Performing Department
(N/A)
Non Technical Summary
The highly pathogenic avian influenza (HPAI) outbreak in U.S. dairy cattle is a major threat to U.S. cattle producers, U.S. food security, and the agricultural sector of the U.S. economy. The outbreak is especially challenging because it is the first time that HPAI has been found in cattle. We know that shipments of dairy cattle have moved HPAI over long distances between states, but we don't know much yet about how to predict or stop the spread. The goal of this project is to create predictions of how HPAI could continue spreading across the U.S. based on cattle shipments and disease data. We have three objectives: 1) we will collect data on cattle shipments and the HPAI outbreak; 2) we will develop predictions about U.S. cattle shipments and HPAI transmission using the collected data (obj. 1) and other information; 3) we will use the HPAI disease predictions to understand how HPAI is spreading in the U.S. and how it can be controlled. This project will give us a way to determine the risk of HPAI spreading into new locations in the U.S. and what can be done to prevent that spread. We will work with the U.S. Department of Agriculture and State Veterinarians to ensure that the results of this project are used to help create surveillance and response strategies for HPAI. The results of this research project will provide an understanding of the ongoing HPAI outbreak and will be used to protect the interests of individual producers and the cattle industry in general from the impacts of HPAI outbreaks now and in the future.
Animal Health Component
100%
Research Effort Categories
Basic
(N/A)
Applied
100%
Developmental
(N/A)
Goals / Objectives
The highly pathogenic avian influenza (HPAI) outbreak in U.S. dairy cattle demonstrates the threat of transboundary animal diseases (TADs) to U.S. livestock producers and food security. A challenge of the HPAI outbreak is the lack of information about transmission dynamics. Mathematical models are invaluable for studying transmission risk factors and ways to minimize those risks. For livestock infectious disease outbreaks, understanding shipment patterns is critical for reducing the risk of long-distance spread. Shipments of infected cattle have moved HPAI over large distances and sparked new chains of transmission in states that previously had no cattle infections. In order to get ahead of the HPAI outbreak and understand potential impacts of response actions, national-scale mathematical models that take into account different aspects of transmission and response strategies are needed. Our goal is to develop data-driven, national-scale cattle shipment (U.S. Animal Movement Model (USAMM)) and HPAI-customized disease transmission (U.S. Disease Outbreak Simulation (USDOS)) models for the U.S. that can be used to understand the drivers and risk factors for HPAI transmission in cattle and the impacts of potential response strategies. These models will allow for large-scale exploration of a current TAD threat to U.S. agriculture and provide vital information to inform rapid response and recovery both from the current HPAI outbreak and potential TAD threats in the future.Objective 1: Collect data to inform HPAI in cattle modeling efforts. In order to make modeling efforts on HPAI and other TADs as useful as possible, all parameters and data inputs must be informed by the best data available. For Obj. 1, we identified three types of data to collect and analyze to inform HPAI modeling efforts. The first type of data is shipment data that will inform the USAMM model. There is no federal database of livestock shipments, but Interstate Certificates of Veterinary Inspection (ICVI) are collected by states when livestock shipments cross state borders and provide the best sample of interstate cattle shipments in the U.S. Here we will focus on collecting data from a subset of states in the contiguous U.S. that are very important to the shipment network. Secondly, we will collect data on non-shipment-based HPAI transmission events that occur over smaller spatial scales, such as within a single county. Preliminary analysis of HPAI transmission patterns in the U.S. shows that non-shipment based transmission must be occurring in order to explain the patterns. By collecting data on HPAI transmission over short distances we will have the information needed to fit a local between-herd transmission kernel, which will parameterize USDOS and other modeling efforts. Finally, we will collect data on prevalence and dynamics of HPAI within herds as it becomes available. These data will inform USDOS parameters including the duration of premises infectiousness and the within-herd dynamics.Objective 2: Develop additional functionality and integrate new data into USAMM and USDOS. Currently, USAMM is the most comprehensive tool available for estimating livestock shipment networks, but we have identified several issues that need to be improved for the scope of this