POBLACIÓN DE MUJERES ZARZALEÑAS
14. ANÁLISIS DEL MERCADO 1 ANÁLISIS DE LA DEMANDA
14.2 ANÁLISIS DE LA OFERTA
Fulfilling the objectives of this study involved many challenges and development of the BSurvE model has been an evolutionary process. This culminated in the European Food Safety Authority (EFSA) being invited by the EC in 2004 to provide advice on
Table 3.2: nj,t and xj,t: The observed number of animals tested and testing positive in hypo- thetical country A at agetin surveillance streamj in 2004
Age Cattle testednj,t Year of Cattle testing positivexj,t (years) Healthy Fallen Casualty Clinical birth of Healthy Fallen Casualty Clinical
slaughter stock slaughter suspect cohort slaughter stock slaughter suspect
0 2004 1 0 0 45 0 2003 0 0 0 0 2 79,562 800 76 12 2002 0 0 0 0 3 35,874 870 154 16 2001 1 0 0 1 4 12,965 753 79 22 2000 0 3 1 3 5 13,856 851 85 25 1999 2 7 0 9 6 11,938 678 76 31 1998 2 8 0 10 7 7195 486 67 20 1997 3 4 1 7 8 30,607 1738 23 17 1996 1 1 0 2 9 20,259 1150 15 12 1995 1 1 0 1 10 12,811 727 10 7 1994 0 0 0 1 11 7593 431 6 4 1993 0 0 0 0 12 4318 245 3 2 1992 0 0 0 0 13 2219 126 2 1 1991 0 0 0 0 14 995 56 1 1 1990 0 0 0 0 15 385 22 0 0 1989 0 0 0 0 16 80 5 0 0 1988 0 0 0 0 Unknown 0 0 85 10 0 0 0 0
the general approach used within BSurvE. The EFSA Scientific Expert Group convened for the evaluation reported its findings in October 2004 (EFSA 2004).
In addition to these discussions, three two-day practical workshops have been held. This allowed veterinary epidemiologists from EU Member States and other countries to use BSurvE to analyse their country’s surveillance data. As a result of these workshops a number of refinements were made to BSurvE and consideration has been given to providing advice where the available data inputs are lacking in various respects. The following provides a summary of the key issues identified.
National data collection procedures vary and in many countries there is limited availability of data suitable for use in BSurvE. For example, some countries do not record the age of tested animals. Such data can be distributed according to an estimated age distribution but this introduces greater uncertainty in the results, particularly when estimation is done for individual age cohorts. Age and surveillance data are commonly collected within multi-age groups, particularly for older animals, and in order to make prevalence estimations on a cohort basis it is necessary to distribute this data according
to an approximation of the age distribution of a cohort, i.e. the estimated count dt.
The estimated count is determined manually, guided by the age data that is available, i.e. the age distribution of the national population for up to 5 years. Automation of the approximation process was attempted but found to be inadequate due to the wide variability in national data and the need for distribution of different multi-age groups. The collection of national data across a wider age range will improve accuracy and
if gathered over a period of time will allow the construction of the current estimated age counts from actual data. BSurvE makes the assumption that the estimated age distribution is applicable to all cohorts considered in the model (which implies that the population is stable over the analysed time period), although a few countries have had wide variation in population size within a short period of time.
Initial versions of BSurvE accepted age data stratified by industry (dairy and beef), because management systems and age structures are very different between the two production sectors and there is a higher prevalence of BSE in dairy animals (Wilesmith et al. 1988; Stevenson et al. 2000, Ducrot et al. 2003). However, unless all input and surveillance data is allocated by production sector it is not possible to estimate the prevalence within each sector; such information was not available to the authors and would in any case be difficult to collect. Hence, we modified the model to accept total population age data without reference to industry. Analyses could be run for each production sector separately if data were available.
Currently, up to 5 years of surveillance data may be entered and analysed by the user. In a situation where it is not necessary to estimate infection prevalence on a per- cohort basis, data from multiple years can be amalgamated. Some countries categorise animals into a larger number of streams or have different eligibility criteria for the streams used in BSurvE. For the analysis of data within a country the user should interpret the results according to their categorisation of the input data.
Country-specific exit probabilities for uninfected animals (dj|t) should in principle not
be difficult for a country to obtain, regardless of national BSE prevalence. Probabilities
for infected animals (cj|t) would be more easily obtained for a high prevalence country
than a low prevalence/uninfected country because case data can be used to indicate the required values. National studies on the disposal of cattle would assist in providing the required data, and are being undertaken in some countries.
BSurvE assumes that most animals are infected in their first year of life. Epi- demiological studies of the BSE epidemic in Great Britain indicate that there is an age-dependent susceptibility, with cattle in their first six months of life being at most risk (Arnold and Wilesmith 2004). Modelling studies of the French BSE epidemic sug- gest that most cattle were infected between six and twelve months of age (Supervie and Costagliola 2004). The distribution of the time until clinical signs develop reflects these findings. However, allowance for an older age of infection can easily be incorporated into BSurvE by adjusting this distribution. BSurvE also allows the user to set the proportion of preclinical animals that would be detectable if culled in the year prior to the onset of clinical signs. As indicated above, we suggest a range of 0.3 to 0.5, based on current evidence from the validity of the three screening tests that have been used to date, and for which performance characteristics are for practical purposes similar. If more-sensitive screening tests are developed then there is the facility to change the
value of this input parameter.
When BSurvE was used by various countries, prevalence estimation by cohort re- sulted in realistic curves, whereby increases or decreases in the estimate could be related to events such as implementation of feed bans or other statutory controls. The wide confidence intervals for recent cohorts will decrease over time as these cohorts age and more data become available.
This study was able to capitalise on the results of the very large amount of surveil- lance carried out within the EU. The use of BSurvE model has been examined for a wide range of scenarios and can accommodate the various demographic and epidemi- ological situations likely to exist. It is an epidemiological tool with which users can become proficient in a relatively short period of time.