Biobehavioral and Environmental Influences on Sleep in Children with Cancer

Disturbed sleep-wake patterns are prevalent in up to 40% of children with cancer; however, sources of sleep-wake pattern disturbances are not well understood. Light and noise disrupt sleep-wake patterns in intensive care settings. Cancer-related symptoms, including fatigue, may impair sleep. Individual temperament characteristics influence symptoms in acutely and chronically ill children. Polymorphisms in the dopamine D4 receptor gene and regulatory region of the serotonin transporter promoter gene are associated with behavioral variation and may influence the child's response to physiologic and environmental stimuli. The purpose of this multiple-case study was to examine sleep-wake patterns and the influences of environmental and biobehavioral factors on sleep-wake patterns among children with cancer. The UCSF Symptom Management Model was the conceptual framework. Participants were 15 school-age children with cancer receiving inpatient chemotherapy. Data were collected during an admission lasting 3 days or longer. Wrist actigraphs and sleep diaries measured sleep. A data logger and sound pressure level meter measured environmental variables. Fatigue was measured using Fatigue Scale: Child and Parent Versions. Temperament was measured using Carey Temperament Scales. Polymerase chain reaction was used to genotype polymorphisms. Data analysis included statistical and visual graphical analyses. Children with cancer slept less and had more fragmented sleep compared with age-related norms. A basic mixed linear model identified a significant main effect for epoch (F = 56.27, g < .01) on sleep minutes within a night shift. Main effects for sound (F = 50.87, p < .01) and light (F = 7.04, p < .01) on sleep minutes were present. A backwards multiple regression model contained five variables (sound, light, number of medications, pain, and nausea) Accounting;; for 57.4% of the variance in sleep minutes within each 2-hour epoch of a 12-hour night shift (F = 62.85, p < .01). Children with less predictable biologic cycles slept longer and had fewer awakenings. Children homozygous for short serotonin polymorphisms, associated with increased anxiety, had poorer sleep efficiency. Clinicians should seek opportunities to minimize nighttime disruptions. Research is warranted to develop and test interventions to promote sleep in the hospital and investigate the role of intraindividual factors on children's cancer-related symptoms.

BIOBEHAVIORAL AND ENVIRONMENTAL INFLUENCES ON SLEEP IN CHILDREN WITH CANCER by Lauri Ann Linder A dissertation submitted to the faculty of The University of Utah in partial fulfillment of the requirements for the degree of Doctor of Philosophy College of Nursing University of Utah August 2009 Copyright © Lauri Ann Linder 2009 All Rights Reserved THE UNIVERSITY OF UTAH GRADUATE SCHOOL SUPERVISORY COMMITTEE APPROVAL of a dissertation submitted by Lauri A. Linder This dissertation has been read by each member of the following supervisory committee and by majority vote has been found to be satisfactory. Chair: BeckyfJ. Ohrhrti J 0 Susan L. Beck Sandra Smith Jj^p q, aooq AtlulJ a^ . CLL^CL^ Pamela Hinds /^iLJQ0! /Yvel THE UNIVERSITY OF UTAH GRADUATE SCHOOL FINAL READING APPROVAL To the Graduate Council of the University of Utah: I have read the dissertation of Lauri A. Linder ;n jts f m a j form and have found that (1) its format, citations, and bibliographic style are consistent and acceptable; (2) its illustrative materials including figures, tables, and charts are in place; and (3) the final manuscript is satisfactory to the supervisory committee and is ready for submission to The Graduate School. Chair: Supervisory Committee Approved for the Major Department Maureen R. Keefe Chair/Dean Approved for the Graduate Council David S. Chapman Dean of The Graduate School ABSTRACT Disturbed sleep-wake patterns are prevalent in up to 40% of children with cancer; however, sources of sleep-wake pattern disturbances are not well understood. Light and noise disrupt sleep-wake patterns in intensive care settings. Cancer-related symptoms, including fatigue, may impair sleep. Individual temperament characteristics influence symptoms in acutely and chronically ill children. Polymorphisms in the dopamine D4 receptor gene and regulatory region of the serotonin transporter promoter gene are associated with behavioral variation and may influence the child's response to physiologic and environmental stimuli. The purpose of this multiple-case study was to examine sleep-wake patterns and the influences of environmental and biobehavioral factors on sleep-wake patterns among children with cancer. The UCSF Symptom Management Model was the conceptual framework. Participants were 15 school-age children with cancer receiving inpatient chemotherapy. Data were collected during an admission lasting 3 days or longer. Wrist actigraphs and sleep diaries measured sleep. A data logger and sound pressure level meter measured environmental variables. Fatigue was measured using Fatigue Scale: Child and Parent Versions. Temperament was measured using Carey Temperament Scales. Polymerase chain reaction was used to genotype polymorphisms. Data analysis included statistical and visual graphical analyses. Children with cancer slept less and had more fragmented sleep compared with age-related norms. A basic mixed linear model identified a significant main effect for epoch (F = 56.27, g < .01) on sleep minutes within a night shift. Main effects for sound (F = 50.87, p < .01) and light (F = 7.04, p < .01) on sleep minutes were present. A backwards multiple regression model contained five variables (sound, light, number of medications, pain, and nausea) accounting for 57.4% of the variance in sleep minutes within each 2-hour epoch of a 12-hour night shift (F = 62.85, p < .01). Children with less predictable biologic cycles slept longer and had fewer awakenings. Children homozygous for short serotonin polymorphisms, associated with increased anxiety, had poorer sleep efficiency. Clinicians should seek opportunities to minimize nighttime disruptions. Research is warranted to develop and test interventions to promote sleep in the hospital and investigate the role of intraindividual factors on children's cancer-related symptoms. v This dissertation is lovingly dedicated to my husband, Wayne, for his endless devotion and support. It seems only fitting that we are celebrating this milestone during the same year as our 20th anniversary. TABLE OF CONTENTS ABSTRACT iv LIST OF TABLES xi LIST OF FIGURES xv ACKNOWLEDGEMENTS xvii Chapter 1. INTRODUCTION 1 Significance of the Problem and Specific Aims 1 Study Purpose 2 Specific Aims 2 Review of the Literature 3 Overview of Childhood Cancer 3 Sleep Physiology in Children and Consequences of Disturbed Sleep 4 Evaluation of Sleep-Wake Disturbances 6 Sleep and Children with Cancer 7 Sources of Disturbed Sleep in Acutely and Chronically 111 Children 9 Temperament and Children's Responses to Illness and Symptoms 11 Genetic Polymorphisms Associated With Variability in Temperament.... 13 Conceptual Framework 15 Summary of Study Background and Significance 15 2. ANALYSIS OF THE UCSF SYMPTOM MANAGEMENT MODEL AND APPLICATION IN RESEARCH WITH CHILDREN 17 Abstract 17 Introduction IB Process of Theory Analysis 18 UCSF Symptom Management Model and Project Aims 19 Analysis of the UCSF-SMM 19 Origins 19 Meaning 23 Logical Adequacy 28 Usefulness 33 Generalizability 34 Parsimony 35 Testability 35 Empirical Support for the UCSF-SMM 36 Relationships Between Dimensions of Symptom Management 37 Relationships Among the Domains of Nursing Science and Dimensions of Symptom Management 38 Directions for Future Research 39 Considerations for Research With Children 39 Report and Measurement of Symptoms in Children 40 Inclusion of Parents and Family Members 40 Application Across Pediatric Populations 41 Summary 41 Acknowledgements 42 3. METHODS .43 Study Design 43 Case Study Design 43 Assumptions of Case Study Research 44 Study Setting 46 Study Participants 47 Inclusion Criteria 47 Exclusion Criteria 48 Study Variables and Measures 48 Demographic and Clinical Variables 48 Sleep-Wake Patterns 49 Environment Variables 51 Fatigue 52 Temperament 53 DRD4 and 5-HTTLPR Genotyping 54 Study Procedures 56 Institutional Review Board Approval 56 Participant Recruitment 56 Data Collection Procedures 57 Data Collector Training and Quality 59 4. RESULTS 60 Data Management and Analysis 60 Data Entry and Management 60 Actigraph Data 61 Environment Data 64 Sample Description 66 Participant Characteristics 66 viii Participant Accrual 69 Nighttime Sleep Characteristics 69 Aim 1 69 Description of Nighttime Sleep Characteristics 70 Comparison of Sample Characteristics With Age-Related Norms 70 Comparison of Sleep-Wake Patterns Across Study Nights 77 Comparison of Sleep-Wake Patterns Within Study Nights 91 Environmental Variables 99 Aim 2 99 Description of Environmental Variables 99 Evaluation of Environmental Variables for Serial Dependency 103 Evaluation of Patterns of Environmental Variables Within Study Nights ..105 Description of Clinical Variables 112 Relationships Between Environmental and Clinical Variables 117 Interrelationships Between Sleep-Wake Patterns, Environmental Variables, Biobehavioral Factors and Fatigue 119 Aim 3 119 Basic Mixed Linear Model 119 Multiple Regression 122 Relationships Between Temperament and Sleep-Wake Patterns 126 Behavioral Genotype 136 Fatigue 140 DISCUSSION, CONCLUSIONS, AND IMPLICATIONS 165 Discussion 165 Summary of Children's Sleep-Wake Patterns 165 Significance of the Hospital Environment 167 Nursing Care and Children's Symptoms 168 Temperament and Behavioral Genotype 171 Fatigue 174 UCSF Symptom Management Model as a Framework for Investigating Children's Cancer Symptoms 177 Case Study Research as Method of Inquiry to Investigate Children's Cancer Symptoms 180 Acceptability and Feasibility of Study Measures 181 Study Limitations 182 Implications for Research 183 Implications for Practice and Policy 186 Implications for Nursing Care 186 Implications for Multidisciplinary Practice and Policy 189 Conclusions 191 ix Appendices A. FATIGUE SCALE - CHILD VERSION 193 B. FATIGUE SCALE - PARENT VERSION 195 REFERENCES 198 x LIST OF TABLES Table Page 1.1. Parameters Recommended for the Evaluation of Sleep-Wake Disturbances... 7 2.1. Proposed Relationships Predicted Within the UCSF Symptom Management Model 29 4.1. Child Participant Characteristics 67 4.2. Parent Perceptions of Children's Sleep Characteristics at Home 68 4.3. Parent Participant Characteristics 69 4.4. Characteristics of Nighttime Sleep Based on Children's Reported Time in Bed 71 4.5 Pearson Correlations Between Nighttime Sleep Characteristics 72 4.6 Spearman Correlations Between Nighttime Sleep Characteristics 73 4.7 Characteristics of Nighttime Sleep Efficiency of Less Than 80% 74 4.8 Characteristics of Nighttime Sleep Latency Greater Than 19 Minutes 75 4.9 Characteristics of Average Nighttime Sleep Episodes of Less Than 90 Minutes Duration 76 4.10 Characteristics of Longest Nighttime Sleep Episodes Lasting Less Than 90 Minutes 77 4.11 Wilcoxon Matched-Pairs Signed-Ranks Comparison of Nighttime Total Sleep Time Between Study Nights 79 4.12 Number of Children Sleeping 600 Minutes or More on Each Study Night 80 4.13 Wilcoxon Matched-Pairs Signed-Ranks Comparison of Nighttime Awakenings Between Study Nights 85 4.14 Correlations Between Total Nighttime Sleep Minutes and Nighttime Awakenings 90 4.15 Minutes of Sleep Within Each Study Night Based on 2-Hour Epoch 92 4.16 Comparisons of Estimates of Fixed Effects of Epoch on Sleep Minutes With Epoch 6 93 4.17 Number of Children Sleeping 60 Minutes or More During the First Two Epochs 96 4.18 Summary of Sound in Decibels by Epoch Across Each Study Night 100 4.19 Summary of Light in Lumens/ft2 by Epoch Across Each Study Night 101 4.20 Summary of Temperature in Degrees Fahrenheit by Epoch Across Each Study Night 102 4.21 Comparisons of Estimates of Fixed Effects of Epoch on Sound Level With Epoch 6 109 4.22 Comparisons of Estimates of Fixed Effects of Epoch on Light Level With Epoch 6 110 4.23 Summary of Reported Pain Ratings by Epoch Across Each Study Night 113 4.24 Summary of Reported Nausea Ratings by Epoch Across Each Study Night... 114 4.25 Summary of Vomiting Episodes Across Study Nights 114 4.26 Summary of Medication Doses Within Each Study Night 115 4.27 Pearson Correlations Between Environmental and Clinical Variables 118 4.28 Spearman Correlations Between Environmental and Clinical Variables 118 4.29 Comparisons of Estimates of Fixed Effects of Epoch on Sleep Minutes, Controlling for the Effects of Sound and Light 121 4.30 Comparisons of Estimates of Fixed Effects of Sound by Epoch Interaction on Sleep Minutes 121 4.31 Pearson Correlations Between Sleep Minutes, Environmental Variables, Number of Medications, Pain, Nausea, and Vomiting Within 2-hour Nighttime Epochs 123 xii 4.32 Spearman Correlations Between Sleep Minutes, Environmental Variables, Number of Medications, Pain, Nausea, and Vomiting Within 2-Hour Nighttime Epochs 124 4.33 Backward Regression of Independent Variables on Sleep Minutes Within Each Epoch 125 4.34 Coefficients of Variables Included in Final Backward Regression Model ... 125 4.35 Interpretation of Standardized Scores for Temperament Dimensions 127 4.36 Summary of Temperament Characteristics of Study Participants 128 4.37 Pearson Correlations Between z-Scores for Temperament Dimensions and Sleep Parameters Across Study Nights 129 4.38 Spearman Correlations Between z-Scores for Temperament Dimensions and Sleep Parameters Across Study Nights 130 4.39 Mann-Whitney Comparison of Mean Ranks of Sleep Parameters by Activity 132 4.40 Mann-Whitney Comparison of Mean Ranks of Sleep Parameters by Rhythmicity 133 4.41 Mann-Whitney Comparison of Mean Ranks of Sleep Parameters by Adaptability 133 4.42 Mann-Whitney Comparison of Mean Ranks of Sleep Parameters by Intensity 134 4.43 Mann-Whitney Comparison of Mean Ranks of Sleep Parameters by Mood ... 134 4.44 Mann-Whitney Comparison of Mean Ranks of Sleep Parameters by Threshold Sensitivity 135 4.45 Kruskall-Wallace Comparison of Mean Ranks of Sleep Parameters Based on DRD4 Genotype 138 4.46 Mann-Whitney Comparison of Mean Ranks of Sleep Parameters Based on 5-HTTLPR Genotype 139 4.47 Correlations Between Child and Parent Fatigue Scales 142 4.48 Correlations Between Fatigue Scales and Child's Admission Hematocrit 143 xiii 4.49 Fatigue Scale Scores at Baseline and Following Each Study Night 144 4.50 Wilcoxon Matched-Pairs Signed-Ranks Comparisons of Child Fatigue Scale Scores Across Study Nights 146 4.51 Wilcoxon Matched-Pairs Signed-Ranks Comparisons of Parent Fatigue Scale (Intensity Component) Scores Across Study Nights 151 4.52 Wilcoxon Matched-Pairs Signed-Ranks Comparisons of Parent Fatigue Scale (Fatigue Sources Component) Scores Across Study Nights 158 4.53 Pearson Correlations Between Daily Fatigue Scores and Nighttime Sleep Variables 162 4.54 Spearman Correlations Between Daily Fatigue Scores and Nighttime Sleep Variables 163 xiv LIST OF FIGURES Figure Page 2.1. UCSF Symptom Management Model 20 2.2. Revised UCSF Symptom Management Model 22 4.1 Actigraph Data File Scored Based on Child's Reported Nighttime Sleep 62 4.2 Actigraph Data File Scored Based on a 12-Hour Night Shift 63 4.3 Graphical Display of Environmental Data Across Three Nights 65 4.4 Profile Plot of Estimated Marginal Means of Sleep Total Sleep Time Across Study Nights 78 4.5 Profile Plot of Child With Decreasing Total Sleep Time Across Study Nights 81 4.6 Profile Plot of Child With Increasing Total Sleep Time Across Study Nights 82 4.7 Profile Plot of Child With Variable Total Sleep Time Across Study Nights ... 