A person is censored when we know they were event-free up to a certain time, but we do not know what happened after.
Why it matters
Handling censoring correctly is the whole point of survival analysis.
Key ideas
- Right censoring: event has not happened by the last follow-up
- Causes: study ended, lost to follow-up, withdrew
- Censored people contribute information until they leave
- Assume censoring is unrelated to the outcome (non-informative)
- Report how many were censored
Real-life research example
A patient moves abroad after 18 months, alive at the last visit. They are censored at 18 months.
When to use
- Whenever follow-up is incomplete
When NOT to use
- Don't treat censored cases as events, nor drop them
Mini example
Times: 5, 8, 12+, 12, 20+, 25 (+ = censored)
Quick takeaway
Censored means "survived at least this long".
Try it yourself
Run the Kaplan–Meier tool with and without censoring (all events = 1). Compare the median survival.
Model answer
Treating censored people as events makes survival look shorter. Handling censoring properly gives more realistic estimates.