The patient says her back is burning. The pain spreads into her buttocks and down both legs, worse than anything the surgery was supposed to cause. Then her legs begin to jerk. The jerking becomes a seizure.

Midazolam doesn’t stop it. A second dose doesn’t stop it. Keppra doesn’t stop it. Her heart rate climbs. Her blood pressure climbs. And somewhere in that sequence, someone finally looks at the syringe.

That is not a description of what happened in Nashville last week.

It is a description of a syndrome that has been appearing in the medical literature since 1988, when the first accidental intrathecal injection of tranexamic acid (TXA) was published.¹

It has happened often enough, in enough countries, that clinicians who have never seen it can recognize it from the first few sentences.

The consequences are catastrophic.

Mortality has been reported near 50 percent.² A 2023 narrative review identified 22 reports of intrathecal TXA toxicity in the four years between July 2018 and September 2022. Of those patients, 36 percent died and another 19 percent suffered permanent harm, including muscle weakness, chronic pain, spinal fractures caused by the seizures, and prolonged disability.³

This is not a new hazard.

And that is what makes it so difficult to accept.

The warning exists. The literature exists. The recommendations exist. The technology exists. And yet the same category of error keeps happening.

We Already Knew the Failure Mode

APSF published one of these cases in detail in 2024.⁴ It is worth walking through because there was nothing exotic about it.

A 67-year-old man came in for a left total knee replacement.

The hospital already had a policy for TXA. It was supposed to be ordered from pharmacy in prefilled infusion bags. But that policy was seldom followed in that particular operating room.

So the anesthesia professional did what the room routinely did: pulled 10 mL vials of both TXA and bupivacaine from the automated dispensing cabinet.

He drew up what he believed was bupivacaine.

The syringe was labeled Marcaine and fentanyl.

The label was correct.

The contents were not.

The block did not work. The patient reported itching in his perineum. The team interpreted that as a failed block and converted to a general anesthetic. There was some minor twitching in his legs.

It was only later in the case, when someone reached for the TXA, that they discovered the TXA vial was already open and the bupivacaine vial was still sealed on the anesthesia cart.

The patient seized in the PACU.

He went to the neurological ICU, underwent cerebral spinal lavage, developed myoclonic status epilepticus, and remained intubated until day 14. He was discharged to a rehabilitation hospital on day 23. He tore a rotator cuff during the seizures.

His neurologist documented a return to baseline roughly 13 months after a knee replacement.

It’s worth reading that again.

The fix was already policy.

The room just didn’t follow it.

So Why Does TXA Keep Ending Up in a Spinal Needle?

TXA belongs in a vein. In the subarachnoid space, it behaves like a neurotoxin. The FDA now says so on the label in a boxed warning.⁵

So the question is not:

How could someone be so careless?

The better question is:

How does a drug with this history keep ending up in a spinal needle?

APSF and ISMP answered that question in a joint statement in November 2025.⁶ And their answer deserves to be read slowly, because almost none of it is about attention.

TXA and local anesthetics are stocked in similar containers, in similar locations. Both are used during the same operations. The majority of these catastrophic errors have occurred during orthopedic arthroplasty. And they are stocked in areas of the hospital where barcode scanning is not routinely used, including operating rooms, labor and delivery, and emergency departments.

So the error is unlikely to be caught before the drug reaches the patient.

The problem is almost painfully simple: two clear, colorless drugs, in look-alike containers, stored side by side, needed minutes apart and used in the same case. And they are being used in the one part of the hospital that has never fully adopted the technology capable of telling them apart.

Then add the connector.

APSF and ISMP were blunt about this:

Education and double checks are not enough on their own.⁶

Then they told us what to do.

First: Eliminate the Dangerous Presentation

Get TXA vials and ampules out of perioperative areas entirely. Stock ready-to-administer premixed bags instead.

Second: Put High-Leverage Safeguards in Place

Three of them:

1. Barcode scanning or another machine-readable check at the point of care before administration.

2. Forcing functions such as NRFit connectors that physically cannot mate with a Luer syringe.

3. Standardized workflows that reduce the need for clinician-prepared syringes.

Third: Eliminate Unsafe Drug Presentations

A drug’s “presentation” is simply the physical form in which it arrives: a vial, an ampule, a prefilled syringe, or a premixed bag.

When APSF and ISMP call a presentation unsafe, they mean the container itself contributes to the error.

A TXA vial that looks like a bupivacaine vial is an unsafe presentation.

And here is the scale of the problem: thirteen companies manufacture TXA for the U.S. market. Eighty-one-point-five percent of what they sell comes in vials. Only 16.9 percent comes in premixed bags.⁴

The safer presentation already exists. It is simply not what most hospitals are buying.

Why?

Because the premixed bag costs more per dose.

So eliminating the vial becomes a purchasing decision made several floors above the operating room, by people who will never meet the patient it protects.

APSF’s authors have an answer for that objection: the price of a premixed bag varies by region, contract, and purchasing group, but it is inconsequential compared with the cost of an error that causes serious injury or death.⁴

Fourth: Protect the People Involved

Build non-punitive reporting for near misses. Provide peer support for clinicians involved when something goes wrong.⁶

That fourth recommendation may ultimately be the most important for the people standing in the room when the system fails.

But most clinicians reading the list will probably move past the second recommendation.

NRFit.

It is the one nobody talks about.

It deserves a few minutes.

The Connector That Says No

The Luer connector is one of the most successful designs in the history of medicine.

It is also one of the most dangerous.

It is a simple friction taper, and for most of the last century it was the connector for nearly everything: IV lines, hypodermic needles, epidural catheters, spinal needles, enteral feeding tubes.

Universal compatibility was the whole point. It made equipment interchangeable across every department in the hospital.

But that same universal compatibility created a vulnerability.

Any syringe fits any port.

The connector cannot know that the drug in your hand was never meant to go where that needle is pointing.

It offers no resistance.

Two deaths in England helped force the issue.

In 2001, an 18-year-old leukemia patient named Wayne Jowett received vincristine into his spine instead of a vein and died.

Years later, a nurse named Mayra Cabrera died when bupivacaine intended for her epidural was infused into a vein instead, and the NHS trust involved was prosecuted.⁷

Two errors running in opposite directions.

Same root cause.

Nothing in the equipment said no.

The International Organization for Standardization responded with the ISO 80369 series, assigning physically distinct connectors to different routes of administration.

ISO 80369-6 is the neuraxial section. It was published on March 15, 2016.

The connector is called NRFit.⁸

It works by geometry.

NRFit is about 20 percent smaller in diameter than Luer. It looks similar enough that you have to be told what it is.

But it will not mate.⁹

An intravenous syringe of TXA physically cannot connect to an NRFit spinal needle.

There are no adapters.

And that is deliberate.

An adapter would restore the very vulnerability the standard was designed to remove.

Many manufacturers use yellow for neuraxial components, but the color is not the safeguard.

The safeguard is that the parts do not fit.

That is what a forcing function actually is.

And most of us have used one this week without thinking about it.

Think About Your Car

You cannot put diesel into a gasoline car.

Not because a sticker warns you. Not because the label on the pump is a different color. Not because someone double-checks you before you squeeze the handle.

The diesel nozzle is physically too large for the filler neck.

The car does not care if you are tired, distracted, talking to someone, or reading the wrong side of the pump.

The parts do not fit.

So the fuel does not go in.

We built that protection into automobiles in 1975, when the EPA began requiring a restrictor in the fuel inlet of catalytic-converter vehicles so that the larger leaded-fuel nozzle could not be inserted.¹⁰

Fifty years ago, the industry decided a warning label was not adequate protection for an emissions system.

NRFit applies the same logic to the spinal cord.

It does not depend on the clinician being rested. It does not depend on nobody walking into the room at the wrong moment. It does not fire an alert that a human being can dismiss.

It simply does not go together.

Japan has already done this.

Every hospital in the country completed the conversion to NRFit, announced in early 2021.¹¹ The United Kingdom has backed the transition nationally. In the United States, California is the only state that has mandated it.¹²

There is no national timeline. Facilities are free to convert at their own pace.

In practice, most have not.

The APSF authors describe the comparable American effort in a single sentence:

It met resistance.⁴

The standard has existed for ten years.

We Know Which Safeguards Work

ISMP ranks safety strategies.

Education, training, and policies come near the bottom. They ask people to try harder.

Forcing functions, barriers, fail-safes, and automation come first.

They remove the hazard.¹³

Now look at the three recommendations again.

Only one of them is free.

Barcode verification at the point of care costs money, integration work, and a fight with a culture that has resisted scanning in operating rooms for years.

NRFit means replacing needles, syringes, catheters, and filters across an entire service line and retraining everyone who touches them.

Moving preparation out of the clinician’s hands and into pharmacy costs essentially nothing.

It is a policy memo. It can be written on a Tuesday and reported to the quality committee on a Wednesday.

So that is the leg many places took.

It is also the only one of the three that ISMP would call low leverage.

And then came Nashville.

August 14, 2026

Four patients arrived at Ascension Saint Thomas Hospital Midtown for outpatient joint replacements.

According to the hospital, a medication error occurred in the pharmacy, and all four received potassium phosphate instead of the intended anesthetic, mepivacaine.¹⁴

One patient, 72-year-old Glenda Dorton, has paralysis at T6.

Another, 74 years old, was placed on a ventilator.

The hospital self-reported the event to state regulators the day it occurred. The Tennessee Bureau of Investigation is now involved.¹⁵

Different drugs than the TXA cases.

Same route.

Same operation.

Same category.

Wrong drug. Wrong route.

But there is an important distinction.

NRFit would not have saved these four patients.

And that matters.

Because it exposes exactly why a safety bundle is necessary.

Go Back to the Fuel Pump

Imagine the tanker driver accidentally drops diesel into the station’s unleaded tank that morning.

You pull up.

The nozzle says unleaded. It fits your car perfectly.

The restrictor has no objection because the restrictor was never designed to check what is in the tank. It only checks what is on the end of the hose.

Every car that pulls in behind you gets the same fuel.

That is Nashville.

NRFit prevents the wrong route. It stops a drug drawn up in an intravenous syringe from ever reaching the neuraxial space. It would have stopped every one of those TXA cases cold.

But it does nothing about the wrong drug.

If a pharmacy loads a neuraxial syringe with the wrong medication, the connector mates perfectly because it is the correct syringe. Only verification of the source container catches that error.

Which is the whole point of the bundle.

The safeguards are not redundant. They close different doors.

NRFit closes wrong route.

Barcode verification closes wrong drug.

Standardized preparation reduces the number of hands in the process and moves the remaining risk upstream, where the other two safeguards have a chance to catch it.

Take away one leg and you have not implemented a bundle.

You have relocated the problem.

And there is another problem hiding inside that relocation.

The vial that came from the manufacturer had a machine-readable barcode: a unique national drug code identifying the product, concentration, and manufacturer.

A scanner can verify it in under two seconds.

The moment those contents are drawn into a syringe, that barcode is gone.

What replaces it is a label generated by a person and applied by a person.

Every check after that point is one human reading what another human typed.

Ask the anesthesiologist in that knee replacement case what the label said.

It said mepivacaine.

We moved preparation into the safest hands in the building and, in the same motion, deleted the only machine-verifiable fact in the chain.

That is the lesson.

Not that pharmacists are unsafe.

Not that anesthesiologists are unsafe.

Not that nurses are unsafe.

That the system is still asking humans to be the last line of defense when technology could have been the last line of defense instead.

We Do Not Need Another Case to Write the Fix

I do not know what happened inside that pharmacy. Nobody outside of it does yet. The investigation will take months, and I am not going to guess.

But the literature said this could happen.

It said so in writing before August 14.

There is a fourth recommendation in the APSF and ISMP statement, and it is the one that will probably get the least attention:

Provide peer support for practitioners involved in adverse events.⁶

That matters.

Because the lesson of Nashville is not that a pharmacy made a mistake.

The lesson is that we were told, in writing, by the two organizations whose entire purpose is preventing exactly this kind of event, what the failure modes were and which safeguards would address them.

We took the one that was free.

The standard for a connector that cannot be misconnected has existed since 2016. Japan converted an entire country. The recommendations are still on the APSF website.

They were there before August 14.

They are there now.

We do not need another patient to write the fix for us.

The fix is already written.

That is the argument.

But it is not what I keep thinking about.

The Second Patient

Glenda Dorton is 72 years old.

She went to the hospital on a Friday morning to have a joint replaced—a procedure people have so they can keep walking.

She is now paralyzed from the chest down.

Three other people went into that hospital that same morning for the same reason. Their families spent the next week in waiting rooms learning words they never expected to know.

Nothing in this article changes what happened to them. I know that.

But I keep thinking about the other people in that building that morning: the clinicians, pharmacists, nurses and anesthesiologists who came to work expecting to take care of four patients and went home carrying something they may carry for the rest of their lives.

If you were one of them, you are probably replaying that morning right now. You are looking for the moment—the vial, the label, the syringe, the conversation, the person who walked into the room, the thing you should have noticed.

You will find one.

You will probably find several.

That is what the mind does after an event like this. It goes backward and searches for the moment when everything could have been different.

But finding that moment is not the same thing as understanding why the system allowed it to exist. And it is not the same thing as proving that you are a bad clinician.

Albert Wu named this phenomenon in 2000: the “second victim.” Susan Scott and others later described what happens to healthcare workers in the months that follow.¹⁶

I prefer another term.

Second patient.

Because that is what you are.

You are a person who was hurt by a patient-safety event too.

Whatever you did or did not do that morning, it is not the whole of your career. It is not the whole of you.

You have taken care of thousands of people who are out there right now, living their lives. Raising their children. Going to work. Walking into grocery stores. Getting on airplanes. Complaining about traffic.

They will never know your name. They will never know what you did for them.

But they are alive because of the care you gave them.

That ledger is real, even when you cannot see it.

So find someone to talk to. Not just your employer. Not just the people conducting the review. Find someone who understands what it means to carry this.

And if you are the person who pushed that syringe, I want you to hear something very clearly:

You administered a preparation that came from the pharmacy. That was not a shortcut. That was not you deciding to ignore a safer process. It was the higher-leverage practice—the one APSF and ISMP recommend over drawing the medication up yourself.

You did the safer thing.

To whoever runs the review, I hope these people get a fair process.

Not a soft one.

A fair one.

Those are not the same.

The severity of an outcome should never determine the severity of our judgment of the person involved. What someone did and what resulted from it are two separate questions. A Just Culture review keeps them separate on purpose.

Put the people involved in the room. Tell them what you find. Tell them again when it is finished.

And then fix what needs to be fixed.

Because four patients went into a hospital in Nashville expecting to walk out with a new knee.

Instead, four families are learning how to live with something nobody chose.

And the people who cared for them are learning how to live with it too.

None of them should have had to teach us this.

So let’s make sure they are the last.

Not because the next clinician will be more careful. Not because the next pharmacist will double-check the label. Not because we will add another policy, another training module, or another reminder to “be vigilant.”

But because we finally built the system to say no.

The syringe should not fit. The barcode should not disappear. The wrong drug should not make it through.

The safest system is not the one that asks good people to make fewer mistakes. It is the one that protects good people—and their patients—when being human inevitably happens.

We already know how to build that system.

Now we have to decide that another family should not have to pay the price before we do it.

Continue the Conversation

What happens when medication scanning becomes mandatory?

This Wednesday, 8/26/26 at 1400 EST, I’ll be joined by leaders from three EMS systems that have implemented ScanSafe and made medication scanning mandatory.

We’ll discuss what happened in the real world; the clinician response, the challenges, what worked, what didn’t, and what they’ve learned since implementation.

Join us Wednesday for a conversation about turning medication safety from a policy into a practice.

REGISTER FOR THE WEBINAR

References

1. Wong JO, Yang SF, Tsai MH. Accidental injection of tranexamic acid (Transamin) during spinal anesthesia. Ma Zui Xue Za Zhi. 1988;26(2):249-252.

2. Patel S, Robertson B, McConachie I. Catastrophic drug errors involving tranexamic acid administered during spinal anaesthesia. Anaesthesia. 2019;74:904-914.

3. Patel S. Tranexamic acid associated intrathecal toxicity during spinal anesthesia: a narrative review of 22 recent reports. Eur J Anaesthesiol. 2023;40:334-342.

4. Lefebvre PA, Meyer T, Lindsey A, Jew RK, Rebello E. Unraveling a recurrent wrong drug-wrong route error, tranexamic acid in place of bupivacaine: a multistakeholder approach to addressing this important patient safety issue. APSF Newsletter. June 2024;39(2).

5. US Food and Drug Administration. FDA provides update to health care professionals about the risk of inadvertent intrathecal (spinal) administration of tranexamic acid injection.

6. Anesthesia Patient Safety Foundation and Institute for Safe Medication Practices. Preventing wrong drug-wrong route errors involving tranexamic acid and local anesthetics. Joint announcement. November 26, 2025.

7. BBC News. Anger as fatal jab doctor freed. September 23, 2003. Tran M. NHS trust prosecuted over new mother’s death from wrong drug. The Guardian. November 18, 2009.

8. International Organization for Standardization. ISO 80369-6:2016, small-bore connectors for liquids and gases in healthcare applications, part 6: connectors for neuraxial applications. Published March 15, 2016. See also American Society of Regional Anesthesia and Pain Medicine. What is ISO 80369-6:2016? ASRA News. July 2018.

9. Global Engineered Device Supplier Association. NRFit frequently asked questions. Stay Connected initiative.

10. US Environmental Protection Agency. Prohibition on gasoline containing lead or lead additives for highway use: fuel inlet restrictor exemption for motorcycles. Federal Register. May 24, 2002. Describes the fuel inlet restrictor requirement adopted for catalytic converter vehicles beginning with the 1975 model year.

11. Global Engineered Device Supplier Association. Japan health system improves patient safety with adoption of NRFit neuraxial connectors. February 25, 2021.

12. Institute for Safe Medication Practices. NRFit: a global fit for neuraxial medication safety. Nurse Advise-ERR. August 2020.

13. Institute for Safe Medication Practices. Implement strategies to prevent persistent medication errors and hazards. ISMP Medication Safety Alert! Acute Care. 2023;28:1-4.

14. Ascension. Ascension Saint Thomas Midtown update. August 21, 2026.

15. CBS News. Drug mix-up at Nashville hospital causes dire complications in 4 patients. August 21, 2026. Additional reporting from WSMV, WKRN, and NewsChannel 5 Investigates, August 19 to 21, 2026.

16. Wu AW. Medical error: the second victim. BMJ. 2000;320:726-727. Scott SD, Hirschinger LE, Cox KR, et al. The natural history of recovery for the healthcare provider second victim after adverse patient events. Qual Saf Health Care. 2009;18:325-330.

17. Antevy P. Every sentinel event has a second patient. They are back on shift tomorrow. August 2026.

18. Institute for Safe Medication Practices. Dangerous wrong-route errors with tranexamic acid. ISMP Medication Safety Alert! Acute Care. 2019;24:1-2.

19. Institute for Safe Medication Practices. Guidelines for safe medication use in perioperative and procedural settings. 2022.

Somewhere tonight, a paramedic, EMT, nurse, or physician is awake at 2 a.m. replaying a call.

Something went wrong.

Maybe they made a mistake. Maybe they did not. Either way, the questions are the same.

Did I miss something? Could I have done something differently? What does everyone think of me now?

I know those questions because I have lived them.

About 15 years ago, a young boy came into my emergency department because his back was itching.

That was it. He was running up and down the hallway. His neurologic exam was normal. His strength, reflexes, sensation and gait were normal. I obtained X-rays, which were also normal. I sent him home.

The next day he was diagnosed with Ewing sarcoma of the spine. What he had described as itching was actually paresthesia.

Thankfully, he did well. He is an adult now, and I still keep in touch with his family.

But the hospital performed a root cause analysis of his case.

I was not invited.

Nobody asked me what I saw that night. Nobody asked why I made the decisions I made. Other people sat in a room and reviewed my care without the one person who had actually been standing in that room with the patient.

Eventually, I learned that some people on the committee defended my care and felt the workup had been appropriate.

I was grateful.

But looking back, I should not have had to be grateful that the right people happened to be sitting in the room.

And nobody ever sat down with me afterward and said, here is what we found, here is what we think happened.

So I carried it myself.

For years.

I have also made a real error.

In 2005, I gave a child ten times the medication dose I intended to give. I have spoken publicly about that mistake many times.

Those two cases were very different. In one, I made an error. In the other, I believe my clinical judgment was reasonable.

But emotionally, they had something in common.

I was left alone with both of them.

