"Medicine, Faculty of"@en . "Population and Public Health (SPPH), School of"@en . "DSpace"@en . "UBCV"@en . "Hedden, Lindsay Kathleen"@en . "2008-10-08T14:46:17Z"@en . "2008"@en . "Master of Science - MSc"@en . "University of British Columbia"@en . "Purpose: A majority of childhood and adolescent cancer survivors face life-long cancer- and treatment-related sequelae. Long-term follow-up is necessary to facilitate timely diagnosis and management of these health conditions. As part of strategic long-term follow-up, provider continuity of care (PCOC) may improve outcomes through appropriate use of surveillance, screening, and coordination of services. The purpose of this thesis was to assess physician services utilization and PCOC among survivors of childhood cancer compared with general population subjects, and to examine factors associated with survivors' use of physician services and PCOC scores.\nMethods: Physician services utilization and PCOC were assessed in a population-based cohort of 1322 five-year cancer survivors diagnosed between 1981 and 1995 under age 20 in British Columbia, and a group of 13,220 age- and gender-frequency matched, randomly selected population-based subjects, whose records were linked to individual-level administrative healthcare datasets. Effects of clinical and sociodemographic modifiers on utilization and PCOC were examined using generalized linear modeling. Changes in utilization and PCOC by age were estimated using a longitudinal, repeated measures modeling approach.\nResults: Survivors incurred an average of 8.94 medical visits per year: 4.82 to primary care physicians, 2.69 to specialists, and 1.43 to non-physician providers. Survivors had more visits than comparators in all visit categories (p<0.0001 for all). As they age, survivors' use of primary care services increases significantly, while their use of specialist services declines, trends that are not mirrored by the comparison population.\nThe average PCOC score for survivors was 0.54 \u00B1 0.22, indicating survivors saw the same primary care provider for only 50% of their primary care visits. Mean score did not differ between survivors and comparators; however, in the population sample scores improved with age (p=0.02), while among survivors, scores worsened (p=0.05).\nConclusions: The dramatic age-related increase in primary care visits observed in the survivor group suggests that primary care physicians play a key role in ensuring quality long-term follow-up care. Survivors are at heightened risk for poor PCOC as they age and transition into adult-oriented community care, raising concerns about whether they are receiving the appropriate follow-up care encompassing screening, surveillance and psychosocial support."@en . "https://circle.library.ubc.ca/rest/handle/2429/2490?expand=metadata"@en . "8186014 bytes"@en . "application/pdf"@en . "HEALTH SERVICES UTILIZATION AND PROVIDER CONTINUITY OF CARE AMONG SURVIVORS OF CHILDHOOD CANCER: A COHORT ANALYSIS by Lindsay Kathleen Hedden B.Sc. Hons, University of Waterloo, 2004 A THESIS SUMBITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in The Faculty of Graduate Studies (Health Care and Epidemiology) THE UNIVERSITY OF BRITISH COLUMBIA April 2008 \u00C2\u00A9 Lindsay Kathleen Hedden, 2008 ABSTRACT Purpose: A majority of childhood and adolescent cancer survivors face life- long cancer- and treatment-related sequelae. Long-term follow-up is necessary to facilitate timely diagnosis and management of these health conditions. As part of strategic long-term follow-up, provider continuity of care (PCOC) may improve outcomes through appropriate use of surveillance, screening, and coordination of services. The purpose of this thesis was to assess physician services utilization and PCOC among survivors of childhood cancer compared with general population subjects, and to examine factors associated with survivors' use of physician services and PCOC scores. Methods: Physician services utilization and PCOC were assessed in a population-based cohort of 1322 five-year cancer survivors diagnosed between 1981 and 1995 under age 20 in British Columbia, and a group of 13,220 age- and gender-frequency matched, randomly selected population-based subjects, whose records were linked to individual-level administrative healthcare datasets. Effects of clinical and sociodemographic modifiers on utilization and PCOC were examined using generalized linear modeling. Changes in utilization and PCOC by age were estimated using a longitudinal, repeated measures modeling approach. Results: Survivors incurred an average of 8.94 medical visits per year: 4.82 to primary care physicians, 2.69 to specialists, and 1.43 to non-physician providers. Survivors had more visits than comparators in all visit categories (p<0.0001 for all). As they age, survivors' use of primary care services increases significantly, while ii their use of specialist services declines, trends that are not mirrored by the comparison population. The average PCOC score for survivors was 0.54 \u00C2\u00B1 0.22, indicating survivors saw the same primary care provider for only 50% of their primary care visits. Mean score did not differ between survivors and comparators; however, in the population sample scores improved with age (p=0.02), while among survivors, scores worsened (p=0.05). Conclusions: The dramatic age-related increase in primary care visits observed in the survivor group suggests that primary care physicians play a key role in ensuring quality long-term follow-up care. Survivors are at heightened risk for poor PCOC as they age and transition into adult-oriented community care, raising concerns about whether they are receiving the appropriate follow-up care encompassing screening, surveillance and psychosocial support. iii AI LE^NDISHCI AGMS Z\"E LEIAIVIIDONd H32IVHS311 dIHSHOAIMMS IIRDNVD rmav 9N110A (NV `INRDSTIOCIV `CIOOHCIIIHD HILL VC LE^A9010(10H.L3141 :\u00C2\u00A3 113.14VHD SEAuvvains S'Z tE^saino31n0 t't'Z EEunoinetias mleaH E'VZ 1\u00C2\u00A3^sapspapento lenPfA!Pul Z177 6Zsapspapento lemaluoD I'V'Z 9Z^Nouvzniin AO 181101AIIVIMOIAVH3EI S,NRSHRUNV :)UOMRIAIVIld TiflidaDNOD 177 91^@JED jo fii!nuPuoD I'E'Z IIS2IOAIA2111S 1104 RIVD dfl-MOTIOd AIIIVIlb E'Z 6^saoyuas ipleaH ieuosiadjo asfl 'T'Z'Z 8samoisfvm RIVD dIHS110AIMMS Z.Z V^SURDNVD IN33STIOUV (INV ClOOHCIIHD dO SHOAIAIIIIS VZ tM3lA311 311111V113.L11 :Z IladifH3 T^NOLDflOOILLNI :I 113.1.dVHD !!PcSINHIA139O3TMONN3V !pc^SNOUVIA311118V dO ISI1 pcS32lf1913 AO ISII !HA^S3111111. dO ISII A!S.I.N3INO3 dO aliPil !!^IDVILLSEIV SINHINOD AO RIEIVI 3.2.2 Objectives and Research Questions^ 39 3.3 IDENTIFICATION OF THE STUDY COHORT 41 3.3.1 Survivors^ 41 3.3.2 Comparators 42 3.4 DATA SOURCES AND ETHICS APPROVAL^ 43 3.4.1 Case Ascertainment^ 43 3.4.2 Treatment Data Collection 44 3.4.3 BC Linked Health Database^ 45 3.4.4 Linkage and Preparation 46 3.5 STUDY VARIABLES^ 47 3.5.1 Outcome Variables 47 3.5.2 Clinical Covariates^ 52 3.5.3 Demographic Covariates 53 3.6 STATISTICAL ANALYSES^ 54 3.6.1 Data Preparation and Descriptive Statistics^ 54 3.6.2 Physician services utilization^ 56 3.6.3 Primary Care Provider Continuity of Care^ 57 CHAPTER 4: DESCRIPTIVE STATISTICS AND UTILIZATION OF PHYSICIAN SERVICES^ 60 4.1 DESCRIPTIVE STATISTICS^ 60 4.1.1 Clinical Characteristics 61 4.1.2 Demographic Characteristics^ 65 4.3 PHYSICIAN SERVICES UTILIZATION 68 4.3.1 Summary Statistics^ 68 4.3.2 Regression Modeling: Overall Visit Patterns^ 71 IA 6E1^saa!puaddv OZIpolo swom 81T^SNOISIVIDNOD 9'9 911HDIIVRSHII 82lf1inA1101 SNOLLDRHIU S'9 9T T^uollealiddd '917'9 SITtproicidv leppspelS '9'19 UT^luatuamseaw pue salqupeA '1717'9 ZTTsamiun pue saamos ma .c.t.9 OTT^sllogoD r17.9 601Asa(' APMS 1'79 601^SNOLLVIIINI1 CINV SHIDNRHIS .17.9 801amp JO filmmluop Jo uolleznenldapuop aql u! sauelip .E.E.9 901^sanssl amp leuomsueld, 'Z'\u00C2\u00A3'9 170Tslapow amp dn-mollog T.E.9 'POT^SNOLLVD11.31/11 \u00C2\u00A3'9 OOTSEMSRII Z.9 86^DNVDIAINDIS I'9 86SNOISf11311103 UM, NOISSIDS113 :9 1131c1VH3 Z6^saloas pcin pue DOD a21 As lenuue!S Z'Z'S 88samps Ddn pue DOD Helano i'Z'S 88^91\u00E2\u0096\u00A0111HCIOIAI NOISSRIMHZ'S L8SDLLSIIVIS AHVIATIAMS T'S L8^31110 AO A.LMNIINO3 113131A0Ild :s 113.141,1-13 8Lsulaped l!s!A lenuur!El :2ullapow uopsaaali E'\u00C2\u00A3'17 APPENDIX A: THE CHILDHOOD, ADOLESCENT, AND YOUNG ADULT CANCER SURVIVORSHIP RESEARCH PROGRAM^ 139 APPENDIX B: CERTIFICATE OF EXPEDITITED ETHICS APPROVAL - ANNUAL REVIEW^ 141 APPENDIX C: REGRESSION MODELS - VISIT PATTERNS^ 144 APPENDIX D: REGRESSION MODELS - CONTINUITY OF CARE 168 vii LIST OF TABLES Table 2.1 Summary of Impacts of Continuity of Care^ 20 Table 3.1 Study variables^ 49 Table 3.2 Summary of ICCC Diagnosis Groups^ 52 Table 3.3 Treatment Classifications^ 52 Table 4.1 Application of Exclusion Criteria 61 Table 4.2 Survivor Diagnosis and Treatment Characteristics^ 61 Table 4.3 Diagnosis-Treatment Combinations^ 63 Table 4.4 Age at and Period of Diagnosis 65 Table 4.5 Case-Control Demographic Characteristics^ 66 Table 4.6 Mean Visits/Year (\u00C2\u00B1 Standard Deviation) for Specific ICCC Classifications 69 Table 4.7 Mean Visits/Year (\u00C2\u00B1 Standard Deviation) for Specific Treatment Classifications^ 70 Table 4.8 Mean Visits/Year (\u00C2\u00B1 Standard Deviation) for Survivors versus Comparators^ 70 Table 4.9 Grouped Mean Visits/Year for Survivors versus Comparators^ 71 Table 4.10 Factors Influencing Visits per Year: Survivors Only^ 73 Table 4.11 Factors Influencing Visits per Year: Case-Control 77 Table 4.12 Factors Influencing BBA Visit Pattern: Survivors Only^80 Table 4.13 Factors Influencing BBA Visit Pattern: Case-Control 85 Table 5.1 Continuity Scores (\u00C2\u00B1 Standard Deviation) for Specific ICCC Classifications 87 Table 5.2 Continuity Scores (\u00C2\u00B1 Standard Deviation) for Specific Treatment Classifications^ 88 Table 5.3 Factors Influencing COC and UPC Scores: Survivors Only^90 viii Table 5.4 Factors Influencing COC and UPC Scores: Case-Control^92 Table 5.5 Factors Influencing BBA COC and UPC Scores: Case-Only^94 Table 5.6 Factors Influencing BBA COC and UPC Scores: Case-Control^97 Table C.1 Survivor-Only, Overall: Total Visits^ 144 Table C.2 Survivor-Only, Overall: Primary Care Visits^ 146 Table C.3 Survivor-Only, Overall: Specialist Visits 148 Table C.4 Case-Only, Overall: Non-Physician Visits^ 150 Table C.5 Case-Control, Overall: Total Visits 152 Table C.6 Case-Control, Overall: Primary Care Visits^ 153 Table C.7 Case-Control, Overall: Specialist Visits 154 Table C.8 Case-Control, Overall: Non-Physician Visits^ 155 Table C.9 Case-Only, Biannual By Age: Total Visits 156 Table C.10 Case-Only, Biannual by Age: Primary Care Visits^ 158 Table C.11 Case-Only, Biannual by Age: Specialist Visits 160 Table C.12 Case-Only, Biannual by Age: Non-Physician Visits^ 162 Table C.13 Case-Control, Biannual By Age: Total Visits 164 Table C.14 Case-Control, Biannual By Age: Primary Care Visits^ 165 Table C.15 Case-Control, Biannual By Age: Specialist Visits 166 Table C.16 Case-Control, Biannual By Age: Non-Physician Visits^ 167 Table D.1 Survivor-Only, Overall: COC Score^ 168 Table D.2 Survivor-Only, Overall: UPC Score 170 Table D.3 Case-Control, Overall: COC Score^ 172 TableC.4 Case-Control, Overall: UPC Score 173 ix Table D.5 Survivor-Only, Biannual by Age: COC Score^ 174 Table D.6 Survivor-Only, Biannual by Age: UPC Score 176 Table D.7 Case-Control, Biannual by Age: COC Score^ 178 Table D.8 Case-Control, Biannual by Age: UPC Score 179 x LIST OF FIGURES Figure 2.1 Andersen's Behavioural Model of Utilization^28 Figure 3.1 Andersen's Behavioural Model of Utilization: Variable Placement^50 Figure 4.1 Primary Care Visits Per Year by Age, Grouped by Age at Diagnosis^83 Figure 4.2 Specialist Visits Per Year by Age, Grouped by Age at Diagnosis^84 Figure 5.1 Continuity of Care Trends by Age, Grouped by Age at Diagnosis ^96 xi LIST OF ABBREVIATIONS BCCA^British Columbia Cancer Agency BCCH^British Columbia Children's Hospital BCCR^British Columbia Cancer Registry BCLHD^British Columbia Linked Health Database CAYACS^Childhood, Adolescent, and Young Adult Cancer Survivors COC^Continuity of Care GEE^Generalized estimating equations GLM^Generalized Linear Models ICD-0^International Classification of Diseases for Oncology \"K\"^Known Provider Continuity LTFU^Long-term Follow-Up MSP^Medical Services Plan PHN^Personal Health Number SECON^Sequential Continuity Index SES^Socioeconomic Status SIR^Standardized Incidence Ratio SMR^Standardized Mortality Ratio UPC^Usual Provider of Care xii ACKNOWLEDGEMENTS This thesis work was inspired by the many gifted scholars and teachers I have encountered in my six years of university education. Specifically, this work would not have been possible without the guidance and support of my mentor, Dr. Sam Sheps, under whose supervision I chose this topic and drafted this thesis. I am also indebted to the other members of my thesis committee - Mary McBride, Anne- Marie Broemeling, and Charlyn Black - who all provided direction and feedback for each of my chapters, and for the many conference abstracts that arose from them. I owe a great deal of the statistical mastery I have earned over the past two years to Maria Lorenzi at the BC Cancer Research Centre, who guided me through many statistical issues, the likes of which I had never encountered in my extensive statistical course coursework. I would also like to acknowledge the valuable contributions made by staff and advisory committee members of the Childhood, Adolescent, and Young Adult Cancer Survivors Research Program, who are too numerous to name here. Thanks are owed to the Western Regional Training Centre (WRTC) for Health Services Research for the master's award, which has supported me during the completion of this project. The other WRTC students and alumni, whose thoughtfulness and intelligence continue to amaze me, were a significant driver in the completion of this project; I continue to aspire to the high academic standard they have set. I also owe a great deal to my parents, who instilled in me the values of education and life-long learning from an early age, values that I endeavor to pass on to my own children. And last, my heartfelt gratitude must go to my husband for proofreading without necessarily understanding; for being the willing recipient of many thesis-inspired tirades; and for the incredible support and patience he has shared with me over the past year. It is to him that I dedicate this thesis. You'll be bothered from time to time by storms, fog, snow. When you are, think of those who went through it before you, and say to yourself, \"What they could do, I can do.\" --Antoine de Saint Exupery Wind, Sand and Stars Funding: Part of the Childhood/Adolescent/Young Adult Cancer Survivor Research Program, funded by the Canadian Cancer Society through the National Cancer Institute of Canada, Grant#PPG016001. xiv \u00E2\u0080\u00A2uNoy JoS CHAPTER 1: INTRODUCTION The overall goal of this study is to evaluate physician services utilization and provider continuity of care among survivors of cancer diagnosed in childhood and adolescence, and to explore how visit patterns and continuity of care vary between survivors of cancer and the general population. Rates of cancer survival have increased dramatically since the 1970s and a population of survivors facing unique long-term health-related challenges is emerging. The necessity of life-long follow-up care for survivors is increasingly being recognized; however, little is currently known about the follow-up care currently received by the survivor population. Although there is limited consensus among the myriad guidelines for long-term follow-up that have emerged in recent years, most include some elements of continuity or coordination of care. Moreover, given the complex care needs of this unique group, provider continuity of care has the potential to improve health outcomes through improved use of surveillance and screening, and coordination of necessary services; therefore, it is an important element of a best-practices model for follow-up care for cancer survivors. Patterns of health services utilization, as well as provider continuity, have not been well characterized in this population. The potential impact of improved provider continuity has also not been examined. This descriptive and analytic research takes the necessary first steps to fill this knowledge gap. The specific objectives of this descriptive and analytic project are as follows: 1 \u00E2\u0080\u00A2 To describe, using both cross-sectional and longitudinal methodologies, long- term physician visit patterns and provider continuity of care among survivors of cancer diagnosed in childhood and adolescence; \u00E2\u0080\u00A2 To analyze the effect of clinical cancer and sociodemographic factors on patterns of physician visits and continuity of care; and \u00E2\u0080\u00A2 To assess differences in unadjusted and adjusted visit patterns and continuity of care between survivors of childhood cancer and a general population of age-matched comparators, using both cross sectional and longitudinal approaches. The design of this study, and the context in which it is placed, were informed by existing empirical literature and conceptual theory. Specifically, results from studies on the health care needs of cancer survivors and studies of the role of continuity in measures of quality of care informed the study questions. Existing access to care and utilization theory provided a conceptual underpinning on which to apply the defined questions. An overview of this literature will be provided in the subsequent chapter, followed by a discussion of access to care theory, study methods and results. The final chapter centers on the research, health care, and health policy implications of this work, and provides targets for future research in the area. This project is part of the Childhood, Adolescent, and Young Adult Cancer Survivorship (CAYACS) Research Program at the BC Cancer Agency. A brief description of the program has been attached as Appendix A. The funding for this 2 program has been provided by the Canadian Cancer Society through the National Cancer Institute of Canada. Student funding has been provided by the Western Regional Training Centre for Health Services Research. 3 CHAPER 2: LITERATURE REVIEW 2.1 SURVIVORS OF CHILDHOOD AND ADOLESCENT CANCERS The successful treatment of childhood cancer is truly one of oncology's greatest success stories. Three decades of multimodal (surgery, chemotherapy and radiation) therapy for the treatment of cancer acquired in childhood or adolescence has increased five-year survival rates to 79.9 percent overall (Ries et al., 2007) 1 - and over 80% in Canada 2 (Canadian Cancer Society/National Cancer Institute of Canada, 2007) - from just over 60 percent in 1975 (Ries et al., 2007). As a result, the population of survivors has grown dramatically to over 270,000 in the United States as of 2002 (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003). One in every 640 adults aged 20 to 39 is a survivor of childhood cancer (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003). As cancer treatments have continued to increase survival rates, the research challenge is shifting from finding therapies that will arrest the disease to minimizing the medical and social issues associated with cancer survivorship. It is now recognized that cancer treatment can be associated with significant long-term adverse events, including mortality (Robertson, Hawkins, & Kingston, 1994). Aptly put by Dr. Philip Rosoff (Rosoff, 2006) \"there is a dark side to being cured of cancer as a young person\". Refers to cases diagnosed under the age of 15. 2 Refers to cases diagnosed under the age of 19 4 Up to 75 percent of childhood cancer survivors have been reported to suffer from at least one chronic treatment-related health problem - referred to in the literature as a \"late effect\" of treatment - and between 30 and 50 percent of survivors have late-occurring or chronic health problems severe enough to require long-term treatment (Garre et al., 1994; Geenen et al., 2007; M. Hudson et al., 2003; Humpl, Fritsche, Bartels, & Gutjahr, 2001; Stevens, Mahler, & Parkes, 1998). Approximately one quarter of cancer survivors experience a late effect that is life threatening (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003). In general, late-effects are complications, disabilities, adverse outcomes, or any combination thereof, that are persistent and result from the disease process, treatment, or both (Meadows, 2003). Ongoing monitoring of overall and disease-specific mortality rates amongst survivors have demonstrated a nine-fold difference in death rates between survivors and the general population (Standardized Mortality Ratio (SMR) = 9.1, 95% CI, 7.8-10.5) (MacArthur, Spinelli, Rogers, Goddard, Abanto et al., 2007). More specifically, relative mortality was significantly increased due to cancer-related causes of death (SMR = 81.7, 95%Cl, 68.6-95.8), circulatory (SMR = 9.7, 95% CI, 4.2- 19.1) and respiratory (SMR = 16.7, 95% CI, 4.6-43.0) causes (MacArthur, Spinelli, Rogers, Goddard, Abanto et al., 2007). Survivors of childhood cancers are also at high risk of developing subsequent malignant neoplasms (Bhatia et al., 2003; Hawkins, Draper, & Kingston, 1987; Jazbec, Ecimovic, & Jereb, 2004; MacArthur, Spinelli, Rogers, Goddard, Phillips et al., 5 2007; C. A. Sklar et al., 2002). Compared to the general population, survivors face an 18.5-fold increased risk of second neoplasms (Standardized Incidence Ratio (SIR) = 18.5, 95% CI, 15.6-21.7) (Bhatia et al., 2003), the most common diagnosis being breast cancer (SIR = 56.7) (Bhatia et al., 2003). Other commonly occurring solid malignancies included thyroid cancer (SIR = 36.4), bone tumors (SIR = 37.1), and colorectal (SIR = 36.4), lung (SIR = 27.3) (Bhatia et al., 2003), and gastric cancers (SIR = 63.9). The cumulative incidence of second malignancies among survivors reached as high as 26% thirty years from the original cancer diagnosis (Bhatia et al., 2003). Survivors also experience an elevated risk of developing other chronic conditions. A recent cohort study of 14,000 cancer survivors and their siblings found that the adjusted relative risk for a chronic condition in a survivor, as compared with siblings, was 3.3 (95% CI 3.0 to 3.5) (K. C. Oeffinger et al., 2006). Even more striking, the relative risk increased to 8.2 (95% CI 6.9 to 9.7) for severe or life-threatening conditions (K. C. Oeffinger et al., 2006). Among survivors, the cumulative incidence of a severe, disabling, life-threatening condition or death due to a chronic health condition reached a staggering 42.4% (95% CI 33.7 to 51.2) (K. C. Oeffinger et al., 2006). The most common late-occurring or chronic health problems related to cancer treatment are first recurrences, as described above, (Bhatia et al., 2003; Hawkins, Draper, & Kingston, 1987; Jazbec, Ecimovic, & Jereb, 2004; MacArthur, Spinelli, Rogers, Goddard, Phillips et al., 2007; C. A. Sklar et al., 2002), cardiopulmonary toxicities (Miller et al., 1986; Steinherz, Steinherz, Tan, Heller, & Murphy, 1991), endocrine dysfunction (C. A. Sklar & Constine, 1995), and 6 musculoskeletal problems (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003). Additionally, many cancer survivors face neurocognitive deficits or psychological problems, including social and educational difficulties (L. K. Campbell et al., 2007; Dickerman, 2007; Lansky, List, & Ritter-Sterr, 1986; Mulhern & Palmer, 2003; National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003; Raymond-Speden, Tripp, Lawrence, & Holdaway, 2000 Mar; Recklitis, Lockwood, Rothwell, & Diller, 2006; Teta et al., 1986; B. J. Zebrack, Gurney et al., 2004; Brad J. Zebrack et al., 2007). Survivors of childhood cancer also tend to self-report poorer health status: A report from the U.S. Childhood Cancer Survivor Study found that survivors are significantly more likely to report adverse general health and mental health as well as greater functional impairment when compared with their siblings (M. M. Hudson et al., 2003). The emergence and severity of late effects depends on age at cancer diagnosis, specific cancer diagnosis, treatment modality, and disease severity, among other factors (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003). Some late effects, typically those that are identified early in follow-up, are minor and transient, resolving without medical intervention (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003). Others may persist or develop in adulthood, and may require long-term or even life-long therapy (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003). 7 2.2 SURVIVORSHIP CARE TRAJETORIES For this research, cancer survivorship is defined as a distinct phase of the cancer trajectory that follows primary treatment and precedes recurrence, onset of subsequent cancer, or death. The full extent of long-term health consequences, and therefore the specific health care needs, of individuals in the survivorship phase of the cancer trajectory is not yet known; however, evidence suggests that survivors suffer from late-occurring or chronic health conditions severe enough to require long-term treatment and medical surveillance (Bleyer et al., 1993; Wendy Landier et al., 2004; W. Landier, Wallace, & Hudson, 2006; National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003; Oeffinger, 2007; K. C. Oeffinger & Hudson, 2004). Given the potential severity of late effects, and the variability in the time of appearance and resolution, both the American Cancer Society and the Pediatric Working Group of the Canadian Strategy for Cancer Control recommend life-long follow-up of cancer survivors (Bhatia & Sklar, 2002; Bleyer et al., 1993; National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003). The rationale for long-term follow-up (LTFU) care, consisting of routine, periodic medical visits and tests, is to reduce or eliminate excess morbidity and mortality among survivors by detecting incipient late effects, diagnosing emerging late effects or new primary cancers promptly, and facilitating timely management of chronic toxicity (Grunfeld, Dhesy-Thind, & Levine, 2005; Skinner, Wallace, & Levitt, 2007; Skinner, Wallace, Levitt, & Uk Children's Cancer Study Group Late Effects, 2006). LTFU also serves many other important roles, notably the provision of ongoing information about potential late effects and relevant health promotion 8 advice, as well as psychological and psychosocial support (Grunfeld, Dhesy-Thind, & Levine, 2005; Skinner, Wallace, & Levitt, 2007; Skinner, Wallace, Levitt, & Uk Children's Cancer Study Group Late Effects, 2006). A secondary benefit of long-term follow-up is the ability to better understand how cancer therapies impact the development of other common health problems associated with aging (K. C. Oeffinger, 2003). There is currently limited evidence surrounding the effectiveness of LTFU in reducing excess morbidity and mortality from late effects. 2.2.1. USE OF PERSONAL HEALTH SERVICES In spite of both the clear need for long-term follow-up and the recommendations for continued screening and surveillance, the study of health services utilization among survivors has been largely neglected. In general, survivors will attend follow-up visits with a paediatric oncologist for either five years following diagnosis or until they reach 19 years of age. Few survivors are actively followed once they transition to adult care (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003). Only two studies, both using case-control methodology and self-report data gathered from cancer centres, have assessed follow-up care and health services utilization among survivors (K. C. Oeffinger et al., 2004; A. K. Shaw et al., 2006). A 2004 report from the US Childhood Cancer Survivor Study indicated that most survivors received care from primary care physicians (K. C. Oeffinger et al., 2004). The report also suggests that health care utilization diminished with increasing age or interval from cancer diagnosis, male gender, lack of health insurance, or lack of concern for future health (K. C. Oeffinger et al., 2004). Importantly, risk-based care - 9 \"care based on proactive and anticipatory periodic evaluation and counseling to reduce the risk and minimize progression of disease, rather than reacting to disease as it occurs (K. C. Oeffinger, 2003)\" - was found to be uncommon (K. C. Oeffinger et al., 2004). A similar analysis from a Canadian cohort demonstrated that although child survivors and their comparators were equally likely to self-report a visit to a general practitioner, the survivors were twice as likely to report four-or-more visits to health care professionals per year (A. K. Shaw et al., 2006) and were more likely to consult with any specialist, including non-paediatric oncologists (A. K. Shaw et al., 2006). A linear relationship between morbidity and annual visits was observed (A. K. Shaw et al., 2006), which is not surprising. Previous studies examining the health care of young adult survivors of childhood cancer are based on case ascertainment from multiple institutions and use of periodic self-reported data on late effects and health utilization. Limitations of this approach include the effective recruitment and retention of a representative sample (Greenland, 1977; Institute of Medicine, 2006), continuous access to health records, minimizing attrition bias (Greenland, 1977; Institute of Medicine, 2006) - particularly in the transition from pediatric to adult care - and the potential biases associated with self-report data (Coughlin, 1990; Institute of Medicine, 2006). The use of this methodology also introduces the potential for biases due to participant self-selection (Bell & Hammal, 2002). 