Artificial intelligence (AI) is rapidly transforming healthcare, yet concerns about gender bias in AI-driven solutions remain a critical issue. A joint virtual meeting of the Women in Leadership and Harnessing Big Data Special Interest Groups brought together global experts to explore how healthcare leaders can navigate these challenges, mitigate bias, and ensure equitable and effective AI implementation. The role of AI in healthcare Sara Siegel, Global Healthcare Lead at Deloitte and chair of the meeting, opened the session by highlighting the importance of addressing gender bias in AI and the need for continued discussions on the topic. Reflecting on past discussions, including a dedicated session at the World Hospital Congress in Rio, she emphasized how crucial it is for healthcare leaders to recognize and tackle bias in AI systems. AI has the power to reduce administrative burdens, giving clinicians more time to focus on patient care. However, Toni Laracuente, Head of Analytics and Digital Health at the College of Health Information Management Executives, stressed that while AI can enhance productivity, it must not introduce or reinforce bias that negatively impacts certain populations. She emphasised that gender bias in AI is reflective of structural bias in healthcare as a whole, which frequently results in unintended consequences and sub-optimal outcomes in women patients. Identifying and mitigating bias One of the major challenges in AI in healthcare is ensuring that algorithms do not reflect or amplify existing biases. “Gender bias is when a patient receives different care or treatment based on their gender. It’s as simple as that” – Toni Laracuente Laracuente also explained that AI models are only as good as the data they are trained on. When critical data is missing, both humans and AI systems make assumptions – often leading to errors in decision-making. These unintended consequences can result in misdiagnoses, delayed or inadequate treatment, and flawed clinical decisions. Oriana Kraft, Founder of FemTechnology, emphasized the critical gap between awareness and action in addressing gender bias in healthcare AI. She highlighted findings from their recent survey of 200 physicians across six countries and five specialties, revealing that while 80% of physicians observe sex and gender differences in treatment progression and response, less than 30% feel equipped to address them in practice. “One of the most important questions to ask is whether the patient population used to build the model is representative of the population it will serve. Women’s quality of life has historically been deprioritized in medicine. So we should be wary of models that accidentally follow this precedent—optimizing for survival metrics or cost reduction while ignoring chronic pain, functional outcomes, and long-term wellbeing” – Oriana Kraft If AI is to truly improve healthcare, it must not perpetuate these blind spots but instead be designed to account for the full spectrum of patient outcomes that matter. Ensuring fair and effective AI implementation Dr Nasibeh Zanjirani Farahani, Assistant Professor of Medicine and AI Scientist at Mayo Clinic, shared insights from her work with startups developing AI solutions for healthcare. She highlighted some of the challenges she has faced when managing projects, including that data might misrepresent certain genders or that a model can amplify existing biases in the data. To address these issues, she highlighted the importance of scrutizing data sources, redefining labels and using multiple evaluation metrics. “Bias in AI rarely exists in isolation, it is often linked with race, class, and disability. Tackling one type of bias means addressing these interconnected issues as well” – Dr Nasibeh Zanjirani Farahani Professor Huiyu Zhou, Professor of Machine Learning at the University of Leicester, presented research on fairness analysis in inflammatory arthritis (IA) detection. He highlighted how AI can accelerate early disease detection of IA while also stressing the need for fairness control and ethical considerations in AI development. He emphasized that healthcare organizations must adhere to established guidelines, validating algorithms rigorously and ensuring they align with both legal and ethical standards. “AI should support, not dominate, healthcare decision-making” – Professor Huiyu Zhou Looking ahead The discussion underscored the need for collaboration between healthcare professionals and managers, AI developers, and policymakers to create AI-driven healthcare solutions that are fair, effective, and beneficial to all patients. Closing the session, Sara Siegel reinforced the importance of keeping these conversations at the forefront. “AI is a powerful tool, but only if we use it responsibly. We must ensure it serves everyone equally and advances healthcare for all.”
