EBME Forums Forums Free discussion Free discussion [poll]What do YOU use to check the accuracy of infusion devices?

K.I.T.T.

Hows about using a capacitive strip sensor as used in fuel tanks. They are seen medically in urodynamics investigations when measuring the total volume and rate of flow of urine. I've no idea what degree of depth acuracy these give.

I know with urodynamic systems the receptacle has to be primed with a small amount of fluid before starting and they must be cleaned and dried after use and not left soaking overnight. I think that is because of the corrosive powers of the urine and the contaminants that will coat the sensor if left soaking rather than any requirement for the sensor to be stored out of fluids.

Accuracy would almost certainly be inadeqaute for pump calibration purposes but would probably be ok for IDA type testing and particularly for checking linearity of delivery. In fact given the continuous nature of the sensor's function it could give a superior graphing function and would eliminate problems with artefacting due to intermittent nature of readings from a conventional IDA.

It seems to me the fundamental problem for you is choice of transducer.

Marc

K.I.T.T.

Hows about using a capacitive strip sensor as used in fuel tanks. They are seen medically in urodynamics investigations when measuring the total volume and rate of flow of urine. I've no idea what degree of depth acuracy these give.

I know with urodynamic systems the receptacle has to be primed with a small amount of fluid before starting and they must be cleaned and dried after use and not left soaking overnight. I think that is because of the corrosive powers of the urine and the contaminants that will coat the sensor if left soaking rather than any requirement for the sensor to be stored out of fluids.

Accuracy would almost certainly be inadeqaute for pump calibration purposes but would probably be ok for IDA type testing and particularly for checking linearity of delivery. In fact given the continuous nature of the sensor's function it could give a superior graphing function and would eliminate problems with artefacting due to intermittent nature of readings from a conventional IDA.

It seems to me the fundamental problem for you is choice of transducer.

Marc

K.I.T.T.

Hows about using a capacitive strip sensor as used in fuel tanks. They are seen medically in urodynamics investigations when measuring the total volume and rate of flow of urine. I've no idea what degree of depth acuracy these give.

I know with urodynamic systems the receptacle has to be primed with a small amount of fluid before starting and they must be cleaned and dried after use and not left soaking overnight. I think that is because of the corrosive powers of the urine and the contaminants that will coat the sensor if left soaking rather than any requirement for the sensor to be stored out of fluids.

Accuracy would almost certainly be inadeqaute for pump calibration purposes but would probably be ok for IDA type testing and particularly for checking linearity of delivery. In fact given the continuous nature of the sensor's function it could give a superior graphing function and would eliminate problems with artefacting due to intermittent nature of readings from a conventional IDA.

It seems to me the fundamental problem for you is choice of transducer.

Marc

I thought we were being asked what methods we use to measure the accuracy of our infusion pumps that's why I gave a faily detailed response! I'm sorry if this has upset anyone.
Clearly, any new method that someone can come up with would be most welcome. In my view, it needs to be simple to operate and maintain and give the required accuracy. Moreover, with pressure on budgets it need to be cheap!
However, let's go back and ask ourselves why we need to measure the accuracy of any item of medical equipment? In the case of an infusion pump we want to be sure that the volume delivered and the rate at which it's delivered is as close as possible to settings carried out by the nursing staff when it was initially set up i.e. "it does what it says on the tin" (I'm sorry if this is stating the bloody obvious!)Again, how accurate does this measurement have to be? "I refer the honorable gentleman to my previous answer" IT NEEDS TO BE WITHIN THE MANUFACTURERS SPECIFICATION. Therefore, we should always use this spec. when we make measurements and/or use the test equipment which the manufacturer specifies. For example, in the case of the Arcomedical Volumed uVP5005 infusion pump (early service manual - I can't find the later one at the moment) the deviation in flow-rate using the correct giving set was specified as +/- 5%. The delivery accuracy in volume was specified using a measuring beaker and measured at 25 ml with a rate of 250 ml/h over a six minute period This should be 25 cm3 or 25 g +/- 5% of water. Hence, if the pump meets this spec. it's OK.
I rest my case.

