The role of healthcare system distrust in shaping patients’ attitudes and beliefs of artificial intelligence (AI) use in oncology.

Abstract

1574 Background: AI offers significant potential to improve cancer care, yet little is known about patients’ attitudes and beliefs around its use and which factors influence acceptance of this new technology. Distrust in healthcare is increasingly prevalent and may hinder patient perceptions of innovations such as AI. This study aimed to evaluate the relationship between healthcare system distrust and acceptance of AI in oncology. Methods: We conducted a cross-sectional survey study with patients at an urban academic cancer center. We developed an 8-item AI Patient Acceptance scale, where patients rated their comfort with AI in different aspects of oncologic care (e.g. diagnosis, treatment planning) on a 5-point Likert scale (range 8-40, higher scores indicate greater comfort; Cronbach’s α = 0.94). The survey also included a 10-item Health Care System Distrust (HCSD) scale (range of 10-50, higher scores indicate greater distrust). Multiple linear regression was performed to evaluate the association between HCSD and AI Patient Acceptance scores, adjusted for demographic and clinical factors. Results: Of 383 patients approached, 330 (86%) participated. Among these, 49.4% were age 65 or older, 55.9% male, 68.1% non-Hispanic white, 77.4% had a college degree or more. The most common tumor types reported were prostate (34.5%) and breast (26.4%) cancer, with 70.6% currently receiving treatment. Patients were most comfortable with AI use in cancer screening (80.2% somewhat or very comfortable), and supportive care applications, such as exercise (78.2%) and diet (74.8%). They were least comfortable with AI use to assist with diagnosis (70.4%) and other clinical decision-making applications, including treatment planning (64.8%) and prognosis (61.5%). Higher levels of distrust measured by the HCSD scale were negatively associated with the AI Patient Acceptance scale scores after adjusting for co-variates ( B = -0.263, p = 0.002). Younger patients (age < 65) were more likely to report lower scores on the AI acceptance scale ( B = -1.996, p = 0.021), while sex, race/ethnicity, and education level were not associated with AI acceptance. Conclusions: Higher distrust in the healthcare system is associated with lower acceptance of AI in cancer care. As we integrate new technologies like AI into oncology, mitigating distrust in the medical community will be essential to ensure patient-centered implementation.