The Right Information in the Right Hands at the Right Time Warfighter Health’s MERIT Artificial Intelligence (AI) tools are designed to deliver speed and accuracy in clinical decisions that can accelerate healthcare interventions and have a direct impact on force readiness
The health of servicemembers is vital to force readiness. The military has always collected member information—starting at recruitment and extending all the way to separation and/or retirement. The challenge resides in harnessing the power of that data to improve decision making—whether for an individual servicemember, a military unit, or the joint force. It is a major task to aggregate and utilize all of the information that exists in an effective and timely manner to measure health status consistently, but this task is becoming easier with the introduction of electronic health records (EHR) and artificial intelligence tools.
The Warfighter Health Mission Initiative within the Joint Artificial Intelligence Center is working on a project known as MERIT—Medical Evaluation Readiness Information Tools. MERIT uses data science and AI capabilities to develop predictive models that analyze electronic healthcare records and other applicable data to provide indicators of a servicemember’s current health status. The initial AI algorithms are designed to be a decision support tool for medical providers. These results can also be summarized to capture unit health and readiness.
Visualization is a critical component to demonstrate AI-enabled decision support capabilities for the consumers. The MERIT team has built a dashboard, representing three primary viewpoints of AI (medical provider, commander, and senior leader) to demonstrate its potential. A MERIT dashboard is currently on the Defense Health Agency’s CarePoint system to begin field testing with functional subject matter experts (medical providers, data scientists, and AI experts) to get consumers comfortable with AI and refine the dashboard based on user feedback.
MERIT and the Medical Provider
Medical providers typically have 15 minutes for each appointment. To make a clinical decision, servicemember information must readily be available to the medical provider that is easy to access and in a usable format. Currently, we have multiple systems that house enormous amounts of data, but providers may have a limited time to review. Therefore, providers may only be looking at a very small portion of the EHR data to make a clinical decision, whereas AI can scan the entire record continuously within seconds. The real value of AI is the speed in which it can process huge amounts of data to give a holistic view of the servicemember that enables the medical provider to use their time effectively with the servicemember. These AI capabilities can operate around the clock, much like a “Check Engine Light” in your vehicle, continually scanning medical records to alert medical providers when an intervention (checkup) is needed.
The MERIT tools mark a foundational step in delivering game-changing mission outcomes that address the National Defense Strategy and improve readiness by delivering AI in healthcare. Trading analytics for time, servicemembers and their providers are afforded more flexibility, gaining efficiencies in business operations. Using historical data, the team has already built and trained AI models to scan records for eight of these conditions identified by DoD. These include asthma, arthritis, limited motion in arm and leg, residuals of traumatic brain injury, intervertebral disc syndrome, and major depression. In time, the number of conditions will be expanded to the 20 conditions seen in more than 90 percent of servicemembers with disability ratings.
“MERIT does not make any clinical decisions,” Michelle Padgett of the JAIC’s Warfighter Health Team notes, explaining that the tools and the information they provide will only be used by medical providers and deployment availability boards in a decision-support capacity to identify “exactly who would benefit from early intervention and treatment” and potentially be more proactive to “limit these types of injuries in the future which may extend servicemembers’ careers and improve medical readiness.”
Impact of AI
Among the eventual benefits of a successful MERIT project are:
- The ability to make data-informed accession and retention decisions.
- Early identification and personalized healthcare for service members.
- The ability to retrain service members into new roles before they suffer career ending medical conditions.
- Improved force readiness through broader AI development.
- Defense against our adversaries using weaponized AI (in humans) in the future.
Future MERIT Use Cases
Although the MERIT project is still in the prototype phase, it has the potential to be used for a variety of use cases. The AI models will evolve from “post diagnosis” to “moving left of diagnosis” to monitor behaviors, exposures, and other data, gaining valuable insights about servicemember health and how to protect it.
For example, MERIT AI predictive models will provide a commander with the very latest information on force health and readiness. “A quick look at the MERIT dashboard could let the commander know, for example: Your unit is currently 85 percent ready, and in three months, with no major staff changes, it’s going to be at 90 percent, or in six months, it will drop to 75 percent,” Padgett explains. “With that information, a commander would have the opportunity to make strategic, data-informed decisions.”
The MERIT dashboard could also alert the commander to overarching health issues that have the potential to negatively impact unit performance. “Imagine if a commander was alerted that individual(s) within the unit are at risk for severe musculoskeletal issues,” Padgett explains. “Well, perhaps it’s the way you’re training. So, do you embed a physical therapist to figure out why and what could be done to reduce the risk of injuries or allow technology to assist? Capabilities such as MERIT simply change the way we look at a situation before it becomes a problem—enabling us to be more prevention-focused.”
There is so much potential with MERIT. “These capabilities could be integrated with GENESIS (the Military Health System’s electronic health record), the Department of Veterans Affairs’ disability rating system, and the insurance industry’s ICD-9 and ICD-10 codes, as examples, to flag potentially disabling conditions earlier in the process and highlight the most effective treatment pathways,” Padgett notes. For example, with the algorithm running in the background, scanning the individual’s records and performing comparative analysis, an alert may allow the provider to tell the patient: “The average length of recovery for your age, size, weight, and condition…a 28-year-old female with musculoskeletal injury, is X number of days, and 92 percent of people with this condition have been referred to the Disability Evaluation System.” With the AI models providing more information and options, the servicemember can take control and make informed decisions, Padgett explains.
Looking Ahead and Charting the Way
The Warfighter Health team continues to test and evolve its MERIT AI models. The next step will be to field-test the models in a real-world environment with servicemembers.
“Having dashboards that show projections of the non-deployable population from a health perspective is the first step toward a common operating picture. Current provider, commander, and senior leader viewpoints will evolve from standalone views to complex perspectives with roll-up (summary) and drill-down (individual) capabilities leveraging near real-time data,” Padgett explains. “These views will represent what force readiness projections may look like today and then 30, 60, 90, and even 365 days from now.”
Moving toward enabling more widespread use of capabilities such as MERIT will require the DoD to address challenges and barriers beyond the technical. The greatest challenge for researchers and developers, according to Padgett, is access to data, with longstanding DoD policies and processes making it difficult to share and access medical information of servicemembers, even when done anonymously. “Data science and AI technology are proven. Technology is not where we struggle…it is the easier part of what we do,” she says. “To utilize AI models, process data at scale, and acquire interconnected, high-performing systems will necessitate an enterprise-wide joint solution.”
Padgett concludes: “We have an opportunity to change the landscape for healthcare delivery. This requires an evolution from current program and product-thinking to an ecosystem of shared data and services. Disruptive innovation and AI in healthcare presents a wholesale shift from where we are today, benefiting health and our nation.”
Join The JAIC on Twitter