key: cord-0805650-5z250q5l authors: ROEDERER, T.; MOLLO, B.; VINCENT, C.; NIKOLAY, B.; LLOSA, A.; NESBITT, R.; VANHOMWEGEN, J.; ROSE, T.; ANNA, F.; TORRE, C.; FOURREY, E.; GOYARD, S.; JANIN, Y.; CHARNEAU, P.; VRATSKIKH, O.; COURY, A.; VANEL, S.; MENDIHARAT, P.; PORTEN, K.; HENNEQUIN, W.; MILLS, C.; LUQUERO, F. title: High seroprevalence of SARS-CoV-2 antibodies among people living in precarious situations in Ile de France date: 2020-10-09 journal: nan DOI: 10.1101/2020.10.07.20207795 sha: 04eab14485a2b74ad18ac382c38f15bb63bcaed3 doc_id: 805650 cord_uid: 5z250q5l Background A nationwide lockdown was implemented in France on 17 March 2020 to control the COVID-19 pandemic. People living in precarious conditions were relocated by the authorities to emergency shelters, hotels and large venues. Medecins sans Frontieres (MSF) then intervened to provide medical care in several of these locations in Paris and in Seine-Saint-Denis, one of its suburbs, between March and June 2020. A seroprevalence survey was conducted to assess the level of exposure to COVID-19 among the population living in the sites. To our knowledge, this is the first assessment of the impact of the pandemic on populations living in insecure conditions in Europe. Methods We conducted a cross-sectional seroprevalence study in the food distribution sites, emergency shelters and workers residences supported by MSF in Paris and Seine-Saint-Denis, to determine the extent of COVID-19 exposure as determined by SARS-CoV2 antibody seropositivity. The detection of SARS-COV2 antibodies in serum was performed at the Institut Pasteur of Paris using two LuLISA (Luciferase-Linked Immunosorbent Assay) assays and a Pseudo Neutralization Test. A questionnaire covering sociodemographic characteristics, living conditions, adherence to sanitary recommendations and symptom manifestations was also completed. We describe here the seroprevalence site by site and identify the risk factors for seropositivity using a multivariable logistic regression model with site random effects. We also investigated associations between seropositivity and symptoms eventually reported. Findings Overall, 426/818 individuals tested positive in the 14 sites investigated. Seroprevalence varied significantly with the type of site (chi2 p<0.001). It was highest at 88.7% (95%CI 81.8-93.2) among individuals living in workers residences, followed by 50.5% (95%CI 46.3-54.7) in emergency shelters and 27.8 % (95%CI 20.8-35.7) among individuals recruited from the food distribution sites. Seroprevalence also varied significantly between sites of the same type. Among other risk factors, the odds for seropositivity were higher among individuals living in crowded sites (medium: adj. OR 2.7, 95%CI 1.5-5.1, p=0.001; high: adj. OR 3.4, 95%CI 1.7-6.9, p<0.001) compared with individuals from low crowding sites and among those who reported transit accommodation in a gymnasium before the lockdown (adj. OR 3.1, 95%CI 1.2-8.1, p=0.023). More than two-thirds of the seropositive individuals (68.3%; 95%CI 64.2-72.2) did not report any symptoms during the recall period. Interpretation The results demonstrate rather high exposure to SARS-COV-2 with important variations between study sites. Living in crowded conditions was identified as the most important explanatory factor for differences in levels of exposure. This study describes the key factors which determine the risk of exposure and illustrates the importance of identifying populations at high risk of exposure in order to orient and adapt prevention and control strategies to their specific needs. care in several of these locations in Paris and in Seine-Saint-Denis, one of its suburbs, between March and June 2020. A seroprevalence survey was conducted to assess the level of exposure to COVID-19 among the population living in the sites. To our knowledge, this is the first assessment of the impact of the pandemic on populations living in insecure conditions in Europe. We conducted a cross-sectional seroprevalence study in the food distribution sites, emergency shelters and workers residences supported by MSF in Paris and Seine-Saint-Denis, to determine the extent of COVID-19 exposure as determined by SARS-CoV2 antibody seropositivity. The detection of SARS-COV2 antibodies in serum was performed at the Institut Pasteur of Paris using two LuLISA (Luciferase-Linked Immunosorbent Assay) assays and a Pseudo Neutralization Test. A questionnaire covering sociodemographic characteristics, living conditions, adherence to sanitary recommendations and symptom manifestations was also completed. We describe here the seroprevalence site by site and identify the risk factors for seropositivity using a multivariable logistic regression model with site random effects. We also investigated associations between seropositivity