FMD and its epidemiology

FMD is a highly contagious animal viral disease in cloven-hoofed animals. FMD is generally considered to be the most contagious of all diseases of farm animals, and it can spread rapidly and unexpectedly on a national and international scale. Therefore, it is also regarded as one of the most serious transboundary animal diseases (Geering et al., Reference Geering and Lubroth2002).

This transmissible disease was described in 1897 by Loeffler and Frosch, who theorized that it could be caused by a distinct agent. With the advancement in microbiological techniques, the causative agent was found to be a virus called FMDV which is a member of the Apthovirus genus of the Picornaviridae family. There are seven serotypes of FMDV and all cause a disease that is clinically similar but has immunologically distinct properties. The three distinct serotypes O, A and C were recovered by scientists in France and Germany in the 1920s. The other four serotypes, SAT 1–3 and Asia 1, were later recognized by the UK Pirbright team in the 1940s and 1950s (Sobrino and Domingo, Reference Sobrino and Domingo2004).

All domestic and wild cloven-hoofed animals are susceptible to FMDV infection, but the severity of the resulting disease varies with the level of immunity, the infectious dose, the virus strain and the host species (Sobrino and Domingo, Reference Sobrino and Domingo2004). Domestic livestock species that are susceptible to FMD are cattle, buffaloes, pigs, sheep, goats and deer, but the disease is generally most severe in cattle and pigs. Wild cloven-hoofed species also are susceptible; however, the clinical disease is rarely observed in those species. There have been some reports of infection in humans but these were rare and the clinical signs and symptoms were mild (Geering et al., Reference Geering and Lubroth2002).

FMDV infection is characterized by vesicular lesions on the mucous membranes of the mouth, muzzle, snout, udder and feet. The first clinical sign is salivation, and then the mucosa of the oral cavity becomes reddened. Shortly after, erosions develop in the interdigital spaces of the feet and on the udder. Infected animals then become lame and show unwillingness to stand (Geering et al., Reference Geering and Lubroth2002; Sobrino and Domingo, Reference Sobrino and Domingo2004).

Susceptible animals generally acquire the infection via the respiratory route. Very small doses of FMDV can initiate infection. Pigs are usually infected via the respiratory route, but this omnivore is more susceptible to infection by the oral route than are ruminants. The virus can also enter through abrasions in the skin or mucosa. After infection, the virus is excreted in large quantities in expired air, secretions, excretions and ruptured vesicles. Excretion of FMDV can begin up to 4 days before clinical signs appear. FMDV can retain infectivity for very long periods in frozen or chilled lymph nodes, bone marrow and residual blood clots. Additionally, the FMDV can remain infective for considerable periods in an environment where it is protected from desiccation, heat and adverse pH conditions (Geering et al., Reference Geering and Lubroth2002).

FMD can be transmitted by direct contact and by indirect transmission. Direct contact between infected and susceptible animals may lead to a faster spread in intensive farm situations than in low stock density areas. In indirect transmission, FMDV can spread mechanically by a variety of fomites including animal foodstuffs, artificial insemination equipment, vehicles, livestock holding areas and livestock equipment that could be contaminated with infected secretions and excretions (Geering et al., Reference Geering and Lubroth2002). Other research has shown that certain farm situations encourage the faster spread of FMD. Bates et al. (Reference Bates, Thurmond and Carpenter2001) found that larger livestock facilities had a higher frequency of direct and indirect contacts. Veterinarians and other people who have close contact with livestock can be at risk of carrying the virus from farm to farm. In temperate climates, infection can be spread over considerable distances by the airborne route. In past outbreaks, airborne spread has generally been from pigs at source to cattle downwind. Windborne transmission requires a slow and steady wind speed and direction, high relative humidity, weak sunlight and an absence of heavy rain (Geering et al., Reference Geering and Lubroth2002).

