Sanford Health will soon start to schedule its more than 10,000 nurses leveraging an augmented intelligence tool it developed that uses the future instead of the past to predict staffing needs.
The rural health care system partnered with Flexwise to incorporate the algorithm and clinical expertise into the tech company’s workforce optimization software. The platform uses predictive analytics and AI-based technology to schedule the right number of people at the right time by predicting staffing needs out 18 months.
Sanford Health has been piloting its internally developed AI tool since April 2021 in Fargo, North Dakota, and will roll it out to 68 departments on April 15 and then other units. Sanford Health’s development is currently being incorporated into Flexwise’s commercial software product and the co-developed solution will roll out across the entire health system in 2022.
“The tool allows our teams to spend less time on spreadsheets and more time with our patients and their family members as well as help grow and mentor our nursing teams,” said Erica DeBoer, R.N., chief nursing officer.
“We have to be able to anticipate what our staffing needs are so that we can make sure we’re taking care of our patients and our communities. And this benefits our people by making sure our managers, our directors and other leaders can spend time and support our frontline teams.”
LAMP: A nod to Nightingale
Emily Buckingham-Carlson, R.N., head of enterprise clinical staffing and scheduling for Sanford Health, started the effort in 2019. Her job is to optimize Sanford Health’s nursing workforce. She thought real-time information could help identify efficiencies.
“As we’re looking towards the future, how can we use data in a different way, to help understand what our needs are going to be? We all are managing the workforce challenges that the pandemic has prompted,” she said.
“This work has brought to light some of the processes that we’ve had in the past that just really don’t cut it. We have grown exponentially over the last several years. One of the things that this tool is trying to do is stay ahead of our patient care needs and be able to have the nurses where they’re needed when they’re needed.”
She called the AI tool Leveraging Analytics to Mobilize (our workforce) and Prepare (for the future), or LAMP. It honors Florence Nightingale, considered to be the founder of modern nursing because of her work in the 1800s that included checking on wounded soldiers at night with a lamp.
“It’s really a nod to Florence Nightingale. She was a disrupter. She was an innovator. She really challenged the norms of her time to move them into a new space,” Buckingham-Carlson said.
Traditionally, nurse managers also looked to history to fill out schedules by factoring in patient numbers from a year ago. LAMP represents a change from that approach, Buckingham-Carlson said.
“We’ve got to look at every single hour. What are our nurses doing? What is the workload? Understanding it at a very different level we ever have in the past will help us inform the future,” she said.
“That’s where predictions and the projections come in because we’re constantly forward facing. We’re using the information to say what’s next, not what happened to help inform us for tomorrow. It’s a big shift.”
That shift to stabilize the workforce helps counteract the numerous destabilizing factors brought about by the COVID-19 pandemic and other unforeseen challenges.
“We have many different things flying at us that we didn’t have three years ago when we started this. The beauty of this, is that it’s meant to be agile. The tool is built to be something that changes with us over time, and it’s not meant to be stagnant,” Buckingham-Carlson said.
Though data drive the AI tool, frontline staff provide valuable input to guide the algorithm, Buckingham-Carlson said. They can assist in informing how many nurses are needed to treat each type of patient.
“We’re trying to engage with the nursing leaders to really understand their operations in a real way to help build this. One of the benefits, even though it’s new and disruptive, is that we have a lot of engagement,” she said.
LAMP helps nurse leaders plan for the right staffing level well ahead of time instead of reacting, which frees them up to lead and mentor, DeBoer said.
“If we don’t have a way to understand and see at the core what our needs are so we can take care of our patients, we have to spend more time on the phone calling and managing and reacting to situations instead of anticipating what those needs might be,” she said.
“LAMP helps us work smarter.”
The tool already has proven itself in testing, DeBoer said. “We’ve been able to show tried-and-true accuracy.”
Happier nurses (and patients)
Besides making sure there are enough nurses during each shift, the AI tool also will help Sanford Health make better decisions about how many nurses should be hired to each unit. That will help ensure they get the days off they request and can maintain healthy life balance.
“If we plan better and reduce turnover, our nurses are happier. If they can come to a shift and they know what to expect and have people to take care of the folks that we have, we reduce that variance and they’re happier. Then we hope to impact nursing turnover and retention and keep them happy with their schedule and important work they do every day to serve our patients,” Buckingham-Carlson said.
“We’re looking to really support our employees. That supports our patients.”
‘Can only consume so much information’
At the beginning of that effort to improve staff satisfaction and patient care is innovation, which is where Robert Menzie comes in. He’s the lead data scientist on Sanford Health’s advanced analytics group who developed the algorithm.
“I’m the math guy,” he quipped.
Anyone who has spent time on a hospital floor as a patient, family member or visitor has seen firsthand how much health care is staff-intensive. And nurse managers who have worked in the industry for any length of time generally find a right balance between staffing and patient demand, Menzie said.
“Their gut’s right 80 percent of the time. But we can solidify that and help make it more accurate,” he said.
Besides improvements in accuracy, the AI tool eases much of the busywork of scheduling, which frees leaders to spend more time with their employees. Managers still must tweak staff schedules, and ultimately are in control. But the computer model can factor in much more information, Menzie said.
“A lot of people making decisions now have lived through the data, so they are the database, essentially, for their own decisions,” he said.
“But they can only see so much of the data that they’ve collected over the years of experience. Whereas the computer, it doesn’t have short-term memory or long-term memory. It has memory in general. So it can look at everything that’s ever happened and apply the right pieces at the right time. Whereas a person, they can only consume so much information before it gets lost in the background.”
