Evaluating ImageTrend Collaborate as a National EMS Dataset: A Cross-Sectional Comparison with the National EMS Information System

Abstract

Objectives: ImageTrend Collaborate (ITC) is a new dataset of emergency medical services (EMS) patient care events. While offering a new option for prehospital research, it is unclear if ITC data accurately reflects EMS event characteristics across the United States. Our objective was to assess the national representativeness of the ITC dataset.

Methods: We performed a cross-sectional study comparing the 2022 ITC with the 2022 National EMS Information System (NEMSIS) research dataset, a federally funded national EMS data repository. We compared 9-1-1 EMS activations across call (e.g., agency level of service, urbanicity, census region), patient (e.g., age category, sex, dispatch complaint, response time, primary symptom, primary impression), and intervention (e.g., procedures) characteristics. We quantified absolute proportional differences (PD, ITC% - NEMSIS%) with 95% confidence intervals (CI) and standardized mean differences (SMD) to evaluate the effect size of observed differences.

Results: We included 9,592,122 EMS events in ITC and 41,997,730 in NEMSIS. ImageTrend Collaborate included a higher proportion of fire-based agencies (PD 8.2, SMD 0.16) and Advanced Emergency Medical Technicians (PD 2.1, SMD 0.12) and lower proportion of volunteer agencies (PD -1.4, SMD -0.12). Census Region differences included West (PD 13.9, SMD 0.31), Northeast (PD -11.0, SMD -0.33), and Midwest (PD -5.2, SMD -0.14). Age, sex, dispatch complaint, and primary impression had absolute proportional differences less than 1.0% with SMD less than 0.10 (i.e., negligible) except for "sick person" dispatch (PD -3.0, SMD -0.08) and "other general" primary symptom (PD -1.5, SMD -0.11). Median response times were identical in both databases. The most common procedures were similar between the datasets, but differed in proportion, including: catheterization of vein (PD -4.8, SMD -0.12), 12 lead electrocardiogram (PD -7.9, SMD -0.22), and evaluation procedure (PD -3.8, SMD -0.15).

Conclusions: ImageTrend Collaborate reflects most national EMS call, patient, and intervention characteristics. Notable and expected differences between ITC and NEMSIS include fire-based agency and census region representation, likely due to self-selection; however, differences in procedures may reflect tradeoffs between granularity of a single vendor dataset versus an aggregated national dataset. Prehospital researchers should understand the characteristics of the ITC dataset to ensure proper use and analysis.