UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Educational and psychological effects of childhood chronic illnesses Martinez, Yvonne Julia 2007

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

Item Metadata

Download

Media
831-ubc_2007-0182a.pdf [ 7.33MB ]
Metadata
JSON: 831-1.0054567.json
JSON-LD: 831-1.0054567-ld.json
RDF/XML (Pretty): 831-1.0054567-rdf.xml
RDF/JSON: 831-1.0054567-rdf.json
Turtle: 831-1.0054567-turtle.txt
N-Triples: 831-1.0054567-rdf-ntriples.txt
Original Record: 831-1.0054567-source.json
Full Text
831-1.0054567-fulltext.txt
Citation
831-1.0054567.ris

Full Text

E D U C A T I O N A L AND P S Y C H O L O G I C A L E F F E C T S O F C H I L D H O O D C H R O N I C ILLNESSES by Yvonne Julia Martinez Bachelor o f Arts (Honours), University o f Waterloo, 2003 A THESIS S U B M I T T E D I N P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F M A S T E R O F A R T S in The Faculty of Graduate Studies (School Psychology) T H E U N I V E R S I T Y OF B R I T I S H C O L U M B I A A p r i l 2007 © Yvonne Julia Martinez, 2007 Abstract The numbers of children affected by a chronic physical illness have steadily increase d over the last few decades with current prevalence estimates at as many as 20% o f school-aged children affected. For children with a chronic illness, the prevalence of educational and psychological problems is nearly doubled in comparison to the general population. The purpose of this research is to investigate the educational and psychological effects of childhood chronic illness among Canadian children. This project w i l l be a retrospective analysis using data from the National Longitudinal Survey of Children and Youth ( N L S C Y ) , taking a cross-sectional look at the relationships between childhood chronic illnesses, performance on a Mathematics Computation Exercise ( M C E ) , and ratings on an Emotional Disorder - Anxiety scale ( E D A ) . A total of 1,644 children (ages 10-15) across Canada were included in the analyses. The results showed that when E D A ratings and educational handicaps were controlled for, children identified with chronic illnesses still have weaker performance on the M C E test. However, illness did not appear to have an impact on children's E D A ratings. The regression analysis indicated that community type and illness were the strongest predictors o f M C E scores. Limitations o f the study are discussed, and implications of the findings and directions for future research are highlighted. Table of Contents Abstract i i List of Tables v i i List o f Figures v i i i Acknowledgments ix Dedication x i C H A P T E R 1 : 1 Introduction 1 Overview 1 Problem Statement 4 Objective of this Research 6 Research Questions One and Two 7 Rationale and Hypotheses for Questions One and Two 7 Research Questions Three and Four 7 Rationale and Hypotheses for Question Three and Four 8 Research Question Five 8 Rationale and Hypotheses for Question Five 9 Summary..... 9 C H A P T E R 2 11 Literature Review 11 Overview 11 Anxiety Disorders in the General Population 13 Anxiety in Children with Chronic Illnesses 16 Theoretical Explanation of the Development o f Anxiety 20 Development of Anxiety in Children with Chronic Illnesses 23 Educational Effects of Childhood Chronic Illness 26 Learning Difficulties and Disabilities in Children with Chronic Illnesses 28 Cognitive Functioning and Achievement in Children with Chronic Illnesses 29 Summary 32 iv C H A P T E R 3 33 Research Methodology 33 Research Questions 33 Data : 33 The National Longitudinal Survey for Children and Youth ( N L S C Y ) 34 Measures 35 Physical (Chronic) Illness 36 Measures o f the Mathematics Computation Exercise ( M C E ) 37 Measure o f Emotional Disorder - Anxiety ( E D A ) 39 Data Preparation 42 Weighting 42 Sampling/Participants for Current Study 43 Descriptive Statistical Analyses 44 M C E Scores 44 E D A scores 45 Educational Handicap 46 Age .- 47 Province of Residence ". 49 Size Area (Community Type) of Residence ; 50 Household Income 50 School Absent Days 51 Reason Absent from School 52 Miss ing Data 52 Statistical Analyses 52 Statistical Analyses for Question One and Two 53 Statistical Analyses for Question Three and Four 53 Statistical Analyses for Question Five 54 Assumptions for Analysis o f Variance ( A N O V A ) , Analysis of Covariance ( A N C O V A ) and Regression Analyses 55 Scatterplot of M C E Scores 59 Summary of Data Preparation and Testing for Statistical Assumptions 59 V C H A P T E R 4 60 Results 60 Analysis o f Variance for M C E Scores by Illness and Gender.. 60 M C E Scores by Illness 61 M C E Scores by Illness and Gender 61 Analysis o f Variance for E D A Scores by Illness, Gender, and Age 62 E D A Scores B y Illness 62 E D A Scores B y Illness A n d Gender 63 E D A Scores B y Illness B y Age A n d Gender 63 Analysis of Variance for E D A scores by Illness, Gender and Age 65 Post Hoc Tests for E D A Scores 65 Comparison of Children Wi th or Without Illnesses in Their Mathematics Computation Exercise ( M C E ) Scores with E D A as a Covariate 68 Comparison o f Children Wi th or Without Illnesses in Their Mathematics Computation Exercise ( M C E ) Scores with Educational Handicaps as a Covariate 69 To What Extent Do Illness A n d E D A Variables Predict Children's M C E Scores, When Gender A n d Socio-Economic Factors A n d Community Type Accounted For? 71 C H A P T E R 5 74 Discussion 74 Summary and Discussion o f Findings 76 Explanation o f Findings for the Educational Effects o f Childhood Chronic Illnesses 76 Explanation of Findings for the Psychological Effects of Childhood Chronic Illnesses 77 Emotional Disorder-Anxiety A s a Moderating Variable for Mathematics Achievement.... 78 Educational Handicaps as a A s a Moderating Variable for Mathematics Achievement . 79 Predictors o f Mathematics Achievement 79 General Implications o f Findings 80 Theoretical Implications 80 Research Implications 81 Clinical and Appl ied Implications 82 General Limitations o f Study and Future Directions 83 Measurement 84 Measures o f illness 84 Mathematics computation exercise test 85 Emotional disorder-anxiety scale 86 School absence days 90 Primary and secondary effects of illness 91 External validity and generalizability 91 Specific Research Needed to Clarify or Extend Findings 92 Measures o f illness 92 Measures o f academic achievement 94 Measure o f anxiety 95 Gender and age effects 95 Primary and secondary effects of illness '. 97 Measures o f community type, household income, etc 97 Conclusion 97 References , 99 Appendix A : Guide for Researchers under Agreement with Statistics Canada 114 Appendix B : U B C Behavioural Research Ethics Board Certificate o f Approval 114 Vll List of Tables Table 3.1 Frequency and Percentages (Unweighted and Weighted) for Children with a V a l i d Report of Illness 44 Table 3.2 Statistics for M C E and E D A scores 45 Table 3.3 Number o f Children with a Chronic Illness and/or an Educational Handicap .. 47 Table 3.4 Age Distribution for Children with a V a l i d Report o f Illness across Canada... 48 Table 3.5 Provinces and Territories of Residence in which Children with a V a l i d Report of Illness Live 49 Table 3.6 Size-Area (Community Type) in which Children with a V a l i d Report of Illness Live 50 Table 3.7 Family Annual Household Income of Children with a V a l i d Report of Illness 51 Table 3.8 School Absent Days of Children with a V a l i d Report o f Illness 51 Table 3.9 Reason that Children with a V a l i d Report of Illness are Absent from School. . 52. Table 3.10 Significance Levels and Effect Sizes for Levene's and Lack o f Fit Tests 58 Table 4.11 Means and Standard Deviations for M C E Scores by Illness and Gender 61 Table 4.12 Analysis of Variances for Mathematics Computation Exercises ( M C E ) Scores 62 Table 4.13 Mean E D A Scores by Illness, Age and Gender Groups 64 Table 4.14 Analysis o f Variances for Emotional Disorder-Anxiety ( E D A ) Scores 65 Table 4.15 Tukey' s Post Hoc Test for E D A scores by Age 66 Table 4.16 Tamhane, Dunnett T3, Games-Howell and Dunnett C Post Hoc Tests for E D A scores by Age 67 Table 4.17 Analysis o f Covariance for M C E Scores, After Removing the Effects of E D A Scores 68 Table 4.18 Mean M C E Scores by Illness and Educational Handicap 69 Table 4.19 Analysis of Covariance for M C E Scores, After Removing the Effects of Identification of an Educational Handicap 70 Table 4.20 Analysis of Variance for Hierarchical Regression Analysis for SES Variables Predicting M C E Scores Above and beyond the effects o f Illness, Gender and E D A scores 72 Table 4.21 Summary o f Hierarchical Regression Analysis for SES Variables Predicting M C E Scores Above and beyond the effects of Illness, Gender and E D A scores 73 ) v i i i List o f Figures Figure 2.1 Model o f the Development of Vulnerability for Anxiety 22 Figure 3.2 Normal P-P Plot o f M C E Scores 56 Figure 3.3 Normal P-P Plot of E D A Scores 57 Figure 3.4 Scatterplot o f M C E Scores 59 Acknowledgments I would like to thank my thesis supervisor, Dr. Kadriye Ercikan who provided continuous support and encouragement at every phase of this project. I am very grateful for having the opportunity to work with her and have gained utmost respect for her because of her knowledge and expertise. She has helped me develop great appreciation towards the area o f measurement, evaluation and research methodology, and I hope that m y interest and knowledge in this area w i l l grow in the future. I would also like to thank Dr. Wi l l i am M c K e e , my academic advisor and committee member, who constantly challenged me to think critically about my research and encourage me to pursue an unconventional path in School Psychology. I would also like to extend my gratitude to Dr. Jennifer Shapka for her valuable insight and expertise in using the National Longitudinal Survey for Children and Youth, and Dr. Lynn M i l l e r who always knew what to say and do to calm my anxiety during stressful times. M y friends and family have been instrumental in supporting me over the course of my magisterial studies at U B C . Their love, encouragement, support and patience was constant, and I sincerely thank them for standing behind me and pushing me to strive for a goal that I was not sure I could ever reach. This is my greatest academic accomplishment to date, and I sincerely thank my friends and family for helping me reach this goal. X Acknowledgments (cont'd) Statistics Canada and the Social Sciences and Humanities Research Council made this project possible by providing access to the National Longitudinal Survey of Children and Youth (NLSCY) databases at the British Columbia Inter-university Research Data Centre (BCIRDC) . I would like to thank Lee Grenon, Cheryl Fu, and James Croal at the British Columbia Inter-university Research Data Centre ( B C I R D C ) for their patience, efficiency, guidance and support. Although the research and analysis are based on data from Statistics Canada, the opinions expressed do not represent the views of Statistics Canada. xi In loving memory of my maternal grandmother and paternal grandfather who both passed away during my magisterial studies at UBC. 1 CHAPTER 1 Introduction Overview Recent advances in the medical field have resulted in an increase of illness survivors over the past few decades (Copeland, 2002). Children, in particular, are surviving illnesses they were never expected to survive, and the recent shift in focus from the medical and physical effects o f chronic illnesses has been in response to the emerging educational and psychological concerns found among illness survivors (Bartel & Thurman, 1992; Bauld, Anderson, & Arnold, 1998; Canadian Paediatric Society [CPS], 1994; Geist, Grdisa, & Otley, 2003; Peterson, Reach, & Grabe, 2003; Sexson & Madan-Swain, 1993; Shaw & Paez, 2002). Chronic illnesses are defined as medical conditions of long duration that require some level of ongoing medical management and accommodations in life functioning that is above and beyond the normal requirements of a child (Geist, Grdisa, & Otley, 2003; Sexson & Madan-Swain, 1993). The definitions of these illnesses include persistence over 3 months in a given year and/or requiring over a 1-month period of continuous hospitalization (Mash & Wolfe, 1999). Some of the most common chronic conditions found in children include asthma, epilepsy, leukemia and diabetes (Clay, 2004). High-incidence conditions (i.e., asthma, diabetes, juvenile rheumatoid arthritis) are more prevalent, but have a better prognosis and treatment regime, and are not as debilitating. Low-incidence conditions (i.e., cystic fibrosis, gastrointestinal disorders, infectious diseases [e.g., hepatitis, HIV/ADDS]) , are less prevalent, 2 yet may require more frequent hospitalizations, intense medical supervision/treatment programmes, and are generally more debilitating for the child (Clay, 2004). Although it is difficult to determine the specific prognosis o f each condition, it is acceptable to say that many o f these children have a decreased life expectancy, and experience some interruptions o f daily life functioning, such as attending school, socializing with peers, and participating in recreational activities (Bartel & Thurman, 1992; C P S , 1994; Pfenning & Fritz, 1983; Katz, Rubinstein, Hubert, & Blew, 1989; Perrin, Lewkowicz, & Young, 2000; Power, DuPaul, Shapiro, & Parrish, 1998; Sexson & Madan-Swain, 1993). Thus, the term "chronic illnesses" w i l l be used to cover the broad range o f chronic health conditions children around the world face today, unless otherwise specified. Current estimates indicate that as many as 20% o f school-aged children are affected by a chronic illness. In North America, it is estimated that up to 12 mi l l ion children are affected (Kubiszyn, 1999; Mash & Wolfe, 1999; Shaw & Paez, 2002). Yet it remains a challenge to determine the exact incidence and prevalence rates of these childhood illnesses as the term is broad and sometimes these children experience more than one health-related concern (Bartel & Thurman, 1992). The various technological and scientific advances in the medical field have significantly contributed to improvement of health-related quality-of-life in children living with a chronic illness (Speith & Harris, 1996). Survival rates are at an all-time high, whereby 80-90% of children suffering from chronic illnesses, which may have been considered fatal at one point, now reach early adulthood (Paone, Whitehouse, & Stanford, 1998; Sexson & Madan-Swain, 1993). 3 A n example of such illnesses is Acute Lymphoblastic Leukemia ( A L L ) ; four decades ago, A L L was the most common childhood cancer, one considered fatal where the average life expectancy was three months (Bartel & Thurman, 1992). However, it is now estimated that 71 to 80% of children with A L L achieve long-term remission (Bartel & Thurman, 1992; Paone et al., 1998; Rynard, Chambers, Kl inck , & Gray, 1998). This translates to 70% of children surviving an illness they were not expected to outlive who now may have to face the educational and psychological remnants of A L L . Preliminary studies have found that medical issues can result in various changes in the life of children, which often carry with them some degree o f stress, emotional or psychosocial distress (La Greca, Swales, Klemp, Madigan, & Skyler, 1995; Serrano-JJdcos, Whitehead, Rees, & Graham, 1999; Shaw & Paez, 2002; Schneiderman, Antoni, Saab, & Ironson, 2001). Children l iving with chronic illnesses are often faced with the fear of death, anticipation o f painful medical procedures, inability to keep up with academic/curricular requirements, physical limitations/deformities (i.e., amputation fatigue, nausea, blurred vision), which are referred to as primary effects of an illness. Changes in lifestyle and social activities, and isolation/separation from peers and family due to an illness are referred to as a secondary effect of an illness (Bauld et al., 1998; Shaw & Paez, 2002). More specifically, primary effects are directly due to the illness or treatment regime, the most common being a change in a child's physical appearance. Other effects include physical abnormalities, delay in physical/maturational development, delayed puberty, nausea, pain and fatigue, all o f which can have implications on a child's psychological, cognitive, physical and mental development (CPS, 1994). For example, a primary effect o f more serious types of asthma would be a deformity in the chest, otherwise known as pigeon chest 4 (pectus carinatum) where the child has a protrusion of the anterior chest wal l (Cataletto, 2006). A s a result o f the physical deformity, the child may be teased and victimized by peers or become embarrassed o f changing in front of other children at gym and have lowered self-esteem (secondary effect). These indirect effects, or secondary effects, are thus a result of the primary effect. Ultimately, secondary effects of an illness can be as devastating as the primary effects (CPS, 1994; Paone et al., 1998; Sexson & Madan-Swain, 1993). Problem Statement A n emerging area o f research with children with chronic illnesses concerns the educational and psychological effects of childhood chronic illnesses. However, these studies typically involve small numbers of participants, are limited to select geographic regions in the United States and other countries around the world, and thus may not be generalizable to Canadian populations. Essentially, studies linking educational, psychological, and physical health remain very limited. Research in this area is still in its infancy. Studies involving children with chronic illnesses are limited to relationships between chronic illness and educational effects, chronic illness and psychological effects; very few studies link the three areas. Specifically, there is a dearth of research exploring the complex relationship between chronic medical conditions in children, psychological effects and how this may affect a child's academic achievement (Bauld et al., 1998; Shaw & Paez, 2002). One approach to study these inter-relationships in the Canadian context would be to explore large-scale assessment data collected in Canada. The National Longitudinal Survey of Children and Youth ( N L S C Y ) is a large-scale survey jointly conducted by Statistics Canada (STC) and Human Resources Development Canada ( H R D C ) in efforts to provide information for policy analysis and program development on factors affecting the 5 development o f children in Canada and to monitor the impact o f these factors on the child's development over time (STC, 2002). The N L S C Y is the only nationally representative, longitudinal study o f children and youth focusing on health, development and well-being to date, and provides us with a detailed look at the development o f children as they make their way through to adulthood (STC, 2002). The nature o f the survey and breadth and depth of sampling (e.g., age range of children and youth surveyed, multiple informants/raters, etc.), promises to be an effective research tool in studying the relationships between physical health, psychological health and educational achievement in Canada. The key characteristics o f the N L S C Y are that surveying includes questions related to a child's physical health (e.g., indication o f asthma, epilepsy, cerebral palsy, kidney condition/disease, etc.), academic achievement (e.g., performance on the Mathematics Computation Exercises [MCE]) , and psychological health (e.g., ratings on the emotional disorders - anxiety factor), collected from the Person-Most Knowledgeable ( P M K ) about the child (usually the mother), the child's teacher and the child. This is a retrospective study that uses the N L S C Y dataset to investigate meaningful relationships between physical health, educational and psychological factors. Particularly, the goal of the study is to determine whether psychological and/or educational outcomes are associated with a child's health status. A major advantage in using retrospective designs is that these types o f studies allow for the analysis o f relationships among many different variables that are not necessarily tied to a specific research question. In addition, such designs facilitate large-scale analyses o f data, which is an important component of exploratory research studies. 6 Studies systematically point to the conclusion that educational and psychological effects of childhood chronic illnesses do exist. There is emerging evidence showing that children with chronic medical conditions experience more problems related to school adjustment and social interaction/development, as wel l , children with illnesses are at a higher risk for developing anxiety (Adewuya & Ola, 2005; Bak i et al., 2004; Chavira, Stein, Bailey, & Stein, 2004; Hommel et al., 2003). Although there have been some mixed findings, differences are often attributed to individual child differences (i.e., personality, coping skills, family support, etc.), and differences between illnesses (i.e., treatment rigor, "severity of symptoms) (Brown & DuPaul, 1999; Frank et al., 1998). Yet despite the research investigating the educational and psychological effects of childhood chronic illnesses, there is a large gap in the literature; the combined educational and psychological effects of childhood chronic illnesses remain unexplored. Objective of this Research The purpose o f this research is to investigate the educational and psychological effects o f childhood chronic illness among Canadian children. This project is a retrospective analysis using data from the N L S C Y . Although the N L S C Y is a longitudinal study, the current study examines the cross-sectional relationships between childhood chronic illnesses, academic achievement and anxiety. For consistency and ease o f reference, this project w i l l refer to school-aged children ages 5 to 18 as a "child", unless otherwise specified even though they are referred to as "children and youth" in the N L S C Y . 7 The current research project aims to answer the following five research questions. Research Questions One and Two Are there differences between children with chronic illnesses and healthy children in their Mathematics Computation Exercise ( M C E ) scores? ( Q l ) . Are there differences between children with chronic illnesses and healthy children in their Emotional Disorder -Anxiety ( E D A ) scores? )(Q2). Rationale and Hypotheses for Questions One and Two Research concerning children with chronic illnesses has found higher rates of school-related problems (e.g., lower achievement, higher learning difficulties/disabilities) and higher reported psychological problems (e.g., anxiety and depression) (Frank et al., 1998; Shaw & Paez, 2002). A s well , there has been substantial research demonstrating the effect of anxiety on a child's school achievement (Albano, Chorpita, & Barlow, 2003; Huberty, 1997; Leibert & Morris, 1967; V a n Ameringen, Mancini , & Farvolden, 2003; Wigfield & Eccles, 1990; Wine, 1971; Wine, 1980). However, there are few studies investigating both educational and psychological effects o f chronic illnesses. Research Questions Three and Four Is the relationship between illness and Mathematics Computation Exercise ( M C E ) scores different when M C E scores are adjusted for the effects of Emotional Disorder -Anxiety ( E D A ) scores? (Q3) In addition, how are chronic illnesses associated with performance on the M C E after removing the effects of various educational handicaps? (Q4) 8 Rationale and Hypotheses for Question Three and Four Based on findings showing a strong association between academic achievement and anxiety (Albano et al., 2003; V a n Ameringen et al., 2003), the relationship between illness and M C E scores w i l l be tested with the variation in M C E scores due to E D A scores (covariate) removed. It is hypothesized that when E D A scores are controlled for, illness w i l l show a stronger effect on M C E scores. The rationale for this hypothesis is that given the literature showing that children who have high anxiety have lower performance on achievement tests (Albano et al., 2003; Huberty, 1997; Leibert & Morris , 1967; V a n Ameringen et al., 2003; Wigfield & Eccles, 1990; Wine, 1971; Wine, 1980), controlling for the possible moderating effect of anxiety (on M C E scores) w i l l help better explain the effect of a physical (chronic) illness on achievement. A s for research question 3, the rationale is that by controlling for and removing the effects of "educational handicaps" (children who are identified by the P M K with mental retardation, a learning disability, attention deficit and/or an emotional disorder in the N L S C Y data) on achievement, differences in performance across different illness groups on the M C E test can be better explained by a child having a physical (chronic) illness versus having a mental handicap, learning disability, attention deficit, and emotional disorder. Research Question Five What is the degree of the association between Mathematics Computation Exercise ( M C E ) scores, and each o f health/illness and Emotional Disorder - Anxiety ( E D A ) scores? How do these associations compare to those between M C E and gender, community type, and household income? 9 Rationale and Hypotheses for Question Five Given evidence showing the associations between chronic illness and academic achievement (Sexson et al., 1993; Taras & Potts-Datema, 2005), illness and anxiety (Adewuya & Ola, 2005; Hommel et al., 2003; Ortega, Huertas, Canino, & Rubio-Stipec, 2002; Rynard et al., 1998; Shaw & Paez, 2002) anxiety and achievement (Albano et al., 2003; Huberty, 1997; V a n Ameringen et al., 2003), the fifth question addresses the degree to which illness and emotional disorders/anxiety predict mathematics achievement, above and beyond typical child background variables such as gender, community type and household income. In addition to looking at the degree to which E D A scores predict M C E scores, other factors that have been found to be associated with performance on achievement tests, such as age, gender, household income, community type (urban vs. rural) were also investigated. Summary For children with a chronic illness, the prevalence of both educational problems and psychological problems (i.e., anxiety disorders) is nearly twice the proportion found in the general population (Frank et al., 1998). Anxiety disorders are the most common, under-treated, yet preventable psychological problem in all children (Albano et al., 2003). Given the increasing numbers o f children affected by a chronic illness and the dearth of research in exploring the complex relationship between chronic illnesses, anxiety, and academic achievement, it is crucial to initiate a line of research to investigate these associations. 10 B y understanding these associations, researchers and practitioners may be able to evaluate the outcomes o f increased health management and medical/technological advances for Canadian children. Perhaps more importantly, our findings may suggest ways to plan strategic allocation of resources (e.g., prevention programs) that w i l l foster the successful and healthy development of children as productive members o f society (Paone et al., 1998; Sexson & Madan-Swain, 1993; Shaw & Paez, 2002; Speith & Harris, 1996). 11 C H A P T E R 2 Literature Review Overview This chapter includes a review of current literature relevant to the relationships between childhood chronic illnesses, academic achievement and anxiety. The first section w i l l describe anxiety disorders in the general population, and then follow with research on anxiety disorders in children with chronic illnesses. Next, various models that attempt to explain the development of anxiety w i l l be reviewed (e.g., psychoanalytic, behavioural-learning, and attachment theories), with a particular focus on Chorpita and Barlow's (1998) Theory o f Uncontrollability and Unpredictability. This particular theory w i l l then be applied to current understanding concerning the development of anxiety in children with chronic illnesses, where current research in the area of anxiety in the pediatric population w i l l be discussed. Following the review of the psychological effects of chronic illnesses, the discussion w i l l then move to a review concerning the educational effects of childhood chronic illnesses. Studies examining educational problems among children with chronic illnesses (e.g., learning disabilities and difficulties, problems with cognitive functioning and academic achievement) w i l l be reviewed, and limitations of these findings w i l l be discussed. The role of gender, socioeconomic status (SES), and culture w i l l be considered throughout these reviews; these issues must be carefully considered, as these variables contribute to the manifestation and development of a host o f problems (Mash & Wolfe, 1999). 12 It is important to note that there is very limited research integrating childhood physical health, mental health, and education. Previous research addressed one of three relationships: (1) the relationship between illness and psychological effects; (2) the effects of illnesses on education; or (3) the relationship between psychological and educational effects. The main gap in the literature concerns the association among these three areas; it is unknown how these factors interact together. Historically, the focus o f treatment in individuals with chronic health conditions centers on addressing medical and physical ailments to ensure survival and treat illness symptomatology (Peterson et al., 2003). There is evidence to support the contention that chronic illnesses not only bear physical consequences, but that these illnesses bear secondary psychological and educational consequences, and have significant impact on individual quality-of-life (Chavira et al., 2004; Peterson et al., 2003; Shaw & Paez, 2002). These concerns include: high levels of negative affect (e.g., anxiety and depression), problems with cognitive ability, academic skills, school performance, learning, adaptive behaviour, peer relationships, social withdrawal and peer conflict (Shaw & Paez, 2002; Taras & Potts-Datema, 2005). Although recent findings suggest that the prevalence o f psychological and educational problems have nearly doubled in recent years for children with a chronic medical condition (Henry & DuPaul , 2004; Perrin et al., 2000), there is an absence o f research in linking these three areas o f study. This shift in addressing other facets of quality-of-life has a central goal of preparing children with illnesses for a successful and healthy adulthood (Paone et al., 1998). 13 Anxiety Disorders in the General Population A common definition of anxiety has been captured by Huberty (1997, p. 305) as "apprehension, distress, or tension about real or anticipated internal or external threats that may be shown in cognitive, behavioural, or physiological patterns." Healthy levels of anxiety have survival purposes (i.e., fight or flight response) and are part o f normal development, however, at extreme levels; anxiety can have deleterious impact affecting a child's personal, social and daily functioning (Albano et al., 2003; and Huberty, 1997). For example, a child who has obsessive worries of being separated from his/her parents can cause extreme distress to parents and teachers, which are effects beyond those experienced by the child. Symptoms o f anxiety are thought to be expressed through three systems: a) the physical/physiological response system; b) the cognitive response system; and c) the behavioural response system. It is crucial to understand how these three systems interplay, as each level of symptomatology is expressed differently in children as a function of age, gender, previous experience, situational variables and coping styles (Huberty, 1997; Mash & Wolfe, 1999). Some common physical/physiological responses include increased heart rate, flushes/chill, nausea, increased respiration, muscle tension, sweating, stomach upset and loss of breath. Examples o f cognitive response systems include thoughts o f being scared, forgetfulness, self-deprecatory/self-critical thoughts, blanking out, and difficulty concentrating. Finally, examples of behavioural response systems include avoidance, screaming, nail biting, clenched jaw, stuttering, fidgeting, and rigid posture (Mash & Wolfe, 1999). 14 Where it was once thought that anxiety in children was a mi ld and transitory period that w i l l dissipate over time, current research findings suggest that anxiety has a chronic course where children who experience anxiety symptoms w i l l continue to display problems in adjustment through to adulthood (Mash & Wolfe, 1999). Groups of anxiety symptoms can co-occur, and children with specific symptoms are likely to experience other symptoms. Hence, these clusters o f symptoms are referred to as dimensions o f anxiety, and are grouped using factor analysis. The diagnosis and classification of anxiety disorders continues to be a debated issue because there is a lack of a common understanding of what is considered a normal and developmentally appropriate anxiety response (Albano et al., 2003; Mash & Wolfe, 1999). Although this research project does not concern the formal, clinical diagnosis o f anxiety, anxiety disorders w i l l be discussed in the next section in order to provide a more comprehensive definition and understanding of anxiety. The diagnostic criteria for the nine anxiety disorders are described in the Diagnostic and Statistical Manual of Mental Disorders, 4 t h Edition, Text Revision ( D S M - I Y - T R ) (American Psychiatric Association [ A P A ] , 1994): separation anxiety disorder (SAD) , panic disorder, agoraphobia, generalized anxiety disorder ( G A D ) , social phobia, specific phobia, obsessive-compulsive disorder (OCD) , posttraumatic stress disorder (PTSD) and acute stress disorder. Anxiety is the predominant feature in each o f these nine disorders, and what distinguishes one disorder from the other is the focus of the anxiety and how anxiety is expressed differently through the physical/physiological, cognitive or behavioural systems (Albano et al., 2003). 