Interventions targeting nonsymptomatic cases can be important to prevent local outbreaks: COVID-19 as a case-study

Background During infectious disease epidemics, a key question is whether cases travelling to new locations will trigger local outbreaks. The risk of this occurring depends on a range of factors, such as the transmissibility of the pathogen, the susceptibility of the host population and, crucially, the effectiveness of local surveillance in detecting cases and preventing onward spread. For many pathogens, presymptomatic and/or asymptomatic (together referred to here as nonsymptomatic) transmission can occur, making effective surveillance challenging. In this study, using COVID-19 as a case-study, we show how the risk of local outbreaks can be assessed when nonsymptomatic transmission can occur. Methods We construct a branching process model that includes nonsymptomatic transmission, and explore the effects of interventions targeting nonsymptomatic or symptomatic hosts when surveillance resources are limited. Specifically, we consider whether the greatest reductions in local outbreak risks are achieved by increasing surveillance and control targeting nonsymptomatic or symptomatic cases, or a combination of both. Findings Seeking to increase surveillance of symptomatic hosts alone is typically not the optimal strategy for reducing outbreak risks. Adopting a strategy that combines an enhancement of surveillance of symptomatic cases with efforts to find and isolate nonsymptomatic hosts leads to the largest reduction in the probability that imported cases will initiate a local outbreak. Interpretation During epidemics of COVID-19 and other infectious diseases, effective surveillance for nonsymptomatic hosts can be crucial to prevent local outbreaks. Funding UKRI-BBSRC, Wellcome Trust, UKRI-MRC, FCDO, EDCTP2, Christ Church (Oxford).


Interpretation 48
During epidemics of COVID-19 and other infectious diseases, effective surveillance for 49 nonsymptomatic hosts can be crucial to prevent local outbreaks. analytically using a branching process model and found that effective isolation of infectious 67 hosts leads to a substantial reduction in the outbreak risk. 68 69 . CC-BY-NC-ND 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(which was not certified by peer review)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 November 7, 2020. ; https://doi. org/10.1101org/10. /2020 to transmit a pathogen while not showing symptoms. For COVID-19, the incubation period has 71 been estimated to last approximately five or six days on average, 9,10 and presymptomatic 72 transmission can occur during that period. [11][12][13][14] Additionally, asymptomatic infected individuals 73 (those who never develop symptoms) are also thought to contribute to transmission. 11,15,16 74 75 Motivated by the need to assess the risk of outbreaks outside China early in the COVID-19 76 pandemic, we show how the risk that imported cases will lead to local outbreaks can be 77 estimated using a branching process model, including nonsymptomatic individuals in the model 78 explicitly. We explore the effects of interventions that aim to reduce this risk. Under the 79 assumption that detected infected hosts are isolated effectively, we consider whether it is most 80 effective to dedicate resources to enhancing surveillance targeting symptomatic individuals, to 81 instead focus on increasing surveillance for nonsymptomatic individuals, or to use a combination 82 of these approaches. 83 84 We show that the maximum reduction in outbreak risk almost always corresponds to a mixed 85 7 138 CC-BY-NC-ND 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 November 7, 2020. ; https://doi. org/10.1101org/10. /2020  Since this research was motivated by the need to estimate outbreak risks outside China in the 167 initial stages of the COVID-19 pandemic, we used a baseline set of parameter values in our 168 analyses that was informed by studies conducted during this pandemic (Table 1). Where 169 possible, these parameter values were obtained from existing literature. However, we also 170 . CC-BY-NC-ND 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 November 7, 2020. Relative isolation rate of nonsymptomatic individuals without intensified surveillance (compared to symptomatic individuals) = 0 · 1 Assumed (for different values, see Figure S7) Rate at which presymptomatic individuals develop symptoms = 0 · 5 days $"

22-24
. CC-BY-NC-ND 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 November 7, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 Recovery rate of asymptomatic individuals = 0 · 1 days $" Chosen so that, in the absence of interventions, the expected duration of infection is identical for all infected hosts ( Upper bound on the fractional reduction in the time to isolation (if no other event occurs) = 0 · 8 Assumed (for different values, see Figure S11) " Surveillance intensification effort targeted at nonsymptomatic hosts " allowed to vary in the range [0,20] N/A -range of values explored # Surveillance intensification effort targeted at symptomatic hosts # allowed to vary in the range [0,20] N/A -range of values explored 185 186 2.4 Probability of a local outbreak 187 The probability that a single imported infectious host initiates a local outbreak was calculated 188 analytically using the branching process model. 189

