key: cord-0845994-6l4hmmaw authors: Reyne, B.; Selinger, C.; Sofonea, M. T.; Miot, S.; Pisoni, A.; TUAILLON, E.; Bousquet, J. J.; Blain, H.; Alizon, S. title: Wearing masks and establishing COVID-19 areas reduces secondary attack risk in nursing homes date: 2020-12-02 journal: nan DOI: 10.1101/2020.11.27.20239913 sha: 19a23227b1a622fd901ac82c9d4fbf771933622a doc_id: 845994 cord_uid: 6l4hmmaw Background: COVID-19 is spreading rapidly in nursing homes (NHs). It is urgent to evaluate the effect of infection prevention and control (IPC) measures to reduce COVID spreading. Methods: We analysed COVID-19 outbreaks in 12 NH using rRT-PCR for SARS-CoV2. We estimated secondary attack risks (SARs) and identified cofactors associated with the proportion of infected residents. Results: The SAR was below 5%, suggesting a high efficiency of IPC measures. Mask-wearing or establishment of COVID-19 zones for infected residents were associated with lower SAR. Conclusions: Wearing masks and isolating potentially infected residents appear to limit SARS-CoV-2 spread in nursing homes. Introduction COVID-19 is spreading rapidly to nursing homes (NH) [13] and recommendations have been issued to reduce new cases in NH where a COVID-19 outbreak has been identied [4] . Guidelines from the European Geriatric Medicine Society (EuGMS) mention a variety of interventions based on testing, mask-wearing, or isolation of people with established or suspected infection and their contacts [5] . The relative impact of the dierent interventions on the magnitude of potential COVID-19 outbreaks in NH remains unclear and requires testing. In France, the COVID-19 epidemic wave is estimated to have started mid-January 2020 [6] . The rst COVID-19 cases in a French NH were detected in the Hérault department (France) on March 10. This rst outbreak triggered an immediate response from the Regional Health Authority (ARS Occitanie). Infection prevention and control (IPC) guidelines were implemented and included mask-wearing, establishment of "COVID-19 units" to isolate exposed or infected residents (Table S1 online), and repeated testing for SARS-CoV-2 [2] . Several studies have assessed the secondary attack risk (SAR) of SARS-CoV-2, showing, for instance, that mask-wearing reduces virus transmission in households [79] . However, we are not aware of similar studies in NH. The goal of this study was to investigate the eciency of IPC measures implemented in the Hérault department (Occitanie region, France) in reducing the spread of SARS-CoV-2 in NHs when a patient was tested positive. We rst estimated the SAR dened as the proportion of individuals infected (positive rRT-PCR) in a NH after an outbreak [10, 11] . We further analysed the data using classical statistical modelling to better understand the relative role of specic factors. In the Hérault department (France), an observational retrospective longitudinal study was carried out in 12 NH which experienced a COVID-19 outbreak between March and May 2020. After clinical identication of a COVID-19 case in any NH, all residents were tested via rRT-PCR on nasopharyngeal swab tests. COVID-19 IPC measures were applied to all residents who were clinically followed up for 6 weeks, with repeated PCR testing. Serum antibodies were assessed at the end of the survey by two clinical laboratories that performed the analyses. Blood testing for IgG antibodies directed against the SARS-CoV-2 nucleocapsid protein used an enzyme-linked immunosorbent assay CE-IVD marked kit (ID screen SARS-CoV-2-N IgG indirect from IDVet, Montpellier, France). In the following analyses, positivity was based on rRT-PCR tests. As reported earlier, there was a 95% match between RT-PCR and the serological result [2] . From March 3, all the sta members and residents of all NH agreed to participate in the study. This observational study was approved by the Internal Review Board from the Montpellier University Hospital (IRB-MTP_2020_06_202000534). Seven main factors associated with each NH were extracted from a survey carried out among the directors of the 12 NH: • number of oors, • number of days between the rst COVID-19 case and the generalisation of mask-wearing, • number of medical sta per resident, • presence or not of a "COVID unit", i.e. isolation of infected patients, • reported sucient or lack of mask availability after the detection of the rst case, • proportion of single rooms, 2 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted December 2, 2020. ; https://doi.org/10.1101/2020.11.27.20239913 doi: medRxiv preprint • presence or not of temporary agency workers before the detection of the rst case. We rst estimated the Secondary Attack Risk (SAR) of the selected outcomes using a model-based approach described by Bailey [10] . Formally, the probability that j persons have been infected among n susceptible persons knowing there were a introduction events in the household during the outbreak is given by the formula Here, the "household" is assumed to be the oor of a NH. The value of the SAR was estimated using a maximum likelihood approach. By treating the proportion of infected residents in each NH after the outbreak (f ) as a nal epidemic size, we estimated the basic reproduction number (denoted R 0 ) in these NHs by solving the following classical equation [12] : To study the eect of the 7 main variables on the proportion of infected residents, we used generalized linear models (GLMs) with a binomial distribution for the response variable weighted by the total number of residents. In these models, the unit of analysis (the "household" in the SAR model) is the NH. We performed GLMs with all possible combinations of our 7 factors (i.e. 2 7 = 128 models) and performed model selection using the Akaike Information Criterion (AIC). An AIC dierence of 2 between models was considered to be signicant following classical statistical practice [13] . To improve statistical power, we also performed the analysis at the level of NH oors. The rationale for accounting for such structure within NHs is that activities between groups of residents were cancelled and that French national guidelines incited NHs to separate dierent structures and sta. To partly correct for non-independence issues, we used a nested structure with the oors being associated to a NH, which was itself treated as a random eect in the model. This was done using Generalised Linear Mixed Models (GLMMs) with a binomial distribution for the response variable and a normal distribution for the random eect [14] . As for GLMs, the model comparison was performed using AICs. Analyses were done using the lme4 package (glmer function) [15] in R version 4.0.2. The study involved 930 residents and 360 medical sta spread over 40 oors from 12 nursing homes (NHs), with 3.3 oors on average per NH. Details regarding the nursing homes can be found in Table S1 . The rst rRT-PCR positive cases were detected on March 10, 2020, while the last NH of the area to be aected by the rst wave reported its rst case on April 21, 2020. We then used GLMMs to identify the factors associated with NH outbreak size. The GLMM with the lowest Akaike Information Criterion (AIC), shown in Table 1 , contained two factors, the delay in maskwearing (in days) and the reported mask-availability. These factors were both signicantly associated with larger outbreaks as shown in Figure S1 . When analysing the 11 GLMMs that were comparable from a statistical point of view (their dierences in AIC with the best model was smaller than 2), we found consistent results. Table 2 shows that the two co-variables identied in the best model were the most often signicant in the 11 models. Another signicant co-factor was the presence of a "COVID unit", which decreases outbreak size. Finally, in some of the models, the presence of temporary agency workers before the rst case was associated with larger outbreaks. When assuming a less detailed model without any structure at the oor level, we identied 11 generalised GLMs that performed comparatively well from a statistical point of view (their AIC dierence with the best model was smaller than 2). The presence of a COVID-unit had a signicant eect in all 11 models. The next signicant eects were the presence of temporary agency workers before the rst case (10 models out of 11), the delay in mask-wearing (6 models out of 11) and, for two models, the number of oors in the NH. The eect of the factors on outbreak size was the same as for the GLMMs. First, we estimated the SAR of COVID-19 outbreaks in NH and found values lower than 5%, which is much lower than earlier estimates that ranged from 13.8% to 19.3% [8] , 23% [9] , or even 35% [7] . This is consistent with the strict IPC measures implemented after the rst outbreak was detected on March 10, 2020. Second, we conducted statistical modelling analysis to identify the relevant cofactors that best explain the heterogeneity of NH spreading of COVID-19. Generalised Linear Models (GLM) were used because the response variable was not continuous (number of infected residents per NH), making classic linear regression such as ANOVA or ANCOVA inadequate. The structure of the data allowed us to gain even more insight by working at the oor level. However, this raises non-independence issues between the Table 2 : Co-factors identied in the 11 best GLMM. We show the proportion of models that contain this co-factor, the proportion of models in which it is signicant (p-value<0.05 in the GLMM), and its eect on the total number of infection if signicant. