Study design and period

We conducted an observational, cross-sectional study using baseline data from the Wellness Hub study. The Wellness Hub study was designed to measure seroprevalence of SARS-CoV-2 antibodies among LTCH and RH staff, residents, and caregivers. It is a prospective cohort study conducted among LTCH/RH populations in Ontario, Canada recruited and enrolled into the cohort between May 17, 2021 and October 25, 2022. In this paper, we use the baseline data from LTCH/RH staff cohort, and restricted to participants residing in the Greater Toronto Area and its surrounding areas (e.g., Hamilton). We report our study findings using the Strengthening the Reporting of Observational Studies in Epidemiology checklist [11].

Sampling frame and sampling strategy

Wellness Hub study sites were recruited using both purposive and convenience sampling. Recruitment strategies included both active (e.g., tailored presentations to LTCH/RH networks) and passive (e.g., website/newsletter advertisements, word of mouth) strategies. We leveraged an extensive network of project partners to disseminate the study materials and facilitate recruitment (S1 Text). Within recruited homes, we recruited staff using convenience and snowball sampling. Recruitment strategies included: emails and phone calls to LTCH/RH managers and staff, advertisements via posters, newsletters, on-site recruitment days, social media, and study self-enrollment via website URLs and QR codes.

Setting

We recruited 72 LTCH/RH to participate in the Wellness Hub study. Homes were eligible to participate in Wellness Hub if they agreed to participate in individual-level recruitment, a home-level needs assessment interview to address challenges facing their home during the COVID-19 pandemic and if they identified a designated ‘point person’ to act as a primary contact for the study. Indigenous LTCH/RH and Indigenous participants were excluded from the study, given resource and experience gaps on our study team to appropriately engage members of Indigenous communities. Within all participating homes, four LTCH/RH populations (staff, household members of staff, residents, and essential care partners of residents [e.g., family, caregivers]) were eligible to participate in the study by providing DBS samples and completing an in-depth questionnaire; in this article, we present the findings from the LTCH/RH staff cohort.

Participants

Participants were eligible for inclusion if they: identified as a staff (including full-time, part-time paid staff and unpaid volunteers) working at a LTCH/RH at the time of consent and enrollment; were comfortable speaking/reading English or French; were at least 18 years old; were willing to complete a study demographic questionnaire and provide a DBS sample.

Data collection

Participants completed a demographic questionnaire (S2 Text) and provided a DBS sample. Demographic questionnaires included multiple choice and open-ended questions, and were completed using one of the following methods: 1) a hard-copy printed version of the questionnaire, 2) an online version of the questionnaire via REDCap platform or 3) a questionnaire completed through an in-person or virtual call led by a research team member. Participants provided DBS samples via one of the following methods: 1) collected in-person with the support of study staff or LTCH/RH staff, or 2) self-collected by participants following detailed instructions. Participants had the option of leaving their self-collected sample at a drop-off site within the study home or mailing the sample back to the study team.

DBS sample storage and testing

DBS were collected on Whatman 903 cards using a blade lancet for finger pricking. No identifying information was listed on the cards, except a unique participant ID. Collection of ≥2 filled spots on the card was requested. Cards were then air dried (according to the manufacturer’s instructions) and placed in individual sealable bags, along with a desiccant pouch. Cards were transferred to the Gingras laboratory at Sinai Health. Upon reception in that lab, the cards were placed at 4°C for short-term storage, or frozen (at -80°C) for long-term storage. Cards were then brought to room temperature before the packages were opened, and a portion of the spot (2 x 3 mm or a 1 x 3 mm diameter circle if the sample was limited as this was sufficient for one pass at the assays if the blood soaked through the Whatman paper) was then extracted with a semi-automated hole puncher directly into a 96-well plate.

