Four Patients Got the Wrong Drug in Nashville. Ask a Better Question.

If you like the post, please share!

On August 14, four patients went to Ascension Saint Thomas Hospital Midtown in Nashville for joint replacement surgery. All four were harmed during the procedure. The hospital has not said how. One family says the drug was potassium chloride instead of the anesthetic. That patient is paralyzed from the chest down.

[Corrected: the hospital later confirmed potassium phosphate and mepivacaine. See the update below.]

Why it matters: Spinal anesthesia for a knee or hip replacement is among the most routine things a hospital does. People go in expecting to walk out. Nothing about this failure was exotic, and nothing about it was unique to this hospital: a nearly identical error has been documented repeatedly with a different drug, and the warnings issued about it have not stopped it from recurring.

Driving the news:

  • Ascension Saint Thomas says four patients were harmed during an event on August 14, that it self-reported to state regulators the same day, and that it has identified the cause and put corrective safeguards in place.
  • The family of Glenda Dorton, 72, of Centerville, Tennessee, told local reporters she was supposed to receive bupivacaine and received potassium instead. Her daughter-in-law, Kristina Buell, says Glenda has no feeling or movement below her breastbone.
  • NewsChannel 5 reports that two of the four patients are paralyzed.
  • The Tennessee Health Facilities Commission has investigators on site. It notified the Tennessee Bureau of Investigation on Friday evening, and the TBI says its investigation is active and ongoing.
  • Metro Police and the Davidson County District Attorney's Office both said they have nothing open at this point, and there are no current indications the event was intentional.

Update, August 21: NewsChannel 5 Investigates reported that the mix-up happened in the pharmacy, where the syringes used for these procedures were filled with potassium instead of bupivacaine. The station attributed that to what it had been told, not to a hospital confirmation. Ascension confirmed it on August 21. It answers the question I raise below about a shared upstream source, and it moves the focus from the operating room to the pharmacy's fill and check steps.

What we don't know: whether the other three patients received the same drug, what quantity was given, what product and what container were involved, and what happened in the fill and check steps that put the wrong drug in a syringe.

[Update: the state report answers all of this except the quantity. All four received potassium phosphate, from vials pulled in the pharmacy, and the fill and check steps are described in detail at the bottom of the post.]

One discrepancy worth flagging: some outlets describe a spinal injection for anesthesia and at least one describes an epidural. Those are different procedures. I have not seen the hospital confirm either.

[Update: the state report describes syringes of a medication intended for use as a nerve block, incorrectly prepared for intrathecal spinal administration. So the route was intrathecal.]

The number that matters is four

One patient getting the wrong drug is an error. Four patients, on one day, in one service line, is a different animal.

Four clinicians independently making the same mistake on the same morning is not a plausible story. Four patients reached through a shared upstream source is very plausible: a bin or tray stocked with the wrong product, a batch drawn from the wrong container, a delivery placed where the anesthetic belongs.

That was inference when I first wrote this. The state report confirms it. Five syringes were filled from the wrong vials in a single preparation session the day before surgery. That is the finding that should drive what other hospitals do this week, because it points at stocking, supply, and preparation rather than at individual attention, or blame.

One sentence in the hospital's statement deserves a closer look. Leaders said they identified the cause within days. Identifying a cause within days is not the same as completing a root cause analysis. Finding the last person to touch the syringe would not explain the conditions that made four patients reachable through a single point of failure.

They may have gone much deeper. Publishing what they found would show us.

This is not a look-alike, sound-alike name problem

“Potassium phosphate” and “mepivacaine” share no letters, no syllables, no stem. Tall man lettering does nothing here. ISMP's look-alike name list does nothing here. None of the safeguards built around look-alike and sound-alike names addresses this failure mode.

What we are looking at is a container problem. Two clear, colorless solutions in small glass containers, sitting somewhere in physical reach of each other.

The packaging matters here, though not the way I first wrote it. When I thought this was potassium chloride, the argument was straightforward. After the deaths of the 1990s, USP required the cap and the overseal on potassium chloride concentrate to be black and to read Must Be Diluted, so I could say the warning existed in the design, in black, and still did not work.

