MA

VA Plans to Exploit AI to Help Veterans

Published: August 15, 2019

Health ITVA

In July, VA appointed Dr. Gil Alterovitz as its first-ever director of Artificial Intelligence (AI). In his role, he is tasked with facilitating the rollout of AI across VA to benefit veterans.

VA is already using AI to combat suicide and pinpoint cancer treatments, but it aims to accelerate its implementation and its use to further improve veteran health care and the veteran experience.

As part of the veteran REACH program, VA uses AI to scan medical records and look for signs of suicide risk among veterans. VA in partnership with IBM Watson Health is also using AI in its precision oncology care to determine the best cancer treatments for veteran patients. The initiative to date has supported 2,700 veterans with cancer and has helped VA physicians identify precision treatments for nearly 30 times more cancer patients than could be previously served.

Alterovitz is based in VA’s Office of Research and Development and is also a professor at Harvard Medical School. He was one of the principal authors of the White House Office of Science Technology and Policy’s National Artificial Intelligence Research and Development Strategic Plan, updated in June.

According to a recent VA article about VA’s AI plans, Alterovitz commented that “much of his work will be in the realm of big data.” One area of concentration will be the Million Veteran Program (MVP) which already contains genomic and health data from more than 750,000 consenting veteran volunteers. Teaching computers to sift through this data to find answers to veteran health problems will be a high priority.

Additionally Alterovitz says, “Applying AI tools to MVP and other large datasets demands the right infrastructure. VA has to build systems that can handle increasingly large amounts of data and make it easier to share those assets—in a secure, confidential manner—with outside experts who can help VA, and Veterans, get the biggest return on their research investment.”

Another area VA wants to use AI is improving customer service and reducing wait times for care. Veterans and their caregivers require well-trained VA agents to handle the breadth and depth of their questions.  This has been a challenge for VA, especially during peak days and times. Because of this, veterans and caregivers often do not receive immediate assistance. VA intends to use AI to help solve this problem.  In the fall of 2018, the Veterans Health Administration began looking for an AI as a service (AIaaS) solution. VA wants features such as natural language processing (NLP) so that callers can hold discussions with the AI technology, similar to a chatbot.

VA also conducted a study with DeepMind Health focused on developing an AI system that could forecast kidney failure or Acute Kidney Injury (AKI).  They published their results last month showing that their “AI model enabled researchers to identify more than 90% of the most severe AKI cases 48 hours sooner than usual care.”

Other uses of AI that hold promise include using natural language processing (NLP) to query clinical notes and using AI to read CT scans.

Alterovitz is currently working on an AI data ecosystem, which involves trying to make federal data more suitable for AI, “AI-able.”  These efforts would make data available to other agencies, academia, industry, etc. allowing them to use it with AI which in turn creates more data, and this becomes an ecosystem.  They've designed a framework that includes voluntary incentivization metrics that incentivize both sides to participate in the ecosystem.  He’s trying to create synergy. 

Another major focus for Alterovitz in developing this ecosystem is the VA TOP (The Opportunity Project) Health AI Sprint. The sprint is modeled after the HHS Health sprint that took place last fall.  But the VA sprint is focused on how to go about creating data sets that are usable for AI.  VA wants to be able to release different data sets over time.  They are hoping to develop a broker platform.

In an interview with Government Matters, Alterovitz admitted there are barriers in working with other agencies and organizations around data. Challenges include processes, people, and policy. In fact last year, VA canceled an agreement with startup Flow Health for an AI pilot, after realizing the agreement would violate current VA policy and regulations around veteran data protection. Alterovitz stated that each barrier will need to be addressed and that each agency has strengths and challenges in the different areas.