By Jonathan Pabalate, DNP, CRNA, APRN | Founder, JPCAIC — JPC Anesthesia Informatics Corp | Nurse Anesthesia Faculty, University of North Florida
There is an exchange I have with every resident on our first day together, and it is worthless.
“Did you check the machine?”
“Yes.”
I ask it. They answer it. Nothing has been established and we both know it. What I want to know is whether they could tell me what the low pressure leak check protects against, and whether they would notice the moment it stopped protecting them. The question I asked cannot reach that. So I ask a question I already know the answer to, I get the answer I expected, and we go to work.
I have been thinking about why I keep asking it anyway.
The check was performed. The patient still could not be ventilated.
Years ago, at a hospital I will not name, I was in the building on a day that still organizes how I teach.
A trainee ran an automated low pressure leak check on a machine that runs the whole sequence for you. You press a button, the machine works through its own tests, and it tells you when it is done. That particular test required an occlusion plug seated in the inspiratory limb. The plug went in. The sequence started. The trainee stepped away to draw up medications.
The sequence finished. The screen asked for the next button. Nobody pressed it.
The patient was induced. Ventilation felt tight, and the team worked the algorithm the way you would want a team to work it. Laryngospasm first, so paralysis. Still tight. Bronchospasm next, so deepen the anesthetic. Still tight. Then intubation, then reintubation by more experienced hands, then a code.
About ten minutes in, someone walking past heard the noise and opened the door. She saw a bright orange plug on the floor, next to a disconnected inspiratory limb.
The patient did not survive.
I have changed or removed every detail that could identify anyone involved, and I am not going to describe the people in that room. The mechanism is the part that belongs to all of us. Here is how I say it to students:
It wasn’t because the automated check failed. In fact, the automated check did exactly what it was supposed to do. It was stuck on the screen that said, push the next button.
The sentence I hear most is the one I trust least.
The machine checks itself now.
I hear that from students, and I hear it from people two decades into practice. It is not a stupid thing to believe. It is close enough to true that it survives most days, which is exactly what makes it dangerous.
The 1993 FDA Anesthesia Apparatus Checkout Recommendations were a task list. Fourteen numbered items, each with lettered substeps, telling you not only what to verify but how to verify it: attach the suction bulb to the common gas outlet, squeeze until fully collapsed, confirm it stays collapsed for at least ten seconds. It read like a procedure because it was one.
In 2008 the American Society of Anesthesiologists replaced it. Jeffrey Feldman, Michael Olympio, Donald Martin, and Adam Striker, writing in the APSF Newsletter as members of the task force that drafted the replacement, gave two reasons. The first was that machines had diverged too far for one procedure to fit them all. The second is the one worth sitting with, and it is their sentence, not mine: of the 1993 checkout, “available evidence suggests that it is not well understood and not reliably utilized by anesthesia providers.”
The same article names the limit of the automation directly. Automated checkout procedures “do not check all of the items that require attention, and vary from machine to machine.”
So the task force built a list of fifteen functions that must be verified and left the how to each machine and each department. That was a defensible engineering decision. It also quietly moved the burden of knowing what a check is for from the document to the clinician, at the same moment the machine started performing the steps on the clinician’s behalf.
One more thing, and it matters more for educators than for anyone else. Go looking for that 2008 document today and you will find it on the ASA site under Resources from ASA Committees, marked member exclusive, listed without a year, under a header that reads: “The following work products and resources have been made available by ASA committees. They have not been approved by ASA’s Board of Directors or House of Delegates and do not represent an ASA Policy, Statement or Guideline.”
The most cited authority in how we teach the pre-use checkout is a member-gated committee work product that disclaims being a guideline. I am not arguing it is wrong. I am arguing that most of us teach it as though it were something it does not claim to be.
Completion was never the thing that was failing.
In 1991, March and Crowley put 188 anesthesiologists in front of a machine with four planted faults. Using their own checkout methods, they found 25.8 percent of them. Handed the FDA checklist, they found 29.9 percent. Their conclusion was blunt: “the mere introduction of the FDA checklist did not improve the ability of anesthesiologists to detect anesthesia machine faults.”
Sixteen years later, Larson and colleagues at Mayo ran 87 providers through five planted faults at a national meeting. Providers with zero to two years of experience found an average of 3.7. Two to seven years, 3.6. More than seven years, 2.3. That is a significant inverse relationship between experience and detection, P less than 0.001.
Read that again, because it is the finding that should reorganize how we teach this. The more anesthesia you have done, the fewer machine faults you found.
Then there is the Israeli board examination, reported by Ben-Menachem and colleagues in 2011, which is the closest thing we have to my thesis measured directly. Nearly every examinee cleared the completion bar on the checkout list. Fourteen of 31 could not explain how to minimize oxygen use during a pipeline failure. Fifteen of 30 could not give the correct differential diagnosis for the failure in front of them. The authors’ summary is that most senior residents recognized the equipment failures and many could not correctly diagnose and manage them.
They did the check. They could not read the machine.
What the automated check cannot see.
The specific blind spots are documented, and they are not obscure.
Michael Dosch tested three current workstations in 2014 by occluding the inspiratory and expiratory limbs and then running each machine’s own automated checkout. On one of the three, the Aisys, the operator was permitted to accept both faults and proceed into simulated patient care. The self-test found the problem, offered it to the human, and the human was allowed to wave it through.
In 2015, Charous and colleagues reported a machine with a small cut in the APL valve bypass tubing that passed the ventilator leak test, the system leak test, the compliance test, and the safety relief valves test, then alarmed the instant mechanical ventilation began. Dräger’s own reply in that article is the cleanest statement of the limit I have found: “No anesthesia system on the market has completely automated all aspects of the checkout procedures and eliminated the need for manual checkout.”
