Rcpa Uncertainty

Guideline
Subject: Uncertainty of Measurement

Approval Date: November 2004, November 2009
Review Date: November 2013
Review By: PPAC
Number: 2/2004
Introduction
This Guideline has been developed to provide some preliminary advice to Fellows on how to
approach Uncertainty of Measurement. In addition, NPAAC have an accreditation
requirement concerning estimation of measurement uncertainty.
It is a requirement of ISO 15189 that laboratories “shall determine the uncertainty of

measurement of results, where relevant and possible”………. (5.6.2)
It should be noted that there is nothing new in the concept of uncertainty of measurement.
However, it is likely that many different terms have been used to describe it.
ISO15189 (3.17): The uncertainty of measurement is a parameter associated with the result
of a measurement that characterises the dispersion of the values that could be reasonably
attributed to the measurement.
For example: If the uncertainty of measurement for a test is 10%, and the result is 100, then
the true result probably lies between 90 and 110.
This is a useful parameter for indicating how reliable the result is, or to indicate whether any
change in test results over time might be clinically significant, or just due to the imprecision of
the analytical method used to produce the result.
Common questions pathologists are asked are:
• Is this result significantly abnormal? (usually when it is close to the reference range)
• Is the most recent result for a test significantly different to the previous result? Does the
change represent a significant change in the patient’s condition?
Generally these questions are answered by trying to remember the “uncertainty of

measurement” for that specific test.
ISO 15189 asks for the degree of uncertainty to be both measured and documented.
IANZ (and NATA) have indicated that as a first step, the measurement of uncertainty should
be determined for tests that report the result as a number. This requires an understanding of
the imprecision of the analytical process used for those tests.
Imprecision and the Calculation of Uncertainty of Measurement:
The calculation is based on the imprecision (CV%) of the test (see next page).
Ideally, this should be the imprecision of the test as it is performed in the specific laboratory.
This means that at least 30 sets of internal QC data should be considered, obtained over a
reasonable period of time (not all in one batch), from which the CV% for that test can be
calculated.
The possible sources of the imprecision data for each method include the following:
• For a new test:

Use the CV% data obtained when the test was first established in the specific laboratory.
However, the imprecision should be re-estimated as soon as 30 data points are obtained
as above, to better reflect the routine performance of the test.
• For a rare or expensive test:
The manufacturer’s data on CV% may be used if it is impractical to gather sufficient data
to calculate the imprecision of the test as performed in a specific setting.
• The use of EQA Programme imprecision data, obtained from a group of laboratories
using the same method as the laboratory wishing to use it, may also be acceptable under
special circumstances, but this should generally be used only until there is sufficient in-
house data to illustrate the performance of the test in the laboratory wishing to use it.
• For tests where it is simply not possible to determine the imprecision of a test by
collecting data, a best estimate based on general experience with the test may suffice.
• There are some laboratory results where it is not relevant to determine uncertainty of
measurement.
The imprecision of tests will change over time. Such changes should be monitored to see if
they justify a recalculation of the uncertainty of measurement. Using common sense, and
rounding off the uncertainty of measurement figures for a particular test, will reduce the need
to keep recalculating it on a regular basis (e.g. quoting 7.63% would require constant review,
whereas <10% may be near enough for clinical use, and give sufficient leeway to remain
unaltered over a longer period).
The recommended calculation of uncertainty of measurement, known as the coverage factor,

is:
CV % x 2
The uncertainty of measurement may have to be calculated at more than one level if there is
a marked change in imprecision across a range of results that could alter clinical
interpretation.
There is little point in calculating uncertainty of measurement at concentrations that are well
away from clinical decision making limits.
The uncertainty of measurement can be expressed as a %, or as a specific number to relate

to a specific test value.
E.g. Albumin: 3 g/L at a concentration of 40 g/L
Potassium: 4%
In determining this process, a number of factors need to be considered:

• It has to be a simple calculation that is fit for purpose. Many documents on standards and
statistics describe complex calculations, usually based on the numerous sources of
possible error in the analytical process. Whereas this is required for the preparation of
standards, controls and components of reagents, such meticulous assessment of the
uncertainty or measurement is both impractical and unnecessary for medical laboratory
tests. Only those factors having a significant influence on the final result need to be
considered.
• The overall imprecision of a test includes most of the significant factors that could
practically be assessed.
• Other factors such as non-specificity, method interferences, instability of specimens,
sample selection from specimens, etc, are variable and generally not measurable for a
particular specimen. They should be noted as potential sources of errors either in the
laboratory’s handbook, or on the report form if the potential error could have significant
clinical consequences.
E.g. Potential interference in the troponin assay from heparin in the collection tube
should perhaps be noted on the report form, whereas ketone interference with the
creatinine assay would be more appropriate to mention in a laboratory handbook.
• Factors such as biological variation and age related values are best considered and
included as part of the reference range/interval.
Clinical Application:
For a single test result:

Uncertainty of Measurement is used to indicate the confidence we have that the reported
figure is correct.
If the uncertainty of measurement is calculated as CV% x 2:

There is a 95% chance that the true result lies within a range covered by result number ± the
uncertainty of measurement.
For two results to be significantly different:

