Impact of a Decision-Support Interface on Mental Workload and Stress in Prehospital Stroke Triage: A Randomized Crossover Trial

Clinical Question (PICO)

For early-career emergency medical technicians (EMTs) managing simulated suspected-stroke cases, does a web-based hospital-selection interface, compared with current dispatch orders, reduce mental workload measured by heart-rate variability and the NASA Task Load Index (NASA-TLX)? Sex, age, education, and service-duration subgroups were exploratory and not prespecified.

Bottom Line

In this 2026 report, 35 EMTs completed both simulations. The primary heart-rate-variability measure, RMSSD, rose from 28.1 to 32.4 milliseconds with decision support (mean paired difference 4.28 milliseconds; 95% CI 1.57–6.99; P < .01), consistent with reduced workload; subjective workload also fell. These are simulation and surrogate outcomes, not evidence of faster treatment or improved patient recovery.

Design

Trial type: Randomized, unblinded, two-period crossover simulation.
N: 35 EMT-intermediates; all completed both conditions. Sequence allocation: current dispatch first, 18; interface first, 17.
Randomization: Two sequences using a Latin square to mitigate order effects; no stratification reported.
Setting: One urban Taiwanese fire-department EMS system; standardized cases with audio distractions.
Enrollment: February–March 2025.
Follow-up: Two 12-minute simulated conditions; no patient follow-up.
Analysis: Within-person paired tests, using paired t-tests or Wilcoxon signed-rank tests as appropriate. All 35 completed both conditions; no analysis population was labeled intention-to-treat.
Primary outcome: Root mean square of successive differences between consecutive heartbeats (RMSSD), measured from a five-minute recording segment beginning with the second scenario. Higher RMSSD was interpreted as lower workload.

Population

Inclusion criteria

Certified EMTs with stroke-management training and fewer than three years of field service, recruited through announcements.

Exclusion criteria

History of arrhythmia or use of cardiac-related medication.

Baseline characteristics

Mean age 27.5 years (SD 5.34); 31 men and four women.
Mean service 1.33 years (SD 0.81); 21 held a university degree and 14 a college degree.
Participants were clinicians, not stroke patients; no patient NIHSS or other patient baseline severity applies.

Interventions

Current dispatch orders: Standard dispatch-based hospital selection using the simulated patient information.
Decision-support interface: A single-page checklist collecting G-FAST findings, onset time and location, then ranking hospitals using predicted treatment probabilities, transport time and hospital treatment capabilities.
Each condition presented up to eight sequential simulated cases within 12 minutes; case-specific location, onset time and G-FAST findings varied.

Outcomes

Primary outcome — RMSSD: Current dispatch versus interface: 28.11 (SD 13.01) versus 32.40 (SD 13.57) milliseconds; paired difference +4.28 milliseconds (95% CI 1.57–6.99; P < .01). This was a continuous physiologic measure, so an odds ratio, hazard ratio or risk ratio is not applicable. The paper did not prespecify subgroups; its exploratory subgroup tests should not be treated as confirmatory.

Secondary outcomes:

Mean heart rate decreased from 77.46 to 73.79 beats/minute, paired difference −3.67 (95% CI −6.43 to −0.90; P < .01). High-frequency power increased, difference +112.46 (95% CI 15.35–209.57; P = .02).
Mean NASA-TLX scores were 62.54 and 35.45. The paper reports a paired change of −26.04 (95% CI −34.23 to −20.02; P < .01), which does not reconcile exactly with the difference of the printed condition means (−27.09). Treat the exact magnitude cautiously.
Interface usability averaged 78.07/100 (SD 16.76); it was assessed only for the interface, not as a comparative score.
Exploratory service-duration subgroup: the 1.0–1.5-year group reported greater usability than the 1.5–3.0-year group (P = .02). This was not prespecified and is hypothesis-generating.

Adverse events / safety: No study adverse events or dropouts were reported. Destination accuracy, time to reperfusion, clinical events and disability outcomes were not established by this simulation.

Figures

*Figure 1.* Participant flow in the randomized crossover study; Trial A used current dispatch orders and Trial B used the proposed decision-support interface.

