Pentagon gets $10M to advance an AI algorithm that detects diseases early via wearables
With new funding from Congress, the publication of a peer-reviewed clinical study and a determination from the Food and Drug Administration, the Pentagon is set to expand an operational effort that’s applying a predictive artificial intelligence algorithm and a growing cache of wearable biometric-tracking devices to enable the early detection of infectious diseases among defense personnel.
New cohorts of officials from across the Defense Department are now poised to participate in the Rapid Assessment of Threat Exposure (RATE) program.
Led by the Defense Innovation Unit (DIU) — and developed in collaboration with multiple government and commercial partners including the Defense Threat Reduction Agency and the technology corporation Philips — the initiative is demonstrating how AI-powered wearables can help predict COVID-19 and other infections days before diagnostic testing.
“RATE goes against our mental model for disease. Normally, you wait until you have symptoms to do anything. With RATE, we have the ability to tell you something’s coming before you feel anything. That is a huge shift for how we think about disease spread as well as prophylactic measures to combat the disease much earlier,” Air Force Lt. Col. Jeff Schneider, DIU’s program manager for RATE, told DefenseScoop on Wednesday in an email.
‘Breaking that mold’
RATE’s early roots trace back to a study initiated in 2018 and completed in 2019, which focused on spotting hospital acquired infections using hospital-grade equipment. Through it, experts demonstrated the capability to detect a wide array of infectious diseases 48 hours, on average, before symptoms appeared.
Then, “when COVID-19 started, we lateralled this technology from the hospital-grade equipment to wearables, and attempted to use the algorithm in real-time,” Schneider explained.
It took a hefty effort to transition a retrospective algorithm based on hospital-grade hardware to a prospective algorithm that used common, commercial wearables — but in the first 6 months, the team had done exactly that.
The powerful algorithm — which continues to advance in sophistication today — was then trained on health data captured by more than 11,000 DOD-affiliated participants wearing bio-sensing, commercial off-the-shelf rings and wristwatches. Officials from the Defense Advanced Research Projects Agency, the Naval Medical Center San Diego, the Air Force Research Lab, and elsewhere, contributed to the pursuit, along with more than 150 Pentagon “investigators.”
“Within a year of the COVID-19 effort, the algorithm was detecting COVID-19 at nearly the same level as the 2018-2019 model detected [hospital acquired infections],” Schneider said. By the “completion of the RATE COVID effort,” he added, the algorithm “even predicted infections up to six days prior — to include asymptomatic cases.”
All of the program data is de-identified, he confirmed.
“The ‘RATE score’ is attached to an anonymized participant ID. Only the DOD participant and the DOD unit’s point of contact knows the participant ID. The RATE score is displayed both at an individual level and at an aggregate unit level via a dashboard,” Schneider noted.
“Philips by nature has access to the de-identified RATE data for the algorithm to work, but has no way to identify the individuals,” he added.
The company has previously been contracted for millions to support this work, but DOD has not disclosed the total value invested in this program to date.
Philips continues to refine the algorithm and its inputs.
DIU initially transitioned RATE to an operational contract after that two-year prototype successfully met its goals for COVID-19 prediction. The program was one of the first 10 pilot projects Congress originally funded through its Accelerate the Procurement and Fielding of Innovative Technologies (APFIT) initiative.
“When APFIT funding was provided to the DOD by [Rep. Ken Calvert, R-Calif.] in 2022 to help remedy ‘Valley of Death’ issues, DIU submitted the project to [the office of the undersecretary of the defense for research and engineering] for additional funding, as our project moved quickly to a Program of Record,” Schneider said. In the acquisition community, the term “Valley of Death” refers to challenges in transitioning promising technologies from research-and-development into production and fielding.
These newly announced APFIT funds are valued at $10 million.
“Beyond providing more DoD warfighters a truly amazing capability, this APFIT effort allows us to continue iterating and receive some great feedback from operators. Too often the DoD’s approach is to wait until a prototype is near the 100% completion rate, and then attempt to field it force-wide. This is both extremely costly if there are changes, as well as historically taking an unacceptable timeline to field to warfighters,” Schneider explained.
“APFIT is breaking that mold by offering an alternative approach where we begin to field technology at a smaller scale to an intermediary before it goes to a large program of record, which in this case is the [joint program executive office] for RATE,” he said.
This expansion follows an FDA determination that the technology would be classified as a “general wellness device,” which — as opposed to medical devices — are unregulated by the agency.
Bring your own device
With that new APFIT funding boost, the Defense Innovation Unit will add 4,500 devices and distribute the wearables to new cohorts within the Defense Department — including Air Combat Command personnel. In association with the command’s new Diamond Care Initiative Plan, 360 1st sergeants will be issued wearables that monitor and capture data reflecting their overall health and vital signs.
Schneider said those involved in this work also “hope to add additional people leveraging a broader number of wearables in the next year and gain additional funding” in fiscal 2024 “to expand the use of RATE when this transitions” to the joint program executive office.
A key objective of this pursuit is to get honest feedback from DOD users to ultimately help enhance the platform before it transitions to that office.
“This allows for an iterative approach and lets RATE improve rapidly at a smaller scale, giving the DOD a better product that’s fielded years earlier when there is a force-wide deployment of this technology,” Schneider told DefenseScoop.
In his view, there’s a “healthy appetite” for wearable devices within the department.
“Beyond the infectious disease prediction, the feedback we’re receiving is that these wearables provide the warfighter better awareness of their fitness and sleep, which often leads to an increase in overall well-being,” Schneider said.
And if they held onto their hardware, the 11,000 officials who previously participated in the RATE efforts will be able to re-affiliate with this work.
Currently, Garmin watches and Oura rings make up the wearables deployed. Through this extension, the DIU-Philips team intends to incorporate three additional health-tracking products. Building on those, the overarching aim is for the evolving RATE algorithm to be device agnostic — meaning it can use biomarker data from any commercial grade, off-the-shelf device.
“We then run those markers against our clinical data sets in the cloud to create a RATE Wellness score. The score has proven to be indicative of onset of infections. We can offer it through a licensing model to anyone who wants to add this capability to their device or as a stand-alone service,” Navin Natoewal, head of integrated technology solutions at Philips, said in a press release.
This “wearable-agnostic” approach makes sense to DIU, too.
“We’ve learned that there is no ‘one size fits all’ for the perfect wearable,” and “DOD warfighters have unique mission sets as well as unique personalities — so one large portion of this effort is to allow three other wearable brands to plug into the RATE algorithm,” Schneider said.
He and his team envision soon allowing for a “bring your own device” or BYOD-based method, so that service members can choose from a slew of wearables matching their uniqueness.
“This BYOD approach also allows us to keep from modifying the normal habit patterns of the warfighters. For example, some people didn’t like the form-fit of the watch, but were comfortable wearing the ring at night — in a few months we’ll have a ring-only option for RATE,” Schneider said.