Controlling bovine paratuberculosis at a regional scale: Towards a decision modelling tool

Leveraging artificial intelligence and software engineering methods in epidemiology for the co-creation of decision-support tools based on mechanistic models

Epidemiological modeling is a key lever for infectious disease control and prevention on farms. It makes it possible to understand the spread of pathogens, but also to compare intervention scenarios even in counterfactual situations. However, the actual capability of decision makers to use mechanistic models to support timely interventions is limited. This study demonstrates how artificial intelligence (AI) techniques can make mechanistic epidemiological models more accessible to farmers and veterinarians, and how to transform such models into user-friendly decision-support tools (DST). By leveraging knowledge representation methods, such as the textual formalization of model components through a domain-specific language (DSL), the co-design of mechanistic models and DST becomes more efficient and collaborative. This facilitates the integration of explicit expert knowledge and practical insights into the modeling process. Furthermore, the utilization of AI and software engineering enables the automation of web application generation based on existing mechanistic models. This automation simplifies the development of DST, as tool designers can focus on identifying users' needs and specifying expected features and meaningful presentations of outcomes, instead of wasting time in writing code to wrap models into web apps. To illustrate the practical application of this approach, we consider the example of Bovine Respiratory Disease (BRD), a tough challenge in fattening farms where young beef bulls often develop BRD shortly after being allocated into pens. BRD is a multi-factorial, multi-pathogen disease that is difficult to anticipate and control, often resulting in the massive use of antimicrobials to mitigate its impact on animal health, welfare, and economic losses. The DST developed from an existing mechanistic BRD model empowers users, including farmers and veterinarians, to customize scenarios based on their specific farm conditions. It enables them to anticipate the effects of various pathogens, compare the epidemiological and economic outcomes associated with different farming practices, and decide how to balance the reduction of disease impact and the reduction of antimicrobial usage (AMU). The generic method presented in this article illustrates the potential of artificial intelligence (AI) and software engineering methods to enhance the co-creation of DST based on mechanistic models in veterinary epidemiology. The corresponding pipeline is distributed as an open-source software. By leveraging these advancements, this research aims to bridge the gap between theoretical models and the practical usage of their outcomes on the field.

Genetic Features of Mycobacterium avium subsp. paratuberculosis Strains Circulating in the West of France Deciphered by Whole-Genome Sequencing

Paratuberculosis is a chronic infection of the intestine, mainly the ileum, caused by Mycobacterium avium subsp. paratuberculosis in cattle and other ruminants. This enzootic disease is present worldwide and has a negative impact on the dairy cattle industry. For this subspecies, the current genotyping tools do not provide the needed resolution to investigate the genetic diversity of closely related strains. These limitations can be overcome by the application of whole-genome sequencing (WGS), particularly for clonal populations such as M. avium subsp. paratuberculosis. The purpose of the present study was to undertake a WGS analysis with a panel of 200 animal field M. avium subsp. paratuberculosis strains selected based on a previous large-scale longitudinal study of Prim'Holstein and Normande dairy breeds naturally infected with M. avium subsp. paratuberculosis in the West of France. The pangenome analysis revealed that M. avium subsp. paratuberculosis has a closed pangenome. The phylogeny, based on alignment of 2,786 nonhomoplasic single nucleotide polymorphisms (SNPs), showed that the strain population is structured into three clades independently of the cattle breed or geographic distribution. The increased resolution of phylogeny obtained by WGS confirmed the homoplasic nature of the markers variable-number tandem repeat (VNTR) and short sequence repeat (SSR) used for M. avium subsp. paratuberculosis genotyping. These phylogenetic data also revealed independent introductions of the different genotypes in two main waves since at least 2003. WGS applied to this sampling demonstrated the presence of mixed infections in herds and at the individual animal level. Collectively, the phylogeny results inferred with French isolates compared to M. avium subsp. paratuberculosis isolates from around the world suggest introductions of M. avium subsp. paratuberculosis genotypes through the animal trade. Relationships between genetic traits and epidemiological data can now be investigated to better understand transmission dynamics of the disease. IMPORTANCE Mycobacterium avium subsp. paratuberculosis causes Johne's disease in ruminants, which is present worldwide and has significant negative impacts on the dairy cattle industry and animal welfare. Prevention and control of M. avium subsp. paratuberculosis infection are hampered by knowledge gaps in strain virulence, genotype distribution, and transmission dynamics. This work has revealed new insights into M. avium subsp. paratuberculosis strains currently circulating in western France and how they are related to strains circulating globally. We applied whole-genome sequencing (WGS) to obtain comprehensive information on genome evolution and discrimination of closely related strains. This approach revealed the history of M. avium subsp. paratuberculosis infection in France, refined the pangenomic characteristics of M. avium subsp. paratuberculosis, and demonstrated the existence of mixed infection in animals. Finally, this study identified predominant genotypes, which allow a better understanding of disease transmission dynamics. This information will facilitate tracking of this pathogen on farms and across agricultural regions, thus informing transmission pathways and disease control points.

