Introduction to Survival or Time-To-Event Analysis Using Stata

This course is designed as an applied introduction to survival analysis, the study of the length of time until an event of interest occurs. Stata will be used for the hands-on computer exercises throughout this course.

 
Level 3 - runs over 5 days
Instructor: 

Dr Mark Griffin is the Director of Insight Research Services Associated (www.insightrsa.com), and holds Adjunct appointments within the School of Public Health, University of Queensland and the Sydney Medical School, University of Sydney. Mark serves on the Executive Committee for the Statistical Society of Australia, and is Chair of their Section for Business Analytics. Mark also serves as the Asia-Pacific Regional Director for the International Institute of Business Analysis, is Chair of their Business Analytics Special Interest Group, and is an IIBA Endorsed Education Provider. He is currently doing research with the Queensland Ambulance Service analyzing their incident reports, where the QAS visits approximately 700,000 incidents per year. To date he has presented over 80 two-day and 10 five-day workshops in statistics around Australia.

About this course: 

Survival analysis is used to study the length of time until an event of interest occurs. Examples of such events include the time when a patient dies (or indeed is classified as disease-free), when a criminal re-offends, and when a customer makes their next purchase. As we study the time until such events researchers may be interested to know whether different interventions (such as medication or marketing) will influence the time until the event. This course is designed to introduce participants to the field of survival analysis where only a basic prior knowledge of statistics will be assumed at the start of the course.

 

Detailed notes with worked examples and references will be provided as a basis for both the lecture and hands-on computing aspect of the course.

 

The target audience for this course is researchers working with large, complex datasets that are using the techniques of survival analysis for the first time.

Course syllabus: 

Day 1 The foundation of survival analysis
On this first day we shall provide a foundation to the topic of survival analysis. We shall describe:

  • Study design and reporting (including Setting goals and objectives, Inclusion and exclusion criteria for participant selection, Data Management and data linkage, Reporting styles including the CONSORT statement, Ethics including privacy and confidentiality)
  • common applications where survival analysis might be an appropriate technique, and the characteristics of datasets where survival analysis may be employed
  • the similarity and differences between survival analysis and longitudinal data analysis
  • simple descriptive and graphical techniques for describing survival data
  • we shall also describing the mechanism of censoring (where not all participants may be followed until the time at which they have their event, eg. A patient who lives for many years after the first symptoms of a disease)

 

Day 2 Regression models for modelling survival data
On this second day we will describe a major approach for assessing whether a set of covariates are predictors for whether study participants will experience a shorter or longer duration until the time of their event (eg. Patients who receive a treatment versus a placebo). During this time we shall discuss:

  • the Proportional Cox Hazards Model
  • how to interpret the results of a Proportional Cox Hazards Model
  • measures of model fit

 

Day 3 - Stratified Proportional Cox Hazards Models
On this third day we will discuss the Proportional Cox Hazards Assumption, how to assess whether this assumption has been met, and what to do if this assumption is not valid. In essence the Proportional Cox Hazards Assumption assumes that the survival function for different groups of participants will have the same shape over time apart from a multiplicative constant dependent on predictors such as whether the participant has received the intervention. The Stratified Proportional Cox Hazards Model is utilized when this Assumption is not valid.

 

Day 4 Time-Varying Covariates
In the simplest Cox Hazards model it is assumed that the predictive covariates are held constant over time (covariates such as whether the participant received the intervention or not). On this day we shall describe how to include covariates that vary over time into the Cox Hazards model.

 

Day 5 Other Topics and Own data
On this last day we shall provide a brief introduction to more advanced topics within survival analysis, and will have time for participants to apply the methods of this course to their own datasets. More advanced topics that we shall briefly discuss include:

  • joint longitudinal-survival models (where we predict a longitudinal outcome and a time-to-event variable in the same model)
  • recurrent event models (where participants can experience more than event, such as a criminal that commits more than one crime)

 

Course format: 

This course will take place in a computer lab. All equipment will be supplied. You are encouraged to bring a data set and/or research problem with you.

Recommended Background: 

Participants must have completed an introductory course in statistics or have equivalent experience. While this workshop will be taught using Stata it is not essential that participants have had prior exposure to Stata.

Recommended Texts: 

Other references include:

  • Applied Survival Analysis: Regression Modeling of Time to Event Data 2nd Edition (2008). David W. Hosmer Jr., Stanley Lemeshow, Susanne May
FAQ: 

Q: Do I have to have any prerequisites to do this course?

A: Yes, see recommended background section for details.

Notes: 

The instructor's bound, book length course notes will serve as the course texts.

Supported by: 

Stata is distributed in Australia and New Zealand by Survey Design and Analysis Services.