Building and Implementing a predictive decision support system based on a proactive full capacity protocol to mitigate emergency overcrowding problem

建立和实施基于主动式全能力协议的预测性决策支持系统 缓解紧急情况下的过度拥挤问题 项目摘要 急诊科(EDs)面临着过度拥挤的重大问题，这构成了重大的患者安全风险，并导致 卫生保健服务质量差，死亡率高。急诊室过度拥挤是一个可以解决的病人流动问题 通过改善从到达到入院或出院的病人流。根据美国紧急情况学院的说法 医生说，全容量协议(FCP)是改善患者流量从而缓解ED的关键方法 人满为患。FCP有不同的级别，这些级别由不同的标准触发，这些标准基于患者流量测量 (全氟甲烷)。FCP的当前实践使用实时(即反应性)信息来决定FCP标准。然而，当它 在实施FCP干预时，使用实时信息效率不高，因为在许多情况下FCP水平 在ED已经人满为患的情况下激活为时已晚。该项目改进了反应性FCP，使其具有主动性 使用人工智能和预测分析。将使用深度学习模型来预测PFM，然后 与反应性FCP集成。将开发一个决策支持系统，以实施拟议的主动FCP。这个 这项计划的整体目标是发展一个架构，以纾缓教育署的挤迫情况。有四个目标(目标1和目标2 R21下的目标3和R33下的目标4)：目标1：开发深度学习模型，以预测不同的PFM值和 将它们纳入积极主动的FCP。许多PFM值代表患者从入院到入院的流程。我们会 构建多个深度学习模型来预测PFM值(例如，寄宿人数)。然后，我们将更新反应性 FCP以包括预测的PFM值。目的2：建立DES模型评价主动FCP的有效性。 在生产中运行主动式FCP之前，我们将比较被动式和主动式FCP的结果 产生例如平均停留时间(LOS)、等待时间和工作人员满意度。目标3：设计、评估和实施 基于主动FCP的决策支持系统(DSS)。我们将设计以用户为中心的决策支持系统，帮助临床医生和 PFCT在实施积极的FCP方面的作用。我们将使用主动式FCP标准作为DSS的输入，以实现关键部件的自动化 积极的FCP干预措施。目标4：通过标准化数据输入和输出，扩大和推广决策支持系统 接口。我们将创建基于FHIR的应用程序编程接口(API)以允许特定于站点的配置， 示范培训、评估和精简实施流程。该项目的成功完成提供了 一种先进的可互操作的决策支持系统，用于实施基于早期、准确的PFM预测的主动式FCP 价值，以允许适当的规划和执行病人流动过程，从而减轻急诊室的过度拥挤。我们的多- 纪律团队处于有利地位，能够成功实现所有目标。

Building and implementing a predictive decision support system based on a proactive full capacity protocol to mitigate emergency overcrowding problem Project Summary Emergency departments (EDs) face a major problem of overcrowding that poses a significant patient safety risk and leads to poor healthcare service quality and high mortality rates. ED overcrowding is a patient flow problem, which can be solved by improving patient flow from arrival to admission or discharge. According to the American College of Emergency Physicians, a full capacity protocol (FCP) is a key approach for improving patient flow and consequently mitigating ED overcrowding. FCP has different levels that are triggered by different criteria, which are based on patient flow measures (PFMs). The current practice of FCP uses real-time (i.e., reactive) information to decide FCP criteria. However, when it comes to implementing FCP interventions, using real-time information is not efficient because in many cases FCP levels are activated too late when ED is already overcrowded. This project improves the reactive FCP to make it proactive through using Artificial Intelligence and predictive analytics. The PFMs will be predicted using deep learning models and then integrated with reactive FCP. A decision support system will be developed to implement the proposed proactive FCP. The overall objective of this project is to develop a framework to mitigate ED overcrowding. There are four aims (Aims 1& 2 under R21; Aims 3& 4 under R33): Aim 1: Develop deep learning models to predict different PFM values and incorporate them in a proactive FCP. Many PFM values represent the patient flow from arrival to admission. We will build multiple deep learning models to predict PFM values (e.g., numbers of boarding). Then, we will update the reactive FCP to include the predicted PFM values. Aim 2: Develop a DES model to evaluate the effectiveness of proactive FCP. Before running the proactive FCP in production, we will compare reactive and proactive FCPs on the outcomes they generate such as average length of stay (LOS), waiting time and staff satisfaction. Aim 3: Design, evaluate, and implement a decision support system (DSS) based on the proactive FCP. We will design user-centric DSS to aid clinicians and the PFCT in implementing the proactive FCP. We will use the proactive FCP criteria as input for the DSS to automate key parts of the proactive FCP interventions. Aim 4: Expand and generalize the DSS by standardizing data input and output interfaces. We will create a FHIR-based application programming interface (API) to allow site-specific configuration, model training, evaluation, and streamlining of implementation processes. Successful completion of this project delivers a state-of-the-art interoperable DSS for the implementation of a proactive FCP based on early, accurate predictions of PFM values to allow proper planning and execution of patient flow processes, thereby mitigating ED overcrowding. Our multi- disciplinary team is well positioned to successfully execute all aims.