Modelling transmission of Mycobacterium avium subspecies paratuberculosis between Irish dairy cattle herds

Bovine paratuberculosis is an endemic disease caused by Mycobacterium avium subspecies paratuberculosis (Map). Map is mainly transmitted between herds through movement of infected but undetected animals. Our objective was to investigate the effect of observed herd characteristics on Map spread on a national scale in Ireland. Herd characteristics included herd size, number of breeding bulls introduced, number of animals purchased and sold, and number of herds the focal herd purchases from and sells to. We used these characteristics to classify herds in accordance with their probability of becoming infected and of spreading infection to other herds. A stochastic individual-based model was used to represent herd demography and Map infection dynamics of each dairy cattle herd in Ireland. Data on herd size and composition, as well as birth, death, and culling events were used to characterize herd demography. Herds were connected with each other through observed animal trade movements. Data consisted of 13 353 herds, with 4 494 768 dairy female animals, and 72 991 breeding bulls. We showed that the probability of an infected animal being introduced into the herd increases both with an increasing number of animals that enter a herd via trade and number of herds from which animals are sourced. Herds that both buy and sell a lot of animals pose the highest infection risk to other herds and could therefore play an important role in Map spread between herds.

A mechanistic model captures livestock trading, disease dynamics, and compensatory behaviour in response to control measures

Trade is a complex, multi-faceted process that can contribute to the spread and persistence of disease. We here develop novel mechanistic models of supply. Our model is framed within a livestock trading system, where farms form and end trade partnerships with rates dependent on current demand, with these trade partnerships facilitating trade between partners. With these time-varying, stock dependent partnership and trade dynamics, our trading model goes beyond current state of the art modelling approaches. By studying instantaneous shocks to farm-level supply and demand we show that behavioural responses of farms lead to trading systems that are highly resistant to shocks with only temporary disturbances to trade observed. Individual adaptation in response to permanent alterations to trading propensities, such that animal flows are maintained, illustrates the ability for farms to find new avenues of trade, minimising disruptions imposed by such alterations to trade that common modelling approaches cannot adequately capture. In the context of endemic disease control, we show that these adaptations hinder the potential beneficial reductions in prevalence suTrade is a complex, multi-faceted process that can contribute to the spread and persistence of disease. We here develop novel mechanistic models ofch changes to trading propensities have previously been shown to confer. Assessing the impact of a common disease control measure, post-movement batch testing, highlights the ability for our model to measure the stress on multiple components of trade imposed by such control measures and also highlights the temporary and, in some cases, the permanent disturbances to trade that post-movement testing has on the trading system.

