Seroprevalence of SARS-CoV-2 antibodies and associated factors in health care workers: a systematic review and meta-analysis.

BACKGROUND: Health care workers (HCWs) represent a high risk population for the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. AIM: To determine the seroprevalence of SARS-CoV-2 antibodies among HCWs, and to find out the factors that are associated with this seroprevalence.
METHODS: The Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines were applied for this systematic review and meta-analysis. Databases including PubMed/MEDLINE and pre-print services (medRχiv and bioRχiv) were searched from inception up to August 24, 2020.
FINDINGS: Forty-nine studies, including 127,480 HCWs met the inclusion criteria. The estimated overall seroprevalence of SARS-CoV-2 antibodies among HCWs was 8.7% (95% CI: 6.7-10.9%). Seroprevalence was higher in studies that were conducted in North America (12.7%) compared to those in Europe (8.5%), Africa (8.2), and Asia (4%). Meta-regression showed that increased sensitivity of antibodies test was associated with increased seroprevalence. The following factors were associated with seropositivity: male gender, Black, Asian, and Hispanic HCWs, work in a coronavirus disease 2019 (COVID-19) unit, patient-related work, frontline health care workers, health care assistants, personal protective equipment shortage, self-reported belief for previous SARS-CoV-2 infection, previous positive polymerase chain reaction test, and household contact with suspected or confirmed COVID-19 patients.
CONCLUSION: The seroprevalence of SARS-CoV-2 antibodies among HCWs is high. Excellent adherence to infection prevention and control measures, sufficient and adequate personal protective equipment, and early recognition, identification and isolation of HCWs that are infected with SARS-CoV-2 are imperative to decrease the risk of SARS-CoV-2 infection.

Introduction

Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) and coronavirus disease 2019 (COVID-19) emerged from Wuhan, Hubei Province, China in December 2019, and the World Health Organization (WHO) declared a pandemic situation on 11th March 2020 [1]. As of 2nd October 2020, WHO reported 34,079,542 cases and 1,015,963 deaths globally due to COVID-19 [2].

Healthcare workers (HCWs) are a high-risk group for infection. A recent meta-analysis with 11 studies found that the proportion of HCWs who were SARS-CoV-2 positive among all patients with COVID-19 was 10.1%, but severity and mortality among HCWs were lower than among all patients with COVID-19 [3]. This proportion varied substantially between countries: China, 4.2%; Italy, 9%; and USA, 17.8% [3]. The lower proportion in China is probably due to immediate implementation of strong public health interventions, such as lockdown measures, home isolation, quarantine measures, wearing masks and social (physical) distancing [4].

SARS-CoV-2 and COVID-19 have significant diagnostic issues, and serological tests aim to identify previous SARS-CoV-2 infection by detecting the presence of SARS-CoV-2 antibodies. It is known that SARS-CoV-2 antibody tests are accurate to detect previous SARS-CoV-2 infection if performed >14 days after the onset of symptoms, but they have very low sensitivity in the first week after symptom onset [5]. Also, rapid diagnostic tests for SARS-CoV-2 antibodies have low pooled sensitivity (64.8) and high pooled specificity (98%), but these data suffer from low power and other significant limitations [6].

Knowledge of the seroprevalence of SARS-CoV-2 antibodies among HCWs is important to understand the spread of COVID-19 among healthcare facilities, and to assess the success of public health interventions. To the authors' knowledge, the overall seroprevalence of SARS-CoV-2 antibodies among HCWs and the associated factors are unknown. Thus, the primary objective of this systematic review and meta-analysis was to determine the seroprevalence of SARS-CoV-2 antibodies among HCWs, and the secondary objective was to identify the factors associated with this seroprevalence.

Methods

Data sources and strategy

The Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines were applied in this systematic review and meta-analysis [7]. The PRISMA checklist is presented in Table S1 (see online supplementary material). PubMed/MEDLINE and preprint services (medRχiv and bioRχiv) were searched from inception to 24th August 2020. In addition, reference lists of all relevant articles were searched, and duplicates were removed. The following search strategy was used: (‘sars-cov-2 antibodies’ OR ‘COVID-19 antibodies’ OR ‘sars-cov-2’ OR ‘COVID-19’ OR antibodies) AND (‘health care personnel’ OR ‘healthcare personnel’ OR ‘health-care personnel’ OR ‘health care workers’ OR ‘health-care workers’ OR ‘healthcare workers’ OR ‘healthcare staff’ OR ‘health care staff’ OR ‘health-care staff’ OR ‘medical staff’).

Selection and eligibility criteria

Two authors undertook study selection independently, and a third (senior) author resolved any disagreements. All studies written in English (except case reports) that reported the seroprevalence of SARS-CoV-2 antibodies among HCWs and associated factors were included. In addition, studies reporting any serological test (e.g. enzyme-linked immunosorbent assay, chemiluminescence immunoassay) used to detect SARS-CoV-2 antibodies (IgA, IgG and IgM) in all HCWs were included. Finally, studies performed under screening conditions where HCWs were not selected for participation based on previous exposure to SARS-CoV-2 or symptoms were also included.

Data extraction and quality assessment

Data collected included authors, location, dates of data collection, sample size, setting, study design, antibody tests, sensitivity and specificity of antibody tests, number of HCWs with SARS-CoV-2 antibodies, factors associated with seroprevalence of SARS-CoV-2 antibodies, and level of analysis (univariate or multi-variate).

