Developing a network for small animal disease surveillance

Text mining for disease surveillance in veterinary clinical data: part two, training computers to identify features in clinical text

In part two of this mini-series, we evaluate the range of machine-learning tools now available for application to veterinary clinical text-mining. These tools will be vital to automate extraction of information from large datasets of veterinary clinical narratives curated by projects such as the Small Animal Veterinary Surveillance Network (SAVSNET) and VetCompass, where volumes of millions of records preclude reading records and the complexities of clinical notes limit usefulness of more "traditional" text-mining approaches. We discuss the application of various machine learning techniques ranging from simple models for identifying words and phrases with similar meanings to expand lexicons for keyword searching, to the use of more complex language models. Specifically, we describe the use of language models for record annotation, unsupervised approaches for identifying topics within large datasets, and discuss more recent developments in the area of generative models (such as ChatGPT). As these models become increasingly complex it is pertinent that researchers and clinicians work together to ensure that the outputs of these models are explainable in order to instill confidence in any conclusions drawn from them.

Using topic modelling for unsupervised annotation of electronic health records to identify an outbreak of disease in UK dogs

A key goal of disease surveillance is to identify outbreaks of known or novel diseases in a timely manner. Such an outbreak occurred in the UK associated with acute vomiting in dogs between December 2019 and March 2020. We tracked this outbreak using the clinical free text component of anonymised electronic health records (EHRs) collected from a sentinel network of participating veterinary practices. We sourced the free text (narrative) component of each EHR supplemented with one of 10 practitioner-derived main presenting complaints (MPCs), with the ‘gastroenteric’ MPC identifying cases involved in the disease outbreak. Such clinician-derived annotation systems can suffer from poor compliance requiring retrospective, often manual, coding, thereby limiting real-time usability, especially where an outbreak of a novel disease might not present clinically as a currently recognised syndrome or MPC. Here, we investigate the use of an unsupervised method of EHR annotation using latent Dirichlet allocation topic-modelling to identify topics inherent within the clinical narrative component of EHRs. The model comprised 30 topics which were used to annotate EHRs spanning the natural disease outbreak and investigate whether any given topic might mirror the outbreak time-course. Narratives were annotated using the Gensim Library LdaModel module for the topic best representing the text within them. Counts for narratives labelled with one of the topics significantly matched the disease outbreak based on the practitioner-derived ‘gastroenteric’ MPC (Spearman correlation 0.978); no other topics showed a similar time course. Using artificially injected outbreaks, it was possible to see other topics that would match other MPCs including respiratory disease. The underlying topics were readily evaluated using simple word-cloud representations and using a freely available package (LDAVis) providing rapid insight into the clinical basis of each topic. This work clearly shows that unsupervised record annotation using topic modelling linked to simple text visualisations can provide an easily interrogable method to identify and characterise outbreaks and other anomalies of known and previously un-characterised diseases based on changes in clinical narratives.

Veterinary Big Data: When Data Goes to the Dogs

Simple Summary

Big data has created many opportunities to improve both preventive medicine and medical treatments. In the field of veterinary medical big data, information collected from companion animals, primarily dogs, can be used to inform healthcare decisions in both dogs and other species. Currently, veterinary medical datasets are an underused resource for translational research, but recent advances in data collection in this population have helped to make these data more accessible for use in translational studies. The largest open access dataset in the United States is part of the Dog Aging Project and includes detailed information about individual dog participant’s physical and chemical environments, diet, exercise, behavior, and comprehensive health history. These data are collected longitudinally and at regular intervals over the course of the dog’s lifespan. Large-scale datasets such as this can be used to inform our understanding of health, disease, and how to increase healthy lifespan.

Abstract

Dogs provide an ideal model for study as they have the most phenotypic diversity and known naturally occurring diseases of all non-human land mammals. Thus, data related to dog health present many opportunities to discover insights into health and disease outcomes. Here, we describe several sources of veterinary medical big data that can be used in research. These sources include medical records from primary medical care centers or referral hospitals, medical claims data from animal insurance companies, and datasets constructed specifically for research purposes. No data source provides information that is without limitations, but large-scale, prospective, longitudinally collected data from dog populations are ideal for further research as they offer many advantages over other data sources.

