key: cord-0307097-xkl6yyhu authors: Kolen, B.; Znidarsic, L.; Voss, A.; Donders, S.; Kamphorst, I.; Rijn, M. v.; Bonthuis, D.; Cloquet, M.; Schram, M.; Scharloo, R.; Boersma, T.; Stobernack, T.; Gelder, P. v. title: SARS-CoV-2 risk taxation model and validation based on large scale Dutch test-events date: 2022-01-10 journal: nan DOI: 10.1101/2022.01.10.21268254 sha: 31a1a1eb5093619be6c315ea90fbe11e4f27d0b2 doc_id: 307097 cord_uid: xkl6yyhu In response to the outbreak of SARS-CoV-2 many governments decided in 2020 to impose lockdowns. Although the package of measures which constitute such lockdowns differs between countries, it is a general rule that contacts between people, and especially in large groups of people, are avoided or prohibited. The main reasoning behind these measures is preventing that healthcare systems become overloaded. As of 2021 vaccines against SARS-CoV-2 are available, but these do not guarantee 100% risk reduction and it will take a while for the world to reach a sufficient immune status. This raises the question whether and under which conditions events like theater shows, conferences, professional sports events, concerts and festivals can be organized. The current paper presents a COVID-19 Risk taxation method for (large scale) events. This method can be applied to events to define an alternative package of measures replacing generic social distancing. took measures, resulting in various lockdowns to reduce the number of contacts between people and the amount of people that can gather in a group. As a consequence, large scale events were generally prohibited. At the same time, Dutch event-organizers affirmed that organizing events on the basis of social distancing would be economically detrimental. The question there is if an alternative package of measures can be defined instead of generic social distancing at these events, while requiring that the average individual risk of becoming infected at an event is equal to the infection risk for staying at home. In this study we distinguish between four types of events which can be representative for almost all events: • Type I: Indoor, passive (theater show or conference), • Type II: Indoor, active (concert or dance events), • Type III: Outdoor, active (public sports events), • Type IV: Outdoor, active festival (festivals). An alternative for physical or biological models is a data driven analysis. Based on available data of infections and contacts among people in this period a model can be developed which relates the risk of infection to the number of contacts and other measures which can be implemented. Because governments directly implemented numerous measures to support social distancing and avoid large groups, the normal social contact data (9) do not apply because the behavior of people had been forced to change. 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 January 10, 2022. ; In this study a SARS-CoV-2 risk taxation method was developed based on National Institute for Public Health and the Environment (RIVM) and Municipal health services (GGD) collected data sets during the SARS-CoV-2 pandemic. The method resulted in the average risk of infection (ܴ ) per hour, whereby ܴ is a function of the prevalence (P) and the number of contacts per hour at a given location. We distinguish two contact-classes: up to 1,5m for droplets ‫ܥ(‬ ଵ ) and up to 10m for aerosols ‫ܥ(‬ ଶ ). The risk can further be reduced by: • • Smart logistics at an event to reduce contacts and gathering of large groups of people. Therefore, specific data were collected at test events. • Personal protection measures such as mouth-nose masks. This reduces the risk with factor ‫ܨ‬ ெ . • The impact of vaccination, which reduces the risk of transmission with a factor ‫ܨ‬ . These measures resulted in an average individual risk of infection per hour, given by: One of the first estimates was the risk per location (see table 1 ). As a consequence of the stay-at-home advice, most people, in absolute numbers, became infected at home. When corrected for the number of people present and duration of contact, the risk of getting infected at these locations can be standardized and the risk estimates be compared with other locations. The hours while sleeping (8 hours) are excluded. In the second step, we estimated the number of contacts between people at a location. To do so, we combined data from a specifically developed questionnaire with statistical data (10) . We focused on the locations at home, at work, visitors at home and leisure, for the other 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 January 10, 2022. ; https://doi.org/10.1101/2022.01.10.21268254 doi: medRxiv preprint locations the amount of available data was not sufficient. In the questionnaire we asked people to estimate: • The time spent at a certain location. • The number of people in a range of 10m. • The number of contacts in a distance < 0,5 m, between 0,5 and 1,5 and between 1,5 and 2,0 m for less than a minute, between 1 and 15 minutes and more than 15 minutes. • The proportion of the time which was indoor, indoor and well-ventilated or outdoor at the location. In the third step, we performed the data-analyses using linear regression via Leastsquares Minimization on function (1) The complete role of virus-laden droplet and aerosol transmission is poorly understood (11) . ‫ܥ‬ ଶ (with a radius of 10m) is 10 contacts for the low contact event and 30 for the 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 January 10, 2022. ; https://doi.org/10.1101/2022.01.10.21268254 doi: medRxiv preprint high contact event. The sensitivity analysis shows that the impact on the results is within a bandwidth of a factor 0.5. The model determines an average risk for infection. The data concerning SARS-CoV-2 infections gathered at the test events can be used to validate the model. The model outcome is based on a skewed probability distribution. For example, consider a large-scale event where 1.000 people will join the event given a prevalence of 0.75%. Without pre-testing 7.5 contagious persons would have attended the event, when ‫ܨ‬ ் ൌ 0 . 