83 4.8 Profile Plot of Estimated Marginal Means of Nighttime Awakenings Across Study Nights 84 4.9 Profile Plot of Child With Low Nighttime Awakenings and Low Variability of Awakenings 86 4.10 Profile Plot of Child With Consistently High Nighttime Awakenings 87 4.11 Profile Plot of Child With a Moderate Number of Nighttime Awakenings With an Increasing Trend 88 4.12 Profile Plot of Child With a Moderate Number of Nighttime Awakenings With a Decreasing Trend 89 4.13 Estimated Marginal Means of Sleep Minutes by Epoch Across Study Night.... 95 4.14 Profile Plot of Sleep Minutes Within Each Epoch for a Child With Few Nighttime Awakenings 97 4.15 Profile Plot of Sleep Minutes Within Each Epoch for a Child With Frequent Nighttime Awakenings 98 4.16 Raw Sound Data Across the Three Study Nights for One Participant 106 4.17 Autocorrelogram for Sound Data for the First Study Night for One Participant 107 4.18 Partial Autocorrelogram for Sound Data for the First Study Night for One Participant 108 4.19 Profile Plot of Estimated Marginal Means of Child Fatigue Scale Scores at Baseline and Following Each Study Night 145 4.20 Child With an Initial Increase in Fatigue Followed by a Decrease to Below Baseline Level 147 4.21 Child With a Moderate Level of Fatigue With Minimal Variation 148 4.22 Child With a Steady Increase in Fatigue With Increased Variability 149 4.23 Child With a Gradual Decline in Fatigue 150 4.24 Profile Plot of Estimated Marginal Means of Parent Fatigue Scale (Intensity Component) at Baseline and Following Each Study Night 152 4.25 Increasing Trend of Parent-Reported Fatigue Intensity Across Study Days ... 154 4.26 Steep Increase in Parent-Reported Fatigue Intensity Followed by a Decrease ..155 4.27 Initial Increase in Parent-Reported Fatigue Intensity Followed by a Decrease Below Baseline 156 4.28 Profile Plot of Estimated Marginal Means of Parent Fatigue Scale (Fatigue Sources Component) at Baseline and Following Each Study Night 157 4.29 Increasing Trend in Sources of Fatigue Across Study Days 159 4.30 Steep Increase in Sources of Fatigue From Baseline Followed by a Decrease.. 160 4.31 Variable, but Decreasing Trend in Sources of Fatigue 161 xvi ACKNOWLEDGEMENTS This dissertation represents the culmination of the efforts of many individuals who have provided encouragement, support, and mentorship throughout my doctoral education. I would like to express my appreciation to: • The members of my supervisory committee for their ongoing support and guidance throughout the development and conduct of my dissertation research. Each performed a vital role in the development and completion of this research. • The children and their parents who participated in this study. I truly appreciate their willingness to support this project and the encouragement they offered me. • My clinical colleagues at Primary Children's Medical Center. I am most grateful for their constant support and celebration of each milestone along this journey. • The Summer Genetics Institute sponsored by the National Institute for Nursing Research which challenged me to explore potential genetic factors related to children's symptoms. • My funding sources: The National Institute for Nursing Research (F31NRO10175- 01A1), American Cancer Society (DSCN-06-204-1), and the Western Institute of Nursing (Dissertation Scholarship Grant). • My husband, Wayne, and my children, Darin, Kyle, and Caitlin for their love and support. • God's sustaining presence (Zechariah 4:6-7). CHAPTER 1 INTRODUCTION Significance of the Problem and Specific Aims Cancer-related symptoms have a significant impact on patient morbidity and mortality both during and following completion of therapy. Although the symptom of disturbed sleep has received attention as a high priority area for research (American Cancer Society, 2001; Berger et al, 2005; National Center on Sleep Disorders Research, 2003; Oncology Nursing Society, 2007), very few studies have addressed sleep in children with cancer. The prevalence of disturbed sleep among children with cancer is estimated at 30-45% (Collins et al., 2002, 2000), yet sources of sleep disturbances among children with cancer are not well understood. Studies of acutely ill children indicate that hospital environmental stimuli impair both sleep quantity and quality (Carno, Hoffman, Henker, Carcillo, & Sanders, 2004; Corser, 1996; Cureton-Lane & Fontaine, 1997). Aspects of the child's temperament, including impulsivity, anxiety, and approach-withdrawal behaviors may impact the child's individual response to stimuli in the hospital environment and affect sleep-wake patterns (Caldwell-Andrews & Kain, 2006; Conte, Walco, & Kimura, 2003; Kain et al., 2002; McClowry, 1990; Vachha & Adams, 2005). Common genetic polymorphisms in the dopamine D4 receptor (DRD4) gene and regulatory region of the serotonin transporter promoter gene (5-HTTLPR) are associated 2 with behavioral variation and may influence the child's response to physiologic and environmental stimuli and impact the child's cancer-related symptom burden (Auerbach, Faroy, Ebstein, Kahana, & Levine, 2001; Auerbach et al., 1999; Ebstein, Levine, Geller, Auerbach, Gritsenko, & Belmaker, 1998; Lakatos et al, 2003; Lesch et al, 1996). Study Purpose The purpose of this study was to examine sleep-wake patterns and the influences of environmental stimuli and biobehavioral (temperament and behavioral genotype) factors on sleep-wake patterns among children with cancer during an admission for inpatient chemotherapy using a multiple-case study approach. Efforts to understand and improve sleep for children with cancer have the potential to improve immune recovery and decrease levels of fatigue which can improve quality of life by decreasing treatment-related morbidity. These goals are consistent with the National Institute of Nursing Research's (NINR) strategic plan and the National Cancer Institute's (NCI) research priorities to improve quality of life for individuals with chronic illness through symptom management (National Cancer Institute, 2007; National Institute of Nursing Research, 2006). Study findings will promote quality of life for children with cancer through identification of factors contributing to disturbed sleep in the hospital and the development of interventions to improve sleep. Specific Aims Specific aims for the study were as follows: 1. Describe nighttime sleep-wake patterns among children with cancer receiving inpatient chemotherapy. 3 2. Describe nighttime patterns of environmental factors (room sound, light, and temperature), and relationships with clinical variables (pain, nausea/vomiting, and medication administration) in children with cancer. 3. Describe relationships between nighttime sleep-wake patterns, environmental I variables, biobehavioral variables (temperament and behavioral genotype), and fatigue in children with cancer. Review of the Literature Overview of Childhood Cancer Incidence. Approximately 12,400 children and adolescents in the United States less than 20 years of age are diagnosed with cancer each year (Ries et al., 2008). This number represents less than 2% of all cancers diagnosed in the United States each year. Of this total, approximately one third of cases are diagnosed in children less than 5 years of age, another one third are diagnosed in children 5 to 14 years of age, and the remaining third of cases are diagnosed in adolescents between 15 to 19 years of age (Ries et al.). Although cancer remains the leading cause of death from disease in children and adolescents, the overall cure rate across all disease types is nearing 80% (Jemal et al., 2008). Role of symptom management. As cure rates have continued to improve, increased attention has been given to addressing children's quality of life, a central component of which is symptom management (Hinds et al., 2004). Cancer-related symptoms may be a consequence of the disease itself or a result of its treatment. Symptoms may continue to persist even following completion of therapy. The severity of the symptoms, and resulting distress, have the potential to influence decision-making 4 related to the child's treatment (Docherty, Sandelowski, & Preisser, 2006; Woodgate, Degner, & Yanofsky, 2003). In recent years, investigation of the symptom of disrupted sleep-wake patterns has received increasing priority for both adults and children with cancer (Berger et al., 2005; Vena et al., 2004). The Oncology Nursing Society (2007) has identified disturbed sleep-wake patterns as a priority area for research with attention to its association with other cancer-related symptoms and the investigation of disturbed sleep-wake patterns across age groups. Delineation of causative mechanisms of symptoms and the development of interventions to improve patient response to symptoms also are current priorities of the National Institute of Nursing Research (2006). Continued research to identify sources and mechanisms of disturbed sleep in children with cancer will guide the development of interventions to lessen the burden of this symptom and improve quality of life. Sleep Physiology in Children and Consequences of Disturbed Sleep Sleep physiology. The National Center for Sleep Disorders Research and the National Sleep Foundation recognize adequate sleep as fundamental for health and well-being (National Center on Sleep Disorders Research, 2003; National Sleep Foundation, 2005). Sleep is a complex, regulated process involving the interaction of ultradian, circadian, and homeostatic processes. Sleep regulation involves ultradian cycling, which refers to movement between phases of rapid eye movement (REM) sleep and non-REM (NREM) sleep. The exact mechanism regulating this process has not been fully identified. Timing of sleep-wake states is regulated by circadian and homeostatic processes. Circadian rhythms are those cycles generally linked to a 24-hour period and are influenced by light-dark cycles. These processes are regulated by the suprachiasmatic nuclei of the 5 hypothalamus. Sleep homeostasis refers to processes by which the individual accumulates a need for sleep during the day and then has that need satisfied during sleep. Slow wave sleep (SWS), which occurs during deep NREM sleep, is essential for normal immune system regulation, tissue healing, and growth (Aldrich, 1999). Age-related normal sleep-wake patterns. Organization of sleep-wake patterns is a component of normal growth and development with most children achieving a bimodal sleep pattern by 18 months of age (Borghese, Minard, & Thoman, 1995; Garcia, Rosen, & Mahowald, 2001; Holditch-Davis, Brandon, & Schwartz, 2003). Daytime napping continues through the early childhood with most children relinquishing naps by 5 years of age (Iglowstein, Jenni, Molinari, & Largo, 2003). Nighttime sleep needs decrease across childhood with 3- to 6-year-old children requiring 11 to 12 hours and children 6 to 12 years of age requiring 10 to 11 hours of sleep per 24 hours (Mindell & Owens, 2003). On average, school-age children awaken briefly approximately 4 to 6 times each night as a result of normal ultradian rhythms associated with sleep cycles (Mindell & Owens). The average sleep cycle lengthens from approximately 40 to 50 minutes during infancy to 60 minutes during the toddler years. By the time children are 5 years of age, the sleep cycle has increased to 90 minutes (Sheldon, 2005). The normal adult sleep cycle is regarded as ranging from 90 to 110 minutes in duration. Consequences of disrupted sleep. Disrupted SWS can alter normal hormonal regulation related to immune function and is associated with other disease-related symptoms. Evidence from adult studies suggests that natural killer cell activity and cytokine activity are altered in the presence of sleep deprivation, resulting in impaired 6 natural host defenses (Irwin, 2002; Irwin, et al., 1996). Cytokine activity can remain impaired even after sleep recovery, suggesting a prolonged impact on immune function. Neuroendocrine processes also are vulnerable to the effects of disrupted sleep. Growth hormone secretion is inhibited by deprivation of SWS (Van Cauter & Spiegel, 1999). Decreased growth hormone secretion can impact glucose and insulin regulation and can also suppress cellular immunity. In women with fibromyalgia, disrupted SWS is associated with increased symptoms, including lowered pain thresholds, increased discomfort, and fatigue (Lentz, Landis, Rothermel, & Shaver, 1999). Consequences of disruption and deprivation of SWS are particularly concerning for children with cancer. During treatment, children must not only recover from the effects of cancer and its treatment-related consequences, but also sustain normal physiologic growth. Sustained SWS disturbances have the potential to contribute to other symptoms, impair normal growth, delay physical recovery, and place children at risk for life-threatening infections. Knowledge gaps related to sleep-wake pattern disturbances in children with cancer include the magnitude of sleep-wake pattern disturbances and factors contributing to disturbances. Understanding sources of disturbances and their significance is essential to develop interventions to prevent SWS disruptions in children with cancer. Evaluation of Sleep-Wake Disturbances The term "sleep-wake disturbances" represents a general term describing sleep-related symptoms that may be experienced by an individual (Berger et al., 2005). It is not a formal diagnostic term used by the American Academy of Sleep Medicine. The range of individual complaints related to sleep-wake disturbances is varied and may include difficulty falling asleep, difficulty staying asleep, or feeling that sleep is inadequate. 7 To facilitate more meaningful assessment of sleep-wake disturbances, Berger and colleagues (2005) have proposed consideration of nine sleep-related parameters. These parameters represent ones that can be measured both objectively and subjectively. The parameters and their definitions are summarized in Table 1.1. Sleep and Children With Cancer Current knowledge regarding sleep-wake pattern disturbances in children with cancer is largely descriptive. Disturbed sleep is a frequently reported cancer-related symptom, yet perceived sources of disturbed sleep have not been clearly identified (Bhatia Table 1.1 Parameters Recommended for the Evaluation of Sleep-Wake Disturbances Sleep Parameter Definition Total sleep time Number of minutes of sleep while in bed Sleep latency Number of minutes between going to bed and falling asleep Awakenings Numeric count of awakenings Wake after sleep onset Number of minutes awake after sleep onset Daytime napping Total minutes of daytime sleep - intentional or unintentional Daytime sleepiness Number of episodes of unintentional daytime sleep Quality of perceived sleep Subjective assessment of sufficiency of sleep Circadian rhythm Biobehavioral phenomenon repeating approximately every 24 hours Sleep efficiency Number of minutes of sleep divided by number of minutes in bed multiplied by 100 Adapted from: Berger, A. et al., (2005). Sleep/wake disturbances in people with cancer and their caregivers: State of the science. Oncology Nursing Forum, 32, e98-el26. 