1st Edition
Multistate Models for the Analysis of Life History Data
Multistate Models for the Analysis of Life History Data provides the first comprehensive treatment of multistate modeling and analysis, including parametric, nonparametric and semiparametric methods applicable to many types of life history data. Special models such as illness-death, competing risks and progressive processes are considered, as well as more complex models. The book provides both theoretical development and illustrations of analysis based on data from randomized trials and observational cohort studies in health research. It features: Discusses a wide range of applications of multistate models, Presents methods for both continuously and intermittently observed life history processes, Gives a thorough discussion of conditionally independent censoring and observation processes, Discusses models with random effects and joint models for two or more multistate processes, Discusses and illustrates software for multistate analysis that is available in R, Target audience includes those engaged in research and applications involving multistate models.
Preface
List of Figures
List of Tables
Glossary
Abbreviations
- Introduction to Life History Processes and Multistate Models
- Event History Processes and Multistate Models
- Multistate Analysis Based on Continuous Observation
- Some examples of analysis with multistate models
- Studies with Intermittent Observation of Individuals
- Heterogeneity and Dependence in Multistate Processes 0
- Process-dependent Sampling Schemes
- Additional Topics
Life History Analysis with Multistate Models
Some Illustrative Studies
Disease Recurrence Following Treatment in a Clinical Trial
Complications from Type Diabetes
Joint Damage in Psoriatic Arthritis
Viral Load Dynamics in Individuals with HIV Infection
Introduction to Multistate Processes
Counting Processes and Multistate Models
Features of Multistate Processes
Marginal Features and Partial Models
Some Aspects of Modeling, Analysis and Design
Objectives
Components of a Model
Study Design and Data
Software
Introduction to Some Studies and Dataframes
A Trial of Breast Cancer Patients with Skeletal Metastases
An International Breast Cancer Trial
Viral Rebounds in HIV-positive Individuals 0
Viral Shedding in HIV Patients with CMV Infection 0
Bibliographic Notes
Problems
Intensity Functions and Counting Processes
Likelihood for Multistate Analyses
Product Integration and Sample Path Probabilities
Time-Dependent Covariates and Random Censoring
Some Important Multistate Models
Modulated Markov Models
Modulated Semi-Markov Models
Models with Dual Time Scales
Models Accommodating Heterogeneity
Linked Models and Local Dependence
Process Features of Interest
Simulation of Multistate Processes
Bibliographic Notes
Problems
Maximum Likelihood Methods for Parametric Models
Markov Models
Semi-Markov Models
Multistate Processes with Hybrid Time Scales
Comments on Parametric Models
Nonparametric Estimation
Markov Models
An Illness-death Analysis of a Metastatic Breast Cancer Trial 0
Semi-Markov Models
Recurrent Outbreaks of Symptoms from Herpes Simplex
Virus
Semiparametric Regression Models
Multiplicative Modulated Markov Models
Regression Analysis of a Palliative Breast Cancer Trial
Multiplicative Modulated Semi-Markov Models
Regression Analysis of Outbreaks from Herpes Simplex Virus
Additive Markov and Semi-Markov Models
Herpes Data Analyses with Additive Model
Nonparametric Estimation of State Occupancy Probabilities
Aalen-Johansen Estimates
Adjustment for Process-Dependent Censoring
Skeletal Complications and Mortality in Cancer Metastatic
to Bone
Model Assessment 0
Checking Parametric Models 0
Semiparametric Models 0
Predictive Performance of Models 0
Consequences of Model Misspecification and Robustness 0
Design Issues 0
Bibliographic Notes
Problems
Competing Risks Analysis
Model Features and Intensity-based Analysis
Methods Based on Cumulative Incidence Functions
Methods Based on Direct Binomial Regression
Models for State Occupancy Based on Pseudo-Values
A Competing Risks Analysis of Shunts in Hydrocephalus
Alternative Methods for State Occupancy Probabilities 0
Estimation Based on State Entry Time Distributions 0
Estimation Based on Binomial Data
A Utility-based Analysis of a Therapeutic Breast Cancer
Clinical Trial
Analysis of State Sojourn Time Distributions
Bibliographic Notes
Problems 0
Introduction
Estimation and Analysis for Markov Models
Model Fitting
Parametric Information and Study Design
Model Checking 0
Illustration: Progression of Diabetic Retinopathy
Nonparametric Estimation of State Occupancy Probabilities 0
Process-Dependent Observation Times
Further Remarks on Dependent Visit Processes
Marginal Features and Inverse-Intensity of Visit Weighting
Estimation of Visit Process Intensities
Nonparametric Estimation of Occupancy Probabilities
Progression to Mutilans Arthritis
Intermittent Observation and Non-Markov Models
Mixed Observation Schemes
Illness-Death Models
General Models
Progression and Progression-Free Survival in Cancer Trials
Bibliographic Notes
Problems
Accommodating Heterogeneity in Life History Processes 0
Frailty Models in Survival Analysis 0
A Progressive Multistate Model With Random Effects 0
Random Effect Models with Recurrent States 0
Analysis of Exacerbations in Chronic Bronchitis
Modeling Correlated Multistate Processes
Dependence Models Based on Random Effects
Intensity-based Models for Local Dependence
Dependence Models Retaining Simple Marginal Properties
The Development of Axial Involvement in Psoriatic Arthritis
Finite Mixture Models
Notation, Likelihood Contribution and Estimation
Modeling Variation in Disease Activity in Lupus 0
Hidden Markov Models
General Introduction
A Hidden Markov Model for Retinopathy in the DCCT
Bibliographic Notes
Problems
History and State-dependent Selection
Types of Selection Schemes and Likelihoods
Empirical Studies of Design Efficiencies for Markov Processes