9 5 on average 0.38 person would have been contagious at the event and would thus be a source. Also the number of contacts per individual person will vary or some will have many close contacts and others will see only a few people. Therefore, it is expected that many events will be organized with no or limited number of infections, and a few with many infections. A model validation would need a large dataset during the SARS-CoV-2 pandemic. This large dataset is expected to cover the skewed probability distribution including events with no infections and events with many infections. Such a database, however, is not available. Data from media and literature of superspreader events are biased as these always attract more attention. For loss of life modelling for natural hazards the limited availability of data also leads to difficulties for conducting model validation. For example, the loss of life models for river and storm surge flooding in The Netherlands is based on the 1953 flood and Katrina in the US (12, 13) . However, despite the limited validation, the model is still used to define the safety standards for Dutch levees which implies an investment program of multiple billions of euros (14) . 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 January 10, 2022. ; Although a perfect validation is not possible, the available data can be used for a first validation of the model. be that people received more visitors than they admitted to in the survey or than allowed within the prevailing COVID-19-rules (two visitors per day until mid-October, and one per day afterwards). Overall, we concluded that the outcomes support the results of the model. The test-events organized by Fieldlab phase 1 can be used for a second validation. The test events occurred while the prevalence was high and a lockdown was still in place. These test events have been organized to measure ‫ܥ‬ ଵ en ‫ܥ‬ ଶ given the implemented measures. Very short, "passing" contacts of less than 10 seconds were not taken into account because these "passing contacts" are assumed not to be significant with regard to transmission. At the test events, the generic measures for social distancing were not in place. Instead, so called "Fieldlab measures" were taken, such as the separation of groups of people in bubbles. During the test events, several packages consisting of variation in occupation rate, catering, routing, instructions on when to wear masks and indoor/outdoor situations have been tested. Persons with COVID-19 (like) symptoms were banned from participation. All visitors and crew needed a negative PCR test taken within 48 hours before the event in order to attend. As the PCR test may pick up low viral loads such as in cases of persons who recently recovered from COVID-19, the ratio of positive tests is higher than the ratio of asymptomatic people only. All Fieldlab participants and crew were asked to get tested on day five after the event, a request that was followed by more than 80%. In addition, all positive cases related to a Fieldlab event, identified by the regional Health Care Services (GGD) were included in the data set. Infections identified after an event, included cases infected just before or after the pre-test or had a PCR test around the cut-off of the PCR, thereby varying in outcome. People can be infected at the event, but also at other locations. The input for the model is defined as: • While in the post-event test results, the crew is also taken into account, the calculated risk for infections applies to visitors only. • ‫ܨ‬ ் ൌ 0 , 9 5 , was set based on expert judgment and the sensitivity of a PCR test (15), combined with the restriction of testing within 48 hours before the event. • Ventilation of all indoor places is according to the building codes, (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 January 10, 2022. when the mask is also used while people move. During type 2 and 4 events, F M was set at 0, as masks compliance was extremely low to non-existent. In Table 4 Depending on the package of measures (as testing, maximum occupation rate, ventilation) the risk at an event without social distancing can be reduced to a risk which is equal or less while staying at home during the lockdown in the Netherlands. A second comparison can be made between the risk at an event and having visitors at home. The risk of visitors at home is about four times higher than the risk at home. The number (and risk of) infections at the event are estimated with the model based on the prevalence and measures at the test-events. The model results can be compared with the confirmed infections after the events. 