and symptoms eventually reported. Seroprevalence also varied significantly between sites of the same type. Among other risk factors, the odds for seropositivity were higher among individuals living in crowded sites (medium: adj. OR 2.7, 95%CI 1.5-5.1, p=0.001; high: adj. OR 3.4, 95%CI 1.7-6.9, p<0.001) compared with individuals from low crowding sites and among those who reported transit accommodation in a gymnasium before the lockdown (adj. OR 3.1, 95%CI 1.2-8.1, p=0.023). More than two-thirds of the seropositive individuals (68.3%; 95%CI 64.2-72.2) did not report any symptoms during the recall period. The results demonstrate rather high exposure to SARS-COV-2 with important variations between study sites. Living in crowded conditions was identified as the most important explanatory factor for differences in levels of exposure. This study describes the key factors which determine the risk of exposure and illustrates the importance of identifying populations at high risk of exposure in order to orient and adapt prevention and control strategies to their specific needs. A novel coronavirus causing a severe respiratory syndrome, severe acute respiratory syndrome coronavirus 2 (SARS-COV-2), emerged at the end of 2019 in Hubei province, China and then spread worldwide (1) . Following this, Europe became the major hotspot of the global pandemic (1) and the first confirmed cases of the novel coronavirus disease were detected in France by 24 January 2020 (2) . As a response to an exponential transmission rate, with hospitalizations and deaths doubling every two to three days, a nationwide lockdown was implemented on the 17th of March 2020. Although the lockdown applied to the entire country, important differences were observed between regions in terms of number of confirmed cases reported and deaths. Although the final number of infections is yet to be established in France, a model-based study estimated a nationwide infection rate of 6% in France during the first wave, ranging from 1.5% infected in Nouvelle Aquitaine to 12% in Ile-de-France (IDF) (3) . A recent seroprevalence study reported similar estimates: 10% in IDF and 3.1% in Nouvelle Aquitaine. (4) Heterogeneity in the risk of exposure to COVID-19 likely exists among different subpopulations. Specific subgroups such as health care professionals or nursing home employees and residents have already been identified as groups at higher risk of exposure than the general population (5) (6) (7) (8) . It is likely that other populations have suffered higher exposure risks to COVID-19 because of their working (9, 10) or living conditions (i.e.: shared housing). In France, an estimated 900,000 people lack a permanent housing, with an estimated 250,000 people experiencing recurrent homelessness, including up to 50,000 in Ile-de-France. It has been estimated that at least 3,500 people are homeless on the streets of Paris and close to 7,000 in the Ile-de-France (11, 12) . People experiencing homelessness or otherwise living in precarious conditions may be particularly vulnerable to exposure to COVID-19. Shared housing, including shelters and encampments, and poor sanitary conditions are factors that can potentially expedite virus transmission. Many of the recommended COVID-19 prevention measures, such as social-distancing and self-isolation if symptomatic, may be challenging or not feasible for a population living under these circumstances. In addition, people experiencing homelessness include older adults who may have underlying medical conditions with higher risks of developing severe COVID-related illness (13, 14) . During the nationwide lockdown in France, known as "confinement", French authorities relocated people experiencing homelessness into emergency shelters, including hotels and large venues, such as gymnasiums. Non-governmental organisations and associations filled several resulting gaps in this relocation, including medical care. Between March and June 2020, the non-profit medical humanitarian organization Médecins sans Frontières (MSF) provided medical care and hygiene . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20207795 doi: medRxiv preprint promotion to populations living in workers' residences and emergency shelters, in Ile-de-France (IDF). Mobile clinic activities included clinical management of non-COVID cases, with hospital referrals when needed, COVID screening, and referrals to COVID isolation centres when necessary. The MSF mobile clinics performed some sample collection for PCR-based COVID assays but many individuals with mild symptoms compatible with COVID were never tested and the actual rate of infection among the homeless population in France remains unknown. Given the variable access to testing, and potentially the substantial proportion of asymptomatic