Presentation Information

[POS-21]Evaluating the accuracy of SARS-CoV-2 viral dynamics estimated from cross-sectional and longitudinal data collection methods

*Jihyeon Kim1, Hyeongki Park2, Keisuke Ejima3, Hyojung Lee1 (1. Kyungpook National Univ. (Korea), 2. Pusan National Univ. (Korea), 3. Nanyang Technological Univ. (Singapore))

Keywords:

COVID-19,mathematical model,SARS-CoV-2,within-host viral dynamics

Viral load is a crucial indicator of disease progression that gives us features of disease such as information about viral shedding and incubation period. Viral load measurements offer valuable insight into infectious disease research, but it is not possible to test every patient almost every day. Then it requires effective monitoring. This study aims to accurately estimate population-level viral dynamics using limited individual-level viral load data that is collected by two methods of cross-sectional or longitudinal sampling.We use the SARS-CoV-2 viral load data collected from NBA players, staff, and vendors between July 2020 and January 2022. Using a mathematical model of viral dynamics to estimate both individual and population-level viral trajectories of patients infected with the Delta variant from data. We synthesize the observed viral load data using the distribution of parameters estimated from the NBA data. Each point was generated by adding sampling error at 3-day intervals starting from the day of symptom onset. We compare the accuracy of viral dynamics estimation according to the total number of data points (e.g., 100, 200, 1000) and the number of tests per participant. If each participant is tested once, it is cross-sectional data and if they are tested more than once, it is longitudinal data. For comparison of estimation accuracy under different study designs, we estimated some key viral dynamics metrics, such as viral shedding duration and peak size, from these simulated trajectories.Our results show that estimated viral load using cross-sectional data was biased as the number of total data points increases. To more accurately estimate viral shedding duration, it is necessary to test at least 3 times per participant. We then found that performing 3 to 4 tests per participant with more than 200 data points yielded better accuracy. And even if data points are limited, peak size and peak timing of viral load can be estimated with reasonable accuracy. Our findings suggest efficient data collection methods based on comparison of the estimated viral dynamics. This allows us to estimate the viral load of future emerging infectious diseases even with limited data. This information can play a crucial role in facilitating the early establishment of infection prevention policies during epidemics.