Rewiring cattle trade movements helps to control bovine paratuberculosis at a regional scale

Paratuberculosis is a worldwide disease mainly introduced through trade. Due to the low sensitivity of diagnostic tests, it is difficult to protect herds from purchasing infected animals. Our objective was to assess if rewiring trade networks to promote risk-based movements could reduce the spread of Mycobacterium avium subsp. paratuberculosis (MAP) between dairy cattle herds at a regional scale. Two levels of control strategies were assessed. At the between-herd scale, trade rewiring aimed to prevent animals from high-risk herds moving into low-risk herds. At the within-herd scale, complementary additional measures were considered based on the herd infection status, aiming to limit the within-herd spread by reducing calf exposure to adult faeces and culling more rapidly after positive test results. We used a stochastic individual-based and between-herd mechanistic epidemiological model adapted to the 12,857 dairy cattle herds located in Brittany, western France. We compared the regional spread of MAP using observed trade movements against a rewiring algorithm rendering trade movements risk-based. All females over two years old were tested. Based on the results, and taking into account the low test sensitivity, herds were annually assigned one of three statuses: A if the estimated true prevalence was below 7%, B if it ranged from 7 to 21 %, C otherwise. We also identified herds with a high probability of being MAP-free (AAA herds that had obtained an A status over three consecutive years) to assess the effect of decreasing their risk of purchasing infected animals on MAP regional spread. We showed that movement rewiring to prevent the sale of animals from high to low-prevalence herds reduces MAP regional spread. Targeting AAA herds made it possible to minimize the control effort to decrease MAP regional spread. However, animals purchased by AAA herds should have a moderate to high probability of being MAP-free, especially if the risk of purchasing animals from herds of unknown status cannot be managed. Improved hygiene and early culling of positive animals were relevant complementary on-farm control options to further decrease MAP spread. Future studies should identify how to define herd statuses to target optimal control measure combinations that could reduce the spread of MAP on a regional scale most effectively.

Towards a Framework for High-Performance Simulation of Livestock Disease Outbreak: A Case Study of Spread of African Swine Fever in Vietnam

Simple Summary

Disease transmission simulation programs in veterinary epidemiology in general and in simulation of African swine fever in particular are often very diverse and require great computing power. However, such programs often share similar workflows from processing input/output data, performing simulations, or storing data. Our paper proposes a common architectural framework for livestock disease transmission simulation programs in order to both improve simulation performance and reduce the effort of developing new simulation programs. Our framework was evaluated with a simulation program of African swine fever transmission currently raging in Vietnam and some other countries around the world. The results from the evaluation experiments not only demonstrate the effectiveness of the framework in terms of performance but also have practical consulting value for decision makers in Vietnam and for international colleagues.

Abstract

The spread of disease in livestock is an important research topic of veterinary epidemiology because it provides warnings or advice to organizations responsible for the protection of animal health in particular and public health in general. Disease transmission simulation programs are often deployed with different species, disease types, or epidemiological models, and each research team manages its own set of parameters relevant to their target diseases and concerns, resulting in limited cooperation and reuse of research results. Furthermore, these simulation and decision support tools often require a large amount of computational power, especially for models involving tens of thousands of herds with millions of individuals spread over a large geographical area such as a region or a country. It is a matter of fact that epidemic simulation programs are often heterogeneous, but they often share some common workflows including processing of input data and execution of simulation, as well as storage, analysis, and visualization of results. In this article, we propose a novel architectural framework for simultaneously deploying any epidemic simulation program both on premises and on the cloud to improve performance and scalability. We also conduct some experiments to evaluate the proposed architectural framework on some aspects when applying it to simulate the spread of African swine fever in Vietnam.

Research perspectives on animal health in the era of artificial intelligence

Leveraging artificial intelligence (AI) approaches in animal health (AH) makes it possible to address highly complex issues such as those encountered in quantitative and predictive epidemiology, animal/human precision-based medicine, or to study host × pathogen interactions. AI may contribute (i) to diagnosis and disease case detection, (ii) to more reliable predictions and reduced errors, (iii) to representing more realistically complex biological systems and rendering computing codes more readable to non-computer scientists, (iv) to speeding-up decisions and improving accuracy in risk analyses, and (v) to better targeted interventions and anticipated negative effects. In turn, challenges in AH may stimulate AI research due to specificity of AH systems, data, constraints, and analytical objectives. Based on a literature review of scientific papers at the interface between AI and AH covering the period 2009-2019, and interviews with French researchers positioned at this interface, the present study explains the main AH areas where various AI approaches are currently mobilised, how it may contribute to renew AH research issues and remove methodological or conceptual barriers. After presenting the possible obstacles and levers, we propose several recommendations to better grasp the challenge represented by the AH/AI interface. With the development of several recent concepts promoting a global and multisectoral perspective in the field of health, AI should contribute to defract the different disciplines in AH towards more transversal and integrative research.

