Prevalence and Correlates of SARS CoV-2 Among a Community-Based Sample Recruited Using Randomized Venue-Based Sampling. Essex County, NJ, 2020.

Disparities in morbidity and mortality related to COVID-19 based on race and ethnicity have been documented in the USA. However, it is unclear if these disparities also exist at the exposure stage. To determine this, studies are needed to document the underlying burden of disease, potential disparities through serologic surveillance. Additionally, such studies can help identify where along the disease spectrum (e.g., exposure, infection, diagnosis, treatment, death) and with regard to the structural factors that necessitate public health and/or clinical interventions. Our objectives in this study were to estimate the true burden of SARS CoV-2 in the community of Essex County, NJ, an early and hard hit area, to determine the correlates of SARS CoV-2 prevalence and to determine if COVID-19 disparities seen by race/ethnicity were also reflected in SARS CoV-2 burden. We utilized venue-based-sampling (VBS) to sample members of the community in Essex County. Participants completed a short electronic survey and provided finger stick blood samples for testing. We sampled 924 residents of Essex County, New Jersey. Testing conducted in this study identified 83 (9.0%) participants as positive for SARS-CoV-2 antibodies. Importantly, our findings suggest that the true burden of SARS-Cov-2 and the pool of persons potentially spreading the virus are slightly more than six times than that suggested by PCR testing Notably, there were no significant differences in odds of testing positive for SARS CoV-2 antibodies in terms of race/ethnicity where we compared Black and Latinx participants to other race participants. Our study suggests that disparities in COVID-19 outcomes stem from potential upstream issues such as underlying conditions, access to testing, and access to care rather than disparities in exposure to the virus.

Introduction

Disparities in the USA based on race and ethnicity with regard to COVID-19 have been documented in terms of morbidity and mortality, but it is unclear if there are similar disparities in exposure to SARS CoV-2 [1-7]. While such patterns parallel other health conditions, the underlying psychosocial and structural mechanisms that enable these disparities continue to be elucidated, minority stressors and social inequities are surely the drivers of the disease as noted in recent conceptualizations of COVID-19 and other diseases [8-10].

To this end, studies are needed to document the underlying burden of disease through serologic surveillance. These types of studies are an important complement to case reporting and data on morbidity and mortality. Additionally, seroprevalence studies empower public health practitioners and others to document potential disparities in the true burden and to identify where along the disease spectrum (e.g., exposure, infection, diagnosis, treatment, death) and with regard to the structural factors that necessitate public health and/or clinical interventions.

For example, if the disparities exist at the exposure stage, interventions might be needed to mitigate the conditions which facilitate exposure. In contrast, if the disparities are only seen in outcomes that are exacerbated by underlying medical conditions or access to care, further upstream health interventions may be needed. Such initiatives are critically important since some 30% infected persons are asymptomatic; thus, our understanding of the burden of disease is informed solely by case reporting and may not capture the actual profile of the diseases particularly in hard to reach and marginalized populations [11].

SARS CoV-2 initially manifested in the locally in the Northeastern region of the USA where morbidity and mortality were extremely high in the second quarter of 2020. Essex County, New Jersey, adjacent to New York City, a COVID-19 epicenter, ranks third in New Jersey as of mid-March 2021 for COVID-19 cases (69,386) and deaths (2,431) translating to a morbidity and mortality rate of approximately 8700 per 100,000 and 305 per 100,000 respectively [12, 13]. By comparison, the rate of the cases is comparable in the USA, although morbidity is lower at the rate of 164 per 100,000 [14]. The extent to which these differences in morbidity are due to temporal factors (i.e., when the pandemic emerged), structural factors or the interaction of both has yet to be delineated.

Essex County is highly diverse along numerous strata with a majority of residents who are Black or Latino (62.7%), a high proportion living in poverty (15.8%), and a densely populated region with 6000 residents per square mile [13]. It is well documented that communities of color, those living in poverty, and in dense residential settings experience disparities in disease burden, morbidity, and mortality, for instance, HIV, diabetes, maternal coronary heart disease, all-cause mortality, breast cancer and hepatocellular carcinoma, alcohol/substance use, and of course COVID-19 to name just a few [7, 15–21].

