Rev. sci. tech. Off. int. Epiz., 16 (1), 207-214 Risks and economic consequences of introducing classical swine fever into the Netherlands by feeding swill to swine H.S. Horst, R.B.M. Huirne & A.A. DLikhuizen Department of Farm Management, Wageningen Agricultural University, Hollandseweg 1, 6706 KN Wageningen, The Netherlands Summary An effective animal disease prevention and eradication programme is of great importance for meat-exporting countries such as the Netherlands. If a serious outbreak of disease w e r e to occur, the eradication measures required by the European Union and a possible ban on meat exports would have severe economic consequences. However, historical and experimental information on which these programmes can be based is scarce. Furthermore, until recently, an integrated approach which combined the various aspects of outbreaks and risks with economic consequences was lacking. This paper describes a project based on such ah integrated approach. The project covered the elicitation of expert knowledge and the development of the virus introduction risk simulation model (VIRiS). VIRiS integrates objective and subjective information concerning risks and consequences of virus introduction, and thus presents policy-makers with a useful tool for the evaluation of existing prevention programmes and possible alternatives. VIRiS is illustrated for classical swine fever. A comparison is made between the current situation and a hypothetical situation where the risk factor 'swill feeding' is completely eliminated. Keywords Classical swine fever (hog cholera) - Economics - Risk analysis - Simulation modelling Swill - Swine - The Netherlands. Introduction control programmes are of great importance government and private industry (14). to both The export of livestock and meat is very important for the Dutch livestock industry. In 1 9 9 4 , the export of livestock products represented a value of over US$4.5 billion ( 1 3 ) . Outbreaks of classical swine fever (CSF) (hog cholera) and other Office International des Epizooties (OIE) List A diseases, such as foot and mouth disease (FMD) and swine vesicular disease (SVD), are likely to cause export bans and are therefore of major concern to the Netherlands and other meat-exporting countries. European Union (EU) regulations demand strict adherence to a prescribed set of actions for the eradication of CSF outbreaks: measures include the establishment of surveillance zones with complete standstill of all animals, stamping-out of infected herds, etc. Outbreaks would clearly have a serious economic impact, even if export bans were not enforced. Hence, adequate prevention and Concerted action by the government and private industry in the Netherlands has resulted in a research project aimed at gaining more insight into the risk and economic consequences of virus introduction into the Netherlands. Extensive research into contagious animal diseases has been conducted, including risk assessment studies aimed at evaluating the risks associated with the introduction of certain single commodities. However, an integrated model which combines the various aspects of outbreaks and risks with economic consequences, and which takes into account a number of risk factors instead of focusing on one single commodity, was still lacking. Therefore, a general outline for such an integrated approach was developed and described (7). Using the integrated approach, a simulation model was developed to evaluate the process of virus introduction and to 208 identify the key factors. This model provides a useful tool for policy-makers to maximise the efficiency of existing disease prevention programmes and to evaluate possible alternatives. The model, called the virus introduction risk simulation model (VTRiS), is based on historical and experimental data, where available and applicable. However, outbreaks of CSF (and other notifiable diseases) occur irregularly in time, place and magnitude, making it very difficult to derive general properties and predictive values. Therefore, expert knowledge was required and was used as additional information. The research project was aimed at the development of a general model, to be applied to several OIE List A diseases. This paper illustrates the application of this approach for CSF and presents an overview of the project, covering both the elicitation of expert knowledge about the factors which influence virus introduction, and the development of the simulation model VIRiS. By way of illustration, VFRiS results for CSF are presented, with special attention given to the influence that swill feeding has on the expected number of outbreaks. The paper concludes with a brief discussion of the approach and application of the results. Materials and methods Input for virus introduction risk