing of each individual predictor demonstrates that they not only contribute to the model as a whole but also independently to the predictiOn of total CRF severity in persons with cancer. Assumptions Met, the Model Confirmed The mean of the predicted value in the final model of total CRF severity was 5.84 (SD = 1.40). Assessment of the histogram of the standardized residuals shows roughly a normal curve as validated by the range of the residual variance from -2.965 to 3.078. The normal probability plot shows a monotonic pattern and reasonable closeness of the cases 71 to the regression line, confirming a normal distribution Additionally, the scatterplots of residuals for each statistically significant predictor variable and criterion variable reveals a pattern that is more dense in the center with decreasing density on the outer edges, thus demonstrating normal homoscedasticity. Summary of Results for Research Question #1 In summary, for persons with cancer, the following predicts greater total CRF and the predictors are in the order of greatest to least strength of prediction: greater total symptom severity of the other unpleasant symptoms; lower PSE for fatigue management; greater number of co-morbid conditions; and women as compared to men. Note that in Step 5 when the two predictor variables, PSE for fatigue management and the other unpleasant symptoms were not accounted for in the model, the younger the age of a person with cancer, the greater the total CRF severity resulted. 72 8:9 «an. age .28: 22. Boga E S 8. 7 om< 53 8a.- 85:55 28.. Basso.— aoo.v SN 82:88 Beoaéo $585.85 5.3: ca: «.2 88.9. 836 8:83.. 02 55 an. 2:85 359:8 :23. A33 «S. gas 8.. .8» 33 an.. as; is 3 8m 2805530 mesa gnaw A23 «2 .- as; see 38. 3.880“. §3 8a.- 3623 865 8:88. oz $23 5 .- basque €282 ASS an $93 3? .85 ....8 $3 an. 852 :83.— A83 5.- €93 an .a> Sun Ego—mam 3.9850830 8 .5: >895 55 ms.-, 3522 8:338? $3 A83 8n- aoaaea Egg 3382 $3 «3.- seem 352 A83 8: 8853 omen 883 an. 83. $3 as. 880.8 25. E. NR .8 Ra 3.. _ at 8383.36 335.80 a? samenesafi .833?“ a a a Na new 3502 gammy—mom nouafizm ugoem 53.5 53828,: a was “850 as, museum a $98 $6 38. 3.8685 382 «am 2: 35:82 3 new .8 338m o 033. 73 .83 :2 85:88 5:853 E . .v S.. 8.8.8 888 8:88. :2 5.... e....- 33.... 28. .83 :8. e033 .3: 8.. can. :8.- 8882 58.. 3885058085....83 ...o: 8:535 :58: .0 2...... .83 :8- 883 88:: 8:88. oz .... 3 8.. 8.8.35 :58: ...o: 3.5:: .8. 5 an. 58.3 8:... 3:5. .55 G. 3 8..- 8:335 5.8: 8:83.... 38.. 8.. 58.3 .8: .8» 8.5 8.8.35 5.3: 38050830 3 3....— .0035 .83 38.- .35 .5552 88.. a: .856 .8 853 .83 ...m. 83. .83 8:. 85.8: 25 m8. ...: .2 o... 8.. a 6...... .8.-. . 38.... 28.. ES 8:- ...882 58.. 6:3 :8.- 0882 58... 28.. 8:33... 58...... 8.... .83 .3. 8:335 :58: ...o: 3.5:: an. .. a... 8m 3. 83.8886 838.80 .5. 8.38.886 8.8.33... 5 x: ... .... 83 83.58 c 8.8.: 74 .38.. N..: 8m 88.. m. .N- :3. ..8. R: 3:55.88 588.8 .30.. ...... .883: :88 8... 8: ... :2 m... e .33.. E. 8m 88.. a. .N- am... ..8. 8.: 85.5.80 58.8.8 .9... 8...- 8.8.8 88:: 8:88. :2 .33 ... .. .993 .8: 5 38:. .55 ...S... :3. . 8:835 :58: ...o: 3.5:: .3... .3... 88.3 .8: 8.: .83... 8...- 8:835 :58: 8338.. 3:85.885 3 S... 9:5: 8.5 883:. 5.8.. .3... a: .830 .8 88m :8. w..: ... 8.: 3.... ... .8... on. 8m ...... .. an. .- :3 .3. 8:388:86 888.80 .3. 838.885 855.38... a x: ... .... . 8.: 8:588 : 2:8. 75 82: Rd aaoafim “5.2%: 25 889%.? 3080352 3938. m2 ENS m3 gum $89 ”3- a? $33." aoueaoo 29.880 o8. 8N.“ v.2 an. o 33 and 5m A23 2”,”. om... 28.55 aoéaoo 29.830 o8. «fix on» So. m E: moafisafio 35:8 3: 835.8% Emsoifi m \u m «m . new 8:580 o 2.5 76 Table 7 Significance Testing of the Predictors to Total CRF Severity In Persons with Cancer Predictor Variable F -Statistic df Co-morbid Conditions 6.0 4, 291 Age 3.0 4, 291 Sex 5.5 4, 291 PSE for Fatigue Management 44.0 4, 291 Other Symptoms Associated with Cancer & 100.5 4, 293 Cancer Treatment All predictors significant at a level of p < .05 Research Question #2: How does having a cancer diagnosis of LC compare with DC diagnoses as a predictor of CRF severity? In answering Research Question #1, the “type of cancer diagnosis”, LC versus 0C diagnoses, was 1 of 25 variables placed in the initial model to identify the predictors that relate to total CRF severity in persons with cancer. Subsequently, in Step 2 of the backward elimination regression to find predictors to total CRF severity in persons with cancer, the “type of cancer diagnosis” was found not to be a statistically significant predictor (t = .980; p = .328). To continue furthering the science of the CRF phenomenon, predictors to total CRF severity were examined separately in both persons with LC and 0C diagnoses. With one exception, similar procedures were used for the separate group analysis of persons with LC and 0C diagnoses as were used for the total sample of persons with cancer. The 77 exception being that “type of cancer diagnoses” was not entered into the initial model for group analysis. Persons with Other Cancer Diagnoses Step I The first step, the calculation of the initial model to find the best of all subset model to predict total CRF severity in persons with DC diagnoses, resulted in a model that was not statistically significant [R2 = .126; F (24, 210) = 1.26; p = .194]. To improve the model, 9 out of 24 predictors with t-values < -/+ 0.50 were eliminated (see Table 8). Step 2 The elimination of 9 out of 24 predictors from the initial model resulted in a statistically significant model in Step 2 explaining 12.4% of the variance [F (15, 219) = 2.06; p = .013]. The model in Step 2 contained only two statistically significant patient predictors to total CRF severity, stage of cancer and co-morbid conditions. In an attempt to improve the model in Step 3, all predictors with t-values > -/+ 0.90 were retained which meant that six predictors were eliminated (see Table 8). Step 3 With the elimination of six predictors, the explained variance in Step 3 decreased from 12.6% to 10.8%, but the model became more statistically significant in predicting total CRF severity in persons with DC diagnoses [F (9, 225) = 3.02; p = .002]. Once again there were only two statistically significant predictors to total CRF severity, stage of cancer and co-morbid conditions. To improve the model predicting total CRF severity, four predictors with t-values 5 -/+1.19 were eliminated (see Table 8). 78 Step 4 The elimination of four predictors in Step 3 resulted in a further decrease in the explained variance to 9.4%. Like the model in Step 3, the model in Step 4 became more statistically significant in predicting tOtal CRF severity in persons with DC diagnoses [F (5, 229) = 4.74; p = .000]. Stage of cancer and co-morbid conditions remained statistically significant The predictor variable, age, converted to statistical significance (see Table 8). Another step, Step 5, was added to improve the model predicting total CRF severity in persons with DC diagnoses which included eliminating the most statistically insignificant predictor variable. Step 5 The elimination of one predictor in Step 4 resulted in a slight reduction of the explained variance from 9.4% to 8.8% (see Table 8). However, the model continued to strengthen with three statistically significant predictors to CRF severity: stage of cancer, co-morbid conditions, and age [F (4, 230) = 5.53; p = .000]. Another step, Step 6, was added to improve the model predicting total CRF severity in persons with DC diagnoses which included eliminating the only nonsignificant predictor variable, sex Step 6 With the elimination of the predictor variable, sex, all three former predictors (stage of cancer, co-morbid conditions, and age) in Step 5 remained statistically significant The model strengthened with an explained variance of 7.6% [F (3, 231) = 33.4; p = .000] (see Table 8). As in the model which included identifying the predictors to CRF severity in the total sample, PSE for fatigue management and the other unpleasant symptoms were accounted for in the final model as depicted in Step 7. 79 Further Analysis, Step 7 After PSE for fatigue management and the other unpleasant symptoms were added to the model in Step 6, the explained variance in the prediction of the total CRF severity in persons with 0C diagnoses increased from 7.6% to 42.4% [F (3, 231) = 33.4; p = .000]. The predictors of stage of cancer and co-morbid conditions remained statistically significant (t > 2.0); but age (t = -1.40) no longer remained a predictor of total CRF severity in persons with DC diagnoses. Both PSE for fatigue management (t = -6. 