proposal as well as for livestock shipment network and disease modeling in the U.S. in general. This project will improve USAMM's predictive performance by incorporating recent shipment data into the model parameterization, while simultaneously expanding the model functionality such that multi-year shipment predictions are feasible. The multi-year USAMM model will then be incorporated into USDOS, which is essential for exploring the role that shipment predictions play in potential TAD outbreaks and how to mitigate those risks. USDOS is currently the only model running at the national scale for the U.S., capturing long range transmission via shipments, and is therefore invaluable for understanding potential TAD outbreaks. A critical piece of managing TAD outbreaks is planning and exploring options for response with accurate information regarding transmission, resources, and potential targets. Therefore, to investigate the current HPAI in dairy response actions, we will expand the USDOS functionality to include premises-level shipment bans and pre- and post-shipment diagnostic surveillance. We will also use the HPAI data collected in Obj. 1 to parameterize USDOS to customize it to the current HPAI outbreak in U.S. dairy cattle.Objective 3: Evaluate HPAI transmission dynamics and response actions in U.S. dairy cattle. To understand transmission and the impacts and efficacy of response actions on HPAI in dairy cattle, mathematical models are needed. USAMM and USDOS are uniquely situated to perform simulation studies on HPAI dynamics in U.S. dairy cattle. We will use the HPAI-specific USDOS with multi-year USAMM (Obj.2) incorporated into it to explore transmission dynamics and risk factors for transmission of HPAI at the national scale. We will then study the risk of HPAI transmission from dairy to beef cattle and potential ways to reduce the risk of spillover. Finally, we will study the impact of response actions including those currently being implemented and those not yet available, such as vaccination.
Project Methods
Objective 1: All data collection activities under Obj. 1 will follow our strict data protocol that has been established and tested by this research team through past data entry initiatives. We have identified a subset of 15 states to focus the Interstate Certificates of Veterinary Inspection (ICVI) data collection efforts on. A sample of 30% of the cattle export ICVI records from two recent years will be collected. We will collect digital ICVI records where they are available and paper records in states that do not have robust digital ICVI databases. Cleaning the ICVI data will follow standard data cleaning protocols established by previous data entry initiatives and the field. We will analyze the shipment data to identify the number of outgoing ICVI records by state and county, the proportion of shipments going to different states, and any anomalies between the years of data. We will perform a comparative network analysis between the new ICVI data and the previously collected 2009 ICVI data. These analyses will be used to incorporate the new data into multi-year USAMM. The efforts from the ICVI data collection and analyses will result in a clean, multi-year, multi-state cattle shipment database that will be shared with the USDA. Additionally, all states participating in the study will be provided with a state-specific report about the shipment patterns and changes since the 2009 ICVI data collection and a clean state-specific shipment database.We will also collect data on the spread of HPAI over small spatial scales, such as a county, and data on the within-herd spread of HPAI. The data on the HPAI spread at small spatial scales will be used to fit a local transmission kernel for HPAI in cattle, which is used in USDOS to model short-distance transmission. The data on within-herd HPAI transmission will be used to inform within-herd epidemic curves that will be used by USDOS. The efforts of this data collection will result in the HPAI transmission data being used to fit HPAI transmission functions for USDOS and these functions will be published and made freely available with HPAI-specific USDOS.The data collection in Obj. 1 will be evaluated through standardized quality control checks throughout the collection process and through the planned analyses. The analyses will show us if there are any biases in the data or missing data that is needed to inform USAMM and USDOS, respectively. For the ICVI database, we will ensure that the coverage of the cattle population is sufficient to inform the USAMM. For the HPAI outbreak data, we will perform analyses that will inform us about the uncertainty in the data and how that uncertainty can be reduced.Objective 2: USAMM will be extended with the capability to analyze a multi-year dataset in a statistically rigorous manner. We will develop USAMM to incorporate a conditional autoregressive (CAR) model structure. This means that when data are sparse for a