15 Anxiety disorders are the most common psychological disorder in children and adults. In children, prevalence rates across the nine disorders are reported to range from 0.6% (Panic Disorders) to 12% (Separation Anxiety Disorder [SAD]) in samples o f children from the United States (Albano et al., 2003; Mash & Wolfe, 1999); lifetime prevalence is estimated to be around 28.8% (Kessler, Berglund, Demler, Jin, & Walters, 2005). Anxiety disorders are equally distributed across socioeconomic status (SES) and ethnic groups and remain quite stable throughout the lifespan (Dadds, James, Barrett, & Verhust, 2004; Huberty, 1997). Research focusing on the effects of anxiety on academic performance has shown that higher levels of anxiety affected academic performance and cognitive functioning, where an increase of irrelevant thoughts o f self-evaluation/deprecation, debilitating off-task thoughts, difficulties with task completion, and problems associated with attention, memory, speech and language have been found in children with higher levels of anxiety (Huberty, 1997; Mash & Wolfe, 1999). To add, researchers have found that anxiety in first grade predicted anxiety in the fifth grade, and anxiety levels in first grade significantly affected achievement at fifth grade (Ialongo, Edelsohn, Werthamer-Larsson, Crockett, & Kel lam, 1994). Though there is sufficient research to demonstrate the relationship between anxiety and academic performance, there has been little research investigating the cause o f lower performance on measures of achievement (e.g., absenteeism, direct interference on cognitive tasks, etc.) (Mash & Wolfe, 1999). In terms o f gender differences, girls have been found to be more l ikely to report anxiety-related symptoms on self-report measures than boys, however, any gender differences detected were generally quite small (Costello, Egger, & Angold, 2004; Huberty, 1997). 16 Anxiety in Children with Chronic Illnesses For children affected with a medical condition, results reported in the literature suggest that they are twice as likely to develop psychological disorders as are their healthy peers, and children with certain health conditions (i.e., asthma, epilepsy and cancer) are at a higher risk of developing anxiety (Adewuya & Ola, 2005; Hommel et al., 2003; Ortega, Huertas, Canino, Ramirez, Rubio-Stipec, 2002; Rynard et al., 1998; Shaw & Paez, 2002). A review by Adewuya and Ola (2005) indicated that the prevalence o f anxiety disorders among children with epilepsy ranged from 16% to 48.3%. Over 30% of children with asthma meet diagnostic criteria for an anxiety disorder, and have overall higher levels o f anxiety compared to individuals who are not asthmatic (Hommel et al., 2003). These numbers suggest that the prevalence rate for anxiety disorders among chronically i l l children is quite high, although the overall prevalence rate for children with chronic illnesses is still not well-established (Shaw & Paez, 2002). New research is suggesting that un-addressed anxiety in childhood/adolescence can lead to secondary disorders in adulthood, disorders such as emotional disturbance/mood disorders and antisocial/criminal behavior (including violence and substance use/abuse) and can affect later life functioning (e.g., unproductive academic and occupational performances, premature death, chronic disability, marital discord and divorce and risk-taking behaviours) (Albano et al., 2003; Huberty, 1997; Perrin et al., 2000; Power et al., 1998; and Shaw & Paez, 2002). There is further evidence demonstrating that anxiety precedes many disorders (i.e., depressive disorders, substance abuse), and may lead to externalizing disorders, followed in turn by dysthymia, major depression and substance abuse. The developmental progression of anxiety is very disconcerting given the increased survival rates o f children 17 with chronic illnesses. Thus, there is a need to curtail problems by identifying and treating anxiety early on in life (Ferdinand, Barrett, & Dadds, 2004). The problem with the identification (and treatment) o f anxiety in children with chronic illnesses is that symptoms may be over-looked and undetected. For example, because o f the internalizing nature of the disorder (anxiety symptoms are within the child), children exhibiting anxiety-related symptoms are less aversive to people or property as compared to disorders o f the externalizing type (e.g., oppositional defiant disorder [ODD], attention deficit hyperactivity disorder [ADHD]) where children can be aggressive, destructive and non-compliant (Albano et al., 2003; Chavira et al., 2004). Thus, children with anxiety disorders are often overlooked in a classroom among children with externalizing types of disorders. Another problem with the identification of anxiety is that some individuals may not have a good understanding about the primary and secondary effects o f a chronic illness to recognize anxiety symptomatology, or may minimize problems experienced by children (Chavira et al., 2004; Masia-Warner, Nangle & Hansen, 2006). The manifestation of anxiety symptoms in a child who is i l l may be mistaken for fatigue from treatment, side effects from drugs, and general inability to keep up academically or socially (Albano et al., 2003). For example, a child with severe asthma may experience anxiety, thus have difficulty concentrating, and be flooded with fear and thoughts o f an asthma attack and perhaps imminent death (cognitive response system). S/he may then respond by fidgeting, crying and immobility (behavioral response system), followed by fatigue and headaches (physical response system). Without a clear understanding of how these anxiety response systems 18 interplay and work, response systems can be mistaken as a side effect o f the illness, behavioural problems and/or inattention. Methods of assessment pose another problem in the identification o f anxiety disorders. For example, the "Tripartite Mode l " (Clark & Watson, 1988) suggests that although anxiety and depression have unique symptoms, the two disorders also have overlapping symptomatology. In fact, Clark, Watson and Beck (1994) argued that symptomatology in self-report measures of anxiety tended to display high correlations with measures of depression. The reason for this is that both anxiety and depression measures are found to assess a broader nonspecific distress coined "negative affect" ( N A ) (Clark et al., 1994). Particularly, N A has been found to be highly correlated with almost all anxiety symptoms and depressive symptoms (Clark & Watson, 1988). For this reason, it may be difficult to decipher whether an individual has anxiety or depression by only using self-reports. One way to address the problem of self-reporting is to use multiple methods of assessment such as structured clinical interviews with multiple raters (parents, teacher and child) and standardized measures of anxiety (e.g., Multidimensional Anxiety Scale for Children [ M A S C ; Marsh, 1997], Revised Children's Manifest Anxiety Scale [ R C M A S ; Reynolds & Richmond, 2000], State-Trait Anxiety Scale for Children [STAIC; Spielberger, Edwards, Lushene, Montuori, & Platzek, 1973].). B y using multiple methods o f assessment, the error and bias involved in measuring anxiety can be reduced (Tarullo, Richardson, Radke-Yarrow, & Martinez, 1995); 19 Some studies have found that parents are the most thorough and reliable in reporting complex details about a child compared to reports from teachers or the self-reports from the child (Schniering, Hudson, & Rapee, 2000). However, other studies have yielded contradictory results, where other researchers have found parent-reports of anxiety to be highly flawed (i.e., under or over-reporting). For example, parents may inflate reports of anxiety, sometimes as a consequence o f their own anxiety, and projecting this anxiety on their children (Dadds et al., 2004). Self-reports of anxiety have also been criticized where a child's ability to read and understand a question has been challenged, especially when methods of questioning may be developmentally inappropriate for young children (Dadds et al., 2004). For example, younger children may not have the cognitive capacity to comprehend questions of complex problems within themselves. Another problem with self-reporting has to do with "faking good" or "faking bad" where some children may provide socially desirable responses (e.g., answering "Yes" when asked "I am always good") to please the examiner (March, 1997). The incorrect/false reporting o f one's feelings, attitudes or behaviours significantly affects the reliability and validity of the report. Unfortunately, it can be very time consuming and economically costly to administer a variety of parent-, teacher- and child-reports of anxiety and conduct structured interviews, and the feasibility of collecting data on anxiety using, multiple measures is sometimes not possible. The risk of emotional difficulty in children with chronic illnesses has been found to be positively associated with age with emotional risk increasing with age. Gershon, Zimand, Pickering, Rothbaum, and Hodges (2004) found that the experience o f distress is quantitatively and qualitatively different for younger and older children. Younger children 20 seem to be less affected (emotionally) by illness, whereas adolescents experience more distress due to their increased understanding o f their illness and other secondary effects (i.e., self-esteem, socialization). Younger children may not have as good an understanding of the impact of their treatment regimens as compared to older children. Older children may also experience greater distress because of a better understanding of the pathology o f the disease (Frank etal., 1998). Theoretical Explanation of the Development of Anxiety There are various theories on the etiology of anxiety including: psychodynamic theories, behavioural/learning theories, attachment theories and cognitive theories, dating back to the early 1900s (Mash & Wolfe, 1999; Sweeny & Pine, 2004). Early psychoanalytic theory was documented by Sigmund Freud's case of Little Hans where anxiety and phobias were viewed as a defense against unconscious conflicts in a child's early upbringing (Mash & Wolfe, 1999). Fol lowing psychoanalytic theories were the behavioral and learning theories (John B . Watson [Watson, 1916]) where anxiety was believed to be learned through classical and/or operant conditioning; otherwise known as the "Two-Factor Theory" (Mash & Wolfe, 1999). Following the Two-Factor Theory, John Bowlby attempted to use an emotional explanation of anxiety, or "Theory o f Attachment", where it was believed that fear in children was biologically rooted in emotional attachments needed for survival in children (Mash & Wolfe, 1999). More recently, Chorpita and Barlow's (1998) "Theory o f Uncontrollability and Unpredictability" has been used to explain the development o f anxiety. This is a conceptual model and gives credit to the negative emotions associated with the development of anxiety, yet also acknowledges the role of early experiences and innate vulnerabilities or 21 predispositions in the experience of anxiety. The notion of the theory is that early experiences o f children where control is lacking (or reduced) can predispose the child in developing anxiety in the future. Chorpita and Barlow (1998) have used research with animal models, familial influences, locus of control in children and biological explanations to link control to the development of anxiety. Essentially it is believed that the experience with lack o f control may play an important role in the development of anxiety, an area supported by diverse areas of research. Additionally, the biological correlates of chronic anxiety seem to be influenced by one's interpretation of a stressful event being controllable or not. They suggest that early experiences with lack o f control (e.g., childhood chronic health conditions) contribute to a schema or template being developed, wherein it becomes fixed and diathetic. Early experiences with uncontrollable or unpredictable stimuli (e.g., unpredictability of epileptic seizures) can influence one's perception of control. Thus, as the individual progresses to a later developmental period, these control-related cognitions may intensify anxiety, inhibition, thus resulting in negative affect (Figure 2.1). 22 Figure 2.1 Model of the Development of Vulnerability for Anxiety1 Early Development Later Development Stored Information L^ow perceived \ c-e- memory) |_ o w perceived control / control (mls)interpreted as uncontrollable / Tempermental stability of / ' hypothalamic and limbic structures (BIS«) • | Negative affect Undifferentiated somatic complaints Increased CRF** and HPA*** activity •BIS = behavioral inhibition system " C R F = corticotropin releasing factor "*HPA = hypothalamic pituitary adrenocortical 1 From "The Development of Anxiety: The Role of Control in the Early Environment", by B. F. Chorpita, and D. H. Barlow, 1998, Psychological Bulletin, 124, p. 16. 23 Although there is not one single theory to explain the development o f anxiety, Chorpita and Barlow's (1998) theory is of interest as it integrates some o f the classic theories (i.e., behavioural/learning, attachment and cognitive theories) and emphasizes the interaction between biological and environmental influences in the development o f anxiety. Development of Anxiety in Children with Chronic Illnesses Preliminary research findings have suggested that the psychological well-being of children suffering from a chronic illness is compromised when a child is sick. It is estimated that these children are twice as l ikely to develop psychiatric disorders compared to their peers (Shaw & Paez, 2002), where rates of any psychiatric disorder range from 9 to 22% (Chavira et al., 2004). Children with chronic medical conditions are at an increased risk for developing affective co-morbidity specifically with anxiety, as medical issues can result in various changes in the life o f children and result in some stress or emotional distress (La Greca et al., 1995; Shaw & Paez, 2002). Chorpita and Barlow's (1998) Theory o f Uncontrollability and Unpredictability, can be applied to better understand why anxiety develops in children with chronic illnesses; particularly, that illness diagnosis can result in a lack of control and compromise the predictability of one's health and life course. Culture is an important area to address in childhood anxiety, given that culture plays a large role in the development, identification and treatment of many childhood disorders. Given that many illnesses are prevalent across cultures (i.e., epilepsy, cancer, etc.), a better understanding of the influence of culture must be understood in order to identify and treat the anxiety symptoms that accompany these health conditions. For example, Adewuya and Ola (2005) found that there is a higher prevalence of anxiety disorders among Nigerian 24 adolescents with epilepsy compared to their Western counterparts. Partially, the authors feel that this can be due to biological factors, stigma and dearth o f information known about epilepsy, as wel l as family support. However there is limited cross-cultural research in this area. Children who are i l l face complicated medical procedures, treatment regimens, fear of death, anticipation o f painful medical procedures, coping with pain, isolation and fear of being teased, change in lifestyle/disruption of normal development, being different from peers, and the inability to participate fully and keep up with academic and social activities and physical deformities resulting from the illness (Bartel & Thurman, 1992; Shaw & Paez, 2002; and Tyrrell , 2005). Four common types of stressors common among children with chronic illnesses include medical, physical, academic, psychological and social stressors (Clay, 2004). Medical stressors an i l l child may face include: integrating medical tasks into daily routine, managing a complex medication regimen and lifestyle modifications (e.g., eating restrictions, exercise restrictions, sleep difficulties). Physical stressors include: growth suppression, changes in physical appearance and intermittent or chronic pain. Academic stressors include: absence from school and social activities, decline in academic performance and development or exacerbation of learning difficulties. Psychological stressors include: somatic responses, losses associated with illness (e.g., eyesight, limbs, hair, uncertain future, and end-of-life concern). Finally, social stressors may include: environmental precautions, restrictions in freedom and activities, changes in the parent-child relationship, and changes in peer relationships (Clay, 2004). Regardless of stressors being real or perceived, the 25 additional stressors faced by children with chronic illnesses are l ikely to exacerbate developmentally appropriate fears and anxieties in children (Frank et al., 1998). For example a study by Henning and Fritz (1983) demonstrated that the most common problem faced in children with cancer were worries and concerns. Concerns were based on the physical effects o f cancer, such as fatigue, nausea and blurred vision, and about safety and management o f treatment. Fear of being teased, feelings o f embarrassment because of changes in appearance, keeping up with schoolwork, difficulties in talking about the illness, and returning to school were very common fears in all participants, but particularly in early adolescence. These findings have been replicated in cases of children with asthma, cystic fibrosis, diabetes, and epilepsy (Adewuya & Ola, 2005; Hommel et al., 2004; Peterson et al., 2003; Pumariega, Pearson, & Seilheimer, 1993). In further support to Chorpita and Barlow's (1998) theory, Hommel et al. (2003) found that the uncertainty associated with having an illness plays a significant role in individuals who have dealt with asthma for several years. In particular, anxiety-levels are associated with higher levels of uncertainty o f the illness (e.g., when an asthma attack would occur) (Hommel et al., 2003). The authors further hypothesize that children with asthma may develop anxiety as a function of increased perceptions o f uncertainty and unpredictability about the frequency, duration and intensity o f attacks, possible hospitalization, use o f medication, steroids and puffers, and other related health-management stressors. However, it is important to note that participants in the study by Hommel et al. (2003) were selected from an undergraduate college population. Therefore, results are not generalizable to children under the age of 18-years. Additionally, most participants were Caucasian, thus not adequately representing children of other ethnic groups. Given the 26 higher incidence and prevalence rates in asthmatic children among the African American population (Henry & DuPaul, 2004), it is important to include sampling from a diverse culture/ethnic sample in studies with asthma. Despite the findings highlighting the role of control and predictability in l iving with chronic illnesses, Frank et al. (1998) report that overall adjustment to chronic illness is influenced more by environmental (e.g., family support) and psychological factors (e.g., coping skills) rather than to the severity of the illness. Unfortunately, this complex relationship between chronic medical conditions and the development o f anxiety has been highly neglected despite anxiety disorders being remarkably common in pediatric settings (Chavira et al., 2004; and Shaw & Paez, 2002). Studies o f anxiety are generally limited to adult populations, and studies with children involve physically healthy children (Ferdinand et al., 2004), resulting in limited knowledge in the relationship between chronic health conditions and anxiety in children. Educational Effects of Childhood Chronic Illness In addition to the psychological problems children with illnesses must deal with, they are also faced with problems with learning and academic achievement. Healthy school adjustment is a keystone developmental accomplishment, one found to be highly significant in achieving a sense of autonomy, independence and future self-sufficiency, essentially, a child's overall adjustment (Henning & Fritz, 1983; Paone et al., 1998). A large part of a child's development (physical, psychological, educational) is fostered within the school setting (Sexson & Madan-Swain, 1993), yet children with chronic illnesses may need to miss school frequently; it is estimated that chronic illnesses result in as many as 40 mil l ion school absence days a year (Sherman, 1995). 27 School absences can be a result of the illness itself (e.g., hospitalization, being too i l l to attend school) or secondary effects (e.g., treatment effects, school refusal, etc.). In effect, children become more vulnerable to experience a host of other problems related to school absenteeism, such as problems with academic achievement and social interaction. U n -addressed educational issues have been found to be related to the development o f school failure, dropout, depression, and unhealthy lifestyles (Pfenning & Fritz, 1983; Paone et al., 1998; Perrin et al., 2000; Rynard et al., 1998). Thus, it is critical to better understand how illnesses are impacting a child's learning and achievements in school, essentially, one's overall adjustment. A detailed review o f past research in the area o f anxiety and school performance was conducted by Huberty (1997). Most of the reviewed research concentrated on the effects of anxiety on academic performance, task performance, and cognitions during performance and task-relevant behavior. For example, Leibert and Morris (1967; as cited by Huberty 1997) researched test performance and learning. Their findings revealed that worry is related to cognitive and attention cues, whereas emotionality is an involuntary conditioned response. Wine (1971, 1980; as cited by Huberty, 1997) and Wigfield and Eccles (1990; as cited by Huberty, 1997) then looked at off-task thoughts in children who scored high in anxiety, and found that thoughts consuming a child were often irrelevant thoughts, ones that are self-evaluative (negatively) and self-deprecatory. In sum, the review by Huberty (1997) revealed that the higher levels of anxiety affected cognitive functioning, specifically, increases of irrelevant thoughts of self-evaluation and self-deprecation, debilitating off-task thoughts, difficulties with task completion, and increased attentional problems. Essentially, healthy individuals who were experiencing anxiety had more difficulty with school-related tasks. 28 Learning Difficulties and Disabilities in Children with Chronic Illnesses In support o f the above findings among the general population, Sexson and Madan-Swain (1993) report that up to 40% of children with a chronic condition experience school-related problems. Learning disabilities and difficulties become a more serious problem during junior and senior high school year for children with chronic illnesses. There are greater demands on homework and studying time and previous existing learning problems may be compounded by illness process, medical interventions and treatments (CPS, 1994; Clay, 2004). For example, in studies of children with asthma, Henry and DuPaul (2004) found a higher number of repeated grades than healthy controls, and are more l ikely to be found in special education settings and are almost two times more l ikely to have a learning disability. These findings are further supported by research with diabetic children; where 40% of males with diabetes were found to have learning problems warranting special instruction or grade retention, and both males and females had more learning difficulties than healthy peers (24% versus 13%) (Holmes, Dunlap, Chen, & Cornwell, 1992). In studies that compare academic performance between children who are i l l and their siblings, Bartel and Thurman (1992) report that one-third to one-half o f children with A L L require specialized academic help (e.g., attend special education classes) and/or repeat grades, compared to only 15% of their siblings and general population requiring specialized assistance. Moreover, Schatz, Brown, Pascual, Hsu, & DeBaun (2001) found statistically significant differences among children with sickle-cell anemia in the number o f grade retention and special education when compared to peers or siblings. 29 Most academic problems have been found in areas such as mathematics and reading; these areas require the student to achieve a level o f mastery in the ski l l , prior to learning subsequent skills and knowledge (CPS, 1994; Fowler et al., 1988; Kaemingk et al., 2004; Paone et al., 1998; Seidenberg et al., 1986). Sexson & Madan-Swain (1993) report the incidence of reading problems (even when cognitive functioning is controlled for) is higher among children with chronic illnesses (14%) compared to healthy controls (4.5%). McCarthy et al. (2002; 2003) found that reading scores and G P A were significantly lower in children with diabetes, especially when metabolic control in these children was poor. To add, Kaemingk et al. (2004) have found differences in math performance between children who are healthy and children with A L L . However, it is unknown as to whether these difficulties are a direct result o f the illness (e.g., hospitalization, pain, fatigue), or secondary effects (e.g., anxiety, school phobia, effects of treatment regiments) or a combination o f these factors. These results are similar to problems with achievement and cognitive functioning that is found in children with anxiety as previously outlined. Cognitive Functioning and Achievement in Children with Chronic Illnesses Declines in general cognitive ability (Intelligence Quotient, or I.Q.) have been found to range up to 20 points in children with chronic illnesses (Bartel & Thurman, 1992). Declines in cognitive functioning and academic ability are quite common among cancer survivors, specially leukemia or brain tumors, and other illnesses where the illness (and treatment) affect the structure and function of the brain (Bartel & Thurman, 1992; Henning & Fritz, 1983; Rausch & Stover, 2001). Moreover, researchers have found that illnesses that impact a child's central nervous system (CNS) have a greater negative effect on learning, 30 disorders of this nature include: cerebral palsy, epilepsy, infectious diseases (Human Irnrnunodeficiency Virus/Autoimmune Deficiency Syndrome [HIV/AIDS]) , and cancer (Clay, 2004). Other areas related to cognitive functioning, such as attention, memory, processing speed, language, visuo-spatial knowledge, concept formation and motor functioning, have been found to be affected in children diagnosed with H I V / A I D S , diabetes, sickle-cell anemia, and epilepsy (Farwell, Dodr i l l , & Batzel, 1985; Hershey, L i l l i e , Sadler & White, 1999; Northam et al., 2001; 1999; Rausch & Stover, 2001; Rovet & Alvarez, 1997; Rovet & Ehrlich, 1999). For example, researchers found that diabetic children with hypoglycemia had more neurological problems, poorer performance on memory tasks, slower processing speed, decreased memory, learning, attention and inhibition (Hannonen, Tupola, Riikonen, & Ahonen, 2003; Hershey et al., 1999; Northam, et al., 1999; Rovet & Ehrl ich, 1999). However, results are mixed as other researchers (Fox, Chen, & Holmes, 2003; Schoenle, Schoenle, Molinar i , & Largo, 2002) did not find hypoglycemia to be related to learning. In longitudinal studies with symptomatic Human Immunodeficiency Virus (HIV) infection, expressive language has been found to be more impaired than receptive language (Rausch & Stover, 2001). Behavioural effects of H I V infection o f the C N S include: loss of interest in others or activities, less goal-directed behaviour, vacant staring and altered affect (Rausch & Stover, 2001), symptoms similar to those of anxiety. In terms of gender effects, Holmes et al. (1992) found that diabetic males perform poorer than diabetic females in attention tasks (Holmes et al., 1992), and there were more reports of boys having learning problems than girls (40% vs. 16% girls). In addition, Fox, Chen, and Holmes (2003) found more difficulties in verbal learning tests for males compared 31 to females; although females still performed poorer than healthy controls. However, Eaton, Haye, Armstrong, Pegelow, & Thomas (1995) did not find differences in performance across standardized academic achievement tests in children with sickle-cell anemia. Age-of-onset of an illness has been found to affect a child's academic and cognitive functioning. Findings reveal that children who are treated earlier in life (<4 years) are more vulnerable to academic and cognitive problems compared to children treated later in life. The rationale is that treatment may affect the developing brain o f a younger child (Bartel & Thurman, 1992). More specifically, Schoenle et al.(2002) found an age effect on learning where declines in learning were negatively associated with age. In follow-up studies, Northam et al. (1999) found that age-of-onset o f diabetes was a strong predictor of performance. Unfortunately, it is not always easy to determine whether cognitive decline is directly affected by the illness (primary effect) or indirectly related to the treatment (secondary effect) (Bartel & Thurman, 1992). For example, although there is evidence showing a decrease in cognitive functioning in children surviving brain tumors, it is unclear whether the decline is due to the tumor itself, or a result of treatment. A s well , in cases o f children with cancer, toxicity of the central nervous system (CNS) may result in confusion, forgetfulness, and impaired school performance, which could be misinterpreted for inattention, problems with memory and school failure, respectively (Henning & Fritz, 1983). It is hypothesized that factors that appear to have an effect on a child's cognitive and academic functioning include: amount/rigor of treatment, structural and functional changes to the brain (i.e., infection to the central nervous system) (Bartel & Thurman, 1992; Clay, 2004). For example, Armstrong et al. (1996), Brown et al. (2000) and Nabors and Freymuth 32 (2002) found that after experiencing strokes, the addition of abnormal M R I s in children with sickle-cell anemia revealed problems in arithmetic, vocabulary, executive functioning, reading, arithmetic, spelling and visual motor speed and coordination. Ultimately the success of the treatment comes in hand with children experiencing academic and cognitive problems. Some researchers have found that at follow up (up to 7-years later) children with diabetes still demonstrated significant declines in verbal IQ compared to their peers. A s well , these children experienced problems with perceptual motor, memory and attention tasks (Rover & Ehrlich, 1999). Although some medications have not shown to affect a child's cognitive functioning, the medication itself may result in the alteration in a child's mood (i.e., anxiety and depression). Another study found that even after 6 years, verbal IQ was significantly lower among children with diabetes especially with earlier onset of diabetes, and had problems with attention, processing speed and executive control (Northam et al., 2001). Summary Given the review of current literature demonstrating the relationships between childhood chronic illnesses, anxiety and academic achievement, and this research w i l l aim to link findings in these three areas. Specifically, the research questions w i l l build on past findings and look at the extent to which chronic illnesses affect children's academic achievement and anxiety. 33 C H A P T E R 3 Research Methodology Research Questions The following five research questions were addressed in the current study: 1) Are there statistically significant differences between children with chronic illnesses and healthy children in their Mathematics Computation Exercise (MCE) scores?; 2) Are there differences between children with chronic illnesses and healthy children in their ratings on the Emotional Disorder - Anxiety (EDA) scale?; 3) How are chronic illnesses associated with performance on the Mathematics Computation Exercise (MCE) after removing the effect of a child's Emotional Disorder - Anxiety score?; 4) How are chronic illnesses associated with performance on the Mathematics Computation Exercise (MCE) after removing the effect of educational handicaps (mental handicap, learning disability, attention deficit or emotional disorder)?; and 5) What is the degree of association between a child's performance on the Mathematics Computation Exercise (MCE), with health/illness, and on the Emotional Disorder - Anxiety (EDA) when other background factors such as gender, size-area of residence (community-type) and household income are considered? Data The data used to address the research questions posed in this study is the National Longitudinal Survey for Children and Youth (NLSCY). The NLSCY is the only nationally representative, longitudinal study of children and youth focusing on health, development and well-being in Canada, and it provides a detailed look at the development of children as they make their way to adulthood (Statistics Canada [STC], 2002). 