Organizations develop within dynamic and constantly changing contexts, which can significantly influence their performance and thus generate uncertainty regarding the fulfillment of established objectives. Clinical laboratories are no strangers to this reality and therefore may have doubts about the level of confidence in the results issued. This uncertainty can be determined by general factors, such as changes in personnel, technological management, information systems, among others, as well as by more specific factors associated with the measurement process used to obtain results from the samples analyzed.

Nowadays, clinical laboratories continue to advance in quality assurance programs to guarantee the reliability of the results issued and that is why statistical models generate new proposals for the analysis of information obtained from quality control strategies, because although, within this progress, the term of measurement uncertainty has taken more and more strength to be included as a strategy that contributes to the interpretation of results.

In the clinical care setting, laboratories function as support services that provide essential information to health care personnel for medical decision making, therefore, the timeliness and certainty of test results are considered essential attributes. In this sense, the estimation of measurement uncertainty (MU) is recognized as a key element attributable to the reliability of results.

As part of quality assurance, compliance with technical quality specifications such as accuracy and trueness is controlled. However, by incorporating a third attribute related to the estimation and evaluation of measurement uncertainty (MU), it allows a better assessment of the accuracy of the results.

While the term measurement uncertainty is related to the initials MU for Measurement Uncertainty and is defined as a non-negative parameter that characterizes the dispersion of the values attributed to a measurand (test or assay), based on the estimate made by[i]. Therefore, it is evident in technical documents that the uncertainty is a numerical value expressed with a +/- sign (it is not an absolute value) associated with the reported data of a test or assay and cannot be calculated, it can only be estimated can only be estimated in an approximate way to be considered as an interval in which the true value is found.

The measurement processes performed in the laboratory contemplate several attributes to evaluate the quality of the results, let’s see the following model that allows to identify the accuracy of the results of quantitative tests through some statistical inducers: calculation of the bias (unidirectional value), calculation of the precision (bidirectional value), and estimation of the measurement uncertainty or MU (bidirectional value), each attribute being interpreted as reflected in the following image:

 

 

 

Figure 1: interpretation of precision, bias and measurement uncertainty (author’s own).

Let’s look at these attributes in terms of data, the following technical performance of the test [Albumin assay] is known with the following data:

1. Accuracy of 1.73% (%hp)
2. Bias of 2.56% (%Bias)
3. ET 5.4% (%ET)
4. Measurement uncertainty of 6.19% (MU%),

This information could be interpreted as follows:

 

Attribute	Inducer	Interpretation from a practical approach.	Application
Accuracy	% CV	Dispersion or variation of the results of a test if the measurement is performed multiple times under defined measurement conditions, given by the measurement process. measurement process.	If the same patient sample is processed more than once, the variation between all data would be 1.73%.
Accuracy	%Bias	Proximity between the result obtained from a test and the reference or «true» value given by the measurement process. measurement process.	The patient’s sample result may be 2.56% away from the true value.
Total Error	% TE	It is the total error that a test result may have considering the sum of the precision and bias obtained from the measurement process.	The result of the patient sample has a total error of 5.4%.
Measurement uncertainty	MU	Range or a range of the reported result issued from a test where the true value is found.	If the patient’s sample result is 4.0 g/dL the true value may be between: 3.8 g/dL and 4.2 g/dL.

Table 1: interpretation of precision, bias, total error and measurement uncertainty

 

Within the quality assurance model for tests or trials, control charts are constructed to calculate and monitor the attributes of trueness and precision, where they are subsequently evaluated against the maximum admissible goals or tolerances (commonly called quality goals), but the estimation and evaluation of measurement uncertainty is rarely considered. Consequently, the term measurement uncertainty has become more relevant in recent times where the importance of not only estimating it but also evaluating the result obtained is projected, as reflected in the updated requirements for quality and competence of clinical laboratories under ISO 15189:2022 in paragraph 7.3.4 whose requirement relates «Evaluation of measurement uncertainty (MU)». [i].

 

In relation to MU estimation, it is important to rely on a methodological design that allows obtaining the most approximate value possible, therefore, it is advisable to take into account the following aspects, the first one the definition of a prudent time that allows contemplating changes in measurement conditions (more than 6 months) and as a second aspect the combination of all contributions or sources of uncertainties accumulated along the entire traceability chain, including the uncertainty of reference materials, assignment of calibrator values and random variability of measurement systems [ii]. ISO/TS 20914:2019 «Practical guide for the estimation of measurement uncertainty» and the CLSI EP29 guide provide information for the estimation and expression of measurement uncertainty.

In relation to the evaluation of MU, 3 models developed and recommended by the EFLM Strategic Conference held in Milan in 2014 can be considered. However, recent studies have identified the importance of previously performing a classification for each of the tests vs. a single model (of those proposed) in such a way that allows to evaluate and consider the acceptable or «allowed» measurement uncertainty in a more appropriate way, [iii] for example:

• For model 1 for disease-specific diagnostic and follow-up testing (e.g., HbA1C), because MU directly assesses the impact of analytical variation in the assay.
• For model 2, we consider tests with strict metabolic control that have a very good homeostatic control or are very stable when the person is not «sick» (e.g. total plasma bilirubin) because the evaluation of MU is estimated from intraindividual biological variability.
• For model 3 tests that do not fall into the previous 2 groups (e.g. hCG), this model is considered based on the state of the art from the highest quality of the technically achievable analytical performance of the test.

Recent studies have also published statistical models for the evaluation of the MU obtained, so that the estimate can be compared with acceptable or «allowed» uncertainty values:

• Model A. – U from the Total Administrative Error (Eta) [iv] [iv].
• Model B – A stricter model suggests that the U of a measurand should be less than 2/3 of the maximum allowable error (Eta), in case it is higher, the different sources of uncertainty should be studied in more detail and corrective actions should be taken to reduce them [v].

Conclusion.

Finally, laboratories have a commitment to define, implement and maintain procedures that contribute to monitor the quality assurance system in the measurement of tests or assays, so it is recommended to carry out processes for the estimation of measurement uncertainty (MU) through the different models proposed and evaluation of the results obtained from this estimate with acceptability criteria or with defined performance goals, so as to contribute to the reliability of the results and contribute to patient safety.

[i] (ICONTEC, 2023)

[ii] (Federica Braga, 2021)

[iii] (Federica Braga, 2021)

[iv] (Carolina Bignone, 2019).

[v] (Carolina Bignone, 2019).

Extensive references

 

1. Metrology, C. E. (JCGM 200:2012). International Vocabulary of Metrology Fundamental and general concepts and associated terms (VIM). Government of Spain. Ministry of Industry, Energy and Tourism.
2. ICONTEC. (2023). NTC-ISO 15189:2022 Clinical laboratories. Requirements for quality and competence. . Bogotá: ICONTEC.

3.4 Federica Braga, M. P. (2021). Performance specifications for measurement uncertainty of common biochemical measurands according to Milan models. Clin Chem Lab Med, 1362-1368.

5.6 Carolina Bignone, E. O. (2019). Evaluation of total and 6Sigma error performance and estimation of measurement uncertainty of 16 clinical biochemistry quantities. Rev Lab Clin, 69-77