Potassium phosphate is a different product. It is on ISMP's high-alert medication list, but that black closure convention was written for potassium chloride, not for this. What the vial does carry, in capitals on the label, is CAUTION: MUST BE DILUTED and For intravenous use only.

So the warning was there. It was printed on a vial that, as the state report describes, would not scan and went into a transfer bin anyway. That is the argument for why labels and warnings sit near the bottom of the hierarchy of controls. Text on a vial only works on somebody who is reading the vial.

It is also entirely ordinary for potassium phosphate to be in a hospital pharmacy, because it goes into parenteral nutrition and electrolyte repletion every day. So the question I asked, what is this drug doing near a neuraxial setup, has a different answer in the IV room than it would in an operating room. What has to be engineered is not its absence but its separation from the workflow that fills spinal syringes.

The same error, with a different drug

There is a recurring wrong-drug, wrong-route error in anesthesia where tranexamic acid gets injected intrathecally instead of bupivacaine.

The FDA alerted clinicians in December 2020, then said again in February 2025 that reports keep arriving, with outcomes including death, disability, and prolonged hospitalization. APSF puts mortality around 50 percent. ISMP documented cases in a knee surgery, a hip replacement, and a bilateral knee replacement, each involving a container mix-up.

Every one of those was a joint replacement. That is not a coincidence. Tranexamic acid is routinely used in joint replacement, so it is physically present in the same room as the spinal tray.

That raises a question about Nashville. With tranexamic acid, the wrong drug is in the room for an obvious reason, and ISMP notes those errors usually trace to storage and to look-alike vials sitting near anesthetics. Potassium phosphate was in the pharmacy for an obvious reason too. That makes this less exotic than I first thought and more transferable, because the drug that harmed these four patients is sitting in every hospital pharmacy in the country, exactly where it is supposed to be.

The FDA alert is also a reminder that an alert is not a control. The notification system fired correctly. It fired twice. Patients kept getting hurt for more than four years.

The investigation may be justified. The chilling effect is still real.

The TBI opening a file here is defensible. Four patients, one day, one wrong drug is exactly the fact pattern where you have to rule out a deliberate act. Tampering and diversion are real. An investigator who did not look would be doing the job badly.

So I am not going to write “they are criminalizing error again.” That argument collapses if nobody is ever charged.

The problem is what Tennessee taught everyone with the RaDonda Vaught prosecution. ISMP called that verdict a threat to patient safety. The possibility of criminal charges can change what people are willing to say long before anybody is actually charged.

Yes, but: compared to Vanderbilt in 2017, this hospital has behaved differently, and I will say so. A named CEO apologized in writing within four days and used the words harm and sorry, without the passive voice and without “if anyone was affected.” Leaders met with each family. The hospital reported to the state the day it happened. In the Vaught case, the initial report did not disclose the error and the cause of death was recorded as natural.

Credit the speed and the naming. Same-day reporting to regulators is largely a compliance obligation, so credit that less. And none of it extends to the silence about mechanism, which is the part that would prevent the next one.

“Patient privacy” is doing three different jobs

When a hospital says it cannot discuss the details, three separate things get bundled into one word.

The system description. Which drug, which storage location, which preparation step, what changed afterward. None of that requires identifying a patient or disclosing their protected health information. “Patient privacy” does not explain why those system details must remain secret.

The individual patient's clinical detail. This is real PHI. It belongs to the patient, not to the hospital. Which is what makes the Nashville situation strange: the Dorton family has gone public, by name, in detail, on television. A signed authorization would remove the patient-privacy barrier entirely. Hospitals almost never ask for one, even when the patient or family is already discussing the case publicly.

Reputation and liability. These concerns can get bundled into “privacy,” too: peer review privilege, brand, market position, and the reflex that saying less is safer.

That asymmetry is hard to defend. The same organizations that cite privacy to withhold will release clinical detail when it helps them, usually to suggest the patient was sicker or an outlier. Privacy turns out to be waivable when it is exculpatory.

Better than “could this happen here”: show me that it cannot

This is the part I actually care about, and none of it requires the investigation to finish.

Walk the physical space. The block cart, the anesthesia workstation drawers, the automated dispensing cabinet in the OR, and any tray or syringe that arrives pre-filled from pharmacy. Ask what concentrated electrolytes are physically reachable from where neuraxial drugs get drawn.