34 The National Longitudinal Survey for Children and Youth (NLSCY) The N L S C Y data is based on probability sampling, and involves random sampling of representative Canadian "households". Households were selected from Statistics Canada's Labour Force Survey (LFS) sample frame of about 60,000 dwellings, and data is collected on a monthly basis. This sample has been deemed representative o f the Canadian Population in previous studies (STC, 2002). The N L S C Y provides comprehensive information on children and youth from all ten provinces and three territories, excluding children l iving on Indian reserves or Crown lands, residents of institutions, full-time members of the Canadian Armed Forces, and residents of some remote regions. The age range of children and youth initially surveyed ranged from newborn to 10/11 year olds. Data collection for the N L S C Y has occurred every two years. Data was first collected in December, 1994 to A p r i l , 1995 (Cycle 1), followed by December, 1996 to A p r i l , 1997 (Cycle 2), October, 1998 to June, 1999 (Cycle 3), September, 2000 to May , 2001 (Cycle 4), September, 2002 to June, 2003 (Cycle 5), and September, 2004 to June, 2005 (Cycle 6). Although both longitudinal and cross-sectional data were available, only cross-sectional data from 5 t h Cycle was used. Cycle 5 was the most recently released data at the start of the current project. The N L S C Y provides opportunities for examining a child's daily functioning from multiple perspectives. The N L S C Y collects data from multiple sources: a) the child/adolescent; b) the Person Most Knowledgeable ( P M K , who typically is the child's mother); c) the child/adolescent's teacher; and d) the child/adolescent's school principal. The majority of data were collected via computer-assisted personal interviews, self-completed 35 instruments, and tests. Surveys that rely on multiple respondents increase the quality o f data collected compared to surveys that only rely on one source o f information (e.g., self-reports) (Schniering et al., 2000). Separate databases are created for data that is collected from each respondent category. In other words, there are four datasets that make up Cycle 5:1) the child/adolescent ( '70 to 19 year" dataset); 2) the P M K ("master" dataset); 3) the child/adolescent's teacher; and 4) the child/adolescent's principal. Each o f the databases can be matched on a P M K unique identifier variable. For this project, only the master and 10 to 19 year old datasets from Cycle 5 were of interest. The master dataset includes information reported by the P M K about the age, gender, province of residence, size of residence (or community type), household income, child's health, educational handicap, and includes the child's score on the M C E test. The 10 to 19year old dataset includes child's score on the E D A scale. Measures Only children with a valid report on the illness questions were included in the study. In addition to illness, data from two measures were of particular interest in this study: Mathematics Computation Exercise ( M C E ) and Emotional Disorders - Anxiety ( E D A ) . These measures are described below. 36 Physical (Chronic) Illness Inclusion in the "Illness Group" required a response of "Yes" by the PMK to one of the questions inquiring whether the child has an illness. The questions related to illness are: 1. "Does your child have asthma?"; 2. "Does your child have bronchitis?"; 3. "Does your child have heart condition?"; 4. "Does your child have epilepsy?"; 5. "Does your child have cerebral palsy?"; 6. "Does your child have a kidney condition?"; and 7. "Does your child have other long-term condition[s]?". Each report of "Yes" to an illness variable was recoded as "1", and each report of "No" was recoded as "0. The rationale for this was so that frequency counts and analyses could be conducted for each illness category. However, following a count of the frequencies for each of the eight illnesses, it was determined that the number of illnesses categories would have to be aggregated as there were low case counts for certain illness types (e.g., Cerebral Palsy and Kidney Condition) and the data would not allow statistical analyses by illness. As a result, a new composite illness variable was created, to adhere to STC's data confidentiality and guidelines for disclosure of data. 37 This composite illness variable grouped children into one of two categories: children who have no illnesses and children who have one or more illnesses. Thus, children with PMK reports of "Yes" to any of the illness variables stated above were placed into the "Identified Illness" group. Children with PMK reports of "No" to all of the illness variables above were be placed into the "No Identified Illness" group. Measures of the Mathematics Computation Exercise (MCE) In the NLSCY, reviews of research literature and consultations with Canadian and international experts were used to identify a battery of measures of learning and achievement. Some of the criteria used in selecting measures were sound psychometric properties, whether they targeted measuring the domains identified in the research framework, and ease of administration (STC, 2002). One of the direct measures of achievement used for children ages 10 to 19-years was the Mathematics Computation Exercises (MCE). The MCE is a shortened version of the Mathematics Computation Test from the Canadian Achievement Test, 2nd Edition (CAT/2) (Canadian Test Centre [CTC], 1992), and the score from this assessment appears in the Master dataset. The CAT/2 is a standardized test designed to measure achievement in basic academic skills (STC, 2002). The shortened version MCE was created specifically for the NLSCY and includes 20 questions for Levels 1 to 8, and 15 questions for Levels 9 and 10. The MCE was administered to children in Grade 2 and above (ages 8 to 15) by the NLSCY interviewer in the child's home, This MCE includes a mathematical operations test and problem-solving test. The mathematical operations test measures the ability to add, subtract, multiply and 38 divide operations using whole numbers, decimals, fractions, negatives and exponents. The problem-solving test involves percentage and order of operations (STC, 2002). Substantial changes were made to some of the Cycle 5 testing components from previous cycles. The M C E is one o f the components that were modified. The first change to the M C E was the number of items in the measure; items were added as anchors in order to link assessment data across future cycles. The second change to the M C E was the collection period of the Cycle 5 data; data collection occurred earlier in the year than in previous cycles. Through analyses and evaluations, S T C determined that these changes to the M C E had resulted in a substantial effect on the M C E raw scores. Therefore, these scores were considered no longer comparable with scores from other cycles and inappropriate to use in analyses (STC, 2002). In particular, the concern was the comparability of raw scores across data cycles. Each child who took the M C E was given a raw score, a Classical scale score (CSS) and an Item Response Theory scale score (ER.T SS). Raw scores were obtained by adding the number of correct answers in the M C E . The CSS and IRT SS are adjusted scores correcting for changes made to the M C E test over time (e.g., different time periods o f data collection periods) (STC, 2002) and are considered more appropriate to be used in the analyses of the N L S C Y data. The IRT approach is widely used in psychometric and educational testing. The probability of a correct response (S-shaped curve over the range o f abilities) is assumed to be a logistic function of one's ability (STC, 2002). The three-parameter logistic model was used to compute the M C E scores. The three-parameter logistic model depends on the difficulty o f the question, discrimination power o f the question and chance o f a hypothetical guessing 39 (STC, 2002). Overall, the IRT SS provides measurements that are comparable across cycles, and provides a greater precision in the estimates of test performance within the N L S C Y population. In addition, IRT SSs allow comparisons to be made across different M C E levels, and age groups. Thus, the M C E IRT score was used in the current study. Measure of Emotional Disorder — Anxiety (ED A) The Emotional Disorder - Anxiety is one component o f the Behaviour Scale in N L S C Y Chi ld Component data. The Behaviour Scale has nine theoretical constructs: 1) Separation Anxiety (ages 2 and 3 years); 2) Opposition (ages 2 and 3 years); 3) Conduct Disorder (ages 2 to 11 years); 4) Hyperactivity (ages 2 to 11 years); 5) Emotional Disorder (ages 2 to 11 years); ) 6) Indirect Aggression (ages 2 to 11 years); 7) Physical Aggression (ages 2 and 3 years and ages 8 to 11 years); 8) Inattention (ages 2 to 11 years); and 9) Prosocial behaviour (ages 8 to 11 years) (STC, 2002). Each of these nine constructs were derived from either one, or a combination of questionnaire items from the Achenbach's Child Behavior Checklist (Achenbach's C B C L ) (Achenbach, Rescorla, McConaughey, Pecora, Wetherbee, & Ruffle, 1991) the Montreal Longitudinal Survey (Dr. R. Tremblay of the University of Montreal), the Ontario Child Health Study (OCHS) (STC, 2000) and scales by K . Weir and G . Duveen and by Lagerspetz, Bjorngvist and Peltonen of Finland (STC, 2002). r .40 Factor and reliability analyses were conducted to assess the psychometric properties of the Behaviour Scale by S T C . Only four scales were determined to be distinct and reliable: 1) Hyperactivity - Inattention; 2) Emotional disorder - anxiety; 3) Physical aggression -conduct disorder; and 4) Indirect aggression (STC, 2002). The Emotional Disorder - Anxiety (EDA) was one of these factors. In N L S C Y , E D A was used to measure emotional well-being o f the children/adolescents. The E D A is comprised of seven questions: 1) "I am unhappy or sad"; 2) I am not as happy as other people my age"; 3) "I am too fearful or nervous"; 4) "I worry a lot"; 5) "I am nervous, high strung or tense"; 6) "I have trouble enjoying myse l f ; and 7) "I cry a lot" (STC, 2002). Each question had three response format options: "Never or not true", "Sometimes or somewhat true", "Often or very true", and assigned values of 0, 1 and 2, respectively. A raw score was then computed by adding up the response options; a score o f 0 would infer that the symptom is not present (or very rare). Scale score were computed by S T C for each of the factors. The E D A scale scores range between 0 and 14. Clark and Watson's (1991) Tripartite Model and Barlow and Chorpita's (1998) Theory of Unpredictability and Uncontrollability provide support for E D A ' s validity as a measure of anxiety in children. Questions 2, 3, 4 and 5, in E D A correspond to traditional factors of anxiety (e.g., "physiological anxiety" [questions 3, 4, 5] and "social 41 concerns/concentration" [question 2]), and questions 1, 6, and 7 correspond to traditional factors of depression ("dysphoria" and "negative mood"). In E D A , more unique features of anxiety are measured by questions 2, 3, 4 and 5, and overlapping features between depression and anxiety ( N A ) are measured in questions 1, 6, and 7. Another line of support for E D A as a measure of anxiety in children is that research has found that children who are anxious often show signs of depression, in particular, these children appear unhappy and report feeling of sadness and loneliness (Beidel & Turner, 1998). Hence, the Theory o f Unpredictability and Uncontrollability, Tripartite Model , and empirical research, provide support for E D A being a measure o f symptoms related to anxiety in children. Factor and reliability analyses were conducted by S T C using a total o f 13,765 children ages 4 to 11 years using the Cycle 4 dataset. The Chronbach's alpha (Cronbach, 1951) was .736 indicating moderate reliability (STC, 2002). However, one o f the limitations of the factor and reliability analyses conducted by S T C is the age range used to conduct these analyses. Developmental level is an important characteristic to consider when making a diagnosis of anxiety, as the focus of the anxiety and how anxiety is expressed can differ across ages (Albano et al., 2003; Warren & Sroufe, 2000). Thus, further analyses were conducted to examine the degree to which results obtained using the Cycle 4 data with children ages 4 to 11, would generalize to Cycle 5 data with older children and adolescents. A n exploratory factor analysis was carried out using the Cycle 5 population data of children ages 10 to 15. In this factor analysis, the seven variables identified by S T C were loaded onto one factor. The correlation between the E D A scale scores in Cycle 4 to the E D A 42 scale scores from Cycle 5 was r = . 881 (p < .001). Item-to-total test score correlations ranged from .893 to .948, and inter-item correlations ranged from .810 to .927, demonstrating strong internal consistency (Hair, Anderson, Tatham, & Black, 1998). Additionally, a reliability analysis was conducted by the researcher to test how well this set of seven items measures the unidimensional construct of " E D A " . Chronbach's alpha for the set of seven items in Cycle 5 was .981, further demonstrating very strong internal consistency. Data Preparation Some data cleanup (e.g., imputation for missing data) was already conducted by S T C . Additional data preparation that involved creating normalized weights, and merging of the master and 10-19 year-old data sets. The two data sets were merged by matching on the P M K unique identifier variable. Weighting Prior to analyses being conducted, a normalized weighting variable was created from the "Standard Longitudinal Weight" in the Cycle 5 data. The mean of the standard longitudinal weighting variable was computed, and then the standard longitudinal weighting variable was divided by its own mean. Thus, the new mean for this normalized weight is equal to 1. The purpose o f weighting the N L S C Y data was to make the N L S C Y sample comparable to the Canadian population. B y weighting variables, the data was corrected for sample non-representativeness that resulted from survey design issues (e.g., over sampling population sub-groups), and data collection issues (i.e., non-response for particular sub-groups). However, by weighting the dataset using the standardized longitudinal weighting 43 variable, there would be an over inflation of power due to the immense increase in the sample size. Thus, using the normalized weighting variable corrected for this over inflation. Sampling/Participants for Current Study A l l o f the N L S C Y data were weighted using a normalized weighting variable to fulfill S T C ' s confidentiality guidelines, as well as for appropriate use o f national data. One exception to this is the overall sample size with valid responses for the key variable, chronic illness. In effect, there were 1,644 children between the ages o f 10 to 15 (effective age at Cycle 5 testing), who had valid responses to the question about whether they had a chronic illness or not, as identified by the P M K (person most knowledgeable about the child). The data was then weighted (using normalized weights), which showed a total of 1512 (N) children who would be included in the statistical analyses. O f these 1512 children, 978 children were identified with no illness(es), and 534 were identified with 1 or more illness(es). The sample included a higher proportion of males (n = 896) than females (n = 616). There were 275 children who were 10-11 years, 391 children who were 12-13 years, and 442 children who were 14-15 years. A total of 1206 children (79.76%) had valid M C E scores and 868 children (57.41%). had valid E D A scores. These results are presented in Table 3.1. 44 Table 3.1 . Frequency and Percentages (Unweighted and Weighted) for Children with a Valid Report of Illness Variable Distribution n (unweighted) n (weighted) No 1018 978 Chronic Illness Yes 626 534 TOTAL 1644 1512 No _ 1320 Educational Yes 193 Handicap TOTAL - 1512 Male - 896 Gender Female - 616 TOTAL - 1512 10-11 years - 275 Age 12-13 years - 391 442 14-15 years -• TOTAL - 1108 MCE - 1206 EDA _ 868 Descriptive Statistical Analyses M C E Scores A child's M C E score was based on his/her IRT scale score on the Mathematics Computation Exercise ( M C E ) test, a standardized mathematical operations and problem-solving test given to all N L S C Y participants aged 8 to 15 years. The M C E score is a continuous variable, and M C E scores for children with a valid illness ranged from 248 to 739 points on a scale that had a population mean of 473.46 and standard deviation of 107.25 (Table 3.2). 45 Table 3.2 Statistics for M C E and E D A scores Variable Mean SD Minimum Maximum MCE 473.46 107.25 248.00 739.00 EDA 3.13 2.55 .00 13.00 E D A scores The Emotional Disorder - Anxiety ( E D A ) scale is purported to capture a child's propensity to develop an emotional disorder, specifically, anxiety-related disorder. The E D A scale score is based on participants' rating on seven questions related to emotional (anxiety) disorders on the N L S C Y questionnaire. Based on a principal component analysis, a simple E D A score was created. Although the scale scores on the E D A scale can range from 0 to 16, scores only ranged from 0 to 11 for children with a valid report of illness, with most children scoring 0, 1 and 4 and the mean E D A score was 3.13 and the standard deviation was 2.55. Higher scale scores on the E D A scale would be indicative of an increased likelihood of the child experiencing emotional disorder that is related to anxiety, and lower scores would be indicative of a decreased likelihood of an emotional disorder (anxiety) being present. The sample E D A scores indicate that the children in the sample were not experiencing high levels o f emotional, anxiety-related disorders. The frequency of E D A scores was plotted on a histogram. The distribution of scores showed a negatively skewed distribution with the highest frequency for scale score of 1, and a sharp decline at a scale score of 5. However, due to S T C ' s Confidentiality Guidelines, this histogram cannot be presented. 46 Educational Handicap Given that children who were identified with a physical illness may also be identified with a mental handicap (mental retardation), or diagnosed with a learning disability, attention deficit disorder, or emotional disorder, it was important to take into consideration the extent to which these difficulties may hinder learning and one's performance on the M C E test. A frequency count was run for each of these four variables (mental handicap, learning disability, attention deficit and emotional disorder) and it was determined that these variables would have to be aggregated into one category due to low case counts, as wel l as to adhere to S T C ' s data confidentiality and guidelines for disclosure of data. Thus, the term "educational handicap" was used to describe children who were identified with one or more of the above illnesses. Children with a response of " Y e s " by the P M K on any one o f the following questions were placed into the "Educational Handicap - Identified" group: 1. "Does child have mental handicap? (diagnosed)"; 2. "Does your child have a learning disability? (diagnosed)"; 3. "Does your child have attention deficit? (diagnosed)"; and 4. "Does your child have an emotional disorder? (diagnosed)". Descriptive analyses were conducted for children with an identification of a chronic (physical) illness and with an identification of an educational handicap (Table 3.3). The rationale for this was to take into consideration the proportion o f children with both a chronic illness as well as an educational handicap that would impact their performance on the M C E test. 47 For children who were not identified with an illness, there were 59 children (5.18%) who had an educational handicap, and 691 children (60.61%) who had an educational handicap. For children who were identified with an illness, there were 64 children (5.61%) who were also identified with an educational handicap, and 326 children (28.60%) who were not identified with an educational handicap. These numbers are presented in Table 3.3. Table 3.3 Number of Children with a Chronic Illness and/or an Educational Handicap Illness Educational Handicap n (weighted) No No 691 Yes 59 No 326 Yes Yes 64 TOTAL TOTAL 1140 Note. An "educational handicap" refers to an identification of a mental handicap, learning disability, attention deficit or emotional disorder by the PMK. Age The N L S C Y uses "effective age" instead of the child's actual age to ensure that the child stays in the specified age group assigned regardless o f whether data collection occurs before or after a child's birthday. For Cycle 5, the effective age was calculated as 2002 minus year of birth. Responses that were coded as " V a l i d Skip", "Don't K n o w " , "Refused" and "Not stated" in the original N L S C Y dataset were treated as missing data. Although the effective age in Cycle 5 ranged from 0 to 19 years-old, only children ages 10 to 15 were of interest in this study. Children with valid illness reports who were 10 to 15 years were extracted from the sample, and then divided into three age composites: 10 to 11 years, 12 to 13 years, and 14 to 15 years. There were two reasons for creating age 48 composites; the first reason was due to the low number of case counts for children of a certain age who are identified with a chronic illness. The use o f low case counts in specific age composites would affect the significance level and power of the statistical tests. Secondly, categorizing children into separate age composites would be appropriate in running statistical analyses as M C E scores are based on a vertical scale with large errors and a wide range o f performance across different age groups. The further apart children are in age along the continuum o f M C E scores (e.g., comparing children who are 10-years-old to children who are 15-years-old), the less comparable the M C E test scores become. Thus, the creation of the three age composites increased the level o f comparability o f the M C E . The sample o f children with valid illness reports ranged between the ages of 10 and 15 years: 24.82% children were 10 to 11 years-old; 35.29% were 12 to 13 years-old; and 38.89% were 14 to 15 years-old (Table 3.4). Table 3.4 Age Distribution for Children with a Valid Report of Illness across Canada Age Categories , . n . . . x a ° (weighted) 10 to 11 years 275 12 to 13 years 391 14 to 15 years 442 TOTAL 1108 49 Province of Residence There were 15 Province of Residence response options which included territories and options other than provinces: "Newfoundland and Labrador", "Prince Edward Island", "Nova Scotia", "New Brunswick", "Quebec", "Ontario", "Manitoba", "Saskatchewan", "Alberta", "British Columbia", "Yukon" , "Northwest Territories", "Nunavut", "United States", "Outside Canada and U S A " . Respondents l iving outside of Canada were collapsed into a variable named "Other". However, the sample o f children with valid illness reports l ived in only eleven provinces across Canada at the time of the Cycle 5 data collection: Ontario (38.49%), Quebec (25.13%), British Columbia (10.98%), Alberta (10.25%), Manitoba (3.24%), Saskatchewan (3.37%), Nova Scotia (3.37%), New Brunswick (2.58%), Newfoundland (2.18%) and Prince Edward Island (.53%) (Table 3.5). Table 3.5 Provinces and Territories in which Children with a Valid Report of Illness Live Province n (weighted) Alberta 155 British Columbia 166 Manitoba 49 New Brunswick 39 Newfoundland and Labrador 33 Northwest Territories 0 Nova Scotia 51 Nunavut 0 Ontario 582 Prince Edward Island 8 Quebec 380 Saskatchewan 51 Yukon 0 TOTAL 1512 Note. No valid reports of illness for children living in the Northwest Territories, Nunavut or Yukon Territories 50 Size Area (Community Type) of Residence There were 5 Size-area (Community Type) of Residence response categories: "Rural area", "Urban Population, less than 30,000", "Urban Population, 30,000 to 99,999", "Urban Population, 100,000 to 499,999", and "Urban Population, 500,000 or more" (Table 3.6). Most children lived in communities that had a population of 500,000 or more, and in populations o f less than 30,000. Table 3.6 Size-Area (Community Type) in which Children with a Valid Report of Illness Live Size-Area (Community Type) n (weighted) Rural Area 164 Urban Population (less than 30,000) 379 Urban Population (30,000-99,999) 124 Urban Population (100,000-499,999) 244 Urban Population (500,000 or more) 602 TOTAL 1512 Household Income. There were 6 Household Income response categories: "Less than $10,000", "$10,000 to 14,999", "$15,000 to 19,999", "$20,000 to 29,999", "$30,000 to 39,999", and "$40,000 or more" (Table 3.7). Most household incomes were between $30,000 to 39,999 per year. 51 Table 3.7 Family Annual Household Income of Children with a Valid Report of Illness Annual Household n Income (weighted) Less than $10,000 7 $10,000 to 14,999 53 $15,000 to 19,999 42 $20,000 to 29,999 101 $30,000 to 39,999 130 $40,000 or more 1178 TOTAL 1512 School Absent Days There were 6 Number o f Days Absent response categories: "0 days", "1 to 3 days", "4 to 6 days", "7 to 10 days", "11 to 20 days", and "more than 20 days" (Table 3.8). Most children missed between 1 to 6 days of school per school year. Table 3.8 School Absent Days of Children with a Valid Report of Illness Number of Absent n Days (weighted) 0 days 70 1 to 3 days 423 4 to 6 days 472 7 to 10 days 262 11 to 20 days 173 More than 20 days 87 TOTAL 1487 Reason Absent from School Children miss school for a myriad of reasons, however, approximately 80.00% of the sample from the N L S C Y missed school due to a health reason (Table 3.9). Other responses, such as "Problems with weather", "Family vacation", etc. made up the other ~20 % of the sample. Although "School Phobia" was listed as one of the reasons for school absence, the case counts were so small that releasing this data would have violated the S T C guidelines for disclosure of data.L Table 3.9 Reason that Children with a Valid Report of Illness are Absent from School Reason for School Absence , . n . . (weighted) Health Reasons 418 Other (weather, vacation, etc.) 105 TOTAL 522 Missing Data. A l l responses that were coded as " V a l i d Skip", "Don't K n o w " , "Refused" and "Not stated" in the original N L S C Y dataset were treated as missing data. Due to the systematic pattern of missing data, and the exploratory nature of this study, imputation methods were not used. Thus, only children with valid reports of the variables of interest were included in the study. Statistical Analyses Due to data confidentiality and adherence to STC's guidelines for disclosure of data, all case counts and statistical analyses were conducted using normalized weighted data (see Appendix A for more details on these guidelines). 53 Statistical Analyses for Question One and Two The first and second research questions were: "Are there statistically significant differences between children with chronic illnesses and healthy children in their Mathematics Computation Exercise ( M C E ) scores?; and "Are there statistically significant differences between children with chronic illnesses and healthy children in their Emotional Disorder -Anxiety ( E D A ) scores?", respectively. Two Analysis of Variance ( A N O V A ) analyses were conducted to address these research questions. For the first A N O V A test, M C E was treated as the dependent variable, and illness and gender were treated as the independent variables. For the second A N O V A model, E D A was treated as the dependent variable, and illness, gender and age were treated as the independent variables. The A N O V A s helped us test whether there were statistically significant differences in M C E and E D A scores for groups o f children who were identified with and without a chronic illness, while also looking at the effect o f gender (on M C E and E D A scores) and age (on E D A scores only). Statistical Analyses for Question Three and Four The third and fourth research questions were: "How are chronic illnesses associated with performance on the Mathematics Computation Exercise ( M C E ) after removing the effect of a child's Emotional Disorder - Anxiety score?; and "How are chronic illnesses associated with performance on the M C E after removing the effects o f various Educational Handicaps (mental handicap, learning disability, attention deficit and emotional disorder)?" Two Analysis o f Covariance ( A N C O V A ) analyses were conducted to examine the effect of illness on M C E after controlling for E D A and educational handicaps. The dependent variable was M C E scores in both of these analyses, and the covariate for the first 54 A N C O V A was E D A , and the covariate for the second A N C O V A was educational handicap. Illness and gender were used to test for interaction effects. Statistical Analyses for Question Five The final research question was: "What is the degree o f association between a child's performance on the Mathematics Computation Exercise ( M C E ) , with health/illness, score on the Emotional Disorder - Anxiety ( E D A ) scale, and when other factors such age, gender, household income, community type (urban vs. rural) are also considered?" A blockwise regression analysis was used with M C E scores as the dependent variable. First, gender and household income were entered as Block 1. The size area (community type) o f residence was included in Block 2. Illness was then added in Block 3. Finally, E D A scores were added in Block 4. 55 Testing Assumptions of Statistical Models The statistical analyses that were proposed for use in this study are the Analysis of Variance ( A N O V A ) , Analysis o f Covariance ( A N C O V A ) and Regression. Prior to running these analyses, the assumptions of the statistical models were tested. Assumptions for Analysis of Variance (ANOVA), Analysis of Covariance (ANCOVA) and Regression Analyses The two assumptions common to an A N O V A , A N C O V A and Regression are the assumptions o f normality and homogeneity of variance. The assumption o f normality posits that the data is normally distributed for the dependent variables. In order for the assumption of normality to be met, the deviation from normality must be small. This can be assessed by looking at the slope o f the data distribution for the variable o f interest (i.e., M C E scores and E D A scale scores) using a normal probability plot (Hair et al., 1998). A diagonal line is indicative of normality in a probability plot. Figures 3.2 and 3.3 are the probability plots for M C E scores and E D A scale scores, respectively. Given that the expected and observed cumulative probability plots are both linear, both the M C E scores and E D A scale scores met the assumption of normality. 56 Figure 3.2 Normal P-P Plot of M C E Scores Normal P-P Plot of Regression Standardized Residual Dependent Variable: Recoded M C E IRT Scores 0.0 0.2 0.4 0.6 0.8 1.0 Observed Cum Prob Cases weighted by longwtn 57 Figure 3.3 Normal P-P Plot of EDA Scores Normal P-P Plot of Recoded EDA Scores Observed Cum Prob Cases weighted by longwtn The second assumption, homogeneity of variance, posits that each population of scores has equal variance. Typically, the Levene's Test of Equality of Error Variances for each of the A N O V A , A N C O V A and Regression analyses tests the null hypothesis that the error variance o f the dependent variable ( M C E scores and E D A scale scores) is equal across the different groups being tested (illness, age and gender). Thus, a Levene's test that is statistically significant is indicative that the groups do not have equal variances (Table 3.10). The significance levels of the Levene's tests were all significant at the p < .001 level. The effect sizes from the Lack o f Fit Tests are presented in Table 10. 58 Table 3.10 Significance Levels and Effect Sizes for Levene's and Lack of Fit Tests Test Variables p-value Levene's Test Effect size /7-vaiue Lack of Fit Test ANOVA1 Dependent Variable = MCE Independent Variable = gender, illness .000 .000 ANOVA 2 Dependent Variable = EDA Independent Variable = gender, age, illness .000 .000 Dependent Variable = MCE ANCOVA1 Independent Variable = gender, illness Covariate = EDA .000 .127 Dependent Variable = MCE ANCOVA 2 Independent Variable = gender, illness Covariate = educational handicap .000 .002 However, given that the effect sizes were very small, despite the significance of the Levene's tests, both A N O V A and A N C O V A are robust to violations o f this assumption. A n y transformation of the dependent variables is expected to complicate the interpretation of findings and a decision was made not to transform these variables. The third and final assumption of the A N C O V A and Regression tests is the assumption of linearity o f the relationship between the dependent variable and each of the independent variables. This assumption requires that the within-group relationship between the covariate and dependent variable ( A N C O V A analysis), and between the predictor and dependent variables (Regression analysis) is linear (Grimm & Yarnold, 1995). Scatterplot was used for the visual identification of nonlinear patterns of the M C E data. M C E data points were plotted against E D A data points, which demonstrated a linear ( relationship. However, due to the S T C confidentiality and data disclosure guidelines, this particular plot was not permitted to be released. Thus, the standardized predicted values 59 were plotted against the standardized residuals (Figure 3.4), and linearity was again detected. Hence, transformations to the M C E data were not necessary. Figure 3.4 Scatterplot of M C E Scores Scatterplot of MCE IRT Scores 2.00000-f 3 n > u 0) T3 N « •o c n 1.00000-0.00000' -1.00000' o o o o o o , 0 „ o O O Q ° ^ O 0 ° o 0 o ° o ° ° o ° o 9) o 0 ° ° o Jo o o o o o o o re-©-p o l ° o . n CK oo o~.o o"*"?o Standardized PredictecfcyalueQ g,qf i -0.0® * ZRE_1 -Square = 0.00 o o o o o 0°o o o o I I I I I -2.00000 -1.00000 0.00000 1.00000 2.00000 Standardized Residual Linear Regression with 95.00% Mean Prediction Interval Summary of Data Preparation and Testing for Statistical Assumptions Following these data preparation and preliminary analyses (e.g., frequency counts, probability plots) the data appeared to meet the assumptions for the data analyses procedures selected as wel l as S T C ' s data confidentiality and disclosure risk guidelines. Thus, the statistical analyses were conducted to answer the five research questions posed. 