Forty manual hospital beds were inspected before shipment, and on six of them the brake pedal travelled the full length of its stroke and stopped 3 mm short of the tyre, because the assembly jig had been left in the wrong position after a changeover and the only check on the line was done with a hand.
Illustrative example This case is written from the patterns we see in furniture export orders, and the figures are representative rather than from a specific client file. It shows how a check works and what it finds. We will replace it with a real, anonymised case as soon as we can publish one with the buyer's permission.
The order, and the clause that was not in it
A European care-home and rehabilitation buyer ordered forty manual hospital beds from a factory in Foshan. Crank-operated backrest and leg sections, side rails, 125 mm medical silent castors, and a central foot brake where one pedal locks all four wheels. The order was placed against a purchase specification the buyer had written, and the specification was long and detailed about almost everything.
It said: central brake pedal, engages all four castors, red and green position indicator at the pedal, locking type castor brakes. It did not say how anyone was to establish that the brake had actually locked the bed. There was no force, no gap, no surface, no load, and no margin.
Thirty per cent had been paid on order. Seventy per cent, $7,420, was due against delivery.
Two of the words in that clause did all the damage. Central brake pedal describes a shape. Engages all four castors describes a quantity. Neither of them says what the bed does when somebody steps on it, and that is the only thing a brake is for.
The check the factory was running
The factory's final check on the brake took about four seconds per bed. The operator stepped on the pedal, watched it travel, confirmed the indicator changed colour, and ticked a box. Forty out of forty passed.
That check is not stupid. It is the check the purchase order described. It asks whether the mechanism is present and whether it moves. It cannot ask whether the shoe reaches the tyre, because that needs a feeler gauge and a force gauge, and neither was specified.
On the six beds that failed, the pedal went down exactly as it should. It travelled the full length of its stroke, it felt firm at the end of the stroke, and the indicator showed locked. Every part was present and every part was moving. That is what a brake that does not lock looks like to a hand.
The test: push the bed, do not press the pedal
The first bed the inspector tested was bed 3. It was loaded to its safe working load, the brake was engaged, and the inspector pushed at the foot end with a force gauge. The bed moved. Not far, and then it stopped, but it moved, at 41 N, with the brake supposedly locked.
That number is worth pausing on. The Chinese national standard for manual hospital beds, YY 0003-2023, sets the force needed to start a mobile bed moving under its safe working load at no more than 160 N, and the force to keep it moving at no more than 85 N. Those limits apply to a bed whose brakes are released. A bed with a working brake should not move at all.
So the inspector stopped using the factory's method, wrote a test that could be read off a gauge, and ran it on every bed in the order rather than on a sample. Forty beds, four measurements each.
| Measurement | 34 beds | 6 beds, numbers 23 to 28 |
|---|---|---|
| Gap between brake shoe and tyre at full pedal travel | 0.2 to 0.5 mm, shoe in contact | 3.0 mm on all four castors |
| Force to start the bed moving, brake engaged, bed at safe working load | did not move under 150 N | 30 to 45 N |
| Movement on a 10 degree ramp, safe working load fitted, brake engaged | 2 to 5 mm, then held | 150 to 300 mm, continuous |
| Pedal travel to the end of its stroke | full stroke, firm | full stroke, firm |
The last row is the one to read twice. On the failing beds the pedal travelled to the end of its stroke and stopped there, exactly as on the good beds. Nothing about the pedal was wrong. Everything about the pedal was working.
Where to go next
Six at once is a shape, not a rate
Six of forty is 15 per cent, and any reasonable sampling plan would have caught that if it had happened to draw a bed from the right part of the order. Put the failures in build order, though, and the number stops being a rate and becomes a shape: the six failing beds were numbered 23 to 28, consecutive, out of forty.
Contiguity is diagnostic. Random defects scatter. They arrive wherever chance puts them, and the pattern tells you the cause is everywhere and nowhere. A defect sitting in one unbroken run has one cause, and that cause was active for exactly the length of the run.
| Build sequence | Beds | Result |
|---|---|---|
| 1 to 22 | 22 | brake held |
| 23 to 28 | 6 | brake did not reach the tyre |
| 29 to 40 | 12 | brake held |
Three millimetres, and a jig left in the wrong place
So the question changed. It was no longer why some beds failed. It was what was different about beds 23 to 28, and the factory's assembly log answered that in about twenty minutes.
The factory ran two bed versions on the same line and the same chassis, each with a different brake lever, and the lever's stop position was set on the assembly jig for whichever lever was on the line. Beds 23 to 28 were built straight after a changeover back from the other version. The jig had not been reset.
