Please note that this program is subject to minor variation.
12:00 - 12:55
12:55 - 1:05 Break
1:05 - 1:55
1:55 - 2:05 Break
2:05 – 3:00
12:00 - 12:55
12:55 - 1:05 Break
1:05 - 1:55
1:55 - 2:05 Break
2:05 – 3:00
The program for the VCA Symposium 2022 will include the following presentations.
Background: Canine and feline patients receiving chemotherapy are frequently treated with antimicrobials. Some expert sources recommend routine prophylactic use of antimicrobials to reduce the risk of febrile neutropenia in all patients or patients receiving specific chemotherapy drugs. Patients that develop neutropenia or fever are frequently treated with antimicrobials as a precaution, due in part to the complexity in differentiating the clinical signs of infection from effects of chemotherapy. The choice of antimicrobial, rationale and frequency of use, and the duration of treatment, have not been characterised or critically appraised. The absence of therapeutic guidelines for patients undergoing chemotherapy leaves clinicians working from first principles rather than by evidence-based population outcomes.
In order to make recommendations for the use of antimicrobials in companion animals being treated with cytotoxic chemotherapy, an understanding of the current use in practices is needed rather than relying on the limited published recommendations. Our hypothesis is that antimicrobial use for chemotherapy patients in practice is used more frequently and for longer periods with inappropriate choices of drugs.
In this presentation the initial results of a partial search of the data will be presented.
Methods: A search of the VetCompass database was performed in 2019 for all chemotherapy drugs used in veterinary practice. For these patients all entries were retrieved and searched to confirm the use of chemotherapy, and then for antimicrobial use, if any. The antimicrobials used were identified, the duration of the course prescribed and the reason for their use if identifiable.
Results: The search identified 209,719 entries for 860 individual patients. To date, 86,796 entries for 360 individuals has been screened. This has identified 57 entries for 184 individuals who received chemotherapy and their data will be presented. Of these individuals, 40 (18%) did not have any antimicrobials dispensed during chemotherapy treatment. The course of the antimicrobials was recorded as ranging from 1 day to 30 days.
Neutropenia was the indicated reason for antimicrobial use in 28 individuals, grade 1 in 11, grade 2 in 5, grade 3 in 4 and grade 4 in 8. Gastroenteric problems were indicated in 37 individuals including anorexia, diarrhoea, haematochezia, mucositis and vomiting. In 28 individuals the problems were indicated as grade 1, with 8 grade 2 and 1 grade 3 toxicity. The reason for antimicrobial use was unclear in 63 individuals and for prophylaxis in 7.
Summary: The preliminary work has shown that antimicrobial use is common in companion animals receiving chemotherapy. There was appropriate use in some cases either related to chemotherapy or tumour complications, and sometimes for concurrent infections. A reasonable number of times the use did not match the current recommendations.
Large databases of electronic healthcare records (EHRs) are now a well-established part of the veterinary research landscape. They are uniquely positioned to provide broad, generalisable insights into conditions in practice; for some sense of scale, VetCompass (UK) currently contains records from over 30% UK practices and over 26 million animals.
Veterinary databases such as VetCompass are predominantly comprised of unstructured clinical notes. Information is captured in blocks of free text rather than as clinical codes or individual measurements. To use these big datasets, we therefore require human interpretation. Annotators with domain knowledge are well placed for this task as they can extract data either directly or inferentially from contextual information. Although this approach is the standard, the scale to which it can be applied is limited by time and the number of suitably qualified individuals.
Natural language processing (NLP) is a branch of computer science that may offer a solution to this problem. Models of language developed using machine learning can be applied to clinical data to complete a number of tasks, relevant examples of which being information extraction and predictive modelling for classification or prognostication. Models provide a scalable solution, meaning that they can generate more results than their human counterparts, however they must be provided with enough data to “learn” the correct response to a specific task. Therefore, training a model to be applied to veterinary data often has similar requirements to projects using human annotation alone.
Until recently, this has been one of the limiting factors in applying NLP in our domain. However, billion-parameter models, so called large language models (LLMs) have overcome this, demonstrating remarkable abilities simply due to their size. Large language models can produce results having been shown very few examples, in some cases, completing tasks without having seen the correct response altogether.
In this presentation, we will introduce LLMs and using real examples, demonstrate how this type of model could be applied to veterinary clinical notes.
Tick paralysis resulting from bites from Ixodes holocyclus and I. cornuatus is a leading cause of emergency veterinary admissions for Australian companion animals, often resulting in death if left untreated. Availability of timely information on periods of increased risk can help modulate behaviours that reduce exposures to ticks and improve awareness of owners for the need of preventative ectoparasite treatment. However, detection of trends in risk is hampered by the lack of clearly annotated historical records of tick paralysis. Natural Language Processing (NLP) of clinical records is required to first ascertain historical cases. Here we describe a platform to perform NLP on VetCompass Australia’s veterinary clinical records to accurately identify historical cases of canine tick paralysis where we make use of combine bespoke spellchecking and tokenization routines with a domain-expertise inspired clinical dictionary to identify important terms in free text indicative of a tick paralysis diagnosis. Resulting time series of tick paralysis cases are then analysed using Dynamic Generalised Additive Models to jointly estimate nonlinear distributed lag effects of environmental predictors and dynamic latent temporal processes that facilitate probabilistic near-term forecasts of tick paralysis risk. Our models forecast tick paralysis cases with exceptional accuracy while preserving epidemiological interpretability. We have designed an interactive online dashboard to showcase our data and modelling results so that we can refine the way we present probabilistic predictions to meet end-user requirements. We expect our data acquisition / modelling pipeline to act as a platform for developing early warning systems that can notify clinicians and pet owners about heightened risks of environmentally driven veterinary conditions.
