In 2015, when Denver-based UCHealth created a Virtual Health Center (VHC) to support smaller community hospitals being added to its system, there were some use cases that were obvious, such as centralized telemetry and a virtual intensive care unit. But UCHealth executives also foresaw that more use cases would develop that weren’t initially apparent. In 2018, clinicians began work on a telemedicine-based sepsis detection and response system, using artificial intelligence to implement a nurse-driven and physician-supported care bundle that has led to a 30 percent decrease in sepsis mortality.
For the impact this homegrown solution has had, the UCHealth team has been named the first-place winner in Healthcare Innovation’s Innovator Awards program for 2021.
Timely and effective care for sepsis, including adherence to evidence-based guidelines, continues to be a challenge and priority for every health system. Upon reviewing their sepsis data, UCHealth executives observed a significant variation in their EHR workflows and outcomes. Best practice alerts would fire frequently with the intent to aid in sepsis identification, but that would take valuable time away from health professionals.
In an effort to reduce that burden and enhance sepsis outcomes, UCHealth designed and implemented a nursing-driven workflow focused on sepsis treatment, using both the local primary nurse and remote critical care nurses. The VHC’s remote teams investigate warnings generated by alerts. Acknowledging the ever-increasing demands on nurses—including increased prevalence of best practice alerts, clinical alert systems and alarm fatigue—the VHC provides remote surveillance for both deterioration and sepsis care in real time.
Besides the 30 percent reduction in mortality, UCHealth also has observed a 59-minute reduction in time obtaining blood cultures, a 33-minute reduction in time administering antibiotics and a 63-minute reduction in time administering fluids.
The project began with a 2018 conversation between Amy Hassell, B.S.N., R.N., director of patient services for the VHC, and the chief quality officer at UCHealth’s academic medical center campus, which was struggling to meet its CMS sepsis metrics. “I said I think we have a way to collaborate with you,” recalls Hassell. “Our challenges are not unique, but we have a lot of novice nursing staff and/or busy nursing staff, and they don’t see a lot of sepsis in the acute care setting,” she explains. “So the thought was that with our attending physician here and the surveillance tools that we were starting to develop, including the algorithms, we could develop an augmentation process for front-line staff where we assist in identifying sepsis in collaborating with them. We can help carry the clinical burden of detection, ordering labs and saying it looks like now is the time to act.”
The remote center uses algorithms as well as two simple physiologic scoring tools, based on data from an acute care wearable device. The wearable allows for the continuous monitoring of vital signs and has the ability to ingest pre-validated data. Experienced critical care clinicians are able to use this multimodal alert system to provide earlier care locally. Per CMS bundle guidelines, if the initial workup results show signs of abnormalities, the remote center physician works in collaboration with the local physician to deliver care accordingly. If needed, the remote center physician is authorized to initiate the sepsis bundle in lieu of the bedside care team.
Using a phased approach, all UCHealth acute and progressive care units implemented the virtual sepsis program by May 2019. The rollout continued across the system, with hospitals in southern Colorado going live in October 2019 and hospitals in northern Colorado going live in January 2020. By March 2020, all UCHealth hospitals were enrolled—with 1,250 acute care inpatient beds available for sepsis monitoring.
Driving care upstream
“We had tried a lot of things in the decentralized operational construct. We were using alerts to the bedside and had an alarm fatigue problem, and we initiated this sepsis alert program to get resources quickly to the bedside, but that was all decentralized,” says Christopher Davis, M.D., emergency medicine physician and medical director for virtual health at UCHealth. “We were still having these outlier cases squeeze through, and they were dragging our performance down as a whole. We realized we have this team at the VHC that is able to add some redundancy, as a type of safety net, and squeeze out those outliers and drive care upstream. So not only are we pushing the care upstream in terms of early identification, but we also have a process to actually intervene if the bedside team isn't available. And that's where the magic comes in.”
Among the tools used are Epic’s sepsis index and deterioration index. They also use a shock index and a respiratory distress index, which is something developed internally at UCHealth. “That’s useful to get upstream, because often the nurses haven't had a chance to even validate that data,” Hassell explains, “but we're seeing that the patient is tachycardic or hypotensive, and we can act on that and start to get ahead of identifying it.”
Davis notes that although they are leveraging Epic’s model, there is added complexity around pulling in demographic information from the EHR — age and past history — to inform the risk model.
