key: cord-0832739-h6d43tzv authors: Kumar, A. title: COVID-19 risk evaluation and testing strategies based on contact tracing network and information analysis date: 2020-12-02 journal: nan DOI: 10.1101/2020.11.30.20240762 sha: 5b2ad463f00ecd74f0cba7f3318af0f2a30aa6c0 doc_id: 832739 cord_uid: h6d43tzv Contact tracing and efficient testing can have an imperative part in containing the COVID-19 spread, with minimal social and economic disruption. Testing serves many purposes: isolating the COVID-19 positive tested individuals, identifying the contacts at the risk, and locating the hotspots and safe zones for administrative planning. However, it is a challenging task to identify the right individuals for the test in view of the high COVID -19 spread, a large number of presymptomatic and asymptomatic cases, and limited testing capabilities. The individuals for COVID -19 are identified based on direct-contact, travel history, and symptoms, which are more individualized and do not explicitly include a group risk assessment, and in turn, do not preclude the transmission from the superspreaders. Policymakers need to limit testing in the shortage of test resources, and focus on gaining the most information from the tests performed. In this work, we introduce a protocol for the identification of the group of individuals to be tested for acquiring maximum risk information of a community with minimum individual tests. Firstly, an algorithm is proposed to determine the risk profile of all the individuals in the community by incorporating serial and parallel pathways of the infection transmission considering multiple steps of transmission. Next, we consider several potential groups that could be tested from the community, and analyze them one by one for their comparison. In a group, few individuals can be positive, and the remaining few can be negative, generating sets of several test-outcomes with unequal probability. The protocol involves the probability calculation and reassessment of the network's risk profile in all the test output cases. Finally, the best group is identified in all the groups studied, in which risk profiles between post and pre-test are maximally different. The analysis shows that in general, information increases with an increase in the group size. Notably, a strategically chosen small group may provide more information from the test results, than a standard larger group. The proposed systematic strategy would help in the selection of the right individuals for the testing, and in extracting far more information from the minimum samples, to effectively aid the epidemic mitigation. The protocol is generic, and can also be applied to any other epidemic spread in the future. COVID-19 is an unprecedented challenge for mankind; affecting health care, the global economy, movement, and social life. Various stringent policies have been placed to mitigate the spread at the cost of shutting business and social events, which means that these initiatives are not permanent and can only be enforced in risky zones. [1] In contrast to all other past epidemics, such as SARS and MERS, asymptomatic cases of COVID-19 in the population are very high, and it is necessary to rely only on specialized testings. [2] In the present situation, contact tracing and well-planned testing are the most important public health interventions available, also recommended by World Health Organization. [3, 4, 5] Various apps are launched by several countries, such as Trace Together by the Singapore govt, Aarogya Setu by the Indian govt, TousAntiCovid by French govt, etc. which are based on Bluetooth based entries to capture the possible interactions, for acquiring the information of interaction with an infected individual. [6] As an example, the TousAntiCovid alerts the individuals who have been in close contact with a reported COVID-19 user for exposure at less than one meter, lasting 15 minutes or more. Most importantly, the contact tracing and testing helps in understanding the statistics of community spread, required for policymakers to contain it. [7, 8, 9, 10] And if vaccine or cure for the COVID-19 is possible in the near future, there is always a need to focus on contact tracing to target individuals for vaccination or meditation. Various testing strategies have been discussed for effectiveness, such as bidirectional contact tracing [11] , pool testing [12, 13] , and testing at the exit of the quarantine period than the entry. [14] Effectiveness of contact tracing depends on the number of contaminated individuals, testing quickness, fraction, and rapidness in contact tracing and quarantine, and tracking the further transmission. [15, 16, 10] The individuals for COVID -19 are primarily identified based on direct-contacts, travel history, symptoms, and random sampling. Mass scale testing in most of the countries is impractical to conduct due to higher testing costs, the flow of new entrants into the community [17] , and the ability of the person tested to become contaminated in the future. The infection spread is largely due to superspreaders in the community who transmit the infection to several individuals, [18] these cases can only be prevented by detailed network