Bullmastiffs are a large breed of dog which were bred to perform as guard-dogs in 19th century England. The breed is the result of a breed mixture; 60% Mastiff to 40% Bulldog, with these two breeds contributing gigantic stature (Australian National Kennel Club weight range of 41-50kg for adult females; 50-59kg for adult males) and brachycephalic features. There is evidence the breed is predisposed to cancer and musculoskeletal disorders although disease investigation of the breed is limited. Owner surveys in the UK recently reported a reduced lifespan of 8.01 years of age for the Bullmastiff. This study reports on the demography, longevity and mortality of Bullmastiffs attending veterinary practices in Australia from 2008-2017. VetCompass Australia collects patient data from veterinary practices across Australia for epidemiological analysis. All patient records of Bullmastiffs available in the VetCompass Australia database during this decade period were reviewed, with demographic information on the breed inclusive of coat colour, sex, neuter status, weight and location collated. VeNom diagnostic codes for the most appropriate cause of death were assigned to deceased dogs. The population comprised of 2,771 Bullmastiffs; 1,259 female dogs (45.4%), 1,491 male dogs (53.8%) and 21 dogs (0.8%) that were of unknown sex or neuter status with an overall median age of 2.8 years. Dogs grew rapidly in their first year, continued to grow in their second year and growth plateaued as adulthood was reached, with a mean body weight of adult dogs (46.6 kg) heavier than bitches (40.5 kg). The median longevity of the Bullmastiff was 8.5 years. The most common causes of death in the breed were mass lesions (28.2%), old age (9.9%), musculoskeletal (9.9%), neurological (5.3%) and behavioural disorders (4.8%). Neutering was protective against mortality from urogenital causes (OR:0.14; CI: 0.02-0.52; P 0.003) of mortality in the study. This study provides demographic and health information on a representative population of Bullmastiffs, which is beneficial due to the dependency of good evidence based veterinary decisions on data analysis. Furthermore, the results will assist owners and breeders to make informed decisions on health risks and selective breeding to reduce heritable disease.
Canine splenic fibrohistiocytic nodules traditionally encompassed benign lymphoid hyperplasia, complex hyperplasia, and malignant fibrous histiocytoma. The latter have been recently re-classified into histiocytic sarcoma, and stromal sarcoma. Reliable indicators of post-splenectomy survival, and demographic factors predisposing to the four types of nodules are not completely understood. This study is divided in two parts, the first part aims to estimate the frequency, survival times, and identify risk factors of splenectomized dogs diagnosed with lymphoid hyperplasia, complex hyperplasia, histiocytic sarcoma, and stromal sarcoma using medical records containing histopathological diagnosis from the VetCompass Australia database (1989-2018), which collects demographic, and clinical information from veterinary clinics. The epidemiological study found that out of 693 dogs, 315 were diagnosed with fibrohistiocytic nodules, mostly lymphoid hyperplasia (169/693, 24.4%), followed by stromal sarcoma (59/693, 8.5%), complex hyperplasia (55/693, 7.9%), and histiocytic sarcoma (32/693, 4.6%). Dogs aged 8-10 years were more likely to be diagnosed with histiocytic or stromal sarcoma, compared to lymphoid hyperplasia. Dogs diagnosed with lymphoid hyperplasia had a longer survival time than those with other diagnoses (median >2 years). Dogs diagnosed with histiocytic sarcoma had longer survival times (median 349 days) than stromal sarcoma (median 166 days). Results suggest that knowledge of the type of splenic fibrohistiocytic nodule, patients’ age and sex can be used to increase prognostic accuracy.
Ear folding in Scottish Fold cats is caused by a variant in the gene TRPV4 and while current evidence suggests that all cats carrying this genetic variant are affected by osteochondrodysplasia (SFOCD), the age of onset and severity of SFOCD varies significantly between individuals. Although well-recognised as a disorder, nothing is known about the prevalence of clinical SFOCD in the general Scottish Fold cat population. This retrospective cohort study aimed to estimate the prevalence of clinically diagnosed and clinically suspected cases of SFOCD in Scottish Fold cats in Australian veterinary clinics.
Electronic patient records (EPRs), collected between 1992 and 2018, were provided by VetCompass Australia. EPRs were analysed using the data analysis software R to identify SFOCD and SFOCD associated clinical key words. EPRs were provided by veterinary clinics located in New South Wales, Queensland, and Victoria, and consisted of 34,926 consultation events, representing 1131 Scottish Fold and 117 Scottish Shorthair/Straight cats. As the causative TRPV4 variant responsible for SFOCD is absent from Scottish Shorthair/Straight cats, these cats were included as a control population. Primary analysis identified 14/1131 Scottish Fold cats clinically diagnosed with SFOCD, representing 1.2% of the studied Scottish Fold population. None of the Scottish Shorthair/Straight cats were diagnosed with clinical SFOCD. When the analysis was extended to also include clinical keywords potentially indicating clinical SFOCD, an additional 194 animals were identified. Clinical records of these cats are now manually assessed to confirm potential case status.