Then change the question you ask. “Could this happen here?” invites a reassuring answer, and you will get one.

Ask instead: show me that it cannot.

That requires evidence, not reassurance. It is checkable in an afternoon, and the answer is either a demonstration or a project.

The durable controls, strongest first:

  • Remove concentrated potassium from procedural areas entirely. This was the answer in the 1990s. Verify it rather than assuming it.
  • Move to manufacturer-prepared, ready-to-administer neuraxial anesthetics so nobody is drawing from a vial at the point of care. Note the limit, which I discuss in the update below: for mepivacaine spinal use there is no such product, because the route is off-label.
  • Close the barcode gap. Anesthesia drug preparation is still the largest area in most hospitals where a clinician selects, draws, labels, and gives a drug without a scan.
  • Independent verification before neuraxial injection of any high-alert drug, with the understanding that a second check catches less than people believe. In the Cardiff insulin case I wrote about last week, the second nurse did her check correctly and the patient still received the wrong drug.

One precision point, because it is about to come up in every discussion of this case. NRFit, the ISO 80369-6 neuraxial connector, is a genuine forcing function and I am in favor of it. But it prevents wrong-route misconnection, like IV tubing joined to a spinal needle. It does not stop anyone from filling a neuraxial syringe with the wrong drug out of a vial, because vials do not have connectors. That is true whether the filling happens at the bedside or in the pharmacy. Nashville was a container selection error. NRFit would not have caught it.

The ask

Ascension operates a large multi-state system. It could push what it learned to every one of its own hospitals within 24 hours, and publish it externally within weeks, without naming a single patient.

That is the concrete thing to ask for. Not more transparency in general. A specific finding, distributed fast, in a form an OR leader can act on.

If this post reaches more operating room leaders this week than the formal safety infrastructure does, that is not a compliment to the blog.

Go deeper

Update, August 28: The State Report Is Out

The Tennessee Health Facilities Commission has released an 80-page survey report on the August 14 event, and it answers most of what I was asking a week ago.

Two corrections first. The drug was potassium phosphate, not potassium chloride. Early family accounts and local coverage described potassium chloride going in instead of bupivacaine, and the hospital later confirmed potassium phosphate and mepivacaine. Second, I aimed the access question at the wrong room. I suggested walking to the block cart and the anesthesia drawers and asking what concentrated electrolytes are physically reachable from where spinal drugs get prepared. That question still matters. It just was not where this one happened.

According to WKRN's account of the report, five syringes were prepared the day before the surgeries. A pharmacy technician retrieved a vial of mepivacaine 2% at 10:14 that morning. Twenty minutes later, needing a second vial, the technician went to a different storage bin and took potassium phosphate. The potassium would not scan. The technician put the mepivacaine vial back and got a second vial of potassium. At 10:37, both potassium vials went into a transfer bin along with unopened syringes and six labels reading Mepivacaine 2%.

Read that sequence twice. The barcode system worked. It refused the drug. What happened next was not a missed alert, it was a response to one: return the correct vial, go get more of the drug the system had just rejected.

Investigators found the pharmacy system let technicians override barcode scanning and enter medication information manually, without pharmacist verification. Three technicians and one pharmacist handled the preparation, verification, and dispensing steps, and none of them caught it. A technician involved had been given a coaching feedback form that same morning, citing difficulty with the IV-room workflow, medication calculations, and a continued need for supervision.

So we have:

  • A system that can be overridden
  • Performance and training concerns
  • Four sets of eyes in the verification and dispensing chain, none of which caught it

These are management problems, not worker problems. Blaming a worker won't fix the system.

So the scanner was there. It fired. It could be bypassed. I wrote last week that an alert is not a control. A scan you are allowed to override is not a control either, and a scan that gets answered by fetching a second vial of the rejected drug is not even an alert. It is a delay.

Why anyone was filling those syringes at all

Go one level further back and ask why a person was drawing mepivacaine into syringes in a pharmacy the day before surgery. Why the batch?

The answer is on the label. Carbocaine's FDA labeling indicates mepivacaine for local infiltration, peripheral nerve block, and central neural techniques including epidural and caudal blocks. Then it says plainly that it is not intended for spinal anesthesia. The same label lists unintended intrathecal injection among the acute emergencies to be managed. Mepivacaine is nonetheless used intrathecally in the United States, off-label, routinely.

