Algorithms and bots are already in hospitals—and they’re doing more than you think.
When the dust settles, the 2017-2018 flu season will go down as the worst in at least a decade, by several measures, according to the CDC. With countless hospital visits for flu-like symptoms and the number of states battling widespread bouts of influenza—the continental 48, plus Alaska and various territories such as Puerto Rico—Americans are battling a challenging foe that can be deadly.
Public health officials, clinicians, and hospital executives are charged with responding to scattered outbreaks. Part of what makes the flu a difficult puzzle to solve each year is the level of required human input. “Everybody who works in primary care over the last couple of months probably encountered the same issues,” said Michael Cantor, MD, associate professor in the Departments of Population Health and Medicine at New York University Langone Health. “Your office is getting flooded with people with upper respiratory infection, who, because of what’s going on in the news, are worried. It’s hard to triage those patients over the phone, and you end up seeing the doctor for something that’s relatively benign.”
The effects of incorrect diagnoses and time invested in patients who don’t need medical intervention can be severe. The most vulnerable patients get drowned out by a chorus of coughing fits. On the macro level, overwhelmed hospital systems find day-to-day operations challenging, if not impossible. In Alabama, for example, health systems got so clogged that Republican Gov. Kay Ivey declared a public health emergency in January, noting that hospitals were “taxed to such an extent that care of patients may now no longer be provided in the traditional, normal, and customary manner.”
But help is arriving, and some is already here. More and more, state-of-the-art diagnostics are being piloted, many using hyper-advanced analytic and artificial intelligence (AI) technologies. But would simply knowing who has the flu and who has a severe cold be enough to save an Alabama hospital system overwhelmed during influenza’s witching hours? Probably not: There is a laundry list of labor-intensive applications carried out in a hospital on a day-to-day basis, many of which rely on human input, which means they may not be running efficiently.
Fortunately, many corners of healthcare are due for an AI-infused upgrade, from pharmaceutical delivery systems to surgery scheduling and even technologies that save doctors time on transcribing patient notes. These innovations will have different looks and applications, but if successful, they will share a common outcome. “The hope is that the field will be more orderly,” Cantor said. “If you’re using these AI systems to help with workflows, there will be more balance. There won’t be [as much] chaos.
“There [are] always going to be emergencies, but the day-to-day operations will be a lot smoother.”
Branded the “hospital of Silicon Valley” and a 5-mile drive from Apple’s headquarters in Cupertino, California, El Camino Hospital is on the cutting edge of many new forms of medical technology. It’s developing AI systems that predict which patients are at risk of falling and robotic delivery modules that hum across hospital corridors. “We are really interested in innovation here,” said Chief Information Officer Deborah Muro, “especially because we’re in the middle of Silicon Valley.” Many of her efforts remain in their infancy, but she made the hospital’s goal clear: “We’re really looking at developing a hospital room of the future, a clinic room of the future.”
Some of the new tech that El Camino is pursuing falls outside the umbrella of AI and analytics. The hospital is partnering with Apple, for instance, to integrate iOS devices into the hospital’s information systems, using customized apps to streamline communications between doctors, nurses, and patients.
But AI systems are slowly rolling out. The hospital recently partnered with Qventus, an AI-based operational decision management platform, to pilot software that attempts to predict patients at risk of falling out of bed. The software uses an algorithm that analyzes 3 risk factors: patients’ medical histories, communication patterns with nurses, and how patients move in bed. El Camino then allocates nursing power accordingly.
So far, the data show, the tech has helped the hospital reduce its fall numbers by 30%.
El Camino and Qventus are also beginning to implement similar programs that identify patients at risk for sepsis, which kills roughly 250,000 Americans per year, and those who are prime for readmission. The systems take into account different factors but are all considered “predictive,” which differs from the “preventive” models of understanding healthcare that have been in place for decades, Muro said. “That’s going to be the future, being able to predict healthcare scenarios for patients rather than waiting for them to come to us when they have a healthcare event,” she added.
El Camino has also pointed its analytical gaze inward to increase efficiency. Muro described a smart surgical system being installed that alerts doctors and surgery teams to open time slots. This technology could help patients undergo surgeries more quickly, and the hospital could then schedule more surgeries. There is no time wasted on human error.
Alongside clinicians, robots putter around El Camino, tackling tasks that are key to care delivery. The machines range in size. Some are as tall as 3 feet and as wide as 4. Visitors and patients love them. The robots aren’t yet “predictive”—that is, they only deliver goods for nurses and doctors, without much adaptive thought.
But these bots could eventually become vital to a much larger AI-driven pharmaceutical branch in hospitals, thanks in part to technology being pioneered by organizations like Omnicell. “The big driving force for us is to automate what can be done by a machine and let the pharmacists focus on patient care,” said Michael Guidry, the company’s senior director and central pharmacy portfolio manager.
