key: cord-339898-ptb6dst8
authors: Bilinski, A.; Mostashari, F.; Salomon, J. A.
title: Contact tracing strategies for COVID-19 containment with attenuated physical distancing
date: 2020-05-08
journal: nan
DOI: 10.1101/2020.05.05.20091280
sha: 
doc_id: 339898
cord_uid: ptb6dst8

Contact tracing has been recommended as a critical component of containment strategies for COVID-19. We used a simple epidemic model to evaluate how contact tracing might enable partial relaxation of current physical distancing restrictions. Testing and tracing coverage need to exceed 50% in order for contact tracing to reduce transmission by at least 10%. With high isolation and quarantine efficacy, contact tracing could reduce overall transmission by >40%, which would allow for substantial loosening of physical distancing measures. Benefits of contact tracing could be enhanced by testing all contacts rather than only those with symptoms and by policies to support high adherence to voluntary isolation and quarantine.

As of May 1, the novel coronavirus SARS-CoV-2 has infected more than 3.2 million people worldwide and caused over 233,000 deaths. 1 Since March, most communities in the United States have been living under physical distancing measures including stay-at-home orders in 40 states. 2 Evidence suggests that these mitigation efforts have slowed the spread of the virus in many jurisdictions. 3, 4 A number of frameworks have been proposed for the safe relaxation of non-pharmaceutical interventions, and most include scaling up testing and contact tracing to support containment. [5] [6] [7] [8] [9] Guidelines for contact tracing call for identifying and monitoring individuals who have been in close contact with confirmed positive cases, facilitation of testing for symptomatic contacts, and counseling and follow-up to encourage voluntary self-isolation, quarantine and symptom monitoring. 10, 11 Several previous papers have considered the role of contact tracing for containment of COVID-19, [12] [13] [14] [15] but important questions remain about potential impact given uncertainty around the extent of presymptomatic and asymptomatic transmission of SARS-CoV-2 and the efficacy of voluntary isolation and quarantine. As decision-makers look toward relaxing current physical distancing measures, there is an urgent need to quantify the degree to which contact tracing programs could allow for partial loosening of restrictions while maintaining control over resurgent infection.

This paper uses a simple model to evaluate different contact tracing strategies to support modification of physical distancing restrictions. We examine the necessary conditions for maximizing benefits of contact tracing. We consider how broadening current testing guidelines from the Centers for Disease Control and Prevention to include testing for contacts without symptoms could amplify the impact of contact tracing programs.

We developed a simple deterministic Markov branching model of COVID-19 ( Figure S1 ). Epidemiological parameters were adapted from prior modeling studies where available (Table  S1 ). Infected individuals generate new infections based on whether they have symptoms, whether disease is detected, and whether they have been identified as a contact of an infected individual (Table S2 ). In our base case analysis we assumed that 40% of infections are asymptomatic, [16] [17] [18] [19] and that confirmed cases have 50% lower rates of transmission than unconfirmed cases; we considered alternatives to both of these assumptions in sensitivity analyses. We assumed that symptomatic cases become infectious prior to emergence of symptoms. 20,21

Estimates on the effectiveness of contact tracing vary considerably. 22,23 We modeled an array of different scenarios in order to characterize prerequisites for effective contact tracing, as well as to evaluate different possible policy priorities. We defined scenarios by the fraction of symptomatic cases detected in the community (not linked to a tracked case), the fraction of contacts successfully traced, the isolation and quarantine efficacy among traced but undetected contacts, and whether testing was restricted to those with symptoms or includes all traced contacts ( Table 1) . Given the likely importance of levels of community testing as a prerequisite condition for contact tracing, we conducted a secondary analysis that quantified the combined benefit of scaling up both testing and contact tracing against a counterfactual in which detection of symptomatic cases remains constant at an assumed current fraction of 20%.

We evaluated the impact of different contact tracing strategies in terms of the percentage reduction in the effective reproductive number Rt (average number of secondary infections produced by each infection) under each contact tracing scenario, compared to a scenario without contact tracing. Assuming that contact tracing strategies would be implemented alongside policy changes to partially relax physical distancing measures, and that the containment phase would begin when Rt was less than or equal to 1.0, reductions in Rt can be used to compute the containment margin for a given strategy. The containment margin signals how much current physical distancing measures could be relaxed in the presence of contact tracing, while maintaining Rt below the critical threshold of 1.0.

