Synthetic Data’s Critical Role in Preventing Suicides in Veterans

As the 12th leading cause of death in the United States, suicide is a major public health concerns and the Department of Veterans Affairs’ top clinical priority. In 2018, VA implemented a 10-year vision to end veteran suicide, as outlined in the 2018 National Strategy for Preventing Veteran Suicide (National Strategy). VA has worked to prevent suicide among Veterans — including those who do not and may never seek care within the VA health care system. Despite the COVID-19 pandemic, the suicide rate among veterans in 2020 continued a decline that began in 2019.  According to VA’s 2022 National Veteran Suicide Prevention Annual Report, the number and rate of suicide deaths among U.S. veterans fell in 2019 and 2020, with 6,146 Veteran suicides in 2020, 343 fewer than in 2019.

VA’s strategy has been an important catalyst helping to drive other major initiatives, such as the President’s Roadmap to Empower Veterans and End a National Tragedy of Suicide (PREVENTS), expanding the Veterans Crisis Line (VCL) to using 988 and, more recently, a nationwide campaign to address firearm safety. In addition, there are multiple efforts underway working in concert with VA including campaigns like ‘Lethal Means Safety,’ Veterans Crisis Line outreach and ‘Don’t Wait. Reach Out,’ plus other programs like the Staff Sergeant Fox Suicide Prevention Grant Program and Mission Daybreak, a suicide prevention grand challenge.

VA understands that veteran suicide is a complex and multifactorial problem, requiring innovative and data-driven approaches that help inform and drive the broader set of suicide prevention efforts.

One such area is the development of “synthetic data,” which is generated health care information that accurately mimics real-world conditions impacting veterans to help “train” computers to identify risk, tailor recommended treatments and engage care as early as possible.

Synthetic data that is based on the real needs of different veteran populations provides an accurate and reliable picture of individual veteran needs, trends across high-risk groups, as well as a basis from which we can proactively identify those who may be escalating towards risk in life-saving supports.

Specifically, VA is applying synthetic data in four areas to predict risk of mental health distress and to test the impact of care pathway modifications for cognitive distress to advance suicide prevention:

1. Predictive modeling improves VA’s understanding of the current state of key data efforts in suicide prevention by confirming and clarifying ongoing challenges veterans face
2. Quality improvement enables VA to assess clinical variation — the overuse, underuse, differential uses and waste of health care practices and services — with varying outcomes in mental health triage and treatment
3. Population health management identifies veterans undertreated for mental health concerns and develops care pathways to engage and connect with veterans at high risk
4. Optimizing diagnosis and treatment uses novel data sets made safe through synthetic data that help identify veterans with undiagnosed physical or mental health concerns increasing their risk of a crisis event or suicidal ideation.

VA is not alone in recognizing the role of synthetic data in reducing and preventing veteran suicide. Early on, VA’s whole government (and patient) approach to synthetic data and veteran suicide involved a broad network of government, industry and academic partners. One of these efforts includes Mission Daybreak. Formerly known as the VA Suicide Prevention Grand Challenge, Mission Daybreak is challenging outside groups to help reduce veteran suicide through real-time innovation efforts that are local and accessible for veterans.

Earlier this summer, VA received more than 1,300 concept submissions from veterans, Veteran Service Organizations, community-based organizations, health tech companies, startups and universities during the first phase of the challenge. The second phase is now underway with the top 30 concept papers selected to move forward and those groups getting eight weeks to carefully refine their ideas. This phase includes the opportunity for the finalists to join a virtual accelerator program designed to help the teams develop roadmaps for prototyping, iteration, testing and evaluation. Plus, they will have access to cloud resources with synthetic data from VA. While using synthetic data isn’t required as part of the challenge, it is available to strengthen and refine solutions that can benefit from supporting health record data.

This month, finalists will present their solutions to key stakeholders, investors and partners during a live pitch event.

Mission Daybreak is one of the many examples of how VA has harnessed the power of partnerships to advance synthetic data science, research, application and impact. VA is also currently using synthetic data to collaborate with partners to develop and test models for heart disease, sleep apnea, Alzheimer’s disease, sarcoidosis and spinal stenosis, among others. VA is planning for more open science challenges in fiscal year 2023 for government, industry and academic organizations to develop solutions that help veterans.

Dr. Carolyn Clancy serves as the assistant under secretary for health (AUSH) for the Office of Discovery, Education and Affiliate Networks (DEAN) at the Veterans Health Administration (VHA). Amanda Purnell, Ph.D., is the director of data and analytics innovation with the Office of Healthcare Innovation and Learning (OHIL).