No definitely not - total worst-case error or uncertainty in the system is the worst-case uncertainty in the measuring device added to the worst-case uncertainties in the device under test (IUT). That's a fundamental concept of measurement and intrumentation theory.

For example if you have a measuring instrument capable of measuring a parameter to an accuracy of +/-20% then the worst-case error will be added to the worst-case accuracy of the IUT, say 2%.

If you have an uncertainty in the measurement instrument much greater than the accuracy of the IUT, say 15% (within its +/- 20% tolerance), how much of this 15% error (from the nominal reading obtained from the IUT) is IUT error? Think about it. The 2% potential error in the IUT is "buried" in the "noise" of the measuring instrument uncertainty.

Limit the measurement instrument error to a fraction of the IUT system error and it becomes negligible - then the majority of error in the reading is IUT error added to a tiny proportion of measurement error. Think why measurement standards always are highly accurate i.e. measure to a high precision and resolution. It's to make the measurement more representative of the actual uncertainty, i.e. error, in the parameter being measured.

In your example worst-case error could possibly +/-10%, overall. If you had a syringe driver with a spec of 2% and got an error of 7% how much of this is actual pump error and how much actual cylinder error?. Both pump error and burette error will vary, statistically speaking, from measurement to measurement (burette more so since error-band is greater relatively speaking) so you can't say the burette error will always be +/-5% or pump error will always be +/-2%. So how do you know the contribution of each to the overall reading?

Take a +/-0.2% accurate measuring device and use it to measure a +/-2% specified accuracy syringe driver and the worst-case error will be +/-2.2% overall error. The actual measured pump accuracy will consistently be within +/-0.2% when measurements are repeated; not potentially anywhere between +/-5% between measurements as in your example.

Overall Err(Ind)/Act Bur Err(5%spec)/Act Pump Err

-3%/-5%/+2%
1%/+2%/-1%
1%/-3%/+2%
1.5%/+0.5%/+0.1%
3%/+5%/-2%


Imagine overall error is what you see and calculate from readings on the burette. This has to be actual burette error plus actual pump error - the errors are additive in the real world.

Now you don't know pump error (that's why you need to measure it) or burette error (it varies in the real world between measurements, of diff volumes, technique, etc) between measurements (you only know worst-case error for sure). Imagine that we do know these unknowns, i.e someone has measured pump and burette measurement accuracy after each measurement (there will be variation).

Cover up the two right-hand columns (the unknowns) and now tell me what information pertaining to the pump accuracy and burette error the overall error gives you. You will probably say that as long as the error is less than the worst-case burette error plus worst-case pump error then everything's ok (+/-7% error acceptable max). This doesn't tell you what the likely pump error is though and the next time the pump is tested the value is likely to be significantly different again since the maginitude of measurement instrument errors are relatively large in comparison to pump error.

Now:


Overall Err/Act Bur err (0.2%spec)/Act Pump err

+2.2%/+0.2%/+2%
1%/-0.2%/1.2%
-1%/-0.05%/-0.95

Cover up the unknowns in the two columns on the right; now using the overall error what can you tell me about the pump accuracy, i.e it's likely value and maximum error? Now, since burette error is never >10% of pump error (at optimum volume delivered of course since accuracy of a burette is down to the ratio of volume delivered to resolution andd ability of operator to read it) the error more closely represents the pump accuracy and will on the next test and the test after that - more than the 5% accuracy instrument will.

We are interested in absolute accuracy, i.e. with specified tolerancesm and repeatability if we are to compare actual performance of a pump to specified performance and get results that are meaningful for the next pedant that comes along. Otherwise it's not worth using test instruments if they produce a significant number of readings that are meaningless.

Oredered 2 burettes yesterday and they arrived this morning (1x100 ml, 1x25 ml = £65 ). Ran a test on them today using a baxter flo guard (cant remember which model.... )

anyway, set it up to infuse at a rate of 150 ml/hr and a total volume of 75 ml. I got 79.00 ml on the burette and 79.80 using the IDA4. Not as i'd had in mind....I thought thered be a larger difference

Anyway, that was just a trial really. Ill start proper measurements on moday using a calibrated tube (on the giving set)...

