Brief
The Duke Emergency Departments did not have an objective approach to support diversion. This led to Emergency Department (ED) overcrowding, which resulted in poor process and clinical outcomes. DIHI and the Duke EDs created an ED capacity monitoring solution, and are coordinating with Durham County Emergency Medical Services (EMS) to implement it into their ED routing approach. In the autumn of 2022, we began implementing and assessing the solution for measurable, high-impact improvement for Duke EDs: time to intervention, ED staff resilience, ED wait times, and rates of patients left without being seen.
The Problem
Emergency Department (ED) overcrowding is a growing national problem affecting the quality of and access to healthcare. Overcrowding can cause the ED to divert incoming emergency medical services (EMS), a decision which is often reactive and subjective. The decision is often made without pre-determined guidelines or sufficient coordination with nearby EDs. Diversion status compromises patient safety and quality of care (e.g., long wait times, high left-without-being-seen rates) and diminishes provider well-being and performance. It also increases downstream transfers to other hospital EDs, further detracting from efficiency and patient care.
Our Solution
We proposed to provide our health system and EMS provider with pre-hospital traffic control using real-time ED census data. We believe this will be an effective approach to ensuring incoming patients are transported to the most appropriate destination to receive timely care. Diversion will be reduced or, at minimum, their effects will be mitigated via the optimization of Duke Health ED capacity.
In 2022, we began implementing an algorithmic tool for capacity management to improve patient care and coordination between the Duke University Hospital (DUH) and Duke Regional Hospital (DRH) EDs. The algorithm was based on hourly census data from the DUH and DRH EDs for three years between 7/1/2018 and 6/30/2021 (26,328 hours in total). We gathered and analyzed data using conventional methods for measuring ED crowding:1 number of ED patients, waiting room status, ventilator use, patient acuity levels (ESI), number of psychiatry boarders, and EMS transports. Figure 1 shows the distribution of patients in the ED by hour over our three-year period. We created new measures such as cumulative time elapsed: the elapsed time spent in the ED summed over all patients. Percent raw capacity was defined as the number of patients out of the maximum count ever observed for that ED over the three-year historical period. We defined ‘imminent ED workload,’ our primary outcome label, to be larger when the number of total patients was high and the median of those patients’ time since arrival was low compared to historical trends (i.e., more work needed to triage newly arrived patients) in Figure 2.
Data granularity was set to the hourly level (26,328 hours summarized per ED over the three-year period), allowing the inclusion of metrics like “arrived or triaged in the past hour” and “time of day comparisons”.
We validated our imminent ED workload outcome label using historical diversion data as well as components of the National Emergency Department Overcrowding Study (NEDOCS) scoring system.1
Our algorithm modifies raw capacity based on the number of severe acuity cases (ESI level 1, 3% capacity increase per case), the number of ventilators in use (10% capacity increase per ventilator), and the percentage of triaged patients (10% capacity increase if less than half of patients have been triaged). We set high workload to be when modified capacity was greater than 55% and medium workload to be when modified capacity was between 33-55%. Critical workload occurred when modified capacity exceeded 55% and the cumulative time elapsed was less than the observed median for that hour of the day between 7/1/2018 and 6/30/2021. Running our algorithm retrospectively over the three-year timeframe, DUH had 1,451 critical hours, 6,222 high workload hours, and 12,418 medium workload hours. We found that, retrospectively, the maximum ratio of high and critical workload labels occurred during the eighteen hours prior to the start of historical diversion events. Future EMS validation of Durham County EMS decisions in the field will be based on real-time ED data inputs.
We developed an algorithm that identifies high and critical ED workload status, which will be incorporated into an AI tool that helps guide EMS transport decisions and hopefully reduces ambulance diversions. The relative ease of implementation was key for timely integration into existing systems and allowed our algorithm to be tailored to the local environment Duke’s Durham EDs operate in. Our algorithm was also highly adaptable to ED-specific protocol changes and patient volume/acuity, both due to and independent of the COVID-19 pandemic.
Anticipated Impact
Measurable high-impact improvement for Duke EDs: clinical outcomes (time to intervention, ED staff resilience) and process metrics (ED wait times, left w/o seen).
Next Steps (as of Dec 2022)
Improved management of EMS transport destinations is a potentially high-value approach to avert downstream overcrowding crises and elevate patient care and safety. We are working with DUH, DRH, and Durham EMS to pilot the solution in early 2023. Then, post-pilot period, we will assess the impact of the solution on time to intervention, ED wait times, rates of patients left without being seen, and ED staff resilience. Future work will include the incorporation of capacity status into DCEMS protocols and the inclusion of patient conditions and patient preferences into ED destination decisions. We also plan to extend this work to support other health system EDs and their patients.
Reference