cases who did not seek such tests, the number of confirmed cases of positive PCR assays does not reflect the true impact of the pandemic. Serological-based assays can accurately identify the number of people exposed to the virus (16) but although several serological/seroprevalence studies have been undertaken nation-wide, few have focused on populations experiencing homelessness (3, 14, 15) . Recommendations from the National Health Authority (Haute Autorité de Santé) have underscored the need for more serological surveys in specific populations, including among vulnerable groups, to evaluate the overall exposure to the virus, raise awareness and improve preparation for the rebound in COVID-19 cases (16), currently observed (17) . This survey was conducted to assess the level of exposure to COVID-19 among the population living in sites served by MSF in IDF. To our knowledge, this is the first survey to assess the pandemic's impact on vulnerable populations in Europe. We conducted a cross-sectional seroprevalence study in sites supported by MSF in Paris and Seine-Saint-Denis, two urban departments of Ile de France region, to determine the extent of COVID-19 infection among the population residing in or frequenting these sites, as determined by SARS-CoV2 antibody seropositivity. A target sample size of 791 individuals was based on the hypothesis that the seroprevalence of anti-SARS-COV2 antibodies among populations living in insecure conditions served by MSF would be two to three fold higher than the modelled estimate for the general population in IDF of 12% (3). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20207795 doi: medRxiv preprint otherwise. To increase participation of those initially selected to participate, sites were visited several times at different times of the day, including on weekends and evenings. In case of refusal or absence, the initially selected person was replaced by the next person in line or another adult sharing the room. After obtaining written informed consent from the participant, a questionnaire was completed faceto-face by a trained interviewer in the participant's language (for French, English, Arabic, Farsi, Spanish and Portuguese interviews were conducted in person and other languages using a translator by telephone). Responses were recorded electronically via a Kobo Collect form on a cell phone or tablet. The data were analysed using Stata V.15 software (StataCorp. 2017. College Station, TX) and R (R 3.6.2). Blood samples were collected on site from each participant, transported and processed within 24 hours for testing. The serology assays for the detection of SARS-COV2 antibodies were performed using the LuLISA (Luciferase-Linked Immunosorbent Assay) technology designed by Institut Pasteur, Paris (18) . The LuLISA uses the nucleoproteins (N) of the SARS-CoV-2 virus, or the Spike (S) protein as target antigens for the detection of specific IgG antibodies in human serum using a variable domain of the IgG single heavy chain from immunized alpaca, specific for human IgG constant domain, expressed as a tandem with a luciferase, NanoKAZ (19) . In the presence of luciferase substrate, hikarazine-Q108, the luminescence intensity (relative light unit/s) yielded by luciferase catalytic activity is related to the number of target-bound IgG (19, 20) . In addition to LuLISA (N), and LuLISA (S), a Pseudo Neutralization Test (PNT) was performed to confirm the presence of antibodies and to assess their potential to neutralise and protect against the SARS-COV-2 virus. The LuLISA technique can be used to assess the incidence of all the antibodies involved in a viral infection response (IgA, IgM and, as used here, IgG) and is considered highly sensitive. The incidence of IgG in patients with a COVID-19 positive PCR test sampled 15 days after onset of symptoms is 100% (21) . The specificity of the LuLISA varied between 97 and 100% in another recent evaluation (21) . is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20207795 doi: medRxiv preprint We describe the study participants using summary measures and estimated the seroprevalence and 95% confidence intervals (95%CI) by site, type of site, and characteristics of participants using the Clopper-Pearson method. A sensitivity analysis of seroprevalence estimates by type of sites, taking into account assumptions about diagnostic test performances, is presented in the supplementary material. To investigate seropositivity risk factors, we first performed univariable logistic regression analysis by type of site and for all sites combined. We subsequently constructed a multivariable logistic regression model with random intercepts for specific sites of recruitment to account for clustering of individuals. In the multivariable model selection, we included variables with