VA Leveraging Synthetic Data to Improve Suicide Prevention Efforts

IT leaders at the Department of Veterans Affairs believe synthetic data could be leveraged to reduce veteran suicide by providing scalable and interoperable suicide risk prevention analytics.

“We have so much data at our disposal about health care and benefits and serving veterans. If we can rip the cultural and structural kryptonite from our chest and integrate that data and manage it, we can better serve veterans and frankly position VA leadership broadly to help lead American health care and wellness,” Kshemendra Paul, VA’s chief data officer, said during ACT-IAC’s Health COI.

Synthetic data is generated from computer simulations or algorithms and provides an alternative to real-world data. It’s gaining traction in the health care space, increasingly being used to create AI models, because it enables rapid experimentation and testing in a secure fashion.

“We need mathematically provable ways of enhanced sharing of information, and I think synthetic data has the potential to help us with that,” Paul said.

Synthetic data could help clarify ongoing veteran challenges to then enable VA to make meaningful changes. Clancy said the VA has internal roles to drive veteran suicide prevention that have helped predict risk factors, however, the agency is looking to expand its support to veterans outside of the VA care system.

“It’s very clear that we need to use reliable and timely data to identify and address issues impacting our veterans, while also ensuring privacy,” Dr. Carolyn Clancy, assistant under secretary for Health for Discovery, Education and Affiliate Networks (DEAN) at the Veterans Health Administration (VHA) said. “We believe synthetic data, modeled to precisely mirror real veteran data, while protecting veteran privacy is a great path forward.”

VA plans to develop a new pathway to generate synthetic data, which will preserve privacy while mirroring real veteran data, and derive meaningful insights. Clancy said the VA has a cloud-based platform where synthetic data can be generated, but faces policy challenges. The VA currently doesn’t have synthetic data policies, which makes it difficult to standardize and use synthetic data across the enterprise.

A newly announced initiative, the VA Suicide Prevention Grand Challenge, aims to reduce veteran suicide by 10%, leveraging external collaborators to rapidly source suicide prevention innovations. The challenge hopes to enhance real-time prevention and responsive interventions that are local and accessible for veterans.

“If we’re going to have an audacious goal of reducing veteran suicide by 10%, we need audacious resources and audacious partnerships, and we need meaningful, novel and secure data as a component for novel solutions for this wildly complex problem” said Amanda Purnell, director of Data and Analytics Innovation at VA.

As part of the Suicide Prevention Grand Challenge, VA needs a dataset that is fit for the right purpose. Purnell explained that synthetic data is significantly more advanced technology compared to removing content from original data. By creating artificial datasets, VA will be able to bolster the privacy and security of veteran data.

“We create and make use of veteran data to serve veterans. The goal of using the data that we have access to is to improve care pathways, to improve proactive efforts, to engage veteran care, improve access to services, improve access to veterans’ benefits. The reason that we pursued synthetic data is a clear signal that deidentified data is insufficient for current computational technologies,” Purnell said. “We’re pursuing synthetic data because it’s easy to reidentify deidentified data.”

One target area the VA will focus on is “digital life data,” novel sources of data collected from devices while preserving privacy and equity. Digital life data could enable VA to develop better early warning systems. Purnell sees potential in identifying at-risk veterans prior to crisis, which would enable a proactive approach to reduce veteran suicide rates.

“In addition to good information and new sources of information, we also need community partners, radical partnerships, collaboration across many government partners, academic and industry partners to really come together to work together in a multi-stage effort to solve an important problem like this,” Purnell said.

VA is on target to launch phase one of this challenge in summer 2022. Purnell says VA plans to maintain this challenge for many phases to come because of the complexity of this long-term issue.