AI in Healthcare Unleashed: Responsible and Ethical Implementation of Large Language Model Chatbots in Clinical Workflows and Patient Care

Research problem and specific questionsFollowing the release of OpenAI’s ChatGPT in November 2022, this artificial intelligence (AI) has never been out of the headlines.

ChatGPT is an example of a Large Language Model (LLM) that can generate content based on training on large data sets & algorithms. LLMs can summarize complex data and write responses in a requested conversational style or literacy level.

This technology is particularly relevant in a healthcare context especially the era of patient online record access (ORA), an innovation Sweden has been at the forefront of worldwide & where overburdened healthcare professionals (HCPs) desire assistance with documentation. However, LLMs also embed errors & biases.

The purpose of this timely interdisciplinary project is to investigate whether LLMs might be used to alleviate documentation burdens & assist patients.

The project has 3 aims:AIM 1: To investigate Swedish HCPs’ experiences & understanding of LLMs in documentation in clinical practiceAIM 2: To investigate patients’ experiences & understanding of LLMs in interpreting their health dataAIM 3: To investigate how healthcare might ethically & safely adopt these tools to benefit HCPs & patientsData and method: The project will survey the experiences & opinions of participants using mixed-methods online surveys for AIM1&2.

AIM1 will leverage a nationwide survey of HCPs that is already being conducted; AIM2 will use a nationwide survey through the national patient portal 1177.se, managed by Inera AB, to solicit patients’ opinions. AIM3 will triangulate results and involve ethical analysis.

AIM3 will also use an online 3-round Delphi methodology to poll experts for recommendations on the ethical, safe adoption of LLMs in health documentation.Societal relevance and utilisation: In December 2023 the EU proposed the world’s first rules for LLM chatbots with makers of generative AI stipulating consumers must be fully informed that they are engaging with AI.

A fundamental problem is we do not know when or how HCPs & patients are using LLMs nor how LLMs can be ethically, safely & effectively integrated into healthcare.Plan for project realisation: The project will be divided into 5 work packages designed to be efficiently & effectively managed by a complement of 7 interdisciplinary research personnel plus 1 Ph.D. student & supported by an international Patient Advisory Group & an LLM Advisory Group.