the potential to epidemiologically explain the differences in seropositivity incidence. We first grouped variables into four categories: (i) sociodemographic characteristics (sex, age, working before lockdown, language barrier), (ii) frequency of leaving the place of residence during the lockdown (for work, cultural activities, use of public transport, time spent outside), (iii) crowding in the place of residence (number sharing room, number sharing sanitary facilities, number sharing kitchen, and number of contacts inside place of residence per day), and (iv) adhering to hygiene recommendations (hand washing, wearing masks, distancing, cleaning, and following recommendations in general). For sociodemographic characteristics and adherence to hygiene recommendations, we selected variables from each category that were most strongly associated with seropositivity. To investigate associations of seropositivity with population density in a residence, we created a cumulative crowding indicator, based on the sum of the levels of each of the 4 questions summarizing the crowding information for sharing (i) the bedroom, (ii) the shower, (iii) the kitchen, and (iv) the number of close contacts (>15 minutes per day closer than 1 meter). This made it possible to categorize low, medium, and high levels of crowding (see Supplementary material for more details on how the composite crowding indicator was calculated). We also constructed a similar score for the frequency of leaving place of residence, combining the answers to the questions 'Gone out to go to work', 'Gone out to go to a distribution/association/church site...' and 'Took public transport' into 3 categories: those who never went out, those who rarely/sometimes went out, and those who went out frequently (every day). Other variables investigated were: tobacco use, awareness of close contact with a COVID case(s), transit through a gymnasium at the beginning of the confinement (as an indicator for exposure to crowding before moving to a specific site), and the type of site. We then performed backwards variable . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20207795 doi: medRxiv preprint selection; variables with p<=0.05 were retained in the final model. A sensitivity analysis of the multivariable risk factor analysis is presented in the supplementary material. This protocol was approved by the MSF ethics review board on 18 June 2020, reference number 2044) and by the Comité de Protection des Personnes (CPP), Ile de France, Paris XI, approved 19 June 2020 (reference number 20050-62628). Between 23 June and 2 July 2020, we conducted a cross-sectional seroprevalence study among 829 people living in 14 facilities: 2 food distribution sites, 2 workers' residences and 10 emergency shelters. Depending on the site, between 1 in 10 and one third of participants randomly selected for participation were replaced with another participant from the same site, due to absence or refusal. After cleaning the data and consolidating the results, 818 people were included in all the sites and received their serological results for these SARS-COV-2 assays. Overall, 79.6% (95%CI 76.7-82.3) of study participants were male with an overall sex ratio of 3.9. There were no females present in the workers' residences. The mean age of participants was 39 years, with 49.0% (95%CI 45.5-52.5) of the population younger than 35 years; the population recruited from food distribution sites was older on average than the one living in shelters (mean age=31. 8 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. Seroprevalence also varied between facilities of the same type: 23-62% in emergency shelters, 18-35% in food distribution sites and 82-94% in workers' residences. The pseudo-neutralization test was positive for 303 out of the 818 individuals. This assay suggests that the seropositive population could therefore be protected against COVID-19 infection, at least at the time of the sample survey (supplementary materials, Table S1 ). Correlations and concordance between the three serology techniques were found to be strong (supplementary materials Figure S2 -3 and Table S1 ). In univariable analysis, the most strongly associated risk factors of seropositivity were those linked to crowded living conditions. The odds of seropositivity was 4.3 times (95%CI 2.2-8.4, p<0.001) higher among individuals sharing a room with more than five individuals compared to those not sharing a room; and 3.1 (95%CI 2.0-5.0, p<0.001) times higher among individuals sharing sanitary facilities with more than five individuals compared to those not sharing facilities ( Table 1 ). The odds of seropositivity increased with the level of crowding with an odds ratio (OR) of 3.6 (95%CI 2.0-6.3; p<0.001) for medium and an OR of 6.7 (95%CI 3.6-12.5; p<0.001) for