simulation model Quantitative information on recent outbreaks of major diseases is stored in databases such as the Animal Disease Notification System of the EU. However, changing circumstances such as free trade within the EU, the new trade contacts with Eastern European countries, ongoing developments in animal health programmes and animal health legislation, etc., make it difficult to base assumptions and estimates on these data. Another possible information source is research. Many researchers (2, 1 1 , 1 5 ) have conducted work in the area of contagious animal diseases, but most of their findings are qualitative only and do not deal with the first introduction of virus into a country (i.e., the cause of a primary outbreak). Despite this lack of 'objective' information, decisions concerning eradication and prevention of outbreaks must still be made. Such decisions rely on the expertise (a combination of experience and understanding of current/future circumstances) of those working in this area, which is a useful and necessary addition to the data available from research and databases. All three sources of information (databases, research and expert information) were used to provide sufficient quantitative information for the simulation model VIRiS. Using sensitivity analysis, the simulation model could be of Rev. sci tech. Off. int. Epiz., 16 (1) use in evaluating the relevance of this uncertain knowledge for the decision-making process. Workshop experiment General To elicit and quantify expert information, workshops were organised and were attended by 4 3 experts: an expert was defined as a person who has working experience with the Dutch prevention and eradication programmes concerning the diseases under study, which included researchers, policy-makers, field veterinarians etc. During the workshops, participants were asked to complete a computerised questionnaire. The questionnaire was self-explanatory in order to minimise the interaction of the participants (with each other or with the workshop facilitators). Questions concerned the relative importance of risk factors which might cause virus introduction, the expected number of outbreaks in European countries, and the efficiency of the disease prevention and eradication systems in these countries. The participants chose the disease on which they thought themselves to be most knowledgeable: questions were answered in relation to this disease only. For pragmatic reasons, European countries were divided according to geographic and trade criteria into the following five groups (all respondents considered all groups): - Neighbouring countries (Belgium, Germany, Luxembourg) - Southern Europe (Greece, Italy, Portugal, Spain) - Central Europe (Austria, France, Switzerland) - Eastern Europe - 'Islands' (Ireland, Scandinavia, United Kingdom) Relative importance of risk factors Risk factors were defined as possible causes for virus introduction: selection was based on literature and earlier in-depth interviews with experts. The technique of conjoint analysis (CA) was used to estimate the relative importance of these factors. CA is widely used in market research to elicit the subjective views and preferences of consumers ( 4 ) . The technique is based on the assumption that a product or event can be evaluated as a package of characteristics or attributes and that the-overall judgement of the package is based on the individual importance of each of these attributes. CA starts with the overall judgements of respondents on a set of alternatives (i.e., combinations of attributes) and breaks these overall judgements down into the contribution of each. attribute, by using ordinary least-squares (OLS) regression analysis. Horst et al. (8) conducted a small experiment using the CA technique to elicit expert knowledge on the relative importance of risk factors concerning the introduction of contagious animal diseases. The experiment yielded promising results, and was used as a guide and framework for the workshop experiment described in this paper. In this 209 Rev. sci tech. Off. int. Epiz., 16 (1) experiment, respondents were presented with a number of profiles on their computer screen (one profile at a time) and were asked to give a score, on a scale of 0 to 100, for the risk of each profile. These profiles were (theoretical) combinations of risk factors which could be either present or not present, as illustrated below. for all country groups, was later omitted for the same reason. According to the participating experts, the import of animal products is only a significant risk factor when considering virus introduction from the countries surrounding the Netherlands and the countries of southern Europe. Expected number of outbreaks Example of a question based on conjoint analysis Imagine the following