UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Do relational differences in demographics and work values result in conflict and burnout in the nursing… Wolff, Angela Christine 2009

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

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
24-ubc_2009_fall_wolff_angela.pdf [ 2.76MB ]
Metadata
JSON: 24-1.0067703.json
JSON-LD: 24-1.0067703-ld.json
RDF/XML (Pretty): 24-1.0067703-rdf.xml
RDF/JSON: 24-1.0067703-rdf.json
Turtle: 24-1.0067703-turtle.txt
N-Triples: 24-1.0067703-rdf-ntriples.txt
Original Record: 24-1.0067703-source.json
Full Text
24-1.0067703-fulltext.txt
Citation
24-1.0067703.ris

Full Text

DO RELATIONAL DIFFERENCES IN DEMOGRAPHICS AND WORK VALUES RESULT IN CONFLICT AND BURNOUT IN THE NURSING WORKFORCE? by ANGELA CHRISTINE WOLFF BScN, McMaster University, 1991 MSN, University of British Columbia, 1998  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Nursing)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  September 2009  © Angela Christine Wolff, 2009  ABSTRACT The consequences of diversity have not been formally considered as contributing to undesirable work environments in healthcare. I sought to address this gap by examining a conceptual model that explains how diversity within the nursing workforce gives rise to interpersonal conflict (relationship and task) within workgroups, which in turn, is linked to burnout (emotional exhaustion, depersonalization, and diminished personal accomplishment). Diversity was defined as the degree of relative difference or dissimilarity between an individual and other workgroup members on select attributes, which in this study were age, education, ethnicity/race, and work values. Using a cross-sectional survey design, data were taken from a population-based sample of 603 nurses (registered nurses and licensed practical nurses) (80% response rate) in two acute care hospitals in British Columbia, Canada. At the individual level of analysis, a two-step approach to latent variable modelling was used: (a) factor analysis techniques to test and establish the validity of the measurement model and (b) structural equation modelling to test the hypothesized model. Partial support for the proposed model was found for both the direct relationships between diversity and burnout as well as the mediating effects of interpersonal conflict. Overall, the results indicated that perceived diversity explained a greater percentage of the variance in burnout compared with the explanatory power of actual diversity. Specifically, perceived work values and educational diversity were the most important explanatory variables of depersonalization (Pratt index = 58% and 21%, respectively) and were similarly predictive of diminished personal accomplishment (Pratt index = 69% and 35%, respectively). Emotional exhaustion was solely (Pratt index = 100%) explained by perceived work values diversity; however, the total variance explained was very minimal. Both individuals’ involvement in relationship and task conflict were the predominant mediating variables of the relationships between perceived work values diversity and emotional exhaustion (59% and 76% total mediation, respectively), depersonalization (57% and 68% total mediation, respectively), and diminished personal accomplishment (28% and 32% total mediation, respectively). The implications of the study relate to nurses and decision-makers at the micro, meso, and macro level of practice to create a climate of support for, and acceptance of, diversity in healthcare workplaces.  ii  TABLE OF CONTENTS Abstract ......................................................................................................................................ii Table of Contents .......................................................................................................................iii List of Tables .............................................................................................................................ix List of Figures...........................................................................................................................xii List of Equations ...................................................................................................................... xiv List of Abbreviations................................................................................................................. xv Acknowledgements .................................................................................................................. xvi Dedication............................................................................................................................... xvii 1  INTRODUCTION ...........................................................................................................1  1.1  Diversity in the Nursing Workforce ...................................................................................1 1.1.1 1.1.2 1.1.3 1.1.4 1.1.5  Age .......................................................................................................................1 Educational Preparation.........................................................................................3 Ethnicity/Race.......................................................................................................4 Work Values .........................................................................................................6 Summary...............................................................................................................6  1.2  Why Study Diversity in the Nursing Workforce? ...............................................................7  1.3  Perspectives of Diversity ...................................................................................................9  1.4  Research Purpose............................................................................................................. 11  1.5  Chapter Summary ............................................................................................................ 12  2  LITERATURE REVIEW.............................................................................................. 13  2.1  The Relational Approach to Diversity .............................................................................. 14 2.1.1  Defining Relational Diversity .............................................................................. 14  2.1.2  Theoretical Foundations ...................................................................................... 15 2.1.2.1 2.1.2.2  2.2  Social Identity Theory ......................................................................... 15 Similarity–Attraction Theory............................................................... 15  2.1.3  Operational Definitions of Relational Diversity ................................................... 16  2.1.4  An Overview of the Diversity Attributes Studied in Previous Research................ 19  2.1.5  Empirical Literature Concerning Relational Diversity in Nursing Workgroups .... 20  2.1.6  Summary ............................................................................................................ 23  Burnout as an Outcome of Relational Diversity................................................................ 24 2.2.1  Defining Burnout ................................................................................................ 25  2.2.2.  Consequences of Burnout .................................................................................... 26  2.2.3  Antecedents of Burnout ....................................................................................... 26  2.2.4  Prevalence of Burnout in Canadian Nurses .......................................................... 28  iii  2.2.5  Empirical Literature Concerning Burnout and the Relational Diversity within Nursing Workgroups ........................................................................................... 29  2.2.6  Empirical Literature about the Relationships between Relational Diversity and Outcomes Interrelated with Burnout .................................................................... 29 2.2.6.1 2.2.6.2 2.2.6.3 2.2.6.4  2.2.7 2.3  Actual and Perceived Age Diversity..................................................... 30 Actual and Perceived Educational Diversity......................................... 32 Actual and Perceived Ethnic/Racial Diversity ...................................... 33 Actual and Perceived Work Values Diversity....................................... 34  Summary ............................................................................................................ 35  Interpersonal Conflict as a Mediator of the Relationship between Diversity and Burnout.. 36 2.3.1  Defining Interpersonal Conflict ........................................................................... 37  2.3.2  Empirical Literature Concerning Nurses’ Interpersonal Conflict .......................... 39  2.3.3  The Empirical Literature Concerning the Relationship between Relational Diversity and Interpersonal Conflict .................................................................... 40 2.3.3.1 2.3.3.2 2.3.3.3  2.3.4  Individuals’ Involvement in Conflict.................................................... 41 Individuals’ Perceptions of Conflict within their Workgroup................ 41 Summary ............................................................................................. 42  The Empirical Literature Concerning the Relationship between Conflict and Burnout ........................................................................................................ 43  2.3.5  Summary............................................................................................................. 45  2.4  Chapter Summary ............................................................................................................ 46  3  CONCEPTUAL MODEL AND HYPOTHESES.......................................................... 48  3.1  Theoretical Foundations................................................................................................... 48  3.2  The Conceptual Link between Relational Diversity and Burnout: Why are Dissimilar Individuals More Likely to Experience Burnout? ............................................ 50  3.3  The Conceptual Link between the Effect of Relational Diversity on Burnout as Mediated by Interpersonal Conflict? ................................................................................ 55  3.4  Chapter Summary ............................................................................................................ 60  4  METHODS .................................................................................................................... 63  4.1  Sample ............................................................................................................................ 63  4.2  4.1.1  Setting and Participants ....................................................................................... 63  4.1.2  Recruitment of Participants ................................................................................. 64  4.1.3  Sample Size ........................................................................................................ 66  4.1.4  Survey Response Rates........................................................................................ 66  Data Collection Process ................................................................................................... 69 4.2.1  Modified Tailored Design Method....................................................................... 70  iv  4.3  4.2.2  Application of the Tailored Design Method ......................................................... 71  4.2.3  Pretesting ............................................................................................................ 75  Operationalization of Study Constructs ............................................................................ 75 4.3.1  Exogenous Variable: Relational Diversity ........................................................... 79 4.3.1.1 4.3.1.2  4.3.2  Mediating Variable: Interpersonal Conflict .......................................................... 84  4.3.3  Endogenous Variable: Burnout............................................................................ 86 4.3.3.1 4.3.3.2 4.3.3.3 4.3.3.4  4.4  Reliability............................................................................................ 87 Content Validity .................................................................................. 87 Construct Validity ............................................................................... 88 Convergent and Discriminant Validity ................................................. 89  Data Analysis Procedures ................................................................................................ 89 4.4.1  Data Preparation and Screening........................................................................... 90  4.4.2  Representation of Ordinal Variables .................................................................... 91  4.4.3  Structural Equation Modelling............................................................................. 91 4.4.3.1 4.4.3.2 4.4.3.3  4.5  Actual Approach to the Measurement of Relational Diversity............... 79 Perceptual Approach to the Measurement of Relational Diversity........ 84  Exploratory Factor Analysis................................................................. 92 Confirmatory Factor Analysis .............................................................. 92 Method of Estimation .......................................................................... 92  4.4.4  Missing Data....................................................................................................... 93  4.4.5  Model Evaluation................................................................................................ 94  Additional Statistical Methods ......................................................................................... 96 4.5.1  Testing the Mediation Models ............................................................................. 97  4.6  Ethical Consideration..................................................................................................... 100  4.7  Chapter Summary .......................................................................................................... 101  5  FINDINGS ................................................................................................................... 103  5.1  Data Screening and Variable Transformation ................................................................. 103  5.2  Descriptive Statistics of the Sample ............................................................................... 104 5.2.1  5.3  Hospital-based Group Differences ..................................................................... 107  Measurement Model for the Exogenous Variables: Actual Diversity .............................. 110 5.3.1  Exploratory Factor Analysis of the Contemporary Work Values Scale............... 110  5.4  Measurement Model for the Exogenous Variables: Perceived Diversity ......................... 114  5.5  Measurement Model for the Mediator Variable: Intragroup Conflict............................... 116  5.6  Measurement Model for the Mediator Variable: Individual Conflict ............................... 120 5.6.1  Initial CFA for Three Factors with all Items ...................................................... 121  5.6.2  CFA of the Task and Relationship Subscales ..................................................... 123  v  5.7  Measurement Model for the Endogenous Variable: Burnout.................................................126  5.8  Examination of Missing Data for the Study Variables ..........................................................133  5.9  Descriptive Statistics for the Exogenous Variables: Relational Diversity ..............................136 5.9.1  Actual Diversity ......................................................................................................136  5.9.2  Perceived Diversity .................................................................................................138  5.10 Descriptive Statistics of the Mediator Variables: Interpersonal Conflict................................139 5.10.1 Intragroup Conflict Scale.........................................................................................139 5.10.2 Individual Conflict Scale.........................................................................................139 5.11 Descriptive Statistics of the Outcome Variable: Burnout ......................................................140 5.12 Bivariate Statistics of the Study Variables ............................................................................141 5.13 Chapter Summary ................................................................................................................144 6  STUCTURAL EQUATION MODELLING FINDINGS..................................................145  6.1  Overview of Methods ..........................................................................................................146  6.2  Organization of the Findings ................................................................................................147  6.3  The Direct and Indirect Effects of Actual Relational Diversity on Burnout ...........................147 6.3.1  Model Fit ................................................................................................................148  6.3.2  Model 1: The Direct Effects of Actual Relational Diversity on Burnout (Condition 1)...........................................................................................................152 6.3.2.1  6.3.3  Summary .................................................................................................153  Model 2: The Effects of Actual Relational Diversity on Burnout as Mediated by Intragroup Conflict .............................................................................................154 6.3.3.1 6.3.3.2 6.3.3.3 6.3.3.4  6.3.4  Condition 2..............................................................................................154 Condition 3..............................................................................................154 Condition 4..............................................................................................155 Summary .................................................................................................155  Model 3: The Effects of Actual Relational Diversity on Burnout as Mediated by Individual Conflict .............................................................................................160 6.3.4.1 6.3.4.2 6.3.4.3 6.3.4.4  6.3.5  Condition 2..............................................................................................160 Condition 3..............................................................................................160 Condition 4..............................................................................................161 Summary .................................................................................................161  Summary of the Direct and Indirect Effects of Actual Relational Diversity on Burnout ..............................................................................................................165  6.4  The Direct and Indirect Effects of Perceived Relational Diversity on Burnout ......................165 6.4.1  Model Fit ................................................................................................................165  vi  6.4.2  Model 4: The Direct Effects of Perceived Relational Diversity on Burnout (Condition 1)...........................................................................................................170 6.4.2.1 6.4.2.2 6.4.2.3  6.4.3  Further Exploration of Unexpected Findings ............................................171 Ordering the Exogenous Variables in Terms of their Importance..............176 Summary .................................................................................................177  Model 5: The Effects of Perceived Relational Diversity on Burnout as Mediated by Intragroup Conflict .............................................................................................179 6.4.3.1 6.4.3.2 6.4.3.3 6.4.3.4  6.4.4  Condition 2..............................................................................................179 Condition 3..............................................................................................180 Condition 4..............................................................................................180 Summary .................................................................................................185  Model 6: The Effects of Perceived Relational Diversity on Burnout as Mediated by Individual Conflict .............................................................................................188 6.4.4.1 6.4.4.2 6.4.4.3 6.4.4.4  6.4.5  Condition 2..............................................................................................188 Condition 3..............................................................................................189 Condition 4..............................................................................................189 Summary .................................................................................................195  Summary of the Direct and Indirect Effects of Perceived Relational Diversity on Burnout ...............................................................................................198  6.5  Chapter Summary ................................................................................................................198  7  DISCUSSION.....................................................................................................................201  7.1  Review of the Findings ........................................................................................................201 7.1.1  Direct Effects of Relational Diversity on Burnout ....................................................202  7.1.2  The Mediating Influence of Interpersonal Conflict...................................................202  7.2  Strengths and Limitations ....................................................................................................203  7.3  A Discussion of the Current Study Findings in Relation to Other Evidence ..........................207 7.3.1  The Direct Effects of Actual Relational Diversity on Burnout ..................................207  7.3.2  The Direct Effects of Perceived Relational Diversity on Burnout.............................209 7.3.2.1 7.3.2.2 7.3.2.3 7.3.2.4 7.3.2.5  7.3.3  Education ................................................................................................209 Work Values............................................................................................210 Age .........................................................................................................212 Ethnicity/Race .........................................................................................212 Theoretical Explanations of the Salience of Education and Work Values in Nursing ..........................................................................213  The Mediating Influence of Interpersonal Conflict...................................................214 7.3.3.1 7.3.3.2 7.3.3.3  The Relationship between Relational Diversity and Interpersonal Conflict ...................................................................................................214 The Relationship between Interpersonal and Burnout ...............................215 The Mediation Model of Diversity on Burnout through Conflict ..............216  vii  7.4  Theoretical Implications ......................................................................................................217 7.4.1  Relational Diversity.................................................................................................217  7.4.2  Burnout...................................................................................................................219  7.4.3  Interpersonal Conflict..............................................................................................220  7.5  Practical Implications ..........................................................................................................221  7.6  Future Research Directions ..................................................................................................225  7.7  Conclusions .........................................................................................................................227  REFERENCES............................................................................................................................229 APPENDICES.............................................................................................................................243 Appendix A:  Summary Tables of the Review of the Relational Diversity Literature ..................244  Appendix B:  Questionnaire.......................................................................................................253  Appendix C:  UBC Behavioural Research Ethics Board Certificate of Approval ........................271  Appendix D:  Fraser Health Authority Research Ethics Board Certificate of Approval ...............272  Appendix E:  Inter-Item Correlation Matrices ............................................................................273  Appendix F:  Additional Factor Analyses for the Individual Conflict Scale................................276  Appendix G:  Frequency Distributions of the Study Variables ....................................................279  viii  LIST OF TABLES Table 4.1  Reasons for the Exclusion of Some Nurses Originally Identified as Eligible ..... 67  Table 4.2  Survey Response Rates .................................................................................... 69  Table 4.3  List of Scales/Items in Final Study Questionnaire............................................. 76  Table 4.4  Normative Scores of the Maslach Burnout Inventory Subscales ........................ 87  Table 5.1  Summary of Scale Items with More than One Response ................................. 104  Table 5.2  Employment Characteristics of the Respondents............................................. 105  Table 5.3  Demographic Characteristics and Hospital-based Group Comparison of the Respondents................................................................................................... 108  Table 5.4  Structure Matrix of the EFA for the 16-item Contemporary Work Values Scale .............................................................................................................. 112  Table 5.5  CFA Results for the Perceived Work Values .................................................. 115  Table 5.6  CFA Results for the Intragroup Conflict Scale with a Three-factor Solution.... 117  Table 5.7  CFA Results for the Intragroup Conflict Scale with a Two-factor Solution ..... 119  Table 5.8  CFA Results for the Individual Conflict Scale with a Three-factor Solution .... 122  Table 5.9  CFA Results for the Individual Conflict Scale with a Two-factor Solution...... 124  Table 5.10  CFA Results for the Maslach Burnout Inventory with Four-factor and Three-factor Solutions.................................................................................... 127  Table 5.11  CFA Results for the Maslach Burnout Inventory with a Three-factor Solution and 8 Cross-loadings ........................................................................ 130  Table 5.12  Summary of the CFAs for the Maslach Burnout Inventory.............................. 131  Table 5.13  Frequency of Missing Data for the Study Variables ........................................ 135  Table 5.14  Summary of Missing Data Patterns................................................................. 136  Table 5.15  Descriptive Statistics and Hospital-based Group Comparisons of the Study Variables.............................................................................................. 137  Table 5.16  Percentage of Nurses Classified as Having High, Moderate, and Low Levels of Burnout for Each Aspect of the MBI ............................................... 141  Table 5.17  Pearson Correlation Matrix for the Diversity and Burnout Latent Variables and the Observed Demographic Variables....................................... 142  ix  Table 5.18  Pearson Correlation Matrix for the Perceived Diversity, Actual Diversity, and Conflict Latent Variables......................................................................... 143  Table 5.19  Pearson Correlation Matrix for the Conflict and Burnout Latent Variables...... 144  Table 6.1  Summary of Variables in Each Model ............................................................ 146  Table 6.2  Summary of the Goodness-of-Fit Indices and Total Variance Explained for the Effects of Actual Relational Diversity on Burnout .................................... 149  Table 6.3  Unstandardized and Standardized Parameter Estimates for Model 1: The Overall Direct Effects of Actual Relational Diversity on Burnout ................... 152  Table 6.4  Unstandardized and Standardized Parameter Estimates for Model 2a: The Direct and Indirect Effects of Actual Relational Diversity on Burnout as Mediated by Intragroup Relationship Conflict ................................................ 156  Table 6.5  Unstandardized and Standardized Parameter Estimates for Model 2b: The Direct and Indirect Effects of Actual Relational Diversity on Burnout as Mediated by Intragroup Task Conflict ............................................................ 157  Table 6.6  Standardized Mediation Effects for Model 2a: The Effects of Actual Relational Diversity on Burnout as Mediated by Intragroup Relationship Conflict .......... 158  Table 6.7  Standardized Mediation Effects for Model 2b: The Effects of Actual Relational Diversity on Burnout as Mediated by Intragroup Task Conflict....................... 159  Table 6.8  Unstandardized and Standardized Parameter Estimates for Model 3a: The Direct and Indirect Effects of Actual Relational Diversity on Burnout as Mediated by Individual Relationship Conflict ................................................ 161  Table 6.9  Unstandardized and Standardized Parameter Estimates for Model 3b: The Direct and Indirect Effects of Actual Relational Diversity on Burnout as Mediated by Individual Task Conflict............................................................. 162  Table 6.10  Standardized Mediation Effects for Model 3a: The Effects of Actual Relational Diversity on Burnout as Mediated by Individual Relationship Conflict ........... 163  Table 6.11  Standardized Mediation Effects for Model 3b: The Effects of Actual Relational Diversity on Burnout as Mediated by Individual Task Conflict ...... 164  Table 6.12  Summary of the Goodness-of-Fit Indices and Total Variance Explained for the Effects of Perceived Relational Diversity on Burnout ............................... 166  Table 6.13  Unstandardized and Standardized Parameter Estimates for Model 4: The Overall Direct Effects of Perceived Relational Diversity on Burnout .............. 170  Table 6.14  Post Hoc Comparisons of Depersonalization Score by Perceived Age Diversity........................................................................................................ 172  x  Table 6.15  Post Hoc Comparisons of Perceived Age Diversity by Observed Age Group.. 173  Table 6.16  Relative Importance of Perceived Diversity Variables in Explaining Burnout.......................................................................................................... 177  Table 6.17  Unstandardized and Standardized Parameter Estimates for Model 5a: The Direct and Indirect Effects of Perceived Relational Diversity on Burnout as Mediated by Intragroup Relationship Conflict ................................................ 181  Table 6.18  Unstandardized and Standardized Parameter Estimates for Model 5b: The Direct and Indirect Effects of Perceived Relational Diversity on Burnout as Mediated by Intragroup Task Conflict ............................................................ 182  Table 6.19  Standardized Mediation Effects for Model 5a: The Effects of Perceived Relational Diversity on Burnout as Mediated by Intragroup Relationship Conflict .......... 183  Table 6.20  Standardized Mediation Effects for Model 5b: The Effects of Perceived Relational Diversity on Burnout as Mediated by Intragroup Task Conflict....................... 184  Table 6.21  Unstandardized and Standardized Parameter Estimates for Model 6a: The Direct and Indirect Effects of Perceived Relational Diversity on Burnout as Mediated by Individual Relationship Conflict ................................................ 191  Table 6.22  Unstandardized and Standardized Parameter Estimates for Model 6b: The Direct and Indirect Effects of Perceived Relational Diversity on Burnout as Mediated by Individual Task Conflict............................................................. 192  Table 6.23  Standardized Mediation Effects for Model 6a: The Effects of Perceived Relational Diversity on Burnout as Mediated by Individual Relationship Conflict ........... 193  Table 6.24  Standardized Mediation Effects for Model 6b: The Effects of Perceived Relational Diversity on Burnout as Mediated by Individual Task Conflict ....................... 194  Table 6.25  Summary of Hypotheses Supported................................................................ 199  Table A.1  Summary of Articles about Actual Relational Diversity within Workgroups ... 244  Table A.2  Summary of Articles about Perceived Relational Diversity within Workgroups ... 249  Table E.1  Polychoric Correlation Matrix for the 16-item Contemporary Work Values Scale.............................................................................................................. 273  Table E.2  Polychoric Correlation Matrix for the Perceived Work Values Diversity Items273  Table E.3  Polychoric Correlation Matrix for the Intragroup Conflict Items..................... 274  Table E.4  Polychoric Correlation Matrix for the Individual Conflict Items ..................... 274  Table E.5  Polychoric Correlation Matrix for the Maslach Burnout Inventory Items ........ 275  xi  LIST OF FIGURES Figure 1.1  Overview of the Postulated Model.................................................................... 11  Figure 3.1  Model 1: The Effect of Actual Relational Diversity on Burnout ........................ 53  Figure 3.2  Model 2: The Effect of Perceived Relational Diversity on Burnout................... 54  Figure 3.3  Model 3: The Effect of Actual Relational Diversity on Burnout as Mediated by Interpersonal Conflict................................................................... 61  Figure 3.4  Model 4: The Effect of Perceived Relational Diversity on Burnout as Mediated by Interpersonal Conflict................................................................... 62  Figure 4.1  Flow Diagram of Participant Recruitment......................................................... 68  Figure 4.2  Data Collection Process.................................................................................... 72  Figure 4.3  Single Mediator Model..................................................................................... 98  Figure 5.1  Final Measurement Model for the Contemporary Work Values Scale ............. 113  Figure 5.2  Final Measurement Model for the Perceived Work Values Scale .................... 116  Figure 5.3  Final Measurement Model for the Two-factor Intragroup Conflict Scale......... 120  Figure 5.4  Final Measurement Model for the Two-factor Individual Conflict Scale ......... 125  Figure 5.5  Final Measurement Model for the Three-factor MBI....................................... 132  Figure 6.1  Model 1: The Effects of Actual Relational Diversity on Burnout..................... 148  Figure 6.2  Model 2a and 3a: The Effects of Actual Relational Diversity on Burnout as Mediated by Relationship Conflict ............................................................. 150  Figure 6.3  Model 2b and 3b: The Effects of Actual Relational Diversity on Burnout as Mediated by Task Conflict ......................................................................... 151  Figure 6.4  Significant Pathways for Model 1: The Effects of Actual Relational Diversity on Burnout...................................................................................... 153  Figure 6.5  Model 4: The Effects of Perceived Relational Diversity on Burnout ............... 167  Figure 6.6  Model 5a and 6a: The Effects of Perceived Relational Diversity on Burnout as Mediated by Relationship Conflict................................................ 168  Figure 6.7  Model 5b and 6b: The Effects of Perceived Relational Diversity on Burnout as Mediated by Task Conflict............................................................ 169  Figure 6.8  Box Plots of Depersonalization Subscale Total Scores by Perceived Age Diversity................................................................................................. 171  Figure 6.9  Box Plots of Perceived Age Diversity by Observed Age Group ...................... 172  xii  Figure 6.10  Box Plots of Perceived Educational Diversity by Observed Nursing Education.....175  Figure 6.11  Significant Pathways for Model 4: The Effects of Perceived Diversity on Burnout................................................................................................................178  Figure 6.12  Significant Pathways for Model 5a: The Effects of Perceived Diversity on Burnout as Mediated by Intragroup Relationship Conflict ................................186  Figure 6.13  Significant Pathways for Model 5b: The Effects of Perceived Diversity on Burnout as Mediated by Intragroup Task Conflict.................................................187  Figure 6.14  Significant Pathways for Model 6a: The Effects of Perceived Diversity on Burnout as Mediated by Individual Relationship Conflict .....................................196  Figure 6.15  Significant Pathways for Model 6b: The Effects of Perceived Diversity on Burnout as Mediated by Individual Task Conflict .................................................197  Figure G.1  Relative Frequency Distribution of the D-Scores for the Actual Age Diversity Variable ...............................................................................................................279  Figure G.2  Relative Frequency Distribution of the D-Scores for the Actual Educational Diversity Variable................................................................................................279  Figure G.3  Relative Frequency Distribution of the D-Scores for the Actual Ethnic/racial Diversity Variable................................................................................................280  Figure G.4  Relative Frequency Distribution of the Average Total Score for the Contemporary Work Values Scale (16 items) .......................................................280  Figure G.5  Relative Frequency Distribution of the Actual Work Values Diversity Variable....281  Figure G.6  Relative Frequency Distribution of the Perceived Age Diversity Variable.............281  Figure G.7  Relative Frequency Distribution of the Perceived Educational Diversity Variable ...............................................................................................................282  Figure G.8  Relative Frequency Distribution of the Perceived Ethnic/racial Diversity Variable ...............................................................................................................282  Figure G.9  Relative Frequency Distribution of the Perceived Work Values Diversity Subscale Scores....................................................................................................283  Figure G.10  Relative Frequency Distribution of the Intragroup Relationship Conflict Scores ...283  Figure G.11  Relative Frequency Distribution of the Intragroup Task Conflict Scores ...............284  Figure G.12  Relative Frequency Distribution of the Individual Relationship Conflict Scores....284  Figure G.13  Relative Frequency Distribution of the Individual Task Conflict Scores................285  Figure G.14  Relative Frequency Distribution of the Emotional Exhaustion Scores ...................285  Figure G.15  Relative Frequency Distribution of the Depersonalization Scores .........................286  Figure G.16  Relative Frequency Distribution of the Personal Accomplishment Scores .............286  xiii  LIST OF EQUATIONS Equation 4.1  Euclidean Distance Measure.................................................................................80  xiv  LIST OF ABBREVIATIONS BC CFA CFI CON CWV CY df E/CFA EFA D-score DIFFTEST DSAge DSEduc DSEth DSVal DP EE EPC LPN MBI MBI-GS MBI-HHS MI ML PAge PA PC PEduc PEth PRS PVal RC REL RMSEA RN SD SEM SRMR TC TDM TLI TSK WLSMV  British Columbia Confirmatory factor analysis Comparative fit index Conflict Contemporary work values Cynicism Degress of freedom Exploratory factor analysis within the confirmatory factor analysis framework Exploratory factor analysis Euclidean distance measure Chi-squared difference test Actual age diversity Actual educational diversity Actual ethnic/racial diversity Actual work values diversity Depersonalization Emotional exhaustion Standardized expected parameter change Licensed practical nurse Maslach Burnout Inventory Maslach Burnout Inventory – General Survey Maslach Burnout Inventory – Human Services Survey Modification indices Maximum likelihood Perceived age diversity Personal accomplishment Process conflict Perceived educational diversity Perceived ethnic/racial diversity Process Perceived work values diversity Relationship conflict Relationship Root mean square error of approximation Registered nurse Standard deviation Structural equation modelling Standardized root mean residual Task conflict Tailored design method Tucker-Lewis index Task Weighed least squares estimation, mean and variance adjusted  xv  ACKNOWLEDGEMENTS Before I started this journey, I was told the key aspects of being successful would be perseverance and intellectual curiosity. While I now recognize that a passion for learning and personal tenacity is required, a project of this magnitude could not have been achieved without the dedication and expertise of my supervisor, Dr. Pamela Ratner, and the other members of my dissertation committee: Dr. John Oliffe, Dr. Sandra Robinson, and Dr. Linda McGillis-Hall. The high degree of respect, support, and commitment I received fostered my learning. Dr. Ratner was also extremely generous in her feedback and guidance in developing my scholarly writing abilities. I feel very fortunate to have had such an outstanding supervisor to guide me through my doctoral program of research. Collectively, my committee has shown me what is possible and has influenced my life in many ways, as well as in ways yet unknown. I thank the many professors that have fulfilled my desire for learning and inspired my ongoing passion for learning and research. Finally, I acknowledge the nurse educators who, during my undergraduate degree at McMaster University, provided a learning environment that fostered independence, self-directedness, and a spirit of inquiry necessary to be a successful graduate student, nurse, and researcher. To the many individuals, in my personal life, who have accompanied me throughout my education journey, there are no words to fully express my gratitude. To begin, a heartfelt thank you to my husband, Blair; daughter, Miranda; my parents; other family members; friends; and colleagues. I value the perseverance all of you demonstrated while encouraging me to complete my studies. I also want to express my gratitude to my sister, Barb, and my friend, Eleanor, who was always willing to listen or assist with editing. A special thanks to Barbara Mildon who offered endless support and encouragement. Next, I appreciate the financial contributions I received from the Social Sciences and Humanities Research Council, the Michael Smith Foundation for Health Research, the Canadian Nurses’ Foundation, NEXUS, the UBC School of Nursing, and the Xi Eta Chapter, Sigma Theta Tau International. Finally, a sincere thank-you to the participants of my study and the in-kind support received from the Fraser Health Authority. With this dissertation, I celebrate what feels right in the world, recognize my lifelong dream has been achieved, and submit myself to the vulnerabilities of being open to new possibilities.  xvi  A Special Message to Miranda, Picture what you want. . . . Pretend you are what you’d like to be. Make a picture in your mind so you can see that what you want can come true. If you believe in your heart, it will come to you. — Dr. W.W. Dyer Thanks for always helping me to remember what is important in life. I hope the perseverance, determination, and creativity that I have role-modelled will be an inspiration to you as your own life unfolds.  