Improving AI Triage Accuracy in GP Practices Using the Patchs System

What this Trial Is About

Researchers are evaluating the use of Artificial Intelligence AI in triaging patients who contact General Practitioner GP practices in England. The study focuses on improving the accuracy and fairness of an AI triage system called Patchs, which assists in determining the urgency and type of care patients need. The research aims to identify challenges in current triage methods and develop enhanced AI models that can better replicate clinicians decisions across various regions and patient backgrounds. The study collects anonymised data from GP practices using the Patchs system, both with AI triage enabled and disabled. It includes four main phases analyzing current triage practices without AI, developing improved AI models using historic data, testing these models in the background without influencing care, and finally implementing the updated AI in real clinical settings. The AI triage suggests urgency levels and care pathways, but final decisions remain with human staff and patients. Participants in this study are GP practices using the Patchs system. Researchers analyze data from online consultations, comparing AI triage decisions with those made by clinical staff to assess accuracy and fairness. The main outcome measured is the F1 score, indicating how well the AI triage matches clinical decisions, tracked over the anticipated four-year study period. The study also monitors how often staff and patients agree with the AI suggestions to ensure safety and fairness.

Improving the Accuracy of Artificial Intelligence Triage in Primary Care