10) and the other unpleasant symptoms (t = 8.95) were statistically significant in the prediction of total CRF severity in persons with DC diagnoses (see Table 8). The Final Model The initial model as identified in Step 1 demonstrated good apriori power at .80 to detect relationships with small effects (0.11) while utilizing 24 predictors with a two- tailed test and a level of significance set at 0.05 (Erdfelder et al., 1996). Therefore, at better-than-chance-levels, the final model as depicted in Step 7 for the prediction of total CRF severity in persons with DC diagnoses accounts for 42.4% of the variance [F (3, 231) = 33.4; p = .000). This prediction of the actual value of total CRF severity is made with small residual variance. The standard error of the estimate shows that on average, the actual value will be +/- 1.77 of the predicted value 68% of the time, and will be within +/- (2) (1.77) 95% of the time. Tolerance levels for the predictor variables ranged from a low of .772 for age, .781 for co-morbid conditions, and .950 and greater for all others. Significance testing. As depicted in Table 9, the significance testing of each individual predictor demonstrates that they not only contribute to the model as a whole 80 but also independently to the prediction of total CRF severity in persons with DC diagnoses. Assumptions Met, the Model Confirmed The mean of the predicted value in the final model of total CRF severity was 5.83 (SD = 1.50). Assessment of the histogram of the standardized residuals shows roughly a normal curve as validated by the range of the residual variance from -3.09 to 2.91. The normal probability plot shows a monotonic pattern and reasonable closeness of the cases to the regression line, confirming a normal distribution. Additionally, the scatterplots of residuals for each statistically significant predictor variable and criterion variable reveals a pattern that is more dense in the center with decreasing density on the outer edges, thus demonstrating normal homoscedasticity. Summary of Results for Research Question #2 (Persons with 0C Diagnoses) In summary, for persons with DC diagnoses, the following predicts greater total CRF and the predictors are in the order of greatest to least strength of prediction: greater total symptom severity from the other unpleasant symptoms; lower PSE for fatigue management; earlier rather than later stage cancer; and greater numbers of co-morbid conditions. Note that in Step 6 when the two predictor variables, PSE for fatigue management and the other unpleasant symptoms were not accounted for in the model, the younger the person is with cancer, the greater the total CRF severity. 81 A83 8a.. 88:3 5.8.2 23 88am 863 3a.- a? A83 2 _ .- 85.35 5.8.. 22. .858 53 42 aoEBoo Eeoaéo suggest: A33 8.. 333 sea... 8:88:82 A33 in. 2:85 35988 33.2 A33 a». .95 92 .8» we; 3..- as: as 3 B: sausage 3.55 e83 $23 «8.- :33 ace 93 588$ $3 as; 3823 86% 3:88. oz A83 28.. gaze 3282 A33 NS. 9?; 3.: 323 ion 33 ”2 .- Beet 83a $5 a. Sea 3.. .aw sun EoEmoEEm 38050830 3 8E bomb—m ii an.. 832% Sesame .33 903 3a- seas»; 8%: 323 A23 §.- 35m 33: 53 S .N 880 mo 8% 53 out 82 ... Base 8218803 25. 2:. 2a «a one as. _ $2 835886 138.80 E: 8988326 acaaoeaa a s a a: new 8502 neaoaom saga uefiam 325 38:28.: a man: ”3°35 .850 use a? 885a a €98 56 Bop magenta 382 sum 2: usage“ 3 new .8 333 w 033. 82 2:3 ”a..- $35 2»: $3 :3. 5m ASS 2.7 8882 so: 53 8.7 a? 20m 8:235 .38: .3 25. 53 NS 8388 neoaéo ES 82 8:835 5.8.. 22. 88am 53 m: 38.8 86% 8:88. oz 38 8:835 .28: A33 03. Sea 3.. .8» some Sm- 08835 5.8: 3838 385820 3E5 .885 32¢ mun: “no? E... 8 3: Gang 3. T .8829. 835 8:88. 02 3.5 208335 6:3 5». 993 BE: Boa. ion an; as... 8522 838m ac .23 $3885 a 3.5 .3me 293 Sn- 5% .352 A83 8.” 888a 8% as. on .2 8a 42. N A83 3.- case.— 2»: 6.3 87 @382 2»: £3 23.. 28?: no: 20m 8555 5.8: .8 25. 53 N3; 88:53 .28: Esteem $53 as. am a: message 38.80 a: aoegaao 88.8352 a e a E new 323.80 a 035. 83 2 «3 n: gum :8; an"- ”3 A83 3..“ aoEBoo 29.3.8 Gas «.2 8.8.3 865 388.. oz 38050830 3 SE bowsm 5.3 E .N .8803 omsm So. an .m v3 v8. v $3 a: new A23 “2. 03 A83 8.” 8328 Beard ARE .2 8.8.3 86% 88%: oz 53.850830 mat—5 bow—am A33 3a.- oaoaozéfim 35.35 80$ :._- 838 335 3:38. oz as; 3. 88.35 .28: 20m 38% ageoefio 3 8E >23 55 «S.- 8535 flaw @8838 am: m . .N 8058.. 03¢ «8. g .a Sn «2. m E: afifiufifi 132280 a: 895836 Emosusm m \n m R 95 BEES a gas. 84 A83 a.» $03.5 “58?: .25 A83 26. 580mg: ”imam é m2 A33 8;. o3 A23 3.” 8388 28360 A33 2..” 886% 09% an .m v.2 v“... ES SN. 03 A83 2 .n 83:88 29.830 A83 8.” 325% 03m 5 .m 3.0 08. 889 m: 5m 6:: 3a- a? 289 «ad coma—Eco 29.58.00 :89 2a bungee owsm on ... «.3 go. a: 8EE§36 2280 E: 898§§6 Eugenia \u m k new 326:8 a 23. 85 Table 9 Significance Testing of the Predictors to Total CRF Severity In Persons with Other Cancer Diagnoses Predictor Variable . F -Statistic 4f Stage of Cancer 8.0 4, 228 Co—morbid Condition 6.5 4, 228 Age 2.5 4, 228 PSE for Fatigue Management 47.5 4, 228 Other Symptoms Associated with Cancer & 102.0 4, 228 Cancer Treatment All predictors significant at a level of p < .05 Persons with Lung Cancer Step I The first step, the calculation of the initial model to find the best of all subset model to predict total CRF severity in persons with LC, resulted in a model that was not statistically significant [R7 = .308; F (23, 39) = .756; p = .759]. To improve the model, 8 out of 23 predictors with t-values < -/+ 0.50 were eliminated (see Table 10). Step 2 The elimination of 8 out of 23 predictors fiom the initial model resulted in another model that was not statistically significant [R2 = .294; F (15, 47) = 1.30; p = .2371. Subsequently, six predictors with t-values < -/+ 1.0 were removed (see Table 10). 86 Step 3 The elimination of 6 out of 15 predictors led to an improved model, but still not statistically significant [R2 = .247; F (9, 53) = 1.93; p = .067]. Next, three predictors were removed with t-values < -/+1.2 (see Table 10). Step 4 After the removal of three predictor variables, the model in Step 4 become statistically significant [R2 = .201; F (6, so) = 2.36; p = .042]. This model contained six variables. Having surgery prior to chemotherapy and sex were statistically significant variables to the prediction of total CRF severity. In an effort to improve the model, one predictor variable with the lowest nonsignificant t-value was eliminated (see Table 10). Step 5 With the elimination of one predictor, the model strengthened with an explained variance fi'om 20.1% to 18.3% [F (5, 57) = 2.56; p = .037]. Having surgery prior to chemotherapy and sex remained statistically significant in predicting total CRF severity. Once again, 1 out of 3 of the predictor variables with the lowest nonsignificant t-value was eliminated (see Table 10). Step 6 The elimination of l of 3 nonsignificant predictor variables improved the model in Step 6 [R2 = .165; F (4, 58) = 2.87; p = .031]. Having surgery prior to chemotherapy remained a statistically significant predictor to total CRF severity, but the variable sex converted to nonsignificance. This model had a total of 4 predictors with 3 of the predictors with nonsignificant t-values. Thus, the predictor with the lowest nonsignificant t-value was eliminated (see Table 10). 