That experience has changed the way I think about quality assurance.

We usually think QA exists to protect patients. Find the error. Review the chart. Fix the protocol. Educate the clinician. Prevent it from happening again.

Of course it should do all of those things.

But there is another patient in the room.

The clinician.

The literature calls this the second victim. After a serious adverse event, clinicians can experience guilt, anxiety, loss of confidence, intrusive thoughts and fear about how their colleagues now see them.^1,2

EMS may be particularly unforgiving, because after the review is over, that clinician goes right back on the truck.

I saw this recently after a sentinel event at one of my agencies.

I will not share the details because they do not matter.

What matters is what happened when we brought the people involved into the room.

We talked about everything.

What happened. What decisions were made. What the system contributed. What could have been done differently. What needed to change.

Nobody minimized anything.

But nobody was looking for someone to blame, either.

And I watched the people in that room change.

The tension disappeared.

Then something happened I did not expect.

The clinicians closest to the event asked if they could help teach the rest of the department about what had happened.

That moment radically changed the way I think about QA.

And we decided not to leave that kind of experience to chance.

We have since put a formal QA policy in place that ensures this process happens. When a serious event occurs, the people involved have a voice, the case is reviewed fairly and promptly, the findings are shared with them, and support is part of the process. Not because someone happens to remember to do it, but because that is now how we do QA.

A good quality process does not just identify what went wrong.

It gives the people involved a voice.

It separates human error from reckless behavior.

It looks at the system instead of automatically blaming the person closest to the patient.

It tells the clinician what was found instead of leaving them wondering. The question stops being asked at two in the morning because it finally got answered at ten.

And when appropriate, it gives them the opportunity to turn one of the worst calls of their career into something that protects the next patient.

That last part matters more than it sounds. When researchers interviewed clinicians about their recovery after adverse events, several described getting involved in improvement work as the thing that let them enjoy the job again.²

I have spent much of my career trying to build systems that prevent someone else from making the medication error I made in 2005.

I could turn that experience into something.

But I could not build my way out of the other case.

What I needed then was not another technology.

I needed a room.

I needed someone to ask me what happened.

I needed to hear what the review concluded.

And I needed to know that one case did not define me as a physician.

It took me years to figure that out on my own.

Some clinicians never do.

They become quieter. They stop volunteering for difficult calls. They lose confidence. They transfer. Sometimes they leave the profession entirely.

We call it burnout without ever asking whether it started with one call that nobody helped them process.

QA absolutely exists to protect the next patient.

But if we do it correctly, it can also take care of the clinician who cared for the last one.

That boy with the itching back is a grown man now. His family and I still talk. They never once made me feel like I had failed them.

The silence afterward did that.

Fifteen years ago, nobody asked me what happened.

A few weeks ago, I watched a group of clinicians get asked that question, answer it honestly, and walk out of the room lighter than they walked in.

None of them are lying awake tonight wondering what their department thinks of them.

There were two patients in that room.

This time, we took care of both.

References

1. Wu AW. Medical error: the second victim. The doctor who makes the mistake needs help too. BMJ. 2000;320(7237):726-727.

2. Scott SD, Hirschinger LE, Cox KR, McCoig M, Brandt J, Hall LW. The natural history of recovery for the healthcare provider “second victim” after adverse patient events. Qual Saf Health Care. 2009;18(5):325-330.

Picture an ED nurse, mid-shift, running a routine medication pass. She scans the patient’s wristband. She scans the barcode on the medication. The system checks it against the order, confirms the match, and logs the dose. It takes four seconds. Just then she gets called out by the charge nurse. Code blue, three doors down. She drops what she is doing and runs. Thirty seconds later she is standing in a resuscitation bay, drawing epinephrine out of a vial, and the exact same system that worked perfectly for her a moment ago has nothing to offer her now. No patient logged in yet. No order sitting in the record. The barcode on that vial means nothing to the software four doors behind her.

To see why, you have to see what barcode scanning actually depends on, on an ordinary day, in that ordinary room she just left. A physician examines the patient, sits down at a computer, and places an order in the electronic health record. A pharmacist reviews that order, and for many medications, prepares it by hand, mixing the drug, printing a label specific to that patient and that dose, and sending it up through a pneumatic tube. Only then does a nurse come to the bedside, log into the chart, scan the barcode on the wristband, and scan the barcode on the medication. That last barcode is worth pausing on. It is not the manufacturer’s barcode. It is not the NDC printed on the vial at the factory. It is a custom label the hospital’s own software generated, built to match an order a pharmacist already verified. The scan itself is the last click in a much longer chain. Physician order, pharmacist verification, custom label, bedside scan. Remove any one link and there is nothing left to scan against.

A code does not remove one link. It removes all of them at once, and it does so by definition, not by accident. There is no order in the EHR, not yet, sometimes not ever. There is no pharmacist in the loop, no custom label, no prepared bag arriving from downstairs. The nurse is drawing up epinephrine, amiodarone, or norepinephrine from a raw vial, in real time, under enormous pressure, off a verbal order called across a crowded room. The barcode infrastructure that hospital spent millions of dollars building is not malfunctioning in that moment. It simply has nothing to check. The safety net has a hole exactly where the patient is most vulnerable, and the hole is not a bug. It is the shape of what a code room is.

Now go one step further. Out the door, into the back of a moving ambulance. It is the identical missing chain, except EMS never had it to lose in the first place. There is no EHR order to write against in a moving rig. There is no pharmacist verifying a dose before a paramedic draws it up. A paramedic pulling epinephrine for a pediatric arrest is standing in the exact same hole as the nurse three doors down from the hospital pharmacy, working off the same kind of verbal order, under the same kind of pressure, with the same nothing to scan against.

That is the concept this whole piece rests on. The code room and the back of an ambulance are not similar problems. They are the same problem. Hospital barcode technology was built for a chain that exists on the general floor and evaporates completely in both places. Twenty years of hospital data on this technology is, almost without exception, data about the setting where the chain is intact. Nobody has good data on what happens once it disappears, not in a hospital, and not in an ambulance.

That is worth sitting with, because it changes what “twenty years of data” actually means. In 2008, Ross Koppel and his team spent three years inside five hospitals watching barcode medication systems in real use, in intensive care and medical surgical units.¹ In 2010, Eric Poon and his colleagues ran their landmark trial across thirty five medical, surgical, and intensive care units, and deliberately left oncology out because its dosing and workflow were too complex to fit the model.² Both studies describe careful, well-resourced hospitals with the full chain in place, an admitted patient, a wristband already on, an order already sitting in the record. Neither one describes what happens once that chain is gone.

It gets more specific than that, and more telling. Barcode scanning has been studied in the emergency department. Bonkowski and colleagues found an 80.7 percent drop in medication administration errors, from 6.3 percent to 1.2 percent, after implementing it in an academic ED.³ But that was routine ED medication administration, an order placed, an order verified, the chain intact in miniature, the same as any other ward. The one study that ever pointed barcode technology at an actual resuscitation ran in 1993, and it compared barcode entry to handwritten notes for accuracy of trauma documentation.⁴ It never touched medication dosing at all. Koppel’s own taxonomy names the reason why the chain breaks down here specifically. Emergencies are listed as a direct cause of workarounds, because stat orders in critical care leave nurses with nothing yet for the system to check against.¹ And when the American Academy of Pediatrics and the American College of Emergency Physicians published their joint safety guideline for pediatric emergency care, they recommended barcode scanning to verify patients and medications in general, and then, for code situations specifically, recommended something else entirely: a second clinician reading the dose back out loud.⁵ That is not a footnote buried in the guideline. That is the two organizations that write the rules for pediatric emergency safety quietly conceding, in the structure of their own recommendation, that barcode scanning does not reach the one room where a wrong dose is most likely to kill a child. Not because nobody tried. Because the technology, built the way it has always been built, was never able to reach that room. There was never a chain there to scan against.

Here is the fact that should stop every one of us in our tracks anyway. Nobody has ever studied barcode medication scanning in an ambulance. Not once. Not one published study, anywhere, has looked at it for error reduction in the prehospital setting.⁶ Hospitals have twenty years of data about the chain-intact world, a handful of pages about the chain-broken world inside their own walls, and nothing at all about EMS. We are not the next chapter in this story. We are the blank page.

That is either terrifying or it is the best opportunity this field has had in a decade, and I choose to see it as the second one. The twenty years hospitals spent are not wasted years. They are a map, even if the map has a hole drawn right where we need it most. All we have to do is read it honestly.

That optimism goes back to 2003, when Bates and Gawande wrote that barcoded wristbands could stop the wrong drug from reaching the wrong patient.⁷ They were right, but their vision assumed the chain, an order, a pharmacy, a record to check against, and nobody was yet asking what happens when that chain isn’t there. Five years later, even in the best case units built for this technology to succeed, nurses overrode the barcode system one in ten times.¹ Two years after that, done right, the same kind of system cut medication errors by 41 percent, and it still missed one in five doses.² None of it improved with time, and this year EMS physicians writing about prehospital medication errors could not cite a single barcode study, because none exists, even with pediatric error rates running as high as 35 percent.⁶ We are walking into a problem at least as large as the one hospitals faced, in a setting the hospital literature never tested.

That is the reason ScanSafe, our newest feature in Handtevy, does not try to recreate the hospital’s chain in the back of an ambulance. It reads the barcode already on the vial, the one the manufacturer printed at the factory, the one every hospital pharmacy and every ambulance drug box already has, rather than waiting on a custom label a proprietary system was never going to generate in a moving rig. It is why it can start working in any ED code room or any ambulance with no interface build and no order to match against first. It scans what is actually in the clinician’s hand, not what a database expected to find there, because in the exact moment we care about most, there is no database that has caught up yet.

None of that solves the physical problem on its own, though. A scanner in a moving ambulance will fail to read for the same reasons a hospital scanner failed, bad lighting, gloves, motion, dead battery, and a paramedic will do exactly what the nurse did: find a way to get the medication to the patient anyway. That is not a hypothetical. That is a guarantee, unless we design for it from day one.

So forget “does it scan.” That was always the wrong bar. A scan confirms identity. It does not confirm the plan, and in a code, the plan is everything. Handtevy Mobile has precalculated and customized medication dosing and equipment for every organization it serves, EMS or hospital, since the day it launched. Handtevy’s ScanSafe feature builds directly on top of that. It matches the scanned vial not just against the medication, but against the indication, push dose or drip, so the same vial can mean two different things depending on what the clinician is actually doing. Scan a vial of epinephrine 1 mg/mL while the clinician has already selected epinephrine IM for anaphylaxis in the app, and it confirms the vial matches what they chose. Scan that exact same vial after the clinician has instead selected an epinephrine drip, and it flags dilution required, because the clinician already told the app which preparation they intend to give, and a push concentration does not match a drip selection. Same vial. Same barcode. Two completely different messages, not because the tool is deciding anything on its own, but because it is checking the vial in the clinician's hand against the choice the clinician already made. That is what "built for exactly this moment" actually means. Not a faster scan. A tool that confirms the clinician's own plan back to them, in the moment they have the least time to double-check it themselves.

Here is exactly what this is and is not. We built the first tool that works in both the code room and the back of an ambulance, reading the barcode already on the vial instead of waiting on an order and a label that were never going to exist in either room. Handtevy already sits in more than 100 hospitals and 3000 EMS systems, and ScanSafe, the barcode piece of it, went live with the same access for hospitals and EMS agencies at the same time. EMS jumped on it immediately. In just over a month, they have logged more than 5,000 scans in the field. Hospitals have that same access right now, and there is no real question that the code room is next. The medication nurse is the highest stress role in that room, by far, pulling drugs from memory against a clock nobody controls, and no hospital that has watched that role up close is going to leave it without a tool built for exactly that moment. Change in healthcare still takes time, and it should. That EMS traction is the leading edge of something bigger. Every one of those 5,000 scans is a data point that has never existed before, in a setting nobody has ever measured before. That is the real opportunity here, not just building the first tool that reaches the code room and the back of an ambulance, but building the evidence alongside it, in real rooms, under real pressure, instead of waiting twenty years to look back and wonder what we missed.

That is the vision behind this piece. Not that we already have every answer. That we get to be the ones who finally go looking, in the back of a moving ambulance and in the code room three doors down, at the exact moments that broke the hospital version of this tool, and that nobody thought to study even inside the hospital where it was born. Do that work well, and the nurse three doors down and the paramedic in the back of that ambulance get the one thing twenty years of hospital technology never gave them. A tool built for the exact moment they needed it, and proof that it works.

References

1. Koppel R, Wetterneck T, Telles JL, Karsh BT. Workarounds to Barcode Medication Administration Systems: Their Occurrences, Causes, and Threats to Patient Safety. J Am Med Inform Assoc. 2008;15(4):408-423. doi:10.1197/jamia.M2616.

2. Poon EG, Keohane CA, Yoon CS, et al. Effect of Bar-Code Technology on the Safety of Medication Administration. N Engl J Med. 2010;362(18):1698-1707. doi:10.1056/NEJMsa0907115.

3. Bonkowski J, Carnes C, Melucci J, et al. Effect of Barcode-Assisted Medication Administration on Emergency Department Medication Errors. Acad Emerg Med. 2013;20(8):801-806. doi:10.1111/acem.12189.

4. Chua RV, Cordell WH, Ernsting KL, Bock HC, Nyhuis AW. Accuracy of Bar Codes Versus Handwriting for Recording Trauma Resuscitation Events. Ann Emerg Med. 1993;22(10):1545-1550. doi:10.1016/s0196-0644(05)81256-9.

5. Joseph MM, Mahajan P, Snow SK, Ku BC, Saidinejad M. Optimizing Pediatric Patient Safety in the Emergency Care Setting. Pediatrics. 2022;150(5):e2022059674. doi:10.1542/peds.2022-059674.

6. DeFilippo M, Patel NT, Yeary J, Jaeger E. Medication Errors in EMS: A Call for Systems-Based Solutions. Emergency Medicine News. 2026;48(7):15-16.

7. Bates DW, Gawande AA. Improving Safety with Information Technology. N Engl J Med. 2003;348(25):2526-2534. doi:10.1056/NEJMsa020847.

I’ve written about this in the past, but the memory came rushing back this week while I was reading the opening chapter of Dr. Marty Makary’s latest book, Blind Spots: When Medicine Gets It Wrong, and What It Means for Our Health.¹ Huge thanks to Steve Wirth for the book recommendation. Makary previewed this chapter a year earlier in the Wall Street Journal, in an essay called How Pediatricians Created the Peanut Allergy Epidemic, published September 19, 2024.¹²

That title bites, and it’s 100% accurate.

Consider that in 1999 only six in a thousand American children had a peanut allergy, most of them mild. The American Academy of Pediatrics wanted to give parents an answer, so in 2000 it told them to withhold peanuts from high-risk infants until age three, and from pregnant and nursing mothers before that.² The data it was based on was very weak. The problem, as the AAP’s own authors acknowledged in the same document, was that nobody could reliably say which infants were high risk. A family history of allergy or asthma qualified under the loosest reading, which meant almost anyone qualified. It led to an entire generation of pediatricians, me included, teaching parents the same mnemonic. Age one, start milk. Age two, start eggs. Age three, start peanuts.

The pediatric allergists knew this wasn’t true all along, yet they were not invited into the room where the recommendation was created. One of them was Dr. Gideon Lack, a world-renowned pediatric allergist and a TIME100 Health honoree, whose LEAP trial showed the opposite was true. He demonstrated very clearly that early exposure cut the risk of peanut allergy by 86%.³ It took years for the guidance from the AAP to change. But the damage from a decade of avoidance was already written into a generation of immune systems, and into a country that now leads the world in peanut allergy. Makary calls this absolutism. Dressing a guess up as certainty because “we don’t know” is bad for public trust. He is right to name it. What bothers me is that we talk about it like it happened once, in one specialty, a long time ago.

Prehospital medicine has its own peanuts, and we are living inside at least three of them right now.

Start with ventilation rate in pediatric cardiac arrest. For years EMS clinicians were taught roughly ten breaths a minute once an advanced airway was placed, a number extrapolated mostly from adult and animal physiology suggesting that faster ventilation raised intrathoracic pressure and choked off coronary perfusion pressure during compressions. In 2020 the American Heart Association moved that number to 20-30 breaths a minute, doubling or tripling it depending on where in the range a clinician lands.⁴ The evidence behind that jump was a single pediatric in-hospital study of roughly fifty children, observational rather than randomized, showing higher rates associated with better survival. But the data were collected from PICU patients, 60% of whom had pre-existing congenital heart disease, 77% who were on pressors at the time of their “arrest,” and 74% of whom still had a pulse.

The AHA and the European Resuscitation Council both classify their own pediatric ventilation guidance as weak, built on very low-quality evidence, and both list optimal ventilation rate as a research gap requiring urgent study.⁵ The two councils do not even agree with each other. Europe still allows a range down to ten breaths a minute for older children. Meanwhile, a study published this year found that field crews are actually delivering an average of 7.5 breaths a minute during pediatric out-of-hospital arrest, well below either recommendation.⁶ We changed a national standard by as much as three hundred percent on the strength of one small observational study, two continents disagree on the number, and the clinicians doing the ventilating in the field have landed somewhere else entirely.

Then there is epinephrine dosing in symptomatic bradycardia. Current PALS guidance gives the identical dose, 0.01 mg/kg, to a child in cardiac arrest and to a child with a pulse who is bradycardic and poorly perfused. In 2020, Mathias Holmberg and colleagues used the American Heart Association’s own Get With The Guidelines-Resuscitation registry to look at what actually happens to these children, and the results were not reassuring. Children who received epinephrine for bradycardia with poor perfusion had lower survival to hospital discharge, lower rates of return of spontaneous circulation, lower 24-hour survival, and worse neurological outcomes at discharge than children who did not.⁷

Adult resuscitation abandoned this dosing logic years ago. When an adult has a pulse but a failing pressure, we no longer push a full code dose of epinephrine. We push a tenth of it, titrated, and call it “push dose.” The American Heart Association’s own 2018 scientific statement on children with cardiac disease said the same principle should apply to pediatrics, a starting dose of one microgram per kilogram, one tenth the arrest dose, for a myocardium that is failing but has not yet stopped.⁸ Eight years later, that guidance still has not made it into the mainstream PALS algorithm most clinicians are taught, despite a registry study showing the current dose is associated with worse outcomes. We give a full cardiac arrest dose of epinephrine to children who are, by definition, not in cardiac arrest, and we call it standard of care.

Then there is epinephrine in shockable rhythms, and this one might be the sharpest example of the three. For decades the reflex was the same no matter what the monitor showed. Push a milligram every three to five minutes. That logic holds for asystole and PEA, rhythms with no organized electrical activity, where vasoconstriction buys the heart time to respond to something. It never made the same sense in ventricular fibrillation, where the heart already has electrical activity and defibrillation is the definitive treatment. Push standard dose epinephrine into a fibrillating heart and you raise myocardial oxygen demand, promote recurrent VF after a successful shock, and impair the microvascular perfusion you are trying to restore.

The evidence has been building for years. PARAMEDIC2, the largest placebo controlled epinephrine trial ever run, found more patients got a pulse back but no improvement in survival with a good neurologic outcome, and whatever benefit existed converged to zero at about twenty minutes of pulselessness.⁹ A 2023 network meta-analysis in Chest found standard dose epinephrine improved survival to discharge for non-shockable rhythms but showed no benefit for shockable ones.¹⁰ A 2025 population based analysis of more than three thousand patients with refractory VF found a favorable neurologic outcome in half of the patients who did not receive epinephrine, compared with eleven percent of those who did.¹¹ That is a fourfold difference, in the patients with the best chance of walking out of the hospital. Palm Beach Gardens Fire Rescue removed epinephrine from its shockable rhythm protocol on the strength of this data, and the early outcomes support the change, though the sample is still small. The AHA’s own guidelines have been moving the same direction since 2020, deferring epinephrine in shockable rhythms rather than pushing it on the same clock as everything else. We are not at a settled national algorithm yet. But unlike the peanut story, the profession is at least looking hard at the data while the old guess is still being defended, instead of waiting fifteen years to admit it.

Makary’s answer for the peanut era is not a new set of absolute rules pointed in the other direction. It is a culture willing to say we do not know yet, and systems that let evidence update practice in years, not decades. That is the work in front of every EMS medical director right now, in ventilation rates, in epinephrine’s dosing, in epinephrine’s role in shockable arrest, and in every protocol we have not questioned recently enough.