10 2.3 QUALITY FOLLOW-UP CARE FOR SURVIVORS There is a paucity of evidence to inform the development of a specific strategy for long-term follow up, and although several sets of guidelines for LTFU have emerged in recent years (Wendy Landier et al., 2004; W. Landier, Wallace, & Hudson, 2006; Skinner, 2005), there is little consensus among them about how often follow-up should occur, where it should occur or what specific procedures it should comprise (Goldsby & Ablin, 2004). Further, these guidelines are not necessarily evidence-informed (Goldsby & Ablin, 2004) and the effectiveness of recommended interventions is often unclear (Institute of Medicine, 2006). These guidelines also fail to address issues related to the most appropriate organizational models for performing LTFU (Skinner, Wallace, & Levitt, 2007). Optimal models of LTFU, including which, when, where, how and by whom survivors should be monitored, are a matter of debate in the literature. There is considerable variation in how LTFU is performed, with several alternative models involving a range of professionals in a variety of locations, and depending on numerous clinical and organization factors (Skinner, Wallace, & Levitt, 2007). And importantly, there are relatively few organized, systematic follow-up programs currently operating in North America (K. C. Oeffinger, 2003\u00E2\u0080\u009E 2004; K. C. Oeffinger & Wallace, 2006). Some experts advocate follow-up care provided by a primary paediatric oncologist on the belief that the initial bond made between this physician and his/her patient will encourage survivors to seek necessary follow-up (Goldsby & Ablin, 2004). Additionally, the primary paediatric oncologist will have the best 11 knowledge of a patient's disease and treatment history and therefore may be in the best position to identify late effects (Goldsby & Ablin, 2004). However, paediatric oncologists are not trained to deal with the clinical conditions seen in the adult population (Goldsby & Ablin, 2004). There is also a lack of capacity for survivor care at most cancer-treating institutions, especially in terms of actively following survivors as and after they transition into adult care (K. C. Oeffinger & Wallace, 2006). Additionally, these institutions often do not have the mandate nor the resources to support ongoing follow-up care (K. C. Oeffinger & Wallace, 2006). Survivor-related barriers to this model may arise because visits to a paediatric oncologist may trigger unpleasant memories of cancer diagnosis and treatment (Goldsby & Ablin, 2004). Avoidance behaviours linked to post-traumatic stress disorders - which occur in a survivor population at a rate of nearly twenty percent (Rourke, Stuber, Hobbie, & Kazak, 1999) - are particularly concerning as survivors may avoid follow-up care in an attempt to distance themselves from their cancer diagnosis and treatment (Ginsberg, Hobbie, Carlson, & Meadows, 2006; Hobbie et al., 2000). Paediatric oncology clinics may also be inaccessible to many patients due to geographical and/or financial limitations (Goldsby & Ablin, 2004). The training of internists or general practitioners specializing in the treatment of late-effects has been recommended as a strategy for the provision of effective follow-up care. Such specialists, located at specialized follow-up care clinics, would have the most up-to-date training on the identification and treatment 12 of late-effects in adolescent, young adult, or adult populations, and would also help survivors more effectively navigate the medical system and transition to adult- oriented care (Goldsby & Ablin, 2004). This approach may not be feasible, however, due to cost and the necessary identification of qualified and interested practitioners. Additionally, unless widely available, survivors' access to such specialized providers could be limited due to personal cost or geography (Goldsby & Ablin, 2004). Follow-up care at the level of the local primary care physician has also been suggested as a potential follow-up strategy (Goldsby & Ablin, 2004). The key advantage of a primary care-based follow-up care program is that the survivor is followed in the local community by a provider who is responsible for their overall health and well-being, as oppose to specializing only in cancer-related issues (Goldsby & Ablin, 2004). A Delphi panel of health policy experts recently identified barriers to young adult survivors receiving appropriate care with a primary care provider (Mertens et al., 2004). Two key barriers identified were that (1) primary care physicians are often not familiar with the health needs of a survivor population (K. C. Oeffinger & Wallace, 2006); and (2) survivors were unaware of potential health risks associated with cancer treatment (Kadan-Lottick et al., 2002). Most primary care physicians have limited exposure to survivors of paediatric cancer: in the US, a typical primary care doctor has only five to seven survivors in their entire practice (K. C. Oeffinger, 2000). Additionally, physicians and do not receive any formal training or education in the area of survivor care (National Cancer Policy, Hewitt, Weiner, & Simone, 13 2003; K. C. Oeffinger, 2003). There is also limited communication between cancer institutions and primary care providers, making accessing up-to-date information on the risks for a specific patient difficult (K. C. Oeffinger & Wallace, 2006). As stated above, survivors themselves may not have the necessary information to seek care appropriately or to guide their primary care physician. Notably, when asked whether past cancer therapies could cause future health problems, only 35% of survivors responded affirmatively (Kadan-Lottick et al., 2002). Also, only 15% of survivors reported receiving a written summary of diagnoses and treatments to keep as a reference or to provide to their family doctor (Kadan-Lottick et al., 2002). Importantly, while approximately 87% of survivors report having a family doctor, only 34% report seeing that doctor for a problem related to follow-up cancer care (Kadan-Lottick et al., 2002). Also, thirty-one percent did not feel that their primary care physician could handle a problem related to a previous cancer (Kadan-Lottick et al., 2002). Much more research is needed to facilitate the development of specific LTFU care strategies that are both widely applicable and demonstrably effective (Friedman, Freyer, & Levitt, 2006). In general, however, quality follow-up care should take \"a risk-based approach to health care with a systematic plan for lifelong screening, surveillance and prevention, incorporating risk based on the original cancer diagnosis and therapy, genetic predispositions, lifestyle behaviours and comorbid conditions\" (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003; K. C. Oeffinger, 2004; K. C. Oeffinger & Hudson, 2004; K. C. Oeffinger 14 et al., 2004). Oeffinger et al. have labeled this concept optimum longitudinal risk- based care (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003; K. C. Oeffinger, 2004; K. C. Oeffinger & Hudson, 2004; K. C. Oeffinger et al., 2004). Further, quality LTFU care for survivors should emphasize the following components, regardless of the setting of care or the specific content of the care strategy: 1) longitudinal care that is considered a continuum from cancer diagnosis to eventual death, regardless of age; 2) continuity of care, consisting of a partnership between the survivor and a single health care provider who can coordinate necessary services; 3) comprehensive, anticipatory, proactive care that includes a systematic plan of prevention and surveillance; 4) a multidisciplinary team approach, with communication between the primary health care provider, specialists of pediatric and adult medicine, and allied/ancillary service providers; 5) health care of the whole person, not a specific disease or organ system, that includes the individual's family and his cultural and spiritual values; 6) sensitivity to the issues of the cancer experience, including expressed and unexpressed fears of the survivor and his family/spouse (K. C. Oeffinger, 2004) (emphasis added). Continuity of care is still emerging as a focus in the literature on survivor care; however, components of optimal care - surveillance for late effects, good communication and clear management roles - are clearly conceptually related to continuity and would constitute generic attributes of high-quality care for any chronic disease, cancer survivors included. Additionally, continuity of care as a partnership between the survivor and a health care provider or group of providers who can coordinate necessary services, is listed as a key component of optimum follow-up care for cancer survivors (K. C. Oeffinger, 2003), and as a basic tenet of 1 5 risk-based health care of childhood cancer survivors (K. C. Oeffinger, 2004). Continuity can also enhance screening, surveillance and prevention efforts targeted by knowledge of a patient's medical history and preferences (D. A. Christakis, Feudtner, Pihoker, & Connell, 2001; D. A. Christakis, Mell, Wright, Davis, & Connell, 2000; Ettner, 1999; G. K. Freeman & Richards, 1994; Lambrew, DeFriese, Carey, Ricketts, & Biddle, 1996; Love, Mainous, Talbert, & Hager, 2000; O'Malley, 1997; A. S. O'Malley & Forrest, 1996; Sturmberg & Schattner, 2001). Despite the hypothesized importance of continuity in optimal follow-up care, measurement of continuity and an assessment of its effects on processes and outcomes of care have not yet been undertaken for a survivor population. 2.3.1 CONTINUITY OF CARE 2.3.1.a Conceptualization The concept of continuity of care is defined and understood differently in different care domains. A systematic search of the literature reveals more than 100 distinct definitions of continuity (R. Reid, Haggerty, & McKendry, 2002). Worse, many investigators fail to explicitly define continuity and instead treat it as a self- evident concept. In a review of 583 published documents, only 32 percent explicitly defined what was meant by \"continuity of care\". The concept was implicitly defined in 48 percent, and in 20 percent, it was impossible to even infer the authors' concept of continuity (R. Reid, Haggerty, & McKendry, 2002). Despite a lack of agreement in the literature about what is meant by the concept of continuity of care, two common themes emerge from the dizzying array of definitions. First, continuity is about the care experience of an individual patient. 16 Second, it refers to the delivery of care over time (R. Reid, Haggerty, & McKendry, 2002). Each of these elements is a necessary but not sufficient condition to define continuity. Thus, for the purposes of this work, continuity of care will be defined as the patient experience of care over time with one or more providers that is coherent and linked (R. Reid, Haggerty, & McKendry, 2002). In addition to the two core elements of continuity, Reid et al. (2002) suggest that the concept of continuity can be split into three aspects: informational, relational, and management continuity. Informational continuity can be described as the availability and use of information on prior health events, patient preferences, and circumstances to enhance the likelihood that current care is appropriate (R. Reid, Haggerty, & McKendry, 2002). The transfer of information between providers or visits has been emphasized most often in the nursing literature. Relational continuity refers to an ongoing relationship between a patient and one or more providers (R. Reid, Haggerty, & McKendry, 2002), and has also been called interpersonal continuity, provider continuity, or personal continuity (J. M. Gill & Mainous, 1998; Saultz & Albedaiwi, 2004). An ongoing relationship with a provider enhances the link between past and present care, and gives patients a sense of predictability and coherence in their care. Relational continuity is emphasized most often in the primary care literature, where it translates to a measurement of patient \"loyalty\" to a provider and a provider's feeling of responsibility to a patient (R. Reid, Haggerty, & McKendry, 2002). Relational continuity has also been emphasized as a key component of a optimum follow-up 17 care for cancer survivors (K. C. Oeffinger, 2003), and will therefore be the focus of this work. Management continuity is the provision of timely and complementary services within the framework of a shared management plan (R. Reid, Haggerty, & McKendry, 2002). Most commonly disease- as opposed to patient-focused management continuity is of primary importance in the mental health literature, where consistent implementation yet flexibility of management plans are emphasized. The principles of management and informational continuity are also emphasized in the chronic disease model of care (Bodenheimer, Wagner, & Grumbach, 2002), which is similar in many respects to LTFU care models for cancer survivors (K. C. Oeffinger, 2003). Given the multiple co-morbidities, both cancer and treatment related, that appear in survivor populations as they age, the principles espoused in the management and informational continuity literature could be critical in providing quality long-term care for aging survivors. It is important to note that although these aspects of continuity are conceptually distinct from each other, in practice they may not be completely independent. For example, a long-term relationship between a patient and a single provider clearly relates to relational continuity; however, the physician in this example is more likely to be aware of the patient's medical history and preferences, and to use this information to guide current care, a measure of informational continuity (R. Reid, Haggerty, & McKendry, 2002). Additionally, a stable patient- 18 provider relationship enhances the incorporation of a patient's values, preferences and social contexts in treatment plans, as well as improves integration of care with other providers, both of which are hallmarks of management continuity. 2.3.1.b Impacts The literature to date on continuity of care has been based on multiple adult populations. In these studies, continuity has been demonstrated to be a critical feature of the processes of care necessary to ensure high-quality outcomes. As a component of quality care, continuity has been linked to dozens of analytical endpoints, which can be classified into process and outcome categories of Donnabedian's Structure, Process and Outcome framework (Donabedian & Suppl, 1966). The literature that examines the impact of continuity of care as it fits into Donabedian's model has been summarized in Table 2.1 below. Process of care refers to differences in the delivery of care. It can include patients' activities in terms of seeking care and adhering to recommendations, as well as a practitioner's ability to make a correct diagnosis and recommend the appropriate treatment (Donabedian & Suppl, 1966). Outcomes of care are the effects of care on a patient's health and wellbeing. Changes in patient knowledge and health behaviour, as well as satisfaction with care are included in this definition. Because many examinations of continuity are undertaken at the level of the primary care provider, utilization of hospital services, including emergency department usage and length of stay, is considered an outcome of care (Cabana & H., 2004) related to a patient's overall health and quality of life. 19 Table 2.1 Summary of Impacts of Continuity of Care Process^ Outcome Increased receipt of preventative services 1 4-8,19 Better patient-provider communication 2-3 Better adherence to medication regimens 14 Improved recognition of patient problems 14, 38, 39 Increased receipt of dental, nutritional and developmental advice 15-16 Improved immunization rates 19 Fewer missed appointments 14,18 Increased physician knowledge of patient's disease and treatment20 Increased patient satisfaction7,10,13,21,22 Reduced hospitalizations2,.8,9,11-13,35,36, 37 Reduced emergency department usage8,9,11,12,15,35,36 Reduced length of stay 13 Improved quality of life (overall & disease specific)23-34 1. (Ettner, 1999); 2. (G. K. Freeman & Richards, 1994); 3. (Love, Mainous, Talbert, & Hager, 2000); 4. (A. S. O'Malley & Forrest, 1996); 5.(0'Malley, 1997); 6. (Sturmberg & Schattner, 2001); 7. (Lambrew, DeFriese, Carey, Ricketts, & Biddle, 1996); 8. (D. A. Christakis, Feudtner, Pihoker, & Connell, 2001); 9. (D.A Christakis, Wright, Zimmerman, Bassett, & Connell, 2003); 10. (Weiss & A., 1989); 11. (J. M. Gill & Mainous, 1998); 12. (Parchman, Pugh, Hitchcock, & Larme, 2002); 13. (Wasson et al., 1984); 14. (Becker, Drachman, & Kirscht, 1974); 15. (Brousseau, Meurer, Isenberg, Kuhn, & Gorelick, 2004); 16. (Bradford, Kaste, & Nietert, 2004); 17. (Irigoyen et al., 2004); 18. (Alpert, 1964); 19. (D. A. Christakis, Mell, Wright, Davis, & Connell, 2000); 20. (Nielsen, Palshof, Mainz, Jensen, & Olesen, 2003); 21. (N. Breslau & Mortimer, 1981); 22. (Chao, 1988); 23. (Saultz & Albedaiwi, 2004); 24. (Fagerberg, Claesson, Gosman- Hedstrom, & Blomstrand, 2000); 25. (Fjaertoft, Indredavik, Johnsen, & Lydersen, 2004); 26. (Grunfeld et al., 1999); 27. (Harrison et al., 2002); 28. (Keitz, Box, Homan, Bartlett, & Oddone, 2001); 29. (Moher, 2001); 30. (Naylor et al., 2004); 31. (Preen et al., 2005); 32. (Samet, 2003); 33. (Smeenk, 1998); 34. (Williams et al., 2001); 35. (J. M. Gill, Mainous, & Nsereko, 2000); 36. (Cree, Bell, Johnson, & Carriere, 2006); 37. (Menec, Sirski, Attawar, & Katz, 2006); 38. (B. C. Reid & Rozier, 2006); 39. (Koopman, Mainous, Baker, Gill, & Gilbert, 2003) Process Endpoints: Continuity of care has been associated with increased utilization of preventative services in general (D. A. Christakis, Mell, Wright, Davis, & Connell, 2000; Ettner, 1999; Sturmberg & Schattner, 2001), as well as specifically with clinical breast exams (Lambrew, DeFriese, Carey, Ricketts, & Biddle, 1996; O'Malley, 1997), mammograms (Lambrew, DeFriese, Carey, Ricketts, & Biddle, 1996; O'Malley, 1997), and Papanicolau Smears (O'Malley, 1997). Similarly, increased continuity has been linked to completeness of immunization coverage at 19-months of age (D. A. Christakis, Mell, Wright, Davis, & Connell, 2000) and utilization of 20 dental, nutritional and developmental advice (Bradford, Kaste, & Nietert, 2004; Brousseau, Meurer, Isenberg, Kuhn, & Gorelick, 2004). The utilization of preventative screening and health promotion advice, enhanced by continuity of care, is important component of a risk-based approach to long-term follow-up of cancer survivors (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003; K. C. Oeffinger, 2004; K. C. Oeffinger & Hudson, 2004; K. C. Oeffinger et al., 2004). More generally, continuity has also been linked to improvements in quality of the physician-provider relationship. Increased continuity is associated with patient perceptions of physician-patient communication (Love, Mainous, Talbert, & Hager, 2000) including the ease a patient feels in talking to a physician (G. K. Freeman & Richards, 1994). It has also been associated with improved physician knowledge of a patient's medical and treatment (Nielsen, Palshof, Mainz, Jensen, & Olesen, 2003) and recognition of new problems (Bradford, Kaste, & Nietert, 2004), as well as patient perceived physician influence over treatment (Love, Mainous, Talbert, & Hager, 2000). In LTFU of cancer survivors, the ability to recognize incipient late effects, diagnose emerging late effects or new primary cancers promptly, and facilitating timely management of chronic toxicity could also therefore be enhanced by continuity of care (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003; K. C. Oeffinger, 2004; K. C. Oeffinger & Hudson, 2004; K. C. Oeffinger et al., 2004). Continuity of care also promotes feelings of responsibility to a single provider, exhibited as a reduction in the number of missed appointments (Alpert, 1964; Bradford, Kaste, & Nietert, 2004). 21 There are some process-level tradeoffs to good continuity, however, such as not being able to see the next available provider in an urgent situation (Love, Mainous, Talbert, & Hager, 2000); however, this limitation can be overcome with a well-functioning group practice or utilization by a practice of other professional groups (e.g. nurse practitioners). Additionally, access to multiple perspectives, or \"second opinions\" can serve as a method for avoiding incorrect diagnoses, while apparently reducing relational continuity (G. Freeman & Hjortdahl, 1997). Similarly, providers with a variety of different expertise may be able to complement others' skills, providing the most comprehensive care possible (Gallagher, Geling, & Comite, 2001; Wachter, 2001). The impact of these process-level tradeoffs on health outcomes and patient satisfaction has not yet been evaluated. Outcome Endpoints: In several cross-sectional and cohort-studies, better continuity led to reduced hospitalizations and use of emergency departments (D. A. Christakis, Feudtner, Pihoker, & Connell, 2001; D.A Christakis, Wright, Zimmerman, Bassett, & Connell, 2003; J. M. Gill & Mainous, 1998; Parchman, Pugh, Hitchcock, & Larme, 2002; Wasson et al., 1984). In the only randomized controlled trial investigating the effects of continuity on hospitalizations, continuity was associated with fewer hospitalizations, less use of emergency services, and shorter lengths of stay (Wasson et al., 1984). Continuity has also been associated with positive health behaviours such as reduced likelihood of drug or alcohol abuse (Ettner, 1999), and better glucose control in diabetics (O'Malley, 1997). One study, however, found that there was no 22 association between increased continuity and positive health behaviours, such as engaging in exercise, quitting smoking or having a healthy body mass index (Ettner, 1999). There is a large body of literature supporting the link between continuity of care and patient satisfaction. According to a systematic review on the topic (Saultz & Albedaiwi, 2004), twenty-two original articles, including four randomized controlled trials (some of which can be considered flawed as they failed to isolate continuity as the only uncontrolled difference between the study groups) found a consistent, positive association between continuity and patient satisfaction. 2.3.1.c Measurement As described, the concept of continuity has many attributes and interpretations, and is therefore notoriously difficult to measure in its conceptual entirety (R. Reid, Haggerty, & McKendry, 2002). In fact, from a quantitative perspective, there is no single measure that can adequately capture even the main aspects or types of continuity (R. Reid, Haggerty, & McKendry, 2002). Multiple measures are therefore required. Provider continuity, also known as relational continuity, predefined as the existence of sustained relationships between a patient and provider reflecting trust, understanding and effective communication (R. Reid, Haggerty, & McKendry, 2002), will be the focus of this research. The validity of measures of relational continuity was assessed by Reid et al. in 2003 (R. Reid et al., 2003). 23 Quantitative measures of provider continuity typically rely on the chronology of a patient's contact with healthcare providers, and are classified as measuring either the duration or intensity of the patient-provider relationship or the concentration of care. The earliest measures of chronology focused on intensity or duration of contact with a provider (R. Reid et al., 2003). Use of these measures is not recommended because they have not been conceptually validated (R. Reid et al., 2003). Measures of concentration of care, including the Usual Provider of Care (UPC) Index (N Breslau & Reeb, 1975), Continuity of Care (COC) Index (Bice & Boxerman, 1977), and Known Provider Continuity (\"K\") Index (Ejlertsson, 1984), measure how care is concentrated across different providers over a specific unit of time. The advantage of these measures lies in their wide use, intuitive appeal and ease of application (R. Reid et al., 2003). Their primary limitation, however, is the fact that they are affected by overall utilization levels, leading to spuriously high scores for low though consistent users of a specific physician (R. Reid et al., 2003; R. Reid, Haggerty, & McKendry, 2002). This weakness can be overcome by restricting the study sample to a higher-use population. The UPC Index is the simplest of the concentration measures, and is defined as simply the proportion of an individual's visits to a \"usual provider\" (the provider most frequently seen). Given that there is a negative correlation between the UPC score and the total number of visits, there is a tendency for individuals who have fewer visits to have higher UPC scores (R. Reid et al., 2003). The UPC score is also 24 limited by a lack of a zero point, and low variability (R. Reid et al., 2003). Despite these weaknesses, the UPC score is a valuable measure because of its wide use, ease of application, and conceptual simplicity (R. Reid et al., 2003). The COC Index was developed to deal with issue of the link between continuity and overall visits. If the number of visits increases, but the distribution of visits to specific providers remains constant, the UPC score will also remain unchanged. In the same situation, the COC Index will increase proportionally as the number of visits increases. The weaknesses of the COC Index, however, include a lack of conceptual clarity, and difficulties in interpretation and calculation (R. Reid et al., 2003). Both the UPC and COC Indices can be measured broadly by examining all visits to a specific clinic, for example (Jee & Cabana, 2006). Such a broad measure would account for continuity experienced across providers working within the same practice, clinic or hospital (R. Reid et al., 2003). Alternatively, these tools measure continuity at the level of the individual provider, which better measures the relationship between a patient and an individual provider (R. Reid et al., 2003). Measures of sequential care, including the Sequential Continuity Index (SECON) (Steinwachs, 1979), have been developed to assess patients' consecutive visits to the same provider. These measures have been conceptually linked to estimating the need for information transfer between practitioners over time (R. Reid et al., 2003). 25 The validity of measures of concentration and sequence of care, specifically the UPC Index, COC Index, \"K\" Index, and SECON, has been assessed for the Canadian context using administrative data (R. Reid et al., 2003). Substantial concurrent and predictive validity was found for these four measures (R. Reid et al., 2003). In general, the measures constructed over a two-year follow-up window outperformed those constructed annually (R. Reid et al., 2003). Additionally, measures constructed with specialist visits attributed back to the originating physician outperformed those that included primary care visits only (R. Reid et al., 2003). This is consistent with the theory that, in a Canadian context, specialty care and primary care are distinct, but connected and thus form a continuum. Additionally, these measures are well correlated with the survey-based measure of provider continuity (R. Reid et al., 2003). They also note that the UPC and \"K\" indices are likely to be preferred by policy makers because of interpretability, minimal data needs, and ease of calculation (R. Reid et al., 2003). Researchers, however, are more likely to prefer either the UPC or COC indices because they are the most commonly used, easing comparison of results across studies (R. Reid et al., 2003). 2.4 CONCEPTUAL FRAMEWORK: ANDERSEN'S BEHAVIOURAL MODEL OF UTILIZATION Andersen's 2001 Behavioural Model of Utilization was used to guide the application of study objectives and research questions (Andersen, 1968\u00E2\u0080\u009E 1995; Andersen & Davidson, 2001) (see Figure 2.1). Specifically, the model was used to provide a conceptual backdrop on which to explain - not predict - utilization. The 26 model centers on Anderson and Davidson's definition of access to health services (Andersen & Davidson, 2001): We define access as the actual use of personal health services and everything that facilitates or impedes their use. It is the link between health services, systems, and the populations they serve. Access means not only getting to service, but also getting to the right service at the right time to promote improved health outcomes (p.3). Access, Andersen and Davidson claim, is linked to health services use, and to the effective and efficient delivery of health services (Andersen & Davidson, 2001). Specifically, they argue that the continuum of access can be broken down into four distinct concepts: potential access, realized access, efficient access, and effective access. Potential access is the presence of absence of enabling resources as part of contextual or individual relationships (Andersen & Davidson, 2001). In contrast, realized access, which was the primary focus of this work, is a measurement of the actual use of health services (Andersen & Davidson, 2001). Efficient access and effective access are outcomes related to the use of health services. Specifically, if access is effective, utilization improves health status and/or patient satisfaction with services (Andersen, 1995). If access is efficient, utilization improves outcomes and/or patient satisfaction relative to the amount of services used (Andersen, 1995). 27 OutcomesHealth Behaviours Individual Characteristics Contextual Characteristics Figure 2.1 Andersen's Behavioural Model of Utilization V Predisposing ---> Enabling \u00E2\u0080\u0094I> Need Demographic^Health Policy^Environment 1 i Social^Financing^Population health indices Beliefs^Organization Predisposing --+ Enabling --+ Need Demographic^Financing^Perceived Social^Organization^Evaluated Beliefs Personal health practices Process of medical care Use of personal health services Perceived health Evaluated health Consumer satisfaction The model focuses on both the recursive nature of the use of health services and how personal health services utilization relates to perceived and evaluated health outcomes, and satisfaction with care, which in turn will affect subsequent utilization (Andersen, 1995; Andersen & Davidson, 2001). This model portrays the multiple influences on health services use: each component may make an individual contribution predicting use, or may participate in a causal ordering (Andersen, 1995; Andersen & Davidson, 2001). The sequence of each model component is an indirect measurement of how influential that particular component is determining use. Thus, need is the most prognosticative and has the highest relative importance of the three components that predict use. Where indicated, the discussion of the various sections of the model are split here between a general description of each factor (\"General\") and description of how that factor may relate to health services utilization specifically in a cancer survivor population (\"Survivors\"). 2.4.1 CONTEXTUAL CHARACTERISTICS 2.4.1.a Predisposing Factors Contextual characteristics are split into predisposing, enabling, and need- related factors. Under predisposing contextual characteristics, Andersen includes demographic, social and beliefs components. These factors capture the influence of community characteristics on health and use of personal health services. Andersen gives the examples the differing availability of specific health services in communities with different population sizes and demographics (Andersen & 29 Davidson, 2001). He also cites differences in availability or access due to differing educational or racial compositions. 2.4.1.b Enabling Factors General: Enabling contextual characteristics focuses on the health care system, including all the policies, resources, and organizational factors therein, recognizing that the structure of health services is itself an important predictor of use and changes in patterns of use over time (Andersen, 1995). Andersen argues that both community and personal \"enabling resources\" must be present in order for use to occur (Andersen, 1995): personnel and facilities must be available and accessible for patients. He breaks down this category into health policy, financing and organization components, reflecting the independent contribution national health policy, and the availability and organization of resources to use patterns (Andersen, 1995; Andersen & Davidson, 2001). Survivors: Structured follow-up programs and the physicians' usage of practice guidelines may affect use patterns within a survivor population. Discussions around optimum follow-up care models for survivors focus on issues related to the availability and accessibility of (1) personnel who are knowledgeable about long-term risks associated with survivorship and willing to provide care; and (2) facilities where services could take place. Lack of accessibility of services may be a weakness of oncologist-based models for follow-up, for example. And, a potential lack of knowledgeable personnel has been considered a weakness in primary care- driven models. Interestingly, however, provider continuity of care has been associated with improved physician knowledge of a patient's medical and treatment 30 (Nielsen, Palshof, Mainz, Jensen, & Olesen, 2003), and thus has the potential to enhance the availability of follow-up care in the primary care setting. 2.4.1.c Need The need criterion emphasizes the influence of health-specific environmental characteristics as well as general population health indices on need use patterns at the individual level. 2.4.2 INDIVIDUAL CHARACTERISTICS 2.4.2.a Predisposing Factors Consistent with the Contextual Characteristics section, Andersen includes demographic factors, social structure and health beliefs among \"predisposing characteristics\". Demographic factors, including gender or age, are known predictors of use in both survivors (A. K. Shaw et al., 2006) and in the general population (Green & Pope, 1999; Hibbard & Pope, 1986). Elements of social structure - education, socioeconomic status (SES), occupation, ethnicity and geography - have also been shown to affect services use in various populations (A. K. Shaw et al., 2006) (Green & Pope, 1999). Health beliefs include knowledge, attitudes and beliefs about health and how use (or lack thereof) of health services will also affect subsequent perceptions of need (Andersen, 1995; Andersen & Davidson, 2001). 2.4.2.b Enabling Factors Enabling factors at the individual level include financial and organizational factors. In order to access care availability in their community, an individual must have the means and knowledge necessary to access these services. Income, health insurance, having a regular source of primary care, as well as issues of travel time and wait time for specific services are important individual-level enabling factors. 31 Although health insurance is not an issue in Canada from the perspective of primary, secondary or tertiary care access, it may affect use of prescription medications and access to non-physician providers, such as physiotherapists or massage therapists, who are not currently covered by provincial health insurance plans in some provinces. 2.4.2.c Need General: Need, both perceived and evaluated, has an obvious effect on utilization: individuals who are, or who perceive themselves to be, sicker will seek more care. Survivors: As presented early in this chapter, a majority of survivors of childhood cancer suffer from late-occurring or chronic health conditions severe enough to require long-term treatment or surveillance (Bleyer et al., 1993; Wendy Landier et al., 2004; W. Landier, Wallace, & Hudson, 2006; National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003; Oeffinger, 2007; K. C. Oeffinger & Hudson, 2004). Their evaluated need for medical care is therefore likely to be higher than that of the general population, resulting in elevated use of health services (A. K. Shaw et al., 2006). Survivors' knowledge of the risks of late effects, and therefore perceptions of personal need for health services, could be an important predictor of use (K. C. Oeffinger & Wallace, 2006). Importantly, survivors themselves may not have the necessary information to seek care appropriately or to guide their primary care physician (A. K. Shaw et al., 2006). Indeed, when asked whether past cancer 32 therapies could cause future health problems, only 35% of survivors responded affirmatively (A. K. Shaw et al., 2006). 2.4.3 HEALTH BEHAVIOURS Andersen splits health behaviours into personal health practices, process of medical care, and use of personal health services. 2.4.3.a Personal Health Behaviours The inclusion of personal health behaviours as a predictor of health services use, acknowledges the role of personal health practices, such as diet, exercise, and self care, in interacting with the use of formal health services to influence health outcomes (Andersen, 1995). There has been some literature suggesting that continuity of care could play a role in better encouraging positive health behaviours, such as reduced likelihood of drug and alcohol abuse (Ettner, 1999). 2.4.3.b Process of Medical Care The process of medical care was included as a measure of provider-patient interactions (Andersen & Davidson, 2001). The quality of the patient-provider relationship can act as an enabling resource to facilitate or impede health services use (Andersen, 1995). Provider continuity of care has a demonstrably positive effect on the quality of a relationship between a patient and survivor, and could therefore affect subsequent utilization. Specifically, increased continuity is associated with patient perceptions of physician-patient communication (Love, Mainous, Talbert, & Hager, 2000) including the ease a patient feels in talking to a physician (G. K. Freeman & Richards, 1994). 