Base-case results are summarized in Figure 1 (see Figure S4 for details). Both community detection of symptomatic cases that are not linked to a tracked case and successful tracing of contacts needed to be at least 50% in order for contract tracing programs to reduce Rt by more than 10% compared to corresponding scenarios without contact tracing.

Testing asymptomatic contacts may substantially increase the impact of contact tracing strategies. Across all scenarios with adequate fractions (≥ 50%) of symptomatic cases detected in the community and contacts traced, testing asymptomatic contacts increased the benefit of contact tracing by a median factor of 1.5, with a range from 1.1 to 2.2. Benefits of asymptomatic testing were substantial in all scenarios except those in which efficacy of isolation and quarantine was already maximized for all contacts.

The overall impact of contact tracing depends strongly on isolation and quarantine efficacy. Median reductions in Rt assuming isolation and quarantine efficacy of 25%, 50% or 75% were 11%, 19% and 27% respectively, for strategies that tested only symptomatic contacts, and 20%, 26% and 31% for strategies that tested all contacts. The contact tracing scenario with the greatest impact overall-defined by high levels of symptomatic detection and successful tracing, high isolation and quarantine efficacy, and testing of all contacts irrespective of symptoms-reduced Rt by 46%.

In a sensitivity analysis (Figures S2/S5 ), we considered how the potential impact of contact tracing strategies might vary if the percentage of cases without symptoms was only 20% rather than 40%, as assumed in our base case. A lower asymptomatic fraction increased effectiveness for all scenarios, by a median factor of 1.2, and a range from 1.1 to 1.3. We also conducted secondary analyses to evaluate the combined effect of scaling up both testing and contact tracing against the counterfactual of persistently limited testing at 20% of symptomatic cases and no contact tracing (Figures S3/S6 ). Accounting for both expanded testing and contact tracing together, the maximum reduction in Rt increased to 57%, and the benefits in many scenarios were at least 10 percentage points greater than the benefits of contact tracing alone. When we further varied the relative transmission rate associated with detected compared to undetected cases from 50% (as in base case) to 25%, results were largely similar, with a maximum reduction of 55%, and higher gains at lower levels of isolation and quarantine efficacy.

To translate results into implications for potential modification of current policies, we used the percentage reductions in Rt from each contact tracing scenario to compute a corresponding containment margin, which indicates how much current physical distancing measures could be relaxed with contact tracing in place, while holding Rt below 1.0. As an example, assume that current physical distancing measures have reduced the reproductive number from R0 = 2.4 to Rt = 1.0, and that a contact tracing strategy could reduce Rt by 40%. Under these parameters, containment would be possible if relaxed physical distancing measures on their own could maintain Rt below 1.67, because the further reduction by a factor of 0.6 due to contact tracing would bring Rt below 1.0. This implies that together with contact tracing, physical distancing measures could be applied at 52% of their current, full implementation effectiveness and still maintain the critical containment threshold of Rt<1.

If Rt has been reduced to levels well below 1.0, the containment margin is greater (i.e. physical distancing measures could be further relaxed); if contact strategies are less effective the margin for loosening physical distancing shrinks. For example, if a contact tracing strategy were half as effective, producing a reduction in Rt of 20%, physical distancing measures could only be reduced to 82% of their current intensity. Further examples of containment margins under different assumptions about R0 and the benefits of contact tracing are provided in Table 2 .

In this study, we computed expected reductions in the effective reproductive number, Rt, under different contact tracing scenarios to quantify the degree to which contact tracing can allow for modification of public health orders and physical distancing restrictions while maintaining containment. To support containment, contact tracing must be implemented in concert with wide-scale community testing and must successfully track a high fraction of infected contacts.

Our results indicate that contact tracing will support a partial relaxation of physical distancing measures but not a complete return to levels of contact prior to physical distancing, consistent with prior studies. 14,15 For example, a recent paper estimated that adding contact tracing to self-isolation could reduce Rt by 35-47%, assuming 90% compliance, 16 which is similar to the ranges estimated in our analysis.