The introduction of new virus to naïve susceptible herds, areas or countries is likely to result in rapid spreading with high morbidity rates. FMD can be rapidly spread by direct contact, but the ability of the virus to survive in the environment means that indirect transmission through fomites may be as important as direct contact. Livestock trading and the movement of infected animals often cause disease transmission between different locations (Geering et al., Reference Geering and Lubroth2002). After an absence of 33 years, FMD was confirmed in pigs in the UK in 2001. This was an example of a serious outbreak of FMD in a country formerly free of the disease. The UK spent 11 months to regain its FMD-free status. Nearly, 4 million animals were slaughtered; the cost of the outbreak was estimated as 3.1 billion pounds sterling from losses to agriculture and food (Sobrino and Domingo, Reference Sobrino and Domingo2004).

This important transboundary disease is endemic in many countries in Africa, Asia, the Middle East and parts of South America; however, the countries in Europe, North and Central America, the Pacific and the Caribbean are mostly free from disease (Geering et al., Reference Geering and Lubroth2002). Paton et al. (Reference Paton, Sumption and Charleston2009) adapted geographical information from the FMD World Reference Laboratory to create a map of the global FMD status in 2008. FMD was listed in many parts of the world. Endemic areas were located in many parts of Asia, the Middle-East, Africa and some northern parts of South America. Sporadic areas of FMDV infection occur in the northern part of Asia and in the northern and some southern parts of Africa. Many countries in South America were observed as free with vaccination or free with multiple zones. North America, the majority of European countries and Australia were listed as free from FMD. Serotypes O and A could be found in almost all FMD areas except in the southern part of Africa. Serotype Asia 1 was located in southeast, south and west Asia and the three serotypes of SAT were found in the African continent (Paton et al., Reference Paton, Sumption and Charleston2009).

FMD in the United States

According to Graves (Reference Graves1979), the United States had nine FMD outbreaks from 1914 to 1929. With the exception of two outbreaks in 1914 and 1924, the disease was eradicated within a few months. The 1914 and 1924 outbreaks involved a considerably longer duration and much national-level effort to finally eradicate the disease. The author did not state the reasons for the difference in the outbreaks of 1914 and 1924 as compared with the other years. With an extensive cooperative eradication program between the United States and Mexico, the disease in that bordering country was eliminated in 1952. The last occurrence of FMD in North America was in 1952 in Saskatchewan, Canada, and eradication was complete in 1953 (Graves, Reference Graves1979).

According to the United States Department of Agriculture (2001 and 2007), cloven-hoofed animals in the United States are highly susceptible to FMD because they have not been exposed to allow for development of immunity to the virus nor have they been vaccinated since 1929. If an outbreak occurred in the United States, the disease could spread rapidly and widely in the country through routine livestock movements, unless the outbreak was detected early and eradicated immediately. Because the disease occurs in many parts of the world, there is always a chance of accidental introduction of the virus into the United States. Therefore, animals and animal byproducts from areas known to be affected are prohibited entry into the United States (US Department of Agriculture, 2001, 2007). A major challenge of FMD prevention in the US is to design an appropriate prevention and control plan which effectively partners capable officials with livestock stakeholders.

FMD control and prevention strategies and their justifications

Disease prevention encompasses all measures designed to exclude disease from an unaffected population of animals. Prevention measures include the exclusion of the causative agent from a given area and protection of uninfected populations from disease that already occurs in the area. Disease control measures are used to reduce the frequency of illness already present in a population by eliminating causes of illness or reducing them to levels of little or no consequence (Schwabe et al., Reference Schwabe, Riemann and Franti1977). The fundamental concepts of prevention and control of FMD or other infectious animal diseases are (i) control access of the virus to a susceptible host, (ii) control contact between infected and susceptible animals, (iii) reduce the number of infected animals and (iv) reduce the number of susceptible animals. To apply these concepts in the field as practical strategies for control of infectious animal diseases such as FMD, they become (i) animal movement control by quarantine or zoning, (ii) planned destruction of infected animals or herds and (iii) vaccination of susceptible animals. Options and requirements for control of FMD in FMD-free countries are different from those in endemic countries. In FMD-free countries or zones, such as the United States, early reaction and rapid containment of disease to the zone of infection and eradication within the shortest time frame are critical to stop the progression to an endemic status (Geering et al., Reference Geering and Lubroth2002; Paton et al., Reference Paton, Sumption and Charleston2009).