KITT

I think BIME use a system where they have a laboratory grade balance scale where they measure the amount of water that is infused based on the assumption that 1KG of water is 1000ml at STP. I think they also use deoxygenated water and also have a layer of parafin/oil in the collection beaker to minimise evapouration of water (not sure how quickly water evaporates but might be an issue in your burette set when its hot i.e. summer. Also might be an issue in the winter when the air is dry and the heating is on). Could explain why your burette reading is lower than your IDA one.

Contact Teresa Dunn at BIME where I'm sure she will share with you what happens there.

Presuming the usual tolerance of +/-5% on volumetric pumps (excluding enteral feed types) - the actual pump accuracy tolerance just passes using the burette +5.3% and fails using the IDA4 at +6% then. Is this the case? Probably not since you need to add/subtract the worst-case calculated/estimated errors of the burette (resolution divided by nominal volume delivered) at the total volume delivered and that specified in the IDA4 manual to these values to arrive at a range of the likely combined pump/giving-set volumetric tolerances (system accuracy).

The difference or error between the actual reading (that measured) and the nominal reading (that expected) consists of the error in the pump plus the error in the burette or IDA4). If burette reading has 5.3% overall error (pump system error +/-burette error) and calculated accuracy is 0.13% (75ml/0.1ml - I asume 0.1ml resolution?) then estimated pump system tolerances are 5.43% - 5.17%. For IDA4 if reading has 6% error overall error (pump system error +/- IDA error) then estimated pump system tolerances are 4% - 8%. Do you think this reasoning is valid?

Limits of accuracy for measuring instruments are indicated so we can do this jiggery-pokery to arrive at an estimated figure of the actual tolerances in the parameter or device we're measuring; which is preferable to providing a figure that includes the measuring instrument uncertainties. Hopefully manufacturers do all this thinking for us when they specify the measuring method, instrument and tolerances on the measuring instrument. However sometimes this is not the case.

There will always be uncertainty but when two sources of error are combined we can at least estimate the limits of accuracy or tolerances in the unknown source of error, with a level of confidence, if the limits of accuracy of the other source (measuring instrument) is known. Particularly if a trial of measurements are taken, i.e. statistically speaking. Single one-off measurements are a bit tricky thus eliminating sources of error in measurements or measurement technique is always preferable under this circumstance.

Hence setup is important - bag-height (@ 30"?), new set, outlet of giving set/burette datum/IDA ports @ mid pump mechanism height, etc, etc. Of course a scientific investigation would perform a statistical analysis and comparison between the two methods. Repeatability and comparability of tests is essential of course.

I think it's likely that using an appropriate choice of burette to measure a particular volume and infusing at a rate and for a time that is indicated in the trumpet and startup curves for the pump (to ensure the driver/pump is infusing at the rate specified, as accurately as it can, for the test to be consistent and to meet manufacturers specification), it is possible to make measurement errors almost negligible for all practical purposes. This includes syringe drivers.

The result of all this waffle is that that the reading on the burette will be very close to the likely pump system delivered volume. Thing is; can the same be said of the IDA-like devices?

Basic Accuracy of a Burette? Minimum volume that can be resolved by eye divided by full-scale value. Remember the accuracy will vary with total volume delivered (since the minimum volume resolved increases as a proportion of the volume delivered as total volume delivered is reduced)so choice of burette, volume delivered, grade of burette and resolution is important. Are you using Grade A or B burettes KITT?

Otherwise, for type-testing and evaluation, gravimetric methods, as described by Bioman, seem to be the "gold-standard" method of measurement since very accurate mass measurements can be achieved and the tests carried out under very carefully controlled conditions (as specified in British standards) I suspect.

The problem is, that the 100 ml burette has a resolution of 0.2 ml Which would most likely have caused it to fail the test on that as well. My 10 ml burette has an resolution of 0.02 ml, so i think using that and comparing it to an IDA would be a good idea, although i find it hard to think of a situatio where an infusion pump wuld be used to deliver such a small amount of fluid.