Extreme environmental events such as heatwaves, push animals to their physiological limits and may be fatal for some species. In dogs, extreme heat conditions can lead to heat related illness (HRI), which is a condition that occurs when core body temperature exceeds homeostatic limits. The most severe form of heat related illness is heat stroke and is potentially fatal in dogs. Heatwaves are expected to increase in frequency and intensity due to anthropogenic climate change (Nairn & Fawcett 2013) and these extreme environmental events push animals to their physiological extremes. This study investigated the incidence and risk factors for heat-related illness in dogs in Australia over a twenty-year period.
Method: The study utilized data from the VetCompass Australia database and included dogs which were registered with a veterinary practice between 1997 and 2017. The Electronic Patient Records (EPRs) were then reviewed using inclusion and exclusion criteria and included if the criteria were met. Risk factor analysis was conducted using multivariable logistic regression to identify potential risk factors associated with heat related illness. Binary logistic regression was used to evaluate potential univariable associations between risk factors (breed, Australian Kennel Council breed (ANKC) groupings https://ankc.org.au/Breed/Index/, coat colour, sex (neuter status), brachycephaly, and age) and HRI diagnosis.
Results and Discussion: There were 119 heat related illness cases recorded between 1997-2017. The incidence of risk of HRI in dogs under primary veterinary care (as defined by criteria) per 1000 dog/years= 0.74 (95% confidence interval by epi.conf =0.61-0.84).
The study identified at-risk breeds for HRI these were the Australian Stumpy Tail Cattle Dog, British Bulldog, French Bulldog, Maremma Sheepdog, Italian Greyhound, Chow Chow, Airedale Terrier, Pug, Samoyed, English Springer Spaniel, Labrador Retriever, Golden Retriever, Cavalier King Charles Spaniel, Border Collie, Staffordshire Bull Terrier and pooled non-Australian National Kennel Council breeds (which included the American and Australian Bulldog)). During the study period, 27 dogs either died from the heat related illness or were euthanased due to the heat related illness, its sequelae, or were euthanased with HRI due to an underlying condition (e.g., laryngeal paralysis). As expected, there were more cases of heat related illness during the warmer summer months compared to other times of the year, as well as in hotter years (e.g., 2016). Interestingly, there were no differences in the risk of heat related illness between males and females nor between desexed or un-desexed dogs; but there was in increased risk of heat related in older dogs. In summary, the current study syntheses evidence-based research findings from the current study along with other published literature, to inform and advance our knowledge on the impact of extreme heat conditions on the health and fatality risk in dogs.
There is increasing evidence that undesirable behaviours (UBs) in dogs can compromise the welfare of both canine companions and their carers. Veterinarians are regularly consulted about affected animals and may be asked to euthanase the more severely affected individuals and this part of their practice clearly compromises their caseload. The objective of two recent related studies was to estimate proportional mortality due to UBs and risk factors for death due to UBs, including death from road traffic accidents, in dogs under three years of age attending primary-care veterinary practices in the UK and Australia. Cases were identified by searching de-identified electronic patient records from primary-care veterinary practices participating in the VetCompass UK programme (2009–2014) and VetCompass Australia (2013 to 2018). The findings reveal that dogs under three years of age are at higher risk of death due to UBs (33.7% in the UK and 29.7% in Australia) than any other cause of death. Male dogs had 1.40 the odds of death from UB compared with females in the UK; and 1.09 the odds in Australia. Neutered dogs had 1.94 the odds of death due to a UB compared with entire dogs in the UK; and 2.5 the odds in Australia. Crossbred dogs had 1.35 the odds of a UB related death compared to purebred dogs in the UK; and 1.43 the odds in Australia. The breeds at highest risk in the UK were Staffordshire bull terriers (odds ratio OR 3.99) whereas in Australia they were Australian cattle dogs (OR 4.77) and American Staffordshire terriers (OR 4.69). In both populations, aggression was the most prevalent UB overall. Veterinarians in the UK had recommended referral to a behaviourist or trainer in 10.3% and 11.0% of cases where dogs died due to exhibiting a UB in the UK and Australia, respectively. They had dispensed nutraceutical, pheromone or pharmacological treatment to 3.0% and 5.9% of cases in the UK and Australia, respectively. The results reveal how often UBs affect dogs and their owners in Australia and the UK, and infer the beneficial impact of educating dog owners and veterinary professionals in modifying and managing UBs. Future reform of the veterinary curricula should reflect the effect of UB on such a significant proportion of the veterinary caseload.