6 0 C H A P T E R 4 Results This study investigated the educational and psychological impact of childhood chronic illnesses on learning and achievement and emotional wel l being using data from the National Longitudinal Survey of Children and Youth ( N L S C Y ) . The objective was to understand whether there are differences in performance on a mathematics achievement measure (Mathematics Computation Exercise [MCE]) and differences in scores on an emotional disorder-anxiety measure ( E D A ) in children who are identified as being healthy versus children identified as having chronic illnesses. Further, this research investigated whether there was an effect of gender and socio-economic status, such as community type and household income on the relationship between illness and academic achievement for these two groups o f children. This chapter contains a summary o f statistical findings to help answer the five research questions posed at the outset of the study. A l l tests o f statistical analyses, assumptions and results were conducted based on normalized weighted data to fulfill S T C ' s confidentiality guidelines, as wel l as for appropriate use of national data (see Appendix A ) . One exception to this is the overall sample size with valid responses for the key variable, chronic illness. Analysis of Variance for M C E Scores by Illness and Gender A univariate Analysis of Variance ( A N O V A ) , with illness and gender as independent variables and M C E scores as the dependent variable, was conducted to examine whether there were statistically significant differences between children with chronic illnesses and healthy children in their M C E scores. 61 M C E Scores by Illness The mean M C E score for children with no identified illness is 485.56, with a SD of 110.07 points, and the mean M C E score for children with an identified illness is 461.90, with a SD of 100.33; this is a 23.66 point difference in SS between children with no identified illness and an identified illness. M C E Scores by Illness and Gender For the analyses on gender, females had a mean M C E score o f 491.45, and males had a mean M C E score o f 466.34. Even though mean M C E scores were very similar for males and females with an identified illness (466.73 and 458.75, respectively), there was a larger difference in M C E scores between females and males (502.29 and 470.84, respectively) with no identified illness. The descriptive statistics for the M C E scores for children by illness and gender are presented in Table 4.11. Table 4.11 Means and Standard Deviations for M C E Scores by Illness and Gender Illnesses Gender Mean MCE SD n (weighted) No Illness Male Female TOTAL 470.84 502.29 485.56 114.83 102.00 110.07 399 351 750 1 Illness Male Female TOTAL 458.75 466.73 461.90 ) 98.38 103.39 100.33 236 154 390 TOTAL Male Female TOTAL 466.34 491.45 477.47 109.09 103.63 107.38 635 505 1140 62 Illness, ( F ( l , 1136)=12.53,/> < .001), and gender, (F(\, 1136)=8.58,/? < .05), were identified as significant factors, however, the interaction effect illness by gender was not significant, (F(l, 1135)= 3.04,p > .05). Overall, females performed higher on the M C E test. Mean M C E scores for children with one or more identified illnesses were lower than the mean for children with no identified illness. However, the effect sizes for illness (n2 = .011) and gender (rj2 = .007) were small. The results are presented in Table 4.12. Table 4.12 Analysis of Variances for Mathematics Computation Exercises ( M C E ) Scores Source df F n P Illness (1) 1 12.53(**) .011 .000 Gender (G) 1 8.58(*) .007 .003 l * G 1 3.04 .003 .082 *p < .05. **p < .001. Analysis of Variance for E D A Scores by Illness, Gender, and Age E D A Scores By Illness The mean score on the E D A scale for children with no identified illnesses is 3.14 (SD = 2.65), and the mean score on the E D A scale for children with an identified illness is 3.50 (SD = 2.25). In other words, there is a mean difference of .36 E D A score points between children with no illnesses and with an identified illness. This suggests that the two groups of children (ones with no identified illnesses and ones with identified illnesses) score very similarly on the E D A scale. 63 E D A Scores By Illness And Gender Males with no identified illness scored lower on the E D A scale (M = 2.85, SD = 2.66) compared to males with an identified illness (M = 2.86, SD = 2.02). A s well , females with no identified illness scored lower on the E D A scale (M = 3.39, SD = 2.62) compared to females with an identified illness (M = 3.53, SD = 2.71). This may suggest that there are some differences between males with no identified illnesses compared to females with no identified illnesses, and between males with an identified illness compared to females with an identified illness. E D A Scores By Illness By Age And Gender The highest mean E D A scores by age, and mean E D A score by gender was found in females. Females who were 14-15 years with an identified illness, reported a mean E D A score of 4.40 (SD = 2.99), and females who were 10-11 years without an identified illness, reported a mean score of 4.31 (SD = 2.96). Interestingly, the lowest scores are also found in females. The lowest scores were found in females who were 12-13 years with an identified illness (M = 2.43, SD = 2.12), and females 12-13 years without an identified illness (M = 2.52, SD = 2.22). Mean E D A scores for children by illness, gender and age are presented in Table 4.13. Table 4.13 Mean E D A Scores by Illness, Age and Gender Groups Health Gender Age Groups Mean EDA SD n (weighted) No Illnesses Male 10-11 3.00 2.79 62 12-13 2.98 3.06 . 82 14-16 2.69 2.29 120 TOTAL 2.85 2.66 264 Female 10-11 4.31 2.96 54 12-13 2.52 2.22 123 14-16 3.89 2.58 117 TOTAL 3.39 2.62 294 TOTAL 10-11 3.61 2.94 116 12-13 2.70 2.59 205 14-16 3.28 2.50 237 TOTAL 3.14 2.65 558 1 Illness Male 10-11 3.26 2.03 34 12-13 . 2.77 1.86 47 14-16 2.71 2.13 63 TOTAL 2.86 2.02 144 Female 10-11 3.81 2.51 26 12-13 2.43 2.12 40 14-16 4.40 2.99 43 TOTAL 3.53 2.71 109 TOTAL 10-11 3.50 2.25 60 12-13 2.61 1.98 87 14-16 3.40 2.64 106 TOTAL 3.15 2.36 253 TOTAL Male 9-11 3.09 2.54 96 12-13 2.90 2.68 129 14-16 2.70 2.23 183 TOTAL 2.86 2.45 408 Female 9-11 4.15 2.82 80 12-13 2.50 2.19 163 14-16 4.03 2.70 160 TOTAL 3.43 2.64 403 TOTAL 9-11 3.57 2.72 176 12-13 2.67 2.42 292 14-16 3.32 2.54 343 TOTAL 3.14 2.56 811 65 Analysis of Variance for E D A scores by Illness, Gender and Age The main effect for illness, 2-way interaction effect (illness by gender, and illness by age), and the 3-way interaction (illness, gender and age) were not significant. Significant effects were found for gender and age ( F ( l , 799)=11.05, p < .001), and F(2, 799)=8.48, jp < .001). However, the effect sizes for these main effects were small. The effect size for gender was rj2 = .014, and the effect size for age was t]2 = .021. Thus, these findings do not seem to have practical implications. Table 4.14 Analysis of Variances for Emotional Disorder-Anxiety (EDA) Scores Source df F n P Illness (1) 1 .00 .000 .988 Gender (G) 1 11.05(*) .014 .001 Age (A) 2 8.48(**) .021 .000 l * G 1 .02 .000 .884 l * A 2 .55 .001 .579 1 * G * A 2 .80 .002 .449 *p<.05. **p<.001. Post Hoc Tests for E D A Scores The Tukey H S D test showed that there were significant differences between children who were 10-11 years from children who were 12-13 years, in addition, there were significant differences detected between children 12-13 years from children who were 14-15 years. Differences were not found between the youngest children o f this sample (10-11) from the eldest children (14-15) (Table 4.15). 66 Table 4.15 Tukey's Post Hoc Test for E D A scores by Age Age Groups Age Group (Comparison) Mean Difference SE P 10-11 12-13 14-15 . .90(*) .26 .24 .23 .001 .511 12-13 10-11 14-15 -.90(*) - .640 .24 .20 ,001 .004 14-15 10-11 12-13 -.26 •640 .23 .20 .511 .004 *p < .05. **p < .001. The Tamhane's T2, Games-Howell, Dunnett's T3 , and Dunnett's C Post Hoc comparisons were also computed, given that the homogeneity o f variance assumption was violated. The pairwise comparisons for illness by age were very similar to what was found in the Tukey's H S D test. Table 4.16 summarizes the results from the Tamhane's T2, Games-Howell , Dunnett's T3, and Dunnett's C Post Hoc tests. Table 4.16 Tamhane, Dunnett T3, Games-Howell and Dunnett C Post Hoc Tests for E D A scores Age Post Hoc Test Age Groups Age Group (Comparison) Mean Difference SE P Tamhane 10-11 12-13 .90(*) .25 .001 14-15 .26 .25 .656 12-13 10-11 -.90(*) .25 .001 14-15 -.640 .20 .004 10-11 -.26 .25 .656 14-15 12-13 •64C) .20 .004 Dunnett T3 10-11 12-13 .900 .25 .001 14-15 • .26 .25 .656 12-13 10-11 -.900 .25 .001 14-15 - .640 .20 .004 10-11 -.26 .25 .656 14-15 12-13 •640 .20 .004 Games-Howell 10-11 12-13 •90O .25 .001 14-15 .26 .25 .553 12-13 10-11 - .900 .25 .001 14-15 -.64(*) .20 .003 10-11 -.26 .25 .553 14-15 12-13 .640 .20 .003 Dunnett C 10-11 12-13 .900 .25 -14-15 .26 .25 -12-13 10-11 - .900 .25 -14-15 -.64(*) .20 -10-11 -.26 .25 _ 14-15 .640 12-13 .20 -*p < .05. 68 Comparison of Children With or Without Illnesses in Their Mathematics Computation Exercise ( M C E ) Scores with E D A as a Covariate This analysis addressed the degree to which differences in M C E scores for children with and without illnesses may be moderated by their emotional health. In other words, do the differences in E D A scores account for differences in M C E scores for children with and without illnesses? To examine differences of M C E scores after removing the effect of E D A scores for children who are healthy and children who are i l l , an Analysis of Covariance ( A N C O V A ) was conducted where E D A scores were used as a covariate. Table 4.17 summarizes the results from the A N C O V A analyses for illness and gender as the independent variables, M C E scores as the dependent variable and E D A scores as the covariate. The main effect o f illness was found to be significant, F(l, 720) = 4.07, p < .05, but the strength of the relationship was very weak with effect size rf = .006. However, the main effect for the covariate ( E D A ) was not significant, F(\, 720) = .12, p >.001, nor were the effect of gender, or any of the 2-way and 3-way interactions significant. Table 4.17 Analysis of Covariance for M C E Scores, After Removing the Effects of E D A Scores Source df F n P Illness (1) 1 4.07O .006 .044 Gender (G) 1 .35 .000 .554 EDA (E) 1 .12 .000 .725 1 * E 1 .13 .000 .719 G * E 1 .25 .000 .617 1 * G * E 1 1.34 .002 .247 *p < .05. Thus, E D A scores do not seem to moderate the effect o f illness. In fact, when the variation in M C E scores by E D A scores is accounted for, the effect size for illness is higher. 69 Comparison of Children With or Without Illnesses in Their Mathematics Computation Exercise (MCE) Scores with Educational Handicaps as a Covariate This analysis addressed the degree to which differences in M C E scores for children with and without illnesses may be moderated by the presence of an educational handicap. A n educational handicap is defined as the identification of at least one o f the following: mental handicap, learning disability, attention deficit and/or emotional disorder. In other words, does a child's educational handicap account for differences in M C E scores? Table 4.18 illustrates the mean M C E scores by children identified with a chronic illness and an educational handicap. Children who were not identified with either a chronic illness or an educational handicap had the highest mean M C E score ( M = 487.39, SD = 110.34). Children with a "dual identification" (identified with both a chronic illness and an educational handicap) had higher mean M C E scores (M = 482.56, SD = 100.96) compared to children with only a single identification. To add, children identified with a chronic illness but without identification of an educational handicap had the lowest mean M C E scores overall (M = 457.85, SD = 99.86). Table 4.18 Mean M C E Scores by Illness and Educational Handicap Chronic Illness Educational Handicaps Mean MCE SD n (weighted) No No 487.39 110.34 691 Yes 464.14 105.38 59 Yes No 457.85 99.86 326 Yes 482.56 100.96 64 70 To examine differences of M C E scores after removing the effect of being identified with an educational handicap for children who are healthy and children who are i l l , an Analysis of Covariance ( A N C O V A ) was conducted where educational handicap status (0 for unidentified, and 1 for identified) was used as a covariate. Table 4.19 summarizes the results from the A N C O V A analyses for illness and gender as the independent variables, M C E scores as the dependent variable and educational handicap status as the covariate. The main effect of illness, gender and educational handicaps were not significant ( F ( l , 1132) = .37,p > .05, F ( l , 1132) = 2.94,p > .05, F(l, 1132) = .06,p > .05, respectively). 2 2 However, the effect sizes o f these relationships were very small (rj = .000 for illness, n = 2 2 .003 for gender, and rj = .000 for educational handicaps). This effect size decreased (from rj = .011 to .000 for illness, and from rf = .007 to .003 for gender) compared to the first analyses where educational handicaps were not used as a covariate. Table 4.19 Analysis of Covariance for M C E Scores, After Removing the Effects of Identification of an Educational Handicap Source df F n P Illness (1) n .37(**) .000 .545 Gender(G) 1 2.94(**) .003 .087 Educational Handicap (EH) 1 .06 .000 .809 l * E H 1 4.25(*) .004 .040 G * EH 1 .04 .000 .849 *p<.05. **p<.001. 71 The interaction effect between illness and the covariate (educational handicap) was significant ( F ( l , 1132) = 4.25, p < .05). This unexpected interaction effect is very interesting. Among the children who did not have a physical (chronic) illness, children with educational handicaps tended to score lower, whereas among the children with a physical (chronic) illness, children with educational handicaps tended to score higher. To What Extent Do Illness And E D A Variables Predict Children's M C E Scores, When Gender And Socio-Economic Factors And Community Type Accounted For? To test the degree of association between a child's performance on the M C E test, illness and E D A , after gender, community type and household income were accounted for, a hierarchical (stepwise) regression analyses was conducted. Tables 4.20 and 4.21 summarize findings from the regression analyses. In the first step, gender and household income were entered into the regression model. These variables were not significant predictors of M C E scores. In the second step, community type (e.g., urban versus rural communities) was added to the model and resulted in a statistically significant R-square change, increasing the adjusted R-square to RadjuSted = -03, F(3 , 749) = 8.84, p < .001. In the third step, illness was added to the model and there was again a significant R-square change, thus increasing the adjusted R-square to RadjustJ ~ -04, F(4, 748) = 8.11, p < .001. In other words, approximately 4% of the variance is accounted for by gender, household income, community type and illness. Finally, in the fourth step, E D A scores were added to the model, however the F-change was not statistically significant. 72 Model 3 appeared to be the best fitting model for this regression analysis. The standardized beta weights (j3standardized) in this step were the strongest for size-area of residence (community type) (fistandardized - -.086, ^(748) = -2.40, p < .05) and illness (^standardized = .16, f(748) = 4.49, p < .001). The negative relationship between M C E scores and community type suggests that children in rural communities have higher M C E scores compared to children from larger communities (communities with a population of 30,000 >), whereas the positive relationship between M C E scores and illness suggests that children with an identified illness have lower M C E scores compared to children with no identified illness(es). Illness and community type appeared to be the only statistically significant predictors of M C E , after accounting for gender, income and E D A scores. However, they are very weak predictors of M C E scores. Table 4.20 Analysis of Variance for Hierarchical Regression Analysis for SES Variables Predicting M C E Scores Above and beyond the effects of Illness, Gender and E D A scores Model df F P 1 2 2.17 .115 2 3 8.84 .000 3 4 8.11 .000 4 5 6.49 .000 *p<.05. **p<.001. 73 Table 4.21 Summary of Hierarchical Regression Analysis for SES Variables Predicting M C E Scores Above and beyond the effects of Illness, Gender and E D A scores Model Variable p2 •» Adjusted fi Standardized fi t P 1 (Constant) .003 496.28 31.211 .000 Gender 10.85 .063 1.732 .084 Household Income 3.53 .047 1.280 .201 2 (Constant) .030(**) 460.83 26.477 .000 Gender 8.35 .049 1.346 .179 Household Income 4.13 .055 1.517 .130 Community Type 9.84 .169(**) 4.696 .000 3 (Constant) .036(**) 467.09 26.621 .000 Gender 7.37 .043 1.190 .234 Household Income 4.25 .056 1.567 .118 Community Type 9.41 .162(**) 4.486 .000 Illnesses -16.02 - . 0 8 6 0 -2.395 .017 4 (Constant) . 0 3 5 D 467.49 26.537 .000 Gender 7.61 .044 1.216 .224 Household Income 4.37 .058 1.590 .112 Community Type 9.34 .161(") 4.459 .000 Illnesses -16.04 - . 0 8 6 0 -2.397 .017 EDA -.34 -.010 -.276 .783 *p < .05. **p < .001. It is interesting to note that although there was a significant effect of gender across the A N O V A tests, gender did not come out as being a statistically significant in the regression models when combined with other predictors. 74 CHAPTER 5 Discussion The numbers of children affected by a chronic illness has been steadily increasing over the last few decades, to the degree that the effect on children and families is regarded as the most serious issue in pediatric medicine (Hobbs, Perrin, & Ireys, 1985; Shaw & Paez, 2002). Despite the medical advancements prolonging the lives o f chronically i l l children, the educational and psychological effects of childhood chronic illnesses continue to be pervasive. Most o f previous research on effects of childhood chronic illnesses used small sample sizes and has mostly been conducted outside of Canada. The National Longitudinal Survey of Children and Youth ( N L S C Y ) provides opportunities for examining such effects in Canada, using a large nationally representative sample. Previous research has systematically found that children with chronic illnesses do experience higher levels of educational and psychological problems. In particular, there has been substantial evidence that children with chronic medical conditions are at a higher risk for developing anxiety and experience more problems related to school adjustment and social interaction/development. The N L S C Y contains a variety o f data from measures of educational, psychological, and physical health that are all relevant to research on childhood chronic illnesses The primary purpose of this study was to utilize data from the N L S C Y , collected by Statistics Canada, to investigate the extent to which children with an identified chronic illness had different mathematics achievement levels and whether they differed in their emotional/anxiety disorder compared to children who are healthy. A second goal of this study was to better understand whether higher levels of emotional/anxiety were associated with lower mathematics achievement. The third goal o f this study was to better understand ' 75 what other types o f factors, such as household income, community type, might predict mathematics achievement, over and above the effect of illness, gender and emotional health/anxiety. The findings from this study are informative to build on the current state o f research in this field. In some cases, the results were consistent with findings from previous research literature. For example, the findings from the current study demonstrated that children who had even one identified chronic illness (asthma, bronchitis, heart condition, epilepsy, cerebral palsy, kidney condition, or other long-term condition), had poorer performance on the M C E test compared to children who were unidentified with an illness. However, other findings were divergent from past literature. For example, children who are i l l were not found to have higher rates o f emotional disorders and anxiety, and factors such as household income and emotional disorders and anxiety did not seem to have a strong effect in predicting math achievement. A summary of each of the four findings and how these findings build on past literature w i l l be discussed. Following this, limitations and future directions specific to each of the findings w i l l be highlighted, and general implications o f the findings, including theoretical, research and practical implications w i l l be discussed to better understand how the findings have extended our knowledge in this current field. Finally, the chapter w i l l conclude with general limitations o f the study and future directions to consider, specifically, issues related to research design/internal validity, external validity/generalizability, statistical analyses/power and measurement. 76 Summary and Discussion of Findings Explanation of Findings for the Educational Effects of Childhood Chronic Illnesses In the first research question, it was hypothesized that children with an identified illness would have poorer performance on the M C E test compared to their healthy peers. The analysis demonstrated that there were indeed statistically significant differences between groups of children who are identified with an illness, and children who are not identified with an illness on their M C E scores. Therefore, this finding rejects the null hypothesis that the two illness groups are the same. In order to rule out the possibility that performance on the math test may be further affected by children having mental handicaps, learning disabilities, attention deficit or emotional disorders, these difficulties were controlled for, however differences in performance on the math test across the two groups of children (illness and no illness) continued to hold true. This main finding is consistent with previous reports in the literature that have found that most academic problems are in areas that require students to achieve a level of mastery in the particular ski l l , prior to learning subsequent skills and knowledge, as it is the case in learning mathematics (CPS, 1994; Fowler et al., 1988; Kaemingk et al., 2004; Paone et al., 1998; Seidenberg et al., 1986). Past researchers have suggested that this difference in academic performance can be attributed to children missing more school days due to their illness compared to their healthy counterparts (Sherman, 1995). Essentially, by attending fewer school days compared to their healthy peers, children with illnesses are missing out on critical learning opportunities, as wel l as the opportunity to master core skills (e.g., basic 77 addition and subtraction) that are required to progress onto more advanced skills (e.g., multiplication and division). Unfortunately, the data concerning absent days (and reason for these absent days) that were made available in the N L S C Y dataset were not informative, as it was unclear how many of the specified absent days were a result of health factors versus other reasons such as family vacation, bad weather, transportation problems, etc. This is an area worth exploring in the future with more refined methods of measuring days absent and reasons for days absent. The findings regarding this first research question has implications for children l iving with a chronic illness, as wel l as for individuals working with such children in educational settings. In particular, the findings inform us that children with identified chronic illnesses tend to perform poorer in math measures compared to their healthy peers. Thus, it is necessary in future research to explore intervention programs to target these discrepancies in math achievement for children with illnesses. The effect of gender was also found to be related to mathematics performance. Given that there were equivalent numbers of males and females in the sample, this difference cannot be attributed to a difference in the sample size. Explanation of Findings for the Psychological Effects of Childhood Chronic Illnesses A s part o f the second research question, it was hypothesized that children with an identified illness would report higher scores on the E D A scale compared to their healthy peers. The findings revealed that illness did not have a statistically significant main effect on E D A scores. Thus, the findings did not provide evidence to support a relationship between illness and symptoms o f anxiety and emotional disorder. 78 In terms o f the effect of age on E D A scores, the findings from this study showed that there were significant differences between the three different age groups, children 10-11,12-13 and 14-15. The findings are difficult to interpret given that children who are 12-13 years appear to score lower on the E D A scale compared to children who are 10-11 and 14-15. The effect of age on how children rate themselves on symptoms related to anxiety was interesting to consider, however, there has already been robust evidence showing the relative consistency in prevalence rates for anxiety disorders across childhood (Albano et al., 2003; Curry, March, & Hervey, 2004; Kessler et al., 2005; Mash & Wolfe, 1999). Though these findings help build on the current understanding of how age, gender and illness contribute to a child's report on a measure of emotional/anxiety, these relationships must be further researched in the future to better understand the interactions among these variables. It is also important to note that all o f the identified statistically significant effects had very small effect sizes. Therefore, even though the findings point to systematic effects, these effects were not identified to be at a level that was o f practical meaning and consequence. This is not to say that these effects should be ignored, since these small effect sizes could be due to a number o f limitations of the study that w i l l be discussed later on in this chapter. Emotional Disorder-Anxiety As a Moderating Variable for Mathematics Achievement When E D A scores were used as a covariate to examine whether controlling for the effects of emotional disorder and anxiety would have an effect on the relationship between illness and mathematics achievement, the results did not provide evidence to support for such a moderating effect. This meant that the emotional disorder and anxiety did not have a 79 confounding effect on chronic illness and differences in achievement could be attributed to the illness. Educational Handicaps as a As a Moderating Variable for Mathematics Achievement The results did not provide supporting evidence for educational handicaps (mental handicap, learning disability, attention deficit or emotional disorder) having a moderating effect on achievement. Thus, we were able to rule out educational handicaps being a confounding factor with chronic illness and the lower achievement scores for the chronic illness group could be attributed to the illness itself. It is noteworthy that there was an interaction detected between illness and educational handicaps. This suggests that the patterns of differences in M C E scores among illness groups varied for different educational handicap groups. While children with illnesses scores were higher for children with handicaps, the reverse was true for children without illnesses. Predictors of Mathematics Achievement The final question for this study was to investigate the degree of association between mathematics achievement and gender, family household income, community type the child lives in, and chronic illness. The findings indicate that although the 4 independent variables (gender, income, community type and illness) are statistically significant predictors of M C E , they are very weak predictors of M C E scores. The regression analysis showed that the "best fitting" model was the one where household income, gender, community type, and illness were used as predictors; 4% of the variance could be explained by these four predictor variables. In this model, community type contributed the most to the overall model in predicting M C E scores. The second variable of considerable strength to the model was illness. However, the results indicated that there are 80 other factors that are affecting a child's performance on the M C E test at a much greater degree (would account for 96% o f the variation in achievement) above and beyond the effects of income, gender, community and illness. These findings provide information to the current literature, in particular, adds to research conducted with Canadian children. To date,vthere have been no published articles with Canadian children, looking at the effects of physical and mental health on educational outcomes. Thus, even the findings that these child background variables are not strong predictors o f achievement is important information in understanding and interpreting the effect chronic illnesses may have on children, in particular their education. General Implications of Findings The findings from the current study provide theoretical, research and clinical implications for individuals who work with children with chronic illnesses, as well as the families where there are children with a chronic illness. Given that this line o f research is still in its infancy, the use of large-scale data, like the N L S C Y , can provide exploratory information to gain better insights in an area with dearth o f research. Theoretical Implications One of the most important contributions of this study was to knowledge about the effects of illness on children's academic achievement. Several theories underlie this research project: a) Children with illnesses perform weaker in school; b) Children with illnesses experience a higher level of anxiety; and c) Children with illnesses experience a higher level of anxiety, which then in turn, impacts their performance on achievement tests; and d) Additional factors, like income, and community type also contribute to a child's overall performance in school. 81 It is not practical to say that our findings confirmed these theories; rather, the findings provided valuable insight to better understand the role of chronic illness and anxiety on learning as demonstrated by performance on a math test. In addition, the study provided insights about the strengths and limitations of the N L S C Y in measuring such variables. Research Implications The core research implications of this study have to do with measurement issues of the N L S C Y variables. To reiterate, this type of large-scale longitudinal assessment is necessary to provide information about how children make their way through to adulthood, by looking at physical and psychological health and educational achievement in Canadian children. To add, the breadth and depth of sampling in the N L S C Y lends itself to be an effective research tool, especially for research in need o f pilot data. What is important for researchers to do in the future is to ensure that effect sizes for all analyses are included in their results, given the large sample sizes. A s wel l , it is important to be cautious about the measures that are used in the N L S C Y . Although there was considerable time and effort put into developing the questions in the N L S C Y by panels of experts, the magnitude o f a Canada-wide longitudinal study is subject to time and economical constraints. A s a result, it would not be feasible to measure every concept extensively. Thus, the scales that were developed capture a general concept of interest, and are not fine grained enough for diagnostic purposes in clinical and applied settings. 82 Clinical and Applied Implications The primary purpose of this type of research is to increase the health-related quality-of-life in children with chronic illnesses, as they are l iving longer and healthier lives than ever before. Given the increasing numbers of children with chronic illnesses attending schools, researchers and practitioners need to understand how such illnesses affect children. Issues concerning a child's psychological well-being and school success are becoming pervasive issues due to the increased medical and technological advances that are increasing the life-span o f children with illnesses, and allowing many o f them to attend schools. So far, research has been able to demonstrate that children with illnesses are having more difficulty functioning in schools and in daily life, and this may be related to the level of anxiety a child experiences due to their illness. However, the results of the current study did not support these past findings. Given the pathogenesis of anxiety disorders in the general population, it is important that this mental health concern is addressed in its early stages so that subsequent mental and physical health problems can be curtailed. This is especially important in vulnerable populations, such as children with chronic health conditions, as their risk for developing other mental health or educational problems are magnified from living and coping with a chronic illness. Another implication of this research is on understanding the effect o f anxiety and stress on illness. This is especially important, because research with adult populations have found that individuals with higher levels of psychological stress have been found to be more susceptible to illness (compromised immune system) (Cohen, Tyrrell , & Smith, 1991). Thus, it is vital to continue this line of research in the future in hopes o f improving health outcomes for children with illnesses who experience high levels of anxiety. 