AI Hospital Software Knows Who’s Going to Fall
(Bloomberg Businessweek) -- El Camino Hospital, located in the heart of Silicon Valley, has a problem. Its nurses, tending to patients amid a chorus of machines, monitors, and devices, are only human. One missed signal from, say, a call light—the bedside button patients press when they need help—could set in motion a chain of actions that end in a fall. “As fast as we all run to these bed alarms, sometimes we can’t get there in time,” says Cheryl Reinking, chief nursing officer at El Camino.
Falls are dangerous and costly. According to the Department of Health and Human Services’ Agency for Healthcare Research and Quality, 700,000 to 1 million hospitalized patients fall each year. More than one-third of those falls result in injuries, including fractures and head trauma. The average cost per patient for an injury caused by a single fall is more than $30,000, according to the Centers for Disease Control and Prevention. In 2015, medical costs for falls in the U.S. totaled more than $50 billion.
Like most other U.S. hospitals, El Camino had invested time and money in fall prevention efforts, such as the call lights, but the various methods hadn’t been effective enough. The parameters for at-risk patients are wide enough that many are tagged as likely to fall at some point. It’s even harder if a hospital has a bigger share of high-risk patients as El Camino does—about 50 percent of its patients are at risk for falls. Effectively monitoring that many people can be tough when nurses are already overworked.
Four years ago, El Camino turned to a health-care technology startup called Qventus Inc., based a few miles away in Mountain View, Calif., to help it prevent falls. The hospital had worked with Qventus the year before to devise a better system of scheduling Cesarean sections. The company created software that would predict the number of women coming in for the surgery to ensure there were enough rooms.
Qventus Chief Executive Officer Mudit Garg and his co-founders, Brent Newhouse and Ian Christopher, quickly began developing a program that predicts falls resulting from what’s known as alarm fatigue—when clinicians experience sensory overload from the many hospital sounds and alerts, leading them to sometimes miss critical alarms altogether. “If I tell you everything is important, nothing is important,” says Garg. “You’re applying the same level of focus to everything.”
Qventus came up with a program that extracts and analyzes data from call lights, bed alarms, and electronic medical records. It also pulled in other information such as a patient’s age, the medication he’s on and when it was last administered, and the vitals last recorded by a nurse. Analysis of the data exposed patterns, such as the time of day when most falls occur or the sequence of events that typically lead to falls. For example, patients who have changed rooms are especially vulnerable.
“I can’t tell you how many late nights we spent trying to figure out how we could, in a mildly usable form, get the call light data” and then write some code to analyze it, says Newhouse.
From the data, Qventus identified several fall indicators used to predict which patients need more monitoring. If a patient meets all the indicators, an alert is sent to a special badge worn by nurses—a “nudge,” as Qventus calls it, reminding them to check on the patient within the next 12 hours. “In the long run, it should cut down on those bed alarms, because they’re intervening earlier,” says Reinking.
Garg, Newhouse, and Christopher founded the company in 2012 to use artificial intelligence and machine learning to help hospitals improve operations. Christopher is a software engineer and Qventus’s chief technology officer. Garg and Newhouse met at McKinsey & Co. in 2008 when both consulted on health-care projects, helping hospitals to become more efficient. Observing doctors and clinicians in hospitals, they say they witnessed a lot of fumbles. “That’s where the focus on the process problems, rather than the clinical problems, came from,” Garg says. They realized, he says, that “data would be a big part of what would anticipate, help get ahead” of problems.
That focus appealed to Steve Kraus, a partner at Bessemer Venture Partners, a fund focused on health-care and technology startups. Machine learning- and artificial intelligence-based health-care solutions have “reached peak hype cycle,” he says. Qventus is a standout, because it’s going after what he calls the “low-hanging fruit in health care, the operational, workflow side,” which is easier to change. In May, Bessemer, along with Norwest Venture Partners, Mayfield Fund, and NewYork-Presbyterian Hospital, invested $30 million in Qventus. Since its founding, the company has raised $43 million.
The fresh cash will enable Qventus to expand to more hospitals, Garg says. The company’s fall prediction platform already is in use in health-care facilities across the country, but Qventus declined to disclose how many. At MedStar Montgomery hospital in Olney, Md., the program has led to a 13.5 percent reduction in patient falls since 2016, according to Qventus.
At El Camino, where the software has been installed since 2014, head nurse Reinking says it took some convincing to get the staff to adapt new procedures. But overall, it wasn’t too hard a sell, given how persuasive the fall prevention results have been. Since 2014, nurses have seen a 29 percent drop in falls. “Once we were able to demonstrate the value of the technology,” Reinking says, “people kind of began to come around.”
To contact the editor responsible for this story: Dimitra Kessenides at firstname.lastname@example.org
©2018 Bloomberg L.P.