Otitis externa is a common health disorder in Pugs and breeds with pendulous ears and brachycephalism. This study used anonymised electronic patient records (EPRs) from participating primary care veterinary practices in the VetCompass Australia (VCA) programme to identify the prevalence of otitis externa diagnosed in Pugs in Australia, associated demographic risk factors and comorbidities during 2017. There were 2,134 Pugs in participating practices between January 1st, 2017, to December 31st, 2017, and a random sample of 1,000 of these Pugs were selected for analysis. The one-year prevalence of otitis externa was 21.80% (95% confidence interval [CI]: 0.19”“0.24). Bodyweight (P < 0.001) and neuter status (P = 0.029) were demographic risk factors associated with otitis. Significant comorbidities associated with otitis externa included stenotic ear canals (odds ratio [OR] 938.42), anal sac impaction (OR 4.42), pyotraumatic dermatitis (OR 8.13), and allergy (OR 4.67). Among Pugs with otitis externa, 63 (29%) had recurring episodes. Otoscopic and cytological examinations were reported in 85.32% (186/218) of Pugs diagnosed with otitis externa. Dogs that were overweight or suffered from a dermatologic disorder exhibited greater risk of otitis externa. Veterinary care to reduce the onset and recurrence of otitis in Pugs may be improved through active management of associated risk factors including excess weight, allergic dermatitis and ear canal stenosis.
The English springer spaniel (ESS) is a popular breed in Australia. The breed has been reported to be predisposed to dermatological diseases including atopy, seborrhoea, Malassezia overgrowth, and otitis externa, in epidemiological studies elsewhere. However, only a few epidemiological studies investigating the common dermatological disorders and their risk factors for English springer spaniels have been reported in Australia. This study aims to investigate the prevalence of common dermatological disorders presenting to primary care veterinary practices and identify the disease risk factors by analysing electronic patient records from VetCompass Australia data. A total of 2834 de-identified patient records from 308 individual dogs identified as English springer spaniels which attended 140 participating veterinary primary care clinics in New South Wales, Victoria, and Queensland from 1 January, 2017 to 31 December, 2017 were reviewed and coded using VeNom coding and analysed in Excel and R. Otitis externa (26.44% prevalence) and Malassezia overgrowth on skin or in the external ear canal (20%) were two of the most prevalent dermatological disorders among Australian ESSs in 2017. Univariate and multivariate generalised linear regression models determined that ESSs identified by veterinarians as overweight/obese had an increased risk of a diagnosis of otitis externa (odds ratio=2.19, p-value=0.014) as were ESSs 9 years or older (odds ratio=5.23, p-value=0.0001). The multivariate linear regression model showed that ESSs 9 years and older (odds ratio=9.93, p-value=0.0046) had higher risks of Malassezia overgrowth. Skin hypersensitivity disorders increased the risk of otitis externa (odds ratio=2.75, p-value=0.03) and Malassezia overgrowth (odds ratio=3.36, p-value=0.01). An additional finding was that neuter status (neutered, p-value=0.01) and coat colour (black and white, p-value=0.02) increased the risk of ESS being overweight. Skin hypersensitivity, which had no predisposing demographic factors, increased the risk of Malassezia overgrowth and otitis externa in ESSs. The current study found that ESSs in Australia had a higher incidence of otitis externa than ESSs in the UK and Finland1,2, while overweight (odds ratio=2.19, p-value=0.014) and older age (odds ratio=5.23, p value=0.00001) were the predisposing factors. The prevalence of Malassezia dermatitis in ESSs was similar to a study conducted in 20063. Otic Malassezia was more frequent in Australian ESSs with otitis externa compared to cocker spaniels with otitis externa in the Finnish study2. ESSs in Australia had a higher risk of being overweight than ESSs studied in VetCompass UK study4, and coat colour was reported as a predisposing factor for overweight the first time. The limitation of this study was the inconsistent, incomplete record-keeping in many primary care practices, and will be a limitation on other uses of this data set. Overall, the findings of this study will provide the most complete record of the high prevalence of otic and dermatological disorders of Australian ESSs. These findings provide evidence for veterinarians to educate the owners of ESSs in Australia on the risks associated with skin hypersensitivity, overweight and older age on predisposition of ESSs to otitis externa and Malassezia overgrowth.
1. Summers JF, O’Neill DG, Church D et al. Health-related welfare prioritisation of canine disorders using electronic health records in primary care practice in the UK. BMC veterinary research 2019;15:163-163.
2. Zur G, Lifshitz B, Bdolah-Abram T. The association between the signalment, common causes of canine otitis externa and pathogens. Journal of small animal practice 2011;52:254-258.
3. Hill PB, Lo A, Eden CAN et al. Survey of the prevalence, diagnosis and treatment of dermatological conditions in small animals in general practice. Veterinary record 2006;158:533-539.
4. Pegram C, Raffan E, White E et al. Frequency, breed predisposition and demographic risk factors for overweight status in dogs in the UK. Journal of Small Animal Practice 2021;62:521-530.