Testing of asymptomatic contacts would substantially magnify potential benefits by extending the coverage of tracing and potentially contributing to improved efficacy of isolation and quarantine. Another potential benefit of testing asymptomatic contacts, not captured in our model, is that negative test results could reduce the number of people needing to quarantine presumptively or could reduce the duration of quarantine, which might produce positive health and economic impacts. 24 For contact tracing to be most effective, broadening testing guidelines to include asymptomatic contacts will be important once testing capacity bottlenecks are addressed. The benefits of contact tracing also depend substantially on levels of adherence to isolation and quarantine among traced cases, which could be enhanced through policies such as providing voluntary out-of-home accommodations and income replacement.

Limitations of this analysis include a simplified modeling framework that lacks network or household structure, and also does not explicitly capture nursing homes, work places, or other potentially high-transmission venues. We furthermore did not model the impact of broader testing of asymptomatic individuals other than traced contacts, which would increase the coverage and impact of a containment strategy. Many uncertainties persist, including the extent of asymptomatic prevalence and transmission. Nevertheless, by examining a range of scenarios that reflect key sources of uncertainty and policy-relevant variables, we provide benchmarks that can aid in developing evidence-based containment strategies to minimize the risk of resurgent COVID-19 spread.

. CC-BY-NC 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 May 8, 2020. . . CC-BY-NC 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 May 8, 2020. . . CC-BY-NC 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 May 8, 2020. . CC-BY-NC 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 May 8, 2020. . CC-BY-NC 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 May 8, 2020. . https://doi.org/10.1101/2020.05.05.20091280 doi: medRxiv preprint Supplemental Information Figure S1 . Model structure and parameters. Parameter definitions: a is the fraction of infections that are asymptomatic; k is the fraction of infections that are detected; r is the number of secondary infections from each infection and p is the fraction of cases that are successfully contact traced. For parameters indexed by subscripts: T is contact traced, N is not contact traced; S is symptomatic, A is asymptomatic; D is detected, U is undetected.

. CC-BY-NC 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 May 8, 2020. 

Fraction of infections that are asymptomatic (a)

Estimates vary across studies. [16] [17] [18] [19] Alternative value of 20% examined in sensitivity analysis ( Figure S2) .

Symptomatic, not contact traced (kNS) varies Values shown in Table 1 Asymptomatic, not contact traced (kNA) 5% Assumed to be negligible based on current US testing guidelines 25

Symptomatic, contact traced (kTS) 90% Assumption, reflecting referral to testing for traced contacts Asymptomatic, contact traced (kTA) 90% Assumption, reflecting referral to testing. Applies only in contact tracing strategies that include testing for asymptomatics (see Table  1 ).

computed See Table S2 for details.

Values shown in Table 1 . Relative number of secondary infections from detected infections compared to undetected infections (q) 0.5 Limited empirical data, rationale for reduced secondary transmission includes: potentially increased likelihood of adherence to selfisolation, targeting of confirmed cases for public health support. 27 Alternative value of 1.0 examined in sensitivity analysis ( Figure  S3 ).

Calibrated Values calibrated to produce baseline Rt=1. Note that relative reductions in secondary infections across program comparisons are . CC-BY-NC 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 May 8, 2020. . https://doi.org/10.1101/2020.05.05.20091280 doi: medRxiv preprint scale invariant.

Values shown in Table 1 . Isolation and quarantine efficacy is approximately the product of how much infectious time remains when the contact is notified, and the degree of adherence to isolation and quarantine measures. Estimates of adherence have ranged considerably in previous studies (0-94%) 22 , including 70% 21 and 90% 14 in previous COVID-19 analyses.

Remaining infectious time is difficult to measure, but likely less than 1. 14,15 A prior modeling study used efficacy estimates of 25% for a 'low-feasibility setting' and 75% for a 'high-feasibility setting.' 12 . CC-BY-NC 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 May 8, 2020. 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 May 8, 2020. . Figure S2 . Results from sensitivity analysis assuming 20% of cases are asymptomatic. The horizontal axis shows the fraction of symptomatic cases that are detected in the community. The vertical axis shows the primary measure of strategy effectiveness: the percentage reduction in Rt in the contact tracing scenario relative to Rt without contract tracing. The color of the lines within each panel indicate the fraction of contacts that are successfully traced. 'Isolation and quarantine efficacy' refers to the level of reduction in transmission rates from traced, undetected contacts.

