Dalla Lana School of Public Health

Advanced Quantitative Methods in Epidemiology

Course Description

• This course will provide students with an overview of the theory and applications of advanced quantitative methods in epidemiology. The purpose of the course is to assist students in answering complex etiological research questions in epidemiology. The course includes three modules: 1) introduction to survival analysis; 2) Cox proportional hazards model and competing risk analysis; and 3) multi-state models for event history data.
• The course follows a modular format.  Each module consists of three lectures and one laboratory. The evaluation for the course will include three assignments (80% of final grade) and an oral presentation (15% of the final grade).  Each assignment will be based on an existing data set collected from a large cohort study. Students will be provided with a specific question to answer and expected to submit a five page paper presenting their analysis and results.  Participation (5% of final grade) will be based on class attendance and active involvement and participation in class discussion.

Course Objectives

• Module 1. Introduction to Survival Analysis
  • Review the basic techniques for survival: definitions (dependent variable, origin of time, study window), censoring, important functions describing survival distribution, life tables and Kaplan-Meier estimation.
  • Convert person data into person-period data, including how to incorporate time-varying covariates.
  • Review the basic techniques for discrete-time survival models.
  • Introduction to time-varying measures.
• Module 2. Cox proportional hazards model and competing risk analysis
  • Apply Cox proportional hazard model to time-to-event data, recognize its assumptions, and relevant syntax. Understand model fit parameters and statistical power and sample size calculations.
  • Understand and conduct competing risk modeling, including model assumptions including relevant syntax & macro design.
  • Evaluate the predictive accuracy of the estimated model, apply optimism correction through bootstrapping for internal validation.
  • Learn the RStudio interface and how to apply Cox proportional hazard modeling.
• Module 3. Multi-state models for event history data
  • Introduce definitions and counting process notation for building multistate models (this includes understanding transition intensity functions, transition intensity matrices, and transition probability matrices).
  • Understand likelihood construction and parameter estimation for multistate models under complete observation.
  • Discuss Markov and Semi-Markov multistate model assumptions.
  • Understand how to structure data for conducting multistate analyses.
  • Understand how to incorporate covariates into a multistate model.
  • Understand how multistate models are related to survival models and competing risks models, and be able to conduct a competing risks data analysis using multistate methods.
  • Know how to perform multistate analyses under the presence of intermittent observation.