Our objectives in this study were to estimate the true burden of SARS CoV-2 in the community, to determine the correlates of SARS CoV-2 prevalence and to determine if COVID-19 disparities seen by race/ethnicity were also reflected in SARS CoV-2 burden.

Methods

We utilized venue-based-sampling (VBS) to sample members of the community in Essex County. VBS is often employed as an alternative sampling method where simple random sampling is not feasible. In VBS, the venues are randomized as a proxy for randomizing the attendees of the venues. For example, VBS is the standard to sample certain populations at high risk for HIV infection where the population is known to attend identifiable venues. VBS is also the standard sampling method used by CDC for men who have sex with men in the USA in National HIV Behavioral Surveillance (NHBS) and is used for other populations internationally but to our knowledge has not been used for SARS CoV-2 [22, 23]. North Jersey Community Research Initiative (NJCRI), which has over 20 years’ experience in VBS as the implementing partner in NHBS for New Jersey, performed the VBS sampling for the present study by utilizing their mobile units, which have medical equipment, survey and phlebotomy space, and a private exam room. These units were also outfitted for COVID-19 by adding additional air filtration units and plexi-glass dividers to keep participants and staff safe. We identified venues using Google.com searches using the key words “Supermarkets in Essex County” and “Groceries in Essex County” to search for all potential locations. We identified 205 potential locations. Days and hours of operation were verified by contacting each venue directly. Venues, days, and times of operation were entered in an Excel spreadsheet. To map the venues, we used R 4.0.2 GUI (The R Foundation for Statistical Computing, Austria) to manipulate and analyze the data. Specifically, we used “tidygeocoder” package 1.0.1 to extract coordinates of each store from US Census and OpenStreetMap (OSM) dataset. For each month of sampling, approximately 42 venue-day-times were randomly sampled and scheduled for a sampling event. NJCRI’s mobile units and staff used the VBS schedule and travelled to these grocery store sites. At the site, potential participants were systematically approached for a standard two-hour period of time, asked to stop for screening (intercept), screened for eligibility, and, if eligible, invited to participate and provide verbal informed consent. Participants who agreed to participate completed a short computer-based survey administered by study staff, using a tablet computer.

The survey included questions on basic self-reported demographics, COVID-19 symptoms (including: fever, dry cough, shortness of breath, chills, headache, sore throat, muscle pain/ aches, new loss of taste or smell, GI issues, neurologic symptoms) since January 2020, seeking viral PCR testing, PCR results, and seeking antibody testing and antibody results. Participants were then asked to provide a blood sample collected by finger stick with the Neoteryx Mitra Collection device to use for SARS-CoV-2 antibody testing [24]. Dried sticks were preserved by the NJCRI collection team or at the testing laboratory at room temperature for up to a week, according to manufacturer’s instructions. At the time of testing, dried blood from the Mitra device was eluted overnight at 4 °C, based on published protocols [24]. The eluate was diluted with 2% non-fat dry milk to1:80 and utilized for detection of IgG antibodies to recombinant receptor binding domain (RBD) of the SARS-CoV-2 Spike (100 ng/well), utilizing standard enzyme-linked immunosorbent assay (ELISA) protocols [25, 26]. To interpret the ELISA data, we calculated a cut off of OD405 = 0.3 (mean + 3 standard deviations) by testing finger-stick eluates from 50 COVID-19 PCR-negative subjects.

Participants were given a result-return card to enable them to call in to the study team for their antibody results a week after sample receipt by the laboratory. Participants received $25 in cash for their participation. This study had ethical approval from the Rutgers IRB. Survey data and lab data were merged in SAS (v9.4). Descriptive, bivariate, and multivariable analyses were conducted in SAS (v9.4).

Results

From September 15, 2020, to December 22, 2020, we conducted 92 randomly selected sampling events from geographically diverse locations in Essex County where we systematically approached patrons for screening. We approached 1349 individuals for screening. Of these, 928 (68.8%) stopped and agreed to be screened, 928 (100%) were eligible and agreed to participate, and 924 (99.6%) consented and had complete survey and laboratory data for analysis.

