The Hallway Consult — Issue #26: AI Chest Pain Triage, HINTS vs. Sudbury Vertigo Risk Score, and Thrombectomy for Medium/Distal Vessel Occlusion Stroke

AI-Assisted Chest Pain Triage: RAPIDx AI

High-sensitivity troponin assays were supposed to simplify chest pain workups. Instead, they made interpretation harder — more borderline elevations, more causes of myocardial injury that aren't type 1 MI, and more cognitive load on ED physicians who don't have cardiology training sorting it out in real time. RAPIDx AI tested whether an AI tool applying the Fourth Universal Definition of MI to hs-troponin results could do that sorting better than clinicians alone.

This was a large cluster-randomized trial across 12 emergency departments in South Australia (6 metropolitan, 6 rural), enrolling 14,131 patients who received hs-troponin testing between April and December 2023. Six hospitals were randomized to standard practice, six to AI-based clinical decision support. The primary intention-to-treat analysis focused on 3,029 patients with myocardial injury and suspected cardiac cause (1,461 control, 1,568 AI-supported).

The primary outcome — cardiovascular death, new/recurrent MI, or unplanned CV readmission within 6 months — showed no difference: 26.0% in the AI-supported group versus 26.4% in the control group. The AI tool did not reduce hard cardiovascular outcomes.

But it wasn't inert. Among patients the AI did not classify as type 1 MI — where coronary angiography isn't known to help — invasive angiography was 47% less likely in the AI-supported group (5% vs. 9.4%). Among patients the AI did classify as type 1 MI, evidence-based prescribing jumped: statins 82% vs. 68%, antiplatelet therapy 56% vs. 44%. And it was safe — among patients discharged directly from the ED, 30-day death or MI was not higher with AI support (0.86% vs. 1.1%, non-inferiority p<0.001).

So the tool changed clinician behavior in the right direction — less unnecessary catheterization, more guideline-directed therapy — without any signal of harm. It just didn't translate that behavior change into fewer hard cardiovascular events at 6 months. The investigators' own read is that incomplete integration into clinical workflow may explain the neutral result, which is a fair hypothesis but not evidence that the tool works — it's a flag for what the next iteration needs to test.

Bottom line: In this 14,131-patient cluster-RCT (3,029 in the primary analysis), AI-based clinical decision support for hs-troponin interpretation did not reduce 6-month cardiovascular death, MI, or readmission (26.0% vs. 26.4%) compared with standard practice. It was safe and measurably shifted practice toward less unnecessary angiography and more evidence-based prescribing — real effects worth having — but on the outcome that matters most, this was a negative trial, not a win.

Lambrakis K, et al. RAPIDx AI trial. Presented at ESC Congress 2024, Hot Line session, London, UK, September 2, 2024.

HINTS vs. Sudbury Vertigo Risk Score: Which Tool Fits the Dizzy Patient?

Dizziness is one of the ED's most reliably over-imaged, under-stratified complaints — roughly a third of patients get a head CT, and the overwhelming majority come back negative, while the real miss rate for serious causes hovers in the low single digits. The two tools clinicians reach for, HINTS and the newer Sudbury Vertigo Risk Score, aren't actually competing for the same job, and conflating them leads to misapplying both.

The Sudbury score was recently put through a proper external validation: a retrospective multicenter cohort across three Canadian university-affiliated EDs (urban and rural academic sites), covering 4,559 patients presenting with vertigo, dizziness, or imbalance over roughly six years. Serious diagnoses — stroke, TIA, vertebral artery dissection, or brain tumor — occurred in 104 patients (2.3%). The score showed excellent discrimination (AUC 0.95), and at a cutoff below 5, the risk of a serious diagnosis was 0%, with 100% sensitivity and 69.2% specificity. That cutoff identified more than two-thirds of the cohort as low-risk, and nearly 45% of all advanced imaging in the study was performed in that low-risk group — imaging that, by the score's own numbers, wasn't needed.

