Handle delayed entry in a survival analysis

Distinguish entry from the time origin

A study may define time from diagnosis but enrol people later. Those enrolled must have remained eligible until entry, and their earlier survival time was not observed in the same way as follow-up after entry. This is delayed entry, often described as left truncation. It is different from left censoring, where an event is known to have happened before an observation time.

The survival package documentation describes start-stop survival intervals. Use that representation only with a clear definition of the origin, entry and exit times and an estimator that forms the appropriate risk sets.

Reconstruct eligibility over time

For every participant, calculate entry and exit on the same time scale. Check that entry precedes exit and that events occur within observed risk intervals. State what makes a person eligible to enter and whether entry timing may be related to prognosis after conditioning on model variables.

Do not add each participant to the risk set from the common origin merely because that origin is available in the record. A diagnosis date can define the analysis clock without implying that the person was under observation from that date. Likewise, resetting everyone to zero at enrolment answers a different time-scale question.

Hypothetical example

A fictional participant is diagnosed on day zero, joins the study on day 90 and has an event on day 150. With diagnosis as the time origin, the observed risk interval is from day 90 to day 150. The participant should not contribute to the study risk set at day 40.

The 60 days between entry and event are observed follow-up, but the event time on the diagnosis clock remains day 150. Keeping both quantities prevents a mistaken mixture of time scales.

Check interpretation and sensitivity

Document the assumptions needed to interpret the delayed-entry analysis, including the relation between entry and the event process. Correct risk-set coding does not automatically make a prevalent cohort representative of everyone diagnosed. Discuss early events or people who never enter when defining the target population.

Use synthetic records with early and late entry to verify the risk-set counts at selected event times. Report the origin, entry rule, follow-up period and exclusions. The survival risk-set worksheet can preserve these checks. A clear timeline often reveals a design problem more effectively than inspecting a fitted hazard ratio after the fact.

Source verification: 6 November 2026. Examples are hypothetical.