Extrapolating incomplete animal population and surveillance data for use in national disease control : examples from Myanmar and New Zealand : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Veterinary Epidemiology, School of Veterinary Science at Massey University, Manawatu, New Zealand

Thumbnail Image
Open Access Location
Journal Title
Journal ISSN
Volume Title
Massey University
The Author
National level databases of animal numbers, locations, and movements provide the essential foundations for disease outbreak investigations, disease control, and disease preparedness activities. These activities are particularly important for managing and mitigating the risks of high impact exotic disease outbreaks like foot-and-mouth disease (FMD) as well as other economically important endemic diseases, which can significantly impact international trade and food security. However, many countries worldwide either lack national animal databases entirely or have multiple, fragmented databases that provide an incomplete picture of animal demographics. Consequently, there has been growing interest in developing novel methods to infer missing information on animal populations from other data sources, to quantify the extent of missing information, and to understand the impacts of missing information on the predictions made from national disease simulation models. This thesis explores these issues in the context of an FMD free country (New Zealand) as well as a country with endemic FMD (Myanmar). In Chapter 3, regression models were used to predict farm-level animal populations in New Zealand based on available data on farm type and location. When the results were compared against a subset of validated animal population data, the predictions at the farm level were found to be inaccurate especially for small-scale farms that keep animals for personal consumption or as a hobby. These properties are of particular interest to animal health authorities as they have been identified as at risk for exotic disease outbreaks. In Chapter 4, the impacts of having inaccurate herd size data on the predictions made by an FMD disease spread simulation model were explored. The results were analysed using cox proportional hazard models and logistic regression models, which showed that simulations run using actual animal population data indicated different optimal control strategies for FMD than models run with imperfect data and these effects differed by the region in New Zealand where the hypothetical disease outbreak was seeded. In Chapter 5, high-resolution local survey data and low-resolution national remote sensor data were used alone and in combination to predict the location of FMD positive villages in Myanmar, which were identified by serological sampling conducted as part of a large OIE funded research project in 2016. The performance of both random forest models and logistic regression models were explored using training and testing data sets. Bovine populations and proximity to cattle markets were found to be significant risk factors for FMD seropositivity and the logistic regression models performed as well as machine learning techniques. Chapter 6 compared verbal reports of FMD outbreaks from village headman and householders against the serological test results from their villages to determine whether using public reports is a viable alternative to conducting resource intensive serological surveys for estimating FMD prevalence in Myanmar. Although village headmen proved to be a better source of FMD reports compared to householders, the verbal reports were still not as accurate as serological tests in an endemic situation where both sensitivity and specificity of observing clinical signs can be complicated by endemic stability and concurrent outbreaks of other diseases. The work in both chapters 5 and 6 was carried out using data from activities of the Livestock Breeding and Veterinary Department and the OIE and as such separate human ethics approval was not required for the surveys described. Chapter 7 addressed the issue of estimating the scale of missing data in a national database by comparing intensively collected interview information with recorded movements at the farm level for farms involved in New Zealand’s Mycoplasma bovis eradication programme. The results showed that dairy farmers often failed to record almost half of high risk movements including leased bulls, calves sent offsite for rearing, and adult cattle sent away for winter grazing. It was also estimated that approximately 60% of animals arriving at abattoirs in New Zealand have multiple movements missing from their life history in the National Animal Identification and Tracing system (NAIT) database. This missing information had a significant impact on the ability of government and industry to effectively respond to the outbreak. However, a positive finding from this study was that the rates of missing data are decreasing over time. Overall, this thesis demonstrated the importance of enhancing efforts to collect accurate and up-to- date national animal population and movement data. For New Zealand, the changes required to improve the national farm animal data landscape include improving compliance with the legislated requirements to record animal movements and modifying the existing databases to record information on the health status of animals against a unique animal identifier. A unique farm identifier is required at the national level and should be agreed upon by industry representatives, government and researchers. The combination of animal health data associated with the unique animal identifier and a single current farm identifier for all farms will result in a robust animal health and biosecurity system.
Livestock, Diseases, Burma, New Zealand, Prevention, Mathematical models, Monitoring