xvii  1  INTRODUCTION Diversity in the workforce is a phenomenon experienced globally. The success of  organizations and the well-being of their members are dependant on understanding the effects of human diversity. When diversity is embraced, differences in backgrounds, perspectives, and skills may provide advantages to individuals and organizations (e.g., job satisfaction, workgroup involvement, commitment, retention, improved problem solving and decision making, and creativity). Conversely, failure to consider the alignment of such human differences may lead to a poor fit between employees and their place of work. It is essential, therefore, to understand how such differences operate. In this study, I sought to advance this understanding by examining the impact of diversity in nursing workplaces.  1.1  Diversity in the Nursing Workforce Historically, the Canadian nursing profession has been fairly homogenous; however,  the attributes of the nursing workforce have changed significantly within the past 20 years. Most noticeable are changes in nurses’ ages, education, ethnicity/race, and possibly work values, although this latter feature has not been well studied. These attributes, in part, reflect trends in the demographic characteristics of the Canadian population, changing policies in the educational requirements for entry to practice as a nurse (i.e., baccalaureate versus diploma preparation for registered nurses), fluctuations in the labour market, and national and organizational policies related to workforce equality and recruitment. 1.1.1  Age There has been a significant shift in the percentage of nurses representing various age  cohorts, with the highest proportion of the Canadian registered nursing workforce (68%) being between 40 and 65+ years of age. More specifically, Canadian registered nurses (RNs) who are 50 years of age and older currently represent 39% of the workforce, which is substantially greater than the1980 average of 16% (Canadian Institute for Health Information, 2008; Canadian Nurses Association, 2002). This means that one in three Canadian RNs is 50 years of age or older (Canadian Institute for Health Information, 2004). In 2007, Canadian RNs under 30 years of age constituted a mere 11% (Canadian Institute for Health Information, 2008), whereas in 1980, this age cohort made up approximately 30% of the workforce. Currently, the average age of RNs in British Columbia is 46.2 years (Canadian Institute for Health Information, 2008), 1  which has gradually increased by 2.2 years since 1999 (Canadian Institute for Health Information, 2004). The trends observed in the licensed practical nursing1 workforce in British Columbia are similar to those observed in the RN population. Almost two thirds (59%) of licensed practical nurses (LPNs) employed in British Columbia are 40 years of age and older (Canadian Institute for Health Information, 2008). The average age of LPNs in British Columbia is 42.3 years, which has declined slightly during the past five years (the average age in 2003 was 45.3 years) (Canadian Institute for Health Information, 2008). This trend is attributed in part to the increase in the number of LPNs entering the workforce who are under the age of 40. Although slightly younger than RNs, the average age of LPNs is still similar to the average age of the Canadian population (approximately 39 years). The increase in the age of the nursing workforce can be attributed to several societal and labour market trends. A significant portion of the Canadian population is 40 years of age or older, resulting from the “baby boom” of the late 1940s to mid 1960s (Statistics Canada, 2007b). From 1970 to 1980, as the baby boomer cohort entered the workforce, the nursing workforce contained a greater percentage (47% to 56%) of nurses under 35 years of age represented the nursing workforce (Canadian Nurses Association, 2002). This cohort is now older and remains the largest group in the population. Although the age of nurses in British Columbia reflect the national population trends, the percentage of RNs aged 45 and older is higher than the national average: 56.3% compared with 26.3%, respectively (Canadian Institute for Health Information, 2008; Statistics Canada, 2007a). Another factor contributing to the age distribution of RNs is the labour trends of the 1990s, when there was significant downsizing in healthcare, and many nurses chose non-nursing careers or sought employment abroad. Accompanying the downsizing in healthcare was a reduction in the number of places in schools of nursing provincially and nationally in the mid-1990s, which was at its lowest in 30 years (Canadian Institute for Health  1  There are three regulated nursing professions in Canada: registered nurses (RNs), licensed practical nurses (LPNs), and registered psychiatric nurses (RPNs). Each provincial and territorial jurisdiction in Canada has its own regulatory body for the regulation and licensure of registrants for each profession. As of 2007, there were 332,794 regulated nurses working in nursing in Canada, with the majority being RNs (78%) and to a lesser extent LPNs (21%). RPNs represent only 2% of regulated nurses. In British Columbia, RNs, LPNs, and RPNs are regulated and educated as separate professions (Canadian Institute for Health Information, 2008). Given the setting of this study in acute care hospitals, RPNs were excluded.  2  Information, 2005; Canadian Nurses Association, 2002). Another contributing factor to the aging trend of the nursing workforce is the age of nursing graduates. The average age of RNs seeking employment after graduation in 2006 was 27 years, compared to 23 years in the early 1980s. Moreover, the percentage of RNs aged 30 or older graduating from their initial nursing program has almost doubled since the 1980s (Canadian Institute for Health Information, 2007b). The same trend has occurred in the LPN workforce. For example, 54% of the LPN workforce who graduated between 2005 and 2007 were aged 30 or older (Canadian Institute for Health Information, 2008). Although nurses have always had to work with colleagues of different ages, in the current nursing workforce, a disproportionately large number of nurses from some age cohorts are represented, particularly those over 40 years of age. Accordingly, younger age cohorts of nurses are underrepresented. Those aged 40 years and over are in the majority and, as such, may be inordinately dominant in shaping the workgroup norms that have greatly influenced nursing practice for the past 30 years. 1.1.2  Educational Preparation The educational demographics of those recently entering the nursing workforce,  relative to those currently practicing, have also changed. Since the 1970s, changes in policies pertaining to entry-level RN education have shifted from hospital-based apprenticeship programs and diploma-based college programs to university baccalaureate degrees (Dussault et al., 1999). The majority of RNs (85% to 98%) who graduated in the 1960s to 1990s were educated in hospital- or diploma-based nursing programs (Ryten, 1997). In 1999, approximately 11% of the Canadian RN workforce had earned a baccalaureate degree before entering practice (Canadian Institute for Health Information, 2004). During the late 1990s, however, most of the Canadian provinces announced that a four-year baccalaureate degree would become the educational requirement for entry-level practice. In January 2006, all basic nursing education programs in British Columbia offered baccalaureate education as the entry-level requirement for practice as a RN. Consequently, 100% of RNs now entering the workforce are prepared at the baccalaureate level. With the changes to entry-level educational requirements in addition to greater numbers of diploma-prepared nurses obtaining their baccalaureate degrees, there has been a gradual increase in the proportion of nurses within the workforce who have obtained a 3  baccalaureate degree as their highest level of education. For example, the percentage of RNs in British Columbia with a baccalaureate degree as their highest education in nursing has increased from 27% in 1999 to 41% in 2007. Approximately 33% of RNs employed in British Columbia hospitals hold a baccalaureate degree (Canadian Institute for Health Information, 2007b). The entry-level educational requirement for LPNs is a diploma or its equivalent. Educational programs for LPNs are offered in postsecondary institutions; however, at one time the training was primarily delivered in hospitals (Canadian Institute for Health Information, 2008). Similar to RNs, the age at which students graduate from a licensed practical nursing program has also increased from 23 years in 1980 to 31 years in 2005. It is anticipated that the educational diversification of the nursing workforce will continue as prospective members of each regulated profession have differing entry-level requirements. The educational diversification of the registered nursing workforce will also continue until the currently employed diploma prepared members of the profession retire. The relative mix of nurses with these varied educational backgrounds may affect the prevailing philosophy of nursing service, the level of professionalism, and the degree of conflict resulting from different perspectives concerning the provision of appropriate nursing care (Dussault et al., 1999). 1.1.3  Ethnicity/Race Another trend that warrants discussion is the increasing participation of ethnic/racial  minorities in the nursing workforce. The exact ethnic/racial composition of the Canadian nursing workforce is unknown because the regulatory bodies and other national nursing groups do not collect these data. The world-wide rates of increasing global migration and immigration have contributed to the changing ethnic/racial demographics of the overall Canadian workforce, including the nursing workforce. The 2001 census indicated that the proportion of foreign-born Canadians was at its highest in 70 years (Statistics Canada, 2005). The visible minority population represents approximately 13% of Canada’s total population (Canadian Council on Social Development, n.d.). In the Lower Mainland region of British Columbia, the number of people self-identifying with a visible minority group increased from 28% in 1986 to 37% in 2006 (Statistics Canada, 2006). In addition to increased migration and immigration rates of visible minority groups, the targeted recruitment of internationally educated nurses has also contributed to the 4  ethnic/racial heterogeneity of the nursing workforce. The nursing workforce in British Columbia, for example, in comparison with those of the other Canadian provinces, has a high proportion of internationally educated nurses. From 2003 to 2007, approximately 15% of the RNs in British Columbia were educated in other countries, compared with the Canadian rate of 8% (Canadian Institute for Health Information, 2008). The source countries of immigration to the British Columbian RN workforce are the Philippines (31%), the United Kingdom (17%), the United States (7%), Hong Kong (5%), India (6%), Poland (3%), France (2%), and other countries (29%) (Canadian Institute for Health Information, 2008). The frequency of immigration from these identified countries has been somewhat consistent during the past 5 years (Canadian Institute for Health Information, 2004). Given the absence of citizenship and immigration data for LPNs, the location of graduation is often used as an indicator of trends in immigration. For LPNs working in British Columbia, only 3.3% graduated from an international nursing program, which is slightly higher than the Canadian average of 1.9%. Of the international graduates, the majority are from the United Kingdom (31%), the Philippines (18%), the United States (12%), and India (5%) (Canadian Institute for Health Information, 2008). Demographic changes in ethnicity/race are expected to continue with the strategy of recruiting internationally educated nurses to deal with the current shortage of nurses (Baumann, Blythe, Kolotylo, & Underwood, 2005) and the need for more ethnically diverse nurses to care for the increasingly ethnically diverse population in Canada (Canadian Nurses Association & Canadian Federation of Nurses Unions, 2004). The evidence also suggests that a greater number of persons of ethnic minority status are enrolled in baccalaureate degree programs in British Columbia. A recent survey of nursing students in Canada indicated that 21% identified their ethnic background as “non-white” (1% First Nations, 3% black, 11% Asian, 1% Hispanic, and 5% other) (Bernard Hodes Group, 2006). In comparison, during the mid-1970s, ethnic minority groups represented about 1% of nursing students in Canada (Wong & Wong, 1980). Although national and provincial statistics of the ethnicity/race distributions of the nursing workforce are not available, given the changing immigration patterns of the general Canadian population, it seems reasonable to surmise that in the past 20 years there has been a gradual increase in the number of ethnic groups represented in the nursing workforce, particularly in large urban centres. At the same time, persons of ethnic minority status are, overall, still underrepresented in the nursing workforce relative to those who identify their ethnicity/race as “white.” 5  1.1.4  Work Values Another trend in the changing attributes of the nursing workforce is the changing  landscape of nurses’ work values. Accompanying the aforementioned demographic changes in the nursing workforce is likely to be variations in nurses’ attitudes toward their work and careers. The work values of nurses may be associated with their age, educational preparation, or ethnicity/race. For example, individuals of different ages and countries of origin may have experienced different approaches to nursing in their educational programs or through legislative or regulatory requirements; consequently, they may hold values different from those of their colleagues (McNeese-Smith & Crook, 2003). The available anecdotal evidence indicates that nurses of different generational cohorts manifest different values in their approach to their work and careers. In the nursing literature, most of the current debate about differences in values among nurses is based on generational cohort theory. Generational cohort theory proposes that different generations hold different work values, and such differences may result in conflict, tension, and poor workgroup outcomes (Hu, Herrick, & Hodgin, 2004; Santos & Cox, 2000; Swearingen & Liberman, 2004); however, there is very little empirical evidence to support this claim. 1.1.5  Summary Workforce diversity is not a phenomenon distinct to nursing; however, historically the  nursing workforce has been relatively homogenous. A review of current trends indicates that the nursing workforce is becoming increasingly diverse in terms of age, educational background, ethnicity/race, and possibly work values. Although the nursing workforce is becoming increasingly diverse, there remains underrepresentation of by those under the age of 40, of ethnic minorities, and of those with baccalaureate degrees. Moreover, there has been speculation that the work values of new graduates are incongruent with those of nurses who have been in the profession for some time. Despite the changing landscape of the attributes of the nursing workforce, there is a lack of research that has critically examined the consequences of this increasing diversification.  6  1.2  Why Study Diversity in the Nursing Workforce? Increasingly, scholars publishing in the general nursing literature support untested  assumptions about the necessity of a workforce that is diverse in culture as a means of improving the quality of care received by ethnically diverse clients (Adams & Price-Lea, 2004; Nugent, Childs, Jones, Cook, & Ravenell, 2002; Shea-Lewis, 2002). In the nursing leadership literature, diversity, particularly in ethnicity/race and gender, is positioned as key to ensuring the success of a workgroup (Matus, 2003; Shea-Lewis, 2002). Although in some instances this may be true, there is substantial evidence from the organizational behaviour literature suggesting that, in certain situations, diversity may be problematic especially if not managed (Riordan, 2000; Tsui & Gutek, 1999; Williams & O'Reilly III, 1998). Despite the growing body of literature calling for further diversification of the nursing workforce, some researchers have emphasized the detrimental effects of diversity attributed to different generations of nurses working together. As a result of the values, beliefs, and attitudes that each generational cohort of nurses brings to the workplace, several researchers have indicated that tension and conflict can result (Duchscher & Cowin, 2004; Hu et al., 2004; McNeese-Smith & Crook, 2003; Swearingen & Liberman, 2004), yet none have tested this hypothesis. Others claim that nurses of different age cohorts vary in their job satisfaction, organizational commitment, intentions to leave the workplace, and frequency of stress and burnout (Apostolidis & Polifroni, 2006; Blythe et al., 2008; LavoieTremblay et al., 2005; Widger et al., 2007); however, the implications of such differences also remain largely unexplored. Finally, whether these differences in values represent a cohort effect, age effect, or time effect has yet to be determined. Although diversity has been researched extensively in the field of organizational behaviour, there is a paucity of research attention paid to the nursing workforce. Of the few studies that examined diversity in the nursing workforce, one concluded that nurses who perceived themselves as different in age, gender, and ethnicity/race were less involved in workgroup discussions and decision making, and did not feel respected, included, or heard (Hobman, Bordia, & Gallois, 2004). Greater diversity in work values also significantly predicts nurses’ job dissatisfaction and lower intentions to stay in their jobs. Lower intentions by nurses to stay in their current jobs are also influenced by greater educational diversity (Gates, 2005).  7  The findings from a qualitative study of nursing teams in acute care hospitals established a strong connection between racial diversity2 and team difficulties with communication and conflict resolution (Dreachslin, Hunt, & Sprainer, 2000). Based on these preliminary findings about the consequences of diversity for nurses, further investigation is warranted to assist in the retention of nurses (within their organizations and profession), create a climate of equity, and improve the quality of nurses’ work-life and possibly client care. Nursing educators, researchers, and administrators agree that problems with retention, particularly of nurses new to the profession, may be a larger issue than is currently recognized. Research about the health and well-being of nurses recognizes job stress, burnout, and mental health issues as key indicators of healthful work environments and as contributors to the retention of nurses (Blythe et al., 2008; Lavoie-Tremblay et al., 2008; Lowe, 2006; Stordeur, D'Hoore, & Vandenberghe, 2001). Overall, 20% of healthcare workers in British Columbia perceived their mental health as “poor,” “fair,” or “good” (Lowe, 2006). Specifically, reports of fair or poor mental health are higher among nurses aged 35 to 44 (7%) than among those aged 55 or older (4%). Almost one-quarter of nurses (22%) in British Columbia report that mental health issues have made it difficult to handle their current workloads (Shields & Wilkins, 2006). Slightly more than one-third of nurses (34%) experience high job strain (Shields & Wilkins, 2006). Given the potential costs associated with nurses’ poor health and stress (e.g., absenteeism, lack of organizational commitment, turnover, and job dissatisfaction), research is necessary to explore the factors contributing to these outcomes, which if addressed, could lead to greater retention of nurses. As the largest cohort of nurses is nearing retirement at a time when a significantly smaller number of nurses exists to replace these individuals, various strategies must be developed to increase the supply and retention of nurses. Specific strategies may include recruiting ethnic minorities, extending the retirement age, attracting younger and second-career individuals, and offering flexible nursing educational programs to increase the number of graduates. Such strategies may result in further diversification of the nursing workforce (Gates, 2005).  2  When reporting published research findings, I use the various authors’ terms applied to the study of racial, cultural, or ethnic/racial differences. When discussing my hypothesis and findings, I use the term ethnic/racial diversity, which is further defined in Chapter 4.  8  Much of the research concerning the retention of nurses has focused on the demands of nurses’ work, characteristics of quality nursing work environments, and the structures that influence the quality of nurses’ work life. For the most part, the social conditions that contribute to undesirable work environments have received limited attention by nursing scholars. Of particular concern are interpersonal conflict among nurses (e.g., “horizontal violence” or “bullying”), unsupportive working relationships, and lack of respect evident in the workplace (Almost, 2006; Farrell, 2001; McKenna, Smith, Poole, & Coverdale, 2003; Stordeur et al., 2001). In Canada, both female and male nurses are exposed to hostility from or conflict with others within their workgroup (44% and 50%, respectively). Most nurses report that their coworkers are helpful in getting the work completed (95%); however, 47% do not feel supported by their coworkers (Shields & Wilkins, 2006). Given how much time people spend at work, being employed in an environment where individuals get along with one another is very important. To design retention strategies that address the root causes of unhealthy work environments and to improve the social aspects of the work environment, research is needed to identify the sources of conflict among nurses and the potentially detrimental effects of conflict on the psychological well-being of nurses. In light of the changing demographics of the nursing workforce, it seems reasonable to speculate that the degree of diversity between an individual and other workgroup members may give rise to interpersonal conflict, and that such conflict, in turn, is linked to burnout. Positioned another way, conflict in workgroups resulting from individuals’ dissimilarities in age, education, ethnicity/race, or values may be a source of stress for nurses, leading to burnout. Understanding the nature of the relationship between conflict and burnout in diverse workgroups is crucial to improving the organizational and professional retention, job satisfaction, and commitment of new graduates and the existing nursing workforce.  1.3  Perspectives of Diversity Broadly defined from a social psychological perspective (Williams & O'Reilly III,  1998), “diversity refers to differences between individuals on any attribute that may lead to the perception that another person is different from self” (van Knippenberg, De Dreu, & Homan, 2004, p. 1008). In principle, diversity refers to an almost infinite number of attributes or characteristics, yet diversity researchers tend to focus exclusively on demographic attributes. 9  Research about differences within workgroups has been approached from either a demographic diversity or relational diversity perspective. Demographic diversity refers to the degree to which an organizational or work unit is heterogeneous with respect to certain demographic attributes or personal characteristics (Jackson et al., 1991; McCain, O'Reilly III, & Pfeffer, 1983; O'Reilly III, Caldwell, & Barnett, 1989). From this perspective, diversity reflects the distributional or composition effects of, in most cases, demographic attributes on organizational units, such as workgroups. Thus, diversity represents a collective property of organizations (Alexander, Nuchols, Bloom, & Lee, 1995; Tsui, Egan, & O'Reilly III, 1992). The demographic diversity perspective, therefore, focuses on the relationship between the collective demographic profile and outcomes such as the work unit’s internal processes and performance as well as the group members’ behaviour and attitudes. At the group level of analysis, organizational diversity researchers focus primarily on demographic attributes such as age, race, gender, tenure, and level of education (Williams & O'Reilly III, 1998), and less on personal attributes such as status, knowledge, and behavioural style (Jackson, Stone, & Alvarez, 1993). In contrast, researchers examining diversity at the individual level of analysis approach diversity from a relational perspective, often termed relational diversity. Relational diversity refers to the degree of relative difference or dissimilarity between an individual and other workgroup members on common attributes (Riordan, 2000; Tsui et al., 1992; Tsui & Gutek, 1999). Relational diversity is similar to demographic diversity in that it measures differences in various characteristics, but dissimilar in that it measures an individual’s distance from other group members, rather than the collective range of diversity (Hobman, Bordia, & Gallois, 2003). The basic premise of the relational perspective is that the similitude of an individual’s attributes with those of all the other members of a particular workgroup has an impact on the individual’s experience within the organizational unit or workgroup (Tsui & Gutek, 1999). The relational perspective thus recognizes that the attributes of individuals may have different effects for each individual in a particular group. The makeup of the members of the group serves as a reference point in terms of the degree to which a particular attribute may be salient for an individual (Tsui & Gutek, 1999). Predominantly objective measures have been employed to study relational diversity; however, researchers have gained increasing support for the use of subjective measures for the study of the effects of an individual’s self-perceptions of 10  difference from others (Garcia-Prieto, Bellard, & Schneider, 2003; Riordan & Holliday Wayne, 2008; Williams, Parker, & Turner, 2007).  1.4  Research Purpose Focusing on the individual level of analysis, the aim of this research was to investigate  whether actual relational diversity and perceived relational diversity contribute to the burnout of nurses. Specifically, I explored the direct and indirect relationships between relational differences in age, education, ethnicity/race, and work values and nurses’ burnout (i.e., emotional exhaustion, cynicism, depersonalization, and diminished personal accomplishment). In consideration of the context in which diversity causes individual burnout, I sought to determine whether an individual’s dissimilarity from others in a workgroup is associated with his or her involvement in conflict, and if diversity is in turn associated with the experience of burnout. The types of interpersonal conflicts examined were task, process, and relationship conflict. Each type of conflict refers to interpersonal disagreements among workgroup members that arise from distinct sources (e.g., conflict generated from personality differences as opposed to differences about the content and goals of the work being performed) and may produce different outcomes. The study constructs are illustrated in Figure 1.1. Figure 1.1  Overview of the Postulated Model  Interpersonal Conflict  Relational Diversity  Burnout  11  1.5  Chapter Summary The purpose of this study was to examine whether, by producing conflict in the  workplace, diversity in age, education, ethnicity/race, or work values is directly or indirectly associated with the burnout of nurses employed in acute care hospitals. Although burnout in nursing has received a great deal of attention, little attention has been paid to the effects of the relative diversity that exists among workgroup members, and whether this relative diversity contributes to much of the burnout that occurs. Very limited information is available, even in the field of organizational behaviour, about the possible conflict experienced between individuals that arises from their relative differences. In the current study, diversity of age, education, ethnicity/race, and work values was examined to determine whether such diversity gives rise to conflict among nurses, and whether that conflict, in turn, is linked to burnout. Understanding the effects of diversity in a nursing context is of both theoretical and practical importance in improving the social aspects of the work environment and lessening the frequency and consequences of burnout (e.g., lower organizational commitment, job dissatisfaction, and higher turnover). In the following chapter, an overview of the literature pertaining to diversity, conflict, and burnout is provided in support of the conceptual models. Based on the literature review, Chapter 3 provides a concise overview of the postulated conceptual model, describing how diversity in the workplace was hypothesized to lead to burnout, and the hypotheses are identified and substantiated. Chapter 4 provides an overview of the sampling strategy, data collection procedures, operationalization of the study constructs, methods for data analysis, and ethical considerations. Details regarding the preparation of the data and the findings of the descriptive statistical procedures and confirmatory factor analyses are described in Chapter 5. Following data preparation and the confirmation of the measurement models of the study variables, the six structural models examining the direct and indirect effects were tested. In Chapter 6 the results of the hypothesis testing are presented. The results of the current study, relative to other evidence, are discussed in Chapter 7, along with their implications and further research directions.  12  2  LITERATURE REVIEW An increase in the levels of job stress and interpersonal conflict experienced by nurses  calls for action to understand the social aspects of their workplaces and their contributions to the quality of healthcare work environments. The aim of the research presented here was to clarify the effects of diversity within the nursing staffs of acute care hospital units and to determine whether, by producing interpersonal conflict in the workplace, such diversity is associated with nurses’ burnout. In this chapter, I provide a review of the literature pertaining to the direct relationship between relational diversity and burnout, and the mediating influence of interpersonal conflict. The literature review is organized around three main themes: relational diversity, burnout, and interpersonal conflict. The literature review regarding relational diversity focuses on the attributes identified as most dominant for, or salient to, nurses as criteria for creating social divisions among individuals in their workgroups: that is, age, education, ethnicity/race, and work values. These attributes were selected in consideration of the demographic trends of the current nursing workforce, the research evidence, and the context of nursing itself. Given the paucity of published research about diversity in nursing populations, I have included a review of literature interrelated to the constructs of interest. When no research was found to be specific to nursing populations, equivalent research conducted in other populations was considered. The majority of literature reviewed was drawn from a comprehensive search of the research literature published in the past two decades, encompassing the fields of nursing, healthcare, psychology, and organizational behaviour (i.e., CINAHL, Medline, PubMed, Web of Science, PsycINFO, ProQuest, SocINDEX, and Business Source Premier). Searches were limited to English-language manuscripts. Key words and phrases searched included diversity in the workplace, demography, diversity, relational diversity, cultural diversity, heterogeneity, differences, dissimilarity, work values, intergenerational relations, conflict, intragroup conflict, intraprofessional relations, interpersonal relations, bullying, horizontal violence, aggression, burnout, occupational stress, and mental health. Search strategies also included manual searches of textbooks, research journals, and journal articles that had been retrieved. Reference lists contained in scientific papers, unpublished dissertations, and books were reviewed as well. The internet was used to search for additional publications by leading authors in relational diversity, 13  intragroup conflict, and burnout. Non-research articles (e.g., editorials, letters to the editors, and opinion articles) included in the review consisted primarily of discussion pieces and provided a theoretical or social context for the variables of interest.  2.1  The Relational Approach to Diversity Studies exploring relational diversity have been marked by a lack of consensus with  regard to the manner in which diversity is defined, operationalized, and categorized. In this section, I provide some conceptual clarity about the use of relational diversity in the research presented here and then provide a review of the literature as it pertains to nursing workgroups. 2.1.1  Defining Relational Diversity The focus of diversity research, in the field of organizational behaviour, is to study the  effects of human differences in organizations above and beyond simple demography. Much of the current thinking in the empirical and theoretical literature about diversity has been influenced by Pfeffer’s (1983) seminal work introducing organizational demography (also referred to as demographic diversity). His research focused on the relationship between the distribution of specific demographic characteristics (e.g., age, race, ethnicity/race, gender, and education) within an organizational unit (e.g., organization, department, or workgroup) and turnover. In contrast to earlier work that focused on diversity at the group level of analysis, researchers have urged others to examine the consequences of diversity on individual outcomes (Riordan, 2000; Tsui & Gutek, 1999). To do so, a relational approach is required. The relational diversity perspective examines the degree of relative difference, or dissimilarity,3 between an individual and other workgroup members on common demographic and nondemographic attributes (Clark, 2001; Hobman, Bordia, & Gallois, 2003; Tsui, Egan, & O'Reilly III, 1992). Tsui and Gutek (1999) noted that “the basic premise of the relational approach is that the relationship of an individual’s own demographic attributes to that of all the other members in a particular unit will have an impact on the individual’s experience in that unit” (p. 23). Thus, the relational approach recognizes that the attributes of individuals may have different effects for each individual in that  3  The terms diversity and dissimilarity are often used interchangeably to reference relational differences between an individual and other workgroup members.  14  group. The makeup of the members of the group serves as a reference point in terms of the degree to which a particular attribute may be salient for an individual (Tsui & Gutek, 1999). 2.1.2  Theoretical Foundations Relational diversity effects have been explained in terms of two complementary  theories: social identity theory and similarity–attraction theory. A brief introduction of these theories is provided in this chapter and further elaborated upon in Chapter 3 with respect to the postulated conceptual model and hypotheses. 2.1.2.1  Social Identity Theory According to social identity theory, individuals categorize themselves and others as a  means of ordering the social environment and locating themselves and others within it (Ashforth, 2001). Individuals tend to perceive themselves and others as either belonging to various categories that share some common identity, or as being members of different categories (Ashforth, 2001; Northcraft, Polzer, Neale, & Kramer, 1995; Tsui, Xin, & Egan, 1995). Similarities and differences are thus employed as a basis for categorizing oneself and others into groups to provide meaningful distinctions between people or subgroups of people (Ashforth, 2001). Through a process of social comparison, individuals define a social category or group according to the most widely shared attributes of category members, specific persons who exemplify the category, or both (Ashforth, 2001); thus, individuals’ social identities are derived from the process of self-categorization and attaching value to particular social categories. Relative to members of other social categories, these categories permit individuals to define themselves in terms of a social identity (Riordan, 2000). Moreover, all social categorizations implicitly involve a distinction between in-groups and out-groups (Ashforth, 2001). 2.1.2.2  Similarity–Attraction Theory The social identity perspective explains identity based on group membership (which  may or may not involve social interaction), whereas the similarity–attraction perspective explains social identity based on attitudes or personal characteristics (Riordan, 2000; Tsui et al., 1992). The initial attraction between oneself and others is based on individuals’ perceptions of similarity about the characteristics and attitudes held by themselves and other individuals. Information obtained about individuals is initially based on visible demographic attributes, which 15  leads to inferences about similarities in values, beliefs, and attitudes (Chuang, Church, & Zikic, 2004; Tsui et al., 1995). Through social interaction, these initial perceptions of others change when detailed information about less visible or nondemographic attributes (e.g., values, beliefs, attitude, and knowledge) is obtained over time and from differing contexts (Harrison, Price, & Bell, 1998). 2.1.3  Operational Definitions of Relational Diversity Researchers in the field of organizational behaviour operationalize relational diversity  from either an objective or subjective perspective. The predominant use of objective, or actual, measures (e.g., the Euclidean distance measure, the use of interaction terms, and polynomial regression) to study relational diversity has been largely influenced by early demography researchers. Objective measures have been used to capture the actual dissimilarity that exists within a workgroup, and that consciously or unconsciously affects individuals’ experiences, attitudes, and behaviour toward others (Riordan & Holliday Wayne, 2008). Other strengths of the use of objective measures are a greater degree of control in terms of what is being measured, a reduction in the bias inherently associated with individuals’ ability to accurately report the degree of diversity, the opportunity to test for nonsymmetrical effects, and the ability to make casual inferences (Riordan & Holliday Wayne, 2008). Despite the widespread use of objective measures to study diversity, the significance of a given attribute is dependent upon the outcome of interest and the population under study. For example, in some workgroups, greater age diversity is associated with work outcomes such as greater intentions to leave one’s job and poorer working relationships among members of the workgroup (Chattopadhyay, 1999; Gonzalez, 2001; Riordan & Holliday Wayne, 2008; Tsui et al., 1992). At the same time, several other researchers have found that greater actual age diversity is not associated with coworker satisfaction, job satisfaction or organizational commitment (Clark, 2001; Liao, Joshi, & Chuang, 2004). In non-nursing populations, actual educational diversity has not been associated with various individual outcomes except for actual turnover (Jackson et al., 1991; Liao, Chuang, & Joshi, 2008). Despite some initial support for the asymmetrical effects for actual ethnic/racial diversity (Tsui et al., 1992), more recent researchers have not identified a negative effect on individuals’ attitudes and behaviour (Clark, 2001; Keller, 2005; Riordan & Holliday Wayne, 2008). These discrepancies may be attributed to 16  methodological shortcomings, such as the use of a variety of sample populations, the sample size, and the nature of the referent groups. As well, objective measures of diversity are less congruent with the theoretical underpinnings of relational diversity. Garcia-Prieto et al. (2003) argued that assessing relational diversity objectively does not consider the dynamic nature of diversity, because individual differences within workgroups change with collegial interactions over time. Viewing diversity as nominal discrete categories, rather than continuous and interdependent, assumes that all individuals are the same within, for example, an ethnic group, and that all individuals share the same identity within a particular social category (Garcia-Prieto et al., 2003). Correspondingly, Tsui and Gutek (1999) acknowledged that approaching relational diversity from an actual perspective is problematic when it assumes that individuals attach the same value to specific attributes. Given the theoretical and methodological limitations of objective measures of relational diversity, other researchers have introduced subjective measures to assess the effects of an individual’s perceptions of his or her differences from others (Clark, 2001; Hobman et al., 2003; Liao et al., 2008; Riordan & Holliday Wayne, 2008; Williams, Parker, & Turner, 2007). The use of subjective, or perceived, measures of relational diversity “is based on the theoretical assumption that individuals assign their own psychological meaning to the differences in demographic characteristics between themselves and others” (Riordan & Holliday Wayne, 2008, p. 571). Individuals’ perceptions of being different often differ from objective reality, and objective differences do not necessarily result in perceptions of dissimilarity (Van der Vegt & Van de Vliert, 2005). Thus, consideration of the perceptual approach to measuring diversity takes into account individuals’ subjective experience of “being different,” recognizes the potential salience of certain demographic and nondemographic attributes to individuals, and considers whether individuals differ in their perceptions of, and reaction to, objective reality (Clark, 2001; Garcia-Prieto et al., 2003; Riordan & Holliday Wayne, 2008). This approach to operationalizing diversity draws on the importance of individuals self-determining which category they subjectively feel they belong to, rather than the researcher determining the objective categories into which individuals fall (Garcia-Prieto et al., 2003). Furthermore, a perceptual approach recognizes the social construction of diversity, in that an attribute defined as  17  very important in one context may be defined as less important in another (Garcia-Prieto et al., 2003). Despite the growing interest in subjective interpretations of individual dissimilarity, research using perceived measures is limited. In the field of organizational behaviour, perceived diversity is negatively associated with helping behaviour, interpersonal conflict, workgroup involvement, turnover intentions, job attitude, work withdrawal, organizational commitment, and collegial interactions among members of workgroups (Hobman et al., 2003; Hobman, Bordia, & Gallois, 2004; Liao et al., 2008; Riordan, 1997; Riordan & Holliday Wayne, 2008; Van der Vegt & Van de Vliert, 2005; Williams et al., 2007). Another key finding from this body of research is that not all diversity attributes are equivalent in terms of their outcomes. For example, Hobman et al. (2003) and Liao et al. (2008) found that perceived similarity in values among colleagues is significantly associated with less conflict, greater workgroup involvement, and a positive job attitude, whereas perceived differences in visible (i.e., age, gender, and ethnicity/race) and information diversity (i.e., education) were not. Kirchmeyer (1995) found no support for a relationship between perceived individual diversity and commitment and turnover intentions. Other studies, however, have indicated that perceived diversity in age, gender, and ethnicity/race is significantly predictive of outcomes such as workgroup identification, organizational commitment, and turnover (Hobman et al., 2004; Riordan & Holliday Wayne, 2008). At this time, measures of perceived diversity may explain some of the consequences of individual dissimilarity; however, no conclusive results have been achieved. Since most studies do not concurrently test actual and perceived relational diversity, a limitation of the aforementioned research is the lack of comparison of the amount of variance in the outcome variables explained by measures of actual and perceived diversity. For example, using a combination of both actual and perceived measures to study various attributes, Jehn et al. (1999) and Liao et al. (2008) reported that perceived values diversity accounts for a significant portion of the variance in task, relationship, and process conflict, in comparison with actual measures of age, gender, and educational diversity. In my review of the literature, only two studies were found that used both an actual and perceived approach to study the attributes of interest (Clark, 2001; Riordan & Holliday Wayne, 2008). The goals of these researchers were to determine whether subjective measures account for a greater percentage of the variance than the 18  more objective measures of relational diversity, and to explore the association between perceived and actual diversity. In a sample of employees working in the public sector, Clark (2001) tested the hypothesis that perceived diversity in age accounts for the variance in various job-related attitudinal outcomes beyond that explained by actual diversity. He revealed that actual age diversity did not predict job satisfaction, turnover intentions, affective commitment, or satisfaction with colleagues; perceived age diversity, however, was a significant predictor of satisfaction with colleagues. Similarly, Riordan and Wayne (2008) found that perceived diversity in age and ethnicity/race accounted for a significant percentage of the variance in individuals’ lack of organizational commitment and limited identification with their workgroup, whereas actual age and ethnic/racial diversity did not. Because research exploring subjective measures of diversity is in its infancy, no conclusive statements can be made that perceived diversity is more strongly predictive than actual diversity. 