Premature mortality analysis of 52,000 deceased cats and dogs exposes socioeconomic disparities

Sean Farrell, Katharine L. Anderson, Peter‐John M. Noble, Noura Al MoubayedScientific 2024 · DOI

TL;DR

This study leverages a novel LLM-based approach (PetBERT-ICD) to analyze the causes of premature death in over 52,000 cats and dogs in the UK. It reveals significant socioeconomic disparities in pet longevity, with animals in less deprived areas having a nearly 50% lower risk of premature mortality, and highlights breed-specific risks, particularly for brachycephalic dogs and behavioral issues in young dogs.

What Problem Does This Paper Address?

Monitoring mortality patterns is crucial for understanding public health and resource allocation, yet national mortality data for companion animals is not systematically collected. This lack of data makes it difficult to identify trends, understand the impact of various factors on pet longevity, and address potential health inequalities. Previous attempts to analyze pet health often rely on insurance data, which can be biased by breed and may not capture all cases, especially severe ones. The paper aims to bridge this gap by developing a robust method to analyze premature mortality in companion animals using veterinary electronic health records (EHRs).

Methods

The core of this research lies in the application of PetBERT-ICD, a specialized large language model (LLM) fine-tuned for classifying veterinary clinical narratives according to the International Classification of Diseases 11 (ICD-11) syndromic framework. This automated approach overcomes the limitations of manual coding and the unstructured nature of free-text veterinary records.

Key methodological steps include:

  • Data Extraction: A generalized regular expression was used to identify records mentioning death or euthanasia. A semi-supervised teacher-student model was then employed to train a binary classifier for case identification, followed by manual verification of a subset of records.
  • Premature Mortality Definition: Breed-specific premature longevity thresholds were established using a bootstrapping method. An animal was considered to have died prematurely if its age at death was below 85% of the lower 95% confidence interval of its breed's median age at death.
  • Cause of Mortality Analysis: ICD-11 chapter labels assigned by PetBERT-ICD were used to calculate Years of Lost Life (YLL) for different age groups and disease categories.
  • Risk Factor Analysis: Univariable and multivariable logistic regression models were used to identify risk factors associated with premature mortality, considering variables such as sex, socioeconomic status (Index of Multiple Deprivation - IMD), and breed characteristics (brachycephalic).

Datasets

The primary dataset used in this study is derived from the Small Animal Veterinary Surveillance Network (SAVSNET). This network comprises electronic health records (EHRs) from 253 volunteer veterinary practices across the UK, collected between March 2014 and March 2024.

  • Total Consultations: 9,281,287
  • Canine and Feline Consultations: 5,772,469 (dogs) and 2,241,767 (cats)
  • Deceased Animals Analyzed: 28,159 dogs and 24,006 cats.
  • Training/Evaluation Data for PetBERT-ICD: 400 professionally annotated records for training and 200 for evaluation.

No new benchmarks were introduced; the study leverages existing ICD-11 classification and the SAVSNET dataset.

Key Insights and Ideas

  • Socioeconomic Disparities are Stark: The most striking finding is the strong correlation between socioeconomic status and pet longevity. Animals in areas with lower deprivation (higher socioeconomic status) have a significantly lower risk of premature death, suggesting that access to resources, including veterinary care, plays a critical role.
  • Behavioral Issues a Major Cause of Premature Death in Young Dogs: For dogs aged 1-6 years, behavioral disorders are a leading cause of premature death, often leading to euthanasia. This points to a need for better behavioral support and early intervention strategies.
  • Brachycephalic Breeds Face Higher Risks: Brachycephalic dog breeds have a higher risk of premature mortality, likely due to inherent health issues associated with their facial structure, such as respiratory problems.
  • LLMs Revolutionize Veterinary Data Analysis: The successful application of PetBERT-ICD demonstrates the power of LLMs in extracting meaningful, standardized information from unstructured veterinary clinical narratives, opening doors for large-scale epidemiological studies in animal health.