Dr. Makary is not writing about EMS. He does not need to. The peanut is never really about the peanut. It is about how easily an institution mistakes its own certainty for the truth, and how long it takes a system built on protocol to admit it was wrong. Ours is no exception.

If this bothers you as much as it bothers me, read the rest of the book. Peanuts are only chapter one. Makary spends the rest of Blind Spots walking through the same pattern again and again, in the hormone replacement therapy panic that scared a generation of women off treatment that helped them, in the low-fat low-cholesterol dogma we handed out for decades, in the years experts insisted opioids were not addictive, in a fourteen-year ban on silicone breast implants with no real evidence behind it. Chapter after chapter, the same story. An institution takes a position, defends it long past the point the evidence supports, and eventually gets around to saying the two words that would have spared everyone years of harm. “We don’t know yet.”

Buy the book. Read one chapter a night. Then go back and read your own protocols with the same skepticism.

References

1. Makary M. Blind Spots: When Medicine Gets It Wrong, and What It Means for Our Health. Bloomsbury Publishing; 2024.

2. American Academy of Pediatrics, Committee on Nutrition. Hypoallergenic infant formulas. Pediatrics. 2000;106(2 Pt 1):346-349.

3. Du Toit G, Roberts G, Sayre PH, et al. Randomized trial of peanut consumption in infants at risk for peanut allergy. N Engl J Med. 2015;372(9):803-813.

4. Topjian AA, et al. Part 6: Pediatric Basic Life Support: 2025 American Heart Association and American Academy of Pediatrics Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. Circulation. 2025;152(suppl 2).

5. Optimal Ventilation for Cardiac Arrest, NIH-funded multicenter trial protocol (NCT07114510); AHA and European Resuscitation Council pediatric ventilation guidance, classified as weak recommendation, very low-quality evidence.

6. Ventilation Rates and Capnography in Pediatric Out-of-Hospital Cardiac Arrest with Advanced Airways. Prehosp Emerg Care. 2025 May 29 [online ahead of print]. doi:10.1080/10903127.2025.2496756.

7. Holmberg MJ, Ross CE, Yankama T, Roberts JS, Andersen LW. Epinephrine in children receiving cardiopulmonary resuscitation for bradycardia with poor perfusion. Resuscitation. 2020;149:200-207.

8. Marino BS, et al. Cardiopulmonary Resuscitation in Infants and Children With Cardiac Disease: A Scientific Statement From the American Heart Association. Circulation. 2018;137(22):e691-e782. [Verify exact citation before publication.]

9. Perkins GD, Ji C, Deakin CD, et al. A Randomized Trial of Epinephrine in Out-of-Hospital Cardiac Arrest. N Engl J Med. 2018;379(8):711-721.

10. Fernando SM, Mathew R, Sadeghirad B, et al. Epinephrine in Out-of-Hospital Cardiac Arrest: A Network Meta-analysis and Subgroup Analyses of Shockable and Nonshockable Rhythms. Chest. 2023;164(2):381-393.

11. Fanet L, Javaudin F, Dumas F, et al. Association of epinephrine and outcome in cardiac arrest with refractory shockable rhythm: a population-based, propensity-score matched analysis. Crit Care. 2025;29:252.

12. Makary M. How Pediatricians Created the Peanut Allergy Epidemic. Wall Street Journal. September 19, 2024.

Over the past two decades, I have run codes in emergency departments. I have stood at the head of the bed and called out drug orders in a room that is loud, fast, and unforgiving. I have watched nurses draw up medications from memory, from habit, and from whatever vial was closest. I have seen the right drug pulled in the wrong concentration. I have seen verbal orders misheard. I have seen the system work, and I have seen it fail quietly, with no alarm, no alert, and no record.

We talk a lot about medication safety in hospitals. We have built extraordinary infrastructure around it. Pharmacy verification. Automated dispensing cabinets. Electronic medication administration records. Barcode scanning at the bedside.

And yet there is one room where almost none of that applies.

The code room.

After two decades, I think we finally have an answer. But to understand why it took this long, you have to understand why the problem was so hard to name.

What barcode scanning actually requires

Modern hospital barcode scanning is built on a specific chain: a physician enters an order in the EHR, pharmacy reviews and verifies it, a pharmacist-prepared product with a hospital-generated label arrives at the bedside, and the nurse scans that label against the active order to confirm the right drug is going to the right patient.

It is a good system. When it works, it works well.

But in a resuscitation, that chain does not exist.

There is no order in the EHR - not yet, sometimes not ever. There is no pharmacy in the loop. There is no pharmacist-prepared bag. The nurse is drawing up epinephrine, amiodarone, or norepinephrine from raw vials, in real time, under enormous pressure, based on a verbal order called across a crowded room.

In that moment, the barcode scanning infrastructure that the hospital spent millions of dollars building is completely irrelevant. There is nothing to scan against. There is no verified order to close the loop on. The safety net has a hole exactly where the patient is most vulnerable.

The RaDonda Vaught case is a symptom, not an anomaly

In 2017, nurse RaDonda Vaught administered vecuronium instead of Versed to a patient at Vanderbilt University Medical Center. The patient died. Vaught was ultimately convicted of criminally negligent homicide.

The case drew fierce debate in the nursing community, and rightly so. Many argued, correctly, that the conditions that led to her error were systemic: an overridden cabinet alert, a broken safety culture, a high-stress environment, and a series of small failures that compounded into a catastrophic one.

What rarely gets named directly is this: in high-acuity, time-compressed clinical moments, human beings make errors. Not because they are careless. Not because they are undertrained. Because the environment exceeds what human cognition can reliably handle alone.

Vaught’s case happened in a hospital. With infrastructure. With systems. And still, a wrong drug reached the patient.

Now ask yourself what happens in a room with less infrastructure, more chaos, and no scanning at all.

The gap no one is solving, and why it has been so hard

I have spent years in emergency medicine and even longer thinking about medication safety in high-acuity environments. And the gap I keep coming back to is not in the pharmacy. It is not in the dispensing cabinet. It is at the bedside, in real time, when the drug is in someone’s hand and no system in the world is checking it.

The reason this gap has persisted is not lack of awareness. It is a technical problem that most people have not fully named.

Barcode scanning in hospitals works because it has something to scan against - a physician order, a pharmacist-verified product, an active entry in the EHR. The scan is a confirmation step at the end of a verified chain. Remove the chain, and the scan has nothing to do. It becomes a camera pointed at a vial with no reference point, no context, and no way to know if what is in that vial is the right drug at the right concentration for this patient right now.

That is why the code room has been unsolvable. Until you have a reference point that does not depend on an EHR order, you cannot do barcode verification in real time resuscitation.

The Breakthrough

The solution, it turns out, was not a scanning problem. It was a data problem.

More than a decade ago I had a single clinical insight that remains at the core of everything we build: every drug used in the code room must be predetermined. Every dose, for every weight, for every patient - push medications, infusions, and equipment - selected, calculated, and clinically verified in advance with the hospital’s clinical and pharmacy team. That is not a small thing.

And it turns out that predetermined data layer is exactly what barcode scanning needs to work without an EHR order. When a clinician opens the app during a code, there is nothing left to compute. That is the entire reason orderless barcode verification is possible, and why no one has been able to do it before.

When a clinician selects a medication in our system, the correct drug and the correct concentration are already known. That becomes the reference point. When they scan the vial, the system checks what is in their hand against what should be in their hand - in real time, at the bedside, with no pharmacy loop and no physician order required.

This is ScanSafe^TM. And it is, to my knowledge, the first-time barcode medication verification has been made available in an orderless environment.

The predetermined data architecture that we built to eliminate dosing errors also eliminated the dependency that made scanning impossible in resuscitation. Those two things, dose calculation and barcode verification, turn out to be the same problem, solved together.

For EMS clinicians who have never had any of this infrastructure, it is a meaningful leap forward. For nurses and emergency physicians running codes, it closes a gap that hospital scanning systems were never designed to address.

The room that needs barcode scanning the most finally has it.

There is a moment in every resuscitation when the room either clicks or it doesn’t. The compressor finds a rhythm. The airway clinician anticipates the next transition. The team leader reads the monitor and calls the next move before anyone has to ask. That moment is the product of rehearsal, performance measurement, structured feedback, and culture built together. And yet the majority of hospitals have no formal mechanism to measure CPR quality after a cardiac arrest. No waveform review. No compression fraction data. No feedback loop of any kind. You cannot improve what you cannot measure.

The New York Knicks just won the NBA Championship. Jalen Brunson is a incredible individual talent (trust me, I wish the Miami Heat had him). But nobody watching that team win a title believes he got there by practicing his jumper alone in a gym and then showing up on game day to meet his teammates. Championships are built on thousands of hours of practice, film review, and shared experience as a unit. The team learns together, fails together, and adjusts together. Individual skill is the entry requirement. Team performance is what wins.

Resuscitation works the same way. And a new study published in JAMA Cardiology suggests that the most widely adopted hospital CPR training program in the country may have forgotten that.

CPR Practice Alone with a Mannequin

See image at the top of the blog post

Researchers evaluated the Resuscitation Quality Improvement (RQI) program across 237 US hospitals participating in the Get With The Guidelines registry, comparing 18 hospitals that adopted RQI against matched controls, encompassing nearly 50,000 in-hospital cardiac arrests between 2017 and 2023.¹ The design matters: by using matched contemporary controls and a difference-in-differences analysis, the study isolated RQI’s incremental contribution to survival rather than simply comparing outcomes before and after adoption. Earlier single-site before/after studies generated positive signals for RQI, but those designs are vulnerable to confounding by secular trends and the overall maturity of a hospital’s resuscitation system. This study was a more honest test. The findings were unambiguous. Hospital adoption of the RQI program was not associated with improved rates of survival to discharge or return of spontaneous circulation in the two years following implementation. Survival rates at RQI hospitals and control hospitals fell in parallel. The program moved nothing.

The science underpinning RQI is legitimate. Low-dose, high-frequency training does improve individual skill retention, and that evidence is well established.^2,3 The AHA built RQI on sound educational principles, and the effort to move beyond biennial certification toward more frequent skill maintenance was a reasonable response to a real problem. It is also worth noting that this study was funded by the AHA and Laerdal Medical, the same organizations that created and operate RQI. In other words, they paid for an honest evaluation of their own program and published the results even when those results showed no benefit. That kind of intellectual honesty deserves acknowledgment.

The question the study forces is not whether the AHA was wrong to invest in training frequency. It is whether training frequency, on its own, was ever going to be enough. RQI may be a necessary foundation for competency maintenance. The evidence now suggests it is not sufficient on its own to move survival.

Resuscitation is a team sport. It always has been. The clinician doing compressions is counting on the person managing the airway to recognize when to pause. The medication nurse is waiting on a clear verbal order. The team leader is synthesizing the monitor, the rhythm, the timeline, and the voices in the room simultaneously. None of that happens at a wagon with individual computerized testing. None of that is trained in isolation. The authors of the Chan study acknowledge this directly, noting that the RQI program does not address leadership, closed-loop communication, or teamwork.¹ Those are not minor gaps. They are the gaps that determine whether a room full of individually competent clinicians functions as a resuscitation team.

There is a related implementation problem worth naming carefully. Across social media platforms including Reddit and TikTok, hospital clinicians have described in considerable detail the workarounds they use to complete RQI requirements: repositioning the manikin to trigger ventilation registration, timing submissions around shift end, completing modules when equipment is malfunctioning. These are not indictments of the clinicians involved. They reflect a predictable behavioral response when a mandatory requirement is perceived as disconnected from real patient care. Implementation science has long recognized that training programs which lack perceived clinical validity struggle to achieve genuine engagement regardless of design quality.⁸ The compliance numbers look good on paper. What happens at the bedside is a different question.

In EMS, we have grappled with this same tension. The answer we arrived at is team-based training with real performance data closing the loop. We train as units. We debrief every cardiac arrest. We use tools like CodeStat and CaseReview to review actual CPR performance data after every code, closing the gap between what we intend and what happens at the bedside. The majority of hospitals have no equivalent process. No waveform review. No compression fraction data. No loop to close. That feedback loop is not optional. It is the mechanism by which teams improve. You cannot improve what you cannot measure.

At Palm Beach County Fire Rescue, we have been running high-performance CPR training sessions with our emergency department partners. The sessions are not about refreshing hand position. They are about building the team. We train together, debrief together, and review the data together. The goal is a room full of clinicians who have rehearsed the hard moments and know each other’s moves.

Florida recognized the broader systems gap from the outside in. When EMS agencies began tracking what happened to cardiac arrest patients after hospital handoff, what we found was not reassuring. The quality of resuscitative care inside the emergency department was variable, inconsistent, and largely unmeasured. We had spent years building high-performance systems in the field, and the standard evaporated at the ambulance bay doors.

So we did what EMS has had to do before. We set the standard and brought it to the hospitals. Florida championed the development of Resuscitation Centers of Excellence for exactly this reason.⁶ The model is direct: access to our cardiac arrest patients is contingent on demonstrated performance. We did the same for stroke care. We did it for pediatric readiness. We are building it now for cardiac arrest. Each time, the motivation is identical: we cannot control what happens inside those walls, but we can control where we bring our patients.

The state has backed this commitment with real resources. The Florida Department of Health developed and funded the Resuscitate Florida program, a multimillion-dollar statewide investment to equip and train EMS agencies and dispatchers and to support the Florida Association of EMS Medical Directors in building out the Resuscitation Centers of Excellence network.⁷ That kind of investment reflects an understanding that survival from cardiac arrest is a systems problem.

The study authors put it plainly: a program focused solely on CPR delivery may not be sufficient to improve in-hospital cardiac arrest survival.¹ The evidence now supports what resuscitation teams on the ground have known for years. Individual skill is necessary. Team performance is what saves lives. And you cannot build the team at a wagon.

Jalen Brunson practiced with his team every day. The Knicks won a championship. The parallel is not subtle.

References

1. Chan PS, Bradley SM, Spertus JA, et al. Resuscitation Quality Improvement Program for CPR Training and Cardiac Arrest Survival in Hospitals. JAMA Cardiol. Published online May 20, 2026. doi:10.1001/jamacardio.2026.1074

2. Sutton RM, Niles D, Meaney PA, et al. Low-dose, high-frequency CPR training improves skill retention of in-hospital pediatric providers. Pediatrics. 2011;128(1):e145-e151. doi:10.1542/peds.2010-2105

3. Cheng A, Nadkarni VM, Mancini MB, et al; American Heart Association Education Science Investigators. Resuscitation education science: educational strategies to improve outcomes from cardiac arrest: a scientific statement from the American Heart Association. Circulation. 2018;138(6):e82-e122. doi:10.1161/CIR.0000000000000583

4. Meaney PA, Bobrow BJ, Mancini ME, et al; CPR Quality Summit Investigators, the American Heart Association Emergency Cardiovascular Care Committee, and the Council on Cardiopulmonary, Critical Care, Perioperative and Resuscitation. Cardiopulmonary resuscitation quality: improving cardiac resuscitation outcomes both inside and outside the hospital: a consensus statement from the American Heart Association. Circulation. 2013;128(4):417-435. doi:10.1161/CIR.0b013e31829d8654

5. Niles DE, Sutton RM, Donoghue A, et al. “Rolling refreshers”: a novel approach to maintain CPR psychomotor skill competence. Resuscitation. 2009;80(8):909-912.

6. Florida Association of EMS Medical Directors. Florida Resuscitation Centers of Excellence (FRCE) Program. Launched 2022. Available at: https://faemsmd.org/FRCE/

7. Florida Department of Health. Resuscitate Florida Program. Available at: https://www.floridahealth.gov

8. Eccles MP, Grimshaw JM, MacLennan G, et al. Explaining clinical behaviors using multiple theoretical models. Implement Sci. 2012;7:99. doi:10.1186/1748-5908-7-99

He survived. Neurologically intact. The case was published this year in JACC Case Reports as the first reported instance of DSED used as the initial defibrillation strategy. He had cardiac sarcoidosis, a substrate that creates patchy myocardial fibrosis and raises defibrillation thresholds. We knew from the history, from the rhythm, and from our protocol that standard defibrillation alone was unlikely to be sufficient. So we did not wait for it to fail.

The standard approach to refractory ventricular fibrillation has always been sequential. Try the standard way first. If it fails, try again. If it keeps failing, escalate. Double sequential external defibrillation (DSED), two defibrillators, four pads, two sequential shocks from AP and AL positions designed to maximize current delivery across the full fibrillating myocardium, was for most of its history reserved for that escalation moment. A last resort. Applied late, to patients who had already failed multiple attempts.

The DOSE VF trial, published in the New England Journal of Medicine in 2022, changed the conversation. In a cluster-randomized trial across six Canadian paramedic services, DSED delivered after three failed standard shocks produced 30.4% survival to hospital discharge versus 13.3% with continued standard defibrillation. That is more than a doubling of survival. For patients in refractory ventricular fibrillation, it was the most compelling evidence the field had seen in years. And critically, it suggested that timing matters: applied earlier, with protocol discipline, the benefit was real and large.

A new observational study from Aotearoa New Zealand, published in Resuscitation in 2026, complicates that picture. Dicker and colleagues evaluated outcomes before and after New Zealand became the first country outside Canada to formally implement DSED into national EMS guidelines. Their conclusion: no survival benefit was observed after implementation. DSD was associated with lower odds of ROSC and 30-day survival, particularly when applied late. The study is being read by some as a challenge to the DOSE VF findings.

It should not be. Not because the authors did anything wrong. Their limitations section is admirably honest. But because the data, read carefully, tell a different story than the headline suggests. The story is not that DSED failed. The story is that this dataset was not built to answer whether DSED works.

Let me explain why.

“Early DSD” That Wasn’t Early

The New Zealand study divided its DSED patients into two groups: early-DSD (three or fewer EMS shocks before the first DSED attempt) and late-DSD (more than three EMS shocks before the first DSED attempt). The distinction matters because the DOSE VF trial was built on the premise that earlier application of an alternate strategy prevents the progressive metabolic deterioration that follows each failed shock.

Here is the problem. The early-DSD group in the New Zealand study had a median of nine total shocks. Nine. The DOSE VF trial reported a mean of 3.9 standard shocks before randomization. The late-DSD group in New Zealand had a median of fourteen total shocks.

At Palm Beach County Fire Rescue, the case published in JACC Case Reports this year demonstrates where our thinking has moved: DSED as the initial shock. Not after three failures. Not after evidence of refractoriness. At first contact with the rhythm, for the right patient. The New Zealand ‘early’ group got DSED within three shocks. That satisfies the definition, but it doesn’t tell us whether shock three came at minute eight or minute forty of a 104-minute resuscitation. Those are not the same patient.

This matters enormously for how to read the results. The secondary analysis of DOSE VF published by Cheskes and colleagues in 2024 stratified patients by whether they were in true shock-refractory VF (continuous fibrillation persisting through all three prior shocks) versus recurrent VF (VF that briefly terminated but returned). Among shock-refractory patients, not a single patient in the standard defibrillation arm survived to hospital discharge. In the DSED arm, 28.6% survived. The intervention works. But the biology is time dependent. Every shock that fails is not a neutral event. It is tissue damage, a longer time in fibrillation, another increment of ischemic injury to a brain and heart that are already running out of tolerance.

The New Zealand data do not disprove DSED. They show what happens when a strategy designed for early application is applied late. The result is predictable, and the authors acknowledge it. The association of late-DSD with lower odds of ROSC (aOR 0.46) most likely reflects resuscitation time bias: DSED was deployed in patients who had already failed multiple interventions, and the longer time to DSED is a marker of a more compromised patient, not evidence of harm from the intervention itself.

One Hundred Minutes on Scene

The median scene time for DSD patients in this study was 104 minutes.

Read that again.

In DOSE VF, the median time from EMS arrival to departure from scene was 26.5 minutes in the DSED arm. The New Zealand DSD cohort spent nearly four times as long on scene. When you are studying a rhythm that kills neurons by the minute, and your exposure and control groups differ by hours of resuscitation time, you are not primarily studying the defibrillation strategy anymore. You are studying the biology of what happens to a heart that has been in sustained arrest for the better part of two hours.

The literature on defibrillation is unambiguous on this point. Holmén and colleagues described survival declining sharply with each additional shock in OHCA. The New Zealand authors themselves cite Harrysson et al., who found that survival drops steeply as the number of defibrillation attempts rises. At some threshold, the rhythm is no longer a rhythm susceptible to defibrillation in any configuration. It becomes an expression of terminal cellular failure. Electrical therapy at that point, regardless of the vector, the energy, or the number of defibrillators, is working against irreversible biology.

