Informatics in Medicine Unlocked (Jan 2022)
Benchmarking performance in emergency medical services for improving trauma care: A data driven approach
Abstract
Objective: We propose a data-driven approach, that can be used by any state or region, to cluster county-level emergency medical services (EMS) and subsequently benchmark them against their peers using over- (OT) and under-triage (UT) rates as performance measures. Methods: The approach consists of three phases: data collection, clustering, and benchmarking. Collected data includes EMS trauma-specific capabilities, volume, and performance improvement activities aggregated at the county level. Clustering approaches include K-means, K-medoids, and CLARANS. Benchmarking of EMS performance in each county against its peer within the same cluster was conducted using guidelines from the American College of Surgeons (ACS) on over- (OT) and under-triage (UT) rates. To illustrate our data-driven approach, we used data collected from the state of Ohio (OH). This included a survey of 318 EMS agencies across 87 counties in the state of OH, and 6002 deidentified patient records from 2012 obtained from the state of Ohio's EMS Division. Results: Our proposed approach can handle both qualitative and quantitative data. It is capable of identifying key factors that allow for grouping of counties as peers, in a cluster, to enable subsequent benchmarking. Application of our approach on the state of OH data revealed that county-type (rural vs urban), number of trauma-related runs, number of paid vs. voluntary employees, and additional resources provided by EMS agencies in the county (to help with well-being and coping mechanisms for EMS providers) were key factors in clustering. A small number of clusters appeared sufficient to group counties based on various EMS attributes. Benchmarking based on ACS guidelines revealed a large variation in UT and OT rates among counties in the same cluster. Rates up to 3–4 times higher were observed in several counties compared to the best performing county in that cluster, even though some of these counties had better capabilities. Conclusion: Our novel, data-driven approach can enable any regional or state EMS agency in the United States (US) to quantitatively compare counties with their peers in the same cluster and help unravel new insights that could be used to target cluster-specific interventions that can achieve improved outcomes.
