13.2 The Toolbox: A Summary of Available Analytical Methods
13.3 The Best Practice Recommendation
13.4 Example Data Analysis Using the REFLECTIONS Study
13.4.1 Array Approach
13.4.2 Propensity Score Calibration
13.4.3 Rosenbaum-Rubin Sensitivity Analysis
13.4.4 Negative Control
13.4.5 Bayesian Twin Regression Modeling
13.5 Summary
References
Chapter 14: Using Real World Data to Examine the Generalizability of Randomized Trials
14.1 External Validity, Generalizability and Transportability
14.2 Methods to Increase Generalizability
14.3 Generalizability Re-weighting Methods for Generalizability
14.3.1 Inverse Probability Weighting
14.3.2 Entropy Balancing
14.3.3 Assumptions, Best Practices, and Limitations
14.4 Programs Used in Generalizability Analyses
14.5 Analysis of Generalizability Using the PCI15K Data
14.5.1 RCT and Target Populations
14.5.2 Inverse Probability Generalizability
14.5.3 Entropy Balancing Generalizability
14.6 Summary
References
Chapter 15: Personalized Medicine, Machine Learning, and Real World Data
15.1 Introduction
15.2 Individualized Treatment Recommendation
15.2.1 The Individualized Treatment Recommendation Framework
15.2.2 Estimating the Optimal Individualized Treatment Rule
15.2.3 Multi-Category ITR
15.3 Programs for ITR
15.4 Example Using the Simulated REFLECTIONS Data
15.5 “Most Like Me” Displays: A Graphical Approach
15.5.1 Most Like Me Computations
15.5.2 Background Information: LTD Distributions from the PCI15K Local Control Analysis
15.5.3 Most Like Me Example Using the PCI15K Data Set
15.5.4 Extensions and Interpretations of Most Like Me Displays
15.6 Summary
References
B
C
D
E
G
H
I
K
L
M
N
O
P
Q
R
S
T
U
V
W
X
Y
Z
About the Book
What Does This Book Cover?
In 2010 we produced a book, Analysis of Observational Health Care Data Using SAS®, to bring together in a single place many of the best practices for real world and observational data research. A focus of that effort was to make the implementation of best practice analyses feasible by providing SAS Code with example applications. However, since that time, there have been improvements in analytic methods, coalescing of thoughts on best practices, and significant upgrades in SAS procedures targeted for real world research, such as the PSMATCH and CAUSALTRT procedures. In addition, the growing demand for real world evidence and interest in improving the quality of real world evidence to the level required for regulatory decision making has necessitated updating the prior work.
This new book has the same general objective as the 2010 text – to bring together best practices in a single location and to provide SAS codes and examples to make quality analyses both easy and efficient. The main focus of this book is on causal inference methods to produce valid comparisons of outcomes between intervention groups using non-randomized data. Our goal is to provide a useful reference to help clinicians, epidemiologists, health outcome scientists, statisticians, data scientists, and so on, to turn real world data into credible and reliable real world evidence.
The opening chapters of the book present an introduction of basic causal inference concepts and summarize the literature regarding best practices for comparative analysis of observational data. The next portion of the text provides detailed best practices, SAS code and examples for propensity score estimation, and traditional propensity score-based methods of matching, stratification, and weighting. In addition to standard implementation, we present recent upgrades including automated modeling methods for propensity score estimation, optimal and full optimal matching procedures, local control stratification, overlap weighting, new algorithms that generate weights that produce exact balance between groups on means and variances, methods that extend matching and weighting analyses to situations comparison more than two treatment groups, and a model averaging approach to let the data drive the selection of the best analysis for your specific scenario. Two chapters of the book focus on longitudinal observational data. This includes an application of marginal structural modeling to produce causal treatment effect estimates in longitudinal data with treatment switching and time varying confounding and a target trial replicates analysis to assess dynamic treatment regimes. In the final section of the book, we present analyses for emerging topics: reweighting methods to generalize RCT evidence to real world populations, sensitivity analyses and best practice flowcharts to quantitatively assess the potential impact of unmeasured confounding, and an introduction to using real world data and machine learning algorithms to identify treatment choices to optimize individual patient outcomes.
Is This Book for You?
Our intended audience includes researchers who design, analyze (plan and write analysis code), and interpret real world health care research based on real world and observational data and pragmatic trials. The intended audience would likely be from industry, academia, and health care decision-making bodies, including the following job titles: statistician, statistical analyst, data scientist, epidemiologist, health outcomes researcher, medical researcher, health care administrator, analyst, economist, professor, graduate student, post-doc, and survey researcher.
The audience