situation. What score would you give this combination of risk factors as a means of introducing CSF virus from Belgium, Germany or Luxembourg into the Netherlands? Give a score between 0 and 100. The higher the score the greater the risk. Risk factor combination: - Import of livestock present - Import of animal products not present - Feeding of imported swill present - Tourists present - Retarning trucks not present - Wildlife not present - Air currents (airborne transmission) not present. Using several statistical techniques (including cluster and regression analysis), the individual importance of the risk factors was calculated. Results for CSF, the disease for which the greatest number of participants responded (19 out of 4 3 participants), are presented in Table I. For all country groups, 'import of livestock' was evaluated as the most important risk factor. Other important factors were returning (uncleaned) trucks and the feeding of imported swill. 'Wildlife' and 'air currents' were only considered for the group of countries bordering the Netherlands (neighbouring countries): 'air currents' proved to be an insignificant factor and was not considered for further analysis. 'Tourists', a factor considered Table I Relative importance of classical swine fever risk factors Risk factor Import of livestock Animal products Swill Returning trucks Wildlife Total Neighbouring countries: Southern Europe: Central Europe: Eastern Europe: ls| ands: NS: Neighbouring Southern Central Eastern 'Islands' countries Europe Europe Europe 0.59 0.07 0.15 0.14 0.05 1.00 0.57 0.06 0.18 0.19 - 1.00 0.63 NS 0.17 0.20 - 1.00 0.67 NS 0.16 0.17 - 1.00 0.65 NS 0.16 0.19 - 1.00 Belgium. Germany and Luxembourg Greece, Italy, Portugal and Spain Austria, France and Switzerland Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Hungary, Poland, Romania, Russia, Slovakia, Slovenia Ireland, Scandinavia and the United Kingdom not significant In simulating virus introduction from European countries into the Netherlands, an important input parameter is the expected number of outbreaks in the exporting countries. It was expected that such estimates would be difficult to make for the experts and therefore a method was chosen which enabled expression of uncertainty. This method, called ELI (elicitation), is a graphically oriented computer programme which facilitates the quantification of subjective knowledge about uncertain quantities. The method is based on the establishment of probability density functions (PDF) which reflect the opinion of a respondent. The top of a PDF indicates the best estimate made by the respondent: the dispersion corresponds with the uncertainty about this best estimate. ELI was developed and described by Van Lenthe ( 1 7 ) . Table II presents some results of this part of the workshop experiment. The experts were asked to estimate the number of outbreaks within a five-year period. Table II gives an indication of their 'best estimates' (in fact, the numbers represent the maximum likelihood estimates [MLE] of these best estimates). According to the experts, Eastern Europe may expect the highest number of outbreaks, followed by the countries surrounding the Netherlands. These results concur with reports on the actual number of outbreaks over the past ten years ( 5 ) . High risk period The high risk period (HRP) defines the period in which the virus is already present in a country but not yet detected or under control. The length of this period is one of the most crucial parameters which determine the magnitude of an outbreak. During the HRP, the virus might be transferred to another country, and a lengthier HRP will obviously increase the total risk of virus introduction from the affected country into another country. The HRP can be divided into two periods, namely: - HRP1 : starts when a first animal is infected, and ends when an infected animal is detected. - HRP2: starts with first detection, and ends when all measures are considered effective. The length of HRP1 depends on the alertness and motivation of farmers and veterinarians. HRP2 is determined by the efficiency of the animal health system of the affected country. After HRP2 has ended, the affected region is no longer of risk to other areas. The participating experts were asked to provide a three-point estimate (minimum, most likely and maximum expected length) for the length of both HRPs. This estimate was 210 Rev. sei tech. Off. int. Epiz, 16 (1] Table II Expected number of classical swine fever outbreaks in a five-year period, and the maximum likelihood estimates for duration of the high risk period Parameter Number of outbreaks MLE for HRP1 (days) MLE for HRP2 (days) Neighbouring countries: Southern Europe: Central Europe: Eastern Europe: Islands: NL: MLE: HRP: Neighbouring countries Southern Europe