Veterinaria Italiana, 43 (3), 541-548 Spatial distribution of brucellosis in sheep and goats in Sicily from 2001 to 2005 Augusta D’Orazi(1), Manuela Mignemi(1), Francesco Geraci(1), Angela Vullo(1), Michele Di Gesaro(1), Stefano Vullo(1), Michele Bagnato(2) & Calogero Di Bella(1) Summary Distribuzione spaziale della Spatial analysis is making an increasingly important contribution to surveillance brucellosi ovi-caprina in Sicilia measures due to its ability to enable immediate dal 2001 al 2005 visualization of information on the phenomenon studied. The authors describe the Riassunto spatial distribution of prevalence and Nell’ambito dell’applicazione delle misure di incidence of brucellosis in small ruminants in sorveglianza l’impiego della analisi spaziale sta Sicily between 2001 and 2005. The study was assumendo un ruolo sempre più rilevante, grazie conducted by integrating geographic alla rapidità di impiego e all’efficacia che deriva information systems (GIS) technology dalla visualizzazione immediata delle informazioni (MapInfo® Professional 7.0) with SaTScanTM sul fenomeno oggetto di studio. In questo lavoro gli software to perform an epidemiological autori descrivono l’andamento spazio-temporale analysis of the municipalities and to locate della prevalenza e dell’incidenza della brucellosi problem areas. A comparison between the ovicaprina in Sicilia nel corso degli anni 2001-2005. thematic maps produced for brucellosis in Lo studio è stato condotto integrando la tecnologia ® small ruminants on the basis of prevalence and GIS (MapInfo Professional 7.0) con il software TM incidence data for each individual year has SaTScan per la caratterizzazione epidemiologica shown that in terms of prevalence, the area dei territori comunali e l’evidenziazione di “aree identified as the secondary cluster in 2001 problema”. Il raffronto fra le mappe tematiche, became the primary cluster from 2002 onwards realizzate per ogni singolo anno per la brucellosi whereas, in terms of incidence, the distribution ovi-caprina sulla base dei dati di prevalenza ed of the clusters was irregular throughout the incidenza, ha evidenziato che, relativamente alla entire region during the years studied. prevalenza, l’area individuata come cluster secondario nell’anno 2001 è diventata dal 2002 in Keywords poi cluster primario. In riferimento all’incidenza, Brucellosis, Geographic information system, nel corso degli anni considerati, la distribuzione dei Incidence, Prevalence, Sicily, Small ruminant, cluster è stata irregolare su tutto il territorio Spatial analysis. regionale. Parole chiave Analisi spaziale, Brucellosi, Incidenza, Ovi- caprini, Prevalenza, Sicilia, Sistema informativo geografico. (1) Area Sorveglianza Epidemiologica – Istituto Zooprofilattico Sperimentale della Sicilia ‘A. Mirri’, Via Gino Marinuzzi 3, 90129 Palermo, Italy [email protected] (2) Assessorato alla Sanità della Regione Siciliana, Ispettorato Veterinario, Piazza Ottavio Ziino 24, 90145 Palermo, Italy © IZS A&M 2007 www.izs.it/vet_italiana Vol. 43 (3), Vet Ital 541 Spatial distribution of brucellosis in sheep and goats in Sicily Augusta D’Orazi, Manuela Mignemi, Francesco Geraci, Angela Vullo, from 2001 to 2005 Michele Di Gesaro, Stefano Vullo, Michele Bagnato & Calogero Di Bella Introduction laboratory tests conducted on sheep and goats by the Istituto Zooprofilattico Sperimentale della Brucellosis in small ruminants caused by Sicilia between 2001 and 2005 as part of the Brucella melitensis is a highly contagious national brucellosis eradication programme. disease that has a considerable impact in the Table I shows annual data for the number of animal husbandry sector. It is also considered flocks tested, the number of flocks tested as a to be the most widespread zoonosis in the percentage of the total present in the region, world and is a serious human health hazard. the number of flocks found to be infected and The prevalence of the disease in Italy is the number of newly positive flocks. currently at very low levels in the northern and central regions of the country but remains Table I critical in some southern regions and in Sicily. Brucellosis in small ruminants: tests conducted in Sicily, 2001-2005 Spatial analysis is making an increasingly No. of important contribution to disease control No. of No. of Flocks newly Year flocks positive covered measures due to its ability to provide positive checked flocks (%) immediate visualisation of the distribution of flocks the phenomena under study within the 2001 7 507 1 802 794 75.60% territory. It is commonly used to determine 2002 8 219 1 854 874 92.40% whether a one-dimensional point process is 2003 8 636 1 777 672 98.73% purely random or whether clusters can be 2004 8 666 1 560 603 99.80% detected (3). Rapid detection of emerging 2005 8 775 1 471 623 99.68% geographic clusters due to unexpectedly occurring risk factors can be of great A flock is considered infected (or positive) if it importance for public health (2). In the specific contains at least one animal that tests positive case of this