• Background

  AI in medicine raises specific concerns regarding social, gender and racial/ethnicity biases. Previous studies have underscored the significant underrepresentation of women and minorities in the datasets likely used for training AI models.

• Objective

  To evaluate the demographic diversity representation in AI-depicted medical leadership compared to real-world data in three hospital organizations.

• Methods

  • Leadership roles based on UEMS, ABMS, ISCO-08 classifications.

  • Real-world data from one American, European, and African hospital

  • AI-Generated Images from closed- and open-source models with standardized prompts.

  • Image Analysis for gender, race/ethnicity, and age assessed by independent reviewers; classification consistency ensured with Chicago Face Database methodology.

  • decentralized management: Virtuous framework/LLaMBA

• Project type: Concept study

• Team status: complete

• Initiators: Sarah Saxena, MD, PhD; Mia Gisselbaek MD

• Co-investigators: Laurens Minsart MD; Mejdeddine Al Barajraj MD, Alexandre Niset, MD; Salim El Hadwe, MD, MSc; Alexandre Englebert, MD; Basak Ceyda Meko, MD, PhD; Francisco Maio Matos MD, PhD; O L. Barreto Chang MD, PhD; Sami Barrit, MD; Joana Berger-Estilita, MD, PhD

• Project status: ongoing