Research Update: Improving Hospital Operations

Update: Improving Hospital Operations through Predictive-Prescriptive Analytics and Behavioral Models 

Prof. Georgia Perakis and her team (Prof. Dessislava Pachamanova from Babson College, Asterios Tsiourvas, a PhD student at the MIT Operations Research Center; and Omar Skali Lami, a recent graduate of the ORC currently employed by McKinsey & Co.) are continuing their operations research at University of Massachusetts Medical School and UMass Memorial Hospital Emergency Department (ED). Initially, their aim was to analyze and develop better methods to improve patients’ length of stay (LOS) in the hospital’s ED.
 
Usually, medical personnel use acuity to triage incoming patients with a rating system such as the Emergency Severity Index (ESI). Perakis’ team, with expertise in operational and predictive analytics, sought to develop a model that would make more efficient use of the ED’s resources and get patients to the right place more quickly.
 
In the previous research update, Perakis and team discovered a new way of approaching this problem by dividing LOS into different phases. With this breakthrough, the team:

1. Could predict the length of stay solely from the incoming patient’s chief complaint
2. Revealed that the ED’s overall LOS was a function of its capacity, flow and resources
3. Highlighted inconsistencies and potential fairness issues with patient routing

In the most recent phase of this project, the team focused on the third point to understand how the inconsistencies and inequity factor into patient treatment decisions. They extended their model to correct for these biases. In addition, using actual hospital data, their model proved capable of directing patients to the right resources, resulting in a 50% increase in patient throughput (number of patients seen) while also decreasing patients’ wait times by more than 50%.

Impressed with these results, their collaborator asked the team to think about extending the operational excellence to other areas of the hospital. These results also present the case that successful models will need to take a granular view of LOS and account for the hospital’s resources and capacities. Perakis and her team’s research have a very practical application that could change how patients are treated and save lives by using what is already available.