From predictive analytics (PA) to algorithms to blockchain technology—they can all improve the lives and the medical care of our warfighters
By Michael Embrich
Picture this. From his smartphone, a veteran logs onto his account on VA.gov to file a claim for hearing loss. The VA’s algorithm initiates a case file for the veteran. The case file begins with a questionnaire. The first question is, “Are you or are you at risk of becoming homeless?” The veteran is homeless, and he has been living in his car for a few weeks. So the veteran answers ‘Yes.’ Except, instead of flagging the veteran’s case file and placing it in a long queue of other veterans’ case files who are similarly at risk, like it does now, the algorithm does something new. This new algorithm geolocates the veteran’s smartphone, pings hosts on a rental platform like Airbnb in that geographical location, and negotiates a price with a host or renter, either by bid or a pre-negotiated rate that abides by federal government regulations under a new End Veteran Homelessness program. The veteran can then stay with one host, or multiple hosts, until he can be placed by a caseworker in permanent housing. This eliminates the long wait for the case file to be reviewed at an unknown date by an unknown caseworker.
By the time the veteran is done filing his claim, a prompt asks if he needs a ride (via a ride-share platform like Uber or Lyft) to his temporary lodging. These prices are also negotiated with the VA by bid or federally approved pre-negotiated price under a new Rides for Vets program. If the veteran selects, “No, I have a ride,” the algorithm then sends the address to the veteran’s Google Maps or Waze app, highlighting the route. The algorithm also automatically places a voucher for one free meal at McDonald’s or any other of the thousands of restaurants that already partner with the VA on multiple fronts. The app asks the veteran if he wants to stop at the closest McDonald’s on his way to his host.
In real-time, the veteran just filed for medical benefits, put in a disability claim, booked an appointment with the local VA for care, and beat homelessness and hunger—all the while never actually dealing with a doctor, caseworker, landlord, federal employee, or taxi driver.
As Director of a homeless veterans program in New York City and someone who worked on the incredibly effective “functional zero” initiative to end veteran homelessness (which involved the Obama White House, New York City, the VA, and the United States Department of Housing and Urban Development), I can tell you it would save the federal government and other veterans service organizations (VSOs) millions of dollars. The VA is plagued by wait-time problems, among others, and a lack of resources to fix such problems. All the technologies I mentioned above exist, and if you use your imagination, you can probably think of a few other ways technology could help fix the VA.
Many of these technologies rely on predictive analytics (PA) and other advanced algorithms, also referred to as machine learning, or deep learning, and the required hardware and infrastructure to make it all work.
The Ford Mark 1A Fire Control Computer was a computer that helped the US Navy destroy and cripple Japanese ships during World War II. Using a primitive computer, the Mark 1 automatically calculated the lead angles to the future position of a target at the end of the shell’s time of flight, adding in corrections for gravity, relative wind, and other effects that projectiles experience during flight. This was a major advantage for the U.S. Navy because the Japanese did not have this technology, and it helped us win the war in the Pacific.
By analyzing data, the Mark 1 was learning about its target and predicting where the target would be by the time the shell reached its destination. These predictions were set in parameters that were created by the gunner’s mates who operated the weapon. Today those parameters have grown so large that only computers can set them. This new form of machine learning is commonly referred to by data scientists as predictive analyticsor PA for short. PA is a multi-billion dollar industry that you interact with every single day. Have you ever opened Facebook, or went on your favorite website only to see an advertisement for something you were just thinking about or discussing with a family member or friend? That is predictive analytics in practice, and in addition to facilitating Black Friday sales, it is currently being used to save lives around the world.
Scientists are already using predictive analytics to do everything from identifying molecular patterns of cancer to finding diamonds in places we never have looked before. These technologies are incredibly powerful tools that have only been limited by the inability of modern computers to process and to calculate the complex mathematical models it takes to be precise in its predictions. That is, until this past fall. Google announced “quantum supremacy,” meaning one or more of its computers has completed quantum computations much more quickly than the best current-day supercomputers in the world, like IBM’s Watson.
Many of those computers used by healthcare networks across the globe—computers that are currently solving great challenges that doctors, hospitals, and others face in providing care to a vast number of patients, all the while collecting data and analyzing patient outcomes—use predictive analytics to improve the system across the network. Therefore, the idea of using predictive analytics to manage administrative and clinical problems isn’t revolutionary.