83 There are also implications for educators of children with illnesses. In particular, educators need to be more sensitive to working with children with illnesses as the numbers of children who attend schools continue to grow. A s well , educators need to learn to be more sensitive to anxiety symptomatology, and not confuse these symptoms with other effects of the illness (e.g., side effects o f medication), so that children can be referred on to a mental health professional as soon as possible. Also , when measuring children on achievement measures, educators need to understand the impact of a chronic illness on a child's performance on such tests. A s previously mentioned, there can be both primary and secondary effects that may impact a child's performance on these measures. Finally, these findings have implications for treatment and intervention for children with chronic illnesses. Aside from addressing physical health concerns in children with chronic illnesses, educational concerns also need to be addressed. Given that children with illnesses are l iving longer and healthier lives, it is important to provide these children with tools to facilitate their development into productive members of society. Research concerning educational intervention programs must be researched so that secondary effects of these chronic illnesses can be curtailed early on. General Limitations of Study and Future Directions There are several limitations of the current study that need to be highlighted, specifically related to survey methodology and design. 84 Measurement Measures of illness. The main limitation o f the N L S C Y dataset in addressing the research questions in this study is the way in which the illness variable was measured and operationalized. Specifically, the N L S C Y lists a subset of illnesses that are commonly found in children (e.g., asthma, diabetes, cerebral palsy, etc.), however, this list is not exhaustive. In other words, children with cancer, blood disorders (e.g., HIV/ADDS, sickle-cell), gastrointestinal disorders, would be aggregated into the "other long term condition" variable. Thus, resulting in a "cloudy" measure of illness. Another problem with the measure of illness is the response format of these illness variables. The Person-Most-Knowledgeable ( P M K ) about the child were asked to identify the presence or absence o f an illness by checking off " Y e s " or " N o " to the question: "Does the child have (illness)?" Given that there were no questions to follow-up on the severity and frequency of these illness(es), children with varying degrees o f illness are categorized into one group. Thus, children with mi ld forms of asthma are treated the same as children with very severe forms o f asthma. Aggregating children with varying grades of an illness into one variable can present problems in interpreting and understanding the results. Past research literature has found that children with more serious types o f chronic conditions show more frequent interruptions to daily functioning and perform poorer on academic achievement measures compared to children with milder variants of the same illness. Hence, it would have been useful to know the severity and frequency of each o f the chronic illnesses so that the relationship between the severity and frequency of illness on academic achievement can also be examined. 85 Another problem with aggregating children with any of the seven illnesses into one category is that children with varying illnesses were treated equally. Given the different educational and psychological outcomes found for children with different types o f illnesses, aggregating all children with any illness into one category increases the variance in the group. In other words, there may have been a stronger effect o f illness on M C E scores, and illness on E D A scores detected i f individual groups of illnesses were measured separately. A s well , analyses o f between-group differences could have been assessed. Mathematics computation exercise test One of the original goals of the study was to research the effect o f illness on academic achievement as wel l as on cognitive functioning. However, the M C E test was the only measure of achievement that was available for the age group of interest (10 to 15 years old) as these children are the only ones who completed both the M C E test and the E D A scale. Other measures that were used in the N L S C Y , such as the Peabody Picture Vocabulary Test (PPVT) (measure of receptive vocabulary) (Dunn, Dunn, Robertson, & Eisenberg, 1981), and the cognitive functioning test, could not be used in the current study. For example, the P P V T was only administered to children 2 to 3 years old (where E D A scores were not collected) and the cognitive functioning test was only administered to adolescents/adults 17 to 19 years old. Thus, the M C E test score was the only measure of achievement that could be used to capture a child's academic achievement, where E D A scores were also collected, and a report of illness by the P M K . A s a result, the dependent variable in this study, M C E scores, is based on performance o f mathematic computation. However, it is important to note that this measure is not a global measure o f achievement, which usually consists of reading, writing, oral 86 language and mathematics (Kaufman & Kaufman, 2004; Flanagan, Ortiz, Alfonso, & Mascolo, 2006). To add global math achievement measures usually comprise of math computation (calculation) and problem solving. Both of these subcomponents assess achievement and math knowledge, however, math calculation additionally measures number facility, and problem solving additionally measures quantitative reasoning. This means that the M C E test is essentially a measure of math achievement, knowledge and number facility, whereas it does not measure quantitative reasoning. Also , the M C E test is comprised of a subset of the math questions asked in the Canadian Achievement Test ( C A T ) in which it was derived. The C A T was originally designed to be administered and interpreted as a global achievement measure. In the case o f the N L S C Y , the math component o f the C A T was extracted from the battery and used as a "stand-alone" measure o f math achievement. Thus, the psychometric properties for this individual subtest are unknown (both reliability and validity), and the use o f and interpretation of the M C E scores must be cautioned as it is not a global measure of academic achievement. Emotional disorder-anxiety scale. The E D A scale is used to measure Emotional Disorders — Anxiety ( E D A ) in the N L S C Y , and in the current study, the E D A scale is used to represent "psychological problems". Although the items were derived out of a sound theoretical and conceptual model (items were drawn from the Ontario Child Health Study [OCHS]) the scale is based on empirical findings (factor analysis), which may not necessarily be supported by theory. This is very different to standardized measures such as the Multidimensional Anxiety Scale for Children ( M A S C ) (March, 1997), the Revised-Children's Manifest Anxiety Scale ( R C M A S ) 87 (Reynolds & Richmond, 2000), or the State-Trait Anxiety Scale for Children (STAIC) (Spielberger et al., 1973), which were developed after careful examination o f other anxiety measures in children, adolescents and adults. A s well , these standardized tests measure various domains of anxiety, and demonstrate sound psychometric properties and. To add, such scales generally undergo an extensive test o f reliability and validity, and selection of items for the measure is based on evidence for reliability (e.g., internal consistency, test-retest reliability) and validity (criterion, construct, and concurrent) (Spielberger et al., 1973). A s well , item development may also be based on the D S M - I V criteria for anxiety disorders and have a clear operationalized definition o f the construct of interest (anxiety) (Spielberger et al., 1973). Another problem with the E D A scale is the number of items used to measure "emotional disorder - anxiety" in children. The E D A scale is comprised o f 7 questions that are believed to measure anxiety disorders in children and adolescents. On more popular standardized measures of anxiety, the number of items generally ranges from 20 items (STAIC) to 39 items ( M A S C ) . Thus, the use of only 7 questions to measure anxiety may not be sufficient enough to measure anxiety reliably and validly. However, due to the scope of the N S L C Y , it is understandable that it would not have been feasible to administer scales that are over 20 items each. Another problem with the E D A scale has to do with the dimensions o f anxiety disorders. The dimensions measured in the R C M A S are: physiological manifestations of anxiety, worry and over-sensitivity, and fear/coricentration. For the M A S C , the dimensions measured are: physical symptoms, harm avoidance, social anxiety and separation/panic. Finally, for the S T A I C , the scale differentiates between how children feel at a particular point 88 in time, and how children generally feel. Thus, given that the E D A scale is made up of 7 questions that purport to measure the wide array o f dimensions found to measure and correlate highly to anxiety, it is unclear which of these dimension(s) the E D A scale measures. Although the name of the E D A scale: "Emotional Disorder - Anxiety" implies that a high score on this scale would equate to a child having an emotional and/or anxiety disorder, it is important to note that the purpose of the scale is not to diagnose or identify children with an emotional and/or anxiety disorder. Despite the reliability (internal consistency) of the E D A scale for Cycles 4 and 5 were quite high (r = .736, and r = .981, respectively), this scale may not be a very sensitive measure of emotional or anxiety disorders, thus, may not have the diagnostic capacity to identify such disorders. To add, some questions may actually be more appropriate as a measure o f negative affect or depression, such as the question: "I cry a lot". However, because this question is part of the E D A scale, it infers that children who cry a lot are also high on emotional disorders - anxiety. i For the purposes o f the current study, the scores on the E D A scale provided insight as to a child's sadness/mood, fear, nervousness, worry, and anhedonia (loss o f interest) at the time of testing, as opposed to a true indication of an emotional and/or anxiety disorder. Finally, given that the data collection for the N L S C Y occurs every two years, the E D A report is ultimately a report o f anxiety at that particular point in time. Without measures over different points in time, it is unclear whether the E D A reports are representative of children's feelings at the time of measurement (state) or in general (trait) (Spielberger, 1973). Another limitation of the E D A scale is that items are self-reported by children. Although parent reports o f the E D A scale are collected for children up to 10/11 years, the current study looked at children from 10 to 15 years, thus the parent reports could not be used 89 in the analyses. For children 10 to 15 years, the N L S C Y only inquired with the P M K s on the amount a child worries about death in the family, worries about divorce, etc., data which was not appropriate to use in the current study. Two potential problems in relying on a single measure are the reliability and validity of the self-report. In terms o f validity of test results, children could potentially "fake good" and report positive responses on the anxiety measure to please the researcher or out of fear that his/her parent may see the results. Alternatively, children may have been tired of the long battery of tests that were administered, thus started randomly responding to the items. Given that there was not an inconsistency index (or a " L i e " scale), it is hard to decipher the validity of the responses. A s well , it is unclear what the child's reading ability/level is, thus, the use o f the E D A scale assumes that the child is able to read and comprehended the questions being asked. Only children over the age o f 10 years old were used in the analyses given the potential problem of younger children not having the reading ability to understand the questions, cognitive capacity to understand the questions being asked, or having the level of introspection required for a self-report of anxiety symptoms. Given the weaker performance on the M C E test in children with a reported illness, there may have been problems in comprehending and validly responding to the questions being asked in the group of children with an illness, hence affecting the results. Overall, i f the E D A scale is subject to too many psychometric problems, the measure may not have been sensitive enough to detect true symptoms of anxiety in the sample of children included in our analyses. Perhaps symptoms of anxiety do have a large effect in children with chronic illnesses, however, without a multi-method assessment approach and a 90 sound measure of anxiety, these effects may not have been detected between groups of children identified with and without a chronic illness. The current analyses did not capture these differences systematically. School absence days. Although data concerning school-absence days were collected as part of the N L S C Y dataset, this variable was not particularly useful in our analyses as it was presented as a range of the number o f school days a child missed in the past school year (e.g., 1 to 3 days, 4 to 6 days, etc.). A s well , there was a separate variable asking the P M K about the child, the reason why children missed these school days. However, there was no way to link the number of school-absence days to the reasons school was missed. Thus, it was unknown how many of the school-absence days were due to weather, transportation, a family vacation, a fear of school, or health reasons. Finally, the category of "health reasons" included both illnesses and injuries, thus there was no way to decipher the exact health reason as to why the child missed a school day. Although these two variables (school absent days and reason for school absence) are informative on a qualitative level, neither variable was appropriate to include in the statistical analyses. The visual inspection of the analyses, however, was not able to confirm the notion that children with chronic illnesses miss more sickness-related absence days compared to children without an identified illness. In fact, it appears that children without identified illnesses have more illness-related school absences (81.5%) compared to children with an identified illness (45.8%). Due to the shortcomings of the data investigating school absent days and reasons for school absences, future research should investigate this relationship so 91 that we are better able to understand the extent to which missing school affects learning, which then impacts performance on academic achievement tests. Primary and secondary effects of illness. Another problem with the data was that it remains unclear whether group differences in MCE scores were associated with the primary or secondary effect of their illness. For example, there was no way to determine whether a child's performance on the MCE test was due to the primary effect of the chronic illness (e.g., difficulty breathing), or to the secondary effect of the illness (e.g., a child not having enough sleep the night before because of difficulty breathing, thus affecting his/her concentration), or both. However, there was no way to gauge such effects using the NLSCY dataset. External validity and generalizability. An issue worth noting that concerns the generalizability of the results is that there were no valid reports of illness for children living in Nunavut, the Yukon Territories or the Northwest Territories in the unweighted sample. Even though, the NLSCY is a nationally representative sample of Canadian children, the way in which the variables were measured restricts the generalizability of the findings. Thus, despite weighting the cases, children from Nunavut, the Yukon and Northwest Territories are not represented in our analyses. As well, there were no children with a report of "Yes" to the chronic conditions of Cerebral Palsy or Kidney Condition who had a valid MCE and EDA score. This is interesting given that we know that the prevalence rate of these two conditions is not zero. In effect, this would deflate our sample of children with chronic conditions. 92 The key limitation o f findings is the very small effect size for all o f the statistically significant effects of factors such as illness and gender. A l l o f the measurement limitations discussed above are expected to contribute to low possible low effect size even when larger, true effects are present. For example, the degree to which illness groups can be compared is critically dependent on whether these categories in the data refer to distinct groups and that this distinction is tied to the degree of illness. In addition, in order to examine the effects of illness on school learning, the measurement of achievement needs to be tied to school learning. The brief (20 questions) mathematics achievement test that focuses on a narrow set of mathematics skills would be expected to be a limited measurement o f school learning at best. Specific Research Needed to Clarify or Extend Findings Measures of illness. When measuring health/illness, it is important to employ multiple measures of health. For example, a more comprehensive list o f common childhood chronic illnesses needs to be included, such as Cancer, Diabetes, Juvenile Rheumatoid Arthritis, Cystic Fibrosis, Gastrointestinal Conditions, Blood Disorders (i.e., Hemophilia), and Infectious Diseases (i.e., H I V / A I D S ) . A multi-method assessment approach would be important to employ when measuring the presence, severity and frequency of an illness. Not only should the parent provide a report of the child's illness, but also other subjective (child self-report, daily diaries) and objective reports (physician) need to be collected. Al so , other objective measures o f health (e.g., blood test results, etc.) would be important to measure based on the research question asked. 93 A set of such objective and subjective measures o f health would be very informative to researchers and practitioners so that they are better able to understand the effects of an illness on a child's daily functioning as well as their mental health and educational achievement. Additionally, these types of information would provide some insight to the within group differences of children diagnosed with an illness. Aside from the use o f reports from the child/youth, parent, physician and other test results, it would be very informative to obtain reports from the child's teacher(s), so that there is another source o f information regarding the effect o f an illness on a child's functioning at school. In addition to asking teachers about a child's illness/physical functioning at school, teachers can also be asked about a child's academic achievement and psychological well-being. Given that teachers interact with the child on a daily basis, she/he would be able to provide insight as to the child's behaviour and performance in school. These supplemental reports from physicians and teachers would provide a better insight to the variables of interest, although, would pose as an economic constraint to the project and extra burden to teachers working with children with illnesses. It is also imperative that future research focuses on the extent to which anxiety affects the healing of an illness (e.g., removal of tumors) or how a child's immune system can be compromised (e.g., the frequency of a child getting bronchitis, or coping with H I V / A I D S ) , given findings demonstrating a strong relationship between stress and physical illness in adults (slowing o f wound healing and weakening o f the immune system) (Cohen et al., 1991; Kiecolt-Glaser, Marucha, Balarkey, Mercado, & Glaser, 1995). 94 Measures of academic achievement A larger battery o f achievement measures (or a global achievement measure) should be used in future research instead o f only one subtest of a global achievement measure. For example, standardized measures o f global academic achievement (e.g., Kaufman Test of Educational Achievement, 2 n d Edition, [KTEA-2] [Kaufman & Kaufman, 2004]), that have demonstrated sound psychometric properties should be employed, so that achievement scores are better representative of a child's overall academic achievement. A s well , measures of achievement should be supplemented with cognitive measures and other qualitative data such as clinical interviews and observations thus providing a more comprehensive look on a child's difficulty with the M C E test. Although the use of a global achievement battery (such as the K T E A - 2 ) would not be practical in large-scale assessments like the N L S C Y (due to the training and administration time of these measures), it is important for the results o f this study to be replicated using more sound measures of achievement. Additionally, it would be very important to include measure o f cognitive functioning, given the emerging literature looking at the effects of illnesses on a child's cognitive ability (e.g., deficits with processing speed, memory). Such a robust measure o f ability and achievement would provide us with more information about how chronic illnesses affect children. 95 Measure of anxiety. Although the E D A scale was a useful tool for looking at anxiety in children sampled for the N L S C Y , the scores on the E D A scale are probably not sensitive enough to detect true emotional disorders or anxiety disorders. Future research examining the effect of anxiety must utilize a multi-method assessment approach that incorporates both objective (e.g., clinical interviews, observations, parent- and teacher reports) and subjective (e.g., self-reports, daily diaries) measures of anxiety. In particular, anxiety measures that should be used are ones that are standardized and ones that demonstrate sound psychometric properties (reliability and validity), such as the M A S C (March, 1997), the R C M A S (Reynolds & Richmond, 2000), or the S T A I C (Spielberger et al., 1973) which are designed to measure anxiety in children and adolescents. These self-reported checklists should also be supplemented with clinical interviews of the child, as wel l , include reports from teachers and parents on a child's anxiety. Another approach to measure anxiety in children is the use of daily diaries where children report daily feelings of worries and fears. This would provide an idea about whether anxiety is a state or trait of the child, as wel l , would provide information about a child's level o f anxiety at multiple time-points. B y using a multi-method assessment approach, a child's anxiety score would be a more valid and reliable representation of a his/her actual level o f anxiety. Gender and age effects. The findings related to gender differences also warrant further research, because i f it is indeed true that chronic illnesses (or certain types of illnesses) impact females differently than males, intervention programs must be designed to address such gender differences in academic settings. Also , because this study did not differentiate between children with 96 different types of illnesses (e.g., asthma versus diabetes versus kidney conditions), it is uncertain whether there may be gender differences between groups of children with varying illness "profiles" (e.g., a child with asthma, a child with asthma and diabetes). There is a need to replicate the current findings with more sensitive measures of illness and anxiety so that we are better able to make better sense of the results. In terms of the findings concerning the effect of age on EDA scores, other qualitative measures need to be incorporated in measuring anxiety so that there can be a more comprehensive understanding of why older children experience higher levels of anxiety compared to younger peers, as well, to understand differences between males and females in their reports of anxiety. Questionnaires concerning the particular illness, how much education a child has received about treatment, management and pathology of their illness, social support, and magnitude of the primary and secondary effects of their illnesses (i.e., self-esteem, socialization) must be explored. With a more comprehensive set of qualitative information, we will be better able to explain why older children may report more symptoms of anxiety compared to their healthy peers, and why females and males may report different levels of anxiety. Although it was hypothesized that children with no reported illness, regardless of age and gender, score lower on the EDA scale compared to children with one reported illness, it was interesting to find that the highest and lowest EDA scores were found among males 12 to 13 years-old. The effect of illness on age and gender need to be more extensively explored in the future to better understand patterns of emotional problems related to anxiety for males compared to females, across different ages. The results from the current study revealed mixed results that are difficult to interpret. A potential reason for the mixed findings of age 97 and/or gender on E D A scores could be due to the sensitively o f the E D A measure. Given that past research has found that anxiety disorders are positively associated with age, and that females report higher levels of anxiety, the effect of age and gender warrants further research. P r i m a r y a n d s e c o n d a r y effects o f i l lness. To address issues concerning the relationship between effects o f an illness and performance in school, future research should utilize self-reports and daily diaries to monitor the primary and secondary effects of an illness so that we are better able to understand the effect(s) o f illness on learning. Thus, children w i l l be able to report on their subjective experiences in how an illness appears to affect their performance in school. M e a s u r e s o f c o m m u n i t y t ype , h o u s e h o l d i n c o m e , e t c Given the mixed findings concerning SES and school achievement and mental health outcomes, there was reason to believe that household income would help predict achievement scores. However, this effect was not found in the current study. In the future, other variables that may affect school achievement and mental health outcomes, such as ethnicity, parental income, mental health history, should be considered. C o n c l u s i o n The prevalence of children with chronic illnesses is increasing as a result of improved health care and services. For children with a chronic illness, the prevalence of psychological problems (i.e., anxiety disorders) and educational problems are nearly double in comparison to the general population. These trends require us to ensure that the educational and psychological effects o f their illness are understood and addressed. 98 This line of research is important to provide information about the educational and psychological outcomes o f children with illnesses. B y understanding such associations, researchers and practitioners w i l l be more capable of providing tools to help children with illnesses increase their quality-of-life. Despite the array o f limitations highlighted in the current study and the small effect sizes detected in the statistical analyses, the secondary use of the N L S C Y dataset was very informative as a research tool. The data provided us with a wealth o f information in researching an area that is still in its infancy. Future studies should be directed to replicating these findings with improved methods of collecting information about a child's physical and psychological health status. A s well , improved methods of measuring children's academic achievement and cognitive functioning in such large-scales studies are warranted. 99 References < Achenbach, T. M., Rescorla, L. A., McConaughey, S. H., Pecora, P. J., Wetherbee, K. M. & Ruffle, T. M. (1991). Achenbach System of Empirically Based Assessment. Burlington: University of Vermont, Department of Psychiatry: ASEBA Research Center for Children, Youth, and Families Adewuya, A. O. & Ola, B. A. (2005). Prevalence of and risk factors for anxiety and depressive disorders in Nigerian adolescents with epilepsy. Epilepsy & Behavior, 6, 342-347. Adolescent Medicine Committee, Canadian Paediatric Society (CPS). (1994). Care of the chronically ill adolescent. Canadian Journal of Paediatrics, Vol. 1(4), 124-26. Albano, A. M., Chorpita, B. F., & Barlow, D. H. (2003). Childhood anxiety disorders. In E. J. Mash and R. A. Barkley (Eds.), Childpsychopathology (2nd Edition, pp. 279-329). New York: Guilford. American Psychiatric Association. (2004). Diagnostic and statistical manual of mental th « disorders, 4 Edition, Text revision (DSM-IV-TR). Arlington, VA: American Psychiatric Association. American Psychological Association. (2001). Publication Manual of the American Psychological Association. Washington, DC: American Psychological Association. Armstrong, F. D., Thompson R. J. Jr., Want, W., Zimmerman, R., Pegelow, C. H., Miller, S. et al. (1996). Cognitive functioning and brain magnetic resonance imaging in children with sickle-cell disease. Pediatrics, 97, 864-870. Baki, O, Erdogan, A., Kantarci, O., Akisik, G., Kayaalp, L., & Yalcinkaya, C. (2004). Anxiety and depression in children with epilepsy and their mothers. Epilepsy & Behavior, 5, 958-964. Barlow, J. H. & Ellard, D. R. (2004). Psycho-educational interventions for children with chronic disease, parents and siblings: an overview of the research evidence base. Child: Care, Health & Development, 30, 637-645. Barron, R. M. & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 57, 1173-1182. Bartel, N. R. & Thurman, S. K. (1992). Medical treatment and educational problems in children. Phi Delta Kappan, 74,51-61. Bauld, C , Anderson, V. & Arnold, J. (1998). Psychosocial aspects of adolescent cancer survival. Journal of Paediatric Child Health, 34, 120-126. Beidel, D. C , & Turner, S. M. (1998). Shy children, phobic adults: Nature and treatment of social phobia. Washington, DC: American Psychological Association. Brown, R. T., Davis, P. C , Lambert, R., Hsu, L., Hopkins, K., & Eckman, J. (2000). Neurocognitive functioning and magnetic resonance imaging in children with sickle-cell disease. Journal of Pediatric Psychology, 25, 503-513. Brown, R. T. & DuPaul, G. J. (1999). Introducing the mini-series: Promoting school success in children with chronic medical conditions. School Psychology Review, 28, 175-182. Canadian Test Centre, Educational Assessment Services. (1992). Canadian Achievement Tests, Third Edition. Ontario, Canada: Canadian Test Centre, Educational Assessment Services. Cataletto, M. E. (2006). Pectus Carinatum. E-Medicine from Web-MD. Retrieved on October 1, 2006 from htt^ ://www.emedicine.com/ped/topicl803.htm on October 1, 2006. Chavira, D. A., Stein, M. B., Bailey, K. & Stein, M. T. (2004). Child anxiety in primary care: prevalent but untreated. Depression and anxiety, 20, 155-164. Chorpita, B. F. & Barlow, D. H. (1998). The development of anxiety: The role of control in the early environment. Psychological Bulletin, 124, 3-21. Clark, L. A. & Watson, D. (1991). Tripartite model of anxiety and depression: Psychometric evidence and taxonomic implications. Journal of Abnormal Psychology, 100, 316-336. Clark, L. A., Watson, D., & Beck, A. T. (1994). Common and specific dimensions of self-reported anxiety and depression: Implications for the cognitive and tripartite models. Journal of Abnormal Psychology, 103, 645-654. Clay, D. L. (2004). Helping schoolchildren with chronic health conditions: A practical guide. New York: Guilford. Clay, D. L., Mordhorst, M. J., & Lehn, L. (2002). Empirically supported treatments in pediatric psychology: Where is the diversity? Journal of Pediatric Psychology, 27, 325-337. Cohen, S., Tyrrell, D. A. J., Smith, A. P. (1991). Psychological stress and susceptibility to the common cold. New England Journal of Medicine, 325, 606-612. Cone, J. D. & Foster, S. L. (2001). Dissertations and Theses From Start to Finish: Psychology and Related Fields. Washington, DC: American Psychological Association. Connell, R., Gauderman, W. J., Gilliland, S. S., Avol, E., & Peters, J. M. (2003). Obesity and the risk of newly diagnosed asthma in school-aged children. American Journal of Epidemiology, 158, 406-415. Copeland, E. P. (2002). Best practices in health promotion for school psychological practice. In Thomas, A. and Grimes, J. (Eds.) Best practices in school psychology TV (pp. 949-961). Washington, DC: National Association of School Psychologists. Costello, E. J., Egger, H. L., Angold, A. (2004). The developmental epidemiology of anxiety disorders. In T. Ollendick & J. March (Eds.), Phobic and Anxiety Disorders in Children and Adolescents, New York: Oxford University Press, pp. 61-91. Cronbach L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16,297-333. Curry, J. F., March, J. S., & Hervey, A. S. (2004) Comorbidity of childhood and adolescent anxiety disorders in T. Ff: Ollendick and J. S. March (Eds.), Phobic and anxiety disorders in children and adolescents: A clinician's guide to effective psychosocial and pharmacological interventions (pp. 116-140). New York, NY: Oxford. Dadds, M R., James, R. C , Barrett, P. M., & Verhust, F. C. (2004). Diagnostic issues, in T. H. Ollendick and J. S. March (Eds.), Phobic and anxiety disorders in children and adolescents: A clinician's guide to effective psychosocial and pharmacological interventions (pp. 3-33). New York, NY: Oxford. Dunn, L. M., Dunn, L. M., Robertson, G. J., & Eisenberg, J. L. (1981). The Peabody