2.1.4  An Overview of the Diversity Attributes Studied in Previous Research What is clear from the empirical and theoretical literature is that diversity is not a  unitary construct. A major difficulty in empirically assessing the impact of diversity on individual and workgroup behaviour is that the number of attributes being studied has broadened significantly during the past 25 years. Comprehensive reviews of discussions about the types or classifications of diversity attributes have been published elsewhere (Harrison et al., 1998; McGrath, Berdahl, & Arrow, 1995; Pelled, 1996a). In general, the diversity attributes researched to date range from discrete demographic categories (e.g., age, gender, race, job tenure, work status, and education) to more broad and varied nondemographic attributes (e.g., task-related capabilities; values, beliefs, and attitudes; personality and cognitive/behavioural styles; and functional background) (Clark, 2001; Garcia-Prieto et al., 2003; Hobman et al., 2004; McGrath et al., 1995; Pfeffer, 1983; Sacco & Schmitt, 2005; Williams & O'Reilly III, 1998). Most diversity researchers focus exclusively on a narrow range of demographic characteristics that are easily observable (e.g., age, gender, and ethnicity/race) without considering other attributes that are not readily apparent (e.g., education) or nondemographic (e.g., values and personality) yet may be most salient to workgroup members (Tsui & Gutek, 1999). In keeping with the theory underpinning relational diversity, researchers have moved toward greater recognition of the importance of attributes that are not readily apparent, especially when investigating perceived 19  dissimilarity. Although demographic attributes may initially be used as part of the categorization and attraction process, those that are less readily apparent are important for shaping social identities within workgroups. Examining a full range of attributes, beyond demographic characteristics, is necessary to capture a wider range of human differences and to consider the complex configuration of such differences (Clark, Ostroff, & Atwater, 2002). Although researchers have sought to provide conceptual clarification regarding the types of diversity attributes, the lack of agreement in defining such attributes, coupled with the incongruence between conceptual and operational definitions of these variables, have added to the complexity of understanding the impacts of diversity. Using a discrete categorical approach, researchers treat each diversity variable as a distinct theoretical concept, based on the argument that different types of diversity may produce different outcomes. Another approach taken in studying diversity has been to treat diversity broadly by grouping diversity attributes (e.g., social category diversity includes age, gender and race attributes; informational diversity includes education, work experience/functional background, and expertise) (Jehn et al., 1999; Pelled, 1996a; Webber & Donahue, 2001) or creating a total composite score of diversity (Chatman, Polzer, Barsade, & Neale, 1998; Chatman & Spataro, 2005). Arguments exist on both sides in support of these approaches to operationalizing diversity. On the one hand, using a broad diversity variable may allow hypotheses or propositions to have greater explanatory power. Conversely, grouping the different types of diversity together may not only increase the error and thus deflate the correlations, but may also cause researchers to overlook important distinctions among them and make inaccurate predictions (Pelled, 1996a; Riordan & Holliday Wayne, 2008). 2.1.5  Empirical Literature Concerning Relational Diversity in Nursing Workgroups Although relational diversity has been researched extensively in the field of  organizational behaviour, actual relational diversity has been the subject of one study sampling from nursing populations (Gates, 2005), while two studies have explored perceived diversity in nursing workgroups (Gates, 2005; Hobman et al., 2004). As part of a larger national study in the United States, Gates (2005) obtained a population sample of 1508 nurses from 248 acute care hospital units. Data were collected three times during a six-month period to examine the direct effect of diversity on nurses’ job satisfaction and intention to stay. He reported that greater actual educational diversity predicted nurses’ intentions to leave their jobs whereas actual age and 20  ethnic/racial diversity did not. He also concluded that greater actual age and educational diversity were not predictive of nurses’ dissatisfaction with their jobs. Counter to his hypothesis, Gates (2005) found that greater actual ethnic/racial diversity predicted greater job satisfaction among nurses. In other words, actual ethnic/racial diversity present in nursing workgroups enhanced job satisfaction. Subsequent analyses revealed that actual ethnic/racial diversity was predictive of greater job satisfaction in older nurses (over 48 years of age) but not younger nurses (under 34 years of age) (Gates, 2005). Given the absence of other studies exploring actual diversity in nursing populations, this study indicates that educational diversity may be important in predicting turnover intentions and that ethnic/racial diversity may be associated with greater job satisfaction. In nursing workgroups, two studies have examined the perceived approach to relational diversity. Hobman et al.’s (2004) study of 119 nurses working in acute care involved two surveys, with 4 weeks between administration, to examine the direct effect of perceived diversity on nurses’ involvement with their workgroup. The first survey contained items regarding the dependent and independent variables, whereas the second survey measured workgroup involvement in conjunction with other variables that were part of a larger study. At different times, she found that nurses’ involvement with their workgroup was associated with perceived visible diversity (i.e., age, gender, and ethnicity/race) and perceived informational diversity (i.e., professional background, work experience, and education) but not perceived values diversity (i.e., work values and motivations). At Time 1 of data collection, regression analyses confirmed that nurses who perceived themselves as different in visible and informational attributes were less involved in workgroup discussions and decision making, and did not feel respected, included, or heard (Hobman et al., 2004). Only perceived visible dissimilarity was a predictor of individuals’ workgroup involvement at Time 2. No significant relationship was found between perceived value diversity and workgroup involvement despite a statistically significant bivariate correlation (Hobman et al., 2004). Conversely, another study (Gates, 2005) showed that perceived differences in work values, in comparison with other demographic diversity attributes, are predictive of both job satisfaction and intention to stay. Specifically, Gates (2005) found that perceived values diversity negatively predicted both job satisfaction and the intention to stay in both older nurses (over 48 years of age) and younger 21  nurses (under 34 years of age). The differences between these two studies in the significance of perceived values diversity may be attributed to studying different outcomes, in addition to methodological issues such as workforce restructuring prior to commencement of the study, and different measures used to assess perceived values diversity. Although not studying relational diversity per se, in a small sample (N = 56) of nurses in Norway, Verplanken (2004) found that values congruence between individuals and their unit was positively associated with job satisfaction and to a lesser extent “ward” attitude. Specifically, greater values congruence with regard to human relations (e.g., empowerment of employees to act, participation and open discussion, and trust and openness) was predictive of a better attitude toward the nursing unit and job satisfaction. Despite the examination of various individual outcomes and the lack of replication studies, collectively these studies suggest that perceptions of differences for observable attributes such as age and ethnicity/race may be associated with negative interactions among members of a workgroup and that perceived differences in work values may influence individuals’ attitudes toward work. External from the diversity research in the field of organizational behaviour, the findings from a qualitative study of nursing teams in acute care hospitals established a connection between racial diversity and team difficulties with communication and conflict resolution (Dreachslin, Hunt, & Sprainer, 2000). Communication processes characterized by conflict and misunderstandings were attributed specifically to racial differences. Members of nursing care teams (i.e., RNs, patient-care technicians, and support associates such as housekeeping and dietary personnel) were said to “see the team’s interactions from different perspectives or vantage points that are strongly influenced by each team member’s racial identity and how he or she experiences that racial identity” (Dreachslin et al., 2000, p. 1408). The differing perspectives provided a framework within which team conflict and miscommunication were interpreted and experienced. Dreachslin et al.’s (2000) study highlighted the importance of approaching diversity from a relational perspective, in that two individuals with different demographic profiles (e.g., race) in the same workgroup may have different experiences within and perceptions of the group. The current debate about differences in values among nurses is based on generational cohort theory. The research regarding value differences has focused on identifying what values 22  the various generational cohorts hold (McNeese-Smith & Crook, 2003) and whether work-related attitudes and behaviour vary among the different generations of nurses (Apostolidis & Polifroni, 2006; Blythe et al., 2008; Santos et al., 2003; Santos & Cox, 2000; Santos & Cox, 2002; Shader, Broome, Broome, West, & Nash, 2001; Stuenkel, Cohen, & de la Cuesta, 2005; Widger et al., 2007). For example, McNeese-Smith and Crook (2003) sought to determine value differences among nurses depending on their age, educational background, and ethnicity/race. They found some statistically significant differences in values among educational levels and ethnicities. Differences in age were positively associated with aesthetics, which means that older nurses scored higher on this value compared with younger nurses. The older nurses, in comparison with the younger nurses, scored lower on three values: economic returns, prestige, and variety. McNeese-Smith and Crook (2003) went on to report that differences in values between the generational cohorts were statistically significant for 2 of the 15 values measured (i.e., variety and economic returns). Investigating whether work-related attitudes and behaviour vary among the generational cohorts, several researchers have identified statistically significant differences in job satisfaction, burnout, job stress, organizational behaviour and intention to leave (Blythe et al., 2008; Lavoie-Tremblay et al., 2008; Santos et al., 2003; Widger et al., 2007). Others such as Hu et al. (2004), however, found no statistically significant differences in the communication styles and job attitudes of the generational cohorts. The mixed results of these studies addressing generational differences among nurses are attributable in part to methodological issues with sampling, measurement, and study design. Studies conducted in the nursing field often suffer from small sample sizes for each generational or age cohort. Moreover, the existing measures used to assess work values are insufficient in that they do not reflect the work-related values or work ethic for each generation. Often researchers conflate generational effects and age effects with respect to various work-related attitudes and behaviour. Given that these studies are of a cross-sectional design, whether these value differences represent a generational cohort effect, age effect, or time effect has yet to be determined. 2.1.6  Summary The majority of the relational diversity research assesses actual differences; however,  the theory underpinning relational diversity within workgroups (e.g., social identity theory and similarity–attraction theory) refers to individuals’ perceptions of similarity and dissimilarity as 23  the major cause of differences in work-related outcomes (Riordan, 2000). Including both objective and subjective measures of relational diversity is important to tap into different aspects of the construct (Riordan, 1997) by capturing the degree of relative diversity as well as the individual’s perception of being different. Research inclusive of both approaches also provides a more comprehensive picture of the complexities of diversity on the attitudes and behaviour of individuals (Riordan, 2000). At the same time, the attributes selected for examination are often unjustified by researchers in terms of their importance to the population being studied and the social context of the referent group. In most instances, more observable attributes are selected without consideration of the context of the workgroups. Despite the examination of various individual outcomes, the absence of replication studies, and the inconsistencies in examining attributes as discrete categories, some emerging findings can be related to the detrimental consequences of diversity in the workplace. Greater perceived age diversity may be associated with negative interactions among members of nursing workgroups, and perceived work-values diversity may influence nurses’ attitude toward work. There also is some indication that actual and perceived ethnic/racial diversity may be associated with greater job dissatisfaction and poorer interpersonal relationships. Finally, actual educational diversity may be important in predicting turnover intentions among nurses. More sophisticated methods of data analysis, beyond multiple linear regression, may be of assistance in the determination of the relative contributions of actual and perceived relational diversity on individual outcomes within nursing workgroups (Riordan & Holliday Wayne, 2008). The following section provides an overview of burnout (i.e., definition, antecedents, consequences, and prevalence) and a review of the literature with regard to burnout as a potential outcome of the diversification of the nursing workforce. The final section illuminates the process by which diversity may lead to burnout, namely interpersonal conflict.  2.2  Burnout as an Outcome of Relational Diversity In organizational behaviour, researchers examining relational diversity at the  individual level of analysis typically examine the relationships between diversity and select outcomes such as absenteeism, commitment, attachment, work performance, satisfaction, and turnover. The outcome variable of interest in the current study was burnout.  24  2.2.1  Defining Burnout Burnout is conceptualized as a psychological syndrome attributed to chronic, everyday  interpersonal stressors and emotional strain experienced on the job. It is one type of job stress arising predominantly from emotionally demanding social interactions between human service providers and their recipients (Cordes & Dougherty, 1993; Duquette, Kerouac, Sandhu, & Beaudet, 1994; Maslach, Schaufeli, & Leiter, 2001). Maslach (1982) conceptualized burnout as a social phenomenon, rather than an individual work-related and situation-specific phenomenon. Four aspects4 of burnout have been identified: emotional exhaustion, cynicism, depersonalization, and diminished personal accomplishment (Maslach et al., 2001). Emotional exhaustion refers to feelings of being emotionally overextended and having one’s emotional resources depleted. Within the human services, a negative, callous, or distant attitude to other people exemplifies depersonalization. Outside of the human services, when individuals feel discouraged and exhausted, they often mentally distance themselves by developing an indifferent attitude toward their work or employer instead of other people, which is referred to as cynicism. Thus, the target of the mental distancing differs. For human service providers, the targets are the recipients of their services; for employees who work with objects or information, the target is the work itself (e.g., the organization at large, the work environment, and people at the job such as other employees) (Maslach et al., 2001). The fourth aspect of burnout, feelings of diminished personal accomplishment, refers to a tendency to evaluate oneself negatively, particularly with regard to feelings of competence and achievement in one’s work with clients (also referred to as personal inefficacy). Accordingly, individuals may feel unhappy about their accomplishments at work and dissatisfied about themselves (Maslach, Jackson, & Leiter, 1996; Salanova et al., 2005). The four aspects of burnout range from low to high degrees of experienced feelings.  4  Traditionally, burnout has been conceptualized as having three dimensions. Based on the results of confirmatory factor analyses, recent disagreements have emerged in the literature about burnout and whether it has a three- or four-factor structure (Maslach et al., 2001; Salanova et al., 2005). The concepts of particular concern in such debates are whether cynicism and depersonalization represent two separate and distinct forms of mental distancing. Within the human services, depersonalization is seen as an attempt to put distance between oneself and one’s clients; outside the human services, when people are exhausted and discouraged, they distance themselves mentally from their work by developing an indifferent or cynical attitude toward their employer or the system (Salanova et al., 2005). For conceptual clarity in understanding the impact of diversity on burnout, I refer to burnout as having four dimensions. It is reasonable to assume that nurses experience one and not the other; that is, they may not depersonalize their patients but may become cynical about the healthcare system in general, or their employer in particular, which may contribute to burnout.  25  Within this study, for ease of communication, the umbrella term “burnout” is used when discussing the phenomenon more generally; however, when discussing the hypotheses and results of the current study, and those found in the literature, the particular aspect of burnout is specified. 2.2.2  Consequences of Burnout As a form of job stress, burnout has been linked to various types of negative individual  and workplace outcomes (Maslach et al., 2001), including both mental stress-related health problems, outcomes such as feelings of poor self-esteem, depression, irritability, helplessness, and anxiety, as well as physical health problems, such as fatigue, insomnia, headaches, and gastrointestinal disturbances (Maslach et al., 2001). The deleterious effects of burnout may also include changes in the nature or frequency of interactions with clients, colleagues, and family members. At the work level, burnout has been linked to attitudinal changes, such as reduced organizational commitment and job dissatisfaction and behavioural consequences, such as poor job performance, absenteeism, and professional turnover (Cordes & Dougherty, 1993). Burnout is a costly concern – not only does the individual’s health and well-being suffer, but the people with whom individuals experiencing burnout come in contact, as well as the organization and the immediate workgroup, bear the cost of this work-related syndrome (Cordes & Dougherty, 1993; Halbesleben & Buckley, 2004; Maslach et al., 2001). Additionally, burnout can have detrimental effects on the quality of care received by clients (Maslach et al., 1996). 2.2.3  Antecedents of Burnout As previously mentioned, burnout is specific to the work context. Thus, a consistent  focus of burnout research during the past 25 years has been the impact of situational sources of work-related, interpersonal stress. The three main sources are: (a) job characteristics (e.g., role conflict and role ambiguity, lack of job resources, role overload, and job demands), (b) occupational characteristics (e.g., care-giving occupations and interpersonal relations with clients), and (c) personal characteristics (e.g., demographics, social support, personal expectations, personality, job expectations, and career progress) (Cordes & Dougherty, 1993). Based on these conceptions of burnout, emotional exhaustion, depersonalization, and cynicism tend to occur as a result of work overload and social conflict, whereas a sense of personal inefficacy arises from a lack of resources necessary for job completion (Cordes & Dougherty, 26  1993; Maslach et al., 2001). Although burnout is conceptualized as a function of the situation, rather than the individual, limited emphasis has been placed on the influence of organizational characteristics (Maslach, 2003). Although a feature of the work context is the provider–client relationship, it is also embedded in layers of the organizational context, for example, the quality of provider interactions with their colleagues at work. More recent work about job engagement, the antithesis of burnout, recognizes the social context of the workplace and other organizational characteristics as important predictors of burnout (Maslach & Leiter, 2008). From this perspective, recognition is given to the implicit influence of organizational processes and structures that shape the work environment, and the social relationships (emotional and cognitive) that people develop in their places of work. Coupled with the stressors associated with the shortage of nurses, organizational changes such as workload and restructuring have greatly influenced the environments in which nurses work, particularly a reduction in the quality of social interaction at work (i.e., the social climate). These less-convivial work environments can create a problem for human service providers; negative collegial interactions are thought to be a source of stress experienced by and affecting human service professionals. Thus, the range of situational factors as antecedents of burnout has been expanded to include organizational characteristics in addition to work and occupational characteristics (Maslach et al., 2001). This more recent theoretical framework of the burnout–engagement continuum incorporates both individual and situational factors occurring simultaneously. This perspective views the person and the environment as interdependent entities in an attempt to explain behaviour by examining the interaction between the two entities (Maslach et al., 2001). Based on research about the organizational factors related to burnout, Maslach et al. proposed a comprehensive model of burnout focusing “on the degree of match, or mismatch, between the person and six domains of his or her job environment” (p. 413): workload, control, reward, community, fairness, and values (Maslach & Leiter, 2008). The greater the chronic mismatch between people and their work settings, in terms of some or all of these six domains, the greater the likelihood of burnout, particularly emotional exhaustion and cynicism (Maslach & Leiter, 2008; Maslach et al., 2001). Conversely, greater congruency represents a high degree of employee engagement. “Despite the close interrelatedness of these six areas, each area brings a 27  distinct perspective to the interactions of people with their work settings” (Maslach et al., 2001, p. 414). This approach is fairly new; however, research about the six areas of the work environment can function together in defining a framework encompassing the major organizational antecedents of burnout (Maslach et al., 2001). Of these six, the two areas relevant to the current research are community and values. 2.2.4  Prevalence of Burnout in Canadian Nurses Several national studies have indicated that the prevalence of high levels of perceived  stress and burnout in the Canadian labour force has increased in the past decade (Duxbury & Higgins, 2003; Jobquality.ca, 2009a; Lowe, 2006). Healthcare providers, including nurses, are at high risk for burnout and work stress. Healthcare workers in British Columbia perceive themselves to have higher levels of work stress than workers in other occupations (Lowe, 2006). More than one third (39%) of healthcare workers report that most days at work are “quite a bit” or “extremely” stressful (Lowe, 2006). The prevalence of burnout among Canadian nurses is typically reported within the norms for medical personnel, as established by Maslach et al. (1996). Although burnout among nurses is within the established norms it is nonetheless striking that in a recent study of registered nurses working in acute care hospitals in Ontario, almost three quarters (69%) reported a moderate to high degree of emotional exhaustion and one half (49%) reported a moderate to high degree of diminished personal accomplishment. Less significantly, 31% of nurses reported a high to moderate degree of depersonalization (Widger et al., 2007). Furthermore, among nurses, the prevalence of burnout is notably higher among 20 to 39 year olds in comparison with their older counterparts. Specifically, those under 40 years of age reported greater emotional exhaustion and depersonalization, whereas nurses 50 years or older reported higher levels of diminished personal accomplishment5 (Blythe et al., 2008). Despite the differences among various age groups, all nurses experience burnout to some degree.  5  With the exception of personal accomplishment, these findings are consistent with other researchers’ reports that burnout is most common in younger individuals who typically have less work experience, in unmarried individuals, and in individuals with higher levels of education (Cordes & Dougherty, 1993; Duquette et al., 1994; Maslach et al., 2001).  28  2.2.5  Empirical Literature Concerning Burnout and the Relational Diversity within Nursing Workgroups No previous published research has investigated the relationship between actual  relational diversity and burnout in nursing workgroups; however, in the field of organizational behaviour, two studies have examined this relationship. In a study of 135 university faculty members, Siegall and McDonald (2004) found a strong association between perceived work-related values and burnout. Specifically, perceived value similarity (i.e., holding work-related values that were similar or congruent with the organization’s values) was positively correlated with personal accomplishment and negatively correlated with emotional exhaustion and depersonalization. In other words, individuals experience more burnout when they perceive their values to be dissimilar from the organization’s. Another study, conducted by Wesolowski and Mossholder (1997), found that greater actual diversity in race within the superior–subordinate dyad was positively associated with subordinates’ burnout; however, actual age diversity was not associated with burnout. Given the limited availability of research examining the linkages between diversity and burnout, no firm conclusions can be drawn as to which attributes are of potential significance to nursing workgroups. Available for comparison are numerous studies that have separately examined actual and perceived relational diversity on individual level outcomes such as absenteeism, work involvement, organizational commitment, job satisfaction, and turnover (Riordan, 2000; Tsui & Gutek, 1999). These attitudinal and behavioural outcomes are often cited as some of the consequences of burnout (Cordes & Dougherty, 1993; Maslach et al., 2001). In the following sections, I discuss the research related to the effects of actual and perceived diversity on outcomes interrelated with burnout. 2.2.6  Empirical Literature about the Relationships between Relational Diversity and Outcomes Interrelated with Burnout As seen in Appendix A (see Tables A1 and A2), the literature related to actual and  perceived diversity was reviewed regarding the attributes of interest for the current study: age, education, ethnic/racial, and work values. Accordingly, the following review is organized by each diversity attribute of interest, speaking first about actual diversity and next about perceived diversity. The outcomes discussed in this literature review are interrelated with the burnout construct; it stands to reason that if diversity is associated with some of these work-related outcomes, then they may also be linked with burnout. 29  Before presenting this body of research, it is worth voicing a word of caution about some methodological issues. Much of what is known is based on samples drawn from a variety of populations (e.g., librarians, manufacturers, university students completing course projects, coaches, banks, restaurants, and salons). Many of the studies are exploratory in nature and suffer from having small sample sizes. Although a few are longitudinal, for the most part the studies are cross-sectional. The referent group is of great importance in influencing the results obtained in relational diversity studies; the referent groups in these studies, however, vary from dyads to entire workgroups (e.g., supervisor–employee relationships, random samples of members of a workgroup, and entire workgroups) and differ in size, level of interaction, and permanence of the workgroup (Riordan, 2000). A limited number of researchers have sought to replicate the findings of earlier studies. Furthermore, the generalizability of some of the findings may be limited because of the use of students as research subjects in artificially constructed workgroup situations and having been set in various countries. Multiple linear regression analysis was primarily used for data analysis in these studies, which may be insufficient to explore adequately the complexities of the diversity attributes that co-occur in the workplace. 2.2.6.1  Actual and Perceived Age Diversity Several researchers have examined actual age diversity in workgroups and have  reported mixed findings, depending on the outcome of interest. For example, several researchers did not find a significant relationship between actual age diversity and work withdrawal behaviour, overall job attitude (i.e., job satisfaction and affective commitment) (Liao et al., 2008), job satisfaction (Clark, 2001), or organizational commitment (Clark, 2001; Gonzalez, 2001; Liao et al., 2004; Riordan & Holliday Wayne, 2008). Both Tsui et al. (1992) and Riordan et al. (2008) reported that the more dissimilar an individual was in age to other members of the workgroup, the greater was the individual’s intentions to leave the organization under study. Counter to this finding, others have reported that age differences are not predictive of actual job turnover (Jackson et al., 1991) or greater intentions to leave (Clark, 2001; Gonzalez, 2001). Although actual age diversity does not seem to have strong relationships with individual work-related outcomes (e.g., job satisfaction) in some instances it does influence individuals’ attitudes and behaviour toward others within the workgroup. For example, Chattopadhyay (1999) reported that actual age diversity was associated with peer relationships 30  within a workgroup. Specifically, the peer relations of older employees were lower when age diversity was found to be greater, and the peer relations of younger employees were better when age diversity was greater. Younger employees were more likely to report better peer relations when there was greater age diversity in the group; however, older employees reported poorer peer relations when they had more dissimilarity with their peers (Chattopadhyay, 1999). Riordon and Wayne (2008) concluded that actual age diversity is predictive of workgroup identification in that the more diversity there is, the lower the identification and attraction among members of a the workgroup. Counter to their hypothesis, Liao et al. (2004) found that when individuals were dissimilar in age to other members of their workgroups they perceived greater support from their coworkers. They did not explore whether differential effects existed between younger and older employees. Riordan and Holliday Wayne (2008) did not find a significant relationship between actual age diversity and the amount of open communication within a workgroup. Using different measurement approaches for actual diversity, Clark (2001) and Liao et al. (2004) both reported that difference in age relative to other workgroup members did not predict individuals’ satisfaction with their coworkers. Finally, Liao et al. (2008) found that actual age diversity affected the helping behaviour of members of a workgroup. Specifically, the more age diversity that existed within a workgroup, the less willing the individuals were to engage in cooperative helping behaviour toward other members of the workgroup (Liao et al., 2008). This finding was not supported by Van der Vegt and Van de Vliert (2005), who did not find a significant relationship between actual age diversity and the helping behaviour of business students who were completing an assigned project. Much of the literature examining the consequences of perceived age diversity focuses on a range of outcomes specific to individuals’ attitudes about their work (e.g., job satisfaction) and the nature of their relationships with members of their workgroup. For example, greater perceived age diversity predicted less identification with the workgroup, lower commitment to the organization (Riordan & Holliday Wayne, 2008), and negative attitudes toward one’s job (i.e., job satisfaction and affective commitment) (Liao et al., 2008). Counter to this finding, Clark (2001) reported that perceived differences in age among members of probation departments did not predict job satisfaction or affective commitment. Clark (2001) concluded that individuals who perceived themselves to be similar in age to others were less satisfied with their coworkers; 31  however, Cunningham (2007) found a negative association between perceived age diversity and coworker satisfaction among track and field coaches. Moreover, perceived age diversity was not a statistically significant predictor of work withdrawal, uncooperative helping behaviour (Liao et al., 2008), perspective taking6 (Williams et al., 2007), or greater actual or intended job turnover (Clark, 2001; Cunningham, 2007; Liao et al., 2008). In summary, the research that has examined actual and perceived age diversity has produced mixed results with regard to individuals’ commitment to their organizations and satisfaction with their jobs and coworkers. The most prominent pattern is that perceived age diversity does not seem influence individuals’ relationship and interaction with others in their workgroups; however, actual age diversity has been show to have some detrimental effects. For example, greater actual age diversity in some workgroups has been associated with greater intentions to leave one’s job, diminished peer relationships, less attraction to other members of the workgroup, and less cooperative helping behaviour. One study indicated that actual age diversity may have differential effects for younger and older individuals in such a way that actual age diversity may positively influence individuals’ behaviour (Chattopadhyay, 1999). 2.2.6.2  Actual and Perceived Educational Diversity Few researchers have examined the salience of educational diversity on individual  outcomes. Four studies were located that examined actual educational diversity. In some instances, individuals were found to be more likely to leave their jobs (actual turnover) if they were dissimilar from their colleagues in terms of their educational level (Jackson et al., 1991; Liao et al., 2008); however, others have reported that actual educational diversity was not found to be predictive of greater intentions to leave (Riordan & Holliday Wayne, 2008; Tsui et al., 1992). A few others have found no statistically significant relationships between actual educational diversity and individual outcomes, such as weaker organizational commitment (Riordan & Holliday Wayne, 2008; Tsui et al., 1992), greater absenteeism (Tsui et al., 1992), more work withdrawal, poorer overall job attitudes, less helping behaviour (Liao et al., 2008), lower workgroup identification, and less open communication within the workgroup (Riordan & Holliday Wayne, 2008).  6  The ability to empathize and make positive attributions about others (Williams et al., 2007).  32  In the context of the aims of the current study, I identified two previous studies that explored perceived educational diversity. Riordan and Holliday Wayne (2008) found that perceived educational diversity was a weak predictor of less identification with the workgroup and diminished communication. Perceived educational diversity, which was grouped with education and lifestyle attributes, was not related to organizational commitment or job turnover (Kirchmeyer, 1995). Given the lack of available studies for comparison and the limitations of the research findings, it is difficult to draw definitive conclusions about the influence of actual and perceived educational diversity. 2.2.6.3  Actual and Perceived Ethnic/Racial Diversity Differences with regard to ethnicity/race are one of the most common diversity  attributes studied in the field of organizational behaviour. Research about relational ethnicity/race has produced mixed and often asymmetrical results. Greater actual diversity in ethnicity/race has been associated with weaker organizational commitment (Liao et al., 2004; Tsui et al., 1992), greater intentions to leave an organization, and a higher frequency of absenteeism (Tsui et al., 1992). Riordan and Shore (1997) reported that the greater the actual ethnic/racial diversity between individuals and others in a workgroup, the more negative were individuals’ attitudes toward the workgroup, weak commitment to the workgroup and lower productivity. Liao et al. (2004), Tsui et al. (1992), and Riordan and Shore (1997) all provided preliminary support for the asymmetrical effects of actual ethnicity/race diversity in that the effects may be greater for individuals who do not represent the majority in a given workgroup. Conversely, more recent researchers have reported that actual ethnic/racial diversity is not associated with less satisfaction with or support from coworkers (Clark, 2001; Liao et al., 2004), job dissatisfaction (Clark, 2001; Cunningham & Sagas, 2004), weak organizational commitment (Clark, 2001; Gonzalez, 2001; Keller, 2005; Riordan & Holliday Wayne, 2008), greater intentions to leave a job (Clark, 2001; Cunningham & Sagas, 2004; Gonzalez, 2001; Keller, 2005; Riordan & Holliday Wayne, 2008), lower workgroup identification, less open communication within the workgroup (Riordan & Holliday Wayne, 2008), or psychological empowerment (Keller, 2005). Several researchers in the field of organizational behaviour have examined perceived individual dissimilarity in ethnicity/race. Two studies identified that perceived ethnicity/race was 33  not statistically significant in predicting job satisfaction, turnout intentions, or affective commitment (Clark, 2001; Cunningham, 2007). Riordan and Holliday Wayne (2008), however, found the opposite in their study, in that greater perceived ethnicity/race predicted weaker identification with the workgroup, diminished communication, lower commitment to the organization, and greater intentions to leave. Two studies found that perceived ethnic/racial diversity was negatively related to individuals’ satisfaction with their coworkers (Clark, 2001; Cunningham, 2007). As with the other attributes previously discussed, the research concerned with actual and perceived ethnic/racial diversity has produced mixed results. When actual differences in ethnicity/race are significant, there is some preliminary support for there being asymmetrical effects of ethnicity/race in that some individuals are affected differently by diversity. The asymmetrical effects of perceived ethnic/racial diversity have not been explored. 2.2.6.4  Actual and Perceived Work Values Diversity Few researchers in the field of organizational behaviour have examined  nondemographic attributes in the study of actual diversity. One challenge with studying nondemographic attributes such as work values is the challenge of creating diversity scores. Available for comparison are two studies that examined work values by assessing the actual value congruence between individuals and members of their workgroup (Gonzalez, 2001) and between supervisor–employee dyads (Gelfand, Kuhn, & Radhakrishman, 1996). Ineffective communication between employees and their supervisor is influenced by actual differences in work values (Gelfand et al., 1996), however, research by Gonzalez (2001) failed to support an association between actual values diversity and organizational commitment or job turnover intentions. In studies of non-nursing samples, there is some support for an association between perceived differences in work values and employees’ attitudes (i.e., job dissatisfaction, greater intentions to leave, and lower commitment to the organization or workgroup) (Clark, 2001; Cunningham, 2007; Cunningham & Sagas, 2004; Gonzalez, 2001; Jehn et al., 1999; Liao et al., 2008). When individuals’ perceive that their values differ from those of the organization or workgroup, they have stronger intentions to quit, are less committed, and are more dissatisfied with their jobs. Although not studying burnout per se, Liao et al. (2008) reported that greater 34  perceived deep-level diversity (which includes differences attributed to work values) was predictive of individuals’ work withdrawal behaviour, helping behaviour, and greater likelihood of leaving the workgroup. At the same time, when employees have work values that differ from those of their colleagues, they are also less likely to be involved in their workgroup (Hobman et al., 2003), unable to see the world from another’s viewpoint (Williams et al., 2007), and dissatisfied with their colleagues (Clark, 2001; Cunningham, 2007). Perceptions of dissimilarity in individuals’ personal attributes (which include work values) were also found to predict job dissatisfaction, greater job turnover intentions, and weaker affective commitment to the organization (Clark, 2001). In summary, because there are very few studies that have examined the effects of actual work-values diversity on individuals, it is difficult to draw conclusions. The body of evidence regarding perceived differences in work values, however, is suggestive of individuals’ holding negative attitudes toward their work and members of their workgroup. Individuals who differ from others within the workgroup are more likely to withdraw, are less involved and accepting of others, and are more dissatisfied with their colleagues. 2.2.7  Summary The research exploring the relationship between diversity and burnout is sparse. A  very tentative conclusion put forth is that perceived work-values diversity is associated with burnout as is actual diversity in ethnicity/race. Although the body of evidence associated with other work-related outcomes interrelated to burnout is more abundant, the findings are inconclusive and in some cases contradictory. In general, the research regarding both age and ethnic/racial diversity has produced mixed findings. The most prominent pattern is that actual age diversity leads to negative outcomes (e.g., turnover intentions and poor interaction among workgroup members) whereas perceived age diversity has shown no effect. No definitive conclusions are drawn with regard to educational diversity given the paucity of literature. Finally, there is evidence of a negative relationship between perceived work-values diversity and individuals’ attitudes toward their work and members of their workgroup. There are several limitations and inconsistencies in the research related to the individual consequences of diversity in the workplace, particularly with regard to age, education, ethnicity/race, and work values. One particular reason for these results might be the lack of 35  attention paid to comparison groups or generalizability. There is also limited use of statistical modelling techniques: only three studies used such techniques (Cunningham, 2007; Hobman & Bordia, 2006; Riordan & Holliday Wayne, 2008). Yet, these techniques allow for the simultaneous testing of the attributes of interest on selected outcomes while controlling for measurement error. Moreover, the theoretical foundations of relational diversity indicate that the effects of dissimilarity between an individual and others within a workgroup are context dependent (Garcia-Prieto et al., 2003; Riordan, 2000). That is, the relevance, importance, and significance attached to an attribute (e.g., age) may yield different work-related attitudes and behaviour in a particular social context. Clearly, more research is needed to acquire a better understanding of the attributes that may determine whether relational diversity leads to burnout, and more specifically whether diversity creates conflict among nurses. The next section provides a review of the literature about interpersonal conflict as a mechanism through which diversity may lead to burnout.  