Why?

Hospitals do that for a reason that holds up. Bupivacaine lasts too long for a patient you intend to send home the same afternoon. A large arthroplasty series found mepivacaine associated with shorter length of stay, more same-day and home discharges, and earlier ambulation.

If your joint replacement program is built on getting people up and out, the short-acting spinal is what makes it work.

But because the route is off-label, no manufacturer sells a ready-to-administer spinal presentation of it. There is no kit. Somebody has to fill the syringe.

That leaves two options and neither is comfortable. Draw it at the point of care, which means a beyond-use date measured in hours, no cleanroom, and an anesthesiologist doing sterile preparation next to a block cart. Or batch it in the pharmacy under a hood, where the sterility is better and the dating lets you work ahead of the schedule and keep it for days.

I want to be careful here, because the easy conclusion is that batching caused this and batching is always bad. It did not and it is not. Compounding in a cleanroom is the better answer to the contamination problem, and the report describes a technician who scanned the vials, entered the lot number, and photographed the medication. That is a more controlled process than most anesthesiologists get on a Friday morning.

What batching does is move the moment of truth. It puts the only person who will ever see that vial twenty-four hours and one building away from the patient. Everything downstream inherits a label.

So the question is not whether to batch. It is why a drug used this way, in this volume, in joint replacement programs across the country, still has to be moved from a vial into a syringe by hand at all. That fix is not available to any single hospital. It is an approval, a manufacturer, and a kit.

The report also covers what happened afterward. Regulators cited three areas at Immediate Jeopardy: the pharmacy, surgery services, and the governing body. That third one is the state saying this was not local to the IV room. The hospital submitted a corrective plan on August 18 that was deemed unacceptable, another the next day that was also unacceptable, and a third that was rejected for lacking evidence that anything had actually been done. When surveyors left, nothing had been accepted. They described the combination of findings as a “culture of overall noncompliance.”

The surgery side was cited too, for what happened after. Investigators found the hospital failed to implement a timely and effective treatment plan, which they tied to progressive neurological, respiratory, and hemodynamic decline in all four patients. The report cites published guidance on managing a wrong-drug spinal injection, where prompt CSF lavage and cardiopulmonary support are the mainstays, that the clinicians did not act on quickly. One patient's family member says a staff member told him they did not know how to handle this.

That is a second failure, separate from the pharmacy, and in some ways a harder one. Nobody had rehearsed the response to a known catastrophic event.

Regulators also faulted the hospital for not immediately removing everyone directly involved from clinical duties, and specifically noted that an involved anesthesiologist was not placed on administrative leave right away.

Worth separating two things that get bundled here. Removing someone from clinical duty during an active investigation is not punishment, and just culture has never said otherwise. It protects patients while the facts are unclear, and it protects the individual from working under suspicion. The state was citing an absence of that. Whether it becomes punishment depends on what happens next.

And what happened next, in this case, is its own finding. Pharmacist #1 was placed on leave on August 14. In his interview with surveyors he said nobody from the hospital followed up with him afterward. He was told to stay home and offered the employee assistance program. He said he did not know whether he needed an attorney before speaking with the survey team, because nobody at the facility had reached out to him. Meanwhile Pharmacy Technician #3, also involved, kept working that day.

So the hospital both removed someone and failed to remove someone, and the one it removed was left alone with it. That is not a just culture problem or a punishment problem. It is the absence of any worked-out approach to either, which is usually what sits underneath both.

I gave this hospital credit earlier in the post for apologizing quickly and by name. I still do. It does not extend to this.

The anesthesiologist finding is worth reading next to the joint statement from the American Society of Health-System Pharmacists and the American Pharmacists Association opposing punitive responses to medication errors. In the same week, about the same event, a state regulator cited a hospital for not removing an individual fast enough, and two national professional bodies argued that individual punishment is the wrong response. I do not think that tension resolves cleanly, and I would rather name it than pretend one side is obviously right.

