Speaker
Description
Survival analysis is the study of time-to-event data. In a clinical trial, this could be the time to death, or the end of the trial, along with an indicator of whether or not death occurred. Common non- and semi-parametric survival models can only predict survival for at most as long time as a clinical trial is run. Here, we implement survival extrapolation models that extend the survival prediction beyond the trial time window. We train and test the models on a dataset consisting of approximately 10 million individuals from the Swedish general population. From these, we model survival for those who were diagnosed with cancer between 2003 and 2018, and who were at least 30 years old at the time of diagnosis. Follow-up data is available until 2023, and we have access to 10 years lookback for previous diagnoses. This dataset is randomly split into training and testing sets. We investigate how the models perform on different subpopulations.