The stop was 3 mm off, which meant the lever rotated 3 mm less than it needed to, which meant the shoe stopped 3 mm short of the tyre. The pedal kept travelling and then pushed against a lever that had nowhere further to go. It felt firm because firm is what a pedal feels like when the lever has reached its stop, whether that stop is in the right place or not.
The line's own changeover check for the brake was a functional click test, and it passed.
This is the part of the case that should worry a buyer most, and it is not the six beds. The other thirty-four were not right because a designer had left room in the brake. They were right because a stop position had been set correctly by hand on a jig, and on this design there is nothing at all between set correctly and the brake does nothing. The pass margin is zero.
A brake with no margin is also a brake with no life. As the tyre wears and the linkage takes up, the same design walks into the same failure without anybody touching a jig. The fix is two separate things and it is worth keeping them apart. Recording and verifying the setting catches the error that happened here. Requiring margin protects against wear. Neither substitutes for the other.
Why thirty-four passing beds did not settle it
The factory's fix was to reset the jig and rework the six beds. The report's job was to say what had to be re-measured afterwards, because the rework itself is the risky part. Adjusting a stop by hand is exactly the operation that can be corrected in one direction and over-corrected in the other.
A stop set 3 mm too far the other way gives a bed that locks perfectly and a pedal with no travel left in it. That bed would hold, it would bite hard into the tyre every time the brake was applied, and it would chew through the castors without anybody noticing. Both faults look identical from the outside: a pedal that goes down and a wheel that stops. Only the measurements tell them apart.
| After rework, all six beds | Reading |
|---|---|
| Brake shoe contact | 0.2 to 0.5 mm gap, contact on all four castors |
| Force to start moving, brake engaged | did not move under 150 N |
| 10 degree ramp, safe working load fitted | 2 to 4 mm, then held |
| Pedal travel remaining at end of stroke | full stroke, firm, no slack at the shoe |
One more thing the report had to say. The force test was carried out on the workshop floor, and the friction between a tyre and a floor is part of what holds a bed. A braked bed that holds on sealed concrete is not automatically a braked bed that holds on hospital vinyl, particularly when it is loaded and the floor is wet. The numbers in the report describe the floor the beds were tested on, and the report says so.
A brake is not an AQL item
The phrase means something specific, and the industry knows it. Characteristics whose failure can injure somebody are inspected on every unit, not sampled, because a sampling plan is built on the assumption that defects are scattered, and this one was in a run.
The arithmetic makes it plain. With six bad beds in forty, a sample of eight has a 76 per cent chance of drawing at least one of them, a sample of three has 39 per cent, and a sample of thirteen has 92 per cent. Even the best of those is close to a coin flip on a safety function, and any of them gets worse when the defects sit in one contiguous block instead of spreading out.
So the brake was checked on all forty beds, and every bed in an order that carries a central brake needs the same treatment. The AQL calculator on this site is the right tool for appearance and dimension defects, which are scattered and where a rate is a fair description of the lot. It is the wrong tool for a brake, and any medical furniture inspection that samples the brake has not inspected the brake.
What the report said, and what it did not say
The report recorded the measurements, the bed numbers, the rate, and the AQL position, and it said plainly that beds numbered 23 to 28 did not meet the clause in the purchase specification, because there was no clause to meet and the brake did not lock.
It also recorded what it could not establish. A medical furniture inspection determines whether the goods match the order. It does not determine whether a bed is fit for clinical use, it does not assess conformity with the EU Medical Device Regulation, and it does not replace the manufacturer's own design file and risk analysis. A brake that holds on a workshop floor on the day of inspection is not a certification.
- Recorded: the gap between brake shoe and tyre, the force to start the bed moving, and the movement on a 10 degree ramp, for every bed tested, with the instrument named.
- Recorded: the failing beds were contiguous in build sequence, and the jig had not been reset after a changeover.
- Recorded: the factory's own final check passed all forty beds and could not have failed any of them, because it tested the pedal rather than the brake.
- Not recorded: whether beds already shipped from this line carry the same stop setting. That is a question for the factory's production records, not for an inspector at the end of one order, and the report says so rather than guessing.
- Not recorded: any judgement on whether the bed is safe to use. That is not an order verification statement, and nobody should read it as one.
What the specification should have said
Here is the specification this order needed and did not have. Six lines, and none of them is advanced.