33 2.4.3.c Use of Personal Health Services Andersen breaks down the use of personal health services into type (primary, secondary or tertiary care), site, purpose and time interval. The quantity and type of use varies based on combinations of predictive characteristics. 2.4.4 OUTCOMES In later iterations of his model, Andersen includes perceived and evaluated health status and patient satisfaction as outcomes of \"effective\" and \"efficient\" access to care. If access is effective, utilization improves health status and/or patient satisfaction with services (Andersen, 1995). If access is efficient, utilization improves outcomes and/or patient satisfaction relative to the amount of services used (Andersen, 1995). Perceived and evaluated health status were added due to the recognition that the function of health services is to maintain and improve the health of users. Although health outcomes related to efficient and effective use are not the subjects of this project directly, there is a substantial body of evidence supporting the relationship between good provider continuity of care and improvements in patient health and satisfaction (D. A. Christakis, Feudtner, Pihoker, & Connell, 2001; D.A Christakis, Wright, Zimmerman, Bassett, & Connell, 2003; J. M. Gill & Mainous, 1998; Parchman, Pugh, Hitchcock, & Larme, 2002; Wasson et al., 1984). Andersen's model supports the hypothesis that continuity - if it successfully improves health status or patient satisfaction with care - can recursively act as an enabling resource, affecting subsequent care-seeking behavior. 34 2.5 SUMMARY For this proposed research, cancer survivorship is defined as a distinct phase of the cancer trajectory that follows primary treatment and precedes recurrence, onset of second cancer, or death. The full extent of long-term health consequences, and therefore the specific health care needs, of individuals in the survivorship phase of the cancer trajectory, after a diagnosis of cancer before age 20, is not yet known; however, recent evidence suggests that survivors suffer from late-occurring or chronic health conditions severe enough to require long-term treatment and medical surveillance (Bleyer et al., 1993; Wendy Landier et al., 2004; W. Landier, Wallace, & Hudson, 2006; National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003; Oeffinger, 2007; K. C. Oeffinger & Hudson, 2004). Up to 75 percent of childhood cancer survivors have been reported to suffer from at least one chronic treatment-related health problem - referred to in the literature as a \"late effect\" of treatment - and between 30 and 50 percent of survivors have late-occurring or chronic health problems severe enough to require long-term treatment (Garre et al., 1994; Geenen et al., 2007; M. Hudson et al., 2003; Humpl, Fritsche, Bartels, & Gutjahr, 2001; Stevens, Mahler, & Parkes, 1998) Given the potential severity of late effects, and the variability in the time of appearance and resolution, both the American Cancer Society and the Pediatric Working Group of the Canadian Strategy for Cancer Control recommend life-long follow-up of cancer survivors (Bhatia et al., 2003; Bleyer et al., 1993; Canadian Strategy for Cancer Control, 2002). The rationale for LTFU care, consisting of routine, periodic medical visits and tests, is to reduce or eliminate excess morbidity 35 and mortality among survivors by detecting incipient late effects, diagnosing emerging late effects or new primary cancers promptly, and facilitating timely management of chronic toxicity (Grunfeld, Dhesy-Thind, & Levine, 2005; Skinner, Wallace, & Levitt, 2007; Skinner, Wallace, Levitt, & Uk Children's Cancer Study Group Late Effects, 2006). Although several sets of guidelines for LTFU have emerged in recent years (Wendy Landier et al., 2004; W. Landier, Wallace, & Hudson, 2006), there is little consensus among them about how often follow-up should occur, where it should occur or what specific procedures it should comprise (Goldsby & Ablin, 2004). Further, these guidelines are not necessarily evidence- informed (Goldsby & Ablin, 2004) and the effectiveness of recommended interventions is often unclear (Institute of Medicine, 2006). The importance of continuity and coordination of care in a quality follow-up care for cancer survivors has also been emphasized by the National Cancer Policy Board (Institute of Medicine, 2006), and is listed as a key component of optimum follow-up care for cancer survivors (K. C. Oeffinger, 2003) and as a basic tenet of risk-based health care (K. C. Oeffinger & Hudson, 2004). As a component of quality care, continuity of care has been linked to dozens of positive process- and outcomes- of-care endpoints in other populations (Donabedian & Suppl, 1966). As a first step in the assessment of follow-up care received by survivors, this study attempts to measure the long-term utilization of physician services among survivors. This study will also assess the continuity of primary provider care, a key component of a quality follow-up care strategy for survivors. 36 CHAPTER 3: METHODOLOGY 3.1 THE CHILDHOOD, ADOLESCENT, AND YOUNG ADULT CANCER SURVIVORSHIP RESEARCH PROGRAM The Childhood, Adolescent, and Young Adult Cancer Survivorship Research Program (CAYACS) is a population-based cohort study of long-term impacts, support, and interventions to maximize quality of life among survivors of cancer diagnosed under the age of 25 years in British Columbia, Canada. CAYACS is also a resource for survivor research and for knowledge translation for policy and practice. The program is supported by the Canadian Cancer Society through the National Cancer Institute of Canada. For further details, please see Appendix A. As part of the CAYACS research program, this project utilized linked registries, and administrative and clinical datasets to describe the extent and patterns of physician services utilization and primary care provider continuity in a cohort of cancer survivors compared to a population-based comparison group. 3.2 STUDY DESIGN Research on the long-term care of cancer survivors necessitates the use of cohort methodology (National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003). Prospective cohort studies requiring long-term prospective data collection are challenging, however, especially in a population of children or young adults. Limitations of this approach include the effective recruitment and retention of a representative sample and ideally comparators; continuous access to health records; minimizing attrition bias, particularly in the transition from pediatric to adult care; and the potential biases associated with self-report data (D. T. Campbell 37 & Stanley, 1963; National Cancer Policy Board (U.S.), Hewitt, Weiner, & Simone, 2003). Additionally, the financial and time commitments necessary to undertake this approach are particularly onerous, to the point of being nearly prohibitive (D. T. Campbell & Stanley, 1963). To overcome the biases and challenges associated with long-term prospective follow-up, this observational study employed a historical cohort time series methodology using registries, administrative data, and clinical data from health records. Population-based, administrative data sets have been used to assess service utilization - among other factors - for many years, and the data linkage and analysis procedures are well-established and have been well-validated (Chamberlayne et al., 1998; Fair, 1997; Newcombe, Fair, & Lalonde, 1997). The use of population-level data sets and record linkage allows for the study of an entire cohort, eliminating selection and sampling biases, and ensuring representativeness. The use of administrative data also eliminates biases associated with self-report data and the attrition problems that are common in studies of long-term follow-up, particularly those using self-reports (D. T. Campbell & Stanley, 1963). Limitations associated with the use of administrative data will be addressed in Chapter 6. This thesis uses a dual methodology approach: each outcome - visit counts and continuity score - was assessed using both cross-sectional and longitudinal methods. Additionally, survivor only and survivor versus comparator models were created for each method-outcome combination. This complex approach was selected to allow for the examination of both an age-trend, and a long-term 38 assessment of outcomes. Case only models were used so that the effect of clinical factors (inappropriate for inclusion in case-control models) on visits and continuity could be modeled. Case-control models were used to examine the difference in visit patterns and continuity between survivors and the general population. 3.2.2 OBJECTIVES AND RESEARCH QUESTIONS Neither patterns of physician services utilization, nor provider continuity, have been assessed in a cancer survivor population. This descriptive and analytic research took the necessary first steps to fill this knowledge gap. The specific objectives and research questions of this descriptive and analytic project are as follows: Objective 1: To describe long-term physician utilization among survivors of childhood cancer compared with general population comparators, and to examine factors associated with survivors' use of physician services. Question 1.1 What is the overall and biannual by age pattern 3 of physician services utilization in the survivor cohort, overall and separately by physician specialty? Question 1.2 How do clinical variables - such as specific cancer diagnosis and treatment - and sociodemographic variables - such as SES, 3 Overall physician services utilization refers to measures of central tendency and variation calculated for the entire study period, while biannual health services utilization refers to measures of central tendency and variation calculated for each two-year age interval (in the biannual file). 39 gender and age - affect the pattern of physician services utilization among survivors? Question 1.3 How does the total and biannual pattern of physician services utilization differ between survivors and their general population comparators? Question 1.4 Is there an age-related trend in the pattern of physician services utilization and does this trend differ between survivors and general population comparators? Objective 2: To assess levels of primary care provider continuity of care among survivors of childhood cancer compared with general population comparators, and to examine factors associated with survivors' continuity scores. Question 2.1 What is the overall and biannual 4 provider continuity of care score in the survivor cohort? Question 2.2 How do clinical variables - such as specific cancer diagnosis and treatment - and sociodemographic variables - such as SES, gender and age - affect provider continuity of care among survivors? 4 As above, overall provider continuity of care refers to measures of central tendency and variation calculated for the entire study period, while biannual provider continuity of care refers to measures of central tendency and variation calculated for each two-year age interval (in the biannual file). 40 Question 2.3 How does overall and biannual provider continuity of care score differ between survivors and their general population comparators? Question 1.4 Is there an age-related trend in the continuity scores and does this trend differ between survivors and general population comparators? 3.3 IDENTIFICATION OF THE STUDY COHORT 3.3.1 SURVIVORS The cancer survivor cohort for analysis of health care utilization and continuity of care in this project (a sub-sample of the overall CAYACS cohort) consisted of 1322 individuals who met the following criteria for inclusion: (1) diagnosed with a first primary cancer Groups I-XII of the International Classification of Childhood Cancers (ICCC) (Kramarova & Stiller, 1998); (2) resident in BC at time of diagnosis; (3) first diagnosed before 20 years of age between 1 January 1981 and 31 December 1995; and (4) survived five years or more after last cancer diagnosis (referred to henceforth as \"five-year survival\"). Of this eligible \"at risk\" population, those who were linked to the BC Medical Services Plan (MSP) Client Registry, which is the registration file of the provincial health insurance system, were assessed for continuity of care. Cases were identified through the British Columbia Cancer Registry (BCCR) as reported to the CAYACS program. Follow-up of the cancer cohort began five years post-last diagnosis, starting in 1986, and continued to the end of 2000, loss of 41 follow-up (due to missing MSP records), or until death, whichever occurred first. The maximum number of years of follow-up data available for each individual was therefore sixteen years. Follow-up of survivors began at five years post-last diagnosis to ensure that long term follow-up care, and not care related to cancer treatment or early side effects of treatment, was measured for physician services utilization and primary care continuity. 3.3.2 COMPARATORS Patterns of utilization and provider continuity of care for survivors was compared with patterns observed in a 10:1 random sample of the BC population identified from the BC MSP Client Registry, frequency-matched by birth year and gender to the survivor cohort. Birth year matching was used to control for period effects. However, because the source of the comparison sample differed from that of the survivor cohort, age at the start of follow-up (and therefore mean age at any time-point under study) was expected to differ between cases and comparators. Survivor follow-up began five-years following last known cancer diagnosis or relapse, or upon entry into MSP, whichever was later; comparator follow-up began at age five-years or entry into MSP, whichever was later. Thus, age at the start of follow-up was expected to be older in the survivor population that in the comparator group. A 10:1 control:case ratio for the comparator sample ensured that statistical power was sufficient to allow for assessment of multiple determinants, and was computationally feasible. 42 3.4 DATA SOURCES AND ETHICS APPROVAL Data requirements of this project included unique personal identifiers (for computerized record linkage), treatment data, information on demographic and other potential outcome determinants, and utilization of physician services, for both the survivor and comparison groups. Access to the required data sources for the purposes outlined in this proposal, as part of the CAYACS Research Program, was approved by the BC Cancer Agency - University of British Columbia (UBC) Ethics Board and the Ministry of Health's Data Access Committee (See Appendix B). A linked, person-level data file was prepared for analysis. Anonymized linked files, with only a study-specific scrambled ID, for both survivors and comparison subjects, were used for analyses. To further maximize confidentiality protection and data integrity, linked files (with study ID only) were maintained on a separate computer network than the treatment records, and were stored in a separate secure file for linkage back to other subject datafiles on an analysis-specific basis. The CAYACS administrative coordinator maintains secured access to this network. 3.4.1 CASE ASCERTAINMENT The BCCR of the BC Cancer Agency (BCCA) (www.bccancer.bc.ca ) identifies all diagnosed cases of cancer to residents of BC, and collects cancer-related data. For this project, the BCCR was used to identify the survivor cohort and to collect the following data: \u00E2\u0080\u00A2 Classification of cancer diagnosis into the International Classification of Diseases for Oncology (ICD-0) (Organization, 1990); 43 \u00E2\u0080\u00A2 Date of cancer diagnosis; \u00E2\u0080\u00A2 Date and underlying cause of death, if applicable (routinely reported to the BCCR by BC Vital Statistics); and \u00E2\u0080\u00A2 Demographic data, including date of birth and gender Oncology charts from the BCCA, containing histories of treatments conducted at the BCCA, were also used for this project. Specifically, treatment data for individuals who received radiation therapy or who received treatment at age 17 or above was abstracted. 3.4.2 TREATMENT DATA COLLECTION Treatment data for cancer survivors was abstracted from health records both at BCCA and British Columbia Children's Hospital (BCCH), into a treatment database within CAYACS. The BCCA has a mandate for provincial care, control, and research, and operates 4 treatment centres treating those aged over 17 years, and also providing all radiotherapy services. Oncology charts from the BCCA, which contain treatment histories for all cases treated at either a BCCA treatment centre or for children treated at the old Health Centre for Children or BCCH, were accessed to abstract treatment data into a treatment database. The BCCH (and its predecessor, the Health Centre for Children at Vancouver General Hospital) is the provincial referral centre for patients diagnosed with cancer aged 17 and younger. Treatment data was abstracted into the CAYACS treatment database from a review of oncology charts from the BCCH. 44 Available treatment data included treatment modality (radiation, chemotherapy, and/or surgery), treatment details such as dates, region and dose of radiation or specific chemotherapy drugs, as well as relapse information, if applicable. 3.4.3 BC LINKED HEALTH DATABASE The Client Registry of the provincial Ministry of Health is the patient masterfile for all health programs in BC. It contains person-specific records for all residents of BC who ever received healthcare in the province covered since 1986. Excepted are individuals with declared First Nations status, the Canadian Forces, veterans, and inmates in federal penitentiaries, whose health services coverage is provided by the Federal Government. 5 Data is available annually and includes Personal Health Number (PHN), demographic data, geographic residence information, enrollment and cancellation dates. Additionally, physician services paid on a fee-for-service basis are also recorded in a claims database. Data available for each claim include type of physician or health practitioner, (scrambled) physician or practitioner identifier, service and indication for service (not always recorded), and date and location of service. The MSP data are maintained in the BC Linked Health Database (BCLHD), which was developed, and is housed, by the Centre for Health Services and Policy Research (CHSPR) at the University of British Columbia. The database contains s Dependents of veterans and Canadian forces members are not included in federal coverage, and thus would be included in the BC MSP files. Of concern, however, is the exclusion of aboriginal cancer survivors and population comparators. 45 linkable administrative health services datasets from the provincial health services plan for the entire population of BC that can be aggregated into individual-level, longitudinal records, making it an ideal data source for health services research. For the purposes of this project, the BCLHD was used to capture the following data for cancer survivors and comparators: \u00E2\u0080\u00A2 Physician visit information (including specific practitioner identifier and specialty, and date of service); and \u00E2\u0080\u00A2 Demographic factors (annual SES quintile, health authority of residence, rural or urban designation). The data are available from 1985 onwards, allowing individuals to be traced forward over time from 1986 to 2000. Through previous assessments, the BCLHD has been found to have high levels of completeness and validity for research purposes (Chamberlayne et al., 1998). 3.4.4 Linkage and Preparation Data from the BCCR, BCCH charts, and the BCHLD were linked at CHSPR using personal health numbers, creating a person-level analytic file containing information on cancer and treatment, physician services utilization, demographics, and vital statistics. Patient and provider identifiers were removed in order to preserve anonymity. Using the linked datasets, a comprehensive patient group, set of treatment and outcome data, and sets of utilization data for publicly-funded services were available at a sufficient level of completeness, detail and reliability for 46 both survivors and comparators, in a patient-based format, providing a unique data set to examine issues of utilization and provider continuity of care. 3.5 STUDY VARIABLES A description of study variables is attached in Table 3.1, and further details can be found below. As described in Chapter 2, utilization and continuity were examined as they pertain to Andersen's Behavioural Model of Utilization (Andersen & Davidson, 2001). A modified model, demonstrating the fit of each variable within the Andersen's framework is attached in Figure 3.1. Text and linkages that are grayed were not measured during this analysis. Red text indicates a study variable. 3.5.1 OUTCOME VARIABLES 3.5.1.a Utilization of Physician Services Utilization of physician services, including the mean and median number of visits to primary care practitioners, specialists and non-physician providers, was described for the duration of follow-up, and on a biannual basis as a precursor to the provider continuity of care calculations. Physicians were categorized as primary care, specialist or non-physician practitioners 6 based on specialty codes declared upon entry into the BC College of Physicians and Surgeons. For the purposes of this study, a physician visit was defined as a claim for a 'face-to-face' encounter with a specific physician in a single day - identified by a scrambled physician billing code - 6 Primary care physicians are those with a specialty code of 00. Specialists are any physician providers with a specialty code other than 00: 1-15, 19, 20, 28, 37, 38, 40-42, and 44-48. Non- physician practitioners included, for example, licensed nurse practitioners, physical therapists, chiropractors, optometrists and naturopaths. Provider codes for this category are 30-32, 34, 39, 43, 80-82, 85, and 87. 47 excluding laboratory, pathology, and radiologic services, as is necessitated by MSP data. Individual visits could therefore be ascribed to an individual provider, which was necessary to compute any measures of continuity. Specifying a limit of one visit to a specific provider per patient per day controlled for physicians billing multiple services for one visit. Visits to primary care practitioners, specialists, and non-physician providers were counted separately and specialist visit counts were stratified according to physician specialty. Data restrictions did not allow for the ascription of specialist visits back to the referring primary care physician, which would have improved the validity of the continuity measures (R. Reid et al., 2003); however, this information will be added to subsequent work. Time and data restraints also did not allow for the assessment of indication for service, and the specific service rendered. This also will be done in future in order to obtain an accurate picture of which visits are, in fact, related to follow-up cancer care. 48 Table 3.1 Study variables Variable^Description Outcome Variables Utilization of Physician Services Primary Care Visits Number of visits physician Specialist Visits^Number of visits specialty Non-Physician^Number of visits to non-physician Visits^practitioners Provider Continuity of Care Continuity of Care Index^n,2 \u00E2\u0080\u0094 N N(N \u00E2\u0080\u00941) Usual Provider of^max(xp...,xk )-1 Care^ N \u00E2\u0080\u00941 to a primary care to specialist, by Type^Source Count^BCLHD Count^BCLHD Count^BCLHD Contin.^Calculated Calculated Where ni is number of visits to provider i and N is total number of visits in a defined period. Where Xi is number of Contin. visits to provider i in a defined time period and N is total no. visits. Clinical Covariates Cancer Diagnosis Group Age at Diagnosis Calendar Period of Diagnosis Treatment Modality Time Since Diagnosis International Classification of Childhood Cancer I-XII Age in years (0-4,5-9,10-14, 15-19) 1986-1990, 1991-1995 Type of primary treatment In years from five year survival Nominal BCCR Interval BCCR Interval BCCR Nominal Chart Review Interval BCCR Demographic Covariates Age^Age in years (0-4,5-9,10-14,15-19,20- 24,25-29,30+) Gender^Male or Female SES Quintile^1-5, derived from on enumeration area- level census data Health Authority^1-5, based on region of residence Rural/Urban^1-4, derived from census categories. Interval Calculated Nominal BCCR Ordinal BCLHD Nominal BCLHD Ordinal BCLHD 49 rs^I Ica pr. Process of medical care Contintnitof care Use of personal health services Primary care. specialist, non physician visit patterns A Figure 3.1 Andersen's Behavioural Model of Utilization: Variable Placement Contextual^ Individual ^ Health ^ Outcomes Characteristics Characteristics Behaviours Predisposing --1\u00E2\u0080\u00A2Enabling Need Predisposing Demographic --* Enabling^Need Financing Demographic Health Policy Environment Gender, Age^SES Perceived Social Financing Populat . on Organization Evaluated health indices Social Urban/ Rural.^ICCC group, Health Authority^treatment Beliefs^uanizat ion Beliefs Perceived health Evaluated health Consumer satisfaction 3.5.1.b Provider Continuity of Care Two different measurement tools assessing concentration of care were used to assess continuity in the context of primary care: the Continuity of Care (COC) Index (Bice & Boxerman, 1977) and the Usual Provider of Care (UPC) Index (N Breslau & Reeb, 1975). These measures are considered to be the most internally and conceptually valid tools for measuring continuity using administrative data (R. Reid et al., 2003). Dual measurement of continuity was deemed desirable to maximize research application from both the researcher and decision maker perspectives (R. Reid et al., 2003) 7 . Although both indices can be measured broadly by examining all visits to a specific clinic (Jee & Cabana, 2006), for the purposes of this work, the COC and UPC measures were identifiable for a particular provider, allowing focus on the potential strength of the relationship between a patient and a specific physician. Continuity measures were calculated in the context of primary care only. To maximize the validity of these measures, continuity was calculated using two-year periods, instead of on an annual basis (R. Reid et al., 2003). To control for the effect of low utilization levels on these measures, and to maximize their validity, calculations were restricted to those who had three or more visits over a two year period (R. Reid et al., 2003). 7 As stated in the previous chapter, the UPC index tends to be preferred by decision makers due to ease of calculation and interpretation (R. Reid et al., 2003). The COC index, alternatively, tends to be preferred by researchers because of its ability to control for the low user, high score issue, and because it is most widely cited (R. Reid et al., 2003). 51 Table 3.2 Summary of ICCC Class II III IV V VI Diagnosis Leukemias Lymphomas CNS Cancers Neuroblastomas Retinoblastomas Renal Tumors Treatment Category Treatment Category 3.5.2 CLINICAL COVARIATES Evidence suggests that particular treatment combinations may place survivors at higher risk of subsequent health problems, and thus it was expected that treatment modality and specific diagnosis would affect the utilization of physician services (K. C. Oeffinger & Hudson, 2004). Therefore, disease and treatment factors were included as independent predictors of patterns of care and provider continuity. To assess the effect of specific cancer diagnosis, cancer survivors were grouped according to the ICCC diagnosis groups (Kramarova & Stiller, 1998) using diagnosis data from the BCCR (see Table 3.2). Diagnosis Groups Class Diagnosis VII^Hepatic Tumors VIII Malignant Bone Tumors IX^Soft Tissue Sarcomas X^Germ Cell Tumors XI^Other Malignant Epithelial Neoplasms XII^Other Unspecified To assess specific treatment effects, primary treatment modalities were also considered. Treatment classifications are summarized in Table 3.3. Table 3.3 Treatment Classifications Radiation only Chemotherapy only Surgery only Radiation and chemotherapy Radiation and surgery ^ Chemotherapy and surgery; Radiation, chemotherapy and surgery; Other (hormone therapy or transplants) No treatment 52 Age at diagnosis and time since diagnosis were also considered, and were treated as both continuous and categorical variables, broken into five-year intervals: <5 years, 5 to <10 years, 10 to <15 years, and 15 to <20 years. The calendar period of diagnosis, 1986-1990 or 1991-1995 was considered as a potential covariate in order to account for any changes in diagnosis and treatment technologies that may contribute to subsequent utilization patterns. Lastly, status at five years post diagnosis was used to determine which patients can be considered as needing follow-up care. Patients who were undergoing cancer therapy at five years post-diagnosis were not receiving follow-up care and thus were excluded from the analysis. 3.5.3 DEMOGRAPHIC COVARIATES Demographic factors, including attained age, gender rural versus urban status, residing health authority and SES, were included as independent predictors of utilization and continuity of care. All demographic covariates were assessed at the following timepoints: for the cross-sectional analyses, demographic covariates were assessed either at the start of five year follow-up or entry to MSP (whichever is later) for survivors, or at the fifth birthday or entry into MSP (whichever is later) for the comparators; for the longitudinal analyses, the covariates were re-assessed at the beginning of each two-year interval. Classification as rural or urban was derived from Statistics Canada Census data. Urban areas were defined using Census Metropolitan Areas and Census 53 Agglomerations. These are further subdivided into four classes as follows (Wilkins, July 2002): \u00E2\u0080\u00A2 Rural: population of less than 10,000 \u00E2\u0080\u00A2 Untracted agglomeration: small community - population of at least 10,000; \u00E2\u0080\u00A2 Tracted agglomeration: large community - population of at least 50,000; \u00E2\u0080\u00A2 Metropolitan area: urban core with a population of at least 100,000. In addition to a rural-urban designation, individuals were grouped into BC's regional health authorities, whose boundaries are defined by the BC Ministry of Health. These five authorities are defined by the BC Ministry of health using geographical boundaries (based on historical local health area boundaries), and were derived from the first three digits of an individual's postal code. Socioeconomic status was defined from Statistics Canada using quintiles of income per person-equivalent derived from postal code and census data, adjusted for neighbourhood household size, of the residence at start of period at risk. 3.6 STATISTICAL ANALYSES 3.6.1 DATA PREPARATION AND DESCRIPTIVE STATISTICS Data were collected into two subject-level files. The first file contained one record per-individual (the \"overall file\"): each case or comparator had one set of visit counts summed over the years of follow-up for each of primary care, specialist and non-physician visits, and one COC and UPC score, calculated from all the years of data they contributed. Individuals were censored upon death, relapse (if relapse occurred later than five years post-original cancer diagnosis) or migration out of BC. 54 The overall file was used to complete the cross-sectional analyses. The second file contained multiple records per-individual, each record covering a two-year age period (the \"biannual file\"). For the survivors, the biannual periods begin at five years post last cancer diagnosis or at entry into MSP, which ever comes later 8 . For their comparators, the biannual periods begin at age five or at entry in MSP, which ever came later. Each survivor and comparator has one visit count for each of primary care, specialist, and non-physician visit, and one COC or UPC score for each two-year age interval. Demographic information, such as urban or rural classification, health authority and SES were reassessed at the beginning of each two year period to account for any changes in these variables over time. The biannual file was used to complete the longitudinal analyses. Prior to conducting any analysis, both files were examined for missing data, potential outliers, and data errors. Transformations of variables were performed as necessary to convert variables into the appropriate form for analysis. Descriptive statistics, including measures of central tendency and variation, were computed for all study variables. Cross tabulations of variables of interest were performed in order to identify both potential confounders and collinearity between variables that may have required consideration in subsequent analyses. Of particular concern was potential correlation between cancer diagnosis group and primary treatment modality. If these clinical covariates were found to be collinear, 8 If a survivor suffered a relapse or second cancer before the five-year mark, their entry into the cohort would be delayed until five years following that relapse or second cancer. 55 only the variable with the stronger association with the outcome of interest (visit count or continuity score) would be retained in models. Continuous variables were assessed for normality and transformations were completed if deemed necessary. 3.6.2 PHYSICIAN SERVICES UTILIZATION The objective of these analyses was to describe physician utilization among survivors of childhood cancer compared with general population sample, and to examine factors associated with survivors' use of physician services. Both cross- sectional and longitudinal methodologies were employed to assess the patterns of physician utilization. 