Design and evaluation of Covid screening strategies in various contexts based on empirical contact data – Coscreen

We decided to first focus on the case of schools, as a context of particular interest: it has indeed been considered as crucial to maintain schools as open as possible during the pandemic, even when other measures were enacted (curfew, teleworking). We thus reviewed available data on contacts in school contexts, and decided to use data from the SocioPatterns collaborations, collected in a pre-pandemic period, and describing contacts in a primary school and in a secondary school.

We devised a systematic way to create synthetic data having the same properties of the empiric ones, for two reasons: (i) the empiric data covers only few days of school, while we needed to simulate a propagation over several months (ii) the empiric data covered only the students of one year in the secondary school. Moreover, we decided on a systematic way to couple the schools simulated in the computer with the outside world, namely simply by tuning the introduction of a fixed number of cases in the school each week.

We then built an agent-based model to simulate the propagation and the protocols in both types of schools. We calibrated the model to have a desired effective reproduction rate R. We built a modular simulation tool in order to be able to simulate and compare the various protocols. The baseline protocol was given by the simple isolation of cases detected because they exhibited symptoms. We then considered the protocol of closing the class (or the level, i.e., 2 or 3 classes) of each detected case, for a period of 7 days. Most importantly, we simulated a new protocol of testing a certain number of times per week (from once every 2 weeks to 4 times per week) a certain fraction of students and teachers (given by the adhesion). We considered adhesion levels ranging from 10 to 90%. We then considered the possible vaccination of the teachers, the possible vaccination of students in the secondary school. We performed sensitivity analysis concerning the infectiousness of children, the value of R, the fraction of teachers of students vaccinated.
For each protocol, we have measured the distribution of epidemic sizes at the end of a three months period: we have thus measured how much each protocol improves the epidemic situation in terms of reduction of epidemic size with respect to the baseline protocol. Moreover, we have measured the cost of each protocol in terms of days of schools lost per student.

Our main conclusion, which is robust across types of schools, is that iterative screening, if frequent enough and with high enough adhesion, allows to reduce the epidemic size at least as much as the current protocols, and in many cases performs actually much better in terms of epidemic size reduction. Moreover, this occurs at a cost in terms of school-days lost that remains much smaller than with protocols that close full classes. This is because isolating only the cases detected avoids sending in isolation many students who actually did not contract covid, while targeting exactly those who have covid. Regular repetition of the tests manages also to find those cases that might have been just contaminated (not yet test positive) at one iteration, and become test positive at the next iteration.

Our early results had already been of very strong interest to the public authorities. The Haute Autorite´ de Sante´ has asked us to intervene at a meeting discussing self-testing on April 22, and the Conseil Scientifique has included our results as part of the “Avis du Conseil scientifique COVID-19” on April 19, 2021 titled “Les autotests: une opportunite´ de sante´ publique” (https://www.vie-publique.fr/rapport/279618-avis-du-conseil-scientifique-covid-19-du-19- avril-2021-les-autotests)

Our consolidated results have been cited by the Conseil Scientifique in their Note d’Alerte www.vie-publique.fr/sites/default/files/rapport/pdf/281331.pdf
and in their avis
(available from solidarites-sante.gouv.fr/actualites/presse/dossiers-de- presse/article/conseil-scientifique-covid-19 )
A press release of INSERM was published on August 26 presse.inserm.fr/a-la-rentree- les-chercheurs-recommandent-lautotest-regulier-pour-controler-lepidemie-dans-les- etablissements-scolaires/43642/

In addition, we have considered an additional protocol, namely the reactive testing of a class when a case is detected; in addition to the evaluation of this protocol (which is currently tested in France in a number of departments), we have compared it to the ones previously simulated. The results are summarized in a note made public on our webpage: www.epicx-lab.com/uploads/9/6/9/4/9694133/reactive_screening_primary_school- 2.pdf

