0%

Book Description

Discover best practices for real world data research with SAS code and examples

Real world health care data is common and growing in use with sources such as observational studies, patient registries, electronic medical record databases, insurance healthcare claims databases, as well as data from pragmatic trials. This data serves as the basis for the growing use of real world evidence in medical decision-making. However, the data itself is not evidence. Analytical methods must be used to turn real world data into valid and meaningful evidence. Real World Health Care Data Analysis: Causal Methods and Implementation Using SAS brings together best practices for causal comparative effectiveness analyses based on real world data in a single location and provides SAS code and examples to make the analyses relatively easy and efficient.

The book focuses on analytic methods adjusted for time-independent confounding, which are useful when comparing the effect of different potential interventions on some outcome of interest when there is no randomization. These methods include:

  • propensity score matching, stratification methods, weighting methods, regression methods, and approaches that combine and average across these methods
  • methods for comparing two interventions as well as comparisons between three or more interventions
  • algorithms for personalized medicine
  • sensitivity analyses for unmeasured confounding

Table of Contents

  1. Contents
  2. About the Book
    1. What Does This Book Cover?
    2. Is This Book for You?
    3. What Should You Know about the Examples?
      1. Software Used to Develop the Book’s Content
      2. Example Code and Data
    4. Acknowledgments
    5. We Want to Hear from You
  3. About the Authors
  4. Chapter 1: Introduction to Observational and Real World Evidence Research
    1. 1.1 Why This Book?
    2. 1.2 Definition and Types of Real World Data (RWD)
    3. 1.3 Experimental Versus Observational Research
    4. 1.4 Types of Real World Studies
      1. 1.4.1 Cross-sectional Studies
      2. 1.4.2 Retrospective or Case-control Studies
      3. 1.4.3 Prospective or Cohort Studies
    5. 1.5 Questions Addressed by Real World Studies
    6. 1.6 The Issues: Bias and Confounding
      1. 1.6.1 Selection Bias
      2. 1.6.2 Information Bias
      3. 1.6.3 Confounding
    7. 1.7 Guidance for Real World Research
    8. 1.8 Best Practices for Real World Research
    9. 1.9 Contents of This Book
    10. References
  5. Chapter 2: Causal Inference and Comparative Effectiveness: A Foundation
    1. 2.1 Introduction
    2. 2.2 Causation
    3. 2.3 From R.A. Fisher to Modern Causal Inference Analyses
      1. 2.3.1 Fisher’s Randomized Experiment
      2. 2.3.2 Neyman’s Potential Outcome Notation
      3. 2.3.3 Rubin’s Causal Model
      4. 2.3.4 Pearl’s Causal Model
    4. 2.4 Estimands
    5. 2.5 Totality of Evidence: Replication, Exploratory, and Sensitivity Analyses
    6. 2.6 Summary
    7. References
  6. Chapter 3: Data Examples and Simulations
    1. 3.1 Introduction
    2. 3.2 The REFLECTIONS Study
    3. 3.3 The Lindner Study
    4. 3.4 Simulations
    5. 3.5 Analysis Data Set Examples
      1. 3.5.1 Simulated REFLECTIONS Data
      2. 3.5.2 Simulated PCI Data
    6. 3.6 Summary
    7. References
  7. Chapter 4: The Propensity Score
    1. 4.1 Introduction
    2. 4.2 Estimate Propensity Score
      1. 4.2.1 Selection of Covariates
      2. 4.2.2 Address Missing Covariates Values in Estimating Propensity Score
      3. 4.2.3 Selection of Propensity Score Estimation Model
      4. 4.2.4 The Criteria of “Good” Propensity Score Estimate
    3. 4.3 Example: Estimate Propensity Scores Using the Simulated REFLECTIONS Data
      1. 4.3.1 A Priori Logistic Model
      2. 4.3.2 Automatic Logistic Model Selection
      3. 4.3.3 Boosted CART Model
    4. 4.4 Summary
    5. References
  8. Chapter 5: Before You Analyze – Feasibility Assessment
    1. 5.1 Introduction
    2. 5.2 Best Practices for Assessing Feasibility: Common Support
      1. 5.2.1 Walker’s Preference Score and Clinical Equipoise