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted December 2, 2020. ; same oor of a NH. To address this, we use Generalised Linear Mixed Models (GLMM), also known as hierarchical generalized linear models, which are commonly used in clinical research [14] . In these model, the NH as a`random' eect thereby partly corrected for the non-independence issue, although care had to be taken in model interpretation [15] . Our main approach was to analyse epidemics at the level of a NH oor using GLMMs. This was motivated by the fact that early guidelines led to contact limitations in NH (e.g. cancellation of group events, or assignment of members of the sta to specic oors). We found that two factors aected the epidemic spread within the NHs: the delay in mask-wearing and the reported mask availability. The earlier mask-wearing was generalised in the NH, the smaller the outbreak. Unexpectedly, reporting a lower mask availability was associated with fewer outbreaks. While this eect should be handled with care because of the subjective dimension of the variable, an explanation could be that a (reported) shortage of masks occurred in the NHs that were using more masks. Interestingly, when using less detailed statistical models that ignored the oor structure (i.e. GLMs), therefore assuming that cases occurred homogeneously in the NH, the main eect we found was the setting up of a "COVID unit", which was associated with smaller outbreaks. This further strengthens our choice to use a detailed GLMM model to analyse the data. The presence of temporary agency workers before the rst case was also associated with larger outbreaks in some of the GLMs. One potential limitation of the analysis could be the presence of over-dispersion in the data. When correcting for this potential bias, only 9 of the 127 potential GLMMs had a signicant factor, which was the presence of a COVID-unit in the NH. Unfortunately, correcting for overdispersion requires the use of quasi-likelihoods, therefore precluding AIC-based model comparison. Another major limitation of this study is that it was conducted retrospectively, which constrained the design and variable choices. In particular, some of the variables are subjective, which can generate unexpected correlations. However, this work can be considered as a pilot and will help to design further prospective studies with improved statistical power and the possibility to potentially include additional factors in the analysis. Overall, these results conrm the eciency of the measures implemented in the South of France to prevent SARS-CoV-2 epidemics in nursing homes. They reveal the importance of within-NH structure. They also show that delays in the generalisation of mask-wearing before the rst case are strongly associated with the magnitude of the outbreaks. Finally, they support the US CDC and European guidelines, which both recommend that NH facing a COVID-19 outbreak dedicate an area of the facility with specic stang and ICP measures to the care of residents with suspected or conrmed COVID-19 infections [2, 4] . Figure S1 : Eect of delay in mask-wearing and mask availability on the proportion of infected residents. The plot is based on the coecients from the GLMM model shown in Table 1. 8 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted December 2, 2020. ; https://doi.org/10.1101/2020.11.27.20239913 doi: medRxiv preprint Table S1 : Nursing home details. The proportion of infected sta and residents were estimated using serological assays. Missing data are All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted December 2, 2020. ; https://doi.org/10.1101/2020.11.27.20239913 doi: medRxiv preprint Epidemiology of Covid-19 in a Long-Term Care Facility in King County, Washington Ecacy of a Test-Retest Strategy in Residents and Health Care Personnel of a Nursing Home Facing a COVID-19 Outbreak Outbreak of COVID-19 in a nursing home in Madrid Center for Disease Control and Prevention. Preparing for COVID-19 in Nursing Homes Interim EuGMS guidance to prepare European Long-Term Care Facilities for COVID-19 Epidemiological monitoring and control perspectives: application of a parsimonious modelling framework to the COVID-19 dynamics in France Secondary attack rate and superspreading events for SARS-CoV-2 Household secondary attack rate of COVID-19 and associated determinants in Guangzhou, China: a retrospective cohort study Reduction of secondary transmission of SARS-CoV-2 in households by face mask use, disinfection and social distancing: a cohort study in Beijing The mathematical theory of infectious diseases and its applications. High Wycombe Secondary Attack Rate The mathematics of infectious diseases A practical information-theoretic approach. Model selection and multimodel inference Using the general linear mixed model to analyse unbalanced repeated measures and longitudinal data Generalized linear mixed models: a practical guide for ecology and evolution No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity Acknowledgements: The authors would like to thank the residents and their families for participating in the study, along with all of the NH sta for taking the time to answer a survey, while helping their residents during the crisis. The authors also thank IdVet for providing serological tests and Anna Bedbrook for her help in improving the writing.