The 100-minute scene time also raises a clinical question the data cannot answer: what was happening during those 100 minutes? Were teams standing by awaiting DSED equipment? Were they attempting other interventions? Were there scene complications, access difficulties, extrication needs? The study cannot tell us, and those factors matter. What we can say is that 100 minutes on scene in cardiac arrest is not a resuscitation. It is an extended death.

Again, this is not a criticism of the authors. It is a limitation of observational implementation science. When a protocol is introduced into a real-world system without the controlled conditions of a randomized trial, the patients who receive the new intervention are often the ones for whom standard therapy has failed most dramatically. They are sicker, more refractory, deeper into their arrest. The New Zealand data hint strongly at this. Despite comprehensive paramedic training before rollout, DSED was applied in only 43% of eligible patients. That gap almost certainly reflects clinical judgment at the patient’s side, and clinical judgment in a prolonged refractory arrest tends to move toward escalation when the situation is most desperate. In other words, the patients most likely to receive DSED were probably the ones least likely to survive regardless of what was done. Adjustment for available covariates helps but cannot fully account for what the authors call unmeasured confounders: comorbidities, estimated weight, community responder presence, precise timing of the first shock. Critical treatment measures such as the amount of epinephrine, time to epinephrine, amount of antiarrhythmic, time to antiarrhythmic, and CPR quality were all measured, controlled for, and reported in DOSE VF. None of them are mentioned in this study. That is a further challenge when observational research attempts to replicate the rigor of a randomized controlled trial. When the selection process itself is unmeasured, and the treatment variables that determine resuscitation success are not captured, no regression model can fully close that gap.

The Denominator Problem

This is the issue that most directly undermines the study’s ability to draw conclusions about DSED, and it deserves careful attention.

The New Zealand study included all cardiac arrests receiving more than three prehospital defibrillations. It did not restrict enrollment to patients presenting in ventricular fibrillation. DOSE VF enrolled only patients presenting in VF or pulseless VT who remained in that rhythm after three consecutive shocks. The biological rationale for DSED is specific: it is designed to overcome a refractory ventricular fibrillation waveform that persists despite adequate energy delivery. There is no mechanistic basis for expecting DSED to benefit a patient in PEA who received defibrillation attempts during rhythm checks that transiently showed a shockable rhythm, or a patient who presented in asystole and developed a secondary shockable rhythm late in a prolonged resuscitation. This has been confirmed by multiple previous studies assessing the impact of defibrillation on survival in those who present in asystole or PEA and become shockable during the resuscitation.

The New Zealand authors acknowledge they could not differentiate between refractory and recurrent VF because defibrillator waveform data were not routinely transmitted into the patient record. They also could not confirm what proportion of their three-plus-shock cohort presented in VF to begin with, and how many of those remained in true refractory VF at the time DSED was applied.

The Cheskes 2024 secondary analysis found that only 17% of DOSE VF patients were in true shock-refractory VF. The remaining 83% had recurrent VF. Both groups benefited from DSED over standard defibrillation. The survival benefit in recurrent VF did not reach statistical significance, but DSED produced a 10% absolute improvement in survival over standard defibrillation in that group. Statistical significance and clinical relevance are not the same thing, and a trial adequately powered for that subgroup may tell a different story. The key point is that the DOSE VF population was selected with precision: VF at presentation, VF confirmed after each of three shocks, a protocol that excluded patients who had converted and re-fibrillated prior to randomization in ways that prevented clean classification. The New Zealand dataset cannot replicate that precision retrospectively.

What the Data Actually Show

None of this means the New Zealand study is without value. It is exactly the kind of honest implementation data the field needs, and the authors should be credited for publishing it with transparency about its limits.

The pre-period to post-period comparison is the cleanest analysis in the paper. It shows that after implementing DSED into national guidelines, 30-day survival moved from 13.4% to 15.5%, a non-significant improvement that nevertheless sits in the same range as the control arm of DOSE VF. The resuscitation landscape did not get worse. It may have gotten marginally better, and the absolute numbers are consistent with a real signal that larger, better-controlled data could detect.

The early-DSD group is also instructive. Among patients who received DSED within three EMS shocks, which the study calls early, adjusted odds of ROSC and survival were not significantly different from those who received no DSED (aOR 1.05 for both). This is neutrality, not harm.

The late-DSD signal, which carries the most dramatic odds ratios, is almost certainly a confounding artifact. The patients who received late-DSD are the ones who failed everything else. They are self-selected by prolonged futile resuscitation. Attributing their poor outcomes to the DSED intervention rather than to the state of their resuscitation at the time of DSED application requires a causal assumption the data cannot support.

The Question the Field Still Needs to Answer

The three randomized trials currently underway in Sweden, Denmark, and Norway (DOUBLE-D, STRAT-DEFI, and DUALDEFIB) will provide better answers than implementation observational data can. But the questions they need to answer are not just ‘does DSED work.’ They are: at what shock count does DSED need to be deployed to preserve its benefit? Does the benefit vary by whether VF is truly refractory versus recurrent? And what happens to survival when the intervention is applied by systems that may not have the training standardization, equipment readiness, and medical oversight that characterized the DOSE VF trial sites?

Early data from the Swedish DOUBLE-D feasibility trial point in a compelling direction. When 80% of patients received DSED after just two shocks, survival in the DSED arm reached 41% compared to 9% in the control arm. The numbers are small and feasibility was the primary focus, but that signal is consistent with the DOSE VF findings and with our own experience at Palm Beach County Fire Rescue. It is a very different picture than what the New Zealand observational data suggest, and the difference almost certainly comes down to when in the resuscitation the intervention was applied.

The most useful interpretation of the New Zealand data is not that DSED does not work. It is that deploying DSED into a real EMS system does not automatically reproduce the outcomes of a controlled trial, and that the conditions under which the intervention is applied, specifically the timing, the patient population, and the system capacity to apply it early and consistently, determine whether the benefit is realized.

That is not a reason to abandon DSED. It is a reason to take implementation seriously. Protocol design matters. Training matters. Equipment availability matters. Medical oversight matters. And the shock count at which you pull the trigger on an alternate strategy may matter most of all.

The man with cardiac sarcoidosis sitting on his couch survived because we did not wait for standard defibrillation to fail. We had the equipment staged, the protocol established, and the mechanistic reasoning in place. When the rhythm appeared on the monitor, we already knew what we were going to do. For the patient in refractory VF, the intervention that works is not the third standard shock. It is a different approach, applied early, by a crew that has practiced it and a system that has set the conditions for success.

That is what DOSE VF showed. The New Zealand study, read carefully, does not contradict it. It reminds us that evidence from a controlled trial does not travel automatically into the real world. It has to be carried there deliberately, with attention to the variables that made the original finding possible.

The story is not that DSED failed in New Zealand.

The story is that implementation is hard. And getting it right matters.

References

1. Cheskes S, Verbeek PR, Drennan IR, et al. Defibrillation strategies for refractory ventricular fibrillation. N Engl J Med. 2022;387(21):1947-1956.

2. Cheskes S, Drennan IR, Turner L, Pandit SV, Dorian P. The impact of alternate defibrillation strategies on shock-refractory and recurrent ventricular fibrillation: a secondary analysis of the DOSE VF cluster randomized controlled trial. Resuscitation. 2024;198:110186.

3. Dicker B, Callejas P, Hutchinson H, et al. Implementation of double sequential defibrillation (DSD): an Aotearoa New Zealand observational study. Resuscitation. 2026. doi:10.1016/j.resuscitation.2026.111138.

4. Duncan TJ, Antevy P, Robles A, Coyle C, Scheppke KA, Cheskes S. Initial double sequential external defibrillation for refractory ventricular fibrillation in cardiac sarcoidosis. JACC Case Rep. 2026;31:105945.

5. Cheskes S, Drennan IR, Turner L, Pandit SV, Walker RG, Dorian P. The impact of alternate defibrillation strategies on time in ventricular fibrillation. Resuscitation. 2025;209:110549.

6. Thomas AJ, Newgard CD, Fu R, Zive DM, Daya MR. Survival in out-of-hospital cardiac arrests with initial asystole or pulseless electrical activity and subsequent shockable rhythms. Resuscitation. 2013;84(9):1261-1266.

7. Goto Y, Maeda T, Nakatsu-Goto Y. Prognostic implications of conversion from nonshockable to shockable rhythms in out-of-hospital cardiac arrest. Crit Care. 2014;18:528.

8. Cournoyer A, Cossette S, Potter B, et al. Prognostic impact of the conversion to a shockable rhythm from a non-shockable rhythm for patients suffering from out-of-hospital cardiac arrest. Resuscitation. 2019;140:43-49.

9. Holmén J, Hollenberg J, Claesson A, et al. Survival in ventricular fibrillation with emphasis on the number of defibrillations in relation to other factors at resuscitation. Resuscitation. 2017;113:33-38.

Last year, our crews at Palm Beach County Fire Rescue told us they were tired. Not the kind of tired that a good night of sleep fixes. The kind that builds up over years of 24 on and 48 off, of holidays missed, of children growing up between shifts, of the slow erosion that no one talks about until someone leaves.

Local 2928 brought it forward. Through good faith negotiations with the department, we are moving this coming January from a 24/48 schedule to 24/72. It is not a small change. It costs money. It requires hiring. It demands a rethinking of how shifts are built and how coverage is maintained. But it is the right change, because the people doing the work asked for it, the union did the hard work of building the case and getting it to the table, and the evidence behind it is sound.

That decision is what EMS Week is really about.

The 2026 theme is Improving Outcomes, Together. Most of the conversation this week will focus, rightly, on what our clinicians do at the bedside. Prehospital ultrasound. Early stroke recognition. Whole blood transfusion. Pediatric resuscitation that looks nothing like it did a decade ago. These advances are real, and they are saving lives every day, in big cities and tribal lands and rural counties and everywhere in between.

But the theme contains a quieter promise, one that is easy to miss. It commits us to investing in the people behind the care. Fair pay. Safe staffing. Mental health resources. Leadership that listens and acts. Opportunities for growth. That promise is the foundation that everything else stands on, and it deserves more than a closing paragraph in a press release.

Here is what I think it can look like when we take that promise seriously.

It looks like a shift schedule built around what the science says about fatigue, not around what is easiest for the budget. The data on long shifts, on circadian disruption, on cumulative sleep debt in first responders is no longer in dispute. When we ask people to make life and death decisions at 4 in the morning after sixteen hours of calls, we are asking them to perform at the edge of what is humanly possible. Building a schedule that respects that reality is not a luxury. It is clinical care.

No one in this country has done more to make that case than James Geering. James is a Florida firefighter and paramedic, the host of the Behind the Shield podcast, and the author of two books, One More Light and Kinder, both of which sit at the intersection of first responder health and the systems that grind it down. For years he has been arguing, with data and with stories, that the 56-hour workweek is unsustainable and that the 24/72 schedule is not a perk but a clinical and operational intervention. He has interviewed military sleep scientists, strength and conditioning experts, and the chiefs of departments that have already made the switch, and he has built the most comprehensive public library on this question that exists.

In November of 2024, Tanner Smida and I had the chance to host James on the NAEMSP Florida Chapter webinar to walk through the case for the 24/72 schedule. It was one of the most useful conversations I have been part of in this space. What stays with me from that hour is not a single statistic but the broader frame James kept returning to: when we treat sleep as a luxury and overtime as a staffing model, we are not saving money. We are deferring a bill that the workforce eventually pays in injury, in illness, in divorce, in early retirement, and in some cases in lives. The schedule is the foundation. Everything else we ask of our clinicians, the resuscitations and the difficult airways and the calls that demand everything they have, gets built on top of it.

The move PBCFR is making this year is one I believe in for clinical reasons. The credit for getting it across the line belongs to Local 2928 and to the members who made the case. But in a broader sense, it is also a downstream effect of the work that people like James have been doing for a long time, often without recognition, to change the national conversation about what we owe the people doing this job.

It looks like a paramedic earning a wage that lets them live in the community they serve. The pay gap between what we ask of EMS clinicians and what we pay them has been a quiet crisis for decades. We train them to run pediatric cardiac arrests, to recognize sepsis, to manage airways in moving ambulances, to make decisions that emergency physicians would not make alone in a fully staffed resuscitation bay. And then we ask them to work a second job. The math does not hold, and the workforce numbers are starting to tell us that.

It looks like mental health support that is real, not symbolic. A peer support program that people actually use. Time off after the calls that stay with you, without having to justify why. A culture where saying you are struggling is met with a phone call, not a quiet reassignment. The clinicians I work with are not asking to be coddled. They are asking to be human, and to be allowed to keep being human after thirty years of seeing what they see.

And it looks like a career, not a job. A path from the back of the ambulance to the front of a classroom, to a research project, to a leadership role, to a place at the table when the next protocol is being written. EMS has been losing too many of our best people to other fields, not because they stopped loving the work, but because they could not see where the work was taking them. The systems that figure out how to grow their own are the systems that will still be standing in twenty years.

And finally, it looks like a federal structure that matches the reality of what EMS has become. Two fights matter most here, and people at the national level have been carrying them for years.

The first is recognition of EMS as an essential service. Federally, EMS still sits under the U.S. Department of Transportation, a structural artifact of how the field was created in the 1960s. Fire and police are guaranteed local funding floors that EMS is not. A handful of states have written EMS into law as essential, and more are trying, but the federal recognition that would change the funding architecture of this entire field has not yet happened. It needs to.

The second is unhooking payment from transport. Under current Centers for Medicare & Medicaid Services policy, an EMS agency is generally paid only when a patient is loaded into the ambulance and taken to a hospital. That rule was written in a different era, when the back of the truck did far less than it does today. It punishes the most clinically appropriate decisions, treating in place, referring to a behavioral health clinic, resolving the call through telemedicine consultation, and it does so most painfully in rural communities where the alternative to a long transport is not less care but better care delivered on scene. Several pieces of legislation are working their way through Congress and through state legislatures right now to fix this, and the national associations have been carrying the case for years. The specific bills will rise and fall. The principle will not. Payment should follow the care, not the wheels.

These are not abstract policy fights. They set the ceiling on everything we have been talking about. A department can negotiate a better schedule, a state can pass an essential service law, a union can win fair pay, but the long term funding picture for this field will be decided by whether Washington decides EMS is a transportation service or a medical one. The clinicians have already answered that question. The federal government needs to catch up.

None of this is separate from clinical outcomes. A rested clinician makes better decisions. A clinician who feels seen by leadership stays in the job long enough to become the experienced hand a new partner will need on a bad call. A workforce that can afford to live where it works builds the kind of community trust that makes a stranger open the door at 2 in the morning. The medicine and the people are not two different conversations. They are the same conversation.

Improving Outcomes, Together means recognizing that the most advanced protocol in the world is delivered by a human being with a family, a body, a mind, and a finite capacity. Honor that capacity, and the outcomes follow. Ignore it, and no amount of technology will close the gap.

This week, we recognize the people behind the patch. Next week, and every week after, we keep building the system they deserve.

Picture a paramedic in the back of a moving ambulance. The patient is a five-year-old with a femur fracture, screaming, scared, and in pain. Fentanyl has already been given and the pain is not controlled. The medic opens an app and voices a text into a chatbot, asking for a pediatric ketamine dose. A paragraph of text comes back. It sounds confident. It sounds clinical. It might even be correct.

But the medical director who signed the protocol document for that agency has never seen those exact words, in that exact order, with that exact reasoning attached. In 21 states, the document is signed not by an agency medical director but by the state medical director, with the force of statute and administrative rule behind it. Either way, in that moment, the legal architecture of EMS quietly comes apart.

This is the question I keep coming back to as I watch large language model chatbots get pitched to EMS leaders. Not whether the technology is impressive. It is. Not whether it will eventually have a role in our field. It probably will. The question is whether the conversational chatbot, as a clinical interface, is the right tool for the back of an ambulance today. I do not believe it is. And I think it is worth saying so plainly, because the gap between a polished demo and a safe deployment in EMS is wider than most people realize.

Let me say up front that I am not a skeptic of artificial intelligence. We use it across our teams every day. It is helping us with research, software development, quality assurance, and protocol review. It is genuinely transformative for non-patient-facing work, and I expect that footprint to keep growing. The question is not whether AI belongs in EMS. It does. It just does not belong in the seat of time-sensitive clinical decisions made on real patients. The reasons are structural, and they are worth walking through.

Start with what an EMS protocol actually is. In the rest of the house of medicine, clinical references are exactly that, references. An emergency physician operates under a license to practice medicine, with the standard of care and individual judgment as the framework. In EMS, the protocol is something different. It is a signed, dated, board-approved document that defines the legal scope of paramedic practice. A paramedic acting within the protocol is operating within their license. A paramedic acting outside the protocol is not. That document is reviewed by counsel, ratified by medical direction, and audited after the fact in QA review and, sometimes, in court.

This is what makes a protocol something more than a reference. It carries regulatory force where states have given it that weight. It defines the standard of care that a clinician will be measured against if something goes wrong. It is admissible as evidence and serves as the basis for expert testimony in court. And it shapes the scope of legal immunity that an EMS provider is entitled to when they act within it.

When a chatbot generates clinical guidance by paraphrasing, summarizing, or recombining protocol content, it produces text that no medical director has approved. The output looks authoritative because it is written in clinical language, but it does not carry the legal weight of the actual protocol. It is a paraphrase masquerading as policy. That distinction matters every time a paramedic acts on it, and it matters even more when something goes wrong and the case is reconstructed line by line.

The second problem is how these systems actually work. Large language models are probabilistic. They generate the next most likely token, then the next, then the next. That means the same question, asked twice, can produce two different answers. Hallucination is not a defect being engineered out. Researchers at the National University of Singapore formalized this mathematically and showed, using results from learning theory, that eliminating hallucination in large language models is not just difficult, it is impossible (Xu, Jain, and Kankanhalli, 2024). OpenAI‘s own 2025 research paper reached a parallel conclusion from a different direction. Hallucinations, they wrote, are a predictable outcome of how these systems are trained and evaluated, not a bug that better engineering will solve. The leading researchers in the field have largely stopped pursuing zero hallucination as a goal. The current consensus is managing uncertainty, not eliminating it.

The common rebuttal to all of this is that grounding the model to a specific document, in this case your agency’s protocol, solves the problem. It does not. Retrieval-augmented generation (RAG) reduces hallucination. It does not eliminate it. The model can still misread a protocol section, blend content from two similar pathways, or generate a clinically wrong sentence with complete confidence and no flag to the clinician. And there is a more dangerous finding tucked inside the OpenAI work. Models do not just get things wrong. They get things wrong confidently, with low expressed uncertainty and no signal that the answer deserves a second look. In a protocol-based system, where uncertainty should trigger a double-check, a tool that projects confidence it has not earned is not a safety net. It is a liability. In a domain where a misplaced decimal kills a child, “usually right” is not a category that exists.

None of this is to say all AI implementations in EMS are equivalent. Applications that surface medical-director-approved protocols through structured retrieval, with the source document directly referenced and the signed content the clinician sees, are doing something different from a chatbot that generates new sentences. The argument here is specifically about the conversational chatbot interface and the generative paraphrase that interface produces.

And there is a deeper problem that is less obvious from a demo. When I write an EMS protocol, I write it from top to bottom with intention. There is a progression. Assessment comes before intervention. Non-pharmacologic options come before drugs. Contraindications come before indications. First-line agents come before second-line. Doses come last, and only after the clinician has been walked through the decision that justifies them. The protocol is not a flat database of facts. It is a structured pathway, and the order is part of the clinical content written and approved by a physician.

A chatbot does not see that. When a medic types “pain dose for a five-year-old,” the model treats the protocol as a searchable text corpus, finds the line that matches, and returns it. Every step that came before that line, the assessment, the contraindications, the alternative agents, the recognition that this child might not need ketamine at all, becomes invisible. The clinician gets the answer to the question they asked, not the answer to the clinical situation they are managing.

Ketamine is the cleanest example. In a typical EMS system, ketamine appears in the analgesia protocol, the agitation protocol, the induction protocol, and sometimes in the bronchospasm pathway. Different doses. Different routes. Different concentrations. Different indications. Different contraindications. Ask a chatbot “what is the ketamine dose” and there is no good answer. Either the model picks one, in which case it is wrong roughly as often as it is right, or it lists all of them and asks the clinician to choose, in which case the burden of decision sits exactly where it should not.

That brings me to the operational point. The pitch for AI in EMS is almost always framed as cognitive load reduction. A paramedic working a critical patient is task-saturated. The argument goes that an intelligent assistant should reduce that load. I agree with the goal. I disagree that a chatbot is the way to get there.