      2. 5.2.2 Standardized Differences in Means and Variance Ratios
      3. 5.2.3 Tipton’s Index
      4. 5.2.4 Proportion of Near Matches
      5. 5.2.4 Proportion of Near Matches
      6. 5.2.5 Trimming the Population
    3. 5.3 Best Practices for Assessing Feasibility: Assessing Balance
      1. 5.3.1 The Standardized Difference for Assessing Balance at the Individual Covariate Level
      2. 5.3.2 The Prognostic Score for Assessing Balance
    4. 5.4 Example: REFLECTIONS Data
      1. 5.4.1 Feasibility Assessment Using the Reflections Data
      2. 5.4.2 Balance Assessment Using the Reflections Data
    5. 5.5 Summary
    6. References
  9. Chapter 6: Matching Methods for Estimating Causal Treatment Effects
    1. 6.1 Introduction
    2. 6.2 Distance Metrics
      1. 6.2.1 Exact Distance Measure
      2. 6.2.2 Mahalanobis Distance Measure
      3. 6.2.3 Propensity Score Distance Measure
      4. 6.2.4 Linear Propensity Score Distance Measure
      5. 6.2.5 Some Considerations in Choosing Distance Measures
    3. 6.3 Matching Constraints
      1. 6.3.1 Calipers
      2. 6.3.2 Matching With and Without Replacement
      3. 6.3.3 Fixed Ratio Versus Variable Ratio Matching
    4. 6.4 Matching Algorithms
      1. 6.4.1 Nearest Neighbor Matching
      2. 6.4.2 Optimal Matching
      3. 6.4.3 Variable Ratio Matching
      4. 6.4.4 Full Matching
      5. 6.4.5 Discussion: Selecting the Matching Constraints and Algorithm
    5. 6.5 Example: Matching Methods Applied to the Simulated REFLECTIONS Data
      1. 6.5.1 Data Description
      2. 6.5.2 Computation of Different Matching Methods
      3. 6.5.3 1:1 Nearest Neighbor Matching
      4. 6.5.4 1:1 Optimal Matching with Additional Exact Matching
      5. 6.5.5 1:1 Mahalanobis Distance Matching with Caliper
      6. 6.5.6 Variable Ratio Matching
      7. 6.5.7 Full Matching
    6. 6.6 Discussion Topics: Analysis on Matched Samples, Variance Estimation of the Causal Treatment Effect, and Incomplete Matching
    7. 6.7 Summary
    8. References
  10. Chapter 7: Stratification for Estimating Causal Treatment Effects
    1. 7.1 Introduction
    2. 7.2 Propensity Score Stratification
      1. 7.2.1 Forming Propensity Score Strata
      2. 7.2.2 Estimation of Treatment Effects
    3. 7.3 Local Control
      1. 7.3.1 Choice of Clustering Method and Optimal Number of Clusters
      2. 7.3.2 Confirming that the Estimated Local Effect-Size Distribution Is Not Ignorable
    4. 7.4 Stratified Analysis of the PCI15K Data
      1. 7.4.1 Propensity Score Stratified Analysis
      2. 7.4.2 Local Control Analysis
    5. 7.5 Summary
    6. References
  11. Chapter 8: Inverse Weighting and Balancing Algorithms for Estimating Causal Treatment Effects
    1. 8.1 Introduction
    2. 8.2 Inverse Probability of Treatment Weighting
    3. 8.3 Overlap Weighting
    4. 8.4 Balancing Algorithms
    5. 8.5 Example of Weighting Analyses Using the REFLECTIONS Data
      1. 8.5.1 IPTW Analysis Using PROC CAUSALTRT
      2. 8.4.2 Overlap Weighted Analysis using PROC GENMOD
      3. 8.4.3 Entropy Balancing Analysis
    6. 8.5 Summary
    7. References
  12. Chapter 9: Putting It All Together: Model Averaging
    1. 9.1 Introduction
    2. 9.2 Model Averaging for Comparative Effectiveness
      1. 9.2.1 Selection of Individual Methods
      2. 9.2.2 Computing Model Averaging Weights
      3. 9.2.3 The Model Averaging Estimator and Inferences
    3. 9.3 Frequentist Model Averaging Example Using the Simulated REFLECTIONS Data
      1. 9.3.1 Setup: Selection of Analytical Methods
      2. 9.3.2 SAS Code
      3. 9.3.3 Analysis Results
    4. 9.4 Summary
    5. References
  13. Chapter 10: Generalized Propensity Score Analyses (> 2 Treatments)
    1. 10.1 Introduction
    2. 10.2 The Generalized Propensity Score
      1. 10.2.1 Definition, Notation, and Assumptions
      2. 10.2.2 Estimating the Generalized Propensity Score
    3. 10.3 Feasibility and Balance Assessment Using the Generalized Propensity Score
      1. 10.3.1 Extensions of Feasibility and Trimming
      2. 10.3.2 Balance Assessment
    4. 10.4 Estimating Treatment Effects Using the Generalized Propensity Score
      1. 10.4.1 GPS Matching
