Challenge
The United States can develop more reliable ways to identify when a veteran’s health or circumstances call for earlier care. Critical signals are distributed across years of clinical record, genetic background, and other health and social data, while existing risk models are static and difficult to validate across large and varied populations. Data integration and new medical and biological applications can advance discovery, diagnosis, and treatment of chronic diseases such as cancer, heart disease, neurologic disease, and mental health conditions. This challenge will expand ongoing efforts to improve risk prediction models in Department of Veteran Affairs (VA) priority areas such as homelessness, suicide, frailty, and chronic diseases prevalent among Veterans, leading to better preventive care and improved health outcomes for those who have served in our nation’s military.
AI Solution
Bringing VA’s vast electronic health record data (EHR) and one of the world’s largest genomic datasets from the Million Veteran Program (MVP), together with DOE’s exascale and supercomputing resources and expertise, provides a foundation for accelerated scientific discovery and the development of trusted biomedical AI models and applications to enable earlier disease detection, prevention, and treatment for Veterans. Secure data integration, large-scale training, and real-time application with continuous validation will ensure robustness, generalizability, and long-term reliability, allowing hypotheses to be generated, tested, and iterated upon to accelerate progress. Additional data from the Department of Health and Human Services (HHS), such as Centers for Medicare and Medicaid Services (CMS), will enhance predictive models.
Justification
VA and DOE have partnered since 2016 to apply secure computing practices to sensitive data, leading to improved health outcomes for Veterans. Copies of VA’s clinical data and MVP genomic data reside in the secure computing enclave at DOE’s Oak Ridge National laboratory (ORNL), with secure access to the Frontier supercomputer. This has enabled collaborative research projects advancing VA’s priority areas, such as suicide prevention, prostate and lung cancers, schizophrenia and bipolar disorder, and heart failure, resulting in over 50 scientific publications and clinical translation of discoveries related to the risk of suicide and prostate cancer. Harnessing VA’s data and Frontier, one of the world’s fastest supercomputers, has produced novel and energy-efficient AI models pre-trained on 1.7 trillion tokens of clinical text and capable of handling contexts exceeding 1 million tokens.
National Impact
Building upon existing infrastructure, resources, and expertise across VA and DOE, this challenge will unlock precision medicine through accelerated scientific discovery, earlier disease detection, and consistent decision support to improve health outcomes for Veterans and, ultimately, all Americans. It will enable the VA to recognize changing needs earlier, connect Veterans with appropriate care and support, preserve independence as Veterans age, and reduce the rate of preventable crises and complications from chronic disease.
Challenge
This was last updated .