The quality of studies was assessed using the Joanna Briggs Institute critical appraisal tools, where a nine-point scale is used for prevalence studies, an eight-point scale is used for cross-sectional studies and an 11-point scale is used for cohort studies [8]. In prevalence studies, a score of 8–9 indicates good quality, a score of 5–7 indicates moderate quality and a score ≤4 indicates poor quality. In cross-sectional studies, a score of 7–8 indicates good quality, a score of 4–6 indicates moderate quality and a score ≤3 indicates poor quality. In cohort studies, a score of 9–11 indicates good quality, a score of 5–8 indicates moderate quality and a score ≤4 indicates poor quality.

Statistical analysis

For each study, the total number of HCWs and the number of HCWs who were positive for SARS-CoV-2 antibodies were extracted. Seroprevalence and 95% confidence intervals (CI) were calculated for each included study. Seroprevalence was transformed with the Freeman–Tukey double arcsine method before pooling [9]. Between-studies heterogeneity was assessed using Hedges Q statistic and I 2 statistic. Statistical significance for Hedges Q statistic is set at P<0.1, while I 2 values >75% indicate high heterogeneity [10]. A random effects model was applied to estimate pooled seroprevalence as heterogeneity between results was very high [10,11]. Study quality, sample size, sensitivity and specificity of antibody tests, publication type (journal or preprint service) and the continent where studies were conducted were considered as prespecified sources of heterogeneity, and explored using subgroup analysis and meta-regression analysis. In addition, leave-one-out sensitivity analysis was performed by removing one study at a time to determine the influence of each study on overall prevalence. A funnel plot and Egger's test were used to assess publication bias. P<0.05 for Egger's test indicates publication bias [12]. Meta-analysis was not performed for factors associated with the seroprevalence of SARS-CoV-2 antibodies as the data were very scarce. Statistical analysis was performed using OpenMeta[Analyst] [13].

Results

Identification and selection of studies

A flowchart of the literature search is summarized in PRISMA format (Figure 1 ). Initially, 3632 potential records were identified through PubMed and 103 records were identified through preprint services for health sciences (i.e. medRχiv and bioRχiv) after removal of duplicates. After screening the titles and abstracts, 3684 records were removed. Twelve additional records were identified and included after searching the reference lists. Finally, 49 studies that met the inclusion criteria were included in this meta-analysis.

Figure 1

Flowchart of the literature search according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines.

Characteristics of the studies

The main characteristics of the 49 studies included in this systematic review and meta-analysis are shown in Table I . In total, 127,480 HCWs were included. Forty-nine studies [[14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48], [49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62]] reported data regarding the seroprevalence of SARS-CoV-2 antibodies among HCWs, and 27 studies [14,15,18,19,[21], [22], [23], [24], [25],[27], [28], [29], [30], [31], [32],[34], [35], [36], [37],39,44,47,52,54,58,60,61] investigated factors for SARS-CoV-2 antibody positivity.

Table I

Main characteristics of studies included in the systematic review and meta-analysis