8 et al., 2002; Collins et al„ 2002, 2000; Phipps, Dunavant, Jayawardene, & Srivastiva, 1999; Varni, Burwinkle, Katz, Meeske, & Dickinson, 2002). Instruments used to measure fatigue in children with cancer include attention to sleep difficulties, yet are not specifically designed to identify actual sleep disturbances (Hinds et al., 2007; Hockenberry et al., 2003). Medications as a source of disrupted sleep. Medications used in cancer treatment and symptom management can alter sleep/wake patterns. Benzodiazepines, antihistamines, and phenothiazines, common adjunct therapies for nausea and vomiting cause drowsiness which can increase daytime sleeping and alter nighttime sleeping patterns. Benzodiazepines decrease SWS and REM sleep which can further alter sleep/wake patterns (Vena, Parker, Cunningham, Clark, & McMillan, 2004). Glucocorticoids also have been implicated in contributing to disturbed sleep (Drigan, Spirito, & Geler, 1992; Vena et al., 2004). Comparison of nighttime sleep among children with acute lymphoblastic leukemia (ALL) for 7 days prior to and during a 7-day period of prednisone administration indicated significantly less intermittency of activity during sleep during prednisone therapy (p < .02), suggesting disturbed sleep patterns. Actual time spent sleeping did not differ significantly between the two time periods (Kline, 1999). Children receiving a 5-day course of dexamethasone as treatment for ALL experienced increases in total nighttime sleep duration, increased daytime napping, and increases in fatigue compared with the 5 days prior to dexamethasone administration (Hinds, Hockenberry, Gattuso, et al., 2007; Hinds, Hockenberry, Srivistava, et al., 2007). Influence of the child's sleep environment. Recently published studies support investigation of the child's sleep environment and identifying interventions to improve 9 sleep quantity and quality. Children and adolescents (8 to 16 years of age) receiving outpatient treatment for acute lymphoblastic leukemia (ALL) had a greater than normal number of nighttime awakenings over a 3-night period (mean = 19.8, SD = 4.1) following treatment, with total nighttime sleep ranging from 6 to 9.5 hours (mean = 8; SD = 1.1) (Gedaly-Duff et al., 2006). All participants (n = 9) reported evening fatigue and both cancer- and noncancer-related pain. A longitudinal case study of cancer symptoms, including disturbed sleep in an adolescent, revealed highly variable sleep efficiency (mean = 69.08%) with no clear pattern across the initial 90 days of treatment for cancer (Docherty, Sandelowski, & Preisser, 2006). In the hospital setting, nighttime awakenings ranged from 0 to 40 (mean = 15.32) with the greatest number occurring on the 3rd of 3 nights of hospitalization (Hinds, Hockenberry, Rai, Zhang, Bassem et al., 2007). For 70% of children and adolescents, the longest uninterrupted nighttime sleep interval during a 2- night period was 1 hour. Room entries and exits by both parents and staff during an 8-hour night shift ranged from 3 to 22 (mean=l 1.30) (Hinds Hockenberry, Rai, Zhang, Bassem et al.). Sources of Disturbed Sleep in Acutely and Chronically 111 Children Environmental influences. Few studies have investigated the role of environmental sources on sleep disturbances in hospitalized children older than neonates. Pediatric intensive care unit (PICU) light and sound levels and caregiver activities were negatively correlated with nighttime sleep (p < .01) (Al-Samsam & Cullen, 2005; Carno et al., 2004; Corser, 1996; Cureton-Lane & Fontaine, 1997). Mean PICU nighttime sound levels are .consistently reported as greater than 50 decibels (dB), with spikes to 103 dB, which is in excess of the 35 dB recommended by the World Health Organization (WHO) for adequate sleep (Al-Samsam & Cullen; Cureton-Lane & Fontaine; World Health Organization, 1980). Fragmented sleep in the PICU has been identified using both polysomnography and observational methods (Al-Samsam & Cullen; Carno et al.; Corser; Cureton-Lane & Fontaine). Mean nighttime sleep as measured by direct observation was 4.7 hours (SD^.49); mean sleep episodes were 27.6 minutes (SD=25.85) with a mean of 9.8 (SD-2.8) interruptions per night (Cureton-Lane & Fontaine). Relationship between sleep and other symptoms. Postoperative pain was associated with a greater than 7.48% decrease in nighttime sleep in previously healthy children undergoing outpatient elective surgery (p < .05) (Kain et al., 2002). Lower sociability temperament (p_< .05), externalizing behaviors, and higher perioperative anxiety (p < .05) also contributed to poor sleep outcomes (Kain et al.; Caldwell-Andrews & Kain, 2006). Children with juvenile rheumatoid arthritis (JRA), which, like cancer, has a chronic symptom profile, have more fragmented sleep as compared with healthy controls (Archbold, Lentz, Brandt, Wallace, & Landis, 2007; Labyak, Bourguignon, & Docherty, 2003; Zamir, Press, Tal, & Tarasiuk, 1998). Self-reports of poor sleep quality and quantity are correlated with increased pain in school-age children (r = .56; p < .05) (Bloom et al., 2002). Disturbed sleep also has been suggested as a predictor of overall symptom severity in children with JRA (Labyak, Bourguignon, & Docherty). These findings support the study of environmental factors as influencing sleep in children with cancer in inpatient settings. A structured nursing care intervention to modify the environment improved sleep outcomes in premature infants (Holditch-Davis, Barham, 11 O'Hale, & Tucker, 1995). No similar interventions have been undertaken in children with cancer. Relationships between sleep and other cancer symptoms warrant further study. Temperament and Children's Responses to Illness and Symptoms Temperament. Identifying subgroups of individuals based on psychologic or behavioral characteristics that may contribute to illness-related symptoms is receiving increased attention (Barsevick, 2007; Parker, Kimble, Dunbar, & Clark, 2005). Variation in temperament, the "how" of a child's behavior, is recognized as influencing symptoms in children with chronic illnesses (Conte, Walco, & Kimura, 2003; McClowry, 1990; Vachha, & Adams, 2005). Normal temperament variation influences children's responses to illness and the environment (Chess & Thomas, 1996). The dimensions of temperament: activity, rhythmicity, intensity, distractibility, persistence, mood, approach/withdrawal, adaptability, and threshold of responsiveness, are stable across childhood and support a biologic basis for temperament (Novosad, Freudigman, & Thoman, 1999; Novosad & Thoman, 1999). Temperament and response to hospitalization and physiologic stressors. Studies involving responses to hospitalization and physiologic stressors in young children have identified significant relationships between selected dimensions of temperament and behavior. In previously healthy school-age children, mood, rhythmicity, intensity, and approach/withdrawal accounted for more than 50% of posthospitalization behaviors 1 week and 1 month following hospitalization (McClowry, 1990). Mood, distractibility, and lower thresholds of responsiveness were associated with increased postoperative pain among previously well 3- to 7-year-olds (Helgadottir & Wilson, 2004). Externalizing behaviors impaired postoperative sleep above and beyond the influence of preoperative 12 sleep and postoperative pain in children undergoing elective outpatient surgery (Caldwell- Andrews & Kain, 2006). Temperament and chronically ill children. Several studies have investigated temperamental characteristics of chronically ill children in an attempt to characterize children with selected conditions and to identify temperamental characteristics that may place the child at risk for "poorness of fit." Children between 5 to 12 years of age with myleomeningocele and hydrocephalus requiring placement of a shunt have been identified as having temperament profiles that vary from a standardized pediatric population. These children were characterized as being less adaptable, more likely to withdraw from or respond negatively to new situations, more distractible, less attentive or persistent, and less predictable as measured by age-appropriate versions of the Carey Temperament Scales (Vacha & Adams, 2005). This study also identified characteristics of children at increased risk for poorer academic and social outcomes. This "at risk" temperament profile included being more likely to withdraw from or respond negatively to new situations, less adaptable to changes in routine, increased distractibility, and poor persistence (Vacha & Adams). Temperamental characteristics of children between 7 and 17 years of age with juvenile primary fibromyalgia syndrome (JPFMS) included decreased mood, decreased regularity of daily habits, lower persistence, and increased distractibility when compared with children with arthritis and healthy controls (Conte, Walco, & Kimura, 2003). Temperament was measured using the Dimensions of Temperament Survey-Revised patterned after the work of Chess and Thomas. These children also were identified as having increased sensitivity to pain. 13 Temperament characteristics and children with cancer. Although no studies investigating temperament characteristics and symptoms in children have been conducted, the role of anxiety has been investigated. In children and adolescents receiving chemotherapy, anxiety was associated with increased cancer symptoms, including vomiting and altered sleep (Docherty, Sandelowski, & Preisser, 2006; Dolgin, Katz, Zeltzer, & Landsverk, 1989). These symptoms influence a child's perceived ability to tolerate cancer treatment. Further exploration of the association between a child's temperament characteristics and cancer symptoms is warranted. Genetic Polymorphisms Associated With Variability in Temperament Genetics and temperament. Genetics plays a role in the underlying biology of symptom profiles as well as temperament characteristics; therefore, investigating potential genetic contributions to temperament is logical. Recent studies have investigated common polymorphisms in genes regulating dopamine and serotonin and their associations with temperament and behavior across the lifespan (Jacobs, Kenis, Peeters, Derom, Vlietinck, & van Os, 2006; Lesch et al., 1996; Okuyama et al., 2000). These normal variations are identified as long or short, based on the number of base repeat units present. Studies of these polymorphisms involving children have included well children or those with behavioral disorders. No published studies involving application of findings among acutely or chronically ill children and their responses to illness were identified. Dopamine D4 receptor gene. Dopamine is associated with individual activation and the intensity of response in situations involving reward (Panskepp, 1986). The exon III region of the DRD4 gene located on chromosome 11 contains variable numbers of a 48 base pair repeat (VNTR). Repeats of 6 or more are regarded as long; repeats of 5 or fewer 14 are regarded as short (Okuyama et al., 2000). The 4-repeat variant is most common, followed by the 7-repeat variant (Lakatos et al., 2003). The 4-repeat variant is two to three times more potent in dopamine-mediated coupling to adenyl cyclase than the 7-repeat variant. Long polymorphisms in the dopamine D4 receptor (DRD4) gene are associated with novelty-seeking behavior: impulsivity, exploratory, or sensation-seeking behavior and are present in approximately 33% of the population (Auerbach et al., 1999, 2001; Ebstein et al., 1998; Lakotas et al., 2003; Lesch et al., 1996). Serotonin transporter gene. The serotonin transporter promoter gene (5-HTT) is located on chromosome 17ql2. Transcription of this gene is modulated by a common insertion or deletion polymorphism in its linked upstream regulatory region (5-HTTLPR) (Heils et al., 1996). The longer, 16-repeat unit polymorphism is more common and has demonstrated a dominant effect in heterozygous individuals (Lesch et al., 1996). Short polymorphisms in the regulatory region of the serotonin transporter promoter gene have been linked to harm avoidance, neuroticism, and anxiety (Auerbach et al., 1999, 2001; Lakotas et al., 2003; Lesch et al.). Approximately 19% of the population is homozygous for short 5-HTTLPR polymorphisms (Lesch et al). Associations between DRD4 and 5-HTT genes, temperament, and response to illness. In the clinical setting, variations in DRD4 and 5-HTTLPR render the child vulnerable to increased symptom distress. Children who are more sensitive to disruptive stimuli in the hospital environment and variations from preestablished home routines may be at increased risk for altered in sleep-wake patterns, which, in turn, can influence other cancer-related symptoms. Genotypic variations in DRD4 and 5-HTTLPR also may 15 predispose children to increased symptom burden and psychologic distress resulting from the illness experience. Conceptual Framework The UCSF School of Nursing Symptom Management Model (UCSF-SMM) (Dodd et al., 2001) was the conceptual framework guiding the study. The domains of nursing science represented in the model, person, environment, and health and illness, provided the context for the study, essential to the study's methodology, a multiple-case study design (Yin, 2003a & 2003b). An analysis of the UCSF-SMM is presented in Chapter 2. The study's scope was limited to the symptom experience dimension occurring in the context of the domains of nursing science. This context provided by the UCSF-SMM directed the selection of study variables contributing to disturbed sleep. An additional benefit of this model in guiding this study is that it recognizes the role of development as influencing the symptom experience, a key consideration for research with children. Study variables located in the person domain included the biobehavioral variables of temperament and behavioral genotype. The environment domain included room sound, light, and temperature. Variables in the health and illness domain were the child's illness, fatigue, number of medications per 12-hour night shift, pain, and nausea/vomiting. Summary of Study Background and Significance Disturbed sleep-wake patterns have the potential to threaten the health and quality of life of children with cancer. This study is among the initial studies to investigate sleep-wake patterns in hospitalized children with cancer and is the first to investigate the role of temperament and genetic variation as potentially influencing symptoms. Understanding 16 such unique differences among children, and their biologic basis, can guide nursing interventions to improve sleep outcomes and decrease the severity of other symptoms. The study also investigated the effect of the hospital environment on sleep-wake patterns. The study aims were consistent with the NINR's agenda to harness technologic advances, including genetics, into research and to improve quality of life for individuals with chronic illness (NINR, 2006). The study's purpose also addressed a priority of the National Center for Sleep Disorders Research to investigate secondary sleep disturbances occurring in the context of chronic illness (National Center on Sleep Disorders Research, 2003). Findings will be used to generate hypotheses for future studies investigating biobehavioral variation of other cancer-related symptoms and to develop individualized interventions to lessen the burden of disturbed sleep and improve quality of life. CHAPTER 2 ANALYSIS OF THE UCSF SYMPTOM MANAGEMENT MODEL AND APPLICATION IN RESEARCH WITH CHILDREN Prepared for submission for publication Abstract The UCSF Symptom Management Model (UCSF-SMM) is a deductive, middle range theory depicting symptom management as a multidimensional process occurring within the domains of nursing science. A theory analysis was undertaken using the process described by Walker and Avant to evaluate its strengths and weaknesses as well as to identify additional areas for development and refinement with attention to its potential application in research involving children with chronic illness. Although largely based on research involving adults, the model has been developed with the intention that it would be applicable in research and clinical practice involving children. To date, the UCSF-SMM has been used as a guiding framework for research involving varied adult illness states as well as in children with cancer. Although agreement among the model components is present among investigators, the complexity and scope of the model has limited its full testability within a given study. Additional considerations for continued use in research involving children include attention to measurement of symptoms in 18 children and clarity regarding the location of the parents and family within the model components. Key words: theory analysis, UCSF Symptom Management Model, children, chronic illness Introduction Theory development in nursing creates conceptual meaning by linking related concepts together in such a way as to illustrate meaning (Chinn & Kramer, 2004). Theory analysis contributes to theory development by providing a systematic process through which strengths and inconsistencies of existing theories are identified. Theory analysis also reveals opportunities for additional theory testing and refinement. Such opportunities could include extension of a theory's applicability across health and illness states or across developmental groups (Walker & Avant, 2005). Process of Theory Analysis Walker and Avant (2005) describe seven key steps in theory analysis. These are 1) identification of the theory's origins, 2) examination of the theory's meaning, 3) analysis of the theory's logical adequacy, 4) determination of the theory's usefulness, 5) definition of the theory's generalizability, 6) determination of the theory's parsimony, and 7) determination of the theory's testability. Utilization of this process facilitates a greater understanding of a theory and stimulates further development and refinement of the theory (Walker & Avant). 