Event-time risk-set control sampling

Sample eligible controls at events

At each incident event, eligible controls are members of the cohort still at risk. Repeated control selection and later case status are allowed by incidence-density sampling. Unique-person counts and sampling appearances must not be added as disjoint groups.

Flow diagram

Event-time risk-set control samplingF1.S1Identify incident casesF1.S2Construct each event-time risk setF1.S3Apply matching rulesF1.S4Sample controls reproduciblyF1.S5Retain matched-set membershipF1.S6Incident casesF1.S7Sampled control appearancesF1.N1Risk-set sampling specificationRuleDefinitionTime anchorReplacementRationaleAt-riskeligibilityMatchingControl reuseSamplingprobabilityF1.N2Assumptions and rationaleF1.N3Method and protocol evidenceF1.S0 Start flowEvent-time risk-set control samplingF1.S999 End flowEnd flow

Steps

Identify incident cases

Specify event definition and ascertainment in the source cohort.

Construct each event-time risk set

Include only cohort members eligible and at risk at that event time.

Apply matching rules

Define matching variables, calipers and eligible control counts.

Sample controls reproducibly

State replacement, reuse across risk sets and random sampling procedures.

Retain matched-set membership

Record event time, case/control role and sampling probabilities as applicable.

Incident cases

Record people using the documented inclusion definition. This count is not a disjoint partition of the other count nodes.

Sampled control appearances

Record control appearances using the documented inclusion definition. This count is not a disjoint partition of the other count nodes.

Event-time risk-set control sampling

End flow

Tables and supporting information

Risk-set sampling specification

Risk-set sampling specificationRuleDefinitionTime anchorReplacementRationaleAt-riskeligibilityMatchingControl reuseSamplingprobability

A control may later become a case; preserve set membership and timing.

Assumptions and rationale

Risk-set sampling differs from case-cohort sampling. Repeated controls and later cases must be handled by the chosen analysis.

Method and protocol evidence

Attach the protocol, methodological sources and versioned software specification supporting these decisions. Record citations and provenance; do not enter participant identifiers.

Sources