Reference	City or state/country	Females (%)	Age (years), mean (SD)	Sample size (N)	Study design	Sampling method	Response rate (%)	Dates of data collection	Setting	Publication
Moscola et al., 2020 [14]	New York/USA	73.7	42.7 (17.1)	40,329	Cross-sectional	Convenience sampling	65.1	20 April–23 June	Primary care facilities and hospitals	Journal
Jeremias et al., 2020 [15]	New York/USA	70.2	42.8 (13.8)	1699	Cross-sectional	Convenience sampling	NR	1 March–30 April	Hospitals	Journal
Houlihan et al., 2020 [16]	London/UK	NR	35.8 (11.2)	181	Cohort	Convenience sampling	NR	26 March–8 April	Hospitals	Journal
Poulikakos et al., 2020 [17]	North West England/UK	73	NR	281	Prevalence	Convenience sampling	NR	NR	Hospitals	Journal
Steensels et al., 2020 [18]	Genk/Belgium	NR	NR	3056	Cross-sectional	Convenience sampling	74	22–30 April	Hospitals	Journal
Blairon et al., 2020 [19]	Brussels/Belgium	72.4	43.9 (1.7)a47.4 (2.1)b	1485	Cross-sectional	Convenience sampling	47.7	25 May–19 June	Hospitals	Journal
Pallett et al., 2020 [20]	London/UK	72.7	39.1 (12.1)	6440	Cohort	Convenience sampling	NR	8 April–12 June	Hospitals	Journal
Korth et al., 2020 [21]	Essen/Germany	NR	NR	316	Cross-sectional	Convenience sampling	65	25 March–21 April	Hospitals	Journal
Martin et al., 2020 [22]	Brussels/Belgium	73	37 (11.3)	326	Cohort	Convenience sampling	87.3	15 April–18 May	Hospitals	Journal
Amendola et al., 2020 [23]	Milan/Italy	83.7	NR	547	Cross-sectional	Convenience sampling	89.4	15 April	Hospitals	Journal
Self et al., 2020 [24]	Washington, Oregon, California, Minnesota, Tennessee, Ohio, North Carolina, New York, Massachusetts, Utah, Colorado, Maryland/USA	65.6	38.5 (12.6)	3248	Cross-sectional	Convenience sampling	NR	3 April–19 May	Hospitals	Journal
Grant et al., 2020 [25]	London/UK	NR	40.3 (11.1)	2004	Cross-sectional	Convenience sampling	54.2	15 May–5 June	Primary care facilities and hospitals	Journal
Mughal et al., 2020 [26]	New Jersey/USA	75	38.5 (15.4)	121	Prevalence	Convenience sampling	NR	1 March–30 April	Hospitals	Journal
Hunter et al., 2020 [27]	Indiana/USA	70.1	43 (NR)	690	Cross-sectional	Convenience sampling	NR	29 April–8 May	Hospitals	Journal
Plebani et al., 2020 [28]	Veneto Region/Italy	71.6	43.2 (11.6)	8285	Cross-sectional	Convenience sampling	NR	22 February–29 May	Primary care facilities and hospitals	Journal
Mansour et al., 2020 [29]	New York/USA	46	38.4 (10.8)	285	Cross-sectional	Convenience sampling	NR	24 March–4 April	Hospitals	Journal
Sotgiu et al., 2020 [30]	Milan/Italy	65.3	44.6 (14.2)	202	Cross-sectional	Convenience sampling	NR	2–16 April	Hospitals	Journal
Garcia-Basteiro et al., 2020 [31]	Barcelona/Spain	72.1	42.1 (11.6)	578	Cross-sectional	Random sampling	74.3	9 March	Hospitals	Journal
Sydney et al., 2020 [32]	New York/USA	NR	NR	1700	Cross-sectional	Convenience sampling	NR	28 April–4 May	Hospitals	Journal
Khalil et al., 2020 [33]	London/UK	NR	NR	190	Prevalence	Convenience sampling	NR	15–28 May	Hospitals	Journal
Stubblefield et al., 2020 [34]	Tennessee/USA	65.5	33.7 (8.7)	249	Cross-sectional	Convenience sampling	NR	3–13 April	Hospitals	Journal
Lackermair et al., 2020 [35]	Bavaria/Germany	83	37.9 (4)	151	Cross-sectional	Convenience sampling	63.7	2–6 April	Primary care facilities	Journal
Paderno et al., 2020 [36]	Brescia/Italy	65.5	41 (NR)	58	Cross-sectional	Convenience sampling	100	NR	Hospitals	Journal
Kassem et al., 2020 [37]	Cairo/Egypt	59.5	32.5 (5.2)	74	Cross-sectional	Convenience sampling	58.7	1–14 June	Hospitals	Journal
Olalla et al., 2020 [38]	Marbella/Spain	80	41.5 (8.9)	498	Prevalence	Convenience sampling	NR	15–25 April	Hospitals	Journal
Iversen et al., 2020 [39]	Capital Region of Denmark/Denmark	78.9	44.4 (12.6)	28,792	Cross-sectional	Convenience sampling	96.3	17–22 April	Hospitals	Journal
Hains et al., 2020 [40]	Indiana/USA	88	41.2 (9.2)	25	Prevalence	Convenience sampling	NR	25 March–11 April	Hospitals	Journal
Solodky et al., 2020 [41]	Lyon/France	NR	NR	244	Prevalence	Convenience sampling	NR	1 March–16 April	Hospitals	Journal
Behrens et al., 2020 [42]	Hannover, Germany	65	36.5 (11.3)	217	Prevalence	Convenience sampling	NR	23 March–17 April	Hospitals	Journal
Brandstetter et al., 2020 [43]	Regensburg/Germany	85.1	18–35 years, 35.8%; 36–50 years, 35.8%; 51–65 years, 28.4%	201	Prevalence	Convenience sampling	NR	NR	Hospitals	Journal
Fusco et al., 2020 [44]	Naples/Italy	49	42.1 (14.6)	115	Cross-sectional	Convenience sampling	95.8	23 March–2 April	Hospitals	Journal
Lahner et al., 2020 [45]	Rome/Italy	63.8	45.2 (11.1)	2115	Prevalence	Convenience sampling	NR	18 March–27 April	Hospitals	Journal
Schmidt et al., 2020 [46]	Hessisch Oldendorf/Germany	80	18–29 years, 14.3%; 30–49 years, 40%; 50–64 years, 44.2%; >64 years, 1.5%	406	Prevalence	Convenience sampling	77.3	20–30 April	Hospitals	Journal
Xu et al., 2020 [47]	Hubei Province, Chongqing, Guangzhou, Guangdong/China	75.2	37.1 (13.3)	4384	Cross-sectional	Convenience sampling	NR	9 March–10 April	Hospitals	Journal
Zhao et al., 2020 [48]	Beijing, Zhejiang province/China	NR	NR	276	Prevalence	Convenience sampling	NR	NR	Hospitals	Journal
Fernández-Rivas et al., 2020 [49]	Barcelona/Spain	76	43.8 (12.4)	7563	Prevalence	Convenience sampling	81.2	4–22 May	Primary care facilities and hospitals	Preprint service
Kammon et al., 2020 [50]	Alzintan/Libya	53	>40 years, 37.4%	77	Prevalence	Convenience sampling	NR	2 April–18 May	Hospitals	Preprint service
Xiong et al., 2020 [51]	Wuhan/China	88.5	31.2 (4.7)	797	Prevalence	Convenience sampling	NR	12 February–17 March	Hospitals	Preprint service
Galán et al., 2020 [52]	Madrid/Spain	73.9	43.8 (11.1)	2590	Cross-sectional	Convenience sampling	90.5	14–27 April	Hospitals	Preprint service
Nakamura et al., 2020 [53]	Iwate/Japan	73.6	40 (11)	1000	Prevalence	Convenience sampling	76.8	18–29 May	Hospitals	Preprint service
Psichogiou et al., 2020 [54]	Athens/Greece	69.7	46.4 (10.3)	1495	Cross-sectional	Convenience sampling	77	13 April–15 May	Hospitals	Preprint service
Chibwana et al., 2020 [55]	Blantyre/Malawi	53	31.4 (7.3)	500	Prevalence	Convenience sampling	NR	22 May–19 June	Hospitals	Preprint service
Tosato et al., 2020 [56]	Padova/Italy	88	47 (10)	133	Prevalence	Convenience sampling	NR	NR	Hospitals	Preprint service
Paradiso et al., 2020 [57]	Bari/Italy	60.6	47.9 (8.6)	606	Prevalence	Convenience sampling	NR	26 March–17 April	Hospitals	Preprint service
Fujita et al., 2020 [58]	Kyoto/Japan	64.1	20–29 years, 32.6%; 30–39 years, 31.5%; 40–49 years, 22.8%; >49 years, 13%	92	Cross-sectional	Convenience sampling	NR	10–20 April	Hospitals	Preprint service
Sikora et al., 2020 [59]	Reading, Newport, Liverpool, Bedlington/UK	50.3	43 (NR)	161	Prevalence	Convenience sampling	NR	14–24 April	Cancer centers	Preprint service
Rudberg et al., 2020 [60]	Stockholm/Sweden	85	44 (12)	410	Cross-sectional	Convenience sampling	100	14 April–8 May	Hospitals	Preprint service
Shields et al., 2020 [61]	Birmingham/UK	75.2	40.9 (15.6)	516	Cross-sectional	Convenience sampling	93.1	25 April	Hospitals	Preprint service
Takita et al., 2020 [62]	Tokyo/Japan	35	20–29 years, 0%; 30–39 years, 9%; 40–49 years, 36%; 50–59 years, 16%; 60–69, 31%; 70–80 years, 7%	55	Prevalence	Convenience sampling	NR	21–28 April	Primary care facilities	Preprint service

