Developing Decision Models in R Using Published Survival Data

28 June 2026, 8:35 AM – 12:00 PM (local time)

Faculty

  • Alexandra Moskalewicz, MSc – Research Analyst, Child Health Evaluative Sciences, The Hospital for Sick Children Research Institute, Toronto, Canada
  • David U. Garibay-Treviño, MSc – PhD Candidate, School of Epidemiology and Public Health, The University of Ottawa, Ottawa, Canada
  • Eline Krijkamp, PhD – Assistant Professor, Health Technology Assessment, Erasmus School of Health Policy & Management, Erasmus University Rotterdam, Netherlands
  • Petros Pechlivanoglou, PhD – Senior Scientist, Child Health Evaluative Sciences, The Hospital for Sick Children Research Institute, Toronto, Canada

Course Overview

Time-to-event outcomes are common inputs for decision models. However, these data are not always available at the individual patient level, and modelers must extract information from aggregate sources such as overall survival and progression-free survival curves from oncology trials. These aggregate data present challenges in informing decision models and determining the level of complexity that can be appropriately modeled.

This course is designed for researchers tasked with building decision models using aggregate information from published survival curves. The course will cover core survival analysis concepts with a focus on non-parametric and parametric survival modeling. Participants will be introduced to methods for digitizing survival curves, including both automated and point-and-click approaches, and the reconstruction of individual patient data from published Kaplan–Meier curves.

The course will cover decision modeling approaches that rely on extrapolations from parametric survival models of reconstructed data, including partitioned survival models. Alternative approaches will also be discussed that aim to represent the underlying disease process using digitized data, such as estimating transition rates for state-transition models. These approaches include approximation methods and methods based on Bayesian calibration.

Participants will also learn how parameter uncertainty can be propagated through these methods to the health economic outputs of decision models. Real-world examples will be used throughout, and participants will engage in hands-on programming exercises in R.

This course will provide participants with a set of tools and resources to apply when individual patient data are not available for decision model development.

Learning Objectives

Participants will learn how to:

  • Recognize the advantages and drawbacks regarding the types of decision models that can be built using information from published survival curves
  • Explain how to implement different approaches to inform inputs for decision models that rely on published survival curves
  • Execute code in R to transform published survival curves into inputs for specific types of decision models

Course Format

The course will combine conceptual instruction with applied programming exercises.

Participants will be introduced to survival analysis concepts, methods for digitizing survival curves, reconstruction of individual patient data, and approaches for extrapolation and model parameterization. Real-world examples will be used throughout, and hands-on coding exercises in R will allow participants to apply the methods discussed during the session and continue using the provided scripts after the course.

Participant Requirements

Participants should have a basic understanding of survival analysis and decision analytic modeling. No pre-course preparation regarding course content is required.

Participants should bring their own laptops with the latest versions of R and RStudio installed and should have administrative rights to install R packages if needed.