geographical area, statistical strength can be borrowed from nearby areas to improve predictions. This will enable multi-year analyses, as it would allow the ICVI dataset from 2009 (which includes 47 states) to be complemented by one or more datasets that cover only a subset of states. This will improve USAMM's predictive performance by incorporating recent shipment data into the model parameterization, while simultaneously expanding the model functionality such that multi-year shipment predictions are feasible. We will evaluate the performance of USAMM by its ability to generate robust predictions over multiple years.We will also be improving the USAMM run times, so that USAMM can be quickly updated in the future. To do this, we will move to the MPI (message parsing interface) standard, which allows communication and synchronization across multiple nodes. With this implementation, the analysis could be spread over an unspecified number of nodes, increasing computational power available and thus shortening run-times significantly. The improvements in run time will be evaluated by quantifying the improvement of run time of the multi-year USAMM over that of previous versions of the model. The efforts of multi-year USAMM development will quantify temporal trends to the shipment patterns that will be published in a scientific journal and communicated to the USDA. Additionally, we will promote the use of USAMM by the USDA, and other modeling groups by making 1,000 replicates of multi-year shipment networks available on the USAMM/USDOS website. We will also update our USAMM Shiny app with the new predictions. These materials will be advertised in the USAMM publication, and directly to the USDA.To fit a local transmission kernel that will be used in USDOS, we will use our previously developed R package that takes spatial outbreak data and fits transmission kernel functions to it. To fit an HPAI kernel we will use the HPAI outbreak data collected in Obj. 1 and combine it with U.S. cattle demographic realizations to assign locations and sizes of infected and susceptible dairies. We will evaluate the kernel using sensitivity analysis to address how parameter uncertainty impacts the predicted outbreak results and our understanding of HPAI transmission dynamics.We will also use the HPAI outbreak data to fit a within-herd epidemic curve function that will then be input into USDOS as we have previously done for FMD. We will evaluate the HPAI within-herd curve using sensitivity analyses to study how uncertainty in the herd prevalence impacts predictions about the outbreak trajectory and duration.To implement the current response strategies to HPAI on dairies, we will make two small modifications to shipment ban and diagnostic testing functionality in USDOS. We will add additional functionality to the shipment ban code so that shipment bans can be implemented at the premises-level for a specified length of time. We will also update the USDOS code so that diagnostic tests can be triggered by either outgoing or incoming shipments. The code additions will be evaluated by ensuring that they arefunctioning correctly and capturing the current response strategy to HPAI.The efforts from the USDOS development will include publishing the HPAI cattle spatial transmission kernel so that it can be used by other modeling groups and the USDA. We will make the updated USDOS code broadly accessible by wrapping the C++ code into the user-friendly USDOS pipeline, which will be available on the USAMM/USDOS website and advertised in the USDOS scientific publication.Objective 3: We will address three main questions involving the dynamics and risk of transmission of cattle HPAI consistent with USDA goals by performing a simulation study using USDOS. Q1) What are the primary drivers of HPAI transmission among dairies in the U.S.? Q2) What is the risk of HPAI infection spilling over from dairy into beef cattle? Q3) What is the impact of response actions on the HPAI outbreak in the U.S.? To run, manage and analyze the USDOS simulations, we will follow the protocols in our fully tested, and successful USDOS pipeline. We will first run simulations in dairy cattle alone and then in beef and dairy cattle. We will also run simulations starting in the county where the HPAI outbreak was found and starting the simulations in every contiguous county and will run a sensitivity analysis. Finally, we will run simulations with and without response strategies in place, which will allow us to explore the differences in HPAI outbreak predictions. The results from the simulation study will be evaluated using statistical analyses and studying how well the predicted patterns match that of the HPAI outbreak. The effectiveness of the response actions will be evaluated by measuring the reduction in outbreak metrics. The efforts from the simulation study will be published in a scientific journal and presented to other researchers at scientific conferences and directly to the USDA.