Animal movement is a high-risk factor for the spread of FMD infection; therefore, it is important to ban the movement of susceptible animal species and animal products within and out of the infected zone. The size and shape of the infected zones are dependent upon control activity plans. It is recommended that the zone be at a minimum 10 km radius around the disease foci and perhaps as much as 50 km in an area of intense livestock raising (Geering et al., Reference Geering and Lubroth2002). A ban of animal movement in some areas could affect many aspects of livestock production and trade. Therefore, it is difficult for the disease control authorities to effectively prohibit at all times 100% of animal movement in the infected area or specified zone during an outbreak.

The slaughtering of a minority of diseased animals to protect the healthy majority has been a control measure option in veterinary medicine for a long time (Schwabe et al., Reference Schwabe, Riemann and Franti1977). The slaughtering of infected animals as well as those in close contact with infected animals, also called stamping-out, is generally operated to control FMD outbreaks in areas previously free from FMD. However, an FMD stamping-out campaign should not be considered unless there are adequate provisions for compensation (Geering et al., Reference Geering and Lubroth2002). An FMD slaughter policy with strict movement controls was first applied in the UK in the late 20th century. The measure was successful in stamping out the disease, but the scale of slaughter at times overwhelmed the financial or organizational capacity (Paton et al., Reference Paton, Sumption and Charleston2009). Stamping-out sometimes can be used along with other control measures such as vaccination and restricted animal movement.

Another notable FMD prevention and control strategic program is vaccination. In South America, vaccination for FMD has been a major component of the national FMD control and eradication program since the 1960s. During an outbreak in Uruguay in 2001, the use of vaccines in an endemic situation and as an anti-epidemic tool showed its effectiveness in creating FMD-free areas in South America. In 2004, over 200 million cattle were vaccinated twice yearly throughout the continent (Schudel and Lombard, Reference Schudel and Lombard2004). In order to be effective, disease control by vaccination must be used in conjunction with zoosanitary measures. Globally, 2 billion animals are vaccinated annually, but the use of vaccination in endemic countries is not uniform. Many African and south and south-east Asian countries use vaccination to a very limited extent. The majority of FMD vaccine is used in large-scale programs in China, South American countries and parts of India and the Middle East (Paton et al., Reference Paton, Sumption and Charleston2009).

According to Paton et al. (Reference Paton, Sumption and Charleston2009), effective options for control of FMD must consider knowledge, capability and policy. The essential knowledge of FMD control could be gathered from publications. However, a good FMD control policy should not only apply knowledge, but must also take capability into consideration. Estimation of capability outside an actual outbreak is difficult. However, epidemiological simulation modeling and spatial analysis can help define the hypothetical FMD spread; this approach can allow the capability for disease control to be estimated.

Epidemiological modeling for FMD and other infectious animal diseases

The aim of modeling infectious animal diseases such as FMD is to predict or to understand the behavior of an epidemic (Gerbier et al., Reference Gerbier, Bacro, Pouillot, Durand, Moutou and Chadoeuf2002) and to assess potential effectiveness of various control and eradication strategies (Bates et al., Reference Bates, Thurmond and Carpenter2003b). Many models have been applied to study the transmission of FMD or other infectious diseases and to predict the impact of control measures. These models are limited, however, by the gap between the data requirements and data availability. This inadequacy could be improved by collection of more data. For example, livestock premises census data were needed to model the FMD spread, but were not available. As a way to reduce the limitation of the model, spatial simulation for livestock farm locations could be applied to estimate the census of livestock premises.

