Efficient and highly effective control of infectious diseases can be achieved by targeting interventions towards farms that are highly connected “super-spreaders” in animal movement networks. However, from an implementation standpoint, it is unclear how much movement data is required to gain an accurate picture of farm connectivity, nor how quickly movement networks change over time. For example, can movement data from last year be used to identify potential super-spreaders this year? How often do such analyses need to be updated? Answering these questions is key to moving from science to practice in terms of successful deployment of network-based targeted control strategies in swine production systems. In this study, Dr Dennis Makau and the VanderWaal lab aim to answer these questions for production systems in the United States.
Using network theory, we analyzed 282,807 pig movements between 2,724 farms from two production systems in the United States between 2014 and 2017. We examined farm-to-farm loyalty in pig movements (i.e., how often is a movement between two specific farms repeated), consistency of farm’s position in the network through time (e.g., number of trading partners), and overall changes in network structure through time. We also simulated the efficacy of targeted interventions at fragmenting between-farm infection chains when using data from different historical timepoints to define potential super-spreaders.
Key findings and their relevance
How many months of data are needed to adequately represent a network?
- Network stability was achieved after aggregating six months of data, and minimal shifts in node-level and global network metrics were observed when adding more data. This indicates that a temporal resolution of six months for pig movement data would be reliable for identifying potential superspreaders and high-risk farms with minimal random variation expected in the network.
How often should network analysis be updated when using network-based targeted control strategies?
- The effectiveness of targeted interventions on high-risk swine farms based on network data from 1-2 years ago was as effective as using current data to identify high-risk farms (see figure below).
- Using previous data is more effective than random interventions on farms (see figure below), although its effectiveness/reliability decreases considerably with data more than 30 months old.
- When simulating the impact of targeted interventions, vaccinating 10-15% of highly connected farms effectively disrupted disease transmission (assuming 100% vaccine effectiveness and disease transmission primarily through animal movements). In contrast, in random vaccination, more than 60% of the farms would have to be vaccinated to achieve a similar impact.
How often are movements between specific farms repeated?
- Repeated movements between specific farms (i.e., loyalty) contributes to the scale of temporal stability in swine movement networks.
- Loyalty in farm-to-farm pig movements was mainly exhibited for movements from sow farms to nurseries and from nurseries to finisher farms. The average proportion of specific farm contacts repeated at least once within a 52-week interval was 51–60 % and 12–22% for weaned feeder pig movements, respectively. These loyalty trends were cyclic; episodes of increased loyalty observed at intervals of 8 and 17–20 weeks for weaned feeder pig movements, respectively. These cycles corresponded to the pig lifecycles and expected time spent at the two stages of production in a multisite production system.
Understanding network stability aids in decisions of disease management, monitoring and surveillance. Moreover, the observed network stability and reliability of historical data indicate that historical data is better than no data.