Data-driven epidemic intelligence: applications for surveillance and control of foot-and-mouth disease in Uganda

Foot-and-Mouth disease (FMD), a highly contagious viral infection affecting livestock and wildlife species, has considerable impacts on animal health, welfare, and productivity. In endemic settings, numerous eco-epidemiological and social factors contribute to a complex local epidemiology where viral genetic diversity and multi-host interactions converge to facilitate viral introduction and spread. In Uganda, FMD poses a substantial economic burden on farmers due to its clinical effects and the indirect impact of livestock movement restrictions and market closures, implemented to curb the outbreaks. Currently, the country is engaged in the Progressive Control Pathway for FMD (PCPFMD), a FAO-led approach aimed at understanding the local epidemiology and identifying control options that align with the local context. Since 2016, Uganda has been on stage 2, focused on implementing risk-based control measures to reduce the impact of the disease. In this context, data mining offers a unique opportunity for proactively tackling FMD and other epidemic-prone diseases in Uganda. By integrating official data and publicly available information, it is possible to amplify their value to identify priority areas for strengthened surveillance and control, leading to more efficient resource allocation. This thesis uses several analytical approaches to study some of the most important epidemiological drivers of FMD to gain strategic understanding of disease risk across the country. Moreover, it documents the experience of designing and fieldtrialling a data-driven strategy for early detection and active monitoring of the virus. Together, these results illustrate the real-life application of disease intelligence methods for evidence-based action. First, an evidence synthesis approach was used to identify the epidemiological factors linked to the risk of FMD outbreaks in endemic settings. For this, spatial and spatiotemporal studies were comprehensively retrieved, classified, and critically appraised. The results highlight the value of using spatial epidemiology tools on outbreak data to enhance the understanding of FMD and map its risk in data-scarce, endemic environments. While the epidemiological factors used for modelling FMD risk varied across studies, it was possible to broadly classify them into 5 themes representing risk pathways relevant for the introduction, spread or survival of FMDv: (a) animal demographics and interactions, (b) spatial accessibility, (c) trade, (d) socioeconomics and (e) environment. A thorough analysis of the risk of FMD outbreaks requires a systemic view as multiple epidemiological factors can contribute to better understand viral circulation. This may improve the accuracy of disease mapping and can guide the design of an all-encompassing strategy for disease control. Next, because animal movements are a major driver for the spread of Transboundary Animal Diseases (TADs) including FMD, Social Network Analysis (SNA) was used to study the structure and architecture of the networks underlying cattle movements in Uganda. This information is essential to detect districts or sub-regions that could benefit from strengthened disease surveillance and control, thereby tackling the spread of FMD across the country. Weighted, directed networks were constructed based on 2019 between-district cattle movements sourced from official records. The purpose of the movement (‘slaughter’ vs. ‘live trade’) was used to subset the network and capture the risks more reliably. A total of 22,024 movements were recorded across 131 districts; these movements involved 234,229 animals, mostly for slaughter (~87%). This analysis suggested that cattle trade can result in local and long-distance disease spread with seasonal variability identified as an important factor influencing cattle mobility in Uganda. These results also show that different centrality measures can be used to target animal health interventions to districts, providing detailed insights for epidemic intelligence. Later, a total 298 FMD outbreaks that were officially reported in Uganda between 2014 and 2019 were analysed using different Bayesian mixed-effects spatial models with the Integrated Nested Laplace Approximation (INLA) method. The results show that cattle density, Enhanced Vegetation Index (EVI), deprivation score and human density were associated with the risk of outbreaks but did not explain most of the spatial distribution of the risk. Adjusted district-specific risk estimates and exceedance probabilities highlighted districts located in the cattle corridor as areas of greater risk. However, while high-risk areas should be prioritized for timely disease management, ‘low risk’ areas should not be marginalised and increased surveillance efforts should be made towards a progressive, long-term disease control. In Uganda, a better understanding of these district-specific factors and establishing a timely, sensitive surveillance system are crucial for prioritizing and supporting FMD control and prevention efforts. Lastly, the results of a pilot study testing a centrality-based monitoring system for FMDv in Uganda are presented. This system was field-trialled in a group of districts located in south-eastern Uganda, characterized by strong commercial ties for livestock trading. A total 17 livestock markets within 11 districts were surveyed. Samples were collected from cattle, the market environment and from vehicles. Evidence of FMDv was detected in five out of 11 surveyed districts (45.5%) and six out of 17 markets (35.3%). Positive markets were typically medium or large-sized, well-connected, enclosed livestock markets that traded higher numbers of cattle per day of operation. These results suggest that data-driven analysis can guide the design of field-based applications and that these could potentially be implemented to strengthen active, risk-based monitoring systems for FMDv in endemic settings like Uganda. The analyses included within this thesis were designed and performed with a strong interest on conducting a comprehensive assessment of FMD risk. Relevant research gaps were addressed as part of this thesis, emphasizing its potential to make a significant contribution to risk-based strategic planning which aligns with Uganda’s endeavours to tackle FMD spread and minimize its impact. In practice, several large-scale disease surveillance and control interventions could be implemented across the most influential districts and regions, as informed by the network analysis and the Bayesian hierarchical model. An approach balancing the use of high-cost measures such as preemptive vaccination or laboratory-based surveillance could be used alongside resource-friendly methods, including community sensitization, for a multifaceted, socially-conscious, and costix effective disease control strategy. However, it is also important to find a middle ground to reconcile the wider goals of disease elimination with the local needs and everyday reality faced by the livestock communities burdened by FMD.