In this prefecture-wide cohort of 1748 patients with prehospital ECG transmission, 38.4% (672 of 1748) achieved transmission within 10 minutes, with a median scene-to-ECG interval of 12 minutes (IQR 9–17). These findings align with prior prehospital studies, in which meeting the ≤ 10-minute target remains uncommon [5, 17, 18, 20].
Greater cumulative EMS-agency experience was associated with shorter scene-to-ECG intervals and higher odds of meeting the 10-minute benchmark. Female sex, older age, and nighttime arrival were linked to lower odds of achieving ≤ 10 minutes. When modelled as a continuous outcome, older age and nighttime arrival corresponded to longer scene-to-ECG intervals, whereas sex did not. Although models adjusting for annual periods and their interaction with agency experience lost statistical significance for the cumulative number of ECG transmissions, the point estimates remained directionally consistent. Results were consistent across sensitivity analyses. Collectively, these results indicate an experience-related improvement in on-scene ECG efficiency.
By examining the prehospital phase and evaluating cumulative agency experience as a process measure, this study underscores system-level factors beyond patient characteristics and identifies modifiable targets within prehospital workflows. A greater cumulative number of ECG transmissions correlated with shorter scene-to-ECG intervals, suggesting that routine ECG use by EMS personnel may enhance on-scene efficiency. Because intermediate workflow elements (for example, task delegation, lead placement, and device setup time) were not measured, the underlying mechanisms remain uncertain, and causal inferences cannot be made. Within this framework, structured onboarding and periodic practice may warrant prospective evaluation to determine whether similar efficiency can be replicated in lower-experience contexts.
After incorporating calendar-period indicators, centring cumulative experience within each period, and including a period–experience interaction, the experience coefficient reflected within-period contrasts rather than overall yearly trends. Allowing the experience effect to vary by year introduced additional parameters and shared variance, reducing precision. Under this specification, the experience term was not statistically significant, although its direction and all sensitivity analyses paralleled the simpler models. This pattern suggests that observed improvements largely reflect system maturation over time, with experience and secular trends evolving together and limiting independent variation. The loss of significance likely reflects temporal overlap rather than absence of an experience effect.
Prior ED studies have linked female sex, older age, and atypical presentations to longer door-to-ECG intervals [12, 22–24]. By contrast, standardized triage protocols, staff education, and workflow redesign have been shown to shorten these intervals [13–16]. Evidence regarding the prehospital FMC-to-ECG interval is more limited; however, existing studies similarly report low achievement of the 10-minute benchmark and identify patient-related contributors to delay [5, 17, 18, 20].
Longer intervals among older patients have been documented in previous prehospital studies [17]. This pattern may reflect additional time required for multiple on-scene procedures, including assessment, positioning, intravenous access, and patient transfer [17]. In the ED, older age has also been associated with prolonged door-to-ECG intervals, although results vary across studies [6, 9, 11, 12].
Prior prehospital reports have attributed longer times in females to a higher frequency of atypical symptoms that delay recognition [17, 20]. In our cohort, females had lower odds of achieving ≤ 10 minutes; however, the sex term was not statistically significant when modelled as a continuous outcome. This discrepancy aligns with the empirical distribution of scene-to-ECG intervals. A small number of males had exceptionally long intervals, influencing linear models and diminishing the statistical significance of the sex difference. In sensitivity analyses excluding values above the 97.5th percentile (Table S2), the sex term retained the same direction of association as in the primary models but yielded smaller p values (0.052–0.062 in Models 2–4). By comparison, full-sample continuous models were nonsignificant, with larger p values (0.671–0.811 in Models 2–4; Table 5). With the ≤ 10-minute outcome, only whether the benchmark is crossed is relevant; values far above 10 minutes do not contribute additional weight. Therefore, a difference by sex is more easily detected with this threshold outcome than with a continuous model encompassing the full distribution, where a few extreme male values can distort estimates. Consistent with prior prehospital findings [17, 20], our results indicate longer intervals in females.
Evidence regarding time-of-day differences in prehospital ECG acquisition remains limited. In one multicenter analysis, the interval from EMS arrival to out-of-hospital ECG was longer during daytime than nighttime, suggesting that delays are not consistently worse after hours [17]. In contrast, our findings associate nighttime arrival with longer scene-to-ECG intervals, implying that time-of-day effects may differ by system and warrant local assessment. Given the sparse and heterogeneous literature, these findings should be generalized cautiously.
This study’s strengths include prefecture-wide coverage of nearly all prehospital ECG transmissions over a four-year period and the use of multivariable mixed-effects models with EMS-agency random intercepts to account for clustering. Directional consistency across prespecified sensitivity analyses supports the robustness of the results.
By shifting the focus from patient characteristics to modifiable operational processes, this study quantifies agency-level cumulative transmission experience in real-world prehospital care. Although intermediate workflow mechanisms were not directly measured, the results identify training and structured onboarding as pragmatic targets for prospective evaluation aimed at reducing scene-to-ECG intervals.
This study has several limitations. First, although the operational policy recommends prehospital 12-lead ECG acquisition for patients with suspected ACS, the final decision rests with individual EMS personnel. In the absence of a standardized triage protocol or validated scoring system to define suspected ACS, patient selection may differ across EMS agencies, and residual selection or indication bias cannot be excluded. Nevertheless, current guidelines recommend prompt acquisition of a prehospital 12-lead ECG whenever ACS is suspected, without requiring a formal triage tool [7, 8]. We sought to mitigate this potential bias by using multivariable mixed-effects models with EMS-agency random intercepts, and the results were consistent in sensitivity analyses that excluded extreme interval values and in complete-case analyses.
Second, the registry included prehospital process measures only and was not linked to in-hospital data. As a result, final diagnoses, treatments received, and clinical outcomes after hospital arrival could not be ascertained. Among patients with prehospital ECG transmission, the proportions ultimately diagnosed with STEMI or receiving reperfusion therapy remain unknown. Consequently, whether shorter scene-to-ECG intervals translate into improved clinical outcomes could not be determined.
Third, this retrospective analysis was conducted within a single Japanese prefecture, which may limit generalizability to systems with different EMS configurations, patient populations, or health care structures. Not all EMS agencies in the prefecture participated (11 of 14), raising the possibility of participation bias and limiting the representativeness of agency-level practices. These findings should therefore be extrapolated cautiously to other settings.
Fourth, because the study period (September 2021–August 2025) overlapped with the COVID-19 pandemic, changes in EMS operations and patient characteristics related to the pandemic cannot be ruled out as potential influences on scene-to-ECG intervals. To account for temporal effects, we included calendar-period indicators and their interactions with the cumulative number of ECG transmissions in the multivariable mixed-effects models. Neither variable reached statistical significance, although modest residual effects may persist.
Fifth, the design cannot fully separate the effects of accumulating experience from time-related system maturation. Thus, the observed association may reflect training-related improvements, temporal system development, or both.
Future work should assess whether targeted training reduces scene-to-ECG intervals and identify the specific on-scene procedures that contribute to this effect. Granular time analyses should decompose the interval into recognition, preparation and role allocation, lead placement, acquisition, and transmission to pinpoint major sources of delay. Based on these findings, a standardized onboarding and training programme focused on rapid ECG acquisition and transmission should be implemented and prospectively evaluated using a prespecified analysis plan comparing pre- and post-intervention performance. Reporting changes not only in total interval length but also in sub-intervals would clarify mechanisms of improvement. Finally, linking prehospital and in-hospital data would enable evaluation of whether process improvements correspond with better clinical outcomes in patients with suspected ACS.