- Brake function shall be verified by measurement, not by operation. With the brake engaged and the bed at its safe working load, a horizontal force of 150 N applied at the foot end shall not start the bed moving. The instrument shall be a force gauge reading in newtons, and the surface shall be the floor of the assembling workshop, recorded with the result.
- Shoe contact shall be measured. With the pedal held fully down, the gap between the brake shoe and the tyre shall not exceed 0.5 mm at any point of the shoe face. A 3 mm feeler gauge shall not pass between shoe and tyre.
- Brake holding shall be verified on an incline. With the bed at safe working load and the brake engaged on a 10 degree ramp, movement shall not exceed 10 mm over one minute. This is the condition in the standards, and it is the condition that matters, because a bed that has to be pushed hard to move is a bed that will move when a resident shifts.
- The brake shall have margin. The design shall provide not less than 5 mm of pedal travel between first shoe contact and the end of pedal stroke, so that stack-up variation and setting error cannot turn a brake into a pedal. The requirement shall be checked on the assembled bed, not on the component.
- Settings shall be recorded and released. The brake lever stop setting shall be recorded per version on the assembly fixture, and the first bed off a changeover shall be verified before the run continues. The setting shall be recorded against the bed's build number.
- The standard in the file shall be the standard that applies. A manual bed is covered by YY 0003-2023 for the Chinese market, ISO 13442 covers wheeled medical beds and their accessories, and ISO 22882 covers hospital bed castors. IEC 60601-2-52 is written for beds with electrical functions and its scope excludes fully manual beds, so naming it on a manual bed order creates a reference the factory cannot test against and nobody can argue about.
A medical furniture inspection can only check against a clause that exists. Every one of these six lines exists to turn a word like engages into a number somebody can write down and a second person can reproduce.
Which standard the order was actually written against
The buyer's specification cited a European standard for medical beds. The bed on this order had no electrical parts at all: cranks, a mechanical brake, no motor, no controller, no battery. IEC 60601-2-52 is written for medical electrical equipment with beds that have electrical functions, and its scope explicitly excludes fully manual beds.
So the clause that named the standard was pointing at a document that did not cover the goods, and the clauses that mattered, the ones on brake force, ramp holding and push force, were all sitting in the wrong standard. They exist in YY 0003-2023 for the Chinese market, and the English-language equivalents for a European buyer sit in ISO 13442 and ISO 22882.
This is worth chasing on every medical furniture order, because it is a defect in the buyer's own paperwork that no amount of factory inspection repairs. The buyer's specification was the only place where the brake could have been defined, and it defined it in a document that did not apply.
What the verification was worth
The order was forty beds at $265, so $10,600. Thirty per cent was paid on order and seventy per cent, $7,420, was due against delivery.
The six beds would have shipped looking perfect. Nothing about them is visible: the pedal moves, the indicator works, the castors are the right castors and the frame is the right frame. They would have passed the buyer's own goods-in check by anyone who pressed the pedal and looked at the indicator, and they would have reached a care home as beds with no brake.
| Line | Amount |
|---|---|
| Order value, 40 beds at $265 | $10,600 |
| Balance due on delivery | $7,420 |
| Inspection, 2 man-days at $199 | $398 |
| Re-inspection of the six reworked beds, 1 man-day at $199 | $199 |
| Total spent before paying the balance | $597 |
The number that is not in the file
The verification cost $597, which is just over 8 per cent of the $7,420 balance being held and under 6 per cent of the order value.
Now the harder half of the question. What does a bed with no brake cost when it is already in a care home? Not the bed, which is replaceable. The incident. A bed that moves when a resident stands on it, or when a nurse transfers somebody onto it, is a fall, and the paperwork that follows is not the buyer's to write. That number is not in any file and nobody in the chain can quote it, which is exactly why it belongs in the decision at the factory gate and not at the hospital door.
Where to go next
Three questions to ask before the next order
Three questions, and all three can be answered on a phone call to the factory before anybody signs anything.
- How is the brake going to be verified, by pressing it or by measuring it? If the answer is that the pedal moves and the indicator changes, you have bought the factory's check, and on this order that check passed all forty beds including the six that had no brake.
- How much margin does the brake have between working and not working? If the answer is that the setting is right because somebody set it, you are buying a stop screw and a jig, and you should be verifying every bed rather than sampling them.
- Which standard is in the file, and does it apply to the bed you are buying? IEC 60601-2-52 does not cover a fully manual bed. A clause that names the wrong standard cannot be tested, and it cannot be argued about either.
Where to go next
Have an order you need verified? Send us your PI and factory address. Three levels, published prices, $149 per man-day from.