Measures of assay performance produce an on-going assessment of assays so it is logical to use this data for MU assessment. Below are some of the most commonly performed assay performance assessments and suggestions how the data produced from such processes can be used for MU.
In most circumstances an assessment of repeatability will be mandatory as a minimum.
Examples of repeatability measures include the Standard Deviation (SD), Standard Deviation of the Mean, Coefficient of Variation (CV) and the Variance of an entire process as assessed by Internal Quality Control.
However, the quantification of individual uncertainty contributors can also be assessed by their repeatability.
Similar to repeatability, reproducibility is assessed statistically using similar information. The difference lies with rather that measuring the same thing, in the same way, a variable should be changed in order to determine that the measurement result is transferable across differing analytical conditions. Examples of variables influential in the determination of reproducibility are:
- Operator to Operator reproducibility
- Reproducibility across equipment (particularly applicable to inter-analyser variability in pathology laboratories)
- Variability between methods (in reality this would be differentiated at the definition of the measurand stage)
- Variability over a prolonged period of time (longer than that assessed for repeatability)
Inter Analyser Comparability
Results of these assessments may be used to calculate the influence on MU for each applicable assay.
Stability is assessed for those assays that use a stored calibration curve only. The stability is the property of (for example) an analyser ensuring the measurements are constant over a period of time. This is not to be confused with drift (systematic error over time) which is assessed separately.
Drift is concerned with the change in the properties of the instrument used to make a measurement over time. It is a systematic uncertainty as the results are drifting in a single direction away from where they were previously.
Resolution is the precision of the result provided by the measurement being made – the smallest change in the quantity being measured and reported. For example, if analysers report results of a Prothrombin Time as 13.1s the resolution is 0.1 seconds. This resolution is not expected to change for the assay, irrespective of the final result.
The resolution will be determined by one (or many) of a number of factors including:
- Specification of the analyser
- International guidelines/consensus on reporting values
- Locally derived criteria
It is recommended to include this in all uncertainty budgets. However, applying this to all pathology assays may not be required as the significance of the contribution of resolution to the final calculation may be expected to be significantly lower than other contributors.
Reference standard uncertainty and stability
Reference standard uncertainty refers to the uncertainty associated with the standard used to calibrate other measurements. In practise this data is obtained from the manufacturer’s specification of calibrators.