87 Step 7 The model in Step 7 depicts all three predictor variables in the model as statistically significant, having surgery prior to chemotherapy, sex, and holding a private health insurance policy. The model as a whole was strengthened from Step 6 explaining 15.1% of the variance in the prediction of total CRF severity in persons with LC [F (3, 59) = 3.51;. p = .021]. The predictor variables, PSE for fatigue management and the other unpleasant symptoms were accounted for in the final model as depicted in Step 8 (see Table 10). Further Analysis, Step 8 After PSE for fatigue management and the total severity from the other unpleasant symptoms were added to the model, the explained variance in the prediction of the total CRF severity in persons with LC increased fi'om 15.1% to 34.8% [F (5, 57) = 6.08; p = .000]. In the following order of greatest to least prediction of total CRF severity in persons with LC, four predictors were statistically significant: total symptom severity fi'om the other unpleasant symptoms; PSE for fatigue management; the patient holding an insurance policy; and, having surgery prior to chemotherapy. The predictor variable of sex, females as compared to males predicts greater total CRF severity, became statistically not significant (t = 1.52) (see Table 10). The Final Model The initial model as identified in Step 1 demonstrated good apriori power at .80 to detect relationships with large effects (0.52) while utilizing 24 predictors with a two- tailed test and a level of significance set at 0.05 (Erdfelder et al., 1996). Therefore, at better-than—chance-levels, the final model as depicted in Step 8 for the prediction of total 88 CRF severity in persons with LC accounts for 34.8% of the variance [F (5, 57) = 6.08; p = .000]. This prediction of the actual value of total CRF severity is made with small residual variance. The standard error of the estimate shows that on average, the actual value will be +/- 1.68 of the predicted value 68% of the time, and will be within +/- (2) (1.68) 95% of the time. Tolerance levels of the individual predictors ranged from .916 to .960. Significance testing. As depicted in Table l 1, the significance testing of each individual predictor demonstrates that they not only contribute to the model as a whole but also independently to the prediction of total CRF severity in persons with DC diagnoses. Assumptions Met, the Model Confirmed The mean of the predicted value in the final model of total CRF severity was 5.88 (SD = 1.18). Assessment of the histogram of the standardized residuals shows roughly a normal curve as validated by the range of the residual variance from -2. 16 to 2.15. The normal probability plot shows a monotonic pattern and reasonable closeness of the cases to the regression line, confirming a normal distribution Additionally, the scatterplots of residuals for each statistically significant predictor variable and criterion variable reveals a pattern that is more dense in the center with decreasing density on the outer edges, thus demonstrating normal homoscedasticity. Summary of the Results for Research Question #2 (Persons with LC) Thus, for persons with LC, the following predicts greater total CRF severity and the predictors are in the order of greatest to least strength of prediction: greater total symptom severity from the other unpleasant symptoms; lower PSE for fatigue 89 management; the patient holding the insurance policy; and, having surgery prior to chemotherapy. Note in Step 7 when the two predictor variables, PSE for fatigue management and the other unpleasant symptoms were not accounted for in the model, females as compared to males predicted greater total CRF severity. Gone 35.. 3835 £32. 22. seam 3.3 so. of. as: .V ”2 8:233 £8: 23 acme.— §3 34 88680 2.55-8 3.5 88.35 5.8: 689 m9..- 3628 cacao 38%... oz E3 22.. 2:85 35:88 :35 a #3 8.7 coma; an new A83 2: once one 2 B: 2.55820 3:5 rowan 8S9 o9. fies ace 23 Season. 53 new- as 8.7 £33.. 5%ch Geog EV. roman can can can 623 SN. Baa cocoa 53 cos roman an as» 8.5 Evie—mam 38050830 8 SE bemusm 3.3 at. 3523 83835,. can 53 e. 6.. seasons nonsense 3283. 63 as... 33m 35: 53 «3.. 8&0 no omen 55 8a.- 82 ..- Bass 62:85 ..o 25. as. an am one. 8m. _ 3a acacoenafi 3338 a: moacocaano .8385 a as m am new 85c: noemoaom conceazm chicam 8&5 £838.: a new: 850 was as, «cocoa a easan ..Eo Bop gnomes 382 com ca 3582 3 new no canoe 2 2an 91 A83 8. .- 3.3.... 2o: 3.... an..- 28.8: ...o: .8: 8.8.8. .28: .c 8.... .8. .0 8... 8.2.80 285.8 9.8.. .... 88.8... 5.8.. ...8 8...... a. .... S.. 8m 8.5 88.3... .28: SN. 8..- 883 ...... .8» .83 8. .- 8.8... 8.....80 .88... 8888.0 8.8: 88am a... .V 2.. .83 ...... c. .8: an... m...- 888 8.2.8 3.88. oz a... .0 mm..- 2.28... 8.283. 98.. on... 88.... ...... .8.» Sun Evie—mam 38050820 8 3...: 99.6 .8... .8. 88.8... 8.88:... .26.. an... .8.- ao8§e 8.8.3. 388... :3... S...- ..35 3.82 :2. mm... 8.80... one... SN. 2. ... o... 8.. N .88 R...- oa>.:.. 2o: 6...... w..... ...8.82 ...o: 38.. a... 8.82 ...»: ...o: 88.8... 28:... 2...... an... ...... 8m 3. 8.8.8.888 8.8.80 .3. 8.8.8.888 88.28... a x... ... .... cam Banana 3 03....- 92 55.0 8. ....m ...80 3.. 88.8... 53.... ...o.. 8...... .8. .0 u... 8.....38 ......oaéo 3.... 88.3... ....8: ......0 .....N .08.... ...... .8» .83 ...... .83338: 895880 38...... .98.... ...... 8858...... 38.0 «a. .- .880... 83 N..... on .0 .2 ...... v .83 «8.- 28.... ...o: ... . .0 ...... 8m .....:8......a.......8:... e..... .2008. 8.....88 2.5.8 .83 ...... 8.88... 5.8.. ...... .88.. .23 8..- 9...... ...... .8» 3R— 88§£ .203. 38050830 ms..—5 .0093 ......0 m... ......» ...... o. ...: 88.0 .8." 5...... ...... 8.. .33 3.... ......s... 8.308.. 1.2.580 98.... 99% ...»: 82......qu ...... .0 ..m... 8.80... 88m S... 8 ... 8.. ....u. m ...0. 88888.0 358.80 .5. 8.8.8.850 8.8.22... .. ...: .. ... .....m 3:50:00 .0. 030... 93 389 2d 5m 92: 2 .N 88.35 .28.. 22. 80E 53 8a 99.3 B .8» 35 8335 .28: 3955820 3 sum rumba so. a a :5 E. 5 $3 84 gem A23 «8. 83:58 zeoaéo 33 o: 85:53 £12 22. Boga 389 :a roman as .8» 33 85:85 5.8: aggogo 2 3E 993 So. mm .v 3.“ m2. 0 639 3.” 5m :2 .v «2 82:28 2.58.8 ASS m3 E?» as. .8» 53 a: 35:35 £8; 22. 823 agoafio 9 33 >33 35 8:835 .28: ASS m3- 385% owsm So. S .m 03. 2:. m a: 89.8636 35:50 a: afifigfi Buosga a «e m E 85 tun—H.580 o— 03am. 94 95 A83 wad nag—Em 8383:: .050 68.. and- 8.8882 8&8". .8 mm.— GMS an. 8m 53 8a 88...»... .28.. 22. 80.8.. 6:: 3a .93.... 3.. .8» 85 8:88... £8: xgfioaoao o. 3.5 88.3 8o. R .m 8.8 88. m 3. 88.8835 1.5.280 3. 888.886 882083.. a \u m ... new 3.3.8 o. 2.3 Table 11 Significance Testing of the Predictors to Total CRF Severity In Persons with Lung Cancer Predictor Variable F -Statistic df Had Surgery Prior to Chemotherapy 4.0 4, 58 Patient Held Health Insurance 4.17 4, 58 Sex 2.17 4, 58 PSE for Fatigue Management 5.0 4, 58 Other Symptoms Associated with Cancer & 10.0 4, 58 Cancer Treatment All predictors significant at a level of p < .05 Research Question #3 .° What is the symptom experience of persons with LC and how does it compare to the symptom experience of 0C diagnoses, including the relationships between CRF and other unpleasant symptoms? Out of 16 possible symptoms including fatigue, the total mean number of symptoms reported by persons with cancer at the baseline interview was 7.43 (SD = 2.60). While not statistically significant (t = -.858; df= 87.50; p = .393), persons with LC had a slightly higher mean number of symptoms (M = 7.71; SD = 2.95) as compared to persons with DC diagnoses (M = 7.37; SD = 2.51). 96 Symptom Frequency At the baseline interview, fatigue was the most frequently reported symptom occurring 100% of the time over a period of the past seven days prior to the baseline interview in both persons with LC and 0C diagnoses. While not statistically different (t = ~1.470; df= 296; p = .143), persons with LC had a slightly higher mean number of days offatigue (M= 5.48; SD = 1.81) than persons with OC (M= 5.05; SD = 2.07). After fatigue, the most frequently reported symptoms in the past seven days prior to the baseline interview by all persons were insomnia (77%), lack of appetite (63%), weakness (60%), dry mouth (60%), pain (55%), and nausea (53%) (see Table 12). However, for persons with LC, dyspnea, weakness, dry mouth, lack of appetite, and cough were the next most common symptoms after fatigue and insomnia, whereas persons with 0C reported lack of appetite, weakness, dry mouth, pain, and nausea more frequently occurring after fatigue and insomnia (see Table 12). Diflerences in Symptom Frequency As compmd to persons with 0C, persons with LC reported a greater frequency in the mean number of days of dyspnea (3 days as compared to 1 day; t = -5.47; (9% 296; p = .000) and cough (3 days as compared to 1 day; t = -4.418; df= 296;p = .000). While not statistically significant, note that persons with LC trended towards experiencing greater frequency of weakness as compared to persons with 0C (3 days as compared to 2 days; t = -l.7l; aY= 296; p = .088). However, persons with OC experienced significantly more days of difficulty remembering as compared to persons with LC (1.9 days as compared to 1.2 days; t = 2.07; df= 116;p = .04). 