I am trying to use cox proportional hazards to identify covariates associated with 30-day readmissions. The event is the occurrence of readmission within 30-days. The time to event is the number of days between first hospital discharge and readmission.
The patients who did not experience readmission within 30-days are right censored. However, these patients have missing time to event and there is no follow-up time (no follow up to study end is available). What should I use as the time to event for right-censored cases in this analysis?
Thank you.
For individuals with no readmission within 30 days, the time should be 30, the number of days followed up. The censoring variable should be coded as censored.
The biggest thing is to make sure the statistical software package that you use accurately censors the data. Otherwise, if you put "30" in the number of days column, the stats program might think that they actually were re-admitted on Day 30. Usually, the censored value (0 or 1) is added as a separate column. In some software packages, '1' should be entered in the separate column to censor the data and '0' should be entered for non-censored data, but in other software packages, the opposite can be true. My main point is to double check with the particular software package that you are using to ensure that you are properly censoring the data.
Sounds good. By putting their time to event as 30 and defining them censored, the model worked. Thank you.
- Badges
- Science topic
- Similar topics
- Psychology