US-UK Collab Linking models and policy: Using active adaptive management for optimal control of disease outbreaks.

Lead Research Organisation: University of Warwick
Department Name: Mathematics

Abstract

In the event of an outbreak of an infectious disease, management strategies to control further spread of infection are generally implemented based either upon strategies employed during previous epidemics or a pre-conceived expectation of the likelihood of success. However, at the onset of an outbreak, there is a great amount of uncertainty regarding the epidemiological properties of the disease and this may have a significant influence upon the ability of the chosen management strategy to contain or control the epidemic. Mathematical models can be developed to simulate spread of disease and evaluate the effectiveness of potential control strategies. However, the effectiveness of these models may be restricted by our limited knowledge of the epidemic as it unfolds.
Extensive analyses of the 2001 Foot and Mouth (FMD) outbreak in the UK have provided valuable information about both the dynamics of disease spread and the implementation of management actions. However, those observations are specific both to the UK setting and to the strain of. A future outbreak in the UK or an outbreak in another country such as the US will not necessarily follow the same pattern. Thus, key aspects of disease spread, and the optimal response, cannot be resolved until an outbreak occurs. Adaptive management (AM) seeks to address this limitation by incorporating monitoring, evaluation, and response into management actions such that management strategies can be modified and updated in response to improved understanding of the outbreak dynamics. The AM framework has previously been applied in conservation management but is yet to be applied to the management of infectious diseases.
AM provides a framework for switching from the early strategy that optimises the average outcome (when uncertainties are yet to be resolved), to the one that optimises the outcome for the specific model (or models) that best matches by the outbreak at hand. Additionally, active adaptive management seeks to make this switch as soon as possible, by initially using sub-optimal controls that allow the specific model to be identified as soon as possible. Thus, early management actions can be used to improve knowledge of the dynamics and more rapidly transition to the strategy that maximizes the global objective.
Although we are interested in the general application of AM to a range of outbreak scenarios; in this project we will use the 2001 FMD epidemic as a detailed, well-defined example. Despite a decade of modelling efforts, key uncertainties concerning optimal control remain, AM will allow us to address these issues. In particular we propose to:
1. Use the observed surveillance from the 2001 outbreak to identify the optimal adaptive strategy and the economic benefit of that strategy relative to a static (fixed) strategy.
2. Simulate the use of active AM to discriminate amongst competing models and selection of the optimal strategy. To that end we will consider the application of management strategies to facilitate learning and rapid updating of control policies.
3. Use AM to determine optimal management strategies for other disease scenarios, helping to generate a more generic understanding.
4. Using the FMD case-study developed in 1 and 2, we will support workshops that engage members of the US and UK policy community in the use of adaptive management for an outbreak.
5. Based on the understanding gained in the workshops, we will develop a US-based outbreak case-study that will be used as the subject of training workshops in the second half of the grant period. This case study would demonstrate the utility of AM in a scenario of extreme uncertainty.
The outputs of this project would elucidate the ability of AM to provide efficient policy advice in the event of future unknown outbreaks of infectious disease. A single, flexible policy that is able to adapt to the observed outbreak would have massive implications in reducing the impact of future outbreaks.

Technical Summary

In the event of an outbreak of infectious disease, it is necessary to make timely, and potentially critical decisions in the face of uncertainty. Detailed models of epidemic dynamics can be valuable in exploring outbreak scenarios and evaluating candidate strategies but are fundamentally limited by our lack of knowledge of the epidemic system before an outbreak has occured. Monitoring the response of epidemic dynamics to management interventions can often reduce the uncertainty encapsulated in competing models. In practice, however, such evaluation and assessment of competing models is often done in retrospect, and is thus of little use to the application of management in real-time. Large scale spatial epidemics require immediate management action. However, careful monitoring of the outbreak response to management allows for the opportunity to learn about the epidemiology of the epidemic and to modify management actions accordingly.
In this project we propose a structured decision-making framework for evaluating and updating management policies in the light of competing, dynamic epidemic models. The key to this adaptive management approach is linking predictive quantitative modelling with sequential evaluation of management objectives. We will develop this framework in the context of a very specific agricultural health setting, Foot and Mouth Disease (FMD). The 2001 UK FMD epidemic provides a comprehensive data set that will allow us to implement detailed analyses of optimal control in the presence of uncertainty. However, we will also use generic, theoretical models of spatial epidemic dynamics to study more generally the impact of uncertainty on the outbreak response and the ability to develop responsive control strategies. In doing so, we will develop algorithms and software to study generic questions about optimal response to spatial epidemics that can be applied to agricultural, veterinary-health (and human health) management in a variety of settings.

