Overview
Much of farm animal surveillance currently relies on passive reporting of notifiable diseases by owners and vets alike, passive surveillance of samples submitted to laboratories, and active surveillance of diseases like bovine tuberculosis. This leaves a gap in our understanding of population-level disease, namely what is being seen in primary veterinary care and on farm.
About this opportunity
iCASE industrial partner web link: https://iechydda.cymru/
Most veterinary surgeons now manage their clinical records digitally; these electronic health records (EHRs) represent a research and surveillance opportunity. In animal health, use of EHRs is best developed in companion animals where digitisation of individual animal health records is most complete.
As part of a wider programme of work focussing on understanding and mitigating AMR in Wales (Arwain DGC), we have been piloting the ethics and usability of collecting EHR data from a sentinel network of farm animal practices in Wales (FAVSNET).
In this PhD, you will leverage these records that are collected in real-time and added to daily. Your specific objectives will be to:
- Use supervised and unsupervised language models to develop methods to extract useful clinical syndromic and diagnostic information from unstructured clinical narratives.
- Test the ability of language model approaches to obtain treatment data at the population level including metrics of dose, frequency and number of animals treated.
- To develop clinically useful and meaningful metrics of accuracy for disease and treatment.
- To work with stakeholders including farmers and practitioners to assess metrics of AI maturity / acceptability in support of real-world practice.
- To combine outputs from 1 and 2 and carry out an interventional trial to assess our ability to improve antibacterial use on high using farms in FAVSNET at the syndrome level.
Placements with lechyd Da will be arranged at several times through the project timed to develop a wider understanding of the context of the project and to facilitate best knowledge transfer from research into practice.
You will have a demonstrable interest in animal health and data science. Whilst a veterinary or bioveterinary-related degree may be desirable it is not essential. But an ability to communicate and work with vets and farmers will be essentials. For those without existing computing-related qualifications, it will be essential to demonstrate a strong desire to learn them.
Further reading
1. Noble PM, Appleton C, Radford AD, Nenadic G. Using topic modelling for unsupervised annotation of electronic health records to identify an outbreak of disease in UK dogs. PLoS One. 2021 Dec 9;16(12):e0260402. doi: 10.1371/journal.pone.0260402.
2. Rodríguez J, Killick DR, Ressel L, Espinosa de Los Monteros A, Santana A, Beck S, Cian F, McKay JS, Noble PJ, Pinchbeck GL, Singleton DA, Radford AD. A text-mining based analysis of 100,000 tumours affecting dogs and cats in the United Kingdom. Sci Data. 2021 Oct 15;8(1):266. doi: 10.1038/s41597-021-01039-x.
3. Radford AD, Singleton DA, Jewell C, Appleton C, Rowlingson B, Hale AC, Cuartero CT, Newton R, Sánchez-Vizcaíno F, Greenberg D, Brant B, Bentley EG, Stewart JP, Smith S, Haldenby S, Noble PM, Pinchbeck GL. Outbreak of Severe Vomiting in Dogs Associated with a Canine Enteric Coronavirus, United Kingdom. Emerg Infect Dis. 2021 Feb;27(2):517-528. doi: 10.3201/eid2702.202452
4. Han L , Gladkoff S, Erofeev G, Sorokina I, Galiano B, Nenadic G. Neural machine translation of clinical text: an empirical investigation into multilingual pre-trained language models and transfer-learning. Frontiers in Digital Health 6 (2024). DOI=10.3389/fdgth.2024.1211564.
5. Bedford C, Galotta ML, Oikonomou G, de Yaniz G, Nardello M, Sánchez Bruni S, Davies P. A mixed method approach to analysing patterns and drivers of antibiotic use and resistance in beef farms in Argentina. Front Vet Sci. 2024 Nov 13;11:1454032. doi: 10.3389/fvets.2024.1454032.