Federal agencies are working with vendors to ensure they have the data, governance and culture in place to deploy artificial intelligence tools that improve healthcare outcomes.
Artificial intelligence is rapidly innovating the healthcare sector, with tools that increase efficiency, improve public health outcomes and inform data-driven decisions. Health organizations are able to leverage AI-powered solutions for early disease detection, precision treatment planning, drug discovery and more – while paving a path to more scalable and cost-effective care.
These tools have transformative implications for the healthcare sector from the lab to the back office. At a recent FedInsider webinar, thought leaders from government and industry discussed how federal agencies can adopt AI tools to innovate while maintaining proper security.
Featured Experts:
AI’s Potential in Federal Healthcare
Federal health agencies are already exploring the benefits that adopting AI and applying it to their data can have on health outcomes. Chris Kinsinger, Assistant Director of Catalytic Data Resources for the National Institutes of Health Common Fund Office of Strategic Coordination, said NIH is looking at addressing challenges in biomedical research with AI.
One of the projects, Voice as a Biomarker, develops data sets of human voices to help models detect five categories of health issues, like respiratory diseases, neurological disorders, mood disorders, pediatric speech/developmental issues and voice disorders.
“Using a biomarker has tremendous potential,” Kinsinger said. “You can transmit voice data from anywhere with cell service quite easily. So, talking about underserved populations in rural areas, that’s a real opportunity.”
NIH is also exploring the ability to apply AI models to data collected from intensive care units at 14 hospitals across the country to predict adverse events like cardiac arrest, strokes or sepsis. “This project aims to collect data from over 100,000 ICU records and make them available for training such models,” Kinsinger said. “We are going beyond just the wires connected to the patients in the beds, but also looking at medical history, neurological imaging and electronic health records.”
These projects would include the general principles that apply to NIH data collection, including ethics, consent, data governance, metadata standards and so on.
Dr. Joseph Ronzio, Deputy Chief Health Technology Officer for the Veterans Health Administration, said that one of the key AI-assisted projects that the VA is working on is called the Medical Smart Home Project. It explores how technology can transform traditional living spaces into proactive health-supporting environments for elderly or disabled veterans.
“The VA spends millions of dollars every year for smart home technologies under a myriad of different funding sources, so we are trying to find a consolidated way to truly see if these devices can help veterans and support their caregivers,” Ronzio said. The VA developed a white paper about smart homes and is working with vendors to try to advance this technology to help disabled veterans while also supporting their caregivers. They are also trying to reduce the cost of care, which can be extremely expensive. While still in progress, part of this work includes determining who handles the veterans’ data, how device data is shared with healthcare providers and refining the algorithm to best make use of AI.
The Applicability of AI
Dr. Roger Klein, Director of the Agency for Healthcare Research and Quality (AHRQ), said AI is already touching all aspects of the healthcare system, and that “it’s a revolution in technology.” Some ways in which AI is being used throughout the healthcare system include streamlining mundane tasks like administrative scheduling and workflows, using large language models to empower note taking and transcriptions, and using predictive analytics to inform clinical decisions.
Part of Klein’s other critical work that touches on AI at AHRQ includes finding places where real-time data can be used to improve diagnosis and treatment, implementing technology that can enhance medication safety, and using AI to leverage technology that advances care delivery.
For instance, “Can you go from one place to the next and give your providers your medical records in a timely fashion?” Klein asked. “Leveraging data and technology can really help healthcare systems improve the quality of care that they deliver.”
According to the speakers at the event, there are several key places where AI can be smartly applied within government. They include things like using predictive analytics to detect patterns that spot fraud or using generative AI like ChatGPT to streamline operations. These apply to the health sector too, and for Tyler Johnson, Principal Sales Engineer at Primer, he’s focused on natural language processing and uncovering the intelligence inside unstructured datasets.
A lot of enterprise data in government is unstructured, Johnson said, and traditional AI utilization today relies on structured data. But putting unstructured data like patient care, doctors’ notes and patient feedback together in a meaningful way beside structured data like electronic health care records can help generate evidence-based decisions for healthcare providers.
“With AI, you’re spending less time on manual tasks and more on making the actual decisions that only you, as a domain expert, can make,” Johnson said. “As a primary care provider, [you are] spending more time on a holistic health care plan and the maintenance of your patients over their entire lifecycle.”
Preparing for an AI Shift in Healthcare
According to Whitney Merrill, Head of Global Privacy and Data Protection Officer at Asana, despite the many potential benefits, resistance is normal with any major technological change. She recommends that teams acknowledge these hesitations and openly communicate who the users are and the problem being addressed.
“The more you can say, ‘your concerns are valid, and we will dig into it more to address it,’ the more it will help with adoption. And an openness to change will develop over time,” Merrill said. Organizations should also first evaluate the workflows they want to change, and the problem they want to solve. Then they should find the right solution for that problem. For example, find a specific use case, dig into it and explore how AI will improve things over time.
“That’s a big, big component of AI, reducing that manual workload or the stuff where you feel like you must be super technical. And AI helps to democratize that in many ways and makes it easier for people,” Merrill added. This, of course, comes with training that will inevitably be needed to help smooth the cultural transition and usability. With the current pace of AI innovation, training will be more frequent too.
Michael Anderson, Chief Strategist for U.S. Public Sector at Informatica, noted that it is important to consider anonymizing data in the healthcare community as it relates to data management platforms if privacy or bias is a concern. Researchers don’t necessarily need to know that type of data to make provider-based decisions.
“Capabilities are there today to do that with technology, and it can be policy-based across an entire environment,” Anderson said. “You can apply policies on who can see what and when.”
Data tagging and governance are also critical when applying AI solutions. “Data definitions and glossaries are important. Even in terms of AI, you must figure out the foundational things before you can train and feed any model doing anything in the healthcare industry,” he added.
Ensuring that the right foundational data management requirements are in place will lead to better outcomes and reduce the risk of common AI concerns like biases and hallucinations.
Ultimately, experts agreed that AI can – and is – making waves in federal healthcare, as agencies shift from exploring and prototyping AI to preparing to deploy AI-based solutions in existing systems that can improve patient care. While this introduces challenges and cultural changes, it’s a cross-agency and cross-sector approach that requires building trust with patients and providers alike.
