This is the first study to use data from a national electronic system for pig traceability to describe, in detail, pig movements occurring between different types of producers including backyard and hobby producers. This study had a clear objective to provide better information on the level of interaction between the different actors, commercial and non-commercial, in the Scottish swine industry. The recent epidemic of African swine fever in the eastern European region has highlighted the involvement of non-commercial swine producers in the spread of swine fevers via animal movements  and the threat that they may pose to both national and regional swine industries . Therefore, a better understanding of these relationships is crucial for identifying likely routes of disease incursion and transmission prior to disease detection, and identifying weaknesses in biosecurity, surveillance and contingency planning. Such knowledge is a key component in the development of a representative, and therefore more robust, biosecurity and surveillance planning which would secure the resilience of the Scottish swine industry to any incursion of exotic notifiable diseases.
Movement of live pigs leading to disease transmission by direct contact is an important route of transmission for swine pathogens [18, 23]. Therefore quantifying the movement patterns of pig keepers, including small producers, provides valuable information to inform disease models and risk assessments. However, other transmission routes can also be important, depending on the pathogen. Transmission via fomites is another important route of disease transmission for a number of pathogens [18, 24], and could include transmission via vehicles or personnel that visit more than one type of pig unit, potentially providing additional links between commercial and non-commercial units. Spread of infectious disease, such as classical swine fever or porcine reproductive and respiratory syndrome virus, via the movement of contaminated semen have been demonstrated in Europe [25, 26]. However, there is little information available on interactions between producers through semen, despite the potential importance of semen in the spread of disease within and between countries. Whilst commercial pig semen producers maintain high biosecurity to reduce the risk of pathogen transfer, it is not known how common use of semen is between smaller producers, who may use other sources of semen. Swill feeding is also thought to have been responsible for the initial incursion, and in some cases ongoing spread, in outbreaks of swine diseases, including classical and African swine fevers and foot-and-mouth disease . However, swill feeding is not currently permitted in UK. Whilst wild boar play a role in transmission of swine pathogens in other areas, the wild boar population in Scotland is thought to be too low to maintain disease, although wild boar could be responsible for local transmission. Knowledge of all important transmission routes is critical for preparing for future incursions of exotic notifiable diseases and for the creation of accurate and reliable computerised disease outbreak models that may be used to inform policy. Quantification of pig movements conducted in this study provides detailed and valuable information on one transmission route, but also highlights a number of other transmission routes for which there are important knowledge gaps, particularly around the potential role of small scale pig producers.
Historically, characterisation of the interactions between commercial and non-commercial swine producers has been difficult and is rare in the scientific literature. Many previous studies were reliant on questionnaire surveys which invariably suffer from reporting bias and/or non-representative sampling of the study population [4, 5, 27]. Here, we used a national movement database where information on movements is electronically reported by farmers themselves. Under Scottish  and European legislation , all pig keepers moving animals are required to provide information on pig movements within Scotland, and between Scotland and England, as well as details on transportation type and location of all deliveries. As such, ScotEID provided detailed information on pig movements in Scotland at sufficient resolution to allow analyses. A potential limitation of using self-reported data is that some movements may not be reported (reporting bias) [28–31] or erroneously labelled (misinformation bias) [31, 32]. Misinformation bias is likely to be minimal in ScotEID because of cross-referencing procedures which are in place to assure accurate reporting; farmers are indeed required to (1) be registered to the national pig keeper database and enter existing postcodes and premises identifier prior to reporting movements, and (2) report any movement that they receive. Although reporting bias may also be (at least partly) minimised by these cross-referencing procedures, underreporting still remains possible and therefore result in an underestimate of the number of movements that occur between premises, particularly between those that do not keep pigs for commercial purposes. It was also difficult to define the production type for all farms trading pigs in ScotEID. Although pig producers that belong to a health quality scheme could be differentiated as a function of their production system (i.e. genetic supplier, farrow-to-finishers, feeders, weaners or breeders), this was not possible with certainty for the remaining producers. As a consequence, the structural level of businesses (i.e. herds associated in so-called 'pyramid' breeding organisations) that policy targets when facing emergency could not be defined nor extracted from other readily available databases for Scotland. Instead, after discussion with the Scottish swine industry, producers were regrouped into three categories according to their pig population size, movement activity and health quality assurance scheme membership. These categories were considered the most representative of different production types, and were also likely to reflect characteristics such as the level of biosecurity carried out on farm. Indeed, the degree of biosecurity on livestock farm has been shown to be related to (1) management practices and size [5, 6, 33–35], and (2) quality health assurance scheme membership . This information is particularly relevant to disease transmission and response planning exercises . However, providing more details on premises production types and the purposes of movements (for example by including additional question in the movement databases register) would drastically increase the potential of such analyses to inform policy.
