key: cord-0841830-560jhnua authors: Wells, K.; Lurgi, M. title: COVID-19 containment policies through time may cost more lives at metapopulation level date: 2020-04-29 journal: nan DOI: 10.1101/2020.04.22.20075093 sha: 21401f48a0c55fbc9623278bf6790db8e11dbe4e doc_id: 841830 cord_uid: 560jhnua The rapid and pandemic spread of COVID-19 has led to unprecedented containment policies in response to overloaded health care systems. Disease mitigation strategies require informed decision-making to ensure a balance between the protection of the vulnerable from disease and the maintenance of global economies. We show that temporally restricted containment efforts, that have the potential to flatten epidemic curves, can result in wider disease spread and larger epidemic sizes in metapopulations. Longer-term rewiring of metapopulation networks or the enforcement of feasible long-term measures that decrease disease transmissions appear to be more efficient than temporarily restricted intensive mitigation strategies (e.g. short-term mass quarantine). Our results may inform balanced containment strategies for short-term disease spread mitigation in response to overloaded health care systems and longer-term epidemiological sizes. from household to city levels (7) . In fact, it is well known that contact network structure and patterns of connectivity among local populations are main drivers of contagion processes from local to regional scale, with contact frequencies among populations largely determining whether local epidemics are coupled or not (8) . However, the extent to which temporarily restricted intensive mitigation strategies (e.g. short-term total containment) in response to the COVID-19 epidemic can successfully lower overall epidemic size, and protect those most vulnerable, has yet to be quantified. and compared across different scenarios of landscape-level connectivity (i.e. metapopulation structure). 65 Theory predicts that metapopulation extinction-colonization dynamics can maintain disease persistence even with frequent local fade-out and extirpation (9) . If local epidemics in different populations are weakly coupled, synchronous but short-term mitigation strategies may fail their target to contain disease spread during epidemic peaks (10). Further, local containment of 70 outbreaks with delayed local extirpation may facilitate disease spread at the metapopulation level. Metapopulation source-sink dynamics, the effect of metapopulation structure, and the role of depleted pools of susceptible individuals (i.e. herd immunity), have been studied in the context of disease spread for decades (11) . Nevertheless, there is a lack of understanding of whether and how temporally restricted intensive mitigation strategies, could result in long-term 75 reduction of overall infection rates on spatially structured metapopulations. Understanding the usefulness of short-term intensive containment strategies constitutes a pressing challenge in times of the ongoing COVID-19 crisis. More generally, this understanding will contribute to the unresolved debate on whether it is possible to establish alternatives to socioeconomically detrimental mass quarantines (6, 12) . 80 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. We address this challenge using an individual-based epidemiological model, that incorporates landscape-scale connectivity structure among local populations (i.e. metapopulation), to explore the outcome of the spread of COVID-19 and equivalent emerging infectious diseases in metapopulation networks after one year and under different mitigation scenarios. The interplay between transmission rates at local scales, and metapopulation dynamics (e.g. source-85 sink), which ultimately results in the possible depletion or replenishment of the pool of susceptible individuals, should determine the outcome of temporal mitigation strategieseither through escape from epidemic waves by accurately timed isolation or through more prolonged exposure induced by temporarily arrested and sustained epidemic spread. We were interested in exploring the extent to which network structure and mitigation strategies may impact the outcomes of disease spread. We created synthetic metapopulations by linking discrete local populations according to three well known connectivity structures: random, scale-free and small world networks (13) . We then modelled temporary mitigation on 95 metapopulations as reduced transmission rates (the likely outcome of physical distancing measures) within certain time windows for the entire metapopulation. We additionally explored the effects of more selective mitigation strategies, in which transmission rates are only reduced for subsets of individuals assumed to be at high risk of severe disease effects (see methods). On each of the three metapopulation structures, we ran numerical simulations of an individual-100 based SEIR model (see methods) across a sensible range of parameter values of (i) transmission rates (), (ii) network connectivity (C), and individual commuter travel to (iii) neighbouring or (iv) distant populations ( and , respectively). We explored scenarios of lowering transmission rates () between 10-90% over time windows of 21-300 days (), which we believe to 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 April 29, 2020. . https://doi.org/10.1101/2020.04.22.20075093 doi: medRxiv preprint represent a range of possible mitigation strategies to prevent COVID-19 spread. This allowed 105 us to compare the outcomes of different mitigation strategies in terms of relative epidemic sizes (the difference in epidemic sizes resulting from different mitigation strategies applied over the same transmission scenario) and spatial spread for different network types and mitigation strategies: (1) no mitigation, (2) whole-population mitigation, and (3) selective mitigation applied to vulnerable individuals only. Across all scenarios, we found strong support for scale-free connectivity structures among local populations being likely to minimize disease spread. Under no mitigation, total epidemic sizes were limited to <66% of the metapopulation infected and <30% local populations affected in 115 scale-free networks (Fig. 1) . Epidemic sizes and fractions of affected populations varied greatly among different transmission scenarios in random and small world networks, ranging from rapid disease extirpation with small epidemic sizes to spread into entire metapopulations (Fig. 1 ). This variation is mostly explained by differences in transmission rates  for all network types and connectivity C in scale-free networks (Table S2) . While constantly low transmission rates (and hence low basic reproductive numbers