Understanding acceptability of AI triage tools amongst underserved populations: lessons from the early phases of co-production with Bangladeshi communities in Birmingham, UK
Abstract
Background: Effective communication is central to safe, high-quality primary care. For Bangladeshi communities in the United Kingdom (UK), linguistic barriers continue to impede access to timely and culturally sensitive healthcare. This study describes an early phase of the co-production pathway of the tool seeking to understand the contextual acceptability of an AI-enabled translation tool designed for general practice, with functionality to capture symptoms, clinical problems, across diverse Bangladeshi dialects and provide guidance on next steps.
Methods: We conducted a series of semi-structured interviews with a sample of Bangladeshi patients from South Birmingham, UK to understand their attitudes towards using the AmarDoctor translation tool. The data were analysed using a directed content analysis to populate Sekhon's theoretical framework of acceptability.
Results: Seven participants, all native Bengali speakers, were recruited. AmarDoctor was viewed positively for supporting appointment booking, guiding appropriate next steps, and offering a safe, anonymous means of discussing sensitive concerns. Noted strengths were its ability to capture symptoms in multiple Bengali dialects and its ease of use by those with limited digital literacy. The most frequently shared concern centred on the potential for translation inaccuracies and the subsequent risks.
Conclusions: Participants expressed optimism about the role of AmarDoctor and similar AI-enabled translation tools in improving access to primary care. To gain wider acceptance, AmarDoctor must maximise the next steps of the co-production process that includes staff and commissioners to ensure translation accuracy meets the needs of all users and that credible pathways for implementation at scale are developed.
Patient or public contribution: Patients and the public have been involved from the beginning of the AmarDoctor initiative, contributing to the design and content of patient facing materials, and informing the content of our topic guide.
Date
2025-12-04
Type
Article
Subject
Language, Triage, Communication, Language, Health Inequalities
Collections
Citation
Litchfield I, Delanerolle G, Harper L, Dunning S. Understanding Acceptability of AI Triage Tools Amongst Underserved Populations: Lessons From the Early Phases of Co-Production With Bangladeshi Communities in Birmingham, UK. Health Expect. 2025 Dec;28(6):e70523. doi: 10.1111/hex.70523.
Journal / Source Title
Health Expectations
DOI
10.1111/hex.70523
PMID
41342429
Publisher
Wiley
Publisher’s URL
https://onlinelibrary.wiley.com/journal/13697625