Timing uncertainty in first reported chronic disease across international aging cohorts

Intermittent interviews record disease reports rather than clinical onset. We evaluated how event-time conventions and later response qualification change five-year estimates across seven ageing cohorts. The initial demographic inventory comprised 260,340 adults; 187,754 entered at least one disease-specific negative-entry risk set. We compared a Turnbull interval-censored reference distribution with Kaplan-Meier estimates using midpoint or first-positive-visit assignment. The reference interpretation requires an absorbing-transition working approximation and suitable inspection and censoring assumptions. We separated retrospective confirmation and observed-report competing mortality from this reference analysis. Additional post hoc analyses examined disease and schedule variation, paired resampling uncertainty and simulations with reporting error and attrition. Across 99 combinations there were 2,277,779 person-disease risk-set records and 245,125 first reports. In 33 core combinations, median first-positive-minus-Turnbull and midpoint-minus-Turnbull differences were − 0.79 and − 0.01% points. Reconstructed confirmation-qualified estimates had a median of 3.64% versus 5.64% under first-report qualification. In 18 new simulation scenarios, midpoint often had smaller root mean squared error, while informative attrition and reporting error affected all estimators; Turnbull nominal interval coverage was 1.5–59.5% in informative-attrition scenarios. Additional paired-interval checks returned 88.0–96.0% Wald coverage under ideal reporting, while a model-centred bootstrap-root alternative performed worse. Timing conventions can materially change short-horizon estimates, but no estimator recovers clinical onset from survey responses alone. Interval-censored estimation is useful when its reporting and observation assumptions are credible; midpoint dating is a sensitivity analysis, and mortality and reporting errors require separate treatment.