Key Results

  • Socioeconomic Impact: Areas with the lowest Index of Multiple Deprivation (IMD) scores showed nearly a 50% reduction in the risk of premature mortality for both cats and dogs compared to the most deprived areas.
  • Brachycephalic Dog Risk: Brachycephalic dogs had a 1.19-fold increase in the odds of premature mortality.
  • Behavioral Disorders in Dogs: Behavioral issues were associated with a 5.69-fold increase in the odds of premature mortality in dogs, compared to 2.13 in cats.
  • Age-Related Causes: In cats, "Neoplasms" were a significant cause of YLL in older age groups, while "Certain infectious or parasitic diseases" were leading causes in younger cats. In dogs, "Mental, behavioural or neurodevelopmental disorders" were prominent in younger ages, with "Neoplasms" increasing with age.
  • Euthanasia Dominance: Euthanasia accounted for over 99% of recorded deaths in both cats and dogs, highlighting its role as a primary end-of-life decision.

Limitations and Future Work

  • Data Quality and Completeness: The quality and completeness of free-text consultation notes varied significantly among clinicians, introducing potential inconsistencies. The study acknowledges that the free-text nature of EHRs can lead to inaccuracies and omissions.
  • ICD-11 Syndromic Labels: The ICD-11 labels provide broad categorizations, not specific causes of death, limiting the granularity of causal analysis.
  • Mixed-Breed Dog Categorization: Mixed-breed dogs were categorized as "other" due to insufficient cephalic data, and their diverse phenotypic features, including potential brachycephalic traits, were not fully accounted for in breed-specific analyses. The study also did not account for size-based classifications in mixed breeds.
  • Underrepresentation of Non-Euthanasia Deaths: The omission of animals that died without euthanasia likely under-represents their numbers.
  • SAVSNET Representativeness: The SAVSNET network, while extensive, may not fully represent the entire companion animal population across the UK.

Future Work: The authors suggest developing more sophisticated methods to extract true causes of death beyond syndromic labels. Further research is needed to explore the socioeconomic disparities in more detail and develop targeted interventions. Investigating the impact of size on lifespan in mixed-breed dogs and exploring more granular categorizations of mixed-breed dogs are also recommended.

Summary updated

Abstract

Monitoring mortality rates offers crucial insights into public health by uncovering the hidden impacts of diseases, identifying emerging trends, optimising resource allocation, and informing effective policy decisions. Here, we present a novel approach to analysing premature mortality in companion animals, utilising data from 28,159 deceased dogs and 24,006 deceased cats across the United Kingdom. By employing PetBERT-ICD, an automated large language model (LLM) based International Classification of Disease 11 syndromic classifier, we reveal critical insights into the causes and patterns of premature deaths. Our findings highlight the significant impact of behavioural conditions on premature euthanasia in dogs, particularly in ages one to six. We also identify a 19% increased risk of premature mortality in brachycephalic dog breeds, raising important animal welfare concerns. Our research establishes a strong correlation between socioeconomic status and premature mortality in cats and dogs. Areas with the lowest Index of Multiple Deprivation (IMD) scores show nearly a 50% reduction in the risk of premature mortality across cats and dogs, underscoring the powerful impact that socioeconomic factors can have on pet health and longevity. This research underscores the necessity of examining the socioeconomic disparities affecting animal health outcomes. By addressing these inequities, we can better safeguard the well-being of our companion animals.
BibTeX
@inproceedings{farrell2024premature,
  author = {Sean Farrell and Katharine L. Anderson and Peter‐John M. Noble and Noura Al Moubayed},
  title = {{Premature mortality analysis of 52,000 deceased cats and dogs exposes socioeconomic disparities}},
  booktitle = {Scientific},
  year = {2024},
  doi = {10.1038/s41598-024-77385-8},
}
Figure