By Pragati Verma, Contributor
Veteran Affairs (VA) has received a lot of scrutiny for its inefficiency in providing timely medical treatment, with stories coming to light of veterans waiting months to receive care. And while there are no immediate solutions to correct the agency’s perennial wait time problem, the VA is turning to artificial intelligence (AI) for at least a few answers.
“AI is very important to us in a lot of different ways, [including] healthcare claims and processing,” explained Dr. Paul Tibbits, program executive officer for the VA’s Financial Management Business Transformation Special Program Office in a media interview.“We envision that [these technologies] will, or in fact already are, improving services to veterans.”
Waiting for Care
The VA’s inefficiencies first came to light in 2014, when managers at a Phoenix VA facility revealed an elaborate scheme to hide the fact that several veterans waited many months to see a doctor. In 2015, the VA Office of Inspector General reported a web of complications in the VA’s management of healthcare enrollment data, including inadequate procedures to oversee records, software glitches, and inconsistencies in identifying veterans who had died waiting for medical treatment.
Now, the latest U.S. Government Accountability Office report finds that the Veterans Choice Program — set up to reduce wait times by allowing veterans to go to a private sector doctor at the VA’s expense — has not helped either. The report states that in many cases veterans could still face wait times of up to 70 days to receive care. “Delays in care have been shown to negatively affect patients’ morbidity, mortality, and quality of life,” report researchers wrote.
According to a statement issued by the VA, it has been a challenge to maintain a large staff of well-trained agents to handle the depth and breadth of questions veterans and caregivers face, especially during peak days and times. As a result, veterans and caregivers are unable to receive immediate assistance because agents are actively assisting other customers.
For Tibbits, what is clear is that AI can improve the VA’s wait-time issue, as well as diagnostic and other therapeutic accuracies, making fewer mistakes and picking the right treatment for the right illness. The process is already underway when it comes to phone service and addressing claims.
The VA is currently looking for an AI as a service (AIaaS) solution — off-the-shelf and cloud-based — with features such as natural language processing (NLP) so that people can make requests and hold discussions with the AI technology (much like a chatbot). Leaders at the VA also want the AIaaS system to evaluate the words of users to detect their emotional state and respond accordingly. Once the AI system is trained and put into production, the VA said, it can use machine learning to expand and improve upon its own capabilities over time.
AI Emerges in Veteran Healthcare
Yet the AIaaS solution won’t be the agency’s first brush with AI. The VA has already partnered with London-based Alphabet subsidiary DeepMind to explore how machine learning can help predict patient deteriorations. Together, the VA and DeepMind will analyze patterns from over 700,000 historical, depersonalized medical records to determine if machine learning algorithms can accurately identify the risk factors of patient deteriorations and predict their onset.
Dominic King, clinical lead at DeepMind Health, emphasized that the data used in the partnership research has been stripped of any identity before DeepMind receives it — an important feature for the VA after a recent privacy concern.
Last year, the VA terminated an agreement with the San Francisco-based startup Flow Health after realizing the agreement would violate current VA policy and regulations, as well as its commitment to protecting veterans’ data. The research would have used genomic data from over 500,000 volunteer participants from the Million Veteran Program.
The Prevention Game
In partnership with DeepMind, the VA first plans to focus its work on identifying the most common signs of acute kidney injury (AKI), a problem that can lead to dialysis or death, but is preventable if detected early.
“Not only is the onset of AKI sudden and often asymptomatic, but the risk factors associated with it are commonplace throughout hospitals,” King explained in a recent blog post. “AKI can also strike people of any age and frequently occurs following routine procedures and operations like a hip replacement.”
The goal, according to King, is to find ways to improve the algorithms currently used to detect AKI and allow doctors and nurses to intervene sooner. Eventually, the VA plans to apply similar approaches to other signs of patient deterioration.
As the VA explores more AI-powered approaches, leaders in the organization expect the growing technology will provide improved care and shorter wait times for many more patients — with fewer people developing serious infections and conditions and, as the VA press release noted, “ultimately saving lives.”