study, this tool was used to to the official serological tests for brucellosis. determine the spatial distribution of ‘Newly positive’ flocks are those that are brucellosis in small ruminants in Sicily in found to be infected or positive after testing order to identify geographic problem areas, negative the previous year. The number of defined as groups of municipalities with a flocks located in each municipality in the higher probability of positive flocks, and to region was defined as the population. For each analyse positivity trends in the areas year, we analysed ‘cases’ within this considered between 2001 and 2005. population (all flocks testing positive for brucellosis, i.e. prevalence) and ‘newly Materials and methods positive’ flocks (those that had become positive during the year, i.e. incidence). SaTScanTM software (M. Kulldorff and Information Management Services, Inc. The principal objective was to establish SaTScanTM v. 3.0: Software for the spatial and whether events (cases of infection) occur space-time scan statistics. Bethesda, National randomly throughout the Sicily region or Cancer Institute, 2002) was used for statistical whether clusters can be identified. For this calculations and MapInfo® software was used purpose, the purely spatial method of to represent the results obtained in Sicily in the SaTScanTM software and the Poisson form of reliable and interpretable spatial probability model were used. The distribution images. In our study, each of the of cases throughout the region was assumed to 390 municipalities making up the region of be random, in other words, there were no Sicily was identified as a centroid. The spatial areas with a higher probability of flocks position of each centroid was identified by becoming positive. geographic coordinates (latitude and longitude To describe the spatial distribution of expressed in decimal degrees). The analysed brucellosis in small ruminants, the Sicily data were obtained from the results of official region was partitioned into cells 542 Vol. 43 (3), Vet Ital www.izs.it/vet_italiana © IZS A&M 2007 Augusta D’Orazi, Manuela Mignemi, Francesco Geraci, Angela Vullo, Spatial distribution of brucellosis in sheep and goats in Sicily Michele Di Gesaro, Stefano Vullo, Michele Bagnato & Calogero Di Bella from 2001 to 2005 corresponding to municipalities. For each µ(z) = expected number of cases inside the municipality, the coordinates of geographic circular window z centroid, number of flocks and number of µ(G) = expected number of cases in the cases of disease were considered. The region. SaTScanTM software created circular windows To solve the equation, we performed a within the area, identifying each municipality simulation using the Monte Carlo technique as the centre of the circle (the centroid). The with 9 999 iterations (3). software included a gradually higher number of municipalities surrounding the centroid by For example, in 2005 each replicate involved increasing the radius from zero (corresponding choosing 1 471 flocks at random from the total to one municipality) to a maximum limit. At number of flocks (8 775) and labelling these as this maximum limit, the window contained a cases. percentage of cases out of the total number of The windows in which the null hypothesis can flocks in the region that was lower than or be rejected (p-value less than or equal to 0.05) equal to the values of regional prevalence or can be defined as clusters. incidence for each year. Each window therefore differed from the others in terms of Results its position and radius. Finally, clusters inside the circles were identified by SaTScan. The prevalence in the region during 2005 was For each window, we constructed a hypothesis found to be 17% (95%, with a confidence system to verify whether the probability of the interval [CI] of 16-18%) and this was taken as disease occurring inside the window was the upper limit for the radius. Table II gives equal to the probability of the disease the results obtained from the simulation of the occurring outside the window (1). To do this, equation. The incidence in the region in 2005 we let G be the geographical space (Sicily was 7% (95%, CI: 6.56-7.64%) and this was region) and Z the set of all circular zones taken as the upper limit for the radius. Table III gives the results obtained from the z ⊂ G (z, p, q) (1). We considered a point in simulation of the equation considering the new p q the parameter space where and vary cases. Tables IV and V give the results of the from 0 to 1 and z is used to denote a vector simulations for 2001-2004. made by the centroid coordinates and the Prevalence in the entire region fell slightly radius of circle and also to denote the zone it over the five year period (Fig. 1). describes (1). The hypothesis system is as Cluster analysis revealed that the first cluster follows: measured in 2001 became the second cluster H : p = q 0 from 2002 to 2005 with a smaller radius (in H : p > q 1 terms of kilometres), covering fewer p : probability of a case occurring in the municipalities.