There have been several studies using epidemiological modeling techniques for FMD in order to advance the development of FMD modeling. For example, a study conducted in California enhanced the current understanding of transmission of FMD by estimating contact rates of FMD in livestock (Bates et al., Reference Bates, Thurmond and Carpenter2001). The authors worked within a 3-county region of California, and estimated the direct and indirect contact rates in livestock facilities and the distance traveled between herd contacts. These researchers found that direct and indirect contacts occurred in livestock facilities over a wide geographic area, with larger facilities having a higher frequency of contacts (Bates et al., Reference Bates, Thurmond and Carpenter2001). One of the conclusions was that the results of their study may be useful for developing biosecurity programs at herd, state and national levels and for modeling transmission potential for FMDV. Later, these researchers used the rates and other information from this study as parameters to study epidemic simulation modeling to evaluate control strategies during an outbreak of FMD (Bates et al., Reference Bates, Thurmond and Carpenter2003a, Reference Bates, Thurmond and Carpenterb).

After the FMD epidemic in the United Kingdom in 2001, French researchers Le Menach et al. (Reference Le, Legrand, Grais, Viboud, Valleron and Flahault2005) developed a farm-based stochastic model to evaluate the consequences of virus introduction into France. This study identified and mapped the high-risk zones for the spread of FMD if the virus was imported. With the standard control policy simulated in the same 50 initially infected farms, the hypothetical outbreak would infect 16,350 of approximately 280,000 susceptible farms in France. The high-risk zones were the regions having high densities of cows and sheep. When farms were tightly clustered, the disease was transmitted quickly within the cluster. This study demonstrated that the epidemic process for FMD depends on the location, size and species type of the initially infected farms (Le Menach et al., Reference Le, Legrand, Grais, Viboud, Valleron and Flahault2005).

Carpenter et al. (Reference Carpenter, Thurmond and Bates2004) used simulation modeling to predict the spread and control of FMDV if it were introduced into a disease-free country. Simulation models have also been used to evaluate strategies and aid decision makers in identifying the optimal disease eradication plan. Carpenter et al. constructed a model using a commercial spreadsheet and a simulation add-in for Monte Carlo sampling. The Monte Carlo simulation technique involves the random sampling of each probability distribution within the model to produce hundreds or even thousands of scenarios (also called iterations or trials) (Vose, Reference Vose2000). Each probability distribution is sampled under a distribution shape, and then the distribution of the values is calculated for the model outcome. The level of mathematics required to perform a Monte Carlo simulation is quite basic, but complex mathematics (e.g. power functions, logs, if statements) can also be included.

The North American Animal Disease Spread Model (NAADSM) is a stochastic, state-transition simulation model for the spread of highly contagious diseases of animals. NAADSM users can establish parameters to define model behavior in terms of disease progression; direct contact, indirect contact, airborne dissemination; and implementation of control measures such as destruction and vaccination while the direct costs associated with these measures are considered. This model is being used to evaluate outbreak scenarios and potential control strategies for FMD and exotic animal diseases in the United States, Canada and elsewhere. NAADSM can define model behavior in terms of disease progression while taking into consideration the implementation of control measures such as destruction and vaccination. Therefore, this model can be used effectively to evaluate outbreak scenarios and potential control strategies for FMD. NAADSM is freely available via the internet at http://www.naadsm.org (Harvey et al., Reference Harvey, Reeves, Schoenbaum, Zagmutt-Vergara, Dubé, Hill, Corso, McNab, Cartwright and Salman2007). Many models mentioned assumed that individuals within the population have equal opportunity to come into contact with any other individual. The NAADSM redefined individuals as the fixed locations of premises or flocks or herds. Contact rates for premises were unequally dependent on the defined production types and the distance of premises from other premise locations. This may be more acceptable for the simulation of FMD spread within a large study area having a large number of premises.