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 January 10, 2022. ; As the number of events was limited related to the expected probability distribution, the data analysis would have profited from a higher number of events. Still, the data can be used to check if the model is plausible. We believe that we included nearly all infected people at the event because we combined the information of regular testing procedures and the after-event tests. In the model, we assumed an average prevalence for the Netherlands. Our assumptions were actually corroborated by the pre-event test results, which were in the range of the nationally reported prevalence of SARS-CoV-2. However, it might be that the prevalence among the visitors was higher than the assumed average. A first argument is that more young people attended the test events, and these age groups contribute relatively more than elderly (as > 60 years old) to the positive PCR tests (18). A second argument is that the test events were during the lockdown, and the non risk-averse people which attend these events also might have more other activities. Because of these other activities it can be expected that the source of infections for some of the cases identified after the event may be unrelated to the event. At day 5 after the event, visitors and crew were tested, but our model exclusively calculates the risk for visitors, as contact data of the crew were not measured. During the events, the crew attempted to keep their distance from the participants, and wore masks continuously. In general, we found more PCR-positive people in the pre-event than in the post-event testing. As the PCR test may pick-up low viral loads, some of the people testing positive, especially in the pre-tests group, may have been recovered from COVID-19, consequently resulting in a higher positive test ratio. This in part explains the positivity rate in the pre-and post-event testing. However, in the after test, also people with COVID-19 symptoms were 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 January 10, 2022. ; included. Only for (one of) thetype 4 events, the ratio of positive tests in the after test was higher than in the pre-test. This corresponds with higher numbers of infections at the event. In the post-event tests, 14 cases have been identified where persons were possibly infected at the events. Other positive tests were excluded during interviews as participants had known contacts with SARS-CoV-2 positive cases at their home or unrelated events around the same time. Even the 14 cases are only "possible" cases, as infections could have occurred in other places and situations at any time from around the pre-test, the day of the event, or even the after the event. If we use the Dutch average infection risk during the event as a control group we can estimate the number of infections which could be expected outside the event. The time spent at the event is about 4-10% of the time a period of, for example, three days where people were at different locations exposed to the risk of infection. If we also consider the risk at other locations, one or two of the possible infections could be (on average) related to the events. During the eight test-events, four persons were confirmed to be infected during the event. Confirmed infections are those infections which can be related to each other for example by sequencing or because of proven contacts with other positive tested people during the event. Two infections occurred while travelling home with a contagious person, which caused two infections at the type IV event. These are not infected at the event (and part of the model) but these are related to the event. A last remark about the model can be made with regard to the data which is used to estimate A 1 and A 2 . The contacts used to train the model were gathered during a period of a (partial) lockdown. Large scale events were already prohibited or regulated. This could cause an underestimation of the risk in dynamic settings. 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 January 10, 2022. ; Based on the available but scarce information, the current risk taxation model results in plausible results, the model cannot be rejected as being invalid. The test events show that the Fieldlab measures at the events, which replace generic social distancing, reduce the risk to a level below the threshold level (which was the risk at home). The model can be extended with the risk for loss of life and hospitalization using the relation with the age of people. However, if more data is available, the model can be improved because of the larger training set. If more data of especially type IV and maybe type II events are available, research can be done on an additional factor for increased transmission at dynamic events. Therefore, it is recommended to collect more data about infections and events while the COVID-19 pandemic continues. voor Volksgezondheid en Milieu -RIVM 6 april 2021. 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 January 10, 2022. 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 January 10, 2022. 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 January 10, 2022. ; https://doi.org/10.1101/2022.01.10.21268254 doi: medRxiv preprint 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 January 10, 2022. ; https://doi.org/10.1101/2022.01.10.21268254 doi: medRxiv preprint Estimate of the Basic Reproduction Number for COVID-19: A Systematic Review and Meta-analysis Estimation of the basic reproduction number, average incubation time, asymptomatic infection rate, and case fatality rate for COVID-19: Meta-analysis and sensitivity analysis SARS-CoV-2, SARS-CoV, and MERS-CoV viral load dynamics, duration of viral shedding, and infectiousness: a systematic review and meta-analysis Incubation period of 2019 novel coronavirus (2019-nCoV) infections among travellers from Wuhan The incubation period during the pandemic of COVID-19: a systematic review and meta-analysis Associations between measures of social distancing and SARS-CoV-2 seropositivity: a nationwide populationbased study in the Netherlands The Netherlands in figures Transmission of COVID-19 virus by droplets and aerosols: A critical review on the unresolved dichotomy Loss of life estimation in flood risk assessment -theory and applications Analyse slachtofferaantallen VNK-2 en voorstellen voor aanpassingen van slachtofferfuncties. HKV lijn in water Rapport Hoogwaterbeschermings programma: van 'sober en doelmatig' naar 'slim en doelmatig'Board of Government Advisors. The Hague Insight into the practical performance of RT-PCR testing for SARS-CoV-2 using serological data: a cohort study