Based on the research paper “Prediction-Driven Surge Planning with Application in the Emergency Department.” 

Key Takeaways

  • Making accurate staffing allocation decisions for nurses in hospital emergency departments is challenging given the high level of uncertainty in patient volume. But by quantifying this uncertainty, it’s possible to make smart bets that guide staffing decisions in a way that balances quality care and staffing costs.  
  • Prediction models for emergency department patient volume that incorporate real-time information provide an opportunity to utilize algorithms that offer an effective way to facilitate prediction-driven base and surge staffing decisions.
  • Even when a predictive model does not forecast with 100 percent accuracy, it can still contribute substantial benefits in terms of guiding smart staffing decisions. 

The paper “Prediction-Driven Surge Planning with Application in the Emergency Department” explores how prediction-driven surge planning can be used to improve hospital emergency department staffing decisions. The study was co-authored by Yue Hu, Carri W. Chan, and Jing Dong of Columbia Business School. 

Determining nurse staffing decisions for hospital emergency departments in a way that balances quality of service and staffing costs is extremely challenging given potential variations in patient volume and demand. However, a prediction-driven staffing framework that integrates real-time information to determine base and surge staffing may reduce staffing costs by up to 16 percent while still guaranteeing timely and quality access to care for patients. 


Managing emergency department nurse staffing decisions in a way that balances costs and quality of service is both complex and challenging. This is especially true when planning for scenarios where patient demand exceeds nursing staff supply and overtime or temporary staff are brought in to help fill gaps. Among the most complicated aspects of this challenge are the high levels of uncertainty in patient demand and the fairly static nature of emergency department full-time staffing decisions, which are often made months in advance. Cost is another complex factor in the mix: Base-stage staffing is cheaper but faces a higher level of uncertainty, whereas surge-stage staffing is more expensive but faces a lower level of uncertainty. 

Interested in determining how predictive analytics could be used to guide emergency department nurse staffing decisions, the researchers built a machine learning algorithm to predict when patients are going to arrive at emergency departments. They separated the predictive algorithm into two different time scales that align with common staffing scenarios: one made weeks in advance to account for base staffing (full-time employees) and another that uses real-time predictions to determine surge staffing needs. 

Seeking to quantify when surge staffing is beneficial, the algorithm uses information across various categories, including time of shift, previous shift arrival counts, patient severity level, Google trends in the week preceding the shift, and weather. 


The researchers found that incorporating real-time information into a two-stage predictive model can improve emergency department staffing decisions. In particular, the research revealed substantial benefits to surge staffing when arrival-rate uncertainty is greater than the underlying system variability. 

The researchers propose a policy that first guides staffing decisions at the base stage by following a specific mathematical model to serve the expected patient volume. Then, to guide staffing decisions for the surge stage, it incorporates a square-root hedging against the system stochasticity. Last, the researchers extend the analysis to account for prediction error at the surge stage. 

Through using data to inform smarter staffing decisions, the researchers estimate that this algorithm can save an estimated 10 to 15 percent of staffing costs while maintaining a high level of quality care and access. 


In a time when many hospitals are operating on negative margins and access to care is growing increasingly limited as hospitals and emergency departments close, any opportunities to realize savings while maintaining quality care offer substantial benefits. This research demonstrates that real-time information, when incorporated into a two-stage prediction model, can effectively quantify uncertainty in emergency department demand and facilitate smart base and surge staffing decisions. As health care institutions increasingly have access to a wealth of high-quality data, they can leverage it to guide smart, well-informed operational decisions that balance cost and quality of care.