19 UCSF Symptom Management Model and Project Aims The UCSF Symptom Management Model (UCSF-SMM) is a deductive, middle range theory depicting symptom management as a multidimensional process occurring in the domains of nursing science. Although the model is based on programs of research working with adult patients, the model also is proposed to be applicable for children (Dodd et al., 2001). The aims of this theory analysis are twofold: 1) to analyze the UCSF Symptom Management Model (UCSF-SMM) using the process described by Walker and Avant (2005) and 2) to evaluate the utility of the UCSF-SMM for potential use in research with chronically ill children. Analysis of the UCSF-SMM Origins Model development. The UCSF-SMM is a middle range model illustrating a multidimensional process of symptom management. It was initially published in 1994 as a collaborative effort by members of the Symptom Management Faculty Group at the University of California at San Francisco School of Nursing (UCSF School of Nursing Symptom Management Faculty Group, 1994) (Figure 2.1). This group of nurse scientists represented a broad scope of programs of research in multiple adult illness states including diabetes, cardiac disease, chronic pulmonary disease, chronic pain, and cancer. The result of their efforts was the development of a deductive, process-focused model addressing three interactive dimensions of symptom management: symptom experience, symptom management strategies, and outcomes. The underlying premise of the model was effective symptom management required consideration of all three dimensions should be considered. 20 SYMPTOM EXPERIENCE SYMPTOM-MANAGEMENT STRATEGIES Perception of -- Patient _ Symptoms - Evaluation of «- Symptoms Healthcare System Response to Symptoms Healthcare ^Provider, SYMPTOM OUTCOMES Functional Status Emotional Status Health Service M-• .Utilization Financial - Status Symptom Status Quality of Life „ Mortality Morbidity & Comorbidity Figure 2.1: UCSF Symptom Management Model From: UCSF School of Nursing Symptom Management Group (1994). A model for symptom management. Image: Journal of Nursing Scholarship 26, 272-276. Reprinted with permission. Copyright: Wiley-Blackwell Publishing. 21 An updated version of the UCSF-SMM was published in 2001 as a result of ongoing research and collegial discussions (Dodd et al., 2001) (Figure 2.2). This revised model places the process of symptom management within the context of the domains of nursing science: the person, environment, and health and illness. The symptom management dimension was revised to reflect components of symptom management strategies. This change was intended to guide the development of interventions which then could be replicated. Changes to the outcome dimension included conceptualization of all of the outcomes of symptom management as being interrelated. Another addition to the model was the concept of adherence, which is depicted as extrinsic to the dimensions of symptom management and having the potential to disrupt the relationship between symptom management strategies and the outcomes dimension. Assumptions. The UCSF-SMM includes six stated assumptions addressing the nature of symptoms, the focus of symptom management strategies, and relationships within the model. These assumptions are as follows: That the gold standard for the study of symptoms is based on the perception of the individual experiencing the symptom and his/her self-report. That the symptom does not have to be experienced by an individual to apply this model of symptom management. The individual may be at risk for the development of the symptom because of the influence (impact) of a context variable such as a work hazard. Intervention strategies may be initiated before an individual experiences the symptom. 22 Person Demographic, psychological, sociological? physiological, developmental Symptom f/ / experience .•Perception -4-+- Evaluation : of of • symptoms symptoms ^ J* * Response r to symptoms • • \ . Components of \ \ \ 'symptom management\ / strategies . I Who? (Delivers) ' I I What? How? 1 I \ When? To Whom? ' / \ Where? How much? Outcomes ^Environment' Physical > Social \ Cultural S S s Functional status 1 Emotional Self-care status Symptom status Mortality , ^.V-Adhei Costs Quality of life -Adherence /Health & Illness/ / Risk factors / / Health status / S Disease & injury^."* Morbidity & j:o-morbidity Figure 2.2: Revised UCSF Symptom Management Model From: Dodd, M., Janson, S., Facione, N., Faucett, N., Froelicher, E. S., Humphreys, J. et al. (2001). Advancing the science of symptom management. Journal of Advanced Nursing, 33, 668-676. Reprinted with permission. Copyright: Wiley-Blackwell Publishing. 23 That nonverbal patients (infants, poststroke aphasic persons) may experience symptoms and the interpretation by the parent or caregiver is assumed to be accurate for purposes of intervening. That all troublesome symptoms need to be managed. That a management strategy may be targeted at the individual, a group, a family, or the work environment. That symptom management is a dynamic process; that is, it is modified by individual outcomes and the influences of the nursing domains of person, health/illness, or environment (Dodd et al., 2001, pp. 669-670). Meaning The second component of theory analysis involves identification of definitions of key concepts within the model or theory as well as relational statements and specified relationships within the model (Walker & Avant, 2005). Dodd et al. define a symptom as "... a subjective experience reflecting changes in the biophysical functioning, sensations, or cognition of an individual (p. 669). Symptom management is recognized as a multidimensional process that requires consideration of each of the three dimensions of the model to be considered effective (Dodd et al.). The authors also maintain that each of the model's dimensions requires attention for effective symptom management to occur. Dimensions of symptom management. The UCSF-SMM includes three dimensions of symptom management: the symptom experience, symptom management strategies, and outcomes. Each dimension is conceptualized with examples from the authors' research. Each dimension is depicted as being related to the other two through the use of bidirectional arrows. 24 The symptom experience dimension, with its three components, is the most thoroughly described aspect of the model and is depicted as the beginning of the symptom management process. This dimension consists of the individual's perception, evaluation, and response to a symptom. Bidirectional arrows are used to depict the relationships among these components. These relationships are recognized as occurring in a repetitious manner, or even simultaneously (Dodd et al., 2001). Application of this dimension of the model in children may pose unique challenges. The authors note that parents and children may ascribe different meaning to perceived symptoms. Such incongruency can lead to difficulty in planning interventions. The authors also acknowledge the benefits of technology as contributing to the measurement of symptoms in children (Dodd et al., 2001). Such measures add a component of objectivity and can complement more subjective measures offered by children and parents. The second dimension of the model is comprised of the individual components of the symptom management strategy. These are described as the "...what, where, why, how much, to whom, and how," which guide the clinician or investigator in selecting appropriate intervention strategies (Dodd et al., 2001, p. 673). No relationships are specified among the components of this dimension. This dimension indicates that symptom management strategies may be targeted toward the individual, family, or community group (Dodd et al.). Such a conceptualization of the components of symptom management strategies supports its application in research with children by acknowledging that family members often are involved in the individual's care and in providing support through the illness experience. This application of the model, however, 25 has not been tested in study populations including children experiencing illness-related symptoms and their families. Adherence is defined as "...whether the intended recipient of the strategy actually receives or uses the strategy prescribed," and is depicted as influencing the relationship between symptom management strategies and outcomes (Dodd et al., 2001, p. 674). Although adherence is recognized within the model as being influenced by health care providers and systems, it is regarded as being controlled by the recipient of the intervention strategy (Dodd et al.). Of the three dimensions, the outcomes dimension is the least completely developed. In addition to the status of the symptom, this dimension includes seven other outcomes that the individual may experience as the result of the symptom experience and/or symptom management strategies. These include functional status, emotional status, self-care, costs, quality of life, morbidity and co-morbidity, and mortality (Dodd et al., 2001). No relationships are specified within this dimension; however the authors posit that each of the identified outcomes may be related to the symptom status and to each other (Dodd et al.). Domains of nursing science. The domains of nursing science were included in the 2001 update of the UCSF-SMM to depict the context in which the symptom management process occurs. These include the person, health and illness, and environment domains, all of which are described as influencing each of the dimensions of symptom management (Dodd et al., 2001). The person domain encompasses variables ".. .intrinsic to the way an individual views and responds to the symptom experience," (Dodd et al., p. 670). These include demographic, psychological, sociological, physiological, and 26 developmental variables. Recognition of the significance of development on the symptom experience and the overall symptom management process is essential for research involving children whose physiologic responses to symptoms and perception of symptoms along with their related distress varies based on normal developmental differences. The health and illness domain includes those variables "...unique to the health or illness state of an individual..." (Dodd et al., 2001, p. 670). These are identified as risk factors, health status, and disease and injury. Previous studies by the authors of the UCSF-SMM have demonstrated both direct and indirect relationships between variables in the health and illness domain and the model's three dimensions. The third domain, the environment, encompasses physical, cultural, and social variables representing the "aggregate of conditions" in which a symptom is occurring (Dodd et al., 2001). Physical variables may include the individual's home, work, or the hospital setting. Social variables could include an individual's interpersonal relationships or sources of social support. Cultural variables represent values, practices, and beliefs which arising from the individual's racial, ethnic, or religious group (Dodd et al.). Relationships within the model. The bidirectional arrows illustrated in the model's diagram depict associational relationships among the three dimensions of symptom management. Associational relationships also are depicted among the three components of the symptom experience. Each of the overlapping domains of nursing science is depicted as influencing each of the model's dimensions. Predicted relationships are described only for the health and illness domain which is identified as having both direct and indirect effects on the model's dimensions. The authors also describe alternate 27 relationships among variables such as the impact of gender on cardiovascular outcomes in which morbidity and mortality are worse for women than for men and the impact of developmental stage on interventions in which premature infants are at increased risk for inadequate analgesia based on health care providers' misinterpretation of behavioral cues. (Dodd et al., 2001). The diagram of the UCSF-SMM also depicts relationships among model components which are not clearly explained. Although the domains of nursing science are illustrated as overlapping ellipses in the model diagram, no relationships among the variables within these domains are specified in the authors' description of the model. The diagram also illustrates each domain as connected to one of the dimensions of symptom management, e.g., the health and illness domain is drawn as being connected to the symptom experience domain. The rationale for diagramming the model in this manner is not specified, nor are any proposed relationships between the interconnected domains and dimensions. Boundaries of the model. The UCSF-SMM is a middle range theory. As such, it is limited to the phenomenon of symptom management, yet it is sufficiently broad in its application to encompass a wide scope of illness states and developmental groups. Smith and Liehr (2003) describe a middle range theory as one that contains a "...usable set of ideas..." and one that is less concretely defined than empirical generalizations and less abstract than grand theories (p. 8). The UCSF-SMM is based on the empirical work of its authors and represents their collaborative efforts to develop a framework to guide research and clinical practice. 