Course 1
Analysis of life history data with multistate models
Presenters: Richard Cook and Jerry Lawless Room: Saanich 1&2
(Sunday, 10 July 2016 from 9:00 am – 5:00 pm)
Summary: Multistate models offer a powerful framework for advancing scientific understanding of disease processes including rates of progression and associated prognostic factors, the risk of particular complications, and the effect and costs of medical interventions. The multistate framework has also proven useful as a way of formulating statistical models for survival analysis when dealing with incompletely observed time-varying covariates, truncation and non-ignorable censoring or selection effects. Despite their relevance and utility, multistate models are not widely used.
This workshop presents a general framework for multistate modeling and highlights the utility of basic processes including Markov models, semi-Markov models and models with hybrid time scales. Settings in which the times of transitions are subject to right-censoring will be considered, as well as settings where exact transition times are unobserved because a person’s status is only known at periodic inspection times, often corresponding to clinic visits; in the latter case transition times are interval-censored. Parametric, nonparametric and semiparametric methods of estimation will be presented. Robust estimation of marginal features of disease processes such as state occupancy probabilities will be discussed along with associated regression techniques. Numerous examples will illustrate how to construct data frames and carry out analyses with software in R and S-PLUS.
Aims
- Introduce a general framework for multistate modeling and statistical analysis of life history data.
- Define Markov, semi-Markov and hybrid processes and illustrate their use in specific problems.
- Describe and illustrate methods for dealing with both continuously and intermittently observed processes.
- Discuss techniques for model assessment.
- Present robust methods for estimating marginal process features.
- Illustrate how to carry out statistical analysis using R/S-PLUS software.
Learning Outcomes
At the end of the day participants should:
- Have an understanding of how multistate modeling can be used for the analysis of a broad range of disease processes.
- Understand the effects of censoring or intermittent inspection times along with problems of dependent censoring and inspection times and how to deal with them.
- Know how to construct data frames for analysis, how to fit models and conduct estimation and testing, and how to interpret findings.
- Have a broad set of useful examples to draw upon.
Topics covered
The course material will be presented in a lecture format. The introduction of many examples from health research will motivate the methods presented and promote class discussion, which the presenters will encourage through question and answer techniques. Attendees will learn how to analyse data using multistate models and in the process will be exposed to a range of important problems in medicine and public health research. They will learn about the assumptions underlying different methods of analysis, and also about practical and theoretical limitations of specific methods.
Topics covered include:
- Introduction to Event History modelling and Multistate Models
- Model Fitting, Estimation and Inference Methods
- Some Illustrative Analyses
- Processes with Intermittent Observation
- Modelling Heterogeneity and Associations
- Dependent Censoring and Inspection
- Planning Studies, Cost Analysis and Other Topics
Learning strategy
The material will be presented using slides and through class discussion. Attendees will be given a booklet containing the slides, which will contain clear descriptions of the methodology, of applications, and of how to implement analyses in R/S-SPLUS.
Pre‐requisites
This workshop will be directed at statisticians in academia, government or industry interested in the analysis of life history data. Some familiarity with survival analysis will be required.
About the instructors
Richard Cook is Professor of Statistics at the University of Waterloo and holder of the Canada Research Chair in Statistical Methods for Health Research. He has published extensively on new statistical methodology in event history analysis, incomplete data and the design of clinical trials. He collaborates with numerous researchers in medicine and public health and has consulted widely with pharmaceutical companies on the design and analysis of clinical trials.
Jerry Lawless is Distinguished Professor Emeritus of Statistics at the University of Waterloo. He has published extensively on statistical models and methods for survival and event history data, life history processes and other topics, and is the author of Statistical Models and Methods for Lifetime Data (2nd edition, Wiley, 2003). He has consulted and worked in many applied areas, including medicine, public health, manufacturing and reliability. Dr. Lawless was the holder of the GM-NSERC Industrial Research Chair in Quality and Productivity from 1994 to 2004.
Drs. Cook and Lawless have co-authored many papers as well as the book “The Statistical Analysis of Recurrent Events” (Springer, 2007). They have also given numerous workshops together.