Planned Impact

In early February 2001, Foot-and-Mouth Disease (FMD) entered the United Kingdom. The subsequent epidemic lasted until the beginning of October and caused huge devastation, not only to the UK livestock industry but for the export market, the tourist trade and the economy in general. When the first case was reported, movements of all animals between farms ceased, whilst all livestock exports where halted. This resulted in huge economic losses for both farmers and the export market. Similarly, access to the countryside was severely restricted, which limited the tourist trade. Many businesses which relied heavily on tourism struggled to survive and the UK government found itself facing a potentially huge economic crisis.

In the early stages of the epidemic, the UK government consulted members of the mathematical modelling community for advice regarding the optimal strategy to control the epidemic. Since 2001, policy makers from around the world have developed their own contingency plans for control of FMD, incorporating control policies developed with the assistance of mathematical modellers with experience from the 2001 epidemic. Whilst culling strategies were modified in 2001 in response to the perceived failure of strategies introduced at the onset of the epidemic, there was little attempt to learn from observations of early epidemic behaviour and modify management strategies in a rigorous way.

This project has huge benefit to a wide range of groups. Mathematical modelling of infectious diseases is a rapidly growing field, but to date there has been little work done on rigorous optimisation of control policies in the event of epidemiological and demographic uncertainties. As such, this work will provide invaluable insights to the epidemiological and wider scientific community into the successful development of models for use with a wide range of diseases and and for a range of farm demographies.

In the event of any future outbreak of FMD in the US or UK livestock industry, policy makers would again seek advice from mathematical modellers regarding optimal control of disease. Prompt, efficient control can have a huge effect on reducing the overall loss of livestock and the economic cost of any epidemic and detailed numerical models are relied upon (in both the UK and elsewhere) to provide informed advice. This project therefore has the potential to provide a huge benefit to the US and UK governments and policy makers around the world.

As a consequence of the implementation of efficient strategies to control disease, there will be a subsequent reduction in the total number of farms affected, the duration of any epidemic and the spatial spread of disease. This research would therefore provide a huge benefit to the farming community, the tourist industry and the export market.

Publications

10 25 50

Related Projects

Project Reference Relationship Related To Start End Award Value
BB/K010972/1 07/01/2013 30/09/2013 £393,693
BB/K010972/2 Transfer BB/K010972/1 01/10/2013 31/07/2014 £358,244
BB/K010972/3 Transfer BB/K010972/2 01/08/2014 31/03/2016 £311,517
BB/K010972/4 Transfer BB/K010972/3 01/04/2016 31/12/2017 £193,973
 
Description The main aim of this grant is to develop models that can be utilised to advise policy makers during the early stages of disease outbreaks or when there is significant uncertainty regarding the potential effectiveness of control strategies. We have investigated the use of adaptive control policies that use currently available information to parameterise the models so that control policies can adapt to take into account the specific state of the outbreak.

