Integrating administrative data-bases, passive surveillance and GIS.

Integration of data from different sources could provide an effective support in the epidemiological analysis of animal diseases, even in small-scale areas. Our need to integrate GIS (the Veterinary Service of the Province of Reggio Emilia has georeferred, among the ohers, all the bovine, swine, ovi-caprine and poultry farms) and the other administrative data-bases (the National Data Bank and our local data bank ) to gather, manipulate and analyze data from different sources found solution in the use of the ‘R’ software (www.r-project.org - 1). ‘R’ is a highly flexible, expandible and customizable statistical open-source software and an environment which provides all the necessary tools for data analysis and manipulation. R has also the capability of accessing data from a PostgreSQL server (where data are daily exported from the administrative data banks), via the ‘RODBC’ library, and provides tools for reading, writing and manipulating shapefiles (via the ‘maptools’ library) and advanced and modern instruments for the analysis of geographical data (i.e. the packages: ‘splancs’, ‘spdep’, ‘spatial’, ‘spatstat’, ‘Dcluster’). A real-life example of a successful application of this integrated approach is represented by the analysis of echinococcosis passive surveillance data. During the 5-year period 2001-2005, 229 cases of bovine echinococcosis (from 180 farms of the province of Reggio Emilia, Emilia-Romagna Region, Northern Italy), were found at slaughtering through meat inspection, and reported to the Veterinary Service of the Local Health Unit of Reggio Emilia. In order to estimate the prevalence of echinococcosis at the moment of slaughtering, all the data about cattle from the province slaughtered in the above-mentioned period (218,726 records from 2781 farms) were retrieved from the national data bank (the system which tracks all movements of the Italian bovine population). Since all the reports of the disease concerned cows over 2 years of age and became from only two big slaughterhouses in Lombardy region, the dataset and all the subsequent analyses were restricted only to the cows slaughtered in these two slaughterhouses (48,390 cows from 2003 farms), considering them as a random sample (about 46%) of all the cows over 2 years of age sent to the slaughterehouse from farms in the province. Results and discussion. The prevalence of echinococcosis in cows at slaughtering was 4.7 (95% c.i.: 4.15-5.4) cases / 1000 animals. Both empirical Bayes estimates and kernel smoothing techniques were used to investigate the pattern of spatial distribution of the cases. The analysis revealed a heterogeneity in the prevalence estimates by municipality wich was not attributable to the effect of chance alone. In particular, a cluster of cases was identified in the north-western area of the province corresponding to 6 municipalities, where the highest prevalence reached 13.9 cases / 1000 cows. It is worth noting that, in the same area and in the same period, uncontrolled and illegal grazing of infected ovine flocks caused an outbreak of Brucella melitensis infection in cattle, sheep and goats. These findings could suggest that uncontrolled flock grazing could have represented a risk factor for echinococcosis in the cattle coming from the area considered, especially due to the possible presence of infected dogs or the abandoned carcasses and viscera of sheep (2). The results presented here demonstrate how the goal of gathering and coherently assembling data from different sources (passive surveillance, geographical information system and administrative data-bases, such as the National Data Bank) can be succesfully achieved through (and from within) the open-source environment ‘R’, that provided not only the necessary statistical tools but also a common interface to GIS and databases.