Uncertainty of measurement is used to indicate that the difference between two results is
unlikely to be due to method imprecision alone, and is either due to normal biological
variation or a change in the patient’s physiological or pathological condition.
If the uncertainty of measurement is to differentiate two figures with 95% confidence, then
the CV% would be multiplied by 2.77. This is generally not practical for every day use. By
multiplying the CV% by 2 (use the same figure as for a single test) then the confidence is
reduced to 84%. By multiplying the CV% by 3, there can be a 97% confidence level that the
figures are different.
It is recommended that CV% be multiplied by 2 in this situation. The term “significantly

different” should not be used, as this term has a specific statistical meaning (95%), and so
terms such as “highly likely to be different” are preferred. By aiming for 97% confidence (3 x
CV%), then “almost certainly different” may be used. Such a high level of confidence is not
required for most clinical uses of laboratory tests, and there are advantages in having only
one (set of) measurement(s) of uncertainty for each test.
ISO 15189
The requirements contained in ISO15189 should be studied carefully.
ISO 15189: 5.6.2: The laboratory shall determine the uncertainty of results, where relevant
and possible. Uncertainty components which are of importance will be taken into account.
Sources that contribute to uncertainty may include sampling, sample preparation, sample
portion selection, calibrators, reference materials, input quantities, equipment used,
environmental conditions, condition of the sample and changes of operator.
In the case where results are reported as either positive or negative, it is expected that what
would be reported is the certainty the result is not a false positive or a false negative.
If it is impossible to gather data to determine uncertainty of measurement, then a reasonable

estimate is acceptable, but this needs to be done only if uncertainty of measurement has
clinical relevance.
Uncertainty of measurement is irrelevant in disciplines such as Anatomical Pathology for

descriptive results in histology reports. However, the “uncertainty of measurement”
(unreliability) of the cervical screening test would be beneficial in educating the public that it
is a screening test and not a diagnostic test, with the associated increased likelihood of
errors.
ISO 15189: 5.8.3 (k) ……… and where applicable, information on detection limit and
uncertainty of measurement should be provided upon request.
This second requirement makes it clear that uncertainty of measurement does not have to be
included on every report. It simply needs to be available should anyone enquire about the
reliability of the result supplied to them.
Calculating Uncertainty of Measurement
The CV% x 2 is a good calculation of uncertainty of measurement, but there are some other
factors that may need to be noted in the laboratory’s handbook and/or as a comment on the
report form.
Clause 5.6.2 (see above) lists a number of factors that should be considered when
calculating the uncertainty of measurement:
• Sampling – for most tests this has a minor effect, and so need not be included. If it is
critical (unstable specimen), then a note should be made on the report or in the
Handbook indicating the effect of unsuitable collection and delivery of the sample.
Sampling is not a constant variable that affects most measurements, and so is difficult to
estimate. If it is a constant factor, then this should be allowed for in the reference
range/interval. If timing of sampling is important, then notes on specimen collection and
the reference range/interval should indicate this.
• Sample preparation – included in the CV%.
• Sample portion selection – not relevant for most tests that report a figure. Variations are
usually accounted for by sub-sampling (e.g. sub-sample of a 24 hr urine collection rather
than a spot sample), or by expressing the result as a ratio (e.g. a urine concentration as a
ratio to urine creatinine output).
• Calibrators – variation in calibration may be included as part of the CV% and may also be
reflected in the reference range/interval.
• Equipment used – effect included in the CV%.
• Environmental conditions – effect included in the CV%.
• Condition of the sample – may affect a proportion of specimens, but collection
procedures should deal with those tests/samples where this is a potential issue.
• Changes of operator – should be a minor effect, but will be included in the CV% over a
period of time.
From the above, it can be seen that CV% derived from internal QC data over a period of time
will cover most of the significant factors that will add to the uncertainty of measurement.
Preparation for ISO 15189
The College recommends the following:
1. Make a list of all of the tests where the result is reported as a number.
2. From this, make a sub-list of every test which is already in the in-house CV% data.
3. Multiply the CV% by 2 and record this figure as the uncertainty of measurement against
each of these tests. This should cover the majority of tests that a particular laboratory
reports as a number.
4. Determine the CV% for the remaining tests, either by:
• reviewing internal QC data (30 sets) and determining the CV%
• obtaining CV% for method employed from an External QA Programme
• obtaining CV% for method employed from the manufacturer (reagent kit insert)
5. Check if it is either impractical or irrelevant to determine the uncertainty of measurement
for the remaining tests, and document your reasons if these are not obvious.
6. Finally, document the policy for determining the uncertainty of measurement, the
calculations used, and the sources of data used for these calculations.
Related Documents:
• NPAAC (2007) Requirements for the Estimation of Measurement Uncertainty,
http://www.health.gov.au/internet/main/publishing.nsf/Content/health-npaac-emu.htm
• Guidance on the number of significant figures may be found in Hawkins, R.C. and
Johnson, R.N. “The significance of significant figures”, Clinical Chemistry (1990)
36(5):824.

Rcpa Uncertainty

Uploaded by

Copyright:

Available Formats

Rcpa Uncertainty

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Rcpa Uncertainty

Uploaded by

Copyright:

Available Formats

Guideline

Subject: Uncertainty of Measurement

It is a requirement of ISO 15189 that laboratories “shall determine the uncertainty of

Generally these questions are answered by trying to remember the “uncertainty of

Imprecision and the Calculation of Uncertainty of Measurement:

• For a new test:

The recommended calculation of uncertainty of measurement, known as the coverage factor,

The uncertainty of measurement can be expressed as a %, or as a specific number to relate

In determining this process, a number of factors need to be considered:

For a single test result:

If the uncertainty of measurement is calculated as CV% x 2:

For two results to be significantly different:

It is recommended that CV% be multiplied by 2 in this situation. The term “significantly

The requirements contained in ISO15189 should be studied carefully.

If it is impossible to gather data to determine uncertainty of measurement, then a reasonable

Uncertainty of measurement is irrelevant in disciplines such as Anatomical Pathology for

Calculating Uncertainty of Measurement

Preparation for ISO 15189

The College recommends the following:

You might also like