Predictive analytics in practice: A novel simulation application for addressing patient flow challenges in today's emergency departments

Objectives: To develop a flexible software application that uses predictive analytics to enable emergency department (ED) decision-makers in virtually any environment to predict the effects of operational interventions and enhance continual process improvement efforts. To demonstrate the ability of the application's core simulation model to recreate and predict site-specific patient flow in two very different EDs: a large academic center and a freestanding ED. To describe how the application was used by a freestanding ED medical director to match ED resources to patient demand.

Methods: The application was developed through a public-private partnership between University of Florida Health and Roundtable Analytics, Inc., supported by a National Science Foundation Small Business Technology Transfer (STTR) grant. The core simulation technology was designed to be quickly adaptable to any ED using data routinely collected by most electronic health record systems. To demonstrate model accuracy, Monte Carlo studies were performed to predict the effects of management interventions in two distinct ED settings. At one ED, the medical director conducted simulation studies to evaluate the sustainability of the current staffing strategy and inform his decision to implement specific interventions that better match ED resources to patient demand. After implementation of one intervention, the fidelity of the model's predictions was evaluated.

Results: A flexible, cloud-based software application enabling ED decision-makers to predict the effects of operational decisions was developed and deployed at two qualitatively distinct EDs. The application accurately recreated each ED's throughput and faithfully predicted the effects of specific management interventions. At one site, the application was used to identify when increasing arrivals will dictate that the current staffing strategy will be less effective than an alternative strategy. As actual arrivals approached this point, decision-makers used the application to simulate a variety different interventions; this directly informed their decision to implement a new strategy. The observed outcomes resulting from this intervention fell within the range of predictions from the model.

Conclusion: This application overcomes technical barriers that have made simulation modeling inaccessible to key decision-makers in emergency departments. Using this technology, ED managers with no programming experience can conduct customized simulation studies regardless of their ED's volume and complexity. In two very different case studies, the fidelity of the application was established and the application was shown to have a direct positive effect on patient flow. The effective use of simulation modeling promises to replace inefficient trial-and-error approaches and become a useful and accessible tool for hea