The application of social network analysis to examine COVID-19 contact tracing networks in a university setting

We reconstructed COVID-19 contact networks to examine individual characteristics and network structures that influenced transmission during the 2020-2021 hybrid-learning year at a US university.

Cohort study.

We used individual-level exposure histories collected through case investigation and contact tracing interviews during the Fall 2020 semester to construct contact tracing networks. We visualized networks and estimated global network statistics and secondary attack rates (SAR). We conducted a bias analysis on the impact of missing cases on these network statistics.

During the Fall 2020 semester, we identified 441 COVID-19 cases, 1121 close contacts, and 1206 links between individuals. Most cases were female (62 %), off-campus students (49 %), and symptomatic (82 %). Individuals had a mean of 2.9 direct contacts, and the maximum number of individuals separating any two persons in a network was 8. The overall SAR was 9.7 % (50/518). Contacts of symptomatic cases had a higher SAR compared to contacts of asymptomatic cases (11.8 % [42/356] vs. 4.9 % [8/162]; p = 0.015). Networks were minimally clustered with the greatest clustering observed in September. Bias analyses indicated that missingness in our observed network was unlikely to have been random or based upon symptomaticity.

In this assessment of COVID-19 contact network structure and transmission characteristics, we found minimal clustering, a low proportion of asymptomatic cases, and higher SAR among contacts of symptomatic cases. Symptomatic cases were unlikely to have been oversampled in our observed network. Our findings suggest that university campuses have unique transmission characteristics, even in the context of a hybrid learning environment in which social interactions may be attenuated.