MEASUREMENT UNCERTAINTY
Build an evidence map before an uncertainty budget
Distinguish a source of uncertainty from the data used to estimate it.
Revised 30 September 2026. The original publication date and address are retained.
Map effects to evidence
| Effect or question | Potential evidence |
|---|---|
| Routine analytical variation | Representative IQC or precision-study observations |
| Assigned calibration value | Relevant certificate and traceability information |
| Lot or instrument differences | Lot-acceptance and paired comparison studies |
| Bias or correction | Suitable reference materials or comparison procedures |
| Patient-material behaviour | Relevant specimen studies and material-suitability evidence |
| Resolution or rounding | Reporting rules and an appropriate distribution model |
A dataset is not automatically a contribution
IQC, EQA and verification are evidence sources. Traceability is a property of a result, and propagation is a calculation process. Do not treat every heading on a checklist as a separate number to add to a budget.
Prevent double counting
If long-term IQC already spans routine reagent-lot changes, a separate lot-variation term may overlap with it. Identify exactly what each dataset covers, where coverage is incomplete, and which additional term the model genuinely needs.
Existing data can be enough—but not by assumption
Reuse well-characterised data where it answers the question. A short verification, a detection limit or a manufacturer acceptance range is not automatically a standard uncertainty for routine patient results. Additional evidence may be necessary when current data do not cover the intended concentration, specimen or operating conditions.
Use judgement transparently
Type B evaluations may draw on certificates, specifications, prior information and justified scientific judgement. Record the evidence, distribution and limitations. Expert familiarity helps define the model; it does not make an unsupported number equivalent to a measured performance estimate.
The minimum useful record
- Define the effect and the result to which it applies.
- Name the evidence and its period or version.
- Explain how the standard uncertainty is obtained.
- Record overlap, correlation, exclusions and missing evidence.
- Set a review trigger for provisional assumptions.
Sources & further reading
Educational material. Examples are synthetic. Consult applicable full guidance and your locally approved procedure for operational decisions.