While we are basking in the sunshine of LLMs being used for generic and mundane work, there are some who are putting into specialised usage, to create one of its kind use cases, to provide the real AI transformation the world intends to achieve. Such use cases can become products in themselves and over a few interactions and testing, it can be taken to a different level – a failsafe one. Suffice to say, it’s not only the AI tools that learn, the learning is on the expert’s side as well, in the present case the doctors too. Penda Health and OpenAI clinical copilot showed promising results, as per a real world implementation in Nairobi, Kenya.
What is AI Consult? It was developed by Penda Health and powered by OpenAI’s GPT-4o. The AI tool has been integrated into electronic health records, running silently during patient visits. The system gets activated only when potential errors are detected, ensuring that the clinicians remain in full control. The tool uses a color-coded alerts system: green means no concerns, yellow connotes moderate concerns and red meaning safety-critical issues requiring attention. What is important to note that currently, adoption towards solving real-world patient and clinician challenges remains slow.
An OpenAI-backed study conducted with Kenya-based Penda Health found that deploying an AI “clinical copilot” inside electronic health records significantly improved the accuracy of frontline medical care. The study was conducted of 39,849 patient visits across 15 clinics in Nairobi. The key findings are startling. The survey showed 31% reduction in diagnostic errors when red alerts were being triggered. Thirty two percent improvement is found in history-taking accuracy and 10% better in investigation ordering.
The clinicians over a period of time triggered fewer red alerts over time, indicating learning and improved decision making. Around 75% of clinicians reported a “substantial” improvement in the care quality. ” What contributed to this outcome? It was a fusion of a capable AI model, clinically-aligned implementation co-developed with users, and active deployment efforts to help clinicians understand and utilize the tool effectively. This marks one of the first large-scale, real world validations of AI in frontline healthcare.
AI CAN BRIDGE HEALTH CARE GAPS SEAMLESSLY.
Sanjay Sahay
Have a nice evening.