After launching a multi-campus phased rollout of the predictive insulin dosing software, Philadelphia-based Temple Health’s University Hospital has reduced hypoglycemia rates two- to threefold.
The tool — called Endo Tool Sub-Q — enables real-time blood sugar monitoring, allowing care teams to respond more proactively, according to an Aug. 15 news release from the health system.
The rollout was led by an integrated team of Temple Health leaders: Chaudron Carter Short, PhD, EdD, MSN, executive vice president and chief nurse executive; Carl Sirio, MD, chief medical officer; Benjamin Slovis, MD, chief medical information officer; Joy Weaver, DNP, MSN, RN, senior clinical performance excellence manager; and David Fleece, MD, former chief medical information officer.
The team spoke to Becker’s about the strategy behind the rollout as well as the outcomes they have seen.
Editor’s note: Responses have been lightly edited for spelling, grammar and punctuation only.
Question: What were the biggest clinical challenges Temple Health faced with diabetes management before EndoTool?
Dr. David Fleece: We were looking for a way to improve our metrics around hyperglycemia and hypoglycemia.
Part of implementing a new way to manage subcutaneous insulin was a lot of operational things, like the timing of the meal trays and assessing how many carbs the patient is going to eat before we give them their insulin. That was a big sea change for us. Instead of reacting to patients’ blood sugars after their meals or dosing them on what their sugar was six hours ago, we’re now anticipating what they’re about to eat and trying to proactively manage their sugars going forward from this meal.
Q: What makes this system a leap forward compared to past approaches to diabetes management?
Dr. Joy Weaver: One thing that is so important is the resolution from a hypoglycemic event time. We went from 120 to 130 minutes from when a patient had an under-70 hypoglycemic event to 75 minutes.
When a patient is starting to decline in their hypoglycemia, the tool says, “Nurse, give juice,” and all of a sudden, we don’t have a hypoglycemic event. That was one of the things that really stuck out.
Dr. Carl Sirio: Through this process, both IV and subcutaneous, we’ve gotten to better control and fewer episodes by a factor of several times.
At one hospital, we’ve reduced hypoglycemia ninefold. At the Temple main hospital, the main University Hospital, it has been two- to threefold.
Q: Why did Temple choose a phased, campus-by-campus rollout instead of a systemwide launch? What lessons can other systems learn from the rollout?
CS: This is a really big lift. It requires a whole bunch of coordinated structural change, lots of education and training, and hands-on at the bedside. When you’ve got a huge cultural change like this, it’s more likely to succeed if you focus on smaller units. It was a logistical decision around the complexity of the rollout.
DF: The nursing lift was probably 80% of the lift, but there was a big lift on the provider end as well. When these patients are coming into the hospital, they’re often pretty sick with a variety of problems that aren’t just diabetes. Blood sugar management often would be second or third tier in the provider’s mind.
We found that we really had to push that up higher in their priorities to get them to spend a little more time in their admission efforts accounting for insulin and blood sugar than maybe they had done historically.
Q: How did you approach building staff buy-in, particularly around shifting long-standing clinical behaviors? What role did ‘Super Users’ play in the rollout?
Dr. Chaudron Carter Short: The buy-in came from the fact that we were able to show the data pretty early. It’s hard to dispute data. When the staff really saw that the data was helping our patients in a safe manner, that got us going. We had the super users who were important to have as our frontline team, because they knew the workflow. They were doing it every day, and were able to present challenges and opportunities for change, as well as some clear answers to those challenges.
JW: One thing about the super user program that we used for operational readiness is that the goal we started with at the beginning, at every campus, was making sure that everybody agreed the super users would get extra time, support and validation. That was a big reason our super users were so prepared for Go Live.
Q: How is the Power BI glycemic dashboard changing the way clinicians monitor and respond to blood sugar trends in real time? How do you view this result in the broader context of patient safety and hospital quality metrics?
JW: The Power BI glycemic dashboard was actually Temple’s first biggest dashboard — and still is the biggest dashboard we have. At the time, our analyst department was just starting to grow. So this was a really good opportunity to build a strong relationship with that team and build this.
What we did was take retrospective data and make it real time up to the night before. That was new for Temple and was a great asset.
My goal when I helped to build that was to ask: What would a nurse or a doctor want to see? How could we do it so they could see that easily in real time? That was the foundation. We built off our current state, made it real time, and then as people got involved and looked at the data, they wanted to see more and have more real-time access.
Q: Looking ahead, what does the future of innovation look like at Temple Health?
Dr. Benjamin Slovis: From this rollout, we’ve established a well-known point: End users don’t want to leave the electronic health record. As we develop integrated tools, we want to make sure they are as streamlined as possible into the workflow, and that the real-life workflow dictates the electronic workflow, not vice versa.
We’re also going to continue to leverage AI and machine learning tools in the future in a meaningful way. We’re going to look for specific use cases and identify tools that will be impactful in either improving efficiency or improving quality.
To that end, we are already leveraging ambient voice technology to improve provider EHR usability. We’re looking at sepsis predictive models and a number of other generative tools. But we are trying to be very thoughtful in our approach in order to solve specific problems, not chase shiny objects.