A conversational interface, by design, asks the clinician to do work. Type a question. Wait for a paragraph. Read it. Parse it. Decide whether it applies to the protocol you are operating under. Reconcile the concentration the chatbot returned with the concentration on the truck. Choose between the three doses it offered. That is more cognitive load, not less, and it is being added at the worst possible moment, in a moving vehicle, with one hand, with bad lighting, with a sick patient who cannot wait. The right interface for the back of an ambulance is the one that gets the clinician to the right page, in the right protocol, in two taps, with no parsing required. Structured navigation removes work. Conversation adds it.

The final piece is accountability. Twenty years of serving as a medical expert in EMS cases has taught me that when something goes wrong, the case gets reconstructed line by line. Documentation is read aloud. Protocols are introduced as exhibits. Every decision is traced back to the person responsible for it. That chain of accountability is not a bureaucratic detail, it is the architecture that protects clinicians and patients alike.

Today, when a paramedic deviates from protocol, that chain holds. The deviation is documented. The medical director reviews the case. QA flags the variance. Education or remediation follows. The system learns. The clinician is supported. The patient is protected. It works because every step is traceable to a person who stands behind it.

When a chatbot produces a clinical recommendation, who is responsible for the output? The vendor disclaims it in the terms of service. The medical director never approved it. The paramedic acted on it in good faith because it appeared inside an app the agency adopted. If the recommendation contributed to a bad outcome, the chain breaks at exactly the moment it matters most. That is not a hypothetical concern. It is the structure of every malpractice case that will eventually be written about this technology.

All of which leads me to the part that bothers me most as a clinician and a scientist. The way we introduce new interventions into the house of medicine is well established. We define the outcome that matters. We run the trial. We look at the results. We let the data tell us whether the intervention helps, harms, or makes no difference. Then we decide. That discipline is not optional, and it does not change because the new intervention is software.

We have a clean recent example of what this looks like when it is done right. The Copenhagen Emergency Medical Services group ran a double-masked randomized trial of a machine learning model designed to help dispatchers recognize out-of-hospital cardiac arrest during 911 calls (Blomberg et al., JAMA Network Open, 2021). The trial enrolled more than 169,000 calls, including over a thousand confirmed cardiac arrests. The model worked. On the algorithmic metric, it outperformed the human dispatchers, with sensitivity of 85 percent compared to 77.5 percent. By the standard a vendor demo would use, that is a win.

And yet, when the trial measured what actually mattered, whether dispatcher recognition of cardiac arrest improved when the AI was in the loop, there was no difference. The model was better at the task, the dispatchers had access to its alerts, and dispatcher performance did not change. The investigators were direct about it. They wrote that almost all decision support tools driven by artificial intelligence or machine learning have so far failed to improve outcomes in practice. The 2025 AHA guidelines reviewed this evidence and reached a measured conclusion. The technology is promising. It is not ready to be relied on.

That is how this is supposed to work. We test our tools the way we test our drugs. We do not deploy them based on how good they look in the conference hall. The chatbot products being marketed to EMS today have not been studied with anything close to that rigor. There are no randomized trials. There are no patient-level outcomes. There are demos, testimonials, and screenshots. That is not enough.

None of this is an argument against artificial intelligence in EMS. There is real, serious work to be done in this space, and there are use cases where machine learning will help us deliver better care. The point is narrower than that. The conversational chatbot, as a clinical interface for paramedics making time-critical decisions on real patients, is the wrong form. The right form is curated, deterministic, auditable, and approved by the medical director, who signs the protocol. The clinician sees the same content the medical director wrote, in the same order, with the same context, every single time.

EMS is not ready for an AI chatbot. Not because we are afraid of the technology, but because we are clinicians and scientists who understand what a protocol is and how new tools earn their place in our field. A protocol is a legal document, written with intention, designed to be navigated, not paraphrased. A new clinical tool is something we study before we deploy it, not after. Until the chatbot can honor both of those principles, the demo is the easy part. The deployment is where people get hurt.

References

Blomberg SN, Christensen HC, Lippert F, et al. Effect of Machine Learning on Dispatcher Recognition of Out-of-Hospital Cardiac Arrest During Calls to Emergency Medical Services: A Randomized Clinical Trial. JAMA Network Open. 2021;4(1):e2032320.

Kupas DF, Schenk E, Sholl JM, Kamin R. Characteristics of Statewide Protocols for Emergency Medical Services in the United States. Prehospital Emergency Care. 2015;19(2):292-301.

Xu Z, Jain S, Kankanhalli M. Hallucination is Inevitable: An Innate Limitation of Large Language Models. arXiv. 2024;2401.11817.

OpenAI. Why Language Models Hallucinate. 2025.

He was twenty-three years old and had no idea what was about to happen.

The motorcycle hit the car at the intersection and he was thrown eighty-five feet. By the time we arrived, he was on the pavement, unconscious. Traumatic brain injury. Multisystem trauma. Active hemorrhage. A young man whose survival would depend on decisions made in the next several minutes.

The crew gave him whole blood and plasma.

That decision did not come from instinct. It came from nearly twenty years of science, science that was on full display last week when Dr. John Holcomb packed grand rounds at St. Mary’s Medical Center on Thursday and Memorial Healthcare System on Friday. Holcomb, now at the University of Alabama at Birmingham , is one of the most consequential trauma surgeons of this generation, a former military physician who helped reshape resuscitation medicine through battlefield experience and decades of rigorous research. His message was not new. That was the point.

For years, the evidence that plasma improves survival in traumatic brain injury has been building. The PAMPer trial, published in the New England Journal of Medicine, showed that prehospital plasma reduced thirty-day mortality in hemorrhagic shock patients.

A secondary analysis identified where the benefit was strongest: patients with CT-confirmed intracranial hemorrhage. Transferred patients saw no benefit, not because plasma failed, but because by the time they arrived at a receiving facility, too much time had passed. The window is prehospital. The question was how narrow.

That question now has a number, and it comes from the most important plasma-TBI paper published to date.

In April 2026, Heejoo Kang , Crisanto Torres, MD, MPH, FACS , and colleagues at Boston Medical Center published a study in the The Journal of Trauma and Acute Care Surgery examining 6,183 adults with traumatic intracranial hemorrhage from the ACS Trauma Quality Improvement Program database. These were severely injured patients. Median Injury Severity Score was 34. Overall thirty-day mortality was 45 percent. Patients who received plasma had a 45 percent lower hazard of death at thirty days compared to those who had not yet received it at the same time point.

But the finding that should change how every EMS system in this country thinks about head injury was the timing curve.

The survival benefit dropped steeply in the first thirty minutes after emergency department arrival and then flattened. Every ten-minute delay within that window was associated with a five percentage point absolute increase in thirty-day mortality. Plasma after thirty minutes carried a 33 percent greater hazard of death, nearly double the risk of craniotomy, and a 38 percent greater risk of intracranial drain placement compared to plasma given within that first half hour. The thirty-minute mark is not a hard cutoff. It is where the curve bends. And for a patient thrown eighty-five feet, that clock starts on the pavement, not in the emergency department.

At grand rounds, Dr. Greg Zorman, former chief of neurosurgery at Memorial Regional Hospital, made a point that landed clearly. Traumatic brain injury has seen as little clinical progress as glioblastoma. The biology is understood but the silos have slowed the translation. Holcomb’s response was direct: if plasma reduces laparotomy rates in liver trauma by repairing the vascular endothelium, it can reduce ICP monitor placements and emergent craniotomies in brain injury. The mechanism is the same. Breaking the silos between EMS, emergency medicine, and neurosurgery is not optional. It is the work.

Palm Beach County Fire Rescue has been carrying cold liquid plasma since December 2024, with Chief Coyle leading the project and Dr. Holcomb as a mentor throughout. Last week’s grand rounds was not a one-time academic visit. In that room were Holcomb and our trauma surgery colleagues at St. Mary’s, Dr. Azar and Dr. Rueda, all now collaborators on a retrospective observational study of prehospital plasma in isolated TBI patients from Palm Beach County Fire Rescue. That study is a deliberate first step, not a final answer.

The TXA story makes this point sharply. The prehospital TXA for TBI trial, published in JAMA in 2020, enrolled 966 patients with moderate or severe TBI across 20 trauma centers and 39 EMS agencies. The primary outcome, favorable neurologic function at six months, showed no significant difference between TXA and placebo. A neutral trial, and a well-designed one. But buried in the subgroup analysis was a signal that should have stopped everyone cold: in the patients with CT-confirmed intracranial hemorrhage, the 2-gram bolus group had an 8.2 percent lower 28-day mortality compared to placebo. The drug was not the problem. The patient selection was. Because fewer than half the patients enrolled in that trial actually had blood on imaging, the signal was diluted beyond detection in the primary analysis.

Here is what the data have made clear: treatments work in patients with CT-confirmed intracranial hemorrhage. But fewer than half of patients who present with severe TBI in the field will have hemorrhage on imaging. That means we are enrolling the wrong mix of patients in every prehospital TBI trial we run, and the signals keep getting buried. The only way to change that is a prehospital, multicenter, randomized controlled trial that selects for the right patients, enrolling them before the CT scan, before the transfer, before the thirty-minute window closes. Our observational work is designed to build the foundation for exactly that. It is what the science demands. It is what our patients deserve.

The twenty-three-year-old motorcyclist survived neurologically intact.

Whole blood and plasma on the pavement. Before the CT scan. Before the neurosurgeon. Before the thirty-minute window had closed. The Torres paper put a number on the timeline. And a young man is alive today because clinicians acted on the evidence when it mattered most.

References

1. Sperry JL, Guyette FX, Brown JB, et al. Prehospital plasma during air medical transport in trauma patients at risk for hemorrhagic shock. N Engl J Med. 2018;379(4):315-326.

2. Gruen DS, Guyette FX, Brown JB, et al. Association of prehospital plasma with survival in patients with traumatic brain injury: a secondary analysis of the PAMPer cluster randomized clinical trial. JAMA Netw Open. 2020;3(10):e2016869.

3. Kang H, Ong C, Kenzik K, et al. Association of plasma transfusion timing with mortality among patients with traumatic intracranial hemorrhage across US adult civilian trauma centers. J Trauma Acute Care Surg. 2026;00:000-000.

4. Rowell SE, Meier EN, McKnight B, et al. Effect of out-of-hospital tranexamic acid vs placebo on 6-month functional neurologic outcomes in patients with moderate or severe traumatic brain injury. JAMA. 2020;324(10):961-974.

Angela Johnson works for the Indiana Department of Transportation. On the morning her life changed, she was driving a dump truck on a highway construction site in Henry County, hauling 1,800 gallons of water. The radio crackled. She glanced down. By the time she looked back up, she was over-correcting, and the truck began to roll.

She was ejected through the windshield, landed face-down on the road, and the truck rolled across her chest before coming to rest in a farmer’s field below.

The 911 call went out as an unknown traffic accident. By the time the Newcastle-Henry County EMS crew arrived, dispatch had upgraded it to a possible ejection with severe blood loss. They pulled up to a scene that any seasoned medic would recognize immediately as someone who was probably not going to make it to the hospital alive.

Newcastle-Henry County had recently begun carrying whole blood on a supervisor’s vehicle, the model that most prehospital blood programs in the United States use. Blood doesn’t usually live on every ambulance. It lives on a chase vehicle that meets the ambulance at the scene, allowing one unit to cover a wider footprint and reduce wastage. Chief Harrison and his crew had been carrying that blood for a few months. They had not used it yet. Angela was the first.

The Chief describes the moment in the video I’ll link below. He tells you what it felt like to make the decision, to spike the bag, to start the transfusion. And then he tells you what every clinician in this field has been saying for years: he watched her vital signs come back in real time. Heart rate down. Blood pressure up. Mid-transfusion. The way you see a diabetic wake up after dextrose, except this was a woman with seven broken ribs, double pulmonary contusions, and a face she would later describe, gently, as “ripped off.” She had been told, before she lost consciousness, that without that blood she would not have made it.

She did make it. Her sixth grandchild was born the night before she told me her story.

I’m starting with Angela because the rest of this post is about money, and money is an abstraction. Angela is not. There are roughly ten thousand Angelas a year in this country who don’t get the unit of blood that would have saved them, because the agency that responded to their crash didn’t carry it, because they couldn’t afford to.

That is the problem we went to Washington to talk about. And the most striking thing I heard there was a number.

A number to remember

The number is $164.

That’s what it costs, on average, to give a trauma patient one additional year of healthy life by putting blood products in front of them at the scene rather than waiting for a hospital.

One hundred and sixty-four dollars. Less than a tank of gas in a fire engine. Less than a shift dinner for the crew. By every standard health economists use to measure value in medicine, it’s not just a good deal. It’s an absurd one.

So you might reasonably ask: if it’s that good, why isn’t it everywhere?

That question, in one form or another, was the heartbeat of the entire conference. The answer has very little to do with science and almost everything to do with a law written when ambulances were essentially station wagons with a red light bolted to the roof.

The economics, in plain English

When health economists want to know whether a treatment is worth the money, they look at what you get for it. The thing you get is measured in something called a quality-adjusted life year, or QALY. Think of it as one full year of healthy living. Not lying in a hospital bed, not bedridden at home, but actually living. Walking your dog. Going to your kid’s soccer game. Showing up at work. Holding your sixth grandchild.

Every medical intervention in the country, from a hip replacement to a chemotherapy regimen, can be evaluated this way. How much does it cost, and how many extra years of healthy life does the average patient gain from it? Divide one by the other and you get a price per QALY.

In the United States, the rough rule of thumb is that anything below about $50,000 per QALY is a good buy. A new cancer drug at $150,000 might still get approved, but with eyebrows raised. Above $200,000, justification gets hard.

Prehospital blood comes in at $164. Roughly three hundred times more cost-effective than the threshold we use to decide whether something is worth doing. It’s not even on the same map as most medical interventions.

That figure comes from a cost-effectiveness analysis presented at the conference by the research group at the University of North Carolina and the University of Alabama at Birmingham, the same team behind the recent nationwide work estimating how many lives prehospital blood would save if it were universally available. The formal paper is in the publication pipeline, but the methodology was rigorous and the result was striking enough to anchor the entire economic discussion at the meeting.

Here’s the part that should reassure any skeptic. That number didn’t come from a back-of-the-napkin calculation. The researchers ran their model ten thousand times, varying every assumption they could think of. What if the blood costs more? What if the survival benefit is smaller? What if you’re in a rural system that only transfuses a handful of patients a year? They tortured the numbers. They tried to make prehospital blood look bad. They couldn’t.

In one sensitivity analysis, they pushed the cost of a transfusion all the way up to eleven thousand dollars per patient, an absurd figure no real program would ever hit. Even at that price, prehospital blood still came in well below the value threshold. In another, they shrunk the mortality benefit down to almost nothing. Still cost-effective. Still a bargain.

If you remember nothing else from this post, remember this. The economics aren’t close. They aren’t debatable. They aren’t waiting for one more study. The economics are done.

Which brings us to the more uncomfortable question.

Why isn’t it everywhere?

The honest answer, delivered repeatedly and from many angles in that DC conference room, is this: the math doesn’t matter if the people writing the checks aren’t allowed to write them.

When a paramedic hangs a unit of blood on a hemorrhaging patient at the side of an interstate, that paramedic has just delivered, in the field, a treatment that ten years ago you could only get in a trauma bay. It’s medicine. It would be billed at thousands of dollars in a hospital. But Medicare doesn’t see it that way. Medicare sees a transportation expense.

In 1965, when the Medicare statute was written, the people drafting the law looked at ambulances and saw exactly what they were at the time: vehicles. The job was to move a patient from point A to point B. The crew weren’t really clinicians; they were drivers with a stretcher. So Medicare wrote ambulance services into the statute as a transportation supplier, paid a flat fee per call, with small bumps for higher levels of service. That language is still on the books. Today. In 2026.

When Dr. Sean Michael, a senior physician leader at CMS, took the microphone, he was as candid as anyone in his position can be. The medicine, he said, is far ahead of the money. CMS pays for the truck and the transportation, not the treatment. The rules they’re working under were written for a world that no longer exists.

So here’s what that looks like on the ground. An EMS agency that staffs an ambulance for around six hundred dollars per shift gets reimbursed about two hundred fifty dollars per call. A higher-acuity call where the paramedic actually administers blood pays roughly five to seven hundred dollars. The blood itself, plus the program infrastructure to deliver it, costs about fifteen hundred dollars per transfusion. Every time we save a life this way, we lose money doing it.

CMS made a quiet but genuinely important change earlier this year, allowing prehospital blood transfusion to qualify a call for the higher reimbursement category. It was the right thing to do. But Dr. Michael himself acknowledged what it actually was: a recognition fix, not a readiness fix. It told the system that this thing we’re doing is real. It did not pay for it.

Real reform requires Congress.

And it gets even more cost-effective from here

One more thing worth flagging, because it affects every economic argument in this post.

When most people in the public hear “prehospital blood,” they picture a unit of low-titer O whole blood riding around on a chase vehicle in a refrigerator. That’s where many of today’s leading programs are. But the future of this field, in my view and in the view of many of the trauma leaders I spoke with in DC, is plasma. And specifically, spray-dried plasma.

Plasma is the fluid component of blood. It carries the clotting factors that bleeding patients are losing fastest, and a growing body of evidence shows that getting plasma into a patient early, sometimes even before red cells, is what bends the survival curve. The randomized PAMPer trial showed a meaningful mortality benefit from prehospital plasma alone in air medical transport. The challenge has always been logistics. Frozen plasma takes time to thaw. Liquid plasma needs cold storage and has a short shelf life. Neither one travels well on a supervisor’s vehicle in rural America.

Spray-dried plasma changes that. It’s a powder. It sits at room temperature for up to two years. It reconstitutes in a couple of minutes. You can carry it in a glove box. You can stockpile it for a mass casualty event without needing a freezer. Velico Medical, one of the companies developing the technology with BARDA funding, completed its first-in-human safety trial in 2025, and the first civilian US blood center installation is up and running in San Antonio. The product is still investigational and not yet FDA-approved, but it’s moving.

Here’s why this matters for the economics. The unit cost of spray-dried plasma will be higher than liquid plasma, roughly $550 per unit compared to about $80 today. That sounds like a step backward until you look at the full picture. The real cost of a prehospital blood program isn’t the unit on the shelf. It’s the cold chain to keep that unit viable, the wastage when it expires before it gets used, the staffing and compliance overhead, and the simple fact that many rural and frontier agencies can’t run a cold-chain program at all. Spray-dried plasma collapses most of those costs. Two-year shelf life at room temperature means dramatic reductions in wastage. No refrigeration on the supervisor vehicle, no telemetry, no power redundancy. Programs that today look at the logistics and walk away will be able to participate. The unit gets more expensive. The program gets cheaper. And the population health benefit, the actual goal, expands to communities that currently get nothing.

So when I tell you prehospital blood comes in at $164 per QALY today, understand that the trajectory is bending downward, not upward. The intervention is going to get cheaper to deliver and easier to scale. The question of whether it’s worth paying for is going to get easier, not harder, every single year from here.

The 1965 problem

If you’re a legislator reading this, or if you work for one, this is the part I’d ask you to lean in on.

The single most important thing the federal government could do for prehospital emergency care is update the EMS authorizing language to reflect what EMS actually is in 2026. We are not a transportation industry. We are a healthcare delivery system that happens to operate out of a vehicle.

That sounds like semantics. It is not. It is the binding constraint on everything else. As long as EMS is classified as transportation, CMS cannot build a real payment model for prehospital treatment. As long as CMS won’t pay for it, commercial insurers won’t either. As long as no one will pay for it, agencies have to choose between staffing ambulances and stocking blood. And patients in the wrong zip code at the wrong moment pay the difference.

There’s a moment from one of the later panels I keep coming back to. A senior physician was reminiscing about the early years of EMS, when paramedics had to call medical control on the radio for permission to defibrillate a patient in cardiac arrest. The medic would describe the rhythm, and the doctor would, in most cases, give the green light. Today that sounds insane. The science settled the argument.

We are in exactly that moment with prehospital blood. The science has settled. The economics have settled. We are waiting on a section of federal statute, and on the political will to update it.

A bargain we keep refusing to take

I want to bring this back to where we started.

Angela Johnson is alive because Chief Harrison and his crew had whole blood on their supervisor vehicle that morning, and because Indiana had agencies brave enough to start carrying it before the reimbursement caught up. The unit that saved her cost about fifteen hundred dollars to deliver. The Medicare reimbursement for her transport, had she been on Medicare, would not have come close to covering it. The economic value of the years she got back, by any conventional measure, runs into the millions.

