Thirteen Stages, 114 Failure Modes: The Safety Work That Happens Before a Chamber Opens

A bolted multiplace hyperbaric chamber hatch in a hospital treatment room, with a coiled oxygen hose, an inspection clipboard and a small electrical-safety pictogram on the steel shell

When a hospital in northern Israel prepared to open a hyperbaric oxygen service, a multidisciplinary team sat down and tried to enumerate everything that could go wrong — not just inside the chamber during a treatment, but in the maintenance bay, in the corridor where patients are transferred, at discharge, and in the electrical supply feeding the room. They mapped thirteen process stages and came up with 114 distinct failure modes.

The highest risk priority number, 504, belonged to something almost banal: the use of a non-approved external electrical device.

That detail is worth sitting with, because it inverts the usual mental picture of hyperbaric risk. The public imagination, and a good deal of clinical anxiety, centres on pressure — barotrauma, oxygen toxicity, a chamber that will not come back down. The failure mode that this team ranked first was a plug, a cable, a piece of equipment nobody had cleared for use in an oxygen-enriched, pressurised steel room. It is also the exact hazard the US Food and Drug Administration chose to write about in its August 2025 letter to health care providers, which mentions fires that caused serious injuries and deaths and warns specifically about prohibited electrical and static devices.

The risk register went outside the chamber

The project, published in the Joint Commission Journal on Quality and Patient Safety in August, is a prospective failure mode and effects analysis — FMEA, the engineering method that assigns severity, occurrence and detection scores to each way a process can break, multiplies them into a risk priority number, and then ranks the fixes. What makes the account useful is not the arithmetic. The authors are unusually explicit that RPNs are prioritisation aids rather than precise risk estimates, and that their single-centre project did not evaluate whether FMEA reduced patient harm or improved clinical outcomes.

What it does document is where the high-priority risks landed. Maintenance. Patient preparation and transfer. Discharge planning. Equipment control. Fire and electrical safety. Of the ten highest-priority failure modes, mitigation actions for seven were finished before launch; three needed reinforcement during early follow-up. All core launch-facing staff completed preparation, three operational checklists went into service, and by February 2026 the unit had run 347 treatment sessions without a patient harm event in the reviewed top-risk categories — an early operational observation, as the authors say, not evidence of anything.

Read that list again and notice how little of it is hyperbaric medicine. A chamber service is a small hospital inside a hospital: it needs oxygen supply, compressed air, fire response, infection control, moving and handling, an emergency plan and a maintenance regime with records. The interesting question for anyone evaluating a clinic is not whether the chamber is certified. It is whether anyone wrote down, in advance, the 114 ways the surrounding system could fail.

A defibrillator is the hardest device to justify in there

If electrical equipment inside a chamber is the top-ranked hazard, the most consequential instance is the one you cannot simply leave outside: a defibrillator. Cardiac arrest during treatment is rare, but the whole point of a chamber that treats carbon monoxide poisoning, gas gangrene and necrotising soft-tissue infection is that some of its patients are critically ill.

A scoping review published in Diving and Hyperbaric Medicine in December 2025 looked for evidence on what to do when one of them arrests. Ten thousand three hundred and forty-eight records were screened and ten included; hand-searching reference lists added twenty-three, for thirty-three publications in the final set. Four of them described actual defibrillation during hyperbaric oxygen therapy, covering five patients in total. The authors found no standardised guidelines. They report strong consensus that defibrillation is strictly contraindicated inside a monoplace chamber, and that in a multiplace chamber the risks and benefits have to be weighed individually, because the oxygen-enriched atmosphere and the increased ambient pressure both raise the fire and explosion risk if a spark occurs.

The Australasian picture is consistent with that vacuum. A survey sent to all fifteen hyperbaric units in the region got fourteen responses, and none reported a clinical case of in-chamber cardiac arrest or defibrillation. Australian and New Zealand standards, the authors note, prohibit the defibrillators currently licensed for in-chamber use because they contain lithium batteries. Six of the fourteen units — 43 per cent — ran regular cardiac-arrest drills. Eleven wanted a uniform regional protocol, though comments suggested that may not be achievable.