Central Europe Eastern Europe Islands' NL 15.3 27.5 8.1 13.4 37.9 22.6 5.1 25.1 17.8 22.1 43.1 26.0 0.9 23.7 15.9 2.5 23.1 4.1 Belgium, Germany and Luxembourg Greece, Italy, Portugal and Spain Austria, France and Switzerland Bosnia and Herzegovina, Bulgaria, Croatia, Czech Republic, Hungary, Poland, Romania, Russia, Slovakia, Slovenia Ireland, Scandinavia and the United Kingdom The Netherlands maximum likelihood estimate high risk period requested for all groups and for the Netherlands. Table II presents the MLEs for the estimate of the most likely duration of both HRPs. The lengthiest HRPs are expected in Eastern Europe. Short HRPs are expected for the United Kingdom, Ireland and Scandinavia ('Islands') and for the Netherlands. To interpret the results, a comparison was made with findings of Terpstra (16), who established HRPs for recent outbreaks in Europe ( 1 9 8 6 - 1 9 9 6 ) . The estimates of the experts were all within the range of values established by Terpstra. However, that range was based on only a small and rather dispersed number of outbreaks, which is one of the reasons why historical data alone were thought to be insufficient as model input data. Virus introduction risk simulation, a Monte Carlo simulation model The results given in Table II describe only the central tendency (by means of MLEs) of the input parameters. But, as described above, information on the variation in these values was also obtained. To incorporate this information into the proposed simulation model, Monte Carlo modelling was used. This type of modelling uses randomly drawn numbers from specified PDFs (probability distributions) and therefore acknowledges the fact that there is variation in the system to be modelled (9). This variation might be due not only to uncertainty of the input parameters but also to the nature of the system itself (i.e., in reality outbreaks may vary in duration, number of outbreaks may differ between years). When such a Monte Carlo simulation model is run repeatedly, the distribution of output values will realistically reflect the possible behaviour of the system under study. A complete VIRiS-run simulates the events of a one-year period. The model starts with sampling for new outbreaks in the groups defined and 'follows' the disease (taking into account HRP, importance and volume of risk factors, etc.), up to a possible introduction of virus into the Netherlands. Distributions based on quantitative information on epidemiology, animal density and trade flows (statistics of the National Product Boards) determine the location of the outbreaks which could occur in the Netherlands (in four regions: north, east, south, west). The number, location and cause (group and risk factor) of outbreaks are stored for each run and, after a number of runs, these present the user with a distribution of possible outcomes with regard to primary outbreaks. Thereafter, the model of Saatkamp (14) was used to simulate the number of secondary outbreaks. Using information originating from databases, research (literature) and experts, a base scenario was constructed. Since some of the information was only available as relative numbers (such as the relative importance of risk factors), the number of expected outbreaks (Table II) was used to calibrate the model, i.e., translate the relative numbers into absolute values while keeping the relative 'distances' intact. Therefore, the base scenario for CSF is calibrated on approximately one outbreak every two years. Next, changing (some) input parameters, and thus simulating alternative situations, provides insight into the influence of these parameters on the number of outbreaks in the Netherlands. Such an alteration, demonstrated by the elimination of the risk factor 'swill' from the calculation, is evaluated below. More insight into the significance of the results provided by VIRiS might be obtained by calculating economic repercussions. Meuwissen et al. (12) calculated the economic effects of CSF outbreaks if stamped out in accordance with EU disease eradication regulations, and concluded that the total loss from such an outbreak in the Netherlands ranged from US$0.7 million to U S $ 1 3 0 million, depending on the region affected (US$0.7 million in the northern region, US$130 million in the southern region of the Netherlands). As could be expected, there is high correlation between total losses and the animal and farm density of the regions involved. The values calculated include losses to all links in the production chain (slaughter of infected herds, production standstill on infected farms, movement restriction inside protection and surveillance zones, market support through government compensation, and organisational costs). The calculation was based on the (optimistic) scenario that the outbreak did not 211 Rev. sci tech. Off. int. Epiz., 16 (11 result in trade bans. The implementation of such trade bans may result in far more serious losses (3). Results of the virus introduction risk simulation Figures 1 and 2 show some results calculated by VIRiS for the basic scenario where the expected number of CSF outbreaks in the Netherlands equals 2 . 