2.3  Interpersonal Conflict as a Mediator of the Relationship between Diversity and Burnout What are the processes occurring in the workplace that cause diversity to lead to poor  outcomes among individuals within a workgroup? Leading diversity researchers call attention to the need to move beyond “black box” studies and to examine the mediating roles of various workgroup processes (e.g., integration, conflict, communication) that explain why or how certain outcomes occur as a result of relational diversity (Riordan, 2000; Tsui & Gutek, 1999; Williams & O'Reilly III, 1998). One particular intermediary process that has been emphasized is interpersonal conflict (Jehn & Chatman, 2000; Pelled, 1996a). Conflict is an important indicator of the quality of nurse–nurse interactions and to some extent reflects nurses’ satisfaction with, and the quality of, the social climate in their workplace. The mechanisms by which diversity might influence the occurrence of burnout are a relatively new area of research, particularly with regard to the mediating role of interpersonal conflict. After defining interpersonal conflict, this portion of the literature review focuses on two central themes: (a) the association between diversity and conflict and (b) the association between conflict and burnout. In examining the research literature related to conflict, some emerging themes are identified that provide a  36  preliminary indication that conflict is a mediator of the relationship between diversity and burnout. 2.3.1  Defining Interpersonal Conflict The term, interpersonal conflict is used frequently to refer to perceived  incompatibilities commonly arising when members of a workgroup hold discrepant views about a particular situation or issue or have personal incompatibilities (Jackson & Joshi, 2004; Jehn & Bezrukova, 2004; Kirkman, Tesluk, & Rosen, 2004; Pelled, Eisenhardt, & Xin, 1999; Sacco & Schmitt, 2005; van Knippenberg, De Dreu, & Homan, 2004; Webber & Donahue, 2001; Williams & O'Reilly III, 1998). Barki and Hartwick’s (2004) work on the conceptualization of interpersonal conflict indicated that the cognitive, affective, and behavioural elements of interpersonal conflict are reflected by three fundamental properties: disagreement, negative emotion, and interference. Several different cognitions (e.g., disagreement, differences in opinion, or divergent viewpoints) can be associated with interpersonal conflict; however, the most common cognition is disagreement. According to Barki and Hartwick (2004), “Disagreements exist when parties think that a divergence of values, needs, interests, opinions, goals, or objectives exists” (p. 232). The predominant affective states associated with interpersonal conflict are negative emotions including anger, distrust, fear, frustration, annoyance, hostility, distress, animosity, and jealousy. Finally, several different behaviours (e.g., debate, argumentation, competition, political manoeuvring, backstabbing, aggression, hostility, and destruction) are linked to interpersonal conflict. Interpersonal conflict is generally thought to exist, however, only when such behaviour exhibited by one person interferes with or opposes another person’s attainment of his or her own interests, objectives, or goals (Barki & Hartwick, 2004). Barki and Hartwick’s (2004) typology for conceptualizing and assessing interpersonal conflict in organizations specifies the need to examine three properties (i.e., cognition/disagreement, behaviour/interference, and affect/negative emotion), and the ability to identify more clearly the nature or types of interpersonal conflict. Organizational scholars suggest that the various types of interpersonal conflict can be classified according to their content and focus (e.g., task versus relationship) (Barki & Hartwick, 2004). Interpersonal 37  incompatibilities or disagreements among workgroup members generated by personality differences, differences of opinions, or nonwork-related preferences are described as relationship conflict (also called emotional or affective conflict) (Barki & Hartwick, 2004; Jehn & Bendersky, 2003; Jehn & Chatman, 2000). Task-centred disagreements, on the other hand, concern either the content or the process of a task (Jehn & Bendersky, 2003). Task conflict (also labelled substantive, cognitive, or content conflict) refers to disagreement about the content and goals of the tasks or work being performed, including differences in viewpoints, ideas, and opinions (Barki & Hartwick, 2004; Jehn & Bendersky, 2003). Conflict arising from the process of a task, or process conflict (also called procedural or distributive conflict), focuses on disagreements about how to accomplish a task, who is responsible for a task, or the delegation of duties and resources (Jehn & Bendersky, 2003). Jehn and Bendersky (2003) explicated the distinction between process and task conflict: “Process conflicts are about the means to accomplish the specific tasks, not about the content or substance of the task itself, but about the strategies for approaching the task” (p. 201). Although each conflict type is distinct, under some circumstances, task-related conflict may evolve into relationship conflict, or vice versa (Jehn & Bendersky, 2003; Jehn & Mannix, 2001); for example, if workgroup members harbour particularly strong feelings about a task issue, they may become emotional about an issue (Jehn, 1997; Jehn et al., 1999). Nonetheless, distinctions between task-related and relationship conflict lead to different predictions about the effect of conflict on individual outcomes such as burnout as well as workgroup outcomes (Jehn & Bendersky, 2003). Identifying the constituent properties of interpersonal conflict, as well as their foci and targets, can provide greater clarity to the meaning of the construct and suggests several theoretical and methodological implications (Jehn, 1997; Pelled & Adler, 1994).  38  2.3.2  Empirical Literature Concerning Nurses’ Interpersonal Conflict The literature in nursing has established that the social climate in which nurses work is  fraught with poor nurse–nurse interpersonal relationships, which include various forms of conflictive interactions (Almost, 2006; McKenna, Smith, Poole, & Coverdale, 2003; Quine, 2001; Sa & Fleming, 2008; Stevens, 2002; Yildirim & Yildirim, 2007). In support of these claims, a few qualitative researchers have identified the presence of conflict among nurses in the workplace, which is often characterized as horizontal violence, bullying, or dysfunctional nurse–nurse interactions (Farrell, 1998; Randle, 2003; Taylor, 2001). Others have established that a common source of workplace stress and worry is poor interpersonal relationships (Jobquality.ca, 2009b), which may affect the care provided (Shields & Wilkins, 2006). Approximately one in seven employed Canadians report that poor interpersonal relations in their workplace are a source of stress or excess worry (Jobquality.ca, 2009b). In 2005, among Canadian registered nurses, almost one half (46%) reported low coworker support (Shields & Wilkins, 2006). At 48%, those between the ages of 45 and 54 years were found to be slightly more likely to report low coworker, but on the whole the differences across age groups were small (younger than 35 years = 44% and 55 years or older = 39%). In this large, national survey, coworker support was determined by assessing nurses’ exposure to conflict and the helpfulness of others at work. Both female and male nurses were found to be exposed to hostility or conflict within their workgroup (44% and 50%, respectively). Moreover, 47% did not feel supported by their coworkers (Shields & Wilkins, 2006). Other researchers have identified similar patterns confirming the presence of conflictive interactions among nurses. For example, Rowe and Sherlock (2005) reported that the most common source of verbal aggression frequently experienced by nurses was from other nurses, specifically their staff-nurse colleagues. A small percentage (13%) reported that a verbally abusive experience contributed to a practice error; in one of six of these cases, the experience remained unresolved (Rowe & Sherlock, 2005). The most common long-term consequences of verbally abusive experiences with other nurses were poor working relationships with the aggressor, job dissatisfaction, a diminished sense of well-being, and a lack of trust and sense of support in the workplace (Rowe & Sherlock, 2005). In a different study, McKenna et al. 39  (2003) found that nurses in their first year of practice frequently experienced covert interpersonal conflict, the most common types were feeling undervalued by other nurses, experiencing a lack of supervision, and being distressed by the conflict occurring among others. Those under the age of 30 years were more likely to experience interpersonal conflict, particularly being undervalued and verbally humiliated (McKenna et al., 2003). Some identified consequences of the conflict experienced by new graduates were absenteeism (14%) and intentions to leave the profession (34%). In summary, interpersonal conflict is prevalent within nursing workgroups and manifests itself in various forms. The following sections each focus on an individual factor that may cause conflict in the workplace, specifically the diversity of nurses relative to others within their workgroups and burnout as an effect of conflict among nurses. 2.3.3  The Empirical Literature Concerning the Relationship between Relational Diversity and Interpersonal Conflict My search of the literature did not identify any published studies examining the link  between relational diversity and conflict in nursing workgroups. In the field of organizational behaviour, however, one study explored conflict as a mediator between diversity and worker morale (i.e., satisfaction, intent to remain, and work commitment) (Jehn et al., 1999). Jehn et al. (1999) reported that the effect of perceived work-values diversity on worker morale was mediated by both relationship and process conflict. No other relational diversity attributes were examined. Furthermore, the mediating role of task conflict was not examined. Given the paucity of research regarding conflict as a mediator, I turned to the existing research regarding the direct effects between select diversity attributes and conflict. Six studies were located that examined the diversity–conflict linkage from a relational diversity approach. Three of the studies examined individuals’ involvement in conflict (Hobman & Bordia, 2006; Hobman et al., 2003; Pelled, Xin, & Weiss, 2001) and the others examined individuals’ perceptions of the amount of conflict that occurred within the workgroup (Jehn, Chadwick, & Thatcher, 1997; Jehn et al., 1999; Pelled, 1996b). I considered the findings of these studies according to each type of conflict: relationship, task, and process conflict. Methodological issues associated with these findings are discussed in the summary portion of this section.  40  2.3.3.1  Individuals’ Involvement in Conflict Individuals’ involvement in conflict has been associated with select relational  diversity attributes. Grouping diversity attributes, Hobman et al. (2003) found that perceived values diversity was predictive of both relationship and task conflict (Hobman et al., 2003); however, the other demographic diversity attributes investigated were not found to be significant. Using objective measures of individual dissimilarity, Pelled et al. (2001) reported that greater age dissimilarity between individuals and other workgroup members increased the likelihood of relationship conflict; however, it was not predictive of task conflict (Pelled et al., 2001). In a recent longitudinal study of 165 university graduate students in business administration, Hobman and Bordia (2006) explored the consequences of relational diversity on individuals’ involvement in conflict and determined whether the effects of actual value diversity strengthened across time. Individuals who differed from other members of the project team with regard to their values were more likely to directly experience relationship and task conflict at Time 2, whereas actual dissimilarity in age and ethnicity/race were not significant. For individuals who reported greater identification with their project team, greater actual age and ethnic/racial diversity resulted in less task conflict. The changes over time in the association between diversity and conflict were not statistically significant (Hobman & Bordia, 2006). This lack of significance could be due to the referent group, which varied in size, the amount of interaction within the group, and the permanence of the group. No studies were located that examined the link between diversity and individuals’ involvement in process conflict. 2.3.3.2  Individuals’ Perceptions of Conflict within their Workgroup The second approach through which conflict has been explored is by measuring  individuals’ perceptions of the amount of conflict occurring within their workgroup. Work-values diversity has been consistently associated with relationship conflict, whereas mixed findings have been reported for other attributes. Jehn et al. (1997; 1999) reported that individuals whose values (actual and perceived) were different from others in their workgroups were more likely to report that members of the workgroup experienced a greater amount of relationship conflict. Actual differences in gender and tenure were also found to be predictive of relationship 41  conflict, whereas education, age, and ethnicity/race were not (Jehn et al., 1997; Jehn et al., 1999; Pelled, 1996b). Other researchers have found that task conflict is a consequence of some diversity attributes. For example, Jehn et al. (1997; 1999) showed that value dissimilarity was predictive of task conflict, whereas the effects of other demographic attributes were less conclusive. Although not studying relational diversity per se, Jehn (1994) also reported that individuals whose work values were not congruent with those of their workgroup experienced greater task conflict. Jehn et al. (1997) reported that actual dissimilarity in education and in work values was found to be positively predictive of task conflict; however, a greater percentage of the variance was explained by workgroups having dissimilar values. Actual age diversity and actual ethnic diversity were not associated with task conflict, whereas perceived value diversity and actual informational diversity (i.e., education, functional area, and position) were (Jehn et al., 1997; Jehn et al., 1999). Congruent with these findings, measures of perceived value diversity, compared with actual measures, have been show to explain more of the variance in task conflict. A third form of conflict, process conflict, has received limited attention by researchers in comparison with relationship and task conflict. Process conflict has been found to be significantly and positively predicted by perceived values diversity (Jehn et al., 1999); a greater perception of individual differences in values relative to other workgroup members resulted in individual perceptions of more process conflict present. In contrast, actual informational diversity (i.e., education, functional area, and position) and social category diversity (i.e., gender and age) were found not to be predictive of individual perceptions of process conflict (Pelled, 1996b). 2.3.3.3  Summary The previous published studies of the association between relational diversity and  conflict have produced mixed findings. Such findings can be partially attributed to differences in study populations (e.g., service and professional workers at a large public sector organization, production workers at an electronics assembly plant in Central Mexico, graduate-level business students, employees of household-goods moving companies, and blue-collar employees of electronics manufacturing facilities), the classification of diversity attributes (e.g., discrete categories versus grouping categories), and the operationalization of relational diversity (e.g., 42  actual versus perceived). Notwithstanding these limitations, this body of evidence, albeit small, suggests that there is a link between relational diversity and conflict. A common theme is that individual differences in values, whether actual or perceived, are indicative of relationship conflict, task conflict, and process conflict. Furthermore, one could cautiously conclude that actual differences in some demographic attributes may be associated with conflict: for example, (a) education with individual perceptions of task conflict and (b) age with individual involvement in relationship conflict. Interestingly, measures of perceived diversity, rather than actual diversity, were found to account for a greater percentage of the variance in conflict. This section of the literature review has provided some evidence that diversity is a plausible cause of conflict among nurses (which is a necessary condition for mediation); however, the study of the linkages between diversity and conflict is in its infancy, and further research is warranted. In the next section, I discuss the literature regarding burnout as a potential consequence of conflict, which is another necessary condition to establish that conflict is a mediator of the relationship between diversity and burnout. 2.3.4  The Empirical Literature Concerning the Relationship between Conflict and Burnout Interpersonal conflict among nurses, which is common in the workplace but often  subtle, is an unpleasant experience that results in negative attitudes and behaviour. The adverse consequences of interpersonal conflict may include, but are not limited to, job dissatisfaction, weak organizational commitment, absenteeism, lack of involvement, burnout, and turnover (Almost, 2006; Ayoko, Callan, & Hartel, 2003; Cox, 2001, 2003; De Dreu, van Dierendonck, & Maria, 2004; Gardner, 1992; Jehn & Bendersky, 2003). Specifically, researchers have shown that in the general population and among healthcare employees, the occurrence of burnout, particularly emotional exhaustion, can be attributed to negative collegial interactions and interpersonal conflict (Giebels & Janssen, 2005; Mulki, Jaramillo, & Locander, 2008; Taris, Le Blanc, Schaufeli, & Schreurs, 2005). Moreover, positive collegial relationships have been associated with less job-related stress (Jobquality.ca, 2009b; Tervo-Heikkinen, Partanen, Aalto, & Vehvilainen-Julkunen, 2008). A preliminary theme identified in the field of nursing is the link between conflict and burnout. As this is a relatively new area of research, only six relevant studies were located. A 43  limitation of this body of research is that many researchers predominantly examine the emotional exhaustion aspect of burnout or, in one instance, burnout as a unitary construct. Hillhouse and Adler (1997) used the Staff Burnout Scale for Health Professionals to examine the characteristics of groups experiencing low, moderate, and high levels of burnout. They found that interpersonal conflict among nurses was one of several stressors that resulted in feelings of burnout (Hillhouse & Adler, 1997). Of five studies that used the Maslach Burnout Inventory, only two examined the effects of conflictive nurse–nurse interactions on all three aspects of burnout (Fujiwara, Tsukishima, Tsutsumi, Kawakami, & Kishi, 2003; Payne, 2001). For this reason, I present the literature with regard to each aspect of burnout: emotional exhaustion, depersonalization, and personal accomplishment. A few research teams have reported that conflictive interactions among nurses were found to be positively correlated with emotional exhaustion (Payne, 2001; Stordeur, D'Hoore, & Vandenberghe, 2001). Similarly, Leiter and Maslach (1988) found that unpleasant relationships with coworkers resulted in individuals’ feeling emotionally exhausted. In two different studies involving Japanese nurses, the research teams reported that coworker conflict was not significantly associated with emotional exhaustion (Fujiwara et al., 2003; Kitaoka-Higashiguchi, 2005). The differences pertaining to the relationship between conflict and emotional exhaustion may be related to the different populations for which the samples were drawn (e.g., North American versus Japan). To a lesser extent, depersonalization also has been addressed as a consequence of interpersonal conflict. Two studies examined the relationship between interpersonal conflict and depersonalization. Both supported the hypothesis that conflict with other nurses resulted in depersonalization (Fujiwara et al., 2003; Payne, 2001). Among hospices nurses, nurse–nurse conflict explained the greater proportion of the variance in depersonalization (21%) relative to other occupational stressors (Payne, 2001). Positive interactions among coworkers were found to be beneficial in lessening the occurrence of depersonalization (Leiter & Maslach, 1988). The other aspect of burnout, personal accomplishment, has received minimal attention from researchers. In one study, Payne (2001) concluded that personal accomplishment was not associated with interpersonal conflict among nurses. In other words, experiencing interpersonal conflict with colleagues at work did not seem to cause these nurses to evaluate themselves in a 44  negative manner. In summary, these studies contribute to the evidence indicating that a potential consequence of interpersonal conflict is burnout, particularly emotional exhaustion and depersonalization. A limitation of this body of research is that it does not acknowledge the types of interpersonal conflict (e.g., relationship, task, and process conflict) occurring between nurses and their colleagues. 2.3.5  Summary Some documentation in the nursing literature addresses the prevalence of interpersonal  conflict within nursing workgroups. What is not clear from this body of research is why such conflict occurs, the particular types of conflict that occur (e.g., task, process, or relationship), and their detrimental effects. Additional research is required to obtain a more thorough understanding of the individual factors that engender such conflict and the impact it has on nurses. No research has been conducted to examine specifically whether conflict, arising from individual differences within the nursing workforce contributes to burnout. The portions of the literature regarding the possible diversity–conflict–burnout linkages is sparse, fragmented, and in a few instances contradictory. Nonetheless, evidence from the field of organizational behaviour provides some preliminary insights into the relationship between diversity and conflict at the individual level. Specifically, work-values diversity, whether actual or perceived, is indicative of relationship conflict, task conflict, and process conflict. Some demographic attributes, for example age and education, are more salient than others in explaining each type of conflict. The evidence presented in the literature also hints toward the possibility that chronic and unresolved conflict is destructive to the individual and the larger work community, which may lead to burnout, particularly emotional exhaustion and depersonalization. Further research is required to understand the complexities of whether interpersonal conflict is salient in explaining how diversity results in burnout.  45  2.4  Chapter Summary Joining the many hints from the three main bodies of literature reviewed, a plausible  conceptual model is presented with regard to the direct effects of actual and perceived relational diversity on burnout. It is plausible that interpersonal conflict may be a mechanism by which diversity leads to burnout. In understanding the complexities of diversity in the workplace some noteworthy methodological shortcomings include the limited use of modelling techniques for data analysis, the correlational nature of most of the studies, and the use of a variety of sample populations and referent groups. Notwithstanding the aforementioned methodological limitations, the various bodies of evidence relevant to diversity in the nursing workforce indicate that each attribute does not necessarily result in the same outcome. For example, differences in observable attributes such as perceived age and ethnicity/race (actual and perceived) may be associated with negative interactions among members of a workgroup, and perceived differences in work values may influence individuals’ attitudes toward their work. A few researchers have examined the link between diversity and burnout in non-nursing populations; however, the findings are fragmented and in some instances inconclusive. Researchers examining the outcomes of diversity have primarily focused on work-related factors that are interrelated with the burnout construct. In such instances, actual age diversity has been associated with turnover intentions and poor collegial relationships. Mixed results are available with regard to the impact of actual ethnic/racial diversity. Very limited information exists with regard to educational diversity (actual and perceived) and perceived ethnic/racial diversity. Work values diversity, however, stands out as a particularly salient attribute that may contribute to individuals’ negative attitudes toward their work and other members of their workgroup. In general, there is indication that individuals working in diverse workgroups may experience some negative consequences. Undoubtedly, research is needed to address the social factors in their work environment that affect the well-being of nurses and contribute to the burnout they frequently experience. In considering the mechanisms by which diversity leads to burnout, the literature relevant to interpersonal conflict reveals that relational diversity attributes such as work values, education, and age may be important predictors of relationship and task conflict. Moreover, 46  several researchers have highlighted that burnout, particularly emotional exhaustion and depersonalization, is a consequence of the conflict that arises between nurses. Given the prevalence of burnout and interpersonal conflict in the work environment, there is a need to obtain a greater understanding of the social aspects of the workplace and their contributions to the quality of healthcare work environments. Building upon the literature review of the direct and indirect effects of relational diversity in the workplace, the following chapter highlights a postulated conceptual model that specifies the means by which diversity is hypothesized to lead to burnout and the mechanisms by which this occurs – that is, the experience of interpersonal conflict.  47  3 CONCEPTUAL MODEL AND HYPOTHESES This chapter provides an overview of a conceptual model specifying how diversity in the workplace is hypothesized to lead to burnout. In addition to articulating direct effects between diversity and burnout, the conceptual model delineates how the degree of diversity between an individual and others within a workgroup leads to interpersonal conflict, and how this conflict, in turn, leads to burnout. In other words, the model reveals how the influence of relational diversity in age, education, ethnicity/race, and work values on burnout (emotional exhaustion, depersonalization, cynicism, and a diminished sense of personal accomplishment) is explained through the mediating effects of relationship, task, and process conflict. The relationships among the constructs of interest are based on two complementary theories: social identity theory and similarity–attraction theory. The conceptual model is illustrated in Figure 3.1 to Figure 3.4.  3.1  Theoretical Foundations Much of the relational diversity research is predicated on the logic of social identity  theory (Tajfel & Turner, 1986; Tajfel, 1978) and its newer extension self-categorization theory (Turner 1982, 1987), which provides a social psychological perspective of group members’ identification with their group as a whole, rather than with individual members within that group (Brewer, 1995; Chattopadhyay, George, & Lawrence, 2004). Embedded in these theories is the notion that the individual’s sense of self is comprised of both a personal identity and a social identity (Ashforth, 2001). Two key premises of social identity theory are that individuals: (a) derive a significant portion of their identity from the social categories to which they belong and (b) have a desire to maintain a high level of self-esteem and a positive self-identity (Riordan, 2000). According to social identity theory, individuals use salient attributes to define themselves and others as either belonging to various social categories that share some common identity, or as being members of different categories (Northcraft, Polzer, Neale, & Kramer, 1995; Tsui, Xin, & Egan, 1995). Through a process of social comparison, or self-categorization, the individual categorizes himself or herself and others into groups by attaching value to particular social attributes that are then used to provide meaningful distinctions between people or subgroups of people (Ashforth, 2001; Riordan, 2000). These social identifiers or identities are 48  relational and comparative in that category membership is defined relative to the members of other categories (Ashforth, 2001). By identifying with a particular social category, “individuals perceive themselves as psychologically intertwined with the fate of the category, sharing its common destiny, and experiencing its successes and failures” (Ashforth, 2001, p. 25). As individuals begin to classify themselves and others, they usually assume the perceived prototypical or exemplary characteristics of the category as their own. Additionally, the unique attributes of individuals are downplayed as they come to see themselves as more or less typical of the social category. When the social identity is salient, individuals think and act as exemplars of the category (Ashforth, 2001). Because individuals want to sustain a high level of self-esteem and a positive self-identity, they tend to accentuate similarities within and differences among categories, and they tend to develop more positive opinions of their own category (the in-group) and negative opinions of those outside of their category (the out-group) (Ashforth, 2001; Webber & Donahue, 2001). Specific persons who exemplify the salient attributes become members of the in-group, while those who are different represent members of the out-group (Ashforth, 2001), thereby creating we–they or us–them distinctions that could potentially affect individual behaviour and result in poor collegial relationships (Hobman, Bordia, & Gallois, 2003; van Knippenberg, De Dreu, & Homan, 2004). The basic premise of the similarity–attraction perspective is that “individuals who possess similar individual characteristics and attitudes will perceive one another as similar and will be attracted to each other” (Chuang, Church, & Zikic, 2004, p. 28). Other researchers hypothesize that this initial attraction between oneself and others, established through perceptions of similarity or dissimilarity in visible demographic attributes, leads to inferences about similarities in values, beliefs, and attitudes (Tsui, Xin, & Egan, 1995). This initial perception of others may change when detailed information about less-visible or nondemographic attributes (e.g., values, beliefs, attitudes, and knowledge) are obtained. Thus, the meanings assigned to attributes are socially constructed as individuals act in, and toward, the world through social interaction, which varies among individuals, from one context to another, and over time. Regardless of the attributes leading to an initial attraction, individuals develop a sense of predictability, comfort, and confidence with similar others (Harrison, Price, & Bell, 1998; Tsui et al., 1995). 49  3.2  The Conceptual Link between Relational Diversity and Burnout: Why are Dissimilar Individuals More Likely to Experience Burnout? In the presence of diversity, there are several reasons why individuals may experience  burnout. Similarity attraction and categorization processes result in “othering” where individuals thought to be different from oneself are marked and named as such (Canales, 2000; Johnson et al., 2004). Those perceived as different from the dominant social category may experience covert forms of social mistreatment, which may cause individuals to have difficulty in relating to their colleagues, and vice versa. The noted dissimilarity among individuals in a workgroup is likely to affect the level of respect and support among members of the group, the ease of their communication, and the degree to which they have a sense of belonging or attachment to their workgroup (Hobman et al., 2003; Pelled, 1996b). Individuals experiencing ongoing, stressful collegial relations may feel emotionally drained and depleted, and, in some cases, may have negative feelings about their colleagues, may feel like they are no longer able to give themselves to others, and may come to view their jobs negatively (Cordes & Dougherty, 1993; Maslach, Schaufeli, & Leiter, 2001). Moreover, working in an environment that is disrespectful and unsupportive may cause individuals to voluntarily isolate themselves and to minimize contact with all people (Maslach, 1982). Such individuals are more likely to develop depersonalized responses (e.g., negative and callous interactions) and to develop indifferent attitudes toward their work and others within the workgroup (and possibly their clients). According to Maslach (1982), “‘Just leave me alone and let me do my job by myself’ is the message that comes from the individual who sits off in a corner, does not socialize with coworkers at lunch or coffee breaks, and leaves immediately when the day is done” (p. 43). If this isolation persists individuals’ feelings of efficacy may diminish (Maslach et al., 2001). Individuals who do not exemplify the dominant majority may become isolated and excluded (Canales, 2000; Hobman et al., 2003; Pelled, 1996b). Individuals working in such an environment may (a) feel emotionally drained, (b) choose to leave by psychological withdrawing from the workgroup and distancing themselves from aspects of their work and the people with whom they work, and (c) experience feelings of personal inadequacy (Maslach et al., 2001). Individuals sharing a social category often are assumed to share similar values and interest, and thus in-group members are often viewed as being more predictable, trustworthy, and likely to reciprocate favours than are members of an out-group (Schneider & Northcraft, 1999). 50  When there is less personal attraction among members of a workgroup, individuals’ core values and beliefs about their work are threatened, and members are less likely to develop a sense of predictability and confidence in each other’s abilities and behaviour (Harrison et al., 1998; Tsui et al., 1995). When individuals are judged or criticized they may develop feelings of inadequacy and may self-impose a verdict of failure. They may experience diminished feelings of competence or achievement in their work (Schaufeli, Maslach, & Marek, 1993). Collectively, social identity theory and similarity–attraction theory suggest that diversity between an individual and others within a workgroup leads to feelings of emotional exhaustion and depersonalization, and a diminished sense of personal accomplishment. Although these theories emphasize people’s perceptions, diversity researchers have highlighted the importance of an individual’s actual difference from other members of the workgroup. Thus, the following hypotheses were proposed (see Figure 3.1 and Figure 3.2): Hypothesis 1: Actual relational diversity between an individual and others within the workgroup is positively associated with emotional exhaustion, depersonalization, and cynicism, and is negatively associated with personal accomplishment. Hypothesis 2: Perceived relational diversity between an individual and others within the workgroup is positively associated with emotional exhaustion, depersonalization, and cynicism, and is negatively associated with personal accomplishment. Researchers approaching diversity from a relational perspective have increasingly recognized the importance of nondemographic attributes, especially when investigating perceived dissimilarity (Clark, Ostroff, & Atwater, 2002). The various bodies of literature relevant to diversity also support the possibility that a range of attributes may lead to burnout within the nursing workforce. Several possible attributes could be used as criteria for creating social divisions among individuals in workgroups; this research focused on four attributes (i.e., age, education, ethnicity/race, and work values) that were identified as being potentially salient to the population of interest.  51  Hypothesis 1: Actual Diversity Hypothesis 1.1: Actual age diversity between an individual and others within a workgroup is positively associated with emotional exhaustion, depersonalization, and cynicism, and is negatively associated with personal accomplishment. Hypothesis 1.2: Actual educational diversity between an individual and others within a workgroup is positively associated with emotional exhaustion, depersonalization, and cynicism, and is negatively associated with personal accomplishment. Hypothesis 1.3: Actual ethnic/racial diversity between an individual and others within a workgroup is positively associated with emotional exhaustion, depersonalization, and cynicism, and is negatively associated with personal accomplishment. Hypothesis 1.4: Actual work values diversity between an individual and others within a workgroup is positively associated with emotional exhaustion, depersonalization, and cynicism, and is negatively associated with personal accomplishment. Hypothesis 2: Perceived Diversity Hypothesis 2.1: Perceived age diversity between an individual and others within a workgroup is positively associated with emotional exhaustion, depersonalization, and cynicism, and is negatively associated with personal accomplishment. Hypothesis 2.2: Perceived educational diversity between an individual and others within a workgroup is positively associated with emotional exhaustion, depersonalization, and cynicism, and is negatively associated with personal accomplishment. Hypothesis 2.3: Perceived ethnic/racial diversity between an individual and others within a workgroup is positively associated with emotional exhaustion, depersonalization, and cynicism, and is negatively associated with personal accomplishment. Hypothesis 2.4: Perceived work values diversity between an individual and others within a workgroup is positively associated with emotional exhaustion, depersonalization, and cynicism, and is negatively associated with personal accomplishment.  52  Figure 3.1  Model 1: The Effect of Actual Relational Diversity on Burnout  EE  H1 (+) DP ACTUAL  H1 (+)  RELATIONAL DIVERSITY • • • •  Age Education Ethnicity/race Work Values  H1 (+)  CY  H1 (-) PA  Note. EE = Emotional exhaustion, DP = Depersonalization, CY = Cynicism, and PA = Personal accomplishment.  53  Figure 3.2  Model 2: The Effect of Perceived Relational Diversity on Burnout  EE  H2 (+) DP PERCEIVED RELATIONAL DIVERSITY • • • •  Age Education Ethnicity/race Work Values  H2 (+)  H2 (+)  CY  H2 (-) PA  Note. EE = Emotional exhaustion, DP = Depersonalization, CY = Cynicism, and PA = Personal accomplishment.  54  3.3  The Conceptual Link between the Effect of Relational Diversity on Burnout as Mediated by Interpersonal Conflict In this section, I explore the reasons why dissimilar individuals are more likely to be  involved in conflict (relationship, task, and process) and subsequently experience more burnout (emotional exhaustion, depersonalization, cynicism, and a diminished sense of personal accomplishment). There are several reasons why individuals who are different from others within a workgroup may experience interpersonal conflict. First, social categorization and similarity attraction processes increase the likelihood of unpleasant working relationships and negative exchanges among members of a workgroup, particularly for those who are different. The differences that separate members of a workgroup may affect the development of empowering relationships between individuals and their “othered” coworkers. Individuals may have trouble understanding or being understood by members of other social categories (Hobman et al., 2003: Pelled, 1996b), and their interactions are likely to be more antagonistic than those among individuals within a category (Pelled, Xin, & Weiss, 2001). In such situations, individuals tend to develop frustrations with, and hostile attitudes towards, others and to think more negatively about their colleagues (Pelled, 1996). These strained relationships can further create interpersonal friction and unconstructive interpersonal exchanges. Through a lens of negative affect, it may be difficult for dissimilar individuals to see their workgroup in a positive light; hence, they may be particularly inclined to describe the group as having conflict, or they may be more likely to be personally involved in disagreements with their coworkers, or both (Hobman et al., 2003; Pelled, 1996). Moreover, the presence of individual diversity may increase the discomfort of the workgroup as a whole and may hamper communication among members; thus, it may make all workgroup members “edgy” and irritable, prompting frequent arguments among those who are similar to each other, in addition to those who are different (Pelled, 1996b). Another underlying mechanism supporting the linkage between diversity and conflict is that individuals representing various social categories have contrasting values, goals, preferences, and opinions about work- and non-work-related activities (Hobman et al., 2003; Pelled et al., 2001). The more different individuals are from other workgroup members on a given attribute, the greater the likelihood of conflict developing (Hobman et al., 2003). This conflict occurs because people who are attracted to those who are similar to themselves often 55  share the same values and world view (Jehn, Chadwick, & Thatcher, 1997). “They also assume that similar others are easier to work and communicate with as well as believing they are more trustworthy” (Jehn et al., 1997, p. 290). People with dissimilar ages, educational backgrounds, or work values have different opinions and perspectives and tend to approach their work differently. These differences may result in greater involvement in disagreements about work-related topics (e.g., the goals of the work or how to accomplish the work) and relationship disagreements (Hobman et al., 2003; Jehn et al., 1997; Pelled et al., 2001). Conversely, individuals with similar values may have smoother interaction processes and more agreement, which minimize misunderstandings and work-related conflict (Jehn, 1994; Jehn et al., 1997). In this instance, “values can act as perceptual filters [where] members with similar values are more likely to prioritize and interpret group problems and events in similar ways” (Jehn et al., 1997, p. 288), which further reduces work-related (task and process) conflict. Dissimilarity of values also increases relationship and task conflict by reducing the degree to which group members identify with one another (Jehn et al., 1997). Congruent with these ideas, researchers have reported that perceived diversity in values among colleagues is predictive of relationship conflict and in some instances task and process conflict (Hobman et al., 2003; Jehn, Northcraft, & Neale, 1999). Others have reported that, to a lesser extent, the likelihood of relationship conflict increases with greater actual diversity on visible attributes, such as age; however, task conflict has been attributed to actual differences in education and work experience (Jehn et al., 1997; Jehn et al., 1999; Pelled et al., 2001). A third possibility regarding why diversity may lead to interpersonal conflict is that through social categorization processes, in-group favouritism and out-group derogation can lead to stereotyping (Hobman et al., 2003). Through this process of “othering” those individuals thought to be different are noted and named as such (Canales, 2000; Johnson et al., 2004). The more dissimilar individuals are excluded and stereotyped, albeit sometimes unintentionally, the more the attributes of the in-group are solidified and reinforced (Canales, 2000; Hobman et al., 2003; Johnson et al., 2004). The greater the stereotyping, the less likely it is that individuals and their “othered” coworkers will compromise on their beliefs and values (Swearingen & Liberman, 2004). Not being able to find some middle ground increases the likelihood that individuals will engage in constant bickering and fighting, starting arguments with one another, and accentuating 56  trivial issues (Maslach, 1982). Moreover, when individuals identify with an in-group, they are much more likely to perceive the out-group as being responsible for any conflict (Garcia-Prieto, Bellard, & Schneider, 2003). The consequences of being involved in, or exposed to, conflict among members of a workgroup can contribute to burnout. Individuals who experience disagreements with their colleagues are inundated with a plethora of negative feelings (e.g., anger, frustration, distress, fear, annoyance, distrust, animosity, and hostility) and may feel like they are not part of the workgroup (Maslach et al., 2001). Being involved in prolonged and unresolved conflict is destructive to individuals, causing them to feel emotionally drained and withdrawn from their work and other people, including their coworkers and clients. Individuals within a community characterized by unpleasant working relationships with members of their workgroup may isolate themselves and minimize contact as a means of reducing their interpersonal stress (Maslach, 1982, 2003). At the same time, individuals involved in conflict may display negative, callous, and indifferent attitudes toward others. Hostility and anger, attributed to conflict, can also result in professional derogations and lead to interference with one another’s work (Maslach, 1982). The consequences of being excluded as a result of being different from others are often alienation, shrinking opportunities, and internalized oppression7 (Canales, 2000). Collectively, this may result in greater tendency for dissimilar individuals of a workgroup to evaluate themselves negatively and, as a result, their feelings of efficacy may diminish (Wesolowski & Mossholder, 1997). Individuals having conflictive interactions with their colleagues may harbour a sense of futility about discussing work issues with their colleagues and are more likely to experience burnout (Wesolowski & Mossholder, 1997). Links between conflict and some aspects of burnout are evident in the research literature. For example, nurses’ conflict with their colleagues at work is positively associated with emotional exhaustion and depersonalization but not a sense of diminished personal accomplishment (Fujiwara, Tsukishima, Tsutsumi, Kawakami, & Kishi, 2003; Payne, 2001; Stordeur, D'Hoore, & Vandenberghe, 2001).  