Put the two papers together and a practical shape emerges. For most units, the emergency plan for a shockable rhythm is not “shock the patient in the chamber”. It is “decompress, extract, then shock” — which makes the speed of an emergency decompression, and the training of the attendant who has to perform it, part of the resuscitation chain rather than an administrative footnote.

Nothing certified at sea level is the same device at pressure

The same logic runs through the equipment literature, which is thicker than the defibrillation evidence and consistently finds that pressure changes what a device does, not merely where it sits.

A bench study of a widely used transport ventilator, the Uni-Vent Eagle 754, tested it at 1.0, 2.4 and 2.8 atmospheres absolute and found it delivered roughly 25 per cent less volume at depth than at the surface, with breath stacking at high respiratory rates that was blunted under pressure. The investigators had specifically been asked about oxygen leaking into the ventilator housing — a fire pathway — and found no rise in housing oxygen; their practical conclusion was that the machine performs safely but needs spirometry to program settings that actually deliver the intended breath. That is a modest finding with a large implication: a ventilator that is correct on the ward is not automatically correct in the chamber.

Intracranial pressure monitoring has the same character. A 2026 bench study of a Codman EDS 3 external ventricular drain, taken from 101 to 283 kilopascals in a multiplace chamber, found the system stayed structurally intact — but pressure readings varied in the closed, compliance-based models, including when closed to drainage, and unclamped drainage behaved inconsistently enough that the authors could not assess it reliably. Their conclusion is a sentence every neuro-intensivist sending a patient into a chamber should read: intracranial pressure measured during hyperbaric treatment should be interpreted with caution.

Not every device fails the test, and the successes are as informative as the failures. A 2026 study of four laptop-style point-of-care ultrasound machines in a multiplace chamber found all of them functioned at every pressure tested, with no spark or fire, no surface temperature above 120 degrees Fahrenheit, and images good enough to run full cardiac, FAST, IVC and thoracic sequences. The reason that paper exists at all is that “electronic device in an oxygen chamber” had been treated for decades as a category rather than as a question. It turns out to be a question, answerable only device by device.

Who is actually in the chamber at two in the morning

There is a staffing dimension to all of this that the clinical literature rarely makes visible, and an eleven-year cohort from Poznań supplies it.

Of 6,410 patients treated with hyperbaric oxygen at a single academic centre between 2013 and 2023, 176 were mechanically ventilated intensive care patients — 2.7 per cent of the total — and they accounted for 1,381 treatment sessions. Just over half of those sessions, 56.6 per cent, were nocturnal.

The case mix drifted in a way that matters for planning. Per additional calendar year, the odds of a ventilated patient having carbon monoxide poisoning fell (odds ratio 0.85, 95 per cent confidence interval 0.76 to 0.95), as did the odds of clostridial myonecrosis or gas gangrene (0.85, 0.76 to 0.96). The odds of a non-clostridial necrotising soft-tissue infection rose (1.23, 1.11 to 1.36), and so did the odds that the patient was on a catecholamine infusion (1.17, 1.04 to 1.31). Mortality was near null across the period, and the renal replacement estimate was too imprecise to interpret.

The authors’ own conclusion is not clinical but logistical: persistent nocturnal activity supports the case for ICU-capable infrastructure and flexible staffing for emergency treatment. That is the same territory the FMEA flagged — preparation, transfer, discharge, training — arriving from a completely different direction, and it explains why a chamber is not a room with a machine in it. It is a critical-care capability that happens to pressurise.