Here's a news story with a summary and a pharmacist talking about the system error dynamics:

The walk I suggested still holds. It needs a second stop. Go to the IV room. Watch a technician compound a neuraxial syringe. Ask what happens when a scan does not match, who is permitted to override it, how often that happens, and who looks at those overrides afterward. Then ask what the pharmacist's final check actually consists of, whether it is a look at the syringe or a look at the vial that filled it, and what the technician is expected to do when a vial will not scan.

One more thing in the timeline, and it cuts both ways.

The hospital says its leaders ordered an emergency halt to all remaining operations scheduled for August 14 after they noticed the adverse reactions in the four patients. Ascension has since clarified that seven patients were on the joint replacement schedule that day and that four of them received the wrong medication. The report says five syringes were filled with the wrong drug and four were used. The line stopped between the fourth syringe and the fifth.

So somebody stopped the line. I do not want to skip past that. Plenty of organizations would have finished the schedule and sorted it out on Monday.

It also stopped after the fourth patient rather than the first. Three of the seven patients on that schedule were not harmed, and at least one of them had a syringe already filled and waiting.

The trigger, by the hospital's own description, was noticing adverse reactions. That is a stop driven by accumulated harm, not by a signal. I had wondered whether the first patient's reaction was even recognizable in time to protect the second. The report suggests it was. The potassium began attacking the spinal cords almost immediately, and at least one patient described burning and excruciating pain. So the signal was there from the first case. The stop still took four.

Look also at what Ascension added afterward: a mandatory hard stop and escalation for any spinal medication scan alert. A hard stop is only worth adding where a soft one already existed. Their own countermeasure describes the prior state better than any summary I could write.

The andon question is not really whether people are willing to pull the cord. It is what the system requires as evidence before a pull counts as legitimate. Four patients is a lot of evidence.

One more finding, and it is the one I would put in front of any executive who reads this. Ascension announced that high-alert medications including mepivacaine would be stored in separate, distinctly marked locations. Regulators returned to that pharmacy two days later, and again six days later. Both times, the potassium was still on the wall with the other medications, unidentified. An announced countermeasure and an implemented countermeasure are not the same thing, and this is what the gap between them looks like with surveyors standing in the room. Which is the whole reason to ask leaders to show you rather than tell you.

The report is said to name human error as the primary cause. That is true, and it is where the work starts rather than where it ends, which is a big enough argument that I gave it its own post.

Update, September 5: The Pharmacy Director Already Knew

WPLN's Catherine Sweeney published a detailed read of the state report on September 3. It fills in what the earlier TV coverage summarized, and a few pieces of it change how I read this case.

Start with the sentence I did not expect to find.

“The Director of Pharmacy stated that this had been a systemic problem in the pharmacy verification process.”

Read that as it is written. The person accountable for the pharmacy told investigators the verification weakness was systemic. Not a lapse, not one bad shift.

I want to be careful about what that quote settles. The report is paraphrasing an interview, and “had been” could mean he knew before August 14, or it could mean he concluded it afterward while looking back. Those are different findings and I have not seen the underlying report language that would separate them. Either version is bad, but they are bad in different ways. If it was known beforehand, then the question is not why the check failed. It is what happened to the concern, who heard it, and what got prioritized instead.

Somebody should ask that question in a deposition, and I hope somebody does.

The verification step could not have caught it

The report describes what the final pharmacist check actually consisted of, and it is worth walking through slowly.

A technician filled the five syringes with potassium phosphate, photographing each step. That technician then put the vials in the discard bin and brought the syringes to the pharmacist. The pharmacist reviewed the photos and records for 29 seconds, signed off, and applied labels reading mepivacaine.

So the vials were gone before the check happened. By design, the verifying pharmacist was looking at pictures of a vial rather than at a vial. That is a real limitation of remote or image-based verification, and it is common. It is not unique to this hospital. Why would you design the process that way?

The report also notes the order called for six syringes and five were made. The pharmacist did not notice. I mean that part could be inspected / counted.

I am not interested in the 29 seconds as an indictment of one person. I am interested in what 29 seconds tells you about how much work that step was designed to do. If the last defense before a spinal injection is a half-minute photo review by someone who cannot see the container, then it was never a barrier. It was a signature.

And that is the part the pharmacy director apparently already understood.

The report says the mix-up was foreseeable

According to WPLN's account, the report describes the mepivacaine and potassium phosphate vials as clear liquid in small glass vials, white labels, black lettering, yellow rectangles. Then it says the confusion was understandable and even foreseeable.

