key: cord-0794525-tv6z6238
authors: Cook, Jonathan D.; Campbell Grant, Evan H.; Coleman, Jeremy T. H.; Sleeman, Jonathan M.; Runge, Michael C.
title: Evaluating the risk of SARS-CoV-2 transmission to bats using a decision analytical framework
date: 2021-05-28
journal: bioRxiv
DOI: 10.1101/2021.05.28.446020
sha: 8748a7c2127b0db5b1b38f028db1124deb99597d
doc_id: 794525
cord_uid: tv6z6238

Preventing wildlife disease outbreaks is a priority issue for natural resource agencies, and management decisions can be urgent, especially in epidemic circumstances. With the emergence of SARS-CoV-2, wildlife agencies were concerned whether the activities they authorize might increase the risk of viral transmission from humans to North American bats but had a limited amount of time in which to make decisions. We provide a description of how decision analysis provides a powerful framework to analyze and re-analyze complex natural resource management problems as knowledge evolves. Coupled with expert judgment and avenues for the rapid release of information, risk assessment can provide timely scientific information for evolving decisions. In April 2020, the first rapid risk assessment was conducted to evaluate the risk of transmission of SARS-CoV-2 from humans to North American bats. Based on the best available information, and relying heavily on formal expert judgment, the risk assessment found a small possibility of transmission during summer work activities. Following that assessment, additional knowledge and data emerged, such as bat viral challenge studies, that further elucidated the risks of human-to-bat transmission and culminated in a second risk assessment in the fall of 2020. We update the first SARS-CoV-2 risk assessment with new estimates of little brown bat (Myotis lucifugus) susceptibility and new management alternatives, using findings from the prior two risk assessments and other empirical studies. We highlight the strengths of decision analysis and expert judgment not only to frame decisions and produce useful science in a timely manner, but also to serve as a framework to reassess risk as understanding improves. For SARS-CoV-2 risk, new knowledge led to an 88% decrease in the median number of bats estimated to be infected per 1000 encountered when compared to earlier results. The use of facemasks during, or a negative COVID-19 test prior to, bat encounters further reduced those risks. Using a combination of decision analysis, expert judgment, rapid risk assessment, and efficient modes of information distribution, we provide timely science support to decision makers for summer bat work in North America.

The RSM infection risk model was calculated from three encounter types: workers handling bats, 183 workers in proximity to bats in a shared enclosed space, and workers in close proximity to bats 184 but not in a shared enclosed space. The expected number of infected bats resulting from research, 185 survey, or monitoring activities is the sum of the expected number of bats infected through each 186 of the three encounter types: 187 is the probability that a bat in an enclosed space within a 6-foot proximity 207 of (but not handled by) a RSM scientist who was actively shedding 208 virus would be exposed to the virus (an "exposure probability") in the 209 absence of any new restrictions, regulations, or protocols; 210

is the probability that a bat not in an enclosed space within a 6-foot 211 proximity of (and not handled by) a RSM scientist who was actively 212

shedding virus would be exposed to the virus (an "exposure 213 probability") in the absence of any new restrictions, regulations, or 214 protocols; and 215 ߪ ௦ is the species-specific probability that a bat exposed to a sufficient viral 216 dose of SARS-CoV-2 would become infected by the virus (the 217 "probability of susceptibility" is the total number of bats exposed, not in an enclosed space or handled, 235 by wildlife rehabilitators during the 2021 active season; 236

is the probability that a bat handled by a WR who was actively shedding 237 virus would be exposed to the virus (an "exposure probability") in the 238 absence of any new restrictions, regulations, or protocols, taking into 239 account the handling time typical of rehab activities; 240

is the probability that a bat not in an enclosed space within a 6-foot 241 proximity of (and not handled by) a WR who was actively shedding 242 virus would be exposed to the virus (an "exposure probability") in the 243 absence of any new restrictions, regulations, or protocols; and 244 ߪ ௦ is the species-specific probability that a bat exposed to a sufficient viral 245 dose of SARS-CoV-2 would become infected by the virus (the 246 "probability of susceptibility"). 247 248

The WC infection risk model is calculated from two encounter types: bat handling, and workers 250 in proximity to bats but not in a shared enclosed space. The expected number of infected bats 251 arising from wildlife control operations over the summer season is the sum of the expected 252 number of bats infected through each of the two encounter types: 253 is the total number of bats exposed, but not handled, by WC during the 260 2021 active season; 261 ߚ ு ௐ is the probability that a bat handled by a WC who was actively shedding 262 virus would be exposed to the virus (an "exposure probability") in the 263 absence of any new restrictions, regulations, or protocols, taking into 264 account the handling time typical of WC activities; 265 ߚ ௐ is the probability that a bat not in an enclosed space within a 6-foot 266 proximity of (and not handled by) a WC who was actively shedding 267 virus would be exposed to the virus (an "exposure probability") in the 268 absence of any new restrictions, regulations, or protocols; and 269 ߪ ௦ is the species-specific probability that a bat exposed to a sufficient viral 270 dose of SARS-CoV-2 would become infected by the virus (the 271 "probability of susceptibility").