5 per five-year period. In this scenario, the southern region of the Netherlands (41%) is the area most likely to be affected by a primary CSF outbreak; fewer outbreaks ( 1 2 % ) are expected in the north (where the density of pigs and farms is lowest). Figure 1 illustrates the relative importance of risk factors responsible for these potential outbreaks. By far the most significant risk factor is the import oflivestock, accounting for 5 5 % of the outbreaks. Other important risk factors are the feeding of swill and insufficient cleaning of (returning) trucks which were used to export pigs from the Netherlands to other European countries. Figure 2 shows that a CSF outbreak in the Netherlands would be most likely to originate from outbreaks in Greece, Italy, Portugal and Spain (Southern Europe). This is mainly due to the large number of pigs imported from these countries, in particular from Spain. Neighbouring countries (Germany, Belgium and Luxembourg) would also be an important source for CSF introduction into the Netherlands, because of the extent of trade contacts (import as well as export) between the Netherlands and these countries. Country group Fig. 2 Sources of primary classical swine fever outbreaks in the Netherlands: country groups calculated the variation in losses between outbreaks (due to variation in the number of farms involved). A small simulation study using the @RISK software package, which was based on data regarding distribution of losses ( 1 2 ) and the VIRiS results relating to the regional distribution of outbreaks, resulted in cumulative probability distribution of annual losses due to CSF outbreaks in the Netherlands (Figure 3 ) . This Figure shows that annual losses do not exceed several million US dollars in approximately 2 5 % of the cases. In these cases, outbreaks only occur in the north. When outbreaks occur in other regions, losses are much higher (indicated by the twist in the graph) but there is a 9 5 % chance that they will not exceed U S $ 1 0 0 million. Figure 3 also shows the distribution of losses in a 'no swill' scenario. The risk factor 'feeding of imported swill' is expected to account for approximately 2 0 % of the CSF outbreaks in the Netherlands, despite the fact that swill feeding is officially forbidden (swill was thought to be the source of the 1 9 9 2 outbreak of CSF). A complete elimination of this factor therefore decreases the number of outbreaks by 2 0 % , to an Fig. 1 Sources of primary classical swine fever outbreaks in the Netherlands: risk factors As mentioned previously, Meuwissen et al. (12) calculated the losses due to CSF outbreaks in each of the four regions. Combining their findings with the likelihood of each region being affected by an outbreak, results in an expected average loss of approximately U S $ 7 6 million per outbreak or U S $ 3 8 million per year. However, Meuwissen et al. ( 1 2 ) also Fig. 3 Cumulative distribution of annual losses due to classical swine fever outbreaks in the Netherlands, base scenario and 'no swill scenario' 212 average of two outbreaks per five-year period. However, as swill is more likely to cause an outbreak in the west than in the other regions of the Netherlands - the busiest harbours and the main airport for the country are located in the west, thus the highest concentration of imported swill is to be found here - the decrease in annual losses does not equal exactiy 2 0 % . Figure 3 shows the resulting cumulative distribution of annual losses for this scenario. The 5 0 % percentile in the swill-free scenario is at U S $ 2 7 million, which is US$6 million less than in the basic scenario. Concluding remarks Important steps involved in risk management are the identification and assessment of risks and the development, analysis and evaluation of alternative strategies ( 1 , 6, 1 0 ) . With the development of VIRiS, a serious and successful attempt has been made to combine these steps into one model. The quality of the results of a model heavily depends on the quality of the input: 'rubbish in = rubbish out'. A major part of the input of VIRiS is based on the subjective opinion of experts. The quality of this input is difficult to judge as there is no standard available for comparison. However, as long as research does not provide better values, these expert estimates represent the best information available. Nevertheless, while using the VIRiS model, continuous effort must be made to enhance the quality