Off-campus WSU users: To download campus access dissertations, please use the following link to log into our proxy server with your WSU access ID and password, then click the "Off-campus Download" button below.

Non-WSU users: Please talk to your librarian about requesting this thesis through interlibrary loan.

Access Type

WSU Access

Date of Award

January 2019

Degree Type

Thesis

Degree Name

M.S.

Department

Industrial and Manufacturing Engineering

First Advisor

Ratna Babu Chinnam

Abstract

Emergency Department (ED) crowding has become common throughout the nation and getting significant attention from healthcare providers and researchers. Prolonged patient Length-of-Stay (LOS) in EDs due to improper patient flow management and coordination leads to overcrowding, which results in adverse effects to the patients, ED staff as well as hospital revenue. Predicting patients’ ED LOS, throughout the multiple caregiving stages, would provide valuable and actionable information to patients and healthcare providers. The goal of this study to predict patients’ LOS in EDs and identify their important features using Electronic Health Record (EHR) data. In particular, our data contains both structured data collected from patients (e.g., vitals collected during patient triage) as well as unstructured raw text data from “chief complaints” collected from patients during triage. We compare the performance of the tree-based machine learning models like Random Forest, XGBoost and LightGBM. The Mean Absolute Error (MAE) for the best performing model was ~73 minutes. A LightGBM was faster, more accurate and outperforms other two models. In addition, it was able to capture influential predictors like patients vital signs, ED patient census, provider staffing levels, and select features from chief complaints. These variables would be helpful for resource allocation and operations management.

Off-campus Download

Share

COinS