He also says that the familiarity of dealing with sepsis regularly improves performance. “Even though there's a sepsis bundle, it's enormously complicated. There are something like 140 discrete tasks to accomplish to get somebody through the six-hour sepsis treatment window when we're following the whole bundle,” he explains. “If you're in a relatively quiet community hospital that doesn't necessarily see a lot of sepsis, you're just not doing it day in and day out. But the VHC team sees 15 to 20 sepsis alerts a day, so they become expert at it and are familiar with the nuances, and that's key. So having a virtualized checklist to handle that complexity is a big value-added.”
Katherine Howell, R.N., chief nursing executive for UCHealth and chief nursing officer for University of Colorado Hospital, says that change management was key to successful roll-out. “Definitely at the beginning, it was challenging, because nurses were used to their own resources and a little more control,” she says, “but through Amy's leadership, and really spending lots of time with those nursing directors, and problem-solving with them, now the Virtual Health Center is just part of the care team. The benefit the nurses see at the bedside from the assistance they get, especially with these deterioration cases like sepsis, has made a huge difference.”
Steve Hess, UCHealth’s chief information officer, notes that the health system’s IT platform was purposefully built to enable this type of innovation in virtual health. “When we set up the Virtual Health Center years ago, we built it on top of the enterprise electronic health record from Epic to make sure that we are doing things in a consistent way both at the point of care and remotely,” he says. “That required EHR integration of some new tools to allow for the remote monitoring of patients.”
In 2015, UCHealth was planning to build several new community hospitals. “We leveraged that need for telemetry and ICU coverage as part of that business plan,” Hess adds. Rather than adding decentralized local resources at those hospitals, they used that as an opportunity to create the VHC and have centralized resources to provide services such as cardiac monitoring and ICU coverage. “We used the building of the new community hospitals as the reason to stand up this virtual health center capability from a cost-neutrality perspective,” he says. “We also knew that those services, once stood up, would go on to serve the entire footprint of UCHealth, which is now 12 hospitals and hundreds of ambulatory clinics.”
‘The single hardest thing to do’
A culture of innovation has been nurtured at UCHealth. “When we set up the Virtual Health Center, we knew that we were going to have to develop areas of competency to enable us to do things that had not yet been conceived of when these technologies were first envisioned — much like our use of smartphones now is fundamentally different than it was in 2012,” says Richard Zane M.D., UCHealth’s chief innovation officer. “The sepsis work was interesting for a lot of reasons, but mainly because it was the single hardest thing to do. We thought that if we could address sepsis, then there are lots of other things we can do.”
Zane describes his job as chief innovation officer as helping identify early adopters, making them successful, and celebrating their wins, so that others will want to emulate them — and that's what happened with the sepsis project. “But that required being able to demonstrate a certain level of competence to the people at the bedside,” he adds. “We were able to demonstrate that we were accurate, and in our first three months we eliminated 180,000 bedside alarms.”
“We were able to show rather quickly that we reduced mortality by 30 percent. And when we presented this to our board, we didn't just say 30 percent, we said this number of people didn't die because we did this,” Zane says. “Then you earn credibility to do other things.”
The Colorado health system is looking at ways to move the sepsis surveillance a little further upstream. One example is if a patient is in the emergency department and has been admitted to the hospital but hasn’t been moved to a room yet. The VHC might bridge that gap in the transition of care, Davis says. “Once they're admitted, and if they have a sepsis signal, then our processes can be brought to bear.”
UCHealth is building on the work it has begun on surveillance. “We're looking at deterioration because we can recognize when a patient on the floor is becoming ill, even though they don't meet a certain specific criteria of a specific disease,” Zane explains. “What we can do is recognize when a chronically ill patient is deteriorating before they're actually symptomatic. Often someone with heart failure or renal failure will get sicker without feeling sick, so we can intervene and prevent a hospitalization or an emergency department admission. That is the spectrum of care that we're going after.”
“We started with sepsis because there is such a discrete playbook,” Davis says. Now we're sort of knocking on the door for other types of deterioration. How do you intervene if it's respiratory, or from blood loss in a surgical patient? Sepsis had to come first —to validate the model. And now we have the ammunition to go back to local teams and start this whole change management cycle again.”
People might ask where the return on investment is around this VHC infrastructure. “It’s true that standing up all of this infrastructure just for virtual sepsis doesn't work,” Davis says. “You need that leadership buy-in to say this infrastructure is going to solve three or four different problems, and amortize that across a whole bunch of different service lines. That's key.”
For other health systems that might want to emulate their work, Hassell recommends making sure that you design a process that's going to be actionable and augmentative and not replacing the front-line clinicians. “It has to work in tandem with them,” she says, “and on a problem that they are experiencing and that they want to solve.”