NR, not reported; SD, standard deviation.

For females.

For males.

The majority of studies were conducted in Europe (N=31), followed by North America (N=9), Asia (N=6) and Africa (N=3). In particular, nine studies were conducted in the USA [14,15,24,26,27,29,32,34,40], eight studies in Italy [23,28,30,36,44,45,56,57], seven studies in the UK [16,17,20,25,33,59,61], five studies in Germany [21,35,42,43,46], four studies in Spain [31,38,49,52], three studies in Japan [53,58,62], three studies in Belgium [18,19,22] and three studies in China [47,48,51]. Twenty-nine studies did not report the response rate [[15], [16], [17],20,24,[26], [27], [28], [29], [30],[32], [33], [34],38,[40], [41], [42], [43],45,47,48,50,51,[55], [56], [57], [58], [59],62], nine studies did not report the ages of HCWs [17,18,21,23,32,33,36,41,48], eight studies did not report the sex distribution of HCWs [16,18,21,25,32,33,41,48] and five studies did not report the dates of data collection [17,36,43,48,56]. The percentage of females ranged from 35% [62] to 88.5% [51], and was higher compared with the percentage of males in 41 studies; in three studies, the percentage of males was higher than the percentage of females. The mean age of HCWs ranged from 31.2 [51] years to 47.9 years [57], while sample size ranged from 25 [40] to 40,329 HCWs [14]. Regarding study design, 26 cross-sectional studies [14,15,18,19,21,[23], [24], [25],[27], [28], [29], [30], [31], [32],[34], [35], [36], [37],39,44,47,52,54,58,60,61], 20 prevalence studies [17,26,33,38,[40], [41], [42], [43],45,46,[48], [49], [50], [51],53,[55], [56], [57],59,62] and three cohort studies [16,20,22] were included in this review. All studies except one [31] used a convenience sample, and the response rate ranged from 47.7% [19] to 100% [36,60]. Forty-two studies were conducted in hospitals [[15], [16], [17], [18], [19], [20], [21], [22], [23], [24],26,27,[29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48],[50], [62]], four studies in primary care facilities and hospitals [14,25,28,49], two studies in primary care facilities [35,62] and one study in a cancer centre [59]. Thirty-five studies were published in journals [[14], [15], [16], [17], [18], [19], [20], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31], [32], [33], [34], [35], [36], [37], [38], [39], [40], [41], [42], [43], [44], [45], [46], [47], [48]], and 14 studies were published in preprint services [[49], [50], [51], [52], [53], [54], [55], [56], [57], [58], [59], [60], [61], [62]].

Validity assessment (sensitivity and specificity) for antibody tests used in the included studies according to the manufacturers' data are presented in Table S2 (see online supplementary material). Sensitivity ranged from 50% to 100%, and specificity ranged from 80.5% to 100%.

Quality assessment

Quality assessments of prevalence studies, cross-sectional studies and cohort studies are shown in Tables S3, S4 and S5, respectively (see online supplementary material). Quality was moderate in 37 studies, good in 10 studies and poor in two studies. Regarding prevalence studies, 16 were at moderate risk of bias, three were at low risk and one was at high risk. Moreover, 20 cross-sectional studies were at moderate risk of bias, five were at low risk and one was at high risk. Two cohort studies were at low risk of bias and one was at moderate risk.

Meta-analysis of the seroprevalence

A random effects model was applied to estimate pooled prevalence as heterogeneity between results was very high (I 2=99.34, P-value for Hedges Q statistic <0.001). The estimated overall seroprevalence of SARS-CoV-2 antibodies among HCWs was 8.7% (95% CI 6.7–10.9%) (Figure 2 ). Seroprevalence among studies ranged from 0% to 45.3%.

