We conducted a cross-sectional study of adults residing in Hillsborough County during October 2020–March 2021. The study was approved by the Advarra Institutional Review Board and Moffitt Cancer Center’s Scientific Review Committee. The University of Florida’s Bureau of Economic and Business Research drew the study population from the greater Hillsborough County by using randomly selected mailing addresses. Adults >18 years of age who were free of fever at the time of interview were eligible for the study. To ensure an adequate sample size of residents across the lifespan, we aimed to enroll relatively equal numbers of persons (balanced on sex) in each of 4 age groups: 18–34, 35–54, 55–64, and >65 years. We contacted potential participants by mail and email to inform them of the study. If they agreed to participate, they were scheduled for an in-person blood draw after completion of a web-based eligibility criteria checklist and informed consent form and a short questionnaire that captured demographic information, SARS-CoV-2 exposure history, underlying conditions, immunosuppression status, and use of immunosuppressive medications. Participants received a $25 gift card after completing the blood draw.

Among 867 COVID vaccine-naive Hillsborough County residents, 19.5% (95% CI 16.9%–22.3%) tested antibody-positive (Appendix Table); adjusted prevalence of 15% did not differ significantly from the crude estimate. The median age of study participants was 50 years (interquartile range 38–61 years), and 65.7% were women. Most participants were White (82.7%), non-Hispanic (83.2%), never smokers (74.2%), and immunocompetent (91.8%). Eighteen percent reported essential worker status (i.e., employed in either a hospital, clinic, grocery store, or in a public services industry). Approximately 60% had either never been exposed to a SARS-CoV-2–positive person or were not sure if they had been exposed. Nearly all respondents (99.5%) indicated they had changed their behavior since the pandemic started (data not shown); 96.1% reported wearing a mask outside of the home fairly often or often, and 89% reported keeping >6 feet away from other persons since the start of the pandemic. Approximately 30% (30.2%) reported ever having COVID-19 symptoms, although 14.5% reported testing virus positive. Only 7 participants had been hospitalized because of COVID-19.

Figure

Figure. Relationship between mean number of hours per week exposed to a SARS-CoV-2–positive person and antibody positivity among residents, Hillsborough County, Florida, USA, October 2020–March 2021. Error bars indicate 95% CIs....

We observed no differences in SARS-CoV-2 seroprevalence by sex, age group, race, or ethnicity (Appendix Table). Seroprevalence was higher among those living with lung disease (27.2%), especially persons who reported having asthma (28.1%), and lower among those living with an autoimmune disease (9.9%). Reported exposure to virus-positive persons was significantly associated with higher seroprevalence; 35.8% of those reporting contact with a documented positive person and 36.7% reporting contact with a presumed positive person were antibody-positive. In addition, 44.9% seroprevalence occurred among those whose virus-positive contact was a family member. More than 96% of those who tested positive for SARS-CoV-2 RNA (97% for those infected <6 months ago and 96.2% if positive >6 months ago) were antibody-positive. Forty-five percent of those who reported having COVID-19 symptoms were antibody-positive, and 8.3% of those who reported never having COVID-19 symptoms were antibody-positive, which we refer to as asymptomatic infection. Seroprevalence did not significantly differ by reported social-distancing or mask-wearing practices. We noted the relationship between hours per week exposed to a virus-positive person and antibody-positivity (Figure). The percentage testing antibody-positive increased with increasing exposure time but plateaued at ≈50% seroprevalence at >48 hours of exposure per week.

We noted factors independently associated with testing positive for SARS-CoV-2 antibodies (Table 1). Potential exposure to a virus-infected person increased the odds of testing antibody-positive, including essential workers employed in a hospital, clinic, grocery store or other public services industry (adjusted odds ratio [aOR] 2.40 [95% CI 1.42–4.07]), contact with a virus-positive family member (aOR 4.62 [95% CI 2.49–8.58]) or friend (aOR 4.22 [95% CI 2.44–7.30]), not avoiding groups of people (aOR 1.71 [95% CI 1.06–2.76]), and mean hours per week exposed to the virus-positive person (adjusted continuous odds ratio 1.01 [95% CI 1.00–1.01]). Odds of testing antibody-positive were high among those who reported ever experiencing COVID-19 symptoms (aOR 9.14 [95% CI 5.93–14.08]). We observed significantly lower odds of testing antibody-positive among divorced, separated, or widowed persons (aOR 0.40 [95% CI 0.20–0.77]) and those living with a chronic illness (aOR 0.56 [95% CI 0.34–0.93]).

