This AI tool promises a 'second sight of eyes' to clinicians. Did patients benefit?

A study published in Nature Medicine explores the use of a large language model to assist clinical officers in primary care settings. The tool, called AI Consult, was tested across 16 clinics in Kenya operated by Penda Health. Nearly 10,000 patient encounters were included in the trial. In this setup, the AI reviewed electronic notes in the background and flagged potential issues using a simple color-coded system.

Clinicians received prompts in real time. A green indicator meant the notes were clear, while yellow suggested a minor oversight and red signaled a critical concern. Registered clinical officer Vyonne Njeri reported that the software helped her identify conditions she might have otherwise overlooked during high-volume shifts. An independent panel of Kenyan family physicians later reviewed the notes, concluding that the AI provided a valuable second pair of eyes that improved both diagnostic accuracy and treatment plans.

The cost of the implementation is notable at four cents per patient. Despite these operational improvements, the researchers did not find statistical evidence of significantly better patient outcomes, such as reduced mortality or faster recovery. Researchers attributed this to the fact that serious treatment failures are relatively rare in primary care, making it difficult to measure impact without a much larger sample size.

Experts acknowledge that this is one of the first trials to move beyond simulations and test AI as a direct support for healthcare staff. While the study leaves the question of long-term patient benefit open, the implementation suggests that AI can serve as a practical feedback mechanism for health workers in lower-resource settings where specialist support is limited. Future work aims to determine if the tool can improve patient access or reduce the administrative burden on providers.