FMD simulation modeling with the aim to assess control measures

Several studies have used simulation models to assess FMD control measures. In 2003, Bates et al. (Reference Bates, Thurmond and Carpenter2003b) published the results from epidemic simulation modeling to evaluate control strategies of FMD in Fresno, Kings and Tulare counties of California; in this study, they used the estimated contact rates that had been calculated in 2001 (Bates et al., Reference Bates, Thurmond and Carpenter2001). The authors used a spatial stochastic model to evaluate a hypothetical outbreak of FMD under control scenarios that included baseline control strategies, vaccination strategies and preemptive herd slaughter strategies. The authors concluded that preemptive slaughter of the highest-risk herds and vaccination of all animals within a specified distance of an infected herd consistently decreased the size and duration of an epidemic, compared with the baseline eradication strategy (Bates et al., Reference Bates, Thurmond and Carpenter2003b).

Schoenbaum and Disney (Reference Schoenbaum and Disney2003) modeled alternative mitigation strategies for a hypothetical outbreak of FMD in the United States. Using 72 different scenarios, the authors compared epidemiologic and economic consequences among simulated FMD outbreaks in a generated population of susceptible herds. They suggested that the choice of an appropriate disease control strategy depends on herd demographics and the rate of contact among the herds. They concluded that preemptive slaughter and early ring vaccination could decrease the duration of an outbreak. Even though these mitigation strategies would initially be costly, they would decrease the duration and overall cost of an outbreak (Schoenbaum and Disney, Reference Schoenbaum and Disney2003).

Le Menach et al. (Reference Le, Legrand, Grais, Viboud, Valleron and Flahault2005) also assessed control policies in their FMD outbreak simulation model. The authors developed a stochastic farm-based model adapted to the French farm structure from models of the 2001 epidemic in the United Kingdom. The livestock data were only available at the town scale, so the farm location and number of animals in each farm were simulated over the boundary area of each French town. They found that preemptive culling and ring vaccination had the greatest impact on reducing the number of FMD cases and the length of the epidemic. The results of this model provide useful information for decision makers planning the response to an epidemic of FMD re-emerging in France.

Results from simulation models suggest that in some situations control measures under current legislation would not be sufficient to control FMD in a particular area. According to Martinez-Lopez et al. (Reference Martinez-Lopez, Perez and Sanchez-Vizcaino2010), results from their simulation model in the Castile and Leon region of Spain suggested that the control measures specified in the Spanish and EU legislations were not the most effective strategies to control FMD spread. The authors modified a stochastic spatial disease state-transition model to simulate the hypothetical spread of FMD in the study area. They concluded that preventive depopulation or vaccination at <1 and <3 km radii, respectively, around infected premises (IPs), was more effective in controlling the spread of FMD epidemics in the Castile and Leon region. The current strategy is limited by Spanish legislation to <3 and <5 km radii for depopulation and vaccination, respectively. When evaluated within the simulation model, the conventional strategy did not result in any significant reduction of the magnitude of the epidemic in the region and was not cost-effective (Martinez-Lopez et al., Reference Martinez-Lopez, Perez and Sanchez-Vizcaino2010).

Epidemiological modeling is not only studied in FMD-free countries but it has been applied in FMD-endemic countries as well. Wongsathapornchai et al. (Reference Wongsathapornchai, Salman, Edwards, Morley, Keefe, Van Campen, Weber and Premashthira2008) used epidemiologic risk modeling to evaluate various FMD control programs in southern Thailand, where FMD is endemic. The authors used a quantitative risk assessment to ascertain the probability of FMD introduction and used an intrinsic model to evaluate impact. Five scenarios were created to assess the estimated cumulative incidence of FMD and the impacts of non-structural protein (NSP) testing, mass vaccination and culling. The results of the study suggested that vaccination has a greater impact than the use of NSP testing (Wongsathapornchai et al., Reference Wongsathapornchai, Salman, Edwards, Morley, Keefe, Van Campen, Weber and Premashthira2008).

Spatial analysis in epidemiological study

The distributions of disease events in space are important because most epidemiological studies attempt to define the geographic limits of disease events or concerns and also the patterns of spatial distribution within those limits. A disease may be limited geographically for many reasons that relate to forces that can act upon the susceptible host populations or upon the disease agents. Spatial patterns of disease distribution are widely divided into random, contagious and regular patterns. There is special interest in the clustering of disease events because it may help in identifying a common environmental factor or source of exposure (Schwabe et al., Reference Schwabe, Riemann and Franti1977).