Probability that a crew member is positive and shedding virus 274

We calculated the probability that a crew member is positive and shedding virus as a function of 275 the prevalence of COVID-19 in the surrounding community (߰), and the sensitivity (ܵ݊) and 276 specificity ‫)ܵ(‬ of COVID-19 testing. Sensitivity is the probability that an individual who has 277 COVID-19 tests positive, whereas specificity is the probability that a healthy individual without 278 COVID-19 tests negative. We selected a sensitivity value of 0.70, and specificity of 0.95 279 (Arevalo-Rodriguez et al. 2020; Watson et al. 2020 ); however, we recognize that these values 280 vary according to the type of test administered. For our risk assessment, we are primarily 281 interested in the probability that a crew member receives a negative test result but is truly 282 infected with SARS-CoV-2. This probability can be calculated using Bayes' Theorem as: 283

If a crew member does not take a test, the probability that a crew member is positive and 287

shedding virus can be estimated by the local prevalence, ߰ , or by some other method that 288 accounts for the crew member's risk behavior (e.g., https://www.microcovid.org/). 289

To calculate the number of bats handled (H), encountered in an enclosed space (E), or in 291 proximity to workers in an unenclosed space (P), we multiplied the total number of bats 292 encountered in a typical season of work by the percentage of each bat encounter type (Table 1) available to directly inform little brown bat SARS-CoV-2 susceptibility. Instead, structured 468 a standard peer-review and publication process. While it is not our intention to criticize any 492 journal, reviewers, or peer-review process, we recognize that the production of decision-relevant 493 science using decision analysis, quantitative modeling, and undergoing a full peer-review 494 process may benefit from shorter timelines to provide information needed for urgent agency 495 decisions . 496

There are likely many options to improve the timely delivery of science to support urgent 498 wildlife disease management decisions moving forward, and we provide a few suggestions that 499 may be useful. First, it may be useful for journals to consider creating alternate production tracks 500 that can expedite the review and publication process and provide timely results at the speed of 501 agency decisions. Alternative options for distribution, such as preprint servers (like bioRxiv and 502 medRxiv) have already become critical avenues for timely release of information during the 503 COVID-19 pandemic; however, these avenues do not address the critical role that peer-review 504 plays in the production of reliable science. Second, for agencies that frequently make urgent 505 decisions and that currently rely only on published results to communicate scientific support for 506 those decisions to the public, it may be beneficial to consider using external science review 507 boards that can provide objective evaluations of unpublished findings for formal consideration in 508 time-sensitive decision-making. Lastly, dedicated risk assessment teams that produce rapid 509 qualitative assessments of wildlife disease risks within hours or days of an identified novel 510 hazard would be helpful. While other, more qualitative, assessments may be based on 511 preliminary results and limited knowledge that is subject to considerable change, they can be 512 effective as a bridge to more rigorous assessments that include agency consultation, quantitative 513 modeling, and the evaluation of management alternatives. Nevertheless, we hope that our risk 514 assessments may serve as a model to assess threats that SARS-CoV-2 continues to present to 515 wildlife, and that a larger discussion be stimulated to identify the best approaches to deliver 516 decision-relevant science for emerging wildlife diseases on timescales that matter. "breakthrough" infections to shed virus. As our knowledge continues to improve surrounding 534 SARS-CoV-2 and the risks it presents to bats, these factors, as well as other relevant information, 535 could be included in future assessments to ensure that agencies have the best available 536 information for making decisions. 537 ACKNOWLEDGMENTS The manuscript was improved with comments from reviewers. Any 538 use of trade, product, or firm names is for descriptive purposes only and does not imply 539 endorsement by the U.S. Government 

False-negative results of initial RT-PCR assays for 545 COVID-19: A systematic review

Prioritization of a livestock transboundary diseases in Belgium using a 548 multicriteria decision analysis tool based on drivers of emergence. Transboundary and 549 Emerging Diseases

The scope and severity of white-nose 556 syndrome on hibernating bats in North America

Decision making in natural resources management: A 559 structured adaptive approach

Risks 561 posed by SARS-CoV-2 to North American bats during winter fieldwork. Conservation 562 Science and Practice

Broad host range of SARS-CoV-2 predicted by comparative and structural analysis 567 of ACE2 in vertebrates

Testing the 570 efficacy of homemade masks: would they protect in an influenza pandemic?