Prevalent Cohort Sampling and Failure Times
Design Based on Probabilistic State-Dependent Sampling
Selection and Initial Conditions with Heterogeneous Processes
Initial Conditions with a Finite Mixture Model
Outcome-Dependent Subsampling and Two-Phase Studies
Introduction
Multistate Processes
Inference for Models with Semiparametric Multiplicative
Intensities
Design Issues
Checks on Ignorable Follow-up Assumptions
Bibliographic Notes
Problems
Analysis of Process-Related Costs and Benefits
Individual-Level Models
Quality of Life Analysis and Breast Cancer Treatment
Individual-Level Decision Making
Population-Level Cost Analysis
Prediction
Viral Rebounds Among Persons with HIV
Joint Modeling of Marker and Event Processes 0
Roles of Markers in Disease Modeling 0
Models for Markers and Life History Processes 0
Intermittent Measurement of Markers and Censoring 0
A Joint Multistate and Discrete Marker Process Model 0
Remarks on Causal Inference with Life History Processes
Bibliographic Notes
Problems
Appendix A Selected Software Packages
A Software for Time to Event Data
A Parametric Analyses
A Semiparametric Analyses
A Selected Software for Multistate Analyses
A Multistate Software
A Methods based on Marginal Features
A Dataframe Construction with the mstate Package
A Software for Intermittently Observed Multistate Processes
A Miscellaneous Functions Useful for Multistate Analysis
A Timeline Plots
A Lexis Diagrams
A Drawing Multistate Diagrams with the Epi R Package
Appendix B Simulation of Multistate Processes
B Generating a Three-State Time-nonhomogeneous Markov Process
B Intensities Featuring Smooth Time Trends
B Processes with Piecewise-Constant Intensities
B Simulating Multistate Processes Under Intermittent Inspection
Appendix C Code and Output for Illustrative Analyses
C Illustrative Analysis of Diabetic Retinopathy
C Fitting the reversible Markov model MB with msm
C Fitting the progressive Markov model MB with msm
C Fitting the Hidden Markov Model with msm
C Code for the Onset of Arthritis Mutilans in PsA
C Dataframe and Fit of Intensity-based Model
C Marginal Model for Time to Entry to the Absorbing State
C Inverse Intensity Weighted Nonparametric Estimation
Appendix D Datasets
D Mechanical Ventilation in an Intensive Care Unit
D Outcomes in Blood and Marrow Transplantation (EBMT)
D A Trial of Platelet Dose and Bleeding Outcomes
D Shedding of Cytomegalovirus in HIV-Infected Individuals 0
D Micronutrient Powder and Infection in Malnurished Children
D The Dynamics of Giardia Lamblia Infection in Children
D The Development of Arthritis Mutilans in Psoriatic Arthritis
D Damage of the Sacroiliac (SI) Joints in Psoriatic Arthritis
D The Incidence of PsA in Individuals with Psoriasis
Bibliography
Index
Biography
Richard Cook is Canada Research Chair in Statistical Methods for Health Research at the University of Waterloo. He has received the Gold Medal of the Statistical Society of Canada and is a Fellow of the American Statistical Association. He collaborates and consults widely on health research and has given many short courses. He and Dr. Lawless previously coauthored the influential book, The Statistical Analysis of Recurrent Events (Springer, 2007).
Jerald Lawless is Distinguished Professor Emeritus at the University of Waterloo. He is a Fellow of the Royal Society of Canada, a Gold Medal recipient of the Statistical Society of Canada and Fellow of the American Statistical Association. He is a past editor of Technometrics and has collaborated and consulted in numerous areas. He has presented many short courses, with Dr. Cook and individually.
"The authors of the book are internationally renowned experts in the field of multi-state modeling and have written an extremely clear and comprehensive book on the topic that covers many different aspects, from the fundamental theory to the practical side of analyzing data and interpreting results. The examples are well chosen to represent the most common types of multi-state processes that public health researchers could encounter. The inclusion of software code to illustrate how the models can be fit and interpreted is especially helpful to readers." (Mimi Kim, Albert Einstein College of Medicine)
"The authors of the book are internationally renowned experts in the field of multi-state modeling and have written an extremely clear and comprehensive book on the topic that covers many different aspects, from the fundamental theory to the practical side of analyzing data and interpreting results. The examples are well chosen to represent the most common types of multi-state processes that public health researchers could encounter. The inclusion of software code to illustrate how the models can be fit and interpreted is especially helpful to readers."
~Mimi Kim, Albert Einstein College of Medicine"This is a very nice book that does not exist on the market yet, and the multistate models for example are not well covered in terms of text books. We here have a book that really takes the multistate aspect seriously and provides many genuine examples that are discussed in depth. I cannot recall seeing examples in such depth in other books that deal with similar topics. This is not easy to do but the authors succeed in this fully."
~Thomas Scheike, University of Copenhagen
"This book includes a wide-ranging review of the use of multistate models for the analysis of longitudinal data arising from healthcare. The presentation is very clear and strikes a good balance between general description and rigorous specification of appropriate statistical models, including assumptions and limitations. Illustration of the methods using some substantive datasets, based on the authors’ experience of monitoring complex diseases, provides further insight into the value of different approaches. Both the level of detail and the pace of the arguments is very good."
~Linda Sharples, London School of Hygiene and Tropical Medicine