Creating the space for mission-driven AI
Connecting state and local government leaders
A data ecosystem using patient information, diagnoses and lab results pulled from multiple databases in government and industry matches veterans with relevant clinical trials.
Gil Alterovitz, the Department of Veterans Affairs’ director of artificial intelligence, wanted to match veterans with clinical trials that would address their health concerns, so he decided to create a data ecosystem using patient information, diagnoses and lab results pulled from multiple databases in government and industry.
AI-Able Data Ecosystem Pilot, Clinical Trial Matching System
Department of Veterans Affairs, Department of Health and Human Services, General Services Administration
Click here for all the 2019 Public Sector Innovation winners
“This work would involve not only open datasets, but also small, closed ones with standardized use agreements for AI in the right format so tools could be scaled quickly later,” Alterovitz said.
He added that “a voluntary incentivization framework” was designed to encourage participation and create an ecosystem of usable and reusable components.
A pilot program involving government and nongovernment organizations gathered information from four databases via application programming interfaces. It used natural language processing to extract information and eligibility criteria from unstructured descriptions of clinical trials.
Users can be authenticated by VA, the Centers for Medicare and Medicaid Services or both to obtain access to their personal data and then have that data used to query the National Cancer Institute’s database to search for relevant clinical trials.
A working prototype for that matching process was developed this past summer, and it is an important goal. But Alterovitz and his team have even bigger ambitions for the data ecosystem. By working through the data architecture, collaboration incentives and privacy controls, they are building a framework that could support a wide range of health initiatives.
The goal is for the system to become more of an “honest broker and testing platform [with] standardized agreements for how to test these protocols,” he said.