Criticisms

Workload is a surrogate; this study does not demonstrate better hospital selection, shorter reperfusion times or better neurological outcomes. Breathing was not controlled, complicating interpretation of HRV as a specific stress marker.
Participants could not be blinded, and subjective ratings may reflect expectations. One urban system and 35 predominantly male early-career EMTs limit generalizability.
Subgroups were selected after observing the sample distribution; several comparisons and small cells make chance findings plausible.
The reported NASA-TLX paired change and printed condition means do not exactly agree. The article also gives inconsistent usability-grade counts in its narrative and table; the overall score is more reliable to quote than the grade breakdown.
Two short simulated sessions cannot capture real-world interruptions, dispatch constraints, incorrect recommendations or patient safety.

Funding

Taiwan National Science and Technology Council grants NSTC 112-2314-B-002-222 and NSTC 113-2314-B-002-305-MY2, and National Taiwan University Hospital grant 114-S0081. The accessible article does not specify the funders' role or a trial registration identifier.

The paper

Authors. Lu YC et al.
Title. Impact of a Decision-Support Interface on Mental Workload and Stress in Prehospital Stroke Triage: A Randomized Crossover Trial.
Journal. Western Journal of Emergency Medicine.
Year. 2026.
DOI. 10.5811/westjem.61934
PMCID. PMC13660630

Deep Dive — click to expand

What this is

A 2026 randomized crossover simulation tests whether a structured destination-selection interface makes stroke triage less demanding for novice EMTs. The primary physiologic measure shifted by 4.28 milliseconds and self-reported workload fell, but neither result establishes benefit for patients.

1. Shadow Audit

The headline speaks of reduced stress, yet the measured change was RMSSD, an autonomic marker influenced by factors beyond stress; respiration was not controlled. The favorable NASA-TLX result is subject to unblinded expectations, and its published paired change does not equal the subtraction of the two reported means. Moreover, the narrative's usability-grade counts conflict with its table. None of these limitations proves the interface ineffective; each limits what can be claimed.

2. Inversion Engine

For the interface to have no average RMSSD advantage, the true paired difference would have to be zero or negative; the reported 95% interval of +1.57 to +6.99 milliseconds excludes zero under the study's model. That statistical contrast does not exclude a clinically harmful effect on destination accuracy: the trial did not measure that outcome. A larger, blinded-outcome field evaluation could therefore overturn a presumed clinical benefit without contradicting this workload finding.

3. Second-Order Catalyst

The first plausible adopters are EMS educators and quality-improvement teams in urban systems with multiple destination options, not an entire region's dispatch service. They could test the checklist in supervised simulations now, while a later prospective field pilot measures override behavior, treatment eligibility, destination errors and time to treatment before operational rollout.

4. Asymmetric Leverage

A modest reduction in decision burden could matter across many repeated calls, especially for less experienced crews, but the trial provides no estimate of benefit per patient or population-level effect. Conversely, even a rare wrong destination could offset a convenient interface; measuring uncommon safety failures requires a much larger operational dataset than 35 simulated participants.

5. Paradigm Destroyer

The useful challenge is to the assumption that a destination aid must be judged only by predictive accuracy: human workload and usability also matter. A cautious protocol update would add supervised interface training without changing mandatory destination rules. Record every recommendation, override, destination and time point during a prospective evaluation before making the software authoritative.

MVP — Minimum Viable Proof

In a prospective multi-system study, demonstrate that the interface improves or at least does not worsen independently adjudicated appropriate destination selection and treatment timing, with a prespecified safety margin, while maintaining the workload benefit. The acceptable margin and sample size must be set before collecting outcomes; this trial does not establish them.

Best Combination

Pair this human-factors result with external validation of the underlying destination model and comparative studies of real-world routing and reperfusion. Evidence for model performance, usability and patient outcomes answers different questions; none substitutes for the others.

Overvalue Warning

Do not translate a +4.28-millisecond RMSSD difference into a proven reduction in stroke disability or mortality.
Do not interpret the post-hoc service-duration finding, or the absence of a significant sex difference in four women, as proof that effects differ or are identical across groups.

Educational use notice. The Time Is Brain Review is dedicated to medical education. Figures from published articles are embedded for teaching purposes, with attribution to the original publication. We do not claim copyright over any embedded figure; all rights remain with the original publisher and authors under the article's published license. If you are a rights holder and would like a figure removed, please contact the editorial team. See our editorial notice for the full policy.