JMIR Publications

ABSTRACT

Background:

As artificial intelligence (AI) becomes increasingly embedded in healthcare, understanding patient priorities for its use is critical.

Objective:

This study aimed to identify the priorities, concerns, and expectations of patients and caregivers regarding the use of AI in healthcare.

Methods:

The project was conducted in iterative phases, with each phase informing the next. The first phase consisted of an environmental scan of frameworks and relevant literature to identify key emerging themes and issues related to implementing AI in healthcare, with a particular focus on patient safeguards. In the subsequent phases, using a multi-methods design, we administered a Canada-wide bilingual survey and conducted semi-structured interviews with patients and caregivers to further explore their perspectives on the use of AI in healthcare. Descriptive statistics were used to analyze rating-scale survey items, while deductive content analysis was applied to open-ended survey responses. Regression analyses examined associations between survey responses and demographic characteristics. Interview data were analyzed using inductive thematic analysis. Findings were synthesized at each stage and shared with a national advisory committee of patient partners and subject-matter experts through co-creation sessions.

Results:

The environmental scan identified recurring themes, including privacy, transparency, consent, accountability, and human oversight, while highlighting the limited incorporation of patient perspectives. The survey included 638 participants representing all age categories from 18 to ≥65 years (49.8% women, 69.3% from Ontario, 64.6% White). Survey respondents endorsed the importance of human oversight and compassion (48%), patient/caregiver involvement and transparency (45%), confidence in AI design and governance (45%), equity and fairness (45%), consent (44%), and protection of personal health information (44%). Fewer participants reported having trust in AI-enabled healthcare (5%) or comfort and understanding regarding AI use in healthcare (10%). Regression analyses identified significant demographic differences in perceptions of AI-enabled healthcare. Older adults, women, and non-White participants placed greater importance on privacy, equity, and patient involvement, while participants with disabilities and those identifying as non-heterosexual reported lower levels of trust in AI-enabled healthcare (all P<.05). Qualitative interviews (n=13) revealed cautious optimism toward AI-enabled healthcare. Trust emerged as a central cross-cutting theme, with acceptance of AI dependent on accountability, transparency, and equitable access. Patients and caregivers emphasized that trust can be fostered through healthcare system practices and behaviours that ensure safeguards related to human oversight, data protection, equity and inclusion, and knowledge sharing.

Conclusions:

Integrating patient and caregiver perspectives into AI-enabled healthcare is essential to ensure that policies, safeguards, care practices and implementation strategies are grounded in their needs, values, and expectations. These findings support patient-centered AI practices and inform the co-creation of a charter for the accountable use of AI in healthcare. Clinical Trial: Not applicable.