‘What’s not to like about a birthday party?’: celebrating 10 years of SAVSNET

‘What’s not to like about a birthday party?’ asked Professor Susan Dawson in her opening speech. Professor Susan Dawson is Head of Liverpool Veterinary School at the Leahurst Campus, where SAVSNET (Small Animal Veterinary Surveillance Network) is based, and members of HeRC had joined the SAVSNET team for the project’s 10^th anniversary celebration.

One of HeRC’s long-time partners, the SAVSNET programme is a shining exemplar of delivering health informatics research using a cross-disciplinary and cross-specialism working model. Focusing on data from participating small animal veterinary practices and laboratories, and with over 4 million clinical records covered, the research programme aims to:

• monitor disease trends and highlight appropriate interventions
• identify populations at risk, monitoring treatments and outcomes
• provide data resources for academics and others
• improve general public awareness of small animal diseases and prevention
• provide a route to clinical benchmarking for vets in small animal practice

Whilst based at the University of Liverpool’s Institute of Infection and Global Health, the SAVSNET team collaborates closely and very successfully with academic, industry, charity partners from both human and animal health domains many of whom were present at the celebrations. These include The Dogs’ Trust, DEFRA, CVS Group plc, MerckSharpeDohme, Bristol University and of course HeRC collaborators from the University of Manchester, providing insight from routinely collected data to improve clinical practice and outcomes.

The collection of veterinary practice data has various uses. It can be used by researchers to monitor disease trends, identify populations at risk and monitor treatments and outcomes. The results of this analysis are then fed back to practices, allowing vets to undertake benchmarking, undertake continuous improvement and react to the risk of potential outbreaks as they occur.

After 10 years, SAVSNET now collects data from 10% of veterinary practices in the UK, gaining a national coverage of data from both corporate and individual practices. Numerous papers have been published, including analysis on antibacterial prescribing patterns, risk factors for blowfly strike and the times of year when dogs are most likely to eat chocolate. (It’s easy to guess the correct answer, which is Christmas – but did you know that when it comes to Valentine’s Day, British dogs are the least likely in Europe to get their paws on their owners’ chocolates?)

So, to celebrate a decade of achievements, the SAVSNET team hosted a two-day event. The first day took the form of a series of talks from SAVSNET researchers, covering topics from the use of big data in preventative health to the changing popularity of pet breeds and the different veterinary challenges this brings.

HeRC researcher Professor Goran Nenadic (The University of Manchester) delivered a presentation on his team’s work into methods to how free text can be used in research. Free text is a way of entering data without a certain structure that can be read by a computer. So, while structured text may look something like this:

SPECIES: Dog

BREED: Labrador crossbreed

SEX: Female

REASON FOR VISIT: Vaccination

VACCINATION TYPE: Kennel cough
Free text may look more like this:

Female Labrador crossbreed in for kennel cough vaccination.
There are various pros and cons to using free text in clinical data entry. On one hand, free text can be more efficient to type, as well as being better for expressing nuances that a series of drop-down menus cannot capture. On the other hand, it can be difficult to use in research, as it needs to be converted into structured text before computers can read and analyse it. This can present problems for researchers: free text can contain spelling and punctuation mistakes that may prevent entries from being classified into the correct categories. When used in a clinical context, it can also be cryptic for those not in the practice. In reality, free text can end up looking like this, which is more challenging to analyse:

F lab X kennal cogh vacc
Therefore, researchers like Goran use a number of different approaches to ensure that each piece of data in a free text entry is correctly identified and categorised. There are ‘regular expressions’ (or ‘regex’) that can help misspelled words to be included in searches – for example, you could use regex to search for both ‘kennel’ and ‘kennal’, allowing the above entry to be picked up. Goran also uses strategies such as neural language models, which can identify terms that are related rather than synonymous.

Goran’s methods of free text analysis can prove invaluable when looking at veterinary data entries, which can contain acronyms, regionalisms and the names of complicated procedures which would be difficult to spell even without being speed-typed in the space between appointments. (‘Enterotoxaemia,’ anyone?)

‘It’s been a great experience working alongside the SAVSNET team,’ said Goran on his involvement with the project. ‘The team has been extremely enthusiastic in unlocking evidence that is buried in unstructured clinical data, and I’m looking forward to continuing to collaborate with them on their project.’

On the second day, delegates rolled up their sleeves and put all the expertise shared on day one into practice with a hands-on big data research workshop. After a tutorial on using text mining techniques such as regex, delegates created their own research projects and used these techniques to search the SAVSNET database.

The projects were highly varied, ranging from the obesity rate in cats to the population of animals wounded by air guns. HeRC delegate Ruth Norris became Principal Investigator of the ‘Krazy Kat Ladies study’, while Nina Hayes-Thompson researched which animals have the most operations to remove foreign bodies they’ve swallowed. (Warning to fellow non-vets: do not do this directly after having lunch.)

By the end of the workshop delegates had successfully used the SAVSNET database to produce findings for their research projects. For example, SAVSNET data suggests that amongst both cats and dogs, it is male animals that require the most operations to remove things they’ve swallowed. Less surprisingly to any Labrador owners, the SAVSNET database suggested that they are the dog breed most likely to swallow something they shouldn’t.

The event was a huge success, and the SAVSNET team are looking forward to their future work. Professor Alan Radford, who is SAVSNET’s Principal Investigator and Professor of Veterinary Health Informatics at the University of Liverpool, said ‘SAVSNET has been a fantastic project to work on for the last ten years with great partnerships between veterinary and human health scientists. It will be really interesting to see what we will achieve together in the future.’

For more information about SAVSNET, click here.