28 Logical Adequacy The next component of theory analysis addresses logical adequacy. Evaluation of the logical adequacy of a theory attends to the ability to make predictions independent of the theory's content, the extent to which scientists agree on predictions that can be made from the theory, whether or not the theory makes sense, and any logical fallacies that may be present. Predictions independent of content. Table 2.1 presents a summary of the predictive relationships among the major concepts in the model. Because the UCSF-SMM is an interactive, and somewhat iterative model, directional relationships among its components are more difficult to predict. The model diagram's use of bidirectional arrows to illustrate proposed relationships guides the ability to make general predictions of association among the model dimensions; however, the direction of each relationship is not specified and could be either positive or negative. Likewise, associations are predicted between each domain of nursing science and the dimensions of symptom management. The model's diagram suggests that interactive relationships are predicted among variables in the domains of nursing; however, these relationships are not addressed by Dodd and colleagues (2001). Adherence is depicted as predicting the relationship between symptom management strategies and outcomes. Agreement of scientists. A review of seven studies utilizing the UCSF-SMM suggests agreement among researchers regarding predictions that can be made from the model. Although researchers have utilized different aspects of the model as the basis for their research questions and hypotheses, similarities are present among the proposed relationships. Relationships between the variables in the symptom experience and 29 Table 2.1 Proposed Relationships Predicted Within the UCSF Symptom Management Model ENV HI SE SMS OUT AD PER +/- +/- +/- ENV +/- +/- +/- HI +/- +/- +/- SE +/- +/- SMS +/- +/- OUT +/- Person domain (PER) Symptom experience dimension (SE) Environment domain (ENV) Symptom management strategies dimension (SMS) Health & Illness domain (HI) Outcomes dimension (OUT) Adherence (AD) 30 symptom outcome dimensions were proposed among adults with HIV (Voss, 2005), cancer (Dodd, Miaskowski, & Paul, 2001), and traumatic brain injury (Bay & Bergman, 2006). Intervention strategies were predicted to impact symptom outcomes in adults with schizophrenia (Kanungpairn, Sitthimongkol, Wattanapailin, & Klainin, 2007) and children with cancer (Van Cleve, Bossert, Beecrofit, Adlard, Alvarez, & Savedra, 2004). Although investigators have posed similar hypotheses among the components of the UCSF-SMM, a lack of consistency in their selection of variables for inclusion within the components of the model has been present. Agreement has been greatest in selection of variables to measure the symptom experience; however, selection of measures to capture the symptom or symptoms of interest has varied. Four of the seven studies utilized instruments including multiple items which were summed to provide a continuous measure of the symptom experience (Bay & Bergman, 2006; Kanungpairn et al., 2007; Kris & Dodd, 2004; Voss, 2005). Gedaly-Duff and colleagues (2006) combined a daily diary of symptoms along with actigraphy to capture the symptom experience in children with leukemia and their families. Van Cleve et al. (2004) utilized age-appropriate pain measures to measure pain across the continuum of treatment for childhood leukemia. Dodd, Miaskowski, and Paul (2001) measured the symptom cluster of pain, fatigue, and sleep insufficiency using three individual items from the Quality of Life-Cancer scale. Selection of variables for inclusion in the outcomes dimension of the model also has varied across studies. Four studies included a measure of functional status in the outcome dimension (Bay & Bergman, 2006; Dodd, Miaskowski, & Paul, 2001; Van Cleve et al., 2004; Voss, 2005). Measures of functional status included the Karnofsky 31 Performance Scale and subscales of the Profile of Mood States (POMS) and the SF-36 scales. Kanungpairn and colleagues (2007) located postintervention auditory hallucinations within the outcome dimension, yet this variable was measured using the Characteristics and Severity of Auditory Hallucinations scale that was used to measure the symptom experience prior to the intervention. Four studies addressed variables in the person domain and were in general agreement regarding the types of demographic variables to be included. Voss (2005) included disability and drug use, both scored categorically, in the person domain although these variables also could have been placed in the health and illness domain. The variables that Voss selected for inclusion in the health and illness domain were very specific to HIV/AIDS, which may explain why these nondisease specific-variables were retained in the person domain. Only two studies selected variables for inclusion in the environment domain, and the selected variables were significantly different. Gedaly-Duff and colleagues (2006) located the parents within the environment domain of children undergoing treatment for leukemia, which could be argued as consistent with the social aspect of this domain. Voss (2005) included income and health insurance as environment variables. The rationale for selection of these two variables as representing the environment domain was not stated. The UCSF-SMM authors' description of the environment domain emphasizes physical, cultural, and social variables; however, the location of variables relating to access to care are not clearly specified. Making sense. Utilization of the UCSF-SMM as a guiding framework for research across illness states and developmental stages provides support for its logical sense. That 32 the model itself is a collaborative effort derived from research and practice further contributes to its ability to make sense across practice settings. Logical fallacies. Despite its strengths in agreement and making sense, several logical fallacies are present within the UCSF-SMM. One of these issues relates to the aspect of time across the illness trajectory (Henly, Kallas, Klatt, & Swenson, 2003; Van Cleve, Bossert, & Savedra, 2002). Time is reflected in the symptom management strategies as a component of "when" an intervention is delivered; however, changes in the symptom experience and process of symptom management across the duration of an illness are not specified (Dodd et al., 2002; Henly et al.). To address this limitation of the model, Van Cleve and colleagues (2002) adapted the model to express the three dimensions of the model as occurring along a spiral across the 1st year of treatment for leukemia. Another potential logical fallacy within the model is the assumption that the interpretation of symptoms by the parent or caregiver of nonverbal patients (including children) is adequate for initiating intervention. Previous studies involving symptoms in children with cancer suggest that more subjective symptoms are less recognized by parents and that children with greater levels of pain are at increased risk for having their pain under-recognized by health care providers (Lo & Hayman, 1999; Miser, Dothage, Wesley, & Miser, 1987). Although parent/caregiver report of a symptom may be regarded as adequate for initiating intervention, report of the absence of a symptom may not indicate the absence of a need for intervention. The authors acknowledge that the UCSF-SMM, in its present state, is limited in its ability to capture the experience of multiple symptoms occurring simultaneously 33 (Dodd et al., 2001). Dodd, Miaskowski, and Paul (2001) utilized the model to explore the impact of the symptom cluster of pain, fatigue, and sleep insufficiency on functional status; however, the symptom cluster itself was treated as a single entity within the model. The model does not guide the researcher or clinician to identify which symptom should be addressed initially. The UCSF-SMM does not explain the interaction among multiple symptoms such as how the management of one symptom may have the potential to exacerbate or relieve other symptoms. The UCSF-SMM does not distinguish between acute and chronic symptoms (Dodd et al., 2001). The model also does not clarify where to locate other symptoms within the model when a given symptom is the focus of attention and intervention. One plausible approach may be to locate chronic symptoms within the health and illness domain that exerts an influence on the process of managing an acute symptom. As the model undergoes further testing and development, attention to multiple symptoms and the nature of these symptoms is warranted. Usefulness As a middle range theory addressing the phenomenon of symptom management, the UCSF-SMM is proposed to be relevant across populations experiencing illness-related symptoms and even those identified as being at risk for symptoms (Dodd et al., 2001). To date, the model has served as a useful framework for research among adults receiving postsurgical care, cancer, HIV/AIDS, mental illness, and traumatic brain injury as well as children with cancer (Bay & Bergman, 2006; Dodd, Miaskowski, & Paul, 2001; Gedaly-Duff, Lee, Nail, Nicholson, & Johnson, 2006; Kanungpairn, Sitthimongkol, Wattanapailin, & Klainin, 2007; Kris & Dodd, 2004; Van Cleve et al., 2004; Voss, 2005). 34 The majority of these studies are descriptive studies that have sought to identify relationships among variables within the model's components. Knowledge gained from these studies adds to further understanding of different aspects of the symptom management process in selected populations which then can lead to the development of interventions that can be tested. Because of its foundation in research and clinical practice, the UCSF-SMM has the potential to influence nursing practice. The model provides a conceptual framework for understanding relationships between factors influencing the symptom experience as well as the larger contextual factors influencing symptom management. It also can guide nursing interventions targeted at influencing the context in which symptoms are occurring and the development of symptom management strategies. This knowledge also has the potential to influence institution-based procedures relating to symptom management and to create care environments conducive to minimizing symptoms. Generalizability A theory's generalizability relates to the focus of its content and the extent of its boundaries. By its definition, the content of the model is limited to the process of symptom management and the context in which it is occurring. As presented in its assumptions, however, the UCSF-SMM is applicable across multiple settings. It is regarded to be applicable in both symptomatic and presymptomatic states as well as across developmental stages. The model also supports the development of intervention strategies that can be targeted at individuals or groups. 35 Parsimony In its simplest sense, the UCSF-SMM can be reduced to three dimensions of symptom management and three domains of nursing science, giving it a semblance of parsimony. The number of variables for consideration within each of the domains of nursing science and the intricate interrelationships within each dimension of symptom management, however, significantly increases the complexity of the model. As a process-oriented model, multiple relationships are encompassed within its scope, with some of these occurring simultaneously. This level of complexity can be regarded as both a strength and a limitation of the model. Each of the model's components and proposed relationships are well described, which supports the researcher or clinician in locating variables of interest in the model for potential research or clinical application. These definitions also serve to protect the integrity of the model as the authors intended. Conversely, the number of relationships and variables within the model require careful review of the model's definitions to ensure correct application of the model. The scope of processes contained within the model also make full application difficult. Testability Support for the testability of the UCSF-SMM is its foundation in the empirical work of its authors (Dodd et al., 2001). Each of the components of the model is conceptualized and operationalized to support selection of variables and relationships to be tested empirically. Since its publication, the model has demonstrated the ability to generate research questions and hypotheses across a range of illness states. These attributes add to the overall validity and strength of the model. 36 The scope and complexity of the UCSF-SMM, however, challenge the full testability of the model. The extent of the model limits the number of proposed relationships that can be investigated with in a single study. Although several studies have addressed multiple components of the model, no studies have tested the model in its entirety. Dodd, Miaskowski, and Paul (2001) investigated the impact of the experience of a symptom cluster (pain, fatigue, and disturbed sleep) on the outcome of functional status, but did not include symptom management strategies. Age was included in the regression analyses, yet no further attention was given to relationships between the domains of nursing science and the symptom experience or outcome. Voss (2005) also explored relationships between the symptom experience and outcomes among individuals with HIV/AIDS. He did not address intervention strategies in his study, although relationships between variables within the domains of nursing science and fatigue were explored. Kanungpairn et al. (2007) explored the impact of an intervention on the frequency and severity of auditory hallucinations in patients with schizophrenia, yet they did not address the domains of nursing science in the study's hypotheses or analyses. Empirical Support for the UCSF-SMM As a deductive model, the UCSF-SMM is derived from the empirical work of its authors (Dodd et al., 2001). Findings reported in the studies selected for review in this paper also provide empirical support of selected relationships within the model. Because each study sought to investigate only a component of the model, comparison of empirical support across studies is more challenging. Investigators' selection of variables for inclusion in each of the model's components also has not been consistent across studies, which also limits comparisons of findings across studies. Although investigators may 37 have proposed similar relationships among model components, the selection of variables for inclusion has not been consistent across studies. Relationships Between Dimensions of Symptom Management Symptom experience and outcomes. Proposed relationships between the symptom experience and outcomes dimensions have been supported in several studies. Dodd, Miaskowski, and Paul (2001) found that initial functional status, pain, and fatigue were predictive of change in functional status in adults undergoing outpatient chemotherapy. Age also was predictive of change in functional status; however, the authors did not specifically locate age within the components of the model selected for their study. Neurobehavioral symptoms in adults who were within two years of experiencing mild to moderate traumatic brain injuries accounted for 40%, 32%, and 51% of the variance in the outcomes of tension and anxiety, anger and hostility, and perceived stress, respectively (Bay & Bergman, 2006). Perceived symptoms in adults with HIV explained 18% and 26.4% of the variance in the respective outcomes of mental and physical health (Voss, 2005). Fatigue was identified as an independent predictor of physical health. Components of symptom management, symptom experience, and outcomes. Kanungpairn and colleagues (2007) reported decreased auditory hallucinations in adults with schizophrenia who received a patient-focused, 10-week behavioral management program that included self-monitoring. Although these authors located the postintervention assessment of auditory hallucinations in the outcomes dimension, they used the same instrument that was used preintervention to measure the symptom experience. Alternately, this study's conclusions could have indicated that effective an 38 effective symptom management strategy resulted in reducing the symptom experience dimension. Perceived effective pain management was associated with decreased pain intensity across the treatment continuum for older children (8-17 years) with acute lymphoblastic leukemia (Van Cleve et al., 2004). This pattern was not observed in younger children (4-7 years), suggesting that further investigation of developmental differences in reporting pain perception and identification of effective pain management may be warranted. Relationships Among the Domains of Nursing Science and Dimensions of Symptom Management Although the domains of nursing science were added in 2001, minimal attention has been given to them in subsequent studies. Proposed relationships between variables located in the domains of nursing science and fatigue (symptom experience) and physical and mental health (outcomes) were tested in patients with HIV (Voss, 2005). Women with HIV experienced greater fatigue than men. Variables in the person and environment domains (paid employment, adequate insurance and income, and living alone) were associated with less fatigue. Adequate income and health insurance did not, however, predict physical or mental health (outcomes). HIV-specific health and illness variables including years since HIV infection, CD4 cell count, and viral load explained 4.7% of the variance in mental health. 