Physical Examination Skills in the Context of Burgeoning Diagnostic Tests and AI

Presentation Description / Summary: Novel tests and AI continue to develop exponentially, enhancing or replacing our work as veterinary internists. As such, identifying, practicing and promoting aspects of our profession that are unlikely to ever be replaced and are of high value is of paramount importance. Diagnostic tests, automated animal health tracking and AI interpretation can be of great benefit to animal health. There are, however, potential pitfalls particularly when applied to the individual animal. AI systems can ‘learn’ but only on data they acquire or that is provided, leading to several problems. Firstly, there are facets of clinical examination that AI would be hard pressed to replicate e.g., palpation per rectum, lymph node palpation, ocular examination. Secondarily, compared with human medicine, veterinary prospective and retrospective case-control studies have lower sample sizes with statistical significance of p < 0.05, often equaling close to 1 in 20 (e.g., p = 0.047). Internists encounter this fraction of outliers – the 5% not uncommonly. Inherent heterogeneity in actual patients and unexpected variables mean errors and bias can multiply - a form of ‘AI drift’. Combine this with aspects of clinical examination that AI cannot obtain or interpret leads to an incomplete assessment - unusual or novel diagnoses or outcomes may be missed. This session will outline the types of AI used in large animal veterinary medicine along with its strengths, weaknesses and sources of bias using case examples. Further, the session will identify and revisit irreplaceable ‘human skills’ such as the dying art of thorough physical examination.