3.6.2a Cross-Sectional Approach For the case-control cross-sectional approach, which utilized the overall file, the count and frequency of visits to primary care, specialist and non-physician providers were compared between survivors and comparators for the duration of follow-up (1986-2000). It was expected that visit counts would fit a unimodal, positively skewed distribution. Linear regressions performed using generalized linear models (GLM) using a negative binomial distribution were thus performed on the visit count for each of primary care, specialist, and non-physician visits, as using a Poisson model caused the counts to be severely overdispersed. Covariates were removed from the model via backwards selection according to strength of association with the outcome variable. Simple linear regression was deemed inappropriate for this work for two reasons: first, some variables have non- continuous distributions - treatment modality, for example - and continuous data are required for linear regression to be used appropriately; second, it is likely that 56 the effect of some of the predictor variables on utilization may not be linear in nature. Survivor-only GLMs (negative binomial distribution) were built using clinical and demographic variables to predict visit patterns. Covariates were removed from the model via backwards selection according to the review of the literature and variable correlations were computed. Variables with significant (p<0.05) associations with visit frequency or that were noted as theoretical confounders in the literature, were retained in the model. 3.6.2b Longitudinal Approach In the longitudinal approach, which used the biannual file, time was considered as in two-year age intervals. Two-year periods were chosen for comparability with other published work (K. C. Oeffinger et al., 2004), and for synergy with the provider continuity of care analyses. Generalized estimating equations (GEE) (Der & Everitt, 2006), (also using negative binomial distribution) which allow for repeated measurements from the same individual, were used to control for within-subject visit pattern correlations. An autoregressive correlation matrix was used. As with the cross-sectional analyses, both survivor-only and case- control log-linear models were prepared using backwards selection methods, and the effects of clinical and demographic covariates, and total follow-up time were examined. 3.6.3 PRIMARY CARE PROVIDER CONTINUITY OF CARE The objective of this analysis was to assess levels of primary care provider continuity of care among survivors of childhood cancer compared with a general 57 population cohort, and to examine factors associated with survivors' continuity scores. Unadjusted primary care utilization rates were converted into continuity of care scores using the UPC and COC Indices, as described in Section 2.5.1.b. The measures were correlated using Pearson Correlation Coefficients to ensure intra- measure consistency. To control for the effect of low utilization levels on these measures, calculations were restricted to those who had three or more visits over a two year period (R. J. Reid et al., 2003). A two-year interval was chosen since Reid et al. (2003) demonstrated that continuity scores measured over two-year periods outperform those measured in one-year intervals only. 3.6.3a Cross-Sectional Approach As with utilization levels, continuity was assessed using both cross-sectional and longitudinal methodologies. For the cross-sectional approach, which used the overall file, the effect of clinical and demographic covariates, as well as the effect of overall length of follow-up, on overall COC and UPC scores was modeled for survivors using GLM with a normal distrubtion. Case-control linear regression models were used to examine differences in continuity scores between survivors and their general population comparators. 3.6.3b Longitudinal Approach In the longitudinal approach, which used the biannual file, UPC and COC scores were calculated for two-year age intervals, as described in section 2.6.1. Consistent with methods used for examining utilization, continuity was assessed using GEE (with a normal distribution) to adjust for within-subject correlation (Der & Everitt, 2006). An autoregressive correlation matrix was also used. Both survivor- 58 only and survivor-comparator models were used. In the survivor cohort, continuity trends associated with increasing age, while controlling for time since diagnosis, were also assessed using GEE methods (Der & Everitt, 2006). 59 CHAPTER 4: DESCRIPTIVE STATISTICS AND UTILIZATION OF PHYSICIAN SERVICES The results from this study are divided into two chapters: this initial chapter provides descriptions of clinical and sociodemographic characteristics of the survivor and comparison populations; compares the crude and adjusted utilization patterns; and describes the results of case-control regression modeling for these same cohorts. The subsequent chapter presents crude and adjusted provider continuity of care scores for the survivor population; and describes the results of the continuity case-control regression modeling. 4.1 DESCRIPTIVE STATISTICS A summary of the application of the exclusion criteria is presented in table 4.1. 1374 (87%) of 1578 total cancer survivors diagnosed under age 20, between 1981 and 1995 in British Columbia (BC), were successfully linked to the MSP client registry. A total of 52(4%) of these survivors relapsed before five-year survival and were lost to the Medical Services Plan (MSP) before five-years following their relapse. These cases were therefore excluded, resulting in a final cohort of 1322. A 10:1 (13220) age and sex frequency-matched population comparison group was randomly selected from the MSP Client Registry. Among the survivor population 42 (3%) were censored during follow-up due to the occurrence of a relapse after the five-year survival point, 48 (3%) were censored due to death, and 153 (11%) due to exit from MSP records. Among comparators, 43 (0.31%) were censored due to death, and 3048 (22%) due to loss 60 to follow-up as indicated by exit from MSP registry records 9 . Mean time from entry into the cohort to censoring (due to death, relapse, or loss to follow-up) or December 31, 2000, whichever occurred first, was 6.38 \u00C2\u00B1 4.07 years for survivors and 9.18 \u00C2\u00B1 5.14 for comparators (p<0.0001). The excess follow-up time in the comparator population was a result of the difference in death rates between cases and comparators (3% and 0.3% respectively) and due to censoring of survivor utilization data upon the occurrence of a relapse (3%). Table 4.1 Application of Exclusion Criteria Exclusion Criteria Survivors Remaining No. % Diagnosed between 1981-1995 before age 20, survived at least five years and included in the BC Cancer 1587 100 Registry Successfully linked to MSP client registry 1347 87 Minimum of five-years follow-up after last cancer diagnosis 1322 83 Final Sample 1322 4.1.1 CLINICAL CHARACTERISTICS Diagnosis and treatment characteristics of the survivor population are summarized in Table 4.2. Consistent with published cancer statistics, leukemia (24%) was the most common ICCC diagnosis among the survivor population, followed by tumors of the central nervous system (CNS) (19%), and lymphoma (17%) (Canadian Cancer Society/National Cancer Institute of Canada, 2007). The majority (61%) of survivors were treated with some form of chemotherapy, either 9 The reasons for the difference in rate of censoring due to exit from MSP records between cases and controls are unknown. It is possible that controls and their families are more mobile, and are therefore more likely to move outside of British Columbia (BC). 61 alone or in combination with radiation or surgical treatments. Twenty-six percent of survivors were treated with radiation therapy, alone or in combination, and 25.12% were treated using some combination of chemotherapy and radiation. Treatment data was unavailable for 5.5% of the cohort.\" As expected, diagnosis and treatment categories were strongly correlated: a chi-square test produced a coefficient of 1819.83 (p<0.0001). Specific diagnosis/treatment combinations are summarized in Table 4.3. A close examination of the other clinical variables revealed possible collinearity between the period of diagnosis and attained age variables. Thus, only the variable with the stronger association with the outcomes of interest was included in the regression modeling. Among the survivor cohort, diagnosis was most common under the age of five years (34% of cases) and over age 15 (28% of cases). More cases were diagnosed overall in the under 10 group than in the over 10 age group (see Table 4.4). In terms of calendar period of diagnosis, the largest proportion of cases was diagnosed between 1991 and 1995 (37%), which corresponds to the slight increases in the incidence of childhood cancer observed over that same period (see Table 4.4) (Canadian Cancer Society/National Cancer Institute of Canada, 2007). 10 Treatment data were unavailable for individuals who received treatments outside of theBC Children's Hospital or BC Cancer Agency (BCCA). For example, excision of a localized melanoma could occur with a community surgeon, and this treatment would not be recorded in the registry. Treatment data was similarly unavailable for those who had missing charts. 62 Table 4.2 Survivor Diagnosis and Treatment Characteristics Survivors (n=1374) Characteristic No. % Diagnosis I (Leukemias) 314 23.75 II (Lymphomas) 228 17.25 III (CNS Cancers) 247 18.68 IV (Neuroblastomas) 52 3.93 V (Retinoblastomas) 79 5.98 VI (Renal Tumors) 33 2.50 VII (Hepatic Tumors) 71 5.37 VIII (Malignant Bone Tumors) 6 0.45 IX (Soft Tissue Sarcomas) 63 4.77 X (Germ Cell Tumors) 102 7.72 XI (Other Malignant Epithelial Neoplasms) 126 9.53 XII (Other Unspecified) 1 0.08 Total 1322 100 Treatment Chemotherapy only 276 20.88 Chemotherapy and radiation 221 16.72 Chemotherapy, radiation and surgery 111 8.40 Chemotherapy and surgery 197 14.90 Unknowna 73 5.52 No treatmentb 12 0.91 Radiation only 24 1.82 Radiation and surgery 103 7.79 Surgery only 305 23.07 Total 1322 100 a Corresponds to survivors who had a BCCR record, but who were not admitted to a BCCA centre for treatment. b Corresponds to survivors who were admitted to BCCA but not treated. 63 Table 4.3 Diagnosis-Treatment Combinations Diagnosis Treatment: Number(% of ICCC group) Total C C&R C,R&S C&S R R & S S NI NTa Leukemias 186 122 1 1 2 0 0 1 1 314 (59.24) (38.85) (0.32) (0.32) (0.64) (0.00) (0.00) (0.32) (0.32) Lymphomas 78 64 16 34 4 22 4 3 3 228 (34.36) (28.19) (7.05) (14.98) (1.76) (9.69) (1.76) (1.32) (1.32) CNS Cancers 4 0 32 4 13 57 119 12 6 247 (1.62) (0.00) (12.96) (1.62) (5.26) (23.08) (48.18) (4.86) (2.43) Neuroblastomas 2 1 7 11 1 2 28 0 0 52 (3.85) (1.92) (13.46) (21.15) (1.92) (3.85) (53.85) (0.00) (0.00) Retinoblastomas 1 14 15 19 0 4 24 2 0 79 (1.27) (17.72) (18.99) (24.05) (0.00) (5.06) (30.38) (2.53) (0.00) Renal Tumors 0 0 1 5 0 2 25 0 0 33 (0.00) (0.00) (3.03) (15.15) (0.00) (6.06) (75.76) (0.00) (0.00) Hepatic Tumors 0 0 28 40 0 0 3 0 0 71 (0.00) (0.00) (39.44) (56.34) (0.00) (0.00) (4.23) (0.00) (0.00) Malignant Bone 0 0 0 6 0 0 0 0 0 6 Tumors (0.00) (0.00) (0.00) (100.0) (0.00) (0.00) (0.00) (0.00) (0.00) Soft Tissue Sarcomas 1 16 4 34 0 1 4 2 1 63 (1.59) (25.40) (6.35) (53.97) (0.00) (1.59) (6.35) (3.17) (1.59) Germ Cell Tumors 4 3 5 43 3 7 33 3 1 102 (3.92) (2.94) (4.90) (42.16) (2.94) (6.86) (32.35) (2.94) (0.98) Other Malignant 0 1 2 0 1 8 65 39 0 126 Epithelial Neoplasms (0.00) (0.79) (1.59) (0.00) (0.79) (6.35) (51.59) (38.89) (0.00) Other Unspecified 0 0 0 0 0 0 0 1 0 1 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (100) (0.00) C - Chemotherapy; R - Radiation; S - Surgery; NI - No Information; NT- No treatment allo treatment refers to survivors who were treated outside the cancer agecy. For example, survivors of melanoma, whose tumors were excised at a fa ily physician's office. Table 4.4 Age at and Period of Diagnosis Characteristic Survivors (n=1322) No. % Age at Diagnosis <5 years 455 34.42 5 - <10 years 236 17.85 10 - <15 years 260 19.67 15 - <20 years 371 28.06 Total 1322 100 Period of Diagnosis 1981-1985 367 27.76 1986-1990 454 34.34 1991-1995 501 37.90 Total 1322 100 4.1.2 DEMOGRAPHIC CHARACTERISTICS Unadjusted case-control comparisons of demographic characteristics are presented in Table 4.5. Both the survivor and comparison cohorts were composed of 46% females and 54% males. Mean age at the start of follow-up was significantly greater for survivors than their comparators (14.42 and 11.73 years respectively, p<0.0001) due to the methodology selected for frequency matching. Although the cohorts were frequency matched for birth year, differences in the time of start of follow-up resulted from the necessary differences in selection criteria for survivors versus comparators. Survivor follow-up began five-years following last known cancer diagnosis or relapse, or upon entry into MSP, whichever was later; comparator follow-up began at age five-years or entry into MSP, whichever was later. 65 Table 4.5 Case-Control Demographic Characteristics Characteristic Survivors (n=1322) Comparators (n=13220) P-value No. % No. % Sex Males Females Total 711 611 1322 53.78 46.22 100 7110 6110 13220 53.78 46.22 100 Attained Age a <10 years 112 8.47 926 7.00 10 - <20 years 507 38.35 4808 36.67 20 - <30 years 523 39.56 5294 40.05 30 + years 180 13.62 2192 16.58 Total 1322 100 13220 100 0.0095 Mean Attained Age 20.93 21.53 <0.0001 Rural Status of Residenceb Rural 290 21.94 3359 25.41 Small Community 191 14.45 1532 11.59 Large Community 181 16.39 1379 10.43 Metropolitan 660 49.92 6856 51.86 Unknown 0 0.00 94 0.71 Total 1322 100 13220 100 <0.0001 Socioeconomic Status b 1- Lowest 237 17.93 2853 21.58 2 250 18.91 2763 20.70 3 278 21.03 2325 17.59 4 256 19.36 2290 17.32 5 - Highest 246 18.61 2325 17.59 Missing 55 4.16 691 5.23 Total 1322 100 13220 100 0.0003 Health Authority of Residence b Interior 224 16.94 2095 15.85 Fraser 448 33.89 3892 29.44 Vancouver Coastal 270 20.42 3271 24.74 Vancouver Island 239 18.08 2001 15.14 Northern 114 8.62 1185 8.96 Missing 27 2.04 776 5.87 Total 1322 100 13220 100 <0.0001 Mortality by December 31, 2000 Alive 1282 96.97 13186 99.74 Dead 40 3.03 34 0.26 Total 1322 100 13220 100 <0.0001 aDefined as age at censoring (from death, relapse, or exit from MSP record) or at December 31, 2000. bSES was missing for 55 survivors and 691 comparators, and rural residence status was missing for 94 comparators, due to missing postal costal or failed linkage using the StatsCan software. 66 Mean attained age, defined as age at censoring (from death, relapse, or exit from MSP record) or at December 31, 2000, whichever was earliest, was also significantly different between cases and comparators (20.93 versus 21.53 years respectively, p<0.0001). Although statistically significant, this small difference is likely of little clinical relevance, and has arisen due to higher death rates observed in the survivor population. Socioeconomic status (SES) groupings 11, defined using Statistics Canada algorithms as described in Chapter 3, were also significantly different between cases and comparators (p=0.0003). Specifically, a larger proportion of comparators than survivors fell into the lowest two SES brackets (42.28% versus 36.84% respectively), and more survivors than comparators fell into the upper three SES brackets (52.5% versus 59.0% respectively). Health Authority of residence and urban/rural status at the start of follow-up also varied significantly between survivors and comparators. More general population comparators than cases resided in rural communities with populations of less than 10,000 (25% versus 22% respectively). Additionally, survivors were more likely to live in small or large communities (31% versus 22% respectively), although survivors and the general population comparators were equally likely to reside in metropolitan areas (50% versus 52%). 11 SES groupings were derived using area-specific income data from the nearest census to diagnosis. 67 Although statistically significant, the clinical relevance of the differences in demographic characteristics between cases and comparators has yet to be examined. Nevertheless, sociodemographic variables, which could relate to access to and quality of care, were included as independent covariates in the regression exercise, controlling for any differences observed between survivors and their general population counterparts. 4.3 PHYSICIAN SERVICES UTILIZATION 4.3.1 SUMMARY STATISTICS 4.3.1.a Survivor Only Survivors incurred an average of 8.94 \u00C2\u00B1 9.53 physician visits per year. Of these, 4.82 \u00C2\u00B1 4.76 were to primary care physicians, 2.69 \u00C2\u00B1 5.68 were to specialists 12 , and 1.43 \u00C2\u00B1 3.25 were to billable non-physician providers. Mean visits to all provider types - with the exception of specialist visits - varied significantly between ICCC diagnosis and treatment modality groups (Tables 4.6 and 4.7 respectively). Survivors diagnosed with malignant epithelial neoplasms had the largest mean visits per year with 11.46 \u00C2\u00B1 11.07 total. Survivors diagnosed with CNS cancers had similarly elevated visit patterns, with 10.26 \u00C2\u00B1 9.76 total visits per year. The retinoblastoma and renal tumor survivors had the lowest annual mean visits at 6.26 \u00C2\u00B1 4.65 and 6.62 \u00C2\u00B1 6.16 visits per year respectively. The highest mean visits per year was observed in survivors who had chemotherapy, radiation and surgery; and radiation and surgery treatment with 12 As stated in Chapter 3, specialist visits do not include oncology visits, who are salaried and thus not included in the MSP claims records. 68 11.15 \u00C2\u00B1 11.88 and 12.03 \u00C2\u00B1 14.45 visits per year respectively. Survivors who were treated with these specific modalities visited a specialist more often than average as well, with 5.24 \u00C2\u00B1 10.03 and 4.05 \u00C2\u00B1 8.48 visits per year. Survivors who were not treated had the lowest average yearly visits at 5.29 \u00C2\u00B1 4.00. Table 4.6 Mean VisitsJYead\u00C2\u00B1 Standard Deviation) for Specific ICCC Classifications ICCC Diagnosis Primary Care Visits/Yeara Specialist Visits/Yeara Non- Physician Visits/Yeara Total Visits/Yeara All Survivors 4.82 (4.76) 2.69 (5.68) 1.43 \u00C2\u00B1 3.25 8.94 (9.53) I (Leukemias) 4.10 (3.80) 2.35 (3.89) 0.89 (2.21) 7.35 (6.67) II (Lymphomas) 5.28 (5.78) 2.24 (6.73) 1.65 (3.04) 9.17 (11.76) III (CNS Cancers) 5.02 (4.62) 3.43 (5.63) 1.81 (3.67) 10.26 (9.76) IV (Neuroblastomas) 4.25 (5.55) 4.17 (7.97) 0.86 (2.40) 9.29 (10.52) V (Retinoblastomas) 3.81 (2.89) 2.12 (2.44) 0.34 (0.84) 6.26 (4.65) VI (Renal Tumors) 4.02 (4.76) 1.96 (2.52) 0.67 (0.93) 6.65 (6.16) VII (Hepatic Tumors) 2.68 (2.14) 1.28 (1.25) 0.89 (1.39) 4.84 (3.83) VIII (Malignant Bone 4.91 (5.14) 3.82 (6.58) 1.05 (1.78) 9.77 (9.83) Tumors) IX (Soft Tissue 4.92 (4.25) 2.67 (4.01) 0.95 (1.54) 8.54 (7.27) Sarcomas) X (Germ Cell Tumors) 5.03 (5.05) 2.87 (9.49) 1.45 (2.94) 9.36 (11.8) XI (Other Malignant 6.18 (4.85) 2.17 (4.49) 3.11 (6.31) 11.46 (11.07) Epithelial Neoplasms) XII (Other 4.73 (N/A) 2.20 (N/A) 0 (N/A) 6.93 (N/A) Unspecified) b F-Statistic (P-Value) 2.37 (0.0067) 1.44 (0.148) 5.56 2.88 (0.001) (<0.0001) aRefers to physician visits divided by length of follow-up. bNo Standard deviations presented for sample of N=1. 69 Table 4.7 Mean Visits/Year (\u00C2\u00B1 Standard Deviation) for Specific Treatment Classifications Treatment Modality Primary Care Visits/Yeara Specialist Visits/Yeara Non- Physician Visits/Yeara Total Visits/Yeara Chemotherapy only 4.33 (4.89) 2.19 (4.15) 0.93 (2.27) 7.45 (7.75) Chemotherapy and radiation 4.81 (4.38) 2.68 (5.20) 1.20 (2.20) 8.69 (8.20) Chemotherapy, radiation and surgery 4.61 (2.38) 5.24 (10.03) 1.31 (2.43) 11.15 (11.88) Chemotherapy and surgery 4.77 (5.31) 2.76 (6.26) 1.12 (2.35) 8.65 (9.65) Unknown 6.09 (4.86) 1.53 (2.25) 1.83 (2.97) 9.46 (8.07) No treatment 3.00 (2.50) 1.77 (2.33) 0.53 (0.99) 5.29 (4.00) Radiation only 4.23 (3.45) 1.55 (1.07) 1.63 (2.79) 7.41 (6.42) Radiation and surgery 5.88 (5.37) 4.05 (8.48) 2.11 (4.25) 12.03 (14.45) Surgery only 4.84 (4.62) 2.11 (3.78) 2.00 (4.72) 8.95 (9.16) F-Statistic (P-value) 1.96 (0.048) 4.83 (<0.0001) 3.25 (0.0011) 3.35 (0.0008) aRefers to physician visits divided by length of follow-up. 4.3.1.b Case-Control The substantial difference in mean follow-up time between the survivors and the population sample rendered a simple comparison of mean physician visits over the course of follow-up meaningless. Instead, a comparison of mean visits divided over the period of follow-up (visits per year) was used. Table 4.8 Mean Visits/Year (\u00C2\u00B1 Standard Deviation) for Survivors versus Comparators Group Primary Care Specialist Non- Total Visits/Yeara Visits/Yeara Physician Visits/Yeara Visits/Yeara Survivors 4.82 (4.76) 2.69 (5.68) 1.43 (3.25) 8.94 (9.53) Comparators 1.98 (3.44) 0.43 (1.31) 0.54 (1.79) 2.98 (5.18) Difference 2.84 (3.58) 2.26 (2.13) 0.87 (1.97) 5.96 (5.71) T-Statisticb (P- 21.12 14.38 9.55 (<0.0001) 22.39 value) (<0.0001) (<0.0001) (<0.0001) aRefers to physician visits divided by length of follow-up. bSatterthwaite (non-pooled) statistic quoted to control for unequal variances between cases and comparators. 70 Survivors had significantly more visits per year than the general population group in all visit categories (Tables 4.8 and 4.9). While only 3.8% of survivors had an average of zero visits per year, a full 50% of the comparator group had no visits per year. Additionally, 29.19% of survivors had an average of 10 or more visits per year. Among this group, 28.50% had an average of 20 or more visits per year. Table 4.9 Grouped Mean Visits/Year for Survivors versus Comparators Groupings: Mean Visits/Year Survivors (n=1322) Comparators (n=13220) No. % No. ok <2 visits/year 166 12.56 8203 62.05 2 - <4 visits/year 227 17.17 1595 12.07 4 - <6 visits/year 232 17.55 1137 8.60 6 - <8 visits/year 186 14.07 733 5.54 8 - <10 visits/year 126 9.53 433 3.28 10 - <12 visits/year 100 7.56 313 2.37 12 - <14 visits/year 69 5.22 214 1.62 14 - <16 visits/year 49 3.71 168 1.27 16 - <18 visits/year 32 2.42 121 0.92 18 - <20 visits/year 25 1.89 70 0.53 20 + visits/year 110 8.32 233 1.76 Total 1322 100 13220 100 F-Statistic (P-value) 1224.35 (<.0001) 4.3.2 REGRESSION MODELING: OVERALL VISIT PATTERNS Negative binomial regression models, constructed using generalized estimating equations (GEE), were used to predict use of primary care, specialist, and non-physician providers among survivors of childhood cancer (labeled the \"overall models\"). The strength of an individual covariate, adjusting for the presence of other predictors, was tested as a whole (using chi square tests) and at individual levels (using relative risk and confidence interval calculations) in comparison with a specific reference population. For example, the overall significance of the impact of 71 the SES variable on the different visit types was tested. The differences in visit patterns between levels of the covariates - comparing visits per year between individuals with low versus high SES, for example - was also tested individually. In case-only and case-control models, the influence of clinical and sociodemographic factors on visits per year in each visit category are presented in the regression results in Tables 4.10 and 4.11 respectively. A detailed analysis of the GEE parameter estimates, and full regression results for case-control comparisons can be found in Appendix C, Tables 1-8. Aside from adjusted differences in visit patterns between cases and comparators, discussions regarding the predictive value of covariates in case-control models will be limited 13 . 4.3.2.a Survivors Only Primary Care Visits: Relapse status (p<0.0001), age at diagnosis (p<0.0001) and sex (p<0.0001) were significant predictors of primary care visits per year. Survivors who suffered a relapse either more- or less-than five years following the original diagnosis used more primary care services than survivors who never relapsed (RR 1.53, 95% CI 1.19-1.97; and RR 1.26, 95% CI 1.14-1.41 respectively). Survivors diagnosed at an older age were more likely to use primary care services (RR1.03, 95% CI 1.02-1.04): a survivor diagnosed at age five would, for example, have an excess of three primary care visits per year over a survivor diagnosed at age four. Female survivors also used more primary care services, using nearly 50% 13 Given the 10:1 ratio of controls to cases, the effect of specific covariates on utilization will primarily reflect the patterns seen in controls, which is not the focus of this work. Case-only models will explore the effect of these covariates on visit patterns as they related to survivors only. 72 more services than male survivors (RR 1.47, 95% CI 1.35-1.61). Interestingly, neither specific diagnosis (p=0.77) nor specific treatment (p=0.53) predicted primary care visits, once the effects of other covariates had been controlled. Table 4.10 Factors Influencing Visits per Year: Survivors Only i. Outcome: Total Visits ii. Outcome: Primary Care Visits Parameter Chi- Pr > Chi Parameter Chi- Pr > Chi Square Sq Square Sq ICCC Diagnosis 22.47 0.021 ICCC Diagnosis 7.31 0.7735 Group Group Treatment 25.26 0.0014 Treatment 7.11 0.5253 Category Category Relapse Status 49.58 <.0001 Relapse Status 28.59 <.0001 Sex 59.34 <.0001 Sex 73.74 <.0001 SES 11.58 0.0409 SES 10.62 0.0595 Health Authority 4.18 0.5243 Health 10.55 0.0611 Authority Ruralness 0.56 0.9051 Ruralness 6.13 0.1056 Age at Diagnosis 12.78 0.0004 Age at Diagnosis 34.41 <.0001 iii. Outcome: Specialist Visits iv. Outcome: Non-Physician Visits Parameter Chi- Pr > Chi Parameter Chi- Pr > Chi Square Sq Square Sq ICCC Diagnosis 73.43 <.0001 ICCC Diagnosis 26.66 0.0052 Group Group Treatment 68.48 <.0001 Treatment 0.42 0.8119 Category Category Relapse Status 43.6 <.0001 Relapse Status 12.5 0.1302 Sex 0.59 0.4432 Sex 13.27 0.0003 SES 25.55 0.0001 SES 11.89 0.0363 Health Authority 18.56 0.0023 Health 1.35 0.9298 Authority Ruralness 10.23 0.0167 Ruralness 2.69 0.4428 Age at Diagnosis 41.82 <.0001 Age at Diagnosis 13.59 0.0002 Primary Care 233.53 <.0001 Primary Care 80.81 <.0001 Visits per Year Visits per Year Although SES, health authority of residence, and rural status of residence were not strong predictors of primary care utilization, survivors living in one health 73 authority region - Fraser Health - used 20% more primary care services than those living in Vancouver Coastal Health (RR 1.20, 95% CI 1.06-1.36). Specialist Visits: Specific ICCC diagnosis (p<0.0001) and treatment classification (p<0.0001) were significant predictors of specialist visits per year. When compared to survivors of leukemia, survivors of bone, CNS, and germ cell tumors used more specialist physician services (RR 2.01, 95% CI 1.42-2.87; RR 1.57, 95% CI 1.16-2.12; and RR 1.42, 95% C11.01-1.99 respectively). Survivors of lymphoma used fewer specialist services (RR 0.80, 95% CI 0.63-1.00). Compared with survivors treated with surgery alone, individuals treated with chemotherapy only (RR 1.37, 95% CI 1.01-1.85), chemotherapy plus radiation (RR 1.69, 95% CI 1.27-2.26), chemotherapy plus surgery (RR 1.47, 95% CI 1.16- 1.88), radiation plus surgery (RR 1.51. 95% CI 1.16-1.96), and chemotherapy, radiation and surgery (RR 2.60 95% CI 2.00-3.38), all used significantly more specialist services annually. Individuals who were not treated used the fewest specialist services, but this effect was not significant (RR 0.90, 95% CI 0.46, 1.77). Similar to the results observed for primary care visits, relapse status (p<0.0001) and age at diagnosis (p<0.0001) were both independent predictors of use of specialist care services. Compared with survivors who never relapsed, individuals who relapsed before five years after the original diagnosis (RR 1.53, 95% CI 1.33-1.77) and those who relapsed more than five years post-diagnosis (RR 1.93, 95% CI 1.32-2.83) used significantly more specialist care services. Unlike the trend observed in primary care utilization however, survivors diagnosed earlier in 74 childhood used more specialist services than those diagnosed later (RR 0.96, 95% CI 0.95-0.97). In term of demographics, SES (p<0.0001), health authority (p=0.0023) and rural status (p=0.0167) were all significant predictors of the use of specialist services. Although individuals in the lowest two SES brackets used less specialist care than those in the middle and upper brackets, this trend was not significant. Individuals living in the region of Vancouver Coastal Health used the most specialist care. Survivors living in Fraser Health (RR 0.75, 95% CI 0.63-0.89), Interior Health (RR 0.69, 95% CI 0.54-0.88), Northern Health (RR 0.57, 95% CI 0.43-0.94) or Vancouver Island Health (RR 0.77, 95% CI 0.62-0.95) authorities all used significantly less specialist care. Survivors living in rural areas or small communities used fewer specialist services than individuals living in metropolitan areas (RR 0.78, 95% CI 0.64-0.96; and RR 0.71, 95% CI 0.58-0.89). Although it was a predictor of primary care utilization, sex was not a significant predictor of the use of specialist services (p=0.4432). Specialist care utilization increased relative to primary care utilization (p<0.0001): individuals who used more specialist care services also used more primary care services (RR 1.11, 95% CI 1.09-1.12), likely resulting from the \"gatekeeper\" function of primary care physicians in the Canadian health system. Non-Physician Visits: Specific ICCC diagnosis was a significant predictor of non-physician services utilization (p=0.0052). Specifically, survivors of retinoblastoma used significantly less non-physician care than survivors of 75 leukemia (RR 0.30, 95% CI 0.14-0.64). Survivors of bone, renal, and soft tissue tumors also used less non-physician care than survivors of leukemia; however, these differences were not significant. Treatment category was not a significant predictor of the use of non-physician services. Interestingly, though, survivors treated with surgery only used more non-physician services than any other treatment group, but this trend was not statistically significant. As seen in the primary care visit analysis, age at diagnosis was a significant predictor of use of non- physician services (p=0.0002): survivors who were diagnosed at an older age saw a non-physician practitioner more often than those who were diagnosed at a younger age (RR 1.04, 95% CI 1.02-1.06). Female survivors used non-physician services more frequently than male survivors (RR 1.40, 95% CI 1.17-1.68). SES was also a significant predictor of the use of non-physician services (p=0.036): individuals in lower SES brackets used these services less frequently than those in the higher SES brackets, although this was not statistically significant. Health authority of residence and rural status of residence status did not have a significant effect on non-physician services utilization. As observed with specialist utilization, the use of non-physician services increased relative to primary care utilization (p<0.0001); individuals who used more non-physician care services also used more primary care services (RR 1.1, 95% CI 1.08-1.13). 76 4.3.2.b Case-Control When demographic differences are adjusted for, survivors use significantly more primary care (p<0.0001), specialist (p<0.0001), and non-physician (p<0.0001) services than the general population sample (Table 4.11). Specifically, survivors see a primary care physician nearly three times as often as comparators (RR 2.94, 95% CI 2.59-3.32), and a non-physician provider twice as often (RR 2.02, 95% CI 1.77- 2.32). Even more striking, they see a specialist more than six times as often (RR 6.42, 95% CI 5.74-7.19). Table 4.11 Factors Influencing Visits per Year: Case-Control i. Outcome: Total Visits ii. Outcome: Primary Care Visits Parameter^Chi- Square Pr > Chi Sq Parameter Chi- Square Pr > Chi Sq Case-Control 568.77 <.0001 Case-Control 361.03 <.0001 Sex 138.4 <.0001 Sex 192.18 <.0001 SES 3.76 0.5842 SES 8.93 0.1118 Health 11.92 0.0359 Health 9.07 0.1062 Authority Authority Ruralness 22.47 <.0001 Ruralness 19.06 0.0003 Attained Age 521.88 <.0001 Attained Age 435 <.0001 iii. Outcome: Specialist Visits iv. Outcome: Non-Physician Visits Parameter Chi- Pr > Chi Parameter^Chi-^Pr > Chi Square Sq Square Sq Case-Control 1355.72 <.0001 Case-Control 119.02 <.0001 Sex 40.4 <.0001 Sex 9.23 0.0024 SES 14.99 0.0104 SES 25.54 0.0001 Health 77.91 <.0001 Health 40.86 <.0001 Authority Authority Ruralness 91.05 <.0001 Ruralness 10.28 0.0164 Attained Age 12.64 0.0004 Attained Age 349.47 <.0001 Primary Care 3659.72 <.0001 Primary Care 1986.77 <.0001 Visits per Year Visits per Year 77 As observed in the case-only modeling, sociodemographic characteristics such as sex (p<0.0001), health authority (p=0.036), and ruralness status (p<0.0001) affect utilization in all visit categories. Health authority and rural status of residence both have a particularly strong effect on specialist services utilization (see Appendix C, Table 7). Additionally, presumably due to the system \"gatekeeper\" role of primary care physicians in Canada 14, use of primary care services is a key predictor of both specialist (p<0.0001) and non-physician services (p<0.0001). 4.3.3 REGRESSION MODELING: BIANNUAL VISIT PATTERNS As opposed to examining all follow-up data available for each individual cross-sectionally, as was done in the above two sections, biannual visit patterns were examined here as a function of age: follow-up was broken into two-year age intervals with the intention of looking at age-related utilization trends, and the differences in age-related trends between survivors and the general population sample. Sociodemographic variables, including SES, rural status, and health authority of residence, were also re-calculated for each of the two-year periods. GEE modeling was used to appropriately adjust for repeated measurements on each subject. In case-only and case-control models, the influence of clinical and sociodemographic factors on visit patterns in each visit category are presented in the biannual-by-age (BBA) regression results in Tables 4.12 and 4.13 respectively. A 14 The population can only access specialist and billable non-physician care via a referral from a primary care provider. 