Patellar luxation is one of the most common musculoskeletal disorders in Cavalier King Charles Spaniels (CKCS) and it can be a high welfare burden on dogs due to the chronic, progressive, painful and disabling effects of instability on joint function. Despite its welfare implications, risk factors and demographic information on patellar luxation in this breed have not been explored. This study aimed to establish risk factors that are associated with the prevalence and severity of patellar luxation in CKCS attending primary-care veterinary practices in Australia. Electronic patient records of CKCS (n = 321,517) from 2008 to 2017 were obtained from VetCompass Australia. Two risk factor analyses were performed for CKCS over 18 months of age: i) to compare case and control groups and ii) to assess risk factors influencing the severity of patellar luxation within the case group. The overall 10-year prevalence rate of patellar luxation in CKCS attending Australian veterinary practices was 12.5% (95% confidence interval (CI) 12-13). The greatest risk of developing patellar luxation was in neutered dogs (odds ratio (OR) = 3.00, 95% CI 1.97-4.71, p<0.001) and dogs weighing between 11 to <13kg (OR = 2.52, 95% CI 1.49-4.39, p<0.01), a weight range that exceeds the breed standard weight of 5.4-8.2kg as suggested by the Australian National Kennel Council. In addition, CKCS that were 10 years or older had the lowest risk (OR = 0.44, 95% CI 0.27-0.69, p<0.01) of being diagnosed with patellar luxation out of all age groups. Ruby-coloured CKCS were most likely to be diagnosed with patellar luxation (OR = 2.04, 95% CI 1.37-3.04, p<0.01). CKCS living in Victoria had the highest risk of severe patellar luxation (OR = 1.68, 95% CI 1.11-2.53, p<0.05) compared to dogs in other states. Moreover, CKCS with bilateral patellar luxation were more likely to be diagnosed with higher grade patellar luxation compared to those with unilateral patellar luxation. Increasing public recognition of patellar luxation as a breed-associated risk is required to reduce this condition in CKCS through early testing for the condition, weight management and avoiding breeding from the affected population.
Pugs are predisposed to corneal disorders due to their disproportionate skull morphology, reduced corneal sensitivity and eyelid anomalies such as distichiasis and entropion. Studies in the UK and USA have demonstrated that the risk of corneal disorders increases with age and conditions such as keratoconjunctivitis sicca (KCS), corneal pigmentation and corneal ulcer, are tightly associated with each other. However, the prevalence of ophthalmological disorders and corneal conditions have not been reported in Australia where the Pug population is young due to its rapid growth in popularity. Anonymised electronic patient records (EPR) of Pugs attending primary care veterinary clinics in the year of 2017 participating in VetCompass Australia (VCA) were used to investigate the demography of Pugs in Australia, the prevalence of ophthalmological and corneal pathologies in categories and specific disorders, the associated demographic risk factors, and the age of onset of specific ophthalmological disorders.
There was a high prevalence of overweight/obesity (prevalence: 20.2%, 95% CI: 18.1-22.4), ophthalmological abnormalities (14.5%, 95%CI: 12.6-16.3), and corneal disorders (12.4%, 95% C 12-15.7). There were 88.6% of Pugs with ophthalmological conditions that had corneal involvement. The most prevalent ophthalmological disorders were corneal ulcers (5.5%, 95%CI: 4.4-6.9), corneal pigmentation (3.6%, 95%CI: 2.8-4.8) and KCS (3.3%, 95%CI: 2.5-4.5). Age was a risk factor associated with the occurrence of corneal disorders. Corneal pigmentation, KCS and keratitis were more prevalent in Pugs that are greater than 7 years.
Based on the high prevalence of corneal disorders and age being a risk factor of corneal conditions, they should be health priorities in this breed which could allow early diagnosis and intervention to reduce the severity and suffering associated with these conditions.
Background: The Cavalier King Charles Spaniel is a breed selected for its favourable juvenile physiological and behavioural traits. There is no known link to the high prevalence of mitral valve disease in CKCS dogs, nor the risk factors associated with the shorter average longevity observed in this breed.
Methods: This retrospective cohort study investigated the longevity and risk factors associated with mortality within the CKCS breed over the 1st January 2008 to 31st December 2017. The aim of this is to report on demographic risk factors associated with euthanasia relative to unassisted death, as well as cardiac disorder incidences. This was achieved by using examination record data from over 200 participating veterinary clinics in the VetCompass Australia consortium.
Results: The most common primary specific cause of death was congestive heart failure (57.5%), followed by mass lesion (4.6%), osteoarthritis (4.0%). There is association of euthanised dogs with musculoskeletal disorders (p = 0.027) and mass lesions (p = 0.049), where incidences of both musculoskeletal disorders and mass lesions resulted in a greater likelihood of euthanasia. There is also an association of cardiac disorders with age at death (yearfrac) (p = 0.001), gastrointestinal disorders (p = 0.019) and respiratory disorders (p < 0.001). This refers to the incidences of both gastrointestinal disorders and respiratory disorders being associated with incidences of cardiac disorders, as well as the older the dog the more likely they have a diagnosis for a cardiac disorder.
Discussion:These results potentially could inform veterinarians, breeders and owners on the longevity and mortality of CKCS dogs. As well as enable the reform of veterinary treatment and breeding protocols to increase the longevity and welfare of CKCS dogs in the future, as there is currently no informed breeding protocol for CKCS dogs in Australia, possibly due to a lack of available data on the morbidities and mortalities of this breed.
