Risk prediction models for patients who leave without being seen

In the U.S., approximately two percent of patients who visit an emergency department (ED) leave the system without being seen by care providers. Leaving without being seen (LWBS) is a serious issue in the ED because appropriate medical treatment may be delayed, and this delay can lead to poor patient health outcomes. Previous studies have examined various factors associated with LWBS in the ED. However, these studies tend to focus on developing descriptive models that estimate a probability of LWBS patients without considering patients' willingness to wait to see a care provider in the ED. The objective of this paper is to develop a new risk index that incorporates patients' ED length of stay (LOS). We compared the performance of this new index with the risk computed through a machine-learning method that does not include the time component. The result showed that the newly-introduced, time-sensitive risk index performs better at identifying LWBS patients compared to other prediction models considered in the study. This new index may help EDs identify patients who tend to leave the system after staying in the ED for a short amount of time and may help intervene promptly before these patients leave the system.