Generative models of network dynamics provide insight into the effects of trade on endemic livestock disease

We develop and apply analytically tractable generative models of livestock movements at national scale. These go beyond current models through mechanistic modelling of heterogeneous trade partnership network dynamics and the trade events that occur on them. Linking resulting animal movements to disease transmission between farms yields analytical expressions for the basic reproduction number R ₀. We show how these novel modelling tools enable systems approaches to disease control, using R ₀ to explore impacts of changes in trading practices on between-farm prevalence levels. Using the Scottish cattle trade network as a case study, we show our approach captures critical complexities of real-world trade networks at the national scale for a broad range of endemic diseases. Changes in trading patterns that minimize disruption to business by maintaining in-flow of animals for each individual farm reduce R ₀, with the largest reductions for diseases that are most challenging to eradicate. Incentivizing high-risk farms to adopt such changes exploits 'scale-free' properties of the system and is likely to be particularly effective in reducing national livestock disease burden and incursion risk. Encouragingly, gains made by such targeted modification of trade practices scale much more favourably than comparably targeted improvements to more commonly adopted farm-level biosecurity.

Modelling transmission and control of Mycobacterium avium subspecies paratuberculosis within Irish dairy herds with compact spring calving

Paratuberculosis is a chronic bacterial infection of the intestine in cattle caused by Mycobacterium avium subspecies paratuberculosis (Map). To better understand Map transmission in Irish dairy herds, we adapted the French stochastic individual-based epidemiological simulation model to account for seasonal herd demographics. We investigated the probability of Map persistence over time, the within-herd prevalence over time, and the relative importance of transmission pathways, and assessed the relative effectiveness of test-and-cull control strategies. We investigated the impact on model outputs of calf separation from cows (calves grazed on pasture adjacent to cows vs. were completely separated from cows) and test-and-cull. Test-and-cull scenarios consisted of highly test-positive cows culled within 13 or 4 weeks after detection, and calf born to highly test-positive cows kept vs removed. We simulated a typical Irish dairy herd with on average 82 lactating cows, 112 animals in total. Each scenario was iterated 1000 times to adjust variation caused by stochasticity. Map was introduced in the fully naive herd through the purchase of a moderately infectious primiparous cow. Infection was considered to persist when at least one infected animal remained in the herd or when Map was present in the environment. The probability of Map persistence 15 years after introduction ranged between 32.2-42.7 % when calves and cows had contact on pasture, and between 18.9-29.4 % when calves and cows were separated on pasture. The most effective control strategy was to cull highly test-positive cows within four weeks of detection (absolute 10 % lower persistence compared to scenarios without control). Removing the offspring of highly test-positive dams did not affect either Map persistence or within-herd prevalence of Map. Mean prevalence 15 years after Map introduction was highest (63.5 %) when calves and cows had contact on pasture. Mean prevalence was 15 % lower (absolute decrease) when cows were culled within 13 weeks of a high test-positive result, and 28 % lower when culled within 4 weeks. Around calving, the infection rate was high, with calves being infected in utero or via the general indoor environment (most important transmission routes). For the remainder of the year, the incidence rate was relatively low with most calves being infected on pasture when in contact with cows. Testing and culling was an effective control strategy when it was used prior to the calving period to minimize the number of highly infectious cows present when calves were born.

