Public evaluations of AI-mediated healthcare when considering people with diverse access and support needs: a mixed-methods survey
Lopullinen julkaistu versio - 374.59 KB
Piispanen, J.-R., Kauttonen, J., Alamäki, A., & Rousi, R. (2026). Public evaluations of AI-mediated healthcare when considering people with diverse access and support needs: a mixed-methods survey. International Journal of Medical Informatics, 220, 106625. https://doi.org/10.1016/j.ijmedinf.2026.106625
© 2026 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
Lataukset6
Pysyvä osoite
Kuvaus
Background
Artificial intelligence (AI) is increasingly being introduced into healthcare, including patient communication, monitoring, triage, decision support, and care-related services. Accessibility is central to these developments, particularly when AI-mediated healthcare is considered in relation to people with diverse access and support needs.
Objective
This study examined how accessibility-related evaluations of AI in healthcare are patterned when respondents are prompted to consider people with diverse access and support needs, and how respondents describe risks and conditions of acceptable AI-mediated care in this context.
Methods
We conducted a mixed-methods questionnaire study with a panel sample (N = 1,151). Quantitative analyses combined Bayesian regression, confirmatory factor analysis, and correlation analysis to examine accessibility-related evaluations of healthcare AI. Qualitative analysis examined open-ended concern elaborations from a subgroup of respondents (n = 117) using codebook thematic analysis with collaborative coding.
Results
Support-oriented evaluations clustered together more strongly than they aligned with concern. In the selected regression models, technological acceptance and age were the clearest correlates: higher technological acceptance was associated with more supportive evaluations, while older age was associated with greater concern and lower endorsement of AI’s supportive potential in some models. Four qualitative themes were identified: Human Recourse and Relational Care; Communicative Accessibility and Misunderstanding; Safety, Reliability, and Override Capacity; and Privacy, Misuse, and Unfair Classification. Respondents articulated concern through questions of relational care, communicative fit, accountability, fairness, safety, and access to human support when AI systems fail or are misunderstood.
Conclusions
The findings suggest that healthcare AI implementation should address intelligibility, communicative accessibility, bounded system roles, accountability, and human recourse as central conditions for acceptable AI-mediated care.
Emojulkaisu
ISBN
ISSN
1872-8243
1386-5056
1386-5056
Aihealue
Kausijulkaisu
International journal of medical informatics|220
OKM-julkaisutyyppi
A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä (vertaisarvioitu)