JMIR Publications

ABSTRACT

Background:

The rapid advancement of generative artificial intelligence (AI) has led to the emergence of a plethora of AI scribes: digital tools designed to record, transcribe, and summarise clinical consultations. To date, the majority of publicity surrounding these technologies has focussed on clinician-held AI scribes (CHAIS), which have seen an uprise in deployment and evaluation in the UK and overseas. Herein we offer a description of a new class of tool, patient-held AI scribes, or PHAIS, with similar functionality to CHAIS in the generation of real-time or post-consultation summaries, but with the additional functionality of language or health literacy adjustments, provision of interactive recall features and further medical information signposting to support ongoing patient engagement and clinical care. To date, CHAIS have shown early potential in mitigating administrative burden, enhancing patient and clinician dynamics, and facilitating shared decision-making. However, despite increasing attention from researchers, policymakers, and industry in patient empowerment, there has been no comprehensive mapping of the evidence base for AI scribes' patient-held counterparts.

Objective:

This scoping review will explore the extent, nature, and characteristics of existing research on PHAIS in clinical consultations. Specifically, we will examine their reported effectiveness in supporting patient comprehension, the contexts in which they have been implemented and the underlying infrastructure supporting their implementation, and the ethical, practical, and regulatory challenges associated with their use.

Methods:

The review will be guided by the Arksey and O’Malley framework, refined by Levac et al, and reported according to the PRISMA-ScR checklist. Databases including Ovid MEDLINE, Embase, Scopus, CINAHL, PsycINFO, IEEE Xplore, and ACM Digital Library will be searched. Grey literature will also be screened. Eligible studies will describe the development or evaluation of patient-held AI scribes for clinical consultations. Two reviewers will independently screen and chart data using a piloted data charting pro forma. A narrative synthesis will integrate quantitative and qualitative findings.

Results:

A precursor data charting pro forma and exemplar search strategy for translation across the selected databases has been formulated. Limited preliminary searches undertaken during search strategy development has indicated a limited, though growing, evidence base surrounding PHAIS.

Conclusions:

This review will provide the first systematic mapping of research on PHAIS, identifying knowledge gaps and informing safe, equitable, and patient-centred innovation and the design of future studies surrounding the development and evaluation of patient-centred AI scribe systems.