Figure 2

Forest plot of the seroprevalence of severe acute respiratory syndrome coronavirus-2 antibodies with corresponding 95% confidence intervals. The size of the black boxes is positively proportional to the weight assigned to studies, and horizontal lines represent the 95% confidence intervals according to random effects analysis.

Subgroup and meta-regression analysis

According to subgroup analysis, seroprevalence of SARS-CoV-2 antibodies was higher for studies of poor quality (11.6%, 95% CI 0.7–32.7%) compared with studies of moderate quality (8.8%, 95% CI 6.0–12%) and good quality (7.9%, 95% CI 4.1–12.8%). Moreover, seroprevalence was higher for studies that had been published in journals (9%, 95% CI 6.7–11.6%) compared with preprint services (7.7%, 95% CI 3.4–13.4%). Seroprevalence was higher in studies conducted in North America (12.7%, 95% CI 8.6–17.5%) compared with those conducted in Europe (8.5%, 95% CI 5.8–11.6%), Africa (8.2%, 95% CI 0.8–22.3%) and Asia (4%, 95% CI 1.8–7.1%). Meta-regression showed that increased sensitivity of antibody tests was associated with increased seroprevalence (coefficient = 0.004, 95% CI 0.0001–0.009; P=0.038). Moreover, seroprevalence was independent of sample size (P=0.65) and specificity (P=0.20).

Sensitivity analysis

Leave-one-out sensitivity analysis showed that no single study had a disproportionate effect on overall seroprevalence, which varied between 8.2% (95% CI 6.2–10.3%) with Hoolihan et al. [16] excluded and 9.0% (95% CI 6.9–11.2%) with Nakamura et al. [53] excluded (Figure S1, see online supplementary material).

Publication bias

Egger's test (P=0.0001) and the asymmetric shape of the funnel plot (Figure S2, see online supplementary material) implied potential publication bias.

Factors associated with SARS-CoV-2 antibody positivity

Twenty-seven studies [14,15,18,19,[21], [22], [23], [24], [25],[27], [28], [29], [30], [31], [32],[34], [35], [36], [37],39,44,47,52,54,58,60,61] investigated factors associated with SARS-CoV-2 antibody positivity, and 13 studies found associations [14,18,[23], [24], [25],28,30,32,34,36,39,47,60] (Table II ). Twenty-four studies [15,19,[21], [22], [23], [24], [25],[27], [28], [29], [30], [31], [32],[34], [35], [36], [37],44,47,52,54,58,60,61] used univariate analysis, and three studies [14,18,39] used multi-variate regression analysis.

Table II

Studies that investigated factors associated with severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) antibody positivity among healthcare workers