high crowding compared to individuals with a low crowding composite indicator. The odds of seropositivity were higher among participants who reported transiting in a gymnasium during the lockdown compared to those who did not (OR 2.8; 95%CI 1.1-7.2; p=0.03). There was no significant difference in seropositivity between individuals who were aware of COVID cases among their close contacts and those who were not aware (OR 1.2; 95%CI 0.8-1.8, p=0.41). The main factor associated with a reduction of exposure to the virus was the frequency of leaving the place of is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. Moreover, the odds of seropositivity was lower among regular smokers compared to those who never smoked (adj. OR 0.4, 95%CI 0.3-0.7, p<0.001) and among female than male participants (adj. OR 0.5, 95%CI 0.4-0.8, p=0.01). More than two-thirds of seropositive individuals (68.3%; 95%CI 64.2-72.2) did not report any symptoms during the recall period since 1 March (implying a high proportion of asymptomatic infections). In contrast, 50.2% (95%CI 46.0-54.3) of individuals who did not report symptoms were found to be seropositive. While not statistically significant, there was a trend in association between the report of symptoms associated with COVID and seropositivity: people who were seropositive had a 30% higher odds of reporting symptoms than those who were not (OR 1.3, 95%CI 1.0-1.8, p=0.09). Among the investigated symptoms, six of the twelve reported symptoms were found significantly associated with seropositivity when comparing individuals with severe versus mild or no symptoms is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20207795 doi: medRxiv preprint participants with symptoms (10.5%) reported being tested for COVID previously, of which nine had a positive result and seven were admitted to COVID treatment centres. This study is the first in Europe to evaluate the exposure to SARS-COV-2 virus of populations who lived in precarious situations during the first months of the COVID-19 pandemic. The results show overall a high exposure to SARS-COV-2 with important variations between sites. This study also identifies overcrowding as the most important factor to explain the variability in exposure rather than reported adherence to the recommended preventive measures. The results further suggest living conditions within the place of residence as the most important risk factor related to exposure, as frequently leaving the place of residence was associated with a lower risk of being seropositive. Coherent with other studies (22,23) evaluating the spectrum of COVID-19 disease, there was a high proportion of asymptomatic or paucisymptomatic infections (up to 68%) in a population that is younger on average than the general European population. Combined with the high proportion of exposed individuals, the high proportion of asymptomatic cases puts into question the pertinence of epidemic surveillance strategies solely based on the identification of symptomatic cases and their contacts. Our study has several limitations. First, due to the cross-sectional study design, it is not possible to determine when participants became seropositive. In relation to the sampling strategy, the selection of the study sites was not random: the locations were determined by MSF's operational activities during the first wave of the pandemic in Ile-de-France and other considerations including security constraints and agreement of the sites to give the survey team access. Therefore, our results cannot be extrapolated to other populations living in similarly precarious situations in France or elsewhere. The selection of participants within the study sites could be potentially subject to bias despite the efforts made by the study team to obtain a representative sample; depending on the site, up to one third of those originally selected for inclusion were replaced by another participant. If those replacing the initially selected individuals had a higher risk of exposure, due to, for example, spending more t more time within the place of residence, this could have led to an overestimate of prevalence in the study population. Conversely, refusal to participate could have been higher among those who had previously tested positive, which would bias the seroprevalence estimate in the opposite direction. In addition to possible selection bias, information bias may have affected the measurement of other selfreported exposures,, including living conditions before and during confinement, COVID symptoms and onset period, and adherence to prevention measures. We also cannot exclude social desirability bias, particularly related to compliance with containment and prevention measures, that potentially underestimates exposure. Our efforts to mitigate information bias