El Camino’s Silicon Valley neighbor, Omnicell develops automated medical technology. One of its models, the XR2, looks to free up time for pharmacists by automating the tracking, sorting, and dispensing of drugs within hospitals. Unlike El Camino’s roaming bots, the XR2 is massive and stationary. Customizable to fit a hospital’s needs and inventory, the robots are 12 feet wide and range in length from 12 to 30 feet, growing in 6-foot increments. From the outside, an XR2 resembles a series of lockers at a gym or a public pool, with a touch screen interface on one end and a doorway inside. A thin, lime-green light hugs the top.
Inside, the XR2 houses a library of pharmaceuticals (Omnicell envisions a future in which partner hospitals put their entire inventory into the XR2 machines), with a series of robotic limbs and apparatuses documenting, scanning, sorting, and distributing the drugs. For years, some combination of these tasks has been automated by similar technologies, such as the XR2’s predecessor, ROBOT-Rx. But what makes XR2 unique is its ability to work within a hospital’s communication system to track and sort inventory and then package medications to pharmacists, who hand them off to patients.
If you’re a pharmacist, Guidry said, “your workday really revolves around all these logistical tasks: pulling inventory, counting inventory, getting it ready for where it needs to go.” Tasks, Guidry said, that limit more creative endeavors the pharmacy might want to undertake. The XR2 eliminates these bits of paper-based, time-consuming labor inherent in traditional hospital pharmacies, allowing pharmacists to spend more time with patients and mitigating the risk of human error.
Another perk: The robots aren’t inhibited by human working hours. They fill orders 24 hours a day, which bolsters anticipatory order filling and accommodates late-night patients.
Five hospitals across America have adopted the XR2, in Illinois, Ohio, Pennsylvania, and Virginia, with more systems rolling out in the summer and fall.
The XR2 isn’t the only link in the pharmaceutical supply chain prime for AI. Various pharmaceutical companies the world over are partnering with advanced AI systems to develop new pharmaceutical cures for diseases.
“High-performance computing is becoming more and more available.…Having that kind of power to predict changes to the different types of flu will help [develop better pharmaceuticals],” said Cantor, of New York University Langone Health, who previously worked for Pfizer while it was planning a partnership with IBM Watson to develop predictive pharmaceuticals.
Because drug development involves a lot of trial and error, advanced AI systems like Watson can expedite the search for success. If you ask Cantor, the ideal manifestation of this tech is an algorithm: You plug in, say, a virus, and the computer spits out chemical cocktails known to weaken it. And the tech can be used for much more than flu vaccines.
Outside patient observation and smart pharmaceuticals, AI can help doctors by cutting back on important but menial tasks.
“My dream would be: I spend 20 minutes talking to a patient, and there is some system recording what we’re talking about,” Cantor said. “It can turn my speech in our conversation, and my electronic stethoscope that can hear heart sounds and breath sounds, all into a well-documented note. And so it saves us from being our own secretaries.”
Forms of this do-it-all scribe technology that Cantor is referring to are in development. One such project is being developed by Google AI in conjunction with Stanford University. “Speech recognition is just starting to get good enough that we can tailor it to handle really tricky cases like scribing a doctor’s visit,” Jason Freidenfelds, a senior communications manager at Google, said via email. “We’re really excited for this project because we want patients to have more focused face-to-face time with their doctors.”
A Google Research Blog post from last fall went into more detail about what problems they aim to address, noting that doctors “often spend about 6 hours in an 11-hour workday” on documentation.
Even if those numbers are liberal, the benefits of a doctor saving a few hours a day should be felt, as Cantor and Freidenfelds said, by patients.
AI in medicine does not come without its drawbacks. The more automation comes into play, the more hospitals will have to pay attention to cybersecurity, Guidry said. As in other fields, job loss will affect certain corners of healthcare. (Cantor noted that radiologists and other doctors tasked with analyzing medical imagery are most likely to feel this). And implementing and perfecting the new technology is expensive and time-consuming, Muro pointed out.
But if you think about what a hospital might look like in 30 years, you might be able to see how Alabama could have avoided its flu-induced state of emergency—and how AI goes far beyond diagnostics, despite those uses’ success in grabbing headlines.
Smart flu vaccines help insulate more patients from the flu, and smart pharmaceuticals help ease the burden of those patients still affected by it. Automated delivery systems free up pharmacists and nurses to tend to more patients. Automated scheduling ensures that no hospital room or doctor shift is put to waste, and transcription devices minimize the clerical work that bogs down orderlies.
These developments will not eliminate public health crises. But they will give professionals more tools and more time to fight them.