Our work in the last 12 months has focused upon 3 key areas that I will outline below:

1. The effective use of vaccination for foot-and-mouth disease in the presence of uncertainty

Vaccination is a control policy that is regularly considered for foot-and-mouth disease outbreaks but there are uncertainties regarding the number of animals that can be vaccinated per day, the time delay from vaccination to immunity and the efficacy of the vaccine. These uncertainties may hamper policy makers' abilities to determine the most effective vaccination strategy. In our work, we have determined that, provided that uncertainty regarding the daily capacity to vaccinate can be resolved, significant cost savings can be made should vaccination be implemented, regardless of uncertainties in time delay and vaccine efficacy. This could allow vaccination to be implemented during the early stages of future outbreaks even when there is uncertainty. This work has been published in PLOS Computational Biology and has received some online coverage, including articles in agricultural press.

2. Implementation of optimal control policies as information is accrued during a foot-and-mouth disease (FMD) outbreak

During a future outbreak of FMD, there will be significant uncertainty regarding the spread of disease which means that any model must be parameterised to this new outbreak. However, in the early stages, this uncertainty will lead to uncertainty in parameters that may have an effect upon the model's ability to predict the best control policy. In this work, we have used data from the 2001 UK FMD outbreak and the 2010 Japan FMD outbreak and fitted our model on a weekly basis, to mimic what would happen in real time during a future outbreak. As we gain more information throughout the outbreaks, we can see that our model parameter predictions improve. Each week, we then simulate the outbreak with a range of control measures, to determine which control policy should be implemented based upon the information that we have. We can then compare this to the "true" best policy, which is determined by using parameter values obtained from the whole outbreak. We find that, in the early stages our model predictions of the size and spatial scale of an outbreak are poor and that these predictions improve through time. However, our model can robustly determine which is the most effective control strategy to implement, even as early as week 1, when there is huge uncertainty. This work has been presented at the EUFMD Open Session, the Animal and Plant Health Agency Modelling Symposium and at the EUFMD NTC 27 real time training workshop in Nakuru, Kenya and has received significant interest from DEFRA. The resultant manuscript is currently in revision at PLOS Computational Biology. 3. Real time decision making for Ebola outbreaks. In 2017, our group adapted the work on FMD to consider the effectiveness of intervention policies for the 2014 Ebola outbreak, utilising ensemble modelling and real time decision making. This work utilised a suite of existing ebola models and established optimal intervention strategies that should be implemented to reduce the impact of the disease. Our work concluded that reducing funeral and community transmission was key to minimising the number of human cases, whilst reducing transmission in hospitals had a lesser effect. This work has been published in Proceedings of the National Academy of Sciences.
Exploitation Route Our work on FMD will be highly informative to agricultural policy makers faced with new outbreaks of livestock disease. We have developed robust methodologies that will enable models to be used even in situations where uncertainty regarding future behaviour is significant. Whilst control for livestock diseases does typically evolve through time, decisions tend to be made retrospectively, taking into account the current situation and the evolution of the outbreak. Our work represents a significant technological advance on this, such that adaptive decisions can be deployed at outbreak onset, taking into account the current uncertainty and how that uncertainty will be resolved during an ongoing outbreak. This work is currently being applied to avian influenza outbreaks in South East Asia by a team led by Dr Tildesley and could be applied to other emerging outbreaks around the world. Our project team is also preparing a grant that builds upon this project, focused upon adaptive control of endemic foot-and-mouth disease - this research area will provide significant impact for some of the poorest countries in the world.
Sectors Agriculture, Food and Drink

 
Description The models developed during this project have been presented to policy makers at DEFRA, the Food and Agriculture Organisation of the United Nations and the US Department of Agriculture and we are in regular communication with these organisations to ensure that models such as the ones developed during this project can be used in the early stages of future livestock disease outbreaks. In addition, since 2017 we have started to adapt the models that have been developed in this grant for use in endemic countries, in collaboration with EuFMD. Through attendance at a real time training exercise in Nakuru, Kenya, in November 2017, Dr Tildesley has disseminated the results from this grant to policy makers and stakeholders working in countries affected by endemic FMD and the models are now being adapted, in collaboration with EuFMD to inform decision makers to reduce the impact in these countries. This will be the focus of an upcoming future grant application.
First Year Of Impact 2016
Sector Agriculture, Food and Drink
Impact Types Societal,Economic,Policy & public services