The global problem of unowned domestic cats, driven by their phenomenal reproductive success, carries significant economic, animal welfare and biodiversity costs. Desexing owned cats prior to puberty prevents unwanted litters that contribute to unowned cat populations. The prevalence, and predictors of desexing, and the age at which surgery was carried out were investigated using anonymized electronic patient records in the VetCompass Australia database of cats presented to veterinary practices. Of 52,941 cats born between 2010 and 2017, 83.6% were desexed. Among 7463 desexed females, 21.5% had been desexed by 4 months of age, 59.8% by 6 months and 85.4% by 1 year. Sex, breed, location and socioeconomic indices significantly influenced desexing status and age at surgery. Cats born between 2010 and 2017 had greater odds of being desexed than cats born between 1995 and 2009 at each age cut-off (≤ 4 months [OR 1.76, CI 95 1.58–1.97], ≤ 6 months [OR 1.50, CI 95 1.38–1.62] and ≤ 1 year [OR 2.33, CI95 2.11–2.57] p < 0.001). Most cats presented to veterinarians in Australia are desexed. Compared with cats born before 2010, cats born later are significantly younger at desexing but, even so, many cats would have reached sexual maturity before surgery. These findings will inform the design of front-line strategies promoting prepubertal desexing and they demonstrate, for the first time, a shift towards desexing younger cats.
Introduction: Dogs with pruritic skin and ear diseases often require the use of anti-pruritic therapies to reduce and control the symptoms. These dogs may also have frequent secondary infections which may require recurring antimicrobial treatments. Using the VetCompassTM Australia database, oclacitinib (ApoquelTM), an anti-pruritic therapy, was shown to reduce antimicrobial usage in dogs with pruritic skin diseases compared to other anti-pruritic treatments. In this study, we aimed to uncover the geographical association between antimicrobial resistance in bacterial isolates cultured from canine skin and ear infections and antimicrobial usage in dogs treated with anti-pruritic therapies for pruritic skin and ear diseases in Queensland.
Methods: For this study, a retrospective secondary data linkage of 1) a veterinary laboratory database with bacterial culture and antimicrobial susceptibility results from skin and ear samples from dogs and 2) a VetCompassTM Australia database containing antimicrobial and anti-pruritic therapy usage was performed. Both databases were cleaned and only included data in Queensland, Australia from the 1st of January 2016 to the 21st of August 2018 such as age, sex, breed, neuter status, clinic and client postcodes and relevant dates.
Duplicates of the same bacterial species with the same antibiogram cultured from the same dog simultaneously were removed. If multiple different bacterial genus or species were cultured from the same body site or different sites, then they were all kept in the dataset. To be able to identify areas of high antimicrobial resistance, intrinsic resistances were firstly removed, and multidrug resistance (MDR) was calculated per isolate. The proportion of MDR isolates per clinic postcodes were calculated. Geographical clustering of the MDR patterns per South-East Queensland (SE-QLD) clinic postcodes was identified using Moran’s I and Local Indicators of Spatial Autocorrelation (LISA) statistics.
The proportion of dogs with pruritic skin and ear diseases treated using various prescribed treatments was calculated from the VetCompassTM Australia database per client postcode. This included, for instance, the proportion of dogs prescribed a) oclacitinib only, b) antimicrobials only (at least one topical or systemic antimicrobial), c) oclacitinib regardless of any other treatment types, d) oclacitinib and antimicrobials, and e) oclacitinib, antimicrobials and glucocorticoids. For each of the treatment options, Moran’s I and LISA statistics were conducted to identify the geographical clustering within SE-QLD.
The use of geographical clustering analysis aids in identifying co-distribution hot-spots (i.e., postcodes with a high proportion of isolates with MDR patterns and antimicrobial usage per anti-pruritic therapy). This will uncover whether there are lower levels of antimicrobial resistance from canine skin and ear infections within the same postcodes of dogs treated using oclacitinib compared with other anti-pruritic therapies. Lastly, we will estimate the sociodemographic and environmental risk factors for the geographical differences in resistance patterns between oclacitinib treated dogs versus other therapies using Bayesian condition autoregressive models.
Conclusion: Overall, as antimicrobial usage is considered to be the primary driver of antimicrobial resistance, we hope to demonstrate a direct association between oclacitinib and antimicrobial usage in pruritic dogs and geographical variation in bacterial resistance in dogs with skin and ear infections.
Understanding antimicrobial usage patterns and encouraging appropriate antimicrobial usage is a critical component of antimicrobial stewardship. Studies using VetCompass Australia and natural language processing have demonstrated antimicrobial usage patterns in companion animal practices across Australia. Doing so has highlighted the many obstacles and barriers to the task of converting raw clinical notes into a format that can be readily queried and analysed.
Some of the most important challenges are technology related. Training machine learning models to understand the reason for an antimicrobial administration relies on specific information being incorporated in the clinical records in a consistent manner. To train these models often requires annotations to made to clinical records so that the model can recognise the information that the clinical is recording as free text. Often such labels are unavailable. Having an expert annotate the records by hand to train such models is oftentimes cost prohibitive and extremely time consuming.
Many challenges are due to the record keeping of veterinary practitioners and the current software systems used to record and report such information. Use of free text, abbreviations and clinical shorthand are all designed to make work easier for the practitioner, but all of these practices mean that extracting meaningful data from the dataset is problematic. In addition, clinical records with missing data (patient weight, antimicrobial dosage and duration of treatment) make the utilising big data sets even more difficult.