78 detailed analysis of the GEE parameter estimates can be found in Appendix C, Tables 9-16. 4.3.3.a Survivors Only Primary Care Visits: Primary care visits increased linearly with the two-year age intervals (p=0.054): for each two-year increase in age, primary care visits increased by 1% (RR 1.01, 95% CI 1.00-1.02). Although statistically significant, the clinical significance of a 1% biannual increase is arguable. As expected, the factors found to be influential in terms of overall visit patterns were similarly influential for biannual age-related patterns. Some interesting differences did emerge, however. Although it was highly significant in models presented in section 4.3.2., stratification by two-year age intervals reduced the influence of age at diagnosis on primary care visits (p=0.14), suggesting possible confounding. Additionally, ICCC diagnosis and specific treatment group were not significant predictors of primary care visits in the BBA model, however, GEE parameter estimates demonstrated that individuals treated with the most complex treatment - chemotherapy, radiation, and surgery - had 24% more primary care visits in each two-year age interval than individuals treated with surgery alone (RR 1.24, 95% CI 1.02-1.52). Specialist Visits: Opposite to the trend observed for primary care visits, specialist visits decreased linearly with the two-year age intervals (p=0.0002): for each two-year increase in age, specialist visits decreased by 4% (RR 0.96, 95% CI 0.93-0.98). As observed in the primary care models, the significance of the 79 covariates in terms of specialist visits varied little between the BBA model and the overall model. Again, the inclusion of the intervals of ages eliminated the significance of the age at diagnosis criteria (p=0.55). Table 4.12 Factors Influencing BBA Visit Pattern: Survivors Only i. Outcome: Total Visits ii. Outcome: Primary Care Visits Parameter Chi- Pr > Chi Parameter Chi- Pr > Chi Square Sq Square Sq ICCC Diagnosis 18.41 0.0726 ICCC Diagnosis 7.97 0.7164 Group Group Treatment 11.87 0.1573 Treatment 7.63 0.4701 Category Category Sex 42.72 <.0001 Sex 50.34 <.0001 SES 2.17 0.8256 SES 4.1 0.5351 Health Authority 2.92 0.7116 Health Authority 4.04 0.544 Ruralness 0.55 0.907 Ruralness 2.45 0.4836 Age at Diagnosis 1.48 0.2231 Age at Diagnosis 2.23 0.1355 Age at Start of 2- 2.32 0.1273 Age at Start of 2- 3.7 0.0544 Year Inverval Year Inverval Relapse Status 12.1 0.0024 Relapse Status 8.15 0.017 iii. Outcome: Specialist Visits iv. Outcome: Non-Physician Visits Parameter Chi- Pr > Chi Parameter Chi- Pr > Chi Square Sq Square Sq ICCC Diagnosis 32.68 0.0006 ICCC Diagnosis 13.07 0.2199 Group Group Treatment 18.1 0.0205 Treatment 9.83 0.2772 Category Category Relapse Status 16.97 0.0002 Relapse Status 3.86 0.1454 Primary Care 35.8 <.0001 Primary Care 6.08 0.0137 Visits During 2- year Interval Visits During 2- year Interval Sex 2.74 0.0977 Sex 8.82 0.003 SES 3.54 0.6171 SES 9.02 0.1082 Health Authority 11.65 0.0399 Health Authority 7.09 0.214 Ruralness 3.22 0.3585 Ruralness 1.59 0.6628 Age at Diagnosis 0.35 0.5523 Age at Diagnosis 0.44 0.5075 Age at Start of 2- 13.8 0.0002 Age at Start of 2- 11.16 0.0008 Year Interval Year Interval 80 Additionally, there were slight differences observed in the significance of individual ICCC diagnoses 15 , but almost no change to the relative risk values. The small changes in significance were therefore likely caused by structural changes in the model, and do not reflect real differences in the predictive weight of the criteria. Interestingly, unlike in the overall model, rural status of residence was not a significant predictor of visit patters in the BBA model (p=0.3585). Although individuals living in non-metropolitan areas - large communities (RR 0.86, 95% CI 0.65-1.13), small communities (RR 0.88, 95% CI 0.65-1.19) and rural areas (RR 0.80, 95% CI 0.62-1.02) - used less specialist care than those living in metropolitan centres, this trend was no longer significant. This change could reflect a more relevant measurement of residential status, since any moves would be accounted for in the reassessment of this variable at the beginning of each two-year interval. Non-Physician Visits: Non-physician visits increased linearly with age (p=0.0008): for each two-year increase in age, non-physician visits increased by 6% (RR 1.06, 95% CI 1.03-1.10). As observed in both the primary care and specialist visit models, the inclusion of the age trend in the model eliminated the predictive value of the age at diagnosis criteria (p=0.51), further supporting the confounding relationship between age at interval and age at diagnosis. 15 Those with bone cell tumors still used more specialist services than those with leukemias; however, this trend was no longer significant (RR 1.40, 95% CI 0.86-0.228). Similarly, lymphoma survivors still used less specialist care than leukemia survivors; however, this trend was also no longer significant (RR 0.89, 95% CI 0.61-1.31). 81 Unlike in the overall model, ICCC diagnosis was not a significant predictor of non-physician visits in the BBA model (p=0.22); however, the GEE factor estimates are remarkably similar between the two (Appendix C, Table 12). The lack of significance may, therefore, be a statistical function of the model structure. The predictive significance of the other clinical and sociodemographic characteristics was consistent between the overall and BBA models. Examining visits in terms of two-year attained age intervals reduced the apparent influence of age at diagnosis on patterns of visits for all visit types. This may indicate that the attained (meaning current) age may be a more important predictor of physician services utilization than the age a survivor was initially diagnosed. This result was explored further in the case control modeling exercise below in section 4.3.3.b. Interaction terms assessing the relationship between the age trend, ICCC diagnosis, and treatment categories were not significant predictors of visit patterns for any visit type. 4.3.3.b Case-Control Consistent with patterns observed in the overall regression, when demographic differences are adjusted for, survivors use significantly more primary care (p<0.0001), specialist (p<0.0001), and non-physician (p<0.0001) services in each two year age interval than the general population sample (Table 4.13). Survivors see a primary care physician over four times as often as comparators (RR 4.39, 95% CI 3.74-5.15), and a non-physician provider three times as often (RR 3.20, 95% CI 2.17-4.74). The largest difference is seen in the use of specialist care 82 services, where survivors see a specialist twenty times as often as comparators (RR 20.03, 95% CI 15.34-26.15). Figure 4.1 Primary Care Visits Per Year by Age, Grouped by Age at Diagnosis 16 Primary Care Visits Per Year by Age, Grouped by Age at Diagnosis 16 14 12 10 0 8 6 Survivors, All Dx Years ---s\u00E2\u0080\u0094 Survivors, Dx Age 0-4 Survivors, Dx Age 5-9 Survivors, Dx Age 10-14 ---*--- Survivors, Dx Age 15-19 Comparison Group 4 2 0 cb 0 \"1,,^tx^to 4) 0^ .^bc^cb O^tx10 10 `1, rb Age In all BBA case-control models, an interaction between case-control status and age at the start of each two year interval was present, indicating that the trend in visit patterns for primary care (p<0.0001), specialist (p<0.0001) and non- physician (p=0.0032) visits by age is different between survivors and general population comparators. For BBA primary care visits, the increase in visits per two- year age interval is 3% smaller in survivors than comparators (RR 0.97, 95% CI 0.96-0.98). However, additional analyses revealed that in the survivor group, the 16Figures 4.1, 4.1 and 5.1 show hypothetical survivor populations created using the regression models. Each point represents visits for individuals included in the reference categories (see Appendices C and D) for all variables except age at diagnosis. 83 Survivors, All Dx Years --0-- Survivors, Dx Age 0-4 Survivors, Dx Age 5-9 Survivors, Dx Age 10-14 - Survivors, Dx Age 15-19 - Comparison Cohort 0)^1, N,^11, Age c ^(0`), \u00E2\u0080\u00A2^ `1, 7 6 5 0 4-, 4 ch ch= 3 a. (Is cip 2 1 0 r\u00CC\u0082 age effect is modified by the age a survivor was diagnosed (Figure 4.1). Survivors diagnosed in the oldest age category (15-19 years) demonstrated no increase in primary care utilization as they age. Those diagnosed between ages five and nine, however, showed the most dramatic increase in utilization of any survivor group. Figure 4.2 Specialist Visits Per Year by Age, Grouped by Age at Diagnosis Specialist Visits Per Year by Age, Grouped by Age at Diagnosis A difference in the observed age trend is also present for both non-physician and specialist services utilization. For non-physician visits, the increase in visits per two-year age interval is also 3% smaller in survivors (RR 0.97, 95% CI 0.95-0.99). The difference in trends is most striking for specialist visits. The case BBA model showed that for each two year age interval, specialist visits decrease by 4% among cases (RR 0.96, 95% CI 0.93-0.98). Among comparators, however, the number of 84 i. Outcome: Total Visits ii. Outcome: Primary Care Visits specialist visits increases for each two-year period at a rate of 1.9% per interval (RR 1.091, 95% CI 1.01-1.03). These trends are demonstrated in Figure 4.2. Table 4.13 Factors Influencing BBA Visit Pattern: Case-Control Parameter Case-Control Sex SES Health Authority Ruralness Age at Start of 2-Year Interval Interaction Term: Age*Case Chi-^Pr > Chi Square^Sq 155.72^<.0001 225.81^<.0001 3.95^0.5563 18.83^0.0021 3.41^0.3324 98.07^<.0001 76.15^<.0001 Parameter Case-Control Sex SES Health Authority Ruralness Age at Start of 2-Year Interval Interaction Term: Age*Case Chi-^Pr > Chi Square^Sq 149.85^<.0001 285.84^<.0001 4.11^0.5341 7.21^0.2056 0.26^0.9671 81.99^<.0001 52.99^<.0001 iii. Outcome: Specialist Visits Parameter^Chi-^Pr > Chi Square^Sq Case-Control^120.78^<.0001 Primary Care^255.92^<.0001 Visits During 2-year Interval Sex^2.28^0.1309 SES 9.43^0.0932 Health^31.28^<.0001 Authority Ruralness^63.62^<.0001 Age at Start of^23.94^<.0001 2-Year Interval Interaction^77.62^<.0001 Term: Age*Case iv. Outcome: Non-Physician Visits Parameter^Chi-^Pr > Chi Square^Sq Case-Control^18.24^<.0001 Primary Care^203.96^<.0001 Visits During 2- year Interval Sex^12.2^0.0005 SES 7.79^0.1684 Health^26.46^<.0001 Authority Ruralness^3.03^0.3874 Age at Start of^58.99^<.0001 2-Year Interval Interaction^8.7^0.0032 Term: Age*Case The difference in rate of change between cases and comparators is nearly 8% (RR 0.92, 95% CI 0.92-0.94). Interestingly, as observed with primary care utilization, the age trend is modified in the survivor cohort by the age of diagnosis. 85 Those diagnosed between ages five and nine, who also had the most dramatic increase in primary care utilization, showed no decline in their utilization of specialist services as they aged, opposing the overall trend. As observed in the case-only modeling, use of primary care services is a key predictor of both specialist (p<0.0001) and non-physician services (p<0.0001) in the case-control models. 86 CHAPTER 5: PROVIDER CONTINUITY OF CARE 5.1 SUMMARY STATISTICS The mean primary care provider Continuity of Care (COC) score for survivors, calculated using all years of available data was 0.35 \u00C2\u00B1 0.25. Mean Usual Provider of Care (UPC) score was 0.54 \u00C2\u00B1 0.22. A UPC score of 0.54 indicates that survivors saw their usual family physician for 54% of their overall primary care visits. The interpretation of the COC score is more complicated, however. It is substantially lower than the observed UPC because it adjusts for individuals who have an overall low number of visits (R. Reid et al., 2003). Both the COC and UPC variables were normally distributed. Table 5.1 Continuity Scores (\u00C2\u00B1 Standard Deviation) for Specific ICCC Classifications ICCC Diagnosis n^COC^UPC All Survivors 1155 0.35 (0.25) 0.54 (0.22) I (Leukemias) 274 0.32 (0.24) 0.52 (0.22) II (Lymphomas) 201 0.34 (0.25) 0.54 (0.22) III (CNS Cancers) 221 0.36 (0.25) 0.55 (0.22) IV (Neuroblastomas) 43 0.31 (0.21) 0.52 (0.20) V (Retinoblastomas) 68 0.36 (0.28) 0.54 (0.25) VI (Renal Tumors) 28 0.34 (0.24) 0.53 (0.23) VII (Hepatic Tumors) 63 0.55 (0.26) 0.73 (0.18) VIII (Malignant Bone Tumors) 4 0.34 (0.28) 0.51 (0.25) IX (Soft Tissue Sarcomas) 52 0.35 (0.25) 0.55 (0.23) X (Germ Cell Tumors) 88 0.39 (0.26) 0.57 (0.23) XI (Other Malignant Epithelial 112 0.38 (0.25) 0.55 (0.23) Neoplasms) XII (Other Unspecified) 1 0.18 (N/A) 0.32 (N/A) F-Statistic (P-Value) 1155a 0.99 (0.4503) 0.86 (0.5706) aRecall that only those with three or more visits over the study period are counted for COC or UPC scores, reducing_the sample from 1322 to 1155. 87 Unlike the trend observed in all visit types, neither COC nor UPC scores varied between diagnosis and treatment groupings (Tables 5.1 and 5.2). It is interesting to note, however, that survivors of hepatic tumors had COC and UPC scores that seem substantially - though not statistically significantly - higher than those of other diagnosis groups. Similarly, individuals who did not receive treatment, and therefore face no long-term morbidity arising from cancer therapy, had higher COC and UPC scores than survivors in other treatment groups. Table 5.2 Continuity Scores (\u00C2\u00B1 Standard Deviation) for Specific Treatment Classifications Treatment Modality n COC UPC Chemotherapy only 230 0.37 (0.27) 0.56 (0.22) Chemotherapy and radiation 205 0.31 (0.22) 0.50 (0.21) Chemotherapy, radiation and surgery 98 0.37 (0.24) 0.57 (0.21) Chemotherapy and surgery 162 0.35 (0.25) 0.55 (0.23) Unknown 69 0.34 (0.25) 0.52 (0.23) No treatment 9 0.43 (0.31) 0.62 (0.26) Radiation only 23 0.38 (0.25) 0.56 (0.23) Radiation and surgery 98 0.38 (0.24) 0.56 (0.22) Surgery only 261 0.34 (0.25) 0.52 (0.23) F-Statistic (P-value) 1155 1.32 (0.20) 1.70 (0.093) aRecall that only those with three or more visits over the study period are counted for COC or UPC scores, reducing the sample from 1322 to 1155. 5.2 REGRESSION MODELING 5.2.1 OVERALL COC AND UPC SCORES Linear regression models, constructed using generalized estimating equations (GEE), were used to predict COC and UPC scores, calculated using all primary care visits among survivors of childhood cancer (labeled the \"overall models\"). Comparable to methods used to predict visit patterns, the strength of an individual covariate, adjusting for the presence of other predictors, was tested as a whole (using chi square tests) and at individual levels (using difference in means 88 (DIM) and confidence interval calculations) in comparison with a specific reference population. In case-only and case-control models, the influence of clinical and sociodemographic factors on COC and UPC scores are presented in the regression results in Tables 5.3 and 5.4 respectively. A detailed analysis of the GEE parameter estimates, can be found in Appendix D, Tables 1-4. 5.2.1.a Survivors-Only COC: As found with the unadjusted F-statistics presented in section 5.1, COC score did not vary between ICCC diagnosis groups. Analysis of the individual GEE parameter estimates, however, shows survivors of epithelial (DIM 0.096, 95% CI 0.01-0.18), germ cell (DIM 0.090, 95% CI 0.01-0.17) and hepatic tumors (DIM 0.25, 95% CI 0.003-0.49) all have higher COC scores than survivors of leukemia. Treatment category was found to be a significant predictor of COC score (p=0.04). Specifically, survivors treated with chemotherapy only (DIM=0.08, 95% CI 0.009-0.15), chemotherapy, radiation and surgery (DIM 0.06, 95% CI 0.003-0.13), and radiation and surgery (DIM 0.06, 95% CI 0.006-0.12), had significantly higher COC scores than individuals treated with surgery only. Other clinical factors such as age at diagnosis (p=0.51) and relapse status (p=0.59) were not significant predictors of COC score. 89 Table 5.3 Factors Influencing COC and UPC Scores: Survivors Only i. Outcome: COC ii. Outcome: UPC Parameter Chi- Square Pr > Chi Sq Parameter Chi- Square Pr > Chi Sq ICCC Diagnosis 11.22 0.4252 ICCC Diagnosis 12.45 0.3307 Group Group Treatment 16.06 0.0415 Treatment 19.85 0.0109 Category Category Sex 0.29 0.5874 Sex 0.99 0.3203 SES 5.34 0.3757 SES 3.93 0.5593 Health 18.65 0.0022 Health Authority 13.19 0.0217 Authority Ruralness 6.81 0.0781 Ruralness 4.07 0.2536 Age at Diagnosis 0.43 0.511 Age at Diagnosis 0.57 0.4496 Length of 26.7 <.0001 Length of 44.07 <.0001 Follow-up Follow-up Relapse Status 0.29 0.5874 Relapse Status 0.25 0.8838 Health authority of residence was interestingly the only demographic factor found to be predictive of COC score (p=0.0022), while sex (p=0.59), socioeconomic status (SES) (p=0.38), and rural residential status (p=0.078) were not predictive. Individuals living in the jurisdiction of Vancouver Coastal or Vancouver Island Health Authorities had the highest COC scores, while survivors living in Fraser (DIM -0.063, 95% CI -0.10- -0.023) , Interior (DIM -0.057, 95% CI -0.11-0.0009) or Northern (DIM -0.042, 95% CI -0.11-0.025) Health Authorities all had lower continuity. Length of follow-up was also a significant predictor of the COC score: each additional year of follow-up was associated with a 0.01 point decrease in COC score (p<0.0001). UPC: Results for the UPC score were consistent with those for COC score: treatment category (p=0.011), health authority of residence (p=0.0217), and length of follow-up (p<0.0001) were all significant predictors of continuity. 90 The inverse relationship between length of follow-up and continuity observed for both the COC and UPC variables is clearly counter-intuitive: a longer relationship with a physician should result in better continuity. It is likely that this finding is an artifact of the measurement tool and the specific research methodology. UPC (and to a lesser extent COC) scores are prone to producing spuriously high scores for low, but consistent users (R. Reid et al., 2003). It is probable that individuals with short follow-up time have few primary care visits overall, but that these visits are to a consistent core provider. Similarly, individuals with longer follow-up may have visited more different primary care providers over the course of several years, lowering their continuity scores. This could result in the observed inverse relationship between COC and UPC scores and length of follow-up. 5.2.1.b Case-Control When sociodemographic differences are adjusted for, survivors have significantly higher COC (p=0.006) and UPC (p=0.016) scores than the comparator population sample. Specifically, survivors' COC scores are, on average, 0.02 points higher than their comparator counterparts (DIM 0.0214, 95% CI 0.0061-0.037). Their UPC scores are also 0.017 point higher (DIM 0.017, 95% CI 0.0032-0.031). Sociodemographic factors - health authority (COC: p<0.0001; UPC: p<0.0001), and rural residential status (COC: p=0.0001; UPC: p=0.0001) - also have a significant effect on COC and UPC scores among the general population comparators. For further detail, please see Appendix D, Tables 3 and 4. Sex and SES do not appear to predict either continuity measure in survivor or comparator populations. 91 Table 5.4 Factors Influencing COC and UPC Scores: Case-Control i. Outcome: COC^ ii. Outcome: UPC Parameter^Chi-^Pr > Chi^Parameter^Chi-^Pr > Chi Square^Sq Square^Sq Case-Control^7.54^0.006^Case-Control^5.78^0.0162 Sex^1.62^0.2029^Sex^0.39^0.53 SES 4.58^0.4694^SES 4.18^0.5235 Health^68.14^<.0001^Health^55.27^<.0001 Authority Authority Ruralness^21.04^0.0001^Ruralness^20.89^0.0001 Length of 160.9^<.0001^Length of 315.13^<.0001 Follow-Up ^Follow-Up 5.2.2 BIANNUAL BY AGE COC AND UPC SCORES Consistent with analyses produced to assess patterns of physician utilization, COC and UPC scores were examined as a function of age by conducting a repeated measures analysis on two-year age-intervals. In case-only and case-control models, the influence of clinical and sociodemographic factors on COC and UPC scores are presented in the biannual by age (BBA) regression results in Tables 5.5 and 5.6 respectively. A detailed analysis of the GEE parameter estimates, can be found in Appendix D, Tables 4-8. 5.2.2.a Survivor-Only COC: BBA regression modeling produced results consistent with those produced for the overall models: ICCC diagnosis (p=0.093) is not a significant predictor of COC; however, survivors of germ cell (DIM 0.072, 95% CI -0.0036-0.15) and hepatic tumors (DIM 0.26, 95% CI 0.013-0.52) still have higher COC scores than survivors of leukemia. Treatment category was found to be a significant predictor of COC score (p=0.019). Survivors treated with combination therapy - chemotherapy, radiation and surgery (DIM 0.11, 95% CI 0.041-0.16) - as well as those treated with 92 radiation alone (DIM 0.17, 95% CI 0.05-0.29), or radiation and surgery (DIM 0.078, 95% CI 0.023-0.13) had the highest COC scores. Sociodemographic factors - SES (p=0.35) and sex (p=0.97) - did not have a significant impact on continuity. As observed in the overall models, health authority of residence (at the start of each two year period, instead of at the start of overall follow-up) was a significant predictor of COC score (p<0.0001), with individuals living outside Vancouver Coastal Health (at the start of each two year period) all having worse continuity scores. Specifically, individuals living in the Fraser (DIM - 0.08, 95% CI -0.12- -0.044) and Interior (DIM -0.080, 95% CI -0.13- -0.027) Health Authorities were at the most risk of low COC scores. When assessed biannually, rural status of residence was borderline significant (p=0.054). Individuals living outside metropolitan centres had worse COC scores; however, this trend was not significant. Age at diagnosis, which was not a significant predictor of COC in the overall model, is highly significant in the BBA model (p=0.0044): individuals diagnosed at an older age tended to have higher COC scores than those diagnosed younger (DIM 0.0072, 95% C10.0023-0.012). An age-related trend also emerged: COC score was inversely related to age (p=0.023). Older survivors have worse COC scores than younger ones (DIM -0.0049, 95% CI -0.0091- -0.0007). UPC: As observed in the overall regression models, BBA regression for UPC score produced comparable results to COC. ICCC diagnosis was not a significant predictor of UPC score (p=0.11); however, some specific ICCC groups had 93 Table 5.5 Factors Influencing BBA COC i. Outcome: COC Parameter ICCC Diagnosis Group Treatment Category Sex SES Health Authority Ruralness Age at Diagnosis Age at Start of 2- Year Interval Primary Care Visits During 2- yr Interval Relapse Status Chi-^Pr > Chi Square^Sq 17.53^0.0933 18.34^0.0188 0^0.9698 5.56^0.3511 27.51^<.0001 7.63^0.0544 8.12^0.0044 5.21^0.0225 0.32^0.5698 5.25^0.0726 significantly higher scores than others 17 . Treatment (p=0.013) and age at diagnosis (0.0041) did have a significant impact on UPC score. and UPC Scores: Case-Only ii. Outcome: UPC Parameter^Chi- Pr > Ch ^ Square iSq ICCC Diagnosis^16.94^0.1095 Group Treatment^19.29^0.0134 Category Sex^0.56^0.4549 SES 4.22^0.5175 Health^23.21^0.0003 Authority Ruralness^7.66^0.0537 Age at 8.22^0.0041 Diagnosis Age at Start of^3.73^0.0534 2-Year Interval Primary Care^1.73^0.1881 Visits During 2-yr Interval Relapse Status^4.34^0.1142 Sociodemographic variables Sex and SES did not impact UPC score, while health authority and rural residential status were significant predictors. The same age-related trend observed in for COC appeared in the UPC model (p=0.053): as survivors aged, their UPC scores worsened (DIM -0.0032, 95% CI -0.0065-0). 17 Germ cell (DIM 0.060, 95% CI 0.0019-0.12) and hepatic (DIM 0.19, 95% CI -000033-0.37) tumor survivors had higher UPC scores than leukemia survivors. 94 The use of BBA models explicated two interesting age-related trends in both COC and UPC models. First, individuals diagnosed younger tended to have worse continuity. Second, as survivors age, their continuity scores tend to worsen. 5.2.2.b Case-Control Using a BBA approach, survivors do not have significantly better or worse COC (p=0.94) or UPC (p=0.95) scores than general population comparators. This runs counter to the findings from the overall model. Running simple descriptive statistics revealed differences in the age distribution of cases and the comparison group. Given that both COC and UPC scores change based on age, it is likely that the difference observed between survivors and comparators in the overall models was an artifact of the difference in age distribution between the groups. The BBA approach controls for the age trend. A linear trend by age is apparent for both COC (p=0.0042) and UPC (p=0.023) scores: scores increase with age. Older individuals tend to have better scores (COC: DIM 0.0012, 95% CI 0.0004-0.002; UPC: 0.0007, 95% CI 0.0001-0.0014). Interestingly, this trend is opposite to the inverse relationship between age and COC or UPC scores observed in the survivor population, suggesting that the relationship between age and continuity may be different between survivors and the general population sample. Similar to patterns observed in the visit analysis, the age observed in the survivor cohort modified by the age a survivor was diagnosed (Figure 5.1). A dramatic drop in continuity with age is observed amongst those diagnosed between 95 ages 10 and 14. The trend observed amongst those diagnosed between ages 15 and 19, however, is the opposite: continuity appears to increase with age, at the same relative pace as observed in the comparison cohort. Figure 5.3 Continuity of Care Trends by Age, Grouped by Age at Diagnosis Continuity Scores by Age, Grouped by Age at Diagnosis 0.7 0.6 :12 0.5 cn u 0.4 \u00E2\u0080\u0094\u00E2\u0080\u00A2\u00E2\u0080\u0094 Survivors, All Dx Years \u00E2\u0080\u0094\u00E2\u0080\u00A2\u00E2\u0080\u0094 Survivors, Dx Age 0-4 Survivors, Dx Age 5-9 Survivors, Dx Age 10-14 Survivors, Dx Age 15-19 \u00E2\u0080\u0094\u00E2\u0080\u00A2\u00E2\u0080\u0094 Comparison Group 0.3 0.2 (o^0)^'1,^<')^h^C) Age Sociodemographic factors including sex (COC: p<0.0001; UPC: p=0.0274), SES (COC: p=0.0039; UPC: p=0.027), health authority of residence (COC: p<0.0001; UPC: p=<0.0001), and rural residential status (COC: p=0.0062; UPC: p=0.015) all have a significant effect on both COC and UPC scores in a predominantly comparator-based population. Individuals living in Vancouver Coastal Health and/or in a major metropolitan area have significantly better scores that those living in other health authorities or outside a metropolitan area (see Appendix D, Tables 7-8). Interestingly, sex and SES were not found to be significant predictors of 96 continuity in the survivor-only model, suggesting that the effects of these variables on COC and UPC may be different between survivors and the general population. Table 5.6 Factors Influencing BBA COC and UPC Scores: Case-Control i. Outcome: COC^ ii. Outcome: UPC Parameter Case-Control Sex SES Health Authority Ruralness Age at Start of 2- Year Interval Primary Care Visits During 2-yr Interval Chi-^Pr > Chi^Parameter^Chi-^Pr > Chi Square^Sq Square^Sq 0.01^0.9381^Case-Control^0^0.9533 16.44^<.0001^Sex^7.82^0.0052 17.34^0.0039^SES 12.61^0.0274 97^<.0001^Health^85.9^<.0001 Authority 12.37^0.0062^Ruralness^10.52^0.0146 8.19^0.0042^Age at Start^5.19^0.0227 of 2-Year Interval 1.73^0.1884^Primary Care^20.2^<.0001 Visits During ^2-yr Interval The number of primary care visits during each two year interval was found to be a significant predictor of the UPC (p<0.0001), but not COC (p=0.19) score. This relationship demonstrates the statistical weakness of the UPC measurement: individuals with fewer visits had higher scores than those with more visits (DIM - 0.001, 95% CI -0.0014- -0.0005). 97 CHAPTER 6: DISCUSSION AND CONCLUSIONS The final chapter will be divided into (1) the significance of the project and research area; (2) a brief summary of the research presented in this thesis; (3) consideration of the implications of the results, with reference back to the original literature review; (4) strengths and limitations associated with the chosen methodology; (5) direction for future work in this area; and finally (6) concluding remarks. 6.1 SIGNIFICANCE The primary aim of this analysis was to describe the extent and patterns of physician services utilization and primary care provider continuity of care in a cohort of cancer survivors compared to a population-based comparison group. Given the long-term health risks associated with cancer diagnosis and treatment in childhood, and given the role of need in predicting utilization of health services (Andersen, 1995), it was hypothesized that survivors would use more health services than their counterparts in the general population; however, this assumption had not been well tested. And despite the numerous recommendations for continuous, coordinated, long-term follow-up, there is very limited information on how much long-term follow-up cancer survivors are currently receiving. Additionally, there is an increasing requirement to understand service utilization to support the development of coordinated follow-up programs to meet surveillance and other needs. This project, part of a larger program of research, has taken the necessary first steps to fill this knowledge gap. 98 This is the first and only study to formally assess provider continuity of care in a population of childhood cancer survivors using standard metrics. Given that continuity has an inverse relationship with the number of visits (a limitation of concentration of care metrics), it was hypothesized that increased need for care in the survivor group would result in lower continuity scores than those observed in the comparator group. Continuity is noted as a key component of optimum follow-up care for cancer survivors (K. C. Oeffinger, 2003), and continuity has the potential to enhance screening, surveillance and prevention efforts targeted to known or suspected cancer or treatment effects, by knowledge of a patient's medical history and preferences (D. A. Christakis, Feudtner, Pihoker, & Connell, 2001; D. A. Christakis, Mell, Wright, Davis, & Connell, 2000; Ettner, 1999; G. K. Freeman & Richards, 1994; Lambrew, DeFriese, Carey, Ricketts, & Biddle, 1996; Love, Mainous, Talbert, & Hager, 2000; O'Malley, 1997; A. S. O'Malley & Forrest, 1996; Sturmberg & Schattner, 2001). Visit continuity could have a particularly dramatic effect on outcomes of care that rely on high quality relationships between patients and providers, specifically patient adherence to treatment and screening recommendations (Rodriguez, Rogers, Marshall, & Safran, 2007). Additionally, high continuity has been associated with positive health behaviors and improved health outcomes in other populations, and thus could play a similar role among survivors (Cabana & H., 2004). Therefore, continuity, as a modifier of need, may also be a predictor of subsequent utilization, according to Andersen's model of access to care (Andersen, 1995). 99 The results generated from this work will serve two important purposes: 1) Contributing new knowledge that can inform the development of optimal, evidence-informed long-term follow-up care models, policy and care guidelines for chronic lifetime care of these cancer survivors in the community; and 2) Contributing unique information to the continuity of care literature that may be generalizable to other cancer survivor populations. 6.2 RESULTS Five key results from this research are discussed below: (1) survivors use more physician services than the general population at all examined age intervals, and as much as 20 years after their original cancer diagnosis; (2) patterns of utilization change with age differently between survivors and the general population; (3) survivors see the same primary care practitioner for only about 50% of their primary care visits; (4) continuity of care changes with age differently between survivors and the general population; and (5) the age a survivor was diagnosed has a substantial, measurable impact on both subsequent service utilization and on continuity of care. Survivors of childhood cancer use more primary care, specialist and non- physician services than their age- and birth-year-matched general population counterparts: while only 3% of the survivor cohort had an average of no recorded medical visits per year, 51% of the comparison group met this criteria. The difference in utilization levels is most dramatic for specialist visits, where survivors, 100 perhaps not surprisingly, visit a provider more than six times as often as the general population. Interestingly these results differ from those reported in the only published study that assessed health care use of survivors compared to that of the general population. In a population of childhood cancer survivors in Ontario, Shaw et al., using self report data, found that survivors did not differ from comparators in terms of use of primary care: 71% survivors versus 73% comparators (OR 0.9, 95% CI 0.8-1.0) report at least one visit per year. The dramatic difference in results between the study by Shaw et al. and this work may be illustrative of the impact of recall and selection biases, or may reflect differences in the survivor populations between BC and Ontario. The latter is unlikely, however, given the commonality of treatment strategies. As they age, survivors' use of primary care and non-physician services increases significantly, while their use of specialist services declines, changes that perhaps reflect the changing nature of their medical needs. These age related trends are not mirrored by the comparison population: the comparator group's use of primary care and non-physician services is increasing at a slower rate than their survivor counterparts, and their use of specialist visits is increasing while survivors' use is declining. This trend suggests that the excess utilization of specialist services is highest five to ten years post-diagnosis, and then declines towards the increasing utilization of specialist services seen in the general population. 