The Batch That Passed Every Test and Still Got Rejected
In Q3 2024, we received a shipment of 1,200 Smiths Medical tracheostomy tubes from the Bivona line. Paperwork was clean. Dimensional checks passed. The connector pull-force sat inside the ISO 5366-1 tolerance band. Three days after the lot hit the ICU floor, our clinical educator called me with a complaint that doesn't fit neatly on a spec sheet: the hubs felt looser than the previous batch. Not loose enough to fail a test. Loose enough that two senior nurses independently noticed.
We pulled the lot. Retested. Everything passed again. And that's the problem I want to talk about, because it isn't really about one batch of Bivona tubes. It's about the gap between "spec-compliant" and "fit-for-use"—and how most QC programs I've worked inside, including my own early version, are built almost entirely around the first one.
What We Usually Think the Problem Is
From the outside, medical device quality looks like a documentation problem. If the certificate of analysis is present, the lot is traceable, the sterilization indicators are valid, and the incoming inspection matches the drawing, the device is good to go. That's the logic most receiving departments run on, and it isn't wrong. It's just incomplete.
People assume the hard part is catching counterfeit or out-of-spec product. What they don't see is that the far more common failure—in my experience, anyway—is product that is technically in-spec and still wrong for the clinical workflow it lands in. The spec caught nothing because nothing was out of spec.
What's Actually Going On
Specifications are a floor, not a ceiling
A specification defines the worst acceptable performance, not the best. For a Smiths Medical tracheostomy tube, the standard covers dimensions, materials, cuff behavior, and connector security. What it doesn't cover is whether the assembly feels the same in the hand as the last five lots. And "same in the hand" matters more than we usually want to admit, because in an airway emergency, a clinician's muscle memory is part of the safety system.
Here's something vendors won't tell you: "within tolerance" covers a range wide enough that two batches can both pass incoming inspection and still feel different to an experienced nurse. The tolerance was designed to protect the patient from a bad device. It was never designed to protect the clinician from a confusing one.
According to ISO 5366-1:2023 (ISO, iso.org), tracheostomy tube connector security is specified as a minimum pull-force threshold—not a target feel. That's a reasonable way to write a standard. It also explains why a passed test tells you less than it seems to.
Consumables and capital equipment fail in different ways
I spend most of my time on two categories that get lumped together under "medical devices" but behave nothing alike.
The first is single-use consumables—the Bivona tracheostomy tubes, IV catheters, infusion sets. These fail at the margins. A connector that's slightly stiff. A cuff that inflates a half-second slower. Packaging that tears on the wrong seam during a code. Individually small. Cumulatively, the kind of thing that erodes clinical trust in a brand faster than any recall.
The second is capital equipment: medical imaging systems, endoscopes, patient monitors. If you're new to the category, medical imaging is broadly the use of X-ray, CT, MRI, ultrasound, or nuclear techniques to visualize anatomy or function—but the procurement implication is that a medical imaging system isn't just a box that produces pictures. It's a workflow tool, and it fails when image processing latency, table ergonomics, or DICOM integration slows a procedure down. An endoscope that meets ISO 8600 optical requirements can still be a bad purchase if the angulation is stiff enough that the gastroenterologist reaches for the older scope instead.
Our own program used one inspection template for both categories until 2023, then wondered why capital equipment got signed off and still disappointed. The checklist that works for a tracheostomy tube lot is not the checklist that works for an endoscope.
Certification is not consistency
ISO 13485:2016 certification tells you the quality management system is in place. It doesn't tell you that batch 14 and batch 15 will behave identically in the OR. ISO 13485 clause 7.3.3 requires design outputs to be verified against inputs, but the inputs are written by the manufacturer, not the end user. FDA's 21 CFR 820 and the EU's MDR 2017/745 set the framework for process control, and process control is a statistical claim—not a guarantee at the individual lot level.
FDA's MAUDE database (openFDA, accessdata.fda.gov) is public and searchable. The fact that it exists at all is a reminder that in-spec devices still generate complaints. I'm not arguing against certification. I'm arguing that "we only buy from certified suppliers" is a policy, not a quality program.
What It Costs When We Miss It
The Bivona lot I mentioned earlier cost us roughly $18,000 in expedited replacement, quarantined inventory, and clinical education time to re-train ICU staff on the "new normal" feel of the connector. It took about three weeks—or rather, closer to five when you count the re-qualification cycle and the second batch we rejected for the same reason.
But the number that actually matters isn't on the invoice. It's that our ICU team stopped trusting the Bivona label for about six months. They started asking for a different vendor's tracheostomy tube whenever they had a choice. We eventually recovered the relationship, but it took a documented consistency study and a lot of conversations a spec sheet can't have.
Capital equipment has a similar failure mode with a longer tail. A medical imaging system that passes acceptance testing but has a clumsy worklist interface will get worked around. An endoscope that meets optical specs but adds two minutes per procedure doesn't show up as a defect—it shows up as a scheduling problem, then a utilization problem, then a budget conversation where nobody remembers why the scope was bought in the first place.
Looking back, I should have added a clinician hand-feel check to our receiving protocol in 2022, when we first started seeing lot-to-lot variation across several vendors. At the time, it felt redundant with the dimensional tests—and in my defense, nobody had complained yet. That's the trap. The gap between spec and workflow doesn't announce itself until someone in a clinical setting is annoyed enough to call.
What I Changed (and What I'd Still Question)
I'm not going to pretend there's a clean framework that solves this. What we did was add a second layer to incoming inspection that runs parallel to the compliance layer:
- A clinician blind check for high-touch consumables. Two nurses or techs handle the lot without seeing the batch number and compare it to the previous lot. If they consistently flag a difference, we investigate before the lot goes to the floor. It isn't statistically rigorous. It's fast, and it catches what the spec sheet can't.
- A separate acceptance track for capital equipment. The imaging and endoscope checklist now includes a workflow observation—we watch a real procedure, not a demo, and log where the clinician hesitates. Hesitation is data.
- Change notification language in every contract. Any material, process, or manufacturing-site change requires notice before shipment. A lot of "feel" problems are actually silent process changes that never triggered a spec violation.
What I'd still question: whether this scale of clinician involvement is realistic for smaller facilities. My experience is based on roughly 200 consumable lots and about 40 imaging or endoscope installations over four years, inside a mid-sized health system with enough staff to run parallel checks. If you're a rural clinic or a single-site surgery center, you probably can't replicate this, and "just trust the certificate" might be a reasonable risk trade-off—though I'd want to see your complaint log before agreeing with that.
The industry has gotten better at documenting quality. I think it's gotten only marginally better at defining what quality means at the point of use. What was best practice in 2020—certificate plus visual inspection—isn't wrong in 2025. It's just insufficient. The fundamentals haven't changed: traceability, valid sterilization, dimensional conformity. But the execution has to account for a fact the spec sheet has never handled and never will—that a tolerance band has never held a tracheostomy tube in an emergency.