Among 605 participants who reported never having COVID symptoms, 50 tested antibody-positive and are referred to as having asymptomatic infection (Appendix Table). Essential worker status (aOR 2.28 [95% CI 1.13–4.60]), interacting with a virus-positive friend (aOR 3.72 [95% CI 1.71–8.11]), and not avoiding groups of people (aOR 2.90 [95% CI 1.53–5.50]) were independently associated with having asymptomatic infection (Table 2).

Overall, ≈20% of study participants in this single Florida county had evidence of past infection with SARS-CoV-2, approximately 2-fold higher than the period prevalence of confirmed SARS-CoV-2 infections reported by the Florida Department of Health (10.6% through June 1, 2021) (11). This finding is not surprising given that molecular testing was not widely available early in the pandemic and, when it was available, not all persons with symptoms sought testing, and some never experienced symptoms; thus, many who were infected were undercounted in public health databases. A key finding of this study is that nearly 100% of persons who had confirmed or suspected infection were antibody-positive and remained antibody-positive even if the infection occurred >6 months before antibody testing. A question that remains unanswered by our analysis and other studies is the duration of the antibody response among those who experienced infection with SARS-CoV-2.

The seroprevalence estimated in our study demonstrates that, by March 2021, 1 in 5 adults residing in Hillsborough County may have been previously infected with SARS-CoV-2. This seroprevalence is higher than that found in other studies conducted during a similar timeframe of the COVID-19 pandemic; however, differences in SARS-CoV-2 seroprevalence are highly influenced by geographic location and the populations included in the study. For example, a seroprevalence study conducted in Virginia during June–August 2020 found that, although the overall prevalence estimated was low at 2.4%, the range by ZIP code varied from 0% to 20% (7). Similar to our study, the seroprevalence of 2.4% in Virginia was 2.8 times higher than the confirmed case counts. This ratio is relatively low compared with previous studies conducted in the United States, which have shown 6–53 times more infections than those ascertained by confirmed case counts (8,14,15). Differences in underascertainment across studies evaluating seroprevalence may be attributed to differences in the population included, timing of the epidemic across regions, and differences in test characteristics of assays used.

A nationwide study conducted by the Centers for Disease Control and Prevention that examined residual clinical samples from inpatients and outpatients found seroprevalence ranged from 0% in South Dakota (August 10–27, 2020) to 23% in New York State (July 27–August 13, 2020) (16). This analysis, which used commercial assays, found a seroprevalence of 8.5% for the state of Florida in September 2020. A prior Florida seroprevalence study of >5,500 healthcare workers and first responders tested in early summer of 2020 observed a seroprevalence of 4.1% (range 2.6%–8.7%) (17). However, considerable heterogeneity in SARS-CoV-2 antibody prevalence among first responders and healthcare workers was observed; those residing in Miami–Dade County and adult members of racial and ethnic minority populations, including Haitian, Creole, non-Hispanic Black, and Hispanic or Latino, were more likely to be seropositive (17). Similarly, a nationwide study of adults who had never had COVID-19 diagnosed found a seroprevalence of 4.6%; higher prevalence was found among adults living in early outbreak locations, Black adults, Hispanic adults, and adults residing in urban areas (18). Although our study did not demonstrate racial disparities, we observed that essential workers, including those working in grocery stores, had higher odds of SARS-CoV-2 antibody positivity. Because persons in minority communities are more likely to hold such occupations (19), our results contribute to the literature demonstrating the disproportionate burden of COVID-19 in the United States among vulnerable populations such as racial and ethnic minorities.

In our study, mask wearing was not associated with antibody status. This lack of an association is likely attributable to several factors. Nearly 100% of respondents reported mask usage, so social desirability likely influenced responses to this question. In addition, we did not assess detailed information regarding mask use, such as type of mask, situations in which mask wearing occurred, and consistency of proper mask usage, so our study cannot adequately assess the protection conferred by mask usage. In contrast, social-distancing behaviors consistently emerged as a factor associated with risk for infection, despite 11% of participants reporting that they never, rarely, or almost never practiced social distancing. This behavior included not avoiding crowds and interacting with a known or suspected virus-positive family member, co-worker, or friend. The percentage seropositive increased with increasing hours per week exposed to an infected person. We did not ask participants to report the physical distance they maintained, so we cannot assess what distance in feet was associated with protection.