of the input data. The model is constructed in such a way that changes in these values can be made easily. Designed as a tool to assist policy-makers, the major task for VIRiS is to illustrate the consequences of alternative strategies and the (potential) impact of uncertain (or unknown) input values. One example of analysis of an alternative strategy was presented in this paper. Eliminating the risk factor 'feeding of imported swill' resulted in a decrease of annual losses of US$6 million or 1 5 % in the 5 0 % percentile (median). Such information might be helpful in the process of deciding what measures need to be taken and how much money should be set aside to minimise the risk of virus introduction in a cost-efficient way. Rev. sci tech. Off. int. Epiz., 16 (1) Classical swine fever outbreak in the Netherlands Since 4 February 1997, the Netherlands has been confronted with an outbreak of CSF. Unfortunately, the outbreak occurred in the southern region of the country, where the animal and farm densities are highest. Simulation results showed that in this area, losses incurred by the disease may exceed U S $ 4 0 0 million, excluding losses due to market disruptions. By April 1 9 9 7 , more than 1 2 0 farms were diagnosed as being infected with CSF, and this number may continue to increase. Preliminary results of tracing studies indicate that the causal virus is of the same type (Chinese) as the virus which caused outbreaks in Paderborn, Germany, at the end of 1996. The HRP 1 is an estimated 4.5 weeks, which is more than one week longer than the MLE reported in the simulation study (but within the variation expected to occur around the reported value). Moreover, the enforcement of movement restrictions proved to b e a problem in the early days, after identification of the first infected premises. Therefore, the Netherlands might be facing a 'worst case scenario'. Apart from the financial consequences of market disruption (the trade ban on live animals), which are not yet known, the highest losses will be caused by government purchase and destruction of pigs from farms located within the protection zones where movement of pigs is prohibited. This measure has been taken to avoid welfare problems, further spread of the virus, and disruption of the internal market due to an oversupply of animals (the bulk of production is normally exported). The Dutch Ministry of Agriculture estimates the costs of this measure at approximately U S $ 2 4 0 million for the entire outbreak. However, total losses cannot be calculated until the outbreak is eradicated completely and all restriction measures have been lifted, which is not to be expected before September 1 9 9 7 . Acknowledgements The authors wish to thank Nell Ahl and Paul Sutmoller for their valuable comments on the first draft of this paper. 213 Rev. sci tech. Off. int. Epiz., 16(1] Risques et conséquences économiques d'une introduction de la peste porcine classique aux Pays-Bas par l'alimentation des porcs avec des eaux grasses H.S. Horst, R.B.M. Huirne & A.A. Dijkhuizen Résumé Il est de la plus haute importance, pour un pays exportateur de viande comme les Pays-Bas, de se doter d'un programme efficace de prévention et d'éradication des maladies animales. Si un foyer grave venait à se déclarer, les mesures d'éradication requises par l'Union européenne et une interdiction éventuelle des exportations de viande auraient des conséquences économiques très graves. Toutefois, les informations historiques et expérimentales sur lesquelles fonder de tels programmes sont plutôt rares. De plus, jusqu'à une époque récente, il n'existait pas d'approche intégrée combinant les différents aspects liés à un foyer et à ses risques avec ses conséquences économiques. Les auteurs décrivent un projet qui s'appuie sur une approche de ce type. Le projet a fait appel aux connaissances des experts mais a eu également recours à une modélisation par ordinateur du risque d'introduction de virus (modèle VIRiS). Ce modèle intègre des informations objectives et subjectives sur les risques et les conséquences de l'introduction de virus, et constitue un outil efficace permettant aux autorités responsables d'évaluer les programmes de prévention existants et les alternatives possibles. Le modèle VIRiS est ici appliqué à la peste porcine classique. Une comparaison est établie entre la situation actuelle et une situation hypothétique où le facteur de risque lié aux eaux grasses dans l'alimentation des porcs serait complètement éliminé. Mots-clés Analyse des risques - Eaux grasses - Économie - Modélisation par ordinateur - Pays-Bas - Peste porcine classique - Porcins. Riesgos y consecuencias económicas de la introducción de la peste porcina clásica en los Países Bajos a través de la alimentación de cerdos con desechos líquidos