39 Directions for Future Research Directions for future research involving the UCSF-SMM include greater clarity, specificity, and consistency in locating variables within specific aspects of the model. Additional primary studies with carefully considered variables and associated measures are warranted. As future studies are undertaken, comparison of types of measures, e.g., single item, complete self-report instrument, direct observation, use of technology, etc. can be made within and across study populations to evaluate their adequacy. Additional considerations for research recognized by the UCSF-SMM authors include attention to symptom clusters and determining when symptom management can be defined as complete or successful (Dodd et al., 2001). In its present state, the model reflects an ongoing, interactive process but does not include a mechanism to identify when intervention strategies would no longer be needed or how they would be modified over the course of an illness. Considerations for Research With Children Although the UCSF-SMM is based largely in research and practice involving adults, it has been developed with the intent of being used in pediatric populations. The impact of development is acknowledged in the person domain of nursing science, and the components of symptom management strategies recognizes that symptom management strategies acknowledges that the child's parent may be targeted as the one intervening on behalf of the child. As the model is applied in research with children, attention to measurement of symptoms in children, consideration of the parents and family, and testing across pediatric populations is warranted. Report and Measurement of Symptoms in Children As previously addressed, the UCSF-SMM assumes self-report is the gold standard for identifying the existence of symptoms and that report of symptoms by a parent or caregiver is accurate for the purpose of intervening (Dodd et al., 2001). Although a parent or caregiver's report of the existence of a child's symptom may be regarded as sufficient to initiate an intervention, reliance on proxy-reporting may result in under-recognition and inadequate management of more subjective symptoms (Linder, 2008). Younger children tend to be more present-oriented and may emphasize the physical rather than mental aspects of symptoms (Hockenberry et al., 1998). These children also tend to have more difficulty responding to multiple individual items and may benefit from being asked to describe overall feeling states (Woodgate, Degner, & Yanofsky, 2003). This understanding of children's perception of symptoms calls for the development of reliable and valid instruments for the assessment of symptoms in children, including the use of less traditional methodological approaches to be able to capture the child's perspective. Utilization of technologic devices such as wrist actigraphy in the assessment of disturbed sleep also may support objective measurement of symptoms in young children (Gedaly- Duffet al., 2006). Inclusion of Parents and Family Members Central to research and practice involving children is consideration of the parents and the impact of the child's illness on the family. Although the model recognizes that parents may be involved in symptom management strategies for the child, it is not fully clear where the parents and family members could be located in other model components. The environment domain includes the social variables, represented as the individual's 41 support network and interpersonal relationships, suggesting that parents and family members could be located in this component (Dodd et al., 2001). Gedaly-Duff and colleagues (2006) placed parents in the environment domain of the model; however, their published analyses did not address this relationship. The current focus of the outcomes dimension is the individual. Although this dimension includes eight significant outcomes for consideration, it does not specifically address the impact of the symptom and its management on other family members, which can be particularly significant in children experiencing chronic illness (Woodgate, 2006). Such considerations raise questions as to how to include child and family outcomes as part of the symptom management process. Application Across Pediatric Populations Application of the UCSF-SMM is needed across pediatric populations experiencing both acute and chronic symptoms. To date, its use has been limited to children with cancer; however, these studies have been largely descriptive and have provided only limited testing of the model's proposed relationships (Gedaly-Duff et al., 2006; Van Cleve et al., 2004). Studies across pediatric populations have the potential to increase understanding of the model's generalizability, usefulness, and testability in children. Such studies also may lead to further refinement of the model for application in children and help to clarify its logical adequacy in this population. Summary The UCSF-SMM is a process-focused middle range theory that locates symptom management in the context of the domains of nursing. As a noncategorical model, it has 42 been applied across a variety of illness states in adults and has demonstrated initial applicability in research involving children with cancer (Gedaly-Duff et al., 2006; Van Cleve et al., 2004). Strengths of the model include its attention to the larger context in which symptoms are occurring and its ability to inform research hypotheses and clinical practices. Limitations of the model are its lack of parsimony limiting the extent to which the model can be tested, and its failure to capture the changing nature of symptoms across the illness trajectory. Although the UCSF-SMM is conceptualized as being applicable in research with children, additional testing is necessary to validate the model's proposed relationships in pediatric populations. Additional refinement is necessary to investigate and clarify the role of proxy-reporting of children's symptoms by parents or other caregivers, including health care professionals. Clarification of the location of parents and family within the model's components is warranted as well. As these areas are explored and addressed across pediatric populations, the UCSF-SMM has the potential to guide nursing research and practice to improve children's symptoms and overall quality of life. Acknowledgements This work was supported by a National Research Service Award from the National Institute for Nursing Research, F31NR010175-01; a Doctoral Scholarship in Cancer Nursing from the American Cancer Society DSCN-06-204-1; and a Dissertation Grant Scholarship from the Western Institute of Nursing and the Council for the Advancement of Nursing Science. The author also would like to recognize Becky J. Christian PhD, RN, for her support in the development and review of this manuscript CHAPTER 3 METHODS Study Design Case Study Design This exploratory study used a multiple-case study design to investigate sleep-wake patterns in children with cancer over a single admission for inpatient chemotherapy. Case study research is a recognized method of scholarly inquiry that is utilized to investigate a given phenomenon of interest within its context (Yin 2003a, 2003b). In contrast to more traditional, experimental research, case study research utilizes a within subjects approach with an emphasis on replication across cases rather than identifying differences between groups. Case study research has been utilized among multiple disciplines, including medicine and nursing. Case study research differs from traditional experimental research in that it seeks to describe relationships within the larger context in which a given outcome variable is occurring rather than focusing on relationships between only a few select variables within a controlled setting. Because of its emphasis on the context in which the phenomenon is occurring, case study research does not require investigator control of behavioral or contextual events (Yin 2003a, 2003b). The focus of analysis is on describing relationships between variables within individual cases and replication across cases. 44 The case study design, with its emphasis on the individual, was well-suited to this study due to its relatively small sample size, and because the magnitude of sleep disturbances and factors contributing to disturbed sleep among hospitalized children with cancer are not well established (Docherty, Sandelowski, & Preisser, 2006; Yin 2003b). Assumptions of Case Study Research Selection of variables for case study research. One of the unique distinctions of case study research is the number of variables of interest. With its emphasis on exploring the larger context in which the phenomenon of interest is occurring, a key underlying assumption is that there will be more variables of interest than data points (Yin 2003a, 2003b). Also central to this methodology is that, because of its complexity, the selection of variables is guided by theory. As previously described, the selection of variables for consideration was guided by the UCSF Symptom Management Model (Dodd et al., 2001). A unique aspect of the UCSF-SMM is its emphasis on context, which makes it well-suited to case study research. The model places the process of symptom management in the context of nursing science: 1) the person, 2) the environment, and 3) health and illness. The UCSF-SMM and review of current literature guided the selection of contextual variables included in the study proposal. Sources of data for case study research. Additional key assumptions of case study research are that multiple sources of evidence and that multiple methods of data collection will be used (Yin 2003a, 2003b). This approach to data collection is essential to the construct validity of the case study through triangulation of data and data collection methods. Triangulation increases the rigor of the case study by increasing the 45 completeness of data to describe the phenomenon of interest and the context (McDonnell, Jones, & Read, 2000). The methods of data collection for this study included direct measures using technologic devices, parent and child subjective reports using established instruments, review of the child's medical record, and a genetic sample. Together, these sources of data served to provide a more thorough description of the environmental, intrapersonal, and health/illness context in which the child's sleep-wake patterns are occurring. Study sample selection for case study research. Selection of the sample size for a multiple-case study is based on a replication logic, rather than a sampling logic which is used for experimental research (Yin 2003a, 2003b). Selection of the number of cases (participants) to be included was based on the ability to replicate findings across cases either literally (similar results) or theoretically (contrasting results, but in a predictable manner). A sample size of 15 was selected to be able to generate a sample that was likely to demonstrate variation with regard to behavioral and disease-related variables, yet be able to demonstrate literal or theoretical replication across cases. Although the study included a heterogeneous sample with regard to specific disease type and treatment plan, these differences were acceptable within a multiple-case study because the emphasis is placed on identifying patterns of relationships among variables within each individual. Despite differences in disease and treatment plan, each participant received a similar standard of nursing care during the admission. All children received similar pre-, intra-, and postchemotherapy nursing cares. Nighttime nursing care routines also were similar across treatment regimens, again adding to similarities within the care environment. 46 A uniqueness of the proposed study was its use of a developmental science perspective. Because of the difficulty to obtain an adequate sample size for an experimental study, many nursing studies involving children with cancer have included broader age ranges which limit the ability to identify uniquenesses based on the child's developmental level (Linder, 2008). Use of a multiple-case study design and sample selection based on replication logic rather than sampling logic supported the opportunity to explore symptom characteristics within a given developmental age group. Generalization of case-study research findings. In contrast to experimental research which seeks to generalize findings to populations, the external validity of case study research is its generalizability to theory through replication across individual cases (Yin 2003a, 2003b). Replication is defined as the ability to replicate findings across cases. Replication does not imply that findings are identical across cases but that relationships between variables are present in a manner that is consistent with theory in each individual case. Such replication of findings contributes to existing theory which in turn is used to guide clinical practice and to inform hypotheses for subsequent, larger intervention studies. Study Setting The study setting was Primary Children's Medical Center (PCMC), a Children's Oncology Group institution, in Salt Lake City, Utah. This 250-bed free-standing children's hospital serves the largest geographic region of any children's hospital in the United States. The Hematology/Oncology/Transplant Service Line treats around 300 children and adolescents for cancer each year, with approximately 150 of those being newly diagnosed patients (Primary Children's Medical Center, 2007). Approximately 25 47 children 5 to 12 years of age receive inpatient chemotherapy annually. The 24-bed inpatient Immune Compromised Unit serves children with immune-compromised conditions including cancer, hematologic disorders, and hematopoietic stem cell and solid organ transplant. All rooms are private with a private bathroom. The majority of children served at PCMC are from the Salt Lake Valley. PCMC also serves children throughout the Intermountain West. Study Participants Inclusion Criteria Inclusion criteria were children between 5 and 12 years of age receiving inpatient chemotherapy for 3 or more days as treatment for cancer. Participants included children receiving treatment for both a primary diagnosis of cancer and recurrent disease. The limited age range was selected to decrease variance in data pertaining to sleep-wake patterns and increase the ability to compare findings across other pediatric centers. Excluding children less than 5 years of age reduced variance related to developmental differences in sleep unique to younger children. Because adolescent sleep patterns are more likely to be influenced by personal choice in determining bedtime hours, children 13 years and older were excluded (Mindell & Owens, 2003). The proposed sample size of 15 supported the likelihood of identifying variability within the biobehavioral variables and providing data to inform the long-term objectives of a larger study. Because the procedural routine for inpatient chemotherapy administration is similar across treatment protocols, and the number of children with each cancer is small; a heterogeneous sample was used. 48 Exclusion Criteria Exclusion criteria included surgery during the admission or active treatment for a secondary medical condition requiring nursing care interventions in addition to those related to chemotherapy administration. Children with central nervous system tumors were excluded because of an increased likelihood of sleep disturbances resulting from pathologic consequences of the tumor itself (Rosen, Bendel, Neglia, Moertel, & Mahowald, 2001). Children with major developmental delays and