Multilevel Modelling for Public Health and Health Services Research (Health in Context) ||

This open access book is a practical introduction to multilevel modelling or multilevel analysis (MLA) – a statistical technique being increasingly used in public health and health services research. The authors begin with a compelling argument for the importance of researchers in these fields having an understanding of MLA to be able to judge not only the growing body of research that uses it, but also to recognise the limitations of research that did not use it. The volume also guides the analysis of real-life data sets by introducing and discussing the use of the multilevel modelling software MLwiN, the statistical package that is used with the example data sets. Importantly, the book also makes the training material accessible for download – not only the datasets analysed within the book, but also a freeware version of MLwiN to allow readers to work with these datasets. The book’s practical review of MLA comprises: Theoretical, conceptual, and methodological background Statistical background The modelling process and presentation of research Tutorials with example datasets Multilevel Modelling for Public Health and Health Services Research: Health in Context is a practical and timely resource for public health and health services researchers, statisticians interested in the relationships between contexts and behaviour, graduate students across these disciplines, and anyone interested in utilising multilevel modelling or multilevel analysis. “Leyland and Groenewegen’s wealth of teaching experience makes this book and its accompanying tutorials especially useful for a practical introduction to multilevel analysis.” ̶ Juan Merlo, Professor of Social Epidemiology, Lund University “Comprehensive and insightful. A must for anyone interested in the applications of multilevel modelling to population health”. ̶ S. (Subu) V. Subramanian, Professor of Population Health and Geography, Harvard University ; For researchers and students with a basic mastery of ordinary least squares and logistic regression Discusses multilevel analysis in context of public health, health services research, and epidemiology Includes an online component where users can download the datasets analyzed in the book, and also a freeware version of the multilevel modelling software MLwiN ​​​​​​​Can be used as part of a course on multilevel modelling, or as a self-training text Preface Acknowledgements Contents About the Authors Part I: Theoretical, Conceptual and Methodological Background Chapter 1: Introduction Importance of MLA for Research in Health and Care The Scope of Public Health and Health Services Research Research and Policy Conclusion References Chapter 2: Health in Context Relationships Between the Macro and Micro Levels Micro Level: Behaviour of Patients and Providers The Behaviour of Healthcare Providers The Behaviour of Patients Patient-Provider Interaction From Macro to Micro Level What Contexts Are Relevant? From Micro to Macro Level The Use of ``League Tables ́ ́ Conclusion References Chapter 3: What Is Multilevel Modelling? Methodological Background Why Use Multilevel Modelling? Aggregate Analysis Individual Analysis Separate Individual Analyses Within Each Higher Level Unit Individual-Level Analysis with Dummy Variables What Is a Multilevel Model? What Is a Level? How Many Units Do We Need at Each Level? Hypotheses That Can Be Tested with Multilevel Analysis Hypotheses About Variation Individual-Level Hypotheses Context Hypotheses Aggregated Individual-Level Characteristics Higher Level Characteristics Cross-Level Interactions Conclusion References Chapter 4: Multilevel Data Structures Strict Hierarchies: The Basic Model Multistage Sampling Designs Evaluating Community Interventions and Cluster Randomised Trials Designs Including Time Multiple Responses Non-hierarchical Structures Cross-Classified Models Multiple Membership Model Correlated Cross-Classified Model Other Multilevel Models Pseudo-levels Incomplete Hierarchies Conclusion References Part II: Statistical Background Chapter 5: Graphs and Equations Ordinary Least Squares (Single-Level) Regression Random Intercept Model Random Slope Model Three-Level Model Heteroscedasticity Fixed Effects Model Rankings and Institutional Performance Conclusion References Chapter 6: Apportioning Variation in Multilevel Models Variance Partitioning for Continuous Responses Variance Partitioning for Multilevel Logistic Regression Variance Partitioning for Models with Three or More Levels Interpretation of Variances Zero Variance Multilevel Power Calculations Software for Multilevel Power Calculations Population Average and Cluster-Specific Estimates Omitting a Level Conclusion References Part III: The Modelling Process and Presentation of Research Chapter 7: Context, Composition and How Their Influences Vary Context or Composition? Using Multilevel Modelling to Investigate Compositional and Contextual Effects Model M0: Null Model Model M1: Individual Social Capital Model M2: Neighbourhood Social Capital Model M3: Individual and Neighbourhood Social Capital Model M4: Individual and Neighbourhood Social Capital and Their Interaction Random Slopes and Cross-Level Interactions Impact of Compositional and Contextual Variables on the Variances Model Specification and Model Interpretation Sources of Error Affecting the Estimation of Contextual Effects Lack of Variation in the Contextual Variable Precision of Estimates and Study Design Selection Bias Confounding Information Bias Model Specification Conclusions References Chapter 8: Ecometrics: Using MLA to Construct Contextual Variables from Individual Data Problems with Simple Aggregation Single Variables Composite Variables: The Traditional Method Composite Variables: A Simple Multilevel Model Ecometric Approach Application of the Ecometric Approach Comparison of the Traditional and Ecometric Approach Further Ecometric Properties of the Scale Conclusions References Chapter 9: Modelling Strategies Define the Data Structure Measurement Level and Distribution of the Dependent Variable The Baseline Model Exploratory Research and Hypothesis Testing Context and Composition Modelling the Effects of Higher Level Characteristics Random Effects at Higher Levels Interpreting the Results in the Light of Common Assumptions Conclusions References Chapter 10: Reading and Writing Critical Reading What Is the Research Question? Which Levels Can Be Distinguished Theoretically? What Is the Structure of the Actual Data Used? What Statistical Model Was Used? What Was the Modelling Strategy? Does the Paper Report the Intercept Variation at Different Levels? Cross-Level Interactions What Are the Shortcomings and Strong Points of the Article? Writing Up Your Own Research The Introduction or Background Section The Methods Section The Results Section The Conclusion and Discussion Section Conclusions References Part IV: Tutorials with Example Datasets Chapter 11: Multilevel Linear Regression Using MLwiN: Mortality in England and Wales, 1979-1992 Introduction to the Dataset Research Questions Introduction to MLwiN Opening a Worksheet Names Window Data Window Graph Window Model Specification Creating New Variables Equations Window Fitting the Model Variance Components A 2-Level Variance Components Model Sorting the Data The Hierarchy Viewer Adding a Further Level Interpreting the Model Residuals Predictions Window Model Building Adding More Fixed Effects Intervals and Tests Window Random Coefficients Random Slopes Variance Function Window Higher-Level Residuals Complex Level 1 Variation A Poisson Model: Introduction Setting Up a Generalised Linear Model in MLwiN The Offset Non-linear Estimation Model Interpretation Predictions and Confidence Envelopes References Chapter 12: Multilevel Logistic Regression Using MLwiN: Referrals to Physiotherapy Multilevel Logistic Regression Model Example: Variation in the GP Referral Rate to Physiotherapy The Data Model Set-Up Non-linear Settings Model Interpretation and Model Building A Note on Estimation Further Exercises References Chapter 13: Untangling Context and Composition The Data Structure of the Analysis Estimating the Null Model Fixed Effects Additional Models References Index