R0) are the most crucial aspect in containing epidemic sizes and disease spread, we found rather intriguing and counterintuitive outcomes when temporary mitigation strategies are enforced. Epidemic size was smaller without any mitigation strategies in 29% of all transmission scenarios compared to scenarios where whole-population or selective mitigation strategies are enforced. In particular, no mitigation was likely to results in smaller relative epidemic sizes for transmission scenarios with large transmission rates  and weak containment efforts  (Table 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 April 29, 2020. . https://doi.org/10.1101/2020.04.22.20075093 doi: medRxiv preprint S3). In contrast, relative epidemic size was smallest in 41% of transmission scenarios with 130 whole-populations mitigation policy, especially for transmission scenarios with small , high commuter travel  and strong containment efforts  (Table S3) . Selective mitigation strategies, in turn, resulted in the smallest relative epidemic sizes and the lowest numbers of vulnerable individuals infected in transmission scenarios with large  in combination with weak containment efforts  and in scale-free networks (Table S3) . 135 Notably, the fraction of populations affected increased significantly, with relative larger epidemic sizes, when mitigation strategies are implemented, compared to scenarios without mitigation, regardless of transmission rates (Fig. 2, Table S4 ). This indicates that mitigation policies may come at the cost of larger epidemic sizes if they do not contain disease within affected populations. As expected, epidemic peak size (relative to total epidemic size) was lowered with wholepopulation mitigation strategies, while increasing considerably for larger transmission rates  (Fig. 3, Table S5 ). There was however no apparent advantage of whole-population versus selective mitigation on epidemic peak sizes for vulnerable individuals only (Table S5 ). This highlights the crucial roles of transmission rates in driving the extents of epidemic peak sizes. Low transmission rates are thus key, along with strong short-term containment efforts, to effectively avoid overwhelming health care systems. Our results highlight that strict short-term disease containment actions designed to halt the 150 rapid increase in cases during peaks of local COVID-19, may have idiosyncratic outcomes and do not necessarily guarantee optimal protection of global populations over larger spatiotemporal scales. The modelled variation in underlying transmission scenarios may well represent the diversity of conditions and uncertainty we encounter in real worlds, resulted in a 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 April 29, 2020. Efficient mitigation strategies are urgently needed to contain the spread of SARS-CoV-2 and save lives, but temporarily restricted intensive mitigation strategies (e.g. strict short-term containment and mass quarantine) in response to overloaded health care systems cannot replace 175 efficient long-term mitigation strategies such as rigorous physical distancing or revisiting contact structure at landscape scale according to our results. Our results suggest that the ease or entire uplifting of existing intensive short-term containment efforts, such as mass quarantine, can be dangerous and may result in unprecedented disease spread. At the same time, achieving 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 April 29, 2020. parameter values explored highlights the importance of accounting for regional conditions in real-world decision making. Both local human settlement density and connectivity structure, along with travel and social behaviour, which together set the stage for disease spread, differ widely across the globe. Informed decision making for rewiring interpersonal contacts or measures aimed at lowering transmission rates thus require taking regional conditions and 190 feasibility into account. This knowledge will help policy makers planning adequate strategies to prevent the overwhelming of health care systems worldwide, and, ultimately, to protect lives. 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 April 29, 2020. . https://doi.org/10.1101/2020.04.22.20075093 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 April 29, 2020. . https://doi.org/10.1101/2020.04.22.20075093 doi: medRxiv preprint Acknowledgments: 270 We thank our families for support while conducting this research in home offices. Funding: Funded by Swansea University Research Lectureships to K.W. and M.L. Author contributions: K.W. conceived the study and model, K.W. and M.L. analysed the data and wrote the paper. Competing interests: The authors declare no competing interests. Data and materials availability: All code, and materials used in the analysis are available as 275 Supplementary Materials. Figs. S1 to S3 280 Tables S1 to S5 285 290 295 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. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (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 April 29, 2020. (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 April 29, 2020. . https://doi.org/10.1101/2020.04.22.20075093 doi: medRxiv preprint World Health Organization Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study. The Lancet How will country-based mitigation measures influence the course of the COVID-19 epidemic? 210 The Lancet Effective containment explains subexponential growth in recent confirmed COVID-19 cases in China Herd immunity'': a rough guide Quantifying SARS-CoV-2 transmission suggests epidemic control with digital contact tracing Social encounter networks: characterizing Great Britain Modeling infectious disease dynamics in the complex landscape of global health Metapopulation dynamics of bubonic plague Optimal control of epidemics in metapopulations Modeling Infectious Diseases in Humans and Animals Feasibility of controlling COVID-19 outbreaks by isolation of cases and contacts. The Lancet Global Health The structure and function of complex networks Epidemic modeling in metapopulation systems with heterogeneous coupling pattern: theory and simulations Social encounter networks: collective properties and disease transmission The igraph software package for complex network research. InterJournal, Complex Systems. 1695 Large sample properties of simulations using latin hypercube sampling R Development Core Team. (R Foundation for Statistical Computing The incubation period of coronavirus disease 2019 (COVID-19) from 250 publicly reported confirmed cases: estimation and application Novel coronavirus 2019-nCoV: early estimation of epidemiological parameters and epidemic predictions. medRxiv