Listening to Patients’ Voices on the Use of AI in Health Care: Cross-Sectional Study
Bibliographic record
Abstract
Background: Artificial intelligence (AI) holds great promise in transforming health care delivery. However, successful implementation of AI projects in health care depends on patients' acceptance and trust. There is limited empirical research examining public perceptions, particularly the use of personal health data in AI applications in health care. Objective: This study examined public knowledge and comfort levels with AI use in health care, including use of personal health data with and without consent, and assessed how sociodemographic factors, digital literacy, and health conditions influence these perceptions. Methods: We analyzed data from 6904 Canadian adults who participated in the 2023 Canadian Digital Health Survey. AI-related knowledge and comfort levels were measured using ordinal scales. Sociodemographic characteristics, digital health literacy, and self-reported chronic health conditions were included as predictors. Ordinal logistic regression models were used to assess associations between these factors and AI-related attitudes. Results: A majority of 2919 (42.3%) reported moderate knowledge of AI; only 7.8% (542) described themselves as very knowledgeable. Overall, 44.6% were comfortable with AI use in health care, increasing to 64.7% when personal health data were used with consent but decreasing when used without consent (52.6% uncomfortable). Respondents were most comfortable with AI use for epidemic tracking and workflow management and less for clinical tasks. Fully weighted ordinal logistic regression models indicated that men (odds ratio [OR]=1.57, P<.001), noncitizens (OR=1.71, P<.001), higher-income respondents (OR=1.29, P<.001), those with graduate education (OR=1.43, P<.001), higher digital health literacy (OR=1.08, P<.001), and more chronic conditions (OR=1.08, P<.001) exhibited greater odds of reporting higher AI knowledge. For comfort with AI use in health care, those aged 65+ years (OR=1.47, P<.001), men (OR=1.50, P<.001), noncitizens (OR=1.49, P<.001), higher-income respondents (OR=1.21, P<.001), and those with higher digital health literacy (OR=1.06, P<.001) or more chronic conditions (OR=1.04, P=.04) exhibited greater comfort. Lower-income (OR=0.87, P=.03) and White respondents (OR=0.77, P<.001) reported lower comfort levels. For comfort with using personal health data in AI with consent, adults aged 35-54 years (OR=0.72, P<.001) were less comfortable than those aged 16-24 years. Men (OR=1.39, P<.001), higher-income respondents (OR=1.16, P=.01), and those with higher digital health literacy (OR=1.05, P<.001) or more chronic conditions (OR=1.07, P<.001) showed greater comfort; White (OR=0.78, P<.001), other racial groups (OR=0.77, P=.03), and lower-income respondents were less comfortable (OR=0.83, P=.01). For comfort with using personal health data in AI without consent, men (OR=1.56, P<.001), noncitizens (OR=1.28, P=.03), and those with higher digital health literacy (OR=1.04, P<.001) exhibited greater comfort. Lower-income respondents (OR=0.86, P=.02), adults aged 35-54 years (OR=0.73, P<.001) or 55-64 years (OR=0.77, P=.01), and White (OR=0.69, P<.001) and Black or African-origin (OR=0.71, P=.02) respondents reported lower comfort levels. Conclusions: The findings point to enhancing transparent policies, digital literacy, and ethical data governance as key to increasing public trust in AI-driven health care.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.012 |
| Meta-epidemiology (narrow) | 0.000 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.002 |
| Science and technology studies | 0.002 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.004 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".