We developed a system using novel artificial neural networks to examine the clinical notes and extract data describing the antimicrobial agent prescribed and the clinical indication for administration. This new system also allows for this to be done with annotations that can be created within two hours.
Our methods were applied to over 4.4 million companion animal clinical records across Australia to help us understand what antibiotics are being given and why on a population level. The key elements to assess appropriate antimicrobial use are clinical indication, antimicrobial agent selection, dose and duration of therapy. All consultations with antimicrobial use were selected, and of these approximately only 40% recorded the reason why antimicrobials were prescribed, along with the dose and duration of treatment.
Novel neural networks might be able to overcome the difficulties of harvesting free text data from clinical records, but when the essential data is not recorded in the clinic records then the human barrier becomes the insurmountable obstacle.
Introduction: Currently, knowledge of how antimicrobials are used or the appropriateness of use in Australian horses is limited. Hoof abscesses are a common cause of acute lameness in horses for which effective treatment is largely dependent on establishing drainage and antimicrobial therapy is not indicated unless the patient is immunosuppressed, or cellulitis is present. This study aimed to identify the frequency of antimicrobial prescribing in horses with a hoof abscess using Australian VetCompass equine clinical records.
Materials and methods:Natural language processing (NLP) methods were used to label free text clinical records for the indication of an abscess. These records were then manually reviewed to identify abscesses involving the hoof and their antimicrobial prescriptions.
Results: The NLP methods extracted 2,132 abscess records (1,742 horses) in which a clinical finding of a hoof abscess was manually identified in 1117 records (928 horses) from 21 clinics. No complicating factors were documented in 940 hoof abscess records, but 54% of these had antimicrobials prescribed. Of the 680 records specifying the hoof abscess drainage status, antibiotics were more frequently prescribed than not in those already draining (70%) and those where drainage was established (52%). In non-draining abscesses, less than half (43%) were prescribed antibiotics. The most frequently prescribed antibiotics were procaine penicillin, trimethoprim sulphonamides and long-acting penicillin.
Relevance to Australian clinical equine practice: Although predominantly low importance antimicrobials were prescribed, most records had no complicating factors documented and thus antimicrobials were likely not indicated. This study identified a need to develop antimicrobial stewardship resources to target improved prescribing for equine hoof abscesses by Australian veterinarians.
The VetCompass Australia (VCA) program collects real-time clinical records from veterinary practices and aggregates them for researchers to interrogate. Bringing together all seven Australian veterinary schools, it is the world’s first nationwide surveillance system collating clinical records on companion-animal diseases and treatments. This represents a significant resource but only offers access to part of the veterinary record.
Building on the VCA collaboration and infrastructure, we are developing the Veterinary and Animal Research Data Commons (VARDC) with co-investment from the Australian Research Data Commons (ARDC). Our platform will deliver big-data opportunities to Australian veterinary and animal science researchers allowing them to mesh clinical data on focal animals with related data, such as pathology reports and imaging. It will offer a series of tools that will enhance the text searching, cleaning and analysis through the Virtual Lab, and will connect with other platforms, such as the Australian Imaging Service (AIS).
The VARDC platform is an Enterprise application leveraging Azure PaaS services. This includes Azure services and serverless features and is coupled with Microsoft’s customer identity access solution Azure AD B2C for authentication - providing a flexible, scalable and cost-effective solution.
When complete in 2023, this platform will offer a unique opportunity to examine animal health by bringing the many benefits of a big data approach to animal health, whilst maintaining a highly detailed view of the individual animal.
Clinical placements inside and outside the university setting are critical in veterinary training. Opportunities to learn and practise technical and communication skills are well-recognised benefits of extramural placements (EMP). Less well-recognised is provision of opportunities to consider and evaluate “real world” clinical thinking, history-taking, problem-solving, management planning and record-keeping, all of which are important Day One Skills. The electronic patient record (EPR) encapsulates the caseload of a practice, displaying its approach to clinical reasoning and capturing the veterinary life history of individual animals.
The University of Adelaide and VetCompass Australia (VCA) have developed a Virtual Online Clinical Placement package to help deliver some of the non-technical learning outcomes of EMPs described above. Recognising the merits of a unified approach across the vet schools, VCA reviewed EMP Learning Objectives for several schools, considering them in combination with RCVS Day One Competences. These were used to guide the development of the VOCP pilot. The VOCP aims are:
- To enhance students’ research skills through clinical data analysis and report writing.
- To provide students an opportunity to reflect on the way records are kept in veterinary practice and how they will approach this after they graduate.
- To provide an alternative to student face to face EMS placements, opportunities for which have been and are likely to continue to be reduced due to COVID-19 restrictions.
The VOCP program was run as two consecutive two-week pilots with small groups of students enrolled in the Doctor of Veterinary Medicine at the University of Adelaide. For the VOCP, students accessed the VCA online application to review between 250 and 500 medical records from veterinary clinics around Australia. Based on the most recent visit, each record was coded by the student according to short, pre-set questions. Students then analysed the records, reviewing for common themes, then selecting one topic of interest for deeper consideration. The students each produced a final report, following a simple, flexible template, uploading the report into Sonia, the student placement platform. Each report was peer-assessed anonymously. Finally, students were asked to complete an evaluation of the VOCP experience, with comments from the first group leading to modified setup and instructions for the second.