Spatial epidemiology concerns the analysis of the geographical distribution of the incidence of disease. The simplest unit that can be analyzed is a map of the locations of disease cases using symbols of sizes that are small compared with the distances between them. Therefore, in epidemiological study, the locations of disease cases are sufficiently represented as dots on a map. The associated issues related to map production and the statistical analysis of mapped data must be applied. Since mapped data are spatial in nature, the application of spatial statistical or geostatistical methods is a core part of the analysis. One of the important stages of map construction that can be associated with spatial information is the choice of scale. A suitable scale for the map must be chosen and the choice leads to a process of averaging spatial information from higher levels of resolution. The small scale, for example, could be executed for disease clustering and applied to explore putative sources of hazard. The large scale would be applied for illustrating disease mapping, ecological analysis, infectious disease modeling and epidemiological surveillance (Lawson, Reference Lawson2006).

In geographical data, the relative risk function over a geographical region can be estimated effectively using kernel density estimation to examine the spatial distribution of disease cases and a sample of controls (Bithell, Reference Bithell1990). Examples of this in the literature include studies by Bithell (Reference Bithell1990) and Mukherjee et al. (Reference Mukherjee, Kumar, Mittal and Saxena2002). Bithell made use of a two-dimensional Poisson process to estimate the density of cases of childhood leukemia in the vicinity of the Sellafield nuclear reprocessing plant in Cumbria, UK. The author defined the relative function as a ratio of the density of cases of childhood leukemia and controls. Other researchers used the spatial technique of density estimation to explain how the risk of developing goiter varied at different villages within the study area from the Indo-Nepal border to the Sitamarhi district headquarters in Muzaffarpur in the Sub-Himalayan belt area of India (Mukherjee et al., Reference Mukherjee, Kumar, Mittal and Saxena2002). The results were presented as a three-dimensional perspective plot with the x- and y-axes representing distances and the z-axis representing the probability density function of the goiter-affected population. These studies applied the kernel density estimation technique to determine spatial distribution of disease, including levels of risk factors. Results from the studies were displayed as plots or maps which help in better visualization for understanding disease patterns.

Geostatistical and spatial statistics methods were utilized for studying highly pathogenic avian influenza (HPAI) subtype H5N1 (Ward et al., Reference Ward, Maftei, Apostu and Suru2008). The authors stated that the purpose of their study was to evaluate the usefulness of statistical and geostatistical methods to determine how HPAI might spread through a national population of village poultry in Romania. Methods in the study included directional statistics, Moran's correlation statistic, variography and krigging. Risk mapping was used to visualize the evolution of the epidemic which could be characterized into two parts: disease introduction, local spread and sporadic outbreaks; and long-distance disease spread with rapid epidemic propagation. The researchers found that the environmental characteristics and the landscape in the Danube River Delta area played a critical role in the introduction and initial spread of HPAI. In later seasons, the movement of poultry might have introduced the infection into central Romania. The authors noted that the outbreaks of HPAI H5N1 in Romania between 2005 and 2006 were not randomly distributed in time and space, and they identified differences in the spatial distribution of the outbreaks during time phases. In this study, HPAI H5N1 could have been introduced into domestic poultry populations from wild waterfowl, but the virus spread within domestic poultry populations was likely to be via the movement of live birds or fomites.

Riley (Reference Riley2007) reviewed studies on four diseases – measles, FMD, pandemic influenza and small pox – to demonstrate the benefits of different methodologies for spatial models. Patch models, distance transmission models, multi-group models and network models were represented as the methodologies of large-scale spatial-transmission models for the four diseases, respectively (Riley, Reference Riley2007). Riley's review helped characterize the large-scale patterns and evaluate the impact of interventions.