That language matters. A state surveyor writing that a wrong-vial selection was foreseeable is describing a design condition, not a character flaw. It sits oddly next to the same report naming human error as the primary cause, which I wrote about separately.

It also confirms the container argument I made earlier and undercuts the look-alike-name framing. Nothing about the words helps. Everything about the glass, the label, and the shelf does.

If it was forseeable, why wasn't it prevented?

The containment response is the harder failure

I wrote earlier that the delayed treatment was a second failure and in some ways a worse one. The detail makes that stronger.

The first patient complained of pain loudly enough that the procedure was canceled. Shortly after, he said he could not feel or move his legs. Around 8 a.m., per the report, the anesthesiologist said they were not concerned at that time and declined support from a hospitalist. By 9:45 the patient had no sensation from the abdomen down. At 11 he was intubated and taken to the ICU. As of investigators' last check on August 21, he was still on life support.

Another patient described being taken to the ICU without an explanation, and said staff appeared not to know what to do and stood around deciding what to do next. She told investigators the chief medical officer and a neurologist said they had found a treatment option online and were going to carry it out.

I do not want to turn that last detail into a joke, because I think it is the most useful thing in the report. Nobody looks something up mid-crisis because they are careless. They look it up because the response was never designed, rehearsed, or written down anywhere they could reach. There is published guidance on managing an accidental intrathecal injection. The FDA and APSF have both written about this exact class of event. It just was not in that building, in a form anybody could act on, at 9 a.m. on a Friday.

All four eventually got a spinal lavage (a washout of the spinal fluid), hours after the injections, after being moved by ambulance to a second hospital. Regulators cited the delay in assessment, reassessment, intervention, coordination, and treatment. As of August 21, none of the four had regained feeling below the waist.

If your organization has a low-frequency, high-consequence event that would require an unfamiliar procedure under time pressure, ask when it was last practiced. Not whether there is a policy. When someone last did it.

Revising what I said about the stop

I wrote that the first patient's reaction was recognizable in time to protect the second. I want to hold that more loosely now.

The second patient entered the operating room just before 8 a.m., which is roughly when the anesthesiologist was saying he was not concerned about the first. Those may have been close to simultaneous. So the window to protect patient two was narrow, possibly nonexistent.

Patients three and four are a different matter. By then, a case had been canceled, a man could not move his legs, and there were still syringes on the schedule.

Nonconformance is a management word

The report cited three Immediate Jeopardies, including the governing body. It described a culture of overall nonconformance. Three corrective action plans were rejected. On August 20, six days after four people were paralyzed, surveyors found the potassium phosphate still on the pharmacy wall with the other vials, with no high-alert label and no visual cue.

Why??? Beyond not being proactive, that's not even reactive.

Ascension has since announced isolated storage for high-alert medications, a mandatory hard stop with escalation for any spinal medication scan alert, and independent dual pharmacist verification.

Those are reasonable countermeasures. Every one of them also describes what was missing.

Meanwhile, the TBI is still assessing, and WPLN reports it is unclear whether criminal charges will follow.

Here is what I would ask anyone watching this. The state's own finding points at governance, at the verification process, at storage, and at a corrective plan the hospital could not get accepted while surveyors were standing there. That is a set of decisions made above the IV room.

I am not aware of a prosecution in American healthcare that has landed there. In Tennessee, five years ago, it landed on the nurse holding the syringe. If a case gets built here, who do you expect to be sitting at the defendant's table, and what would that choice say the problem was?

Go Deeper:

Get New Posts Sent To You

Select list(s):
Tags
If you like the post, please share!
Mark Graban
Mark Graban

Mark Graban is an internationally-recognized consultant, author, and professional speaker, and podcaster with experience in healthcare, manufacturing, and startups.

Mark's latest book is The Mistakes That Make Us: Cultivating a Culture of Learning and Innovation, a recipient of the Shingo Publication Award.

He is also the author of Measures of Success: React Less, Lead Better, Improve More, Lean Hospitals and Healthcare Kaizen, and the anthology Practicing Lean.

Mark is also a Senior Advisor to the technology company KaiNexus.

Articles: 5921

Related Posts