Like the Pandemic Response Challenge, Veteran Affairs (VA) is utilizing AI to optimize care strategies for COVID and other urgent healthcare needs.
By Dr. Thomas Osborne
With large numbers of people suffering and dying from COVID every day, there’s no time to waste. As a result, we’ve been driven to meet this challenge with the most advanced care strategies and latest technological tools available. Key factors for success that we’ve identified include collaboration and the ability to understand complex interrelated factors to guide data-informed decision making to enhance care.
In this sense, our approach is similar to that of the Pandemic Response Challenge, to which I’ve been invited as a senior advisor. Cognizant and XPRIZE launched the challenge to bring experts from around the world together to combine data, artificial intelligence (AI), collaborative partnerships and human-led innovation to understand and predict how policy interventions will impact the trajectory of COVID-19.
The VA National Center for Collaborative Healthcare Innovation (NCCHI) collaborates not only with the U.S. Department of Veterans Affairs but also with other experts from around the world. In this catalytic environment, we’re accelerating our combined understanding of advanced technologies such as AI, virtual reality, augmented reality, sensor technologies and 5G.
One of these important areas involves big data. At VA, we are fortunate to have a robust longitudinal electronic health record (EHR) database, which we’ve been utilizing to better understand and inform optimal care strategies for our Veterans. Importantly, the insights from VA have also serendipitously been beneficial to others who are working to solve similar challenges.
Predicting Needs and Healthcare Resources
Prior to the pandemic, NCCHI tested and verified a way to use advanced analytics and data in our EHR dataset to predict risk and the likelihood of different postsurgical outcomes. The goal was to see if routinely collected historical data could help guide the best care options for individual patients. In our published peer reviewed research, we found our EHR system could do this better than other tools available and do it automatically because the structured data is gathered routinely. We’ve collaborated with our VA colleagues as we leverage big data to address critical issues such as hospital readmission risk.
Once COVID hit, our NCCHI team switched gears and quickly utilized the same analytics strategy to understand and improve COVID-19 care. One of our early objectives was to identify the patients at greatest risk for serious COVID-19 outcomes. These projects were developed with the intention to help both front-line staff with direct care decisions and executive leaders working to tactically distribute resources. In one of our published peer reviewed publications, we demonstrated that demographic factors, comorbidities and other data in our EHR could predict the likelihood of COVID-19 mortality, hospital admission, long hospital admission, ICU stay, long ICU stay and intubation.
As we worked to optimize our COVID-19 predictive models, one of our data analysts uncovered an unexpected association related to COVID-19 mortality. Specifically, our recently published paper reveals that taking aspirin prior to COVID-19 diagnosis was correlated with a dramatic decrease in mortality. Although prospective assessment is required before incorporating this finding into care, the potential association between aspirin’s anti-inflammatory and anti-coagulation properties could have a significant impact on COVID-19 treatment strategies in the near future. In parallel, we are also closely assessing the impact of COVID-19 in relationship to social determinants of health.
We’ve also collaborated on a COVID-19 wearables research project with the Department of Defense (DoD). This DoD study utilizes wearable sensor data and AI to detect changes in vital signs that can predict the likelihood of early COVID-19 infection. The goal is that early detection can prevent the unintended spread of the virus before symptoms arise.
We are also utilizing AI and other advanced technologies to enable remote care, which has become more important during COVID, and to provide insights that offer the potential to improve outcomes. Here’s a quick look at just three of these initiatives:
- We are working with Podimetrics as part of our initiative to end diabetic limb loss with the use of an AI-driven, sensor-based “smart mat” to detect patients at risk of diabetic foot ulcers from their home. If a significant enough change is identified in the person’s foot heat signature, action can be taken before a potential infection develops. Therefore, the system reduces the risk of osteomyelitis and amputation.
- A partnership with Verizon, Microsoft, Medivis and NCCHI has resulted in our VA facility being one of the first 5G hospitals in the world. This lays the groundwork for advanced AI point of care solutions empowered by 5G edge computing. We are also using 5G-enabled augmented reality tools to power advancements in medical training and teaching at a distance, through the use of interactive 3D holographic models.
- A partnership with Google/Deep Mind has demonstrated that our EHR data can be utilized to predict onset of acute kidney injury days in advance, with the potential for earlier and more successful intervention. This VA-Google collaboration has also uncovered additional important insights about AI for the general scientific community.
The Human Component
Although advanced tools such as AI are integral to innovations that improve outcomes, there are three decidedly non-technology-oriented factors that are also critically important:
- Collaboration. Just as collaboration is a key strength of the Pandemic Response Challenge, we at VA recognize the power of collaborations with partners in the business, academic and government realms, with the goal to create solutions to advance vVeteran health. A major premise of NCCHI is that by bringing diverse people together, in an appropriate way, we can achieve much more than we would ever be able to do individually.
- Putting people first. We also know the pitfalls of using technology for technology’s sake. We are all most successful if we start with a well-defined problem that people want solved, and then look at the portfolio of the best tools available to solve the challenge. We then prioritize solutions with the best chance of succeeding and the greatest potential positive impact, so we can then diligently evaluate their potential before sharing with the rest of the VA.
- Making sure humans lead technology such as AI -- not the other way around. Advanced technology, such as AI and predictive analytics, can be an important component of groundbreaking solutions. However, AI is only as good as the data available, how it is curated, and the way it is analyzed. We need to be diligent, not just in terms of the mechanics of the algorithm but also, and perhaps most importantly, in how we acquire and curate data so that we can create unbiased answers and produce the best value to our patients. As such, it is very much about human-machine collaboration and finding the best way to combine skills and assets to enhanced healthcare.
Unfortunately, the challenges from COVID-19 will not be over when the virus is eradicated. For example, “long haulers” and others with longer term health problems caused by the virus will need to be better understood, and specific customized treatments developed for those fighting prolonged illness. In addition, we will also be faced with the needs of those who have avoided medical treatment or did not have access to typical routine care as a result of COVID. We will need to develop and deploy advanced solutions that can predict gaps and develop advanced ways to avoid and address long-term challenges.
With large numbers of people suffering and dying from COVID every day, there’s no time to waste. The fact that XPRIZE and its partners have come together with such agility is inspirational – there’s no doubt we can do more, more quickly, by working together with the best tools at our disposal at a critical time.
About the Author
Thomas Osborne, MD, is the Director of the National Center for Collaborative Healthcare Innovation (NCCHI) and Chief Medical Informatics Officer (CMIO) at VA Palo Alto Health Care System. As Director of NCCHI, he is leading multidisciplinary teams in the co-development and implementation of pioneering healthcare solutions throughout VA, with other government agencies, and with leading technology companies. His teams overarching goal is to deliver the best and most advanced healthcare solutions to our Veterans. He is the inaugural recipient of the VA Under Secretary for Health Robert L. Jesse Award for Excellence in Innovation. His work has been published in numerous medical journals and textbooks on topics such as predictive analytics, artificial intelligence, sensor technology, augmented reality, virtual reality, and the future of health care. Dr. Osborne received his medical degree from Dartmouth Medical School, completed his residency at Harvard Mount Auburn Hospital, and fellowship at Harvard Massachusetts General Hospital. He is board certified in both Diagnostic Radiology and Neuroradiology. Dr. Osborne is also a staff neuroradiologist at VA Palo Alto Healthcare System and Clinical Associate Professor (affiliated) at Stanford University School of Medicine.