AbstractBackgroundCohort studies of people with a history of COVID-19 infection and controls will be essential to understand the epidemiology of long-term effects. However, clinical diagnosis requires resources that are frequently restricted to the severely ill. Cohort studies may have to rely on surrogate indicators of COVID-19 illness. We describe the prevalence and overlap of five potential indicators: self-reported suspicion, self-reported core symptoms, symptom algorithm, self-reported routine test results, and home antibody testing.MethodsAn occupational cohort of staff and postgraduate students at a large London university who participated in surveys and antibody testing. Self-report items cover March to June 2020 and antibody test results from ‘lateral flow’ IgG/IgM antibody test cassettes sent to participants in June 2020.ResultsValid antibody test results were returned for 1882 participants. Of the COVID-19 indicators, the highest prevalence was core symptoms (770 participants positive, 41%), followed by participant suspicion of infection (n=509, 27%), a symptom algorithm (n=297, 16%), study antibody positive test (n=124, 6.6%) and self-report of a positive external test (n=39, 2.1%). Study antibody positive result was rare in people who had no suspicion they had experienced COVID-19 (n=4, 0.7%) or did not experience core symptoms (n=10, 1.6%). When study antibody test results were compared with earlier external antibody results in those who had reported them, the study antibody results agreed in 88% cases (kappa= 0.636), with a lower proportion testing positive on this occasion (proportion with antibodies detected 15% in study test vs 24% in external testing).DiscussionOur results demonstrate that there is some agreement between different COVID indicators, but that they a more complete story when used together. Antibody testing may provide greater certainty and be one of the only ways to detect asymptomatic cases, but is likely to under-ascertain due to weak antibody responses to mild infection, which wane over time. Cohort studies will need to review how they deal with different and sometimes conflicting indicators of COVID-19 illness in order to study the long-term outcomes of COVID-19 infection and related impacts.What is already known on this subject?Research into the effects of COVID-19 in the community is needed to respond to the pandemic. Objective testing has not been widely available and accuracy may not be high when carried out in retrospect. Many cohort studies are considering how best to measure COVID-19 infection status.What this study adds?Antibody testing is feasible, but it is possible that sensitivity may be poor. Each indicator included added different aspects to the ascertainment of COVID-19 exposure. Using combinations of self-reported and objectively measured variables, it may be possible to tailor COVID-19 indicators to the situation.