🩺 THIS WEEK IN EM
IO vs IV in Out-of-Hospital Cardiac Arrest: The Drill Is a Backup, Not an Upgrade
If you work prehospital or receive enough arrest patients from the field, you already know why IO became so attractive. It is fast in theory, often easier than a collapsed peripheral vein, and it gets epinephrine moving without the visual chaos of a difficult stick during CPR. That practical appeal led a lot of clinicians to start treating IO as if it were not just a fallback, but maybe the better first move.
The problem is that the modern randomized trials do not support that upgrade. PARAMEDIC-3, a pragmatic UK trial that enrolled 6,082 adults with out-of-hospital cardiac arrest, compared an intraosseous-first strategy with an intravenous-first strategy. At 30 days, survival was 4.5% in the IO group and 5.1% in the IV group, with no meaningful difference. Return of spontaneous circulation was actually a bit lower in the IO-first group. Then came IVIO, which randomized 1,479 adults in Denmark. IO achieved first-pass assigned access more often than IV, but that procedural win did not translate into the outcomes that matter: sustained ROSC was 30% versus 29%, 30-day survival was 12% versus 10%, and time from emergency call to epinephrine was the same in both groups.
That is the part worth slowing down for. IO may be technically easier, but what we care about is not how elegant the access looks. What we care about is whether drugs get in fast enough, reliably enough, and in a way that changes survival or neurologic outcomes. The current evidence says IO is very good at solving a procedural problem. It has not proved that it solves the clinical one better than IV.
The guideline language has now moved to match that evidence. The 2025 AHA update recommends that clinicians first attempt IV access for drug administration in adult cardiac arrest. IO remains reasonable if IV attempts are unsuccessful or not feasible. That is a subtle but important shift away from “either is fine” and back toward “IV first when you can, IO when you need to.” In other words, the drill still belongs on the truck. It just does not belong on autopilot.
The practical lesson is not that IO is bad. It is that we should stop pretending faster access automatically means better outcomes. If your crew can get an IV without creating delay, the evidence supports starting there. If IV is failing or the situation is ugly, go IO quickly and do not romanticize the choice.
Bottom line: The best current evidence does not show an outcome advantage for an IO-first strategy in adult out-of-hospital cardiac arrest. Start with IV when it is feasible without delay, and use IO as the rapid backup when IV is unsuccessful or impractical, which is exactly where the 2025 AHA guideline now places it.
Couper K, Ji C, Deakin CD, et al; for the PARAMEDIC-3 Collaborators. A Randomized Trial of Drug Route in Out-of-Hospital Cardiac Arrest. N Engl J Med. 2025;392(4):336-348.
Vallentin MF, Granfeldt A, Klitgaard TL, et al. Intraosseous or Intravenous Vascular Access for Out-of-Hospital Cardiac Arrest. N Engl J Med. 2025;392(4):349-360.
Wigginton JG, Duff JP, Andersen LW, et al. Part 9: Adult Advanced Life Support: 2025 American Heart Association Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. Circulation. 2025;152(suppl 2):S374-S423.
AI-Assisted Chest X-Rays in the ED: Useful Second Reader, Dangerous Autopilot
Every emergency physician has had the same chest x-ray moment: you are juggling six things at once, the film looks mostly fine at first glance, and then twenty minutes later you realize there was a pneumothorax, a subtle infiltrate, or a widened mediastinum sitting there the whole time. That is the exact kind of cognitive miss AI is supposed to help with. The real question is whether it actually improves decisions, or just adds one more noisy box to click through.
The best current ED-specific paper is a 2025 Emergency Medicine Journal multi-reader, multi-case study by Lyell and colleagues. Two hundred Australian emergency doctors worked through 18 chest x-ray clinical vignettes, with half randomly assisted by a commercial AI tool capable of flagging 124 findings. AI assistance improved correct diagnosis by 5.9% and improved patient management decisions by 3.2%. Confidence increased, and interpretation time barely moved. The biggest gain was seen in senior residents, where diagnostic accuracy improved by 11.8%.