190
The probability of a local outbreak not occurring, starting from presymptomatic hosts, 191 symptomatic hosts, and asymptomatic hosts was denoted by +,-,. . Starting from one 192 presymptomatic host (so that = 1 and = = 0), there are four possibilities for the next event. ), or; iv) be isolated (with probability ). Consequently, 196 . CC-BY-NC-ND 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 November 7, 2020. occurring beginning from a single presymptomatic host is given by 222 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10.1101/2020.11.06.20226969 doi: medRxiv preprint The funders had no role in study design, preparation of the manuscript or the decision to publish. 234 We considered the effect of & and the duration of the presymptomatic and asymptomatic 238 periods on the probability of a local outbreak when a nonsymptomatic host enters a new host 239 population ( Figure 2). We considered presymptomatic periods of length 1/ = 1 day, 1/ = 2 240 days and 1/ = 4 days; in each case, the duration of the asymptomatic period (1/ days) was 241 adjusted so that the relative proportion of infections arising from asymptomatic hosts compared 242 to presymptomatic hosts remained fixed ( ( / ' = 0 · 218, as in the baseline case). If instead 243 nonsymptomatic infections are not accounted for, the infectious period follows an exponential 244 distribution and the probability of a local outbreak is given by = 1 − 1/ & (red dash-dotted 245 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10.1101/2020.11.06.20226969 doi: medRxiv preprint increased risk of a local outbreak in the absence of surveillance intensification ( Figure 2A). 247

248
We then considered the dependence of the probability of a local outbreak on the intensity of 249 surveillance targeting nonsymptomatic and symptomatic hosts ( Figure 2B-D). The maximum 250 value of the surveillance intensification effort that we considered (given by ! or " values of 251 20) corresponded to a 76% reduction in the expected time to isolation (blue line in Figure 1B The length of the presymptomatic and asymptomatic periods significantly affected the 255 dependence of the probability of a local outbreak on the level of surveillance targeted at 256 nonsymptomatic and symptomatic hosts. In Figure 2B, in which the duration of the 257 presymptomatic period was 1 day, increasing surveillance targeted at nonsymptomatic hosts ( ! ) 258 had a limited effect on the probability of a local outbreak, while increasing surveillance targeted 259 at symptomatic hosts ( " ) had a more significant effect. For example, increasing the surveillance 260 effort targeted at nonsymptomatic hosts to ! = 5 (a 67% reduction in the time to isolation) only 261 reduced the probability of a local outbreak from 0·730 to 0·716 (3 s.f.), whereas the equivalent 262 effort targeted at symptomatic hosts ( " = 5) reduced the probability to 0·630 (3 s.f.). As shown 263 in Figures 3C and D, however, when the presymptomatic and asymptomatic periods were longer, 264 the benefit of directing surveillance resources towards detecting nonsymptomatic individuals 265 increased. This was because longer presymptomatic and asymptomatic periods increased the 266 proportion of infections generated by nonsymptomatic individuals ( ' + ( , see equations (1)  267 . CC-BY-NC-ND 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.

286
. CC-BY-NC-ND 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.

Optimising surveillance enhancement 287
We next considered in more detail the impact of surveillance targeted at nonsymptomatic hosts 288 ( ! ) relative to the impact of surveillance targeted at symptomatic hosts ( " ). For our baseline 289 parameter values, we considered the probability of a local outbreak starting from a single 290 imported nonsymptomatic individual for a range of values of ! and " . We calculated the 291 steepest descent contours (white lines in Figure 3A) numerically using a gradient maximisation 292 approach, in which at each point the contour direction was determined by minimising the local 293 outbreak probability over a fixed search radius (see Supplementary Material). These contours 294 indicate how ! and " should be altered to maximise the reduction in the probability of a local 295 outbreak. In this case, enhancing surveillance targeting both symptomatic and nonsymptomatic 296 hosts is always optimal (the steepest descent contours are neither horizontal nor vertical). 297

