MetroHealth to Test Conversational AI With Cancer Patients
Cleveland based academic medical center MetroHealth is partnering with Dallas-based Pieces Technologies Inc. to study a conversational AI agent designed to allow cancer patients to ask questions in real-time about any aspect of their care, while simultaneously ascertaining information about social determinants of health (SDOH).
With a $2 million Small Business Innovation Research (SBIR) contract from the National Cancer Institute (NCI) within the National Institutes of Health (NIH), Pieces and MetroHealth will deploy and study how PiecesChat converses with patients.
“One barrier to advancing cancer care is the material challenge of getting real, actionable data from patients. Having an AI tool that's able to converse directly with patients, spend the necessary time with them and demonstrate empathy will not only help clinicians obtain SDOH
information, but also help categorize findings into actionable data,” said R. Douglas Bruce, M.D., senior vice president, chief clinical integration officer for MetroHealth System, in a statement. “MetroHealth has food for medicine programs, transportation and a number of other resources that we can offer if we truly understand what patients need in a nuanced way. We are eager to collaborate with Pieces to leverage the unique capabilities of AI to help clinicians move one step closer to resolving the SDOH disparities that our patients face.”
A key innovation of the project involves extending the patent-pending Pieces SafeRead platform to support conversational AI. The company said its SafeRead system employs highly-tuned adversarial AI alongside human-in-the-loop (HITL) oversight to minimize errors of communication.
This project will be one of the first rigorous research demonstrations of HITL-based conversational AI in the healthcare domain, the organizations said.
“The technology being studied has potentially far-reaching implications in multiple domains, including cancer care, SDOH management and patient empowerment. For the first time patients will have broad ability to ask any question or detail about their care to a highly supervised AI,” said Ruben Amarasingham, M.D., chief executive officer of Pieces, in a statement.
Pieces said it is leveraging both passive and active screening to capture SDOH information during this clinical trial. Working with Amazon Web Services (AWS), Pieces Technologies will use Amazon Bedrock, a fully managed service that offers a choice of high-performing foundation models (FMs) to responsibly integrate generative AI capabilities.
The study will measure utilization, effectiveness, reliability, accuracy, empathy and patient perceptions of the AI tool. The study also will help expand Pieces’ hallucination risk classification framework for use in conversational AI, which the NIH evaluation panels identified as an opportunity to advance AI safety protocols in clinical care delivery.