That is the optimistic half of the story, and it is real. But the paper also built in an important stress test: four vignettes used chest x-rays known to trigger incorrect AI findings. When the AI was wrong in the false-positive direction, diagnosis barely changed. When it was wrong in the false-negative direction, correct diagnosis worsened, a finding the authors conclude carries meaningfully more risk than false-positive AI errors. That is the whole issue in one sentence. AI is most helpful when it highlights something you might have missed. It is most dangerous when it reassures you about something that is actually there.
So the question is not whether AI can help. It can. The question is what role you assign it. Used as a second reader, especially for less experienced clinicians or for high-throughput chest x-ray decisions, it probably has real value. Used as a shortcut to stop thinking, it becomes another source of overconfidence. That distinction matters more than the software brand.
This also explains why AI chest x-ray tools may be more useful in some environments than others. They are probably best at tightening up the middle of the distribution: the busy resident, the overnight attending with too many parallel tasks, the subtle but not invisible miss. They are much less trustworthy as final authority when the stakes are high or the film and clinical picture do not match.
Bottom line: AI assistance can modestly improve chest x-ray interpretation and downstream management decisions in the ED, but the benefit comes from using it as a second reader, not a substitute reader. When the algorithm misses a finding, blindly trusting it can make your diagnosis worse.
Lyell D, Dinh M, Gillett M, et al. Evaluating the impact of AI assistance on decision-making in emergency doctors interpreting chest X-rays: a multi-reader multi-case study. Emerg Med J. 2025;42(12):774-782.
Intranasal vs Subcutaneous Ketamine for Acute Traumatic Pain: Pick the Route That Fits the Room
The appeal of non-IV ketamine in trauma is obvious. Pain is severe, IV access may not exist yet, opioids are not always the right move, and sometimes the fastest way to get analgesia started is the route you can use immediately. What has been less clear is which non-IV route actually gives you the better tradeoff between speed, pain relief, and side effects.
Dhaoui and colleagues tried to answer that in a 2025 Annals of Emergency Medicine randomized trial involving 1,194 adults with acute musculoskeletal trauma. Patients received fixed-dose ketamine 20 mg either intranasally or subcutaneously. At 30 minutes, the intranasal group had a larger reduction in pain score than the subcutaneous group, but the difference was only 0.72 points on the NRS, statistically significant, yes, but below the prespecified 1.3-point threshold for clinical importance. The same basic pattern held across the other time points. There was no meaningful difference in the secondary outcomes, except that minor adverse events were more common in the subcutaneous group.
That is a very EM result. One route was a little better on paper, but not enough to matter at the bedside in most cases. So now the question becomes operational rather than ideological. If the patient is cooperative, not bleeding from the nose, and you want the least invasive option, intranasal ketamine makes plenty of sense. If nasal delivery is impractical or poorly tolerated and a quick subcutaneous dose is easier in the situation you are standing in, that remains a perfectly defensible choice.
The more important lesson is what not to do with this trial. Do not oversell the intranasal route as some major analgesic breakthrough, and do not read the paper as proof that subcutaneous ketamine is obsolete. This was a fixed-dose comparison, not a universal route hierarchy. The trial tells you these two approaches are broadly comparable, with a slight convenience and side-effect argument in favor of the nose rather than the needle.
That is useful, because traumatic pain management in the ED is often constrained less by pharmacology than by logistics. The patient in front of you does not care which route wins a conference debate. They care how fast you can give something that helps.
Bottom line: In acute musculoskeletal trauma, intranasal and subcutaneous ketamine produced very similar pain relief, with intranasal showing only a small numerical advantage that did not reach clinical importance. If you do not have IV access yet, either route is reasonable. Choose based on speed, practicality, and patient factors rather than hype.
Dhaoui R, Kouraichi C, Toumia M, et al. Intranasal Versus Subcutaneous Ketamine for the Treatment of Acute Traumatic Pain in the Emergency Department: A Randomized Clinical Trial. Ann Emerg Med. 2025.
🔭 Next Week
IV saline for acute migraine in the ED: does the liter actually help, or are we ritualizing fluids?
Sphenopalatine ganglion block for ED headache: does a bigger bupivacaine dose buy anything?
Difficult IV access: does venous transillumination improve first-stick success enough to matter?
The Hallway Consult is built for EM clinicians who want the useful version of the literature. Forward it to a colleague if it helped.
— The Hallway Consult team