A separate piece of work from the same research group, this one already published, asked how many lives we lose each year because prehospital blood isn’t widely available. The answer, drawn from four years of national EMS data, is on the order of ten thousand. Annually. Most of those patients are young. Most are in the first hour after injury, the window where the trajectory can still be bent. Many are on Medicare or Medicaid, which means the federal government is, in a quiet and indirect way, already paying for the worst possible outcome. We’re paying for funerals when we could be paying for transfusions.

This is fixable. It requires Congress to act. It requires statehouses to keep funding programs while the federal picture catches up. It requires all of us in EMS to keep showing the data, telling the patient stories, and refusing to let this issue go quiet.

A hundred and sixty-four dollars per year of life saved. Ten thousand lives a year on the table. One Angela Johnson holding her grandchild who almost wasn’t.

If you want to understand what’s at stake, watch Angela tell the story herself. Then watch Izzy Niemczyk’s story, a young woman from Palm Beach who is alive for the same reason. Multiply by ten thousand. Then ask your representative why we aren’t paying it.

Watch their stories

Angela Johnson:

Izzy Niemczyk:

References and further reading

Hrebinko KA, Sperry JL, Guyette FX, Brown JB, Daley BJ, Miller RS, Harbrecht BG, Claridge JA, Phelan HA, Neal MD, Zuckerbraun BS, Yazer MH, Nicholson KJ. Evaluating the Cost-effectiveness of Prehospital Plasma Transfusion in Unstable Trauma Patients: A Secondary Analysis of the PAMPer Trial. JAMA Surgery. 2021;156(12):1131-1139. doi:10.1001/jamasurg.2021.4533

Lammers DT, Betzold R, Henry R, Dilday J, Conner JR, Williams JM, McClellan JM, Eckert MJ, Jansen JO, Kerby J, Holcomb JB, Hashmi ZG. Nationwide estimates of potential lives saved with prehospital blood transfusions. Transfusion. 2025;65(Suppl 1):S22-S29. doi:10.1111/trf.18174

Hashmi ZG, Jansen JO, Kerby JD, Holcomb JB, et al. Nationwide estimates of the need for prehospital blood products after injury. Transfusion. 2022;62(Suppl 1):S203-S210. doi:10.1111/trf.16991

Hurson T, Schaefer R, Carico C, Griffin R, Bank E, Krohmer J, Jenkins D, Holcomb J, Hashmi Z. Evaluating reimbursement for prehospital blood transfusions: A nationwide survey. Transfusion. 2025;65(Suppl 1):S6-S13. doi:10.1111/trf.18217

Schaefer RM, Bank EA, Krohmer JR, Haskell A, Taylor AL, Jenkins DH, Holcomb JB. Removing the barriers to prehospital blood: a roadmap to success. Journal of Trauma and Acute Care Surgery. 2024;97(2S):S138-S144. doi:10.1097/TA.0000000000004378

Levy MJ, Garfinkel EM, May R, Cohn E, Tillett Z, Wend C, et al. Implementation of a prehospital whole blood program: lessons learned. Journal of the American College of Emergency Physicians Open. 2024;5(2):e13142. doi:10.1002/emp2.13142

Sperry JL, Guyette FX, Brown JB, Yazer MH, Triulzi DJ, Early-Young BJ, et al. Prehospital Plasma during Air Medical Transport in Trauma Patients at Risk for Hemorrhagic Shock (PAMPer). New England Journal of Medicine. 2018;379(4):315-326. doi:10.1056/NEJMoa1802345

Pusateri AE, Moore EE, Moore HB, Le TD, Guyette FX, Chapman MP, et al. Association of Prehospital Plasma Transfusion With Survival in Trauma Patients With Hemorrhagic Shock When Transport Times Are Longer Than 20 Minutes: A Post Hoc Analysis of the PAMPer and COMBAT Clinical Trials. JAMA Surgery. 2020;155(2):e195085. doi:10.1001/jamasurg.2019.5085

Brown JB, Yazer MH, Kelly J, Spinella PC, DeMaio V, Fisher AD, Cap AP, Winckler CJ, Beltran G, Martin-Gill C, Guyette FX. Prehospital Trauma Compendium: Transfusion of Blood Products in Trauma. A Position Statement and Resource Document of NAEMSP. Prehospital Emergency Care. 2025. doi:10.1080/10903127.2025.2476195

Peng HT, Singh K, Rhind SG, da Luz L, Beckett A. Dried Plasma for Major Trauma: Past, Present, and Future. Life. 2024;14(5):619. doi:10.3390/life14050619

Velico Medical. FrontlineODP Spray Dried Plasma System. First-in-human Phase I clinical trial completed 2025. Currently investigational; not FDA-approved. https://www.veli.co

Imagine your crew is working a pediatric cardiac arrest. It’s 0200. The screen on their dosing tool glows in the dark. It reads: Epinephrine 1.0 mL.

The decimal point is small. The screen is cracked. Everyone is moving fast. Someone reads “10 mL.”

That child receives ten times the intended dose.

I know this scenario because I’ve lived a version of it. In 2005, I made a tenfold dosing error on a pediatric patient. I was the physician. I knew the rules. And it still happened.

That experience changed how I think about medication safety. Good intentions are never enough. The systems and tools around us either make the right thing easier, or they don’t.

This isn’t theoretical. The Institute for Safe Medication Practices (ISMP) has documented trailing zero errors for decades, identifying them as one of the most preventable causes of tenfold medication mistakes in healthcare.

The good news is this is fixable.

The Rule Isn’t Complicated

Never write a trailing zero after a decimal point in medication dosing.

Write 1 mg, not 1.0 mg. Write 5 mL, not 5.0 mL.

This isn’t a formatting preference. It’s a patient safety standard.

It’s taught in training. It’s tested on licensing exams. It’s embedded in hospital policy. The Joint Commission placed it on the “Do Not Use” list over twenty years ago, and it has been reinforced ever since by the FDA, WHO, and U.S. Pharmacopeia.

The reason is simple.

Decimal points disappear.

Under stress, in a moving ambulance, on a dim or damaged screen, “1.0” becomes “10” in an instant. In pediatrics, where every dose is weight-based, that difference can be catastrophic.

Why This Matters More in the Digital Era

Technology has raised the standard for what’s possible in EMS. It has also raised the responsibility to get the basics right.

Clinicians trust what they see on a screen. When a clinical system displays “1.0 mL,” it does more than introduce the risk of misreading. It teaches that this is acceptable. It normalizes the format. That number gets spoken, documented, and handed off. The risk travels with it.

Well-designed systems don’t just deliver information. They model best practice.

When they do, they reduce cognitive load and protect clinicians in high-stress moments. When they don’t, they quietly widen the gap between intention and outcome.

A Simple Standard Worth Checking

If your system uses a digital dosing tool, it’s worth taking a direct look at how doses are displayed.

• Does it show 1 mg or 1.0 mg?

• Does it include a leading zero for doses like 0.5 mg, or display .5 mg, which can be misread as 5?

These are straightforward questions. They’re the same ones any pharmacist, safety officer, or accreditation surveyor would ask.

I’ve shared what a tenfold error costs the person who makes it.

I still think about the one I made.

I remember the moment. I remember the weight of it. That doesn’t go away.

I don’t share that story for sympathy. I share it because I don’t want anyone else carrying that with them.

We ask our clinicians to perform under pressure, in imperfect conditions, often in the middle of the night. They deserve systems designed to protect them when it matters most.

The decimal point is small.

But the consequences aren’t.

I’ve made this mistake.

You don’t have to.

References

• The Joint Commission. “Do Not Use” List (first published 2004; reaffirmed annually)

• Institute for Safe Medication Practices (ISMP). List of Error-Prone Abbreviations, Symbols, and Dose Designations (2021)

• U.S. Food and Drug Administration (FDA). Safety Considerations for Product Design to Minimize Medication Errors: Guidance for Industry (2016)

• World Health Organization (WHO). Medication Safety in High-Risk Situations (2019)

• U.S. Pharmacopeia (USP). General Chapter <7> Labeling and General Chapter <17> Prescription Container Labeling

Figure 1: Pad Placement in the Three Defibrillation Strategies

Alan is not a statistic. But his survival is also not a miracle. It is the result of a system that had asked the right questions and built protocols around the best available answers. One of those questions is the subject of a new paper by Cheskes and colleagues, just published in Resuscitation. And the answer it offers should change how every EMS system in the country thinks about where to place defibrillation pads.

The study examined the relationship between transthoracic current delivery, pad position, and return of spontaneous circulation in 342 patients with refractory ventricular fibrillation, analyzing over 1,800 individual shocks across multiple defibrillator platforms. What they found upends a foundational assumption about how defibrillation works.

Anteroposterior (AP) pad placement delivered higher transthoracic current than anterolateral (AL) placement across every shock analyzed. AP pads consistently demonstrated lower impedance and greater current through the chest wall. If current is what terminates VF, AP should win by every measure.

But current wasn’t what predicted ROSC. Pad position was.

In the generalized estimating equation model pooling all interventional shocks, AP positioning was independently associated with ROSC at more than double the odds of AL positioning (aOR 2.01; 95% CI: 1.12 to 3.59). Delivered current, despite the measurable difference between positions, was not associated with ROSC at all. While current is important in defibrillation success, the signal wasn’t in the amps. It was in the path.

Cheskes explained the mechanism directly in a recent interview with the St. Emlyn’s podcast. “The vast majority of refibrillation that occurs in the ventricle happens in the posterior inferior aspect of the ventricle,” he said. “That’s the part that actually is defibrillated more thoroughly by AP defibrillation and may be missed by AL defibrillation.” When pads are placed anteroposteriorly, current traverses more of the myocardium, including those posterior regions that standard placement may not adequately reach. The shock fires, the monitor shows termination, and somewhere in the posterior wall, fibrillation persists just long enough to trigger refibrialltion in the the whole ventricle. The patient goes back into VF, and no one fully understands why. As Cheskes put it simply: the critical question is not the absolute value of current you deliver, but what path that current travels through the heart.

This mechanism helps explain both the benefit of vector change and, more importantly, why DSED outperforms it. Both strategies place pads in the AP position, which is the shared ingredient. But Cheskes was emphatic in the podcast that DSED and vector change are not the same therapy. The original DOSE VF trial, published in the New England Journal of Medicine in 2022, showed that DSED produced roughly double the neurologically intact survival of standard defibrillation. Vector change improved survival but did not reach significance for neuro-intact survival, and the rate of rearrest was substantially higher with vector change than with DSED. When you add AP to AL in DSED, Cheskes explained, “you get a much more homogeneous distribution of current across the whole ventricle.” The second shock in DSED does something additional: it arrives when the myocardium is in a different electrical state after the first shock, making it more susceptible to termination. AP pad position is a key mechanism. DSED leverages it fully.

Cheskes himself described this new pad position paper on LinkedIn as likely “the most impactful” mechanistic work to emerge from the DOSE VF series. The doubling of probability of ROSC for AP versus AL positioning, he noted, “will certainly generate plenty of discussion on where pads should be placed during cardiac arrest.” That is an understatement. It shifts the question from “should we use DSED?” to “should every cardiac arrest patient receive AP pad placement from the start, with DSED as the primary strategy?”

The guideline landscape around this question is more fractured than it should be. The European Resuscitation Council and ILCOR currently recommend anterolateral placement for initial defibrillation, citing faster application as the rationale. The American Heart Association endorses either position as acceptable. These recommendations were formed without data directly examining the comparative efficacy of AP versus AL placement on ROSC. When asked on the podcast about the speed argument for AL, Cheskes was direct: “Speed does not equal efficacy. If you’re defibrillating quickly using AL, but it’s not as effective as AP, it should give people food for thought.” He went further: “Trained responders can take a couple more seconds, put the pads in a more efficacious position (AP) and deliver a shock that’s much more likely to produce ROSC than the AL position.”

And he was explicit about what his own systems are already doing. “We start with pads in the AP position. I think that’s warranted by the science, even pending a randomized trial. Sometimes you don’t need a randomized trial. Sometimes you need common sense.”

This is observational work, and Cheskes acknowledges it. This is a secondary analysis of a randomized trial, and the findings are hypothesis-generating rather than confirmatory. To change practice at scale, a randomized trial is needed, and that trial is coming. Cheskes confirmed the team is preparing the POSITION RCT, designed specifically to compare AP versus AL as the initial pad position in cardiac arrest. The science is ready to be tested definitively.

Until that answer arrives, the more instructive question may be what high-performing systems are already doing. The emerging consensus among resuscitation scientists who have followed the DOSE VF literature closely looks something like this: AP pads from the start, DSED deployed early rather than as a last resort, minimizing epinephrine in shockable rhythms, antiarrhythmics given earlier rather than waiting for shock three or five, IV over IO for medication delivery, and in some systems the addition of esmolol as an adjunct. These are not radical departures from guidelines. They are protocol decisions made by systems that stopped waiting for perfect guideline alignment and started building around the best available science.

This is not the first time EMS has found itself holding strong observational evidence while guideline bodies deliberate. In the epinephrine debate, in the antiarrhythmic literature, in the evolving DSED data, we keep encountering the same pattern: the physiologically intuitive assumption doesn’t always survive contact with patient-centered outcome data. The assumption that more current to the chest equals better defibrillation is the latest entry in that line.

Alan Nierow went home. In fact, he recently invited me to play doubles on the same tennis court he collapsed on. He is alive because an EMS system asked better questions, built better protocols, and placed the pads where the science said to put them. There are other Alans out there. The question is whether they will be lucky enough to collapse in a system that has done the same.

Pictured (Left to right): Peter Antevy, Joyce Nierow, Alan Nierow, Adam Nierow (11/2025)

References:

1. Cheskes S, Drennan IR, Turner L, et al. The impact of defibrillation current and pad position on return of spontaneous circulation during refractory ventricular fibrillation. Resuscitation, 2026.

2. Cheskes S, Verbeek PR, Drennan IR, et al. Defibrillation strategies for refractory ventricular fibrillation. N Engl J Med. 2022;387:1947-56.

3. St. Emlyn’s Ep. 289: Refractory VF, Double Sequential Defibrillation, and the Future of Cardiac Arrest

We are grateful to Alan Nierow and his family for allowing us to share his story

Pediatric medication dosing has never had that same discipline. Published error rates in the prehospital setting range from 36% to 49%. In 2020, the National Association of EMS Physicians published evidence-based guidelines to address the problem: confirm patient weight at the time of care, document in kilograms, use precalculated volume-based dosing tools, and train clinicians with the same equipment they carry in the field. The recommendations were clear. What nobody had shown was whether implementing them would produce sustained, measurable improvement. Until now.

A paper recently published in Prehospital Emergency Care begins to close that gap.

Maia Dorsett, MD, PhD, and colleagues at the University of Rochester describe a multi-year quality improvement initiative across the Monroe-Livingston EMS System in upstate New York. This is a large, complex system: 12 ALS transport agencies, 9 BLS transport agencies, 38 BLS first response fire departments, roughly 350 paramedics and 1,500 EMTs operating under a common set of protocols. About 190,000 calls for service annually.

Using state registry data, the authors calculated baseline pediatric medication dosing accuracy for patients aged 13 and younger. From 2020 to 2022, only 79% of first-dose medications were correctly dosed. That means roughly one in five pediatric patients received the wrong dose. The worst performer was midazolam for seizure, correctly dosed only 65% of the time.

The team implemented a protocol-integrated, volume-based dosing platform, distributed length-based tapes for weight estimation, and built a train-the-trainer education program with hands-on case scenarios. This was not a passive rollout. They surveyed medication concentrations across every ALS agency in the system. They configured the platform to flag the one medication (ketamine) available in more than one concentration. They piloted with a small group, incorporated feedback from a clinician with color vision deficiency, and secured formal policy approval before going live. Training began in mid-2023, and by September of that year, more than 350 active users were on the platform.

The initial results were promising. Sustaining them required a deeper commitment.

Initial improvements were real. Overall correct dosing climbed from 79% to a median of 91% by fall 2023. But it plateaued short of the 95% target. Then, in May 2024, a protocol change doubled the IM/IN midazolam dose from 0.1 mg/kg to 0.2 mg/kg while leaving the IV dose unchanged. That introduced a new category of error: clinicians giving the old IV dose via the IM route, effectively underdosing by half.

Education did not fix it. The team tried retraining. They tried emphasizing the seizure indication within the platform. They changed font colors. None of it worked.

What worked was redesigning how the dose was displayed. They highlighted the IM/IN dose in black with yellow text and listed it first, making the correct action the most visually prominent action. That is a human factors intervention, not an educational one. And it moved the needle where education alone could not.

This is exactly what high-performing agencies do in cardiac arrest. They adopt high-performance CPR choreography and refine it using CPR Assist. They review every arrest through CodeStat or CaseReview, measuring CPR fraction, compression rate and depth, time to first shock, time to first epinephrine for non-shockable rhythms. They shift to anteroposterior pad placement as default. Others adopt dual sequential external defibrillation for refractory VF. None of those changes happened all at once. Each was measured, tested, and refined through iterative cycles. That is the same process Dorsett’s team applied to pediatric medication dosing.

By December 2025, the system had achieved 97% correct dosing for all first-dose pediatric medications across the entire region, exceeding their prespecified aim of 95%. The improvement was sustained, not a single-month spike.

There are a few things about this paper that matter beyond the topline number.

First, the authors explicitly demonstrate that implementation alone did not solve the problem. They cite a Michigan study by Kazi and colleagues showing no reduction in dosing errors after implementation of a pediatric dosing reference without surrounding QI infrastructure.

Dorsett’s team built the infrastructure. Monthly data review. Case-by-case error analysis. Categorization of every dosing error into one of three buckets: no access to the platform, had access but didn’t use it, or used it and still made an error. Each bucket demanded a different intervention. That kind of granularity is the difference between having a tool and having intelligence about how your system is actually performing. It is not enough to deploy technology. You have to know whether it is being used, how it is being used, and where it is falling short.

Second, the improvements extended across the entire medication formulary. Dexamethasone went from 88% to 99% correct. Epinephrine for anaphylaxis went from a volatile baseline to 97%. Diphenhydramine from 74% to 97%. Fentanyl hit 100% in 2025. The midazolam story gets the attention because it was the hardest problem to solve, the one that required iterative redesign even after the platform was in place. But the broader finding is that a system-level approach lifted every medication.

Third, this is an independent academic study published in a peer-reviewed journal, with no industry funding and no competing interests declared. The authors used SQUIRE 2.0 methodology and Shewhart statistical process control charts. They state plainly that the platform was one component of a comprehensive approach. That is exactly the kind of rigorous, honest evaluation this field needs more of.

This is the lesson the field needs to internalize. An agency can believe it is dosing pediatric patients correctly and have no idea that one in five kids is getting the wrong dose until someone actually measures it. Rochester measured it. And then they did the hard, iterative work to fix it.

This is not the first time this kind of evidence has emerged for the Handtevy platform specifically. In 2023, Rappaport and colleagues at Denver Health published a study in the same journal showing that implementation of Handtevy’s precalculated, volume-based field guide improved correct pediatric dosing from 51% to 89% across their EMS system.

That was an important finding. What Dorsett’s paper adds is the next chapter: what happens when you pair the platform with sustained QI infrastructure, when you refuse to accept the initial plateau, and when you iterate through protocol changes, retraining, and human-factors redesign over multiple years. The result is 97%. The trajectory from Rappaport to Dorsett is the trajectory from proof of concept to proof of sustainability. Two independent research teams. Two different EMS systems. The same platform. The same direction.

One final detail worth noting. The local EMS chiefs’ association, which collectively purchases the platform, funded its expansion to every BLS clinician and every BLS first response fire apparatus in the system. By the end of 2025, there were 766 licensed users, with 89% having accessed the platform within the prior six months. The two primary paramedic training programs in the region have now integrated the platform and cognitive offload strategies into their curricula. That is what sustainability looks like.

This matters because not all tools in this space are supported by the same evidence. The Kazi study out of Michigan demonstrated that implementing a pediatric dosing reference without surrounding QI infrastructure produced no reduction in errors. The tool itself was not enough. But it is also true that the design of the tool matters. A platform built for rapid access, cognitive offload, and volume-based dosing integrated directly into local protocols is fundamentally different from a static reference that still requires calculation at the bedside. Dorsett’s results did not happen by accident. They happened because the platform was designed to reduce cognitive burden and the QI process was designed to catch what the platform alone could not.