 
Description ESPRC Institutional Award (Global Challenges research)
Amount £36,993 (GBP)
Funding ID EP/P511079/1 
Organisation Engineering and Physical Sciences Research Council (EPSRC) 
Sector Academic/University
Country United Kingdom
Start 08/2016 
End 03/2017
 
Description Engineering and Physical Sciences Institutional Award: EP/R512916/1
Amount £29,979 (GBP)
Funding ID EP/R512916/1 
Organisation Engineering and Physical Sciences Research Council (EPSRC) 
Sector Academic/University
Country United Kingdom
Start 08/2017 
End 03/2018
 
Description University of Warwick Chancellor's Scholarship
Amount £100,000 (GBP)
Organisation University of Warwick 
Sector Academic/University
Country United Kingdom
Start 10/2016 
End 03/2020
 
Title Framework for adaptive disease control 
Description As a direct output of this project, we have developed a mathematical model that can be utilised in the early stages of disease outbreaks to determine the optimal multi-phase control policy to be implemented at the onset of a new outbreak when there is significant uncertainty regarding how an epidemic will spread. This has been demonstrated for foot-and-mouth disease through publications and is now being adapted to avian influenza and ebola. 
Type Of Material Computer model/algorithm 
Provided To Others? No  
Impact These adaptive frameworks have been presented at meetings involving policy makers from DEFRA in the UK, the US Department of Agriculture and the Food and Agriculture Organisation of the United Nations. These approaches have generated significant discussion and resulted in further communication with these agencies regarding optimal strategies for disease control. 
 
Description Participation and training of vets at real time training workshop organised by EuFMD in Nakuru, Kenya 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Professional Practitioners
Results and Impact In November 2017 I participated in a real time training workshop run by EuFMD in Nakuru, Kenya. The purpose of this workshop was to train vets in the identification of animals infected with foot-and-mouth disease and to determine risk factors for transmission of the disease. I attended both as a participant as an expert trainer in epidemiological modelling. I presented my work on adaptive control from this grant and educated the workshop participants in the role of models to determine strategies for disease control. In addition, I went into the field and was educated regarding the considerations for disease control for farmers in Kenya. The workshop was useful both in communicating the work from this grant and in educating me in how to translate the outputs of this grant to different demographic environments, which will be highly informative for planning future grants building upon this research.
Year(s) Of Engagement Activity 2017
 
Description Presentations at EUFMD Open Session 
Form Of Engagement Activity A talk or presentation
Part Of Official Scheme? No
Geographic Reach International
Primary Audience Policymakers/politicians
Results and Impact The Open Session of the EUFMD meeting is a two-yearly meeting attended by policy makers and stakeholders who work on foot-and-mouth disease (FMD) in both endemic and disease free countries. Dr Michael Tildesley and Dr Will Probert (the PDRA on this project) attended this meeting and presented their work on adaptive control strategies for FMD. The presentations were aimed at guiding policy makers regarding strategies for disease control during ongoing outbreaks when there is little information available in order to make control decisions. These presentations received interest from policy makers working on FMD control and discussions are ongoing regarding how models such as those from this work can be utilised during future outbreaks.
Year(s) Of Engagement Activity 2016
 
Description Public Science Evening at the University of Warwick 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Public/other audiences
Results and Impact I led a public science event on February 13th 2018 at the University of Warwick, to showcase epidemiological modelling to the general public. This event was attended by approximately 80 people from across the Coventry and Warwickshire region, including two school groups. I presented work on adaptive control that was developed from this grant and ran several interactive activities to engage the general public in the decision making process during disease outbreaks. The school groups in particular reported that the evening was very successful and that the science communication was pitched as exactly the right level for the audience.
Year(s) Of Engagement Activity 2018