How to calculate Median time-to-event based on only the subset of patients who actually experienced an event (rather than of the total number of patients)

[chrispoptic]

How to calculate Median time-to-event based on the total number of patients who actually experienced an event (rather than of the total number of patients)?

The median_survival_time_ attribute of the KaplanMeierFitter class returns the estimated median time to event.

But this median survival time is evaluated for all patients, regardless of whether the experienced an event or not.
For Kaplan Meier curves where survivor probability never drops below 50%, this leads to returning numpy.inf.

Instead, how to calculate the median_survival_time_ for only the subset of patients who actually experienced an event?

[CamDavidsonPilon]

Hi @chrispoptic,

I suppose you would just fit the KaplanMeier to only the patients that experienced the event, then call median_survival_time_.