Trying to predict a hospital’s inpatient nurse staffing needs is a tough day-to-day challenge that involves matching often unpredictable patient demand to the nurse resources needed to care for those patients. The difficulty of that task increases rapidly with the size and complexity of the hospital organization. Case in point: the Chesterfield, Mo.-based Mercy health system, a 31-hospital network serving Missouri, Kansas, Oklahoma and Arkansas, a hospital system that ranges from critical care hospitals with an average census of perhaps five to 10 patients a day to large tertiary care facilities.
In March, 2011, Mercy implemented enterprise-wide web-based scheduling software (supplied by Avantas, Omaha Neb.) that has significantly reduced the guesswork involved in its inpatient nurse staffing allocation. Bruce Weinberg, Mercy’s executive director of nursing resource management, notes that nursing “accounts for a high amount of labor dollars; is critical to patient care, so is a vital resource; but it’s scarce. It’s incumbent on us to utilize that resource as effectively as possible.” That made inpatient nursing the department of choice for implementing the system, which is as of August was running in 23 of Mercy’s 31 hospitals.
Weinberg says that like most healthcare provider organizations, Mercy had been flying somewhat blind, using limited information to match staffing resources and patient requirements. Missing the mark was costly both in terms of being under-staffed or having excess staff, he says. Being short-staffed meant bringing in someone per diem—an expensive alternative—and can also be costly in terms of quality, safety and service, he says. Having too much staff meant possibly sending staff home, thereby incurring the cost of hiring people who are not used. Either situation will eventually lead to unnecessary turnover, he says. Mercy’s key goal was to get the right staff to the right place at the right time.
Predictive Modeling Helps to Hit Targets
Weinberg says the software provides dynamic predictive modeling, giving a sense of not only staffing needs, but how staffing looks compared to the predicted demand. The model uses regression analysis to consider various indices, such as historical census and staffing levels and announcements by the Centers for Disease Control and Prevention (CDC), which have been chosen as relevant to the provider organization. Weinberg says that at Mercy, the program considers over 100 variables and determines the influence a factor has on the census in a particular unit. The information is fed electronically to the application, which forecasts what the census will be and determines potential staffing needs.
“It’s all very scientific,” Weinberg says, adding that although the system is not perfect, it is far more efficient than trying to figure out staffing needs manually—something he knows about from personal experience. Prior to implementing the present system, he recalls pulling up 52 weeks of Mondays on a single nursing unit, checking the census at a certain hour of the day, and trying to help a manager figure out how many nurses to schedule based on that history. “I have that data and I can dig into it, but the amount of time it took to do that for one single unit you could not replicate unless you had a whole army of people,” he says. He says the model is 97-percent accurate across all of the facilities where it is implemented.
According to Weinberg, one of the biggest learnings in using the system, which is ongoing, is how to use the information effectively. “Suddenly you are presented with all of this information, and you have to start learning not only how to use it, but what structures and processes do we have in place. There’s lots of improvement, lots of focus on improving staffing so we have what is needed to care for our patients all of the time.
Expanding to Other Departments
Weinberg says Mercy is exploring additional departments to leverage the software, including where there is 24/7 staffing, variability in staffing needs, or there is significant labor expense. Potential areas include operating rooms, emergency departments and pharmacy. (While the predictive model may not be a driving force in non-inpatient areas such as pharmacy, which has relatively fixed staffing needs, having a macro-level view of the census is helpful there as well, Weinberg says.)
Overall, automating staff scheduling has helped Mercy to get a better handle on its core staff that is hired to work in a specific unit, as well as contingency staff that floats among several different units, Weinberg says. “When you don’t have a good grasp of those needs, it’s hard to recruit, hire and retain for different kinds of positions,” he says.
He adds that the predictive modeling has provided insight into where staffing needs exist; but the system also provides analytical insights, such how much overtime is being used during a certain week, shift or unit to better allocate expensive staff resources. “If we know that we are using a lot of contingency resources at a certain date and time, then maybe it’s time to hire some additional core staff,” he says.
Mercy currently has 250 units on the system, scheduling about 9,000 staff. Although the main users are managers, co-workers are regular users as well; many of the units are “self scheduling,” where co-workers sign up for a particular time slot and communicate electronically with their colleagues and manager for approval. That’s an area where the predictive model has proved helpful, because it can predict a need in a certain area or shift, Weinberg says.
Overall, Weinberg says, the system has set itself apart as a comprehensive tool with predictive capability, both of which has led to better management of the hospital system’s staff resources.