Backyard poultry cases in UK small animal practices: Demographics, health conditions and pharmaceutical prescriptions

BACKGROUND

Backyard poultry ownership is of keen interest in the United Kingdom. However, despite this, little is known about veterinary care engagement and outcomes of visits in this group of species.

METHODS

This study described and characterised veterinary practice-visiting backyard poultry, utilising electronic health record data supplied by veterinary practices voluntarily participating in the Small Animal Veterinary Surveillance Network between 1st April 2014 and 31st March 2019.

RESULTS

In total, 4424 recorded poultry consultations originating from 197 veterinary practices (352 sites) were summarised. Chicken consultation (n = 3740) peak incidence was in early summer (April-June), relative to all recorded species. More chickens resided in rural (incident rate ratio = 2.5, confidence interval [CI] 2.3-2.6, p <0.001) or less deprived areas. Non-specific clinical signs were commonly recorded (17.6% of chicken consultations, CI 15.9-19.2), as were those indicative of advanced disease. This latter finding was reflected in prescribed management strategies, with euthanasia comprising 29.8% (CI 27.0-32.6) of consultations. Antimicrobials were commonly prescribed (33.0% of consultations, CI 29.8-36.2), 43.8% of which included antimicrobials considered 'highest priority critically important' by the World Health Organisation.

CONCLUSION

Our findings indicate a need to tailor antimicrobial prescription guidance to the backyard poultry setting. In addition, late presentation of disease, vague clinical descriptions in clinical narratives and high euthanasia rates show that disease identification, management and knowledge of poultry health and welfare among owners and veterinary surgeons can be improved.

What drives antimicrobial prescribing for companion animals? A mixed-methods study of UK veterinary clinics

Antimicrobial use in companion animals is a largely overlooked contributor to the complex problem of antimicrobial resistance. Humans and companion animals share living spaces and some classes of antimicrobials, including those categorised as Highest Priority Critically Important Antimicrobials (HPCIAs). Veterinary guidelines recommend that these agents are not used as routine first line treatment and their frequent deployment could offer a surrogate measure of 'inappropriate' antimicrobial use. Anthropological methods provide a complementary means to understand how medicines use makes sense 'on-the-ground' and situated in the broader social context. This mixed-methods study sought to investigate antimicrobial use in companion animals whilst considering the organisational context in which increasing numbers of veterinarians work. Its aims were to i) to epidemiologically analyse the variation in the percentage of antimicrobial events comprising of HPCIAs in companion animal dogs attending UK clinics belonging to large veterinary groups and, ii) to analyse how the organisational structure of companion animal practice influences antimicrobial use, based on insight gained from anthropological fieldwork. A VetCompass^TM dataset composed of 468,665 antimicrobial dispensing events in 240,998 dogs from June 2012 to June 2014 was analysed. A hierarchical model for HPCIA usage was built using a backwards elimination approach with clinic and dog identity numbers included as random effects, whilst veterinary group, age quartile, breed and clinic region were included as fixed effects. The largest odds ratio of an antimicrobial event comprising of a HPCIA by veterinary group was 7.34 (95% confidence interval 5.14 - 10.49), compared to the lowest group (p < 0.001). Intraclass correlation was more strongly clustered at dog (0.710, 95% confidence interval 0.701 - 0.719) than clinic level (0.089, 95% confidence interval 0.076 -0.104). This suggests that veterinarians working in the same clinic do not automatically share ways of working with antimicrobials. Fieldwork revealed how the structure of the companion animal veterinary sector was more fluid than that depicted in the statistical model, and identified opportunities and challenges regarding altering antimicrobial use. These findings were organised into the following themes: "Highest priority what?"; "He's just not himself"; "Oh no - here comes the antibiotics police"; "We're like ships that pass in the night"; and "There's not enough hours in the day". This rigorous mixed-methods study demonstrates the importance of working across disciplinary silos when tackling the complex problem of antimicrobial resistance. The findings can help inform the design of sustainable stewardship schemes for the companion animal veterinary sector.