Picture Vocabulary Test - Revised (PPVT-R). Nashville, TN: American Guidance Service. Eaton, M. L., Haye, J. S., Armstrong, F. D., Pegelow, C. H. and Thomas, M. (1995). Hospitalizations for painful episodes: Association with school absenteeism and 103 academic performance in children and adolescents with sickle-cell anemia. Issues in Comprehensive Pediatric Nursing, 18, 1-9. Farwell, J. R., Dodrill, C. B., & Batzel, L. W. (1985). Neuropsychological abilities of children with epilepsy. Epilepsia, 26, 395-400. Ferdinand, R. F., Barrett, P. M. & Dadds, M. R. (2004). Anxiety and depression in childhood: Prevention and intervention in T. H. Ollendick and J. S. March (Eds.), Phobic and anxiety disorders in children and adolescents: A clinician's guide to effective psychosocial and pharmacological interventions (pp. 459-475). New York: • Oxford. Flanagan, D. O., Ortiz, S. O., Alfonso, V. C. & Mascolo, J. T. (2006). The Achievement Test Desk Reference: A Guide to Learning Disability Identification, 2nd Edition, Hoboken, NJ: Wiley. Fowler, M. G , Whitt, J. K., Lallinger, R. R., Nash, K. B., Atkinson, S. S., Wells, R. J., et al. (1988). Neuropsychologic and academic functioning of children with sickle cell anemia. Journal of Developmental Pediatrics, 9, 213-220. Journal of Developmental Pediatrics, 9, 213-220. Fox, M. A., Chen, R. S. & Holmes, C. S. (2003). Gender differences in memory and learning in children with insulin-dependent diabetes mellitus (IDDM) over a 4-year follow-up interval. Journal of Pediatric Psychology, 28, 569-578. Frank, R. G., Thayer, J. F., Hagglund, K. J., Vieth, A. Z., Schopp, L. H., Beck, N. C, et al. (1998). Trajectories of adaptation in pediatric chronic illness: The importance of the individual. Journal of Consulting and Clinical Psychology, 66, 521-32. Gershon, J., Zimand,.E., Pickering, M., Rothbaum, B. O., & Hodges, L. (2004). A pilot and feasibility study of virtual reality as a distraction for children with cancer. Journal of the American Academy of Child & Adolescent Psychiatry, 43, 1243-1249. Gosch, E. A., Flannery-Schroeder, E., Mauro, C. F & Compton, S. N. (2006). Principles of Cognitive-behaviour therapy for anxiety disorders in children. Journal of Cognitive Psychotherapy: An International Quarterly, 20, 247-262. Geist, R. Grdisa, V. & Otley, A. (2003). Psychosocial issues in the child with chronic conditions. Best Practice & Research Clinical Gastroenterology, 17, 141-152. Hair, J. F., Anderson, R. E., Tatham, R. L. & Black, W. C. (1998). Multivariate data analysis, Fifth Edition. New Jersey: Prentice Hall. Hannonen, R., Tupola, S., Riikonen, R., & Ahonen, T. (2003). Neurocognitive functioning in children with type-1 diabetes with and without episodes of severe hypoglycaemia. Developmental Medicine and Child Neurology, 45, 262-268. Henning, J. & Fritz, G. K. (1983).. School reentry in childhood cancer. Psychosomatics, 24, 261-269. Henry, C. N. & DuPaul, G. J. (2004). Pediatric asthma among African American children: The emerging role of the School Psychologist. NASP Communique, Vol. 33(2), p. 1-8. Hershey, R., Lillie, R., Sadler, M., & White, N. H. (1999). The effect of severe hypoglycemia on cognition in children and adolescents with type 1 diabetes mellitus. Diabetes Care, 22, 1318-1324. Hobbs, N., Perrin, J. M., & Ireys, H. T. (1985). Chronically ill children and their families. San Francisco: Jossey-Bass. Retrieved on June 10, 2006 from http://www.findarticles.eom/p/articles/mi m2248/is .132 33/ai 53870307. Holmes, C , Dunlap, W. P. Chen, R. S„ & Cornwell, J. M. (1992). Gender differences in the learning status of diabetic children. Journal of Consulting and Clinical Psychology, 60, 698-704. Hommel, K. A., Chaney, J. M., Wagner, J. L., White, M. M., Hoff, A. L. & Mullins, L. L. (2003). Anxiety and depression in older adolescents with long-standing asthma: The role of illness uncertainty. Children's Health Care, 32, 51-63. Howell, D. C. (1999). Fundamental Statistics for the Behavioral Sciences. Pacific Grove, CA: Brooks/ColePublishing. Huberty, T. J. (1997). Anxiety. In Bear, G. G, Minke, K. M. & Thomas, A. (Eds.) Children's needs II: Development, problems and alternatives (pp. 305-314). Betheseda, MD: National Association of School Psychologists. Ialongo, N., Edelsohn, G., Werthammer-Larsson, L., & Kellam, S. (1994). The significance of first graders' self-reports of anxious symptoms. Journal of Abnormal Child Psychology, 22,441-456. Kaemingk, K. L., Carey, M. E., Moore, I. M., Herzer, M. & Hutter, J. J. (2004). Math weaknesses in survivors of acute lymphoblastic leukemia compared to healthy children. Child Neuropsychology, 10, 14-23. Katz, E. R., Rubinstein, C. L., Hubert, N. C, & Blew, A. (1988). School and social reintegration of children with cancer. Journal of Psychosocial Oncology, 6, 123-140. Kaufman, A. S. & Kaufman, N. L. (2004). Kaufman Test of Educational Achievement, 2nd Edition (KTEA-2). Circle Pines, MN: AGS Publishing. Kelly, A. F. & Hewson, P. H. (2000). Factors associated with recurrent hosptiatlization in chronically ill children and adolescents. Journal of Paediatric Child Health, 36, 13-18. Kessler, R. C , Berglund, P., Dernier, O., Jin, R. & Walters, E. E. (2005). Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the national comorbidity survey replication. Archives of General Psychiatry, 62, 593-602. Kiecolt-Glaser, J. K., Marucha, P. T., Balarkey, W. B., Mercado, A. M. & Glaser, R. (1995). Slowing of wound healing by psychological stress. The Lancet, 346, 1194-1196. Koopman, H. M., Baars, R. M., Chaplin, J. & Zwinderman, K. H. (2004). Illness through the eyes of the child: The development of children's understanding of the causes of illness. Patient Education and Counseling, 55, 363-370. Kronenberger, W.G. & Meyer, R. G. (2001). The child clinician's handbook, 2nd edition. Needham Heights, MA: Allyn & Bacon. Kubiszyn, T. (1999). Integrating health and mental health services in schools: Psychologists collaborating with primary care providers. Clinical Psychology Review, 19, 179-198. La Greca, A. M., Swales, T., Klemp, S., Madigan, S. & Skyler, J. (1995). Adolescents with diabetes: Gender differences in psychosocial functioning and glycemic control. Children's Health Care, 24, 61-78. Leibert, R. M. & Morris, L. W. (1967). Cognitive and emotional components of test anxiety: A distinction and some initial data. Psychological Reports, 20, 975-978. 107 Lynch, E. W., Lewis, R. B. and Murphy, D. S. (1992). Educational services for children with chronic illnesses: Perspectives of educators and families. Exceptional Children, 59, 210-221. March, J. S. (1997). The multidimensional anxiety scale for children. North Tonawanda, NY: Multi-Health Systems. Mash, E. J. & Wolfe, D. A. (1999). Abnormal child psychology. Belmont, CA: Wadsworth Publishing. Masia-Warner, C, Nangle, D. W., and Hansen, D. J. (2006). Bringing evidence-based child mental health services to the schools: General issues and specific populations. Education and treatment of children, 29, 165-172. McCarthy, A. M., Lindgren, S., Mengeling, M. A., Tsalikian, E. & Engvall, J. C. (2002). Effects of diabetes on learning in children. Pediatrics, 109, e9. McCarthy, A. M., Lindgren, S., Mengeling, M. A., Tsalikian, E. & Engvall, J. C. (2003). Factors associated with academic achievement in children with type 1 diabetes. Diabetes Care, 26, 112-117. Marien, W. E. & Bell, D. J. (2004). Anxiety- and depression-related thoughts in children: Development and evaluation of a cognition measure. Journal of Clinical Child and Adolescent Psychology, 33, 717-730. Mitru, G., Millrood, D. L., & Mateika, J. H. (2002). The impact of sleep on learning and behaviour in adolescents. Teachers College Record, 104, 704-726. Nabors, N. A. & Freymuth, A. K. (2002). Attention deficits in children with sickle-cell disease. Perceptual Motor Skills, 95, 57-67. Northam, E. A., Anderson, P. J., Jacobs, R. (2001). Neuropsychological profiles of children with type 1 diabetes 6 years after disease onset. Diabetes Care, 24, 1541-1546. Northam, E. A., Anderson, P. J., Werther, G. A., Warne, G. L., & Andrewes, D. (1999). Predictors of change in the neuropsychological profiles of children with type 1 diabetes 2 years after disease onset. Diabetes Care, 22, 1438-1444. Ortega, A. N., Huertas, S. E., Canino, G., Ramirez, R. & Rubio-Stipec, M. (2002). Childhood, asthma, chronic illness, and psychiatric disorders. The Journal of Nervous and Mental Disease, 190, 21'5-281. Shaw, S. R. & Paez, D. (2002). Best practices in interdisciplinary service delivery to children with chronic medical issues. In Thomas, A. and Grimes, J. (Eds.) Best practices in school psychology IV (pp. 1473-1483). Washington, DC: National Association of School Psychologists. Paige, L. Z. (1993). The identification and treatment of school phobia. Bethesda, MD: National Association of School Psychologists. Paone, M. C , Whitehouse, S., & Stanford, D. (1998). The challenges of transition: Coping with a chronic condition. BC Medical Journal, 40, 73-75. Pen-in, E. C , Lewkowicz, C. & Young, M. H. (2000). Shared Vision: Concordance Among Fathers, Mothers, and Pediatricians About Unmet Needs of Children With Chronic Health Conditions. Pediatrics, 105, 277-285. Peterson, L., Reach, K. & Grabe, S. (2003). Health-related disorders. In E. J. Mash and R. A. Barkley (Eds.), Childpsychopathology (2nd Edition, pp. 716-744). New York: Guilford. Politis, C. (1998). The psychosocial impact of chronic illness. Annals of the New York Academy of Sciences, 850, 349-354. Power, T. J., DuPaul, G. J., Shaprio, E. S., & Parrish, J. M. (1998). Role of the school-based professional in health-related services. In L. Phelps (Ed., Health-Related Disorders in Children and Adolescents: A Guidebook for Understanding and Educating (pp. 15-26). Washington, DC: American Psychological Association. Pumareiga, A. J., Pearson, D. A. & Seilheimer, D. K. (1993). Family and childhood adjustment in cystic fibrosis. Journal of child and family studies, 2, 109-118. Rausch, D. M. and Stover, E. S. (2001). Neuroscience research in AIDS. Progressive Neuro-Psychopharmacological & Behavioral Psychiatry, 25, 231-257. Ray, W. J. (2000). Methods: Toward a science of behavior and experience. Belmont, CA: Wadsworth. Reynolds, C. R. & Richmond, B. O. (2000). Revised children's manifest anxiety scale (RCMAS). Los Angeles, CA: Western Psychological Services. ' Rovet, J. & Alvarez, M. (1997). Attentional functioning in children and adolescents with EDDM. Diabetes Care, 20, 803-810. Rovet, J. F. & Ehrlich, R. M. (1999) The effect of hypoglycemic seizures on cognitive function in children with diabetes. A 7-year prospective study. Journal of Pediatric Psychology, 134, 503-506. Rynard, D. W., Chambers, A., Klinck, A. M. & Gray, J. D. (1998). School support programs for chronically ill children: Evaluating the adjustment of children with cancer at school. Children's Health Care, 27, 31-46. Sarason, S. B., Lighthall, F. F., Davidson, K. S., Waite, R. R. & Ruebush, B. K. (196). Anxiety in elementary school children. New York: Wiley. Schatz, J., Brown, R. T., Pascual, J. M., Hsu, L., & DeBaun, M. R. (2001). Poor school and cognitive functioning with silent cerebral infarcts and sickle-cell disease. Neurology, 56, 1109-1111. Schneiderman, N., Antoni, M. H., Saab, P. G., & Ironson, G. (2001). Health psychology: Psychosocial and biobehehavioral aspects of chronic disease management. Annual Review Of Psychology, 52, 555-580. Schniering, C. A., Hudson, J. L., & Rapee, R. M. (2000). Issues in the assessment and diagnosis of anxiety disorders in children and adolescents. Clinical Psychology Review, 20, 453^78. Schoenle, E. J., Schoenle, D., Molinari, L. & Largo, R. H. (2002). Impaired intellectual development in children with Type 1 diabetes: Association with HbAlc, age at diagnosis and sex. Diabetologia, 45, 108-104. Seidenberg, M., Beck, N., Geisser, M., Giordani, B., Sackellares, J. C., Berent, S., et al. (1986). Neuropsychological correlates of academic achievement of children with epilepsy. Journal of Epilepsy, 1, 23-29. Serrano-Ddcos, E., Lask, B., Whitehead, B. Rees, P. & Graham, P. (1999). Heart of heart-lung transplantation: Psychosocial outcome. Pediatric Transplantation, 3, 301-308. Sexson, S. B. & Madan-Swain, A. (1993). School reentry for the child with chronic illness. Journal of Learning Disabilities, 26, 115-125. Shaw, S. R. & Paez, D. (2002). Best practices in interdisciplinary service delivery to children with chronic medical issues. In Thomas, A. and Grimes, J. (Eds.) Best practices in school psychology IV (pp. 1473-1483). Washington, DC: National Association of School Psychologists. Spielberger, C. D., Edwards, C. D., Lushene, R. E., Montuori, J. & Platzek, D. (1973). State-Trait Anxiety Inventory for Children ("How I Feel Questionnaire"). Palo Alto, CA: Consulting Psychologists Press Inc. Spieth, L. E. & Harris, C. V. (1996). Assessment of health-related quality of life in children and adolescents: An integrative review. Journal of Pediatric Psychology, 21, 175-193. Statistics Canada website. (2002). National Loneitudinal Survey of Children and Youth: Cycle 4 Survey Instruments 2002/03 Book 1 - Parent, Child&Youth 2005. Retrieved on October 1,2006 from http://www.statcan.ca/english/sdds/instrument/4450 02 V4 E.pdf. Statistics Canada website. (2000). Ontario Child Health Study (Second Follow-up). Retrieved on March 20, 2007 from http://www.statcan.ca/english/sdds/document/3824 Dl T9 VI E.pdf Sweeny, M. & Pine, D. (2004). Etiology of fear and anxiety in T. H. Ollendick and J. S. March (Eds.), Phobic and anxiety disorders in children and adolescents: A clinician's guide to effective psychosocial and pharmacological interventions (pp. 34-60). New York: Oxford. Szabo, M. & Lovibond, P. F. (2006). Anxiety, depression, and tension/stress in children. Journal of Psychopathology and Behavioral Assessment, 28 195-205. Taras, H. & Potts-Datema, W. (2005). Chronic health conditions and student performance at school. Journal of 'School Health, 75,255-266. Tarullo, L. B., Richardson, D. T., Radke-Yarrow, M., & Martinez, P. E. (1995). Multiple sources in child diagnosis: Parent-child concordance in affectively ill and well families. Journal of Clinical Child Psychology, 24, 173-183. Tyrrell, M. (2005). School phobia. The Journal of School Nursing, 27, 147-151. Van Ameringen, M., Mancini, C. & Farvolden, P. (2003). The impact of anxiety disorders on educational achievement. Journal of Anxiety Disorders, 17, 561-571. Vera, L., Nollet-Clemencon, C , Vila, G., Mouren-Simeoni, M. C. & Robert, J. J. (1997). Social anxiety in insulin-dependent diabetic girls. European Psychiatry, 12, 58-63. Warren, S. L. & Sroufe, L. A. (2004). Developmental issues in T. H. Ollendick and J. S. March (Eds.), Phobic and anxiety disorders in children and adolescents: A clinician's guide to effective psychosocial and pharmacological interventions (pp. 92-115). New York: Oxford. Watson, J. B. (1916). Behavior and the concept of mental disease. Journal of Philosophy, Psychology, and Scientific Methods, 13, 589-597. Wigfield, A., & Eccles, J. S. (1990). Test anxiety in the school setting. In M. Lewis & S. m. Miller (Eds.), Handbook of developmental psychopathology (pp. 237-250). New York: Plenum Press. Wilmshurst, L. (2005). Essentials of child psychopathology. Hoboken, NJ: John Wiley & Sons. Wine, J. D. (1971). Test anxiety and direction of attention. Psychological Bulletin, 76, 92-104. 113 Wine, J. D. (1980). Cognitive-attentional theory of test anxiety. In I. G. Sarason (Ed.), Test anxiety: Theory, research, and applications (pp. 349-385). Hillsdale, NJ: Lawrence Erlbaum. Witt, W. P., Kasper, J. D. & Riley, A. W. (2003). Mental health services use among school-aged children with disabilities: The role of sociodemographics, functional limitations, family burdens, and care coordination. Health Services Research, 38, 1441-1466. 114 Appendix A: Guide for Researchers under Agreement with Statistics Canada Statistics Canada Research Data Centres (RDCs) Guide for Researchers under Agreement with Statistics Canada ROC CDR October, 2005 October, 2005 PREFACE This guide applies to researchers working under contract to the federal department Statistics Canada accessing data in the Research Data Centres (RDCs) located at universities throughout the country. The RDC Program intends for this guide to assist researchers working or who plan to work in an RDC as well as summarizing the role of the RDC staff, so that researchers understand all facets of the RDC. The guide is organized as follows: • Chapter 1: Introduction to the Research Data Centres and the general application process for accessing the centres. • Chapter 2: Procedures to follow at each phase of the research process, starting from the project proposal to submission of the final product. • Chapter 3: Describes the procedures for protecting the confidentiality of respondents in statistical output. And, explains how to remove statistical outputs from the RDC. • Chapter 4: Provides website addresses if researcher requires additional information. • Appendices providing more detailed information on some of the guidelines for disclosure analysis, proposal and research paper guidelines and contract information. Any comments or questions can be directed to the RDC staff or you may contact: Dr. Gustave Goldmann, RDC Program Manager Room 1710 Main Building Tunney's Pasture Statistics Canada Ottawa, ON K1A 0T6 Phone: (613) 951-1472 Fax: (613) 951-4942 Email: rdc-cdr@statcan.ca October, 2005 116 TABLE OF CONTENTS PREFACE 1 CHAPTER 1 - INTRODUCTION TO THE RESEARCH DATA CENTRES AND APPLICATION PROCESS 3 1.1 Background 3 1.2 Research Data Centres: Application Process 4 1.2.1 Research Proposal 4 1.2.2 Approval of proposals based on four criteria 5 1.2.3 Project Proposal Template 5 1.2.4 Granting Access to an RDC for students 5 1.2.5 Masters degree students 6 1.2.6 PhD students : 6 1.2.7 Undergraduate Students 6 1.3 Proposal Review Process -. 6 1.3.1 Committee Decision.. 7 1.3.2 Clarification of Proposal is Requested by the Committee 7 1.3.3 The Steps for Researchers to Take if the Strategic Programs and Joint Initiatives program officer requests clarification of your proposal 7 1.3.4 What Happens if the Committee Members Request that You Revise and Resubmit your Proposal? • 7 1.3.5 Are there Circumstances where a Proposal has been Rejected? 7 1.4 Proposal Approved 8 1.4.1 Security Clearance Procedures 8 1.4.2 RDC Microdata Research Contract 9 1.4.3 Orientation 9 1.4.4 Oath (Oath of Office and Secrecy) 9 1.4.5 Deemed Employee Status 10 CHAPTER 2 - THE LIFE CYCLE OF RESEARCH AT THE RDC ...11 2.1 Beginning your Project at the RDC 11 2.1.1 Physical Security of the Centre 11 2.1.2 The Computer Facilities 11 2.2 Guidelines for Statistical Analysis of Statistics Canada Surveys 12 2.3 Your Use of the RDC - Security. 13 2.3.1 Getting Support 13 2.3.2 Accessing the RDC 13 2.3.3 System and Electronic Devices 14 2.3.4 Data.... .' 14 2.4 Your Use of the RDC - Procedures 15 2.4.1 Confidentiality and Release of Research Output 15 2.4.2 Obtaining New Data 16 2.4.3 New Research Project with Same Data 16 2.4.4 Add or Remove a Team Member 16 2.4.5 Amendment of the Contract Completion Date 17 2.4.6 Contract Extensions 17 2.4.7 Revision Contracts 17 2.4.8 Transferring of Results 18 October, 2005 1 1 7 2.5 Completing the Contract 18 2.5.1 Producing a Product for the Project 18 2.6 What happens to theproduct(s) when submitted? 20 2.6.1 Seminars and Conference Presentations 21 2.6.2 Related Papers and Research Reports 21 2.7 Exit Process 21 2.8 Re-entering the RDC under a New Agreement. 21 CHAPTER 3 - DATA ACCESS, CONFIDENTIALITY, DISCLOSURE ANALYSIS POLICIES AND PROCEDURES 22 3.1 What is Disclosure? 22 3.1.1 Types of Disclosure 22 3.1.2 Some Examples of Disclosure in Survey Data 22 3.2 How Can Researchers Protect Against Disclosure? 23 3.3 General Risks for Confidentiality Problems 23 3.4 Disclosure Analysis 24 ZA.2 Disclosure Analysis for Parametric Model Output 24 3.4.3 Disclosure for.Non-parametric, Semi-parametric or Hierarchical Data 25 3.4.4 Disclosure Analysis for Variance-Covariance Matrices 25 3.5 Use of Weights 25 3.6 Quality Guidelines for Releasing Data 26 3.7 Responsibilities • 26 3.7.1 The Researcher's Responsibility 26 3.7.2 The Analyst's Responsibility 26 3.8 How to Remove Research Output from the RDC 26 3.8.1 Disclosure Requests 26 3.8.2 Some Quick Tips for How Researchers Can Check for Disclosure 27 CHAPTER 4 - Acquiring More Information 28 APPENDIX 1 - MORE ON DISCLOSURE AND DISCLOSURE RISK ...29 APPENDIX 2 - SAMPLE MICRODATA RESEARCH CONTRACT 30 APPENDIX 4 - THE OATH 38 APPENDIX 5 - SAMPLE RESEARCH CONTRACT AMENDMENT ....39 APPENDIX 6 - SAMPLE DISCLOSURE REQUEST FORM 41 October, 2005 CHAPTER 1 - INTRODUCTION TO THE RESEARCH DATA CENTRES AND APPLICATION PROCESS 1.1 Background Decision-makers need an up-to-date and in-depth understanding of Canadian society to help them respond not only to today's needs, but to anticipate the needs of tomorrow as well. This need underlies the growing demand for analytical output from the rich sources of data collected by Statistics Canada. The Research Data Centres (RDC) program is part of an initiative lead by Statistics Canada, the Social Sciences and Humanities Research Council (SSHRC) and a university consortia to help strengthen Canada's social research capacity and to support the policy research community. RDCs provide researchers with approved access, in a secure university setting, to microdata from population and household surveys. The centres are staffed by Statistics • Canada employees. They are operated under the provisions of the Statistics Act in accordance with all the confidentiality rules and are accessible only to researchers with approved projects who have been sworn in under the Statistics Act as 'deemed employees.' RDCs are located throughout the country, so researchers do not need to travel to Ottawa to access Statistics Canada microdata. Each Research Data Centre (RDC) employs an RDC Analyst who would be happy to advise you during the preparation of your proposal. The Analyst can assess your proposal to determine whether the required elements are present. The Analyst can also advise whether your project is appropriate for access to the detailed microdata. Also, the RDC Analyst can provide you with information about conducting research in the RDC environment and the data holdings of the RDC. Please contact your local RDC Analyst for more information. You may also find it helpful to access documentation for the surveys located on the Research Data Centre's website: • http://www.statcan.ca/english/rdc/whatdata.htm • or read about tips on proposal writing located at http://www.statcan.ca/english/rdc/apply.htm or http://www.statcan.ca/english/rdc/applv.htm#guides) Researchers guide - October 4 2005.doc October, 2005 119 1.2 Research Data Centres: Application Process Any researcher (or team) from academic organizations with legitimate research interests related to public policy development is eligible to apply to the RDC program. The detailed microdata kept by Statistics Canada in an RDC are protected under the Statistics Act. Only employees or deemed employees of Statistics Canada may have access to these data. A process has been established to permit academic researchers with approved proposals to work with these data. Researchers must become deemed employees of Statistics Canada and produce a research paper that follows the Statistics Canada mandate (Statistics Act sections 3, 19, 20, 22). To be granted access to the RDCs, a researcher must submit a research project proposal to SSHRC for peer and institutional review, undergo a security evaluation, and take an oath promising to protect confidentiality. The proposed project, once approved, will be the basis of a contract between the researcher and Statistics Canada. 1.2.1 Research Proposal Research proposals are submitted to an adjudicating committee operating under the auspices of SSHRC and Statistics Canada. It is done electronically by accessing the submission form on the SSHRC website: http://www.sshrc.ca/web/applv/application/rdc application e.asp Researchers must submit a separate proposal for each new project that they wish to carry out in the RDC. In the case of significant changes in scope of the approved project, researchers may be asked to re-submit another proposal incorporating the changes. To apply for RDC access, you also obtain personal security clearance from Statistics Canada (refer to the section Security Clearance Procedures). SSHRC invites applications from individual researchers or research teams led by a principal applicant. The principal applicant is responsible for submitting application forms on behalf of the team. Applicants must complete the Web-based CV < (https://webapps.nserc.ca/sshrc/logon ciss e.htm) and application forms and include the attachments listed below. Before you begin your on-line application, please ensure that you have prepared all required attachments. These attachments include research contributions (maximum five pages) and the Project Proposal (maximum five pages). To ensure a swift review of your proposal, applicants are encouraged to include all elements of the Project Proposal Template. 4 October, 2005 120 1.2.2 Approval of proposals based on four criteria • scientific merit and viability of the proposed research; • viability of the methods to be applied— the data to be analyzed; • demonstrated need for access to detailed microdata; and, • expertise and ability of the researchers to carry out the work.. 1.2.3 Project Proposal Template The Project Proposal is a maximum of five pages and must include the following: • Title of the Project o The project title • Rationale and objectives of the study o Clearly identify the specific questions or objectives of the project o State how the research will contribute to the knowledge in the field of study • Proposed data analysis and software requirements o What is the proposed statistical methodology? How is it suitable for this project? o What software will you use? • Data Requirements o An explanation of why access to the confidential data (as opposed to public use microdata files) is necessary, o Which survey file/files or cycles are to be used? o Provide a statement that the confidential data file(s) identified is (are) in fact suitable for the proposed research, o What is the specific population of interest in the required data set(s)? o What are the variables to be used? • Expected project start and end dates o Expected project start and end dates • References o Sources used to cite your quotes used in the proposal or for specific analytical methods employed 1.2.4 Granting Access to an RDC for students Access will be granted based on the nature of the proposal. For example if a proposal requires analysis of cross-sectional data sets this would not on its own be considered in depth enough to warrant access to RDC data. In general, students listed as either principal or co-investigators on proposals (for thesis or dissertation research) should include in their application a letter of recommendation from their supervisor stating that they are capable of utilizing statistical software such as SPSS, STATA, or SAS and that the supervisor will be an active member of the research team supporting the student throughout the process. 5 October, 2005 121 Students are to send a copy in electronic format to lle@sshrc.ca, and a signed hard copy to: Luc Lebrun, C/O SSHRC RDC Program Officer 350 Albert Street P.O. Box 1610 Ottawa, Ontario KIP 6G4 '/.2.5 Masters degree students Master degree students can apply as Principal Investigator (PI) or co-investigator, but their supervisor must be part of the research team and be included on the contract. 1.2.6 PhD students All PhD students who request access to the RDC must include their Academic Advisor as the co-investigator on their research team. 1.2.7 Undergraduate Students Access will only be granted to Honour's year undergraduate students. The Academic Supervisor/Advisor must be the Principal Investigator of the research team in order to grant access. The student then becomes a co-investigator (the PI does not necessarily have to access the RDC and can leave this data research solely to the student). 1 . 3 Proposal Review Process The formal review process is coordinated by SSHRC. A three-member peer-review committee selected by SSHRC and Statistics Canada is formed for each proposal submitted. Two of the peers are from academic institutions with expertise in the field of the proposal. The third member is chosen from Statistics Canada. The committee evaluates each proposal based on its scientific merit and recommends to Statistics Canada whether or not the project meets the established criteria (refer to the criteria outlined above). As such, the task of the committee differs fundamentally from other selection committees run by SSHRC. The committee does not select winners from a range of applicants, and there is no financial grant involved in this process. The committee acts as a gateway. Research proposals that have undergone peer review from one of the tri-council funding agencies (SSHRC, CIHR, NSERC or FQRSC) will bypass peer review and move directly to institutional review. 6 October, 2005 122 1.3.1 Committee Decision Within eight weeks of the date of application (sooner if possible), SSHRC will communicate the adjudication committee's decision to the project's principal applicant. The review process will produce one of two outcomes - acceptance or rejection; and the decision must be unanimous among all peers. The applicant is advised of the results via written correspondence. The decision rationale is included in the report of the evaluation results for the projects that are rejected. Researchers are free to address these concerns and resubmit the proposal. 1.3.2 Clarification of Proposal is Requested by the Committee On occasion, committee members request clarification from the principal applicant if the proposal does not demonstrate, clearly, for example that their project requires access to detailed micro-data (i.e., that their analyses cannot be accomplished with publicly available micro-data). If clarification is requested of the principal applicant, then the Strategic Programs and Joint Initiatives program officer, will act as a liaison between the committee members and the principal applicant, and request the necessary information. 1.3.3 The Steps for Researchers to Take if the Strategic Programs and Joint Initiatives program officer requests clarification of your proposal Provide the necessary clarification to the Strategic Programs and Joint Initiatives program officer as soon as possible so that the committee members can review your comments and continue to evaluate your proposal. 1.3.4 What Happens if the Committee Members Request that You Revise and Resubmit your Proposal? Some proposals require substantial modifications and it is simpler to request that the principal applicant revise and resubmit their proposal in order to avoid many rounds of committee member evaluations. If the Strategic Programs and Joint Initiatives program officer contacts you for a revise and resubmit, you are invited to consult with the local RDC analyst if you are unclear about the revisions requested of you. Once you have modified your proposal, resubmit it online (http://www.sshrc.ca/web/apply/application/rdc application e.asp) to the SSHRC application page. 1.3.5 Are there Circumstances where a Proposal has been Rejected? If a proposal is rejected by all committee members then the researcher is invited to resubmit their revised proposal. The Strategic Programs and Joint Initiatives program officer, Luc Lebrun, will inform the researcher that they can modify their proposal and re-submit through the proper SSHRC-channels. All rejected applications 7 October, 2005 must be officially re-submitted through the RDC electronic system and be subject to review by the overall evaluation committee. Researchers can expect that this process will take the same estimated time as their original proposal submission to SSHRC. The only exceptions to the overall evaluation process apply to the CIHR, SSHRC or FQRSC-funded projects which need only be sent to back to the Strategic Programs and Joint Initiatives program officer, Luc Lebrun, who will then forward the revised proposal directly to the relevant Statistics Canada Institutional reviewer for evaluation (omitting the Academic peer review step). 1.4 Proposal Approved If your proposal is approved and you are given access to a Research Data Centre, your contract with Statistics Canada allows you to access only the microdata specified in your approved research project and only for the purpose of completing that project. You must submit a proposal for any subsequent research project that you wish to carry out at an RDC. In addition, SSHRC and Statistics Canada may ask for a new proposal if the scope of your research changes significantly. Once the proposal is approved, researchers are required to get personal security clearance from Statistics Canada. 1.4.1 Security Clearance Procedures Once a project is approved a number of security procedures must be followed: • Statistics Canada will perform an Enhanced Reliability Check on any researcher who needs to access its data. • Researchers will have to complete the security clearance forms within the presence of the Statistics Canada analyst at the RDC where the research will be conducted. • The RDC analyst will send this form to Statistics Canada in Ottawa to be processed and will contact researchers to inform them of the results of the security check. • The RDC analyst will invite the researcher, or group of researchers, for an orientation session to explain procedures at the RDC. (All team members accessing the data: 1. require security clearance; 2. must sign the contract; 3. must attend an orientation session; and, 4. must sign The Oath or Affirmation of Office and Secrecy. 