those previously diagnosed with sleep disorders also were excluded. Study Variables and Measures Consistent with a case study design, multiple sources of evidence were collected to describe sleep-wake patterns and the environmental context (Yin, 2003b). This approach supported data triangulation to increase reliability of findings. Data sources included the child's medical record, parent report, child self-report, and physiologic measures. Demographic and Clinical Variables Demographic and clinical variables were collected through parent/guardian interview and review of the child's medical record. Demographic variables included the child's gender, age, grade, family composition, and race/ethnicity. Clinical variables included cancer classification and stage, illness severity, date of diagnosis, treatment protocol, chemotherapy medications, other medical conditions, and current medications. Admission absolute neutrophil count (ANC), hemoglobin, and hematocrit values and subsequent values were identified. The child's sleep patterns at home were identified by 49 parent report for comparison with sleep-wake patterns identified in the hospital. Vomiting episodes and ratings for nausea and pain were recorded from the nursing flowsheet. Nausea and pain were rated per institutional guidelines. Antiemetic and analgesic doses along with the total number of medication doses per 12-hour shift were recorded. Sleep-Wake Patterns Wrist actigraphs. Wrist actigraphs and sleep diaries were used to measure sleep-wake patterns. MicroMini Motionlogger® actigraphs (Ambulatory Monitoring, Inc., 2006) were worn continuously on the child's nondominant wrist during the data collection period. The MicroMini Motionlogger® is about the size of a wristwatch and contains a piezoelectric sensor that generates a voltage when the actigraph is moved. The Zero Crossing Mode, which is a measure of frequency of movement and is the most frequently used mode of operation in published literature, was used (Ambulatory Monitoring, Inc.; Ancoli-Israel, Cole, Alessi, Chambers, Moorcrofit, & Pollak, 2003). Sleep/wake states are identified by sampling the individual's movements several times per second. Data are transformed digitally and stored in 1-minute epochs for future analysis. The American Academy of Sleep Medicine (AASM) recognizes actigraphy as a reliable and valid measure to detect sleep and assess disturbed sleep patterns, circadian rhythm disorders, and intervention effects (Ancoli-Israel et al., 2003; Littner et al., 2003). Studies of adults and children report actigraphy as highly correlated with polysomnography, the gold standard for measuring sleep. Actigraphs continuously measure motion - a proxy measure of sleep-wake periods, based on the assumption that fewer movements occur during sleep than during periods of wakefulness (Ancoli-Israel et 50 al.). Sleep-wake states are identified by sampling the individual's movements several times per minute. Actigraphy has been used to measure sleep-wake patterns objectively for over 20 years in neonates to adults and is reported to be more accurate and yield greater rates of adherence than self-report diaries alone (Berger et al., 2002, 2003; Docherty, Sandelowski, & Preisser, 2006; Eissa, Poffenbarger, Portman, 2001; Gedaly-Duff et al., 2006; Hinds et al., 2007b; Kline, 1999; Korte, Wulff, Oppe, & Siegmund, 2001; Kushida, Chang, Gadkary, Guilleminault, & Dement, 2001; Landis, Frey, Lentz, Rothermel, Buchwald, & Shaver, 2003; Tworoger, Davis, Vitiello, Lentz, & McTiernan, 2005). The AASM recommends conducting sleep measurements over at least 3 consecutive 24-hour periods, and studies of infants and children suggest that using actigraphy for 3 to 5 days is well-tolerated (Ancoli-Israel et al. 2003; Docherty, Sandelowski, & Preisser, 2006; Gedaly-Duff et al., 2006; Hinds, Hockenberry, Rai, Zhang, Bassem et al., 2007; Kline, 1999; Littner et al., 2003). Actigraphs do not limit normal activity and may be removed to keep the device from getting wet. A major advantage of actigraphy is that it allows continuous ambulatory monitoring of sleep-wake pattern and is not limited to nighttime hours or a laboratory setting. Actigraphy facilitates objective measurement of sleep-wake patterns in very young children who are unable to self-report sleep disturbances and, as a result, often are excluded from studies. A limitation of actigraphy is that it is more specific for detecting sleep rather than wake states (Kushida et al., 2001). This poses a risk of overestimating sleep time, which is of particular significance with fragmented nighttime sleep. 51 Sleep diaries. Sleep diaries were maintained for comparison with actigraphy data and to support analyses as recommended by the American Academy of Sleep Medicine (Littner et al., 2003). Diaries identified the child's time to bed and morning waking time. The parent/guardian assumed responsibility for completion of the diaries with children encouraged to participate as much as possible. Studies of chronically ill children indicate that children as young as 6 years of age are capable of completing a self-report diary over at least a 3-day period successfully (Franck et al., 1999; Shapiro et al., 1995). Environment Variables Light and temperature. Light and temperature levels in the child's room were measured continuously using the HOBO® U12-012 Temp/RH/Light/External Channel Data Logger (Onset Computer Corporation, 2005). This device is 2.3 x 2.9 x 0.85 inches. The internal sensors for temperature and light were used. The data logger recorded sensor measurements at 30-second intervals and stored the measurements in its internal memory. The data logger temperature sensor measures from 0 to 50 degrees centigrade (°C) with a resolution of 0.03°C and accuracy of ± 0.35°C. The light sensor measures light intensity with a range of 1 to 3,000 lumens/ft . The recommended light level for reading printed material is 30 lumens/ft2; the recommended light level for close work in 9 9 laboratories is 100 lumens/ft . Complete darkness is measured as zero lumens/ft . Sound. Sound was measured using Extech Instruments model number 407736 digital sound pressure level (SPL) meter (Extech Instruments Corporation, 2005). The SPL meter is 9.54 x 2.68 x 1 inches and was connected to the data logger using the external channel input. Measurements were obtained at 30-second intervals. 52 The device measures sound pressure level ranges of 35 to 100 decibels (dB) with 0.1 dB resolution and an accuracy of ±1.5 dB. The human ear can detect a sound level difference of approximately ldB in laboratory conditions. Among most individuals under 60 years of age, the minimum detectable difference is closer to 3dB. Sound levels of 35 dB or less are recommended by WHO for adequate sleep (World Health Organization, 1980). Average conversational noise is about 40 to 60 dB. A busy street or orchestra measures around 70dB, and a subway registers at 100 dB. Fatigue Fatigue was measured daily using 24-hour versions of the Fatigue Scale: Child Version (FS-C) (Appendix A) and the Fatigue Scale: Parent Version (FS-P) (Appendix B) developed by Hockenberry, Hinds and colleagues using their theoretical framework based on their research on cancer-related fatigue in children and adolescents (Hinds et al., 1999; Hinds, Hockenberry, Rai, Zhang, Razzouk, et al., 2007; Hockenberry et al., 2003; Hockenberry-Eaton et al, 1998). These scales take approximately 5 to 8 minutes to complete. The FS-C is a 14-item self-report scale addressing the child's report of fatigue-related symptoms in the past 24 hours. Scale items are rated from 1 = "not at all" to 5 = "a lot." Individual item scores are summed to provide a total fatigue score with a possible range of 14 to 70. The FS-P has 17 items assessing the parent's perception of the intensity of the child's fatigue in the past 24 hours. Scale items are rated from 1 = "not at all" to 5 = "always." Individual item scores are summed to provide a total fatigue score with a possible range of 17 to 85. The second component of the FS-P contains 18 items 53 addressing perceived sources of the child's fatigue. Individual scale items are rated from 1 = "not at all" to 5 = "always" and then summed with possible total scores ranging from 18 to 90. Cronbach's alpha coefficients for the FS-C and FS-P reported by Hinds, Hockenberry, Rai, Zhang, Razzouk, et al. (2007) ranged from 0.64 to 0.72 and 0.76 to 0.96, respectively. Previously reported completion rates for 2- to 4-day inpatient admission were 96% for the FS-C and 100% for the FS-P (Hinds, Hockenberry, Rai, Zhang, Razzouk, et al.). Temperament Temperament was measured using age-appropriate versions of the Carey Temperament Scales (CTS), the Behavioral Style Questionnaire (BSQ) for children 5 to 7 years of age, and the Middle Childhood Temperament Questionnaire (MCTQ) for children 8 to 12 years of age (Behavioral-Developmental Initiatives, 2005). The CTS are norm-referenced, age-based scales assessing the nine dimensions of temperament based on the work of Chess and Thomas (Chess & Thomas, 1996; Hegvik, McDevitt, & Carey, 1982; McDevitt & Carey, 1978). Scales were completed by the parent, contained 100 items, and took approximately 15 to 20 minutes to complete. CTS items are rated on a 6-point scale from which a continuously measured score for each temperament dimension is obtained and reported. Scores are generated using computer-based software (Behavioral- Developmental Initiatives) and are reported as both standardized scores and z-scores. The CTS have been widely used in studies of healthy and chronically ill infants and children in the United States and internationally (Helgadottir & Wilson, 2004; Jorm et al., 2000; McClowry, 1990; Suskauer, Cintas, Marini, & Gerber, 2003; Vacha & Adams, 54 2005; Wilson, Megel, Fredrichs, & McLaughlin, 2003). Previously reported split-half reliabilities of the nine temperament dimensions measured by the BSQ ranted from 0.47 to 0.80 with a median of 0.70, and test-retest reliabilities were 0.67 to 0.94 (median 0.81) (McDevitt & Carey, 1978). Cronbach's alpha coefficients for the dimensions of the MCTQ ranged from 0.71 to 0.87 (median 0.81), and test-retest reliabilities ranged from 0.89 to 0.93 (median 0.88) (Hegvik, McDevitt, & Carey, 1982). DRD4 and 5-HTTLPR Genotyping DNA was extracted from children's cheek swab samples using the protocol and reagents supplied with the Qiamp kit from Qiagen Corporation (Qiagen Inc., 2005). Samples were obtained at admission, a time when oral mucosa were likely to be less vulnerable to trauma. Specimens were stored in a locked container and frozen until processing. PCR amplification for genotyping was completed under the supervision of Yvette Conley, PhD, (molecular biologist and dissertation committee member) who provided laboratory support at the University of Pittsburgh School of Nursing. PCR amplification of the region containing the variable number of tandem repeats (VNTR) in exon III of the DRD4 gene followed the protocol described by Ronai et al. (2000). Their procedure expanded genotyping to the identification of heterozygous and homozygous alleles rather than merely distinguishing between the presence of a 7 x 48 base pair repeat versus a non- 7 x 48 base pair repeat. Genotyping was accomplished using the HotStarTaq polymerase kit using 1 pM of each primer (sense primer: 5'-TGC TCT ACT GGG CCA CGT TC-3' and antisense primer: 5'-TGC GGG TCT GCG GTG GAG TCT-3'), 1 ng of DNA template, 200 pM dATP, dTTP, dCTP, and 100 pM dITP in a final volume of 20 pL (Ronai et al., 2000). Thermocycling was initiated at 95°C for 10 minutes 55 to activate the enzyme and to denature genomic DNA. This was followed by 35 cycles of 1 minute denaturation at 95°C, 1 minute annealing at 60°C, and 1 minute extension at 72°C. Final extension was carried out for 10 minutes at 72°C followed by cooling the samples to 4°C. The amplified products were separated by automated ultrathin-layer gel electrophoresis using 1% agarose - 2% linear polyacrylamide composite matrix, containing 50 nM ethidium bromide. Consistent with previous studies, short DRD4 alleles were defined as having 2 to 5 repeat units, and long DRD4 alleles were defined as having 6 to 8 repeat units (Ronai et al., 2000). PCR amplification of the 5-HTTLPR region followed the protocol described by Lesch et al. (1996). Oligonucleotide primers flanking the 5-HTTLPR and corresponding to the nucleotide positions -1416 to -1397 (stpr5, 5'-GGC GTT GCC GCT CTG AAT GC) and -910 to -888 (stpr3 5'-GAG GGA CTG AGC TGG ACA ACC AC) of the 5-HTT gene 5'-flanking regulatory region were used to generate 484- or 528- base pair fragments. PCR amplification was carried out in a final volume of 30 |il consisting of 50 ng of genomic DNA, 2.5 raM deoxyribonucleotides (dGTP/7-deaza-2'-dGTP= III) 0.1 |ig of sense and antisense primers, 10 mM tris-HCl (pH 8.3), 50mM KCL, 1.5 mM MgC12, and 1U of Taq DNA polymerase. Annealing was carried out at 61°C for 30 seconds, extension at 72°C for 1 minute, and denaturation at 95°C for 30 seconds for 35 cycles (Heils, Teufel, Petri, Stoeber, Riederer, Bengel, & Lesch, 1996; Lesch et al., 1996). Amplified products were separated on 2.5% agarose gel stained with ethidium bromide and visualized under UV light. Consistent with previous studies, short 5-HTTLPR alleles were defined as having 14 repeat units present. Long 5-HTTLPR alleles were defined has having 16 repeat 56 units present (Auerbach et al, 2001, 1999; Ebstein et al., 1998; Lakatos et al., 2003; Lesch et al., 1996). Study Procedures Institutional Review Board Approval Because this study included patients with cancer, obtaining Institutional Review Board Approval was a multistep process with approval first being granted by the Clinical Cancer Investigation Committee at the Huntsman Cancer Institute. Institutional Review Board approval for the study also was granted by the University of Utah and Primary Children's Medical Center. Participant Recruitment Children scheduled for admission for inpatient chemotherapy were reviewed weekly to evaluate initial eligibility. An introductory letter explaining the study was distributed to families of eligible children. Families also were informed about the study during a routine preadmission phone call by the service line admission coordinators. During this phone call, families were asked if the investigator had permission to approach them about the study upon their admission to the inpatient unit. No families declined to be approached. The investigator approached interested parents/guardians and children after the child's admission to the inpatient unit to explain the study purpose and obtain informed consent. Parental permission was obtained for all child participants with written assent obtained from children 7 years of age and older. Informed consent also was obtained from parents providing data regarding their children. The data collection process was explained to participants in developmentally appropriate terms, as well as to their parents or guardians. Because data collection occurred continuously over several days, attrition was a risk. The most likely source of attrition was anticipated to be the child's inability or unwillingness to wear the actigraph for the duration of the data collection period. Previous studies suggest that actigraphy is well-tolerated in infants and children, and that adequate measurements over multiple consecutive days can be obtained (Ancoli-Israel et al., 2003; Docherty, Sandelowski, & Preisser, 2006; Gedaly-Duff et al., 2006; Hinds, Hockenberry, Rai, Zhang, Bassem et al., 2007; Hinds, Hockenberry, Rai, Zhanr, Razzouk et al., 2007; Kline, 1999; Littner et al., 2003. The AASM recommends that sleep measurements be conducted over at least three consecutive 24-hour periods (Ancoli-Israel et al., 2003). Only one child withdrew after initial enrollment because she was unwilling to wear the actigraph. Data Collection Procedures Data were collected