As noted above, both the Dorsett and Rappaport studies used Handtevy. We have made it a point since the platform’s inception to stay out of the evaluation. We do not design the studies, fund them, or participate in the analysis. We believe that if the platform works, others will demonstrate it on their own terms, through rigorous, independent research. That is what happened here.

It is also worth noting that Dr. Dorsett directs the NAEMSP Year-Long Quality Improvement and Safety Course, training the next generation of QI leaders in EMS. The methodology in this paper reflects that expertise. And it is not limited to pediatrics. This is a model for how any clinical domain in EMS can move from assumption to evidence.

Pediatric medication dosing errors are a system problem. System problems require system solutions. A platform without QI infrastructure is a product. A platform embedded in a culture of measurement, feedback, redesign, and collaboration is a safety system. That is what Clinical Intelligence looks like in practice. Not a single tool, but a closed loop between the point of care and the data that drives improvement.

EMS learned how to improve cardiac arrest outcomes through measurement, iteration, and a refusal to accept the status quo. This paper proves the same discipline works for pediatric medication safety.

Seventy-nine percent was the starting line. Ninety-seven percent is what happens when you refuse to stop there.

References

1. Dorsett M, O’Donnell S, Strickland W, Cheng W, Sensenbach B, Cannuli M, Eisenhauer AM, Cushman JT. Improving Pediatric Medication Safety Through Implementation of Evidence-Based Guidelines. Prehospital Emergency Care. Published online March 18, 2026. DOI: 10.1080/10903127.2026.2645140

2. Rappaport LD, Markowitz G, Hulac S, Roosevelt G. Medication Errors in Pediatric Patients after Implementation of a Field Guide with Volume-Based Dosing. Prehospital Emergency Care. 2023;27(2):213–220. DOI: 10.1080/10903127.2022.2025962

3. Cicero MX, Adelgais K, Hoyle JD, Lyng JW, Harris M, Moore B, Gausche-Hill M. Medication Dosing Safety for Pediatric Patients: Recognizing Gaps, Safety Threats, and Best Practices in the Emergency Medical Services Setting. A Position Statement and Resource Document from NAEMSP. Prehospital Emergency Care. 2021;25(2):294–306. DOI: 10.1080/10903127.2020.1794085

The paramedic crew stood in the doorway of a small apartment, looking at a woman in her seventies who had been vomiting since midnight. She was pale, mildly dehydrated, and scared. The standard play was obvious. Load her up, drive her to the closest emergency department, hand off the paperwork, and move on to the next call. That is what EMS has always done. That is what the system has always paid for.

But what if, instead, a physician who actually knew this woman, who had treated her cancer for months, could direct a critical care paramedic to her bedside within thirty minutes? What if that paramedic could run labs, start fluids, administer antiemetics, and call the physician back with real-time results, all without the patient ever leaving her home?

That is not a hypothetical. We did it.

A recently published commentary in Prehospital Emergency Care by Michael Levy offers a clear-eyed analysis of the challenges faced during the CMS Emergency Triage, Treat, and Transport (ET3) program, and it reads like a cautionary tale about the distance between good ideas and successful implementation. Levy was responding to a study by Solazzo and colleagues describing the rollout of a treat-in-place (TIP) program within a large urban EMS system in New York City. The results were sobering. Of the patients who were eligible for a TIP encounter, 73% were never even offered one. Among those who were offered the option, 92% refused. The entire program produced just 31 completed treat-in-place encounters.

Levy pulls no punches in his analysis. He raises critical questions about whether the investigators anticipated the cultural sea change required for success, whether video-based training was sufficient to win over the hearts and minds of clinicians, and whether course corrections were made as problems became apparent. He also zooms out to the national picture: the ET3 program was projected to run five years but was canceled two years early. Of the 185 enrolled ambulance suppliers and hospital-based providers, only 70 delivered even one paid ET3 intervention. The TIP component accounted for 90% of those interventions, but 75% came from just eight agencies. Nationwide, the program produced just over 3,000 treat-in-place encounters across its entire lifespan.

The failure was not in the concept. It was in the execution.

Levy writes that agencies which successfully implemented TIP shared common traits: strong organizational commitment, clear protocols requiring clinicians to offer TIP to eligible patients, close partnerships with alternate care options, and robust quality improvement infrastructure. That description matches, almost word for word, the program we built in South Florida in 2015.

In collaboration with ChenMed and American Ambulance (a Falck subsidiary), we launched Transitional Health Solutions (THS), a mobile integrated healthcare program designed around a fundamentally different premise than what most ET3 participants attempted. Our paramedics did not operate in isolation. Every call was bookended by the patient’s own primary care physician, who directed care via telephone or secure video before the paramedic arrived and again after the encounter was complete. Our clinicians were all critical care certified. They carried i-STATs, PT/INR testing, urine dipsticks, and a full medication cache that included antibiotics, diuretics, insulin, bronchodilators, and antiemetics. They could run a basic metabolic panel in a patient’s kitchen and transmit the results to the directing physician in real time.

The training was not a video module. It was an intensive, hands-on program that fundamentally reoriented paramedics from the speed of 9-1-1 response to the slow, methodical approach required for in-home patient care. We brought in national experts. We established protocols and formalized operational decisions before the first patient was seen.

And the results spoke for themselves. Over 506 acute episodic MIH encounters in a 12-month period, 84% of patients stayed home. Only 2% required 9-1-1 transport. Forty-two percent received oral or parenteral medications, 33% received an EKG, 14% received point-of-care lab testing, and 11% received IV fluids. There were zero adverse events reported by patients or field clinicians. Estimated cost savings exceeded $1.4 million. The data was compelling enough that we were invited to present our findings directly to CMS.

The contrast with the national ET3 experience could not be more striking. Where ET3 participants struggled to produce 3,000 total interventions over three years, our single program generated 506 encounters in its first year alone. Where the NYC program managed 31 completed treat-in-place encounters, we kept 426 acutely ill Medicare patients safely at home.

So what made the difference?

Levy identifies the core issue precisely: broad agreement on a sentiment is insufficient to cause an outcome. The EMS community has long agreed that treat-in-place is a good idea. Everyone nods along at conferences. Everyone acknowledges that transporting every low-acuity patient to an emergency department is wasteful and often harmful. But agreeing that something should work and building a system that actually works are entirely different undertakings.

Our program succeeded because it was physician-directed from the start. Not physician-adjacent. Not physician-available-if-needed. The physician was on the call before the paramedic walked through the door and was on the call again when the paramedic was ready to leave. That model eliminated the ambiguity and operational risk that plagued so many ET3 participants. Clinicians in the field were never left wondering whether it was safe to treat in place, because the patient’s own physician was making that determination in real time.

The financial model mattered too. ChenMed operated as a full-risk managed services organization, meaning they bore the complete financial burden of patient care. Hospitals, as Levy implicitly acknowledges, have complicated incentives. They may publicly support reducing unnecessary ED visits while privately benefiting from full waiting rooms. Full-risk organizations like ChenMed had every reason to keep patients healthy and out of the hospital. The incentives were aligned from day one.

And the patient selection was deliberate. We focused on the sickest 20% of the Medicare population, patients with multiple comorbidities who were already known to their physicians, not on undifferentiated 9-1-1 callers with no established care relationship. That made the clinical decision-making cleaner, the communication pathways shorter, and the trust between patient and clinician stronger.

Levy closes his commentary by calling for adoptable, flexible, scalable models of TIP that include training, education, quality guidelines, and effective change management. That call is exactly right. The models exist. The data exists. What the field needs now is the institutional will to stop treating mobile integrated healthcare as a pilot project and start treating it as a standard of care.

The woman in the apartment that night was named Mildred. Her cancer would ultimately take her life months later. But on that particular evening, a physician-directed paramedic kept her comfortable, kept her home, and kept her son from spending his brief visit watching his mother sit in an emergency department hallway.

That is what treat-in-place looks like when it is done right.

References

1. Levy M. Lessons from the implementation of emergency medical services treat-in-place programs. Prehosp Emerg Care. Published online January 21, 2026. doi:10.1080/10903127.2025.2588672

2. Castillo L, Solazzo EM, Blutinger E, et al. ET3 treat in place program implementation in a large urban EMS system. Prehosp Emerg Care. Published online October 28, 2025. doi:10.1080/10903127.2025.2576564

3. Centers for Medicare and Medicaid Services. Emergency Triage, Treat, and Transport (ET3) model: final evaluation report. January 2025. https://www.cms.gov/priorities/innovation/data-and-reports/2025/et3-model-final-eval-rpt

4. Centers for Medicare and Medicaid Services. Emergency Triage, Treat, and Transport (ET3) model. https://www.cms.gov/priorities/innovation/innovation-models/et3

5. Antevy P, Goldman H. Mobile integrated healthcare in an acutely ill Medicare cohort: a descriptive study. Poster presented at: NAEMSP Annual Meeting; January 2017.

6. Antevy P. How care collaboration is improving patient outcomes. EMS World. 2017;46(4):26-33.

It is 2 AM, and your crew is two minutes out from a witnessed cardiac arrest. Bystander CPR is in progress. The patient is 58, collapsed at a restaurant, and the 911 caller says the AED advised a shock. Your partner is already prepping the monitor. You are running through the algorithm in your head. HP-CPR, Defibrillate. Epinephrine. Amiodarone. Repeat.

But what if that third step, the one drilled into every paramedic and every ACLS provider on the planet, is actually making things worse?

That is not a rhetorical question anymore. It is a data-driven one. And the answer is becoming difficult to ignore.

Palm Beach Gardens Fire Rescue ran this protocol in 2025. They removed epinephrine from their shockable rhythm algorithm, prioritized anterior-posterior pad placement, early dual sequential external defibrillation, early amiodarone, and relentless focus on high-quality CPR. The results, reported through CARES, are hard to look away from.

Among 61 cardiac arrest patients, PBGFR achieved a 41% ROSC rate, nearly double the national average of 22.1%. Survival to hospital discharge reached 31.1%, compared to 9.7% nationally, a 3.2-fold difference. Neurologically intact survival hit 23%, more than three times the national rate of 7.5%. And Utstein survival, the benchmark for witnessed shockable arrests, reached 66.7%, compared to 30.9% nationally.

Those are small numbers. Sixty-one patients. But 3.1 times the national average for neurologically intact survival is not a statistical artifact. Something is working.

The honest answer is that we do not know exactly which variable is driving these outcomes. Is it the removal of epinephrine? The anterior-posterior pad placement? Early dual sequential defibrillation? The early push of amiodarone? Or is it simply what we have always known matters most, excellent BLS care and relentless chest compressions? It is likely some combination of all of these, working in concert within a system that has decided to let the evidence guide its protocols rather than tradition.

But here is what we do know. The published evidence on epinephrine in shockable rhythms has been accumulating for years, and it all points in the same direction.

A 2025 propensity-matched analysis published in Critical Care examined 3,163 patients with refractory ventricular fibrillation, meaning their VF persisted after three or more defibrillation attempts. Among patients who did not receive epinephrine, 50% achieved a favorable neurological outcome. Among those who did receive it, that number dropped to 11%. Across multiple analytical approaches, epinephrine was consistently associated with worse neurological outcomes, with an adjusted odds ratio of 0.24. That is not a marginal difference. That is a fourfold reduction in the likelihood of walking out of the hospital neurologically intact.

A 2023 network meta-analysis in Chest reinforced this pattern from a different angle. Standard-dose epinephrine improved survival to discharge in non-shockable rhythms, with an odds ratio of 2.10. But in shockable rhythms, the odds ratio was 0.85, crossing the null line and suggesting no benefit at all. A drug that helps one type of arrest may be neutral, or worse, in another.

Observational data from in-hospital cardiac arrest tells a similar story. Hospitals with the highest rates of epinephrine administration before initial defibrillation had 10% lower risk-adjusted survival compared to hospitals with the lowest rates. More epinephrine, worse outcomes. The American Heart Association’s own 2025 guidelines acknowledge that systematic reviews consistently show the benefit of epinephrine in shockable rhythms is less than in non-shockable rhythms.

Then there is the PARAMEDIC2 trial, the largest randomized placebo-controlled study of epinephrine in cardiac arrest ever conducted. Published in the New England Journal of Medicine, it demonstrated that epinephrine increased return of spontaneous circulation. More patients got a pulse back. But when researchers looked at the outcome that actually matters, survival at three months with favorable neurological function, there was no significant difference. Getting a pulse back is not the same as getting a life back. Post-hoc analysis further revealed that whatever benefit epinephrine might offer converges to zero at approximately 20 minutes of pulselessness.

The pathophysiology makes sense once you step back and look at it. VF and pulseless ventricular tachycardia are fundamentally different from asystole and PEA. In shockable rhythms, the heart still has electrical activity. It is chaotic, but it is there, and it is amenable to electrical conversion. Defibrillation is the definitive treatment. Epinephrine’s vasoconstriction, so useful for generating coronary perfusion pressure in a heart with no electrical activity, becomes counterproductive when the heart is fibrillating. It increases myocardial oxygen demand. It promotes recurrent VF after successful defibrillation. It impairs microvascular perfusion. In other words, it fights the very recovery that defibrillation is trying to achieve.

This is why Palm Beach County Fire Rescue has removed epinephrine from its shockable rhythm standing orders. The updated protocol focuses on what the evidence supports: high-quality CPR, early and repeated defibrillation, and antiarrhythmic therapy. Epinephrine remains fully indicated for non-shockable rhythms, where the data continues to support its use. This is not an abandonment of pharmacology. It is a refinement based on the recognition that different cardiac arrest presentations require different approaches.

It is also possible that the answer is not a complete removal of epinephrine but a recalibration. A fraction of the current dose. A single dose instead of repeated pushes every three to five minutes. The final answer may be more nuanced than a binary yes or no. But the current standard of 1 mg every three to five minutes in a patient whose heart is ready to be defibrillated is not supported by the evidence we have today.

The AHA’s 2020 guidelines, 2024 focused update, and 2025 full guidelines have all moved in this direction, recommending that epinephrine be deferred in shockable rhythms until initial CPR and defibrillation attempts are unsuccessful. Our protocol takes that trajectory to its logical conclusion. If the evidence shows no neurological benefit and potential harm, the responsible move is to remove the intervention rather than simply delay it.

This is a conversation the entire EMS community needs to have. We have spent decades building algorithms around a drug that may be doing more harm than good in the very patients with the highest potential for meaningful survival. Shockable rhythms carry the best prognosis in cardiac arrest. These are the patients we are most likely to save. The question is whether we are willing to update our approach based on where the evidence now points.

Palm Beach Gardens is running this protocol. Their patients are surviving at more than three times the national rate with brains intact. The sample is small. The questions about which variables matter most remain open. But when you see numbers like that, you do not wait for a perfect answer before you start asking better questions.

The best resuscitation drug for a patient in VF has always been electricity. The evidence just finally caught up.

References:

1. Association of Epinephrine and Outcome in Cardiac Arrest With Refractory Shockable Rhythm: A Population-Based, Propensity-Score Matched Analysis. Critical Care. 2025. Fanet L, Javaudin F, Dumas F, et al.New

2. Epinephrine in Out-of-Hospital Cardiac Arrest: A Network Meta-Analysis and Subgroup Analyses of Shockable and Nonshockable Rhythms. Chest. 2023. Fernando SM, Mathew R, Sadeghirad B, et al.

3. Part 9: Adult Advanced Life Support: 2025 American Heart Association Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. Circulation. 2025. Wigginton JG, Agarwal S, Bartos JA, et al.GuidelineNew

4. 2023 American Heart Association Focused Update on Adult Advanced Cardiovascular Life Support: An Update to the American Heart Association Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. Circulation. 2024. Perman SM, Elmer J, Maciel CB, et al.Guideline

5. Part 3: Adult Basic and Advanced Life Support: 2020 American Heart Association Guidelines for Cardiopulmonary Resuscitation and Emergency Cardiovascular Care. Circulation. 2020. Panchal AR, Bartos JA, Cabañas JG, et al.Guideline

6. A Randomized Trial of Epinephrine in Out-of-Hospital Cardiac Arrest. The New England Journal of Medicine. 2018. Perkins GD, Ji C, Deakin CD, et al.

A seven-year-old collapses at school. The teacher calls 911. Paramedics arrive within minutes. The child is postictal, confused, not following commands. The EMS crew documents a seizure, starts an IV, and transports to the nearest pediatric emergency department.

Nobody considers stroke.

Twelve hours later, an MRI reveals an acute ischemic infarct in the left middle cerebral artery territory. The child has a dense right hemiparesis. The window for intervention has closed.

This scenario is not hypothetical. It is the norm.

A new study from the University of California system analyzed 3,888 children transported by EMS for acute neurological symptoms over four years. The findings put numbers behind something many of us have suspected: EMS almost never considers stroke in kids. Among nearly 4,000 activations for pediatric neurological emergencies, only 13 patients (0.3%) had an EMS primary impression of “stroke.” And none of those 13 actually had one.

But here’s what caught my attention.

When EMS documented “stroke” as the primary impression, 100% of those patients received neuroimaging in the ED. All 13 had CT scans. Eleven had MRIs. In other words, when EMS framed the case as a potential stroke, the emergency department responded accordingly. The EMS impression shaped the ED’s diagnostic approach, even when the suspicion was ultimately incorrect.

This is anchoring bias working in our favor.

For the 695 children who received neuroimaging, the median time from EMS activation to CT was 2.3 hours. The median time to MRI was 26.8 hours. That delay to MRI is the critical finding. Unlike adults, where CT suffices to rule out hemorrhage and guide thrombolytic therapy, MRI is essential for diagnosing pediatric ischemic stroke. CT misses most acute ischemic strokes in children. Among the 125 patients who received MRI, 9 had acute strokes. Another 31 had actionable findings: CNS infections, tumors, vascular malformations. These children needed MRI to establish the diagnosis and guide treatment.

The study also found disparities in imaging decisions. Children under 5 were less likely to receive neuroimaging, despite being at higher stroke risk than older children. Hispanic children were significantly less likely to be imaged compared to other races and ethnicities (15.5% vs. 20.8%). And most pediatric stroke patients didn’t arrive by EMS at all. During the study period, 43 children were admitted with out-of-hospital stroke. Only 21% arrived by ambulance. A third were walk-ins brought by family. Nearly half were transfers from other facilities.

None of this is surprising when you think about it. Pediatric stroke doesn’t look like adult stroke. Kids present with seizures, altered mental status, severe headache with vomiting, loss of consciousness. The classic adult stroke tools we rely on, CPSS, FAST, RACE, were designed for focal deficits and aphasia, symptoms that are often subtle or absent in children. Layer in a much higher rate of stroke mimics (migraine, seizure, trauma), and it becomes clear why EMS rarely puts stroke on the differential for a pediatric patient.

That may be starting to change. The 2026 AHA/ASA Acute Ischemic Stroke Guidelines now include the first-ever pediatric-specific stroke screening tools, and one in particular deserves attention: the PedRACE.

The Pediatric Rapid Arterial Occlusion Evaluation scale was adapted from the validated adult RACE scale by the Catalan Stroke Network. It is a simple 5-item assessment: facial palsy, arm motor function, leg motor function, head and gaze deviation, and aphasia or agnosia. It can be administered in about two minutes. In a prospective study of 50 children evaluated through the Catalan Pediatric Stroke Code, prehospital and emergency staff scored the PedRACE comparably to child neurologists, with 82% full agreement across all items and a weighted kappa of 0.920. The scale correlated strongly with the PedNIHSS for both child neurologists (Spearman ρ 0.852) and prehospital/ED staff (Spearman ρ 0.781). A PedRACE score of 5 or higher was significantly associated with large vessel occlusion, the subset of pediatric strokes most likely to benefit from thrombectomy.

This is a tool designed for non-neurologists. It does not require advanced training. And it gives EMS and ED clinicians a structured way to quantify what they are seeing in a child with acute neurological symptoms. We recently added the PedRACE to our platform’s Clinical Scores library so that clinicians can access it at the point of care, right alongside the tools they already use every day.

The sample size is small. The sensitivity and specificity for pediatric stroke screening have not yet been established. But this is the first prehospital tool specifically designed to detect large vessel occlusion in children. Its inclusion in the 2026 AHA/ASA guidelines signals that the field is ready to move beyond awareness and toward action.

And we already have proof that action matters. This study shows that EMS documentation influences ED decision-making. When EMS raises the possibility of stroke, hospitals respond. Clear documentation in the EMS report, direct verbal handoff highlighting stroke concern, prehospital notification for suspected large vessel occlusion when protocols support it. These are not expensive interventions. They are communication. And they change the trajectory of care.

