Estimation of serial intervals and time-varying reproduction numbers across SARS-CoV-2 variants in childcare clusters: symptom onset vs. reporting dates

Abstract Background Respiratory virus transmission in childcare centers is facilitated by dense contact patterns, making accurate estimation of serial intervals (SI) and time-varying reproduction numbers (Rt) essential for outbreak management. Although symptom onset dates are epidemiologically preferred, reporting dates are frequently used in routine surveillance due to operational constraints. Reliance on reporting dates may introduce systematic bias in transmission parameter estimates. This study aimed to evaluate differences in SI and Rt estimates based on symptom onset versus reporting dates in COVID-19 childcare clusters across SARS-CoV-2 variant periods. Methods Nationwide data from the Korea Disease Control and Prevention Agency (January 2020–February 2022) were analyzed. Four genotype-confirmed clusters meeting the predefined eligibility criteria were included: one Before Delta, two Delta, and one Omicron cluster, each comprising at least 50 epidemiologically linked cases. Serial intervals were fitted using gamma distributions. Time-varying Rt values were estimated using the EpiEstim package separately based on symptom onset and reporting dates. Results Onset-based mean SI declined from 9.85 days (95% CI: 1.97–23.95) in the Before Delta period to 5.34 days (95% CI: 2.00–10.28) during Omicron. In contrast, reporting-based SI estimates were generally longer, reaching 8.87 days (95% CI: 0.71–27.08) during Omicron. Across all variant periods, reporting-based Rt estimates were consistently lower than onset-based estimates. For example, during the Delta period, mean Rt was 1.90 (95% CI: 0.79–3.76) using onset dates compared with 1.08 (95% CI: 0.41–2.39) using reporting dates. Conclusion The choice of temporal reference substantially affects SI and Rt estimation in childcare clusters. Reliance on reporting dates may underestimate real-time transmission intensity and distort epidemic risk assessment. Prioritizing symptom onset data in surveillance systems may improve the accuracy and timeliness of public health responses in high-risk settings.