DrShushen
commented
1 year ago I think the way you explained it should be the right way to do it. If you can frame your problem as a a time-series problem, you could also try: our time-series library: https://github.com/vanderschaarlab/temporai
Question
How to run a RiskEstimationStudy with fixed evaluation time points?
Further Information
I'm currently trying to use Autoprognosis to develop a risk estimation model on a clinical study dataset. The study has evaluation time points at 3, 6 and 12 months. The presence of disease is determined at each of these time points for each subject. Patient characteristics are gathered at baseline.
I have looked at the survival analysis example and see that in the example dataset, the duration/time-to-event column seems to contain the time-to-event (or general survival), Y contains the presence of an event and time_horizons seems to be evaluations? How do you recommend coding the duration column in this case? The easiest way would be to just use the presence of disease at the different timepoints (ie if someone has the disease at the 3 month evaluation assign 3 months as the time-to-event value). however, this would that there are only 3 different values in the time-to-event column: 3, 6 and 12.
Maybe you have encountered this problem before and can give a recommendation on how to handle such cases.