Outcome measure

Our primary outcome was seropositivity, defined as having a SARS-CoV-2 infection. SARS-CoV-2 infection status was determined using DBS, based on anti-nucleocapsid total IgG antibodies by a validated chemiluminescent ELISA. In alignment with another seroprevalence study conducted in LTCH during this time period [12], we performed our analyses at a specificity threshold of 90% (a relative ratio to a synthetic standard of 0.341; 11.34 binding antibody units [BAU/mL], corresponding to 88% sensitivity in a vaccinated cohort, which provides the best predictive value (negative predictive value * positive predictive value) when seroprevalence is 40% or greater [13]. As noted in several studies, many individuals who have breakthrough infection after vaccination either do not seroconvert for nucleocapsid or had lower antibody levels to it [14]. Thus, we used the relative ratio nucleocapsid threshold of 0.341 (BAU of 11.34 BAU/mL) to reduce false negatives. DBS passing the threshold of seropositivity for anti-nucleocapsid IgG antibodies were identified as having a natural COVID-19 infection and were marked as ‘seropositive for natural infection’; DBS that did not pass the threshold were considered ‘seronegative for natural infection’ (see S3 Text for more detail).

Conceptual framework and variables

Our conceptual framework in selecting variables to examine patterns of seroprevalence was informed by literature on epidemiological and socioeconomic factors associated heightened risks of SARS-CoV-2 acquisition in Canada; and literature more specific to exposure risks in occupational settings [9,15–17]. We classified variables of interest according to four domains conceptualized based on potential contexts that shape exposure risks for SARS-CoV-2 among LTCH/RH staff: sociodemographic, household characteristics, neighbourhood characteristics, and occupational characteristics. Sociodemographic characteristics included: age [18–20], gender [21], educational attainment [22]; and race as a proxy for individual, institutional, and structural racism [9,22–24]. Household characteristics included: household income [8], household density [9,22], and housing type [22,25,26] (S1 Table). Neighbourhood-level characteristics refer to the residential neighbourhood of the participant (not the LTCH/RH) and include: socioeconomic indicators of neighbourhood-level median household income (measured as tertile) and proportion essential workers (measured as tertile). The tertiles for neighbourhood-level median household income and proportion essential workers were based on data at the level of the dissemination area from the 2016 Census from Statistics Canada [27], with details included in S4 Text. We also included a neighborhood-level measure of community transmission, defined as “hotspot versus non-hotspot” using the neighbourhood-level cumulative SARS-CoV-2 diagnoses. We used the first three digits of participant postal code of residence (called the forward sortation area) to create the neighbourhood-level hotspot variable as a dichotomous measure. A hotspot referred to neighbourhoods (defined using the forward sortation area) in Ontario that comprised 20% of the population with the highest per-capita incidence of diagnosed SARS-CoV-2 cases between January 23, 2020, and January 16, 2021 [28,29]. We defined the remaining geographical areas as non-hotspots. S4 Text details the derivation of neighbourhood-level variables including hotspots, dissemination area characteristics, and forward sortation area levels.

Finally, our last domain centered on occupational-related characteristics, and included: occupational role, employment status, type of home (LTCH or RH), transportation used to attend the home for work, and current paid sick leave. We classified occupational role using the self-reported answer to “what best describes your job title at this LTCH/RH” and regrouped open text responses based on best fit to the categories. We classified employment status using the self-reported answer to: “what is your employment status in this LTCH/RH” with open text responses coded into existing categories. We grouped modes of transportation to the home based on anticipated magnitude of contacts during occupation-related movement from/to households from workplaces: few contacts (i.e., working from household or remote work) to greater number of contacts (i.e., use of rideshares or carpooling; use of public transportation). The variable of current paid sick leave at the home of employment was generated using self-reported answer to: “do you have paid sick leave at this facility”. If individuals answered “no”, “do not know/ unsure”, or if responses were missing, we categorized their response as “no” for current paid sick leave.