Reference	Factors investigated for SARS-CoV-2 antibody positivity	Factors associated with SARS-CoV-2 antibody positivity	Level of analysis
Moscola et al., 2020 [14]	Age, sex, race/ethnicity, borough/county of residence, type of occupation, previously diagnosed with COVID-19 by PCR test, self-reported high suspicion of SARS-CoV-2 exposure, primary location of clinical work, direct patient care, working in a COVID-19 unit	– Previous positive PCR test (OR=1.52, 95% CI 1.44–1.60; P<0.001) – Self-reported high suspicion of SARS-CoV-2 exposure (OR=1.23, 95% CI 1.18–1.23; P<0.001)	Multi-variate
Jeremias et al., 2020 [15]	Sex, ethnicity, type of occupation, primary location of clinical work	– None	Univariate
Steensels et al., 2020 [18]	Age, sex, involvement in clinical care, work during the lockdown phase, involvement in care of patients with COVID-19, exposure to COVID-19-positive coworkers, household contact with suspected or confirmed cases of COVID-19	– Household contact with suspected or confirmed cases of COVID-19 (OR=3.15, 95% CI 2.33–4.25; P<0.001)	Multi-variate
Blairon et al., 2020 [19]	Age, sex, type of occupation, level of exposure to patients with COVID-19	– None	Univariate
Korth et al., 2020 [21]	Age, sex, type of occupation, level of exposure to patients with COVID-19	– None	Univariate
Martin et al., 2020 [22]	Age, sex, type of occupation, level of exposure to patients with COVID-19	– None	Univariate
Amendola et al., 2020 [23]	Age, sex, type of occupation, primary location of clinical work	– Surgery department (OR=6.47, 95% CI 2.37–17.63; P=0.0003) and paediatric intensive care unit (OR=3.77, 95% CI 1.44–9.89; P=0.007)	Univariate
Self et al., 2020 [24]	Age, sex, race/ethnicity, chronic medical conditions, substance use, type of occupation, primary location of clinical work, participants' self-reported belief of previous SARS-CoV-2 infection, face covering for all clinical encounters, participants who reported a shortage of personal protective equipment	– Males (OR=1.39, 95% CI 1.03–1.86, P=0.029) – Other participants (Black, Asian, Hispanic etc.) compared with White participants (OR=2.30, 95% CI 1.71–3.10; P<0.001) – Participants' self-reported belief of previous SARS-CoV-2 infection (OR=5.67, 95% CI 4.21–7.63; P<0.001) – Did not use a face covering for all clinical encounters (P=0.012)a – Reported a shortage of personal protective equipment (P=0.009)a	Univariate
Grant et al., 2020 [25]	Prolonged direct contact with patients, working in a COVID-19 unit	– Prolonged direct contact with patients (OR=1.57, 95% CI 1.27–1.93; P<0.005) – Working in a COVID-19 unit (OR=1.67, 95% CI 1.40–1.99; P<0.001)	Univariate
Hunter et al., 2020 [27]	Age, sex, type of occupation, level of exposure to patients with COVID-19	– None	Univariate
Plebani et al., 2020 [28]	Age, sex, type of occupation	– Aged ≥40 years (OR=1.36, 95% CI 1.09–1.60; P=0.006) – Healthcare assistants (OR=1.39, 95% CI 1.05–1.84; P=0.02)	Univariate
Mansour et al., 2020 [29]	Age, sex	– None	Univariate
Sotgiu et al., 2020 [30]	Age, sex, type of occupation, contact with patients with COVID-19	– Males (OR=3.21, 95% CI 1.43–7.19; P=0.003)	Univariate
Garcia-Basteiro et al., 2020 [31]	Age, sex, type of occupation, daily contact with patients, working in a COVID-19 unit, close contact with confirmed or suspected cases of COVID-19, previously diagnosed with COVID-19 by PCR test, comorbidity, household size, flu vaccine	– None	Univariate
Sydney et al., 2020 [32]	Age, sex, race/ethnicity, primary location of clinical work	– African American participants compared with other ethnic groups (P<0.05)a	Univariate
Stubblefield et al., 2020 [34]	Age, sex, race/ethnicity, comorbidity, smoking, primary location of clinical work, type of occupation, previously diagnosed with COVID-19 by PCR test, face covering for all clinical encounters, participants' self-reported belief of previous SARS-CoV-2 infection	– Participants' self-reported belief of previous SARS-CoV-2 infection (P=0.02)a – Previous positive PCR test (P<0.001)a	Univariate
Lackermair et al., 2020 [35]	Age, sex, contact with patients with COVID-19, temporary residence in a high-risk SARS-CoV-2 region	– None	Univariate
Paderno et al., 2020 [36]	Age, sex, type of occupation, hospital and household contacts without personal protective equipment	– Household contacts without personal protective equipment (P=0.008)a	Univariate
Kassem et al., 2020 [37]	Age, sex, type of occupation	– None	Univariate
Iversen et al., 2020 [39]	Age, sex, comorbidity, smoking, alcohol consumption, type of occupation, working in a COVID-19 unit, patient contact	– Males (OR=1.49, 95% CI 1.31–1.68; P<0.001) – Aged <30 years (OR=1.40, 95% CI 1.22–1.60; P<0.001) – Working in a COVID-19 unit (OR=1.65, 95% CI 1.34–2.03; P<0.001) – Front-line healthcare workers (OR=1.38, 95% CI 1.22–1.56; P<0.001) – Regular patient contact (OR=1.22, 95% CI 1.03–1.45; P=0.02)	Multi-variate
Fusco et al., 2020 [44]	Age, sex, type of occupation, primary location of clinical work, working in a COVID-19 unit, participation in training event on personal protective equipment	– None	Univariate
Xu et al., 2020 [47]	Age, sex, type of occupation	– Aged ≥65 years (P<0.001)a	Univariate
Galán et al., 2020 [52]	Age, sex, comorbidity, type of occupation, primary location of clinical work	– None	Univariate
Psichogiou et al., 2020 [54]	Sex, country of birth, education, household size, front-line or second-line HCWs, personal protective equipment	– None	Univariate
Fujita et al., 2020 [58]	Age, sex, type of occupation, primary location of clinical work, history of seasonal common cold symptoms, history of regular contact with children, history of exposure to a viral infection	– None	Univariate
Rudberg et al., 2020 [60]	Age, sex, type of occupation, patient-related work, contact with patients with COVID-19	– Patient-related work (OR=2.9, 95% CI 1.9–4.5; P<0.001) – Contact with patients with COVID-19 (OR=1.4, 95% CI 1.1–1.8; P=0.003) – Assistant nurses (OR=3.8, 95% CI 2.3–6.1; P<0.001)	Univariate
Shields et al., 2020 [61]	Age, sex, ethnicity	– None	Univariate

COVID-19, coronavirus disease 2019; PCR, polymerase chain reaction; CI, confidence interval; OR, odds ratio.

Data not available to calculate OR and CI.

Three studies [24,30,39] found that SARS-CoV-2 antibodies were more frequently detectable in males, with odds ratios (OR) ranging from 1.39 to 3.21. Results regarding age were controversial as SARS-CoV-2 antibody positivity was associated with HCWs aged <30 years (OR=1.40, 95% CI 1.22–1.60) [39], HCWs aged ≥40 years (OR=1.36, 95% CI 1.09–1.60) [28] and HCWs aged ≥65 years (P<0.001) [47]. Significantly higher percentages of SARS-CoV-2 antibodies were found among African American HCWs (P<0.05) [32] and Black, Asian and Hispanic HCWs compared with White HCWs (OR=2.30, 95% CI 1.71–3.10; P<0.001) [24].

Three studies [25,39,60] found a significantly higher probability of a positive SARS-CoV-2 antibody test in HCWs working in a COVID-19 unit, with ORs ranging from 1.4 to 1.67. Similar results were found for HCWs with patient-related work (OR 1.22–2.9) [25,39,60] and front-line HCWs (OR=1.38, 95% CI 1.22–1.56) [39]. Moreover, Self et al. [24] found that HCWs working in a surgery department (OR=6.47, 95% CI 2.37–17.63) and a paediatric intensive care unit (OR=3.77, 95% CI 1.44–9.89; P=0.007) had a significantly higher percentage of SARS-CoV-2 antibodies. Two studies [28,60] found that SARS-CoV-2 antibody positivity was higher among healthcare assistants (OR=1.39, 95% CI 1.05–1.84; OR=3.8, 95% CI 2.3–6.1). Self et al. [24] found that not using a face covering for all clinical encounters (P=0.012) and a shortage of personal protective equipment (P=0.009) increased the probability of a positive SARS-CoV-2 antibody test in HCWs.

