The Unit Record, Not the Checklist: What the Hospital Signs

What a hospital signs off is not a form; it is a record per unit, and that distinction is what makes the difference between an acceptance that survives an audit and one that is re-opened at the first fault. The factory supplies the test values for each serial number, the site supplies the installation and integration tests, and the buyer assembles both into one file per unit, along with the configuration, the software version, the accessories and the training record. Nothing is finally accepted, paid for or handed over while a unit’s file is incomplete, which is the same gate logic that acceptance testing uses in other industries.

What belongs in a unit record

A usable record has one page per serial number and a small, fixed set of fields. The identifiers: model, serial number, the order line and the configuration code. The factory values: the tests performed before packing, with the measured result and the tolerance for each, not a tick. The electrical safety test, recorded as values rather than as a pass. The battery test where the device has one, with the load it was tested under. The software version as shipped. The accessory list, so that what was delivered can be compared with what was ordered. The site tests: installation, level and stability where relevant, the electrical connection and earth, and the room integration checks. The training record, with who was trained, on what and when. And the dates that matter commercially: delivery, acceptance and the start of the warranty.

The last field is the one that gets argued about. A warranty that starts when the equipment was manufactured is a warranty that has quietly expired while the equipment sat in a warehouse, and a warranty that starts at clinical use is a warranty that starts on a date nobody recorded. Either convention is defensible if it is written down in the order; it is indefensible if it is discovered during a claim.

Which tests can only happen where

The factory can measure what is intrinsic to the device: output, function, the performance of the mechanism, the electrical safety of the assembly, the battery runtime under a defined load, the accuracy of a control system. The site can measure what depends on the building and the installation: the level and stability of a table on the actual floor, the anchoring and the ceiling structure that carries a light or a pendant, the gas supply at the outlet, the electrical supply and its earth, and the clearances and access that make the equipment usable in that room.

Confusing the two is how acceptance goes wrong. A hospital that expects the factory to prove that a pendant suits the theatre’s ceiling has bought a promise; a factory that is asked to certify a gas installation it did not perform is being asked for something it cannot give. Divide the list explicitly, in the order, and write next to each test who performs it, who witnesses it and where the result is filed. That column alone removes most of the arguments that happen at handover, because most of them are about who was supposed to have done a test that nobody did.

The tests that only mean something with a number

Three test types are routinely recorded as a tick and should never be. The first is the electrical safety test, where the measured leakage and earth continuity values, taken under the stated conditions, are the only evidence that means anything later. The second is a battery or backup runtime, which is a number of minutes under a stated load, not a statement that the battery works. The third is anything with a tolerance, from the accuracy of a position to the height range of a table, because a value outside the tolerance is a decision rather than a failure, and the decision has to be made by someone who can see the number.

Ask the factory for those three as numbers per serial in the order confirmation, not as a promise to test. A supplier who can produce them already has a production test regime; a supplier who cannot will discover one while you wait, or will send a form with ticks that a biomedical engineer will reject on the day of acceptance, which is the most expensive day to discover a documentation gap.

The acceptance gate, and how to write it

Acceptance should be a gate with a definition, not a mood. In the order: the delivery is accepted unit by unit when the record for that unit is complete and the site tests for it have passed; the payment schedule releases against batches of accepted units rather than against shipment dates; and units with open items are listed on a punch sheet with a description, an owner and a date, which is the same discipline a construction project applies to a commissioning checklist. Where an item cannot be closed before use, the record should say so and say who accepted the risk, in writing.

The factory’s part in the gate is the factory test file and the configuration record; the buyer’s part is the site tests and the assembly of the file. A distributor who builds this habit delivers a shipment and a file in the same week, and a hospital that receives both accepts equipment rather than promises. The file is also what makes the second order easier, because it is the reference for what “the same as last time” actually means.

Writing the site test protocol

The site test protocol is the buyer’s document, and it can be assembled from three sources: the factory’s manual, which says what the device should do; the project’s requirements, which say what this room needs; and the hospital’s own commissioning practice, which says what its engineering department will sign. Each line in the protocol should have four fields: what is being tested, the acceptance criterion, who performs it, and where the result is recorded. Thirty to fifty lines are usually enough for a theatre set, and the exercise of writing them exposes the assumptions that otherwise surface at handover.

Where the project has a formal acceptance process of its own, the same content can be presented as a site acceptance test with the factory’s results attached as the counterpart, which is the pattern described in the note on factory and site acceptance testing. What matters is not the label but the pairing: for every factory value there is a site confirmation, and for every project requirement there is a test that demonstrates it in the room where the equipment will work.

Three practical habits keep the gate honest. The record travels with the unit rather than with the shipment, so a single missing page does not hold up the other nineteen. The site tests are scheduled with the people who perform them, which means the electrical contractor and the hospital’s biomedical team are booked before the equipment arrives rather than after. And the file is reviewed by someone who did not run the tests, because a record assembled by the person who performed the work tends to contain what was done rather than what was required, and the difference between those two lists is exactly where an audit finds the gap.

What it protects, six months later

Two events make the value obvious. A fault appears, and the engineer’s first question is whether this unit behaved differently from the start, which the record answers with the values it was accepted against. Or a dispute arises about who supplied what, and the accessory list and the configuration code settle it without a search through emails. In both cases the cost of the record was an hour per unit at delivery, and the cost of not having it is a fault investigation that starts from zero, with the equipment already in clinical use and the supplier on the other side of the world. There is a third case, quieter and more common: an inspector or an accreditation reviewer asks to see the equipment file for a theatre, and a department with a page per serial number answers in ten minutes, while a department with a folder of delivery notes spends a week reconstructing what it can and explaining what it cannot.

Write it into the order in seven lines: the per-serial factory test values; the electrical safety values; the battery runtime under load, where applicable; the software version; the accessory and configuration list; the site test list with the responsible party for each; and the acceptance gate with its payment consequence. Then file the pages by serial number as they arrive, because a record assembled in the week of delivery is a record that exists, and a record promised for assembly later is a record that will be reconstructed from memory. The test is simple enough to apply at the next delivery: pick one unit at random, and ask how long it takes to produce its complete page from the file.