Use of Cattle Movement Data and Epidemiological Modeling to Improve Bovine Tuberculosis Risk-based Surveillance Scott Wells College of Veterinary Medicine University of Minnesota
Minnesota Bovine TB, 2005-2009 M. bovis identified in cow at slaughter plant in Feb 2005 and traced to Minnesota beef herd 12 infected beef cattle herds 27 deer with lesions M. bovis in positive cattle and deer linked to isolates from cattle in SW US and Mexico
Epidemiologic links between cattle herds in Minnesota Bovine TB outbreak, 2005-2009 Index herd purchased many cattle from other states All but 2 Minnesota herds connected through known cattle movements Hypothesis: Cattle movements key transmission factor, with spillover to free-ranging deer Linda Glaser, MN Board of Animal Health
Bovine TB Surveillance in US Limitations Slaughter surveillance 5 year delay in detection of infected herds (Fischer, 2005) Herd sensitivity influenced by herd size and cattle type Herd testing Not possible to perform cost-effective surveillance in all herds How to target surveillance? Disease does not spread randomly, spread depends on underlying risk factors
Targeted (risk-based) surveillance Focus on strata of the population more likely to have disease based on risk profile Herd testing among herds at higher risk M. bovis risk factors Herd size and cattle movements (Bessell 2012; Brooks-Pollock 2009; Gilbert 2005; Gopal 2006; Picasso 2017; USDA-APHIS-VS, 2011)
Pathogens move between herds via animal movements Herd 1 Herd 2 Endemic infection Disease Free
Uruguayan Cattle Production Pasture-based production system 11 million cattle on 52,000 farms Complete individual-based animal traceability system for cattle All cattle uniquely identified with electronic ear tag ID All cattle movements recorded by date, number animals, origin/destination Cattle movement database electronically available Ministry of Livestock Agriculture & Fisheries (MGAP) Statistic Division (DIEA). 2014 data
Bovine TB-positive farms in Uruguay 30 25 20 15 10 5 0 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Current BTB Eradication Program Annual CFT Testing in all dairy herds annually Slaughter surveillance
Use of Cattle Movement Data and Epidemiological Modeling to Improve Bovine Tuberculosis Risk-based Surveillance Study Objectives 1) Identify Bovine TB risk factors including between-herd interactions 2) Develop simulation model of spread of Bovine TB within and between cattle herds 3) Assess cost-effectiveness of targeted surveillance strategies for early detection of Bovine TB in cattle herds USDA NIFA AFRI Foundational Program
Bovine TB Research Team University of Minnesota Uruguay Julio Alvarez Meggan Craft Eva Enns Andres Perez Federico Fernandez Kim VanderWaal Catalina Picasso Szu-Yu (Zoe) Kao Yuanyuan Wang Andres Gil
Variable Category (N) OR 95% CI P-value Number cattle in herd <116 (76) 116-360 (73) >360 (79) Incoming steers No (198) Yes (30) Number cattle introduced Risk factors for Bovine TB in Uruguayan dairy herds Picasso, PVM, 2017 1 (79) >1-44 (71) >44 (78) 1.00 5.79 14.38 1.58-21.21 4.06-50.90 0.008 <0.001 1.00 2.88 1.12-7.37 0.027 1.00 0.77 1.90 0.29-2.06 0.81-4.44 0.608 0.138 No significant two-way interactions
13 Integrating within- and between-farm bovine TB transmission models VanderWaal, Sci Reports, 2017 12 15 4 16 9 6 5 8 7 11 1 14 Between-herd model 2 3 Susceptible 10 S Occult (Exposed) Infectious: No Reactive to testing: No O Reactive Infectious: No Reactive to testing: Yes Within-farm bovine TB mo R Infected Infectious: Yes Reactive to testing: Yes Within-herd model ~41 days I
Alternative BTB Surveillance Strategies in Uruguay Limit surveillance to only targeted (high risk) farms, based on known risk factors (Picasso 2017) Number of cattle received in 3-year period Number of cows Simulation model Seeded infection in 10 randomly selected dairy herds 1000 simulations performed per scenario for 5.5 years
Cost-effectiveness of Surveillance Strategies for Bovine TB in Uruguay VanderWaal et al, Sci Reports, 2017 Option D. Maintain similar surveillance sensitivity as baseline with 40% fewer farms sampled