298
We then considered a scenario in which, at any time, it is only possible to direct resources 299 towards enhancing surveillance of either nonsymptomatic individuals or symptomatic 300 individuals (e.g. tracing and testing of nonsymptomatic contacts of known infectious individuals, 301 or screening for symptomatic individuals at public events). In Figure 3B, the blue region 302 represents values of ! and " for which enhancing surveillance targeting symptomatic hosts (i.e. 303 increasing " ) leads to a larger reduction in the local outbreak probability than enhancing 304 surveillance targeting nonsymptomatic hosts (i.e. increasing ! ). In contrast, in the green region, 305 enhancing surveillance of nonsymptomatic individuals is more effective than enhancing 306 surveillance of symptomatic individuals. The white line represents the steepest descent contour 307 starting from ! = " = 0, under the constraint that surveillance can only be enhanced for 308 symptomatic or nonsymptomatic hosts at any time. 309 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10. 1101/2020 Practical deployment of surveillance is often subject to logistical constraints, and policy-makers 311 may wish to design surveillance strategies to achieve a specific objective -for example, to 312 maximise the effectiveness of limited resources or to minimise the cost of achieving a desired 313 outcome. We therefore also considered the following two examples of such objectives. 314

315
Objective 1: Minimise the probability of a local outbreak for a fixed total surveillance effort. 316 First, we considered the question: given a fixed maximum surveillance effort ( ! + " = ), 317 how should surveillance be targeted at nonsymptomatic and symptomatic hosts? This involves 318 setting the values of ! and " to minimise the local outbreak probability. The optimal strategies 319 in this case are shown in Figure 3C. The red dotted lines represent contours along which the total 320 surveillance effort ! + " is held constant (i.e. different values of ). On each contour, the red 321 circle indicates the point at which the local outbreak probability is minimised. 322 323 If surveillance resources are increased (i.e. increases), a further question is how surveillance 324 should then be increased. In Figure 3C, the white line represents the contour of steepest descent, 325 under the constraint that the total change in surveillance effort ( ! + " ) is held constant at each 326 step (rather than a constant search radius, as in Figure 3A -for more details, see the 327 Supplementary Material). This contour coincides exactly with that shown in Figure 3B. 328

329
These results indicate that, if surveillance resources are such that is greater than 2·8 330 (corresponding to a 59% reduction in time to isolation of symptomatic hosts), the optimal 331 surveillance strategy involves both enhanced surveillance of symptomatic individuals and 332 . CC-BY-NC-ND 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.

335
Objective 2: Minimise the total surveillance effort to achieve a pre-specified reduction in the 336 probability of a local outbreak 337 Second, we considered the question: given a pre-specified acceptable risk level (i.e. probability 338 of a local outbreak), how should the surveillance level targeted at nonsymptomatic and 339 symptomatic hosts be chosen? This involves choosing ! and " to minimise ! + " along a 340 given contour corresponding to a fixed local outbreak probability (red dotted lines in Figure 3D). 341 On each contour, the red circle indicates the point along that contour at which the total 342 surveillance effort ! + " is minimised. These optimal points also lie exactly along the line on 343 which enhancing surveillance targeted at symptomatic hosts is equally effective compared to 344 enhancing surveillance targeted at nonsymptomatic hosts. 345 346 As long as the target local outbreak probability was less than 0.69, optimal surveillance involved 347 enhanced surveillance of nonsymptomatic individuals as well as symptomatic individuals. For 348 example, in order to reduce the local outbreak probability to 0.6, the optimal approach was to 349 deploy resources such that ! = 12 · 4 (a 74% reduction in time to isolation of nonsymptomatic 350 individuals) and " = 18 · 0 (a 76% reduction in time to isolation of symptomatic individuals). 351 352 Plots analogous to Figure 3D in which the parameters were varied from their baseline values are 353 shown in the Supplementary Material. In each case we considered, our main message was 354 unchanged. There always existed a threshold local outbreak probability such that, below this 355 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10.1101/2020.11.06.20226969 doi: medRxiv preprint threshold, the optimal strategy for further reduction in the local outbreak probability involved 356 enhancing surveillance targeting both nonsymptomatic and symptomatic individuals.  CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10.1101/2020.11.06.20226969 doi: medRxiv preprint effort required to achieve a pre-specific risk level (an "acceptable" local outbreak probability). Red dotted 375 lines indicate contours of constant local outbreak probability (i.e. lines on which the probability of a local 376 outbreak takes the values shown); red circles indicate the points along these contours at which the total 377 surveillance effort " + # is minimised. The white line indicates the optimal strategy to follow if the pre-378 specified risk level is increased or reduced.

