Dr. Gagandeep Sukhija
Gagan is a Consultant Rheumatologist in London, UK, with postgraduate qualifications in Internal Medicine and Rheumatology and an academic background spanning India and the UK. His interests include lupus, vasculitis, myositis, rheumatology-associated interstitial lung disease, mental health and equitable access to care.
He is active in medical education, research and quality improvement, and is involved with the British Society for Rheumatology Mental Health Special Interest Group and the EMEUNET Social Media Sub-Committee.
| Poster 0515 Sunday, 08.11.26 10:30 Research Methodology Poster A (0512-0530) Presenting author: L. Bischof (Austria) Title: AI-assisted Screening: Evaluating ASReview for Systematic Literature Reviews for the Development of Rheumatology Recommendations  This study evaluated AI-assisted screening across 45 systematic review datasets. Empirically derived stopping rules maintained approximately 95% recall while reducing screening to around 25–35% of records. Five independent validation datasets achieved 95.6% sensitivity. These findings suggest a practical way to reduce the workload of evidence synthesis for rheumatology recommendations.. |
| Poster 0869| Sunday, 08.11.26 10:45 Plenary 1 (0867-0871) Presenting author: : A. Pitliya (USA) Title: Artificial Intelligence in Rheumatology: A Comparative Evaluation of Guideline Concordance and Hallucination Risk Across Platforms This study compared five AI platforms across 60 rheumatology scenarios. OpenEvidence and ChatGPT achieved the highest guideline concordance, at 91.2% and 89.4%, respectively. Hallucination rates ranged from 3.8% to 18.1%. Clinically relevant errors occurred across all platforms, highlighting the continued need for clinician oversight when using AI for decision support. |
| Poster 1878| Monday, 09.11.26 14:00 Abstracts: Measures & Measurement of Healthcare Quality (1874–1879) Presenting author: I. Hernández Pérez (Spain) Title: Can a Guideline-Grounded Artificial Intelligence Scribe Support Rheumatology Consultation Prioritization from Patient Narratives? A Feasibility Study This feasibility study assessed an AI scribe in 21 simulated rheumatology consultations. The tool structured patient histories, flagged alarm features and generated justified priority scores. It exactly matched expert prioritisation in 62% of cases, with a mean absolute error of 0.62 on a 0–10 scale. These findings suggest potential to support referral workflows, pending evaluation in real-world practice. |
| Oral 1897| Monday, 09.11.26 14:15 Abstracts: Vasculitis – Non-ANCA-Associated & Related Disorders – Clinical (1892–1897) Presenting author: C.-J. Bauer (Germany) Title: Artificial Intelligence for Ultrasound-Based Detection of Giant Cell Arteritis This study developed a deep learning model using 3,800 vascular ultrasound images from 244 individuals. The best model achieved AUCs of 0.94 for axillary and 0.87 for temporal artery image classification. AI-assisted interpretation could improve access to GCA ultrasound assessment where expertise is limited. International validation across populations, examiners and ultrasound systems is underway. |
| Poster 2243| Tuesday, 10.11.26 16:15 Systemic Lupus Erythematosus – Diagnosis, Manifestations, & Outcomes Poster C (2228–2283) Presenting author: M. J. Sandino Bermudez (USA) Title: Deep Learning Prediction of Autoantibodies From ANA Immunofluorescence Images This study trained deep learning models to predict specific autoantibodies directly from ANA immunofluorescence images, with patient-level training, validation and test splits. Discrimination was strongest for antimitochondrial, anticentromere and anti-dsDNA antibodies, with AUCs of 0.993, 0.950 and 0.902, respectively. The findings suggest potential for image-based laboratory triage, pending further validation. |
| Oral 2897| Tuesday, 10.11.26 16:15 Professional Education (2893–2898) Presenting author: E. Balczewski (USA) Title: Artificial Intelligence (AI) Curriculum for Continuing and Graduate Medical Education: A Multi-Center Pilot This multicentre pilot delivered a four-hour AI curriculum across five healthcare systems and nine training programmes, including rheumatology fellows. Knowledge scores increased from 32.8% to 57.7%, alongside higher self-reported confidence. The study supports practical AI education for clinicians, although substantially lower post-survey participation limits interpretation: 183 participants completed some pre-survey items, compared with 51 post-survey participants. |
| Oral 2965| Wednesday, 11.11.26 11:30 Abstracts: Patient Education/Community Programs: Patient Knowledge and Awareness (2965–2970) Presenting author: M. Juárez-García (Spain) Title: Artificial Intelligence, Rheumatologist, or Both? A Blinded Comparative Evaluation of Responses to Patient Questions in Rheumatology This blinded study compared rheumatologist-written, AI-generated and clinician-reviewed AI answers to 20 patient questions. Twenty patients and 12 healthcare professionals evaluated the responses. Patient ratings were similar across approaches, while clinician review improved several professional-rated domains, including readability, utility and completeness. The findings support collaborative development of patient information with expert review before use. |