In this study, although the majority of movements were to a slaughterhouse, we found that a high connectivity between commercial and non-commercial swine producers still remains when considering non-slaughter movements, particularly between small and non-assured producers (Figure 1). This finding is in clear contrast to what has been previously considered and reported in Europe ; although based on a combination of (1) expert opinion and (2) published literature from countries where non-commercial pig farming is limited. Alternatively, a questionnaire survey targeting all known pig holdings in New Zealand  showed an outcome similar to what was found in Scotland. Nevertheless, not all movements have the same impact on the risk of pre-detection disease spread. Professional producers are responsible for the majority of the animals moved (71.9%), frequently transporting a large number of animals either for slaughter or to other premises and travelling larger distances to get their pigs marketed (Table 1, Figure 5). In contrast, Scottish small producers are characterised by a low number of movements, with a low number of pigs per batch and shorter distances travelled (Table 1, Figure 5). Thus if not detected, an incursion occurring among small producers would be limited and less likely to involve wide geographic spread but a disease incursion spreading via professional producers could feasibly generate geographically widespread epidemics through exposure of a large number of farms and animals. Given that three times more non-commercial producers move pigs than commercial ones (Table 1, Figures 1 and 6), this is consistent with results from modelling exercises showing that epidemic take-offs are unlikely to occur in the Scottish swine industry, should a highly infectious disease be introduced . However, in that study, the model parsimoniously assumed that all primary outbreak farms would be reported, on average, in four weeks, regardless of the heterogeneity in the farm population size and biosecurity measures occurring on farm. While these figures on detection period are consistent to what was observed during epidemics of classical swine fever in the UK and the Netherlands [37, 38], the latter only involved commercial farms where veterinary inspection are routinely carried out. However, there are likely to be characteristics of small producers, such as less regular visits by veterinarians and lower standards of biosecurity , which could negatively impact on the efficiency of surveillance systems, thereby allowing incursions in small producers to remain undetected for significant periods of time. In this case, the risk of pre-detection spread of disease could increase and magnify the potential of the Scottish swine industry to initiate epidemics of swine diseases.
It was interesting to note that only small producers showed seasonal variations in their movements, with more outward movement to other producers in spring/summer followed by a peak of movement to slaughter in autumn (Figure 2D and E). This pattern is likely to reflect smallholders who buy weaners in spring to finish over the summer months. Seasonal variations in small scale production have been previously observed in the Caucasus where a peak of movement to slaughter has been reported for the end of the year celebrations . Quantification of this seasonal pattern is useful as movement peaks are likely to influence disease transmission and therefore be of importance when managing disease incursions or considering disease control options, particularly the impact of movement restrictions. Whilst movements to slaughter present less of a risk of onward transmission, increased movement of young pigs between producers would increase the risk of spread of any pathogens present.
According to our results, both small and professional producers demonstrated tendencies to trade animals with farms of similar features (Table 3). This finding suggests that the likelihood of spillover of disease from small producers into the commercial pig sector is low. This finding was surprising. Although we anticipated that professional producers (who adhere to quality assurance scheme guidelines on risks associated with animal trading) would trade with other professionals, we expected small producers to buy pigs indiscriminately from commercial and non-commercial producers, because they do not produce sufficient pigs to enable the sustainability of their production system. We therefore anticipated that small producers would show a high degree of interaction with other producer types. Although such a hypothesis may partly explain the observed difference between the selling (E-I Index of −0.79) and buying (E-I Index of −0.26) behaviour in this sector (Table 3), this clearly did not outweigh the importance of within-type trade. In contrast, non-assured commercial producers showed a high degree of trade (selling and buying) with other producer types (Table 3). Together with similar level of production outputs but possibly reduced biosecurity standards than professional producers, it is therefore plausible that non-assured producers may represent a bridge between the commercial and non-commercial sector of the industry, potentially allowing undiagnosed pathogen to spread throughout the industry. To test this hypothesis, network analysis can be applied to further explore the connectivity between producers via the movement of pigs [10, 11, 13, 14], identify characteristics that may increase the influence of some producers on the spread of pathogen [10, 11] and whether this influence is permanent or restricted in time . Such an analysis will be the subject of another paper.
The role of livestock haulage in the spread of pathogens within and between sectors of the industry is likely to be important since animal transporters may act as fomites and spread pathogens onto farm premises and their surroundings [19, 41–43]. In the Scottish swine industry, haulage is mostly used by commercial professionals and, to a lesser extent, by non-assured commercial producers, whereas small producers rarely use hauliers for transporting their pigs (Figure 5). We also showed that haulage has the potential to increase the interconnectivity between farms within the commercial sector of the industry, without showing any evidence of direct exchanges of animals. Indeed, during the study period, a single haulage company may have connected 132 farms, of which, only a fraction were directly trading animals. These results are relevant when preparing for future incursions of exotic notifiable diseases as this degree of connectivity could increase the rate of pathogen dissemination across both the Scottish and GB swine industries. Although regulations are in place for assuring that livestock haulage vehicles are regularly cleansed and disinfected in an attempt to limit the risk of widespread dissemination of pathogens [44, 45], effective cleansing and disinfection can be difficult to achieve, both logistically [46–48] and in terms of compliance . Together with improving disease awareness and biosecurity knowledge among producers and haulage companies, ensuring that procedures are optimised to reduce the risk of transmission via contaminated haulage vehicles is of high importance, particularly within the commercial sector.
During the study period, half of the haulage companies that moved pigs between producers operated within a single region. This result is in agreement with the local clustering of haulage use observed in a survey of quality assured producers in GB  and may create some geographical limitations to disease spread. In contrast, nearly 20% of movements to slaughter were sent across the border to England, especially in regions close to the border and in the East of England (Figure 6B, Tables 1 and 2). Although this pattern of pig movements to England highlights the reduced capacity of the Scottish slaughterhouses to process pigs, particular attention should be paid to its implication in the spread of diseases between England and Scotland. Increased cross-border movements of live animals and vehicles between England and Scotland is likely to significantly increase the risk of disease incursion from one country to another, either directly or due to returning vehicles [19, 50]. From the point of view of livestock disease regulation agency, surveillance should therefore be modulated as a function of the frequency of incoming movements and the density of movements within Scotland. As such, targeting producers in “South Scotland” and “Scotland Central Belt”, as the likely entry point from cross border movements would likely be the most cost-efficient surveillance strategy for cross-border incursions. In contrast, farms in “Mid Scotland and Fife” may be tagged with a lower priority as it seems relatively isolated from farms in the other regions in terms of trade (Figure 6).