DISCUSSION 381
A key component of infectious disease epidemic management is inferring the risk of outbreaks in 382 different locations. 3-6,25 Different surveillance and control strategies can be introduced to reduce 383 the risk that imported cases will lead to local outbreaks. surveillance should be targeted at nonsymptomatic or symptomatic hosts in order to reduce the 395 probability that cases imported to new locations will trigger a local outbreak ( Figure 3A,B). We 396 also showed how the optimal surveillance level targeting these two groups can be assessed when 397 surveillance resources are limited and policy-makers have specific objectives ( Figure 3C,D). In 398 each case, our main conclusion was that enhanced surveillance of nonsymptomatic hosts ( ! > 399 0) can be an important component of reducing the local outbreak risk during epidemics. 400 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10.1101/2020.11.06.20226969 doi: medRxiv preprint Our goal here was to use the simplest possible model to explore the effects of surveillance of 402 nonsymptomatic and symptomatic individuals on the risk of local outbreaks. However, for our 403 modelling approach to be used to make precise quantitative predictions during epidemics, it 404 would be necessary to update the model to include the range of different specific surveillance 405 and control interventions that are in place. Detection of nonsymptomatic individuals is facilitated 406 by contact tracing and testing, which are carried out routinely during epidemics and can be 407 included in models explicitly. 7,26,27 Reductions in contacts due to social distancing strategies and 408 school or workplace closures could also be accounted for, although such interventions are often 409 introduced after a local outbreak has begun rather than in the initial phase of a potential local 410 outbreak as considered here. We modelled the level of surveillance targeted at nonsymptomatic 411 and symptomatic hosts in a simple way, using a function describing the relationship between 412 surveillance effort and effectiveness ( Figure 1B). If different public health measures are included 413 in the model explicitly, then it would be possible to increase the accuracy of assessments of the 414 relative public health benefits of specific interventions that only target symptomatic individuals 415 (e.g. screening for passengers with heightened temperatures at airports) compared to 416 interventions that also target nonsymptomatic hosts (e.g. travel bans or quarantine of all inbound 417 passengers). Of course, this would require data from which the relative effectiveness of different 418 measures could be inferred. 419 420 In this article, we chose a baseline set of parameter values that were consistent with findings of 421 studies conducted during the COVID-19 pandemic, although constructing a detailed transmission 422 model for this pandemic was not our main focus. For example, we set the relative rates at which 423 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10.1101/2020.11.06.20226969 doi: medRxiv preprint symptomatic individuals so that 48·9% of transmissions arise from presymptomatic infectors, 425 and 10·6% arise from asymptomatic infectors. 11 While this is in line with other reported 426 estimates, 28,29 there is substantial variation between studies. We therefore also conducted 427 sensitivity analyses in which we explored a range of different values of model parameters 428 . CC-BY-NC-ND 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 November 7, 2020. CC-BY-NC-ND 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 November 7, 2020. prior to symptom onset. eLife 2020; 9: e57149. 492 . CC-BY-NC-ND 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.

17.
Li . CC-BY-NC-ND 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 November 7, 2020.

535
In the Methods section, we outlined an approach for deriving the probability of a local outbreak starting from a 536 single infectious host in either the presymptomatic, symptomatic or asymptomatic classes. Here, we provide 537 more details about that derivation. The probability of a local outbreak not occurring, starting from 538 presymptomatic hosts, symptomatic hosts and asymptomatic hosts, is denoted by +,-,. . If we consider the 539 . CC-BY-NC-ND 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.