analysis. [19] Thus, it is very important to post-process the contact tracing data, and make strategies for testing and isolating individuals at the risk. Policymakers should have strategies to acquire maximum spread information using minimum tests conducted in the community. However, the detailed risk evaluation based on a network level with several multi serial and parallel connections is still an ongoing challenge. Currently, the goal of a test is largely on an individual level, and to obtain the positivity rate of the community, but does not provide a detailed risk assessment of the community. Apps such as Aarogya Setu inform the users of the exposure to COVID-19, by testing their proximity to known patients, [20] while considering only single step transmission. As per our knowledge, only Guttal et al. [21] have looked at the risk determination of the contacts, however, it uses the Gillespie simulations, which may not be easy to perform and does not consider the weighted connections. The present work introduces the strategies to evaluate the risk profile of the community, identify individuals for tests, and propose testing protocols in the direction of acquiring the maximum information with a minimum number of tests. Consider a social network with individuals represented as nodes, and edges representing the transmission probability T i→j of infection between the individuals. Figure 1a represents a typical network with 13 individuals, represented as alphabets from A-K, with red and blue nodes representing the infected and the susceptible individuals respectively. The corresponding connections values (0-1), represent the probability of disease transmission between the individuals. The transfer probability can be obtained from app-based data or manual contact tracing questionnaire considering the time of mutual interaction between the individual, possibility of spread, and decaying the weight over time. In this work, we do not focus on the network construction based on contact tracing, but rather on post-analysis of the network, which would be important for the risk assessment and suggested testing strategies. For the first step of transmission, R 1 i shown in Figure 1b , nodes immediately linked to the infected nodes may be at risk of infection, the risk value proportional to the total of transmission strengths from neighboring infected nodes. The risk value (0-1) is the risk that the transmission from the reported cases in the community could potentially infect a person. In this manuscript, the risk value is based only on an overview of the contact tracing network taking into account the network's direct and indirect transmission route. As an example, there are 2 infected individuals connected to node B, i.e. node J and node M, with interaction weights of 0.2 and 0.26 respectively, thus the probability of infection of node K is 0.46. Figure 1 (b) displays all individuals' risk profiles, calculated similarly, with the node values and color intensity as the risk. Nodes with higher risk values can be potentially infected nodes, and in turn, may pass the infection to their respective neighbors in the next step. The risk of i individual for s step transmission (R S i ) considering all neighbors j with their corresponding risk values R S−1 j at S − 1 steps can be calculated as, Note that a variable threshold value dependent on age and medical history may also be incorporated in the realistic case. One can easily make a histogram of risk values to quantify the spread in the community. Furthermore, the simulations can be re-run for long-term forecasts of several transmission phases as seen in Figure 1d for three transmission steps demonstrating maximum risk at the nodes: B, K, H, C and F. Algorithm 1 shows the detailed steps of the computation, the parallel connections 2 . CC-BY 4.0 International license It is made available under a 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. ; terminating at a node are added, while the serial connections are multiplied, constraining the number of steps, and total risk value not exceeding unity. Note that, the algorithm, avoids self-loops in accordance with the physical situation, that is, infection by i to j and back j to i (by any mode) is not allowed. The time scales of transmission of epedemic can be integrated in order to predict the time evolution of infected population. Although the code can anticipate multi-step predictions, we have confined the transmission to only two stages in further studies, for the simplicity of the readers. It is important to test the potentially infected individuals for controlling the epidemic spread. Nevertheless, in a shortage of test resources, policymakers need to minimize the tests and focus on acquiring maximum information from the tests. Herein, we try to identify the group of individuals, who should be tested for the maximum information. (Figure 1c ) and E being positive is 0.06 (Figure 1c) , thus the probability of C and E both being positive is the product of probability i.e. 0.018. The individual probability for C and E to be negative is 0.7 (1-0.3) and 0.94 (1-0.06), which is 0.658, shown in Figure 2 in two decimal values. Thus, the probability of an