between March 2008 and January 2009. Data were collected from children with cancer and their parents during a scheduled hospital admission for chemotherapy. No additional travel time related to data collection was required of participants. Study participation did not interfere with routine care delivery of care or delay treatment. Study participation also did not increase the child's length of hospitalization. Investigator responsibilities. The investigator conducted initial interviews with the child's parent and collected demographic data. The investigator was responsible for calibration and maintenance of study equipment throughout the study. At the onset of data 58 collection, the investigator placed the actigraph on the child's wrist, placed the data logger in the child's hospital room, and obtained the cheek swab sample for genotyping. The data logger was placed on the bedside supply cart in each child's room. The bedside supply cart was maintained in a consistent location in each child's room and was approximately 48 inches from the child's bedside. The investigator assessed the child's wrist for evidence of irritation from the actigraph each day. No evidence of irritation or skin breakdown was observed during each data collection period. The investigator reviewed and entered clinical variables from the child's nursing flowsheet and medication administration record (MAR) each day. Child participant responsibilities. Child participants wore the actigraphs on their nondominant wrists continuously during data collection. Actigraphs were able to be removed for 1 hour each day to prevent the device from getting wet during personal cares. Actigraph data were downloaded into a laptop computer at completion of data collection. Children completed the FS-C daily starting on the day of admission. Children were allowed the option of completing the scale independently or having items read aloud. Children were encouraged to maintain sleep diaries with parent support. Parent participant responsibilities. Parents maintained daily sleep diaries during the data collection period with input from the child. Parents completed the CTS appropriate for their child's age at the time of admission. Parents also completed both components of the FS-P daily starting with the day of admission. Staff nurse responsibilities. Staff nurses did not have unique study responsibilities other than routine institutional protocols and procedures. Staff nurses assessed and documented nausea, vomiting and pain per institutional procedure. Staff nurses also 59 documented medication administration, including chemotherapy per institutional procedure. Data Collector Training and Quality The investigator provided inservices to service line nursing staff to review the study protocol and staff nurse responsibilities. The investigator also provided one-on-one education regarding the study procedures to other multidisciplinary service line team members. Parents and children received written instructions regarding actigraph use and for completing self-report scales. The investigator reviewed the scales daily with parents and children for accuracy and completeness and to identify and review missing data. Healthy community volunteers and 2 children with cancer were recruited to pilot test study procedures to identify and address challenges in administering the study protocol. Because the study procedures did not change after pilot testing with the 2 children with cancer, these 2 children were included as study participants. CHAPTER 4 RESULTS This chapter presents a summary of the study results. It begins with a discussion of data management and analysis, including data retrieval from actigraphs and data loggers as well as organization of data files and management of missing data. A summary of child and parent participant characteristics is provided. Results of analyses addressing the study's aims also are presented and are organized by each individual aim. Data Management and Analysis Data Entry and Management SPSS Version 16.0 for the Macintosh was used for data entry and analysis (SPSS, 2006). Initially, each participant's data were entered in a single row to facilitate within-subjects repeated measures analyses. Following data entry, individual variables were reviewed to identify missing data, aberrant data, and distribution of data. Distribution of continuously measured variables was reviewed for evidence of a normal distribution using standard errors of skewness and kurtosis as well as visual examination of histograms. To support analyses using the basic mixed linear model, a data file with a row for each epoch within each night for each participant was developed. 61 Actigraph Data Data management and scoring. Actigraphs were initialized using Act Millennium Version 3.10.13.1 software at the initiation of data collection (Ambulatory Monitoring, Inc., 2006). The Zero Crossing Mode (ZCM), which measures the frequency of motion, was used for data collection. Digitized actigraph data were uploaded into a laptop computer for scoring and analysis at the conclusion of data collection using Act Millennium software. Data were saved into individual files for scoring and analysis. The Action-W Version 2.6 Software was used to score digitized actigraph data and to generate quantitative sleep variables (Ambulatory Monitoring, Inc., 2006). Files were scored in three different ways to support analysis of study aims. Files were scored to identify total nighttime sleep guided by the child's sleep diary (See Figure 4.1). Secondly, actigraph files were scored to identify the 12-hour night shift period (7:00 PM to 6:59 AM) (See Figure 4.2). Thirdly, files were scored to identify 2-hour epochs within the night shift period (e.g., 7:00 PM to 8:59 PM) to be able to evaluate trends across the night shift and to facilitate comparison of sleep data with environment data. Data obtained from the actigraph files were saved into Excel files and then transferred into SPSS. Missing data. On one occasion, an actigraph failed to capture data, possibly because of human error during initialization. Of the remaining 14 actigraph data files, two had data on the first night that was scored as "bad" per instructions provided by Ambulatory Monitoring, Inc. In both of these cases, children had removed the actigraph to wash their hands and then had forgotten to replace the actigraph. One child had two epochs with missing data on the first night of data collection. The 2nd child had five epochs with missing data on the first night of data collection. Another child was Figure 4.1: Actigraph Data File Scored Based on Child's Reported Nighttime Sleep ON to Frj 06120108 Sat 06121108 Sun 00122108 1200 1800 0000 0600 1200 Figure 4.2: Actigraph Data File Scored Based on a 12-Hour Night Shift UONJ Fli 06120108 s.t 06/21/08 ---r- ,~ Sun 06122/08 1200 +~.11-1---~---L--41--~- -+"--·H---~-I ' 1800 0000 0600 1200 64 discharged from the hospital earlier than was initially expected at the time of his enrollment in the study and had only 2 complete nights of data collection. Because of variability in the individual children's sleep patterns across the study nights, missing data were treated as missing. Environment Data Data management. Data loggers were initialized using HOBOware Pro Version 2.4.2 software (Onset Computer Corporation, 2007). Stored data were downloaded into a laptop computer for analysis using HOBOware Pro software, which supports graphical and numeric display of data (See Fig 4.3). Numeric data were downloaded directly into Excel and then transferred into SPSS. The SPSS file for environment data was initially organized with each participant's sound, light, and temperature data organized into individual columns. Because measurements were obtained at 30-second intervals, each 12-hour night shift included 1440 individual measurements for each environment variable. Means of each variable for each 2-hour epoch across each night were used to support analysis of relationships between environment variables and sleep variables. Missing data. Data files of environmental variables were reviewed for missing and aberrant data. On one occasion, the 9-volt battery in the sound pressure level meter failed, resulting in missing sound data for 1 participant on the 3rd night of data collection. Because the patterns of nighttime sound were similar across individual participants and within each study night, a mean substitution of the sound levels at each epoch for the previous two nights of data collection were used to generate a complete data set. No other cases of equipment failure were observed. 75.G n IQO-i S109 o-r-r 120 -100 -SO k60 UO r 20 12:00:00 AM 12:00:00 AM 12:00:00 AM Figure 4.3: Graphical Display of Environmental Data Across Three Nights a-\ 66 Sample Description Participant Characteristics Child participants. Child participants were 15 children with cancer ranging from 5.4 to 12.3 years of age (mean = 8.8; SD = 2.3; median = 8.3) who were receiving inpatient chemotherapy at PCMC. Child participant characteristics are summarized in Table 4.1. The sample included children who receiving treatment for a primary diagnosis of cancer and those with relapsed disease. Diagnoses included in the study sample were: high-risk and very-high risk pre-B cell acute lymphoblastic leukemia (ALL), T-cell ALL, acute myelogenous leukemia (AML), osteosarcoma, undifferentiated sarcoma, and neuroblastoma. During the time of study participation, no children were receiving corticosteroids as part of their treatment plan. At the time of study enrollment, all child participants had been scheduled for an inpatient hospital admission that would include at least 3 nights. One child was discharged from the hospital early on the 3rd evening, resulting in only 2 nights of complete data. The remaining 14 children were hospitalized for at least 3 nights. Three were hospitalized for 4 nights. Another 3 were hospitalized for 5 nights. Parents' perceptions of children's sleep characteristics at home are summarized in Table 4.2. All children had sleep quality rated as "good," "very good," or "excellent." Sixty percent of participants were perceived as sleeping through the night without awakenings. Only 2 participants were reported as taking daily naps at home. No children were reported as taking medication to assist with sleep at home. Two children (13%) were taking pain medications at home at the time of enrollment in the study. 67 Table 4.1 Child Participant Characteristics Variable N(15) % Mean Standard Median Deviation Age (years) 8.8 2.3 8.3 Gender Male 10 67 Female 5 33 Race/Ethnicity African-American 1 7 White 14 93 Diagnosis Leukemia 8 53 Lymphoma 2 13 Sarcoma 3 20 Other Solid Tumor 2 13 Primary vs. Relapsed Disease Primary Disease 12 80 Relapsed Disease 3 20 Time from Initial Diagnosis (months) Primary Disease 3 2 2.4 Relapsed Disease 64 28 52.5 68 Table 4. Parent Perceptions of Children's Sleep Characteristics at Home Variable NO 5) % Sleep quality Excellent 3 20 Very good 8 53 Good 4 27 Number of nighttime awakenings None 9 60 One or two 6 40 Daytime naps No 13 13 Yes 2 87 Children's average hematocrit at admission was 32.8% (SD=4.7; median = 32.3%). Normal hematocrit values for school-age children are 35 to 45% (Brugnara, 2009). Mean absolute neutrophil count (ANC) at admission was 2,960/^L (SD = 4,392; median = 1,600/fiL). Neutropenia is defined as an ANC of less than l,000/|aL (Wolff, Altman, Berkow, & Johnson, 2004). Children were required to have an ANC of at least 750/|iL to meet eligibility criteria to receive chemotherapy. Some treatment protocols required an ANC of 1,000/pL. Parent/caregiver participants. Parent/caregiver participant characteristics are summarized in Table 4.3. All were parents of child participants and included 4 fathers and 15 mothers. In four cases, parents rotated days during the child's admission in which they stayed with the child, and therefore, both provided study-related data pertaining to their child. In each of these cases, informed consent was obtained from each parent. 69 Table 4. Parent Participant Characteristics Variable Number Mean Standard Median Deviation Age (years) 35.8 6.6 34.8 Gender Male Female 4 15 Race/Ethnicity African-American White 1 18 Participant Accrual A sequential approach to sampling was used in which all families of eligible children were informed about the study and invited to participate. Eighteen total families were offered the opportunity to participate. Two families declined to participate. One child declined to participate because she does not like to wear anything on her wrist. The second family declined for reasons unrelated to the study procedures. One child withdrew less than 24 hours from the start of data collection because she no longer wished to wear the wrist actigraph. Nighttime Sleep Characteristics Aim 1 Describe nighttime sleep-wake patterns among children with cancer receiving inpatient chemotherapy 70 Description of Nighttime Sleep Characteristics Because the majority of study participants were hospitalized for only 3 nights, reported findings will reflect data obtained from these 3 nights. Nighttime sleep characteristics obtained from wrist actigraphs based on children's reported time in bed for nighttime sleep are summarized in Table 4.4. Reported characteristics are based on parameters recommended for the evaluation of sleep-wake disturbances by Berger and colleagues (2005). Correlations between nighttime sleep characteristics are presented in Tables 4.5 and 4.6. Comparison of Sample Characteristics With Age-Related Norms Total time in bed and total nighttime sleep. Mindell and Owens (2003) report that school-age children (6 to 12 years of age) require 10 to 11 hours of sleep over a 24-hour period. In this sample, average nighttime time in bed exceeded 10 hours on Night 1 only (mean = 10.92 hours). Children's total nighttime sleep time did not reach 10 hours on any of the 3 study nights. A clinically significant decrease in total sleep time for adults, based on an 8-hour period of nighttime sleep is 30 to 60 minutes which represents a decrease of 6.25 to 12.5% (Carskadon & Dement, 1981). Applying these percentages to the average school-age child's 10-hour nighttime sleep requirement, a 6.25 to 12.5% decrease would represent a loss of 37.5 to 75 minutes of sleep. These estimates suggest that total nighttime sleep on both the 2nd and 3rd study nights represents a clinically significant decrease based on school-age children's nighttime sleep requirements. 71 Table 4. Characteristics of Nighttime Sleep Based on Children's Reported Time in Bed Night of Hospitalization Variable Night 1 (N=13) Night 2 (N=14) Night 3 (N=13) Total Time in Bed (minutes) Mean 655.31 583.57 597.46 SD 126.94 95.33 92.76 Median 621.00 590.50 587.00 Range 526-853 368-733 406 - 742 Total Sleep Time (minutes) Mean 572.15 504.64 506.38 SD 101.21 77.17 91.21 Median 553.00 506.50 510.00 Range 449 - 784 335 -648 342-619 Sleep Efficiency (percentage) Mean 88.51 89.11 85.94 SD 4.53 8.10 8.25 Median 89.58 92.72 90.34 Range 80.68-96.74 72.01 -95.98 69.58-93.79 Sleep Latency (minutes) Mean 12.69 15.79 9.23 SD 15.73 15.51 7.76 Median 7.00 10.50 7.00 Range 0 - 5 8 4 - 6 1 4 - 3 4 Number of Awakenings Mean 13.38 11.00 12.69 SD 6.69 4.31 4.91 Median 13.00 10.50 12.00 Range 5 - 2 5 5 - 1 7 7 - 2 2 Wake after Sleep Onset (minutes) Mean 76.46 64.86 83.00 SD 42.76 55.37 55.26 Median 60.00 43.00 54.00 Range 17-164 22-201 41 -223 Average Sleep Episode (minutes) Mean 54.03 57.12 48.95 SD 25.63 26.07 24.58 Median 43.31 51.81 37.42 Range 26.36- 100.80 26.27- 107.25 23.32-88.43 Longest Sleep Episode (minutes) Mean 129.85 139.21 112.85 SD 25.30 49.88 34.55 Median 125.00 127.00 103.00 Range 88-161 73 - 260 65-176 Table 4.5 Pearson Correlations Between Nighttime Sleep Characteristics Sleep Sleep Sleep latency Wake after minutes efficiency sleep onset Number of Average Longest wake sleep episode si e e p episode episodes duration Sleep duration .88" -.31