Contrasting rule and machine learning based digital self triage systems in the USA

Area of Science:

• Health Informatics
• Artificial Intelligence in Healthcare
• Digital Health

Background:

• Patient smart access and self-triage systems have been developed for decades.
• Expert systems and computational models have been widely deployed.
• No large language models (LLMs) for processing self-reported patient data have been published by health systems.

Purpose of the Study:

• To review progress in patient self-triage systems.
• To analyze the current landscape of self-triage solutions offered by US health systems.
• To identify the absence of LLM-based solutions in this domain.

Main Methods:

• Systematic review of scientific literature on patient self-triage.
• Analysis of self-triage tools available on the websites of 647 US health systems.
• Assessment of the types of technologies employed in existing systems.

Main Results:

• Decades of development in patient access and self-triage systems.
• Widespread use of expert systems and computational models.
• Absence of published LLM-based self-triage solutions by health systems.
• Identification and categorization of existing self-triage solutions across 647 US health systems.

Conclusions:

• The field of patient self-triage has seen significant evolution, but lacks integration of advanced LLM technology.
• Existing systems primarily rely on older computational models and expert systems.
• Further research and development are needed to explore LLM applications for patient self-reported data processing.