97 Table 12 Rank, Frequency, and Percentage of Symptoms Reported by the Total Sample and by Group in the Past Seven Days fi'om the Baseline Interview Symptom Total Sample Lung Cancer Other Cancer N = 298 N = 63 N = 235 Rank N (%) Rank N (%) Rank N (%) Fatigue 1 298 (100) l 63 (100) 1 235 (100) Insomnia 2 229 (77) 2 44 (70) 2 185 (79) Lack of Appetite 3 187 (63) 6 37 (59) 3 150 (64) Weakness 4 180 (60) 4 4O (63) 4 140 (60) Dry Mouth 5 178 (60) 5 39 (62) 5 139 (59) Pain 6 ' 164 (55) 7 31 (49) 6 133 (57) Nausea 7 159 (53) 7 31 (49) 7 128 (54) Difficulty Remembering 8 133 (45) 10 21 (33) 8 112 (48) Numbness/1‘ ingling 9 124 (41) 9 22 (35) 9 102 (43) Constipation 10 118 (40) 8 24 (38) 10 94 (40) Dyspnea 11 115 (39) 3 41 (65) 12 74 (31) Cough 12 105 (35) 6 36 (57) 14 69 (29) Alopecia 13 100 (34) 9 22 (35) 11 78 (33) Diarrhea 14 93 (31) 9 22 (35) 13 71 (30) Vomiting 15 45 (15) ll 12 (19) 15 33 (14) Fever 16 30 (10) 12 9 (14) 16 21 (9) 98 Presence of Symptoms on All Seven Days For all persons with cancer, the most frequently reported symptoms present on all seven days prior to the baseline interview were fatigue (47%) followed by dry mouth (27%), insomnia (27%), weakness (24%), alopecia (22%), and cough (21%) (see Table 13). The presence of symptoms occurring on all seven days varied by cancer group (see Table 13). For persons with LC, fatigue (52%), cough (38%), weakness (30%), dyspnea (29%), pain (29%), and dry mouth (29%) were the top ranked symptoms occurring on all seven days prior to the baseline interview. Whereas for persons with DC diagnoses, fatigue (45%), insomnia (29%), dry mouth (27%), weakness (22%), alopecia (21%), and numbness and tingling (20%) were reported as the top ranked symptoms occurring on all seven days prior to the baseline interview. 99 Table 13 Rank, Frequency, and Percentage of Symptoms Reported by the Total Sample and by Group On All Seven Days Prior to the Baseline Interview Symptom Total Sample Lung Cancer Other Cancer N=298 N=63 N=235 Rank N(%) Rank N(%) Rank N (%) Fatigue 1 139 (47) 1 33 (52) 1 106 (45) Insormria 3 80 (27) 6 11 (18) 2 69 (29) Lack oprpetite 8 57 (19) 5 15 (24) 7 42 (18) Weakness 4 71 (24 3 19 (30) 4 52 (22) DryMouth 2 81 (27) 4 18 (29) 3 63 (27) Pain 7 59 (20) 4 18 (29) 8 41 (17) Nausea 12 23 (8) 8 8 (13) 11 15 (6) Difiiourty Remembering 10 41 (14) 10 4(6) 9 37 (16) Numbness/Tingling 9 56 (19) 7 10 (16) 6 46 (20) Constipation 13 14 (5) 9 5 (8) 13 9 (4) Dyspnea 11 36 (12) 4 18 (29) 10 18(8) Cough 6 61 (21) 2 24 (38) 9 37 (16) Alopecia 5 65 (22) 5 15 (24) 5 50 (21) Diarrhea 13 14(5) 11 1 (1.6) 12 13(5) Vomiting 14 2 (.7) 11 1 (1.6) 14 2 (.9) Fever 15 1 (.3) 11 1 (1.6) 15 1 (.4) 100 Symptom Severity Out of a total of 16 symptoms, alopecia was the most severe (M = 5.77; SD = 3.24) reported symptom by 32% of all persons with cancer (see Table 14). After alopecia, insomnia (M = 5.40; SD = 2.63), vomiting (M = 5.29; SD = 2.97), constipation (M = 5.26; SD = 2.81), and fatigue (M = 5.23; SD = 2.33) were the next most severe symptoms reported in persons with cancer. Mean Symptom Severity Scores The range of symptom severity scores varied by cancer diagnosis (see Table 14). For persons with OC, the most severe symptom was alopecia (M = 5.67; SD = 3.23), and the least severe symptom was cough (M = 3.61; SD = 2.49). However, the range of the symptom severity scores ran higher for persons with LC, with constipation being the most severe symptom (M = 6.50; SD = 2.67), and difficulty remembering things being the least severe symptom (M = 4.05; SD = 2.42). Consequently, the total symptom severity score for persons with LC was significantly worse (M = 4.99; SD = 1.43) as I compared to persons with OC (M = 4.54; SD = 1.60) (t = -1.99; df= 294; p = .047). Weighted Mean Symptom Severity Scores When assessing. the rank order of the mean symptom severity report, one needs to take into account the mean was computed based on those that reported the symptom. Therefore, the mean severity calculations did not take into consideration the fiequency of occurrence of the symptom relative to the total in the corresponding sample. When the mean symptom severity scores were weighted by frequency of occurrence, fatigue and insomnia were the most severe symptoms in both persons with LC (fatigue MW = 5.38; insomnia MW = 3.79) and OC diagnoses (fatigue MW = 5.10; insomnia MW = 4.31). 101 This is in contrast to the mean symptom severity calculation in persons with LC where fatigue was rated the 7th and insomnia the 5th most severe symptoms, and in persons with OC fatigue was rated the 4th and insomnia the 2m most severe symptom. Similar to the mean total symptom severity score, the weighted mean total symptom severity score remained higher for persons with LC (MW = 5.29) than persons with OC (MW = 4.67) (see Table 15). Dijferences in Symptom Severity As compared to persons with OC, persons with LC had greater symptom severity in pain (t = -2.403; df= 162;p = .017), dry mouth (t = -2.l94; df= .035;p = 58.17) and constipation (t = -2.465; df= 116; p = .015). Persons with LC not only trended towards greater report of frequency of weakness, but also more in the severity of weakness (t = -1.823; af= 178;p = .070) as compared to persons with OC. 102 Table 14 Rank and Mean Symptom Severity of the Total Sample and by Group Symptom Total Sample Lung Cancer Other Cancer N=298 N =63 N=235 Rank M (SD) Rank M (SD) Rank M (SD) Alopecia 1 5.77 (3.24) 2 6.20 (3.32) 1 5.66 (3.23) Insomnia ‘ 2 5.54 (2.63) 5 5.56 (2.15) 2 5.53 (2.74) Vomiting 3 5.29 (2.97) 12 4.83 (3.01) 3 5.46 (2.99) Constipation 4 5.26 (2.81) 1 6.50 (2.67) 6 4.95 (2.78) Fatigue 5 5.24 (2.33) 7 5.47 (2.23) 4 5.17 (2.37) Diarrhea 6 5.10 (2.56) 9 5.27 (2.35) 5 5.04 (2.63) Lack oprpetite 7 5.04 (2.48) 3 5.61 (2.18) 7 4.91 (2.53) Weakness 8 4.84 (2.42) 8 5.45 (2.56) 8 4.66 (2.36) Dry Mouth 9 4.80 (2.60) 4 5.61 (2.63) 9 4.57 (2.55) Pain 10 4.62 (2.27) 6 5.48 (2.23) 11 4.41 (2.23) Dyspnea 11 4.59 (2.14) 11 4.95 (1.99) 13 4.39 (2.20) Nausea 12 4.54 (2.64) 10 5.13 (2.83) 12 4.40 (2.59) Numbness/Tingling 13 4.44 (2.55) 13 4.50 (2.50) 10 4.43 (2.58) Fever 14 4.07 (2.32) 14 4.44 (2.92) 14 3.90 (2.05) Cough 15 3.89 (2.36) 15 4.39 (2.03) 16 3.62 (2.49) Diffimity 16 3.71 (2.44) 16 4.05 (2.42) 15 3.65 (2.45) Remembering Total Symptom - 4.64 (1.58) - 4.99 (1.43) .- 454 (1.60) Severity N= 296 N= 63 N= 233 103 Table 15 Rank and Weighted Mean Symptom Severity Score for Persons with Lung Cancer and Other Cancer Diagnoses Symptom Lung Cancer Other Cancer N = 63 N = 235 Rank My Rank My Alopecia 10 1.97 10 1.83 Insomnia 2 3.79 2 4.31 Vomiting 15 0.92 15 0.77 Constipation 9 2.48 8 1.98 Fatigue 1 5.38 1 5.10 Diarrhea 11 1.84 12 1.48 Lack oprpetite 5 3.21 3 3.09 Weakness 3 3.46 4 2.78 DryMouth 4 3.38 5 2.63 Pain 7 2.70 6 2.50 Dyspnea 6 3.14 13 1.35 Nausea 8 2.52 7 2.38 Numbness/Tingling 13 1.57 9 1.89 Fever 16 0.63 16 0.32 Cough 14 1.00 14 1.00 Dificulty Remembering 12 1.62 1 l 1.62 Total Symptom Severity -- 5.29 -- 4.67 104 Correlations Among Symptoms Correlations were calculated for both the total CRF severity and the individual symptoms, and the total CRF severity and total symptom severity in the total sample and by group (see Table 16). The total symptom severity associated with each symptom was calculated by summing each subject’s response to severity scores for each symptom reported (i.e., a reported symptom is a symptom with a severity score greater than zero) and dividing by the total number of symptoms. For the total sample (r = .512), and for persons with LC (r = .441) and OC (r = .530) diagnoses, the total symptom severity positively significantly correlated with total CRF severity (p = .000). Eleven individual symptoms were also statistically significant in positively correlating with total CRF severity in the total sample and persons with OC diagnoses. Out of the 11 individual symptoms, weakness showed the greatest strength in correlating with total CRF severity for both the total sample (r = .565; p = .000) and persons with 0C (r = .597; p = .000). For persons with LC, five individual symptoms, nausea, dyspnea, lack of appetite, dry mouth, and weakness significantly positively correlated with total CRF severity. These five symptoms not only significantly correlated with total CRF severity in persons with LC, but also for persons with OC diagnoses. Further, similar to the total sample and persons with OC diagnoses, weakness showed the most strength of any of the individual symptoms in positively correlating with total CRF severity in persons with LC (r = .487; p = .001). 