Military Algorithm Can Predict Illness 48 Hours Before Symptoms Show

U.S. service members are strong but they’re still people, and people get sick sometimes. But when one gets sick at the last minute, that can have serious repercussions on their unit’s ability to execute critical missions.

The Defense Threat Reduction Agency, or DTRA, is trying to get ahead of this problem by developing a predictive algorithm that knows whether a service member is falling ill—due to anything from a cold to exposure to biological weapons—up to 48 hours before they start to show any symptoms.

“Think of it as a check-engine for the human body,” Edward Argenta, science and technology manager for DTRA’s Joint Science and Technology Office, told Nextgov.

DTRA partnered with the Defense Innovation Unit to leverage the latter’s other transaction authority—a special procurement method outside of the Federal Acquisition Regulations—to develop the algorithm hand-in-hand with health IT company Royal Philips.

Using its own globally-collected data sets, Philips was able to develop a unique algorithm for the Defense Department. Using 165 distinct biomarkers across 41,000 cases, the Philips team was able to create the Rapid Analysis of Threat Exposure, or RATE, algorithm, which the company says can “predict infection 48 hours before clinical suspicion” with better than 85% accuracy.

“For comparison, this performance lies in between blood-based breast and prostate cancer screening tests, and an enzyme immunoassay based first-tier Lyme disease test,” according to a company release.

“By coupling large-scale data, with our experience in AI and remote patient monitoring with DTRA’s drive for innovation, we were able to develop a highly predictive early-warning algorithm based on non-invasively collected biomarkers,” Joe Frassica, chief medical officer and head of research for Philips North America, said in the release. “While the RATE data is derived from acute care settings, we believe that is adaptable to active duty personnel.”

While Philips will hold onto its data—with initial testing done, the government doesn’t need that proprietary data, Argenta said—the military now holds the intellectual property rights to the algorithm that was created, allowing DTRA to further develop the program to work for all service members, from the back-office to the battlefield.

The solution procured through Philips is “the first step in that journey,” Argenta said. “At the end of the day we do want to have it as a whole-forces, all-warfighter solution. But we started in a clinical setting to evaluate the state of the science and see if we could actually get there.”

Testing the algorithm in a clinical setting was instrumental, Argenta explained, as developers were able to monitor patient vitals before symptoms began to show. As clinicians saw things like infections and sepsis develop in patients, they were able to compare data prior to symptoms showing and create an algorithm that would predict the onset of illness 48 hours before any outward signs.

“In some of these clinical studies, you’re waiting for these overt signs and symptoms and then you can start monitoring,” he said. “Since we’re trying to be predictive, we need to be ahead of any event that we might think is a biological or chemical exposure or illness.”

From here, Argenta’s team plans to refine the model at military hospitals and clinics managed by the Veterans Affairs Department.

“But in the future, we’re looking to bring it to the tactical edge and do this at the individual soldier level,” he said.

The algorithm developed by Philips currently uses two data models: one using lab results and the other only using vitals obtained through non-invasive means. Argenta said his team is focused on the latter.

“The goal of my portfolio is to never prick or stick or take a sample out of you,” he said. “If I can do everything remotely, digitally, from monitoring you with a wearable device that might sit on your body—like a watch-based one or a chest strap one—that’s where the portfolio has been focused on.”

The non-invasive model currently uses biomarkers such as heart rate, saturation of oxygen levels in the bloodstream, blood pressure and temperature.

“Really, routine measurements that are easily captured through some of the rapid advancements in the wearable technology space,” Argenta said.

Argenta said it is too early in the process to say whether troops will have to opt-in to the monitoring program or be opted-in by virtue of enlisting. He said, as the program matures, department leaders will have to look at how the program fits with doctrine, organization, training, materiel, leadership and education, personnel and facilities—a military assessment process known as DOTMLPF. Before any new technology is deployed with forces, leadership first has to amend the DOTMLPF to account for the new capability.

However, Argenta said the program has another two to three years of testing in operational environments before it even gets to that state.

“This first effort is really about is the science mature enough,” he said. “During the next two years, we’ll be going out to the tactical level and seeing if this [early] success warrants going through some DOTMLPF changes.”

That is a soft timeline, Argenta added, though he thinks it is on the conservative side. With how fast this technological area is advancing, he said a few innovations in the private or public sector could push the program along much faster.

DTRA plans to continue working with Philips, using the follow-on option to extend the OTA contract as the department begins operational testing in the field. Those first field tests will be held at military hospitals and academies before being deployed in combat.

As the program develops, Argenta said he hopes the algorithm can be expanded to be forward-looking, as well, predicting how long it will take troops to recover from an illness and giving commanders hard data on force readiness.

“We think this first, initial capability is really exciting,” he said. “And it does open up the door for us to continue resourcing it and continue funding the research and development of it to see how far we can go.”