AI systems capable of clinical reasoning and dialogue have the potential to dramatically expand access to medical expertise and care, while giving physicians the opportunity to spend time with their patients where it really matters. However, developing these technologies responsibly requires a rigorous, evidence-based approach. Over the past few years, our teams have explored the “art of the possible” through research systems that demonstrate physician-level capabilities in simulated settings. While we have begun testing the safety and feasibility of these systems in clinical settings, additional rigor and scale are required to move to the next phase of evaluation of these systems. This involves studying the usefulness and impact of AI in virtual care involving more patients, in different geographic areas and conditions, and with controlled comparisons.
Today, we’re announcing an important step in that ongoing research journey: in partnership health includedA leading US healthcare provider, we will initiate a prospective consensual nationwide randomized study to assess AI in a real-world virtual care setting, pending Institutional Review Board (IRB) approval. This new research will build on our foundational research on the use of AI for diagnosis and management reasoning, personalized health insights, and navigating health information.
This work represents an important development in our research. Preliminary studies published Nature First evaluated our AI system clinical reasoning abilityWhich also includes its helpful effect for physicians. Then we compared the systems Conversational Diagnostic Capabilities For primary care physicians in simulated settings with patient actors. In addition to understanding capabilities, we also explored a physician-centric paradigm with asynchronous supervision of AI. Our initial step toward testing conversational AI in real-world clinical settings was a single-center feasibility study in partnership with Beth Israel Deaconess Medical Center. The goal of the study was to demonstrate the safety of the system outcome measures Such as the number of interruptions by a security supervisor in response to security concerns. We have seen strong signals of safety in this early study and look forward to sharing the results when completed.