included the use of standardized is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. Congregate housing conditions, whether long term or as temporary emergency measures, carry risks during an infectious disease outbreak that should be weighed against the risks in remaining unsheltered. The extent to which these risks can be mitigated, using masks and hand hygiene in crowded living environments and or situations with high seroprevalence as seen in our study, remains uncertain. It may be that the good reported adherence to hygiene measures and frequently reported mask wearing contributed to the high proportion of asymptomatic infections seen in this study (26) . Lack of housing is a known risk factor for poor physical and mental health, and overcrowding for infectious diseases (27, 28) . This poses a complex dilemma for public health policy makers who aim to balance individual health risks and public health in the face of a global pandemic. While ours is the first study to identify overcrowding specifically as a risk factor for SARS-CoV-2 seropositivity among an already vulnerable population in IDF, , large household size has been identified as a risk factor for COVID-19 in a recent study in the general population of France (4) as well as in a study focusing on a cohort of pregnant women tested for SARS-CoV-2 RNA in New York City (29) . Many others have shown that the COVID-19 pandemic highlights existing socioeconomic and racial inequalities and inequities (30) (31) (32) . In conclusion, the results presented here highlight a high level of exposure to SARS-COV-2 virus, with substantial variability among our sampled population. This underscores the importance of identifying populations at high risk of exposure, understanding the factors that determine this risk to orient prevention and control strategies, and adaptation according to the needs of the affected population. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20207795 doi: medRxiv preprint Relocations of individuals living in insecure conditions, particularly those at highest risk of severe disease (i.e. elderly individuals and/ or individuals with certain comorbidities), should be implemented; this process should limit the number of people per room and ensure access to hygienic shared spaces. Based on a high seroprevalence, older average age and other risk factors, adapted strategies for the workers' residence should be prioritized. Finally, considering the lack of existing data globally, further epidemiological studies in populations experiencing similar vulnerable conditions should be considered, including qualitative studies, in order to properly protect at-risk populations. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . CC-BY-NC-ND 4.0 International license It is made available under a perpetuity. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20207795 doi: medRxiv preprint Figure S1 shows a density plot of observed relative light units per second (RLU/S) among study participants. Positivity for LULISA assays was defined by a cut-off of 10,291 RLU/s; however as the distribution of RLU/s is continuous (and not bimodal); misclassification of participants may occur at values close to the used cut-off. While the main peak in density occurs at RLU/s values lower than the cut-off, q second small peak in density occurs at higher values (50,000-75,000 RLU/s) potentially reflecting individuals with multiple exposures. Table S1 . PNT can be considered more specific than LuLISA; seroprevalence estimates obtained by PNT can therefore be considered as conservative estimates of seroprevalence. Misclassification based on LuLISA may be related to the transmission setting, where in low prevalence settings seroprevalence estimates by LuLISA may be more strongly overestimated than in high prevalence settings, which is reflected by a higher relative prevalence of LuLISA compared to PNT. This tendency is also visible in the relative prevalences of LuLISA compared to PNT by recruitment sites ( Figure S2 ). Overall, concordance between the techniques was pretty strong ( Figure S3 ). is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20207795 doi: medRxiv preprint is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20207795 doi: medRxiv preprint Here we present results of a sensitivity analysis of the multivariable analysis of risk factors for seropositivity. In addition to the main model which had the lowest Akaike Information Criterion (AIC) of 938-we additionally tested models without site random effects, without adjusting for the type of site, or with number of people sharing the room as alternative crowding indicator. Model 2: Including a random effect for sites but not adjusting for the type of site -crowding indicator: cumulative crowding indicator is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . https://doi.org/10.1101/2020. 