7  Internalized oppression is the process by which a member of an oppressed group comes to accept and live out the inaccurate myths and stereotypes applied to the group (Urban Dictionary, n.d.). External oppression becomes internalized oppression when a person comes to believe and act as if the oppressor's beliefs system, values, and life way constitute reality (Women's Rural Advocacy Programs, n.d.).  57  The degree to which an individual differs from other workgroup members on select attributes can have profound effects on the amount of conflict experienced. Being different not only shapes individuals’ perspectives of the workgroup as having more conflict, but also influences individuals’ involvement in conflict with their coworkers. The following hypotheses specify the indirect relationship between relational diversity and burnout as being mediated by the interpersonal conflict experienced by an individual (see Figure 3.3 and Figure 3.4). Hypothesis 3: The effects of actual relational diversity on burnout are mediated by individuals’ perceptions of conflict within the workgroup. Hypothesis 3.1: The effects of actual age diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ perceptions of relationship, task, and process conflict within the workgroup. Hypothesis 3.2: The effects of actual educational diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ perceptions of relationship, task, and process conflict within the workgroup. Hypothesis 3.3: The effects of actual ethnic/racial diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ perceptions of relationship, task, and process conflict within the workgroup. Hypothesis 3.4: The effects of actual work values diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ perceptions of relationship, task, and process conflict within the workgroup. Hypothesis 4: The effects of actual relational diversity on burnout are mediated by individuals’ involvement in conflict within the workgroup. Hypothesis 4.1: The effects of actual age diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ involvement in relationship, task, and process conflict within the workgroup. Hypothesis 4.2: The effects of actual educational diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ involvement in relationship, task, and process conflict within the workgroup. Hypothesis 4.3: The effects of actual ethnic/racial diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are 58  mediated by individuals’ involvement in relationship, task, and process conflict within the workgroup. Hypothesis 4.4: The effects of actual work values diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ involvement in relationship, task, and process conflict within the workgroup. Hypothesis 5: The effects of perceived relational diversity on burnout are mediated by individuals’ perceptions of conflict within the workgroup. Hypothesis 5.1: The effects of perceived age diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ perceptions of relationship, task, and process conflict within the workgroup. Hypothesis 5.2: The effects of perceived educational diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ perceptions of relationship, task, and process conflict within the workgroup. Hypothesis 5.3: The effects of perceived ethnic/racial diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ perceptions of relationship, task, and process conflict within the workgroup. Hypothesis 5.4: The effects of perceived work values diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ perceptions of relationship, task, and process conflict within the workgroup. Hypothesis 6: The effects of perceived relational diversity on burnout are mediated by individuals’ involvement in conflict. Hypothesis 6.1: The effects of perceived age diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ involvement in relationship, task, and process conflict within the workgroup. Hypothesis 6.2: The effects of perceived educational diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ involvement in relationship, task, and process conflict within the workgroup. Hypothesis 6.3: The effects of perceived ethnic/racial diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ involvement in relationship, task, and process conflict within the workgroup. 59  Hypothesis 6.4: The effects of perceived work values diversity on emotional exhaustion, depersonalization, cynicism, and personal accomplishment are mediated by individuals’ involvement in relationship, task, and process conflict within the workgroup.  3.4  Chapter Summary People may experience burnout in response to being different from others within a  workgroup. Being different from others, they may feel emotionally exhausted, display distant, negative, or cynical attitudes toward others, or experience a diminished sense of personal accomplishment. The influence of diversity on burnout is explained, in part, through the effects of relationship, task, and process conflict. These effects are explicated within the framework of social identity theory and similarity–attraction theory. In the next chapter, I provide an overview of the methods used to test the hypotheses explicated here.  60  Figure 3.3  Model 3: The Effect of Actual Relational Diversity on Burnout as Mediated by Interpersonal Conflict  CONFLICT • Relationship • Task • Process  H3 & H4 (+) EE  H3 & H4 (+)  H3 & H4 (+)  ACTUAL RELATIONAL DIVERSITY • • • •  Age Education Ethnicity/race Work Values  DP  H3 & H4 (+) CY  H3 & H4 (-) PA  Note. EE = Emotional exhaustion, DP = Depersonalization, CY = Cynicism, and PA = Personal accomplishment  61  Figure 3.4  Model 4: The Effect of Perceived Relational Diversity on Burnout as Mediated by Interpersonal Conflict  CONFLICT • Relationship • Task • Process  H5 & H6 (+) EE  H5 & H6 (+)  H5 & H6 (+) DP  PERCEIVED RELATIONAL DIVERSITY • • • •  Age Education Ethnicity/race Work Values  H5 & H6 (+) CY  H5 & H6 (-) PA  Note. EE = Emotional exhaustion, DP = Depersonalization, CY = Cynicism, and PA = Personal accomplishment  62  4  METHODS This research study used a cross-sectional, correlational design to test the conceptual  model presented in Chapter 3. Self-administered questionnaires were distributed to the nursing staff working at two acute care hospitals in British Columbia (BC), Canada. Structural equation modelling was used to test the theoretical model of relational diversity and its relationships with conflict and burnout. The following chapter provides an overview of the sampling strategy, data collection procedures, operationalization of the study constructs, methods for data analysis, and ethical considerations.  4.1  Sample  4.1.1  Setting and Participants The population of interest was practicing nurses who provided direct client care in two  acute care hospitals in the Lower Mainland of BC. The selection of nurses working in acute care facilities was based on the observations, in 2006, that 63% (n = 18,109) of the registered nursing workforce in BC identified their primary place of employment as a hospital and that 67% of registered nurses (RNs) working in the hospital setting were 40 years of age or older (Canadian Institute for Health Information, 2007b). For RNs employed by hospitals, the number of years that had passed since their graduation was distributed as: 0-10 years = 25%, 11-20 years = 28%, 21-30 years = 24%, and 31+ years = 21%). In BC, in 2006, about one third of RNs working in hospitals (n = 18,109) were employed full-time (34%). A smaller percentage worked part-time (18%) or on a casual basis (10%) (Canadian Institute for Health Information, 2007b). In addition, 90% of these RNs lived in urban areas (Canadian Institute for Health Information, 2007b). At the same time, slightly more than one half of the province’s licensed practical nurses (LPN) were employed by hospitals (n = 2,945, 54%) and most provided direct care (n = 5,313, 98%) (Canadian Institute for Health Information, 2007a). About one half of the LPNs employed by hospitals were employed full-time (47%) and to a lesser extent casually (39%) or on a part-time basis (13%). Almost one half of these LPNs had completed their education 10 or more years in the past (Canadian Institute for Health Information, 2007a). In general, large metropolitan areas such as the Lower Mainland are more ethnically diverse than the rural areas of BC. The selected health authority, in the Lower Mainland, 63  provides a wide range of healthcare services to approximately 1.5 million people; it has more than 8,000 nurses providing services in 12 acute care hospitals and various community programs. In consideration of the sample requirements and feasibility, two tertiary care hospitals were selected to initiate data collection. Each hospital has approximately 350 beds and collectively employed approximately 1,476 nurses (830 full-time equivalent) (Bennington, 2006). In each hospital, the nurses (RNs and LPNs) who provided direct client care on the medical, surgical, pediatric, perinatal, and neonatal intensive care nursing units were targeted for inclusion in the study. The use of a relational approach to the study of diversity required the enrolment of close to the entire population of nurses in each work unit, as opposed to the drawing of a random sample. Some nursing units employed both RNs and LPNs, and both types of nurses were included. A recruitment plan was implemented to ensure that the maximum number of nurses from each work unit were included in the current study. 4.1.2  Recruitment of Participants Relational diversity is more likely to influence and be affected by interpersonal  relationships with people among whom interaction is frequent (Riordan, 2000), such as nursing unit colleagues, patient care leaders, unit-specific nursing educators, and clinical nurse specialists. The demographic composition of the people with whom a particular member interacts most within a workgroup has a stronger potential to shape the workgroup image he or she constructs. Actual diversity, therefore, was computed at the individual level of analysis. Nurses (RNs and LPNs) employed on a particular nursing work unit constituted the sample workgroup. To measure relational diversity and to enhance the interpretability of the findings, participants were recruited by sampling entire work units so that the study sample was represented by as many individuals as possible from each workgroup (Riordan, 2000). The work units were groups of nurses working in specific nursing units or departments that primarily provided direct client care to a specific population (e.g., medical, surgical, or obstetrical patients). All nurses in each work unit, regardless of their employment status, shift schedule,  64  position, 8 or employment status9 (i.e., full-time, part-time, or casual), were invited to participate on a voluntary basis. Colleagues were defined as the people with whom the participants had the most contact within their nursing work unit, including coworkers (RNs and LPNs), clinical educators (CNEs), clinical nurse specialists, and clinical practice leaders (e.g., patient care coordinators (PCCs) and clinical resource nurses (CRNs). If the recruited nurse worked on more than one unit (e.g., casual employment), she or he was asked to complete the questionnaire in reference to the workgroup with whom she or he interacted the most. The participants were required to be registered as a regulated nurse and to work on a regular basis. “Regular” employment on the nursing unit (e.g., working casual on a frequent basis) was required so that the participants could answer questions that required them to make comparisons between themselves and their coworkers, on their nursing unit, and to assess the amount of conflict on the nursing unit. Excluded from the study were nurses on leave (e.g., parental, education, sick, or disability leave), on gradual return to work, or those assigned to a float pool.10 One month before distribution of the study materials, the nursing unit managers generated lists of nurses11 employed on the  8  Nurses whose area of responsibility was coordinating client care (e.g., patient care coordinators or clinical resource nurses) or education (clinical nurse educators) were eligible for inclusion if they spent a significant amount of time interacting with members of a particular workgroup and, in some instances, their role may have included the provision of direct client care. 9  Because 60% of the new graduates were not employed in regular (permanent) nursing positions, nurses who were employed in regular, permanent (i.e., full-time or part-time), temporary full- or part-time, or casual positions were invited to participate (College of Registered Nurses of British Columbia, 2005). Regular full-time employees were those who were scheduled regularly to work at least 35 hours of work per week, on average, and regular part-time employees were those scheduled to work a minimum of 14.4 hours or equivalent per week but less than the full hours. Casual employment refers to those employees whose employment schedule did not guarantee a fixed number of hours of work per pay period and who were usually pre-booked or called in to relieve employees on short-term vacation, or sick leave, or to assist with workload demands. Temporary full-time or part-time employment refers to a temporary position in which the employment schedule guaranteed a regular number of hours of work per period for a specific time period (or until return of the incumbent). This usually applied to employees who were relieving other employees on a long-term leave or maternity leave or employees working in term positions (e.g., time defined projects or summer relief positions). 10  A float pool refers to a list of casual employees who were usually pre-booked or called in to relieve employees on short-term vacation, or sick leave, or to assist with workload demands. The float pools were usually hospital-wide float pools in that the nurses were assigned to work on several different units. 11  Not all lists were up-to-date.  65  nursing units from Sites A and B. A total of 879 nurses were initially identified as being eligible to participate (see Figure 4.1). 4.1.3  Sample Size No formula was available for an a priori determination of the sample size required to  produce sufficient statistical power because the solution of the structural equation models depended on the reliability of the indicators included, the number of parameters estimated, the distributions of the indicators, the extent to which collinearity was present, the extent to which the model was identified, and factors unknown at the beginning of the study (Hancock, 2006). Two approaches were considered during the planning phase of the study to ensure that an adequate sample size was obtained. First, the root mean square error of approximation (RMSEA) test of “not close fit” was used to estimate the sample size required to test overall model fit (Hancock, 2006; Hancock & Freeman, 2001; MacCallum, Browne, & Sugawara, 1996). To test a model with df = 250 and desired power (π) of 0.80, the requisite sample size depends on the level of noncentrality (ε) anticipated. For a perfectly specified model (ε = 0), a minimum of n = 101 individuals was estimated to be required (Hancock, 2006; Hancock & Freeman, 2001; MacCallum et al., 1996). Increasing the levels of possible noncentrality resulted in a corresponding increase in the required sample size to 125 and 403 for ε = 0.02 and 0.04, respectively. The model ultimately specified in this study had df = 777; consequently, a sample as small as 125 or as large as 403 would have resulted in sufficient power for the two levels of noncentrality, respectively. In consideration of the RMSEA test, a maximum n of 403 was considered sufficient to test for overall model fit with π = 0.80 (or greater) and ε = 0.04. Based on the experiences of other researchers who have surveyed nurses (Borkowski, Amann, Song, & Weiss, 2007; Cho, Laschinger, & Wong, 2006; Gregory, Way, LeFort, Barrett, & Parfrey, 2007; Zeytinoglu et al., 2007), a 30% to 58% response rate was anticipated. Based on a conservative response rate estimate (35%), the recruitment of 1,151 nurses was viewed as feasible and appropriate to obtain a final sample of about 400 nurses. 4.1.4  Survey Response Rates The nursing workforce on some nursing units changed frequently during the data  collection period. When visiting the work units, I assessed the nurses’ eligibility, particularly for 66  the nurses with whom I did not have direct contact or who had not received the study materials. In these situations, follow-up occurred with the nursing unit manager to determine the nurses’ eligibility (e.g., current employment status). In some instances, the original staff lists provided by the nursing unit managers, to guide recruitment efforts, were not up-to-date. Ongoing discussions revealed that some nurses were not reachable for a variety of reasons (see Table 4.1). Table 4.1  Reasons for the Exclusion of Some Nurses Originally Identified as Eligible  Reason Terminated employment (e.g., resigned or retired) On leave (e.g., parental, education, sick, long-term disability, or gradual return to work) Recently initiated employment (e.g., new hire or completing orientation) or employment pending Self-identified or unit manager-identified as not being a regular member of the nursing unit (e.g., infrequent casual or float nurse on multiple units) TOTAL  Frequency (%) 25 (17.5) 34 (23.8) 31 (21.7) 53 (37.0) 143 (100%)  Nurses who initiated or resumed their employment (e.g., recently hired or returned from leave) were asked to participate if they self-identified as having adequate contact with their coworkers on the nursing unit to ensure accurate responses to the questionnaire. Some nurses who had recently initiated their employment were not reachable. The nurses employed on a casual basis were eligible to participate if they self-identified as working “regularly”12 on a nursing unit and were familiar with the other staff members of the unit. These atypical situations were considered on an individual basis to determine the nurses’ eligibility status. The nurses deemed unreachable were deleted from the recruitment list. Of the 879 participants that were initially identified as being eligible, a final sample of 736 met the inclusion criteria (see Figure 4.1).  12  Not formally defined because each situation was considered on an individual basis and participants were allowed the opportunity to determine whether they were sufficiently familiar with their colleagues on the unit to answer the questions. In general, “regular” employment meant a minimum of 4 shifts a month.  67  Figure 4.1  Flow Diagram of Participant Recruitment All possible participants (N = 879)  Total eligible participants (N = 736) Site A, n = 340 Site B, n = 396 Participants excluded (n = 143): • Terminated their employment before or during the study • On leave or retired • Infrequent employment on nursing unit (e.g., casual or float nurse) • Self-identified or unit manager identified as not being a regular member of the nursing unit.  Questionnaire returned (N = 606) 130 nurses did not return a completed questionnaire.  Completed questionnaires (N = 603)  3 partial questionnaires excluded from analyses  68  Site A consisted of 8 nursing units and 340 nurses eligible to participate. Site B consisted of 9 nursing units and 396 nurses eligible to participate.13 Table 4.2 illustrates the response rates of the potential participants. A total of 606 eligible nurses (Site A = 284 and Site B = 322) returned completed surveys. The gross completion rate, considering the 606 completed surveys, of the 879 participants identified by the managers as actively working as regulated nurses on the selected nursing units, was 68.9%. However, of the original potential participants, 143 (16.3%) were excluded because they were not actively employed on the nursing unit during the data collection period. Accordingly, a more reasonable calculated response rate is 81.9% (603 completed surveys used for data analyses from 736 eligible participants). On average, 83% of the nurses from each unit participated (range = 61% to 97%). Table 4.2  Survey Response Rates  Type of rate  Comparison  Numerator/ denominator 606/882  Gross response rate All completions/All possible participants a Most reasonable response All completions/Total eligible participants 606/736 rate (liberal) b Most reasonable response All completions used for analyses /Total 603/736 rate (conservative) eligible participants a Excludes 143 participants that were not reachable due to inactive employment on the nursing unit.  Rate (%) 68.7% 82.3% 81.9%  b  Excludes 3 partially completed questionnaires with less than 55% of the questions answered. Site A, n = 282 and Site B, n = 321.  4.2  Data Collection Process The data were collected through a self-administered questionnaire that was distributed  to the nurses on their unit of primary employment. The following section describes the distribution of the questionnaire based on a modified version of Dillman’s (2000) Tailored Design Method, using multiple points of contact combined with a foot-in-the door approach. Included in this section is a brief description of the pilot test of the study questionnaire.  13  In some instances, the nurses worked equal amounts of time on more than one unit (e.g., hired by one cost centre for logistical purposes but worked on another unit(s) on a regular basis). Each situation was considered on an individual basis to determine the appropriate nursing unit to determine response rates and for data analysis.  69  4.2.1  Modified Tailored Design Method Because the participant completion rate per nursing unit was critical to the success of  the project (e.g., calculating actual diversity), and given the nature of the work schedules of the nurses, various strategies were used to optimize the response rates and to address the various reasons for nonresponse (Dillman, 2000). Some of the reasons why people do not respond to questionnaires can be explained by social exchange theory, which forms the basis of the Tailored Design Method (TDM) (Dillman, 2000). Dillman proposed that participants are motivated by considering the rewards of responding to a survey with the costs associated with participating. Embedded in the TDM are attempts to provide rewards by offering appreciation, providing positive regard, and having a questionnaire that captures the interest of potential participants. The costs of participating in the survey can be reduced by making the task appear manageable, preventing embarrassment, and eliminating any direct monetary expenses (Dillman, 2000). The element of trust between the participant and the researcher is another key aspect of the TDM. Strategies such as establishing legitimacy by identifying with a known organization and providing an advanced token of appreciation are ways that trust can be achieved (Dillman, 2000). Although attending to issues of reward, cost, and trust can facilitate the achievement of response rates ranging from 58% to 92%, researchers must also tend to the detailed organization of survey administration to ensure that questionnaires and follow-ups are received in a timely and accurate manner (Dillman, 2000). To achieve this, Dillman (2000) recommended five necessary elements for achieving high response rates: (a) respondent-friendly questionnaires, (b) a minimum of five points of contact during a 3-week period (i.e., a prenotification letter, questionnaire package, thank-you reminder postcard, replacement questionnaire package, and a final reminder using special procedures such as certified mail), (c) return envelopes with paid postage, (d) personalized correspondence, and (e) prepaid tokens of appreciation. These five elements are the structural features of the TDM that facilitate potential participants’ understandings of what is being requested, and provide several opportunities to motivate participants to respond. Past survey research using the TDM has reported an average response rate of 74% (Dillman, 2000).  70  4.2.2  Application of the Tailored Design Method Once the hospitals were selected, members of the leadership groups (i.e., hospital  directors, program directors, and nursing unit managers) were interviewed to determine the most feasible methods of sampling the nursing work units and of distributing the self-administered questionnaires (e.g., in-person during departmental meetings, mailout through human resources, or via e-mail). To arrange convenient times for the data collection and to modify the distribution process to accommodate each nursing unit, I liaised with the managers of each nursing unit on a regular basis. For various reasons, it was determined that the most feasible method of distributing the data collection materials was to affix the materials to the nurses’ paycheque statements that were distributed through the human resources department every two weeks and delivered to nurses on their respective nursing units (e.g., use of staff mailboxes). The leadership groups believed that it was important to use both electronic and paper modes of communication as means of contacting all the eligible nurses.14 Numerous strategies from the TDM were used to incorporate the principles of reward, cost, and trust. To schedule the distribution of the research materials with the organizations’ payroll systems, a minimum of six points of contact with the potential study participants were completed during a 7- to 11-week period (see Figure 4.2). At both sites, the start date for distribution of the research materials was staggered over two weeks; four units began the study at the same time and then, two weeks later, the remaining nursing units received their research materials. All participants were assigned the same end date to return their completed questionnaires; they had a minimum of 50 days to a maximum of 78 days to complete the survey (on average, the nurses had 64 days to complete the questionnaire).  14  Most of the study correspondence was distributed by e-mail in addition to paper and online. A web-based version of the questionnaire was provided. All electronic correspondence was sent by the nursing unit managers, and an online survey management company, Zoomerang, was used.  71  Figure 4.2  Data Collection Process Replacement questionnaire package • Same content as initial package • Received by non-respondents  Pre-notification letter Thank-you reminder letter  Posters on units  Final thankyou (e-mail) and $2 gift certificate  Contact #1, Wk 0 | Contact #2, Wk 2 | Contact #3, Wk 4 | Contact #4, Wk 6 | Contact #5, Wk 8 | Contact #6, Wk 10 | Wk 11 and beyond  Reminder for early bird draw Questionnaire package • Cover letter • Additional consent information • Online survey instructions • Questionnaire • Self-addressed return envelope with paid postage • Draw entry form  Site A Final Reminder (non-respondents) Site B Final reminder and thank-you letter  Researcher visited units, attended meetings, and distributed gift certificates.  72  For each nursing unit, six in-hospital mailings occurred. A one page prenotification letter enclosed in a personalized envelope was the first piece of correspondence mailed two weeks before the questionnaire distribution. This preletter introduced the study, emphasized the importance of the survey for the health region, and indicated that a questionnaire would arrive in two weeks’ time and that the nurse’s response would be greatly appreciated. An influential person within the organization (i.e., the hospital director at Site A and program directors at Site B) and representatives of the nurses’ union (i.e., the president of the British Columbia Nurses’ Union and the secretary-business manager of the Hospital Employees’ Union) signed the introductory letter. During the distribution of the introductory letter, study posters were placed in central locations on the nursing units. When appropriate to do so, staff meetings were attended as another way of introducing the study, answering questions, and inviting staff participation. Two weeks following the prenotification letter, the questionnaire package was mailed to all eligible participants. Each questionnaire package included a cover letter explaining the study, additional consent information, directions about how to access the online survey, a paper version of the questionnaire with instructions for its return, a prize draw entry form and envelope, and a stamped, self-addressed envelope for return of the completed questionnaire to the School of Nursing, University of British Columbia. The cover letter was written on agency letterhead and conveyed important information about the study. To promote candid responses, the nurses were assured of complete confidentiality and anonymity. The 15-page questionnaire was presented in a respondent-friendly format that could be completed in 20 to 30 minutes. At the same time that the questionnaire was distributed, each nursing unit received a file containing extra questionnaire packages and a draw box for prize draw entry forms. As well, two formal meeting times were established for each nursing unit to launch the study and to offer food, as a way of showing positive regard. With follow-up contacts, Dillman (2000) indicated that response rates will usually be 20 to 40 percentage points higher than those normally attained. Two weeks following the distribution of the questionnaire, a personalized thank-you and reminder letter was distributed to all nurses in the target sample. In this letter, people who had already returned their questionnaire were thanked and those who had not were asked to do so as soon as possible. Two weeks following this letter, a flyer reminding the nurses of the early bird prize deadline was distributed 73  to all nonrespondents. Because of the staggered start dates, nurses in the second group at Site A did not receive a paper version of the flyer; instead, it was e-mailed and conveniently posted on the nursing units. As the prize draw entry forms were received, gift certificates were distributed to respondents on the nursing units. Six weeks after the distribution of the questionnaire, a personalized replacement questionnaire package was mailed to all nonrespondents. This package contained the same study materials included in the initial distribution with minor modifications made to the cover letter. Two weeks following the replacement questionnaire package, a final letter with a reminder of the final deadline was mailed. At Site A, this letter was followed with e-mail correspondence thanking all the participants. At Site B, this letter was written as a thank-you reminder letter and sent to all nurses in the sample. Following the final deadline and receipt of all survey packages, outstanding gift certificates were mailed to nurses who completed and returned the questionnaire. The entry forms were used as a means of acquiring mailing addresses for the gift certificate. This distribution process was used in combination with a foot-in-the-door approach (Dillman, 2000). This approach involved a brief conversation with the nurses inviting them to participate and showing the contents of the questionnaire package, which was meant to increase the perceived salience of the research, to establish a sense of value in them participating, and to increase the likelihood of storing the request in long-term memory (Dillman, 2000). Frequent visits (one to three times a week for the duration of the study) were conducted on each nursing unit during the data collection period to inform the potential participants of the study, to answer their questions, and to encourage their participation. Approximately 80% to 95% of the nurses, on each nursing unit, were approached in person. The nurses that completed and returned the questionnaire were eligible for several small individual incentives, including an early bird draw prize (one per nursing unit), a bonus prize at the end of the data collection period, and a $2 gift certificate from a local coffee supplier. A grand prize draw ($360 or the equivalent of the nursing registration fee with the provincial regulatory body) was also awarded to one individual. Finally, a group reward was given to the nursing unit with the greatest proportion of nurses participating. Nurses could enter their name for the incentives by completing a detachable random draw entry card. The participants were 74  instructed to put their name and contact information on the entry form and to place the card in a smaller sealed envelope. Participants could either return the envelope to the researcher by placing the envelope in a draw box conveniently located on the nursing unit or they could place it in the larger envelope with the completed questionnaire to be mailed to the School of Nursing. 4.2.3  Pretesting Pretesting was conducted to evaluate the study questionnaire (Bourgque & Fielder,  2003; Dillman, 2000). Some of the researcher’s colleagues who were not involved with its development reviewed the study questionnaire. Feedback was provided on response categories for scalar questions, clarity of instructions, and questionnaire aesthetics. Potential respondents at a hospital site not involved in the study were asked to respond to the questionnaire to obtain an understanding of how each question was interpreted and whether the intent of each question was realized. Information about the length of time taken to complete the questionnaire was obtained. Finally, feedback was provided regarding the questions that were likely to be of most interest to the participants, the quality of the information presented in the cover letter and poster, and the clarity of the questionnaire instructions. Planned data collection protocols were not tested; however, they were discussed at length with the unit managers and program directors to ensure their success.  4.3  Operationalization of Study Constructs The 15-page study questionnaire included numerous self-report items about the  constructs of interest: relational diversity (the derived exogenous variables), interpersonal conflict (the mediator variables), and burnout (the endogenous variables). A thorough review of instruments used to measure the constructs was carried out to determine their psychometric properties and feasibility of use. Table 4.3 provides an overview of the study constructs used in the structural equation modelling component of the analysis, the associated instruments or items, and the model in which the study constructs were included. The final questionnaire consisted of 138 self-report items with Likert-type responses (see Appendix B).  75  Table 4.3  List of Scales/Items in Final Study Questionnaire  Study constructs  Observed indicator (Questionnaire subscale and items)  EXOGENOUS VARIABLES – RELATIONAL DIVERSITY D-score calculated based on responses to the question: ACTUAL AGE In what year were you born? DIVERSITY DIVERSITY  D-score calculated based on responses to the question: What is your highest educational qualification in nursing?  ACTUAL ETHNIC/RACIAL  D-score calculated based on responses to the question: Are you . . . [list of responses] (for examples see Appendix B)  ACTUAL EDUCATIONAL  DIVERSITY  ACTUAL WORK  D-score calculated based on summed responses to the questions below.  VALUES DIVERSITY  Contemporary Work Values Scale  I expect work to be a meaningful and fulfilling part of my life. When working, I have high expectations of receiving both intrinsic and extrinsic rewards. Work provides a channel for expressing myself and my opinions. I need to be listened to by my superiors; work should be a two-way communication process. Work is worth doing only when it makes a meaningful contribution to society. I would like to work less in order to have more free time for personal interests. My input should be considered before decisions are made that affect my work situation. I desire work that provides opportunities for personal growth and allows me to “feel good inside.” I want to have control over my work assignments and how work tasks are done. Work has to be meaningful for me to do it well. It is important to me that my job provides opportunities to strengthen my abilities and talents. A worker should have some direct “say” in nursing unit operations. Being held in high regard by others is important to me. I am very concerned that I receive personal satisfaction from my work. Responsibility for high-quality patient care should be placed upon workers and not solely on managers. Work provides many opportunities for “personal growth” experiences. I enjoy work assignments that are challenging and require extensive use of thought processes. Only when it earns me self-respect is my work worthwhile. Work assignments should provide sufficient rewards for me; in other words, I would not accept just any job unless I have to. Work is beneficial in helping me to become a “whole” person. Work has value only because it is strictly a means to an end. I want more say over what will be assigned to me and how it is to be completed. I must be given a high degree of freedom to accomplish work in the best way possible.  76  Study constructs  PERCEIVED AGE  Observed indicator (Questionnaire subscale and items) Work contributes to my understanding and development of my character and capabilities. Work should provide me with a high degree of self-satisfaction or selffulfillment. I accept total responsibility for the successful completion of my work. I wish I could find interesting work. I want to be informed about the activities and plans of my nursing unit. I seek work experiences that help me expand and use my potential to the fullest extent possible. I would like variety in my work. Work provides individuals with an opportunity to “grow” and realize their full potential. I seek various emotional and psychological rewards from working in addition to my pay cheque. A person can effectively integrate work and other interests. Work should be an extension of one’s lifestyle and not merely a means to obtain subsistence. A need exists for more openness and better communication in work relationships. In my nursing unit, the other nurses are similar to me in terms of their age.  DIVERSITY  PERCEIVED EDUCATIONAL DIVERSITY  PERCEIVED ETHNIC/RACIAL  In my nursing unit, the other nurses are similar to me in terms of their educational background in nursing (e.g., diploma or degree). In my nursing unit, the other nurses are similar to me in terms of their ethnicity or culture.  DIVERSITY  PERCEIVED WORK VALUES DIVERSITY  In my nursing unit, the other nurses are similar to me in terms of their work ethic (values).  Perceived Work Values Diversity Scale  In my nursing unit, the other nurses are similar to me in terms of the principles that guide their work. In my nursing unit, the other nurses are similar to me in terms of their attitudes about work. In my nursing unit, the other nurses are similar to me in terms of their beliefs about work. MEDIATOR VARIABLES – INTERPERSONAL CONFLICT RELATIONSHIP CONFLICT Intragroup Conflict Relationship Subscale  Individual Conflict Relationship Subscale  How much friction is there among members in your nursing unit? How much are personality clashes evident among members in your nursing unit? How much tension is there among members in your nursing unit? How much rivalry is there among members in your nursing unit? How much anger is there among members in your nursing unit? How much friction is there between you and your coworkers? How much are personality clashes evident between you and your coworkers? How much tension is there between you and your coworkers? How often do you get angry with your coworkers?  77  Study constructs TASK CONFLICT Intragroup Conflict Task Subscale  Individual Conflict Task Subscale  Observed indicator (Questionnaire subscale and items) How often do members in your nursing unit disagree about the work being done? How frequently are there conflicts about work ideas among members in your nursing unit? How much conflict about the work you do is there among members in your nursing unit? To what extent are there differences of opinion among members in your nursing unit? To what degree do you and your coworkers have diverging opinions about the work being done? How much conflict about work ideas exists between you and your coworkers? How often do you and your coworkers disagree about what things should be done? To what extent do you and your coworkers have disagreements about work?  PROCESS CONFLICT Intragroup Conflict Process Subscale  How often do members in your nursing unit disagree about who should do what? How frequently do members in your nursing unit disagree about the way to complete a task? How much conflict is there about delegation of tasks among members in your nursing unit? Individual Conflict How often do you disagree with your coworkers about who should do what? Process Subscale How frequently do you disagree with your coworkers about the way to complete a task? How much conflict do you have with your coworkers about delegation of tasks on your nursing unit? ENDOGENOUS VARIABLES – BURNOUT 15 Maslach Burnout Inventory Emotional Drained Exhaustion Used up Fatigued Work strain Burned out Frustrated Work hard People stressful End of rope Impersonal Depersonalization Callous Hardening Not care Patients blamed  15  Reproduction of items was prohibited without the publisher’s written consent.  78  Study constructs Personal Accomplishment  Cynicism  4.3.1  Observed indicator (Questionnaire subscale and items) Understand patients Deal with problems Positively influence Energetic Create atmosphere Exhilarated Accomplish Deal calmly Less interested Less enthusiastic Not be bothered Cynical Doubt  Exogenous Variable: Relational Diversity Currently, there are no established instruments to measure relational diversity in  nursing; as a result, I drew heavily from researchers in the field of organizational behaviour. Two approaches were used to operationalize the relational diversity construct (Riordan, 2000). First, the Euclidean distance score approach was used to measure relational diversity. The second approach used was the perceptual approach, which measures how different individuals think they are from others in their workgroup (or similar) on specific attributes (Riordan, 2000). The four relational diversity attributes of interest for this study were: age, education, ethnicity/race, and work values. Each attribute was treated as a distinct theoretical concept and analyzed separately. 4.3.1.1  Actual Approach to the Measurement of Relational Diversity The most common measure of examining relational diversity, from an objective  standpoint, is the Euclidean distance measure (D-score). This approach provides a measure of an individual’s “actual” difference from (or similarity to) other workgroup members. Relational diversity measured with the D-score approach is usually referred to as actual relational diversity (herein referred to as actual diversity). Computationally, the D-score “is the square root of the average squared distance of an individual relative to all other members of the group” (Liao, Joshi, & Chuang, 2004, p. 982) (see Equation 4.1).  