Porcine reproductive and respiratory syndrome virus dissemination across pig production systems in the United States

Porcine reproductive and respiratory syndrome virus (PRRSV) remains widespread in the North American pig population. Despite improvements in virus characterization, it is unclear whether PRRSV infections are a product of viral circulation within production systems (local) or across production systems (external). Here, we examined the local and external dissemination dynamics of PRRSV and the processes facilitating its spread in three production systems. Overall, PRRSV genetic diversity has declined since 2018, while phylodynamic results support frequent external transmission. We found that PRRSV dissemination predominantly occurred mostly through transmission between farms of different production companies for several months, especially from November until May, a timeframe already established as PRRSV season. Although local PRRSV dissemination occurred mainly through regular pig flow (from sow to nursery and then to finisher farms), an important flux of PRRSV dissemination also occurred in the opposite direction, from finisher to sow and nursery farms, highlighting the importance of downstream farms as sources of the virus. Our results also showed that farms with pig densities of 500 to 1,000 pig/km² and farms located at a range within 0.5 km and 0.7 km from major roads were more likely to be infected by PRRSV, whereas farms at an elevation of 41 to 61 meters and surrounded by denser vegetation were less likely to be infected, indicating their role as dissemination barriers. In conclusion, our results demonstrate that external dissemination was intense, and reinforce the importance of farm proximity on PRRSV spread. Thus, consideration of farm location, geographic characteristics and animal densities across production systems may help to forecast PRRSV collateral dissemination.

How mechanistic modelling supports decision making for the control of enzootic infectious diseases

Controlling enzootic diseases, which generate a large cumulative burden and are often unregulated, is needed for sustainable farming, competitive agri-food chains, and veterinary public health. We discuss the benefits and challenges of mechanistic epidemiological modelling for livestock enzootics, with particular emphasis on the need for interdisciplinary approaches. We focus on issues arising when modelling pathogen spread at various scales (from farm to the region) to better assess disease control and propose targeted options. We discuss in particular the inclusion of farmers' strategic decision-making, the integration of within-host scale to refine intervention targeting, and the need to ground models on data.

Implications of Cattle Trade for the Spread and Control of Infectious Diseases in Slovenia

The objectives of this study were to gain insight into the structure of the cattle trade network in Slovenia and to evaluate the potential for infectious disease spread through movements. The study considered cattle movements between different types of premises that occurred between August 1, 2011 and July 31, 2016 with the exclusion of the movements to the end nodes (e.g., slaughterhouses). In the first part, we performed a static network analysis on monthly and yearly snapshots of the network. These time scales reflect our interest in slowly spreading pathogens; namely Mycobacterium avium subsp. paratuberculosis (MAP), which causes paratuberculosis, a worldwide economically important disease. The results showed consistency in the network measures over time; nevertheless, it was evident that year to year contacts between premises were changing. The importance of individual premises for the network connectedness was highly heterogeneous and the most influential premises in the network were collection centers, mountain pastures, and pastures. Compared to random node removal, targeted removal informed by ranking based on local network measures from previous years was substantially more effective in network disassociation. Inclusion of the latest movement data improved the results. In the second part, we simulated disease spread using a Susceptible-Infectious (SI) model on the temporal network. The SI model was based on the empirically estimated true prevalence of paratuberculosis in Slovenia and four scenarios for probabilities of transmission. Different probabilities were realized by the generation of new networks with the corresponding proportion of contacts which were randomly selected from the original network. These diluted networks served as substrates for simulation of MAP spread. The probability of transmission had a significant influence on the velocity of disease spread through the network. The peaks in daily incidence rates of infected herds were observed at the end of the grazing period. Our results suggest that network analysis may provide support in the optimization of paratuberculosis surveillance and intervention in Slovenia. The approach of simulating disease spread on a diluted network may also be used to model other transmission pathways between herds.

Neighbourhood contacts and trade movements drive the regional spread of bovine viral diarrhoea virus (BVDV)

To explore the regional spread of endemic pathogens, investigations are required both at within and between population levels. The bovine viral diarrhoea virus (BVDV) is such a pathogen, spreading among cattle herds mainly due to trade movements and neighbourhood contacts, and causing an endemic disease with economic consequences. To assess the contribution of both transmission routes on BVDV regional and local spread, we developed an original epidemiological model combining data-driven and mechanistic approaches, accounting for heterogeneous within-herd dynamics, animal movements and neighbourhood contacts. Extensive simulations were performed over 9 years in an endemic context in a French region with high cattle density. The most uncertain model parameters were calibrated on summary statistics of epidemiological data, highlighting that neighbourhood contacts and within-herd transmission should be high. We showed that neighbourhood contacts and trade movements complementarily contribute to BVDV spread on a regional scale in endemically infected and densely populated areas, leading to intense fade-out/colonization events: neighbourhood contacts generate the vast majority of outbreaks (72%) but mostly in low immunity herds and correlated to a rather short presence of persistently infected animals (P); trade movements generate fewer infections but could affect herds with higher immunity and generate a prolonged presence of P. Both movements and neighbourhood contacts should be considered when designing control or eradication strategies for densely populated region.