105 Table 16 Correlations Among Total CRF Severity, Individual Symptom Severity, and Total Symptom Severity for the Total Sample and by Cancer Group Total Sample Lung Cancer Other Cancer Total CRF Severity Total CRF Severity Total CRF Severity Pain .367" .133 .434“ Nausea .268”I .363“ 245""I Vomiting .181 .566 .091 Insomnia .344” .271 .358“I Dyspnea .314"I .318“ .336” Diarrhea .219" .248 .211 Lack of Appetite .347" .436" .325“ Fever .284 .222 .318 Cough .188 .230 .185 Dry Month .391“ .441" .379“ Constipation .386" .242 .409" Difficulty Remembering .268“ .224 .271" Numbness/Tingling .238” .188 .246‘I Weakness .565” .487“ .597" Alopecia .178 -.159 .251“ Total Symptom Severity .512" .441" .530" Note. *p< .05, 3* p < .01 Summary of Results for Research Question #3 Data indicate a high number of concurrent symptoms in all persons with cancer with a mean of 7.4 symptoms with no significant difi‘erences found between the LC and OC diagnoses groups. Fatigue was the most frequently reported symptom occurring for both persons with LC and OC diagnoses. With no significant differences found between groups, fatigue ranked as the 7th most severe symptom in persons with LC and the 4‘11 most severe symptom for persons with OC diagnoses. However, persons with LC as compared to persons with OC diagnoses had a statistically significant higher total symptom severity score as compared to persons with OC diagnoses. For persons with LC and OC diagnoses, total CRF severity positively significantly correlated with the total symptom severity of the other unpleasant symptoms. Five individual symptoms, nausea, dyspnea, lack of appetite, dry mouth, and weakness significantly positively correlated with total CRF severity for both persons with LC and OC diagnoses, with weakness showing the greatest strength of any individual symptom correlating with total CRF severity. Research Question #4: Does PSE for fatigue management mediate the relationship between CRF severity and PF S in persons with cancer (LC and 0C) ? Mediation was tested using a series of three regression analyses as specified by Baron and Kenny (1986) and Kenny (2005). Further analysis was performed which involved significance testing of the mediation pathway via the Sobel Test (Dudley et al., 2004). For persons with cancer, results indicated that PSE for fatigue management partially mediated the relationship between CRF and PFS (see Table 17). Significant paths were demonstratw from CRF to PSE for fatigue management (b = -.40; p = .000), from CRF 107 to PFS (b = -6.06; p = .000), from PSE for fatigue management to PFS while controlling for CRF (b = 1.81; p = .006), and fiom CRF to PFS while controlling for PSE for fatigue management (b = -5.34; p = .000). Note that the previously direct relationship between CRF and PFS was reduced after PSE for fatigue management was controlled Results of the Sobel Test indicated that this indirect effect was significant (t = -2.59; p = .009). The results support a partial mediation model accounting for 12% of the total mediated efi‘ect. 108 Table 17 Mediation Analysis for the Total Sample (N = 298) Paths b se t p R2 F df CRF to PSE for -.40 .05 -7.32 .000 .15 53.64 1, 296 Fatigue Management CRF to PFS ' -6.06 .62 -9.81 .000 .25 96.30 1, 296 PSE for Fatigue 1.81 .66 2.77 .006 .27 53.07 2, 295 Managanent to PF S while Controlling for CRF CRF to PFS while -5.34 .66 -8.05 .000 .27 53.07 2, 295 Controlling for PSE for Fatigue Management Sobel Test Partial mediation model accounting for 12% of the total mediated effects (1 = -2.59; p = .009). b = Unstandardized regression coefficient. 109 Research Question #5: Does mediation difl‘er in persons with LC as compared to OC when evaluating PSE for fatigue management as a mediator between the relationship of CRF severity and PF S ? Persons with Other Cancer Diagnoses Similar to the total sample, the results from the regression analyses for the OC group indicated that PSE for fatigue management partially mediated the relationship between CRF and PFS (see Table 18). Significant paths were demonstrated from CRF to PSE for fatigue management (b = -.43; p = .000), fiom CRF to PFS (b = -5.93; p = .000), from PSE for fatigue management to PFS while controlling for CRF (b = 1.68; p = .015), and from CRF to PF S while controlling for PSE for fatigue management (b = -5.209; p = .000). The previously direct relationship between CRF and PFS for persons with OC was reduced after PSE for fatigue management was controlled demonstrating partial mediation The results of the Sobel Test indicated that this indirect effect was significant (t = -2.32; p = .020) supporting a partial mediation model accounting for 12% of the total mediated effect. 110 Table 18 Mediation Analysis for Persons with Other Cancer Diagnoses (N = 235) R’ Paths b se t p F df p CRF to PSE for -.43 .06 -7.06 .000 .18 49.78 1, 233 .000 Fatigue Management CRF to PFS -5.93 .64 -9.28 .000 .27 86.04 1, 233 .000 PSE for Fatigue 1.684 .67 2.45 .015 .29 46.96 2, 232 .000 Management to PFS while Controlling for CRF CRF to PFS while -5.21 .70 -7.48 .000 .29 46.96 2, 232 .000 Controlling for PSE for Fatigue Management Sobel Test Partial mediation model accounting for 12% of the total mediated effects (t = -2.32; p = .020). b = Unstandardized regression coefficient. 111 Persons with Lung Cancer After performing the three regression analyses as specified by Baron and Kenny (1986) and Kenny (2005) for persons with LC, the results indicated that PSE for fatigue management partially mediated the relationship between CRF and PFS (see Table 19). Significant paths were demonstrated from CRF to PSE for fatigue management (b = -.26; p = .033), from CRF to PFS (b = -6.60; p = .000), from PSE for fatigue management to PFS while controlling for CRF (b = 3.70; p = .017), and from CRF to PFS while controlling for PSE for fatigue management (b = -5.62; p = .000). The previously direct relationship between CRF and PFS was reduced after PSE for fatigue management was controlled demonstrating partial mediation. The Sobel Test showed that this indirect effect was not significant (t = -1.63; p = .104). However, the results of the Sobel Test are tenuous since the LC group had a shortfall of the number of cases recommended for calculating this test, 63 persons with LC as opposed to 200 persons. The Sobel Test is calculated from the unstandardized regression coefficients and standard errors of these coefficients from the path evaluating CRF to PSE for fatigue management and the path evaluating PSE to PFS. The observed power to detect an effect within these two paths for the total sample (N = 298) and for the OC group was statistically powerful, with an observed power of 1.0. Conversely, for the LC group (N = 63) the observed power to detect an effect was lower, with the observed power in the path from CRF to PSE for fatigue management .574 and the observed power in the path from PSE to PFS .893. 112 run—H . Summary of Results for Research Question #5 For the LC and OC groups separately, the hypothesis of mediation from CRF to PFS through PSE for fatigue management was tested showing significant support for partial mediation. Table 19 Mediation Analysis for Persons with Lung Cancer (N = 63) Paths b se t p R2 F df CRF to PSE for -.26 .12 -2. 182 .033 .07 4.76 Fatigue Management CRF to PFS -6.60 1.49 -4.43 .000 .24 19.62 PSE for Fatigue 3.70 1.52 2.44 .017 .31 13.60 Management to PFS while Controlling for CRF CRF to PFS while -5.62 1.49 -3.78 .000 .31 13.60 Controlling for PSE for Fatigue Management Sobel Test t= -1.63; p = .104 1, 61 l, 61 2, 60 2,60 .033 .000 .000 .000 b = Unstandardized regression coefficient. 