Poznań Provincial Hospital in western Poland fitted RFID tags to 11,000 surgical instruments in November 2025, the first hospital in Poland to track its surgical inventory that way. Nine months later it was in the newspapers for the wrong reason. On 30 July 2026 Gazeta Wyborcza reported that theatre staff at the hospital’s ul. Lutycka site were living with constant nerves about whether a chip would fall off, that chips had come away from instruments during use and sterilisation, and that chips had been lost in patients’ wounds.
Several hundred tags were damaged over those nine months. The headline that travelled was that RFID chips are falling off surgical instruments in a Polish hospital, but that is not really a story about RFID. It is a story about tag specification, attachment method and process control, and those three things would have failed an adhesive label or an etched DataMatrix plate just as readily. What makes Poznań worth studying by anyone running an asset tracking programme is the consequence. In a warehouse a detached tag costs you a lost pallet. In an operating theatre it is a retained foreign object.
What Poznań Provincial Hospital deployed, and why
Szpital Wojewódzki w Poznaniu is a regional hospital owned by the Wielkopolska (Greater Poland) region. Its supplier was Caretag, a Danish company selling an AI-supported UHF RFID platform and durable tags built for surgical instrument tracking and sterile processing. UHF RFID is the long-range passive flavour of the technology, working in Europe around the 865 to 868 MHz band, and a reader can identify many tags at once from a distance rather than one at a time in contact. That is what makes tray-level counting practical, and why the supplier claims packing time falls by more than a third against scanning DataMatrix codes item by item.
The estate grew from 11,000 instruments at go-live to 12,500 by August 2026, and around 3,000 instruments that appeared on no inventory turned up during the marking exercise. The system cost 3 million PLN, funded in full by the Wielkopolska Marshal’s Office, with the hospital’s director putting the return at roughly 1 million PLN a year and an 18-month payback. It was procured openly, as tender SZW/DZP/64/2025, published on 2 June 2025 and closing to bids on 17 July 2025.
Most of the expected benefit sat in central sterilisation, the department that cleans, inspects, packs and autoclaves reusable instruments between operations. At launch Prof. Dawid Murawa, the hospital’s director, projected a 30 per cent efficiency gain across the operating block and central sterilisation. By August 2026 the reported figure was narrower and should be quoted precisely: 40 per cent of the time previously spent on manual instrument counting had been saved. That is a saving on counting, not a 40 per cent gain across the theatre block. The driver is the EU Medical Device Regulation, MDR 2017/745, whose Unique Device Identification rules oblige devices to carry a machine-readable identifier so individual items can be traced. May 2027 has been cited as the deadline for most Polish hospitals, so Poznań was early rather than eccentric.
What actually went wrong
Radio Poznań reported on 11 August 2026 that the cause of the damage lay in Polish sterilisation practice, specifically the excessively rapid heating of instruments to high temperature. The hospital and the manufacturer responded on 12 August 2026, with Głos Wielkopolski publishing its own follow the same day and crediting Gazeta Wyborcza’s original reporting. The hospital said staff had since resolved the heating problem and that it intends to extend the system to further departments.
Read that cause again, because it is the whole article in one line. The tags were not defeated by sterilisation. They were defeated by a ramp rate.
The RFID was not at fault. The tag choice and the attachment were
A surgical instrument is one of the more hostile places a passive tag can be asked to live. Before it reaches an autoclave it goes through ultrasonic cleaning, where cavitation hammers every edge and joint, and enzymatic detergents that attack organic residue and, incidentally, adhesive chemistry. The steam cycle runs at 121 to 134 degrees C at up to around 2.1 bar, followed by a drying phase in which any moisture that has worked into a housing wants out. Standard RFID labels delaminate, crack or lose read sensitivity within a handful of cycles. Purpose-built high-temperature tags are rated for continuous exposure at around 200 degrees C, and the tags used at Poznań are specified to survive more than 2,500 sterilisation cycles without affecting performance or the usability of the instrument.
So the tag was rated. The catch is what a cycle rating means. A tag qualified for a number of cycles is qualified against an assumed cycle profile: a ramp rate, a hold temperature, a pressure curve, a drying regime. Change the ramp and you are outside the qualification, and the datasheet no longer describes your process.
The other half of the problem is attachment. There are three broad ways to fix a tag to a surgical instrument, in rough order of permanence: high-temperature medical-grade adhesive bonding, which is least invasive and quickest to apply across a large existing inventory; mechanical attachment by rivet or weld, which is considerably more robust; and embedding the tag in a recess machined or drilled into the instrument, the most durable and the most expensive because it means touching the instrument’s geometry. Which one is appropriate depends on the size and shape of the item. Delicate instruments generally need a small form factor tag, often housed in PPS with rounded edges so it cannot snag on tissue, glove or drape.
The failure mechanisms worth designing against are thermal shock from rapid ramp rates, moisture ingress during drying, pressure cycling, ultrasonic cavitation, and enzymatic chemistry working on the bond line. Notice how many of those attack the join rather than the electronics. The chip is rarely the weak point. The bond is. Tag retention is a mechanical engineering problem wearing an RFID badge, and it tends to be specified by people thinking about read rates.
ISO 10993 answers a different question
Caretag’s co-founder and chief commercial officer, Michael Leibig, responded that the chips are safe for humans, comply with ISO 10993, and that testing shows contact with patients presents no danger. On its own terms that is very probably accurate. ISO 10993 is the biocompatibility standard for medical devices, governing whether a material is safe in contact with the body: cytotoxicity, sensitisation, irritation.
It says nothing at all about whether a tag stays attached. Biocompatibility and retention are different properties, tested differently, and answering a retention question with a biocompatibility certificate does not close it. A perfectly biocompatible tag lying loose in a surgical wound is still a retained foreign object, and that is a never event in surgery whatever the object is made of. Any hospital procuring instrument-level RFID in healthcare should ask for retention data against its own validated sterilisation cycle, in writing, alongside the biocompatibility file.
The audits nobody was talking about
The detachment story was not the first sign of trouble. On 1 April 2026, four months earlier, Gazeta Wyborcza reported that officials from the Marshal’s Office, the body that paid for the system, had run two audits at the Lutycka hospital and found “a series of irregularities” in the rollout of the instrument-marking programme. A deputy director lost her job over it.
Prof. Murawa also said several instruments showed signs of chips having been picked off by hand, something the hospital treats as a kind of sabotage, and that cameras were installed in response. He said all chips were recovered. That should be read as what it is, a claim by the institution under criticism, and no investigation outcome has been reported. Two silences are worth recording for anyone researching this later. The hospital’s own website has published nothing about the RFID system, at launch or during the controversy, so every hospital statement has reached the public through journalists. The supplier has published no case study or customer reference for Poznań either, despite it being its flagship Polish deployment.
What every RFID project should take from this
The lessons are not exotic. Qualify the tag against your process, not the vendor’s assumed process, with ramp rate, hold temperature and chemistry written into the test plan. Choose the attachment method for the instrument rather than for the speed of the rollout. And instrument the failures, because several hundred damaged tags over nine months should have surfaced as a trend in week three, not as a newspaper story in month nine.
Poznań has not shown that UHF RFID does not belong in an operating theatre. The savings on counting are real, the MDR clock is running, and 3,000 instruments nobody knew they owned is its own argument. What it has shown is how narrow the margin gets between a good idea and a clinical incident when the least glamorous line in the specification, the bit that holds the tag on, is treated as a detail.