101 The average UPC score recorded for survivors was 0.54 \u00C2\u00B1 0.22 (on a scale from zero to one), meaning that they saw the same primary care provider for only about 50% of their primary care needs. Work completed by Rodriguez at al. using a similar methodology and the same metric 18 found an average 0.69 in a healthy population, a full one third higher than the survivors' mean score (Rodriguez, Rogers, Marshall, & Safran, 2007; Wilson, Rogers, Hong, & Safran, 2005). Although higher than the score observed in the survivor group, this value does fall within the 95% confidence interval, suggesting that the measured survivor mean is not significantly different from the general population average. Clinically, however, a 33% difference in visit continuity seems substantial. When assessed cross-sectionally, provider continuity of care scores were higher amongst survivors than those observed in the general population sample; however, this difference disappeared once age is controlled appropriately using the biannual (longitudinal) method, suggesting that the age difference between the survivors and comparators was the main factor causing the difference in scores observed in the cross-sectional work. And indeed, age is a significant predictor of COC and UPC scores amongst both survivors and comparators. In the comparator population, scores improved significantly, for both measures, as age increased; among survivors, however, scores declined significantly with age. 18 Rodriguez et al. used administrative claims data to assess continuity using the UPC metric in a sample of 14,835 patients with 2 or more primary care visits over the 6 months before being surveyed (Rodriguez, Rogers, Marshall, & Safran, 2007). 102 The age a survivor was diagnosed had a significant impact on subsequent utilization and on continuity of care, even when specific diagnosis and treatment combinations were controlled. Although primary care visits increased with age overall, visits per year declined slightly among those diagnosed between ages 15 and 19. Similarly, while specialist visits per year declined as age increased, visits patterns were stable amongst those who were diagnosed between ages five and nine, while there was a substantial drop in visits per year amongst those diagnosed between ages 10 and 14. These age at diagnosis effects may result from a combination of a number of clinical, developmental, and social issues. First, specific diagnosis-treatment combinations tend to occur within specific age groups. These combinations may have a dramatic effect on need for treatment, changing subsequent utilization patterns. However, the diagnosis-treatment effect was controlled using regression modeling, such that the graphs above demonstrate age-related effects over and above the impact of diagnosis and treatment 19 . Second, cancer therapies result in the development of late effects of varying frequency and severity depending on the developmental age of the child receiving treatment, thus altering subsequent need for care. Lastly, social issues involved in follow-up care-seeking behaviour on behalf of a survivor and his/her family may be different depending on the age of child when he/she is diagnosed. Changes in care-seeking behaviour could therefore also 19 And diagnosis-treatment interaction terms were not significant predictors of utilization or continuity in any model. 103 be affecting subsequent use. These hypothesis should be investigated in future work. Like utilization patterns, continuity scores were affected by age at diagnosis. Individuals diagnosed between ages 10 and 14 experience a dramatic drop in scores as they age. Perhaps counter-intuitively, survivors diagnosed between ages 15 and 19 have high scores, which continue to increase as they get older (at nearly the same rate as the comparator population, an observation of considerable interest). It is possible that issues around the transition between paediatric and adult care are the reasons for this disparity: follow-up for survivors diagnosed aged 15 to 19 begins well after the transition to adult-oriented care. The follow-up period for those diagnosed between ages 10 and 14, in contrast, is centered around the time of transition, perhaps causing the some of the observed decline in continuity. This hypothesis corresponds with current thought about the losses to follow-up that tend to occur during the transition from paediatric to adult care. 6.3 IMPLICATIONS 6.3.1 FOLLOW-UP CARE MODELS As they age and the time since their cancer diagnosis increases, survivors' utilization patterns change. Importantly, visits to specialists decline, and care becomes concentrated with primary care physicians, suggesting that family physicians provide most of the follow-up care for the survivor population in British Columbia. This project did not assess the use of oncologist for follow-up; however, there are restrictions on the level of oncology care provided during the post- treatment phase, or to adult-age survivors of a paediatric cancers, and evidence 104 (from Ontario and the United States) suggests that survivors do not regularly see an oncologist once risk of tumour recurrence has been thought to diminish(A. K. Shaw et al., 2006; Vaughn & Meadows, 2002), supporting the assertion that primary care providers are the major source of follow-up care for this growing population. Primary care physicians are ideally situated to be provide follow-up, and have the advantage of being able to coordinate necessary visits to specialists. They are also more likely to provide care that is focused on overall health and well-being, as oppose to specializing only in cancer-related issues (Goldsby & Ablin, 2004). Currently, however, it is unlikely that primary care physicians are able to provide the most up-to-date, evidence-informed care for survivor populations. It is doubtful that primary care providers have the necessary awareness of survivor health issues, and information on surveillance and prevention to effectively identify and address health care needs of survivors (K. C. Oeffinger et al., 2004; Vaughn & Meadows, 2002; B. J. Zebrack, Eshelman et al., 2004). Although there are many screening and surveillance guidelines for highly specific diagnosis and treatment (and even dosage) combinations, it is unlikely that these lengthy and complex guidelines will be particularly useful to a busy family doctor who only has five to seven survivors in their entire practice (K. C. Oeffinger, 2000). Additionally, there is currently no evidence that (1) these guidelines are currently being used outside of childhood cancer centres, or (2) that following them would, in fact, reduce long term morbidity and mortality in the survivor population. 105 It has been recommended that instead of focusing on better informing family practitioners regarding the specific risks associated with cancer survivorship through the use of guidelines, linkages between oncologists and family doctors should be enhanced. There is currently minimal contact between oncologists (or paediatric cancer centres) and family doctors, minimizing information transfer between these groups (K. C. Oeffinger & Wallace, 2006). More importantly, although most survivors have general knowledge of their diagnosis and treatment (i.e. whether they received chemotherapy or radiation), few have a detailed treatment summary or are aware of the risks for late effects (Kadan-Lottick et al., 2002). Upon treatment completion, survivors and their families should be provided with a detailed treatment summary and suggestions for follow-up screening and surveillance, information that can easily be shared with the family doctor who is providing follow-up care. 6.3.2. TRANSITIONAL CARE ISSUES The issue of the transition from a paediatric cancer centre to adult-based community health care, which tends to occur around the age of 18 years, has been raised frequently in the pediatric survivorship literature (Ginsberg, Hobbie, Carlson, & Meadows, 2006; Konsler & Jones, 1993; K. C. Oeffinger & Wallace, 2006; William E. MacLean Jr, 1996); however, there is only anecdotal evidence to support the assertion that issues of transition are faced by, and are a challenge for, a survivor population. Nor is there currently evidence supporting the assertion that transitional issues could be damaging to the future health of a survivor. 106 The continuity of care analyses completed as part of this thesis, however, does, in fact support that issues of transition to adult care may be salient in this group. Continuity scores were examined over time. As survivors age, continuity declines, which is opposite to the trend of continuity improvement with age that was observed in the general population. This assertion is strengthened by the observed age at diagnosis effect. Individuals diagnosed between ages 10 and 14 experience a more dramatic drop in continuity scores than other survivor groups as they age and transition from one care setting to another. This age effect could be caused by the fact that follow-up care in this age group is centered right around the time of transition. In fact, given the average length of follow-up used in this study, continuity calculations in this group are dominated by the time of transition. Similarly, and further strengthening this argument, survivors diagnosed between ages 15 and 19 have high continuity scores overall, which continue to increase as they get older. Since follow-up care in this group begins between ages 20 and 24, well after the transition to adult-oriented care (in most cases), declines in provider continuity resulting from the transitions from pediatric to adult care are not an issue in this population. Transitional issues among survivors warrant further study. It is important to note that in addition to being recommended as part of high quality long-term follow- up care strategy for survivors, enhanced continuity could certainly help to combat the issue of loss to follow-up occurring as survivors transition into adult care (K. C. Oeffinger, 2003). The health outcomes associated with the failure to smoothly transition to adult and community care also warrant investigation. 107 6.3.3. CHANGES IN THE CONCEPTUALIZATION OF CONTINUITY OF CARE Continuity of care is recommended as a basic tenet of quality follow-up care for survivors (K. C. Oeffinger, 2004); however, what is meant by \"continuity\" in this context, and how this elusive concept links with quality is not yet clear. Similarly, which elements of the broad and somewhat ill-defined concept of continuity are valuable for a survivor population is unknown. For example, perhaps the sharing of information between providers (informational continuity) involved in a survivors' health care is more critical than having a single coordinating provider (provider continuity). Alternatively, the shared use of a screening and disease management plan (management continuity) could be more important. The concept of continuity overall is beginning to shift away from a provider- centric definition, which assumes benefit arises from consistent contact with a single provider. This conceptual shift is necessary as the health care system begins to shift preferentially to a team-based, interdisciplinary care approach, particularly in the context of complex post-cancer follow-up. Provider continuity, defined as having a continuous relationship with a single provider, is at odds with concepts of interdisciplinary, collaborative, team-based care. This raises questions about whether the relationship itself is the critical factor, or whether it is the specific knowledge of patient history and preferences that creates a positive impact on health and health care. These issues warrant investigation from not only the survivor perspective, but also in the general and health services research literature. 108 6.4 STRENGTHS AND LIMITATIONS 6.4.1 STUDY DESIGN This project used a historical cohort methodology to assess physician services utilization and primary care provider continuity among survivors of childhood cancer. Visit patterns and continuity were assessed using both cross sectional and longitudinal methodologies, with survivor-only and survivor- comparator models. The inclusion of a longitudinal approach allowed for the examination of changes to physician services utilization or provider continuity over time, as well as the comparison of observed trends between survivors and the general population. Thus far, the only two studies that examined health services utilization among survivors utilized a \"snapshot\" cross-sectional self-report approach only (K. C. Oeffinger et al., 2004; A. Shaw et al., 2006), and were thus unable to examine changes in actual utilization as survivors age and as time since diagnosis increases. The feasibility of cohort studies is often limited by cost, due to the need for continued monitoring of populations of interest, and time necessary to obtain outcome estimates (Rothman & Greenland, 1998). An Institute of Medicine monograph on methods conducting research on cancer survivors also suggests that conventional cohort methodologies, using clinical samples, have introduced the potential for bias in subject selection, participation, and follow-up (Institute of Medicine, 2006). The issues of cost and time, as well as issues related to selection biases were dealt with via the use of existing disease and utilization monitoring 109 systems (the registry and MSP databases respectively), and by relying on an inception historical cohort. 6.4.2 COHORTS The selection of the survivor and comparison cohorts is a significant methodological strength of this project for several reasons. First, and most importantly, problems associated with selection bias 20 are greatly reduced. The inclusion of all individuals who had a diagnosis of cancer under age 20 in the years of interest eliminates the possibility that survivors who use more services are more or less likely to chose to participate in this research. Work by Reijneveld and Stronks demonstrated that individuals who chose to respond in self-report based studies tend to have higher utilization of almost all types of care, leading to an overestimate of mean and median visits (Reijneveld & Stronks, 1999). This issue is of particular concern given that the only two published studies assessing utilization both relied on self-reported physician utilization data (K. C. Oeffinger et al., 2004; A. Shaw et al., 2006), potentially resulting in the introduction of selection bias (Rothman & Greenland, 1998), and overestimates of service utilization (Reijneveld & Stronks, 1999). An additional benefit of the cohort selection source, as a comprehensive registry, is the optimization of external representativeness. Results from this study are generalizable to survivors of childhood cancer within Canada. Results may also 20 If a sampling via recruitment techniques had been used, selection bias could have been introduced. In this case, survivors who elected to participate may have had particular visit or continuity patterns not observed among the survivors who chose not to participate. 110 be applied to, or compared with, other survivor populations outside of Canada, or to other chronic disease populations (Rothman & Greenland, 1998), provided that treatment strategies are comparable. The use of a registry to select a survivor cohort necessitated the use of a different sampling strategy for survivors versus comparators. One weakness of this differential sampling strategy 21 is the potential for non-differential misclassification bias: because the comparison cohort is a random sample of the MSP-enrolled population, there exists the potential some survivors may have, by chance, been included in the control cohort. Non-differential misclassification of the exposure of interest, in this case surviving a childhood cancer, would bias results towards the null, reducing the apparent differences between survivors and comparators (Rothman & Greenland, 1998). However, since childhood cancer is a rare disease (Canadian Cancer Society/National Cancer Institute of Canada, 2007), non- differential misclassification is not likely to have produced a discernable effect on the results. Another weakness resulting from the differential sampling strategy is the potential for confounding by age. Mean age at the start of follow-up was significantly greater for survivors than their comparators (14.42 and 11.73 years respectively, p<0.0001). To minimize the potential confounding effect of age on physician 21 Survivors were selected from the BC Cancer Registry; comparison subects were selected from the MSP registration file. Survivor follow-up began five-years following last known cancer diagnosis or relapse, or upon entry into MSP, whichever was later; comparator follow-up began at age five-years or start of MSP coverage, whichever as later. 111 services utilization or continuity, age was included as a covariate in all regression models. Matching by birth-year (as oppose to age at the start of follow-up, which would have removed the potential confounding by age) had the benefit of controlling for period effects, however. One additional limitation of this sampling strategy is the automatic exclusion of individuals with declared First Nations status, the Canadian Forces, veterans, and inmates in federal penitentiaries from the comparison and survivor groups. Health coverage for these individuals is provided by the Federal Government, and thus they have no possibility of being randomly selected. 6.4.3. DATA SOURCES AND LINKAGES Population-based, administrative data sets have been used to assess service utilization - among other outcomes - for many years, and the data linkage and analysis procedures have been validated and are well-established (Chamberlayne et al., 1998; Fair, 1997; Newcombe, Fair, & Lalonde, 1997). As discussed above, the use of administrative data also eliminates biases associated with the use self-report data and the attrition problems that are common in studies of long-term follow-up (D. T. Campbell & Stanley, 1963). Thus, the use of administrative data sources is a significant study strength, eliminating recall and non-response biases prevalent in self-report studies. There are certainly, however, limitations associated with the secondary use of administrative data for health care research. The selection, quality and data collection methods are not under the researchers' control, and are difficult to 112 validate (SORensen, Sabroe, & Olsen, 1996). Specifically, the prevalence of random or systematic data entry errors can be difficult to estimate; however, systematic, differential (between survivors and comparators) errors are unlikely, and random errors should be evenly distributed between the study groups. Additionally, although less an issue with the use of secondary data than primary data collection, subject attrition from datasets can be problematic. In this project, nearly one quarter of the control sample was censored due to loss to follow-up as indicated by an exit from MSP records, likely due to migration out of the province. 6.4.4. VARIABLES AND MEASUREMENT 6.4.4.a. Outcomes The optimum method for measuring the concept of continuity of care in its entirety remains elusive. To maximize construct validity, this study was limited to the assessment provider continuity of care, as informational and management continuity, though certainly important, cannot yet be accurately measured using administrative data. Two measurement tools, both previously validated for use on administrative data in a Canadian context, were selected and correlated, enhancing internal validity as well as generalizability. Both tools operate on the assumption that provider continuity is promoted when patients concentrate care with a particular physician or see the same physician sequentially. This assumption, although widely held, has yet to be properly validated. Similarly, the link between provider continuity of care and overall care quality continues to be tenuous 22 . Thus, 22 The link between continuity and overall care quality is tenuous because it has not yet been studied effectively. 113 no conjectures or assessments of care quality can, or should, be made as a result of this work. Utilization of physician services was assessed using long-term primary care, specialist and non-physician visit patterns. The use of the MSP claims database ensures the virtually complete ascertainment of fee-for-service billed patient visits. Visits to salaried providers and oncologists, who do not make claims to MSP, could not, however, be assessed using this same data source. The exclusion of oncology visits is certainly a limitation of this analysis (K. C. Oeffinger & Wallace, 2006). The role of the paediatric oncologist in coordinating or providing follow-up is still being debated, and an assessment of oncology visit patterns would be a valuable addition to this discussion. The CAYACS program has plans to abstract oncology visit data from BC Children's Hospital charts for future projects in this area. It should be noted, however, that during the period under study, there was no long-term follow-up program in place in BC that actively coordinated care between oncologists and primary or specialty care providers. Additionally, BC Children's Hospital is not mandated to provide care for individuals over 17 years of age, making the long-term use of oncology services unlikely. 6.4.5.a. Clinical and Demographic Predictor Variables Observational studies are particularly vulnerable to problems with internal validity; however, this project has been designed to minimize bias wherever possible. A comprehensive set of sociodemographic, disease and treatment variables 114 has been included in the analysis to identify potential modifiers and minimize the potential for confounding. Clinical variables included in this analysis included specific cancer diagnosis, treatment, age at diagnosis, and period of diagnosis 23 . The treatment variables were obtained using direct chart abstraction. Like other secondary data sources, there is potential for the introduction of human error during the data abstraction process. To minimize random error, the validity of the data was verified by quality checks by two data coordinators. Sociodemographic covariates used include area-specific SES, gender, age, regional health authority and rural residential status. Data for SES, health authority and ruralness were missing for a small portion of both the survivors and comparator populations. 6.4.5. STATISTICAL APPROACH The data analysis strategy was developed in consultation with two experienced statisticians, attempting to maximize the validity of statistical conclusions. Due to a large sample, the study was well powered, allowing for the detection of small, but potentially important differences between survivor groups and the general population. Although the sample size is a primary advantage of the use of administrative data and the cohort methodology, statistical overpowering is a concern. Very small differences that may be statistically significant could be 23 Period of diagnosis was not included in the final models due to collinearity with attained age and a lack of significant effect on either utilization or continuity outcomes. 115 clinically meaningless. Using regression models allows for the assessment of a multitude of predictor variables; however, the probability of making a type I error increases with each prediction. As a result, observed differences should be examined with an eye to both statistical and clinical relevance. 6.4.6. APPLICATION The use of a population-based registries and an administrative data strategy has maximized the external validity of this research. The application of these findings outside of Canada must be considered with caution, however, as methods of cancer therapy and recommendations for long-term follow-up of survivors may vary between countries. However, since approximately 90% of childhood cancer case in North America are treated using standardized clinical protocols (and similar protocols are in place in western Europe), treatment-related effects can be considered comparable in these jurisdictions. Moreover, the issues raised in the US, for example, seem to mirror those of our survivors re continuity. Additionally, generalization of the assessments of provider continuity of care to the over-arching concept of continuity should not be made as it is not possible to measure informational or management continuity in the context of this work. 6.5 DIRECTIONS FOR FUTURE RESEARCH In addition to issues of optimum follow-up care models (encompassing which, who, what, where and how often), the reconceptualization of continuity, and issues of transition to adult follow-up care discussed earlier in this chapter, a more complete study of health services utilization among survivors should be conducted. Specifically, examining oncology visits, visits to salaried providers, and use of 116 hospital services should be assessed. Additionally, a study of what proportion of visits relate directly to a late-effect of therapy, or to a screen for late-effects, new cancers or relapses is necessary to identify actual follow-up care. A more in-depth analysis of high- and low-risk survivors, and high- and low- users should also be completed, in order to aid in the development of more targeted follow-up guidelines. More work is also needed to assess the implications of the transition from paediatric- to adult-oriented care which we observed here and which is linked to reductions in provider (and perhaps informational or management) continuity. Informational and management continuity could both play a key role in enhancing the quality of follow-up care by insuring that necessary patient information is shared between providers, and a that an appropriate managed care plan is followed by all providers who have a role in the healthcare of a survivor. Studies of current informational and management continuity practices should be undertaken, and the potential value of good informational and management continuity in the context of survivor are should be assessed. Investigations regarding the link between levels of provider, informational and management continuity and health outcomes are also needed. Finally, and most importantly, studies investigating patient, family and provider perspectives on follow-up and on what constitutes quality follow-up care should be developed. To date, there is limited evidence on which providers would 117 be the best coordinators of follow-up care from either the patient or provider perspective. 6.6 CONCLUSIONS Clinical research suggests that the majority of survivors of childhood cancer face considerable risk of long-term morbidity and mortality related to their original cancer diagnosis and treatment. The necessity of life-long follow-up is increasingly being recognized; however, there has been limited research into the care they are currently receiving in the survivor phase of the cancer trajectory. This research uniquely attempted to address that knowledge gap by assessing physician services utilization and continuity of primary care among a total population of survivors over a very long period of follow-up. Survivors use significantly more primary care, specialist and non-physician services compared to the general population, even up to 20 years after their original diagnosis. As they age, their use of primary care increases dramatically (while their use of specialist services declines) suggesting that the primary care physician provides the most health care for this growing population, and thus has a key role in ensuring quality long-term follow-up. On average, survivors see the same primary care provider for only about 50% of their primary care. Of great concern, their continuity scores worsen with age, suggesting that a serious loss of care continuity may be occurring as they transition from paediatric-to adult-oriented care. The prevalence of this seemingly fragmented care approach raises questions regarding whether survivors are 118 receiving the appropriate follow-care, encompassing relevant screening, surveillance, treatments and psychosocial support. Whether this follow-up should be provided by a primary care provider, specialist or oncologist is still a subject of debate in the literature; however, it is clear that a coordinated, continuous, follow- up care strategy will critical to maintaining the health and quality of life of childhood and adolescent cancer survivors in Canada as they, in increasing numbers, reach adulthood. 119 WORKS CITED Alpert, J. J. (1964). Broken Appointments. Pediatrics, 34, 127-132. Andersen, R. M. (1968). A Behavioral Model of Families' Use of Health Services. Chicago: Centre for Health Administration Studies, University of Chicago. Andersen, R. M. (1995). Revisiting the Behavioral Model and Access to Medical Care: Does it Matter? Journal of Health and Social Behavior, 36(1), 1-10. Andersen, R. M., & Davidson, P. L. (2001). Improving Access to Care in America: Individual and Contextual Indicators. In R. M. Andersen, T. H. Rice & G. F. Kominski (Eds.), Changing the US Health Care System (2nd Edition ed., pp. 3- 30). San Francisco, CA: Jossey-Bass. Becker, M. H., Drachman, R. H., & Kirscht, J. P. (1974). Continuity of pediatrician: new support for an old shibboleth. Journal of Pediatrics, 84(4), 599-605. Bell, C. L., & Hammal, D. M. (2002). Case-control versus cohort studies in the investigation of childhood cancer. Pediatric Hematology & Oncology, /9(5), 289-293. Bhatia, S., & Sklar, C. (2002). Second cancers in survivors of childhood cancer. Nat Rev Cancer, 2(2), 124-132. Bhatia, S., Yasui, Y., Robison, L. L., Birch, J. M., Bogue, M. K., Diller, L., et al. (2003). High risk of subsequent neoplasms continues with extended follow-up of childhood Hodgkin's disease: report from the Late Effects Study Group. J Clin Oncol, 21(23), 4386-4394. 120 Bice, T. W., & Boxerman, S. B. (1977). A quantitative measure of continuity of care, Medical care (Vol. 15, pp. 347). [Philadelphia]: Lippincott Williams Wilkins. Bleyer, W., Smith, R., Green, D., DeLaat, C., Lampkin, B., Coltman, C., et al. (1993). Long-term care and lifetime follow-up: American Cancer Society Workshop on Adolescents and Young Adutls with Cancer, Atlanta, Georga, Oct 2-3, 1992. Cancer, 71(7), 2413. Bodenheimer, T., Wagner, E. H., & Grumbach, K. (2002). Improving Primary Care for Patients With Chronic Illness. JAMA, 288(14), 1775-1779. Bradford, D. W., Kaste, L. M., & Nietert, P. J. (2004). Continuity of medical care, health insurance, and nonmedical advice in the first 3 years of life, Medical care (Vol. 42, pp. 91). [Philadelphia]: Lippincott Williams Wilkins. Breslau, N., & Mortimer, E. A. (1981). Seeing the same doctor: determinants of satisfaction with specialty care for disabled children, Medical care (Vol. 19, pp. 741). [Philadelphia]: Lippincott Williams Wilkins. Breslau, N., & Reeb, K. (1975). Continuity of care in a university-based practice. J Med Educ, 50(965-969). Brousseau, D. C., Meurer, J. R., Isenberg, M. L., Kuhn, E. M., & Gorelick, M. H. (2004). Association between infant continuity of care and pediatric emergency department utilization, Pediatrics (Vol. 113, pp. 738): American Academy of Pediatrics. Cabana, M. D., & H., J. S. (2004). Does continuity of care improve patient outcomes?, The Journal of family practice (Vol. 53, pp. 974): Dowden Health Media, Inc. 121 Campbell, D. T., & Stanley, J. C. (1963). Experimental and Quasi-Experimental Designs for Research on Teaching. In N. L. Gage (ed.), Handbook of Research on Teaching. Campbell, L. K., Scaduto, M., Sharp, W., Dufton, L., Van Slyke, D., Whitlock, J. A., et al. (2007). A meta-analysis of the neurocognitive sequelae of treatment for childhood acute lymphocytic leukemia. Pediatr Blood Cancer, 49(1), 65-73. Canadian Cancer Society/National Cancer Institute of Canada. (2007). Canadian Cancer Statistics 2007. Toronto, Canada. Canadian Strategy for Cancer Control. (2002). Pediatric Working Group: Final Report: Canadian Strategy for Cancer Control. Chamberlayne, R., Green, B., Barer, M. L., Hertzman, C., Lawrence, W. J., & Sheps, S. B. (1998). Creating a population-based linked health database: a new resource for health services research. Canadian Journal of Public Health Revue Canadienne de Sante Publique, 89(4), 270-273. Chao, J. (1988). Continuity of care: incorporating patient perceptions, Family medicine (Vol. 20, pp. 333). Kansas City, Mo.: Society of Teachers of Family Medicine. Christakis, D. A., Feudtner, C., Pihoker, C., & Connell, F. A. (2001). Continuity and quality of care for children with diabetes who are covered by medicaid, Ambulatory pediatrics (Vol. 1, pp. 99). Lawrence, KS: Alliance Communications Group. Christakis, D. A., Mell, L., Wright, J. A., Davis, R., & Connell, F. A. (2000). The association between greater continuity of care and timely measles-mumps- 122 rubella vaccination, American journal of public health (Vol. 90, pp. 962). New York, N.Y.: American Public Health Association. Christakis, D. A., Wright, J. A., Zimmerman, F. J., Bassett, A. L., & Connell, F. A. (2003). Continuity of care is associated with well-coordinated care, Ambulatory pediatrics (Vol. 3, pp. 82). Lawrence, KS: Alliance Communications Group. Coughlin, S. S. (1990). Recall bias in epidemiologic studies. Journal of Clinical Epidemiology, 43(1), 87-91. Cree, M., Bell, N. R., Johnson, D., & Carriere, K. C. (2006). Increased continuity of care associated with decreased hospital care and emergency department visits for patients with asthma. Disease Management, 9(1), 63-71. Der, G., & Everitt, B. (2006). Statistical Analysis of Medical Data Using SAS. Boca Raton, FL: Chapman & Hall/CRC. Dickerman, J. D. (2007). The Late Effects of Childhood Cancer Therapy. Pediatrics, 119(3), 554-568. Donabedian, & Suppl. (1966). Evaluating the quality of medical care, The Milbank quarterly (Vol. 44, pp. 166). New York, N.Y.: Published for the Milbank Memorial Fund by Cambridge University Press. Ejlertsson, G. B. S. (1984). Continuity-of-care measures. An analytic and empirical comparison, Medical care (Vol. 22, pp. 231). [Philadelphia]: Lippincott Williams Wilkins. 