113 Research Question #6: What is the unique contribution of physiological and contextual characteristics, CRF, other unpleasant symptoms, and PSE to the PF S of persons with cancer (LC and 0C)? Hierarchical multiple regression analysis was performed to examine how most of the variance in PF S can be explained by blocks of independent variables, over and above that explained by an earlier block of independent variables. The theoretical fi'amework for this dissertation project guided the statistical model and the decision-making of when and which independent variables were to be entered into the hierarchical model. In sequential order, the independent variables were entered in six separate blocks: M: Contextual Patient Characteristics [race; marital status; level of education achieved; employment data including whether a person was retired, receiving disability, was on a temporary leave from employment, and whether they had to quit employment; annual combined household income; health insurance data including whether or not a person with cancer had health insurance and if so, who held the policy (patient or spouse), and the type of health insurance policy held (private, Medicare, or Medicaid)]; _1_3_l_9_9_l_<__I_I_: Physiological Patient Characteristics [type of cancer; stage of cancer", co-morbid conditions; age; and sex]; M: Type of Treatment Requiring Statistical Control [receiving radiation therapy treatment; surgery prior to chemotherapy which includes four groups (yes had surgery; don’t know if had surgery; this response choice was not selected; no surgery and this group served as the reference group when the groups were dummy coded, i.e., group left out); surgery during chemotherapy which includes three groups (yes had surgery; this response choice was not selected; no surgery and this group served as the reference group when the groups were dummy coded, i.e., group left out)]; 114 Block IV: PSE for Fatigue Management; M: Other Unpleasant Symptoms Total Severity; and W: Total CRF Severity. Impact of Variables on PFS for Persons with Cancer The results of hierarchical multiple regression analysis is shown in Table 20. An apriori power analysis was conducted showing that when all 28 predictors were entered into Block V1 with a power set at 0.80 and an alpha level at .05, small effects (.09) could be detected in a sample of 298 persons with cancer. Block I (Contextual Patient Characteristics). The first block of the independent variables, contextual patient characteristics, were entered to create Model 1. These variables explained 10.4% of the variance in PPS of persons with cancer [F (14, 281) = 2.32, p = .005]. Receiving disability was the only contextual patient characteristic that demonstrated a statistically significant relationship with PFS. Consequently, persons receiving disability predicted a lower PFS. Block [1 (Physiological Patient Characteristics). When the physiological patient characteristics were entered in the second block, type of cancer and co-morbid conditions demonstrated a statistically significant relationship with PFS, and the percent of explained variance increased by 13.7% [F (5, 276) = 10.03, p = .000] to 24.1% [F (19, 276) = 4.62, p = .000]. None of the contextual patient characteristics made a significant contribution to PFS in this second model. Thus, in Model 2, persons with LC as opposed to OC diagnoses, and persons with greater co-morbid conditions predicted lower PFS. Block 111 (Type of Treatment Requiring Statistical Control). Type of cancer and co- morbid conditions remained statistically significant in predicting lower PFS after entry of the third block of variables. There was no statistically significant increase in the percent 115 of explained variance of PFS by 1.9% [F (6, 270) = 1. 12, p = .353] to 26% [F (25, 270) = 3.79, p = 000]. None of the treatment variables requiring statistical control entered in the third block were statistically significant in predicting PFS. Block W (PSE for Fatigue Management). Entering PSE for fatigue management in the fourth block resulted in a statistically significant positive relationship with PFS. Also found in the fourth block was a model which predicted lower PFS for persons who had surgery prior to chemotherapy. Type of cancer and co—morbid conditions also remained statistically significant predictors to lower PF S as well. From Model 3 to Model 4 there was a 9.1% [F (1, 269) = 37.64, p = .000] increase in the explained variance of PFS to 35.1% [F (26, 29) = 5.59, p = .000]. Block V (Other Unpleasant Symptoms Total Severity). When the other unpleasant symptoms as measured by their total symptom severity score was added in the fifth block to create Model 5, the variables that were in Model 4 continued to remain statistically significant. Additionally, the results showed that the greater the severity of the other unpleasant symptoms, the lower the PFS. The explained variance in PPS in Model 5 increased by 6.9% [F (1, 268) = 31.88, p = .000] to 42% [F (27, 268) = 7.18, p = .000]. Block VI (Total CRF Severity). The sixth and final model commenced with the addition of total CRF severity and resulted in an increase in the explained variance of PFS by 5.7% [F (1, 267) = 29.04, p = .000] to 47.7% [F (28, 295) = 8.68, p = .000]. Like Models 4 and 5, the following variables continued to remain statistically significant in the sixth and final model: type of cancer, co-morbid conditions, having surgery prior to chemotherapy, PSE for fatigue management, and the total symptom severity of the other unpleasant symptoms. 116 Summary of the Results for Research Questions #6 Through blockwise, hierarchical multiple regression, similar levels of CRF severity were found to significantly worsen the PFS of persons with LC as compared to OC diagnoses (t = -3.78). In addition to type of cancer diagnoses, five other factors in the total sample were identified through blockwise, multiple hierarchical regression as the most important factors accounting for 47. 7% of the explained variance in PFS [F (28, 295) = 8.68, p = .000]. Specifically, higher levels of PSE for fatigue management (t = 3.55) were found to be one of the strongest predictors of greater PFS, while lower levels of PFS were predicted by greater total CRF severity (t = -5 .39), greater number of co-morbid conditions (t = -4.20), greater total symptom severity (t = -2.46), and having surgery prior to chemotherapy (t = -2.31). 117 so. on.- o? m. .- 2.. E.- 828 .26.. seesaw-88:82 .28: N... 2. 8.- 2.. on. 3.- 8:8 .22 2.888882 .28: 8.- 3.- R. S. 2.. e. .2 88.8... 3888-8882 .28m 8. so. 2. cm. mm. 3.. 6882 22.88.. 82.2.8 82.5.. o. .. 2.. n. .. e... .... 2. one; 2.... 8 8m assignam 8.- 2.. N. .- 2. on. 2..- one: .88 381888282828”... 8.. 8.. a... o. .. 3.. 3.." 82.82 antenna-888825 so. 2 .- 2.. 8.- a. .- 2.. 8.8m..8§o_nam m: 8.. 8.. S.. 8.. 82 8828 88888 .23 am. 3.. .2 CS 2.. on. 388 8.82 2... .2 e.. t.. E. 2: 8.2 5 82m > 285 2 285 E 285 a 262m 2 262m 282 2F 2 8228.» 3.56580 .828 cm 8on 3mm Sea mos—«>4 6% is .880 .23 3.88.. co 32% 38.385 33923 no 838.28» cocoa—om mo flop-mm noun-H. ofi me 29:92 mommy—mom baa—=2 Roach-Boa a «o 338% cm Baa-H. 118 a. me. on. G. 888.. 83888 22.2. 88% 2 .7 3w.- oer Y“.- 8828 82.68qu 8 more >8me 2. 2. 2..- G.- 325 ream-cacao 3 SE bomb—m gm."- nmdi med. em; .. ©0332 weer-2:05 2 SE 3.6me S. 8N. mm. 8.- .8865 newsman 2.388: mm... m:- 2- 2:- 2:- 8m 8.:- mm.. on- 3.“.- t..- om... 8.... m5.- G.... 8...- 36- 882.8 2288.8 :.. on. S. S.- R.- .ooeao .8 83 2.....- 32- e..-....- 38. 8....- 886 we 25 m... an. on. 8. 2. 8: 26.: 6:8 8.382.888... .28: mm. 8. mm. 2.. we. 8.- 26.. 8:8 682628.88... .28: E. 2.. on. a: mo: 2... 26.. .828 6822885.... .28: 5 8.85 > 285 >~ Moo—m E 285 n :85 H goo—m .682 6.; 2 82825 832.60 8 use... 119 .Amo. v 3 «anomawmm 93 Son 5 mos—ax, $5... $9 33...- .\.£ .33” $12 R B 88E? 8:335 mo 382 393:30 3.....- bcga 56 Bob 2&- . m3.- bfizm .33 333% $3.5 550 and 3m :6 .83; 2&3 é mmm 2 .- a..- a..- 2 ._- 38.3. 