Throughout the pandemic, much discussion has occurred about the dangers of infection for those with autoimmune disease and their greater susceptibility to illness. Study participants with autoimmune disease or on immunosuppressant medication had lower rates of antibody positivity, roughly half that of the overall study population. This finding is likely attributable to extra precautions taken to avoid infection as opposed to a reflection of actual susceptibility to illness.

Despite some study participants reporting a large number of hours per week exposed to a person with COVID-19 symptoms, only ≈50% became infected themselves. Infection rates among adults with high exposure to COVID-19 is likely multifactorial and varies from population to population on the basis of contextual factors, which can be demonstrated by seroprevalence studies conducted in large healthcare systems of healthcare workers with high exposure to COVID-19. For example, the seroprevalence of healthcare personnel tested during June–August 2020 throughout all Mayo Clinic facilities was only 0.6%, and areas undergoing greater community disease transmission and burden were associated with higher seroprevalence among healthcare providers (20). Notably, the Mayo Clinic in Florida had a seroprevalence of 0.8%. In contrast, the seroprevalence of healthcare workers in New York City during the same period was 13.7% (21). The marked difference in seroprevalence among these highly exposed adults may be caused by differences in hotspots or outbreaks of SARS-CoV-2 infection in the regions each health system is located, as well as differences in access to personal protective equipment and adherence to precautions including wearing masks. Similarly, differences we observed in our study may be caused by variability in adherence to preventive behaviors, such as prompt isolation from persons with COVID-19, or the household setting and environment. We observed that the odds of SARS-CoV-2 seroprevalence was highest among those with a family member as a known virus-positive contact. In shared family spaces where social distancing may not be possible, risk for household transmission is high (22). Asymptomatic transmission before the onset of symptoms in a household is also highly probable.

A low percentage of antibody-positive persons never had COVID-19 symptoms, what we refer to as the asymptomatic infection prevalence. Surprisingly, this percentage was relatively low (≈8%) and was not associated with age. The only factors significantly associated with asymptomatic infection were those related to social distancing, whether that was not avoiding crowds, contact with a friend who was virus-positive, or repeated contact with community members as an essential worker.

A strength of this study is the relatively large sample size and inclusion of a broad range of ages. We captured participant information regarding factors that may be associated with susceptibility to infection, protective behaviors practiced, and exposure and length of exposure to SARS-COV-2–infected persons. Although invitations were sent at random to county residents, a small rate of participation resulted from this recruitment method. Some residents may have received multiple invitations (i.e., email and letter, or email and postcard). Because of the low rate of participation, we were not able to obtain a representative sample of the underlying county population. However, the study enrolled persons from 53 (96%) of the 55 ZIP codes associated with Hillsborough County. The final study sample included a higher proportion of women than men, and participants were predominantly non-Hispanic White. Many reported exposure to an infected person, so the seroprevalence we report may be an overestimate of the actual period prevalence. Hillsborough County includes the city of Tampa as well as both rural and suburban communities. Although the area is relatively densely populated, few residents use public transportation; this county is less densely populated than other urban counties in Florida, such as Miami–Dade County. Therefore, the period seroprevalence we have reported may be lower than what would be observed in city centers.

The estimates of seroprevalence from our study demonstrate that the cumulative case numbers confirmed through molecular RNA–based testing likely underrepresent the actual number of cases of SARS-CoV-2 infection in the United States and Florida. Frequency of contact with family or friends with confirmed COVID-19 diagnoses was strongly associated with being SARS-CoV-2 antibody–positive, indicating the importance of social distancing, particularly from friends or family with confirmed COVID-19. The availability of vaccination should help alleviate disparities in SARS-CoV-2 positivity observed for higher risk groups because of structural and occupational factors, such as among essential workers and those with frequent contact with persons with confirmed COVID-19. This analysis should inform the broader ongoing policy in the United States regarding the relative benefits of recommended mitigation strategies against the spread of SARS-CoV-2.