Uruguay Targeted Surveillance Study 1. Relaxing surveillance on low-risk farms would reduce testing effort with no apparent increase in number of infected farms 2. Adapt Uruguay model for US surveillance context (with uncertainty about cattle movements)
United States Limited Cattle Movement data Simulated county-level network from US Animal Movement Model (Lindström, 2013) County-level out-movements from Minnesota County-level in-movements to Minnesota
Premise-level Movements Simulated movements for 10 years with cycle length of 3 months Production types: beef cow, dairy, and heifer raisers Data Simulated farm location and initial herd size for dairy and beef cow farms in Minnesota 1,000 simulated county-level movements and annual number of in/outshipments for Minnesota (provided by USDA-APHIS-VS) Assigned county-level movements to farms using simulated seasonal shipments for each farm from USDA-APHIS data
Integrating within- and between-farm bovine 13 12 15 4 16 9 6 5 8 7 11 1 TB transmission models 14 Between-herd model 2 3 Susceptible 10 Z. Kao S Occult (Exposed) Infectious: No Reactive to testing: No O Reactive Infectious: No Reactive to testing: Yes Within-herd model Within-farm bovine TB mo R ~41 days Infected Infectious: Yes Reactive to testing: Yes I
Bovine TB transmission model for US Simulated btb transmission using movements for 10 years with 1 index farm seeded among beef or dairy farms Baseline assumptions Spatial risk kernel and transmission rate (Uruguay model estimates) Perfect contact tracing Evaluated alternative targeted bovine TB testing strategies By herd size, number in-shipments last year, and number incoming cows in previous year
Minnesota btb Model Results Baseline prevalence (single herd seed) Slaughter surveillance alone Larger outbreak size and higher risk of between-herd transmission in dairy scenario than beef Time to detect the first case was longer in beef scenario than dairy Probability of disease freedom in 10 years was close to 90% Time to reach disease freedom was longer in beef scenario than dairy Alternative risk-based surveillance was very costly More than $2 million over 10 years Most efficient alternative strategy Beef scenario - Top 40% of herds by herd size and herd additions, cost $16 million more than status quo for one infection averted over 10 years Dairy scenario: top 40% of herds by herd size and in-shipments, cost $18 million more than status quo for one infection averted over 10 years
Minnesota btb Model Results Higher Initial Prevalence (5 herds seeded) Slaughter surveillance alone (compared to baseline prevalence) Higher number of additionally infected herds Lower probability of detection Lower probability of disease freedom in 10 years and longer time to disease freedom Alternative risk-based surveillance (compared to slaughter surveillance alone) Improved probability of disease freedom in 10 years (from 45% to 84% in beef scenario, and 62% to 82% in dairy scenario) Time to disease freedom could be reduced up to 16 or 19 months Most efficient alternative strategy cost $3.8 million (beef scenario) and $3.5 million (dairy scenario) more than slaughter surveillance alone per infection averted over 10 years
Conclusions from btb Modeling Advantages of slaughter surveillance alone Small bovine TB outbreak sizes Probability of detecting infected herds relatively high ( 80%) Less costly than other surveillance strategies Disadvantages of slaughter surveillance alone Not as effective at higher disease prevalence Longer time to detecting the first case Longer time to disease freedom and less likely to reach disease freedom in 10 years If disease status changed in affected state, economic loss due to movement restriction and testing requirement could outweigh the cost of risk-based testing
Key to disease control is understanding disease transmission Animal movement data (Social network analysis) Pathogen genotype data (Phylodynamics)