82.8.20 3.96 Emsm S “.85 > “.88 Z x85 5 x85 a #85 H #85 $82 2; a 83%; 6035280 om oSd-H 120 Research Question # 7: Considering the unique contribution of physiological and contextual characteristics, CRF, other unpleasant symptoms, and PSE for fatigue management to PF S, how does having a cancer diagnosis of LC compare with DC diagnoses as a predictor of PFS? * In Research Question #6 it was found that after entering the block of physiological patient characteristics from Model 2 through Model 6 that the type of cancer diagnoses (LC versus 0C diagnoses) made a statistically significant contribution to the explained variance of PFS. Specifically, it was found that having LC as compared 0C diagnoses predicted lower PFS. As a result of this finding, hierarchical multiple regression analysis was performed separately on groups of cancer patients, those with LC and those with DC diagnoses. The procedures used for hierarchical multiple regression analysis with the total sample was also used for persons with LC and OC diagnoses. Impact of Variables on PF S of Persons with DC Diagnoses An apriori power analysis was conducted which revealed that in Block VI small effects (.11) could be detected with a sample size of 235 persons with DC diagnoses, the use of 27 predictors from an alpha level of .05 with power set at .80 (Erdfelder et al., 1996). The results of hierarchical multiple regression analysis is shown in Table 21. Block I (Contextual Patient Characteristics). Model 1 was not statistically significant with the entry of the fust block of variables, the contextual patient characteristics [F(14, 218) = 1.48,p = .119]. Block II (Physiological Patient Characteristics). Unlike Model 1, with the addition of the second block of variables, Model 2 was statistically significant to explain 18.5% [F (18, 214) = 2.70, p = .000] of the variance in the PFS of persons with 0C diagnoses. 121 Co-morbid condition was the only variable out of the contextual and physiological patient characteristics that demonstrated a statistically significant relationship with PF S. Thus, the greater the number of co-morbid conditions, the lower the PFS of persons with DC diagnoses. Block 111 (T we of Treatment Requiring Statistical Control). With the addition of the third block of variables which created Model 3, there was not a statistically significant increase in the explained variance [F (6, 208) = .815, p = .560]. None of the new variables entered, the type of treatment variables requiring statistical control, were statistically significant in predicting PFS. Co-morbid condition sustained statistical significance from Model 2 to Model 3 in accounting for 20.3% of the variance in PFS of persons with DC diagnoses [F (24, 208) = 2.21, p = .002]. Block IV (PSE for Fatigue Management). With the entry of the fourth block, PSE for fatigue management, there was a significant increase by 9.4% [F (1, 207) = 27.84, p = .000] in the explained variance of PFS in persons with DC diagnoses fiom Model 3. Consequently, Model 4 explained 29.8% [F (25, 207) = 3.51, p = .000] ofthe variance in PPS in persons with DC diagnoses. Specifically, in Model 4, PFS was found to be lower in persons who had a greater number of co-morbid conditions, and higher in persons with greater PSE for fatigue management. Block V (Other Unpleasant Symptoms Total Severity). With the addition of the other unpleasant symptoms as measured by the total symptom severity score in the fifth block to create Model 5, the variables that were in Model 4 continued to remain statistically significant. Additionally, the results showed that the greater the severity of the other unpleasant symptoms, the lower the PFS. The explained variance in PFS in Model 5 122 increased by 7.4% [F (1, 206) = 24.37, p = .000] from Model 4 to 37.2% [F (26, 206) = 4.70, p = .000]. Block VI (Total CRF Severity). In the final model, Model 6, with the addition of the sixth block consisting of the variable total CRF severity, the explained variance of PFS increased by 6.3% [F (1, 205) = 22.71, p = .000] to 43.5% [F (27, 205) = 5.84, p = .000]. Similar to Models 4 and 5, the following variables continued to remain statistically significant in the final model: co-morbid conditions and PSE for‘fatigue management. The other unpleasant symptoms as measured by the total symptom severity was not statistically significant t = -1.96, p = .051). 123 e.. 8.- on- a. .- a. .- 8. .- 8...... 8.2. omega-885.... .28: cm. 8. 8.- .... .... 2.- 8...... 8.2. Ewan-88.8... .28: a..- ....- om. an. em. 8. 885...... newsman-885...... .28: .N. 8.- 3.- mm. 8. R. neon... 22.88.. 8.2.2.8 .35.... .m. 8.. an. on. 8.. on. .83 ...... a. ...: .8553: an. E. . R. E. .... S. . .83 so... 88. 588....8Ee...am m... 8. 8. em. a... N.. 2.28... 83.82.8323: ... a. 8. mm. mm. mm. 8.8:-.8Eo.aam a... a... 8.. e5 .2 .n. 88.8.. 8.888 ..e .28. m... 8.. o... N... 8.. .e. 8a.. .382 8.. . "N.. S. 3.. a... 2. 38. .> 28.: > U.82: >. V.8... ... 28.: .. 28.: . U.22: .28.). e..-w ... 8228> 3:20E30 «.328 3. 28m 33. :8... 3:13.. Ammm n 2. 88:25 .350 .050 5.3 2.8.0: ..o 253m 3552...... 30.93.. .8 832:; 382% we flue-mm .80.-mo..- o... ..o 2ng ce..-6&3. 29:32 30.58805 5 me 338% .N 035. 124 on.- 3.. me. mm. mm..- we. 3.. 3 .T unfit mn- ow... an. 3.. am.- me. $- mn- no.7 an. N..—- mm.- ran-n. NN. 2... am. mm.- co..- 3. we. on... man..- mm. 2... #0.. «ad. ma.- 2... a... 2. G..- 2. vw. mm.- 3... t.. 2..- N...- :4... an.- 5. MN; co. mm...- 3.- and. 8.- to. coo. 2: No. on. NN. v0.8.8 .02-2.8.... ms..... bow-5m .8380. mo>iofioao 2.2.... 2.525 cote—om .0262.vo o. .3... bowam Bo...— ruoaéaoao 8 .3... EwSm 3330. mo>iofivo 8 8.2.. .535 .88... 88...... 2.288.. xom ow... macaw—2.8 29.08.60 .088 ..o 03¢ 22. 8...... eaoeozooaaaa... .28: 2a.. 8...... 28.8.2888... .28: 22. 8...... neat-38.8... ....8: 5 :85 > :85 a .305 a .305 m :85 2.85 .032 2E. 5 33.3.3, 82.2.8 .N 2......- 125 .3. v s .8828... 2.. 2.... ... 8...... $3... $2... 2...... 8...... $3. 8... ... 888...... 8.8.98... .822. 9.2.2.26 R... 8.88.. ...... .5... ca..- 3.... 8.55... .89 8.328% .5329... .850 m..." and ”a... .8583... 22...... 8.. mm: .> .85 > 28:. >. 28:. ... 285 .. 202m . 28:. .252 2.... ... 838...; 90.0330 .828 9m 0.8m 33. So... 838... 8......80 .N 8...... 126 Impact of Variables on PFS of Persons with LC Unlike the hierarchical models produced in the total sample and persons with DC diagnoses, the sample size for persons with LC was substantially smaller. As a result, the apriori power and subsequent ability to detect effects was devalued With the inclusion of 26 predictors, a power set at .80 and an alpha level at .05, a large effect of .57 can be detected with a sample size of 63 (Erdfelder et al., 1996). Nonetheless, a decision was made to continue with the hierarchical multiple regression modeling analysis knowing that the lower sample size would contribute to a lack of precision. The results of hierarchical multiple regression analysis for persons with LC is shown in Table 22. Block I - Block 111 (Contextual and Physiological Patient Characteristics, and Type of Treatment Requiring Statistical Control). The blocks of variables found in Model 1 [F(13, 49) = 1.28,p = .256], Model 2 [F(17, 45) = 1.33,p = .218], and Model 3 [F (23, 39) = 1.71, p = .069] were found to have no overall relationship with the PFS in persons with LC. Block IV (PSE for Fatigue Management). With the entry of the fourth block, PSE for fatigue management, Model 4 explained 36.3% of the variance in PFS in persons with LC [F (24, 38) = 2.48, p = .000]. Specifically, this model predicted lower PFS for persons who were Caucasian as compared to other races; persons who said they had surgery prior to chemotherapy; and, persons who did not report whether or not they had surgery prior to chemotherapy. This model also predicted greater PFS in persons who reported having a private or Medicaid insurance policy; and for those persons who had greater PSE for fatigue management. Unlike the models produced for the total sample and persons with 127 0C diagnoses, these findings which impact PF S are new with one exception, the influence of PSE for fatigue management. Block V (Other Unpleasant Symptoms Total Severity). When the other unpleasant symptoms as measured by the total symptom severity score was added in the fifth block to create Model 5, the variables that were in Model 4 continued to remain statistically significant. In Model 5, lower PFS was also predicted if a person had a greater number of co—morbid conditions, was receiving disability, had to quit work, or had a greater total symptom severity score. Similar to the total sample and persons with DC diagnoses, the other unpleasant symptoms experienced by persons with LC were statistically significant in predicting PFS in Model 5. Consequently, the variables in Model 5 account for 48.3% of the variance in PPS in persons with LC [F (25, 37) = 3.31, p = .000]. Block VI (Total CRF Severity). In the final model, after total CRF severity was entered to create Model 6, the following variables continued to remain statistically significant since Block 4 in predicting PFS in persons with LC: race; holding a private or a Medicaid insurance policy; persons who did not report whether or not they had surgery prior to chemotherapy; and PSE for fatigue management. Unlike the total sample, the other unpleasant symptoms (t = -1.69) and total CRF severity (t = -1.87) were not significant variables in predicting PFS in persons with LC. Model 6 accounted for 51.5% of the variance in the prediction of PFS in persons with LC [F (26, 36) = 3.53, p = .000). Summary of the Results for Research Question # 7 Perceived self-efficacy for fatigue management significantly contributed to greater PFS in both persons with LC and OC diagnoses. A greater number of co-morbidities and a greater total severity of the other unpleasant symptoms were found to lower PFS in both 128 groups, persons with LC and 0C diagnoses. Greater cancer-related fatigue severity was found to predict lower PF S only in persons with DC diagnoses. The LC group had race, insurance, and employment related variables identified as predictors of PFS that were unique to this group alone. 