10 is the author/funder, who has granted medRxiv a license to display the preprint in (which was not certified by peer review) preprint The copyright holder for this this version posted October 9, 2020. . https://doi.org/10.1101/2020.10.07.20207795 doi: medRxiv preprint Weekly Epidemiological Update Coronavirus disease 2019 (COVID-19) 21 Clinical and virological data of the first cases of COVID-19 in Europe: a case series Estimating the burden of SARS-CoV-2 in France Seroprevalence of SARS-CoV-2 among adults in three regions of France following the lockdown and associated risk factors: a multicohort study Risk of COVID-19 in health-care workers in Denmark: an observational cohort study Pandemic peak SARS-CoV-2 infection and seroconversion rates in London frontline health-care workers. The Lancet High seroprevalence of SARS-CoV-2 in elderly care employees in Sweden Presymptomatic Transmission of SARS-CoV-2 Amongst Residents and Staff at a Skilled Nursing Facility: Results of Real-Time PCR and Serologic Testing Meat plants-a new front line in the covid-19 pandemic COVID-19 Outbreak Among Employees at a Meat Processing Facility -South Dakota On comptait 3 552 SDF dans les rues de Paris à la fin de janvier. Le Monde.fr The health of homeless people in high-income countries: descriptive epidemiology, health consequences, and clinical and policy recommendations People experiencing homelessness: Their potential exposure to COVID-19 SARS-CoV-2 seroprevalence in COVID-19 hotspots Les projets de recherche (coronavirus) Connaître le statut immunitaire de la population pour guider la décision publique A highly sensitive bioluminescent method for measuring allergen-specific IgE in microliter samples Gram-scale synthesis of luciferins derived from coelenterazine and original insights into their bioluminescence properties Bioluminescence Profiling of NanoKAZ/NanoLuc Luciferase Using a Chemical Library of Coelenterazine Analogues. Chemistry A comparison of four serological assays for detecting anti-SARS-CoV-2 antibodies in human serum samples from different populations High Proportion of Asymptomatic SARS-CoV-2 Infections in 9 Long-Term Care Facilities COVID-19 among people experiencing homelessness in England: a modelling study Facial Masking for Covid-19 -Potential for "Variolation" as We Await a Vaccine Risk factors and risk factor cascades for communicable disease outbreaks in complex humanitarian emergencies: a qualitative systematic review Tuberculosis among the homeless Associations Between Built Environment, Neighborhood Socioeconomic Status, and SARS-CoV-2 Infection Among Pregnant Women Racial Disparity of Coronavirus Disease 2019 in African American Communities Racisme systémique et inégalités de santé, une urgence sanitaire et sociétale révélée par la Homeless people hospitalized with COVID-19 in Brussels We then created a composite indicator of the sum of levels of the categorical variable and created categories as low (values ≤5), medium (values 6-9), and high (values ≥10). For example, an individual sharing the room with no other person (level 1), the kitchen and sanitary facilities with 1 other person (each of level 2), and who reported on average 1 close contact per day We performed a sensitivity analysis of seroprevalence estimates taking uncertainty about diagnostic test performances into account as described in "Seroprevalence of anti-SARS-CoV-2 IgG antibodies in Geneva, Switzerland (SEROCoV-POP): a population-based study ) (i) 70 true positive samples out of 100 positive controls (Sens 70/100) and 70 true negative samples out of 100 negative controls (Spec 70/100) (ii) 100 true positive samples out of 100 positive controls (Sens 100/100) and 70 true negative samples out of 100 negative controls (Spec 70/100) (iii) 70 true positive samples out of 100 positive controls The estimated seroprevalence for emergency shelters ranged from 34.7-62.0% (compared to the main estimate of 50.5%); for food distribution sites it ranged from 16.2-38.5% (compared to the main estimate of 27.8%); and for workers' residences it ranged from 82 Authors want to thank all the participants to the survey, as well as the managers of the hostels and emergency shelters who made the survey possible.The authors also thank the team from the Mission France Project Team: Arielle Calmejane, Jean-François Véran, Marianne Viot, Cécile Arondel, Flora Boirin and Vanessa Lalouelle. We also thank the nurses for their hard work and availability: Ophélie Cahu, Aurélie Rawinski, Valérie Tacussel and Elodie Tarral. The interviews have been thoroughly conducted by Rachid Kaddour, Marine Labatut, Saeed Shairzad, Tara Singh, Juliette Dechaux, Margot Dupe, Diane Goulas, Prisla Mogango, Sarah Perretti and Ciamony Mohammadi.