A systematic comparison of methodologies for the estimation of the serial interval

Serial interval (SI) is a crucial indicator for characterizing the course of epidemic transmission and provides guidance in designing the intervention strategy for preventing epidemic spreading. Many approaches have been developed to estimate the serial intervals, while their effectiveness evaluations are rarely exhibited due to the unavailable of the real SI values. Here, we developed the data-driven based transmission model to simulate the “realistic” epidemics and kept the entire micro-transmission chain to obtain the real SI values for comparison. We firstly found the direct statistic of serial interval obtained from the infector-infectee pairs within households can serve as an accurate representative of the entire population. The households monitored during the early and peak phases of an epidemic provide precise estimations, while the serial interval would be underestimated if the households are collected during the decreasing phase of epidemic. This study reveals that both the ML method using the contact tracing data and the BY method imputing with the household follow up data outperform that the TM method which employing the aggregated data during the early and decreasing phases of epidemic, while the TM method provides a more precise estimation of mean value of SI during the peak phase of epidemic. Our study also presents that sampling 100 households would be enough to give an accurate estimation of SI value. Other sensitiveness on both the epidemic transmission parameters and the noise in the date of symptom onset demonstrate the robustness of the above results. This study provides a systematic comparison of SI estimation methods and gives the public health authority the evidence when they are using the relevant SI estimating methods to inform the policy.