Combined spatial and network analysis of Great Britain pig movement and abattoir surveillance data

Aims and objectives:

• Provide a high-quality movement dataset;
• Obtain epidemiologically relevant network and spatial properties;
• Pig industry characterization regarding the animal movements and network / spatial relations;
• Identification of relationships between three example diseases and network characteristics;
• Characterisation of the weight between spatial and network properties affecting disease prevalence and preventive/control measures;
• Modelling network and spatial structure in relation to prevalence of other key pathogens affecting our pig industry;
• Study the contribution of spatial and network analysis in the definition of compartmentalisation;
• Assess the impact that preventive and control measures have on networks;
• Recommendations on the targeting of surveillance and intervention measures based on network structure and spatial characteristics;
• Development of generic research skills – including planning research, acquiring & assessing data quality, communicating scientific results;
• Development of specific research skills – including data cleaning and manipulation, spatial, network and statistical analyses

Summary of findings 

In the last decade, the British pig industry has taken a proactive stance concerning the health and welfare of pigs. On a national level, the industry implemented two major disease monitoring tools (BPHS and WPS) to provided producers and veterinarians with valuable information on farm disease prevalence. Pig production is not a closed system restricted by geographical boundaries, however, not only neighbouring factors but also animal movements may contribute to disease transmission. As a consequence of the Foot and Mouth Disease outbreak in 2001, recording of pig movements has become a requirement and these records have become a source of surveillance data available for epidemiological analysis. This information allows the identification of the “active” pig holdings and provides information about the inter-holding contacts and their geographical location. Through the use of pig movement data and spatial and network analysis, this project aimed to identify and characterise the “active” British pig holdings regarding their demographic, spatial and network structure.

The datasets used in the project were: a) Rapid Analysis and Detection of Animal-related Risks (RADAR) dataset (government-held pig movement records, combined movements from Animal Movement Licensing System (AMLS) and Scottish Animal Movement System (SAMS), subset of movements from years 2006-2011); b) Sprod and Eprod datasets (pig movement records from 2008 from one Scottish and one English pig producer, respectively); and c) BPEX Pig Health Scheme (BPHS) and Wholesome Pigs Scotland (WPS) datasets (abattoir monitoring data, from 2006-2009, for England/Wales and Scotland, respectively). The movement data was obtained from RADAR, the producer’s datasets were obtained directly from the producers, BPHS data was obtained from BPEX and WPS data from Quality Meat Scotland (QMS). The records provided by RADAR were submitted to a cleaning process and explored both for illogical and inconsistent information (e.g. correction of inconsistencies in the holding type, renaming of “Unknown”, “Non-AMLS” and “Common Land” County/Parish/Holdings (CPHs) as “Farm”; removal of movements with the same CPH number in departure and destination; removal of “illogical” or irrelevant movements such as Abattoir-to-Farm or Market and Abattoir-to-Abattoir).

The datasets were compared with the objectives of quantifying: 1) the number of comparable movements between datasets; 2) the agreement between the numbers of animals reported; and 3) the number of non-comparable records. For this comparison a subset of RADAR dataset was used (2006-2009). During this exercise, systematic errors were identified and reasons for non-matching and non-agreement were found, e.g. the comparison of different dates, the problem of aggregated data and data entry errors. The comparison was possible for only 7,724 (1%) of all RADAR movements and 74 movements were found to be absent from the RADAR dataset provided. There was an agreement in the number of animals reported in 5,605 movements (73%), the overall mean difference of animals obtained was -0.16 and 96% of the non-agreement records had differences less than 52 animals. This highlighted that the majority of the comparable movements show an agreement in the numbers reported and that only a small amount of movements were identified as under- reported for the RADAR dataset. Therefore, the RADAR dataset was considered a useful source for analysing the pig movements in GB and was therefore used for the descriptive analysis.