      2. 10.4.2 Inverse Probability Weighting
      3. 10.4.3 Vector Matching
    5. 10.5 SAS Programs for Multi-Cohort Analyses
    6. 10.6 Three Treatment Group Analyses Using the Simulated REFLECTIONS Data
      1. 10.6.1 Data Overview and Trimming
      2. 10.6.2 The Generalized Propensity Score and Population Trimming
      3. 10.6.3 Balance Assessment
      4. 10.6.4 Generalized Propensity Score Matching Analysis
      5. 10.6.5 Inverse Probability Weighting Analysis
      6. 10.6.6 Vector Matching Analysis
    7. 10.7 Summary
    8. References
  14. Chapter 11: Marginal Structural Models with Inverse Probability Weighting
    1. 11.1 Introduction
    2. 11.2 Marginal Structural Models with Inverse Probability of Treatment Weighting
    3. 11.3 Example: MSM Analysis of the Simulated REFLECTIONS Data
      1. 11.3.1 Study Description
      2. 11.3.2 Data Overview
      3. 11.3.3 Causal Graph
      4. 11.3.4 Computation of Weights
      5. 11.3.5 Analysis of Causal Treatment Effects Using a Marginal Structural Model
    4. 11.4 Summary
    5. References
  15. Chapter 12: A Target Trial Approach with Dynamic Treatment Regimes and Replicates Analyses
    1. 12.1 Introduction
    2. 12.2 Dynamic Treatment Regimes and Target Trial Emulation
      1. 12.2.1 Dynamic Treatment Regimes
      2. 12.2.2 Target Trial Emulation
    3. 12.3 Example: Target Trial Approach Applied to the Simulated REFLECTIONS Data
      1. 12.3.1 Study Question
      2. 12.3.2 Study Description and Data Overview
      3. 12.3.3 Target Trial Study Protocol
      4. 12.3.4 Generating New Data
      5. 12.3.5 Creating Weights
      6. 12.3.6 Base-Case Analysis
      7. 12.3.7 Selecting the Optimal Strategy
      8. 12.3.8 Sensitivity Analyses
    4. 12.4 Summary
    5. References
  16. Chapter 13: Evaluating the Impact of Unmeasured Confounding in Observational Research
    1. 13.1 Introduction
    2. 13.2 The Toolbox: A Summary of Available Analytical Methods
    3. 13.3 The Best Practice Recommendation
    4. 13.4 Example Data Analysis Using the REFLECTIONS Study
      1. 13.4.1 Array Approach
      2. 13.4.2 Propensity Score Calibration
      3. 13.4.3 Rosenbaum-Rubin Sensitivity Analysis
      4. 13.4.4 Negative Control
      5. 13.4.5 Bayesian Twin Regression Modeling
    5. 13.5 Summary
    6. References
  17. Chapter 14: Using Real World Data to Examine the Generalizability of Randomized Trials
    1. 14.1 External Validity, Generalizability and Transportability
    2. 14.2 Methods to Increase Generalizability
    3. 14.3 Generalizability Re-weighting Methods for Generalizability
      1. 14.3.1 Inverse Probability Weighting
      2. 14.3.2 Entropy Balancing
      3. 14.3.3 Assumptions, Best Practices, and Limitations
    4. 14.4 Programs Used in Generalizability Analyses
    5. 14.5 Analysis of Generalizability Using the PCI15K Data
      1. 14.5.1 RCT and Target Populations
      2. 14.5.2 Inverse Probability Generalizability
      3. 14.5.3 Entropy Balancing Generalizability
    6. 14.6 Summary
    7. References
  18. Chapter 15: Personalized Medicine, Machine Learning, and Real World Data
    1. 15.1 Introduction
    2. 15.2 Individualized Treatment Recommendation
      1. 15.2.1 The Individualized Treatment Recommendation Framework
      2. 15.2.2 Estimating the Optimal Individualized Treatment Rule
      3. 15.2.3 Multi-Category ITR
    3. 15.3 Programs for ITR
    4. 15.4 Example Using the Simulated REFLECTIONS Data
    5. 15.5 “Most Like Me” Displays: A Graphical Approach
      1. 15.5.1 Most Like Me Computations
      2. 15.5.2 Background Information: LTD Distributions from the PCI15K Local Control Analysis
      3. 15.5.3 Most Like Me Example Using the PCI15K Data Set
      4. 15.5.4 Extensions and Interpretations of Most Like Me Displays
    6. 15.6 Summary
    7. References
  19. Index
    1. A
      1. B
      2. C
      3. D
      4. E
      5. G
      6. H
      7. I
      8. K
      9. L
      10. M
      11. N
      12. O
      13. P
      14. Q
      15. R
      16. S
      17. T
      18. U
      19. V
      20. W
      21. X
      22. Y
      23. Z
107.20.123.227