Treating Veterans' Complex Health Problems with New Technology

Members of the armed services are regularly exposed to health risks that few civilians could ever imagine: Eardrum-bursting blasts of artillery fire. Airborne toxins from burn pits in Iraq. The threat of death or dismemberment in combat zones. The hazards of toting 50-pound packs across boulder fields.

The chance of injury or impairment is almost always present.

"There's just a lot more going on in your physical environment that can impact your health in ways that civilians don’t experience--we’ve had terrible things happen to us," says Mary Lowe Mayhugh, a health systems strategist at MITRE who is also a 36-year veteran of the U.S. Army. “And, of course, on top of physical injuries, there are so many ways in which these experiences can affect service members emotionally later, from post-traumatic stress disorder [PTSD] to addiction and suicide."

As a result of these complicated, far-reaching, and unique health issues, more than 9 million veterans turn to the Veterans Health Administration (VHA) for services each year. At MITRE, helping the government to provide veterans with reliable access to the customized care they need is a priority in both our direct work with the Department of Veterans Affairs (VA) and through our independent research and development program.

"Veterans consistently put the needs of the country above their own needs," says Sybil Klaus, M.D., M.P.H., who leads MITRE's health research area. "So it's vitally important that we support them after they leave the service in the same way that they supported us during their time of service."

MITRE's health research program focuses on three domains:

• Health systems are the networks of organizations, people, and activities designed to promote health. These go beyond actual services to include information systems, infrastructure, and other entities. MITRE researchers use systems thinking and applied mathematics to tackle issues in this realm.
• Health data science applies advanced analytics and artificial intelligence to generate data-driven solutions to complex health problems.
• Health futures explores the emerging technologies that might have a significant impact on health issues and helps organization prepare for change.

Klaus says her portfolio of projects strives to reimagine healthcare delivery in an age in which extraordinary advances and cutting-edge technologies are becoming available. She says that all of the health research projects she manages either directly or indirectly benefits veterans. These projects support the top priorities of the VHA.

We take a look at a few of the projects here.

Using Speech Data for PTSD Predictive Analysis

Thousands of veterans suffer from PTSD, but only about half of them seek out and receive the treatment they need, according to a National Academies of Sciences, Engineering, and Medicine report.

This disparity is due to many factors, including lack of regional clinical services and insufficient staff, which can cause long wait times. Patients may also not be able to take time off for a doctor's appointment, particularly if care is two or more hours away from where they live.

Addressing these gaps is behind a MITRE research project in the health systems domain that provides a method for rapid, multi-modal screening and prediction for PTSD and other mental health conditions.

The MITRE Multi-modal Predictive prototype combines a systems approach with new technologies to accelerate the assessment and treatment of people at risk of mental disorders. People answer a series of questions about their lives and experiences during a PTSD interview, which can be recorded with voice data. Based on the machine learning models—which are built on real PTSD recordings and clinical information—the system then determines the likelihood of PTSD using their answers to PTSD assessment questions and voice characteristics. One of the biggest benefits: The interviews and predictive analysis can be done remotely and systematically.

"Technology such as this has the potential to increase the number of veterans who can receive timely screening," says Qian Hu, Ph.D., chief scientist for speech technology and principle investigator of this project. "For example, they can have the PTSD interview over telephones or Web services, without making a trip anywhere."

"The end result is an objective assessment that can be shared with multiple providers to help develop the best treatment plan," she adds.

Hu's team has spent thousands of hours listening to actual interviews of more than 1,200 veterans talking about what they experienced in the line of duty. By mapping certain speech characteristics, such as rate, energy, timing, contour, and vocal effort, the researchers identified patterns that signal nuances associated with PTSD symptoms. They apply audio signal processing to extract emotion data from voice signal, time-series analysis to collate emotion detection and change patterns, and supervised machine learning to enhance the PTSD predictive algorithms.

"The voice analysis adds an objective dimension to the PTSD assessment," Hu says. “The way people speak reveals the true state of their emotions or mental state."

Her hope is that the multimedia and multi-modal predictive system could one day become a standard diagnostic tool, similar to X-rays or bloodwork, to facilitate diagnosis and treatment decisions and progress tracking, which could ultimately improve the health and lives of veterans.

Can We Help Prevent Suicides?

Another project in the health data science domain addresses the critical problem of suicide. Every day, about 20 veterans take their own lives. What if there were a way to determine when someone is at risk of suicide?

A team from MITRE partnered with the VA and the University of California, Los Angeles, to build on the VA's mobile sensing application which monitors mental health indicators, such as sociability, physical activity, and sleep quality. It’s all done passively—the individual doesn’t have to do anything after downloading the app.

The collection of indicators can accurately reflect extreme variability in a person's mood. This marks a significant advance over traditional static measures, such as filling out a questionnaire in the doctor's office, which only captures a single moment in time.

Researchers are now refining the sensor suite on the basis of early feedback. For instance, they are working to determine the nuances between "normal" stress, such as a student facing an upcoming exam, and "unique" stress, such as a recent breakup. The former would be appropriate, while the latter could signal a potential emotional upheaval.

"We are learning a lot about how to label mental health, which is by nature very subjective," says principal investigator Lionel Levine. "We're showing that you can assess an individual passively and correlate that data to mental health. Our long-term goal is to turn that process into a tool that individuals and families can use to achieve wellness."

Ultimately, having an additional method to quantify mental health could have a significant impact on treatment, enabling more accurate diagnoses and more customized treatment plans, including when to make critical interventions in time to save lives.

Helping the VA Make the Right Decisions to Support Veterans' Health

MITRE researchers in the health futures domain are exploring the possibilities and requirements for the near- and long-term use of alternative methods for providing care.

A team led by Matt Henchey and Deborah Ercolini is developing a modeling and simulation tool to help the VA understand the implications of expanding access to telehealth (long-distance communications) capabilities for veterans suffering from PTSD. The tool can analyze data to help answer such questions as: How could telehealth services affect wait times? Travel times and costs? Would patients access care more often?

"We want to help decision makers identify potential capacity requirements and gaps in resources, such as personnel, equipment, and key technologies needed for use of alternative methods, including telehealth," Henchey says. "And we want to make this totally patient-centric. We want to better understand how the patient would use these services and what the outcomes would be."

The tool maps the patient journey from the life event that caused the trauma, through screening, diagnosis, treatment, and monitoring. It uses a variety of publicly available data to analyze service supply and demand, then combines the patient journey with a demand model and a supply model to create a realistic scenario.

The approach offers a framework and repeatable process that can help the VA identify gaps, allocate resources, and determine effective policies and procedures—based on the actual needs of the veterans.

Ercolini recently demonstrated the research at the VHA's 2019 Innovation Experience, a showcase for new advances in veteran-centric healthcare.

"Focusing on veterans' well-being is very much what MITRE does best—deal with complex problems that stretch across many domains," says Mayhugh, who works with Henchey and Ercolini.

"Applying innovative approaches to big challenges like healthcare is part of MITRE's mission to make the world safer."

—by Twig Mowatt