Three studies [14,24,34] found an association between a HCW's self-reported belief of previous SARS-CoV-2 infection (OR 1.23–5.67) and SARS-CoV-2 antibody positivity. Similar results were found for HCWs with a previous positive polymerase chain reaction (PCR) test (OR=1.52, 95% CI 1.44–1.60 in one study [14] and P<0.001 in another study [34]). Also, two studies [18,36] found that household contact with suspected or confirmed cases of COVID-19 increased the probability of a positive SARS-CoV-2 antibody test in HCWs (OR=3.15, 95% CI 2.33–4.25 in one study [18] and P=0.008 in another study [36]).

Discussion

To the authors' knowledge, this is the first systematic review and meta-analysis to estimate the overall seroprevalence of SARS-CoV-2 antibodies among HCWs in screening settings. Overall seroprevalence was 8.7%, ranging from 0% to 45.3% between studies. Population-based and community-based studies in the USA showed high variability in the seroprevalence of SARS-CoV-2 antibodies, ranging from 1.1% to 14.4% [[63], [64], [65], [66], [67]]. Similar studies in Europe [[68], [69], [70]] and China [71] found very different seroprevalence in the general population, ranging from 0.23% to 10.9%. These differences in seroprevalence among studies may be attributable to several reasons, such as different study populations, different antibody tests with variation in sensitivity and specificity, different study designs, different lockdown and quarantine measures, and different dates of data collection. Moreover, according to the subgroup analysis, the seroprevalence of SARS-CoV-2 antibodies was higher for studies of poor quality (11.6%) compared with those with moderate quality (8.8%) and good quality (7.9%), indicating that a difference in study quality could also represent a significant reason for a difference in seroprevalence.

Subgroup analysis identified that seroprevalence was higher in studies conducted in North America (12.7%) compared with those conducted in Europe (8.5%), Africa (8.2%) and Asia (4%). This finding is in accordance with a meta-analysis [3] which found that the overall proportion of HCWs who were SARS-CoV-2 positive among all patients with COVID-19 was lower in China (4.2%) than in the USA (17.8%) and Europe (9%). This might be explained by good adherence to infection prevention and control measures and appropriate use of personal protective equipment among HCWs in China. Also, the USA and Europe seemed to be unprepared to handle the surge of patients which led to severe shortages in personal protective equipment, and the USA and most countries in Europe (with significant exceptions such as Germany and Greece) took action too late [72]. For example, according to reports in the UK and Italy, HCWs experienced extreme situations during the COVID-19 pandemic, wearing paper face masks and plastic aprons instead of appropriate masks, visors and gowns [73,74]. In this meta-analysis, seroprevalence in studies in the UK (N=7) and Italy (N=8) was higher (10.3%) compared with overall seroprevalence (8.4%), and seroprevalence in studies in Germany (N=5) and Greece (N=2) was lower (2.2%) than overall seroprevalence. On the other hand, China controlled the severe acute respiratory syndrome (SARS) epidemic that broke out in 2003 rapidly and efficiently [75,76], and immediately adopted the lessons learned from the SARS epidemic in the case of the COVID-19 pandemic by applying effective measures (e.g. early case identification and isolation; active large-scale surveillance of individuals including smartphone application, tracing and quarantining of COVID-19 contacts; temperature screening in public places; physical distancing; traveller screening; and street camera system for identification of individuals without a mask or showing symptoms) [71,77,78]. Moreover, some hospitals in China implemented a tactical training protocol for all aspects of COVID-19 that resulted in a very low infection rate among HCWs, including front-line HCWs in Wuhan [79].

Seropositivity was higher for HCWs performing patient-related work [25,39,60] and front-line HCWs [39]. Grant et al. [25] and Rudberg et al. [60] found that seropositivity of HCWs was much higher compared with the general population in London and Stockholm, respectively, indicating an occupational health risk among HCWs. Several studies emphasized the risk of occupational transmission of SARS-CoV-2 among HCWs, as HCWs are at the front-line response to COVID-19 and are more prone to viral transmission [73,[80], [81], [82], [83], [84]]. Increased HCW exposure to SARS-CoV-2 may be attributable mainly to patient-to-HCW transmission and HCW-to-HCW transmission due to shortages of personal protective equipment, poor adherence to infection prevention and control measures, and space constraints in hospitals. Additionally, SARS-CoV-2 antibody positivity was higher among healthcare assistants [28,60], which supports patient-related transmission of SARS-CoV-2 to HCWs as these HCWs are involved in most near-patient work.