123 Ettner, S. L. (1999). The relationship between continuity of care and the health behaviors of patients: does having a usual physician make a difference?, Medical care (Vol. 37, pp. 547). [Philadelphia]: Lippincott Williams Wilkins. Fagerberg, B., Claesson, L., Gosman-Hedstrom, G., & Blomstrand, C. (2000). Effect of acute stroke unit care integrated with care continuum versus conventional treatment: A randomized 1-year study of elderly patients, Stroke (Vol. 31, pp. 2578). [Philadelphia, PA]: Lippincott Williams Wilkins. Fair, M. (1997). Record Linkage in an Information Age Society. Paper presented at the Record linkage Techniques: Proceedings of an International Workshop and Exposition, Arlington, VA. Fjaertoft, H., Indredavik, B., Johnsen, R., & Lydersen, S. (2004). Acute stroke unit care combined with early supported discharge. Long-term effects on quality of life. A randomized controlled trial, Clinical rehabilitation (Vol. 18, pp. 580). London: Edward Arnold. Freeman, G., & Hjortdahl, P. (1997). What future for continuity of care in general practice?, BMJ (Vol. 314, pp. 1870). London: BMJ Pub. Group. Freeman, G. K., & Richards, S. C. (1994). Personal continuity and the care of patients with epilepsy in general practice, The British journal of general practice (Vol. 44, pp. 395). London: The College. Friedman, D. L., Freyer, D. R., & Levitt, G. A. (2006). Models of Care for Survivors of Childhood Cancer. Pediatr Blood Cancer, 46(2), 159-168. 124 Gallagher, T. C., Geling, 0., & Comite, F. (2001). Use of multiple providers for regular care and women's receipt of hormone replacement therapy counseling, Medical care (Vol. 39, pp. 1086). [Philadelphia]: Lippincott Williams Wilkins. Garre, M. L., Gandus, S., Cesana, B., Haupt, R., De Bernardi, B., Comelli, A., et al. (1994). Health status of long-term survivors after cancer in childhood. Results of an uniinstitutional study in Italy. Am J Pediatr Hematol Oncol, 16(2), 143-152. Geenen, M. M., Cardous-Ubbink, M. C., Kremer, L. C. M., van den Bos, C., van der Pal, H. J. H., Heinen, R. C., et al. (2007). Medical Assessment of Adverse Health Outcomes in Long-term Survivors of Childhood Cancer. JAMA: The Journal of the American Medical Association, 297(24), 2705-2715. Gill, J. M., & Mainous, A. G. (1998). The role of provider continuity in preventing hospitalizations, Archives of family medicine (Vol. 7, pp. 352). Chicago, IL: American Medical Association. Gill, J. M., Mainous, A. G., & Nsereko, M. (2000). The effect of continuity of care on emergency department use, Archives of family medicine (Vol. 9, pp. 333). Chicago, IL: American Medical Association. Ginsberg, J. P., Hobbie, W. L., Carlson, C. A., & Meadows, A. T. (2006). Delivering Long-Term Follow-Up Care to Pediatric Cancer Survivors: Transitional Care Issues. Pediatr Blood Cancer, 46(2), 169-173. Goldsby, R. E., & Ablin, A. R. (2004). Surviving childhood cancer; now what? Controversies regarding long-term follow-up. Pediatric Blood & Cancer, 43(3), 211-214. 125 Green, C. A., & Pope, C. R. (1999). Gender, psychosocial factors and the use of medical services: a longitudinal analysis. Social Science & Medicine, 48(10), 1363-1372. Greenland, S. (1977). Response and follow-up bias in cohort studies Am. J. Epidemiol., 106(3), 184-187. Grunfeld, E., Dhesy-Thind, S., & Levine, M. (2005). Clinical practice guidelines for the care and treatment of breast cancer: follow-up after treatment for breast cancer (summary of the 2005 update). Cmaj, 72(10), 1319-1320. Grunfeld, E., Gray, A., Mant, D., Yudkin, P., Adewuyi-Dalton, R., Coyle, D., et al. (1999). Follow-up of breast cancer in primary care vs specialist care: results of an economic evaluation. Br J Cancer, 79(7-8), 1227-1233. Harrison, M. B., Browne, G. B., Roberts, J., Tugwell, P., Gafni, A., & Graham, I. D. (2002). Quality of life of individuals with heart failure: a randomized trial of the effectiveness of two models of hospital-to-home transition, Medical care (Vol. 40, pp. 271). [Philadelphia]: Lippincott Williams Wilkins. Hawkins, M. M., Draper, G. J., & Kingston, J. E. (1987). Incidence of second primary tumours among childhood cancer survivors. Br J Cancer, 56(3), 339-347. Hibbard, J. H., & Pope, C. R. (1986). Age Differences in the Use of Medical Care in an HMO: An Application of the Behavioral Model. Medical Care, 24(1), 52-66. Hobbie, W. L., Stuber, M., Meeske, K., Wissler, K., Rourke, M. T., Ruccione, K., et al. (2000). Symptoms of Posttraumatic Stress in Young Adult Survivors of Childhood Cancer. J Clin Oncol, 18(24), 4060-4066. 126 Hudson, M., Mertens, A., Yasui, Y., Hobbie, W., Chen, H., Gurney, J., et al. (2003). Health Status of Adult Long-term Survivors of Childhood Cancer: A Report from CCSS.JAMA, 290(12), 1583-1592. Hudson, M. M., Mertens, A. C., Yasui, Y., Hobbie, W., Chen, H., Gurney, J. G., et al. (2003). Health status of adult long-term survivors of childhood cancer: a report from the Childhood Cancer Survivor Study. JAMA, 290(12), 1583-1592. Humpl, T., Fritsche, M., Bartels, U., & Gutjahr, P. (2001). Survivors of childhood cancer for more than twenty years. Acta Oncol, 40(1), 44-49. Institute of Medicine. (2006). From Cancer Patient to Cancer Survivor: Lost in Transition (An American Society of Clinical Oncology and Institute of Medicine Symposium). Washington, D.C.: National Academy Press. Irigoyen, M., Findley, S. E., Chen, S., Vaughan, R., Sternfels, P., Caesar, A., et al. (2004). Early continuity of care and immunization coverage, Ambulatory pediatrics (Vol. 4, pp. 199). Lawrence, KS: Alliance Communications Group. Jazbec, J., Ecimovic, P., & Jereb, B. (2004). Second neoplasms after treatment of childhood cancer in Slovenia. Pediatr Blood Cancer, 42(7), 574-581. Jee, S. H., & Cabana, M. D. (2006). Indices for continuity of care: a systematic review of the literature. Medical Care Research & Review, 63(2), 158-188. Kadan-Lottick, N. S., Robison, L. L., Gurney, J. G., Neglia, J. P., Yasui, Y., Hayashi, R., et al. (2002). Childhood cancer survivors' knowledge about their past diagnosis and treatment: Childhood Cancer Survivor Study. JAMA, 287(14), 1832-1839. 127 Keitz, S. A., Box, T. L., Homan, R. K., Bartlett, J. A., & Oddone, E. Z. (2001). Primary care for patients infected with human immunodeficiency virus: a randomized controlled trial, Journal of general internal medicine (Vol. 16, pp. 573). Malden, MA: Blackwell Science. Konsler, G. K., & Jones, G. R. (1993). Transition Issues for Survivors of Childhood Cancer and their Healthcare Providers. Cancer Practice, 1(4), 319-324. Koopman, R. J., Mainous, A. G., III, Baker, R., Gill, J. M., & Gilbert, G. E. (2003). Continuity of Care and Recognition of Diabetes, Hypertension, and Hypercholesterolemia. Arch Intern Med, 163(11), 1357-1361. Kramarova, E., & Stiller, C. (1998). International Classification of Childhood Cancer. In D. Parkin, E. Kramarova, G. Draper, E. Masuyer, J. Michaelis, J. MNeglia, S. Qureshi & C. Stiller (Eds.), International Incidence of Childhood Cancer (Vol. IARC Scientific Pub No 144, pp. 15-19). Lyon, France: International Agency for Research on Cancer. Lambrew, J. M., DeFriese, G. H., Carey, T. S., Ricketts, T. C., & Biddle, A. K. (1996). The effects of having a regular doctor on access to primary care, Medical care (Vol. 34, pp. 138). [Philadelphia]: Lippincott Williams Wilkins. Landier, W., Bhatia, S., Eshelman, D. A., Forte, K. J., Sweeney, T., Hester, A. L., et al. (2004). Development of risk-based guidelines for pediatric cancer survivors: the Children's Oncology Group Long-Term Follow-Up Guidelines from the Children's Oncology Group Late Effects Committee and Nursing Discipline. Journal of Clinical Oncology, 22(24), 4979-4990. 128 Landier, W., Wallace, W. H. B., & Hudson, M. M. (2006). Long-Term Follow-Up of Pediatric Cancer Survivors: Education, Surveillance, and Screening. Pediatr Blood Cancer, 46(2), 149-158. Lansky, S. B., List, M. A., & Ritter-Sterr, C. (1986). Psychosocial consequences of cure, Cancer (Vol. 58, pp. 529). New York, NY: John Wiley Sons. Love, M. M., Mainous, A. G., Talbert, J. C., & Hager, G. L. (2000). Continuity of care and the physician-patient relationship: the importance of continuity for adult patients with asthma, The Journal of family practice (Vol. 49, pp. 998): Dowden Health Media, Inc. MacArthur, A. C., Spinelli, J. J., Rogers, P. C., Goddard, K. J., Abanto, Z. U., & McBride, M. L. (2007). Mortality among 5-year survivors of cancer diagnosed during childhood or adolescence in British Columbia, Canada. Pediatr Blood Cancer, 48(4), 460-467. MacArthur, A. C., Spinelli, J. J., Rogers, P. C., Goddard, K. J., Phillips, N., & McBride, M. L. (2007). Risk of a second malignant neoplasm among 5-year survivors of cancer in childhood and adolescence in British Columbia, Canada. Pediatr Blood Cancer, 48(4), 453-459. Meadows, A. T. (2003). Pediatric cancer survivors: Past history and future challenges, Current Problems in Cancer (Vol. 27, pp. 112). Chicago, 111.: Year Book Medical Publishers. Menec, V. H., Sirski, M., Attawar, D., & Katz, A. (2006). Does continuity of care with a family physician reduce hospitalizations among older adults? Journal of Health Services & Research Policy, 11(4), 196-201. 129 Met-tens, A. C., Cotter, K. L., Foster, B. M., Zebrack, B. J., Hudson, M. M., Eshelman, D., et al. (2004). Improving health care for adult survivors of childhood cancer: recommendations from a delphi panel of health policy experts. Health Policy, 69(2), 169-178. Miller, R. W., Fusner, J. E., Fink, R. J., Murphy, T. M., Getson, P. R., Vojtova, J. A., et al. (1986). Pulmonary function abnormalities in long-term survivors of childhood cancer, Medical and pediatric oncology (Vol. 14, pp. 202). New York, NY: Wiley-Liss. Moher. (2001). Cluster randomised controlled trial to compare three methods of promoting secondary prevention of coronary heart disease in primary, BMJ (Vol. 322, pp. 1338). London: The Association. Mulhern, R. K., & Palmer, S. L. (2003). Neurocognitive late effects in pediatric cancer. Curr Probl Cancer, 27(4), 177-197. National Cancer Policy, B., Hewitt, M., Weiner, S. L., & Simone, J. V. (2003). Childhood Cancer Survivorship: Improving Care and Quality of Life. Washington, DC: National Academies Press. National Cancer Policy Board (U.S.), Hewitt, M., Weiner, S. L., & Simone, J. V. (2003). Childhood Cancer Survivorship: Improving Care and Quality of Life. Washington, DC: National Academies Press. Naylor, M. D., Brooten, D. A., Campbell, R. L., Maislin, G., McCauley, K. M., Schwartz, J., et al. (2004). Transitional care of older adults hospitalized with heart failure: a randomized, controlled trial, Journal of the American Geriatrics Society (Vol. 52, pp. 675). Baltimore, MD [etc.]: Williams Wilkins [etc.]. 130 Newcombe, H., Fair, M., & Lalonde, P. (1997). The Use of Names for Linking Personal Records. Paper presented at the Record Linkage Techniques: Proceedings of an International Workshop and Exposition, Arlington, VA. Nielsen, J. D., Palshof, T., Mainz, J., Jensen, A. B., & Olesen, F. (2003). Randomised controlled trial of a shared care programme for newly referred cancer patients: bridging the gap between general practice and hospital. Qual Saf Health Care, 12(4), 263-272. O'Malley. (1997). Continuity of care and the use of breast and cervical cancer screening services in a multiethnic community, Archives of internal medicine (Vol. 157, pp. 1462). Chicago, IL: American Medical Association. O'Malley, A. S., & Forrest, C. B. (1996). Continuity of care and delivery of ambulatory services to children in community health clinics, Journal of community health (Vol. 21, pp. 159). [New York: Human Sciences Press]. Oeffinger. (2007). Childhood cancer survivors, late effects, and a new model for understanding survivorship, JAMA (Vol. 297, pp. 2762). Oeffinger, K. C. (2000). Childhood cancer survivors and primary care physicians./ Fam Pract, 49(8), 689-690. Oeffinger, K. C. (2003). Longitudinal risk-based health care for adult survivors of childhood cancer. Curr Probl Cancer, 27(3), 143-167. Oeffinger, K. C. (2004). Longitudinal Cancer-Related Health Care for Adult Survivors of Childhood Cancer: National Cancer Policy Board. 131 Oeffinger, K. C., & Hudson, M. M. (2004). Long-term complications following childhood and adolescent cancer: foundations for providing risk-based health care for survivors. CA Cancer] Clin, 54(4), 208-236. Oeffinger, K. C., Mertens, A. C., Hudson, M. M., Gurney, J. G., Casillas, J., Chen, H., et al. (2004). Health care of young adult survivors of childhood cancer: a report from the Childhood Cancer Survivor Study. Ann Fam Med, 2(1), 61-70. Oeffinger, K. C., Mertens, A. C., Sklar, C. A., Kawashima, T., Hudson, M. M., Meadows, A. T., et al. (2006). Chronic health conditions in adult survivors of childhood cancer. New England Journal of Medicine, 355(15), 1572-1582. Oeffinger, K. C., & Wallace, W. H. B. (2006). Barriers to Follow-up Care of Survivors in the United States and the United Kingdom. Pediatr Blood Cancer, 46(2), 135-142. Organization, W. H. (1990). International Classification of Diseases for Oncology. Second Edition. Parchman, M. L., Pugh, J. A., Hitchcock, P., & Larme, A. C. (2002). Continuity of care, self-management behaviors, and glucose control in patients with type 2 diabetes, Medical care (Vol. 40, pp. 137). [Philadelphia]: Lippincott Williams Wilkins. Preen, D. B., Bailey, B. E. S., Wright, A., Kendall, P., Phillips, M., Hung, J., et al. (2005). Effects of a multidisciplinary, post-discharge continuance of care intervention on quality of life, discharge satisfaction, and hospital length of stay: a randomized controlled trial, International Journal for Quality in Health Care (Vol. 17, pp. 43). Oxford: Pergamon. 132 Raymond-Speden, E., Tripp, G., Lawrence, B., & Holdaway, D. (2000 Mar). Intellectual, neuropsychological, and academic functioning in long-term survivors of leukemia. Journal of Pediatric Psychology, 25(2):59-68. Recklitis, C. J., Lockwood, R. A., Rothwell, M. A., & Diller, L. R. (2006). Suicidal ideation and attempts in adult survivors of childhood cancer. Journal of Clinical Oncology, 24(24), 3852-3857. Reid, B. C., & Rozier, R. G. (2006). Continuity of care and early diagnosis of head and neck cancer. Oral oncology, 42(5), 510-516. Reid, R., Barer, M., McKendry, R., McGrail, K., Prosser, B., Green, B., et al. (2003). Patient-focused care over time: Issues related to measurement, prevalence and strategies for improvement among patient populations (No. RC2-0276-10). Ottawa, ON: Canadian health Services Research Foundation. Reid, R., Haggerty, J., & McKendry, R. (2002). Defusing the Confusion: Concepts and Measures of Continuity of Healthcare. Ottawa, ON: Canadian Health Services Research Foundation. Reid, R. J., Barer, M. L., McKendry, R., McGrail, K. M., Prosser, B., Green, B., et al. (2003). Patient-focused care over time: Issues related to measurement, prevalence, and strategies for improvement among patient populations: Canadian Health Services Research Foundation. Reijneveld, S. A., & Stronks, K. (1999). The impact of response bias on estimates of health care utilization in a metropolitan area: the use of administrative data. Int. J. Epidemiol., 28(6), 1134-1140. 133 Ries, L. A. G., Melbert, D., Krapcho, M., Mariotto, A., Miller, B. A., Feuer, E. J., et al. (Eds.). (2007). SEER Cancer Statistics Review, 1973-2004. Bethesda, MD: National Cancer Institute. Robertson, C. M., Hawkins, M. M., & Kingston, J. E. (1994). Late deaths and survival after childhood cancer: implications for cure. Bmj, 309(6948), 162-166. Rodriguez, H. P., Rogers, W. H., Marshall, R. E., & Safran, D. G. (2007). The effects of primary care physician visit continuity on patients' experiences with care. Journal of General Internal Medicine, 22(6), 787-793. Rosoff, P. M. (2006). The two-edged sword of curing childhood cancer. New England Journal of Medicine, 355(15), 1522-1523. Rothman, K. J., & Greenland, S. (1998). Modern Epidemiology (Second Edition ed.). Philadelphia, PA: Lippincott Williams and Wilkins. Rourke, M. T., Stuber, M. L., Hobbie, W. L., & Kazak, A. E. (1999). Posttraumatic Stress Disorder: Understanding the Psychosocial Impact of Surviving Childhood Cancer Into Young Adulthood. Journal of Pediatric Oncology Nursing, /6(3), 126-135. Samet. (2003). Linking alcohol- and drug-dependent adults to primary medical care: a randomized controlled trial of a multi-disciplinary health intervention in a detoxification unit, Addiction (Vol. 98, pp. 509). Abingdon, Oxfordshire, UK: Carfax Pub. Co. Saultz, J. W., & Albedaiwi, W. (2004). Interpersonal Continuity of Care and Patient Satisfaction: A Critical Review. Ann Fam Med, 2(5), 445-451. 134 Shaw, A., Pogany, L., Speechley, K. N., Maunsell, E., Barrera, M., & Mery, L. S. (2006). Use of health care services by survivors of childhood and adolescent cancer in Canada. Cancer, 106(8), 1829-1837. Shaw, A. K., Pogany, L., Speechley, K. N., Maunsell, E., Barrera, M., & Mery, L. S. (2006). Use of health care services by survivors of childhood and adolescent cancer in Canada. Cancer, 106(8), 1829-1837. Skinner, R. (2005). Therapy based long-term follow up: A practice statement. Leicester: United Kingdom Childrens' Cancer Study Group (Late Effects Group). Skinner, R., Wallace, W. H., & Levitt, G. (2007). Long-term follow-up of children treated for cancer: why is it necessary, by whom, where and how? Archives of Disease in Childhood, 92(3), 257-260. Skinner, R., Wallace, W. H., Levitt, G. A., & Uk Children's Cancer Study Group Late Effects, G. (2006). Long-term follow-up of people who have survived cancer during childhood. Lancet Oncology, 7(6), 489-498. Sklar, C. A., & Constine, L. S. (1995). Chronic neuroendocrinological sequelae of radiation therapy, International journal of radiation oncology, biology, physics (Vol. 31, pp. 1113). New York: Pergamon Press. Sklar, C. A., Mertens, A. C., Mitby, P., Occhiogrosso, G., Qin, J., Heller, G., et al. (2002). Risk of disease recurrence and second neoplasms in survivors of childhood cancer treated with growth hormone: a report from the Childhood Cancer Survivor Study./ Clin Endocrinol Metab, 87(7), 3136-3141. 135 Smeenk, F. W. (1998). Transmural care. A new approach in the care for terminal cancer patients: its effects on re-hospitalization and quality of life, Patient Education and Counseling (Vol. 35, pp. 189). Princeton, N.J.: Excerpta Medica. SORensen, H. T., Sabroe, S., & Olsen, J. (1996). A Framework for Evaluation of Secondary Data Sources for Epidemiological Research. Int. J. Epidemiol., 25(2), 435-442. Steinherz, L. J., Steinherz, P. G., Tan, C. T., Heller, G., & Murphy, M. L. (1991). Cardiac toxicity 4 to 20 years after completing anthracycline therapy,JAMA (Vol. 266, pp. 1672). Steinwachs, D. M. (1979). Measuring provider continuity in ambulatory care: an assessment of alternative approaches, Medical care (Vol. 17, pp. 551). [Philadelphia]: Lippincott Williams Wilkins. Stevens, M. C., Mahler, H., & Parkes, S. (1998). The health status of adult survivors of cancer in childhood. Eur J Cancer, 34(5), 694-698. Sturmberg, J. P., & Schattner, P. (2001). Personal doctoring. Its impact on continuity of care as measured by the comprehensiveness of care score, Australian family physician (Vol. 30, pp. 513). Sydney: Australian College of General Practitioners. Teta, M. J., Del Po, M. C., Kasl, S. V., Meigs, J. W., Myers, M. H., & Mulvihill, J. J. (1986). Psychosocial consequences of childhood and adolescent cancer survival. J Chronic Dis, 39(9), 751-759. Vaughn, D. J., & Meadows, A. T. (2002). Cancer Survivorship Research: The Best Is Yet to Come. J Clin Oncol, 20(4), 888-890. 136 Wachter, R. M. (2001). Does continuity of care matter? No: discontinuity can improve patient care, The Western journal of medicine (Vol. 175, pp. 5). San Francisco, Calif. [etc.]: BMJ Pub.. Wasson, J., Sauvigne, A., Mogielnicki, R., Frey, W., Sox, C., & Gaudette, C. (1984). Continuity of outpatient medical care in elderly men. JAMA, 252, 2413-2417. Weiss, G. L., & A., R. C. (1989). Regular source of primary medical care and patient satisfaction, Quality review bulletin (Vol. 15, pp. 180). Chicago, Ill.: Joint Commission on Accreditation of Hospitals. Wilkins, R. (July 2002). PCCF+ Version 3] User's Guide (Geocodes/PCCF). Automated Geographic Coding Based on the Statistics Canada Postal Code Conversion Files, Including Postal Codes to May 2002: Health Analysis and Measurement Group, Statistics Canada, Ottawa. William E. MacLean Jr, G. V. F. K. R. C. S. (1996). Transitions in the care of adolescent and young adult survivors of childhood cancer. Cancer, 78(6), 1340-1344. Williams, J. G., Cheung, W. Y., Chetwynd, N., Cohen, D. R., El-Sharkawi, S., Finlay, I., et al. (2001). Pragmatic randomised trial to evaluate the use of patient held records for the continuing care of patients with cancer, Quality in health care (Vol. 10, pp. 159). London: BMJ Pub. Group. Wilson, I. B., Rogers, W. H., Hong, C., & Safran, D. G. (2005). Cost-related Skipping of Medications and Other Treatments Among Medicare Beneficiaries Between 1998 and 2000.IGIM: Journal of General Internal Medicine, 20(8), 715-720. Zebrack, B. J., Eshelman, D. A., Hudson, M. M., Mertens, A. C., Cotter, K. L., Foster, B. M., et al. (2004). Health care for childhood cancer survivors: insights and 137 perspectives from a Delphi panel of young adult survivors of childhood cancer. Cancer, 100(4), 843-850. Zebrack, B. J., Gurney, J. G., Oeffinger, K., Whitton, J., Packer, R. J., Mertens, A., et al. (2004). Psychological outcomes in long-term survivors of childhood brain cancer: a report from the childhood cancer survivor study.] Clin Oncol, 22(6), 999-1006. Zebrack, B. J., Zevon, M. A., Turk, N., Nagarajan, R., Whitton, J., Robison, L. L., et al. (2007). Psychological distress in long-term survivors of solid tumors diagnosed in childhood: a report from the childhood cancer survivor study. Pediatric Blood & Cancer, 49(1), 47-51. 138 APPENDICES APPENDIX A: THE CHILDHOOD, ADOLESCENT, AND YOUNG ADULT CANCER SURVIVORSHIP RESEARCH PROGRAM The Childhood, Adolescent, and Young Adult Cancer Survivorship Research Program (CAYACS) is a population-based cohort study of long-term impacts, support, and interventions to maximize quality of life among survivors of cancer diagnosed under the age of 25 years in British Columbia, Canada. CAYACS is also a resource for survivor research and for knowledge translation for policy and practice. The program is supported by the Canadian Cancer Society through the National Cancer Institute of Canada. The objectives of this research program are as follows: \u00E2\u0080\u00A2 To identify a survivor cohort and a comparison group, and develop a database as an ongoing resource for surveillance and research into the survivorship experience of this patient group; \u00E2\u0080\u00A2 To carry out a set of studies of treatment impacts including late health effects and educational effects; \u00E2\u0080\u00A2 To examine health care and resource issues, in particular health care utilization and continuity of care, relating to these late effects; \u00E2\u0080\u00A2 To build the capacity for ongoing collaborative research of survivorship issues; and 139 \u00E2\u0080\u00A2 To develop tools to translate and transfer research results to policymakers and care providers. This research program will address existing gaps in knowledge of risks and resource issues for this group. Research findings will be transferred to policymakers and care providers for development of long term strategies. Methodologies are innovative and cost-effective, and results will be comprehensive. These cohorts and databases can be maintained into the future to identify new risks, evaluate longer- term outcomes in these domains and to assess compliance with, and effectiveness of, strategies for long-term management of survivors. They will also serve as a surveillance and research resource for ongoing investigation of survivorship issues. 140 APPENDIX B: CERTIFICATE OF EXPEDITITED ETHICS APPROVAL - ANNUAL REVIEW 141 Page 1 of 2 BC Cancer Agency University of British Columbia - British Columbia Cancer Agency Research Ethics Board (UBC BCCA REB) UBC BCCA Research Ethics Board Fairmont Medical Building (6th Floor) 614 - 750 West Broadway Vancouver, BC V5Z 1H5 Tel: (604) 877-6284 Fax: (604) 708-2132 E-mail: reb@bccancer.bc.ca Website: httplIwww.bccancer.bc.ca > Research Ethics RISe: http://rise.ubc.ca UBc 9671^tit Certificate of Expedited Approval: Annual Renewal PRINCIPAL INVESTIGATOR: Mary L. McBride INSTITUTION / DEPARTMENT: BCCA/Cancer Control Research (BCCA) REB NUMBER: H05-60113 INSTITUTION(S) WHERE RESEARCH WILL BE CARRIED OUT: Institution^ I^ Site BC Cancer Agency^ Vancouver BCCA Other locations where the research will be conducted: BC Cancer Research Centre PRINCIPAL INVESTIGATOR FOR EACH ADDITIONAL PARTICIPATING BCCA CENTRE: Vancouver:^N/A^ Vancouver Island:^N/A Fraser Valley: N/A Southern Interior:^N/A SPONSORING AGENCIES AND COORDINATING GROUPS: National Cancer Institute of Canada PROJECT TITLE: Childhood / Adolescent / Young Adult Cancer Survivorship Research Program APPROVAL DATE: August 14, 2007 EXPIRY DATE OF THIS APPROVAL: August 14, 2008 CERTIFICATION: 1. The membership of the UBC BCCA REB complies with the membership requirements for research ethics boards defined in Division 5 of the Food and Drug Regulations of Canada. 2. The UBC BCCA REB carries out its functions in a manner fully consistent with Good Clinical Practices. 3. The UBC BCCA REB has reviewed and approved the research project named on this Certificate of Approval including any associated consent form and taken the action noted above. This research project is to be conducted by the provincial investigator named above. This review and the associated minutes of the UBC BCCA REB have been documented electronically and in writing. The UBC BCCA Research Ethics Board has reviewed the documentation for the above named project. The research study as presented in documentation, was found to be acceptable on ethical grounds for research involving human subjects and was approved for renewal by the UBC BCCA REB. UBC BCCA Ethics Board Approval of the above has been verified by one of the Following: https://rise.ubc.ca/rise/Doc/0/8ESUJM7721(134N9VCIEVDIRC9F9/fromString.html^9/18/2007 142 Page 2 of 2 Dr. George Brownian,^Dr. Joseph Connors,^Dr. Lynne Nakashima Chair^ First Vice-Chair Second Vice-Chair f you have any questions, please call: Bonnie Shields, Manager, BCCA Research Ethics Board: 604-877-6284 or e-mail: reb@bccancer.bc.ca Dr. George Browman, Chair: 604-877-6284 or e-mail: gbrowman@bccanceribc.ca Dr. Joseph Connors, First Vice-Chair: 604-877-6000-ext. 2746 or e-mail: jconnors@bccancer.bc.ca Dr. Lynne Nakashima, Second Vice-Chair: 604-707-5989 or e-mail, Inakas@bccancer.bc.ca https://rise.ubc.calrise/Doc/0/8ESLUM772KB4N9VCIEVDIRC9149/fromString.html^9/18/2007 143 APPENDIX C: REGRESSION MODELS - VISIT PATTERNS Table C.1 Survivor-Only, Overall: Total Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 4.5476 3.5410 5.8410 <.0001 ICCC Diagnosis VIII (Malignant Bone Tumors) 1.0685 0.8322 1.3722 0.6032 III (CNS Cancers) 1.2499 1.0187 1.5339 0.0326* XI (Other Malignant Epithelial 1.2115 0.9334 1.5724 0.1492 Neoplasms) X (Germ Cell Tumors) 1.0696 0.8470 1.3509 0.5720 VII (Hepatic Tumors) 0.8129 0.4083 1.6185 0.5556 II (Lymphomas) 0.9332 0.7919 1.0999 0.4099 IV (Neuroblastomas) 1.1773 0.8793 1.5762 0.2731 XII (Other Unspecified) 1.3256 0.2829 6.2102 0.7205 VI (Renal Tumors) 0.8060 0.6197 1.0481 0.1076 V (Retinoblastomas) 0.8814 0.6252 1.2423 0.4708 IX (Soft Tissue Sarcomas) 0.9972 0.7890 1.2602 0.9811 I (Leukemias)R 1.0000 1.0000 1.0000 . Treatment Modality Chemotherapy only 1.0333 0.8406 1.2702 0.7555 Chemotherapy and radiation 1.1567 0.9483 1.4109 0.1509 Chemotherapy and surgery 1.1367 0.9571 1.3497 0.1441 Chemotherapy, radiation and surgery 1.4240 1.1840 1.7127 0.0002* No Info 0.9497 0.7608 1.1855 0.6482 No treatment 0.6835 0.4248 1.0998 0.1169 Radiation only 0.9227 0.6631 1.2838 0.6328 Radiation and surgery 1.2430 1.0286 1.5020 0.0244 Surgery only R 1.0000 1.0000 1.0000 . Sex Female 1.4192 1.2994 1.5502 <.0001* Male R 1.0000 1.0000 1.0000 . SES Unknown 0.7817 0.6033 1.0130 0.0625 1 (Lowest) 1.0737 0.9309 1.2383 0.3289 2 1.1052 0.9582 1.2747 0.1697 3 1.0575 0.9198 1.2157 0.4322 4 1.1770 1.0217 1.3561 0.024* 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority Unknown 1.0527 0.9316 1.1897 0.4097 Fraser Health 1.0123 0.8493 1.2065 0.8919 Interior Health 0.8806 0.7180 1.0802 0.2224 144 Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Northern Health 1.1184 0.7855 1.5923 0.5349 Vancouver Island Health 1.0062 0.8663 1.1686 0.9356 Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 0.9666 0.8266 1.1302 0.6704 Rural 1.0130 0.8802 1.1658 0.8572 Small community 1.0261 0.8830 1.1924 0.7362 Metropolitan R 1.0000 1.0000 1.0000 . Age at Diagnosis 1.0166 1.0075 1.0258 0.0003* Relapse Status Relapse <5yrs 1.3742 1.2379 1.5256 <.0001* Relapse >5yrs 1.7213 1.3326 2.2235 <.0001* No relapse R 1.0000 1.0000 1.0000 . 