Enhanced sensitivity and fast turnaround time in laboratory diagnosis for bovine paratuberculosis in faecal samples

Bovine paratuberculosis (Johne's disease in cattle) is caused by the pathogen Mycobacterium avium subsp. paratuberculosis (MAP) and is a widespread chronic bacterial infectious disease in cattle. Due to the peculiarities of the pathogen, detection of MAP in faeces remains difficult. DNA extraction and real-time PCR for detection of MAP in bovine faeces (direct PCR) have been refined and feasible procedures for rapid, sensitive and automatable detection of the pathogen agent have been developed. Accordingly, in a first step we tested 20 faecal samples using two MAP complete kits (DNA extraction kits based on magnetic beads combined with real-time PCR assays) and six other DNA extraction kits for faeces. MAP-specific DNA was detected by real-time PCR assays. Cultivation of MAP on the solid medium HEYM and in the liquid medium M7H9C served as reference standards. The two complete kits detected significantly more MAP-DNA positive samples than the other procedures applied (p < 0.04). Ct values of 37 and 38 served as cut-off for the respective real-time PCR assays calculated on the basis of standard curves and droplet digital PCR (ddPCR). In a second step, the two MAP complete kits were employed for a comprehensive study including 107 positive and 50 negative faecal samples which had been previously tested on HEYM cultivation. The MAP complete kits yielded sensitivity values of 86% and 89% and specificity values of 100% compared to cultivation of MAP in the liquid medium M7H9C. In detail, cultivation of MAP in M7H9C detected the pathogen in 97% and 100% of the samples tested after an incubation period of six and twelve weeks, respectively. However, the cultivation of MAP on HEYM succeeded in only 74% after twelve weeks of incubation. In all these solid culture positive samples, MAP was also detected using the two complete kits. Additionally, the impact of repeated freezing and thawing of samples on re-cultivation of MAP was tested using 20 faecal samples and resulted in a reduction to 75% and 25% of bacterial growth when using liquid medium M7H9C and solid medium HEYM, respectively. The results of this study show that complete kits with refined automatable protocols for DNA extraction in combination with real-time PCR assays for detection of MAP can compete with sensitive cultivation of the pathogen in liquid medium.

Control measures to prevent the increase of paratuberculosis prevalence in dairy cattle herds: an individual-based modelling approach

Paratuberculosis, a gastrointestinal disease caused by Mycobacterium avium subsp. paratuberculosis (Map), can lead to severe economic losses in dairy cattle farms. Current measures are aimed at controlling prevalence in infected herds, but are not fully effective. Our objective was to determine the most effective control measures to prevent an increase in adult prevalence in infected herds. We developed a new individual-based model coupling population and infection dynamics. Animals are characterized by their age (6 groups) and health state (6 states). The model accounted for all transmission routes and two control measures used in the field, namely reduced calf exposure to adult faeces and test-and-cull. We defined three herd statuses (low, moderate, and high) based on realistic prevalence ranges observed in French dairy cattle herds. We showed that the most relevant control measures depend on prevalence. Calf management and test-and-cull both were required to maximize the probability of stabilizing herd status. A reduced calf exposure was confirmed to be the most influential measure, followed by test frequency and the proportion of infected animals that were detected and culled. Culling of detected high shedders could be delayed for up to 3 months without impacting prevalence. Management of low prevalence herds is a priority since the probability of status stabilization is high after implementing prioritized measures. On the contrary, an increase in prevalence was particularly difficult to prevent in moderate prevalence herds, and was only feasible in high prevalence herds if the level of control was high.