Our sample population was mostly over 39 years of age (n = 593, 64.2%) with a mean age of 44.7 years, mostly male at birth (n = 584, 63.2%), predominantly Black (n = 656, 71.0%), and Latino (n = 194, 21.0%). A majority had a high school diploma or some college education (n = 652, 70.5%). Only 343 (37.1%) were employed full or part-time, while 398 (43.1%) reported being unemployed. Personal annual income reported by participants was low, with almost half, 442 (47.8%), reporting a personal annual income of less than $15,000 while another 300 (32.5%) reported an income of between $15,000 and $35,000. Homelessness was common in our sample, with 251 (27.2%) reporting being homeless sometime in the 12 months preceding the survey and 181 (19.6%) reporting being homeless at the time of the study. A majority of the sample reported living in Newark (n = 704, 76.2%). Finally, 77.9% (n = 720) of the sample reported having some sort of health insurance at the time of the study (Table 1).

Table 1

Demographics, symptoms, testing seeking, and risk indicators and prevalence of SARS-CoV-2 antibody prevalence in a community sample, Essex County, NJ, 2020

Variable	N (924)	% (100)
Age
18–21	63	6.8
22–29	111	12.0
30–39	157	17.0
40–49	185	20.0
50 +	408	44.2
Sex at birth
Female	340	36.8
Male	584	63.2
Race/ethnicity
Native American	2	0.2
Asian	6	0.6
Black	656	71.0
White	44	4.8
Latino	194	21.0
Other	11	1.2
Mixed	6	0.6
Missing	5	0.5
Educational attainment
None	2	0.2
Grades 1–8	29	3.1
Grade 9–11	170	18.4
HS diploma/GED	466	50.4
Some college	186	20.1
Bachelors	53	6.4
Any grad school	12	1.3
Employment status
Full time	248	26.8
Part time	95	10.3
Homemaker	11	1.02
Student	9	1.0
Retired	42	4.6
Disabled	120	13.0
Unemployed	398	43.1
Missing	1	0.1
Personal annual income
≤ 14,999	442	47.8
15,000–35,000	300	32.5
36,000–50,000	72	7.8
51,001–80,000	34	3.7
> 80,000	76	8.2
Homeless 12 months prior to survey	251	27.2
Homeless at time of survey	181	19.6
Residential locality within Essex County
Belleville	1	0.1
Bloomfield	15	1.6
East Orange	56	6.1
Elizabeth	19	2.1
Hillside	2	0.2
Irvington	59	6.4
Maplewood	2	0.2
Montclair	3	0.3
Newark	704	76.2
Orange	52	5.6
Roselle	1	0.1
Short Hills	1	0.1
Verona	2	0.2
West Orange	6	0.7
Homeless	1	0.1
Has health insurance	720	77.9
Any COVID-19 symptoms	60	6.5
Sought a polymerase chain reaction (PCR) virus test	424	45.9
Got PCR test	407	44.1
Got PCR result	383	41.5
Reported a positive previous PCR test
Yes	13	1.4
No	911	98.6
Sought Antibody (AB) test	33	3.6
Got Ab test outside study	29	3.1
Got Ab result outside study	23	2.5
Reported a previous positive AB test
Yes	5	0.5
No	919	95.5
SARS CoV-2 Antibody result
Negative	841	91.0
Positive	83	9.0
Household size (including respondent)
1	106	11.5
2	373	40.4
3	193	20.9
4	118	12.8
5 +	134	14.5
Any household COVID-19 diagnosis
Yes	46	5.0
No	878	95.0
How often respondent leaves the house each week
Everyday	778	84.2
Weekly	111	12.0
Once a month	21	2.3
Less than once a month	14	1.5

In terms of SARS CoV-2 risks, just over half of the sample reported living alone (11.5%) or with just one other person (40.4%) and only 14.5% reported living in households with five or more persons. Only a small proportion of the sample (5.0%) reported having had someone in their household diagnosed with COVID-19. The majority of the sample reported leaving their house daily (84.2%), and only 4.8% reported only leaving home once a month or less than once a month. Only a few participants reported a previous positive PCR test result (13/924, 1.4%).

Only 6.5% of the sample reported any COVID-19 like symptoms since January 2020, while 45.9% reported having sought out a COVID-19 test. In total, 13 (1.4%) participants reported receiving a positive SARS-CoV-2 PCR test result. While 33 participants (2.6%) sought a SARS-CoV-2 antibody test outside of the study, only 0.5% of the sample reported a positive antibody result. Testing conducted in this study identified 83 (9.0%) participants as positive for SARS-CoV-2 antibodies (Fig. 1).

