When NorthShore University Health System did an analysis of needed staffing levels, they found, to their surprise, that Tuesdays were as busy as Mondays, said Dr. Daniel Chertok, senior data scientist for the Chicago-area health system.
They decided to build a predictive model that predicts staffing needs four hours in advance so they can call nurses in to work before they find themselves understaffed.
“Not all nurses are happy about on-demand scheduling,” Chertok said.
According to Chertok, the example is just one reason why utilizing predictive analytics can be a smart business investment. Staffing needs to be ramped up before busy periods, and analytics can uncover when this will be, he told the crowd at the Predictive Analytics forum in Boston on Tuesday.
There is also a correlation between staffing levels and patient wait times in the emergency department. The hospital system made graphs of the median patient waiting time versus turnaround time.
Drawbacks to using such models are that the data must be clean and valid for up to six weeks, he said. Inertia also plays a role in changing how things are done.
What’s needed is a commitment from management, clean data, support for stakeholders such as the emergency room, a good data warehouse infrastructure and a support team.
Efforts are underway to develop a real-time prediction system that’s integrated with Epic, the system’s electronic health record software.