The description of British pig holdings regarding their demographic, spatial and network structure, over 2006 to 2011, showed that the number of holdings included in the movement database varied over the years but, overall, the farm CPHs accounted for 98.6% of all CPHs in the movement dataset. When looking at movements over the years, 70% of the movements were to abattoirs, 23% to farms and 7% to market. There was an average of 200,000 movements per year. In terms of pigs being moved, on average there were approximately 5 million pigs moved towards farms, 8 million to abattoirs and only 160,000 to markets, resulting in a total of around 14 million pigs moved per year in GB. The majority of pig farms (82%) were located in England, while Wales account for 10% and Scotland for 8%. The farm distribution over the different GB NUTS II regions showed some regions of England had the highest density of farms per region, e.g. Cornwall and Isles of Scilly (30 farms/100 km²) and Dorset and Somerset (24 farms/100 km²). In Wales, the region of West Wales and The Valleys (11 farms/100 km²), and in Scotland the northeast of the country (5 farms/100 km²) had the highest density of farms per region. In GB, 27% of the farms had less than 5 other farms in a 3km radius, 43% had between 5 to 10; and 30% more than 10 surrounding farms. This highlights the risk for spreading airborne diseases. However, the majority of farms in Scotland (84%) had only between 1 to 5 farms in the vicinity, showing a decrease risk in terms of local spread.

From the total of 1,161,406 pig movements that occurred in GB, from 2006 to 2011 to which was possible attribute a geographic location, the departure and destination of 56% were in the same region and 97% in the same country. Movements between regions were largely observed when pigs were moved from a market to an abattoir, whereas the opposite happened when pigs were moved between markets, i.e. they tend to happened within the same region. The maximum distances travelled were observed in movements between farms (792km) and from farm to abattoir (767km). The median distance of pig movements was approximately 31km for the majority of movement types, except for movements market-to- abattoir (median=53km) and market-to-market (median=71km). Only 14% of the movements involved distances of over 100km. The large majority of within-country pig movements occurred in England (91%) with a median of 30km and a maximum of 563km travelled. Movements between countries accounted for only 3% of the total pig movements. The majority of across GB country movements occurred from Wales to England (38%), Scotland to England (34%) and England to Wales (24%). Only a small fraction of movements occurred from England to Scotland and between Scotland and Wales. The median batch size of the between-farm moments was equal within countries. Scotland sent larger batches to markets (8 pigs per batch) than England and Wales; abattoirs in Scotland also received larger batches (median of 18 and maximum of 668) than England (median of 10 and maximum of 299).

The number of sources (in-degree) and destinations (out-degree) was assessed for each type of CPH. The median in and out degree for farm and markets was one; the median in- degree for abattoirs was 24 for English abattoirs, 14 for Welsh abattoirs and 4 for Scottish abattoirs. Regardless of the type of movement, England always had the highest in and out degree, followed by Scotland and Wales.

With the 2008 RADAR data, the farms were categorised as “Large” when they had moved more than 1000 animals; they were categorised as “Small”, when their total number of animals moved was less than 35. The farms having the total number of animals moved in the range [35 - 1000] were considered “Medium” producers. The proportions for the totality of farm CPH (18,565) were: 2,470 (13%) “Large”, 3,486 (19%) “Medium” and 12,609 (68%) “Small”.

The higher percentage of farm to farm movements was observed between large producers and at the same time more animals were moved by these producers, with a higher median number of animals transported per movement. The large producers were also responsible for the majority of the number of movements and number of animals moved from farm to abattoir. The medium producers, on the other hand, had the higher proportion of movements between farm and gathering holdings. The small producers did not have a large number of movements or animals moved; however, there were a small number of movements from small to medium and large producers.

These types of unexpected movements should be explored in order to see if they might pose a higher risk of disease transmission, or whether they are genuine movements or just errors in the database. The movements between large producers or from large to abattoir were the ones involving the greatest distance. The small producers had the smallest distances travelled, irrespective of the type of movement