Streaming Media

Name of Corresponding Author

Paula de la Pena

Credentials of Corresponding Author



Emergency Department (ED) over-utilization and crowding hinders patient care and is associated with increased morbidity and mortality. Additionally, ED length of stay (LOS) is associated with both quality of care delivered as well as patient care perceptions. There is an inverse relation that exists between occupied beds in the ED and available beds in the hospital. Traditional forecasting of ED patient length of stay and discharge disposition has been based on previous experience of the provider and existing practice.


Natural language processing (NLP) is a highly promising method to aid in early decision making in the ED. Free-text data in clinical notes may lend invaluable insight into judgements based on provider knowledge expertise that have significant effects on patient length of stay and ultimate discharge disposition (Chen et al., 2020). The purpose of this literature review is to explore the existing utilization of NLP methods to address overcrowding in the ED.

Search strategy

A literature search was conducted in CINAHL, PubMed, Ovid Medline. The search was limited to a period of the past five years. Search terms were limited to the concepts of “emergency department disposition” and “emergency department length of stay”. Keywords searched included: “natural language processing”, “emergency department”, “disposition”, “length of stay”, “prediction modeling”, “forecasting”, and “machine learning”.

Results of literature search

Forty-two articles were assessed for eligibility. Articles were excluded if natural language processing was not a primary method of statistical analysis (n= 28) or were irrelevant to the emergency department (n=4). Six studies met inclusion criteria and were included for final analysis.

Synthesis of evidence

Natural language processing remains a novel approach to address systems of care in the ED. There is evidence to support that NLP is a more apt method of predicting and explicating relations that exist between the non-linear and complex predictors of illness in the ED.

Implications for practice

While limited, there is a growing body of evidence to support that the addition of natural language processing techniques and the inclusion of free-text data to predictive modeling is useful in predicting discharge disposition and length of stay in the ED.



The Utilization of Natural Language Processing in Predicting Emergency Department Overcrowding: A Literature Review