129 ma." 5...». 34 a...“ v: 8." so: 3:8 035.8853 .23: «2. OS. :4- 3.- :.- 8.- 3:8 £2. oaaméasma £8: a. on. on- 2. 9.. E. 3:8 fiatuofiaaosaaa 5.8: a. 8. E. «m. 8.- 3.. 2:85 238:3 35:88 12:2. 5.7 S..“- ”E. 8.7 84. n2- as; :3. a B: ..8853m a..- an- E . 7 3w- 8.7 «5. as; 89a 28. Songhaeioaam S; S." m2 mg a: a." 338% gfiéafiififim mm. mm. mm. 3. S.- 2 .- Biomafiaaoaam 2.. we. 2. 2.- t .- «N.- 3:38 8E8? .3 33 S. . 8. ma. 2: 8. 3. mean 3%: a." E." 3." $4 5: 8." 8am 5 v.85 > 385 2 :85 5 #85 = #85 a x85 332 2p 5 Egg, $56580 comma 0% 28m 23m Sea 3:33 Ame u 3 8230 mesa 55 885m we £53m 385:5 323.5 no "flu—nuts, vogue—om .3 floohm 338. 05 mo main—SE :BmmboM 29232 3032805 a .«o 333% NN 2an 130 we. NN. mo. VN. :6. 2.0. 2: an. 8.7 2.7 3. no. 7 3. 5. mo. 3. Nb. 7 8."- 3. . me. an." cm." 3.7 3.7 hm. 9. agn- mm. 3‘."- N~.7 S. 8.- 8.7 mm. nnfi 3.- 2 .- 8; en."- 5. 5.7 v.7 E .- 57 S .- mm; and- 8.7 me. an.- as; St.— mu.“ 3.7 3... ms.- 3828 82.0805 mat—6 >6me @0388 8.723% mg @035 @3028 82.0805 8 3E >8me 305 #896823 3 BE ©0me 3332 8.70826 3 8E Ewe-5 355 8:52 méaoom now ow< 886:8 2808.00 8058 mo 03% 22 $8 afieozéga is: 23 £8 oaofizéaéa sac: H> Moo—m > goo—m a goo—m gflowwooo gamma 0% Boom Bum 88m mos—SE a x85 fl moo—m 3.02m Ego—2 2E. E 833.5» 82:80 mm 2.3 131 A3. v 3 68mg“ 2.. 28 a .33, k 33?. 33m $3.. $3; 5.8 $3 $3 a: 88E, 3396 .8 385a ogaaao 5.7 55,8 56 as... am. 7 . and- 5.536 .32 8.89:? «Sumac—mg 860 a." 3.... m3 aoaowafia 2&3 3.. m2 S “—85 > :85 a moo—m E x85 = “—85 :85 .032 2:. E 83523 way—06580 nommmflwom Boom 3mm 88m 8395 82:80 2 2.3 132 Research Question #8: Through the employment of a Path Model, is the PF S of persons with cancer (LC and 0C) predicted through physiological and contextual patient characteristics, CRF, other unpleasant symptoms, and PSE for fatigue management? Testing of the exogenous-endogenous initial model started with 16 exogenous variables for the prediction of total CRF severity. The 16 exogenous variables included physiological patient characteristics (type of cancer; stage of cancer; surgery prior to chemotherapy; surgery during chemotherapy; co-morbid conditions; sex; and age) and contextual patient characteristics (marital status, level of education, employment data including whether a person was retired, receiving disability, and whether they had to quit employment; annual combined household income; and the type of health insurance policy held [private, Medicare, and Medicaid]). The model was initially tested as conceptually depicted in Figure 2 to examine the overall fit. While a converged solution was obtained, the fit of the model was not acceptable (Satorra-Bentler Scaled Chi-Square 185.5; p = 0.00; df= 50; RMSEA = .098). Driven by model fitting measures which included evaluation of parameter estimates, modification indexes, goodness-of-fit—tests, and following the theoretical framework of the study, modifications were made to attain a parsimonious final solution First, the model was simplified by trimming 13 out of 16 nonsignificant paths one at a time fi'om the exogenous variables to the total CRF severity. This resulted in three significant paths, the effect of age (t = -2.01), co-morbid conditions (t = 3.14), and sex (t = 2.42) on the total CRF severity. Two insignificant paths were retained, the effects of stage of cancer (t = 1.95) and having surgery prior to chemotherapy (t = 1.60) on the total CRF severity since the elimination of their effects revealed that the fit of the model was 133 significantly worse without the two paths. Subsequently, the model fit was improved (Satorra-Bentler Scaled Chi-Square = 116. 14; p = 0.00; af= l7; RMSEA = .14; CFI = .76). Next, the effect of having surgery. prior to chemotherapy on the total severity of the other unpleasant symptoms was added based upon model fitting measures, particularly the modification index (5.93). The addition of this effect, having surgery prior to chemotherapy (t = ~2.67) on the total severity of the other unpleasant symptoms improved model fit (Satorra-Bentler Scaled Chi-Square = 109.50; p = 0.00; df= 16; RMSEA = .14; CFI = .77). Furthermore, the initial model depicted the total severity from both CRF and the other unpleasant symptoms as non-recursive. The effect of total CRF severity on the total severity of the other unpleasant symptoms was significant (t = 2.07), but the effect of the total severity of the other unpleasant symptoms was not significant (t = 1.16). However, the model fitting measures indicated that when eliminating the effect of the other unpleasant symptoms on CRF while retaining the effect of CRF on the other unpleasant symptoms, the model fit was improved (Satorra-Bentler Scaled Chi-Square 109.5; p = 0.00; df= 16; RMSEA = .14; CFI = .77). Consequently, the non-recursive relationship between total CRF severity and the total severity of the other unpleasant symptoms was modified to a single direct path from total CRF severity to the total severity of the other unpleasant symptoms (t = 9.69). Last, model fitting measures indicated an improved model fit with the addition of three paths. First, the effect of total CRF severity on PFS was added based upon a high level modification index (54.16) which resulted in improvement of the model fit 134 (Satorra-Bentler Scaled Chi-Square 64.73; p = 0.00; df= 16; RMSEA = .10; CFI = .88). Adding this path from total CRF severity on PF S significantly contributed to the model (t = -8.01). The second path, the effect of the total symptom severity on PFS, was added based upon a high level modification index (6.72). The addition of this path (I = -2.69) also improved the model fit (Satorra-Bentler Scaled Chi-Square 58.49; p = 0.00; df= 15; RMSEA = .10; CFI = .89). The last path, the effect of co-morbid conditions on PFS, was added based upon a high level modification index (40.26). The addition of this significant path (t = -7.47) resulted in the final parsimonious model (Satorra—Bentler Scaled Chi- Square 17.76; p = .22; df= 14; RMSEA = .03; the lower bound 90% CI = 0.00; and CFI = .99) (see Figure 3). Consequently, the direct paths in the final model demonstrate the following for persons with cancer (LC and 0C diagnoses): younger age (t = -2. 18), greater co—morbid conditions (t = 3.36), and being female as compared to male'(t = 2.11) predicts greater total CRF severity. Having surgery prior to chemotherapy (t = -2.85) predicts greater total severity from the other unpleasant symptoms. Greater total CRF severity predicts both greater total severity fi'om the other unpleasant symptoms (t = 9.69) and lower PSE for fatigue management (t = -7.02). Greater PSE for fatigue management predicts greater PFS (t = 2.87). Lastly, lower PFS was predicted by greater number of co-morbid conditions (t = -7.47), greater total CRF severity (t = -5.30), and greater total severity from the other unpleasant symptoms (t = -2.71) (see Figure 3). 135 35m unease 323$ Eon—ow; ozwuam 5 80833 33023 mash—Sam 3883:: 550 05 89a mow—gum 33. tow—gem osmnem 83345050 33. 2880 a. 3032335 a. 8:39.850 Sousa .\ $333 53 com 382 339250 .N enema 136 mo. .I. E0 98 Mcod .I. H0 $3 953 .532