JMIR Publications

ABSTRACT

Background:

The ethical risks of medical artificial intelligence (AI) have become a global focus of attention. Countries around the world have successively introduced relevant ethical policies. However, understanding of the ethical risks of medical AI remains fragmented, making it difficult to comprehensively present the overall framework of medical AI ethical risks and their internal connections. Moreover, physicians’ awareness of these ethical risks is still unclear.

Objective:

Based on the Chinese context, this study constructs a framework for identifying ethical risks of medical AI, conducts a survey on physicians' cognition of AI ethical risks, and provides theoretical support for AI risk governance.

Methods:

In the first phase, we conducted semi-structured interviews with 36 experts. Grounded theory was applied to construct a framework for the ethical risks of medical AI. NVivo 11 software was utilized for coding textual data. In the second phase, based on the theoretical framework, a questionnaire was designed to survey physicians' perceptions of ethical risks of medical AI . Convenience sampling was used to survey 600 Chinese physicians. After verifying the reliability and validity of the questionnaire, descriptive statistics and multiple linear regression analysis were employed to examine physicians' perceptions of ethical risks and their influencing factors.

Results:

his study constructs a multidimensional framework for ethical risks of medical AI, including 5 primary categories--physical risks, psychological risks, data and privacy risks, social risks, and economic and sustainability risks --and 15 subcategories. It was found that physicians were most concerned about data and privacy, liability attribution, and physician-patient relationships, while having the lowest rate of agreement with economic and sustainability risks. Differences in physicians' risk cognition were mainly influenced by multiple factors, such as training experience, familiarity with AI, and whether their hospitals had established standardized ethical review procedures for medical AI .

Conclusions:

This study not only enriches the theoretical framework of ethical risks in medical AI but also provides empirical evidence for their targeted governance. It is recommended that future efforts should focus on enhancing the ethical training of medical professionals, improve the ethical review mechanisms for AI in healthcare institutions, and clarify the division of liabilities and accountability. These measures will promote the robust development of medical AI within an ethically compliant framework. Clinical Trial: NA,The study was approved by the Clinical Research Ethics Committee of China-Japan Friendship Hospital (Approval No.: [2024-KY-254]). All participants were fully informed, and the data were stored anonymously , encrypted, and strictly limited to academic use.