Fig. 1

COVID-19 and SARS-CoV-2 indicators among a community sample, Essex County, NJ, 2020

In our bivariate analysis, only age (Χ2 = 16.1, p < 0.01), employment status (X2 4.8, p = 0.03), household diagnosis of COVID-19 (p = 0.01), and reporting a previous positive PCR test (p < 0.01) were significantly associated with testing positive for SARS CoV-2 antibodies. Antibody prevalence across age groups was not uniform nor was prevalence higher in older participants. Indeed, 18–21-year olds had the highest prevalence at 19.1%. Interestingly, reporting homelessness in the 12 months preceding the survey was significantly negatively correlated with testing positive for SARS CoV-2 antibodies (X2 = 4.9, p = 0.03) (Table 2).

Table 2

Bivariate associations with SARS CoV-2 prevalence among a community sample, Essex County NJ, 2020

	SARS CoV-2 negative (841)		SARS CoV-2 positive (83)
Variable	n	%	n	%	X2	p
Age					16.1	< 0.01
18–21   22–29   30–39   40–49   50 +	51 106 137 175 372	81.0 96.0 87.3 94.6 91.2	12 5 20 10 36	19.1 4.5 12.7 5.4 12.1
Sex at birth					1.1	0.3
Female   Male	305 536	89.7 91.8	35 48	10.3 8.2
Race/ethnicity						1.0*
Black   Latinx   Other	596 177 68	90.9 91.2 91.9	60 17 6	9.2 8.8 8.1
Educational attainment					3.4	0.2
Less than high school   High school   More than high school	189 423 229	94.0 90.8 89.1	12 43 28	6.0 9.2 10.9
Employment status					4.8	0.03
Employed   Not working	303 538	88.3 92.6	40 43	11.7 7.4
Personal annual income					0.6	0.4
< 35,000   ≥ 35,000	678 163	91.4 89.6	64 19	8.6 10.4
Homeless in the 12 months prior to survey					4.9	0.03
Yes   No	237 604	94.4 89.6	14 69	10.4 8.6
Homeless at time of survey					2.3	0.1
Yes   No	170 671	93.9 90.3	11 72	6.1 9.7
Residential locality within Essex County					0.4	0.8
Newark   Other	640 201	90.9 91.4	64 19	9.1 8.6
Household size (including respondent)					7.3	0.1
1   2   3   4   5 +	101 945 172 102 121	95.3 92.5 89.1 86.4 90.3	5 28 21 16 13	4.7 7.5 10.9 13.6 9.7
Any household COVID-19 diagnosis						0.01*
Yes   No	37 304	80.4 91.6	9 74	19.6 8.4
How often respondent leaves the house each week					0.4	0.5
Everyday   Less than everyday	706 135	90.8 92.5	72 11	9.3 7.5
Reported a positive previous PCR test						< 0.01*
Yes   No	7 834	53.9 91.6	6 77	46.2 8.5

*Fisher exact test

In our multivariable analysis, we included age, race/ethnicity, sex assigned at birth, employment status, personal annual income, previous PCR test, household COVID-19 diagnosis, and homelessness in the 12 months prior to the study. While controlling for all of these variables simultaneously only age and household COVID-19 diagnosis were significantly associated with higher odds of testing positive for SARS CoV-2 antibodies. In terms of age, younger persons (18–21) had significantly higher odds (adjusted odds ratio [AOR] 5.6, 95% confidence interval [95% CI] 1.8, 17.2, p < 0.01) of testing positive for SARS CoV-2 antibodies compared to those aged 22–29. Those who previously had a positive PCR test had higher odds (AOR 11.3, 95% CI 3.2, 38.9, p < 0.01) of testing positive for SARS CoV-2 antibodies. Notably, there were no significant differences in odds of testing positive for SARS CoV-2 antibodies in terms of race/ethnicity where we compared Black and Latinx participants to other race participants.