Actionable insights are direct, meaningful actions that are taken as the result of analysis of raw data collected by the VA. The VA has recently created an Office of Artificial Intelligence as part of President Trump’s America AI Initiative. Presumably designed to explore how VA patient data can be used by AI to create actionable insights, the details on the program are scant at best. Artificial Intelligence could be used for a lot of things, but enhancing telemedicine to reach VA patients in rural areas where VA-friendly doctors or VA facilities don’t exist seems like the best use for this technology. This office was just created and is in the very early phases of development, a story we often hearfrom federal agencies. This is federal-speak for maybe full implementation will come in 5-10 years. Nevertheless, this technology is being used now by many companies here in the United States.
AI could be used to assess patients, diagnose and prescribe medicine for minor ailments, as well as flag patients for further intervention or treatment by a real doctor; it can find the appropriate doctor, make an appointment, and arrange for the veteran to be taken to that VA-friendly doctor or facility using Uber or Lyft rideshare technology, all in real-time. This would drastically cut wait times and free up precious VA resources for other important and essential tasks that the VA currently struggles with, due to the enormous number of patients with minor medical issues that it deals with daily.
The patient overload sometimes hurts veterans who are seriously injured or have chronic issues, as we have seen in the past, who fall through the cracks due to the volume of patients the VA treats. But a system like the one described above would certainly solve some of the VA’s biggest problems right now.
American digital health companies like Ro and others are currently virtually treating patients for minor ailments by having patients fill out an online survey about their health, lifestyle, medical history, and symptoms. Then they have a real physician review the results and prescribe a treatment or medication if appropriate medications are required. The medication is shipped to the patient (in the future, by drones), and the visit is completed, all online, without travel, appointments, or face-to-face interactions. (Some states require the physician to call the patient to confirm the information received online, a minor blip in the process, that doesn’t really affect the end result or wait times, as these calls can be scheduled by the patient). These types of patient-client interactions are happening every day legally in the United States, yet the VA has consistently found legislative barriers to telemedicine for itself, and it hasn’t asked Congress to fix them. This has been going on for so long that telemedicine, while a good alternative to in-person visits, has largely become irrelevant due to the rise of virtual doctors.
The VA conducted a study on how predictive analytics could be used to assess patient risk. The Care Assessment Needs (CAN) predictive analytics tool under the REACH VET: Applying Predictive Analytics to Clinical Practice (an umbrella program) to prevent Veterans Suicide program. Participants did well under study conditions.
• 79% of patients said the CAN system was easy to use.
• 73% found it helpful.
• 72% said they are confident in their ability to use the CAN score for clinical practice.
• 69% thought the CAN score accurately represents their patient population.
• 68% reported they use the CAN score regularly.
So ramping-up efforts to implement PA at the VA isn’t purely fantasy, and on a controlled level is working quite well to identify veteran patients who are statistically at the highest risk of hospitalization or mortality. Imagine what the VA could do if a program like this was implemented system-wide?
For example, the VA is plagued with recurring hepatitis outbreaks, a virus that veterans suffer from in greater numbers than the general public. Although the VA has made great strides in eradicating the virus in our veteran population, a predictive analytics tool like CAN could easily process VA-collected data to predict when and where the next outbreak might occur and to give VA administrators a ‘heads-up,’ giving them time to free up beds, stockpile vaccines and medications, and reallocate resources to deal with such a health crisis. These tools could also be used to predict things like staffing shortages, machines breakdowns, which hospitals will reach maximum capacity, and even things like when a veteran is at greater risk of suicide(which is already being explored by the VA on a limited basisunder REACH VET). PA could also be used at the VA to assess the need for a number of other interventions that are not currently being explored by the VA, like virtual doctors, as discussed before.
The VA says that this study and its findings on using predictive analytics under the REACH VET program are “preliminary and limited.” In federal-speak, this means that we are years away from any potential implementation, even if they decide to keep using the tool or a VA version of it.
Systems similar to REACH VET are currently being used successfully at hospital networks across the globe and within other agencies within the federal government. A common problem the VA faces in adopting technology from the private sector or other agencies is the constant legislative battles the VA’s own lawyers consistently wage against itself. These bureaucratic hurdles make systematic changes (which are required for such a program to be implemented agency-wide)impossible. Either way, it may be too early to evaluate the impact of the REACH VET initiative, as it was only started in 2018 and it relies on the VA’s own reporting of veteran suicide. The total number of suicides among veterans increased four out of the last five years. Outside experts note that from 2007 to 2017, the rate of suicide among veterans jumped almost 50 percent. And they fear that this year the veteran suicide rate is only getting worse. But without proper metrics and the infrastructure to assess them, the VA won’t be very accurate in tracking and translating this data in the near future.