79  Equation 4.1 Euclidean Distance Measure n   2  1 n − 1∑ (S i − S j )  j =1    1  2  (Tsui, Egan, & O'Reilly III, 1992) The measure takes “the square root of the summed squared differences between an individual Si’s value on a specific demographic variable and the value on the same variable for every other individual Sj in the sample for the work unit, divided by the total number of respondents in the work unit” (Tsui et al., 1992, p. 562) minus the focal individual (n - 1) (Tsui & Gutek, 1999). Using the D-score to measure relational diversity at the individual level of analysis allows for consideration of the focal individual’s score on a specific attribute and all other workgroup members’ scores on the same attribute (Tsui et al., 1992). For the purposes of this study, all D-scores were scaled in such a way that larger values refer to greater individual diversity (difference) on a specific attribute (e.g., focal individuals with higher age D-scores are more different from others within the workgroup than those with lower age D-scores) (Tsui et al., 1992; Wagner, Pfeffer, & O'Reilly III, 1984). The D-score has been used reliably by other researchers; however, it is not without several shortcomings: (a) fails to account for any effects beyond the linear plane (e.g., quadratic functions), (b) measures only magnitude, rather than directional effects, (c) treats nominal classifications as if they were interval data (e.g., each ethnic classification was thought to be equally distant from each other), and (d) ignores the possibly that the separate components of the Si – Sj score (i.e., focal individual’s score for a given attribute and all other members’ scores on the same attribute) may disproportionately contribute to the prediction of individual outcomes (Clark & Ostroff, 2003; Edwards, 1994; Riordan, 1997, 2000; Wagner et al., 1984). Several approaches for operationalizing relational diversity (e.g., polynomial regression) have been described thoroughly elsewhere (Edwards, 1994; Edwards & Parry, 1993; Riordan & Holliday Wayne, 2008). The questionnaire included several questions (see Table 4.3) about the four attributes of interest (age, education, ethnicity/race, and work values), which were used to calculate D-scores for each individual. Each diversity variable was treated as a distinct theoretical concept. 80  Age diversity simply refers to differences in years of age between an individual and other workgroup members. To measure their age, the respondents were asked to report their year of birth which was subtracted from 2007. As an attribute that is not easily observable, education reflects an individual’s cognitive ability, knowledge, training, and skill (Liao, Chuang, & Joshi, 2008). In the current study, educational diversity refers to differences in levels of nursing education, namely diploma or baccalaureate degree preparation, between an individual and other workgroup members. In British Columbia, both registered nurses and licensed practical nurses may have earned a diploma before entering practice. Some registered nurses may initially graduate with a diploma to enter practice as a registered psychiatric nurse before earning their credentials as a registered nurse. No baccalaureate degree education exists for licensed practical nurses. Education was assessed through responses to the question, “What is your highest educational qualification in nursing?” Ethnic/racial diversity refers to individual differences relative to other workgroup members based on ethnicity or “race,” a multidimensional and dynamic construct that is in a constant state of flux (Statistics Canada, 2003). Ethnicity is thus not a fixed label, but is socially constructed and refers to a sense of belonging and group identity (Gerrish, 2000). Some common aspects comprising ethnicity are race, origin or ancestry, identity, language, and religion; however, other more subtle dimensions such as culture, the arts, customs, and beliefs may also be viewed as informing one’s sense of ethnicity (Statistics Canada, 2003). Ethnicity/race in the current study is viewed as a multidimensional and dynamic construct, referring to the sharing of common and subtle features (e.g., culture (which includes shared origin, shared genetic characteristics, and shared language), religion, cultural traditions, and skin colour) (Ford & Kelly, 2005). Ethnicity/race was measured by asking the participants how they self identified (“Are you . . .” [a list of ethnic or cultural groups was provided]). Work values diversity refers to differences in a constellation of attitudes and beliefs pertaining toward work-related activity in general and the work environment (McNeese-Smith & Crook, 2003; Miller, Woehr, & Hudspeth, 2002; Smola & Sutton, 2002). For the current study, the following definition was adopted: “Work values are the evaluative standards relating to work or the work environment by which individuals discern what is ‘right’ or assess the importance of 81  preferences” (Dose, 1997 as cited by Smola & Sutton, 2002, p. 366). Work values are often viewed as an attitudinal construct that determines an individual’s personal norms, preferences or choices, and behaviour related to work and the work environment (Verplanken, 2004); thus, the work values held by individuals determine their work attitudes, work standards, and work ethic (e.g., preferences and behaviour). Such values may also influence what members of a workgroup think the group’s task, goal, and mission should be (Jehn, Northcraft, & Neale, 1999). Work values diversity was measured with the 35-item Contemporary Work Values (CWV) Scale (Wayne, 1989) (see Table 4.3). According to Wayne, work values refer to “the usefulness, importance, or general worth that a person assigns to some behaviour or conception of work (e.g., physical effort and length of time on task/job) and nonwork activities (e.g., leisure, benefits, and rewards)” (p. 793). Originally, this instrument was developed to identify a collection of “newer” work values that influence a person’s attitudes or orientation toward work in general, as opposed to a specific job. These new, or contemporary, work values differ from traditional work values (e.g., the Protestant Ethic) in that they include a collection of principles of conduct and values that place less emphasis on dependence and commitment to one’s work, obedience, and respect for authority combined with a desire for more work-life balance (Wayne, 1989). An in-depth review of the literature about work values was undertaken by Wayne (1989) to develop questions pertaining to the Protestant Ethic and contemporary work values. Before Wayne undertook pilot testing, the questions were reviewed by several expert panels to establish content validity and to assess the readability of the instrument. Based on the findings from the pilot testing, the final instrument consisted of 111 questions, of which 35 captured contemporary work values. This instrument uses a 4-point Likert scale: strongly agree (4), agree (3), disagree (2), and strongly disagree (1). The 111-item questionnaire was then tested on a sample of 688 individuals (Wayne, 1989). The 35-item CWV scale achieved a Cronbach’s alpha of 0.91 and test-retest reliability of 0.74. Also, paired t-test, item analysis, and discriminate function analyses were conducted indicating that the CWV instrument distinctly measures one particular type of work values. Higher item and total scores reflect a stronger contemporary work values orientation (Wayne, 1989). Further details about the reliability and validity of this instrument, and its development, are described in detail by Wayne (1989). In the current study, the items were summed to create a total score. The average total score, which takes into consideration the 82  number of items answered by each respondent (i.e., some respondents had missing data), was used to calculate the D-score for participants to determine actual work values diversity. The D-score has been computed reliably for observed variables that are categorical and continuous (Liao et al., 2004; Riordan, 1997). In the current study, age and work values were treated as continuous variables. For example, assume Person A (focal individual) is 51 years old and works in a group of four other individuals (age in years for Person B = 36, C = 51, D = 31, and E = 37). The age difference between Person A (focal individual) and Person B, who is 36 years old, is 15. The squared distance between A and B is 225. The age difference and subsequent squared distances for the remaining members relative to the focal individual is as follows: C = -1 (1), D = 20 (400), and E = 14 (196). The squared distances for all members within the group are summed (225 + 1 + 400 + 196 = 822) and then divided by the total number of coworkers in the group minus the focal individual (5 - 1) (822/4 = 205.5). The square root of 205.5 = 14.3, which indicates the focal individual’s age relative to other members of the workgroup (example adapted from Liao et al., 2004). The range of scores for the continuous variables varied depending on the degree of difference. In the current study, the ethnicity/race and highest level of nursing education variables were treated as categorical variables for the calculation of the D-scores. The scores were computed by comparing the focal individual’s ethnicity/race, for example, with all the other ethnic/race backgrounds represented within the workgroup (Tsui et al., 1992). If two workgroup members belonged to the same ethnicity/race the value for Si – Sj was assigned a score of 0. A score of 1 was assigned to focal individuals if they belonged to a different ethnicity/race. For example, a “white” individual (focal individual) in a workgroup with one other “white” member and three others representing different ethnic/race groups (e.g., Chinese, Black, and South Asian), is assigned a score of 0 for being the same as the other “white” member and three scores of 1 for being different from each of the three workgroup members that are not “white” (Tsui et al., 1992). Next, the squared distance is calculated for each focal individual relative to all members in the group and then summed. In this example, the sum squared distances equal 3. Next, these squared distances are divided by n-1, which in this case is 3/4 = 0.75, and then the square root of the result is calculated. The square root of the final number indicates the focal individual’s ethnic/racial diversity relative to the workgroup (Liao et al., 2004). The focal 83  individual, who identified as “white,” receives a D-score of 0.87. The three individuals whose ethnicity/race is not “white” receives a D-score of 1.00, which would indicate that they are the sole members of an ethnic/race group (Tsui et al., 1992). The D-scores for categorical variables range from 0 to 1, with higher scores indicating greater differences between an individual and other workgroup members on a specific attribute (Riordan & Shore, 1997; Tsui et al., 1992). 4.3.1.2  Perceptual Approach to the Measurement of Relational Diversity The perceptual approach to measuring relational diversity represents individuals’  perceptions of how different they are from (or similar to) other workgroup members (Riordan, 2000). Relational diversity, measured from the perceptual approach, is referred to as perceived diversity. Based on the work of several researchers (Jehn et al., 1999; Kirchmeyer, 1995; Riordan, 1997), the participants were asked to indicate how similar they were to other members of their workgroups for each diversity attribute. One question was used to measure their perceived age, education, and ethnic/racial diversity (see Table 4.3). For the work values attribute, four items asked the individuals how similar they were to others in their workgroup regarding their work-related values, beliefs, and goals (Perceived Work Values Diversity Scale) (Cronbach’s alpha = 0.85) (Jehn et al., 1999). A 6-point Likert scale with anchors at 1 (not at all similar) to 6 (very similar) was used. Explicit instructions were given to the participants to make comparisons between themselves and their nursing coworkers that worked regularly on their immediate nursing unit. Items were reverse scored so that higher scores reflected greater individual diversity on a specific attribute. 4.3.2  Mediating Variable: Interpersonal Conflict Based on Barki and Hartwick’s (2004) conceptualization, interpersonal conflict was  defined for my purpose as a “phenomenon that occurs between independent parties as they experience negative emotional reactions to perceived disagreements and interference with the attainment of their goals” (Barki & Hartwick, 2004, p. 234). In the current study, the presence and intensity of interpersonal conflict is viewed as individuals’ perceptions of conflict, formed by their perceptions of disagreement, negative emotion, and interference present in the situation (Barki & Hartwick, 2004). In the field of organizational behaviour, there are two common approaches to measuring relationship, task, and process conflict: (a) the individual’s perceptions 84  of conflict within the workgroup (Jehn, 1994; Jehn, Chadwick, & Thatcher, 1997; Jehn & Chatman, 2000; Pelled, 1996b) and (b) the individual’s reported involvement in conflict (Hobman, Bordia, & Gallois, 2003; Pelled, Xin, & Weiss, 2001). These measures were based on the original work of Rahim (1983) and were modified to suit the context of this study. For both scales a 5-point Likert scale with anchors at 1 (none) and 5 (a lot) were used. Higher scores indicated greater amounts of conflict. To measure individuals’ perceptions of conflict within the workgroup (herein referred to as the Intragroup Conflict Scale), the respondents were asked 12 questions about the extent of disagreement evident among the members of their primary unit of employment (see Table 4.3). To answer the conflict items, the respondents were instructed to refer to all nurses (RNs and LPNs) that regularly worked on their particular unit. The items were based on the work of several researchers (Jehn, 1994; Jehn et al., 1997; Jehn & Chatman, 2000; Pelled, 1996b), which were modified slightly to be consistent with the other employed conflict scale. The number of items used to measure relationship, task, and process conflict were, respectively, 5, 4, and 3 items. Reported Cronbrach’s alpha for the scales of relationship conflict (range from 0.81 to 0.94), task conflict (range from 0.78 to 0.94), and process conflict (range from 0.78 to 0.93) were within acceptable range (Jehn, 1995; Jehn, Chadwick, & Thatcher, 1997; Jehn & Mannix, 2001; Jehn et al., 1999; Pelled, Eisenhardt, & Xin, 1999). To measure individuals’ involvement in conflict (herein referred to as the Individual Conflict Scale), the respondents were asked 11 questions about the degree (“how much”) of conflict they had with their nursing coworkers on their primary unit of employment (see Table 4.3). To answer the conflict items, the respondents were instructed to refer to all nurses (RNs and LPNs) that regularly worked on their particular unit. The Individual Conflict Scale consisted of four items to measure relationship conflict, four items to measure task conflict, and three items to measure process conflict. The relationship and task items were based on the work of Pelled et al. (2001) and modified to parallel the Intragroup Conflict Scale. In consideration of the items used to measure intragroup process conflict, individual conflict process items were developed by the researcher. Reported reliability coefficients for the relationship (α = 0.79) and task (α = 0.79) subscales (Pelled et al., 2001) were slightly lower than the subscales measuring intragroup conflict. 85  In reference to Barki and Hartwick’s (2004) typology of interpersonal conflict, the relationship conflict items for both scales assess negative emotions resulting from disagreements attributed to nonwork-related preferences, whereas task and process conflict measure disagreements about work. A 5-point Likert response format, anchored by 1 (none) and 5 (a lot), was used for all subscales; a higher value represents a greater amount of conflict. Although the conflict variables were highly correlated, discriminant validity testing indicated that each subscale measured a distinct aspect of conflict (Hobman et al., 2003; Jehn & Chatman, 2000; Jehn et al., 1999). Confirmatory factor analysis with oblique rotation supported a three-factor structure (Jehn, 1995; Jehn & Mannix, 2001); however, because reliability has not been established for the nursing population, a confirmatory factor analysis was completed as part of the current study (see Chapter 5). 4.3.3  Endogenous Variable: Burnout The predominant measure used by researchers to operationalize the construct of  burnout has been the original Maslach Burnout Inventory (MBI) (see Table 4.3). The MBI was originally designed to assess burnout among human service providers (e.g., nurses) who had direct relationships with clients, and has since been revised to measure burnout across occupations (e.g., nonhuman service fields and educational settings) and nationalities (Maslach, Jackson, & Leiter, 1996). It is important to note that the original MBI (1986) is equivalent to the MBI-Human Services Survey (HSS) sometimes referred to in the measurement literature. This study used the HSS scale, which consists of three subscales, to assess the frequency of emotional exhaustion (EE), depersonalization (DP), and a sense of diminished personal accomplishment (PA), along with the five cynicism (CY) items from the MBI-General Survey (GS). The Likert responses for included questions ranged from 0 (never) to 6 (every day). Examples of the burnout items for each subscale are:16 EE – “I feel like I’m at the end of my rope,” DP – “I feel I treat some recipients as if they were impersonal objects,” PA – “I feel I’m positively influencing other people’s lives through my work,” and CY – “I doubt the significance of my work.” Rather than  16  Reproduced with special permission of the publisher, CPP, Inc., Mountain View, CA 94043 from Maslach Burnout Inventory – Human Services Survey by Christina Maslach and Susan E. Jackson. Copyright 1986 by CPP, Inc. All rights reserved. Further reproduction was prohibited without the publisher’s written consent, which was not sought.  86  measuring the presence or absence of burnout, the levels of burnout experienced fall on a continuum. Individuals who experience higher scores in EE, DP, and CY and lower scores on PA, experience a higher degree of burnout. The scores for each subscale are norm referenced (see Table 4.4) and are typically considered separately; therefore, a composite score is not usually calculated (Maslach et al., 1996). To reflect the multidimensional structure of burnout (which evidence supports as containing conceptually distinct components), and to gain a more precise understanding of the relationships among the variables and the particular components of burnout (Cordes & Dougherty, 1993), each component was measured, analyzed, and reported separately. Instrument reliability and validity are well established (Maslach et al., 1996). The license to use the MBI and the sample items was obtained from Consulting Psychologists Press, Inc. Table 4.4  Normative Scores for the Maslach Burnout Inventory Subscales  MBI subscales  Mean  SD  Range of experienced burnout  Low Average High (middle third) (lower third) (upper third) EE 22.19 9.53 18 or less 19-26 27 or greater DP 7.12 5.22 5 or less 6-9 10 or greater PA 36.53 7.34 40 or greater 39-34 33 or less CY 1.80 1.24 1.00 or less 1.01-2.19 2.20 or greater Note. The EE, DP, and PA scores are based on a sample of medical workers (physicians and nurses), N = 1,104. The range of CY scores are based on a North American sample (N = 3,727), whereas the sample mean and SD are based on a Canadian sample of nurses, N = 1,257 (Maslach et al., 1996).  4.3.3.1  Reliability The MBI has demonstrated an adequate degree of internal consistency; inter-item  correlation values for the 22-item measure were within the recommended range of 0.30 and 0.70. The reported Cronbrach’s alpha for each subscale were: EE = 0.90, DP = 0.71, CY = 0.84, and PA = 0.71 (Maslach et al., 1996; Salanova et al., 2005). The test-retest reliability coefficients for the subscales range from 0.82 to 0.54 at one month to one year intervals, with the emotional exhaustion subscale demonstrating the highest degree of consistency (Maslach et al., 1996). 4.3.3.2  Content Validity In the 1970’s, findings from exploratory research were used to formulate ideas about  the attitudes and feelings that distinguished individuals experiencing burnout because of their 87  working with people. The initial version of the MBI-HSS included 47 items to assess both the intensity and frequency of each component of burnout (Maslach et al., 1996). This version was administered to a sample of human service providers (e.g., police officers, nurses, agency administrators, teachers, counsellors, social workers, probation officers, mental health workers, physicians, psychologists and psychiatrists, attorneys, and others). Following several exploratory and confirmatory factor analyses, the measure now consists of 22 items measuring the frequency of burnout across a wide range of occupations (e.g., MBI-GS for nonhuman service providers, the MBI-HSS for professionals in the human services, and the MBI-Educators Survey for those in the teaching profession) (Maslach et al., 1996). 4.3.3.3  Construct Validity Previous research regarding the MBI-HSS has demonstrated acceptable construct and  predictive validity (Maslach et al., 1996). This measure originally consisted of 47 items, and, through a series of studies using factor analysis, was pared down to the current 22-item version. The populations sampled represented a variety of health and service occupations dealing directly with people. Both men and women were sampled. Other demographics of the samples were not specified. A factor analysis on the 22 items of the MBI-HSS, using principal factoring and orthogonal rotation, produced a three-component structure (i.e., EE, DP, and PA) (Maslach et al., 1996). Studies have consistently found cross-loadings for item 12 (“energetic”) and item 16 (“people stressful”); nonetheless, these items have been retained, and the final three-component factor structure consists of nine items in the EE subscale, five items in the DP subscale, and eight items in the PA subscale (Maslach et al., 1996; Schaufeli & Van Dierendonck, 1993). The HSS-GS originated as a 28-item version and was reduced to 16 items through a series of regression analyses and factor analyses. Confirmatory factor analyses have resulted in a three-factor structure consisting of five items each for EE and CY, and six items for professional efficacy (Maslach et al., 1996). The MBI-HSS measuring the three aspects of burnout has been used in numerous nursing studies (Ilhan, Durukan, Taner, Maral, & Bumin, 2008; Laschinger & Leiter, 2006; Sahraian, Fazelzadeh, Mehdizadeh, & Toobaee, 2008). In many instances, the emotional exhaustion subscale has been used as the most prominent and robust measure of burnout among 88  nurses (Cho et al., 2006; Janssen, Jonge, & Bakker, 1999; Lang, 2007; Laschinger, Shamian, & Thomson, 2001; Stordeur, D'Hoore, & Vandenberghe, 2001). Although the authors of the burnout inventory report that confirmatory factor analysis confirmed a three-factor model of burnout (Maslach et al., 1996; Schaufeli & Van Dierendonck, 1993), others have demonstrated mixed results on the fit of this measurement structure (Beckstead, 2002; Lang, 2007; Salanova et al., 2005). Beckstead (2002) found that the hypothesized three-factor model, allowing items to load on only one latent factor, did not fit the observed data, and subsequently recommended four different measurement models be tested in future structural equation modelling studies using the MBI. In an analysis of the burnout inventory with all four subscales (i.e., EE, DP, CY, and PA), cynicism and depersonalization were both found to be distinct manifestations of mental distancing; consequently, Salanova et al. (2005) recommended including cynicism in addition to the three traditional subscales when studying human services. 4.3.3.4  Convergent and Discriminant Validity Maslach et al. (1996) reported substantial evidence demonstrating the convergent  validity of the MBI-HSS. Further evidence has been obtained to distinguish the burnout inventory from other psychological constructs that might be confounded with burnout (Maslach et al., 1996; Schaufeli & Van Dierendonck, 1993). A negative correlation between the subscales of the MBI-HSS and job satisfaction (ranging from 0.40 to 0.52) has been documented (Beckstead, 2002; Maslach et al., 1996). Burnout subscales for the HHS and GS have been differentiated from anxiety, depression, mental and physical strain, organizational commitment, job involvement, and occupational stress (Cordes & Dougherty, 1993; Maslach et al., 1996). Correlations between the MBI and the Crowne-Marlowe Social Desirability Scale were reported as not statistically significant (Maslach et al., 1996).  4.4  Data Analysis Procedures This section provides an overview of the data analysis procedures carried out to test  the direct and mediating relationships between the exogenous variables (the relational diversity variables), the mediator variables (intragroup conflict and individual conflict), and the endogenous variables (emotional exhaustion, depersonalization, cynicism, and diminished personal accomplishment) while controlling for measurement error. Processes followed for data 89  screening and the handling of missing data are described. Next, the details of a two-step approach taken to conduct the statistical analysis of the data using structural equation modelling (SEM) are provided. This approach uses confirmatory factor analysis (CFA) to test and establish the validity of the measurement model before testing the structural model (Schumacker & Lomax, 2004). Included is a description of the estimation used for modelling with ordered categorical (ordinal) data, the criteria used for evaluating model fit, the process applied for model respecification, and the statistical method employed to determine the relative importance of the exogenous variables (the Pratt Index) (Thomas, Hughes, & Zumbo, 1998). Last, the procedures undertaken to test the mediation models are explained. The data analyses were conducted using the SPSS 12.0 for Windows and Mplus version 5.1 software programs. For all statistical procedures, the level of significance was set at a minimum of p = 0.05 and corresponding 95% confidence intervals. 4.4.1  Data Preparation and Screening Before any data analysis was undertaken the raw data were screened for incorrect  responses, data entry errors, or missing responses. Using SPSS 12.0, the distributions of the demographic and study variables were examined by using frequency and simple cross-tabulations. The chi-squared statistic was used to examine differences in the employment and demographic characteristics of the Site A and Site B respondents. Differences between responses from Site A and Site B respondents for the study variables were explored through use of the Mann Whitney U test.17 Assumptions for multivariate analysis (i.e., normality, linearity, and homoscedasticity) were tested and the data were screened for the presence of outliers. Outliers in the data were examined to ensure that the data were entered correctly, that the outlier was a member of the intended sample population, and that the extreme values were within the acceptable range of the variables (Tabachnick & Fidell, 2007). If the outliers were from the intended sample population but represented more extreme values than a normal distribution for a given variable, they were retained for data analysis.  17  The appropriate statistic for bivariate analysis was determined by dividing the skewness value by the standard error of skewness. Values above or below ± 1.96 were considered significantly skewed (p = 0.05) and thus required a nonparametric test, such as the Mann Whitney U test (Munro, 2001).  90  4.4.2  Representation of Ordinal Variables According to Finney and Distefano (2006), a general rule of agreement among  researchers is that “when ordinal data are approximately normal and have a least five ordered categories that the ordered categorical data may be treated as if they were continuous without great distortion in the fit indices” (p. 276). At the same time, according to Brown (2006), The potential consequences of treating categorical variables as continuous variables in CFA are multifold, including that it can (1) produce attenuated estimates of the relationships (correlations) among indicators, especially when there are floor or ceiling effects; (2) lead to ‘pseudofactors’ that are artifacts of item difficulty or extremeness; and (3) produce incorrect test statistics and standards errors. (p. 387) For the data analysis undertaken in this study, the distributions of the observed variables were examined to determine whether the variables should be treated as categorical or continuous. The data were collected using Likert scales with four to seven ordered categories. Based on the univariate analyses, most of the distributions of the study variables were skewed (range from -7.50 to 14.13) and displayed some kurtosis (ranged from -4.01 to 12.96). These values were not within an acceptable range for CFA with maximum likelihood estimation (e.g., skewness < |2.0| and kurtosis < |7.0|) (Finney & DiStefano, 2006). Deviations from normality may lead to exaggerated chi-squared statistics and distorted fit indices, underestimated parameter estimates, and biased standards errors, which result in increased Type I error rates (Finney & DiStefano). Accordingly, the data were treated as severely non-normal and the indicators of all the observed variables were treated as ordered categorical (ordinal) for the analyses. For additional information about modelling with categorical data see Finney and DiStefano (2006) as well as Brown (2006). 4.4.3  Structural Equation Modelling A multi-step process was used to determine the extent to which the hypothesized  conceptual models (see Chapter 3) were actually consistent with the sample data (Schumacker & Lomax, 2004). The steps to SEM are an iterative process where problems or modifications determined at a later step may require modifications to earlier steps (Kline, 2005). The theoretical models were specified as structural models to ensure that the models were identified. 91  Next the data were collected, screened, and prepared for data analysis. Using the Mplus 5.1 software program, data analysis was carried out using CFA, exploratory factor analysis (EFA), and SEM techniques (Kline, 2005). Factor analysis techniques (CFA and EFA) were used to assess the relationships between the observed variables (i.e., indicators or scale items) and the latent variables or factors (referred to as the measurement model) before assessing the structural model (Schumacker & Lomax, 2004). All factor analyses were conducted first to establish the measurement models before specifying and testing the structural model (Schumacker & Lomax, 2004). During the model estimation step the following events occurred: (a) the model fit was evaluated; (b) the parameter estimates were inspected for direction, magnitude, and significance; and (c) alternative models were considered. If necessary, the models were respecified and evaluated accordingly (Kline, 2005). 4.4.3.1  Exploratory Factor Analysis In this study, EFA was used in instances when there was limited evidence available  about the factor structure of an instrument (Munro, 2001). EFAs were conducted based on the guidelines and recommendations provided by Tabachnick and Fidell (2007) and others (Brown, 2006; Fabrigar, Wegener, MacCallum, & Strahan, 1999). 4.4.3.2  Confirmatory Factor Analysis CFA was conducted to ensure the appropriate loading of the indicators on their  corresponding concepts and to determine the validity of the study measures. The metric of the latent variables was set to be the same as the marker or reference indicator, which was the observed variable with the highest reported parameter estimate (Brown, 2006). In the case of multidimensional constructs (e.g., burnout), the factors were allowed to covary. 4.4.3.3  Method of Estimation According to Brown (2006), when some of the indicators are ordered categorical  (ordinal) an alternative to maximum likelihood (ML) should be used. The best estimation method for categorical indicators and non-normal data was identified to be robust mean and variance adjusted weighted least squares (WLSMV), which is included in the Mplus 5.1 software package (Brown, 2006; Finney & DiStefano, 2006; Flora & Curran, 2004; Muthen & Muthen, 2007). The WLSMV estimator was used for all modelling. 92  The WLSMV estimator provides “weighted least square parameter estimates using a diagonal weight matrix (W) and standard errors and a mean- and variance-adjusted chi-squared test statistic that uses a full weight matrix” (Muthen & Muthen, 2007, p. 484). Thus, use of the WLSMV estimator with categorical data results in more reliable fit statistics, parameter estimates that are less biased, and reduced Type 1 error rates (Beauducel & Herzberg, 2006; Finney & DiStefano, 2006). When using the WLSMV estimator the W is not required to be positive definite (Brown, 2006). CFAs with ordinal indicators use a tetrachoric correlation matrix for binary indicators, a polychoric correlation matrix for polytomous indicators, and a polyserial correlation matrix for a combination of continuous and ordinal variables (Brown, 2006; Schumacker & Lomax, 2004). The degrees of freedom (df) for WLSMV are estimated in a manner that differs from maximum likelihood. According to Muthen and Muthen (2006, Jan 20), with the WLSMV estimation method “the chi-square and degrees of freedom are adjusted until a correct p-value is found.” Accordingly, when reporting the chi-squared statistic, it is the p-value that should be interpreted as opposed to the degrees of freedom. For more information see the online technical appendices of the Mplus User’s Guide (Muthen & Muthen, 2008). 4.4.4  Missing Data Returned surveys with substantial missing data on the variables central to the study  (n = 3) were excluded from the raw data file. Of the remaining cases (n = 603), the SPSS 12.1 software program was used to examine the missing responses for the demographic variables. Information from the demographic variables was used, when possible, to logically infer answers for missing responses on other demographic variables. The study variables were examined to determine the frequency and pattern of missing data using the Mplus 5.1 software program. To maintain a large sample size while at the same time minimizing the influence of missing data, procedures for handling missing data were determined after the raw data were screened. If less than 5% of the data were missing, and the missing patterns appeared to be random, then the cases were retained for analysis (Tabachnick & Fidell, 2007). The WLSMV estimation method uses pairwise deletion, that is, all cases are included and covariances are calculated using only available pairs of observations (Brown, 2006). Pairwise deletion uses “limited information” from pairs of variables and therefore uses all 93  individuals with observations on that pair (Muthen & Muthen, 2006, Feb 6). Although the use of pairwise deletion can result in a nonpositive definite covariance matrix and differing sample sizes for the CFAs and SEMs, Muthen and Muthen (2006, Feb 6) indicated that the use of pairwise deletion is why weighted least squares is more robust than maximum likelihood. 4.4.5  Model Evaluation Model fit for both the CFA models and SEMs was evaluated using the chi-squared  statistics, residual correlation matrix, and global fit indices. Statistical non-significance of a chi-squared statistic (χ2) indicates that an observed (sample) matrix (S) and an implied (hypothesized) matrix (3) are similar and hence the difference between the two matrices is minimal. Chi-squared values close to zero indicate perfect fit between the matrices.18 The values of the differences between the observed and implied matrices are displayed in a residual correlation matrix. When a specified theoretical model fits the sample data there is little difference between the implied and observed correlation matrices, hence the residual values in the residual correlation matrix are close to zero (Brown, 2006). To further assess model fit (and explore potential areas of misfit for specific variables), the overall pattern of residual correlations was inspected and the residual correlations were examined for values greater than |0.1| (Brown, 2006; Sawatzky, 2007). Based on Hu and Bentler (1999), the global fit indices used to assess the model fit with categorical data were the comparative fit index (CFI), the Tucker-Lewis index (TLI), and the root mean square error of approximation (RMSEA). Acceptable model fit was defined by the following criteria: CFI ≥ 0.95, TLI ≥ 0.95, and RMSEA ≤ 0.08 with an ideal value of 0.06 as being indicative of a well-fitting model. For the EFAs, the standardized root mean residual (SRMR) was considered with a value less than 0.08 being desirable (Brown, 2006; Hu & Bentler, 1999; Schumacker & Lomax, 2004). These widely accepted criteria for evaluating model fit were initially based on models using continuous variables (Beauducel & Herzberg, 2006). Beauducel and Herzber (2006) conducted a simulation study to evaluate the applicability  18  The chi-squared statistic may be statistically significant because it is sensitive to increased degrees of freedom, larger sample size (>200), and deviations from multivariate normality (Schumacker & Lomax, 2004).  94  of the traditional global fit indices with ordinal variables (two to six categories). They found that the SRMR was the same when using WLSMV and ML estimation, whereas the RMSEA was slightly larger for models estimated with WLSMV for four to six categories. For variables with two and three categories, the reverse was found for both the SRMR and RMSEA. The CFI values for WLSMV estimation, for variables with five or six categories, were the same as the CFI based on ML estimation; but, for variables with two and three categories, the CFI values were larger with WLSMV estimation. The TLI values for variables with two and three categories resulted in larger values with WLMSV estimation; however, for variables with five and six categories the TLI values were smaller with WLMSV estimation in comparison with ML estimation (Beauducel & Herzberg, 2006). For the current study, the criteria for model fit were applied in a cautious manner recognizing the need for further simulation studies exploring global fit indices and WLSMV estimation, particularly with variables that are skewed and kurtotic (Beauducel & Herzberg, 2006). Multiple indices were used in this study because they provided different information about model fit (i.e., absolute fit, effect of model complexity, and fit adjusting for model parsimony) and a conservative and reliable evaluation of the solution (Brown, 2006). Model evaluation and localized areas of strain (when one or more of the global fit indices were outside the acceptable range) were assessed using additional statistics: (a) the residual correlation matrix was examined, which provides specific information about the difference between the observed and implied matrices (values ≥ |0.10|), (b) the standardized residuals were examined, which can be interpreted as z scores (values ≥ |1.96| are statistically significant at p < 0.05), (c) the modification indices (MI) and the standardized expected parameter change (EPC) were examined, which reflect an approximation of how much the overall model χ2 would decrease and the parameter estimates would change if a given parameter was freely estimated (values ≥ 10.0), (d) direction, magnitude, and statistical significance of the parameter estimates (factor loading ≥ 0.40 and unstandardized parameter estimates/standard error ≥ |1.96|), and (e) R-squared of the factor loadings (R2 > 0.49) (Brown, 2006; Hu & Bentler, 1999; Schumacker & Lomax, 2004). Model respecification, where indicated, was guided by prior theory and evidence to add or remove parameters (Brown, 2006). Consideration was given to the original measurement 95  structure and substantive justification of the established instruments. In some instances, model fit may have improved with further respecification; however, attempts were made to keep the instruments as similar as possible to their original structure (e.g., the MBI). To assess the overall model fit between competing models, the chi-squared difference test (DIFFTEST) was used. This test compares the chi-squared statistics of two nested models to evaluate the significance of the difference between the models (Brown, 2006). CFAs estimated with WLSMV have their df calculated differently (for further information see Muthen and Muthen (2007)) from those estimated with ML. Accordingly, the DIFFTEST in Mplus 5.1 software was used to compare the chi-squared statistics of two nested models. The pattern and range of the residual correlations for the competing models were also examined for values greater than |0.10| (Brown, 2006; Sawatzky, 2007).  4.5  Additional Statistical Methods To test internal consistency reliability, researchers have traditionally used Cronbach’s  coefficient alpha. However, in the context of a CFA measurement model, Cronbach’s alpha does not provide a dependable estimate of multiple-item measures when items cross-load (Brown, 2006). A suggested alternative to estimate scale reliability has been the composite reliability score. For continuous variables, the composite reliability of each latent variable takes into consideration the parameter estimates and error variances of items measuring the construct (Brown, 2006). In factor analysis with ordered categorical outcomes the composite reliability score is not interpretable and would require use of item response theory techniques (Muthen & Muthen, 2006, May 4), which is beyond the scope of this dissertation. After model fit was determined, the Pratt Index was used to examine the proportion of variation accounted for by each construct in the model (Thomas et al., 1998). In other words, once good model fit was obtained for the structural models, the Pratt Index (d) was calculated to determine the relative importance of each of the perceived and actual diversity variables in explaining burnout. For additional information about the Pratt Index refer to Thomas, Hughes, and Zumbo (1998) and Zumbo (2007).  96  4.5.1  Testing the Mediation Models A mediating variable is the variable that provides a possible causal explanation as to  how or why an independent (exogenous) variable causes a dependent (endogenous) variable. In other words, following a causal sequence, an independent variable leads to a mediator variable which in turn leads to a dependent variable (Baron & Kenny, 1986; MacKinnon, 2008). Mediation models are concerned with the overall direct effect, direct effect, and indirect (mediated) effect. The relationship between an exogenous variable and an endogenous variable (without controlling for a mediator variable) is referred to as the overall direct effect (c) (Wu & Zumbo, 2008). Although Baron and Kenny (1986) identified a statistically significant overall direct effect as a required condition for mediation, more recently, researchers have argued against its necessity (MacKinnon, 2008; Wu & Zumbo, 2008). For example, the requirements that Baron and Kenny proposed do not allow for suppression or inconsistent mediation (MacKinnon, Krull, & Lockwood, 2000). The direct effect (c′) refers to the relationship between an exogenous variable and an endogenous variable, while controlling for a mediator variable (see Figure 4.3). The total indirect effect (ab) refers to the product of the indirect effect (a) between an exogenous variable on the mediator variable and the indirect effect (b) between a mediator variable on an endogenous variable. The quantity of the total indirect effect reflects how much a one unit change in an exogenous variable affects an endogenous variable indirectly through a mediator variable. Another effect often examined in mediation models is the total effect, which is the sum of the direct effect (c′) and the total indirect effect (ab) (MacKinnon, 2008).  97  Figure 4.3  Single Mediator Model  Mediator Variable (η2)  a (γ21)  ζ2 b (β12)  ζ1 Exogenous Variable (ξ1 )  c′ (γ11)  Endogenous Variable (η1)  Note. Notations in brackets represent the Greek symbols used to represent matrices and parameters for SEM.  (adapted from MacKinnon, 2008)  The various methods used in testing for mediation can be grouped into three categories: (a) the casual step based on testing hypotheses consistent with mediation (also referred to as the Baron and Kenny method), (b) the differences in coefficients (c – c′), and (c) the product of coefficients (ab) (MacKinnon, 2008). MacKinnon et al. (2002) conducted a simulation study to determine the Type I error rates and statistical power for the various methods used to test for mediation. They concluded that the most important conditions for mediation are that the indirect effects (a and b) and the product of these effects (ab) are statistically significant. Building on the work of Baron and Kenny (1986) and MacKinnon (2008) the following four-step procedure for testing for mediation effects within a SEM framework was followed: 1. The first step was to test the fit of the relevant model and to examine the overall direct effects for the actual and the perceived diversity variables (age, education, ethnicity/race, and work values) on burnout (EE, DP, and PA). This step is referred to as Condition 1. The next step 98  was to test the mediator model. Although, statistically significant parameter estimates for the overall direct effects may provide information as to whether a mediation effect is in place (Baron & Kenny, 1986), given recent work by others (MacKinnon, 2008; Wu & Zumbo, 2008), the mediator models were analyzed with all exogenous variables in the model regardless of their significance. If no statistically significant mediated effects were identified for a particular variable, then it was removed, the model was respecified, and subsequent analyses were conducted. 