Author: Christopher Lew
Emergency departments (ED) are constantly flooded with patients from various situations, including 911 calls, interfacility transports, and packed waiting rooms. This overcrowding not only places strain on the doctors that must constantly work around the clock to aggressively treat patients to clear space in the ED, but also impacts patient care as doctors must work more quickly on patients, with many patients are left waiting longer to receive care. Many hospitals have taken a step towards alleviating the overcrowding of the ED by utilizing an emergency department observation unit (EDOU) for patients who require further observation before discharge or hospital admission. Currently, many patients often enter the ED only to enter the EDOU immediately, with the EDOU often far from maximum capacity. The goal of this project is to use a neural network (NN) to study patients that are sent to the EDOU, with the NN trained using basic patient information, vitals, chief complaints, time of day, and other parameters to assess whether patient would be successful in the EDOU. Assuming the learner has access to EMR records to train, creation of a training and test set using patients who are both sent and not sent to the EDOU can allow the learner to be trained adequately before assisting physicians. Along triage assistance, the learner could use test results that are obtained while a patient is in the ED along with updates to patient parameters to determine when a patient’s health is stable enough to be sent to the EDOU. Furthermore, the learner could reduce the number failed EDOU patients (patients who are end to the EDOU but are later admitted), further increasing the efficiency of the ED and increasing the quality of patient care. Improving triage is a huge step to increase the efficiency and patient care of the patients in the ED, with a learner that assists with the process able to find the most efficient method achievable.