Decision making in oncology: Unleashing the power of artificial intelligence—Beyond the question of feasibility.

Abstract

9021 Background: This study examines how consulting artificial intelligence (Chat GPT, Bard) or electronic tools (Apps, Journal or Guidelines) affects the decision-making process in Oncology Fellows. Methods: In the OSCE IOHM-USAL 2023, Oncology Fellows (2022 and 2023 cohorts) individually answered stations I and II, both spontaneously and after AI consultation. Group responses for stations III and IV were also gathered. Sign tests compared pre- and post-consultation responses in stations I and II while the impact of AI in stations III and IV was individually assessed on a five-point scale (0= Null and 5= Extremely useful). Results: 15 Oncology Fellows participated in OSCE 2023- 14 Females. Median age: 30 (r=28-36). A) Stations I - II Cases: Case1: 70 y. Non smoker. PS 0. Stage IV NSCLC. PS 0. EGFR mutated E746 _A750. PDL1 90%. Case 2: 52 y. Postmenopausal Stage II Luminal B Her2 negative Breast cancer (pT2, pN1, M0). Case 3: 35 yo. Advanced colon cancer. Stage IV. BRAF p(V600E) c.1799 A>C. No Micorsatellite Instability. B) Station III (Ethics Committee), the following cases were addressed: 4. “BRCA1: Insured, not excluded”. 5. “Smoke vs. oath? Tough choice” 6. “Follow Lynch path, or HMO exit”. In all three cases, 3/15 students reported finding AI useful in the discussion. C) Station IV (Tumor board), the following cases were addressed: 7. “Postmenopausal NOS IBC (pT1c pNmic Luminal B Her2 negative)”. 8. “Oncology emergency in a 22 y male patient testicular tumor pending pathology”. In Case 7, 3/15 students reported finding AI useful in the discussion. In Case 8, 4/15 students reported finding AI useful. Conclusions: 1. A study of 15 oncology fellows found no appreciable influence of AI on clinical decision-making. 2. Analysis of eight cases (individual/collective) identified anchor value, optimistic denial, and confirmation bias in reasoning and processing. 3. Fellow training in critical AI analysis and outcome evaluation is imperative. [Table: see text]