Evaluation of artificial intelligence for online assessment of health problems: Recommended level of care, patient outcomes, cost-effectiveness, and user experience

Purpose and aimsPrevious research on triage tools based on artificial intelligence (AI) is insufficient for concluding whether the systems are accurate and safe.

An online patient-operated AI-tool for triage of health problems, i.e. assessment and recommendation of level of care, will be implemented in Swedish primary healthcare in 2023.

Our aim is to assess the performance of the AI-based triage tool compared to care-as-usual; nurses’ telephone triage at the national advisory service 1177.This is a collaboration between three universities, regions, Inera and the company providing the tool.

We will perform a randomised controlled trial to assess if there are differences in the proportion of patients advised to self-care with the AI-based triage tool compared to nurses’ triage.

We will also assess patients’ health and safety outcomes, the cost-effectiveness of the triage options, and patients’ and healthcare professionals’ experiences with the AI-based triage.AI-based triage has the potential to increase access to equitable professional healthcare advice in a cost-effective way, while enhancing patient satisfaction and reducing the need for visits in primary care.

However, there are concerns about the diagnostic accuracy and safety of the tools, and there is a lack of randomised head-to-head comparisons in real-life settings. This project will provide evidence relevant for most Swedish citizens when seeking health care advice.