Enhancing time-varying reproduction number estimates for COVID-19 with behavior and surveillance data in South Korea, 2020–2022

Accurate estimation of the time-varying effective reproduction number, R(t), is essential for interpreting transmission dynamics and informing public health actions. Incidence-based approaches can be biased when behavioral change and surveillance performance alter realized infectiousness and the timing of observed cases. We developed a behavior- and surveillance-informed framework tailored to the Korean context (Feb 2020–Jan 2022). National epidemiological data (20,155 linked infector–infectee pairs after quality control) and Google mobility indicators were used to construct setting-specific behaviors—residential mobility as a proxy for household contact duration and a composite non-residential signal for non-household activity. Infection-to-diagnosis delays were incorporated via a surveillance-adjusted generation-interval kernel that links recent incidence to current infectiousness. A context-specific transmission measure was mapped to R(t) and connected to daily cases using a count model that accounts for reporting variability, with full technical details described elsewhere. The estimated R(t) captured phase-specific swings in transmissibility and responded to shifts in mobility and detection timing. Household transmission provided a relatively stable baseline, whereas non-household activity drove episodic surges. Surveillance adjustment shortened effective generation times during periods of faster detection and improved calibration of R(t) relative to naïve incidence-based estimates. Forecast evaluation demonstrated consistent short-term skill with appropriate empirical coverage. Combining routinely available mobility and surveillance summaries improves the interpretability and responsiveness of R(t) estimation in densely connected settings. The workflow is transparent and reproducible, supporting near-term assessment and communication of transmission risk, and can be adapted to other surveillance systems where behavioral and diagnostic conditions evolve over time.