Discussion

We successfully sampled a large number of residents of Essex County, New Jersey, for serological testing for SARS CoV-2 antibodies. Our sample was predominantly Black and Latino and residing in Newark. Importantly, our findings suggest that the true burden of SARS-Cov-2 and the pool of persons potentially spreading the virus are slightly more than six times than that suggested by PCR testing. This burden from our robustly sampled community-based study is higher than most estimates obtained through studies of remnant blood samples from hospitals (4.2% antibody prevalence), samples from staff at a public-school system (2.9% antibody prevalence), and residents of a California county recruited with targeted Facebook ads (1.5% antibody prevalence). These differences underscore the importance of conducting robust community-based sampling to determine the true burden of disease [27-29]. Moreover, with only 6.5% of the sample reporting any COVID-19-like symptoms, our finding suggests that the number of asymptomatic persons may be close to 1.5 times greater than anyone reporting symptoms (it is noted however that the symptoms listed might include other conditions).

In our bivariate analysis, we found that antibody prevalence was higher among younger and older persons, among those who were employed, those that were homeless sometime in the 12 months preceding the survey, and those who reported someone in their household being diagnosed with COVID-19. We did not find significant differences in SARS CoV-2 antibody prevalence by race/ethnicity, or frequency of leaving home. Other studies have also noted ongoing employment as potential risk factor serving as proxy for the risks experienced by front line workers. Indeed, our multivariable analysis suggests that only younger age and having previously received a positive PCR test for the virus were associated with SARS CoV-2 antibodies (Table 3).

Table 3

Adjusted multivariable associations with SARS CoV-2 antibody prevalence among a community sample, Essex County NJ, 2020

Variable	AOR	95% CI	p
Age
18–21	5.6	1.8, 17.2	< 0.01
22–29	Ref	-	-
30–39	3.7	1.3, 10.5	0.1
40–49	1.4	0.14, 4.3	0.6
50 +	2.5	0.9, 6.8	0.07
Race/ethnicity
Black	1.2	0.5, 3.3	0.7
Latinx	1.1	0.4, 3.1	0.9
Other	Ref	-	-
Birth sex
Female	1.2	0.7, 1.9	0.5
Male	Ref	-	-
Employment
Employed	1.4	0.8, 2.3	0.2
Not working	Ref	-	-
Personal annual income
< 35,000	1.1	0.6, 1.9	0.8
≥ 35,000	Ref	-	-
Reported a positive previous PCR test
Yes	11.3	3.2, 38.9	< 0.01
No	Ref	-	-
Any household COVID-19 diagnosis
Yes	2.1	0.9, 4.9	0.08
No	Ref	-	-
Homeless in 12 months prior to the study
Yes	Ref	0.9, 3.2	0.1
No	1.7	-	-

Our study has limitations. VBS may not reach individuals who do not do their own shopping or live in settings where food is provided, thus introducing some form of sampling bias. Secondly, while Essex County is predominantly populated by Black/Latino individuals, our study oversampled these groups. This may be due to lower enthusiasm for participation among other race/ethnicity groups that may not be as altruistic or interested in the incentive provided by the study. Despite these limitations, we were able to sample a large number of individuals and estimate SARS-CoV-2 antibody prevalence in a county highly impacted by the current pandemic. Our findings underscore the importance of conducting seroprevalence studies in addition to case reporting to understand the true magnitude of SARS-CoV-2 impact on communities and plan for mitigation strategies to reduce transmission.

In conclusion, our study suggests that disparities in COVID-19 outcomes stem from potential upstream issues such as underlying conditions, access to testing, and access to care rather than disparities in exposure to the virus. In effect, these patterns manifest a biopsychosocial understanding of COVID-19, namely, the structural and social inequities that undermine health and interfere with access to effective healthcare by communities of color and other marginalized populations [9]. Our findings also suggest that other factors may be influencing these disparities. These factors may include hesitancy to get tested or hesitancy to seek care due to concerns about health care coverage/costs of care and/or fear and mistrust of the medical profession [8]. This bears further investigation particularly in relation to whether persons of color present for COIVD-19 care at more advance stages of the disease than other populations. SARS CoV-2 has exposed, in a dramatic way, the underlying systemic issues that influence health disparities in regard to race and ethnicity in the USA and point to ongoing vigilance in creating conditions that foster health equity both locally and internationally [8].