2. If the mediator model fit the data, then the indirect (mediating) effects of the actual and perceived diversity variables on the conflict latent variables and the conflict latent variables on the burnout latent variables were tested. The relationship between the exogenous variables (actual and perceived diversity in age, education, ethnicity/race, and work values) and the mediator variables (intragroup and individual relationship, task, and process conflict) are referred to as Condition 2. Condition 3 refers to the relationship between the mediator variables with the endogenous latent variables (EE, DP, and PA), while controlling for the exogenous variables. The relationships for both Conditions 2 and 3 must be statistically significant to establish that a mediation effect is in place (MacKinnon, 2008). 3. The total indirect effect (ab) was examined to determine the statistical significance of the mediation effect (whether an observed effect was due to chance). In a single mediator model, the value for the total indirect effect is the effect of the exogenous variable on the endogenous variable that is indirect, through the mediating variable. The total indirect effect must be statistically significant to conclude that a mediation effect exists (Condition 4) (MacKinnon, 2008). In Mplus, significance tests for indirect effects are calculated using the parameter estimate of the indirect effect and its standard error to provide p-values and confidence intervals (MacKinnon, 2008). According to MacKinnon (2008), “Confidence intervals provide more information about a mediated effect because a range of possible values for the mediated effect are considered rather than one single value” (pp. 79-80). 4. To determine whether an effect was partially or completely mediated, the direct effect (c′) of an exogenous variable on an endogenous variable, while controlling for a mediator variable, is compared to the overall direct effect (c). To establish that a variable completely mediates a 99  relationship, the direct effect should be close to zero and nonsignificant (MacKinnon, 2008). In other words, once a mediator is included to explain the variation in an endogenous variable, the overall direct effect disappears and the mediation hypothesis is supported (Wu & Zumbo, 2008). If the indirect effect is significant and the direct effect is reduced (but not zero), then the relationship is said to be partially mediated (Wu & Zumbo, 2008). These steps in the analysis of a single mediator can be applied to a model with two or more mediators in combination with more than one exogenous and endogenous latent variables (MacKinnon, 2008). For additional information about mediator models with ordered categorical (ordinal) variables see MacKinnon (2008). Further statistical methods were used to determine the degree of mediation (effect size) (MacKinnon et al., 2000), that is, the degree to which the relationships between diversity and burnout were mediated by relationship and task conflict. This was calculated by dividing the total indirect effect (ab) by the total effect (and multiplied by 100%) to create a percentage.  4.6  Ethical Considerations Before conducting the study, ethical approval was obtained from the University of  British Columbia, Behavioural Research Ethics Board (see Appendix C) and the appropriate Health Authority (see Appendix D). The principles outlined in the Canadian Tri-Council Policy Statement for Research Involving Human Subjects were adhered to. Nurses agreeing to participate in this study were assured that their participation was voluntary, that they had the right to refuse to participate, and that their responses would remain anonymous and have no effect on their employment. Consent to participate was implied by completion of the questionnaire. The participants’ confidentiality was protected by ensuring that no identifying information was entered on the questionnaire. The participants were advised not to place their names on the questionnaires. To further maintain confidentiality, the participants were asked to return their completed questionnaires directly to the researcher, using the stamped return envelope. Completed study questionnaires were stored in a locked room in the School of Nursing. All data were treated as confidential and were only accessible to the researcher and the dissertation supervisor. Only aggregate data were reported. The nature of the study presented minimal risks to the participants. A contact number for the researchers was provided for those 100  who had questions about their participation or the content of the questionnaire. No known benefits were anticipated as a direct consequence of participating in the study; however, all participants had the option of having their name entered for a gift of funds equivalent to the cost of annual practicing registration with the College of Registered Nurses of British Columbia, as well as minor incentives (e.g., $2 coffee cards and draw prizes) that were distributed as tokens of appreciation for taking the time to complete the study questionnaire. The nurses who returned the questionnaire were given the option of receiving a brief report of the study results. To calculate the D-scores, the fourth page of the questionnaire was labelled with an identification letter and number code to indicate a particular work unit within a specific hospital. To assure anonymity and confidentiality of the online responses, the survey was distributed to the participants by generating a URL link (and not through the Zoomerang mail server), which was included in the study correspondence (e.g., letters, e-mail correspondence, and an online instruction sheet). The participants were informed that the data from their completed online questionnaires would be stored on the Zoomerang servers, or those of their agents, which reside outside of Canada. There was no personally identifiable information on the online questionnaire.  4.7  Chapter Summary This chapter has provided the details of the methods for this study, including  sampling, data collection, instrumentation, data analysis, and ethics. A total of 603 practicing nurses from two acute care hospitals in an urban setting completed the 138-item self-administered questionnaire. The data collection process took approximately 10 to 12 weeks in each hospital site and was guided by Dillman’s Tailored Design Method, which involved the use of multiple points of contact combined with a foot-in-the-door approach. Two approaches were used to operationalize the relational diversity construct (exogenous variable); the D-score to measure “actual” diversity and the perceptual approach to measure “perceived” diversity. The mediator variable, interpersonal conflict (relationship, task, and process) was measured using both the Intragroup Conflict Scale and the Individual Conflict Scale. The endogenous latent variable, burnout, was measured using the Maslach Burnout Inventory. At the individual level of analysis, structural equation modelling was used to model the main and mediating relationships between relational diversity, interpersonal conflict, and burnout while controlling for 101  measurement error. Ethical approval was obtained from the University of British Columbia Behavioural Research Ethics Board and the appropriate Health Authority.  102  5  FINDINGS This chapter provides the details of the preparation of the data prior to analysis and the  findings of the descriptive statistical procedures and confirmatory factor analyses (CFA). The first section of this chapter focuses on the preparation and screening of the data. Next, the employment and demographic characteristics of the sample are presented to provide an overview of the participants involved in the study. The third section presents the results of the CFA portion of the data analysis, which examined the measurement models of the latent variables. The findings are presented for the exogenous variables (actual and perceived relational diversity for age, education, ethnicity/race, and work values) first, followed by the hypothesized mediator variables (relationship, task, and process conflict), and then the exogenous variables (three aspects of burnout). Given the iterative nature of structural equation modelling, once the measurement models were established then the missingness of the study variables was examined and the descriptive statistical procedures completed. The final section of this chapter provides a summary of the descriptive statistics for all study variables.  5.1  Data Screening and Variable Transformation The data were screened for questionable response patterns (e.g., length of service as a  nurse shorter than length of service on a nursing unit), multiple responses, incorrect data entry and coding errors, and missing responses. Inconsistencies between years worked as a nurse, years worked on a nursing unit, and years worked at a hospital were visually inspected and corrected accordingly. Information from the demographic variables was used to deduce answers for some of the missing responses on other demographic variables. For example, if a respondent indicated his/her job title as “Staff Nurse – Registered Nurse,” the information was used to conclude that the type of licensure held, if missing, was “Registered Nurse.” To create an ethnicity/race variable with only one response, which was necessary to calculate the Euclidean distance scores (D-score), missing, implausible, and multiple responses about how the respondents perceived their ethnicity/race (“Are you . . . ” [list of responses such as White, Chinese, South Asian, Black, Filipino, see Appendix B]) were modified based on their answers to the questions about the ethnic/cultural background of their ancestors (“To which ethnic or cultural group did your ancestors belong?”), languages spoken at home, and country where they 103  completed their first/initial nursing education program. The “language spoken” variable was treated in a similar manner. For example, 113 (19%) participants reported that they spoke English and another language at home and their language was noted to be a language other than English. Approximately 5 participants reported speaking more than 2 languages and they were reassigned to the language most consistent with their self-identified ethnicity/race. For all Likert scale items, if the respondents selected more than one response (e.g., “2” and “3”) the data were entered as an average value (i.e., 2.5), to one decimal place. After data entry, all scale items with midpoints were reviewed and reassigned a whole number score (e.g., either a “2” or “3”) using random assignment software. Table 5.1 provides an overview of the number of items per scale that required rescaling for midpoint values. Table 5.1  Summary of Scale Items with More than One Response Variable  Contemporary Work Values Scale (35 items) Perceived Similarity Scale (14 items) Intragroup Conflict Scale (12 items) Individual Conflict Scale (11 items) Maslach Burnout Inventory (27 items)  Items frequency (%)  Respondents frequency (%)  Section A: 3, 6, 8, 9, 10, 14, 16, 17, 20, 22, 23, 26, 29, 31, 32, 34 Section C: 1, 4, 5, 6, 7, 10, 11, 13, 14  16 (45.7)  20 (3.3)  9 (64.3)  14 (2.3)  Section E: 1, 4, 8, 12  4 (33.3)  4 (0.7)  Section G: 1, 3, 5, 6, 7, 8, 9, 10, 11  9 (81.8)  18 (3.0)  Section B: 9, 10, 13, 14, 16, 19, 20, 22, 23, 26  10 (37.0)  13 (2.2)  Question/item number  Note. N = 603. Random assignment software was used for all scale items with midpoints to round up or down the value to a whole number.  5.2  Descriptive Statistics of the Sample This section provides an overview of the employment and demographic characteristics  of the sample. The sample (N = 603) consisted of 282 nurses (46.8%) from Site A and 321 nurses (53.2%) from Site B (see Table 5.2) (82% response rate, see Table 4.2). Of the entire sample, 456 (75.6%) were employed on medical, surgical, or medical/surgical combination nursing units and 147 (24.4%) were employed on non-medical/surgical nursing units (i.e., perinatal, paediatric, or neonatal intensive care units). Registered nurses represented 86.6% (n = 522) of the sample and licensed practical nurses (LPN) represented 13.4% (n = 81). The predominant job title of the respondents was “RN staff nurse” (n = 482, 79.9%). The respondents ranged in length of service as a nurse from 1 month to 45 years (M = 14.1 years; SD = 11.9), 104  their length of service at the current hospital ranged from 1 month to 37 years (M = 8.9 years; SD = 9.2), and length of service with the current nursing unit ranged from 1 month to 36 years (M = 6.9 years; SD = 8.0). Table 5.2 further delineates the length of service of the nurses in their profession as well as with their current units and hospitals. More than one half of the nurses (n = 331, 54.9%) were employed full-time, 26.0% were employed part-time (n = 157), and the remainder (n = 115, 19.1%) were employed on a casual or temporary basis.  Table 5.2  Employment Characteristics of the Respondents Characteristic  Site A frequency (%)  Site B frequency (%)  Total sample frequency (%)  213 (75.5)  269 (83.8)  482 (79.9)  13 (4.6)  14 (4.4)  27 (4.5)  8 (2.8)  5 (1.6)  13 (2.2)  48 (17.0)  33 (10.3)  81 (13.4)  Job Title Registered nurse (RN) Clinical resource nurse or patient care coordinator (RN) Clinical nurse educator or clinical nurse specialist (RN) Licensed practical nurse Years Worked as a Nursea  Between group comparison statistic 2 = 7.52 (df = 3)  2 = 11.88  Less than 1 year 1 to 2 years  29 (10.4) 45 (16.1)  26 (8.1) 32 (10.0)  55 (9.2) 77 (12.8)  3 to 5 years  31 (11.1)  39 (12.1)  70 (11.7)  6 to 10 years  52 (18.6)  47 (14.6)  99 (16.5)  11 to 15 years  27 (9.6)  37 (11.5)  64 (10.7)  16 to 20 years  24 (8.6)  27 (8.4)  51 (8.5)  21 to 25 years  21 (7.5)  29 (9.0)  50 (8.3)  26 to 30 years  22 (7.9)  40 (12.5)  62 (10.3)  Greater than 30 years  29 (10.4)  43 (13.4)  72 (12.0)  Years Worked as a Nurse on Nursing Unitb Less than 1 year  67 (24.2)  68 (21.3)  135 (22.7)  1 to 2 years  82 (29.6)  58 (18.2)  140 (23.5)  3 to 5 years  30 (10.8)  46 (14.4)  76 (12.8)  6 to 10 years  49 (17.7)  56 (17.6)  105 (17.6)  11 to 15 years  18 (6.5)  26 (8.2)  44 (7.4)  16 to 20 years  17 (6.1)  24 (7.5)  41 (6.9)  Greater than 20 years  14 (5.1)  41 (12.9)  55 (9.2)  Years Worked as a Nurse in Hospitalc Less than 1 year  47 (16.7)  55 (17.2)  102 (17.0)  1 to 2 years  68 (24.4)  47 (14.7)  115 (19.2)  3 to 5 years  40 (14.3)  40 (12.5)  80 (13.4)  (df = 8)  2 = 21.01* (df = 6)  2 = 12.58* (df = 6)  105  Site A frequency (%)  Site B frequency (%)  Total sample frequency (%)  6 to 10 years  46 (16.5)  59 (18.4)  105 (17.5)  11 to 15 years  22 (7.9)  35 (10.9)  57 (9.5)  16 to 20 years  24 (8.6)  30 (9.4)  54 (9.0)  Greater than 20 years  32 (11.5)  54 (16.9)  86 (14.4)  Characteristic  Current Employment Status on Unit Full-time  167 (59.2)  164 (51.1)  331 (54.9)  Part-time  59 (20.9)  98 (30.5)  157 (26.0)  Temporary full- or part-time  12 (4.3)  8 (2.5)  20 (3.3)  Casual  44 (15.6)  51 (15.9)  95 (15.8)  Between group comparison statistic  2 = 8.54* (df = 3)  Note. N = 603; Valid percent used. a Total missing = 3. bTotal missing = 7. cTotal missing = 4. * p  0.05  The demographic characteristics of the sample are found in Table 5.3. The average age of the respondents was 40.3 years (SD = 11.1, n = 585) and their ages ranged between 22 and 65 years. All age groups were represented fairly equally. Most of the nurses (94.3%, n = 567) were female. The nurses’ education ranged from LPN diploma 13.0% (n = 78) to graduate education at the master’s level (1.2%). An equivalent number of nurses reported their highest level of nursing education as a RN diploma (n = 267, 44.4%) or a baccalaureate (in nursing) (n = 250, 41.5%); however, fewer nurses (n = 207, 34.4%) reported their first educational qualification as a baccalaureate (in nursing). The year when these nurses completed their first educational qualification ranged between 1951 and 2007. Almost three quarters (73.1%, n = 441) of the nurses completed their first nursing educational qualification in Canada. Of those that completed their first nursing education in another country, most reported an Asian country. For those educated outside of Canada, the average length of time they had lived in Canada was 14.6 years (SD = 10.7, range = 2 months to 42 years). Of those nurses who reported their ethnicity/race (n = 598), slightly more than one-half self-identified as “white” (56.5%, n = 338). The two other predominant ethnicity/race categories were “Filipino” (16.7%, n = 100) and “South Asian” (13.2%, n = 79). When asked what language was spoken at home, about two-thirds (67.7%, n = 407) reported speaking “English;” the other predominant languages included “Tagalog” (10.8%, n = 65) and “Punjabi/Hindi” (8.7%, n = 52).  106  5.2.1  Hospital-based Group Differences Differences in the employment and demographic characteristics of the Site A and Site  B respondents were explored through the use of the chi-squared statistic (see Table 5.2 and Table 5.3). Based on the criterion of p  0.05, group differences of primary concern were those demographic variables in the model used to calculate the D-scores, namely, age, highest level of education, and ethnicity/race. Site A and Site B nurses did not significantly differ with respect to their age. A significant difference between the hospitals was found for the nurses’ ethnicity/race and their highest nursing education. To account for the small cell counts for the different ethnicity/race categories in each hospital, the variable was recoded into two categories: “white” and “all other.” The 2 statistic was significant (2 = 3.74, df = 1, p  0.05); Site B had significantly more “white” nurses in the sample and their educational backgrounds differed (which may have been influenced by the inclusion of non-medical surgical units at Site B) in comparison with Site A. However, the focus of this study was the ethnicity/race and highest level of nursing education of the individual nurse relative to other members of his or her nursing unit (relational diversity). Accordingly, these significant group differences between Site A and B are likely addressed through the use of the D-score measure.  107  Table 5.3  Demographic Characteristics and Hospital-based Group Comparisons of the Respondents Site A frequency (%)  Site B frequency (%)  Total sample frequency (%)  20 to 29 years  61 (22.2)  72 (23.2)  133 (22.7)  30 to 39 years  86 (31.3)  79 (25.5)  165 (28.2)  40 to 49 years  60 (21.8)  72 (23.2)  132 (22.6)  50 to 59 years  58 (21.1)  78 (25.2)  136 (23.2)  60 years plus  10 (3.6)  9 (2.9)  19 (3.2)  Characteristic Agea  Genderb Female Male  RPN Diploma RN Diploma (hospital program)  258 (91.8)  309 (96.6)  567 (94.3)  23 (8.2)  11 (3.4)  34 (5.7)  63 (22.4)  48 (15.0)  111 (18.4)  1 (0.4)  5 (1.6)  6 (1.0)  57 (20.3)  78 (24.3)  135 (22.4)  RN Diploma (community college program)  57 (20.3)  86 (26.8)  143 (23.8)  Baccalaureate in Nursing  103 (36.7)  104 (32.4)  207 (34.4)  Year Completed First Educational Qualification in Nursinge 1950-1969  15 (5.4)  8 (2.5)  23 (3.9)  1970-1979  38 (13.7)  72 (22.4)  110 (18.4)  1980-1989  46 (16.5)  61 (19.0)  107 (17.9)  1990-1999  76 (27.3)  73 (22.7)  149 (24.9)  2000-2007  103 (37.1)  107 (33.3)  210 (35.1)  Highest Educational Qualification in Nursingf LPN Diploma  46 (16.4)  32 (10.0)  78 (13.0)  RN Diploma  108 (38.4)  155 (48.3)  263 (43.7)  Bachelor of Nursing  124 (44.1)  130 (40.5)  250 (42.2)  3 (1.1)  4 (1.2)  7 (1.2)  Canada  209 (74.1)  232 (72.3)  441 (73.1)  Other Country  73 (25.9)  89 (27.7)  162 (26.9)  43 (15.4)  25 (7.9)  68 (11.4)  14 (5.0)  11 (3.5)  25 (4.2)  3 (1.1)  16 (5.0)  19 (3.2)  1 (.4)  16 (5.0)  17 (2.9)  Master of Nursing Country of First Education in Nursingh  Southeast Asia (Malaysia and Philippines) North/West/South Asia (Iran, India, Pakistan, Russia and United Arab Emirates) Northern Europe (United Kingdom) East Asia (China, Hong Kong, and Japan)  (df = 4)  2 = 6.32*  First Educational Qualification in Nursingc LPN Diploma  Between group comparison statistic 2 = 3.21  (df = 1)  2 = 11.24*d (df = 4)  2 = 11.85* (df = 4)  2 = 9.99*g (df = 2)  2 = 0.26 (df = 1)  108  Characteristic Eastern and Western Europe (Romania, Czechoslovakia, Poland, Bulgaria, Moldova, Netherlands, and France) Australia and Pacific Ocean (Fiji, New Zealand, and Australia) United States and Central America (Mexico and Nicaragua) Africa  Site A frequency (%)  Site B frequency (%)  Total sample frequency (%)  4 (1.4)  8 (2.5)  12 (2.0)  3 (1.1)  6 (1.9)  9 (1.5)  2 (0.7)  2 (0.6)  4 (0.7)  0  1 (0.3)  1 (0.2)  Ethnicity/Racei White  146 (52.3)  192 (60.2)  338 (56.5)  All other  133 (47.7)  127 (39.8)  260 (43.5)  6 (2.2)  30 (9.4)  36 (6.0)  50 (17.9)  29 (9.1)  79 (13.2)  4 (1.4)  7 (2.2)  11 (1.8)  61 (21.9)  39 (12.2)  100 (16.7)  Latin American  2 (0.7)  2 (0.6)  4 (0.7)  Southeast Asian  3 (1.1)  4 (1.3)  7 (1.2)  West Asian  2 (0.7)  8 (2.5)  10 (1.7)  Japanese  2 (0.7)  3 (0.9)  5 (0.8)  Korean  1 (0.4)  2 (0.6)  3 (0.5)  Pacific Islander  2 (0.7)  0 (0)  2 (0.3)  0 (0)  3 (0.9)  3 (0.5)  Chinese South Asian Black Filipino  First Nations, Aboriginal, or Métis Language Spokenj English  186 (66.4)  221 (68.8)  407 (67.7)  All other  94 (33.6)  100 (31.2)  194 (32.3)  Tagalog (Filipino)  38 (13.6)  27 (8.4)  65 (10.8)  Punjabi/Hindi  35 (12.5)  17 (5.3)  52 (8.7)  Mandarin/Cantonese  4 (1.4)  26 (8.1)  Other (e.g., Taiwanese, French, Punjabi, Hindi, Farsi, Vietnamese, Polish, Romanian, Spanish)  17 (6.1)  30 (9.3)  30 (5.0) 47 (7.8)  Between group comparison statistic  2 = 3.74* (df = 1)  2 = 0.40 (df = 1)  Note. N = 603, Site A = 282 and Site B = 321; Valid percent used. a Total missing data = 18. bTotal missing data = 2. cTotal missing data = 1. d 2 cells (RPN diploma) had expected counts of less than 5. The chi-squared statistic had RPN diploma combined with RN diploma (community college). eTotal missing data = 4. fTotal missing data = 1. g2 cells (Master of Nursing) had expected counts less than 5. The chi-squared statistic had Master’s combined with baccalaureate. hTotal missing data = 7. iTotal missing data = 5; 7 respondents indicated multiple responses which were recoded as one response; 2 white participants also indicated Romanian and Filipino; 2 Chinese respondents also indicated Latin American and Caribbean, respectively; 2 South Asian respondents also indicated Southeast Asian and West Indian, respectively; West Asian participant also indicated Arab. jTotal missing data = 2 (Site A); 113 respondents indicated English and another language, which were coded according to the non-English language and approximately 5 respondents indicated more than 2 languages, which were recoded to one language according to ethnicity/race background. * p  0.05  109  5.3  Measurement Model for the Exogenous Variables: Actual Diversity As discussed in Chapter 4, one approach to measuring relational diversity is the use of  the Euclidean distance score (D-score). The age, education, and ethnicity/race attributes were measured with one item for each attribute, 19 which were subsequently used in the calculation of the D-score for each respondent (see Equation 4.1). To calculate the D-score for the work values attribute the Contemporary Work Values (CWV) Scale was used. This scale asked respondents to rate their agreement with 35 statements about their work beliefs and attitudes (Wayne, 1989). This first section examines the measurement structure of this scale. Once the measurement structure was finalized, the items were summed and averaged, to create an average total score, which was then used to create a D-score for each respondent. The actual diversity variables were modelled as manifest variables in the structural modelling portion of the analysis, which examined the influence of actual diversity on burnout, as mediated by conflict. 5.3.1  Exploratory Factor Analysis of the Contemporary Work Values Scale Using the Mplus 5.1 software, the 35-item CWV Scale was subjected to a series of  explanatory factor analyses (EFA) to confirm the claims published about the scale and its unidimensionality. Data analysis occurred for the purpose of identifying several items, which taken together, represent the work values construct. All data for the EFA were treated as ordinal and non-normal. Oblique geomin rotation was used as the default for the WLSMV estimator method. Cases with missing values were excluded pairwise from the analyses. This systematic process guided the factor analysis: (a) the factor structure was restricted to one factor, (b) the items with the lowest factor loadings ( 0.55) were removed serially from the analysis, and (c) the global fit indices were reviewed and acceptable limits were established as CFI  0.95, TLI  0.95, SRMR  0.08, and RMSEA  0.08 with a preferred value of 0.06 (Brown, 2006; Hu & Bentler, 1999; Schumacker & Lomax, 2004). The initial step was to conduct an EFA specifying a one-factor structure with the entire 35 items. The overall goodness-of-fit indices suggested that the one-factor model with all  19  Age in years was calculated based on year of birth. For their education, respondents indicated their highest level of education in nursing (LPN Diploma, RPN Diploma, RN Diploma, Bachelor of Nursing, Master of Nursing). To determine their ethnicity/race, the respondents were asked to self-identify the category that best represented their background.  110  of the indicators did not fit the data well, 2 (171) = 1616.90, p  0.001, CFI = 0.71, TLI = 0.86, RMSEA = 0.12 and SRMR = 0.10. The factor loadings ranged from 0.16 to 0.73. Based on the above analysis plan, several other EFA models were estimated with one poor indicator removed serially. The model with the best fit was a one-factor structure with 16 indicators, 2 (65) = 328.10, p  0.001, CFI = 0.93, TLI = 0.97, RMSEA = 0.08, and SRMR = 0.06. Table 5.4 provides the factor loadings for each indicator (range from 0.55 to 0.76) and Figure 5.1 depicts the specification of the final model. The eigenvalue for the one factor was 7.33 and the inter-item correlations ranged from 0.24 to 0.60 (see Appendix E1). The residual correlations ranged from -0.12 to 0.14.  111  Table 5.4 Variable name A1WE A4WE A8WE A11WE A14WE A16WE A17WE A20WE A24WE A25WE A28WE A29WE A30WE A31WE A32WE A34WE  Structure Matrix of the EFA for the 16-item Contemporary Work Values Scale Item I expect work to be a meaningful and fulfilling part of my life. I need to be listened to by my superiors; work should be a two-way communication process. I desire work that provides opportunities for personal growth and allows me to “feel good inside.” It is important to me that my job provides opportunities to strengthen my abilities and talents. I am very concerned that I receive personal satisfaction from my work. Work provides many opportunities for “personal growth” experiences. I enjoy work assignments that are challenging and require extensive use of thought processes. Work is beneficial in helping me to become a “whole” person. Work contributes to my understanding and development of my character and capabilities. Work should provide me with a high degree of self-satisfaction or selffulfillment. I want to be informed about the activities and plans of my nursing unit. I seek work experiences that help me expand and use my potential to the fullest extent possible. I would like variety in my work. Work provides individuals with an opportunity to “grow” and realize their full potential. I seek various emotional and psychological rewards from working in addition to my pay cheque. Work should be an extension of one’s lifestyle and not merely a means to obtain subsistence. Chi-squared (df; p) CFI; TLI RMSEA; SRMR  Factor loading 0.59  0.35  0.56  0.31  0.68  0.46  0.74  0.54  0.57 0.64  0.33 0.40  0.65  0.42  0.68  0.46  0.74  0.55  0.73  0.54  0.59  0.35  0.71  0.51  0.62  0.38  0.76  0.57  0.66  0.44  0.55  0.31  2  R  328.10 (65; 0.001) 0.93; 0.97 0.08; 0.06  Note. N = 603; Oblique geomin rotation and WLSMV estimation; Eigenvalue for one factor = 7.33. p < 0.001.  The construct of contemporary work values was best explained with 16 indicators, which reflect intrinsic attitudes and behaviour. These indicators were used to calculate a total score for each participant. Because the total score did not take into consideration the number of items missing for each respondent, an average of the total score was calculated for each respondent. To compute the average total score for the CWV Scale, a criterion was established that at least 14 of the 16 items (88% of items answered) had to be completed for the case to be included in further analyses. The average total score was then used to calculate the D-score for the actual diversity in work values variable. This variable was treated as a manifest variable, which was used in the structural equation modelling portion of the analysis to examine the influence of actual diversity on burnout, as mediated by interpersonal conflict.  112  Figure 5.1  Final Measurement Model for the Contemporary Work Values Scale  1  A1WE  2  A4WE  3  A8WE  4  A11WE  5  A14WE  6  A16WE  7  A17WE  8  A20WE  9  A24WE  10  A25WE  11  A28WE  12  A29WE  13  A30WE  14  A31WE  15  A32WE  16  A34WE  Contemporary Work Values  λ standardized parameters for relationships between the latent factor and the observed, p < 0.001.  113  5.4  Measurement Model for the Exogenous Variables: Perceived Diversity The second approach to measuring relational diversity was the perceptual method,  which asked the respondents to indicate how similar to them other members of their workgroups were on each diversity attribute. The age, education, and ethnicity/race attributes were measured with one item for each attribute. The perceived work values attribute was measured using 4 items (Jehn, Northcraft, & Neale, 1999), which were subjected to confirmatory factor analysis. For the structural equation modelling portion of the analysis, all of these exogenous variables were treated as latent factors. 20 The work of Jehn et al. (1999) was used to specify the measurement model for perceived diversity in work values (Perceived Work Values Diversity Scale). The factor structure specified for this scale was a one-factor model with 4 indicators, which were measured on 6-point Likert scales. With the exception of the RMSEA, the overall goodness-of-fit indices for the CFA suggested that the one-factor model fit the data well, 2 (2) = 19.11, p < 0.001, CFI = 0.99, TLI = 0.99, and RMSEA = 0.12. Table 5.5 lists the parameter estimates for each latent variable and Figure 5.2 depicts the specification of the final model. No modification indices were noteworthy. All of the standardized parameter estimates were greater than 0.80 and the R-squared values ranged from 0.67 to 0.69. The item-to-item correlations ranged from 0.64 to 0.69 (see Appendix E2,), which were indicative of construct validity.  20  In Mplus software the exogenous variables (latent and manifest) are automatically treated as continuous variables. However, in this study the variables were treated as categorical. To rectify this limitation in the software, thresholds were determined for each variable and the syntax was specified in such a way that these variables could be modelled as categorical.  114  Table 5.5 Variable name C6SPRI C2SWE C8SATT C13SBEL  CFA Results for the Perceived Work Values Diversity Scale Question    In my nursing unit, the other nurses are similar to me in terms of the principles that guide their work. In my nursing unit, the other nurses are similar to me in terms of their work ethic (values). In my nursing unit, the other nurses are similar to me in terms of their attitudes about work. In my nursing unit, the other nurses are similar to me in terms of their beliefs about work.  One-factor model SE    2  R  1.00*  0.00  0.83  0.69  0.99  0.02  0.82  0.67  0.99  0.02  0.82  0.67  0.99  0.02  0.82  0.67  Chi-squared (df; p) CFI; TLI RMSEA  19.11 (2; 0.001) 0.99; 0.99 0.12  Note. N = 602. All fixed parameter estimates statistically significant, p < 0.001. * Fixed to equal 1.0.  To further examine the misfit attributed to a larger than acceptable RMSEA, the residual correlations were inspected (ranging from -0.03 to 0.02) and additional analyses were conducted. 21 A 3-item model was just-identified and thus a goodness-of-fit evaluation did not apply; however, the magnitudes of the factor loadings were similar to the 4-item model. Given the theoretical and prior psychometric evaluation of a one-factor structure (Jehn et al., 1999) and failure to identify any areas of localized strain, the original one-factor structure with 4 items was retained (see Figure 5.2).  21  In an EFA, the SRMR = 0.01 and eigenvalue for one factor was 3.02. When the variables were treated as continuous and ML estimation was used, the RMSEA was 0.08. Based on the work of Beauducel and Herzber (2006) the RMSEA may be slightly larger for models estimated with WLSMV using four to six categories.  115  Figure 5.2  Final Measurement Model for the Perceived Work Values Scale  ε1  C6SPRIR  ε2  C2SWER  ε3 ε4  C8SATTR  Perceived Work Values  C13SBELR  λ standardized parameters for relationships between the latent factor and the observed variables, p < 0.001.  5.5  Measurement Model for the Mediator Variable: Intragroup Conflict Prior theory and evidence (Barki & Hartwick, 2004; Jehn, 1994, 1995; Jehn &  Chatman, 2000) were used to specify the measurement model that assessed how much conflict there was in a specific workgroup (Intragroup Conflict Scale). The Intragroup Conflict Scale consisted of three latent variables that were composed of several indicators: relationship conflict (5 items), task conflict (4 items), and process conflict (3 items). Several steps were followed to confirm the factor structure of this scale. The initial factor structure examined was a three-factor model with 12 indicators. Table 5.6 lists the indicators loaded onto their respective latent variable. With the exception of the RMSEA, the overall goodness-of-fit indices suggested that the three-factor model fit the data well, 2 (36) = 202.96, p  0.001, CFI = 0.97, TLI = 0.995, and RMSEA = 0.09. To further examine the misfit attributed to a larger than acceptable RMSEA, the residual correlations (ranging from -0.05 to 0.05) and modification indices were inspected (one value at 10.10). All of the standardized parameter estimates were greater than 0.71 and statistically significant, and the R-squared values ranged from 0.51 to 0.81 (see Table 5.7). The item-to-item correlations ranged from 0.53 to 0.78 (see Appendix E3). The correlations for the latent variables were greater than 0.70: relationship and task = 0.96, relationship and process = 0.91, and task and process = 1.0. As a result of the large correlations between the latent variables, the matrix was not positive definite. 116  Table 5.6  CFA Results for the Intragroup Conflict Scale with a Three-factor Solution  Variable name  Three-factor model   R  1.00  *  0.85  0.72  1.03  0.88  0.77  1.06  0.90  0.81  0.98  0.83  0.69  0.92  0.78  0.61  How often do members in your nursing unit disagree about the work being done? How frequently are there conflicts about work ideas among members in your nursing unit? How much conflict about the work you do is there among members in your nursing unit? To what extent are there differences of opinions among members in your nursing unit?  1.00*  0.84  0.70  1.02  0.86  0.73  0.85  0.71  0.51  0.98  0.82  0.67  How often do members in your nursing unit disagree about who should do what? How frequently do members in your nursing unit disagree about the way to complete a task? How much conflict is there about delegation of tasks among members in your nursing unit?  1.00*  0.84  0.71  0.99  0.84  0.70  1.00  0.84  0.71  Question    2  RELATIONSHIP How much friction is there among members in your nursing unit? How much are personality clashes evident among members in your nursing unit? How much tension is there among members in your nursing unit? How much rivalry is there among members in your nursing unit? How much anger is there among members in your nursing unit?  E1GFRIC E4GPERS E7GTENS E10GRIV E12GANGR TASK E2GWRK E5GIDEA E8GWRKDO E9GOPIN PROCESS E3GWHO E6GTASK E11GDELG  Chi-squared (df; p) CFI; TLI RMSEA  202.96 (36; 0.001) 0.97; 0.995 0.09  Notes. N = 602, latent variable covariance matrix was not positive definite. Correlations for the latent variables were: Relationship  Task = 0.96, Relationship  Process = 0.91, Task  Process = 1.01. * Fixed to equal 1.0.  To be able to test the hypothesized structural models it was important to be able to distinguish between the types of conflict; however, the large correlations among the conflict latent variables resulted in concerns about discriminant validity. From a theoretical standpoint, it is reasonable to expect that types of conflict may overlap, particularly the task and process items, which capture disagreements about work. At the same time, task-related conflict may evolve into relationship conflict, or vice versa (Jehn & Bendersky, 2003; Jehn & Mannix, 2001). Other researchers have reported a strong relationship between task and relationship conflict (range = 0.39 to 0.99) (Simons & Peterson, 2000). Prior psychometric evaluation of the conflict scale, using exploratory factor analysis techniques with oblique rotation, has supported both a three-factor structure (Jehn & Chatman, 2000; Jehn & Mannix, 2001; Jehn et al., 1999) and a 117  two-factor structure (Jehn, 1994, 1995; Pearson, Ensley, & Amason, 2002). Measuring relationship conflict, using principal component analysis with varimax rotation, Pelled (1996b) reported that 7 items loaded onto a single factor. However, no published studies have used these measures in a structural equation modelling context. After considering the model fit for CFAs with each latent variable (one factor structures) and models with variation in the number of factors (i.e., one-factor and two-factor) the source of model misspecification was attributed to the process indicators. Guided by theory, prior evaluation of the Intragroup Conflict Scale, and the overall aim to achieve model parsimony, the process indicators were removed and the model was respecified as two factors: relationship conflict and task conflict. The model fit for the 9-item two-factor solution was 2 (20) = 87.62, p  0.001, CFI = 0.99, TLI = 0.996, and RMSEA = 0.08 with standardized factor loadings ranging from 0.71 to 0.90. Table 5.7 lists the indicators loaded onto each latent variable and Figure 5.3 depicts the specification of the final two-factor model. The residual correlations were less than or equal to |0.05| and no modification indices greater than 10.0 were identified. The correlation between relationship and task conflict was 0.96 (p  0.001).  118  Table 5.7 Variable name  CFA Results for the Intragroup Conflict Scale with a Two-factor Solution Question  Two-factor model SE      2  R  RELATIONSHIP E7GTENS E1GFRIC E4GPERS E10GRIV E12GANGR  How much tension is there among members in your nursing unit? How much friction is there among members in your nursing unit? How much are personality clashes evident among members in your nursing unit? How much rivalry is there among members in your nursing unit? How much anger is there among members in your nursing unit?  1.00  *  0.00  0.90  0.82  0.94  0.02  0.85  0.73  0.96  0.01  0.87  0.76  0.92  0.02  0.83  0.69  0.86  0.02  0.77  0.60  1.00*  0.00  0.87  0.75  0.96  0.02  0.83  0.70  0.82  0.03  0.71  0.50  0.95  0.02  0.82  0.67  TASK E5GIDEA E2GWRK E8GWRKDO E9GOPIN  How frequently are there conflicts about work ideas among members in your nursing unit? How often do members in your nursing unit disagree about the work being done? How much conflict about the work you do is there among members in your nursing unit? To what extent are there differences of opinions among members in your nursing unit? Chi-squared (df; p) CFI; TLI RMSEA  87.62 (20; 0.001) 0.988; 0.996 0.08  Note. N = 602. All parameter estimates are statistically significant p < 0.001. Correlation for Relationship  Task = 0.96. * Fixed to equal 1.0.  119  Figure 5.3  Final Measurement Model for the Two-factor Intragroup Conflict Scale  ε1  E7GTENS  ε2  E1GFRIC  ε3 ε4 ε5  E4GPERS E10GRIV E12GANGR  ε6  E5GIDEA  ε7  E2GWRK  ε8 ε9  Relationship Conflict  E8GWRKDO  Task Conflict  E9GOPIN  λ standardized parameters for relationships between the latent factor and the observed variables, p < 0.001.  5.6  Measurement Model for the Mediator Variable: Individual Conflict The Individual Conflict Scale measured individuals’ perceptions of their involvement  in conflict (relationship, task, and process) with their coworkers. The measurement model for this scale was based on prior theory and evidence for both this scale (Hobman, Bordia, & Gallois, 2003) and the Intragroup Conflict Scale (Jehn, 1994, 1995; Jehn & Chatman, 2000). This scale consisted of three latent factors that were composed with several indicators: relationship conflict (4 items), task conflict (4 items), and process conflict (3 items). Four steps were followed to confirm the factor structure of this scale. 120  5.6.1  Initial CFA for Three Factors with all Items The initial factor structure examined was a three-factor model with 11 indicators.  Table 5.8 lists the indicators loaded onto each latent variable. With the exception of the RMSEA, the overall goodness-of-fit indices suggested that the three-factor model fit the data well, 2 (27) = 143.26, p  0.001, CFI = 0.98, TLI = 0.99, and RMSEA = 0.09. To further examine the misfit attributed to a larger than acceptable RMSEA, the residual correlations were inspected (ranging from -0.08 to 0.10) and the modification indices were reviewed (3 indices ranging from 18.63 to 22.64 involved items G10IANGR and G8ITASK). All of the standardized parameter estimates were greater than 0.70 and statistically significant. The R-squared values ranged from 0.50 to 0.91. The item-to-item correlations ranged from 0.52 to 0.87 (see Appendix E4). The correlation between relationship and task conflict was 0.90, between relationship and process conflict was 0.86, and between process and task conflict was 0.95. The large correlations between the latent variables (> 0.85) raised concerns about their discriminant validity (Brown, 2006), particularly between task and process conflict. When the overall fit of an initial model is found to be satisfactory, but there is a significant amount of overlapping among latent variables, a more parsimonious solution similar to the initial structure may be achieved by combining factors (Brown, 2006).  121  Table 5.8  CFA Results for the Individual Conflict Scale with a Three-factor Solution  Variable name  Question    Three-factor model 2 R   RELATIONSHIP G1IFRIC G4IPERS G7ITENS G10IANGR TASK G2IOPIN G3IIDEA G5IWRKDO G9IWRK  How much friction was there between you and your coworkers? How much are personality clashes evident between you and your coworkers? How much tension was there between you and your coworkers? How often do you get angry with your coworkers? To what degree do you and your coworkers have diverging opinions about the work being done? How much conflict about work ideas exists between you and your coworkers? How often do you and your coworkers disagree about what things should be done? To what extent do you and your coworkers have disagreements about work?  1.00  *  0.90  0.81  0.94  0.83  0.70  1.06  0.95  0.91  0.79  0.71  0.50  1.00  *  0.76  0.58  1.12  0.85  0.72  1.19  0.91  0.82  1.18  0.90  0.81  1.00  *  0.86  0.74  0.97  0.83  0.69  0.98  0.85  0.72  PROCESS G6IWHO G8ITASK G11DELG  How often do you disagree with your coworkers about who should do what? How frequently do you disagree with your coworkers about the way to complete a task? How much conflict do you have with your coworkers about delegation of tasks on your nursing unit? Chi-squared (df; p) CFI; TLI RMSEA  143.26 (27; 0.001) 0.98; 0.99 0.09  Notes. N = 602, all parameter estimates statistically significant, p < 0.001. Correlations for latent variables were: Relationship  Task = 0.90, Relationship  Process = 0.86, Task  Process = 0.96. *Fixed to equal 1.0.  To be able to test the hypothesized structural model, it was important to discriminate between the three types of conflict. From a theoretical standpoint it was reasonable to expect that the constructs overlapped, particularly the task and process items, which captured disagreements about work. No psychometric evaluation of a three-factor structure has been published. However, using a 6-item scale to measure individual involvement in task (4 items) and relationship conflict (2 items), Hobman et al. (2003) conducted a principal component factor analysis with varimax (oblique) rotation and found some support for a two-factor structure with one item loading on both factors. The correlation between task and relationship conflict was statistically significant (r = 0.49) (Hobman et al., 2003). No published studies have used the Individual Conflict Scale in a structural equation modelling context. To examine further the  122  measurement model of the Individual Conflict Scale, additional analyses were conducted prior to confirming a final structure (see Appendix F). 5.6.2  CFA of the Task and Relationship Subscales A final confirmatory factor analysis was conducted with the 8 items from the  relationship and task subscales and one cross-loading. The fit indices for the two-factor structure were: 2 (15) = 28.13, p < 0.001; CFI = 1.0, TLI = 1.0, and RMSEA = 0.04. Table 5.9 lists the parameter estimates for each latent variable and Figure 5.4 depicts the specification of the final two-factor model. All of the standardized parameter estimates were greater than 0.78, with the exception of G10IANGR. The R-squared values ranged from 0.45 to 0.92. The residual correlations were inspected (ranging from -0.04 to 0.03). There were no modification indices greater than 10.0. The correlation between task and relationship conflict was 0.89 (p < 0.001) and the item-to-item correlations ranged from 0.49 to 0.88. Given the overall goodness-of-fit indices, the overall aim to achieve model parsimony, and consideration of available theory, an 8-item two-factor structure with one cross-load was accepted as a suitable measurement structure for the Individual Conflict Scale.  