Blockchain and Records Management
There is the story of the forgotten Filipino World War II veterans, who for 60 years fought the VA for the recognition and the benefits they had earned. The VA simply couldn’t pay them because their names weren’t on some list that made it to the United States after the war. With only a few of them still alive, those forgotten heroes finally received the $15,000 promised to them by the VA, closing another sad chapter in the history of records fiascos at the Department of Defense and the VA.
Consider those forgotten veterans the lucky ones. There are so many more American heroes who were denied VA compensation altogether due to the National Personnel Records Center fire of 1973, in which approximately 16–18 million official military personnel records were lost. Or veterans like my grandfather,who was denied his Purple Heart and VA benefits until his dying day because his doomed mission was classified until the 1990s,when the treasure that was on his sunken ship was found in the Indian Ocean by a team of salvagers.
Luckily today these types of record-keeping snafus can be avoided by using blockchain technology. The same technology that fuels anonymous cryptocurrencies like BitCoin is also being used by bankers and contract attorneys to create bulletproof,binding agreements known as “smart contracts” and other legal documents. Because the contract or legal document exists everywhere on the blockchain, it cannot be forged, stolen, or destroyed anywhere on or off the blockchain. The company Ethereum is leading the industry in these smart-documents and could revolutionize the way the VA and other federal agencies keeps their records.
Office of Personnel Management Implementation
The other looming problem that the VA faces in implementing new and emerging technology systems is the United States Office of Personnel Management (OPM). OPM is an independent agency of the United States Federal Government that manages the government’s civilian workforce. This summer OPM announced that Data Scientist would be a new job title for potential federal employees. But many other federal agencies have already been performing the work of data scientists outside the scope of an official job title. Therefore we can assume that other essential job titles, like data collection specialists, and a number of other tech-centric titles, needed to fill the positions that will drive these technologies can be achieved without an official action from a slow-moving bureaucracy.
Partnering with the Department of Defense and the National Security Agency
The National Security Agency (NSA) and the United States Department of Defense (DOD) have some of the largest intelligence datasets in the world and the computing power to make that data into actionable insights. They have been developing actionable insights and avoiding government regulations in a few different ways. For example, they assume they have the authority to subvert federal regulations and norms that can stifle agency or organizational innovation, all in the name of National Security. If a war were to break out, anyone can tell you that veterans would be the first ones drafted, so veterans affairs issues are also national security Issues—and Congress, the Federal Government, and the White House all acknowledge this fact publically, yet there is no way to float, lend, or mix the National Defense Authorization Act money to VA Agency coffers. In fact, in the last decade the VA and the DOD have worked together to streamline medical records and the transition from soldier to civilian. There is no reason why these initiatives cannot be pushed further using PA and other VA programs. This would fully fund the VA and leave America with a healthy and strong reserve military force.
Who will take action?
Right now, the IRS is using predictive analytics to learn and find people who are cheating on their taxes. This technology is also being used by other non-intelligence federal agencies, so whathas been discussed in this article isn’t exactly revolutionary. The problem lies in the massive bureaucracy that lies between veterans and this technology, which could save and improvetheir lives. It is a common misconception in government that politicians are ‘too old’ to get the implications of PA and these other technologies, but many politicians use these very same technologies in their political campaigns and in their daily lives.
Data scientist, data analyst, computer scientist, IT specialist—all these job titles and others currently exist in the Office of Personnel Management, the federal government’s human resources agency. All that remains is to hire these folks at the VA and let them go to work, doing what they already know how to do. There are also potential partnerships with companies that are lining up to work with the VA, IBM’s Watson, LinkedIn, and others are already working with the VA, but without a strong and robust tech policy and tech infrastructure at the VA, that help isn’t being fully optimized.
It is now up to us to make sure that the technologies mentioned, and others, are implemented in order to provide the best care possible to our veterans. The ancillary benefits to using these technologies are many, and probably require another policy paper to fully explore, but it is safe to assume that billions of U.S. tax dollars, federal work hours, and countless lives will be saved if we act decisively and swiftly to take advantage of all the tools at our disposal to remain the greatest nation in the world—a nation that will always “care for him who shall have borne the battle and for his widow, and his orphan.”