Wisconsin Real-time Emergency Department Surveillance and Responsive Training (WIRED-RT)

At a Glance

The COVID-19 pandemic highlighted the need for detailed and timely information in making public health and operational decisions. This project aimed to develop an advanced COVID-19 syndromic surveillance system using electronic health records (EHR) and predictive analytics to provide just-in-time training for health care providers. Researchers successfully developed data pipelines from UW Health emergency department (ED) EHR data and validated them at Marshfield Clinic. Analysis of the data revealed an increase in acute respiratory infection (ARI) cases during the COVID-era, and predictive models performed well in predicting ARI arrivals and boarders ​(patients admitted to the hospital who remained in the ED for a minimum or four hours after admission). The Wisconsin Real-time Emergency Department Surveillance and Responsive Training (WIRED-RT) COVID-19 Simulation Curriculum Toolkit was developed to help health care teams prepare for surges in cases, focusing on key care scenarios.

The Challenge

The COVID-19 pandemic highlighted the critical role that detailed and timely information plays in public health and operational health system decisions during public health crises. Health care providers in the emergency care system, including emergency departments (ED) and the ambulance-based emergency medical services (EMS), are the first to notice the presence of emerging illnesses and provide care. As a result, these services create a rich source of data, including patient symptoms, care delivery and outcomes, which provide insight into the location and extent of emerging illnesses and injuries. While state-wide and national syndromic surveillance programs exist, they are primarily focused on population level health patterns and are not integrated into a just-in-time training program which delivers information to health care providers precisely when they need it. As a result, there was a need to develop a local COVID-19 specific surveillance report using patient level electronic health record (EHR) data to identify surges in cases real-time which affords decision makers the opportunity to prescribe interventions that can tailor allocation of resources thereby improving health care delivery and improve outcomes.

Project Goals

The objective of this project was to expand and test an advanced EHR-based surveillance system using predictive analytics linked to just-in-time training infrastructure to facilitate early and targeted health system operational responses in Wisconsin for the COVID-19 pandemic and future public health crises. This objective was addressed through three project phases:

1. Development of a pipeline to obtain care metrics from UW Health ED electronic records.
2. Validation data pipeline at Marshfield Clinic.
3. Responsive Training Development.

Results

The project team was successful in developing a pipeline to extract key care metrics from UW Health ED electronic records and validating that pipeline at Marshfield Clinic. Two unique datasets were extracted from each of the health systems, one pre-COVID and one COVID-era. Analysis of the data from ED admissions in these time periods indicated a significant increase in the proportion of patients with acute respiratory infection (ARI) during the COVID-era compared to the pre-COVID period. Further, the number of boarders (patients admitted to the hospital who remained in the ED for a minimum or four hours after admission) with ARI was consistently higher during the COVID-era at both health systems, although total boarders were higher during the pre-COVID era at UW Health.

The predictive models performed well in predicting ARI boarders and arrivals with ARI, with better performance observed for models trained on pre-COVID data at UW Health and COVID-era data at Marshfield Clinic. These results suggest that the models’ performance improved as the COVID-era progressed. The transferability of the models between locations was better during the pre-COVID period compared to the COVID era which was possibly due to the differences in patient behavior and ARI patient volume between the health systems. Overall, these results indicate that model performance can vary based on training data period and locations, with some models showing better transferability during the pre-COVID period. Results were published in the Journal of the American Medical Informatics Association.

The Wisconsin Real-time Emergency Department Surveillance and Responsive Training (WIRED-RT) COVID-19 Simulation Curriculum Toolkit was developed in the final phase of this project to help health systems prepare for surges in COVID-19 cases. The toolkit included four responsive training simulation cases, each of which include a case overview, a set-up guide, pre-briefing materials, a facilitation guide and PowerPoint, a facilitation checklist, a debriefing guide and a video example of the case. The cases cover emergent intubation, use of high-flow oxygen delivery, cardiac arrest care and safe intra- and inter-hospital transport protocols for patients with COVID-19. This curriculum prepared interprofessional health care teams to safely and effectively care for their patients, improve communication and obtain deliberate practice in high-stakes, high acuity, low-frequency care events.