outcome case (P c ) can be written in terms of individual risk values of tested positive and negative individuals as: The risk value of the tested positive individual is updated as 1, and for the negative tested individual as zero. Since all the cases are having unequal probabilities (p c ) and risk values (R iC ), the risk variation of a node, ∆R i can be written as 3 . CC-BY 4.0 International license It is made available under a 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. ; Figure 1b . The individual probabilities of the test outcomes are multiplied to obtain group test probability. The tested positive individual is represented by T+ and tested negative is represented by T-. The risk profile for all these cases is re-calculated and represented as node values. The risk variation for all group tests is compared, considering all the above cases with corresponding probabilities of occurrence. The best group is found to be [D, F], having maximum risk variation with the tests, which implies maximum information with the tests. . CC-BY 4.0 International license It is made available under a 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/10.1101/2020.11.30.20240762 doi: medRxiv preprint The objective of a test in general is to acquire information about a system, which the observer does not know before carrying out the test. In this particular case, if the risk profile variation with a test is negligible, it means there is no benefit of a test. 5 . CC-BY 4.0 International license It is made available under a 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/10.1101/2020.11.30.20240762 doi: medRxiv preprint CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) A systematically selected smaller group can offer more information, than an unplanned selected bigger group. As the computation considers all the attributes of the network, thus one need not rely on betweenness, degree, or any such network-feature for selecting the nodes. The protocol is automated using the proposed algorithm and thus can be used at different institutional levels, with minimal training of the policy-user. Though we have used COVID-19 as a specific challenge to handle, nevertheless, the approach is generic and can be applied to any similar epidemic, without any specific modification. In the manuscript, we have developed a model for the evaluation of the epidemic risk profile in the community and discussed effective testing strategies for transmission chain mitigation. A simple model is proposed to determine the risk profile of the population with the multi-step transmission, incorporating serial and parallel transmission pathways in the network. We address the strategies for identifying individuals for COVID-19 tests, based on risk assessment in the community. On testing a group of individuals, few of them can be positive, and the remaining can be negative, generating a set of possible test outcomes. The group for testing is identified with maximal variation in the risk profile, based on all possible test outputs. Further, the individuals at the risk are identified for different test outcomes that could be useful for alerting the right individuals, without their individual tests. Using the proposed testing framework, the strategist shall obtain much more prediction information, with a minimal number of tests. On systematic applicability, the present strategy can break the transmission chain, which in turn, can control the epidemic despite limited testing capacity. 7 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted December 2, 2020. ; https://doi.org/10.1101/2020.11.30.20240762 doi: medRxiv preprint Long-term strategies to control covid-19 in low and middle-income countries: an options overview of community-based, non-pharmacological interventions Quantifying sars-cov-2 transmission suggests epidemic control with digital contact tracing Feasibility of controlling covid-19 outbreaks by isolation of cases and contacts. The Lancet Global Health A survey of covid-19 contact tracing apps Modelling the impact of testing, contact tracing and household quarantine on second waves of covid-19 Covid-19 mass testing programmes Sampling designs for epidemic prevalence estimation COVID-19 testing strategies and objectives Bidirectional contact tracing dramatically improves covid-19 control. medRxiv Simulation of pool testing to identify patients with coronavirus disease 2019 under conditions of limited test availability Optimally pooled viral testing Optimal COVID-19 quarantine and testing strategies. medRxiv Effective contact tracing for COVID-19: A systematic review. medRxiv Impact of delays on effectiveness of contact tracing strategies for covid-19: a modelling study Modeling geographical spread of COVID-19 in India using network-based approach. medRxiv Impact of superspreaders on dissemination and mitigation of covid-19. medRxiv Exploration of superspreading events in 2015 mers-cov outbreak in korea by branching process models CA review of India's contact-tracing app Risk assessment via layered mobile contact tracing for epidemiological intervention. medRxiv The author acknowledges Prof. Fabien Alibart and Prof. Giridhar U. Kulkarni for the encouragement.