One more finding worth highlighting. Adult stroke care uses “door-to-needle” and “door-to-puncture” times as quality metrics. Those don’t work for pediatric stroke. The authors propose “door-to-diagnosis” as the key metric for children: time from ED arrival to CT for suspected hemorrhagic stroke, and time from ED arrival to MRI for suspected ischemic stroke. This accounts for the reality that MRI is essential for pediatric stroke diagnosis, while CT suffices for adults. If we are going to improve outcomes, we need a way to measure progress. This is a good start.

EMS clinicians almost never suspect stroke in kids. But when they do, it changes everything that happens next.

The opportunity here is not perfect stroke recognition in the field. That is unrealistic given how rare pediatric stroke is and how often it mimics other conditions. The opportunity is making sure stroke stays on the differential when a child presents with acute neurological symptoms. A structured tool like the PedRACE. A clear EMS impression. A verbal handoff that names the concern.

In pediatric stroke, where median time to diagnosis is still approximately 24 hours from symptom onset, even small improvements save brain.

Sources:

Pearson R, et al. Striving toward quality metrics for pediatric stroke: time from door to diagnosis. Front. Stroke. 2025.

https://www.frontiersin.org/journals/stroke/articles/10.3389/fstro.2025.1718355/full

Turón-Viñas E, et al. Design and interrater reliability of the pediatric version of the RACE scale: PedRACE. Stroke. 2024;55:2240–2246.

https://doi.org/10.1161/STROKEAHA.124.046846

A 24-year-old woman is pinned in the driver’s seat of a rolled SUV on a rural highway. Her pulse is thready. Her systolic blood pressure is in the 70s. The flight crew is 18 minutes out. Your ground unit carries low-titer O-positive whole blood.

She’s bleeding to death.

You reach for the cooler. And then someone asks the question that has paralyzed EMS systems across the country:

“What if she’s RhD-negative?”

That question, asked at the wrong time, in the wrong way, has cost women their lives. Not because the concern isn’t valid. It is. But because the answer has been so poorly understood that clinicians have defaulted to hesitation when the science says act.

A landmark paper just published in Transfusion finally puts the best practices in writing.

The Paper

The THOR (Trauma, Hemostasis, and Oxygenation Research) Network, in collaboration with the Allo Hope Foundation and supported by the U.S. Department of Defense, has published a comprehensive best-practices document for managing life-threatening hemorrhage in females of childbearing potential (FCP). The paper, authored by leaders from Johns Hopkins, DC Fire and EMS, Harris County Emergency Services, the University of Pittsburgh, and several other institutions, was published in Transfusion in 2026 (Levy MJ, et al. Transfusion. 2026. doi:10.1111/trf.70119).

This isn’t a theoretical exercise. It’s an operational playbook.

The Problem We Created

Prehospital blood programs are expanding across the United States. By 2024, over 300 trauma centers had incorporated low-titer O-positive whole blood (LTOWB) into clinical practice. Roughly 60 to 75 percent of air medical transport systems now carry blood products, and a growing number of ground units are following.

Most of these programs carry O-positive products because O-negative supply is critically limited. Only about 3 to 4 percent of the U.S. population can donate low-titer O-negative blood. The math doesn’t work. The demand far exceeds the supply.

Here is where the tension lives. When an RhD-negative woman of childbearing potential receives RhD-positive blood, there is a risk of alloimmunization. Her immune system can develop antibodies against the RhD antigen, and those antibodies can cross the placenta in a future pregnancy and attack fetal red blood cells. The result is hemolytic disease of the fetus and newborn, or HDFN.

HDFN ranges from mild jaundice to severe fetal anemia, hydrops, and death.

That is a real risk. But here is what the fear-based narrative gets wrong: the probability of that outcome is far lower than most clinicians believe, and the systems we now have to manage it are far better than most clinicians realize.

The Math That Should Change the Conversation

The paper presents a Swiss cheese model illustrating the sequential events that must all occur for a case of severe HDFN to develop. The patient must be of childbearing age and potential. She must be RhD-negative. She must survive the trauma. She must become D-alloimmunized. The anti-D must reach high titer. She must become pregnant. The fetus must be RhD-positive. And even then, the fetus must develop a clinically significant adverse event from HDFN.

One model estimated one case of fetal death or permanent disability per 670 RhD-positive transfusions in RhD-negative FCP. That is roughly 0.15 percent. Another model estimated a 0.04 percent chance of perinatal death and a 0.24 percent chance of death or other serious adverse event.

Compare that to the near certainty of death from uncontrolled hemorrhagic shock without transfusion.

The risk-benefit calculation is not close.

The Equity Problem

Here is a finding from the paper that should make every EMS medical director uncomfortable: on adjusted analysis, females of childbearing potential were 40 percent less likely to receive LTOWB than males.

Forty percent.

That means women in hemorrhagic shock are being treated differently because of a theoretical future risk, while men with the same injury and the same physiology get the blood that could save them. The paper makes the point clearly: withholding prehospital blood products from women, or defaulting to scarce RhD-negative red-cell-only transfusions while men receive LTOWB, intentionally widens survival gaps.

We would never accept a protocol that said “withhold defibrillation from women in cardiac arrest because it might affect a future pregnancy.” Yet that is functionally what some systems are doing with blood.

What the Best Practices Say

The THOR Network’s guidance is clear, practical, and built for the field. The key principles center on a few critical themes.

First, life-threatening hemorrhage takes priority. The immediate risk of death from hemorrhagic shock outweighs the potential future risk of HDFN. Do not delay transfusion when RhD-negative blood is unavailable. Use RhD-positive LTOWB or RBCs to preserve life.

Second, communication is everything. EMS must document and communicate every RhD-positive transfusion to the receiving hospital. The handoff is a critical juncture. Hospitals need to know what was given so blood banks can act. The paper highlights the STRAC prehospital blood product transfusion record as an example of how to standardize this process.

Third, Rh immunoglobulin (RhIg) has a role, but a limited one. RhIg is recommended only after small-volume exposure of one unit or less to RhD-positive products in RhD-negative FCP. It is not recommended after large-volume resuscitation during exsanguinating hemorrhage. When indicated, it can be given up to 72 hours after exposure. There is no role for RhIg in males or postmenopausal females.

Fourth, follow-up and counseling are non-negotiable. All RhD-negative FCP who receive RhD-positive blood products should receive standardized informational materials before discharge, including clear documentation of what was transfused and why. Antibody screening should be scheduled between 14 weeks and 6 months post-exposure, and results should be shared with outpatient clinicians and maternal-fetal medicine specialists.

What This Means for Systems

This paper gives EMS and hospital leaders a framework to build around. It is not enough to carry blood on an ambulance or helicopter. Systems need protocols that address the entire continuum: field recognition, rapid initiation of balanced resuscitation, EMS-to-hospital handoff, post-exposure testing, counseling, and long-term follow-up.

The paper also points toward future directions that matter. Multicenter trials including MATIC-2 and TROOP are ongoing and will generate critical data on LTOWB outcomes, alloimmunization rates, and adverse events in both adult and pediatric populations. Registries need to track the cumulative burden of alloimmunization and the proportion of patients who develop clinically significant high-titer antibodies. Digital hemovigilance linking EMS records, hospital transfusion services, and regional blood banks can enable real-time detection, targeted counseling, and transparent reporting.

And on the horizon, the development of RBC antigen-agnostic blood products, including stem cell-derived and artificial red cells, could one day eliminate this dilemma entirely.

The Bottom Line

The science has caught up to the scenario every prehospital clinician has faced or will face. A woman is hemorrhaging. The blood on the shelf is O-positive. The question is no longer whether to give it.

The question is whether your system is ready for what comes after.

Give the blood. Save her life. Document it. Communicate it. Follow up.

That is the standard. And now we have the evidence-based framework to hold ourselves to it.

Citation: Levy MJ, O’Byrne H, Bank EA, Cannon JW, Cap AP, Glassberg E, et al. The THOR network best practices for females of childbearing potential with life-threatening hemorrhage: Guidance for emergency medical services and hospitals. Transfusion. 2026. doi:10.1111/trf.70119

Subscribe now

It’s 2 AM in a rural county. A two-year-old is seizing. The crew that rolls up hasn’t run a pediatric call in six weeks. The last child they treated was a broken arm.

Now they’re standing in a hallway with a 12-kilogram child turning blue, and they need to push a benzodiazepine. The dose is weight-based. The concentration varies by formulation. The math has to be right, right now, under the worst possible conditions.

This isn’t a training scenario. It’s Tuesday night in most of America.

For the first time, we have national data that tells us exactly how prepared EMS systems are for this moment. The answer should make every EMS leader pay attention.

The First National Baseline

This week, Annals of Emergency Medicine published the results of the National Prehospital Pediatric Readiness Project, the most comprehensive assessment of EMS pediatric capabilities ever conducted in the United States. Nearly 7,000 EMS and fire-rescue agencies participated, representing roughly half of all 911-response services in the country and covering more than 2 million pediatric emergencies per year.

The researchers, led by Dr. Kathleen Adelgais at the University of Colorado, evaluated agencies across eight domains: equipment, training, protocols, quality improvement, care coordination, family-centered care, safety policies, and more.

The median Pediatric Readiness Score: 65.5 out of 100.

Think about that number. If a student brought home a 65 on an exam, you wouldn’t call it a passing grade. And this isn’t a pop quiz. It’s our collective ability to keep children alive in the field.

To be fair, there were bright spots. Agencies scored well on equipment, safety policies, and clinical protocols. The tangible, checkable things. The gaps showed up in the areas that require sustained organizational commitment: quality improvement, hospital coordination, and family-centered care. In other words, we’ve bought the gear. We haven’t built the systems.

The Volume Problem Nobody Can Fix with Training Alone

Here’s the finding that should reshape how we think about pediatric readiness: three out of four EMS agencies see fewer than eight pediatric patients per month.

Eight patients. Per month. Across an entire agency.

That means the average paramedic might go weeks between pediatric calls. And within that already rare exposure, a truly critical pediatric emergency, like a cardiac arrest, a severe allergic reaction, or a status seizure, might come once or twice a year.

You can’t train your way out of that. You can run every simulation in the book, but when the real thing shows up at 2 AM and it’s been months since the last one, proficiency degrades. Memory fades. Confidence erodes. The data confirms it: agencies with higher pediatric call volumes scored higher on readiness. Experience matters, and most agencies don’t get enough of it.

This is the fundamental paradox of pediatric EMS. The patients who need us to be sharpest are the ones we see least.

The Single Variable That Changes Everything

If there’s one finding from this study that deserves a red circle around it, it’s this: only 38% of EMS agencies have a designated Pediatric Emergency Care Coordinator. But the agencies that do outperform their peers in every single category.

Not slightly better. Five times more likely to achieve top scores in quality improvement. Five times more likely to excel in health system coordination. Significantly stronger in family-centered care.

And here’s what makes this actionable: the advantage held true even for low-volume, low-resource agencies. You don’t need a big budget or a high call volume. You need someone whose job it is to care about this.

A Pediatric Emergency Care Coordinator keeps pediatric readiness on the agenda when the daily demands of running an agency push it to the margins. They drive training, track performance, coordinate with hospitals, and advocate for the systems that make the 2 AM seizure survivable. Without that role, pediatric readiness becomes everyone’s responsibility and nobody’s priority.

From Readiness Scores to Survival Rates

This isn’t an academic exercise. We already know what happens on the hospital side. Research published in Pediatrics has demonstrated that higher pediatric readiness in emergency departments correlates with significantly lower mortality in critically ill children. The next phase of research will examine whether the same relationship holds in the prehospital environment.

I’d bet heavily that it does. The principle is the same: when systems are designed to support clinicians during high-stakes, low-frequency events, outcomes improve. When they’re not, clinicians are left relying on memory and adrenaline. And that’s when errors happen.

Building Systems for the Calls That Rarely Come

Aviation doesn’t expect pilots to memorize every emergency procedure for rare mechanical failures. They build checklists, decision trees, and automated systems that activate precisely when the situation exceeds routine human performance. Pediatric EMS should work the same way.

Weight-based dosing errors remain one of the most common and preventable sources of harm in pediatric prehospital care. Every dose for every child requires a calculation that depends on weight, age, concentration, and clinical context. That’s not a task you want someone doing from memory at 2 AM after six weeks without a pediatric call.

This is where point-of-care technology becomes a force multiplier, particularly for the agencies this study identified as most vulnerable. When our platform puts age- and weight-specific protocols directly in a clinician’s hands at the point of care, dosing becomes automatic rather than a cognitive burden. It doesn’t replace clinical judgment. It supports it at the exact moment when the gap between experience and exposure is widest.

More importantly, this creates consistency. Every clinician, whether they’re a 20-year veteran or a newly certified paramedic, has access to the same evidence-based guidance when it matters most. That’s not a luxury. For low-volume agencies, it’s the infrastructure that makes readiness possible.

What Comes Next

This study gives us a mirror. For the first time, EMS leaders can see exactly where their pediatric systems stand, and where the cracks are.

The path forward isn’t complicated. Designate a Pediatric Emergency Care Coordinator. Build quality improvement processes that actually track pediatric outcomes. Invest in the technology that bridges the gap between rare exposure and reliable performance.

Because that crew at 2 AM, the one that hasn’t seen a sick kid in six weeks, they’re not going to get more pediatric calls. The volume isn’t changing. The only thing that changes is whether the system they work in was built to support them when the call finally comes.

A 65 out of 100 is where we are. It doesn’t have to be where we stay.

References

1. Adelgais KM, et al. The National Pediatric Prehospital Readiness Project: First Comprehensive Assessment of United States Emergency Medical Services Agencies. Annals of Emergency Medicine. 2026.

2. National study finds strengths, gaps in EMS pediatric readiness before children reach the hospital. University of Colorado Anschutz Medical Campus. February 26, 2026.

3. Remick K, et al. Pediatric readiness in emergency departments and mortality in critically ill children. Pediatrics. 2024;154(4):e2024066001.

4. Lammers R, Byrwa M, Fales W, Hale R. Simulation-based assessment of paramedic pediatric resuscitation skills. Prehosp Emerg Care. 2009;13(3):345-356.

Imagine you’re the charge nurse in the ED. It’s 2 a.m. A six-year-old arrives seizing. You need lorazepam and levetiracetam (Keppra) now. Normally, you pull up the order in Epic, and the system calculates the dose by weight, flags the range, and walks you through administration. Tonight, Epic is down. The pharmacy system is down. The dose-range checker is down. You have a calculator, a paper formulary, and a child who is still seizing.

This is not a hypothetical. This is Mississippi right now.

On Thursday, the University of Mississippi Medical Center, Mississippi’s only Level 1 trauma center and only children’s hospital, was hit with a ransomware attack that took down Epic, their phone system, and their entire IT network. All 35 clinics across the state shut down. Chemotherapy was canceled. Elective surgeries halted. The FBI is investigating.

The hospital stayed open. The EDs kept running. But every electronic safeguard for medication safety disappeared overnight.

Inside that building right now:

• Paper trauma admission order sets: 10-page handwritten forms replacing digital workflows

• Walkie-talkies with a 16-channel assignment chart taped to the wall, substituting for the phone and paging system

• A printed sheet of surgeons’ personal cell phone numbers, because there is no other way to reach them

• Rows of two-way radio charging stations staged in conference rooms throughout the building

This is a major academic medical center operating without any of the technology it was built to depend on. And the most dangerous part is not the disruption to scheduling or billing. It is what is happening at the bedside.

The Medication Safety Crisis Nobody Is Talking About

When ransomware attacks on hospitals make the news, the coverage centers on data breaches, financial losses, and operational chaos. What rarely gets discussed is the patient safety crisis embedded in every one of these events: when your EHR goes down, your entire medication safety infrastructure goes with it.

Modern hospitals rely on electronic systems for weight-based dose calculations, dose-range checking that flags dangerous orders before they reach the patient, allergy and interaction cross-referencing, automated dispensing tied to verified orders, and barcode medication administration at the bedside. All of it vanishes in a ransomware attack. Every single layer.

What’s left is a physician doing mental math, writing a dose on paper, and a nurse trying to verify it manually under the most operationally stressful conditions imaginable.

A 2025 JAMA study found a 21% increase in in-hospital mortality for time-sensitive conditions during ransomware-related system downtime [1]. In the past year, 72% of healthcare organizations reported that at least one cyber incident disrupted patient care [2]. And for one in four affected organizations, recovery stretches beyond a month, meaning every medication dose across that entire period is being calculated and administered without the electronic safety net [3].

But perhaps the most sobering data point concerns hospitals that were not even attacked. A 2024 peer-reviewed study published in Critical Care Explorations examined what happened at neighboring hospitals during a month-long ransomware event at a regional health system. Cardiac arrests at those adjacent, untargeted facilities nearly doubled. And survival with favorable neurologic outcome for out-of-hospital cardiac arrest patients collapsed from 40% before the attack to just 4.5% during it [4]. Those patients never set foot in the hospital that was hacked. They went to the one down the street. The attack found them anyway.

Every Patient. Every Dose. Every Time.

Weight-based dosing errors are not a pediatric problem. They are a patient safety problem. Every high-acuity patient, whether a 12-kilogram child receiving lorazepam and levetiracetam for a seizure or a 90-kilogram adult in septic shock on vasopressors, requires precise, individualized dosing. A miscalculated dose, a trailing zero misread, a concentration confused: the consequences are the same. Immediate. Severe. Potentially fatal.

In a functioning hospital, these errors are caught by dose-range alerts, auto-calculations, and pharmacy verification. During a ransomware downtime, none of those catches exist. Every medication order is being calculated by hand, verified by memory, and administered on trust.

Right now, that is the reality for every patient at UMMC. That is not a backup system. That is an exposure.

What Hospital Leaders Must Do Before the Next Attack

Most hospital downtime plans are written to preserve operations. Few are written to preserve safety. Those are not the same thing. Keeping the ED open during a ransomware attack is an operational achievement. Ensuring that every patient in that ED receives a correctly calculated, safely verified medication dose during that same event is a safety achievement, and it requires a completely different kind of preparation.

The gap between those two things is where patients get hurt.

Hospital leaders need to ask hard questions of their current downtime protocols: Does your plan include weight-based dosing references that providers can actually use at the bedside without logging into anything? Have you drilled medication safety specifically, not just communications and logistics? Do your providers know what they would reach for if Epic disappeared tonight?

This is precisely the problem Handtevy was built to solve. Deployed in more than 100 hospitals today, Handtevy provides real-time, weight-based dosing and dose-range safety checking across the full spectrum of patients, from the smallest pediatric case to the highest-acuity adult. It includes a dedicated contacts section where critical phone numbers are always current and accessible, so staff are never hunting for a surgeon’s cell number or taping a channel chart to a wall. Any update across the platform, whether to a medication protocol, equipment reference, or contact, is pushed instantly across the entire system so every user is always working from the same current information. And the built-in notification center gives hospital leaders a direct channel to keep clinical staff informed in real time, ensuring critical updates reach the right people instantly.

Critically, all of it operates completely independently of the EMR and internal hospital networks. When Epic goes down, Handtevy keeps working. When the pharmacy system is unreachable, Handtevy keeps working. It was never built to depend on your IT infrastructure, which means it does not fail when your IT infrastructure does.

That independence is not just a feature. In a ransomware event, it is the difference between having a safety net and not having one.

Ransomware attacks on hospitals are no longer rare events. They are predictable, recurring, and increasing in sophistication. The question is no longer whether your organization will face a period of full electronic downtime. It is whether your medication safety systems are robust enough to survive it.

UMMC is navigating that reality right now, with real patients, in real time. Every hospital in the country should be watching and asking whether they are genuinely ready, or simply assuming they are.

Because the next attack won’t give you time to find out.

References

1. Neprash HT, et al. Trends in Ransomware Attacks on US Hospitals, Clinics, and Other Health Care Delivery Organizations, 2016–2021. JAMA Health Forum. 2022;3(12):e224873. doi:10.1001/jamahealthforum.2022.4873

2. Cybersecurity and Infrastructure Security Agency (CISA). Healthcare and Public Health Sector Cybersecurity Report. 2024.

3. American Hospital Association. Cybersecurity and the Healthcare Sector: 2024 Report. AHA; 2024.

4. Pham TT, Loo TM, Malhotra A, et al. Ransomware Cyberattack Associated With Cardiac Arrest Incidence and Outcomes at Untargeted, Adjacent Hospitals. Critical Care Explorations. 2024;6(4):e1079. doi:10.1097/CCE.0000000000001079