Is it time to stop sweeping data cleaning under the carpet? A novel algorithm for outlier management in growth data

All data are prone to error and require data cleaning prior to analysis. An important example is longitudinal growth data, for which there are no universally agreed standard methods for identifying and removing implausible values and many existing methods have limitations that restrict their usage across different domains. A decision-making algorithm that modified or deleted growth measurements based on a combination of pre-defined cut-offs and logic rules was designed. Five data cleaning methods for growth were tested with and without the addition of the algorithm and applied to five different longitudinal growth datasets: four uncleaned canine weight or height datasets and one pre-cleaned human weight dataset with randomly simulated errors. Prior to the addition of the algorithm, data cleaning based on non-linear mixed effects models was the most effective in all datasets and had on average a minimum of 26.00% higher sensitivity and 0.12% higher specificity than other methods. Data cleaning methods using the algorithm had improved data preservation and were capable of correcting simulated errors according to the gold standard; returning a value to its original state prior to error simulation. The algorithm improved the performance of all data cleaning methods and increased the average sensitivity and specificity of the non-linear mixed effects model method by 7.68% and 0.42% respectively. Using non-linear mixed effects models combined with the algorithm to clean data allows individual growth trajectories to vary from the population by using repeated longitudinal measurements, identifies consecutive errors or those within the first data entry, avoids the requirement for a minimum number of data entries, preserves data where possible by correcting errors rather than deleting them and removes duplications intelligently. This algorithm is broadly applicable to data cleaning anthropometric data in different mammalian species and could be adapted for use in a range of other domains.

Syndromic surveillance by veterinary practitioners: a pilot study in the pig sector

Traditional indicator-based livestock surveillance has been focused on case definitions, definitive diagnoses and laboratory confirmation. The use of syndromic disease surveillance would increase the population base from which animal health data are captured and facilitate earlier detection of new and re-emerging threats to animal health. Veterinary practitioners could potentially play a vital role in such activities. In a pilot study, specialist private veterinary practitioners (PVP) working in the English pig industry were asked to collect and transfer background data and disease incident reports for pig farms visited during the study period. Baseline data from 110 pig farms were received, along with 68 disease incident reports. Reports took an average of approximately 25 minutes to complete. Feedback from the PVPs indicated that they saw value in syndromic surveillance. Maintenance of anonymity in the outputs would be essential, as would timely access for the PVPs to relevant information on syndromic trends. Further guidance and standardisation would also be required. Syndromic surveillance by PVPs is possible for the pig industry. It has potential to fill current gaps in the collection of animal health data, as long as the engagement and participation of data providers can be obtained and maintained.

Validation of text-mining and content analysis techniques using data collected from veterinary practice management software systems in the UK

Electronic patient records from practice management software systems have been used extensively in medicine for the investigation of clinical problems leading to the creation of decision support frameworks. To date, technologies that have been utilised for this purpose such as text mining and content analysis have not been employed significantly in veterinary medicine. The aim of this research was to pilot the use of content analysis and text-mining software for the synthesis and analysis of information extracted from veterinary electronic patient records. The purpose of the work was to be able to validate this approach for future employment across a number of practices for the purposes of practice based research. The approach utilised content analysis (Prosuite) and text mining (WordStat) software to aggregate the extracted text. Text mining tools such as Keyword in Context (KWIC) and Keyword Retrieval (KR) were employed to identify specific occurrences of data across the records. Two different datasets were interrogated, a bespoke test dataset that had been set up specifically for the purpose of the research, and a functioning veterinary clinic dataset that had been extracted from one veterinary practice. Across both datasets, the KWIC analysis was found to have a high level of accuracy with the search resulting in a sensitivity of between 85.3-100%, a specificity of between 99.1-99.7%, a positive predictive value between 93.5-95.8% and a negative predictive value between 97.7-100%. The KR search, based on machine learning, was utilised for the clinic-based dataset and was found to perform slightly better than the KWIC analysis. This study is the first to demonstrate the application of content analysis and text mining software for validation purposes across a number of different datasets for the purpose of search and recall of specific information across electronic patient records. This has not been demonstrated previously for small animal veterinary epidemiological research for the purposes of large scale analysis for practice-based research. Extension of this work to investigate more complex diseases across larger populations is required to fully explore the use of this approach in veterinary practice.