8 October, 2005 124 • During the orientation session, researchers will sign their contract with Statistics Canada and take the Oath or Affirmation of Office and Secrecy (refer to Appendix 4k Note: Researchers with no active research contracts with Statistics Canada for more than a year are required to update their security clearance. 1.4.2 RDC Microdata Research Contract The accepted proposal will become part of the contract between the researcher(s) and Statistics Canada. The contract specifies the following terms of access: • Data sets to be provided by Statistics Canada (please note the contract grants researchers access only to the microdata specified in the approved research proposal). • Project start and completion date. • Agreement that the researchers abide by the RDC security and confidentiality requirements. • In fulfillment of the contractual obligations, the researcher will provide a product to Statistics Canada once the contract end date has arrived. 1.4.3 Orientation The orientation session is a presentation designed to familiarize researchers with the privilege and duty of becoming deemed employees of Statistics Canada and working in the RDC. It includes a discussion of the policies and procedures related to data confidentiality, disclosure analysis, and other information regarding the operation of the RDC. Procedures for obtaining the release of research output will also be discussed. Researchers will also sign the Oath of Office during the orientation. The orientation session is not and should not be considered a substitute for a thorough reading of this document (Guide for Researchers). 1.4.4 Oath (Oath of Office and Secrecy) Before accessing data, researchers are asked to take an Oath of Office (oath of secrecy) to become a deemed employee of Statistics Canada. This is the legal requirement to maintain data confidentiality. A copy of the oath can be found in Appendix 4. As part of this process the researcher(s) agree(s) to: "... not disclose or knowingly cause to be disclosed, by any means, any information obtained under the Statistics Act in such a manner that it is possible from the disclosure to relate the particulars obtained from any individual return to any identifiable individual person, business or organization " (Statistics Act). 9 October, 2005 This oath requires that researchers be personally accountable to uphold the confidentiality provisions of the Statistics Act, and not to reveal anything about individual respondents, either directly or indirectly. The oath is binding for life; hence even after researchers have completed their research contract, they may not reveal any confidential information. A violation of confidentiality, whether intentionally or accidentally, would put the RDC endeavour at risk, and reduce research opportunities. In this document, the majority of the discussion is intended to help prevent an accidental breach of confidentiality. It is also important to remember that this oath is legally binding with repercussions if violated. If a researcher violates the confidentiality provisions of the Statistics Act, the researcher will lose all privileges at this and any other Statistics Canada RDC indeterminately. Other consequences and penalties are outlined in Appendix 4. Note that Statistics Canada employees are subject to the same penalties when a violation of the oath occurs. 1.4.5 Deemed Employee Status Once you have signed the Oath, you are a deemed employee of Statistics Canada for the duration of the microdata research contract. All deemed employees within the RDC Program are legally responsible for upholding the Statistics Act and all policies and procedures of Statistics Canada. Deemed employees are not permitted to take away with them any sensitive statistical information. They remain subject to the oath/affirmation of secrecy even after their project is terminated. The contractual arrangements between Statistics Canada and deemed employees do not involve payment to the researcher but require: • Statistics Canada is to provide access to confidential microdata (sensitive statistical information); and, • The researcher is to deliver a product at the end of the contract. 10 October, 2005 C H A P T E R 2 - T H E L I F E C Y C L E O F R E S E A R C H A T T H E R D C 2.1 Beginning your Project at the RDC All Statistics Canada's RDCs have particular elements in common that researchers should be aware of and adhere to. All RDCs are physically secure environments with an isolated computer network (i.e., no internet capabilities or other external capabilities). Statistics Canada staff, RDC analysts, are on site to assist researchers and to approve all output leaving the RDC. This chapter is designed to help researchers through each phase of the life cycle of a research project conducted in the RDC environment. 2.1.1 Physical Security of the Centre The RDC is a physically secure facility to the extent that the researchers use it properly. • Each researcher will be issued a security pass. One cannot gain access to the RDC without the pass. • Do not share or lend your security pass to anyone else. • Report lost or stolen security passes immediately to the RDC Analyst. • No visitors. Researchers may not escort individuals who do not have security clearance into the centre. Only RDC Analysts can permit entry of any individual into the premise. • Do not open the door for anyone (i.e. do not answer a knock/doorbell indicating that someone without a pass would like entry into the RDC). 2.1.2 The Computer Facilities Researchers are not assigned specific workstations in the RDC. It is important to book a time in the RDC in advance to minimize conflicts over resources and workstations. As a courtesy to other users, please adhere to the booking as much as possible. Contact the RDC Analyst to request the purchase and installation of new software or software upgrades. Researchers may not install their own copies of software onto any RDC workstation. • Each computer in the RDC is a workstation to which researchers can log on and connect to the server. • Data • Do not operate laptop computers, PDAs, cell phones or any other storage devices in the lab area. • Workstations do not have CD/floppy disk-write capabilities. Disk-writing can only be conducted by the RDC Analyst through secured terminals. It is strictly prohibited to connect any portable or mass storage device to a workstation. 11 October, 2005 2.2 Guidelines for Statistical Analysis of Statistics Canada Surveys Statistics Canada household surveys, such as those whose data are available in the RDCs, are based on complex sample designs that include stratification, multiple stages of ^ selection, and unequal probabilities of selection. Generally, commercially available statistical software packages ignore these complexities. (Some of these packages can make use of the weights, which do contain some design information, but do not compute correct variances.) However, software that accounts for the design complexities is available. Ignoring the design complexities could have an impact on the results of the analysis of the data. As an example, because of special interest in certain subpopulations, a survey might sample some units at higher rates than other units; an analysis of the population as a whole that ignores these differential rates will generally produce biased results. As another example, clusters of units are often included in a sample, such as several people from the same household; this clustering may lead to correlation between the observations, which, if ignored, could lead to underestimation of the standard errors of parameter estimates. Many standard analytical procedures have been adapted to incorporate the sample design information. However, there are some analytical procedures for which there is currently no recommended design-based approach. Research is ongoing into appropriate analytical methods for such cases. In the meantime, approaches developed for non-survey data would have to be used; however, for these cases, there still may be some facility for making use of the survey weights. 12 October, 2005 2.3 Your Use of the RDC - Security 2.3.1 Getting Support RDC Analyst The first line of support while working at the RDC comes from the RDC Analyst. Researchers can direct questions about the dataset to the RDC analyst. The analyst will either have the answer or will direct the question to other Statistics Canada employees. When managed properly, this type of knowledge exchange benefits both the researchers and the Statistics Canada data programs. Data Analysis Resource Centre (DARC) The second line of support is available through the Data Analysis Resource Centre (DARC) at Statistics Canada. There has been considerable research carried out by this group on the appropriateness of various statistical analyses for complex sample designed datasets produced by Statistics Canada. DARC can provide suggestions on suitable methods and software tools. DARC also conducts seminars and research on various methods specific to longitudinal data. These services are offered on a cost-recovery basis. For more details, please contact your RDC analyst. 2.3.2 Accessing the RDC As a Statistics Canada's deemed employee, it is your responsibility (as well as the RDC analyst's) to maintain the RDCs' security: • Breaches in security or lost door keys should be reported to the RDC analyst immediately. • RDC entrance information must not be shared. • Consult with the RDC analyst for the sign-in procedures for the RDC. • No visitors are permitted into the secure RDC facilities. 13 October, 2005 129 2.3.3 System and Electronic Devices • Do not operate any personal electronic device such as laptop computers, palm pilots, cellular phones or other devices with optical beam capability inside the RDC secure premises. In case of emergency, cellular phones, or similar device may be used inside the RDC but away from the workstation. • You are accountable for your working files and how you use the system. With this accountability comes the responsibility of protecting your log-on information. • Make sure to log off after every work session. If you need to be away from your workstation during the session, all work should be saved and the computer locked to prevent others from accessing it. • Save all work in your assigned electronic folder. This aides in the disclosure avoidance that the RDC analyst must do before your output can be released, project folder backup and system maintenance. 2.3.4 Data • Do not discuss confidential information at any time outside the RDC, whether over the telephone, by e-mail (or other electronic media) or in person. • Do not ask other researchers at the RDC for access to their data sets. Researchers are granted access only to the data for which access has been approved. Written requests or a new proposal is required for access to other data sets. • Do not give other researchers, other than members of your team, access to any of your secondary datasets or output files. See the RDC Analyst for ways to share files with team members electronically without sharing passwords. • Do not carry out analyses for colleagues or other researchers who do not have approved RDC research projects or who are not deemed employees. • Do not use RDC data files to conduct data analysis outside the mandate of the approved project. • Please place any output containing confidential data that you wish to discard into a designated receptacle, shredder, or return to the RDC Analyst. 14 October, 2005 130 2.4 Your Use of the RDC - Procedures 2.4.1 Confidentiality and Release of Research Output Statistics Canada considers it important not only to avoid disclosure of confidential information, but also to avoid the perception of disclosure. It is the trust of the respondents that makes it possible for Statistics Canada to provide valuable data on the socio-economic condition of Canadian society. The following is a brief summary of Statistics Canada policies regarding confidentiality and obtaining release of the research output. • Absolutely no detailed microdata may leave the facilities. • Absolutely no research results may leave the facilities without being examined and approved by the RDC analyst. • All output leaving the RDC must undergo disclosure analysis, first by the researcher and then by the RDC Analyst. Disclosure analysis is the means by which researchers ensure that any material removed from the.RDC does not pose a disclosure risk of data confidentiality. See Chapter 3. Appendix 1 and Appendix 6 for a complete discussion of the topic. Researchers must complete a Disclosure Request Form each time they make a disclosure request. Researchers must provide description of variables used, new variables created, documentation of data sets and programs used in producing output to help the RDC Analyst better understand the materials requested for release. Working closely with the RDC Analyst during the disclosure analysis process will help to avoid misunderstandings and speed up the final approval for release. Disclosure of individual cases in any manner - in research papers, via email, phone or fax, or in casual conversation - with persons outside designated facilities is prohibited. This includes researchers working on other projects who have deemed employee status. Research output that emphasizes model output rather than tabular (descriptive) output are encouraged in order to reduce disclosure risk. Removal of "intermediary output "from the RDC is discouraged. Intermediary output is typically produced during the exploratory and model development stages of the analysis. These outputs often consist of detailed tables of descriptive statistics. Such outputs often present more detailed information than final model outputs; their removal can greatly increase the risk of disclosure, especially residual disclosure. Within the secure premises of the RDC, output of any type can be printed, but it must not be removed from the RDC before being examined and approved by the RDC analyst. Output without disclosure control will be printed on coloured paper (usually green) to avoid accidental confusion with outside material brought into the centre or output already approved to leave the centre. Coloured paper (usually green) DOES NOT LEAVE THE RDC. At the end of a 15 October, 2005 131 session, materials should be returned to the RDC analysts for secure storage until the next visit. This material will be shredded by the researcher or when the project is completed. Researchers must comply with Statistics Canada rules and regulations on data confidentiality when removing output from the RDC. 2.4.2 Obtaining New Data On rare occasions, it may be appropriate to access another data set to complete the project. An appropriate case may be that the sample size in the originally selected data set is not adequate but another data set contains similar questions and more respondents. The researcher may also want to verify their findings from one data set using another dataset. The nature of the project or objective of the research questions must not change significantly. If the researcher wishes to add a data set to their project under these conditions, they must submit a written request to the RDC Analyst detailing the rationale for accessing a new data set and a commitment to maintain the objectives of the original proposal. Provide a detailed discussion when requesting new data be added to your project. This request will be subject to the formal project review process. If the proposed use of the new data does not fit within the original proposal as judged by Statistics Canada, you will be asked to begin a new project by submitting a new proposal. The contract does not have to be amended to add a new cycle of a longitudinal data set, only for a new data set or if there are changes to the research questions. 2.4.3 New Research Project with Same Data While working on a project a researcher can begin to develop new ideas for other analyses. If these analyses do not fit under the original objective of the initial proposal a new proposal will be required. Do not start a new project without submitting a proposal to the review process. Violation of this rule will result in immediate withdrawal of data access privileges including data access to projects currently under contract. 2.4.4 Add or Remove a Team Member After approval of the research proposal, researchers may make changes to the research team. To remove a research team member from the project, please provide a written memorandum to the RDC analyst. Both the principal investigator and the team member being removed from the contract need to sign the amendment. To add a new researcher, please provide a written memorandum along with a curriculum vitae of the researcher to the RDC Analyst. The new member will be required to complete the PERSONNEL SCREENING, CONSENT AND AUTHORIZATION FORM; part of the process for the researcher to undergo a Reliability Check. 16 October, 2005 132 And they will also need to take the oath/affirmation of secrecy and attend an RDC Program orientation session. 2.4.5 Amendment of the Contract Completion Date There are two ways to grant RDC data access after contract expiry: by contract extension or by a revision contract. 2.4.6 Contract Extensions This type of extension applies when the principal investigator realizes that they have significantly underestimated the time needed to complete the project or other factors may be preventing them from completing their project by the date specified in the contract. Contract extensions occur when a project reaches its contract end date but the research team has not yet completed enough analysis to prepare their product. There are valid reasons for a project extension including: • The data analysis is in the final stages and the extension will cover the completion of the data analysis; • Funding or staffing shortages have prevented progress on the contract to date (including leaves of absence, long-term illnesses, etc.); • There have been complications in obtaining data to complete the analysis. The contract can be extended for up to one year past the contract due date at the request of the principal investigator. Anticipate the need for an extension of your project and provide your RDC analyst with a written justification well in advance of the contract expiring. 2.4.7 Revision Contracts A revision contract for data access may be obtained only in the following two cases: (1) A researcher has submitted to a scholarly journal a manuscript based on results obtained from their RDC research; o The researcher has received a "revise and resubmit" from the editors of that journal that requires them to return to the data; and, o The researcher's original project contract has expired. (2) A student has submitted a dissertation or thesis to their academic review committee; 17 October, 2005 133 o The student has received comments, usually through marked up copies of the text, and must now submit a written request to the RDC analyst for permission to re-enter the RDC in order to perform additional work with the data. o The student's original project contract has expired. Access under a Revision Contract is provided with the understanding that the research team is undertaking only the work suggested by the referees and not undertaking a new research project. They are generally used when a project has expired and their manuscript has already been submitted to Statistics Canada in fulfilment of their contractual obligations. • Revision contracts usually last up to six months from start to finish. A product is required at the end of the contract even if the content is very similar to their original manuscript. • In addition to providing the information required on the Contract Extension/Revision Contract Request Form, the Principal Investigator must provide a copy of the referee reports as evidence that the project continues to have scientific merit and requires re-access to the microdata. If you are uncomfortable sending these reports in their entirety, a summary of these comments will suffice. 2.4.8 Transferring of Results Results may be transferred to another RDC without disclosure analysis when used by other team members in another RDC. Transfers require that the results be encrypted by the RDC analyst to encrypt the results before they are recorded on the transportation medium (CD or DVD). Please schedule time for transferring files with the RDC Analyst. 2.5 Completing the Contract RDC Program at Statistics Canada has a mandate to provide a publicly available document from all research conducted on its data sets and this requirement extends to research conducted by deemed employees in exchange for data access. Researchers are responsible for producing a product from their research. See the contract, Appendix 3. section (5), "Limitations on uses of the microdata file and proposed output". Approximately one to two months prior to contract expiry date a reminder letter (signed by the Program Manager) will be sent by email to the Principal Investigator. 2.5.1 Producing a Product for the Project Researchers with approved projects in the RDCs are granted access to the detailed micro data as deemed employees of Statistics Canada under the terms of the Statistics Act. The Act stipulates that only employees and deemed employees of Statistics Canada may have access to the detailed 18 October, 2005 134 (non-public use) micro data. To be a deemed employee of Statistics Canada implies that some product or service must be rendered to the agency. The product is defined in the research proposal that was submitted as part of the application process and it is stipulated in the contract that is entered into between the researcher(s) and Statistics Canada. A project is complete1 when the Principal Investigator submits the product to Statistics Canada thus fulfilling their contractual obligations. The Research Data Centre Program has redefined its policy on the type(s) of products they would like researchers to submit in fulfilment of their contractual obligations for data access. For the most part, it means that instead of submitting an RDC working paper, researchers now have the option to submit a number of different products such as a journal article as their final product for the contract. Researchers are always encouraged to submit all final products to us so that we track these as measures of success for the RDC program. There are five types of products that a researcher can produce as a product for their project. 1. RDC working paper 2. Peer-reviewed journal article 3. Book or book chapter 4. Graduate level thesis or dissertation2 5. Commissioned report (e.g., a government commissioned report) If a researcher identifies an alternate possibility, this should be discussed with an RDC Analyst supervisor before the contract is written. Changes to contract wording of the acceptable product will not be made retroactively, only proactively for new contracts: If a researcher would like to submit a product other than a Statistics Canada working paper as the main product from their already existing contract, they should discuss this with their RDC Analyst who will be submitting the product to the RDC Manuscript Coordinator.. 1 Note: the contract end date does not always coincide with the project completion date. The contract end date dictates when data access must stop. The researchers may continue to work on the project outside the RDC environment to complete the writing phase. A project ends when the main product is submitted. 2 If a member of a research team wishes to use the analysis for a thesis or dissertation, a separate contract should be written to cover their analysis, even if it was not indicated separately on the original proposal. The student should submit a one-page description of the analysis plan explaining how their work is consistent with the original group proposal and sign a contract along with their supervisor. 19 October, 2005 135 2.6 What happens to the product(s) when submitted? This is an important new development: Not all products will be peer and institutionally reviewed.3 Instead, the product will be given to a Triage committee. The Triage committee will identify the 'gems' that we would like to showcase through a Statistics Canada flagship publication venue. What is a gem? It is a product that has wide appeal to a Canadian audience with a clear message about social trends or policy. There are three outcomes from the Triage process: 1. The product is not appropriate for a Statistics Canada flagship publication, for one reason or another. For example, the results of the analysis did not identify any conclusions or the analysis was not feasible given the data. The Principal Investigator will be notified by email from Statistics Canada head office that the product was received and that their contractual obligations are fulfilled. The local RDC Analyst will receive a carbon copy of this message. 2. The product is a good product but it is not of general interest to a Canadian audience. The Principal Investigator will be notified by email from Statistics Canada head office that the product was received and that their contractual obligations are fulfilled. This product will be included in a bibliography of research conducted in the RDC. The local RDC Analyst will receive a carbon copy of this message. 3. The product is a 'gem' and Statistics Canada is interested in publishing it in a flagship publication. The Principal Investigator will be notified by written letter from Statistics Canada head office that the product was received and that their contractual obligations are fulfilled. That letter will also invite to the Principal Investigator to consider a Statistics Canada publication under one of three possible scenarios dependent upon their other intentions for publication. The three publication scenarios are: (a) A joint release between a peer-reviewed journal and the Statistics Canada Daily. On the day the journal publishes the article, an advertisement for this will be listed in the Daily along with a brief description of the research. (b) A release in the Statistics Canada Daily after a journal publication has occurred. (c) A publication in a Statistics Canada flagship publication such as Canadian Social Trends, Health Reports, Perspectives on Labour and Income. The timing of this publication can be postponed to accommodate a peer-review journal publication first if requested by the Principal Investigator. If the Principal Investigator is interested in pursuing one of these publication options and re-access to the data is required, either an extension or a revision contract will be prepared depending on the circumstances. 3 Some researchers who are new to the data set utilized in their project may have concerns about how they have analyzed the data. The Principal Investigator has the right to request an institutional review of their paper at any time after the submission of the final product to Statistics Canada. If a researcher would like an institutional review, the RDC Analyst should notify the RDC Manuscript Coordinator who will make the necessary arrangements. 20 October, 2005 136 2.6.1 Seminars and Conference Presentations Results may be presented at seminars or conferences. All work presented must reflect the proposal. And, all results must undergo disclosure analysis by the RDC Analyst. Please give the RDC Analyst a copy of all presentations so an inventory can be maintained. 2.6.2 Related Papers and Research Reports Beyond the RDC product produced, you are requested to provide a reference, a copy of the abstract, or the full document of any subsequent papers produced from the original RDC product and analysis. Statistics Canada hopes to maintain a bibliography of publications resulting from the RDC program. Papers, abstracts or references can be sent to the RDC analyst or to the Program Manager. 2.7 Exit Process When submitting the final paper, researchers are asked to provide secondary data sets, model programs, relevant documentation, output, etc used in creating the product. This information will be archived and can be accessed during the revision contract. At the time the project is complete, the Academic Director of your RDC will be interested in posting your product (or a reference to it) on the local RDC website. This is negotiable and is done on a case by case basis. See your RDC Analyst about cleaning out your storage area, archiving files and returning your security card and/ or locker keys. 2.8 Re-entering the RDC under a New Agreement To access the RDC to begin a second or subsequent research project, researchers will follow the application process outlined earlier in this document by submitting a research proposal through the Social Science and Humanities Research Council (SSHRC) web site. After the proposal is approved, the research team members will sign a new contract and may be asked to review the orientation material with the RDC Analyst and reaffirm the oath before accessing the data. 21 October, 2005 137 CHAPTER 3 - DATA ACCESS, CONFIDENTIALITY, DISCLOSURE ANALYSIS POLICIES AND PROCEDURES The material in this chapter applies only to output that you intend to remove from the RDC. Any amount of output can be produced and stored (up to available capacity) for use on-site. This chapter presents policies and procedures to follow to clear research results for release. Appendix 1 provides examples of disclosure problems and suggested approaches in correcting them. The goal of disclosure avoidance is to protect the information provided by respondents while presenting the least possible hindrance to research pursuits. Ultimately, users of confidential micro data files are responsible for avoiding disclosure of confidential data and minimizing the risk of disclosure. The RDC Analyst and researchers will work together to find solutions to data confidentiality problems. 3.1 What is Disclosure? Disclosure occurs when data that can be attributed to individual respondents (e.g., persons, households, businesses, other organizations), are released. 3.1.1 Types of Disclosure There are three types of disclosure; Identity, Attribute and Residual. Identity disclosure occurs when an individual can be identified from the released output, leading to information being provided about that identified subject. Attribute disclosure occurs when confidential information is revealed and can be attributed to an individual. It is not necessary for a specific individual to be identified or for a specific value to be given for attribute disclosure to occur. For example, publishing a narrow range for the salary of persons exercising a particular profession in one region may constitute a disclosure. Residual disclosure can occur when released information can be combined to obtain confidential data.). Care must be taken to examine all output to be released. While a table on its own might not disclose confidential information, disclosure can occur by combining information from several sources, including external ones, (e.g., suppressed data in one table can be derived from other tables). 3.1.2 Some Examples of Disclosure in Survey Data • A well-known personality, e.g., a professional athlete, is selected in a survey and; • Information published about her community, such as the highest reported income in that community, almost certainly was reported by her. (Identity disclosure.), 22 October, 2005 138 • Results from a longitudinal survey highlight one household with a highly unusual migration pattern, leading to its identification. (Identity disclosure.) • The parents of a 16-year-old selected in the sample see a table showing that all sampled 16-year-old respondents in their region have tried drugs. (Attribute disclosure.) • A newspaper article relates a 37-year-old widower's complaints about being surveyed, and there are only two sampled 30 to 39-year-old widowers in survey cross-tabulations. (Eventually leading to identity and/or attribute disclosure.) • By combining several results a person identifies information that was purposely excluded from the Public Use Microdata File (PUMF) because it presented too high a disclosure risk (e.g., the country of birth of recent immigrants). Note that even the appearance of disclosure can tarnish a statistical organization's reputation with respect to confidentiality. Damage could occur even if it turned out that the wrong person or household had been identified in the first two examples. Refuting a mistaken identification may . increase the risk of exposing the real respondents. 3.2 How Can Researchers Protect Against Disclosure? Disclosure analysis is the term applied to the examination of output researchers would like to remove from the RDC site. It involves a careful look not only at whether obvious identification of individual cases has occurred, but whether information about individual cases can be inferred or deduced from the output. This chapter outlines some broad approaches to disclosure analysis that researchers can use to prepare output for release from the RDC. Appendix 1 presents some examples of specific problems and approaches to solve them. The most important policy is: Absolutely no data and/or research results may leave the facilities without being examined and approved by the RDC Analyst. 3.3 General Risks for Confidentiality Problems Researchers should keep in mind that data confidentiality is primarily a problem for descriptive data, tables of magnitude, individual statistics, etc. It tends not to be a problem for multivariate analysis results (as regression coefficients) except for a few special cases. Confidentiality is also a problem when you are using small geographical identifiers or narrowly defined population groups. 