Analysis plan

Survey data (S2 Text) with responses fewer than n=5 were collapsed in order to protect participant anonymity. Open-ended responses were independently themed by a Research Scientist (CF) and Research Coordinator (NP) to facilitate analysis. Continuous data were reported using means, standard deviations (SD) and medians, and inter-quartile range (IQR). We conducted descriptive analyses and reported seropositivity (i.e., seroprevalence) by the covariates of interest. We then generated age-adjusted prevalence ratios (PR) with 95% confidence intervals (95% CI) for each covariate using normal approximation (Wald), and using a priori selection of the reference category as appropriate. For age-adjustment, we applied logistic regression with the log link function for the PR, where the outcome was seroprevalence and included the age as a continuous variable. Data cleaning, analyses, and visualization were conducted in R 4.3.0. We used the “stats” and “crosstable” packages to conduct our analysis (Table 1) and the “ggplot2” package to generate Fig 1.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Demographic Characteristics.

https://doi.org/10.1371/journal.pgph.0004294.t001

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Age-adjusted prevalence ratios.

https://doi.org/10.1371/journal.pgph.0004294.g001

Research ethics

This study was approved by the Unity Health Toronto Research Ethics Board (REB 20-347). Participants provided formal consent between May 6 2021 and June 26 2023. All participants or their shared decision maker provided either written (hard copy) or online (captured via REDCap) consent. There were no minors involved in the study (participant age range 18-79 years).

Participant characteristics

Of the n=797 LTCH/RH staff who participated in the Wellness Hub study and completed a demographic questionnaire, n=603 provided a DBS and were included in this analysis. Among the 603 participants, 235 (39%) were enrolled in period 1 and 368 (61%) in period 2. Nearly all (97.7%) participants reported receipt of at least 2 doses or equivalent of SARS-CoV-2 vaccination (37.1% of participants in period 1 and 60.8% in period 2). Table 1 summarizes participant characteristics under the conceptual framework. Eighty-three percent of the participants identified as women, 12% as men, and 4% chose not to specify. Nearly half (44%) of participants identified as White, 24% as East or Southeast Asian, 10% as Black, and 16% identified as a member of a non-Black, non-East/Southeast Asian racialized community (Table 1). The crude SARS-CoV-2 seroprevalence in our sample was 24% in period 1 and 44% in period 2 (Table 1, S2 Table).

Age-adjusted pattern of seroprevalence

Seroprevalence was nearly 2-fold higher in period 2 compared to period 1 [PR 1.87; CI 1.44-2.43]. Seropositivity was nearly 2-fold higher among racialized participants compared with White participants: the prevalence ratio for seropositivity among Black participants was 1.78 [CI 1.28-2.48], 1.55 [CI 1.18-2.04] among East/Southeast Asian participants, and 1.42 [CI 1.03-1.96] among other racialized participants. Although there were few men in our study, there was a trend towards a 1.5-fold higher seroprevalence among women [PR 1.48, CI 0.97-2.25] and similar seroprevalence by levels educational attainment (S2 Table).

We did not observe a pattern across household characteristics. There was a meaningful trend towards higher seropositivity among participants residing in hotspots compared with non-hotspots [PR 1.25, CI 1.00-1.57]. Prevalence ratios were lower for participants living in higher income neighbourhoods [PR 0.72, CI 0.58 – 0.98].

We did not observe variability in seroprevalence across occupational characteristics, with the exception of paid sick leave. Seropositvity was higher among participants with paid sick via the home at the time of the survey [PR 0.58, CI 0.45-0.75].

Limitations

Our study is limited to staff who worked in homes that participated in the Wellness Hub study, and geographic scope was restricted to the Greater Toronto Area (a largely urban setting, with a larger proportion of racialized population, and where SARS-CoV-2 transmission during the study period was also higher than in more rural settings). As such, our findings may not be generalizable to less urban settings. We were also limited by grouping our examination of associations across wide ranges of time because participants were recruited over 17 months. Another limitation is measurement bias in our outcome of interest, as we did not have data on polymerase-chain-reaction SARS-CoV-2 testing data to confirm and validate the results from the DBS samples. We may have underestimated past infections because some vaccinated individuals may not have produced sufficient anti-nucleocapsid antibodies to pass the positivity threshold or their antibody levels may have waned below the level of detection at the timing of testing. Finally, the results of our descriptive analyses should be interpreted as observations around variability in patterns of seroprevalence; future study is needed to conduct risk factor and potentially causal analyses.