Artificial intelligence (AI) is rapidly transforming numerous fields, and healthcare is no exception. AI-powered tools are emerging with the promise of enhancing diagnostic accuracy and treatment options. However, the adoption of these technologies hinges significantly on patient acceptance. A recent study conducted by researchers at the University of Arizona Health Sciences delves into patient attitudes towards AI in medical diagnosis, revealing a nuanced perspective on the human versus machine dynamic in healthcare.
Published in PLOS Digital Health, the study titled “Diverse Patients Attitudes Towards Artificial Intelligence (AI) in Diagnosis,” led by Dr. Marvin J. Slepian, a Regents Professor at the UArizona College of Medicine – Tucson, and Christopher Robertson from Boston University, explored whether patients would favor an AI system or a human doctor for diagnosis and treatment. The findings indicate that a slight majority still lean towards human doctors, highlighting the critical role of trust and the human element in patient care as AI integrates into medical practices.
The research involved placing participants in hypothetical patient scenarios, asking them to choose between AI and a human physician for their medical needs. The study was conducted in two phases. The first phase involved structured interviews to gauge initial reactions to AI technologies in healthcare. The second phase broadened the scope, surveying 2,472 participants from diverse backgrounds to quantify preferences and identify influencing factors.
The results showed that approximately 52% of participants expressed a preference for a human doctor over an AI for diagnosis and treatment, while about 47% were inclined towards AI. This near even split underscores a significant point: patient acceptance of AI in healthcare is not yet a given. Interestingly, the study also found that when participants were informed that their primary care physicians considered AI to be a valuable and helpful tool, their acceptance of AI increased. This highlights the powerful influence of the physician-patient relationship in shaping attitudes towards AI in medicine.
Disease severity, surprisingly, did not significantly alter patient trust in AI. Whether the scenario involved leukemia or sleep apnea, the preference ratio between human doctors and AI remained relatively consistent. However, demographic factors did play a role. Black participants showed less inclination towards AI compared to white participants, while Native Americans were more likely to choose AI. Age and political ideology also correlated with preferences, with older and politically conservative participants showing more resistance to AI in diagnostic roles. The importance of religion was another factor associated with lower AI preference.
These disparities across racial, ethnic, and social groups emphasize the need for tailored communication and culturally sensitive approaches when introducing AI into healthcare. Building trust and ensuring equitable access to AI benefits will require careful consideration of diverse patient populations and their unique perspectives.
Dr. Slepian emphasizes the crucial takeaway: “While many patients appear resistant to the use of AI, accuracy of information, nudges and a listening patient experience may help increase acceptance.” The study suggests that the “human touch” remains vital. To effectively integrate AI into clinical practice and gain patient trust, healthcare providers must focus on clear communication, demonstrating the accuracy and reliability of AI tools, and ensuring that the patient experience remains empathetic and patient-centered. Future research should focus on optimizing physician integration of AI and understanding patient decision-making processes in this evolving landscape of digital health.
The research team also included Andrew Woods, JD, Jess Findley, JD, PhD, and Cayley Balser, JD from UArizona James E. Rogers College of Law, and Kelly Bergstrand, PhD, from the University of Texas at Arlington. This study was supported in part by the National Institutes of Health, award no. 3R25HL126140-05S1.