23 October, 2005 139 3.4 Disclosure Analysis 3.4.1 Disclosure Analysis for Tabular Output Tabular output is what is often called 'descriptive statistics.' Researchers should limit the requests for disclosure clearance of tabular output to the minimum necessary to describe the sample used in models, and how it might compare to an underlying population. Tables of summary statistics (means, variances, and related statistics) intended to describe the distributions of variables present certain special issues: • Tables should not contain low frequency cells. Unless otherwise specified in the survey documentation, this means less than 5 observations in a cell. • Take care with "full cells" (In two-way tables these occur when all the respondents in a particular row or column are concentrated in a single cell) or "zero cells". Cells with no respondents can pose a particular problem with sensitive variables. These are variables such as income or health problems not generally known by the public. • For many quantitative variables the maximum and minimum values must not be released. This applies particularly to sensitive variables related to income or consumption. Maximum values may also present a risk for apparently harmless variables, such as household size or age, as extreme values can lead to identity disclosure. • Care must be taken in presenting other statistics intended to describe the shape of distributions. Where possible, present general statistics on distributional shape (e.g., skewness, kurtosis measures) rather than showing anything closely related to individual observations (e.g., the 99 percentile if small sample). Tables or graphs that report quintiles of distributions should be discussed with the RDC analyst. If the researcher or the RDC Analyst finds a disclosure risk, modification of the output will be necessary (e.g., collapsing categories). This will avoid disclosure risk at the expense of output detail, and is the preferred course of action. 3.4.2 Disclosure Analysis for Parametric Model Output Parameter estimates from some models, or the accompanying diagnostics generated by analysis packages, provide sufficient information to enable the re-creation of simpler statistics such as means and totals for cross-classes of variables. This is particularly true of so-called "saturated models" using binary coded variables, i.e., models that include all main effects and possible interaction terms. If the models or accompanying diagnostics provide much detail, it may be necessary to produce the underlying (un-weighted) summary results or tables for purposes;of disclosure analysis. 24 October, 2005 140 If the researcher or the RDC analyst finds disclosure risk, modification of the output and possibly the model specification may be required. For example, researchers may be asked not to report the values of the affected coefficients (a good example would be the estimates of the fixed-effects in a fixed effects model). Or researchers may report that the coefficient is in a certain range (e.g., positive and significant). The RDC analyst and the researcher will work together to ensure, as much as possible, that the research results remain meaningful after any such modifications. 3.4.3 Disclosure for Non-parametric, Semi-parametric or Hierarchical Data Non-parametric or semi-parametric estimates (for example, those used in survival analysis or kernel estimators used in discrete choice models) may pose special disclosure risks. An example is the now increasingly popular hierarchical linear model. This is likely to happen if the clusters contain small numbers of observations. Special care must be taken with output such as these. The RDC analyst and researcher will work together to avoid disclosure risk in these cases. 3.4.4 Disclosure Analysis for Variance-Covariance Matrices While Statistics Canada prefers that all data analysis be completed in the RDC, some researchers will prefer to leave the RDC with variance-covariance matrices in order to conduct statistical inference and hypothesis testing. This is permitted, provided that the variance-covariance matrix adheres to the data confidentiality guidelines for tabular output and tables of magnitude. See Appendix 1 for guidelines. 3.5 Use of Weights Statistics Canada household surveys like those found in the RDC, are based upon complex sample designs with stratification, multiple stages of selection, and unequal probabilities of selection of respondents. Using data from such complex surveys presents problems to analysts because the survey design and the selection probabilities affect the estimation and variance calculation procedures that should be used. In order for survey estimates to minimize bias, the survey weights must be used. Weighting brings results from a sample to the level of the population. Even when population estimates are not the main interest weighting corrects for sample non-representativity arising from the survey design (as over-sampling population sub-groups) and from data collection and processing operations (e.g., higher non-response among young adults). Weighting also assists in the protection of confidentiality because one case may no longer represent one respondent. In very limited circumstances un-weighted results can be released, such as analytic procedures that can not assimilate the use of weights. Researchers have to provide a written justification for requesting un-weighted results to be released. It is important to include a statement in research manuscripts and presentations informing your audience that estimates may be influenced by the survey design and that population inferences are not valid when based on un-weighted results. 25 October, 2005 141 Please consult the documentation accompanying the survey for more information on the use of weights. 3.6 Quality Guidelines for Releasing Data While data confidentiality is closely related to data quality, it is the responsibility of the researcher as RDC analysts do not screen output for data quality. Documentation for each survey contains guidelines on what the minimum cell size should be to obtain acceptable estimates and maintain data quality. Before releasing and/or publishing any estimate, users should first determine the quality levels of the data set they are using. The quality levels are acceptable, marginal and unacceptable. All estimates can be considered releasable. However, those of marginal or unacceptable quality level must be accompanied by warning to caution subsequent users. Please consult the documentation accompanying the survey for more information on data quality. 3.7 Responsibilities 3.7.1 The Researcher's Responsibility Researchers are responsible for applying the rules and regulations for disclosure analysis as specified by Statistics Canada. Limiting the number of requests will reduce the risk of disclosure and facilitate the timeliness of processing. Disclosure analysis involves a certain amount of processing time, researchers should consider this when planning their activities and submitting output for release. A three day turn around time is typical; however, this time can vary depending on staffing resources and the number of other submitted disclosure requests. 3.7.2 The Analyst's Responsibility A major responsibility of the RDC Analyst is to control the output released and ensure that confidentiality is not breached. When a problem of potential disclosure of confidential information is discovered in the results presented for review, RDC Analysts will do their best to find a solution. All attempts will be made in collaboration with the researchers. But overall, the RDC Program must maintain confidentiality and sometimes no release may be the outcome. 3.8 How to Remove Research Output from the RDC You will be asked to provide materials that allow Statistics Canada to document all approved releases of research output. The documentation must indicate that the output was produced as part of an approved project (and which project), and it must allow Statistics Canada to document that published output does not disclose confidential information. 3.8.1 Disclosure Requests At the beginning of your project you will be assigned a project computer directory on the network where all data sets, programming syntax, logs, and output should be stored. The RDC 26 October, 2005 Analyst will provide you with two folders ('to be vetted' and 'vetted'4) in your project directory. The 'to be vetted' folder is where you will copy the output for disclosure by the RDC Analyst as well as the Disclosure Request Form. Once the output has been vetted a copy will be placed in the 'vetted' folder and e-mailed to the researcher. When you have output to be removed from the RDC: 1. Make an appointment with the RDC analyst to discuss the clearance request. 2. Fill out a Disclosure Request Form (see Appendix 6 for a copy of the form). You are required to provide: • The survey name, weighting information, types of methods and sub-sample characteristics (or location of the files) • Supporting documentation files (or location of the files) • Information about possible residual disclosure Please properly document the output. This can be done by ensuring variables are clearly labelled and cell sizes below the minimum requirement are collapsed or re-categorized. The researcher is to ensure that mistakes in the output have been corrected and specific disclosure issues for the survey are addressed. Please be prepared to discuss the clearance request with the analyst performing disclosure analysis. This may require changes in the output to satisfy clearance requirements. Working interactively with the analyst performing disclosure analysis will maximize learning on both sides, avoid costly misunderstandings, and speed the review process. The clearer and more complete the request is, the sooner it can be released. 3.8.2 Some Quick Tips for How Researchers Can Check for Disclosure Verifying these elements can speed up the vetting process: • Provide un-weighted frequency counts (in an identical separate file). • Do not submit tables with small cell sizes. • Restrict cross-tabular analysis to two or three dimensions. • Do not submit listings of cases or show graphs with outliers (i.e. no minimum or maximums). • Provide supporting information for the correlation matrices. • Be cautious when using small subgroups or small areas. • Limit the release of tabular output before the end of your project. • Schedule time with an RDC analyst well in advance. • Check documentation for survey specific rules. Refer to Appendix 1 for further instructions on approaching disclosure problems. 4 The names of these folders may vary between centres. 27 CHAPTER 4 - Acquiring More Information If more information is required about the RDC Network please log on to: http://www.statcan.gc.ca/english/rdc/network.htm If more information is required about the RDC Program please log on to: http://www.statcan.gc.ca/english/rdc/index.htm If more information is required about the Statistic Canada's data sets or research tools please log on to: http://www.statcan.gc.ca/english/rdc/whatdata.htm For an up-to-date list of the projects currently being conducted at the various centres or for abstracts and citation details of research papers published by researchers involved in the program, please log on to: http ://www. statcan. gc .ca/english/rdc/rdcproi ects.htm We are pleased to welcome you to the RDC program and facilities. We hope that your experience with the Program is going to be very profitable and rewarding for you and your research team. 28 144 APPENDIX 1 - MORE ON DISCLOSURE AND DISCLOSURE RISK All variables on a database can be categorized according to their importance to data confidentiality: Direct identifiers: Variables such as name, address or telephone number, etc. that provide an explicit link to a respondent. These are all stripped from the master files to which you will have access. Indirect identifiers: Variables such as age, sex, marital status, area of residence, occupation, type of business etc. that could be used to identify an individual. Sensitive variables: Characteristics relating to respondents' private lives, or business, which are not usually known. These variables seem harmless on their own but used together could reveal information about individuals. For example, consider the case of drug experimentation during adolescence. The parents of a 16 year-old respondent may see a table showing that all sampled 16 year olds (indirect identifier) in their socio-economic group (indirect identifier) have experimented with drugs (sensitive variable). They thus know that their 16-year-old has experimented with drugs. Data confidentiality is primarily a problem for frequency data and tables. It tends not to be a problem for correlative or causal analysis results such as regression coefficients. But it can occur in any kind of data output. General rules to apply at all times: • Use weighted data: the disclosure risk is reduced when weights are used to generate output. • Do not report statistics or cells with fewer than five respondents (or the number reported in the documentation of the survey you are using, whichever is less). • No anecdotal information may be given about specific respondents. 29 APPENDIX 2 - SAMPLE MICRODATA RESEARCH CONTRACT Project ID number: Revised, March 16, 2005 MICRODATA RESEARCH CONTRACT BETWEEN: HER MAJESTY THE QUEEN IN RIGHT OF CANADA, represented by the Minister of Industry, designated as the Minister for purposes of the Statistics Act, (hereinafter referred to as "Statistics Canada"), AND: (Name and affiliation of Principal Investigator), ' (hereinafter referred to as the "Principal Investigator"), and, (Name and affiliation of Co-investigators) (hereinafter referred to as the "Co-investigators"), WHEREAS Statistics Canada requires the services of the Principal Investigator to undertake statistical research and analysis on (Name of the Microdata File) to fulfill its mandate under the Statistics Act; AND WHEREAS to perform these services and to have access to confidential information, the Principal Investigator and the Co-investigators must become "Deemed Employees" of Statistics Canada and are required to take the Oath of Secrecy; AND WHEREAS Statistics Canada wishes to make clear the terms and conditions under which access to the microdata will be granted; NOW THEREFORE the Parties agree as follows: SERVICE PROVIDED BY PRINCIPAL INVESTIGATOR 1. (1) The Principal Investigator will carry out the research project set out in Appendix "A" and provide the report described under "Proposed Output". 30 (2) It is understood that this is a contract for the performance of a service and the Principal Investigator and Co-investigators are engaged for the sole purpose of providing that service. CONDITIONS OF ACCESS T O T H E MICRODATA 2. The Principal Investigator and the Co-investigators must undergo an enhanced reliability check satisfactory to Statistics Canada and take the oath of secrecy in order to obtain access to the non identifiable microdata file required to perform the analysis pursuant to this contract. 3. (1) Access to the non identifiable microdata file (no names, addresses or identifying numbers) and associated documentation shall be provided on Statistics Canada premises, which includes Headquarters and the Statistics Canada Regional Offices during normal hours of operation, Monday to Friday, and the Research Data Centres. (2) The Principal Investigator and Co-Investigators will be provided with the necessary computing facilities, software and documentation as is reasonably necessary to complete the research and analysis pursuant to this contract. D E P A R T M E N T A L REPRESENTATIVE 4. The Director of the (name of the division) or the Manager of the Research Data Centre Program is the designated Statistics Canada representative responsible for the administration of this contract. LIMITATIONS ON USES OF T H E MICRODATA FILE AND PROPOSED OUTPUT 5. (1) The Principal Investigator and the Co-investigators shall not use or disclose any of the information obtained or produced pursuant to this contract for any administrative or regulatory purposes. (2) Access to the microdata file is being provided for the statistical and research purpose outlined in the proposal attached as Appendix A and the microdata file shall not be used for any other purposes without the prior written consent of Statistics Canada. (3) The Principal Investigator and the Co-investigators shall not disclose any of the information from the individual records obtained or produced 31 147 pursuant to this contract to anyone other than current Statistics Canada employees. (4) The Principal Investigator and the Co-investigators shall ensure that no attempts are made to link the microdata file to any other files in order to relate the particulars to any identifiable individual person, business or organization. (5) The Proposed Output must meet the requirements of both peer and institutional review prior to being released by Statistics Canada, for example, in one of its publications or in a research paper. (6) Thereafter, the Principal Investigator may, subject to subsection 6(5), carry out a secondary analysis, but such analysis shall be based solely on the approved "Proposed Output" produced pursuant to this contract and be related to the analytical work undertaken to produce the "Proposed Output". (7) The Principal Investigator agrees to work with Statistics Canada in trying to meet the requirements of peer and institutional review required for the publication or research paper. For the Research Data Centres, a timetable for conducting the peer and institutional review is available in the guidelines for producing the "Proposed Output". • In the event the "Proposed Output" fails a peer or institutional review and Statistics Canada decides not to publish it, Statistics Canada will give the Principal Investigator written notice to this effect within thirty days of having made the final decision. • Subject to subsections 6(5) and 10(2), in the event Statistics Canada notifies the Principal Investigator in writing that the proposed output will not be published, the Principal Investigator will not be prevented from: (a) Publishing the "Proposed Output" elsewhere, and/or (b) Using the "Proposed Output" for purposes of the attainment of an educational degree. O W N E R S H I P 6. (1) The microdata file and related documentation shall at all times be and remain the sole and exclusive property of Statistics Canada, it being mutually agreed that this contract pertains to the use of the microdata file and related documentation to produce a "Proposed Output" for Statistics 32 Canada and that nothing contained herein shall be deemed to convey any title or ownership interest in the microdata file or the related documentation to the Principal Investigator or the Co-investigators. The computer equipment provided for use by the Principal Investigator and the Co-investigators must never be removed from the premises of Statistics Canada and remains the sole and exclusive property of the access facility. (2) Statistics Canada reserves the right to publish in whole or in part, to publish an amended version or not to publish at all, as Statistics Canada deems appropriate, the "Proposed Output" produced by the Principal Investigator pursuant to this contract. (3) Copyright in the "Proposed Output" produced by the Principal Investigator pursuant to this contract shall vest in Her Majesty the Queen in Right of Canada. The Principal Investigator shall provide to Statistics Canada at the completion of the contract or at such other time as Statistics Canada may require a written permanent waiver of Moral rights from every author who contributed to the aforementioned material. Statistics Canada (Her Majesty the Queen in Right of Canada) hereby grants to the Principal Investigator a non-exclusive license to use, reproduce, publish and distribute the "Proposed Output" for any purpose, including, without limitation, research, teaching and publication in any medium. (4) Secondary releases of the "Proposed Output" may be considered by Statistics Canada subject to obtaining consent from the Principal Investigator. (5) In publishing the "Proposed Output" elsewhere, using the "Proposed Output" in the attainment of an educational degree or carrying out any secondary analysis, any reports, documents, or materials which are subsequently prepared by the Principal Investigator which use, incorporate or are in any way based on any material produced under this agreement, shall prominently display the following notice: "The research and analysis are based on data from Statistics Canada and the opinions expressed do not represent the views of Statistics Canada." C O N F L I C T O F I N T E R E S T 7. (1) All persons engaged in the course of carrying out this contract shall conduct themselves in accordance with the principles and spirit of the Conflict of Interest and Post-Employment Code for the Public Service. 33 149 (2) Should a conflict exist prior to the commencement of this contract or be acquired or develop during the life of this contract, the person with the conflict engaged in carrying out this contract shall be responsible for discussing the conflict with the Director of the Division sponsoring the project or the Manager of the Research Data Centre Project and, should it be determined that a conflict exists, for completing-the Confidential Report as required by the Conflict of Interest and Post-Employment Code for the Public Service. (3) No person engaged in the course of carrying out this contract may use any of the information gained by accessing the confidential data for any other purpose except that which was agreed upon in this contract. (4) Notwithstanding subsection 7(3), it is understood that the principles of the Conflict of Interest and Post-Employment Code for the Public Service were not meant to prohibit the persons engaged in this contract from accomplishing any secondary analysis as permitted by the contract. SECURITY REQUIREMENTS 8. (1) Any material to be removed from the Statistics Canada premises by the Principal Investigator or Co-investigators must first be screened by Statistics Canada to ensure that there is no risk of disclosure of confidential information or information that may lead to the identification of an individual respondent. It is the responsibility of the Principal Investigator or Co-investigators to take all precautions to avoid disclosure of confidential information or information that may lead to the identification of an individual respondent. The Principal Investigator or Co-investigators may remove summary files, tabulations and analytical output under the terms of this subsection. (2) The Principal Investigator and the Co-investigators shall not remove any of the original data set or copies of subsets of the microdata file or any confidential sensitive statistical information provided pursuant to this contract from the premises of Statistics Canada. (3) The Principal Investigator and the Co-investigators shall be provided with copies of all relevant Statistics Canada policies related to confidentiality, privacy and security and the standard operating procedures of the appropriate access facility and shall acknowledge in writing their compliance with all of the policies and operating procedures. 34 150 TERM 9. This contract comes into force when signed by both parties and shall continue in force until ; unless revoked or terminated at an earlier date. TERMINATION 10. (1) Where the Principal Investigator is in default in carrying out any of its obligations under this Contract, Statistics Canada may, upon giving written notice to the Principal Investigator, terminate the Contract immediately. (2) The Contract may, by providing 30 days written notice, be terminated by mutual written consent between the Principal Investigator and Statistics Canada. (3) Any notice to be given to Statistics Canada or the Principal Investigator shall be sent by registered mail to: (Address of Statistics Canada) (Address of the Principal Investigator) (4) Any notice shall be deemed to be effective on the day it is received at the address set out above. PENALTIES 11. (1) As a Deemed Employee of Statistics Canada, the Principal Investigator and the Co-investigators are subject to all the applicable penalties provided for in the Statistics Act for contravention of any of the confidentiality provisions and is liable on summary conviction to any of the applicable fines or imprisonment terms. (2) Subsection 11(1) survives indefinitely the completion of this contract or the termination of this Agreement pursuant to subsections 10(1) or 10(2). AMENDMENT 12. No amendment to this contract shall be valid unless it is reduced to writing and signed by the Parties hereto. 35 CONSIDERATION 13. The Parties agree that consideration for this agreement shall be the mutual promises and covenants of the Parties included in this contract. ENTIRE A G R E E M E N T 14. This contract constitutes the entire agreement between the Parties listed below and Statistics Canada with respect to the subject matter described herein and supersedes all previous negotiations, communications and other agreements on the same issue with Statistics Canada unless they are specifically incorporated by reference in this contract. IN WITNESS WHEREOF, this Agreement has been executed in duplicate on behalf of Statistics Canada and the Principal Investigator by: FOR STATISTICS CANADA: Witness Date Director of (Name of Division) or Manager of the Research Data Centre Program FOR T H E PRINCIPAL INVESTIGATOR: Witness Date (Name of Principal Investigator) Witness Date (Name of Co-investigator) 36 152 Appendix "A" - Resea rch P r o p o s a l T i t l e Submitted By: Name of Division and Director Name of Principal Investigator and Co-investigator Address of Principal Investigator Telephone Numbers Affiliation Proposal Detailed description of the proposed research. Data Requirements List of files, additional data fields or variables required for the analysis. Proposed Output Indicate the type of output to be generated from the list below (delete those which do not apply) • RDC working paper: A paper for the RDC working paper series authored by the Principal Investigator for Statistics Canada. • Peer-reviewed journal article: A journal article authored by the Principal Investigator for a peer-reviewed journal. • Book or Book chapter: A book or book chapter authored by the Principal Investigator. • Thesis or Dissertation: A graduate level thesis or dissertation. • Commissioned Report: A commissioned report authored by the Principal Investigator for (insert name of commissioner and proposed title of the report). Completion Date Proposed Completion Date: (Enter Dates and expected product) Research Location Indicate where the Co-investigator is to be located for purposes of this project. Source of Funding List any agencies which may be providing funds related to this project. 37 153 APPENDIX 4 - THE OATH Statistics Statistique Canada Canada Note: This to be taken by all team members listed on the contract. OATH OR AFFIRMATION OF OFFICE AND SECRECY faithfully and honestly fulfill my duties as an employee of Statistics Canada in conformity with the requirements of the Statistics Act, and of all the rules and instructions thereunder and that I will not without due authority in that behalf disclose or make known any matter or thing that comes to my knowledge by reason of my employment. Sworn (affirmed) before me at this day of 20. I, ., do solemnly swear (or affirm) that I will (Authorised to administer oaths and affirmations pursuant to the Statistics Act) Source: Statistics Act. 38 Project ID number APPENDIX 5 - SAMPLE RESEARCH CONTRACT AMENDMENT R E S E A R C H D A T A CENTRES MICRODATA R E S E A R C H C O N T R A C T A M E N D M E N T (Could be used as an extension of contract, addition of team members or addition of dataset) B E T W E E N : HER MAJESTY THE QUEEN IN RIGHT OF CANADA, represented by the Minister of Industry, designated as the Minister for purposes of the Statistics Act, (hereinafter referred to as "Statistics Canada"), AND: (NAME OF INVESTIGATOR) (hereinafter referred to as the "Principal Investigator") (NAME OF CO-INVESTIGATOR OR RESEARCHER) (hereinafter referred to as the "Co-investigator") WHEREAS Statistics Canada and the Principal Investigator named above have signed an On-Premises Microdata Research Contract (copy attached) to undertake a study AND WHEREAS the Co-Investigator named above will be working on this project as a 'deemed employee' of Statistics Canada and by signing this amendment will be subject to the same terms, obligations and penalties as outlined in the attached research contract; OR AND WHEREAS the Investigator wishes to extend the completion date of the contract from to in order to properly complete the research and analysis as outlined in the attached research contract; OR AND WHEREAS the Investigator wishes to have access to the following datasets: - (Name of dataset or datasets) as a new data set to be used for the data output in the original contract in order to properly complete the research and analysis as outlined in the attached research contract; 39 Project ID number IN WITNESS WHEREOF, this amendment has been signed on the dates indicated below. F O R S T A T I S T I C S C A N A D A Witness Date Manager, Research Data Centres Program F O R T H E I N V E S T I G A T O R Witness Date Investigator (could indicate: signing for the team) Witness Date Co-Investigator 40 Project ID number APPENDIX 6 - SAMPLE DISCLOSURE REQUEST FORM Research Data Centre Disclosure Request Form Name: Date: Email address: Phone number: Project number or title: These few simple checks can facilitate the release of disclosure requests: Check that the following is complete: Yes / Not Applicable Variables are clearly labelled, particularly recoded variables. * Cell sizes below the minimum requirement are collapsed or re-categorized. Mistakes in the output are corrected. Specific disclosure issues for the survey (e.g., rounding of EDS output) are addressed. * If Applicable, please provide variables information file: S:\Project nu/nbeADisclosure\Pendinq\Fo/der name Directory of files to be vetted: S:\Pro7ecf /n/mber\Disclosure\Pendinq\Fo/o'er name Output for Release: File Name Survey Name and cycle(s) used Weighting (e.g., population, fractional/adjusted/sa mple, bootstrap) Types of Methods (Insert number(s) from list below or specify) Subsample Characteristics (e.g., gender, age range, regional distribution, etc.) 41 Project ID number File Name Survey Name and cycle(s) used Weighting (e.g., population, fractional/adjusted/sa mple, bootstrap) Types of Methods (Insert number(s) from list below or specify) Subsample Characteristics (e.g., gender, age range, regional distribution, etc.) Types of Methods: 1. Descriptive methods (e.g., Frequencies, Cross-tabular analysis, means and distributions) 2. Scaling methods (e.g., Factor Analysis) 3. Graphs (e.g., histograms) 4 . Regression methods (e.g., O L S , A N O V A , Logistic, Probit, Tobit) 5. Complex methods of modeling (e.g., Structural equation modeling, Hierarchical Linear Model ing, Growth analys is, Survival analys is , Event History Analys is , Simultaneous-equations Models, Fixed Effects Models, Random Effects Models) Support ing Documentat ion Fi les: Includes information to address disclosure issues but is not for release: • A clearly labelled un-weighted version of the tabular data. • The necessary supporting cross-tabulations of binary variables for variance-covariance matrices. • Clear indication of the source of the data if contextual information has been merged with the dataset. 42 Project ID number Directory File Name S:\Project number\Disclosure\Pending\Folder name S:\Project number\Disc losure\Pending\Folder name Residua l Disc losure: A re-release of the same output after slight modifications greatly increases the risk of residual disclosure. Statistics C a n a d a strongly recommends that researchers submit as few versions of the output as possible for release from the R D C . Check: Yes / No / Not Applicable Has a version of this output has been previously re leased? Have var iables been recoded or modified affecting only a few respondents s ince a previous re lease? Have individual outliers or c a s e s have been dropped from the analysis since a previous re lease? H a s the sub-sample or target population has changed slightly from a previous re lease? If the answered is 'yes' to any one of these quest ions, the output is at risk for residual disclosure. P lease consult with the RDC Analyst for solutions and strategies. Please list any other comments or concerns you may have for your disclosure request: 43 1 5 9 Appendix B: UBC Behavioural Research Ethics Board Certificate of Approval The University of British Columbia Office of Research Services Behavioural Research Ethics Board Suite 102, 6190 Agronomy Road, Vancouver, B.C. V6T 1Z3 CERTIFICATE OF APPROVAL - MINIMAL RISK PRINCIPAL INVESTIGATOR: INSTITUTION / DEPARTMENT: UBC BREB NUMBER: UBC/Education/Educational & Kadriye Ercikan Counselling Psychology, and Special H06-03719 Education INSTITUTIONS) WHERE RESEARCH WILL BE CARRIED OUT: I n s t i t u t i o n I S i t e N/A N/A O t h e r l o c a t i o n s w h e r e t h e r e s e a r c h w i l l b e c o n d u c t e d : British Columbia Inter-university Research Data Centre (BCIRDC) CO-INVESTIGATOR(S): Jennifer Shapka SPONSORING AGENCIES: N/A PROJECT TITLE: Psychological and Educational Effects of Childhood Chronic Illnesses CERTIFICATE EXPIRY DATE: February 6, 2008 DOCUMENTS INCLUDED IN THIS APPROVAL: DATE APPROVED: February 6, 2007 D o c u m e n t N a m e I V e r s i o n I D a t e Protocol: Proposal_BREBs Application 1 December 5, 2006 The application for ethical review and the document(s) listed above have been reviewed and the procedures were found to be acceptable on ethical grounds for research involving human subjects. \ Approval is issued on behalf of the Behavioural Research Ethics Board and signed electronically by one of the following: \ Dr. Peter Suedfeld, Chair Dr. Jim Rupert, Associate Chair Dr. Arminee Kazanjian, Associate Chair Dr. M. Judith Lynam, Associate Chair 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0054567/manifest

Comment

Related Items