Ph.D. Defense: Wanyi Chen

The Department of

Statistics and Operations Research

The University of North Carolina at Chapel Hill

Ph.D. Thesis Defense

Public Presentation

Monday, August 13th, 2018

130 Hanes Hall
1:00 PM

Wanyi Chen

A data-driven approach for operational improvement in
emergency departments.

Emergency departments (EDs) in the US are experiencing significant stress from crowding, of which one of the main contributors is the lengthy boarding process, which is the process of to-be-admit patients waiting in the ED for the hospital to ready beds for them. We explored ways to reduce crowding by initiating hospital bed request (BeRT) early on for likely to-be-admit patients. In Chapter 2, we modeled the ED patient flow as a Markov decision process. With the objective of balancing the tradeoff between waiting cost and the cost of false early BeRTs, we found the optimal early BeRT policy to be of threshold type, where the threshold is a function of census and patients probability of admission. Chapter 3 built a fluid model, where patients flow into the ED (a fluid tank) as continuous fluid flowing at a time-dependent deterministic rate. To control the number of false early BeRTs, we imposed a constraint on the length of time for the early BeRT option. The optimal policy that minimizes the fluid level (congestion level) in the ED dictates that when ED is under heavy traffic regime, one should BeRT early as early, and as long, as allowed. In chapter 4, we looked at several early BeRT heuristics that are inspired by the theoretical optimal policies found previously. We tested and compared their performances in terms of length-of-stay and waiting time using a simulation model built for the UNC ED based on 2012 patient data. We observed that as the admission probability distributions of the patient population became less variable, the heuristics that take more information into account performed better. Lastly, we offered a different perspective on ED crowding in Chapter 5, where we explored the association between ED cencus and providers’ triage and admission decisions. We found that the more crowded the ED was, the more conservative providers were, in that nurses tend to triage more patients as critical, and physicians tend to admit more patients into the hospital.