With the COVID-19 pandemic, EMPs have been difficult to secure and prone to cancellation. The VOCP is not intended to substantially replace face to face experiences. However, this virtual experience provides time for students to focus on the value of clinical records, enhancing their appreciation of common complaints in veterinary practice and variations in their management. At the same time, it engages students in simple research activity and critical thinking. A common student comment was that the VOCP highlights the valuable role of good record keeping in veterinary practice. In sum, the VOCP package engages students in authentic veterinary activities with a managed risk of plagiarism and a minimal assessment load, and is a valuable addition to the variety of available in-person EMPs.
With stable annual registrations around 1000 dogs over the past decade, West Highland White Terriers (WHWTs) are considered the third most popular terrier breed in Australia. Despite its popularity and well-known predisposition towards allergic skin conditions, few studies were conducted to overview the breed’s current status. By accessing and analysing electronic patient records from the VetCompass Australia (VCA) database, we were able to describe the population profile (demography and mortality) of WHWTs attending Australian primary-care clinics from 2008 to 2017, as well as estimating the period prevalence of health conditions identified by veterinarians in the same breed in 2017. The LMS method was used to plot WHWT growth curves. The impact of risk factors (sex, neuter status and overweight or obesity) on longevity was assessed by Log Rank test and Kaplan-Meier curves. The correlation between disease occurrence and risk factors (age, sex and neuter status) were tested in multi-variate logistic regression models.
The spatial distribution of the 4372 WHWTs included in our study was basically consistent with the human population distribution, as most of them resided in the major urbanised areas along the eastern coast. The sex ratio was approximately 1 to 1, and 71.2% (n = 3113) of the population was neutered. The rapid juvenile musculoskeletal growth of WHWTs generally ceased around one year of age (50 weeks), and the median adult body weight was 8.5 kg (female median adult body weight was 8 kg, and male median adult body weight was 9 kg). The median longevity of WHWTs deceased during the study period (n = 511) was 13 years, and their mortality was most frequently subsequent to disorder and health condition (n = 317, 62.04%), led by mass lesion (n = 70, 21.9%) and urogenital conditions including renal failure (n = 37, 11.6%). There was no significant difference in longevity between the two sexes, while neutered WHWTs appeared to live a longer life than their entire counterpart (P < 0.001). Overweight and obesity did not affect the dogs’ longevity as well. Regarding the 2017 prevalence of common health conditions in WHWTs under primary veterinary care in Australia, the population (n = 1060) was challenged by the age-related dental, integumentary, musculoskeletal and urogenital conditions (P < 0.001). In senior population (at least 11 years old), females were more prone to urogenital conditions compared to males (P < 0.001), regardless of their neuter status. The most common conditions reported in the study population were dental conditions (36.8%, 95% CI 33.9 – 39.7%), integumentary conditions (33%, 95% CI 30.2 – 35.9%; eg. otitis externa, hypersensitivity skin disorder, and unspecified dermatitis), and musculoskeletal conditions (12.8%, 95% CI 10.8 – 14.8%; eg. patellar luxation, osteoarthritis and unspecified arthritis). Our findings further confirmed the overrepresentation of this breed for allergy-related conditions, and emphasised the importance of senior care, as well as management of chronic conditions in WHWTs. This stronger, breed-specific evidence base provided by the current study could help veterinarians to predict medical concerns in this breed, and facilitate breeder’s planning in Australia.
|Prevalence and risk factors for otitis externa in Pugs in Australia, Leanne Leong, University of Sydney|
Are large language models the future of clinical natural language processing?
Jenny Wilshaw BVetMed MRCVS PhD, Royal Veterinary College
Large databases of electronic healthcare records (EHRs)are now a well-established part of the veterinary research landscape. They are uniquely positioned to provide broad, generalisable insights into conditions in practice; for some sense of scale, (UK) currently contains records from over 30% UK practices and over 26 million animals.
Veterinary databases such as are predominantly comprised of unstructured clinical notes. Information is captured in blocks of free text rather than as clinical codes or individual measurements. To use these big datasets, we therefore require human interpretation. Annotators with domain knowledge are well placed for this task as they can extract data either directly or inferentially from contextual information. Although this approach is the standard, the scale to which it can be applied is limited by time and the number of suitably qualified individuals.
Natural language processing (NLP) is a branch of computer science that mayoffer a solution to this problem. Models of language developed using machine learning can be applied to clinical data to completea number of tasks,relevant examples of which being information extraction and predictive modelling for classification or prognostication. Models provide a scalable solution, meaning that they can generate more resultsthan their human counterparts, however they must be provided with enough data to “learn” the correct response to a specific task. Therefore, training a model to be applied to veterinary data often hassimilar requirements to projects using human annotation alone.
Until recently, this has been one of the limiting factors in applying NLP in our domain. However, billion-parameter models, so called large language models (LLMs)have overcome this, demonstrating remarkable abilities simply due to their size. Large language models can produce results having been shown very few examples, in some cases, completing tasks without having seen the correct response altogether.
In this presentation, we will introduce LLMs and using real examples, demonstrate how this type of model could be applied to veterinary clinical notes.