Cumulative Incidence Function (CIF) calculator

Calculate event probabilities in the presence of competing risks. Fit the Aalen-Johansen estimator online.

What it does

Answers "out of everyone who started, what fraction had THIS event by now?" when more than one kind of event can happen and only the first one counts. Each event type gets its own curve, and the curves add up sensibly because every subject is counted once, in whichever state they ended up.

When to use it

When one event can prevent another from ever happening. On a transplant waiting list, a patient who is transplanted can no longer die-before-transplant: they have not "dropped out of the study", they have left the risk of your event entirely. Kaplan-Meier has only one way to handle someone leaving — censoring — and censoring assumes they would have gone on to have your event at the usual rate. For a transplanted patient that is not merely doubtful, it is impossible, which is why the answer comes out too high.

Cautions

• Treating a competing event as censoring and reporting 1−KM systematically overstates risk, always in the same direction.
• Cumulative incidences for all event types sum to the probability that something happened, and never exceed 1 — a consistency check 1−KM fails.
• A difference in one event's incidence can be driven entirely by the rate of the other. Always show both curves.

Alternatives

• Kaplan-Meier — no competing event exists, or your endpoint is all-cause

How to read the output

The cumulative incidence curves

Each rises from 0 and shows the probability of having had THAT outcome by each time, counting the competing outcome as what it is — an event that ended the possibility, not a censoring. These curves are not 1 − Kaplan-Meier, and they are lower. Treating a competing event as censoring assumes the patient could still have had your event afterwards, which is precisely what a competing event rules out; the overstatement grows with how common the competing event is.

The curves together

All the cumulative incidences plus the event-free share account for everyone, so they can be read as a partition of the cohort at any time. Read each curve against the others. A treatment that lowers one incidence purely by raising a competing one has not helped anybody.

The outcome definitions you entered

Each value in the event column is mapped to a named outcome; a value you did not define is treated as censored. Check the mapping covers every value actually present. A miscoded level silently becomes censoring — the exact bias this view exists to remove.

The number-at-risk table

How many patients remain under observation at each time point. It is the honest scale of the curve above it. Read the curve only as far as the numbers support. The right-hand tail is estimated from a handful of patients — a step there moves the line a long way and means almost nothing. If the table reads 4, 2, 1, do not quote survival at that time.

How this calculator is validated · Which statistical test should I use?