ChatGPT's Epoch in Rheumatological Diagnostics: A Critical Assessment in the Context of Sjögren's Syndrome

Abstract

Introduction: The rise of artificial intelligence in medical practice is reshaping clinical care. Large language models (LLMs) like ChatGPT have the potential to assist in rheumatology by personalizing scientific information retrieval, particularly in the context of Sjögren's Syndrome. This study aimed to evaluate the efficacy of ChatGPT in providing insights into Sjögren's Syndrome, differentiating it from other rheumatological conditions. Materials and methods: A database of peer-reviewed articles and clinical guidelines focused on Sjögren's Syndrome was compiled. Clinically relevant questions were presented to ChatGPT, with responses assessed for accuracy, relevance, and comprehensiveness. Techniques such as blinding, random control queries, and temporal analysis ensured unbiased evaluation. ChatGPT's responses were also assessed using the 15questionnaire DISCERN tool.Results: ChatGPT effectively highlighted key immunopathological and histopathological characteristics of Sjögren's Syndrome, though some crucial data and citation inconsistencies were noted. For a given clinical vignette, ChatGPT correctly identified potential etiological considerations with Sjögren's Syndrome being prominent.Discussion: LLMs like ChatGPT offer rapid access to vast amounts of data, beneficial for both patients and providers. While it democratizes information, limitations like potential oversimplification and reference inaccuracies were observed. The balance between LLM insights and clinical judgment, as well as continuous model refinement, is crucial. Conclusion:LLMs like ChatGPT offer significant potential in rheumatology, providing swift and broad medical insights. However, a cautious approach is vital, ensuring rigorous training and ethical application for optimal patient care and clinical practice.

ChatGPT Performance in Diagnostic Clinical Microbiology Laboratory-Oriented Case Scenarios

“…This included intra-subject variability with the creation of case scenarios based on the intentional focus to represent distinct aspects of this healthcare field. Specifically, 10 topics were selected as follows: (1) The implications of isolating a nonpathogen parasite in ova and parasite (O&P) examination, with possible indication of fecal contamination [22]; (2) The adoption of minimum inhibitory concentration (MIC) determination via broth microdilution as the standard approach for assessing colistin susceptibility [23]; (3) Resistance of methicillin-resistant Staphylococcus aureus (MRSA) to all beta-lactam antibiotics [24]; (4) The intrinsic resistance of Enterococci to clindamycin [25]; (5) The utilization of simple direct techniques (colonial morphology and germ tube testing) for identifying Candida albicans [26]; (6) The interpretation of urine culture results for the diagnosis of urinary tract infection (UTI) [27]; (7) The identification of Brucella spp. infection in blood specimens through biochemical and serotyping testing, with special consideration of the safety issues [28]; (8) Interpretation of threshold cycle (Ct) values in multiplex real-time polymerase chain reaction (PCR) testing for viral respiratory pathogens, along with associated QC issues [29]; (9) The significance of assessing sample quality before sputum culture [30]; and (10) The identification of Salmonella enterica and its serotyping based on the Kaufmann-White classification [31].…”

“…Despite the great potential and perspectives of AI-based models such as ChatGPT in healthcare, one of the challenges that needs to be investigated is the variability in performance across different tested subjects [7][8][9][10][11]. This variability in performance across various medical fields can be related to several factors.…”