Spatial analysis for FMD

The spread of FMD in the United Kingdom was analyzed using a point-pattern model (Gerbier et al., Reference Gerbier, Bacro, Pouillot, Durand, Moutou and Chadoeuf2002). The farm-to-farm process of infection transmission was examined by point-pattern methodology. In the model description, two broad types of risk factors were classified regarding local transmission and long-distance transmission. From this study, the authors stated that the spatial spread of FMD in the 1967–1968 UK epidemic was influenced by heterogeneity of exposure to the virus, animal density and the networks formed by contacts between farms.

Spatial clustering of disease may help in identifying a common environmental factor or source of exposure. For example, this technique has been used to quantify the associations between hypothesized epidemiological factors and the spatial distribution of FMD in Nepal (Chhetri et al., Reference Chhetri, Perez and Thurmond2010). Methods in this study included cluster analysis implemented using SatScan software, and Bayesian mixed-effects Poisson regression to model the association between FMD cases and factors hypothesized to be associated with the risk of having an FMD-positive farm in a district. Spatial scan statistics and cluster analysis techniques identified two significant clusters of FMD reports. The clusters were identified risks of FMD reports, such as size of human, buffalo populations and number of technicians. This finding increased the knowledge on the epidemiological dynamics of FMD and improved the efficiency of resource allocation and control efforts in Nepal.

The issues relating to the epidemiology of the 2001 FMD epidemic in the UK, including analysis of spatially referenced data, were examined by Lawson and Zhou (Reference Lawson and Zhou2005). They discussed the use of exploratory statistical tools such as density estimation and nonparametric regression, and considered the need for descriptive models for space, time, and space-time epidemic dynamics. They also discussed the advantages of using Bayesian models for disease spread and applied them to the UK outbreak. From several analyses in this study, the authors concluded that spatial statistical and GIS (Geographical information system)-based methods have an important place in the analysis of animal diseases because they can lead to insights in exploratory settings, ecological studies and surveillance.

Spatial analysis can be combined with epidemiologic methods such as a retrospective case−control study to investigate the potential geographic risk factors of disease transmission. Such a study was done on the geographic and topographic determinants of FMD transmission in the 2001 UK epidemic (Bessell et al., Reference Bessell, Shaw, Savill and Woolhouse2008). Methods in this study were univariate and multivariate generalized linear mixed model analyses of predictor variables and risk factors for FMD transmission. This study investigated features of the landscape and their impact on transmission during the period of the national movement ban during the outbreak. The results indicated that the presence of rivers and railways had an additional protective effect to reduce the probability of transmission between holdings.

Another study using the spatially resolved farm census data from the 2001 FMD outbreak in the UK was conducted by a research team to formulate a spatially explicit distance-transmission model of FMD, with farms as the individual units of infection (Riley, Reference Riley2007). The researcher applied the distance-transmission model to consider the vaccination priority in the UK outbreak. They concluded that prioritization of farms for vaccination based only on their proximity to IPs reported in previous 10 days, or to dangerous contact with those IPs, was better than other vaccination plans. This new priority was shown to be the most effective in terms of reducing the number of animals culled to eradicate the disease.

The FMD epidemic control area may be more effectively identified when local and regional georeferenced data are considered. Using data from Uruguay in 2001, Rivas et al. (Reference Rivas, Smith, Sullivan, Gardner, Aparicio, Hoogesteijn and Castillo-Chavez2003) explored whether early analysis of spatial data may result in identification of variables associated with epidemic spread of FMD. The authors created a georeferenced database and performed a retrospective analysis. They compared mean or median results of day 1 to day 3 versus day 4 to day 6 of the epidemic and the results of correlation analysis. They found that as time progressed, disease spread was significantly associated with increases in road density, cattle density, and dairy cattle production, but decreased with smaller farm size and greater distances between the case farm and the nearest road. It was concluded that the direction of an epidemic can be assessed on the basis of road density and spatial variables as early as 6 days into the epidemic (Rivas et al., Reference Rivas, Smith, Sullivan, Gardner, Aparicio, Hoogesteijn and Castillo-Chavez2003).