And in 2016 the APSF Committee on Technology wrote, about foreign material and misconnections in breathing circuits, that these are “a problem that cannot be detected by performing a simple leak check.”
The 2024 AHRQ case of a wall outlet marked oxygen that was piped to nitrous oxide makes the boundary explicit. Among the contributing factors Bohringer and colleagues list, one is that the machine’s self-calibration did not check the gas source. The machine verified itself honestly and thoroughly, and it had no way to know what was coming down the wall.
None of that is a story about bad machines. Lisanne Bainbridge named the shape of it in 1983, writing about industrial process control: “the designer who tries to eliminate the operator still leaves the operator to do the tasks which the designer cannot think how to automate.” Parasuraman and Manzey, reviewing three decades of the human factors evidence in 2010, defined automation complacency operationally as “poorer detection of system malfunctions under automation compared with under manual control,” and reported that it appears in experts as readily as in novices and cannot be trained away with simple practice.
The closed claims data points the same direction. Caplan and colleagues found in 1997 that among gas delivery claims, misuse of equipment (75 percent) was three times more common than equipment failure (24 percent). When Mehta and colleagues updated that analysis in 2013, gas delivery claims had fallen to about one percent of the database, and 85 percent of the remaining claims involved provider error. Thirty-five percent were judged preventable by a pre-anesthesia machine check.
The equipment got safer. The proportion attributable to us went up.
We assess the half that was never the problem.
Here is where I think anesthesia education is stuck, and I will mark clearly that this next part is my argument rather than a finding.
The COA standards for nurse anesthesia programs require that a graduate “conduct a comprehensive equipment check.” They also require, separately, that a graduate “identify and take appropriate action when confronted with anesthetic equipment-related malfunctions.” Both are there. Only the first has an instrument attached to it in most programs, and that instrument is a signature in a log.
That is not a small design flaw. Assessment steers learning, and it has been documented steering it the wrong way since Newble and Jaeger described, in 1983, a change in final-year assessment that produced learning behavior that was “the exact opposite” of what was intended. Cilliers and colleagues later mapped the mechanism, and quote a student saying the quiet part out loud: “you leave things out that you think they will not ask. So it’s maybe big things or maybe important things that could save a patient’s life 1 day, but you don’t swot it because you have to pass the test now.”
There is also a specific reason to distrust the checklist as the instrument here. Regehr and colleagues found in 1998 that expert global ratings outperformed checklists on reliability and validity, and that adding the checklist contributed nothing on top of expert judgment. A year later, Hodges and colleagues found something stranger: on global scales experienced clinicians scored better than residents and clerks, and on checklists they scored worse. Binary checklists, they concluded, “may not be valid measures of increasing clinical competence.”
Set Hodges next to Larson and look at the shape. In the OSCE, checklists scored experts below novices. On the anesthesia machine, experts detected fewer faults than novices. I cannot prove those two inversions share a cause, and I am not going to pretend otherwise. But if the instrument rewards step completion, and expertise expresses itself as selective attention rather than step completion, you would expect exactly this, and you would expect it to get worse as the machine absorbs more of the steps.
The man who wrote the book saw this coming.
Long before any of this was my argument, it was someone else’s warning, and I did not understand it at the time.
Jerry Dorsch wrote the first real reference text on anesthesia equipment. Understanding Anesthesia Equipment, with Susan Dorsch, reached a fifth edition in 2007 at better than a thousand pages on construction, care, and complications. Other books have been written since, and some of them are arguably better. His was first, and for a long stretch it was the only one.
He taught several of my classes, and he taught me directly, in conversations that were one on one and were never meant for publication. What he wanted us to carry was not the sequence. It was the machines themselves, where they were weak, and why, knowing that, you could still depend on them.
He was wary of the newer equipment. Not because it was worse. Because it automated the parts he believed the clinician should know.
I am repeating what he said to me privately, so read it as my recollection rather than as his published position. I did not have the experience to argue with him then, and I do not have the standing to disagree with him now.
What I would assess instead.
Not whether the check was done. Whether the resident can tell me what each check is for.
Four questions I actually use, and they take under two minutes: What does the low pressure leak check protect against, and which component sits between you and that failure? What did the automated sequence just verify, and what did it not touch? If this machine failed right now, what is your backup, and where is it? And, the one that catches the most people, what would you see first?
Chiu and colleagues introduced a single experiential machine check session for PGY-1 residents, then tested them three weeks later against a machine with ten planted faults. The controls were PGY-5s who had received the standard didactic lecture. The juniors scored higher on the checklist and found more faults, both at P less than 0.001, and 21 of them were retested in their senior year and still outperformed the controls. The authors’ conclusion is one experiential session beating five years of residency.
Two hours, once, at the beginning.
I also want to name the thing that makes this teachable rather than punitive. When I ask a resident whether they checked the machine, the answer I most want to hear is the honest one, and I tell them so on day one. “I only checked the circuit and the suction. I didn’t do the full check.” That is a usable answer. It tells me what I am standing on. Yes is not an answer, it is a social reflex, and I have been rewarding it for years by asking a question that accepts it.
The question I should be asking instead.
I have started asking a different one, and it is the question I want the residents carrying long after they have forgotten which vaporizer tips at what angle.
Ten years from now, will you still do a high quality machine check every single day?
Not on the first day of clinical, when everyone does the best machine check in the history of anesthesia. On a Tuesday in your ninth year, on the fourth room of the day, on a machine you have used four hundred times, when the screen says the sequence is complete and there is a plug in your hand.
The mindset I want you to build toward is not “I can’t believe this is happening.”
It is “I knew this would happen one day.”