123  Table 5.9 Variable name  CFA Results for the Individual Conflict Scale with a Two-factor Solution Two-factor model  Question  2    SE     R  How much tension is there between you and your coworkers? How much friction is there between you and your coworkers? How much are personality clashes evident between you and your coworkers? How often do you get angry with your coworkers?  1.00*  0.02  0.96  0.92  0.95  0.00  0.91  0.82  0.88  0.02  0.84  0.71  0.41  0.11  0.39  0.45  How often do you and your coworkers disagree about what things should be done? To what degree do you and your coworkers have diverging opinions about the work being done? How much conflict about work ideas exists between you and your coworkers? To what extent do you and your coworkers have disagreements about work? How often do you get angry with your coworkers?  1.00*  0.00  0.89  0.79  0.87  0.03  0.78  0.60  0.97  0.02  0.86  0.75  1.00  0.02  0.89  0.80  0.11  b  RELATIONSHIP G7ITENS G1IFRIC G4IPERS G10IANGR TASK G5IWRKDO G2IOPIN G3IIDEA G9IWRK G10IANGR  Chi-squared (df; p) CFI; TLI RMSEA  0.33  0.30  a  **  28.13 (15; 0.01) 0.998; 0.999 0.04  Notes. N = 602, all parameter estimates statistically significant, p < 0.001 unless otherwise specified. Correlation of Relationship  Task = 0.89. * Fixed to equal 1.0. a Combined R2; bp < 0.01; **Indicator R2 is reported for intended latent variable.  124  Figure 5.4  Final Measurement Model for the Two-factor Individual Conflict Scale  ε1  G7ITENS  ε2  G1IFRIC  ε3 ε4  ε5 ε6  G4IPERS G10IANGR  Relationship Conflict  G5IWRKDO G2IOPIN  ε7  G3IIDEA  ε8  G9IWRK  Task Conflict  λ standardized parameters for relationships between the latent factor and the observed variables, p < 0.001. λ standardized parameters for relationship between the latent factor and the observed variable that crossloaded, p < 0.01.  125  5.7  Measurement Model for the Endogenous Variable: Burnout The hypothesized measurement model for the Maslach Burnout Inventory (MBI) was  specified according to prior theory and evidence (Beckstead, 2002; Maslach, Jackson, & Leiter, 1996; Schaufeli, Bakker, Hoogduin, Schaap, & Kladler, 2001). The MBI Health Services Survey (HSS) consists of three latent variables (factors), which were composed with several indicators: emotional exhaustion (EE) (9 items), depersonalization (DP) (5 items), and personal accomplishment (PA) (8 items). Five items that represent the cynicism (CY) latent variable, which was part of the MBI-General Survey, were also included in the study. The factor structure first examined was a four-factor CFA model with 27 indicators. Table 5.10 provides the parameter estimates of the indicators that were loaded onto each latent variable and summarizes the model fit. Except for the TLI, the overall goodness-of-fit indices suggested that the four-factor model did not fit the data well, 2 (103) = 1096.62, p  0.001, CFI = 0.86, TLI = 0.95, and RMSEA = 0.13. Inspection of the residual correlations (ranging from -0.24 to 0.19) and modification indices (ranging from 10.02 to 158.46) indicated localized points of ill fit in the solution. Appendix E5 provides the polychoric correlation matrix for the items in this measurement model.  126  Table 5.10 Variable name  MBI Item No.  CFA Results for the Maslach Burnout Inventory with Four-factor and Three-factor Solutions Four-factor model Question  EMOTIONAL EXHAUSTION B1EE 1 Drained B2EE 2 Used up B3EE 3 Fatigued B6EE 6 Work strain B10EE 8 Burned out B15EE 13 Frustrated B16EE 14 Work hard B18EE 16 People stressful B25EE 20 End of rope DEPERSONALIZATION B5DP 5 Impersonal B12DP 10 Callous B13DP 11 Hardening B17DP 15 Not care B27DP 22 Patients blamed PERSONAL ACCOMPLISHMENT B4PA 4 Understand patients B7PA 7 Deal with problems B11PA 9 Positive influence B14PA 12 Energetic B19PA 17 Create atmosphere B20PA 18 Exhilarated B21PA 19 Accomplish B26PA 21 Deal calmly CYNICISM B8CY 8 Less interested B9CY 9 Less enthusiastic B22CY 13 Not be bothered B23CY 14 Cynical B24CY 15 Doubt Chi-squared (df; p) CFI; TLI RMSEA  Three-factor model 2      R      R2  1.00* 1.03 1.06 0.89 1.12 1.00 0.80 0.91 1.08  0.76 0.78 0.80 0.67 0.85 0.76 0.61 0.69 0.82  0.57 0.60 0.64 0.45 0.72 0.57 0.37 0.47 0.67  0.97 * 1.00 1.00 0.74 1.03 0.91 0.76  0.63 0.66 0.66 0.37 0.71 0.55 0.38  0.98  0.79 0.82 0.81 0.61 0.84 0.74 0.62 removed 0.80  1.00 1.17 1.21 1.01 0.81  *  0.69 0.80 0.83 0.69 0.55  0.47 0.64 0.69 0.48 0.31  0.88 1.00* 1.06 0.85 0.70  0.69 0.79 0.84 0.67 0.55  0.48 0.63 0.70 0.45 0.31  1.00 1.93 1.60 2.71 2.30 1.66 2.43 1.96  *  0.28 0.55 0.45 0.77 0.65 0.47 0.69 0.55  0.08 0.30 0.20 0.59 0.42 0.22 0.47 0.31  0.50 0.81 0.65  0.38 0.61 0.49 Removed 0.75 0.52 0.71 0.59  0.14 0.37 0.24  1.00* 1.02 0.60 0.84 0.88  0.92 0.94 0.55 0.77 0.81  0.85 0.89 0.31 0.59 0.65  1096.62 (103; 0.001) 0.86; 0.95 0.13  1.00* 0.68 0.95 0.78  0.64  0.56 0.26 0.51 0.34  Removed Removed Removed Removed Removed 462.53 (64, 0.001) 0.92; 0.95 0.10  Notes. N = 603, all parameter estimates are significant, p < 0.001.  = Unstandardized parameter estimates;  = Standardized parameter estimates. Correlations of latent variables for the four-factor model were: EE  DP = 0.72; EE  PA = -0.42, EE  CY = 0.77; PA  DP = -0.49; DP  CY = 0.75; PA  CY = -0.47. Correlations for the latent variables for the three-factor model were: EE  DP = 0.71; EE  PA = -0.26, PA  DP = -0.46. * Fixed to equal 1.0.  The factor structure was respecified according to the original three-factor model for the MBI-HSS with items B14PA (energetic indicator for personal accomplishment) and B18EE (people stressful indicator for emotional exhaustion) removed as suggested by Maslach, Jackson, and Leiter (1996). Table 5.11 provides the parameter estimates for the 20 indicators that were loaded onto the EE, DP, and PA latent variables. Each of the overall goodness-of-fit indices 127  suggested that the three-factor model fit the data marginally, 2 (64) = 462.53, p  0.001, CFI = 0.92, TLI = 0.95, and RMSEA = 0.10. Inspection of the residual correlations (ranging from -0.18 to 0.16) and modification indices (ranging from 10.61 to 60.39) indicated localized points of ill fit in the solution. After considering the model fit of a separate CFA for each latent variable (i.e., EE, DP, and PA) and models with variation in the number of factors (i.e., one-factor and two-factor) the source of model misspecification was attributed to incorrect designation of the relationships between indicators and the latent variables. Guided by prior evidence of the psychometric evaluation of the MBI-HSS (Beckstead, 2002) and theory (Maslach, 1982; Schaufeli, Maslach, & Marek, 1993), the model was respecified. In consideration of the residual correlations (values  |0.10|), the largest modification indices (values  10.0), and the largest value for the expected parameter change (EPC), a systematic process was followed to respecify the model (e.g., allow an indicator to load on two or more factors) (Brown, 2006). For example, the three-factor model without item B14PA and B18EE was respecified so that indicator B6EE loaded on its intended latent variable (i.e., EE latent variable) and the DP latent variable (modification index = 63.39 and expected parameter change = 0.48). The overall goodness-of-fit indices for this three-factor model with one cross-load (i.e., B6EE loading on DP) suggested better fit although the values were still not within the recommended ranges (2 (64) = 406.22, p  0.001, CFI = 0.93, TLI = 0.96, and RMSEA = 0.09), the residual correlations ranged from -0.16 to 0.16, and the modification indices ranged from 12.91 to 28.98. The 2 statistic was used to statistically compare the fit of the model with one cross-loading with the original model (Muthen & Muthen, 2007). The difference test was statistically significant (∆2 (1) = 53.93, p  0.001), indicating that the three-factor model with one cross-load provided significantly better fit to the data; however, the overall fit was still marginal. This process of allowing indicators to load onto two or more factors continued serially by selecting the largest value for the modification indices and expected parameter changes until a model that demonstrated acceptable fit was identified.  128  The final model that fit the data reasonably well was a 20-item three-factor model (excluding B14PA and B18EE) and 8 indicators loading on two or more factors (2 (68) = 320.77, p  0.001, CFI = 0.95, TLI = 0.97, and RMSEA = 0.08, residual correlations ranging from -0.16 to 0.15 and no MI greater than 15. The three-factor model with 8 cross-loadings was a statistically significant better fit than the original three-factor model with no cross loadings (∆2 (7) = 131.89, p  0.001). Table 5.11 lists the parameter estimates for indicators that loaded onto each latent variable and Figure 5.5 depicts the specification of the final three-factor model. A comparison of the model fit indices for the three and four factor models is found in Table 5.12.  129  Table 5.11  CFA Results for the Maslach Burnout Inventory with a Three-factor Solution and 8 Cross-loadings  Variable name  Question   EMOTIONAL EXHAUSTION (EE) B2EE Used up B1EE Drained B3EE Fatigued B6EE Work strain B10EE Burned out B15EE Frustrated B16EE Work hard B25EE End of rope B4PA Understand patients B13DP Hardening B17DP Not care DEPERSONALIZATION (DP) B12DP Impersonal B5DP Callous B13DP Hardening B17DP Not care B27DP Patients blamed B6EE Work strain B25EE End of rope B26PA Deal calmly B15EE Frustrated PERSONAL ACCOMPLISHMENT (PA) B19PA Create atmosphere B4PA Understand patients B7PA Deal with problems B11PA Positive influence B20PA Exhilarated B21PA Accomplish B26PA Deal calmly B10EE  Burned out  Three-factor model SE    1.00* 0.00 0.98 0.03 1.00 0.02 0.23 0.05 0.97 0.03 0.74 0.03 0.77 0.03 0.62 0.05 Added Cross-loadings for EE 0.17 0.05 0.23 0.05 -0.28 0.08 * 0.00 1.00 0.86 0.04 0.81 0.06 1.10 0.08 0.69 0.05 Added Cross-loading for DP 0.61 0.06 0.46 0.06 -0.19 0.06 0.24 0.05  1.00* 0.00 0.63 0.06 0.82 0.06 0.66 0.05 0.68 0.06 0.95 0.06 0.59 0.07 Added Cross-loadings for PA -0.21 0.04  Chi-squared (df; p) CFI; TLI RMSEA  Notes 2  R  0.83 0.81 0.83 0.19 0.81 0.61 0.64 0.51  0.69 0.66 0.70 0.39a a 0.73 0.55a 0.41 0.63a  0.14 0.19 -0.24  ** ** **  0.80 0.69 0.65 0.89 0.55  0.65 0.48 0.61a 0.59a 0.31  0.49 0.37 -0.15 0.19  ** ** ** **  0.75 0.48 0.62 0.49 0.51 0.72 0.45  0.56 a 0.22 0.38 0.24 0.26 0.51 0.28a  -0.16  **  PA indicator DP indicator DP indicator  EE indicator EE indicator PA indicator EE indicator  EE indicator  320.77 (66; 0.001) 0.95; 0.97 0.08  Note. N = 603, all parameter estimates statistically significant, p ≤ 0.01 unless otherwise specified. Correlations for the burnout latent variables were: EE  DP = 0.59; EE  PA = -0.16, PA  DP = -0.47. * Fixed to equal 1.0. a Combined R2; **Indicator R2 is reported for intended latent variable.  130  Table 5.12  Summary of the CFAs for the Maslach Burnout Inventory Model  Four factor model (27 items) – EE, DP, PA, CY Three factor model (20 items) – EE, DP, PA Three factor model (20 items) – EE, DP, PA with 8 cross loadings  2  df  CFI  TLI  RMSEA  1096.62  103  0.86  0.95  0.13  Residual correlations range -0.24 to 0.19  462.53  64  0.92  0.95  0.10  -0.18 to 0.16  320.77  68  0.95  0.97  0.08  -0.16 to 0.15  Note. N = 603 for all models; 2 and df were based on WLSMV estimation; EE = emotional exhaustion, DP = depersonalization, and PA = personal accomplishment.  131  Figure 5.5  Measurement Model for the Three-factor MBI  ε1 ε2  B2EE  ε3  B3EE  ε4 ε5  B6EE B10EE  ε6  B15EE  ε7  B16EE  ε8  B25EE  B1EE  ε9  B5DP  ε10  B12DP  ε11  B13DP  ε12 ε13  B17DP  Emotional Exhaustion  Depersonalization  B27DP  ε14  B4PA  ε15 ε16  B7PA B11PA  ε17  B19PA  ε18  B20PA  ε19  B21PA  ε20  B26PA  Personal Accomplishment  λ standardized parameters for relationships between the latent factor and the observed variables, p  0.01. λ standardized parameters for relationships between the latent factor and the observed variables that cross-loaded, p  0.01.  132  The theory underpinning the MBI was reviewed to determine whether the respecified cross-loadings were reasonable. Upon inspection of the three items that cross-loaded on the EE factor, two were original items from the depersonalization factor (B13DP – hardening and B17DP – not care) and one represented the personal accomplishment factor (B4PA – understand patients). These depersonalization items represent some of the feelings or thoughts that may result from being emotionally overextended. Similarly, when individuals’ emotional resources are depleted they may be less empathic toward patients. Three items from the emotional exhaustion factor (B6EE – work strain, B25EE – end of rope, and B15EE – frustrated) and one item from the PA factor (B26PA – deal calmly) cross-loaded onto the depersonalization factor. When individuals feel discouraged and exhausted they may distance themselves from others by developing indifferent or callous attitudes. Individuals feeling this way may also feel frustrated and stressed in their work, as displayed by the items from the emotional exhaustion factor. At the same time, individuals having negative or callous attitudes may not feel that they deal with emotional problems in a calm manner (Maslach, 1982). Finally, for the personal accomplishment factor, one additional item from the emotional exhaustion factor (B10EE – burned out) was cross-loaded. This item represents a general feeling of being “burned out” that may result from a reduced sense of personal competence and efficacy. In other words, some of the thoughts and feelings that accompany being emotionally overextended may result in attitudinal changes that result in feelings of inadequacy in the ability to provide care (Maslach, 1982). Accordingly, it would make sense that the aforementioned items could represent the emotional exhaustion, depersonalization and personal accomplishment constructs as identified by significant parameter estimates that were cross-loaded on these three factors. The final measurement model for the MBI-HSS was a three-factor structure with 20 items and 8 cross-loads.  5.8  Examination of Missing Data for the Study Variables Given the iterative nature of the modelling process, the examination of missing data  occurred during the initial data screening process as well as during the testing of the measurement models for the study variables. As previously indicated, the work values attribute was the only construct that required the use of a scale to measure actual relational diversity. Eighty-six percent of the respondents (n = 518) completed all the items of the Contemporary Work Values Scale and 66 respondents (10.9%) had a missing value for only one of the items. 133  Based on the findings of several factor analyses the structural model with the best fit was a one-factor structure with 16 indicators. Ninety-six percent of the respondents (n = 576) completed the 16 items of the Contemporary Work Values Scale and 25 respondents (4.1%) had a missing value for only one of the items (see Table 5.13). To measure perceived relational diversity, the only attribute that required the use of more than one item was work values. Missing items on the perceived diversity variables were minimal and within the normally acceptable range (less than 5.0%) (see Table 5.13). This was also the case with the measures used to assess the mediator conflict variables, that is, the Intragroup Conflict Scale and the Individual Conflict Scale. Of the 603 respondents that completed the survey, 571 to 582 (94.7% to 96.5%) completed the various MBI items with slight differences observed for the subscales (see Table 5.13). For additional information about the process followed to handle the missing data see Chapter 4 (pages 93 and 94).  134  Table 5.13  Frequency of Missing Data for the Study Variables Complete data  Frequency (%) 1 Item More than 1 a missing Item missing  ACTUAL DIVERSITY Age Education Ethnicity/Race Work Values  585 (97.0) 602 (99.8) 598 (99.2) 603 (100)  18 (3.0) 1 (0.2) 5 (0.8) 0 (0)  0 0 0 0  18 (3.0) 1 (0.2) 5 (0.8) 0 (0)  CONTEMPORARY WORK VALUES SCALE Revised 16-item  576 (95.5)  25 (4.1)  2 (0.3)  27 (4.4)  PERCEIVED DIVERSITY ITEM/SCALE Age Education Ethnicity/Race Work Values  600 (99.5) 602 99.8) 601 (99.7) 596 (98.8)  3 (0.5) 1 (0.2) 2 (0.3) 6 (1.0)  0 0 0 1(0.2)  3 (0.5) 1 (0.2) 2 (0.3) 7 (1.2)  MASLACH BURNOUT INVENTORY (MBI) Emotional Exhaustion Subscale Depersonalization Subscale Personal Accomplishment Subscale  571 (94.7) 582 (96.5) 576 (95.5)  28 (4.6) 18 (3.0) 23 (3.8)  4 (0.7) 3 (0.5) 4 (0.8)  32 (5.3) 21 (3.5) 27 (4.6)  INTRAGROUP CONFLICT SCALE Relationship Conflict Task Conflict  589 (97.7) 597 (99.0)  12 (2.2) 4 (0.7)  1 (0.2) 2 (0.4)  13 (2.4) 6 (1.1)  INDIVIDUAL CONFLICT SCALE Relationship Conflict Task Conflict  601 (99.7) 594 (98.5)  1 (0.2) 8 (1.3)  1 (0.2) 1 (0.2)  2 (0.4) 9 (1.5)  Variable  Total missing  Note. N = 603. a Maximum number missing is 4 items, except for the Intragroup and Individual Conflict Scales, which had 1 person missing 11 and 12 items, respectively.  In preparation for the evaluation of the structural models, the study variables were examined to determine the incidence and pattern of missing data using the Mplus 5.1 software program. No variables of concern were identified and no prominent missing data patterns emerged (see Table 5.14). The missing data were either missing completely at random (MCAR) or missing at random (MAR). The proportion of data present for covariance coverage ranged from 0.96 to 1.00 for the MBI scale and 0.99 to 1.00 for all other scales, meaning all the variables and pairs of variables had 96% or more data present for analysis. The percentage of missing data, at  4%, was within the acceptable range of no more than 5%.  135  Table 5.14  Summary of Missing Data Patterns Scale  Contemporary Work Values (16-item)  Total # of Patterns 13  Missing Data Pattern Frequency Less Than or Equal to 5 6 to 10 Respondents Respondents 13 0  Perceived Work Values Scale  3  3  0  Intragroup Conflict Scale (2 factor)  10  10  0  Individual Conflict Scale (2 factor)  5  5  0  Maslach Burnout Inventory (3 factor)  27  24  3  5.9  Descriptive Statistics of the Exogenous Variables: Relational Diversity Relational diversity was measured with items about “actual” diversity and “perceived”  diversity in a workgroup. This section provides the descriptive statistics for the diversity variables and compares group differences between Sites A and B (see Table 5.15). 22 5.9.1  Actual Diversity To measure actual diversity in age, education, and ethnicity/race, the respondents  were asked one question for each attribute. All diversity scores were scaled such that a large value referred to greater diversity (i.e., focal individuals with higher age D-scores were more different from others within their workgroup, than were those with lower age D-scores) (Tsui & Gutek, 1999). Observed age in years was treated as a continuous variable resulting in D-scores that ranged from 8.98 to 29.25 (see Appendix G1). Highest level of education and ethnicity/race were treated as categorical. Accordingly, the education D-scores ranged from 0.56 to 0.98 (see Appendix G2) and ethnicity/race D-scores ranged from 0.50 to 1.00 (see Appendix G3). Site A and Site B were significantly different on the actual educational diversity and actual ethnic/racial diversity variables (see Table 5.15).  22  The appropriate statistic for bivariate analysis was chosen by whether there was a significant amount of skew present, which was determined by dividing the skewness value by the standard error of skewness. Values above or below  1.96 were considered significantly skewed (p = 0.05) and thus required a nonparametric test, such as the Mann Whitney U Test (Munro, 2001).  136  Table 5.15  Descriptive Statistics and Hospital-based Group Comparisons of the Study Variables  Variable  Between group comparison statistic  SD  Median Site A  Median Site B  Median Total sample  Skewness  3.9 0.1 0.2 0.1  14.4 0.8 0.7 0.4  14.0 0.7 0.8 0.4  14.2 0.8 0.7 0.4  0.8* 0.0 0.1 1.2*  0.2* -0.8* -1.6* 2.0*  Z = -0.97 Z = -5.05*** Z = -3.29*** Z = -1.17  5.7 0.4  52.0 3.3  53.0 3.3  53.0 3.3  0.0 0.0  -0.4* -0.4*  Z = -0.25 Z = -0.14  3.5 3.0 3.9 11.8  1.3 1.4 1.3 4.2  4.0 3.0 4.0 11.0  4.0 3.0 4.0 12.0  4.0 3.0 4.0 12.0  -0.1 0.2 -0.4* 0.3*  -0.4* -0.8* -0.3 -0.1  Z = -1.58 Z = -0.59 Z = -0.07 Z = -1.31  12.8 10.1  4.2 3.2  11.0 9.0  12.0 10.0  12.0 9.0  0.7* 0.6*  0.3 0.0   = -3.13***  = -2.98***  2.5 2.4  6.0 8.0  7.0 8.0  7.0 8.0  1.3* 0.6*  3.0* 0.8*   = -1.49  = -2.08**  10.9  22.0  21.0  22.0  0.3*  -0.5*  Z = -0.12  5.6 6.3  4.0 38.0  5.0 38.0  4.0 38.0  1.2* -0.7*  1.1* 0.9*   = -0.57  = -0.31  Mean  EXOGENOUS VARIABLES Actual Diversity Ageb 14.8 Educationc 0.8 d 0.8 Ethnicity/race Work Values 0.5 Contemporary Work Values 53.1 Total score (16 item) Average total score 3.3  a  Kurtosis  (16 item)  Perceived Diversity e Age Educationc f Ethnicity/race Work Valuesc MEDIATOR VARIABLES Intragroup Conflict Scale Relationshipc f Task  Individual Conflict Scale Relationshipc 6.9 Taskc 7.7 ENDOGENOUS VARIABLES Maslach Burnout Inventoryg Emotional 22.4 Exhaustion Depersonalization 5.7 Personal 37.1 Accomplishment  Note. N = 603 unless otherwise specified. a The measure of skewness and kurtosis was divided by the standard error of skewness and kurtosis, respectively. This calculation resulted in a number interpreted in terms of the normal curve (z-score). Values above +1.96 or below -1.96 were considered significant at p = 0.05 (Munro, 2001). bTotal missing = 18. cTotal missing = 1. dTotal missing = 5. e4 items, subscales total; Total missing = 3. fTotal missing = 2. gTotal missing = 4 on all subscales, except depersonalization which had 3 missing. Values reported for subscale totals. * Significantly skewed or kurtotic based on measure of skewness and kurtosis (z-score). The Mann Whitney U statistic was used to compare group responses for Site A and Site B. ** p  0.05. ***p  0.01.  To calculate a D-score for actual work values diversity, the respondents were asked 35-items using the Contemporary Work Values Scale. For this instrument, a 4-point Likert scale (1 = strongly disagree and 4 = strongly agree) was used. Higher scale scores indicated stronger contemporary work values. Based on the findings from several factor analyses the measurement model with the best fit was a one-factor structure with 16 indicators (possible range = 16 to 64). The 16 items were summed to create a total score. The mean of the total score for the 16-item 137  scale was 53.1 (SD = 5.7, n = 603) (see Table 5.15) and the possible score values ranged from 32 to 64. Because the total score did not take into consideration the number of items missing for each respondent, an average of the total score (herein referred to as the average total score) was calculated for each respondent. To compute the average total score for the CWV scale, a criterion was established that at least 14 items (88% of items answered) had to be completed for the case to be included in the analyses. The average total score was then used to calculate the D-score for the work values attribute. As previously indicated, 95.5% (n = 576) of respondents answered all 16 items and 100% (n = 603) answered at least of 14 items. Thus, all respondents (n = 603) were included in the analyses. The mean of the average total score for the 16-item scale was 3.3 (SD = 0.4, n = 603) (see Table 5.15) and the scores ranged from 2.0 to 4.0 (see Appendix G4). The D-scores for work values, which were treated as a continuous variable, ranged from 0.29 to 1.24 (see Appendix G5). The actual work values diversity scores did not significantly differ between Sites A and B (see Table 5.15). 5.9.2  Perceived Diversity One item for each variable was used to measure perceived diversity in age, education,  and ethnicity/race. Perceived diversity in work values was measured with 4 items. All items used a 6-point Likert scale where 1 = not at all similar and 6 = very similar. Scale items were reverse scored so that a higher score indicated a greater degree of perceived diversity. A criterion was established that at least three of the four items (75%) were necessary for the case to be included in the analyses. Consequently, only one case was excluded (see Table 5.13). The items were summed to create a total score (possible score range = 3 to 24). The mean for perceived age diversity was 3.5 (SD = 1.3, range = 1 to 6), for education was 3.0 (SD = 1.4, range = 1 to 6), and ethnicity/race was 3.9 (SD = 1.3, range = 1 to 6) (see Table 5.13 and Appendices G6 to G8, respectively). The mean of the total score for the perceived diversity in work values variable was 11.8 (SD = 4.2) (see Table 5.15 and Appendix G9). Site A and Site B were not significantly different from each other on the perceived diversity variables (see Table 5.15).  138  5.10  Descriptive Statistics of the Mediator Variables: Interpersonal Conflict This section provides the descriptive statistics for the mediator variables: individual  perception of workgroup conflict and individual involvement in conflict. Group differences between responses from Site A and Site B respondents were examined. 5.10.1  Intragroup Conflict Scale Using a 5-point scale (1 = none and 5 = a lot), the nurses were asked 12 items about  their perceptions of relationship, task, and process disagreements occurring among members of their nursing unit (Intragroup Conflict Scale); however, only 9 items were retained as part of the final measurement model that was comprised of two factors. Higher scale scores indicated more conflict among members of the nursing unit. Conflict items in each subscale were summed to create a total score. The mean of the total score for the relationship conflict subscale (5 items, minimum of 4 items required) was 12.3 (SD = 4.2) (possible score range = 4 to 25) (see Table 5.15 and Appendix G10). This was slightly higher than the total score for the task conflict subscale (4 items, minimum 3 items required), which had a mean of 10.1 (SD = 3.2) (possible score range = 3 to 20) (see Table 5.15 and Appendix G11). Site A and Site B were statistically significantly different from each other for both intragroup relationship and task conflict. 5.10.2  Individual Conflict Scale The nurses were asked 11 questions about their involvement in conflict with their  coworkers (Individual Conflict Scale). The final measurement structure was comprised of two subscales, relationship and task conflict, with eight items and one item cross-loading. Higher scale scores indicated a higher level of conflict. Only one item was allowed to be missing for the case to be included in the analyses. Based on a 5-point scale (1 = none and 5 = a lot), the possible score range for relationship conflict was 3 to 20 and task conflict was 4 to 25. The relationship and task conflict scores had means of 6.9 and 7.7, respectively (SD = 2.4 and 2.5, respectively) (see Table 5.15 and Appendix G12 and G13). The means for the individual conflict subscales were slightly lower than the intragroup conflict scores. Site A and Site B were significantly different from each other on the individual task conflict measure.  139  5.11  Descriptive Statistics of the Outcome Variable: Burnout This section provides the descriptive statistics and group differences for the outcome  variable burnout. The final measurement structure resulted in a model with three factors (emotional exhaustion, depersonalization, and personal accomplishment) and 20 items with 8 cross-loadings. The Maslach Burnout Inventory uses a 7-point scale (0 = never and 6 = every day). Excluded from the analyses were individuals with more than one item missing. High scores on the emotional exhaustion and depersonalization subscales and low scores on the personal accomplishment subscale are suggestive of a high degree of burnout (i.e., the upper third of the normative distribution). “Average” and “low” scores are represented by the middle third and lower third of the normative distribution, respectively (Maslach et al., 1996). Based on the established normative criteria for medical workers (see Table 4.4, page 88), the total scores for each burnout subscale in this study were within the average range (see Table 5.16). Table 5.15 shows the descriptive statistics for the burnout variables (also see Appendices G14 to G16). The mean of the total score for the emotional exhaustion subscale was 22.4 (SD = 10.9, n = 599). However, 34.6% of the nurses (n = 207) scored “high” (normative range 27 or more) on the emotional exhaustion subscale while 38.7% (n = 232) and 26.7% (n = 160) were within the “low” and “average” ranges, respectively. The total scores for the depersonalization subscale had a mean of 5.7 (SD = 5.6, n = 600). Almost two thirds (59.8%) of the nurses (n = 359) scored “low” (normative range 5 and less) on the depersonalization subscale while 19.5% (n = 117) and 20.6% (n = 124) were within the “average” and “high” ranges, respectively. The mean for the personal accomplishment subscale total scores was 37.1 (SD = 6.3, n = 599). On the personal accomplishment subscale, only one quarter (24.9%) of the nurses scored “high” (normative range 33 or less) while 37.9% (n = 227) and 37.2% (n = 223) were within the “average” and “low” ranges, respectively. Sites A and B were not significantly different from each other on the burnout scores.  140  Table 5.16  Percentage of Nurses Classified as Having High, Moderate, and Low Levels of Burnout for Each Aspect of the MBI  MBI subscales Low (lower third)  Emotional exhaustion Depersonalization Personal accomplishment  5.12  38.7% 59.8% 37.2%  Range of experienced burnout Average High (middle third) (upper third)  26.7% 19.5% 37.9%  34.6% 20.6% 24.9%  Bivariate Statistics of the Study Variables All the perceived diversity variables were statistically significantly, albeit very  modestly correlated, with their corresponding actual diversity variable (e.g., perceived age diversity with actual age diversity r = 0.17), except for work values diversity, which was r = -0.01 (see Table 5.17). The low correlations between actual and perceived measures have been documented elsewhere (Riordan, 1997; Riordan & Holliday Wayne, 2008; Williams, Parker, & Turner, 2007). Correlations between the diversity variables and the burnout variables are found in Table 5.17. Perceived work values diversity had the largest correlations with the burnout variables (r = -0.23 to 0.19). Table 5.18 shows the correlations between the conflict variables and the diversity variables. Perceived educational diversity and perceived work values diversity were significantly correlated with relationship and task conflict. Age was the only “actual” diversity attribute significantly correlated with individual relationship conflict. Both intragroup conflict subscales were significantly correlated with emotional exhaustion, depersonalization, and personal accomplishment (see Table 5.19).  141  Table 5.17  Pearson Correlation Matrix for the Diversity and Burnout Latent Variables and the Observed Demographic Variables 1  1  Emotional exhaustion Depersonalization Personal accomplishment Perceived age diversity Perceived educational diversity Perceived ethnic/racial diversity Perceived work values diversity Age Level of education Ethnicity/race  2 3 4 5 6 7 8 9 10 11 12 13 14 15 *  Work values Actual age diversity Actual educational diversity Actual ethnic/racial diversity Actual work values diversity **  p < 0.05. p< 0.01.  2  3  4  5  6  7  8  9  10  11  12  13  14  15  1.00 0.53  1.00 ***  -0.13  -0.46***  1.00  -0.01  -0.09  -0.02  1.00  0.03  0.13**  -0.18***  0.32***  0.01  0.01  -0.03  0.22  ***  0.25***  1.00  0.14***  0.19***  -0.23***  0.32***  0.38***  0.28***  1.00  ***  ***  ***  1.00  -0.04 0.05 -0.08*  -0.14 0.09 -0.10*  0.13 -0.09 -0.12**  0.20 -0.12** -0.01  0.00 -0.10* -0.05  0.03 -0.06 0.05  0.06 0.04 0.02  1.00 -0.43*** -0.04  1.00 0.13***  1.00  -0.04  -0.17***  0.32***  0.00  -0.07  0.01  -0.08*  0.04  0.10*  0.13***  1.00  *  *  *  *  ***  -0.09  -0.09  0.11  0.17  -0.02  -0.06  -0.11  -0.04  -0.03  0.03  0.03  0.14***  0.05  -0.12  -0.04  -0.10*  0.00  -0.03  0.01  0.10*  0.09*  -0.02  -0.06  **  ***  p< 0.001.  **  ***  0.17  0.08  -0.10  0.03  1.00  0.04  -0.09*  -0.23***  0.02  -0.01  0.06  1.00  0.15***  -0.02  -0.03  0.01  0.64***  0.17***  -0.10*  0.11**  1.00  0.06  -0.01  0.06  -0.11**  0.04  0.03  -0.04  0.04  0.03  1.00  142  Table 5.18 1 2 3 4 5 6 7 8 9 10 11 12 *  Pearson Correlation Matrix for the Perceived Diversity, Actual Diversity, and Conflict Latent Variables  Individual relationship conflict Individual task conflict Intragroup relationship conflict Intragroup task conflict Perceived age diversity Perceived educational diversity Perceived ethnic/racial diversity Perceived work values diversity Actual age diversity Actual educational diversity Actual ethnic/racial diversity Actual work values diversity  p < 0.05.  **  p< 0.01. ***p< 0.001.  1 1.00 0.89*** 0.64*** 0.62*** 0.03 0.19*** 0.08 0.31*** -0.10* 0.03 -0.02 0.04  2  3  4  5  6  7  1.00 0.66*** 0.75*** 0.03 0.17*** 0.07 0.36*** -0.06 0.03 -0.04 0.08  1.00 0.97*** 0.05 0.16*** 0.14*** 0.40*** -0.10* 0.00 0.00 0.01  1.00 0.03 0.14*** 0.14*** 0.43*** -0.08 0.00 0.02 0.07  1.00 0.32*** 0.22*** 0.32*** 0.17*** 0.03 0.00 -0.02  1.00 0.25*** 0.38*** -0.02 0.14*** -0.03 -0.06  1.00 0.28*** -0.06 0.05 0.15*** 0.06  8  1.00 -0.11*** 0.04 -0.02 -0.01  9  1.00 0.06 -0.10* -0.04  10  11  12  1.00 0.11** 0.04  1.00 0.03  1.00  143  Table 5.19: 1 2 3 4 5 6 7  Pearson Correlation Matrix for the Conflict and Burnout Latent Variables  Emotional exhaustion Depersonalization Personal accomplishment Individual relationship conflict Individual task conflict Intragroup relationship conflict Intragroup task conflict  1 1.00 ** 0.51 -0.12* ** 0.31 0.34** ** 0.32 0.36**  2 1.00 -0.45** 0.39** 0.41** 0.25** 0.28**  3  1.00 -0.21** -0.21** -0.07 -0.08  4  1.00 0.89** 0.64** 0.62**  5  1.00 0.66** 0.75**  6  1.00 0.97**  7  1.00  **  *p < 0.01. p < 0.001.  5.13  Chapter Summary This chapter provided the details of the measurement model specification of the  exogenous, mediator, and endogenous latent variables. The Contemporary Work Values Scale was reduced to 16 items, which were subsequently summed and averaged to calculate a D-score for actual work values diversity. The perceived work values diversity variable had a measurement model that consisted of 4 items and each of the other exogenous attributes were measured with a single item. The measurement structure of the Intragroup Conflict Scale included two latent variables: relationship and task conflict. Similarly, the Individual Conflict Scale consisted of two factors: relationship and task conflict with the anger item cross-loading on both factors. The measurement model for the Maslach Burnout Inventory was a three-factor model with 20 items and 8 cross-loadings. The measurement models described in this chapter were used in the structural modelling portion of the analysis described in the next chapter.  144  6  STRUCTURAL EQUATION MODELLING FINDINGS Following data preparation and confirmation of the measurement models of the study  variables, the structural models were tested. This chapter describes the evaluation (i.e., fit and parameter estimates) of the six structural models that specified the direct and indirect effects of relational diversity on burnout as mediated by interpersonal conflict (see Table 6.1):23   Model 1: Actual diversity  burnout    Model 2: Actual diversity  intragroup conflict  burnout  Model 2a: Actual diversity  intragroup relationship conflict  burnout  Model 2b: Actual diversity  intragroup task conflict  burnout    Model 3: Actual diversity  individual conflict  burnout  Model 3a: Actual diversity  individual relationship conflict  burnout  Model 3b: Actual diversity  individual task conflict  burnout    Model 4: Perceived diversity  burnout    Model 5: Perceived diversity  intragroup conflict  burnout  Model 5a: Perceived diversity  intragroup relationship conflict  burnout  Model 5b: Perceived diversity  intragroup task conflict  burnout    Model 6: Perceived diversity  individual conflict  burnout  Model 6a: Perceived diversity  individual relationship conflict  burnout  Model 6b: Perceived diversity  individual task conflict  burnout  23  These models speak to the hypotheses delineated in Chapter 3; however, the components of the hypotheses have changed (i.e., the removal of process conflict and cynicism) because the measurement models did not permit further exploration of some hypotheses. As well, the numbering of the hypotheses was changed to establish continuity in the reporting of the models.  145  Table 6.1  Summary of Variables in Each Model  Study constructs  Instrument/item  EXOGENOUS VARIABLES – RELATIONAL DIVERSITY Actual Age Diversity D-score for Age (DSAge) Actual Educational diversity D-score for Education (DSEduc) Actual Ethnic/racial diversity D-score for Ethnicity/race (DSEth) Actual work values diversity D-score for Work Values (DSVal) Contemporary Work Values Scale Perceived age diversity Perceived Age Diversity (PAge) Perceived educational Perceived Educational Diversity diversity (PEduc) Perceived ethnic/racial Perceived Ethnic/racial Diversity diversity (PEth) Perceived work values Perceived Work Values Diversity Scale diversity (PVal) MEDIATOR VARIABLES – INTERPERSONAL CONFLICT Relationship Conflict Intragroup Conflict Relationship Subscale (REL CON) Individual Conflict Relationship Subscale Task Conflict Intragroup Conflict Task Subscale (TSK CON) Individual Conflict Task Subscale ENDOGENOUS VARIABLES – BURNOUT Burnout MBI – Emotional Exhaustion Subscale (EE) MBI – Depersonalization Subscale (DP) MBI – Personal Accomplishment Subscale (PA)  1  2  Model 3 4 5  X  X  X  X  X  X  X  X  X  X  X  X  6  X  X  X  X  X  X  X  X  X  X  X  X  X  X X  X  X  X X  X X X  X X X  X X X  X  X X X  X X X  X X X  6.1 Overview of Methods As explained in Chapter 4, the data were treated as non-normal, and the indicators for all of the observed variables, except for the actual diversity variables, were treated as ordered categorical (ordinal) latent variables for the analyses. The actual diversity variables were treated as continuous manifest variables and the perceived diversity variables were modelled as categorical latent variables. The Mplus 5.1 software program with robust mean and variance adjusted weighted least squares (WLSMV) estimation was used for the structural equation modelling portion of the analyses. To evaluate model fit, the same criteria applied for the confirmatory factor analyses were applied: CFI  0.95, TLI  0.95, and RMSEA  0.08 with a preferred value of 0.06 being indicative of a well-fitting model (Brown, 2006; Hu & Bentler, 1999; Schumacker & Lomax, 2004).  146  A four-step process was used to test the mediation models (see pages 98 to 100). Given the number of variables included in the mediator model (and the possibility of interactions) (MacKinnon, 2008), single–mediator models were tested and reported separately (i.e., one mediator model for relationship conflict and one for task conflict). An omnibus model of mediation was estimated, which resulted in the same findings. All of the hypothesized pathways were included in the initial mediator analyses to determine nonsignificance of the total indirect effects. After nonsignificance was confirmed, these pathways were removed (e.g., perceived ethnic/racial diversity).  6.2 Organization of the Findings In this chapter, the initial discussion of the findings for the actual relational diversity models (Models 1, 2, and 3) is followed by a discussion of the perceived relational diversity models (Models 4, 5, and 6). The analysis begins with a summary of the goodness-of-fit indices for the direct and indirect models. Next, the tests of the hypotheses are reported (e.g., values, direction, and significance of the parameter estimates), first for the direct effects and then for the indirect (mediation) effects. For example, the direct, unmediated paths between the actual diversity attributes and the three aspects of burnout (Condition 1) were estimated before the inclusion of relationship and task conflict into the model. Next, the significance of the indirect effects (Conditions 2, 3, and 4) was tested by including relationship and task conflict into the model. Once significance of these conditions was determined, the direct effect (c’) was compared with the overall direct effect (c). If mediation was established, the effect size of the total effect was calculated (see page 100).  6.3 The Direct and Indirect Effects of Actual Relational Diversity on Burnout This section reports the findings of the examination of the four actual diversity attributes on emotional exhaustion, depersonalization, and personal accomplishment (Model 1 see Figure 6.1). Next, the conclusions of the four single–mediator models are presented. Model 2 tests intragroup relationship and task conflict as mediators of the association between actual diversity and burnout (see Figure 6.2 and Figure 6.3). Model 3 tests the mediator models with individual relationship and task conflict (see Figure 6.2 and Figure 6.3).  147  Figure 6.1  Model 1: The Effects of Actual Relational Diversity on Burnout DS Age  EE  DS Educ  DP DS Eth  PA DS Val  DSAge = Actual age diversity DSEduc = Actual educational diversity DSEth = Actual ethnic/racial diversity DSVal = Actual work values diversity  6.3.1  EE = Emotional exhaustion DP = Depersonalization PA = Personal accomplishment  Model Fit Table 6.2 summarizes the goodness-of-fit indices and total variance explained in the  actual diversity models. Model 1 demonstrated acceptable fit with the data. The total variance explained for each endogenous latent variable was minimal, ranging from 2% to 3%. Model 2 hypothesized that the effects of each actual diversity attribute on the various aspects of burnout were mediated by intragroup relationship conflict (Model 2a, see Figure 6.2) and task conflict (Model 2b, see Figure 6.3). Model 3 examined the indirect effects of individual relationship conflict (Model 3a, see Figure 6.2) and task conflict (Model 3b, see Figure 6.3). All the mediator models demonstrated acceptable fit with the data, and the total variance explained ranged from 1% to 19% (see Table 6.2).  148  Table 6.2  Model  Summary of the Goodness-of-Fit Indices and Total Variance Explained for the Effects of Actual Relational Diversity on Burnouta Fit indices  Total variance explained for endogenous latent variables  DIRECT EFFECT MODEL EE = 2% Model 1 2 (97) = 388.22, p  0.001 DP = 2% CFI = 0.95 PA = 3% TLI = 0.97 RMSEA = 0.07 SINGLE–MEDIATOR MODELS (INTRAGROUP CONFLICT) EE = 12% Model 2a 2 (116) = 346.05, p  0.001 DP = 9% CFI = 0.97 PA = 4% TLI = 0.98 Intragroup REL conflict = 1% RMSEA = 0.06 EE = 15% Model 2b 2 (118) = 352.78, p  0.001 DP = 10% CFI = 0.96 PA = 4% TLI = 0.98 Intragroup TSK conflict = 1% RMSEA = 0.06 SINGLE–MEDIATOR MODELS (INDIVIDUAL CONFLICT) EE = 13% Model 3a 2 (123) = 400.09, p  0.001 DP = 19% CFI = 0.96 PA = 8% TLI = 0.98 Individual REL conflict = 1% RMSEA = 0.06 2 EE = 14% Model 3b  (128) = 414.10, p  0.001 DP = 19% CFI = 0.95 PA = 8% TLI = 0.97 Individual TSK conflict = 1% RMSEA = 0.06 Notes. N = 603. WLSMV estimator. EE = Emotional exhaustion, DP = Depersonalization, PA = Personal accomplishment, REL = Relationship, and TSK = Task. a Fit indices and total variance explained for the multiple–mediator models:  Actual relational diversity on burnout as mediated by intragroup relationship and task conflict, 2 (120) = 320.93, p  0.001, CFI = 0.97, TLI = 0.99, and RMSEA = 0.05; EE = 17%, DP = 11%, PA = 4%, intragroup relationship conflict = 1%, and intragroup task conflict = 1%.  Actual relational diversity on burnout as mediated by individual relationship and task conflict, 2 (134) = 383.34, p  0.001, CFI = 0.96, TLI = 0.98, and RMSEA = 0.06; EE = 14%, DP = 19%, PA = 8%, individual relationship conflict = 1%, and individual task conflict = 1%.  149  Figure 6.2  Model 2a and 3a: The Effects of Actual Relational Diversity on Burnout as Mediated by Relationship Conflict  REL CON DS Age  EE DS Educ  DP  DS Eth  PA  DS Val  DSAge = Actual age diversity DSEduc = Actual educational diversity DSEth = Actual ethnic/racial diversity DSVal = Actual work values diversity  REL CON = Relationship conflict (the model is the same for both Intragroup and Individual conflict)  EE = Emotional exhaustion DP = Depersonalization PA = Personal accomplishment  150  Figure 6.3  Model 2b and 3b: The Effects of Actual Relational Diversity on Burnout as Mediated by Task Conflict  TSK CON DS Age  EE DS Educ  DP  DS Eth  PA  DS Val  DSAge = Actual age diversity DSEduc = Actual educational diversity DSEth = Actual ethnic/racial diversity DSVal = Actual work values diversity  TSK CON = Task conflict (the model is the same for both Intragroup and Individual conflict)  EE = Emotional exhaustion DP = Depersonalization PA = Personal accomplishment  151  6.3.2  Model 1: The Direct Effects of Actual Relational Diversity on Burnout (Condition 1)  H1.1:  Actual age diversity between an individual and others within a workgroup is positively associated with EE and DP, and is negatively associated with PA.  H1.2:  Actual educational diversity between an individual and others within a workgroup is positively associated with EE and DP, and is negatively associated with PA.  H1.3:  Actual ethnic/racial diversity between an individual and others within a workgroup is positively associated with EE and DP, and is negatively associated with PA. For Hypotheses 1.1 to 1.3, act