During March 3–April 22, 2020 in the Faroe Islands, 854 close contacts of COVID-19 patients were identified and quarantined, including 132 who were later confirmed to have SARS-CoV-2 infection. As a result, 722 close contacts that had not tested positive by RT-PCR were eligible for participation in this study (Figure); 584 participated, a participation rate of 80.9%. The mean participant age was 36 years (range 0–84 years), and 58% were women. Most participants were in the younger age groups; only 5.5% were >67 years of age (Table 1).

A total of 32 (5.5% [exact binomial 95% CI 3.8%–7.7%]) persons, comprised of 17 women and 15 men who had not previously tested positive by RT-PCR, were seropositive for total antibody against SARS-CoV-2. After adjustment for test sensitivity and specificity, we calculated the prevalence of SARS-CoV-2–specific antibodies as 5.3% (95% CI 3.5%–7.5%). A total of 43.8% of the seropositive close contacts retrospectively reported symptoms, mainly fever, running nose, and loss of taste or smell. Most seropositive participants were in the youngest age group (0–9 years, mean 4.5 years, range 1.4–9.6 years) (Table 2). Among seropositive participants, 21 had received negative RT-PCR results during their quarantine; 11 participants recalled symptoms, whereas 10 did not. The other 11 seropositive participants did not have RT-PCR during quarantine, of whom 3 retrospectively recalled symptoms. Median time between quarantine and RT-PCR was 0 days (range –1 to 15 days); 11 participants had RT-PCR upon or before going into quarantine. RT-PCR was most prevalent among the youngest age group, of which all 8 participants had RT-PCR (Table 2). The presence of symptoms was not significantly different among the age groups (p = 0.9).

According to the multivariable logistic regression analysis, type of contact (p<0.01) and age group (p = 0.02) were significantly associated with seropositivity. The risk for seropositivity was significantly higher for household contacts compared with other contacts (adjusted odds ratio [aOR] 5.4, 95% CI 1.9–15.2). We did not find a statistically significant difference for workplace contacts compared with other contacts (p = 0.8). Overall, age was significantly associated with seropositivity (p = 0.02). Most age groups had lower aORs than the youngest age group (0–9 years), although these associations were statistically significant only for the 35–49-year (aOR 0.18, 95% CI 0.04–0.9) and 50–66-year (aOR 0.16, 95% CI 0.03–0.8) age groups. Participants >67 years of age had an increased aOR compared with the youngest age group; however, this association was not statistically significant (aOR 2.4, 95% CI 0.6–9.9).

In this study, we identified 32 secondary SARS-CoV-2 infections by later serologic assay in addition to the 132 identified through initial RT-PCR, indicating a secondary attack rate of 19.2%. Most (67.5%) seropositive persons had been in quarantine with their families. In total, 65% of seropositive persons were infected by household members; when including noncohabitating close family members, this total rose to 71%. The other persons were infected by extended family, workplace, or social contacts. We identified 3 sibling pairs, 2 sets of spouses, and 3 parent–child pairs among the seropositive persons.

In this retrospective cohort study of close contacts of COVID-19 patients, we found a 5.5% seropositivity rate among contacts who were not previously identified as positive by RT-PCR. Seroprevalence was highest among household contacts. We found the highest seropositivity rate among children who were infected by their parents. The risk for seropositivity among household contacts was 5-fold that of the risk posed by other close contacts, probably because household members might have closer and more prolonged interactions than work or social contacts. In total, 56% of secondary infections were asymptomatic.