564
[0, 2 ⁄ ] ( Figure S1A). In practice, this range was discretised into 33 search directions evenly spaced between 0 565 and /2. We then selected the pair of Δ 7 , Δ 8 values for which the local outbreak probability evaluated at 7 + 566 . CC-BY-NC-ND 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 November 7, 2020. ; https: //doi.org/10.1101//doi.org/10. /2020 Δ 8 ( Figure S1B). The white line in Figure 3B, which divides the region in which increasing 7 has a greater 568 effect on the local outbreak probability from the region in which increasing 8 has a greater effect on the local 569 outbreak probability, was computed in an analogous way, with the additional restriction that we only considered 570 the search directions = 0 and = /2 (i.e. intensifying only surveillance of nonsymptomatic or symptomatic 571 hosts).

573
In Figure 3C, the white line represents the contour of steepest descent under the constraint that the total change 574 in surveillance effort (Δ 7 + Δ 8 = ) is held constant at each step, rather than fixing the search radius 575 (Δ 7 ) 8 + (Δ 8 ) 8 = 8 as in Figure 3A. Therefore, instead of scanning over a circular arc, at each step we scan 576 along the line Δ 7 = , Δ 8 = − , where varies in the range [0, ] ( Figure S1D). Otherwise, the process is 577 completely analogous to that described above. 578 579 . CC-BY-NC-ND 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 November 7, 2020. ;https://doi.org/10.1101/2020 doi: medRxiv preprint Figure S1. Computation of the steepest descent contours shown in the main text. A. To compute the steepest 581 descent contours shown in Figure 3A of the main text, we increment 7 and 8 by scanning over a constant 582 search radius (Δ 7 ) 8 + (Δ 8 ) 8 = 8 (blue arc), and moving to the point ( 7 !"# , 8 !"# ) along that arc at which 583 the local outbreak probability is minimised. B. The process shown in A is repeated to generate the complete 584 contour (red dashed line). C. The analogous figure to B, in which the search direction is limited to directly to 585 the right ( = 0) or directly upwards ( = /2). This procedure is used to generate the contour in Figure   CC-BY-NC-ND 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.

593
For each of these, we present plots analogous to Figure 3D for six different values of the relevant parameter. In 594 each case we considered, our qualitative message was unchanged -for any reduction in the local outbreak 595 probability below a particular threshold value, the optimal strategy involved surveillance targeting both 596 nonsymptomatic and symptomatic individuals. 597 598 . CC-BY-NC-ND 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 November 7, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 599 Figure S2. Varying the basic reproduction number 0 from its baseline value ( 0 = 3). Plots are analogous to 600 Figure 3D, showing strategies for minimising the surveillance effort required to achieve a pre-specified risk 601 level (an "acceptable" local outbreak probability). Red dotted lines represent contours along which the 602 probability of a local outbreak is constant, as labelled; red circles indicate the points along these contours at 603 which the total surveillance effort 7 + 8 is minimised. The white line indicates the optimal strategy to follow 604 if the pre-specified risk level is reduced. Apart from 0 and (which is changed in each panel to set the value 605 of 0 ), all parameters are held fixed at their baseline values given in Table 1

608
. CC-BY-NC-ND 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 November 7, 2020. ;https://doi.org/10.1101https://doi.org/10. /2020 Figure S3. Varying the proportion of infections from asymptomatic infectors, , from its baseline value ( = 610 0.2). Plots are analogous to Figure 3D, showing strategies for minimising the surveillance effort required to 611 achieve a pre-specified risk level (an "acceptable" local outbreak probability). Red dotted lines represent 612 contours along which the probability of a local outbreak is constant, as labelled; red circles indicate the points 613 along these contours at which the total surveillance effort 7 + 8 is minimised. The white line indicates the 614 optimal strategy to follow if the pre-specified risk level is reduced. Apart from and (which is changed in 615 each panel to set 0 = 3), all parameters are held fixed at their baseline values given in Table 1. A. = 0. B.

