JMIR Publications

ABSTRACT

Background:

Large language models (LLMs), such as ChatGPT, are rapidly evolving, and their applications in healthcare are increasing. This umbrella review examines concerns related to the use of LLMs in healthcare and healthcare research.

Objective:

To map the concerns associated with the use of LLMs in healthcare and healthcare research.

Methods:

Searches were conducted in seven databases in February 2025. Screening was conducted in two stages, with independent screening by two reviewers. Included studies were quality appraised using AMSTAR-2 and GRADE. Data was extracted using a piloted form and narratively synthesised following SWiM guidelines.

Results:

The search retrieved 381 systematic reviews, of which 39 met the inclusion criteria. Three main themes emerged from the narrative synthesis. In order of most to least frequently discussed: 1) technical capability, 2) ethical, legal & societal and 3) costs. Twelve distinct populations were identified, including researchers and clinicians in various medical specialities. The included reviews were assessed to be of low quality.

Conclusions:

A wide variety of concerns are raised, which overlap and interlink, consistently affecting many populations. It is widely considered that LLMs should be used with caution, and only as a supplementary tool. It can be argued that research into the responsible use of LLMs should be an integral part of healthcare. New areas of research should focus on LLMs other than ChatGPT. A focus on addressing the needs of disadvantaged groups and thoroughly addressing ethical considerations is proposed.