Safety culture: the Nottingham Veterinary Safety Culture Survey (NVSCS)

Safety culture is a vital concept in human healthcare because of its influence on staff behaviours in relation to patient safety. Understanding safety culture is essential to ensure the acceptance and sustainability of changes, such as the introduction of safe surgery checklists. While widely studied and assessed in human medicine, there is no tool for its assessment in veterinary medicine. This paper therefore presents initial data on such an assessment: the Nottingham Veterinary Safety Culture Survey (NVSCS). 350 pilot surveys were distributed to practising vets and nurses. The survey was also available online. 229 surveys were returned (65 per cent response rate) and 183 completed online, resulting in 412 surveys for analysis. Four domains were identified: (1) organisational safety systems and behaviours, (2) staff perceptions of management, (3) risk perceptions and (4) teamwork and communication. Initial indications of the reliability and the validity of the final survey are presented. Although early in development, the resulting 29-item NVSCS is presented as a tool for measuring safety culture in veterinary practices with implications for benchmarking, safety culture assessment and teamwork training.

Inventory of veterinary syndromic surveillance initiatives in Europe (Triple-S project): current situation and perspectives

Within the current context that favours the emergence of new diseases, syndromic surveillance (SyS) appears increasingly more relevant tool for the early detection of unexpected health events. The Triple-S project (Syndromic Surveillance Systems in Europe), co-financed by the European Commission, was launched in September 2010 for a three year period to promote both human and animal health SyS in European countries. Objectives of the project included performing an inventory of current and planned European animal health SyS systems and promoting knowledge transfer between SyS experts. This study presents and discusses the results of the Triple-S inventory of European veterinary SyS initiatives. European SyS systems were identified through an active process based on a questionnaire sent to animal health experts involved in SyS in Europe. Results were analyzed through a descriptive analysis and a multiple factor analysis (MFA) in order to establish a typology of the European SyS initiatives. Twenty seven European SyS systems were identified from twelve countries, at different levels of development, from project phase to active systems. Results of this inventory showed a real interest of European countries for SyS but also highlighted the novelty of this field. This survey highlighted the diversity of SyS systems in Europe in terms of objectives, population targeted, data providers, indicators monitored. For most SyS initiatives, statistical analysis of surveillance results was identified as a limitation in using the data. MFA results distinguished two types of systems. The first one belonged to the private sector, focused on companion animals and had reached a higher degree of achievement. The second one was based on mandatory collected data, targeted livestock species and is still in an early project phase. The exchange of knowledge between human and animal health sectors was considered useful to enhance SyS. In the same way that SyS is complementary to traditional surveillance, synergies between human and animal health SyS could be an added value, most notably to enhance timeliness, sensitivity and help interpreting non-specific signals.

Antibacterial prescribing patterns in small animal veterinary practice identified via SAVSNET: the small animal veterinary surveillance network

In this study, data from veterinary clinical records were collected via the small animal veterinary surveillance network (SAVSNET). Over a three-month period, data were obtained from 22,859 consultations at 16 small animal practices in England and Wales: 69 per cent from dogs, 24 per cent from cats, 3 per cent from rabbits and 4 per cent from miscellaneous species. The proportion of consults where prescribing of antibacterials was identified was 35.1 per cent for dogs, 48.5 per cent for cats and 36.6 per cent for rabbits. Within this population, 76 per cent of antibacterials prescribed were β-lactams, including the most common group of clavulanic acid-potentiated amoxicillin making up 36 per cent of the antibacterials prescribed. Other classes included lincosamides (9 per cent), fluoroquinolones and quinolones (6 per cent) and nitroimidazoles (4 per cent). Vancomycin and teicoplanin (glycopeptide class), and imipenem and meropenem (β-lactam class) prescribing was not identified. Prescribing behaviour varied between practices. For dogs and cats, the proportion of consults associated with the prescription of antibacterials ranged from 0.26 to 0.55 and 0.41 to 0.73, respectively.