11 Reference category 145 Table C.2 Survivor-Only, Overall: Primary Care Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 2.1107 1.6522 2.6961 <.0001* ICCC Diagnosis VIII (Malignant Bone Tumors) 0.8794 0.6856 1.1281 0.3119 III (CNS Cancers) 1.1252 0.9203 1.3759 0.2503 XI (Other Malignant Epithelial 0.9881 0.7651 1.2760 0.9266 Neoplasms) X (Germ Cell Tumors) 0.9524 0.7563 1.1991 0.6778 VII (Hepatic Tumors) 0.9019 0.4434 1.8349 0.7758 II (Lymphomas) 0.9208 0.7825 1.0836 0.3205 IV (Neuroblastomas) 1.0609 0.7925 1.4201 0.6912 XII (Other Unspecified) 1.1958 0.2651 5.3935 0.8160 VI (Renal Tumors) 0.9303 0.7182 1.2049 0.5839 V (Retinoblastomas) 1.0434 0.7427 1.4658 0.8063 IX (Soft Tissue Sarcomas) 1.0236 0.8107 1.2925 0.8445 I (Leukemias)R 1.0000 1.0000 1.0000 . Treatment Modality Chemotherapy only 1.0778 0.8406 1.2702 0.7555 Chemotherapy and radiation 1.1116 0.9483 1.4109 0.1509 Chemotherapy and surgery 1.1171 0.9571 1.3497 0.1441 Chemotherapy, radiation and surgery 1.0976 1.1840 1.7127 0.0002* No Info 1.1375 0.7608 1.1855 0.6482 No treatment 0.7176 0.4248 1.0998 0.1169 Radiation only 0.8851 0.6631 1.2838 0.6328 Radiation and surgery 1.1109 1.0286 1.5020 0.0244* Surgery only R 1.0000 1.0000 1.0000 . Relapse Status Relapse 5yrs 1.5305 1.1895 1.9693 0.0009* No relapse R 1.0000 1.0000 1.0000 . Sex Female 1.4724 1.3495 1.6064 <.0001* Male R 1.0000 1.0000 1.0000 . SES Unknown 1.0462 0.9106 1.2023 0.5235 1 (Lowest) 1.1343 0.9852 1.3060 0.0797 2 1.1589 1.0067 1.3343 0.04* 3 1.0318 0.8992 1.1840 0.6551 4 1.0462 0.9106 1.2023 0.5235 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority 146 Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Unknown 1.2015 1.0642 1.3565 0.003* Fraser Health 1.1346 0.9526 1.3515 0.1570 Interior Health 1.0753 0.8777 1.3176 0.4835 Northern Health 1.3610 0.9619 1.9257 0.0817 Vancouver Island Health 1.1052 0.9525 1.2824 0.1876 Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 1.0189 0.8711 1.1916 0.8154 Rural 1.1260 0.9813 1.2920 0.0908 Small community 1.1698 1.0069 1.3589 0.0404* Metropolitan R 1.0000 1.0000 1.0000 . Age at Diagnosis 1.0272 1.0181 1.0363 <.0001* R Reference category 147 Table C.3 Survivor-Only, Overall: Specialist Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 1.4703 1.0168 2.1261 0.0405* ICCC Diagnosis VIII (Malignant Bone Tumors) 2.0148 1.4164 2.8657 <.0001* III (CNS Cancers) 1.5699 1.1633 2.1183 0.0032* XI (Other Malignant Epithelial 1.3513 0.9328 1.9580 0.1114 Neoplasms) X (Germ Cell Tumors) 1.4181 1.0123 1.9864 0.0423* VII (Hepatic Tumors) 0.7648 0.2866 2.0409 0.5923 II (Lymphomas) 0.7998 0.6349 1.0075 0.0579 IV (Neuroblastomas) 1.3925 0.9223 2.1024 0.1153 XII (Other Unspecified) 2.9606 0.3590 24.4151 0.3133 VI (Renal Tumors) 0.6188 0.4239 0.9031 0.0128* V (Retinoblastomas) 0.8714 0.5314 1.4290 0.5855 IX (Soft Tissue Sarcomas) 1.0243 0.7393 1.4194 0.8851 I (Leukemias)R 1.0000 1.0000 1.0000 . Treatment Modality Chemotherapy only 1.3712 1.0138 1.8547 0.0405* Chemotherapy and radiation 1.6945 1.2710 2.2594 0.0003* Chemotherapy and surgery 1.4746 1.1553 1.8821 0.0018* Chemotherapy, radiation and surgery 2.6044 2.0091 3.3764 <.0001* No Info 0.8153 0.5920 1.1228 0.2111 No treatment 0.8983 0.4569 1.7665 0.7561 Radiation only 1.1379 0.7181 1.8033 0.5821 Radiation and surgery 1.5065 1.1590 1.9580 0.0022* Surgery only R 1.0000 1.0000 1.0000 . Relapse Status Relapse <5yrs 1.5371 1.3320 1.7738 <.0001* Relapse >5yrs 1.9303 1.3184 2.8267 0.0007* No relapse R 1.0000 1.0000 1.0000 . Sex Female 1.0511 0.9255 1.1936 0.4432 Male R 1.0000 1.0000 1.0000 . SES Unknown 0.9771 0.6780 1.4082 0.9012 1 (Lowest) 0.8948 0.7312 1.0949 0.2806 2 0.8273 0.6742 1.0153 0.0695 3 1.1427 0.9368 1.3939 0.1883 4 1.2949 1.0565 1.5869 0.0128* 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority 148 Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Unknown 0.7489 0.6301 0.8899 0.0010* Fraser Health 0.6885 0.5357 0.8846 0.0035* Interior Health 0.5736 0.4272 0.7702 0.0002* Northern Health 0.5781 0.3553 0.9404 0.0273* Vancouver Island Health 0.7714 0.6234 0.9548 0.0171* Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 0.8142 0.6499 1.0202 0.0741 Rural 0.7829 0.6392 0.9589 0.0180* Small community 0.7163 0.5768 0.8897 0.0026* Metropolitan R 1.0000 1.0000 1.0000 . Age at Diagnosis 0.9575 0.9451 0.9702 <.0001* Primary Care Visits Per Year 1.1119 1.0949 1.1292 <.0001* R Reference category 149 Table CA Case-Only, Overall: Non-Physician Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 0.6022 0.3658 0.9917 0.0463* ICCC Diagnosis VIII (Malignant Bone Tumors) 0.6962 0.4169 1.1628 0.1665 III (CNS Cancers) 1.2411 0.8069 1.9090 0.3254 XI (Other Malignant Epithelial 1.3237 0.7625 2.2979 0.3191 Neoplasms) X (Germ Cell Tumors) 1.0591 0.6602 1.6989 0.8119 VII (Hepatic Tumors) 1.4572 0.3805 5.5812 0.5827 II (Lymphomas) 1.1859 0.8388 1.6765 0.3346 IV (Neuroblastomas) 1.0435 0.5756 1.8919 0.8883 XII (Other Unspecified) 0.0000 0.0000 0.0000 0.9969 VI (Renal Tumors) 0.8367 0.4865 1.4391 0.5194 V (Retinoblastomas) 0.3016 0.1422 0.6401 0.0018* IX (Soft Tissue Sarcomas) 0.7018 0.4337 1.1355 0.1492 I (Leukemias)R 1.0000 1.0000 1.0000 . Treatment Modality Chemotherapy only 0.6378 0.4175 0.9743 0.0375* Chemotherapy and radiation 0.6932 0.4628 1.0382 0.0754 Chemotherapy and surgery 0.8513 0.5942 1.2196 0.3800 Chemotherapy, radiation and surgery 0.8465 0.5736 1.2492 0.4012 No Info 0.6217 0.3976 0.9720 0.0371* No treatment 0.2440 0.0826 0.7212 0.0107* Radiation only 0.6875 0.3596 1.3146 0.2572 Radiation and surgery 0.7454 0.5137 1.0817 0.1220 Surgery only R 1.0000 1.0000 1.0000 . Relapse Status Relapse <5yrs 1.0626 0.8596 1.3135 0.5747 Relapse >5yrs 0.9375 0.5633 1.5606 0.8042 No relapse R 1.0626 0.8596 1.3135 0.5747 Sex Female 1.4031 1.1698 1.6830 0.0003* Male R 1.0000 1.0000 1.0000 . SES Unknown 0.5477 0.3037 0.9877 0.0454* 1 (Lowest) 0.7426 0.5544 0.9947 0.0460* 2 0.8692 0.6512 1.1600 0.3410 3 0.9683 0.7254 1.2925 0.8270 4 1.1164 0.8344 1.4935 0.4585 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority 150 Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Unknown 1.2443 0.5697 2.7177 0.5835 Fraser Health 0.9867 0.7660 1.2710 0.9173 Interior Health 0.9437 0.6594 1.3509 0.7519 Northern Health 0.8767 0.5752 1.3362 0.5404 Vancouver Island Health 1.0463 0.7701 1.4215 0.7723 Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 1.1219 0.8242 1.5270 0.4648 Rural 0.8770 0.6549 1.1746 0.3786 Small community 0.8939 0.6464 1.2360 0.4973 Metropolitan R 1.0000 1.0000 1.0000 . Age at Diagnosis 1.0367 1.0171 1.0565 0.0002* Primary Care Visits Per Year 1.1035 1.0778 1.1299 <.0001* R Reference category 151 Table C.5 Case-Control, Overall: Total Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 0.6471 0.5561 0.7531 <.0001* Case-Control Case 3.8915 3.4182 4.4300 <.0001* Control R 1.0000 1.0000 1.0000 . Sex Female 1.5784 1.4630 1.7027 <.0001* Male R 1.0000 1.0000 1.0000 . SES Unknown 1.0807 0.8734 1.3372 0.4753 1 (Lowest) 1.0968 0.9718 1.2379 0.1345 2 1.1142 0.9871 1.2576 0.0802 3 1.0454 0.9226 1.1847 0.4862 4 1.0767 0.9498 1.2205 0.2481 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority Unknown 0.7561 0.6018 0.9500 0.0164* Fraser Health 0.9551 0.8612 1.0594 0.3854 Interior Health 1.0145 0.8673 1.1867 0.8574 Northern Health 0.8422 0.7052 1.0059 0.0581 Vancouver Island Health 0.9295 0.8147 1.0605 0.2773 Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 0.8675 0.7497 1.0039 0.0565 Rural 0.7472 0.6591 0.8471 <.0001* Small community 0.7671 0.6617 0.8893 0.0004* Metropolitan R 1.0000 1.0000 1.0000 . Attained Age 1.0597 1.0545 1.0650 <.0001* R Reference category 152 Table C.6 Case-Control, Overall: Primary Care Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 0.4586 0.3955 0.5318 <.0001* Case-Control Case 2.9353 2.5891 3.3281 <.0001* Control R 1.0000 1.0000 1.0000 . Sex Female 1.6916 1.5711 1.8216 <.0001* Male R 1.0000 1.0000 1.0000 . SES Unknown 1.1148 0.9062 1.3715 0.3040 1 (Lowest) 1.1427 1.0155 1.2858 0.0267* 2 1.1745 1.0438 1.3214 0.0075* 3 1.0785 0.9549 1.2181 0.2231 4 1.0580 0.9365 1.1953 0.3650 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority Unknown 0.7667 0.6139 0.9574 0.0191* Fraser Health 0.9932 0.8979 1.0987 0.8946 Interior Health 1.0242 0.8785 1.1940 0.7600 Northern Health 0.9300 0.7818 1.1061 0.4118 Vancouver Island Health 0.9359 0.8225 1.0650 0.3152 Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 0.8789 0.7620 1.0138 0.0764 Rural 0.7635 0.6752 0.8635 <.0001* Small community 0.8053 0.6972 0.9303 0.0033* Metropolitan R 1.0000 1.0000 1.0000 . Attained Age 1.0531 1.0479 1.0581 <.0001* R Reference category 153 Table C.7 Case-Control, Overall: Specialist Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 0.1233 0.1071 0.1420 <.0001* Case-Control Case 6.4199 5.7351 7.1865 <.0001* Control R 1.0000 1.0000 1.0000 . Sex Female 0.7977 0.7440 0.8552 <.0001* Male R 1.0000 1.0000 1.0000 . SES Unknown 1.0949 0.8966 1.3371 0.3741 1 (Lowest) 0.9063 0.8125 1.0110 0.0775 2 0.8297 0.7435 0.9259 0.0009* 3 0.8947 0.7984 1.0025 0.0553 4 0.9173 0.8184 1.0281 0.1378 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority Unknown 0.6132 0.4972 0.7562 <.0001* Fraser Health 0.7037 0.6405 0.7731 <.0001* Interior Health 0.7163 0.6184 0.8295 <.0001* Northern Health 0.5451 0.4616 0.6437 <.0001* Vancouver Island Health 0.7183 0.6362 0.8110 <.0001* Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 0.7390 0.6450 0.8468 <.0001* Rural 0.6198 0.5515 0.6965 <.0001* Small community 0.5418 0.4721 0.6217 <.0001* Metropolitan R 0.7390 0.6450 0.8468 <.0001* Attained Age 1.0084 1.0038 1.0132 0.0004* Primary Care Visits per Year 1.5512 1.5252 1.5774 <.0001* R Reference category 154 Table C.8 Case-Control, Overall: Non-Phician Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 0.0517 0.0435 0.0615 <.0001* Case-Control Case 2.0245 1.7683 2.3180 <.0001* Control R 1.0000 1.0000 1.0000 . Sex Female 0.8824 0.8140 0.9565 0.0024* Male R 1.0000 1.0000 1.0000 . SES Unknown 1.0206 0.8151 1.2780 0.8587 1 (Lowest) 0.8630 0.7590 0.9814 0.0246* 2 0.7835 0.6896 0.8902 0.0002* 3 0.8527 0.7468 0.9738 0.0186* 4 1.0236 0.8961 1.1693 0.7311 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority Unknown 0.5746 0.4550 0.7256 <.0001* Fraser Health 0.8966 0.8025 1.0018 0.0538 Interior Health 0.8971 0.7619 1.0564 0.1930 Northern Health 0.6289 0.5206 0.7598 <.0001* Vancouver Island Health 0.9274 0.8096 1.0623 0.2763 Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 1.1282 0.9698 1.3125 0.1182 Rural 0.8971 0.7888 1.0204 0.0985 Small community 0.9116 0.7774 1.0690 0.2547 Metropolitan R 0.7390 0.6450 0.8468 <.0001* Attained Age 1.0539 1.0480 1.0597 <.0001* Primary Care Visits per Year 1.4829 1.4528 1.5135 <.0001* R Reference category 155 Table C.9 Case-Only, Biannual By Age: Total Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 6.9845 4.8254 10.1087 <.0001* ICCC Diagnosis VIII (Malignant Bone Tumors) 1.0271 0.6874 1.5346 0.8962 III (CNS Cancers) 1.3412 0.9748 1.8456 0.0714 XI (Other Malignant Epithelial 1.3592 0.9411 1.9631 0.1018 Neoplasms) X (Germ Cell Tumors) 1.1650 0.7925 1.7126 0.4372 VII (Hepatic Tumors) 0.9869 0.6137 1.5871 0.9567 II (Lymphomas) 0.9422 0.7444 1.1927 0.6209 IV (Neuroblastomas) 1.3925 0.9139 2.1215 0.1233 XII (Other Unspecified) 1.9024 1.2817 2.8233 0.0014* VI (Renal Tumors) 0.8746 0.6188 1.2360 0.4477 V (Retinoblastomas) 0.8237 0.5816 1.1667 0.2749 IX (Soft Tissue Sarcomas) 1.0189 0.7790 1.3327 0.8916 I (Leukemias)R 1.0000 1.0000 1.0000 . Treatment Modality Chemotherapy only 1.2014 0.8772 1.6454 0.2528 Chemotherapy and radiation 1.3564 1.0109 1.8198 0.0421* Chemotherapy and surgery 1.2331 0.9282 1.6379 0.1483 Chemotherapy, radiation and surgery 1.4780 1.1268 1.9389 0.0048* No Info 0.9272 0.7030 1.2227 0.5921 No treatment 0.8456 0.4533 1.5776 0.5981 Radiation only 0.9117 0.6405 1.2980 0.6082 Radiation and surgery 1.1287 0.8994 1.4167 0.2959 Surgery only R 1.0000 1.0000 1.0000 . Sex Female 1.5563 1.3811 1.7537 <.0001* Male R 1.0000 1.0000 1.0000 . SES Unknown 0.9137 0.6994 1.1936 0.5080 1 (Lowest) 1.0119 0.8746 1.1707 0.8739 2 1.0749 0.9162 1.2609 0.3755 3 1.0484 0.9081 1.2103 0.5191 4 1.0605 0.9114 1.2340 0.4473 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority Unknown 1.0766 0.8454 1.3709 0.5499 Fraser Health 1.0366 0.8775 1.2245 0.6729 Interior Health 0.9665 0.7819 1.1947 0.7528 Northern Health 0.9204 0.7311 1.1586 0.4798 156 Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Vancouver Island Health 0.9243 0.7688 1.1114 0.4028 Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 1.0427 0.8892 1.2227 0.6067 Rural 1.0550 0.8969 1.2410 0.5183 Small community 1.0627 0.8842 1.2771 0.5171 Metropolitan R 1.0000 1.0000 1.0000 . Age at Diagnosis 1.0114 0.9941 1.0288 0.1972 Age at Start of 2-year Interval 1.0116 0.9976 1.0258 0.1044 Relapse Status Relapse <5yrs 1.1263 0.9814 1.2925 0.0905 Relapse >5yrs 0.5887 0.3924 0.8832 0.0105* No relapse R 1.0000 1.0000 1.0000 . R Reference category 157 Table C.10 Case-Only, Biannual by Age: Primary Care Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 3.9079 2.7243 5.6058 <.0001* ICCC Diagnosis VIII (Malignant Bone Tumors) 0.8210 0.5527 1.2196 0.3287 III (CNS Cancers) 1.1428 0.8324 1.5689 0.4092 XI (Other Malignant Epithelial 1.0609 0.7456 1.5094 0.7427 Neoplasms) X (Germ Cell Tumors) 0.9698 0.6939 1.3554 0.8575 VII (Hepatic Tumors) 0.7260 0.3598 1.4646 0.3712 II (Lymphomas) 0.9276 0.7231 1.1899 0.5542 IV (Neuroblastomas) 1.1787 0.7664 1.8128 0.4540 XII (Other Unspecified) 1.5561 1.0623 2.2796 0.0232* VI (Renal Tumors) 0.8690 0.6274 1.2037 0.3984 V (Retinoblastomas) 0.9457 0.6465 1.3835 0.7737 IX (Soft Tissue Sarcomas) 0.9882 0.7296 1.3384 0.9388 I (Leukemias)R 1.0000 1.0000 1.0000 . Treatment Modality Chemotherapy only 1.1118 0.8028 1.5399 0.5235 Chemotherapy and radiation 1.1871 0.8971 1.5708 0.2300 Chemotherapy and surgery 1.2730 0.9683 1.6735 0.0837 Chemotherapy, radiation and surgery 1.2417 1.0162 1.5174 0.0342* No Info 1.0360 0.8032 1.3362 0.7853 No treatment 0.8168 0.4427 1.5068 0.5171 Radiation only 0.9019 0.6460 1.2590 0.5439 Radiation and surgery 1.0206 0.8332 1.2502 0.8435 Surgery only R 1.0000 1.0000 1.0000 . Sex Female 1.5247 1.3645 1.7035 <.0001* Male R 1.0000 1.0000 1.0000 . SES Unknown 0.9787 0.7458 1.2844 0.8767 1 (Lowest) 0.9678 0.8409 1.1138 0.6479 2 1.1310 0.9563 1.3378 0.1504 3 1.0419 0.9134 1.1883 0.5417 4 1.0148 0.8861 1.1622 0.8318 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority Unknown 1.2512 0.9824 1.5936 0.0694 Fraser Health 1.1022 0.9417 1.2902 0.2257 Interior Health 1.0560 0.8642 1.2906 0.5939 Northern Health 1.1158 0.8933 1.3936 0.3342 158 Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Vancouver Island Health 1.0578 0.8886 1.2591 0.5276 Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 1.1138 0.9334 1.3291 0.2319 Rural 1.1386 0.9604 1.3499 0.1350 Small community 1.1138 0.9150 1.3557 0.2825 Metropolitan R 1.1138 0.9334 1.3291 0.2319 Age at Diagnosis 1.0141 0.9961 1.0325 0.1247 Age at Start of 2-year Interval 1.0146 1.0007 1.0287 0.0396* Relapse Status Relapse <5yrs 1.0277 0.8989 1.1749 0.6895 Relapse >5yrs 0.5840 0.3839 0.8883 0.0120* No relapse R 1.0000 1.0000 1.0000 . R Reference category 159 Table C.11 Case-Only, Biannual by Age: Specialist Visits Parameter^ Relative^95% Confidence^Pr > ChiSq Risk^Interval Intercept 2.5710 1.4673 4.5051 0.0010* ICCC Diagnosis VIII (Malignant Bone Tumors) 1.7551 1.0752 2.8651 0.0244* III (CNS Cancers) 1.8855 1.2726 2.7935 0.0016* XI (Other Malignant Epithelial 1.4000 0.8582 2.2837 0.1778 Neoplasms) X (Germ Cell Tumors) 2.0522 0.9299 4.5290 0.0751 VII (Hepatic Tumors) 1.7362 0.8573 3.5163 0.1254 II (Lymphomas) 0.8932 0.6094 1.3093 0.5628 IV (Neuroblastomas) 1.6251 0.8978 2.9414 0.1087 XII (Other Unspecified) 3.3750 1.6695 6.8230 0.0007* VI (Renal Tumors) 0.8600 0.4985 1.4835 0.5877 V (Retinoblastomas) 0.9326 0.5635 1.5434 0.7861 IX (Soft Tissue Sarcomas) 1.1899 0.8198 1.7272 0.3604 I (Leukemias)R 1.0000 1.0000 1.0000 . Treatment Modality Chemotherapy only 1.7405 1.1687 2.5922 0.0064* Chemotherapy and radiation 2.2571 1.4427 3.5318 0.0004* Chemotherapy and surgery 1.3486 0.7821 2.3257 0.2820 Chemotherapy, radiation and surgery 2.3205 1.3360 4.0301 0.0028* No Info 0.9801 0.6574 1.4611 0.9213 No treatment 1.2336 0.6529 2.3305 0.5179 Radiation only 1.0708 0.7500 1.5288 0.7067 Radiation and surgery 1.7843 1.2645 2.5176 0.0010* Surgery only R 1.0000 1.0000 1.0000 . Sex Female 1.2119 0.9835 1.4935 0.0712 Male R 1.0000 1.0000 1.0000 . SES Unknown 1.0091 0.7815 1.3030 0.9445 1 (Lowest) 1.0608 0.8564 1.3140 0.5888 2 1.0472 0.8465 1.2954 0.6713 3 1.0807 0.8598 1.3583 0.5060 4 1.3160 0.9863 1.7559 0.0620 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority Unknown 0.8932 0.5690 1.4017 0.6231 Fraser Health 0.7848 0.6143 1.0028 0.0526 Interior Health 0.6631 0.4555 0.9651 0.0319* Northern Health 0.5033 0.3472 0.7295 0.0003* 160 Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Vancouver Island Health 1.0578 0.8886 1.2591 0.5276 Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 0.8564 0.6507 1.1272 0.2687 Rural 0.7989 0.6238 1.0232 0.0753 Small community 0.8792 0.6475 1.1940 0.4097 Metropolitan R 1.1138 0.9334 1.3291 0.2319 Age at Diagnosis 1.0109 0.9757 1.0472 0.5499 Age at Start of 2-year Interval 0.9550 0.9328 0.9778 0.0001* Relapse Status Relapse <5yrs 1.4131 1.0803 1.8484 0.0116* Relapse >5yrs 0.5827 0.3921 0.8659 0.0075* No relapse R 1.0000 1.0000 1.0000 . Primary Care Visits During Two-Year 1.0493 1.0432 1.0554 <.0001* Interval R Reference category 161 Table C.12 Case-Only, Biannual by Age: Non-Physician Visits Parameter^ Relative 95% Confidence Pr > ChiSq Risk^Interval Intercept 0.4609 0.1983 1.0711 0.0718 ICCC Diagnosis VIII (Malignant Bone Tumors) 1.1920 0.6775 2.0970 0.5423 HI (CNS Cancers) 1.2094 0.6834 2.1402 0.5140 XI (Other Malignant Epithelial 1.2707 0.7075 2.2826 0.4226 Neoplasms) X (Germ Cell Tumors) 1.0505 0.5013 2.2014 0.8961 VII (Hepatic Tumors) 1.7056 0.7860 3.7014 0.1768 II (Lymphomas) 1.1270 0.7516 1.6898 0.5629 IV (Neuroblastomas) 1.5716 0.6506 3.7970 0.3151 XII (Other Unspecified) 1.2857 0.6582 2.5113 0.4619 VI (Renal Tumors) 0.2730 0.1010 0.7376 0.0105* V (Retinoblastomas) 0.8487 0.4841 1.4877 0.5666 IX (Soft Tissue Sarcomas) 1.1920 0.6775 2.0970 0.5423 I (Leukemias)R 1.0000 1.0000 1.0000 . Treatment Modality Chemotherapy only 0.8243 0.4755 1.4289 0.4912 Chemotherapy and radiation 0.7908 0.4498 1.3903 0.4149 Chemotherapy and surgery 0.6873 0.4457 1.0598 0.0897 Chemotherapy, radiation and surgery 0.6754 0.4187 1.0894 0.1076 No Info 0.7888 0.5082 1.2241 0.2899 No treatment 0.2386 0.1152 0.4942 0.0001* Radiation only 0.6278 0.3324 1.1858 0.1514 Radiation and surgery 0.7503 0.4842 1.1626 0.1986 Surgery only R 1.0000 1.0000 1.0000 . Sex Female 1.4683 1.1575 1.8625 0.0015* Male R 1.0000 1.0000 1.0000 . SES Unknown 0.6866 0.4562 1.0334 0.0715 1 (Lowest) 1.0358 0.7482 1.4342 0.8319 2 0.7778 0.5581 1.0840 0.1379 3 1.0656 0.7884 1.4401 0.6795 4 0.9373 0.6919 1.2696 0.6756 5 (Highest) R 1.0000 1.0000 1.0000 . Health Authority Unknown 0.3948 0.1583 0.9843 0.0462* Fraser Health 1.0413 0.7825 1.3858 0.7810 Interior Health 0.9513 0.6478 1.3972 0.7993 Northern Health 0.7568 0.4800 1.1932 0.2303 162 Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Vancouver Island Health 0.8462 0.5834 1.2275 0.3790 Vancouver Coastal Health R 1.0000 1.0000 1.0000 . Ruralness Large community 1.1852 0.8551 1.6426 0.3076 Rural 0.9696 0.7238 1.2989 0.8359 Small community 1.0779 0.7996 1.4532 0.6226 Metropolitan R 1.1138 0.9334 1.3291 0.2319 Age at Diagnosis 0.9869 0.9517 1.0234 0.4753 Age at Start of 2-year Interval 1.0631 1.0313 1.0959 <.0001* Relapse Status Relapse <5yrs 1.0718 0.8036 1.4292 0.6372 Relapse >5yrs 0.5932 0.3173 1.1090 0.1019 No relapse R 1.0000 1.0000 1.0000 . Primary Care Visits During Two-Year 1.0606 1.0540 1.0673 <.0001* Interval R Reference category 163 Table C.13 Case-Control, Biannual By Age: Total Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 1.5928 1.4382 1.7640 <.0001* Case-Control Case 6.2658 5.3010 7.4053 <.0001* Control 1.0000 1.0000 1.0000 . Sex F 1.6364 1.5408 1.7381 <.0001* M 1.0000 1.0000 1.0000 . SES Unknown 1.0178 0.9341 1.1089 0.6873 1 (Lowest) 1.0241 0.9646 1.0872 0.4354 2 0.9794 0.9224 1.0401 0.4980 3 1.0059 0.9503 1.0649 0.8383 4 1.0184 0.9620 1.0779 0.5308 5 (Highest) 1.0000 1.0000 1.0000 . Health Authority Fraser Health 0.9626 0.8999 1.0297 0.2677 Interior Health 0.9413 0.8556 1.0355 0.2140 Northern Health 0.7906 0.7079 0.8829 <.0001* Unknown Health 0.9288 0.8318 1.0371 0.1890 Vancouver Island Health 0.9656 0.8910 1.0464 0.3929 Vancouver Coastal Health 1.0000 1.0000 1.0000 . Ruralness Large community 0.9881 0.9031 1.0809 0.7931 Rural 0.9436 0.8767 1.0156 0.1216 Small community 0.9570 0.8807 1.0398 0.2986 Metropolitan 1.0000 1.0000 1.0000 . Age at Start of 2-Yearr Interval 1.0624 1.0585 1.0662 <.0001* Interaction Term Age*Case (Case) 0.9579 0.9493 0.9667 <.0001* Interaction Term Age*Case (Control) 1.0000 1.0000 1.0000 . R Reference category 164 Table C.14 Case-Control, Biannual By Age: Primary Care Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 1.1667 1.0537 1.2919 0.0030* Case-Control Case 4.3894 3.7430 5.1474 <.0001* Control 1.0000 1.0000 1.0000 . Sex F 1.7454 1.6439 1.8532 <.0001* M 1.0000 1.0000 1.0000 . SES Unknown 0.9908 0.9120 1.0765 0.8272 1 (Lowest) 1.0488 0.9893 1.1118 0.1104 2 1.0228 0.9660 1.0830 0.4398 3 1.0248 0.9673 1.0857 0.4051 4 1.0098 0.9576 1.0648 0.7183 5 (Highest) 1.0000 1.0000 1.0000 . Health Authority Fraser Health 0.9943 0.9278 1.0656 0.8718 Interior Health 0.9643 0.8751 1.0625 0.4618 Northern Health 0.8711 0.7813 0.9713 0.0130* Unknown Health 0.9717 0.8727 1.0821 0.6016 Vancouver Island Health 0.9946 0.9195 1.0757 0.8919 Vancouver Coastal Health 1.0000 1.0000 1.0000 . Ruralness Large community 0.9983 0.9048 1.1014 0.9731 Rural 0.9847 0.9096 1.0659 0.7029 Small community 0.9931 0.9093 1.0847 0.8774 Metropolitan 1.0000 1.0000 1.0000 . Age at Start of 2-Year Interval 1.0511 1.0472 1.0550 <.0001* Interaction Term Age*Case (Case) 0.9680 0.9597 0.9764 <.0001* Interaction Term Age*Case (Control) 1.0000 1.0000 1.0000 . R Reference category 165 Table C.15 Case-Control, Biannual By Age: Specialist Visits Parameter^ Relative 95% Confidence^Pr > ChiSq Risk^Interval Intercept 0.3096 0.2404 0.3987 <.0001* Case-Control Case 20.0334 15.3482 26.1487 <.0001* Control 1.0000 1.0000 1.0000 . Sex F 1.0877 0.9688 1.2214 0.1546 M 1.0000 1.0000 1.0000 . SES Unknown 1.1333 0.9434 1.3614 0.1812 1 (Lowest) 0.9716 0.8373 1.1275 0.7045 2 0.9331 0.8151 1.0683 0.3158 3 0.8989 0.7975 1.0132 0.0809 4 1.0241 0.8901 1.1782 0.7393 5 (Highest) 1.0000 1.0000 1.0000 . Health Authority Fraser Health 0.7351 0.6238 0.8662 0.0002* Interior Health 0.7495 0.6197 0.9065 0.0030* Northern Health 0.4991 0.3715 0.6706 <.0001* Unknown Health 0.7772 0.5728 1.0547 0.1057 Vancouver Island Health 0.7047 0.5960 0.8333 <.0001* Vancouver Coastal Health 1.0000 1.0000 1.0000 . Ruralness Large community 0.8631 0.7301 1.0204 0.0849 Rural 0.6618 0.5769 0.7591 <.0001* Small community 0.6062 0.5263 0.6982 <.0001* Metropolitan 1.0000 1.0000 1.0000 . Age at Start of 2-Year Interval 1.0206 1.0124 1.0289 <.0001* Interaction Term Age*Case (Case) 0.9239 0.9118 0.9359 <.0001* Interaction Term Age*Case 1.0000 1.0000 1.0000 . (Control) Primary Care Visits During 2-Year 1.1138 1.1089 1.1187 <.0001* Interval R Reference category 166 Table C.16 Case-Control, Biannual By Age: Non-Physician Visits Parameter Relative Risk 95% Confidence Interval Pr > ChiSq Intercept 0.1064 0.0908 0.1248 <.0001* Case-Control Case 3.2046 2.1658 4.7417 <.0001* Control 1.0000 1.0000 1.0000 . Sex F 1.1945 1.0747 1.3278 0.0010* M 1.0000 1.0000 1.0000 SES Unknown 1.0868 0.8783 1.3449 0.4438 1 (Lowest) 0.9326 0.8282 1.0500 0.2485 2 0.9256 0.8189 1.0461 0.2160 3 1.0720 0.9433 1.2181 0.2868 4 1.0247 0.9135 1.1496 0.6770 5 (Highest) 1.0000 1.0000 1.0000 . Health Authority Fraser Health 0.9137 0.8133 1.0268 0.1296 Interior Health 1.0136 0.8395 1.2238 0.8880 Northern Health 0.7036 0.5708 0.8674 0.0010* Unknown Health 0.8228 0.6197 1.0924 0.1775 Vancouver Island Health 1.0488 0.8712 1.2625 0.6150 Vancouver Coastal Health 1.0000 1.0000 1.0000 . Ruralness Large community 1.0268 0.8629 1.2218 0.7660 Rural 0.9260 0.7919 1.0827 0.3351 Small community 0.9204 0.7762 1.0917 0.3410 Metropolitan 1.0000 1.0000 1.0000 . Age at Start of 2-Year Interval 1.0820 1.0745 1.0894 <.0001* Interaction Term Age*Case (Case) 0.9690 0.9508 0.9875 0.0011* Interaction Term Age*Case 1.0000 1.0000 1.0000 . (Control) Primary Care Visits During 2-Year 1.1123 1.1085 1.1161 <.0001* Interval R Reference category 167 APPENDIX D: REGRESSION MODELS - CONTINUITY OF CARE Table D.1 Survivor-Only, Overall: COC Score Parameter Estimate 95% Confidence Interval Pr > ChiSq Intercept 0.374 0.2855 0.4625 <.0001* ICCC Diagnosis VIII (Malignant Bone Tumors) 0.025 -0.0577 0.1076 0.5536 III (CNS Cancers) 0.0599 -0.0082 0.1279 0.0846 XI (Other Malignant Epithelial 0.0956 0.0106 0.1805 0.0275* Neoplasms) X (Germ Cell Tumors) 0.0902 0.0135 0.1669 0.0211* VII (Hepatic Tumors) 0.2472 0.0028 0.4916 0.0474* II (Lymphomas) 0.0199 -0.0336 0.0734 0.4665 IV (Neuroblastomas) 0.0313 -0.0627 0.1252 0.5141 XII (Other Unspecified) -0.0555 -0.5333 0.4223 0.8199 VI (Renal Tumors) 0.0371 -0.0481 0.1223 0.3934 V (Retinoblastomas) 0.0684 -0.0431 0.1799 0.2291 IX (Soft Tissue Sarcomas) 0.0518 -0.0247 0.1282 0.1843 I (Leukemias)R 0 0 0 . Treatment Modality Chemotherapy only 0.0775 0.0089 0.1461 0.0267 Chemotherapy and radiation 0.0236 -0.0414 0.0885 0.4769 Chemotherapy and surgery 0.027 -0.0308 0.0847 0.3599 Chemotherapy, radiation and surgery 0.0638 0.0026 0.1251 0.0411 No Info -0.0335 -0.1036 0.0366 0.349 No treatment 0.0973 -0.0641 0.2587 0.2374 Radiation only 0.0849 -0.0195 0.1894 0.1111 Radiation and surgery 0.0649 0.0056 0.1241 0.0319 Surgery only R 0 0 0 . Relapse Status Relapse <5yrs 0.0097 -0.0245 0.044 0.5778 Relapse >5yrs 0.0221 -0.0581 0.1023 0.589 No relapse R 0 0 0 . Sex Female -0.0078 -0.036 0.0204 0.5874 Male 0 0 0 . SES 1 (Lowest) 0.0283 -0.0184 0.0749 0.2349 2 0.034 -0.0115 0.0794 0.1429 3 0.0092 -0.036 0.0543 0.6899 4 0.0051 -0.0402 0.0504 0.8246 Unknown -0.0475 -0.1352 0.0402 0.2883 168 Parameter Estimate 95% Confidence Interval Pr > ChiSq Highest R 0 0 0 . Health Authority Fraser Health -0.0627 -0.1026 -0.0228 0.0021* Interior Health -0.0569 -0.1146 0.0009 0.0535 Northern Health -0.0412 -0.1077 0.0254 0.2253 Unknown Health -0.0028 -0.1176 0.112 0.9614 Vancouver Island Health 0.0101 -0.0392 0.0593 0.6883 Vancouver Coastal Health R 0 0 0 . Ruralness Large community -0.0329 -0.0843 0.0185 0.2095 Rural -0.0293 -0.0748 0.0163 0.2077 Small community -0.0672 -0.1177 -0.0168 0.009* Metropolitan R 0 0 0 . Age at Diagnosis 0.001 -0.002 0.004 0.511 Length of Follow-Up -0.0101 -0.0139 -0.0063 <.0001* R Reference Category 169 Table D.2 Survivor-Only, Overall: UPC Score Parameter Estimate 95% Confidence Interval Pr > ChiSq Intercept 0.5613 0.4818 0.6408 <.0001* ICCC Diagnosis VIII (Malignant Bone Tumors) 0.0037 -0.0706 0.0779 0.923 III (CNS Cancers) 0.0587 -0.0024 0.1198 0.0598 XI (Other Malignant Epithelial 0.085 0.0087 0.1613 0.0291* Neoplasms) X (Germ Cell Tumors) 0.0796 0.0107 0.1485 0.0236* VII (Hepatic Tumors) 0.2248 0.0053 0.4444 0.0448* II (Lymphomas) 0.0168 -0.0312 0.0649 0.493 IV (Neuroblastomas) 0.0496 -0.0348 0.134 0.2493 XII (Other Unspecified) -0.1024 -0.5316 0.3269 0.6403 VI (Renal Tumors) 0.0262 -0.0504 0.1028 0.5022 V (Retinoblastomas) 0.0581 -0.0421 0.1583 0.2557 IX (Soft Tissue Sarcomas) 0.048 -0.0207 0.1166 0.1709 I (Leukemias)R 0 0 0 . Treatment Modality Chemotherapy only 0.0803 0.0187 0.1419 0.0107* Chemotherapy and radiation 0.0342 -0.0241 0.0926 0.25 Chemotherapy and surgery 0.0419 -0.01 0.0938 0.1133 Chemotherapy, radiation and surgery 0.0788 0.0238 0.1339 0.005* No Info -0.0291 -0.0921 0.0339 0.3651 No treatment 0.111 -0.034 0.256 0.1335 Radiation only 0.0801 -0.0138 0.174 0.0946 Radiation and surgery 0.0663 0.013 0.1195 0.0147* Surgery only R 0 0 0 . Relapse Status Relapse <5yrs 0.0027 -0.0281 0.0335 0.8635 Relapse >5yrs 0.0178 -0.0542 0.0899 0.6281 No relapse R 0 0 0 . Sex Female -0.0128 -0.0382 0.0125 0.3202 Male 0 0 0 . SES 1 (Lowest) 0.0182 -0.0237 0.0601 0.3949 2 0.0263 -0.0145 0.0671 0.2063 3 0.0123 -0.0283 0.0528 0.5538 4 0.0085 -0.0322 0.0492 0.6817 Unknown -0.0437 -0.1225 0.0351 0.277 Highest R 0 0 0 . 170 Parameter Estimate 95% Confidence Interval Pr > ChiSq Health Authority Fraser Health -0.0413 -0.0772 -0.0055 0.0238* Interior Health -0.0467 -0.0985 0.0052 0.0779 Northern Health -0.0388 -0.0986 0.021 0.203 Unknown Health 0.0108 -0.0924 0.1139 0.8378 Vancouver Island Health 0.0106 -0.0336 0.0549 0.6372 Vancouver Coastal Health R 0 0 0 . Ruralness Large community -0.0198 -0.066 0.0263 0.3999 Rural -0.0234 -0.0643 0.0176 0.2633 Small community -0.0464 -0.0917 -0.0011 0.0447* Metropolitan R 0 0 0 . Age at Diagnosis 0.001 -0.0016 0.0037 0.4495 Length of Follow-Up -0.0117 -0.0151 -0.0083 <.0001* R Reference Category 171 Table D.3 Case-Control, Overall: COC Score Parameter Estimate 95% Confidence Interval Pr > ChiSq Intercept 0.4383 0.4184 0.4582 <.0001* Case-Control Case 0.0214 0.0061 0.0367 0.006* Control R 0 0 0 . Sex Female 0.0073 -0.0039 0.0185 0.2029 Male R 0 0 0 . SES Unknown -0.0086 -0.0425 0.0253 0.6194 1(Lowest) 0.0114 -0.0068 0.0296 0.2183 2 0.0139 -0.0043 0.032 0.1349 3 0.0101 -0.0089 0.0291 0.2996 4 0.0011 -0.018 0.0202 0.9126 Highest R 0 0 0 . Health Authority Unknown -0.0068 -0.0449 0.0314 0.7275 Fraser Health -0.0523 -0.0674 -0.0372 <.0001* Interior Health -0.0659 -0.09 -0.0418 <.0001* Northern Health -0.0635 -0.0915 -0.0355 <.0001* Vancouver Island Health -0.0629 -0.0827 -0.043 <.0001* Vancouver Coastal Health R 0 0 0 . Ruralness Large community -0.0247 -0.0471 -0.0024 0.0301* Rural -0.043 -0.0627 -0.0233 <.0001* Small community -0.0411 -0.0641 -0.0181 0.0005* Metropolitan R 0 0 0 . Length of Follow-Up -0.0082 -0.0095 -0.007 <.0001* R Reference Category 172 TableC.4 Case-Control, Overall: UPC Score Parameter Estimate 95% Confidence Interval Pr > ChiSq Intercept 0.6388 0.6208 0.6569 <.0001* Case-Control Case 0.017 0.0032 0.0309 0.0162* Control R 0 0 0 . Sex Female -0.0033 -0.0134 0.0069 0.53 Male R 0 0 0 . SES Unknown -0.006 -0.0367 0.0248 0.7043 1(Lowest) 0.0106 -0.0059 0.027 0.2091 2 0.0092 -0.0073 0.0257 0.2743 3 0.0119 -0.0054 0.0291 0.1776 4 0.0005 -0.0168 0.0179 0.9516 Highest R 0 0 0 . Health Authority Unknown 0.0072 -0.0274 0.0418 0.6849 Fraser Health -0.0419 -0.0555 -0.0282 <.0001* Interior Health -0.0514 -0.0733 -0.0296 <.0001* Northern Health -0.0528 -0.0782 -0.0274 <.0001* Vancouver Island Health -0.0465 -0.0645 -0.0285 <.0001* Vancouver Coastal Health R 0 0 0 . Ruralness Large community -0.0173 -0.0376 0.003 0.0946 Rural -0.0406 -0.0584 -0.0227 <.0001* Small community -0.0302 -0.0511 -0.0093 0.0045* Metropolitan R 0 0 0 . Length of Follow-Up -0.0105 -0.0116 -0.0093 <.0001* R Reference Category 173 Table D.5 Survivor-Only, Biannual by Age: COC Score Parameter Estimate 95% Confidence Interval Pr > ChiSq Intercept 0.4652 0.3706 0.5599 <.0001* ICCC Diagnosis VIII (Malignant Bone Tumors) -0.0341 -0.1193 0.0511 0.4323 III (CNS Cancers) 0.0003 -0.0681 0.0687 0.9937 XI (Other Malignant Epithelial 0.0507 -0.0378 0.1392 0.2613 Neoplasms) X (Germ Cell Tumors) 0.0723 -0.0036 0.1482 0.062 VII (Hepatic Tumors) 0.2646 0.0132 0.5161 0.0391* II (Lymphomas) -0.0449 -0.0986 0.0088 0.1012 IV (Neuroblastomas) 0.0013 -0.086 0.0886 0.976 XII (Other Unspecified) -0.0444 -0.1543 0.0654 0.4279 VI (Renal Tumors) -0.0174 -0.1067 0.0719 0.7026 V (Retinoblastomas) 0.0182 -0.1094 0.1459 0.7794 IX (Soft Tissue Sarcomas) -0.0059 -0.0785 0.0668 0.8743 I (Leukemias)R 0 0 0 . Treatment Modality Chemotherapy only 0.0406 -0.0286 0.1097 0.2504 Chemotherapy and radiation 0.0309 -0.0329 0.0948 0.3418 Chemotherapy and surgery 0.0352 -0.0228 0.0932 0.2348 Chemotherapy, radiation and surgery 0.1019 0.0405 0.1634 0.0011* No Info 0.0025 -0.0795 0.0845 0.952 No treatment 0.112 -0.0464 0.2705 0.1659 Radiation only 0.171 0.0503 0.2918 0.0055* Radiation and surgery 0.0775 0.0226 0.1324 0.0057* Surgery only R 0 0 0 . Relapse Status Relapse <5yrs 0.0319 -0.0027 0.0666 0.071 Relapse >5yrs 0.0764 -0.0139 0.1668 0.0972 No relapse R 0 0 0 . Sex Female -0.0006 -0.0301 0.0289 0.9698 Male 0 0 0 . SES 1 (Lowest) -0.0031 -0.0669 0.0608 0.9249 2 -0.012 -0.0542 0.0301 0.5751 3 -0.0016 -0.0437 0.0405 0.9411 4 -0.0368 -0.0781 0.0046 0.0816 Unknown -0.0309 -0.0754 0.0136 0.174 Highest R 0 0 0 . 174 Parameter Estimate 95% Confidence Interval Pr > ChiSq Health Authority Fraser Health -0.0834 -0.1231 -0.0437 <.0001* Interior Health -0.0799 -0.1333 -0.0266 0.0033* Northern Health -0.0574 -0.1205 0.0056 0.0742 Unknown Health -0.0683 -0.1653 0.0288 0.1682 Vancouver Island Health -0.0019 -0.0507 0.0469 0.9395 Vancouver Coastal Health R 0 0 0 . Rural ness Large community 0.0131 -0.0371 0.0634 0.6085 Rural -0.0231 -0.068 0.0218 0.3129 Small community -0.0472 -0.0921 -0.0023 0.0395* Metropolitan R 0.0131 -0.0371 0.0634 0.6085 Age at Diagnosis 0.0072 0.0023 0.0121 0.004* Age at Start of 2-Year Interval -0.0049 -0.0091 -0.0007 0.0219* Primary Care Visits During 2-Year 0.0004 -0.0008 0.0015 0.5293 Interval R Reference Category 175 Table D.6 Survivor-Only, Biannual by Age: UPC Score Parameter Estimate 95% Confidence Interval Pr > ChiSq Intercept 0.6303 0.556 0.7046 <.0001* ICCC Diagnosis VIII (Malignant Bone Tumors) -0.0331 -0.1012 0.035 0.3407 III (CNS Cancers) 0.0101 -0.044 0.0641 0.7151 XI (Other Malignant Epithelial 0.044 -0.0271 0.1152 0.2253 Neoplasms) X (Germ Cell Tumors) 0.0604 0.0019 0.1188 0.043* VII (Hepatic Tumors) 0.185 -0.0033 0.3734 0.0542 II (Lymphomas) -0.0332 -0.0762 0.0098 0.1298 IV (Neuroblastomas) 0.0223 -0.046 0.0907 0.5216 XII (Other Unspecified) -0.0361 -0.124 0.0519 0.4215 VI (Renal Tumors) -0.0086 -0.0797 0.0625 0.8129 V (Retinoblastomas) 0.0184 -0.0822 0.119 0.7201 IX (Soft Tissue Sarcomas) 0.0069 -0.0509 0.0646 0.8151 I (Leukemias)R 0 0 0 . Treatment Modality Chemotherapy only 0.0466 -0.0077 0.1008 0.0928 Chemotherapy and radiation 0.037 -0.014 0.088 0.155 Chemotherapy and surgery 0.0373 -0.0101 0.0847 0.123 Chemotherapy, radiation and surgery 0.0837 0.0339 0.1334 0.001* No Info -0.0045 -0.0708 0.0619 0.895 No treatment 0.0911 -0.0452 0.2274 0.1903 Radiation only 0.1299 0.0444 0.2154 0.0029* Radiation and surgery 0.0672 0.024 0.1104 0.0023* Surgery only R 0 0 0 . Relapse Status Relapse <5yrs 0.0252 -0.0018 0.0523 0.0674 Relapse >5yrs 0.047 -0.0292 0.1231 0.2264 No relapse R 0 0 0 . Sex Female -0.0091 -0.0328 0.0147 0.4539 Male 0 0 0 . SES 1 (Lowest) 0.0054 -0.0475 0.0583 0.8427 2 -0.0009 -0.034 0.0321 0.9551 3 0.0081 -0.0252 0.0415 0.632 4 -0.0184 -0.0511 0.0144 0.2713 Unknown -0.0159 -0.0502 0.0184 0.363 Highest R 0 0 0 . 176 Parameter Estimate 95% Confidence Interval Pr > ChiSq Health Authority Fraser Health -0.0616 -0.0928 -0.0304 0.0001* Interior Health -0.0537 -0.0964 -0.0111 0.0135* Northern Health -0.0468 -0.0966 0.003 0.0653 Unknown Health -0.0399 -0.1185 0.0387 0.32 Vancouver Island Health -0.0006 -0.0385 0.0373 0.9758 Vancouver Coastal Health R 0 0 0 . Ruralness Large community 0.0185 -0.0212 0.0581 0.361 Rural -0.021 -0.0572 0.0152 0.2555 Small community -0.0292 -0.0647 0.0062 0.1056 Metropolitan R 0.0185 -0.0212 0.0581 0.361 Age at Diagnosis 0.0058 0.0019 0.0097 0.0039* Age at Start of 2-Year Interval -0.0032 -0.0065 0 0.0527 Primary Care Visits During 2-Year -0.0006 -0.0018 0.0005 0.2906 Interval R Reference Category 177 Table D.7 Case-Control, Biannual by Age: COC Score Parameter Estimate 95% Confidence Interval Pr > ChiSq Intercept 0.4545 0.432 0.4769 <.0001* Case-Control Case -0.0006 -0.0167 0.0154 0.9381 Control R 0 0 0 . Sex Female 0.0252 0.0131 0.0374 <.0001* Male R 0 0 0 . SES Unknown 0.0012 -0.0264 0.0288 0.9336 1(Lowest) 0.0052 -0.0109 0.0214 0.5259 2 0.0041 -0.0121 0.0203 0.6203 3 -0.0094 -0.0257 0.0069 0.2601 4 -0.0223 -0.0383 -0.0064 0.006* Highest R 0 0 0 . Health Authority Unknown -0.0654 -0.0811 -0.0497 <.0001* Fraser Health -0.09 -0.1122 -0.0679 <.0001* Interior Health -0.0925 -0.1188 -0.0663 <.0001* Northern Health -0.0442 -0.0852 -0.0032 0.0344* Vancouver Island Health -0.0775 -0.097 -0.058 <.0001* Vancouver Coastal Health R 0 0 0 . Ruralness Large community 0.0113 -0.0086 0.0312 0.2673 Rural -0.01 -0.028 0.008 0.2755 Small community -0.0239 -0.0436 -0.0042 0.0176* Metropolitan R 0 0 0 . Age at Start of 2-Year Interval 0.0012 0.0004 0.002 0.0038* Primary Care Visits During 2-Year 0.0003 -0.0002 0.0008 0.1853 Interval R Reference Category 178 Table D.8 Case-Control, Biannual by Age: UPC Score Parameter Estimate 95% Confidence Interval Pr > ChiSq Intercept 0.6551 0.638 0.6722 <.0001* Case-Control Case 0.0004 -0.0124 0.0132 0.9533 Control R 0 0 0 . Sex Female 0.0136 0.0041 0.0231 0.0051* Male R 0 0 0 . SES Unknown 0.007 -0.0146 0.0286 0.5254 1(Lowest) 0.0054 -0.0071 0.0179 0.3932 2 0.0038 -0.0087 0.0163 0.5516 3 -0.0039 -0.0166 0.0088 0.5478 4 -0.0136 -0.026 -0.0011 0.0332* Highest R 0 0 0 . Health Authority Unknown -0.0467 -0.0588 -0.0347 <.0001* Fraser Health -0.064 -0.0814 -0.0466 <.0001* Interior Health -0.0711 -0.0916 -0.0505 <.0001* Northern Health -0.0308 -0.0621 0.0004 0.0532 Vancouver Island Health -0.0555 -0.0707 -0.0404 <.0001* Vancouver Coastal Health R 0 0 0 . Ruralness Large community 0.0125 -0.003 0.028 0.1153 Rural -0.0075 -0.0217 0.0068 0.3045 Small community -0.0121 -0.0278 0.0035 0.1285 Metropolitan R 0 0 0 . Age at Start of 2-Year Interval 0.0007 0.0001 0.0014 0.0215* Primary Care Visits During 2-Year -0.001 -0.0014 -0.0005 <.0001* Interval R Reference Category 179"@en . "Thesis/Dissertation"@en . "2008-05"@en . "10.14288/1.0066692"@en . "eng"@en . "Health Care and Epidemiology"@en . "Vancouver : University of British Columbia Library"@en . "University of British Columbia"@en . "Attribution-NonCommercial-NoDerivatives 4.0 International"@en . "http://creativecommons.org/licenses/by-nc-nd/4.0/"@en . "Graduate"@en . "Health services utilization and provider continuity of care among survivors of childhood cancer : a cohort analysis"@en . "Text"@en . "http://hdl.handle.net/2429/2490"@en .