Abstract
Emergency department (ED) overcrowding is a worldwide problem that undermines hospitals’ ability to provide timely care to patients who need it urgently. One of the primary reason for ED overcrowding is long ED boarding time due to the lack of coordination between ED and inpatient units. The objectives of this study were to develop models that use patient information collected during triage at EDs to predict the inpatient admission decisions, and to test how the information derived from free-text triage notes by text mining can increase the prediction power. We developed a Lasso Logistic regression model to predict the admission decision for an individual patient using the patient information collected during triage. Such predictive information can potentially reduce ED boarding time by initiating early admission process. Text mining techniques are applied to extract useful information from the free-text triage notes. Our study demonstrated that statistical models can predict ED patient admissions with reasonable accuracy using the routine patient data collected during triage: c-statistics 0.862 (95% CI: 0.857-0.867). Text mining can extract additional useful information from triage notes and significantly increase the prediction power. With the predictive information available in advance, hospitals can be proactive on the admission process, discharge process and bed allocations so as to reduce ED overcrowding.

Biography
Dr. Zhankun (Kevin) Sun is an Assistant Professor of Management Sciences in the College of Business, City University of Hong Kong. He holds a bachelor degree in Industrial Engineering from Tsinghua University, an M.Sc. and a Ph.D. in Statistics and Operations Research from the University of North Carolina Chapel Hill. He is interested in the research area of modeling, analysis, and control of stochastic systems with applications that arise from healthcare operations. He is a recipient of George E. Nicholson Award from Department of Statistics & Operations Research at UNC Chapel Hill.

Photo Album