An interactive web-based dashboard to track COVID-19 573 in real time

Experimental challenge of a North American bat 577 species, big brown bat (Eptesicus fuscus), with SARS CoV 2. Transboundary and 578 emerging diseases

2017. I nvestigate D iscuss E stimate A ggregate 581 for structured expert judgement

Field trial of a probiotic bacteria to protect bats 584 from white-nose syndrome

On information and sufficiency

Efficacy 588 of face shields against cough aerosol droplets from a cough simulator

Measurement of 591 filtration efficiencies of healthcare and consumer materials using modified respirator fit 592 tester setup

Evaluating the accuracy and 594 calibration of expert predictions under uncertainty: Predicting the outcomes of ecological 595 research

CoV-2: A review of viral, host, and environmental factors

Possibility for reverse zoonotic 601 transmission of SARS-CoV-2 to free-ranging wildlife: A case study of bats

The Advisory Committee on Immunization 604 Practices' Interim Recommendation for Use of Janssen COVID-19 Vaccine

R: A language and environment for statistical computing. R Foundation for 608 Statistical Computing

Assessing the 611 risks posed by SARS-CoV-2 in and via North American bats-Decision framing and 612 rapid risk assessment

SARS-CoV-2 in 616 fruit bats, ferrets, pigs, and chickens: an experimental transmission study. The Lancet 617 Microbe

Structured 619 decision making for managing pneumonia epizootics in bighorn sheep. The Journal of 620 Wildlife Management

622 Reducing overconfidence in the interval judgments of experts

Interim Estimates of 625 Vaccine Effectiveness of BNT162b2 and mRNA-1273 COVID-19 Vaccines

Preventing SARS-CoV-2 Infection Among Health Care Personnel, First Responders, and 627 Other Essential and Frontline Workers -Eight

Expert elicitation on the uncertainties associated 631 with chronic wasting disease

Interpreting a covid-19 test result

2020) worked with state and federal 704 decision makers to frame the decision and produce risk estimates that were useful to guide 705 management actions (Runge et al. 2020). Based on new knowledge and data that evaluated 706 critical uncertainties from Runge et al. (2020) (gray dashed arrows and central gray outlined 707 polygon), we revisited several steps (boxes outlined in blue) to rapidly re-evaluate the risk of 708 SARS-CoV-2 transmission during summer RSM, WC, and WR activities. Frequently updating 709 risk assessments using the best available science may help decision makers implement actions 710 that best achieve management objectives

Comparison of probability of susceptibility estimates for little brown bat from Runge 722 et al. (2020) (blue line) and from Cook et al. (2021) (black line)

RSM=research, survey, monitoring, and management activities; WR= 758 wildlife rehabilitation; WC= wildlife control operations. Boxplot whiskers represent 99% 759 prediction interval. For comparisons, we used the same assumed ratio of encounter modes 760 (handling, enclosure, and proximity) and probability of worker shedding SARS-CoV-2 (median: 761 0.057; 80% interval: 0.022-0.112) from Runge et al. (2020). (A) Results reproduced based on 762 expert-elicited data on probability of bat susceptibility from Runge

RSM=research, survey, monitoring, and management activities; WR= 780 wildlife rehabilitation; WC= wildlife control operations. Boxplot whiskers represent 99% 781 prediction interval. We used the same assumed ratio of encounter modes (handling, enclosure, 782 and proximity) from Runge et al. (2020). Results based on expert elicited data on probability of 783 bat susceptibility from the Cook et al. (2021) assessment. (A) Effectiveness of PPE compared 784 against baseline estimates for RSM activities. (B) Effectiveness of PPE compared against 785 baseline estimates for WR activities

Number of bats per 1,000 exposed to and infected by SARS-CoV-2 by the three 793 transmission pathways with and without pre-survey COVID-19 testing. RSM=research, survey, 794 monitoring, and management activities; WR= wildlife rehabilitation; WC= wildlife control 795 operations. Boxplot whiskers represent 99% prediction interval. We used the same assumed ratio 796 of encounter modes (handling, enclosure, and proximity) from Runge et al