H.S. Horst, R.B.M. Huirne & A.A. Dijkhuizen Resumen Para los países exportadores de productos cárnicos como los Países Bajos, disponer de un programa eficaz de prevención y erradicación de las enfermedades animales reviste la mayor importancia. En caso de brote de enfermedad de cierta gravedad, las medidas de erradicación exigidas por la Unión Europea y la posible prohibición de exportar productos cárnicos no dejarían de entrañar severas consecuencias económicas. Sin embargo, es escasa la información histórica y experimental sobre la que puede basarse un programa de prevención y erradicación. Por añadidura, hasta hace poco tiempo se carecía de un enfoque integrado que combinara los diversos aspectos relativos a los brotes infecciosos y a sus riesgos con sus consecuencias económicas. Los autores describen un proyecto basado en un enfoque integrado de este tipo. Dicho proyecto fue llevado a cabo gracias a la consulta de expertos y al 214 Rev. sci tech. Off. int. Epiz, 16(1) desarrollo de un modelo de simulación del riesgo de introducción de un virus (modelo VIRiS). VIRiS integra informaciones de índole tanto objetiva como subjetiva acerca de los riesgos y consecuencias de la introducción de un virus, y constituye por ello una herramienta muy útil gracias a la cual las autoridades públicas pueden evaluar los programas de prevención existentes y las posibles alternativas. El funcionamiento de VIRiS viene ilustrado mediante el ejemplo de la peste porcina clásica. A tal efecto se lleva a cabo una comparación entre la situación actual y una hipotética coyuntura en la que se hubiera eliminado por completo el factor de riesgo «alimentación con desechos líquidos». Palabras clave Análisis de riesgos - Cerdo - Desechos líquidos - Economía - Modelos de simulación Países Bajos - Peste porcina clásica. • References 1. Ahl A.S., Aeree J.A., Gipsern P.S., McDowell R.M., Miller L. & McElvaine M.D. (1993). - Standardization of nomenclature for animal health risk analysis. In Risk analysis, animal health and trade (R.S. Morley, ed.). Rev. sci. tech. Off. int. Epiz., 12 (4),1045-1053. 2. Becker Y. (1987). - Epidemiology of African swine fever virus. In African swine fever (Y. Becker, ed.). Martinus Nijhoff Publishers, Boston, 145-150. 3. Buijtels J.A.A.M. (1997). - Computer simulation to support policy-making in Aujeszky's disease control. PhD thesis. Wageningen Agricultural University (in press). 4. Cattin P. & Wittink D.R. (1982). - Commercial use of conjoint analysis: a survey. J. Mark., 4 6 , 44-53. 5. Davies G. (1994). - An overview of recent epidemics of animal disease. In Proc. Workshop on 'Animal health and related problems in densely populated livestock areas in the Community'. European Community, Brussels, 35-40. 6. Hardaker J.B., Huirne R.B.M. & Anderson J.R. (1997). Coping with risk in agriculture. CAB International, Wallingford, United Kingdom, 288 pp. 7. Horst H.S., Dijkhuizen A.A. & Huirne R.B.M. (1996). Outline for an integrated modelling approach concerning risks and economic consequences of contagious animal diseases. Neth.J. agric. Sci., 44, 89-102. 8. Horst H.S., Huirne R.B.M. & Dijkhuizen A.A. (1996). Eliciting the relative importance of risk factors concerning contagious animal diseases using conjoint analysis: a preliminary survey report. Prev. vet. Med., 27, 183-195. 9. Kleijnen J.P.C. & Van Groenendaal W. (1992). - Simulation: a statistical perspective. John Wiley & Sons Ltd, Chichester, England, 241 pp. 10. MacDiarmid S.C. (1993). - Risk analysis and the importation of animals and animal products. In Risk analysis, animal health and trade (R.S. Morley, ed.). Rev. sci. tech. Off. int. Epiz., 12 (4), 1093- 1107. 11. Mann J.A. & Sellers R.F. (1989). - Foot-and-mouth disease virus. In Virus infections of porcines (M.B. Pensaert, ed.). Elsevier Science Publishers, Amsterdam, 251-258. 12. Meuwissen M.P.M., Horst H.S., Dijkhuizen A.A. & Huime R.B.M. (1997). - Risks and consequences of outbreaks of notifiable diseases in the Netherlands (in Dutch). Wageningen Agricultural University, 164 pp. 13. Productschap Voor Vlees En Eiren (PVE) (Products Boards for Livestock, Meat and Eggs) (1995). - Livestock, meat and eggs in the Netherlands (in Dutch). PVE, Rijswijk, 49 pp. 14. Saatkamp H.W. (1996). - Simulation studies on the potential role of national identification and recording systems in the control of classical swine fever. Mansholt Studies 2, Backhuys Publishers, Leiden, 120 pp. 15. Terpstra C. (1987). - Epizootiology of swine fever. Vet. Q., 9, 50s-60s. 16. Terpstra C. (1996). - Analysis of high risk period. Internal publication (in Dutch). Wageningen Agricultural University, 3 pp. 17. Van Lenthe J. (1993). - Eli; the use of proper scoring rules for eliciting subjective probability distributions. DSWO Press, Leiden University, The Netherlands, 140 pp.