High Explanatory Power Model of FMD by Diwya Alok

Diwya completed his thesis at NPS in March 2013.   His work seeks to find relatively simple models for FMD spread that capture most of the variability in a high-fidelity, state-of-the-art FMD simulator.  Bellow is an executive summary of the thesis and the thesis itself.

Executive Summary

Foot and mouth disease (FMD) has a devastating impact on a country’s economy. The FMD containment process demands considerable efforts in vaccination, monitoring, trade restrictions, quarantines, and, historically, the elimination of millions of animals. Although, no incidence of FMD has been reported in the United States (U.S.) since 1929, there is a chance of its introduction as the disease is prevalent in two-thirds of the world. The United Kingdom (U.K.) had been free of FMD for more than 30 years before a major epidemic occurred in 2001. The epidemic resulted in the slaughter of approximately 7 million animals and had an estimated economic impact of $11.9–$18.4 billion by direct and indirect losses. A study conducted by Carpenter, O’Brien, Hagerman and McCarl in 2011, estimate the economic impact of a Foot and mouth disease (FMD) epidemic in the U.S. to be $2.3–$69.0 billion.

The U.S. Department of Agriculture is the lead agency for coordinating the response plan during an FMD outbreak. The national response plan is detailed in the Red Book (2012). The planning for the containment of the disease involves prior investments in control options, which determine the availability of response measures.

We simulate an outbreak of FMD across central California using the state of the art disease modeling Interspread Plus simulation package. We explore the epidemic’s response to varying disease and control parameters using an experimental design. We carry out 50 replications of 2,048 design points to produce 102,400 epidemic simulation runs. We execute the simulation runs on a cluster of computers. Using the data from the simulations, we identify 16 critical disease and control parameters that have the greatest effect on the spread of FMD. A statistical model based on these 16 parameters and their interactions captures approximately 85% of the variability of the simulation model.

The main takeaways of our analysis of FMD spread are as follows:

  • Initial Condition: The initial condition of the disease plays a significant role in the spread of the disease. We consider four starting scenarios: high animal density region, high premise density region, market, and port of San Francisco. Among the scenarios, the disease spread is almost twice as high when the infection originates in high animal or high premise dense areas. Detection time is, however, 28% shorter if the initial infection originates in high premise or high animal dense areas.
  • Local Spread: The local spread parameter captures the proximity-based spread of FMD between premises. Out of all disease and control parameters, epidemic progression has the highest sensitivity to local spread. Interaction and non- linear effects are significant for this parameter. Restricting local spread to less than 4,000 meters results in a 1.42 fold reduction in the mean number of cattle infected; however, the extent to which we can restrict local spread in a real-world scenario is unknown.
  • Market Movement: Market movement of cattle is a major contributor towards the spread of the disease. Interaction and non-linear effects for market movement are significant for this parameter. Our experimental results indicate that if a typical premise sends an animal to market every 2.2 days instead of every day, we will see a 25% reduction in the mean number of cattle infected.
  • Surveillance: Surveillance measures at dairy-like facilities are highly significant. We observe high positive interactions between surveillance and other control measures such as tracing and depopulation. Among the control measures, surveillance has the maximum impact towards reducing the spread of the disease. If there is less than a three day delay between suspecting an FMD outbreak and declaring an FMD outbreak at dairy-like facilities, we see a reduction in mean detection time for a novel epidemic of 32%. A delay of less than two days in the same parameter reduces the average number of infected cattle by half.
Control measures cannot be taken in isolation. Our models show significant interaction effects between the most effective control measures—market movement, and surveillance—and other control measures such as tracing, vaccination and depopulation. In addition, our model suggests that restricting local spread and controlling direct, indirect and market movements can be decisive towards controlling the spread of the disease in California. Furthermore, surveillance measures and movement control in adjoining zones, in addition to the primary outbreak zone, may help in reducing disease spread.

Thesis Presentation

Complete Thesis

Alok, D. (2013).  A High Explanatory Power Model of Foot And Mouth Disease Spread in Central California(Masters Thesis).  Naval Postgraduate School.

© Copyright 2004-2017 - Ned Dimitrov