Concordance between SARS-CoV-2 index individuals and their household contacts on index individual COVID-19 transmission cofactors: a comparison of self-reported and contact-reported information

Background Following the outbreak of the COVID-19 pandemic, several clinical trials have evaluated postexposure prophylaxis (PEP) among close contacts of an index individual with a confirmed SARS-CoV-2 infection. Because index individuals do not directly inform the efficacy of prevention interventions, they are seldom enrolled in COVID-19 PEP studies. However, adjusting for prognostic covariates such as an index individual's COVID-19 illness and risk behaviors can increase precision in PEP efficacy estimates, so approaches to accurately collecting this information about the index individual are needed. This analysis aimed to assess whether surveying household contacts captures the same information as surveying the index individual directly.Methods REGN 2069/CoVPN 3502, a randomized controlled trial of COVID-19 PEP, enrolled household contacts of SARS-CoV-2 index individuals. CoVPN 3502-01 retrospectively enrolled and surveyed the index individuals. We compared responses to seven similar questions about the index individuals' transmission cofactors that were asked in both studies. We estimated the percent concordance between index individuals and their household contacts on each question, with 50% concordance considered equivalent to random chance.Results Concordance between index individuals and contacts was high on the most objective questions, approximately 97% (95% CI: 90-99%) for index individual age group and 96% (88-98%) for hospitalization. Concordance was moderate for symptoms, approximately 85% (75-91%). Concordance on questions related to the index individual's behavior was only slightly better or no better than random: approximately 62% (51-72%) for whether they received COVID-19 treatment, 68% (57-77%) for sharing a bedroom, 70% (59-79%) for sharing a common room, and 49% (39-60%) for mask wearing at home. However, while contacts were surveyed within 96 h of the index individual testing positive for SARS-CoV-2, the median time to enrollment in CoVPN 3502-01 was 240 days, which may have caused recall bias in our results.Conclusions Our results suggest a need to survey index individuals directly in order to accurately capture their transmission cofactors, rather than relying on their household contacts to report on their behavior. The lag in enrolling participants into CoVPN 3502-01 also highlights the importance of timely enrollment to minimize recall bias.