Title: Using the Cox model for the three tasks of health data science
Authors: Ruth Keogh - London School of Hygiene and Tropical Medicine (United Kingdom) [presenting]
Abstract: The Cox model was described by David Cox in 1972 as a model for the analysis of survival data, specifically to relate explanatory variables to the hazard of an event via hazard ratios. It has since become the most widely used model for analysis of health data and beyond. In several papers, it has been discussed how the research questions addressed in medical statistics, and health data science can be broadly classified into three tasks: description, prediction, and causal investigation. we will discuss the use of the Cox model in performing these three tasks. We will focus especially on the use of the Cox model in causal investigations, including addressing criticisms of the hazard ratio as a measure for describing causal relationships. Examples will be given from medical applications.