Objectives

Many important epidemiological designs collect data on the development of new events (e.g. illness, death). These include cohort studies and RCTs. Traditional epidemiological analysis of these studies involves estimation of incidence, based either on count data or person-time. However, these methods are based on assumptions that can be limiting (e.g. that the incidence rate is constant across all follow-up time). More powerful analysis approaches make use of the time-to-event and employ survival methods. The purpose of this course is to introduce students to the underlying approaches of survival analysis and to provide them with the basic analytical methods used for non-regression and regression based analysis methods. These methods will be linked to epidemiological research designs.

By the end of the course, the student will be able to use survival analysis methods to analyze real data and will be able to produce meaningful interpretations of the results in epidemiological terms. They will also be aware of situations which require more advanced statistical help and when the results of an analysis should not be taken on face value. The course will not attempt to turn the student into a mathematical statistician. We will not spend time on the detailed underlying theory or derivations of the statistical methods to fit and explore models. Rather, the course will concentrate on the practical aspects of the application of the models and the interpretation of the results. The one theory area we will address is maximum likelihood estimation since this is essential to understanding the results of the analyses. In addition, students will develop further their ability to analyze data sets using SAS.

More specific objectives include:

1. Understand how to perform Actuarial life table and Kaplan-Meier analyses;
2. Understand the application and interpretation of Cox modeling to survival data;
3. Be able to relate the statistical techniques to the appropriate epidemiological concepts (e.g. confounding and effect modification);
4. Be able to interpret regression coefficients in terms of epidemiological parameters (e.g. relative risk);
5. Understand approaches to variable selection and how these relate to interpretation of the model in biological/epidemiological terms;
6. Be able to use the computer to analyze data sets.

TOPICS NOT COVERED

As in any advanced course, there are many more topics that would be of interest than could be included in the course. I have had to omit a number of topics which are useful in some areas of application. My decision was based on an assessment of the techniques that I believe will be most useful to you in a career in epidemiology. Important topics that we will not cover include:

1. Accelerated Failure models;
2. Competing events;
3. Relative survival methods;
4. Multi-level applications of survival analysis;
5. Random effect models;
6. Longitudinal data analysis;
7. Clustering effects, Design effects, etc.;
8. Life events modeling (e.g. when an outcome is a repeating event such as influenza rather than a single event);

We will also not be covering computer techniques for preparing large data files for analysis or techniques for data management in large-scale studies.