. CC-BY-NC 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 May 8, 2020. . Figure S3 . Results from secondary analysis on combined effects of scaling testing and contact tracing. The horizontal axis shows the fraction of symptomatic cases that are detected in the community. The vertical axis shows the primary measure of strategy effectiveness: the percentage reduction in Rt in the increased testing plus contact tracing scenario relative to Rt without increased testing or contract tracing. The color of the lines within each panel indicate the fraction of contacts that are successfully traced. 'Isolation and quarantine efficacy' refers to the level of reduction in transmission rates from traced, undetected contacts. The first two columns show results that maintain the base-case assumption that detected infections produce 50% as many secondary infections as undetected infections in the same category (presymptomatic, symptomatic, asymptomatic). The second two columns show results for an alternative in which detected cases produce 75% as many secondary infections as undetected infections.

. CC-BY-NC 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 May 8, 2020. . Figure S4 . Heatmap representation of Figure 1 . The horizontal axis shows the fraction of symptomatic cases that are detected in the community. The vertical axis shows the contacts that are successfully traced. The values and shading in each cell indicate the percentage reduction in Rt in the contact tracing scenario relative to Rt with no contract tracing. 'Isolation and quarantine efficacy' refers to the level of reduction in transmission among traced, undetected contacts.

. CC-BY-NC 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 May 8, 2020. . https://doi.org/10.1101/2020.05.05.20091280 doi: medRxiv preprint Figure S5 . Heatmap representation of Figure S2 : sensitivity analysis assuming 20% of cases are asymptomatic. The horizontal axis shows the fraction of symptomatic cases that are detected in the community. The vertical axis shows the contacts that are successfully traced. The values and shading in each cell indicate the percentage reduction in Rt in the contact tracing scenario relative to Rt with no contract tracing. 'Isolation and quarantine efficacy' refers to the level of reduction in transmission among traced, undetected contacts.

. CC-BY-NC 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 May 8, 2020. . https://doi.org/10.1101/2020.05.05.20091280 doi: medRxiv preprint Figure S7 . Heatmap representation of Figure S3 : secondary analysis on combined effects of scaling testing and contact tracing. The horizontal axis shows the fraction of symptomatic cases that are detected in the community. The vertical axis shows the contacts that are successfully traced. The values and shading in each cell indicate the percentage reduction in Rt in the increased testing plus contact tracing scenario relative to Rt without increased testing or contract tracing. 'Isolation and quarantine efficacy' refers to the level of reduction in transmission among traced, undetected contacts. The first two columns show results that maintain the base-case assumption that detected infections produce 50% as many secondary infections as undetected infections in the same category (presymptomatic, symptomatic, asymptomatic). The second two columns show results for an alternative in which detected cases produce 75% as many secondary infections as undetected infections.

. CC-BY-NC 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 May 8, 2020. . https://doi.org/10.1101/2020.05.05.20091280 doi: medRxiv preprint

Coronavirus Map: Tracking the global outbreak. The New York Times

Incidence, clinical outcomes, and transmission dynamics of hospitalized 2019 coronavirus disease among 9,596,321 individuals residing in California and Washington, United States: a prospective cohort study

The effect of control strategies to reduce social mixing on outcomes of the COVID-19 epidemic in Wuhan, China: a modelling study. The Lancet Public Health

COVID-19 US state policy database

Roadmap to Pandemic Resilience

National coronavirus response: A road map to reopening

Resource estimation for contact tracing, quarantine and monitoring activities for COVID-19 cases in the EU/EEA

Comparative Impact of Individual Quarantine vs. Active Monitoring of Contacts for the Mitigation of COVID-19: a modelling study

Feasibility of controlling COVID-19 outbreaks by isolation of cases and contacts. The Lancet Global Health

Effectiveness of isolation, testing, contact tracing and physical distancing on reducing transmission of SARS-CoV-2 in different settings

Quantifying SARS-CoV-2 transmission suggests epidemic control with digital contact tracing

Getting a handle on asymptomatic SARS-CoV-2 infection

Spread of SARS-CoV-2 in the Icelandic Population