The rapid spread of COVID-19 is partly attributable to transmission by asymptomatic or presymptomatic persons; many of these cases remain undetected because patients might not seek healthcare or undergo testing (19). As a result of the large-scale testing and tracing protocols used in the Faroe Islands, the risk for transmission from persons with undetected cases is probably low. A study of 1,075 randomly selected persons from the Faroe Islands in April 2020 found a 0.6% seropositivity rate (exact binomial 95% CI 0.2%–1.2%), which corresponds to 313 COVID-19 patients (2), somewhat higher than the observed 0.4% crude prevalence of confirmed cases in the Faroe Islands [https://corona.fo]). Our results for close contacts of patients with confirmed COVID-19 are in accordance with the seroprevalence study (2) indicating the existence of some undetected (i.e., not documented in the official records) cases of SARS-CoV-2 infection in the Faroe Islands. Our results highlight the importance of tracing close contacts, who have a much higher seroprevalence than the general population, despite a high proportion of asymptomatic seropositive persons. These findings underscore that testing only symptomatic contacts will miss infections and underestimate the true number of cases. In addition, we emphasize that a negative RT-PCR result might not rule out SARS-CoV-2 infection in a household contact. The close contacts who underwent RT-PCR were tested at a median 0 days (range −1 to 15 days) from start of quarantine (Table 2), possibly indicating that most patients were tested before symptoms developed. Because most of the seropositive contacts in this study were infected by household members, our findings emphasize the importance of isolating infected persons.

Our overall prevalence estimate is lower than that of a study comprising 100,000 participants in England, which found an 18.8% (742/3,946) seroprevalence among those who had unspecified contact with a COVID-19 patient (15). Our estimate is more comparable with that of a study in Singapore that found a 5.5% (29/524) seroprevalence among household contacts who did not have a COVID-19 diagnosis (13). In a seroprevalence study of household members of COVID-19 outpatients in Norway, Cox et al. (14) found that 24/77 (31%) household members were seropositive 6 weeks after the index patient had first tested positive by RT-PCR. Similarly, a study of 61,075 participants in Spain found a 10.6% seroprevalence among coworkers, compared with 31.4% among household members of COVID-19 patients (16). These prevalence estimates are higher than our overall prevalence of 5.5% among close contacts who were not previously identified as positive by RT-PCR. However, comparing seroprevalence studies can be challenging because of differences in the nature and closeness of contacts, classification of contact types, methods of measuring antibodies, and characteristics of eligible study participants. In addition, the level of community transmission in each country would affect prevalence.

Among the 32 seropositive persons we identified, 21 had tested negative by RT-PCR, including 11 (52.4%) who had reported symptoms. The sensitivity of RT-PCR is dependent on multiple factors, primarily the presence of viral RNA on the test swab, which can vary by swabbing technique and whether viral RNA is present at the anatomic site of the test. Therefore, the timing of testing in relation to infection onset is critical. Among participants who underwent RT-PCR, most were tested during the early stage of quarantine, probably before symptom onset, and thus perhaps also before infection onset (e.g., among contacts who were in quarantine with the infected family member), which might explain their negative RT-PCR results (Table 2). In addition, many of the quarantined close contacts who were tested by RT-PCR were young children, in whom collecting adequate swab samples might be difficult.

The major strengths of this study are its nationwide nature, in which all close contacts of confirmed COVID-19 patients from the first wave in the Faroe Islands were directly contacted, and the high participation rate. As a result, the likelihood of selection bias is low. Because the Faroe Islands eliminated COVID-19 for 104 days after the last patient in this wave was identified, there was little risk for exposure from sources other than the index cases.

One limitation of our study is that contacts were asked retrospectively about symptoms during quarantine, introducing the possibility for recall bias. The risk for bias is especially relevant for children, whose questionnaires were answered by parents. However, because most index cases occurred in a family member, parents were probably vigilant for potential symptoms in children. Another limitation might be the classification of participants as household, workplace, or other contacts. Because this classification was based on whether participants knew the identity of the index patient, this measure might be imprecise. However, we were able to classify most participants into 3 groups on the basis of available information. We found that the primary exposure route is within families, and we believe this information is valuable, even if the data might be flawed. Finally, RT-PCR was not routinely conducted for quarantined persons. Because symptomatic persons were probably prioritized for RT-PCR, this selection might have introduced bias.

In conclusion, our study found that seroprevalence among close contacts of COVID-19 patients is higher than that among the general population. Close contacts, especially household members, of COVID-19 patients are at higher risk for infection. Thus, routinely testing household contacts of COVID-19 patients, regardless of symptoms, might improve detection of SARS-CoV-2 infection. Our results also indicate that close contacts should maintain quarantine even if they receive negative RT-PCR results early in quarantine.