In this systematic review, seroprevalence was higher among HCWs working in COVID-19 units [25,39,60]. It is clear that HCWs with contact with patients with COVID-19 represent a high-risk group for SARS-CoV-2 infection, and this was particularly true during the first months of the COVID-19 pandemic where knowledge, control measures and personal protective equipment were limited. Also, Self et al. [24] found that not using a face covering for all clinical encounters and shortages of personal protective equipment increase the probability of a positive SARS-CoV-2 antibody test in HCWs. Thus, personal protective equipment supplies for HCWs in hospitals are a necessary tool against COVID-19, and universal masking is of utmost importance as it decreases the rate of SARS-CoV-2 infection among HCWs [85]. Optimal personal protective equipment is still unknown, but rigorous application of personal protective equipment measures and absolute adherence to all infection prevention and control measures are crucial to reduce nosocomial transmission of SARS-CoV-2 [[86], [87], [88], [89]]. Interestingly, Grant et al. [25] found that seropositivity was lower among HCWs in ICUs. Several reasons could explain this finding, such as the enhanced personal protective equipment for HCWs in ICUs, the fact that intubated patients are ventilated on a closed circuit, and the fact that patients with COVID-19 who require ICU admission are often admitted around day 10 of the natural history of their illness [90], by which point the viral load has usually decreased [91].

According to this review, household contact with a suspected or confirmed case of COVID-19 is associated with a positive SARS-CoV-2 antibody test in HCWs [18,36]. Also, a HCW's self-reported belief of previous SARS-CoV-2 infection was found to be associated with SARS-CoV-2 antibody positivity [14,24,34]. HCWs are exposed to SARS-CoV-2 not only in clinical settings but also at home, in social situations, during joint meals and in office spaces with friends or colleagues. In fact, as community transmission increases, the risk of SARS-CoV-2 exposure for HCWs is higher outside clinical settings through household contacts with cases of COVID-19 or interaction with others in areas with active, unmitigated transmission [[92], [93], [94]].

This review found that a previous positive PCR test increases the probability of a positive SARS-CoV-2 antibody test in HCWs [14,34]. SARS-CoV-2 antibody tests identify previous SARS-CoV-2 infection, but many issues remain controversial. For example, the sensitivity of these tests is low in the first week after symptom onset but increases ≥15 days after the onset of symptoms [5]. Also, the duration of antibody increases is unknown as data >35 days after symptom onset are very scarce [5]. Moreover, it is currently unknown whether antibody titres correlate with protective immunity against re-infection, and if antibody responses differ significantly in asymptomatic individuals and individuals with mild or severe COVID-19 [95,96]. Variation in the validity of commercial SARS-CoV-2 antibody tests, cross-reactivity between SARS-CoV-2 and other coronaviruses, and confusion regarding the possible role of SARS-CoV-2 antibodies as biomarkers of protective immunity or past infection increase uncertainty about the utility of SARS-CoV-2 antibody tests in clinical practice [5,97,98]. However, SARS-CoV-2 antibody tests are an additional tool against COVID-19, and their utility will be expanded as additional data provide a better understanding of the pros and cons of these tests. Also, universal screening for SARS-CoV-2 in high-risk units in hospitals could help to identify asymptomatic HCWs, resulting in self-isolation for the appropriate time [22].

This review found that seropositivity was higher among African American [32], Black, Asian and Hispanic HCWs compared with White HCWs [24]. This finding was confirmed by studies in general populations where a higher percentage of SARS-CoV-2 antibodies was found among Black [67,99] and Hispanic [67] HCWs. According to the preliminary analysis of Cook et al. [100], until 12th April 2020, 106 HCWs died in the UK with COVID-19 and 64.2% (N=68) of them were Black, Asian and Minority Ethnic communities. Moreover, Gould and Wilson [101] found that Black HCWs experienced higher SARS-CoV-2 seroprevalence than White HCWs. Several reasons may be given for this disparity, including work conditions, economic inequality, high population density, limited access to healthcare services, and health insurance. There is a need for strategies tailored to the culture of minority groups and organized by local minority leaders who can mobilize individuals to participate in screening tests, and tracing and quarantining of COVID-19 contacts to avoid additional SARS-CoV-2 infections in minority groups [102].

This review has several limitations. First, 14 of the 49 included studies were published in preprint services which do not apply a peer-review process. Nevertheless, study quality was assessed, and subgroup analysis was performed according to publication type (journal or preprint service) and study quality. Second, the heterogeneity between results was very high. A random effects model and subgroup analysis were applied to overcome this limitation. Third, seroprevalence reported in studies could be underestimated or overestimated depending on the antibody test used. Validity (sensitivity and specificity) of the antibody tests were not reported in most of the included studies. A meta-regression analysis using sensitivity and specificity of the antibody tests according to the manufacturers' data as the moderator variables was performed in order to overcome this limitation. Fourth, the time between exposure and antibody testing in studies is unknown, and seropositivity may have been missed if testing was too early. This systematic bias could result in underestimation of seroprevalence. Finally, data regarding factors associated with seroprevalence of SARS-CoV-2 antibodies were very scarce and it was not possible to perform a meta-analysis; as such, a qualitative approach was applied to assess these factors.

In conclusion, seroprevalence of SARS-CoV-2 antibodies among HCWs is high, indicating that HCWs represent a population at considerable risk of contracting COVID-19. Absolute adherence to infection prevention and control measures; sufficient and adequate personal protective equipment; and early recognition, identification and isolation of HCWs infected with SARS-CoV-2 are imperative to decrease the risk of SARS-CoV-2 infection. Moreover, seroprevalence studies among HCWs could add significant information regarding the level of exposure among HCWs, identification of high-risk departments in hospitals, measurement of the spread of COVID-19, success of interventions, and understanding of asymptomatic transmission of SARS-CoV-2 in clinical settings. Given the limitations of this review and the included studies, and that the COVID-19 pandemic is still evolving, there is a need for further high-quality studies.

Conflict of interest statement

None declared.

Funding sources