We have moreover started working on a more theoretical aspect of the project: namely, the amount of detail needed in the description of the contacts between individuals in specific contexts such as schools, hospitals or offices in order to be able to evaluate and rank possible containment strategies and protocols. Indeed, while very detailed contact data have been collected in some cases, this is not the general case, and often one has only a summarized coarse-grained knowledge of the actual contacts occurring between individuals. In previous works, we had considered the issue of whether such limited knowledge could be enough to predict the impact of a schematic disease, and shown that some overestimation effects could occur. We want however now to consider the problem from the point of view of (i) a more complex and realistic model for the disease (ii) the modeling of intervention protocols and the question of deciding whether they are effective and which protocols are the best.
We have thus defined several data representations for the contacts in various contexts of interest, and have started to evaluate the impact of various protocols when simulated on these various data representations, which encode a variable amount of detail on the contacts.

Self-testing and vaccination against COVID-19 to minimize school closure
Elisabetta Colosi, Giulia Bassignana, Diego A Contreras, Canelle Poirier, Simon Cauchemez, Yazdan Yazdanpanah, Bruno Lina, Arnaud Fontanet, Alain Barrat, Vittoria Colizza
medRxiv:2021.08.15.21261243v1

In most European countries, the first wave of the COVID-19 pandemic has been fought with nationwide lockdowns with huge costs for the society, economy, and well-being of individuals. The measures were lifted before summer and, with the start of normal activities in September, associated with changes in weather conditions, a second wave is currently hitting our countries. The test-trace-isolate strategies have been overwhelmed and renewed restrictive measures are being taken in several countries, including France, to try to limit the spread. There is however a strong political effort to maintain schools open and focus on restrictions targeting other settings. The key to make this possible and avoid new disruptions of educational activities is systematic and efficient testing, along with contact tracing. Improvements in this context are expected from new generations of tests, such as antigenic tests, which lead to fast results. They have a lower sensitivity than PCR tests, but it has been argued that high sensitivity might not be crucial in the context of the screening of a population, especially if the screening can be repeated regularly or the sampling can be made more comfortable (eg saliva sample). New strategies could then become possible, with systematic wide testing repeated at regular intervals in specific populations, coupled with contact tracing and adequate reactive measures whenever a certain number of positive cases is detected. This is currently being tested as a pilot project by APHP in the Medicine Faculties of the Paris Universities and in schools in Switzerland. COSCREEN addresses the urgent and crucial need to design, validate and optimize such new strategies of repeated testing and screening in contexts such as schools, offices, universities, to guide the implementation in the field. As these strategies will address and be tailored to specific contexts, their investigation and validation need to be based on mathematical models that can quantify the advantage of certain strategies depending on the epidemic context, implemented measures, size, structure and contacts of the population. In particular, these models need to include the complex properties of human contacts in various settings, revealed in the last decade by the analysis of detailed contact data. We will leverage such high-resolution data, collected by the PI over the years, and the COVID-19 modeling experience of the co-PI, to design detailed realistic numerical simulations of the spread of COVID-19. In particular, we will consider different epidemic contexts (incidence, effective reproductive number, immunity in the population) estimated from hospitalization and epidemiological data in France, based on the work performed in these months. We will design a set of strategies based on regular screening of the population in the studied settings and on reactive temporary closures, evaluate how the impact of each strategy on the spread depends on the various parameters, with emphasis on (i) the compliance to the testing strategy, (ii) the type of test and its sensitivity, (iii) the delay between test and its result. We will evaluate the cost of the strategies in terms of the number of days in which remote teaching or working has to be implemented. This will allow us to propose optimized tunable strategies depending on contexts and on the cost-benefit balance, achieving different objectives depending on the setting (e.g. maximizing school in session for younger children). The project will deliver preliminary and consolidated reports to public health authorities, which could be disseminated to advise the administrative bodies of schools and companies on optimal strategies for testing. The resulting guidelines could be rapidly implemented depending on the evolution of testing capacities and properties of the available tests. Our framework will be able to be rapidly updated whenever additional knowledge on risk factors, tests, and protocols become available.