HINTS is a different instrument for a different population. It only applies to patients with acute vestibular syndrome — continuous, persistent vertigo with nystagmus — which is roughly 10-20% of all ED dizziness presentations. Within that narrow population, performed by a trained examiner, it can outperform early MRI for detecting posterior circulation stroke. But its accuracy is examiner-dependent: pooled sensitivity and specificity across studies run 83.4%/88.9%, and inter-rater reliability between differently trained examiners has been poor in some cohorts. GRACE-3 guidance already reflects this — it recommends HINTS for nystagmus-positive patients, but only when performed by a trained clinician.

The practical takeaway isn't "pick one." It's that the Sudbury score is the broader, bedside-calculable tool that applies to any dizzy patient and doesn't require a specialized exam — useful for deciding who needs no further workup at all. HINTS is the focused tool for the acute-vestibular-syndrome subset, and it's only as good as the person performing it.

Bottom line: The Sudbury Vertigo Risk Score, validated in a 4,559-patient Canadian multicenter cohort, identifies roughly two-thirds of ED dizziness patients as having a 0% risk of serious diagnosis (100% sensitivity, 69.2% specificity at a cutoff <5) — a tool usable on any dizzy patient without specialized exam skills. HINTS remains valuable but only for the narrower acute-vestibular-syndrome subgroup and only when performed by a trained examiner. These are complementary tools for different patients, not competing answers to the same question.

van Patot ET, Roy D, Baraku E, et al. Validation of the Sudbury Vertigo Risk Score to risk stratify for a serious cause of vertigo. Acad Emerg Med. 2025;32:863-873.

Thrombectomy for Medium/Distal Vessel Occlusion Stroke: DISTAL 12-Month Outcomes

Thrombectomy's success in large vessel occlusion stroke created an obvious next question: does it also help in medium or distal vessel occlusions, where the affected territory is smaller but still clinically meaningful? The original DISTAL trial answered that at 90 days. This year's follow-up publication checks whether a longer observation window changes the story.

DISTAL was an international, pragmatic, open-label RCT across 55 centers in nine European countries, randomizing 543 patients with medium or distal vessel occlusion stroke to endovascular treatment plus best medical treatment versus best medical treatment alone. The original 90-day results, published in NEJM, showed no benefit of adding thrombectomy — functional outcomes were not improved, all-cause mortality was similar between groups, and symptomatic intracranial hemorrhage was numerically higher with endovascular treatment (5.9% vs. 2.6%).

The 12-month outcomes, now published in Lancet Neurology, confirm that result holds: endovascular treatment plus best medical treatment was not associated with a reduction in disability or death at 12 months compared with best medical treatment alone. The investigators frame this explicitly as consistent with the 90-day findings — longer follow-up didn't surface a benefit that was simply slow to appear, and it didn't resolve into harm being outweighed by delayed functional gains either. The null result is durable.

This lands alongside two other trials in the same space — ESCAPE-MeVO and the early-terminated DISCOUNT trial — that reached similar conclusions: in broadly selected medium/distal vessel occlusion populations, routine thrombectomy doesn't clear the bar, and the procedural hemorrhage risk is real. That's three trials now pointing the same direction, which is about as settled as this question gets without a dramatic change in patient selection criteria.

Bottom line: In this 543-patient international RCT, endovascular treatment plus best medical treatment showed no benefit over best medical treatment alone for medium or distal vessel occlusion stroke at 90 days, and the 12-month follow-up confirms that null result persists — no delayed benefit emerged. Combined with ESCAPE-MeVO and DISCOUNT, the evidence now consistently argues against routine thrombectomy in this population outside of patient selection strategies not yet tested.

Fischer U, Brehm A, Ribo M, et al; DISTAL investigators. Endovascular treatment for medium or distal vessel occlusion stroke (DISTAL): 12-month outcomes of a multicentre, open-label, randomised trial. Lancet Neurol. 2026;25(6):571-580.

The Hallway Consult is built for EM clinicians who want the useful version of the literature. Forward it to a colleague if it helped.

www.hallwaymedicine.com

— The Hallway Consult team