Where that capability is missing, improvisation fills the gap. A case report from rural Japan describes a 68-year-old woman and her 42-year-old daughter pulled from a bakery oven with carboxyhaemoglobin of 22.6 and 15.6 per cent; transfer to the nearest multiplace facility would have taken about ninety minutes by emergency vehicle, so both were treated at once in a monoplace chamber. Symptoms resolved, carboxyhaemoglobin fell to 1.4 per cent, the mother was admitted overnight. The authors are candid that treating two people in a one-person chamber requires close physical contact and would realistically only work for relatives. Improvisation that succeeds is still improvisation, and the published record of it is a map of where capacity is thin.

What the rules actually say, and what failure looks like

The regulatory floor is lower than most patients assume. The FDA’s August 2025 letter does not mandate a risk register or a drill schedule. It reminds providers to follow the manufacturer’s instructions for use, to keep to recommended cleaning procedures, maintenance intervals and safety checks, to avoid prohibited items including electrical and static devices, to add grounding where required, and to dress patients in hyperbaric-compatible fabric such as cotton rather than wool or synthetics, which generate more static. It notes that hyperbaric devices are Class II, cleared through the 510(k) process under product code CBF, and points to NFPA 99-2024 and ASME PVHO-1-2023 as the relevant engineering standards. It also states plainly that the root cause of the recent fires is not known.

Coverage does the rest of the shaping. Medicare’s national determination for hyperbaric oxygen therapy limits reimbursement to treatment delivered in a chamber for a fixed list — acute carbon monoxide intoxication, decompression illness, gas embolism, gas gangrene, acute traumatic peripheral ischaemia, crush injuries and suturing of severed limbs, progressive necrotising infections, acute peripheral arterial insufficiency, preparation and preservation of compromised skin grafts, chronic refractory osteomyelitis, osteoradionecrosis and soft-tissue radionecrosis as adjuncts, cyanide poisoning, actinomycosis, and diabetic lower-extremity wounds meeting three criteria — and requires documented failure of standard therapy for at least thirty days before the diabetic-wound indication applies, with reassessment every thirty days. The Undersea and Hyperbaric Medical Society’s fifteenth-edition list carries fifteen indications, from air or gas embolism to avascular necrosis. Almost every patient in the papers above sits inside one of those entries. That is the point of them.

And the failure mode that no checklist fully removes looks like this. An Images in Emergency Medicine report from Zhanjiang, China, describes a 27-year-old man who had been comatose for ten hours after carbon monoxide poisoning, was transferred for hyperbaric treatment, and was ventilated with oxygen saturation at 99 per cent. After he was removed from the chamber he vomited and had nosebleeds; within a minute his heart rate fell to 42 and his saturation to 69 per cent. He developed cyanosis of the head and neck, generalised subcutaneous emphysema, absent breath sounds bilaterally and abdominal distension. Resuscitation began, a needle was inserted into the left chest, and whole-body computed tomography showed bilateral pneumothorax, pneumoperitoneum, pneumomediastinum, systemic arterial gas embolism and extensive subcutaneous emphysema. He died after ninety minutes of resuscitation.

A single case report cannot establish mechanism, timing or preventability, and this one is published as an imaging vignette rather than an investigation. What it establishes is that the catastrophic end of the failure list is real, that it can present within a minute of leaving a chamber, and that its signature — vomiting, nosebleed, then bradycardia and desaturation — is one a chamber team has to recognise instantly. It is also, uncomfortably, the kind of event a pre-launch hazard analysis exists to imagine before it happens.

None of this is an argument against hyperbaric oxygen therapy within its cleared indications; the populations in every study above are inside the FDA-cleared and Medicare-covered list, and the FMEA unit’s own record after 347 sessions was uneventful. It is an argument about what to ask. Whether a service has a written hazard analysis, whether its drill schedule survives a busy month, whether its ventilators have been spirometry-checked at depth, whether anyone has approved the devices in the room. Those questions do not appear in advertising for wellness chambers, and the FDA’s position remains that most wellness and longevity uses are not cleared — with the same fire and electrical hazards present, and often less engineering behind them. The fires the agency describes did not happen in imaginary rooms.

Sources

← Back to blog