That information includes patient numbers on the same date from the previous year as well as much more detail, some of it going back up to six years.
“It’s not just the hour of the day. It’s not just what day of the week it is. It’s not just the month. It’s not just the year. It’s a culmination of all of those and also what’s currently happening,” Menzie said.
“So if you take an average of last year, you’re going to be short-staffed in some departments and over-staffed in others because last year was COVID. And if you go back two years, you’re still going to be wrong because now we’re having an influx post-COVID because everyone who missed all of their appointments now has to go to the doctor. So we can take into account what’s happening in real time over the last couple of days but then also incorporate what happened last year, two years ago, three years ago, and how does all of that play a role in what’s going to happen tomorrow or six weeks from today?”
Some departments, such as labor and delivery, also could use leading indicators. For example, if that manager knows how many Sanford Health patients will be 40-weeks pregnant in six weeks, they can schedule accordingly.
“They’re not all-telling, but they at least give the computer a chance to find those patterns and implement them and in a much faster way than someone can do manually,” Menzie said.
Prescribe, don’t just predict
Beyond predicting staffing needs, the AI algorithm goes one step further: prescribing what could be done to solve a scheduling challenge.
“We predict months out. But then we also prescribe, which is the scheduling piece. And that prescribe piece is a true optimization,” Menzie said. “Now you don’t have to take the time to find the solution. The solution is provided.”
Another benefit is even longer-term planning. For example, game-planning for how many nurses a new 20-bed unit will need when it opens or what happens to staff levels if turnover decreases or the organization reduces the amount of time to hire a nurse.
“We can finagle all of those switches on and off and it gives the people in charge of managing the operations time to plan for it. So we could tell you if in five years you wanted to open 20 beds they’d be used, rather than taking a swing at the fences and hoping you didn’t waste the investment of resources,” he said.
Menzie said that while he delivered the algorithm, Flexwise “provides a modern software user interface rather than a spreadsheet” for nurse leaders to use when scheduling.
“Flexwise Health was identified to be the best-in-class precision staffing company,” said Braden Bills, the member of Sanford Health’s innovation and commercialization team leading the project. “Flexwise values the Sanford Health partnership because we’re doing it internally, so we can provide extensive clinical, operations and data science feedback. That’s a big chunk of what our collaboration is on.”
Flexwise, based in Nashville, Tennessee, helps organizations optimize scheduling for permanent clinical staffers and also manage a pool of float nurses who can quickly help where they’re needed most.
“Sanford Health was really a market leader in developing this concept of predictive analytics to improve staffing and scheduling activities and had developed an in-house application,” said Kevin Godsey, Flexwise CEO. “They did a proof-of-concept and saw value but determined that to realize it systemwide would take more.”
That’s when the organization engaged Flexwise to integrate the algorithm into its software platform as an “intelligence layer” onto Sanford Health’s existing electronic medical record system and applications that handle nurse scheduling, time and attendance, Godsey said.
The pandemic obviously prompted severe staffing strains and shortages. But several factors already were in play that required a new, predictive approach to nurse scheduling, Godsey said.
Among them, he said:
- The average age of nurses is over 50, and not enough people are entering the workforce to replace them.
- Younger nurses want to be more in control of their work life. They are digital natives and want more of a digital connection to their work experience, including scheduling.
- More health care workers want to join the gig, or freelance, economy and choose when and where they work.
“And then the pandemic created massive disruption” that prompted many clinicians to leave the workforce, which further strains the system, Godsey said.
“Part of our mission is how do we engage them and build bridges that let us re-engage the workforce on their terms. They might be willing to work a few shifts a week. Our model opens that up in a way that’s very scalable and easy to manage. Our belief is we can help rebuild a certain part of the labor market by presenting a new way for these disenfranchised workers to come back and contribute again,” he said.
Inpatient departments in the major medical centers across the Sanford Health footprint will start using LAMP in April, followed by support services, DeBoer said. Leaders also plan to use the predictive analytics tool to work with human resources and plan for future staffing and hiring needs. And it will also help inform higher education demands, she said.
“When you stop to think about how we’re using data and innovation to help us anticipate staffing and patient care needs, it helps us manage not only the short-term but also look ahead so that we can shift resources, both HR with hiring but also how we guide our colleges of nursing and other expertise to manage what those needs might be for staffing,” DeBoer said.
The collaboration with Flexwise also could help other health care organizations with their use of the AI tool, Buckingham-Carlson said.
If it works at a sprawling, integrated, rural system like Sanford Health – with 48,000 employees, two major medical centers, 44 other hospitals, 224 clinics and 233 senior living communities – it can work elsewhere.
“It feels small now, but it could be huge in how we support health systems across the nation,” she said.
Sanford Health is focused on innovation to bring new ideas to market that improve the health and well-being of its patients, people and communities. It values the ideas and problem-solving ability of nurses, physicians, researchers, clinical workers and support staff. Any employee with an idea for a device, therapy, software, tool or other method that helps patients is encouraged to contact the innovation and commercialization team and join the people at Sanford Health who are already inventing.
“The work that Braden (Bills) has been doing on this project is a great example of our innovation department helping accelerate innovative ideas and scale them across the enterprise to benefit our caregivers and patients,” said Katie Pohlson, senior director of innovation and commercialization.
- AI ethical principles already in place at Sanford Health
- Algorithm spots patients’ potential breast cancer risk
- Artificial intelligence helps spot diabetes risk