618 619
. CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10.1101/2020.11.06.20226969 doi: medRxiv preprint 620 Figure S4. Varying the proportion of infections arising from presymptomatic hosts in the absence of intensified 621 surveillance ( 9 , given by expression (1) in the main text) from its baseline value ( 9 = 0.489). In each case, 622 the proportions of infections arising from asymptomatic and symptomatic hosts are adjusted so that they remain 623 in the same ratio as in the baseline case. Plots are analogous to Figure 3D, showing strategies for minimising the 624 surveillance effort required to achieve a pre-specified risk level (an "acceptable" local outbreak probability).

625
Red dotted lines represent contours along which the probability of a local outbreak is constant, as labelled; red 626 circles indicate the points along these contours at which the total surveillance effort 7 + 8 is minimised. The 627 white line indicates the optimal strategy to follow if the pre-specified risk level is reduced. Apart from 628 9 and : , as well as and (which are changed in each panel to set the values of 9 and : ), and (which is 629 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10.1101/2020.11.06.20226969 doi: medRxiv preprint 9 = 0 · 2. B. 9 = 0 · 3. C. 9 = 0 · 4. D. 9 = 0 · 5. E. 9 = 0 · 6. F. 9 = 0 · 7. 631 632 633 Figure S5. Varying the proportion of infections arising from asymptomatic hosts in the absence of intensified 634 surveillance ( : , given by expression (2) in the main text) from its baseline value ( : = 0.106). In each case, 635 the proportions of infections arising from presymptomatic and symptomatic hosts are adjusted so that they 636 remain in the same ratio as in the baseline case. Plots are analogous to Figure 3D, showing strategies for 637 minimising the surveillance effort required to achieve a pre-specified risk level (an "acceptable" local outbreak 638 probability). Red dotted lines represent contours along which the probability of a local outbreak is constant, as 639 labelled; red circles indicate the points along these contours at which the total surveillance effort 7 + 8 is 640 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10. 1101/2020 (which is changed in each panel to set 0 = 3), all parameters are held fixed at their baseline values given in 643 Table 1. A. : = 0 · 01. B. : = 0 · 05. C. : = 0 · 1. D. : = 0 · 15. E. : = 0 · 2. F. : = 0 · 25. 644 645 646 Figure S6. Varying the expected time period to isolation conditional on isolation occurring during the 647 symptomatic period, 1/( + ), from its baseline value (1/( + ) = 4.6 days). This is achieved by varying 648 the parameter , whilst holding the recovery rate of symptomatic individuals equal to its baseline value ( = 649 1/8 days ;7 ). Plots are analogous to Figure 3D, showing strategies for minimising the surveillance effort 650 required to achieve a pre-specified risk level (an "acceptable" local outbreak probability). Red dotted lines 651 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10.1101/2020.11.06.20226969 doi: medRxiv preprint changed in each panel to set 0 = 3), all parameters are held fixed at their baseline values given in Table 1

671
Plots are analogous to Figure 3D, showing strategies for minimising the surveillance effort required to achieve a 672 pre-specified risk level (an "acceptable" local outbreak probability). Red dotted lines represent contours along 673 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10. 1101/2020 to set 0 = 3), all parameters are held fixed at their baseline values given in Table 1  parameter is varied simultaneously such that 1/( + ), the expected time period to isolation conditional on 682 isolation occurring during the symptomatic period, remains at its baseline value (1/( + ) = 4.6 days). Plots 683 are analogous to Figure 3D, showing strategies for minimising the surveillance effort required to achieve a pre-

684
. CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10. 1101/2020 strategy to follow if the pre-specified risk level is reduced. Apart from and , and (which is changed in each 688 panel to set 0 = 3), all parameters are held fixed at their baseline values given in Table 1

693
Plots are analogous to Figure 3D, showing strategies for minimising the surveillance effort required to achieve a 694 pre-specified risk level (an "acceptable" local outbreak probability). Red dotted lines represent contours along 695 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10. 1101/2020 to set 0 = 3), all parameters are held fixed at their baseline values given in Table 1  occurs), , from its baseline value ( = 0.8). Plots are analogous to Figure 3D, showing strategies for 704 minimising the surveillance effort required to achieve a pre-specified risk level (an "acceptable" local outbreak 705 probability). Red dotted lines represent contours along which the probability of a local outbreak is constant, as 706 . CC-BY-NC-ND 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 November 7, 2020. ; https://doi.org/10. 1101/2020